diff options
Diffstat (limited to 'fs/ocfs2')
107 files changed, 27545 insertions, 10015 deletions
diff --git a/fs/ocfs2/Kconfig b/fs/ocfs2/Kconfig index 701b7a3a872..77a8de5f711 100644 --- a/fs/ocfs2/Kconfig +++ b/fs/ocfs2/Kconfig @@ -1,11 +1,11 @@ config OCFS2_FS tristate "OCFS2 file system support" - depends on NET && SYSFS - select CONFIGFS_FS + depends on NET && SYSFS && CONFIGFS_FS select JBD2 select CRC32 select QUOTA select QUOTA_TREE + select FS_POSIX_ACL help OCFS2 is a general purpose extent based shared disk cluster file system with many similarities to ext3. It supports 64 bit inode @@ -50,7 +50,7 @@ config OCFS2_FS_USERSPACE_CLUSTER config OCFS2_FS_STATS bool "OCFS2 statistics" - depends on OCFS2_FS + depends on OCFS2_FS && DEBUG_FS default y help This option allows some fs statistics to be captured. Enabling @@ -74,12 +74,3 @@ config OCFS2_DEBUG_FS This option will enable expensive consistency checks. Enable this option for debugging only as it is likely to decrease performance of the filesystem. - -config OCFS2_FS_POSIX_ACL - bool "OCFS2 POSIX Access Control Lists" - depends on OCFS2_FS - select FS_POSIX_ACL - default n - help - Posix Access Control Lists (ACLs) support permissions for users and - groups beyond the owner/group/world scheme. diff --git a/fs/ocfs2/Makefile b/fs/ocfs2/Makefile index 01596079dd6..ce210d4951a 100644 --- a/fs/ocfs2/Makefile +++ b/fs/ocfs2/Makefile @@ -1,6 +1,6 @@ -EXTRA_CFLAGS += -Ifs/ocfs2 +ccflags-y := -Ifs/ocfs2 -EXTRA_CFLAGS += -DCATCH_BH_JBD_RACES +ccflags-y += -DCATCH_BH_JBD_RACES obj-$(CONFIG_OCFS2_FS) += \ ocfs2.o \ @@ -28,6 +28,9 @@ ocfs2-objs := \ locks.o \ mmap.o \ namei.o \ + refcounttree.o \ + reservations.o \ + move_extents.o \ resize.o \ slot_map.o \ suballoc.o \ @@ -35,19 +38,16 @@ ocfs2-objs := \ symlink.o \ sysfile.o \ uptodate.o \ - ver.o \ quota_local.o \ quota_global.o \ - xattr.o - -ifeq ($(CONFIG_OCFS2_FS_POSIX_ACL),y) -ocfs2-objs += acl.o -endif + xattr.o \ + acl.o ocfs2_stackglue-objs := stackglue.o ocfs2_stack_o2cb-objs := stack_o2cb.o ocfs2_stack_user-objs := stack_user.o +obj-$(CONFIG_OCFS2_FS) += dlmfs/ # cluster/ is always needed when OCFS2_FS for masklog support obj-$(CONFIG_OCFS2_FS) += cluster/ obj-$(CONFIG_OCFS2_FS_O2CB) += dlm/ diff --git a/fs/ocfs2/acl.c b/fs/ocfs2/acl.c index fbeaec76210..7e8282dcea2 100644 --- a/fs/ocfs2/acl.c +++ b/fs/ocfs2/acl.c @@ -21,15 +21,17 @@ #include <linux/init.h> #include <linux/module.h> +#include <linux/slab.h> #include <linux/string.h> -#define MLOG_MASK_PREFIX ML_INODE #include <cluster/masklog.h> #include "ocfs2.h" #include "alloc.h" #include "dlmglue.h" #include "file.h" +#include "inode.h" +#include "journal.h" #include "ocfs2_fs.h" #include "xattr.h" @@ -49,10 +51,6 @@ static struct posix_acl *ocfs2_acl_from_xattr(const void *value, size_t size) return ERR_PTR(-EINVAL); count = size / sizeof(struct posix_acl_entry); - if (count < 0) - return ERR_PTR(-EINVAL); - if (count == 0) - return NULL; acl = posix_acl_alloc(count, GFP_NOFS); if (!acl) @@ -63,7 +61,20 @@ static struct posix_acl *ocfs2_acl_from_xattr(const void *value, size_t size) acl->a_entries[n].e_tag = le16_to_cpu(entry->e_tag); acl->a_entries[n].e_perm = le16_to_cpu(entry->e_perm); - acl->a_entries[n].e_id = le32_to_cpu(entry->e_id); + switch(acl->a_entries[n].e_tag) { + case ACL_USER: + acl->a_entries[n].e_uid = + make_kuid(&init_user_ns, + le32_to_cpu(entry->e_id)); + break; + case ACL_GROUP: + acl->a_entries[n].e_gid = + make_kgid(&init_user_ns, + le32_to_cpu(entry->e_id)); + break; + default: + break; + } value += sizeof(struct posix_acl_entry); } @@ -89,7 +100,21 @@ static void *ocfs2_acl_to_xattr(const struct posix_acl *acl, size_t *size) for (n = 0; n < acl->a_count; n++, entry++) { entry->e_tag = cpu_to_le16(acl->a_entries[n].e_tag); entry->e_perm = cpu_to_le16(acl->a_entries[n].e_perm); - entry->e_id = cpu_to_le32(acl->a_entries[n].e_id); + switch(acl->a_entries[n].e_tag) { + case ACL_USER: + entry->e_id = cpu_to_le32( + from_kuid(&init_user_ns, + acl->a_entries[n].e_uid)); + break; + case ACL_GROUP: + entry->e_id = cpu_to_le32( + from_kgid(&init_user_ns, + acl->a_entries[n].e_gid)); + break; + default: + entry->e_id = cpu_to_le32(ACL_UNDEFINED_ID); + break; + } } return ocfs2_acl; } @@ -98,15 +123,11 @@ static struct posix_acl *ocfs2_get_acl_nolock(struct inode *inode, int type, struct buffer_head *di_bh) { - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); int name_index; char *value = NULL; struct posix_acl *acl; int retval; - if (!(osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL)) - return NULL; - switch (type) { case ACL_TYPE_ACCESS: name_index = OCFS2_XATTR_INDEX_POSIX_ACL_ACCESS; @@ -139,40 +160,68 @@ static struct posix_acl *ocfs2_get_acl_nolock(struct inode *inode, return acl; } - /* - * Get posix acl. + * Helper function to set i_mode in memory and disk. Some call paths + * will not have di_bh or a journal handle to pass, in which case it + * will create it's own. */ -static struct posix_acl *ocfs2_get_acl(struct inode *inode, int type) +static int ocfs2_acl_set_mode(struct inode *inode, struct buffer_head *di_bh, + handle_t *handle, umode_t new_mode) { - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - struct buffer_head *di_bh = NULL; - struct posix_acl *acl; - int ret; + int ret, commit_handle = 0; + struct ocfs2_dinode *di; + + if (di_bh == NULL) { + ret = ocfs2_read_inode_block(inode, &di_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + } else + get_bh(di_bh); + + if (handle == NULL) { + handle = ocfs2_start_trans(OCFS2_SB(inode->i_sb), + OCFS2_INODE_UPDATE_CREDITS); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto out_brelse; + } - if (!(osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL)) - return NULL; + commit_handle = 1; + } - ret = ocfs2_inode_lock(inode, &di_bh, 0); - if (ret < 0) { + di = (struct ocfs2_dinode *)di_bh->b_data; + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { mlog_errno(ret); - acl = ERR_PTR(ret); - return acl; + goto out_commit; } - acl = ocfs2_get_acl_nolock(inode, type, di_bh); + inode->i_mode = new_mode; + inode->i_ctime = CURRENT_TIME; + di->i_mode = cpu_to_le16(inode->i_mode); + di->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec); + di->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec); + ocfs2_update_inode_fsync_trans(handle, inode, 0); - ocfs2_inode_unlock(inode, 0); + ocfs2_journal_dirty(handle, di_bh); +out_commit: + if (commit_handle) + ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle); +out_brelse: brelse(di_bh); - - return acl; +out: + return ret; } /* * Set the access or default ACL of an inode. */ -static int ocfs2_set_acl(handle_t *handle, +int ocfs2_set_acl(handle_t *handle, struct inode *inode, struct buffer_head *di_bh, int type, @@ -192,14 +241,19 @@ static int ocfs2_set_acl(handle_t *handle, case ACL_TYPE_ACCESS: name_index = OCFS2_XATTR_INDEX_POSIX_ACL_ACCESS; if (acl) { - mode_t mode = inode->i_mode; + umode_t mode = inode->i_mode; ret = posix_acl_equiv_mode(acl, &mode); if (ret < 0) return ret; else { - inode->i_mode = mode; if (ret == 0) acl = NULL; + + ret = ocfs2_acl_set_mode(inode, di_bh, + handle, mode); + if (ret) + return ret; + } } break; @@ -230,250 +284,29 @@ static int ocfs2_set_acl(handle_t *handle, return ret; } -int ocfs2_check_acl(struct inode *inode, int mask) -{ - struct posix_acl *acl = ocfs2_get_acl(inode, ACL_TYPE_ACCESS); - - if (IS_ERR(acl)) - return PTR_ERR(acl); - if (acl) { - int ret = posix_acl_permission(inode, acl, mask); - posix_acl_release(acl); - return ret; - } - - return -EAGAIN; -} - -int ocfs2_acl_chmod(struct inode *inode) -{ - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - struct posix_acl *acl, *clone; - int ret; - - if (S_ISLNK(inode->i_mode)) - return -EOPNOTSUPP; - - if (!(osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL)) - return 0; - - acl = ocfs2_get_acl(inode, ACL_TYPE_ACCESS); - if (IS_ERR(acl) || !acl) - return PTR_ERR(acl); - clone = posix_acl_clone(acl, GFP_KERNEL); - posix_acl_release(acl); - if (!clone) - return -ENOMEM; - ret = posix_acl_chmod_masq(clone, inode->i_mode); - if (!ret) - ret = ocfs2_set_acl(NULL, inode, NULL, ACL_TYPE_ACCESS, - clone, NULL, NULL); - posix_acl_release(clone); - return ret; -} - -/* - * Initialize the ACLs of a new inode. If parent directory has default ACL, - * then clone to new inode. Called from ocfs2_mknod. - */ -int ocfs2_init_acl(handle_t *handle, - struct inode *inode, - struct inode *dir, - struct buffer_head *di_bh, - struct buffer_head *dir_bh, - struct ocfs2_alloc_context *meta_ac, - struct ocfs2_alloc_context *data_ac) -{ - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - struct posix_acl *acl = NULL; - int ret = 0; - - if (!S_ISLNK(inode->i_mode)) { - if (osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) { - acl = ocfs2_get_acl_nolock(dir, ACL_TYPE_DEFAULT, - dir_bh); - if (IS_ERR(acl)) - return PTR_ERR(acl); - } - if (!acl) - inode->i_mode &= ~current_umask(); - } - if ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) && acl) { - struct posix_acl *clone; - mode_t mode; - - if (S_ISDIR(inode->i_mode)) { - ret = ocfs2_set_acl(handle, inode, di_bh, - ACL_TYPE_DEFAULT, acl, - meta_ac, data_ac); - if (ret) - goto cleanup; - } - clone = posix_acl_clone(acl, GFP_NOFS); - ret = -ENOMEM; - if (!clone) - goto cleanup; - - mode = inode->i_mode; - ret = posix_acl_create_masq(clone, &mode); - if (ret >= 0) { - inode->i_mode = mode; - if (ret > 0) { - ret = ocfs2_set_acl(handle, inode, - di_bh, ACL_TYPE_ACCESS, - clone, meta_ac, data_ac); - } - } - posix_acl_release(clone); - } -cleanup: - posix_acl_release(acl); - return ret; -} - -static size_t ocfs2_xattr_list_acl_access(struct inode *inode, - char *list, - size_t list_len, - const char *name, - size_t name_len) -{ - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - const size_t size = sizeof(POSIX_ACL_XATTR_ACCESS); - - if (!(osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL)) - return 0; - - if (list && size <= list_len) - memcpy(list, POSIX_ACL_XATTR_ACCESS, size); - return size; -} - -static size_t ocfs2_xattr_list_acl_default(struct inode *inode, - char *list, - size_t list_len, - const char *name, - size_t name_len) -{ - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - const size_t size = sizeof(POSIX_ACL_XATTR_DEFAULT); - - if (!(osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL)) - return 0; - - if (list && size <= list_len) - memcpy(list, POSIX_ACL_XATTR_DEFAULT, size); - return size; -} - -static int ocfs2_xattr_get_acl(struct inode *inode, - int type, - void *buffer, - size_t size) -{ - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - struct posix_acl *acl; - int ret; - - if (!(osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL)) - return -EOPNOTSUPP; - - acl = ocfs2_get_acl(inode, type); - if (IS_ERR(acl)) - return PTR_ERR(acl); - if (acl == NULL) - return -ENODATA; - ret = posix_acl_to_xattr(acl, buffer, size); - posix_acl_release(acl); - - return ret; -} - -static int ocfs2_xattr_get_acl_access(struct inode *inode, - const char *name, - void *buffer, - size_t size) -{ - if (strcmp(name, "") != 0) - return -EINVAL; - return ocfs2_xattr_get_acl(inode, ACL_TYPE_ACCESS, buffer, size); -} - -static int ocfs2_xattr_get_acl_default(struct inode *inode, - const char *name, - void *buffer, - size_t size) +int ocfs2_iop_set_acl(struct inode *inode, struct posix_acl *acl, int type) { - if (strcmp(name, "") != 0) - return -EINVAL; - return ocfs2_xattr_get_acl(inode, ACL_TYPE_DEFAULT, buffer, size); + return ocfs2_set_acl(NULL, inode, NULL, type, acl, NULL, NULL); } -static int ocfs2_xattr_set_acl(struct inode *inode, - int type, - const void *value, - size_t size) +struct posix_acl *ocfs2_iop_get_acl(struct inode *inode, int type) { - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_super *osb; + struct buffer_head *di_bh = NULL; struct posix_acl *acl; - int ret = 0; + int ret = -EAGAIN; + osb = OCFS2_SB(inode->i_sb); if (!(osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL)) - return -EOPNOTSUPP; - - if (!is_owner_or_cap(inode)) - return -EPERM; - - if (value) { - acl = posix_acl_from_xattr(value, size); - if (IS_ERR(acl)) - return PTR_ERR(acl); - else if (acl) { - ret = posix_acl_valid(acl); - if (ret) - goto cleanup; - } - } else - acl = NULL; + return NULL; - ret = ocfs2_set_acl(NULL, inode, NULL, type, acl, NULL, NULL); + ret = ocfs2_read_inode_block(inode, &di_bh); + if (ret < 0) + return ERR_PTR(ret); -cleanup: - posix_acl_release(acl); - return ret; -} + acl = ocfs2_get_acl_nolock(inode, type, di_bh); -static int ocfs2_xattr_set_acl_access(struct inode *inode, - const char *name, - const void *value, - size_t size, - int flags) -{ - if (strcmp(name, "") != 0) - return -EINVAL; - return ocfs2_xattr_set_acl(inode, ACL_TYPE_ACCESS, value, size); -} + brelse(di_bh); -static int ocfs2_xattr_set_acl_default(struct inode *inode, - const char *name, - const void *value, - size_t size, - int flags) -{ - if (strcmp(name, "") != 0) - return -EINVAL; - return ocfs2_xattr_set_acl(inode, ACL_TYPE_DEFAULT, value, size); + return acl; } - -struct xattr_handler ocfs2_xattr_acl_access_handler = { - .prefix = POSIX_ACL_XATTR_ACCESS, - .list = ocfs2_xattr_list_acl_access, - .get = ocfs2_xattr_get_acl_access, - .set = ocfs2_xattr_set_acl_access, -}; - -struct xattr_handler ocfs2_xattr_acl_default_handler = { - .prefix = POSIX_ACL_XATTR_DEFAULT, - .list = ocfs2_xattr_list_acl_default, - .get = ocfs2_xattr_get_acl_default, - .set = ocfs2_xattr_set_acl_default, -}; diff --git a/fs/ocfs2/acl.h b/fs/ocfs2/acl.h index 8f6389ed4da..3fce68d0862 100644 --- a/fs/ocfs2/acl.h +++ b/fs/ocfs2/acl.h @@ -26,33 +26,14 @@ struct ocfs2_acl_entry { __le32 e_id; }; -#ifdef CONFIG_OCFS2_FS_POSIX_ACL - -extern int ocfs2_check_acl(struct inode *, int); -extern int ocfs2_acl_chmod(struct inode *); -extern int ocfs2_init_acl(handle_t *, struct inode *, struct inode *, - struct buffer_head *, struct buffer_head *, - struct ocfs2_alloc_context *, - struct ocfs2_alloc_context *); - -#else /* CONFIG_OCFS2_FS_POSIX_ACL*/ - -#define ocfs2_check_acl NULL -static inline int ocfs2_acl_chmod(struct inode *inode) -{ - return 0; -} -static inline int ocfs2_init_acl(handle_t *handle, - struct inode *inode, - struct inode *dir, - struct buffer_head *di_bh, - struct buffer_head *dir_bh, - struct ocfs2_alloc_context *meta_ac, - struct ocfs2_alloc_context *data_ac) -{ - return 0; -} - -#endif /* CONFIG_OCFS2_FS_POSIX_ACL*/ +struct posix_acl *ocfs2_iop_get_acl(struct inode *inode, int type); +int ocfs2_iop_set_acl(struct inode *inode, struct posix_acl *acl, int type); +int ocfs2_set_acl(handle_t *handle, + struct inode *inode, + struct buffer_head *di_bh, + int type, + struct posix_acl *acl, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_alloc_context *data_ac); #endif /* OCFS2_ACL_H */ diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c index 678a067d925..9d8fcf2f3b9 100644 --- a/fs/ocfs2/alloc.c +++ b/fs/ocfs2/alloc.c @@ -29,8 +29,8 @@ #include <linux/highmem.h> #include <linux/swap.h> #include <linux/quotaops.h> +#include <linux/blkdev.h> -#define MLOG_MASK_PREFIX ML_DISK_ALLOC #include <cluster/masklog.h> #include "ocfs2.h" @@ -49,10 +49,22 @@ #include "super.h" #include "uptodate.h" #include "xattr.h" +#include "refcounttree.h" +#include "ocfs2_trace.h" #include "buffer_head_io.h" +enum ocfs2_contig_type { + CONTIG_NONE = 0, + CONTIG_LEFT, + CONTIG_RIGHT, + CONTIG_LEFTRIGHT, +}; +static enum ocfs2_contig_type + ocfs2_extent_rec_contig(struct super_block *sb, + struct ocfs2_extent_rec *ext, + struct ocfs2_extent_rec *insert_rec); /* * Operations for a specific extent tree type. * @@ -79,18 +91,30 @@ struct ocfs2_extent_tree_operations { * that value. new_clusters is the delta, and must be * added to the total. Required. */ - void (*eo_update_clusters)(struct inode *inode, - struct ocfs2_extent_tree *et, + void (*eo_update_clusters)(struct ocfs2_extent_tree *et, u32 new_clusters); /* + * If this extent tree is supported by an extent map, insert + * a record into the map. + */ + void (*eo_extent_map_insert)(struct ocfs2_extent_tree *et, + struct ocfs2_extent_rec *rec); + + /* + * If this extent tree is supported by an extent map, truncate the + * map to clusters, + */ + void (*eo_extent_map_truncate)(struct ocfs2_extent_tree *et, + u32 clusters); + + /* * If ->eo_insert_check() exists, it is called before rec is * inserted into the extent tree. It is optional. */ - int (*eo_insert_check)(struct inode *inode, - struct ocfs2_extent_tree *et, + int (*eo_insert_check)(struct ocfs2_extent_tree *et, struct ocfs2_extent_rec *rec); - int (*eo_sanity_check)(struct inode *inode, struct ocfs2_extent_tree *et); + int (*eo_sanity_check)(struct ocfs2_extent_tree *et); /* * -------------------------------------------------------------- @@ -109,8 +133,17 @@ struct ocfs2_extent_tree_operations { * it exists. If it does not, et->et_max_leaf_clusters is set * to 0 (unlimited). Optional. */ - void (*eo_fill_max_leaf_clusters)(struct inode *inode, - struct ocfs2_extent_tree *et); + void (*eo_fill_max_leaf_clusters)(struct ocfs2_extent_tree *et); + + /* + * ->eo_extent_contig test whether the 2 ocfs2_extent_rec + * are contiguous or not. Optional. Don't need to set it if use + * ocfs2_extent_rec as the tree leaf. + */ + enum ocfs2_contig_type + (*eo_extent_contig)(struct ocfs2_extent_tree *et, + struct ocfs2_extent_rec *ext, + struct ocfs2_extent_rec *insert_rec); }; @@ -121,19 +154,22 @@ struct ocfs2_extent_tree_operations { static u64 ocfs2_dinode_get_last_eb_blk(struct ocfs2_extent_tree *et); static void ocfs2_dinode_set_last_eb_blk(struct ocfs2_extent_tree *et, u64 blkno); -static void ocfs2_dinode_update_clusters(struct inode *inode, - struct ocfs2_extent_tree *et, +static void ocfs2_dinode_update_clusters(struct ocfs2_extent_tree *et, u32 clusters); -static int ocfs2_dinode_insert_check(struct inode *inode, - struct ocfs2_extent_tree *et, +static void ocfs2_dinode_extent_map_insert(struct ocfs2_extent_tree *et, + struct ocfs2_extent_rec *rec); +static void ocfs2_dinode_extent_map_truncate(struct ocfs2_extent_tree *et, + u32 clusters); +static int ocfs2_dinode_insert_check(struct ocfs2_extent_tree *et, struct ocfs2_extent_rec *rec); -static int ocfs2_dinode_sanity_check(struct inode *inode, - struct ocfs2_extent_tree *et); +static int ocfs2_dinode_sanity_check(struct ocfs2_extent_tree *et); static void ocfs2_dinode_fill_root_el(struct ocfs2_extent_tree *et); static struct ocfs2_extent_tree_operations ocfs2_dinode_et_ops = { .eo_set_last_eb_blk = ocfs2_dinode_set_last_eb_blk, .eo_get_last_eb_blk = ocfs2_dinode_get_last_eb_blk, .eo_update_clusters = ocfs2_dinode_update_clusters, + .eo_extent_map_insert = ocfs2_dinode_extent_map_insert, + .eo_extent_map_truncate = ocfs2_dinode_extent_map_truncate, .eo_insert_check = ocfs2_dinode_insert_check, .eo_sanity_check = ocfs2_dinode_sanity_check, .eo_fill_root_el = ocfs2_dinode_fill_root_el, @@ -156,40 +192,53 @@ static u64 ocfs2_dinode_get_last_eb_blk(struct ocfs2_extent_tree *et) return le64_to_cpu(di->i_last_eb_blk); } -static void ocfs2_dinode_update_clusters(struct inode *inode, - struct ocfs2_extent_tree *et, +static void ocfs2_dinode_update_clusters(struct ocfs2_extent_tree *et, u32 clusters) { + struct ocfs2_inode_info *oi = cache_info_to_inode(et->et_ci); struct ocfs2_dinode *di = et->et_object; le32_add_cpu(&di->i_clusters, clusters); - spin_lock(&OCFS2_I(inode)->ip_lock); - OCFS2_I(inode)->ip_clusters = le32_to_cpu(di->i_clusters); - spin_unlock(&OCFS2_I(inode)->ip_lock); + spin_lock(&oi->ip_lock); + oi->ip_clusters = le32_to_cpu(di->i_clusters); + spin_unlock(&oi->ip_lock); } -static int ocfs2_dinode_insert_check(struct inode *inode, - struct ocfs2_extent_tree *et, +static void ocfs2_dinode_extent_map_insert(struct ocfs2_extent_tree *et, + struct ocfs2_extent_rec *rec) +{ + struct inode *inode = &cache_info_to_inode(et->et_ci)->vfs_inode; + + ocfs2_extent_map_insert_rec(inode, rec); +} + +static void ocfs2_dinode_extent_map_truncate(struct ocfs2_extent_tree *et, + u32 clusters) +{ + struct inode *inode = &cache_info_to_inode(et->et_ci)->vfs_inode; + + ocfs2_extent_map_trunc(inode, clusters); +} + +static int ocfs2_dinode_insert_check(struct ocfs2_extent_tree *et, struct ocfs2_extent_rec *rec) { - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_inode_info *oi = cache_info_to_inode(et->et_ci); + struct ocfs2_super *osb = OCFS2_SB(oi->vfs_inode.i_sb); - BUG_ON(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL); + BUG_ON(oi->ip_dyn_features & OCFS2_INLINE_DATA_FL); mlog_bug_on_msg(!ocfs2_sparse_alloc(osb) && - (OCFS2_I(inode)->ip_clusters != - le32_to_cpu(rec->e_cpos)), + (oi->ip_clusters != le32_to_cpu(rec->e_cpos)), "Device %s, asking for sparse allocation: inode %llu, " "cpos %u, clusters %u\n", osb->dev_str, - (unsigned long long)OCFS2_I(inode)->ip_blkno, - rec->e_cpos, - OCFS2_I(inode)->ip_clusters); + (unsigned long long)oi->ip_blkno, + rec->e_cpos, oi->ip_clusters); return 0; } -static int ocfs2_dinode_sanity_check(struct inode *inode, - struct ocfs2_extent_tree *et) +static int ocfs2_dinode_sanity_check(struct ocfs2_extent_tree *et) { struct ocfs2_dinode *di = et->et_object; @@ -229,8 +278,7 @@ static u64 ocfs2_xattr_value_get_last_eb_blk(struct ocfs2_extent_tree *et) return le64_to_cpu(vb->vb_xv->xr_last_eb_blk); } -static void ocfs2_xattr_value_update_clusters(struct inode *inode, - struct ocfs2_extent_tree *et, +static void ocfs2_xattr_value_update_clusters(struct ocfs2_extent_tree *et, u32 clusters) { struct ocfs2_xattr_value_buf *vb = et->et_object; @@ -252,12 +300,11 @@ static void ocfs2_xattr_tree_fill_root_el(struct ocfs2_extent_tree *et) et->et_root_el = &xb->xb_attrs.xb_root.xt_list; } -static void ocfs2_xattr_tree_fill_max_leaf_clusters(struct inode *inode, - struct ocfs2_extent_tree *et) +static void ocfs2_xattr_tree_fill_max_leaf_clusters(struct ocfs2_extent_tree *et) { + struct super_block *sb = ocfs2_metadata_cache_get_super(et->et_ci); et->et_max_leaf_clusters = - ocfs2_clusters_for_bytes(inode->i_sb, - OCFS2_MAX_XATTR_TREE_LEAF_SIZE); + ocfs2_clusters_for_bytes(sb, OCFS2_MAX_XATTR_TREE_LEAF_SIZE); } static void ocfs2_xattr_tree_set_last_eb_blk(struct ocfs2_extent_tree *et, @@ -277,8 +324,7 @@ static u64 ocfs2_xattr_tree_get_last_eb_blk(struct ocfs2_extent_tree *et) return le64_to_cpu(xt->xt_last_eb_blk); } -static void ocfs2_xattr_tree_update_clusters(struct inode *inode, - struct ocfs2_extent_tree *et, +static void ocfs2_xattr_tree_update_clusters(struct ocfs2_extent_tree *et, u32 clusters) { struct ocfs2_xattr_block *xb = et->et_object; @@ -309,8 +355,7 @@ static u64 ocfs2_dx_root_get_last_eb_blk(struct ocfs2_extent_tree *et) return le64_to_cpu(dx_root->dr_last_eb_blk); } -static void ocfs2_dx_root_update_clusters(struct inode *inode, - struct ocfs2_extent_tree *et, +static void ocfs2_dx_root_update_clusters(struct ocfs2_extent_tree *et, u32 clusters) { struct ocfs2_dx_root_block *dx_root = et->et_object; @@ -318,8 +363,7 @@ static void ocfs2_dx_root_update_clusters(struct inode *inode, le32_add_cpu(&dx_root->dr_clusters, clusters); } -static int ocfs2_dx_root_sanity_check(struct inode *inode, - struct ocfs2_extent_tree *et) +static int ocfs2_dx_root_sanity_check(struct ocfs2_extent_tree *et) { struct ocfs2_dx_root_block *dx_root = et->et_object; @@ -343,8 +387,54 @@ static struct ocfs2_extent_tree_operations ocfs2_dx_root_et_ops = { .eo_fill_root_el = ocfs2_dx_root_fill_root_el, }; +static void ocfs2_refcount_tree_fill_root_el(struct ocfs2_extent_tree *et) +{ + struct ocfs2_refcount_block *rb = et->et_object; + + et->et_root_el = &rb->rf_list; +} + +static void ocfs2_refcount_tree_set_last_eb_blk(struct ocfs2_extent_tree *et, + u64 blkno) +{ + struct ocfs2_refcount_block *rb = et->et_object; + + rb->rf_last_eb_blk = cpu_to_le64(blkno); +} + +static u64 ocfs2_refcount_tree_get_last_eb_blk(struct ocfs2_extent_tree *et) +{ + struct ocfs2_refcount_block *rb = et->et_object; + + return le64_to_cpu(rb->rf_last_eb_blk); +} + +static void ocfs2_refcount_tree_update_clusters(struct ocfs2_extent_tree *et, + u32 clusters) +{ + struct ocfs2_refcount_block *rb = et->et_object; + + le32_add_cpu(&rb->rf_clusters, clusters); +} + +static enum ocfs2_contig_type +ocfs2_refcount_tree_extent_contig(struct ocfs2_extent_tree *et, + struct ocfs2_extent_rec *ext, + struct ocfs2_extent_rec *insert_rec) +{ + return CONTIG_NONE; +} + +static struct ocfs2_extent_tree_operations ocfs2_refcount_tree_et_ops = { + .eo_set_last_eb_blk = ocfs2_refcount_tree_set_last_eb_blk, + .eo_get_last_eb_blk = ocfs2_refcount_tree_get_last_eb_blk, + .eo_update_clusters = ocfs2_refcount_tree_update_clusters, + .eo_fill_root_el = ocfs2_refcount_tree_fill_root_el, + .eo_extent_contig = ocfs2_refcount_tree_extent_contig, +}; + static void __ocfs2_init_extent_tree(struct ocfs2_extent_tree *et, - struct inode *inode, + struct ocfs2_caching_info *ci, struct buffer_head *bh, ocfs2_journal_access_func access, void *obj, @@ -352,6 +442,7 @@ static void __ocfs2_init_extent_tree(struct ocfs2_extent_tree *et, { et->et_ops = ops; et->et_root_bh = bh; + et->et_ci = ci; et->et_root_journal_access = access; if (!obj) obj = (void *)bh->b_data; @@ -361,41 +452,49 @@ static void __ocfs2_init_extent_tree(struct ocfs2_extent_tree *et, if (!et->et_ops->eo_fill_max_leaf_clusters) et->et_max_leaf_clusters = 0; else - et->et_ops->eo_fill_max_leaf_clusters(inode, et); + et->et_ops->eo_fill_max_leaf_clusters(et); } void ocfs2_init_dinode_extent_tree(struct ocfs2_extent_tree *et, - struct inode *inode, + struct ocfs2_caching_info *ci, struct buffer_head *bh) { - __ocfs2_init_extent_tree(et, inode, bh, ocfs2_journal_access_di, + __ocfs2_init_extent_tree(et, ci, bh, ocfs2_journal_access_di, NULL, &ocfs2_dinode_et_ops); } void ocfs2_init_xattr_tree_extent_tree(struct ocfs2_extent_tree *et, - struct inode *inode, + struct ocfs2_caching_info *ci, struct buffer_head *bh) { - __ocfs2_init_extent_tree(et, inode, bh, ocfs2_journal_access_xb, + __ocfs2_init_extent_tree(et, ci, bh, ocfs2_journal_access_xb, NULL, &ocfs2_xattr_tree_et_ops); } void ocfs2_init_xattr_value_extent_tree(struct ocfs2_extent_tree *et, - struct inode *inode, + struct ocfs2_caching_info *ci, struct ocfs2_xattr_value_buf *vb) { - __ocfs2_init_extent_tree(et, inode, vb->vb_bh, vb->vb_access, vb, + __ocfs2_init_extent_tree(et, ci, vb->vb_bh, vb->vb_access, vb, &ocfs2_xattr_value_et_ops); } void ocfs2_init_dx_root_extent_tree(struct ocfs2_extent_tree *et, - struct inode *inode, + struct ocfs2_caching_info *ci, struct buffer_head *bh) { - __ocfs2_init_extent_tree(et, inode, bh, ocfs2_journal_access_dr, + __ocfs2_init_extent_tree(et, ci, bh, ocfs2_journal_access_dr, NULL, &ocfs2_dx_root_et_ops); } +void ocfs2_init_refcount_extent_tree(struct ocfs2_extent_tree *et, + struct ocfs2_caching_info *ci, + struct buffer_head *bh) +{ + __ocfs2_init_extent_tree(et, ci, bh, ocfs2_journal_access_rb, + NULL, &ocfs2_refcount_tree_et_ops); +} + static inline void ocfs2_et_set_last_eb_blk(struct ocfs2_extent_tree *et, u64 new_last_eb_blk) { @@ -407,80 +506,78 @@ static inline u64 ocfs2_et_get_last_eb_blk(struct ocfs2_extent_tree *et) return et->et_ops->eo_get_last_eb_blk(et); } -static inline void ocfs2_et_update_clusters(struct inode *inode, - struct ocfs2_extent_tree *et, +static inline void ocfs2_et_update_clusters(struct ocfs2_extent_tree *et, u32 clusters) { - et->et_ops->eo_update_clusters(inode, et, clusters); + et->et_ops->eo_update_clusters(et, clusters); +} + +static inline void ocfs2_et_extent_map_insert(struct ocfs2_extent_tree *et, + struct ocfs2_extent_rec *rec) +{ + if (et->et_ops->eo_extent_map_insert) + et->et_ops->eo_extent_map_insert(et, rec); +} + +static inline void ocfs2_et_extent_map_truncate(struct ocfs2_extent_tree *et, + u32 clusters) +{ + if (et->et_ops->eo_extent_map_truncate) + et->et_ops->eo_extent_map_truncate(et, clusters); } static inline int ocfs2_et_root_journal_access(handle_t *handle, - struct inode *inode, struct ocfs2_extent_tree *et, int type) { - return et->et_root_journal_access(handle, inode, et->et_root_bh, + return et->et_root_journal_access(handle, et->et_ci, et->et_root_bh, type); } -static inline int ocfs2_et_insert_check(struct inode *inode, - struct ocfs2_extent_tree *et, +static inline enum ocfs2_contig_type + ocfs2_et_extent_contig(struct ocfs2_extent_tree *et, + struct ocfs2_extent_rec *rec, + struct ocfs2_extent_rec *insert_rec) +{ + if (et->et_ops->eo_extent_contig) + return et->et_ops->eo_extent_contig(et, rec, insert_rec); + + return ocfs2_extent_rec_contig( + ocfs2_metadata_cache_get_super(et->et_ci), + rec, insert_rec); +} + +static inline int ocfs2_et_insert_check(struct ocfs2_extent_tree *et, struct ocfs2_extent_rec *rec) { int ret = 0; if (et->et_ops->eo_insert_check) - ret = et->et_ops->eo_insert_check(inode, et, rec); + ret = et->et_ops->eo_insert_check(et, rec); return ret; } -static inline int ocfs2_et_sanity_check(struct inode *inode, - struct ocfs2_extent_tree *et) +static inline int ocfs2_et_sanity_check(struct ocfs2_extent_tree *et) { int ret = 0; if (et->et_ops->eo_sanity_check) - ret = et->et_ops->eo_sanity_check(inode, et); + ret = et->et_ops->eo_sanity_check(et); return ret; } -static void ocfs2_free_truncate_context(struct ocfs2_truncate_context *tc); static int ocfs2_cache_extent_block_free(struct ocfs2_cached_dealloc_ctxt *ctxt, struct ocfs2_extent_block *eb); - -/* - * Structures which describe a path through a btree, and functions to - * manipulate them. - * - * The idea here is to be as generic as possible with the tree - * manipulation code. - */ -struct ocfs2_path_item { - struct buffer_head *bh; - struct ocfs2_extent_list *el; -}; - -#define OCFS2_MAX_PATH_DEPTH 5 - -struct ocfs2_path { - int p_tree_depth; - ocfs2_journal_access_func p_root_access; - struct ocfs2_path_item p_node[OCFS2_MAX_PATH_DEPTH]; -}; - -#define path_root_bh(_path) ((_path)->p_node[0].bh) -#define path_root_el(_path) ((_path)->p_node[0].el) -#define path_root_access(_path)((_path)->p_root_access) -#define path_leaf_bh(_path) ((_path)->p_node[(_path)->p_tree_depth].bh) -#define path_leaf_el(_path) ((_path)->p_node[(_path)->p_tree_depth].el) -#define path_num_items(_path) ((_path)->p_tree_depth + 1) - +static void ocfs2_adjust_rightmost_records(handle_t *handle, + struct ocfs2_extent_tree *et, + struct ocfs2_path *path, + struct ocfs2_extent_rec *insert_rec); /* * Reset the actual path elements so that we can re-use the structure * to build another path. Generally, this involves freeing the buffer * heads. */ -static void ocfs2_reinit_path(struct ocfs2_path *path, int keep_root) +void ocfs2_reinit_path(struct ocfs2_path *path, int keep_root) { int i, start = 0, depth = 0; struct ocfs2_path_item *node; @@ -509,7 +606,7 @@ static void ocfs2_reinit_path(struct ocfs2_path *path, int keep_root) path->p_tree_depth = depth; } -static void ocfs2_free_path(struct ocfs2_path *path) +void ocfs2_free_path(struct ocfs2_path *path) { if (path) { ocfs2_reinit_path(path, 0); @@ -607,13 +704,13 @@ static struct ocfs2_path *ocfs2_new_path(struct buffer_head *root_bh, return path; } -static struct ocfs2_path *ocfs2_new_path_from_path(struct ocfs2_path *path) +struct ocfs2_path *ocfs2_new_path_from_path(struct ocfs2_path *path) { return ocfs2_new_path(path_root_bh(path), path_root_el(path), path_root_access(path)); } -static struct ocfs2_path *ocfs2_new_path_from_et(struct ocfs2_extent_tree *et) +struct ocfs2_path *ocfs2_new_path_from_et(struct ocfs2_extent_tree *et) { return ocfs2_new_path(et->et_root_bh, et->et_root_el, et->et_root_journal_access); @@ -626,10 +723,10 @@ static struct ocfs2_path *ocfs2_new_path_from_et(struct ocfs2_extent_tree *et) * I don't like the way this function's name looks next to * ocfs2_journal_access_path(), but I don't have a better one. */ -static int ocfs2_path_bh_journal_access(handle_t *handle, - struct inode *inode, - struct ocfs2_path *path, - int idx) +int ocfs2_path_bh_journal_access(handle_t *handle, + struct ocfs2_caching_info *ci, + struct ocfs2_path *path, + int idx) { ocfs2_journal_access_func access = path_root_access(path); @@ -639,15 +736,16 @@ static int ocfs2_path_bh_journal_access(handle_t *handle, if (idx) access = ocfs2_journal_access_eb; - return access(handle, inode, path->p_node[idx].bh, + return access(handle, ci, path->p_node[idx].bh, OCFS2_JOURNAL_ACCESS_WRITE); } /* * Convenience function to journal all components in a path. */ -static int ocfs2_journal_access_path(struct inode *inode, handle_t *handle, - struct ocfs2_path *path) +int ocfs2_journal_access_path(struct ocfs2_caching_info *ci, + handle_t *handle, + struct ocfs2_path *path) { int i, ret = 0; @@ -655,7 +753,7 @@ static int ocfs2_journal_access_path(struct inode *inode, handle_t *handle, goto out; for(i = 0; i < path_num_items(path); i++) { - ret = ocfs2_path_bh_journal_access(handle, inode, path, i); + ret = ocfs2_path_bh_journal_access(handle, ci, path, i); if (ret < 0) { mlog_errno(ret); goto out; @@ -696,17 +794,9 @@ int ocfs2_search_extent_list(struct ocfs2_extent_list *el, u32 v_cluster) return ret; } -enum ocfs2_contig_type { - CONTIG_NONE = 0, - CONTIG_LEFT, - CONTIG_RIGHT, - CONTIG_LEFTRIGHT, -}; - - /* * NOTE: ocfs2_block_extent_contig(), ocfs2_extents_adjacent() and - * ocfs2_extent_contig only work properly against leaf nodes! + * ocfs2_extent_rec_contig only work properly against leaf nodes! */ static int ocfs2_block_extent_contig(struct super_block *sb, struct ocfs2_extent_rec *ext, @@ -732,9 +822,9 @@ static int ocfs2_extents_adjacent(struct ocfs2_extent_rec *left, } static enum ocfs2_contig_type - ocfs2_extent_contig(struct inode *inode, - struct ocfs2_extent_rec *ext, - struct ocfs2_extent_rec *insert_rec) + ocfs2_extent_rec_contig(struct super_block *sb, + struct ocfs2_extent_rec *ext, + struct ocfs2_extent_rec *insert_rec) { u64 blkno = le64_to_cpu(insert_rec->e_blkno); @@ -747,12 +837,12 @@ static enum ocfs2_contig_type return CONTIG_NONE; if (ocfs2_extents_adjacent(ext, insert_rec) && - ocfs2_block_extent_contig(inode->i_sb, ext, blkno)) + ocfs2_block_extent_contig(sb, ext, blkno)) return CONTIG_RIGHT; blkno = le64_to_cpu(ext->e_blkno); if (ocfs2_extents_adjacent(insert_rec, ext) && - ocfs2_block_extent_contig(inode->i_sb, insert_rec, blkno)) + ocfs2_block_extent_contig(sb, insert_rec, blkno)) return CONTIG_LEFT; return CONTIG_NONE; @@ -797,8 +887,7 @@ static int ocfs2_validate_extent_block(struct super_block *sb, struct ocfs2_extent_block *eb = (struct ocfs2_extent_block *)bh->b_data; - mlog(0, "Validating extent block %llu\n", - (unsigned long long)bh->b_blocknr); + trace_ocfs2_validate_extent_block((unsigned long long)bh->b_blocknr); BUG_ON(!buffer_uptodate(bh)); @@ -847,13 +936,13 @@ static int ocfs2_validate_extent_block(struct super_block *sb, return 0; } -int ocfs2_read_extent_block(struct inode *inode, u64 eb_blkno, +int ocfs2_read_extent_block(struct ocfs2_caching_info *ci, u64 eb_blkno, struct buffer_head **bh) { int rc; struct buffer_head *tmp = *bh; - rc = ocfs2_read_block(inode, eb_blkno, &tmp, + rc = ocfs2_read_block(ci, eb_blkno, &tmp, ocfs2_validate_extent_block); /* If ocfs2_read_block() got us a new bh, pass it up. */ @@ -868,7 +957,6 @@ int ocfs2_read_extent_block(struct inode *inode, u64 eb_blkno, * How many free extents have we got before we need more meta data? */ int ocfs2_num_free_extents(struct ocfs2_super *osb, - struct inode *inode, struct ocfs2_extent_tree *et) { int retval; @@ -877,13 +965,12 @@ int ocfs2_num_free_extents(struct ocfs2_super *osb, struct buffer_head *eb_bh = NULL; u64 last_eb_blk = 0; - mlog_entry_void(); - el = et->et_root_el; last_eb_blk = ocfs2_et_get_last_eb_blk(et); if (last_eb_blk) { - retval = ocfs2_read_extent_block(inode, last_eb_blk, &eb_bh); + retval = ocfs2_read_extent_block(et->et_ci, last_eb_blk, + &eb_bh); if (retval < 0) { mlog_errno(retval); goto bail; @@ -898,7 +985,7 @@ int ocfs2_num_free_extents(struct ocfs2_super *osb, bail: brelse(eb_bh); - mlog_exit(retval); + trace_ocfs2_num_free_extents(retval); return retval; } @@ -907,9 +994,8 @@ bail: * sets h_signature, h_blkno, h_suballoc_bit, h_suballoc_slot, and * l_count for you */ -static int ocfs2_create_new_meta_bhs(struct ocfs2_super *osb, - handle_t *handle, - struct inode *inode, +static int ocfs2_create_new_meta_bhs(handle_t *handle, + struct ocfs2_extent_tree *et, int wanted, struct ocfs2_alloc_context *meta_ac, struct buffer_head *bhs[]) @@ -917,17 +1003,17 @@ static int ocfs2_create_new_meta_bhs(struct ocfs2_super *osb, int count, status, i; u16 suballoc_bit_start; u32 num_got; - u64 first_blkno; + u64 suballoc_loc, first_blkno; + struct ocfs2_super *osb = + OCFS2_SB(ocfs2_metadata_cache_get_super(et->et_ci)); struct ocfs2_extent_block *eb; - mlog_entry_void(); - count = 0; while (count < wanted) { - status = ocfs2_claim_metadata(osb, - handle, + status = ocfs2_claim_metadata(handle, meta_ac, wanted - count, + &suballoc_loc, &suballoc_bit_start, &num_got, &first_blkno); @@ -939,13 +1025,14 @@ static int ocfs2_create_new_meta_bhs(struct ocfs2_super *osb, for(i = count; i < (num_got + count); i++) { bhs[i] = sb_getblk(osb->sb, first_blkno); if (bhs[i] == NULL) { - status = -EIO; + status = -ENOMEM; mlog_errno(status); goto bail; } - ocfs2_set_new_buffer_uptodate(inode, bhs[i]); + ocfs2_set_new_buffer_uptodate(et->et_ci, bhs[i]); - status = ocfs2_journal_access_eb(handle, inode, bhs[i], + status = ocfs2_journal_access_eb(handle, et->et_ci, + bhs[i], OCFS2_JOURNAL_ACCESS_CREATE); if (status < 0) { mlog_errno(status); @@ -958,7 +1045,9 @@ static int ocfs2_create_new_meta_bhs(struct ocfs2_super *osb, strcpy(eb->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE); eb->h_blkno = cpu_to_le64(first_blkno); eb->h_fs_generation = cpu_to_le32(osb->fs_generation); - eb->h_suballoc_slot = cpu_to_le16(osb->slot_num); + eb->h_suballoc_slot = + cpu_to_le16(meta_ac->ac_alloc_slot); + eb->h_suballoc_loc = cpu_to_le64(suballoc_loc); eb->h_suballoc_bit = cpu_to_le16(suballoc_bit_start); eb->h_list.l_count = cpu_to_le16(ocfs2_extent_recs_per_eb(osb->sb)); @@ -968,11 +1057,7 @@ static int ocfs2_create_new_meta_bhs(struct ocfs2_super *osb, /* We'll also be dirtied by the caller, so * this isn't absolutely necessary. */ - status = ocfs2_journal_dirty(handle, bhs[i]); - if (status < 0) { - mlog_errno(status); - goto bail; - } + ocfs2_journal_dirty(handle, bhs[i]); } count += num_got; @@ -985,8 +1070,8 @@ bail: brelse(bhs[i]); bhs[i] = NULL; } + mlog_errno(status); } - mlog_exit(status); return status; } @@ -1013,8 +1098,54 @@ static inline u32 ocfs2_sum_rightmost_rec(struct ocfs2_extent_list *el) } /* + * Change range of the branches in the right most path according to the leaf + * extent block's rightmost record. + */ +static int ocfs2_adjust_rightmost_branch(handle_t *handle, + struct ocfs2_extent_tree *et) +{ + int status; + struct ocfs2_path *path = NULL; + struct ocfs2_extent_list *el; + struct ocfs2_extent_rec *rec; + + path = ocfs2_new_path_from_et(et); + if (!path) { + status = -ENOMEM; + return status; + } + + status = ocfs2_find_path(et->et_ci, path, UINT_MAX); + if (status < 0) { + mlog_errno(status); + goto out; + } + + status = ocfs2_extend_trans(handle, path_num_items(path)); + if (status < 0) { + mlog_errno(status); + goto out; + } + + status = ocfs2_journal_access_path(et->et_ci, handle, path); + if (status < 0) { + mlog_errno(status); + goto out; + } + + el = path_leaf_el(path); + rec = &el->l_recs[le16_to_cpu(el->l_next_free_rec) - 1]; + + ocfs2_adjust_rightmost_records(handle, et, path, rec); + +out: + ocfs2_free_path(path); + return status; +} + +/* * Add an entire tree branch to our inode. eb_bh is the extent block - * to start at, if we don't want to start the branch at the dinode + * to start at, if we don't want to start the branch at the root * structure. * * last_eb_bh is required as we have to update it's next_leaf pointer @@ -1023,9 +1154,7 @@ static inline u32 ocfs2_sum_rightmost_rec(struct ocfs2_extent_list *el) * the new branch will be 'empty' in the sense that every block will * contain a single record with cluster count == 0. */ -static int ocfs2_add_branch(struct ocfs2_super *osb, - handle_t *handle, - struct inode *inode, +static int ocfs2_add_branch(handle_t *handle, struct ocfs2_extent_tree *et, struct buffer_head *eb_bh, struct buffer_head **last_eb_bh, @@ -1038,9 +1167,7 @@ static int ocfs2_add_branch(struct ocfs2_super *osb, struct ocfs2_extent_block *eb; struct ocfs2_extent_list *eb_el; struct ocfs2_extent_list *el; - u32 new_cpos; - - mlog_entry_void(); + u32 new_cpos, root_end; BUG_ON(!last_eb_bh || !*last_eb_bh); @@ -1055,6 +1182,30 @@ static int ocfs2_add_branch(struct ocfs2_super *osb, new_blocks = le16_to_cpu(el->l_tree_depth); + eb = (struct ocfs2_extent_block *)(*last_eb_bh)->b_data; + new_cpos = ocfs2_sum_rightmost_rec(&eb->h_list); + root_end = ocfs2_sum_rightmost_rec(et->et_root_el); + + /* + * If there is a gap before the root end and the real end + * of the righmost leaf block, we need to remove the gap + * between new_cpos and root_end first so that the tree + * is consistent after we add a new branch(it will start + * from new_cpos). + */ + if (root_end > new_cpos) { + trace_ocfs2_adjust_rightmost_branch( + (unsigned long long) + ocfs2_metadata_cache_owner(et->et_ci), + root_end, new_cpos); + + status = ocfs2_adjust_rightmost_branch(handle, et); + if (status) { + mlog_errno(status); + goto bail; + } + } + /* allocate the number of new eb blocks we need */ new_eb_bhs = kcalloc(new_blocks, sizeof(struct buffer_head *), GFP_KERNEL); @@ -1064,16 +1215,13 @@ static int ocfs2_add_branch(struct ocfs2_super *osb, goto bail; } - status = ocfs2_create_new_meta_bhs(osb, handle, inode, new_blocks, + status = ocfs2_create_new_meta_bhs(handle, et, new_blocks, meta_ac, new_eb_bhs); if (status < 0) { mlog_errno(status); goto bail; } - eb = (struct ocfs2_extent_block *)(*last_eb_bh)->b_data; - new_cpos = ocfs2_sum_rightmost_rec(&eb->h_list); - /* Note: new_eb_bhs[new_blocks - 1] is the guy which will be * linked with the rest of the tree. * conversly, new_eb_bhs[0] is the new bottommost leaf. @@ -1089,7 +1237,7 @@ static int ocfs2_add_branch(struct ocfs2_super *osb, BUG_ON(!OCFS2_IS_VALID_EXTENT_BLOCK(eb)); eb_el = &eb->h_list; - status = ocfs2_journal_access_eb(handle, inode, bh, + status = ocfs2_journal_access_eb(handle, et->et_ci, bh, OCFS2_JOURNAL_ACCESS_CREATE); if (status < 0) { mlog_errno(status); @@ -1114,12 +1262,7 @@ static int ocfs2_add_branch(struct ocfs2_super *osb, if (!eb_el->l_tree_depth) new_last_eb_blk = le64_to_cpu(eb->h_blkno); - status = ocfs2_journal_dirty(handle, bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } - + ocfs2_journal_dirty(handle, bh); next_blkno = le64_to_cpu(eb->h_blkno); } @@ -1129,20 +1272,20 @@ static int ocfs2_add_branch(struct ocfs2_super *osb, * journal_dirty erroring as it won't unless we've aborted the * handle (in which case we would never be here) so reserving * the write with journal_access is all we need to do. */ - status = ocfs2_journal_access_eb(handle, inode, *last_eb_bh, + status = ocfs2_journal_access_eb(handle, et->et_ci, *last_eb_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); goto bail; } - status = ocfs2_et_root_journal_access(handle, inode, et, + status = ocfs2_et_root_journal_access(handle, et, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); goto bail; } if (eb_bh) { - status = ocfs2_journal_access_eb(handle, inode, eb_bh, + status = ocfs2_journal_access_eb(handle, et->et_ci, eb_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); @@ -1165,17 +1308,10 @@ static int ocfs2_add_branch(struct ocfs2_super *osb, eb = (struct ocfs2_extent_block *) (*last_eb_bh)->b_data; eb->h_next_leaf_blk = cpu_to_le64(new_last_eb_blk); - status = ocfs2_journal_dirty(handle, *last_eb_bh); - if (status < 0) - mlog_errno(status); - status = ocfs2_journal_dirty(handle, et->et_root_bh); - if (status < 0) - mlog_errno(status); - if (eb_bh) { - status = ocfs2_journal_dirty(handle, eb_bh); - if (status < 0) - mlog_errno(status); - } + ocfs2_journal_dirty(handle, *last_eb_bh); + ocfs2_journal_dirty(handle, et->et_root_bh); + if (eb_bh) + ocfs2_journal_dirty(handle, eb_bh); /* * Some callers want to track the rightmost leaf so pass it @@ -1193,7 +1329,6 @@ bail: kfree(new_eb_bhs); } - mlog_exit(status); return status; } @@ -1202,9 +1337,7 @@ bail: * returns back the new extent block so you can add a branch to it * after this call. */ -static int ocfs2_shift_tree_depth(struct ocfs2_super *osb, - handle_t *handle, - struct inode *inode, +static int ocfs2_shift_tree_depth(handle_t *handle, struct ocfs2_extent_tree *et, struct ocfs2_alloc_context *meta_ac, struct buffer_head **ret_new_eb_bh) @@ -1216,9 +1349,7 @@ static int ocfs2_shift_tree_depth(struct ocfs2_super *osb, struct ocfs2_extent_list *root_el; struct ocfs2_extent_list *eb_el; - mlog_entry_void(); - - status = ocfs2_create_new_meta_bhs(osb, handle, inode, 1, meta_ac, + status = ocfs2_create_new_meta_bhs(handle, et, 1, meta_ac, &new_eb_bh); if (status < 0) { mlog_errno(status); @@ -1232,7 +1363,7 @@ static int ocfs2_shift_tree_depth(struct ocfs2_super *osb, eb_el = &eb->h_list; root_el = et->et_root_el; - status = ocfs2_journal_access_eb(handle, inode, new_eb_bh, + status = ocfs2_journal_access_eb(handle, et->et_ci, new_eb_bh, OCFS2_JOURNAL_ACCESS_CREATE); if (status < 0) { mlog_errno(status); @@ -1245,13 +1376,9 @@ static int ocfs2_shift_tree_depth(struct ocfs2_super *osb, for (i = 0; i < le16_to_cpu(root_el->l_next_free_rec); i++) eb_el->l_recs[i] = root_el->l_recs[i]; - status = ocfs2_journal_dirty(handle, new_eb_bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } + ocfs2_journal_dirty(handle, new_eb_bh); - status = ocfs2_et_root_journal_access(handle, inode, et, + status = ocfs2_et_root_journal_access(handle, et, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); @@ -1274,11 +1401,7 @@ static int ocfs2_shift_tree_depth(struct ocfs2_super *osb, if (root_el->l_tree_depth == cpu_to_le16(1)) ocfs2_et_set_last_eb_blk(et, le64_to_cpu(eb->h_blkno)); - status = ocfs2_journal_dirty(handle, et->et_root_bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } + ocfs2_journal_dirty(handle, et->et_root_bh); *ret_new_eb_bh = new_eb_bh; new_eb_bh = NULL; @@ -1286,7 +1409,6 @@ static int ocfs2_shift_tree_depth(struct ocfs2_super *osb, bail: brelse(new_eb_bh); - mlog_exit(status); return status; } @@ -1307,9 +1429,7 @@ bail: * * return status < 0 indicates an error. */ -static int ocfs2_find_branch_target(struct ocfs2_super *osb, - struct inode *inode, - struct ocfs2_extent_tree *et, +static int ocfs2_find_branch_target(struct ocfs2_extent_tree *et, struct buffer_head **target_bh) { int status = 0, i; @@ -1319,27 +1439,27 @@ static int ocfs2_find_branch_target(struct ocfs2_super *osb, struct buffer_head *bh = NULL; struct buffer_head *lowest_bh = NULL; - mlog_entry_void(); - *target_bh = NULL; el = et->et_root_el; while(le16_to_cpu(el->l_tree_depth) > 1) { if (le16_to_cpu(el->l_next_free_rec) == 0) { - ocfs2_error(inode->i_sb, "Dinode %llu has empty " + ocfs2_error(ocfs2_metadata_cache_get_super(et->et_ci), + "Owner %llu has empty " "extent list (next_free_rec == 0)", - (unsigned long long)OCFS2_I(inode)->ip_blkno); + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci)); status = -EIO; goto bail; } i = le16_to_cpu(el->l_next_free_rec) - 1; blkno = le64_to_cpu(el->l_recs[i].e_blkno); if (!blkno) { - ocfs2_error(inode->i_sb, "Dinode %llu has extent " + ocfs2_error(ocfs2_metadata_cache_get_super(et->et_ci), + "Owner %llu has extent " "list where extent # %d has no physical " "block start", - (unsigned long long)OCFS2_I(inode)->ip_blkno, i); + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci), i); status = -EIO; goto bail; } @@ -1347,7 +1467,7 @@ static int ocfs2_find_branch_target(struct ocfs2_super *osb, brelse(bh); bh = NULL; - status = ocfs2_read_extent_block(inode, blkno, &bh); + status = ocfs2_read_extent_block(et->et_ci, blkno, &bh); if (status < 0) { mlog_errno(status); goto bail; @@ -1374,7 +1494,6 @@ static int ocfs2_find_branch_target(struct ocfs2_super *osb, bail: brelse(bh); - mlog_exit(status); return status; } @@ -1388,20 +1507,18 @@ bail: * * *last_eb_bh will be updated by ocfs2_add_branch(). */ -static int ocfs2_grow_tree(struct inode *inode, handle_t *handle, - struct ocfs2_extent_tree *et, int *final_depth, - struct buffer_head **last_eb_bh, +static int ocfs2_grow_tree(handle_t *handle, struct ocfs2_extent_tree *et, + int *final_depth, struct buffer_head **last_eb_bh, struct ocfs2_alloc_context *meta_ac) { int ret, shift; struct ocfs2_extent_list *el = et->et_root_el; int depth = le16_to_cpu(el->l_tree_depth); - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); struct buffer_head *bh = NULL; BUG_ON(meta_ac == NULL); - shift = ocfs2_find_branch_target(osb, inode, et, &bh); + shift = ocfs2_find_branch_target(et, &bh); if (shift < 0) { ret = shift; mlog_errno(ret); @@ -1413,13 +1530,15 @@ static int ocfs2_grow_tree(struct inode *inode, handle_t *handle, * another tree level */ if (shift) { BUG_ON(bh); - mlog(0, "need to shift tree depth (current = %d)\n", depth); + trace_ocfs2_grow_tree( + (unsigned long long) + ocfs2_metadata_cache_owner(et->et_ci), + depth); /* ocfs2_shift_tree_depth will return us a buffer with * the new extent block (so we can pass that to * ocfs2_add_branch). */ - ret = ocfs2_shift_tree_depth(osb, handle, inode, et, - meta_ac, &bh); + ret = ocfs2_shift_tree_depth(handle, et, meta_ac, &bh); if (ret < 0) { mlog_errno(ret); goto out; @@ -1444,8 +1563,7 @@ static int ocfs2_grow_tree(struct inode *inode, handle_t *handle, /* call ocfs2_add_branch to add the final part of the tree with * the new data. */ - mlog(0, "add branch. bh = %p\n", bh); - ret = ocfs2_add_branch(osb, handle, inode, et, bh, last_eb_bh, + ret = ocfs2_add_branch(handle, et, bh, last_eb_bh, meta_ac); if (ret < 0) { mlog_errno(ret); @@ -1519,8 +1637,9 @@ static void ocfs2_rotate_leaf(struct ocfs2_extent_list *el, } insert_index = i; - mlog(0, "ins %u: index %d, has_empty %d, next_free %d, count %d\n", - insert_cpos, insert_index, has_empty, next_free, le16_to_cpu(el->l_count)); + trace_ocfs2_rotate_leaf(insert_cpos, insert_index, + has_empty, next_free, + le16_to_cpu(el->l_count)); BUG_ON(insert_index < 0); BUG_ON(insert_index >= le16_to_cpu(el->l_count)); @@ -1615,9 +1734,9 @@ set_and_inc: * * The array index of the subtree root is passed back. */ -static int ocfs2_find_subtree_root(struct inode *inode, - struct ocfs2_path *left, - struct ocfs2_path *right) +int ocfs2_find_subtree_root(struct ocfs2_extent_tree *et, + struct ocfs2_path *left, + struct ocfs2_path *right) { int i = 0; @@ -1633,10 +1752,10 @@ static int ocfs2_find_subtree_root(struct inode *inode, * The caller didn't pass two adjacent paths. */ mlog_bug_on_msg(i > left->p_tree_depth, - "Inode %lu, left depth %u, right depth %u\n" + "Owner %llu, left depth %u, right depth %u\n" "left leaf blk %llu, right leaf blk %llu\n", - inode->i_ino, left->p_tree_depth, - right->p_tree_depth, + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci), + left->p_tree_depth, right->p_tree_depth, (unsigned long long)path_leaf_bh(left)->b_blocknr, (unsigned long long)path_leaf_bh(right)->b_blocknr); } while (left->p_node[i].bh->b_blocknr == @@ -1653,7 +1772,7 @@ typedef void (path_insert_t)(void *, struct buffer_head *); * This code can be called with a cpos larger than the tree, in which * case it will return the rightmost path. */ -static int __ocfs2_find_path(struct inode *inode, +static int __ocfs2_find_path(struct ocfs2_caching_info *ci, struct ocfs2_extent_list *root_el, u32 cpos, path_insert_t *func, void *data) { @@ -1664,15 +1783,14 @@ static int __ocfs2_find_path(struct inode *inode, struct ocfs2_extent_block *eb; struct ocfs2_extent_list *el; struct ocfs2_extent_rec *rec; - struct ocfs2_inode_info *oi = OCFS2_I(inode); el = root_el; while (el->l_tree_depth) { if (le16_to_cpu(el->l_next_free_rec) == 0) { - ocfs2_error(inode->i_sb, - "Inode %llu has empty extent list at " + ocfs2_error(ocfs2_metadata_cache_get_super(ci), + "Owner %llu has empty extent list at " "depth %u\n", - (unsigned long long)oi->ip_blkno, + (unsigned long long)ocfs2_metadata_cache_owner(ci), le16_to_cpu(el->l_tree_depth)); ret = -EROFS; goto out; @@ -1695,10 +1813,10 @@ static int __ocfs2_find_path(struct inode *inode, blkno = le64_to_cpu(el->l_recs[i].e_blkno); if (blkno == 0) { - ocfs2_error(inode->i_sb, - "Inode %llu has bad blkno in extent list " + ocfs2_error(ocfs2_metadata_cache_get_super(ci), + "Owner %llu has bad blkno in extent list " "at depth %u (index %d)\n", - (unsigned long long)oi->ip_blkno, + (unsigned long long)ocfs2_metadata_cache_owner(ci), le16_to_cpu(el->l_tree_depth), i); ret = -EROFS; goto out; @@ -1706,7 +1824,7 @@ static int __ocfs2_find_path(struct inode *inode, brelse(bh); bh = NULL; - ret = ocfs2_read_extent_block(inode, blkno, &bh); + ret = ocfs2_read_extent_block(ci, blkno, &bh); if (ret) { mlog_errno(ret); goto out; @@ -1717,10 +1835,10 @@ static int __ocfs2_find_path(struct inode *inode, if (le16_to_cpu(el->l_next_free_rec) > le16_to_cpu(el->l_count)) { - ocfs2_error(inode->i_sb, - "Inode %llu has bad count in extent list " + ocfs2_error(ocfs2_metadata_cache_get_super(ci), + "Owner %llu has bad count in extent list " "at block %llu (next free=%u, count=%u)\n", - (unsigned long long)oi->ip_blkno, + (unsigned long long)ocfs2_metadata_cache_owner(ci), (unsigned long long)bh->b_blocknr, le16_to_cpu(el->l_next_free_rec), le16_to_cpu(el->l_count)); @@ -1764,14 +1882,14 @@ static void find_path_ins(void *data, struct buffer_head *bh) ocfs2_path_insert_eb(fp->path, fp->index, bh); fp->index++; } -static int ocfs2_find_path(struct inode *inode, struct ocfs2_path *path, - u32 cpos) +int ocfs2_find_path(struct ocfs2_caching_info *ci, + struct ocfs2_path *path, u32 cpos) { struct find_path_data data; data.index = 1; data.path = path; - return __ocfs2_find_path(inode, path_root_el(path), cpos, + return __ocfs2_find_path(ci, path_root_el(path), cpos, find_path_ins, &data); } @@ -1796,13 +1914,14 @@ static void find_leaf_ins(void *data, struct buffer_head *bh) * * This function doesn't handle non btree extent lists. */ -int ocfs2_find_leaf(struct inode *inode, struct ocfs2_extent_list *root_el, - u32 cpos, struct buffer_head **leaf_bh) +int ocfs2_find_leaf(struct ocfs2_caching_info *ci, + struct ocfs2_extent_list *root_el, u32 cpos, + struct buffer_head **leaf_bh) { int ret; struct buffer_head *bh = NULL; - ret = __ocfs2_find_path(inode, root_el, cpos, find_leaf_ins, &bh); + ret = __ocfs2_find_path(ci, root_el, cpos, find_leaf_ins, &bh); if (ret) { mlog_errno(ret); goto out; @@ -1842,7 +1961,8 @@ static void ocfs2_adjust_adjacent_records(struct ocfs2_extent_rec *left_rec, * immediately to their right. */ left_clusters = le32_to_cpu(right_child_el->l_recs[0].e_cpos); - if (ocfs2_is_empty_extent(&right_child_el->l_recs[0])) { + if (!ocfs2_rec_clusters(right_child_el, &right_child_el->l_recs[0])) { + BUG_ON(right_child_el->l_tree_depth); BUG_ON(le16_to_cpu(right_child_el->l_next_free_rec) <= 1); left_clusters = le32_to_cpu(right_child_el->l_recs[1].e_cpos); } @@ -1907,12 +2027,12 @@ static void ocfs2_adjust_root_records(struct ocfs2_extent_list *root_el, * - When we've adjusted the last extent record in the left path leaf and the * 1st extent record in the right path leaf during cross extent block merge. */ -static void ocfs2_complete_edge_insert(struct inode *inode, handle_t *handle, +static void ocfs2_complete_edge_insert(handle_t *handle, struct ocfs2_path *left_path, struct ocfs2_path *right_path, int subtree_index) { - int ret, i, idx; + int i, idx; struct ocfs2_extent_list *el, *left_el, *right_el; struct ocfs2_extent_rec *left_rec, *right_rec; struct buffer_head *root_bh = left_path->p_node[subtree_index].bh; @@ -1932,7 +2052,7 @@ static void ocfs2_complete_edge_insert(struct inode *inode, handle_t *handle, left_el = path_leaf_el(left_path); right_el = path_leaf_el(right_path); for(i = left_path->p_tree_depth - 1; i > subtree_index; i--) { - mlog(0, "Adjust records at index %u\n", i); + trace_ocfs2_complete_edge_insert(i); /* * One nice property of knowing that all of these @@ -1950,13 +2070,8 @@ static void ocfs2_complete_edge_insert(struct inode *inode, handle_t *handle, ocfs2_adjust_adjacent_records(left_rec, left_el, right_rec, right_el); - ret = ocfs2_journal_dirty(handle, left_path->p_node[i].bh); - if (ret) - mlog_errno(ret); - - ret = ocfs2_journal_dirty(handle, right_path->p_node[i].bh); - if (ret) - mlog_errno(ret); + ocfs2_journal_dirty(handle, left_path->p_node[i].bh); + ocfs2_journal_dirty(handle, right_path->p_node[i].bh); /* * Setup our list pointers now so that the current @@ -1980,13 +2095,11 @@ static void ocfs2_complete_edge_insert(struct inode *inode, handle_t *handle, root_bh = left_path->p_node[subtree_index].bh; - ret = ocfs2_journal_dirty(handle, root_bh); - if (ret) - mlog_errno(ret); + ocfs2_journal_dirty(handle, root_bh); } -static int ocfs2_rotate_subtree_right(struct inode *inode, - handle_t *handle, +static int ocfs2_rotate_subtree_right(handle_t *handle, + struct ocfs2_extent_tree *et, struct ocfs2_path *left_path, struct ocfs2_path *right_path, int subtree_index) @@ -2002,10 +2115,10 @@ static int ocfs2_rotate_subtree_right(struct inode *inode, left_el = path_leaf_el(left_path); if (left_el->l_next_free_rec != left_el->l_count) { - ocfs2_error(inode->i_sb, + ocfs2_error(ocfs2_metadata_cache_get_super(et->et_ci), "Inode %llu has non-full interior leaf node %llu" "(next free = %u)", - (unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci), (unsigned long long)left_leaf_bh->b_blocknr, le16_to_cpu(left_el->l_next_free_rec)); return -EROFS; @@ -2021,7 +2134,7 @@ static int ocfs2_rotate_subtree_right(struct inode *inode, root_bh = left_path->p_node[subtree_index].bh; BUG_ON(root_bh != right_path->p_node[subtree_index].bh); - ret = ocfs2_path_bh_journal_access(handle, inode, right_path, + ret = ocfs2_path_bh_journal_access(handle, et->et_ci, right_path, subtree_index); if (ret) { mlog_errno(ret); @@ -2029,14 +2142,14 @@ static int ocfs2_rotate_subtree_right(struct inode *inode, } for(i = subtree_index + 1; i < path_num_items(right_path); i++) { - ret = ocfs2_path_bh_journal_access(handle, inode, + ret = ocfs2_path_bh_journal_access(handle, et->et_ci, right_path, i); if (ret) { mlog_errno(ret); goto out; } - ret = ocfs2_path_bh_journal_access(handle, inode, + ret = ocfs2_path_bh_journal_access(handle, et->et_ci, left_path, i); if (ret) { mlog_errno(ret); @@ -2050,16 +2163,12 @@ static int ocfs2_rotate_subtree_right(struct inode *inode, /* This is a code error, not a disk corruption. */ mlog_bug_on_msg(!right_el->l_next_free_rec, "Inode %llu: Rotate fails " "because rightmost leaf block %llu is empty\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci), (unsigned long long)right_leaf_bh->b_blocknr); ocfs2_create_empty_extent(right_el); - ret = ocfs2_journal_dirty(handle, right_leaf_bh); - if (ret) { - mlog_errno(ret); - goto out; - } + ocfs2_journal_dirty(handle, right_leaf_bh); /* Do the copy now. */ i = le16_to_cpu(left_el->l_next_free_rec) - 1; @@ -2078,14 +2187,10 @@ static int ocfs2_rotate_subtree_right(struct inode *inode, memset(&left_el->l_recs[0], 0, sizeof(struct ocfs2_extent_rec)); le16_add_cpu(&left_el->l_next_free_rec, 1); - ret = ocfs2_journal_dirty(handle, left_leaf_bh); - if (ret) { - mlog_errno(ret); - goto out; - } + ocfs2_journal_dirty(handle, left_leaf_bh); - ocfs2_complete_edge_insert(inode, handle, left_path, right_path, - subtree_index); + ocfs2_complete_edge_insert(handle, left_path, right_path, + subtree_index); out: return ret; @@ -2097,8 +2202,8 @@ out: * * Will return zero if the path passed in is already the leftmost path. */ -static int ocfs2_find_cpos_for_left_leaf(struct super_block *sb, - struct ocfs2_path *path, u32 *cpos) +int ocfs2_find_cpos_for_left_leaf(struct super_block *sb, + struct ocfs2_path *path, u32 *cpos) { int i, j, ret = 0; u64 blkno; @@ -2175,12 +2280,14 @@ static int ocfs2_extend_rotate_transaction(handle_t *handle, int subtree_depth, int op_credits, struct ocfs2_path *path) { + int ret = 0; int credits = (path->p_tree_depth - subtree_depth) * 2 + 1 + op_credits; if (handle->h_buffer_credits < credits) - return ocfs2_extend_trans(handle, credits); + ret = ocfs2_extend_trans(handle, + credits - handle->h_buffer_credits); - return 0; + return ret; } /* @@ -2239,7 +2346,7 @@ static int ocfs2_leftmost_rec_contains(struct ocfs2_extent_list *el, u32 cpos) * * The array is assumed to be large enough to hold an entire path (tree depth). * - * Upon succesful return from this function: + * Upon successful return from this function: * * - The 'right_path' array will contain a path to the leaf block * whose range contains e_cpos. @@ -2248,8 +2355,8 @@ static int ocfs2_leftmost_rec_contains(struct ocfs2_extent_list *el, u32 cpos) * *ret_left_path will contain a valid path which can be passed to * ocfs2_insert_path(). */ -static int ocfs2_rotate_tree_right(struct inode *inode, - handle_t *handle, +static int ocfs2_rotate_tree_right(handle_t *handle, + struct ocfs2_extent_tree *et, enum ocfs2_split_type split, u32 insert_cpos, struct ocfs2_path *right_path, @@ -2258,6 +2365,7 @@ static int ocfs2_rotate_tree_right(struct inode *inode, int ret, start, orig_credits = handle->h_buffer_credits; u32 cpos; struct ocfs2_path *left_path = NULL; + struct super_block *sb = ocfs2_metadata_cache_get_super(et->et_ci); *ret_left_path = NULL; @@ -2268,13 +2376,15 @@ static int ocfs2_rotate_tree_right(struct inode *inode, goto out; } - ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, right_path, &cpos); + ret = ocfs2_find_cpos_for_left_leaf(sb, right_path, &cpos); if (ret) { mlog_errno(ret); goto out; } - mlog(0, "Insert: %u, first left path cpos: %u\n", insert_cpos, cpos); + trace_ocfs2_rotate_tree_right( + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci), + insert_cpos, cpos); /* * What we want to do here is: @@ -2303,10 +2413,12 @@ static int ocfs2_rotate_tree_right(struct inode *inode, * rotating subtrees. */ while (cpos && insert_cpos <= cpos) { - mlog(0, "Rotating a tree: ins. cpos: %u, left path cpos: %u\n", - insert_cpos, cpos); + trace_ocfs2_rotate_tree_right( + (unsigned long long) + ocfs2_metadata_cache_owner(et->et_ci), + insert_cpos, cpos); - ret = ocfs2_find_path(inode, left_path, cpos); + ret = ocfs2_find_path(et->et_ci, left_path, cpos); if (ret) { mlog_errno(ret); goto out; @@ -2314,10 +2426,11 @@ static int ocfs2_rotate_tree_right(struct inode *inode, mlog_bug_on_msg(path_leaf_bh(left_path) == path_leaf_bh(right_path), - "Inode %lu: error during insert of %u " + "Owner %llu: error during insert of %u " "(left path cpos %u) results in two identical " "paths ending at %llu\n", - inode->i_ino, insert_cpos, cpos, + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci), + insert_cpos, cpos, (unsigned long long) path_leaf_bh(left_path)->b_blocknr); @@ -2343,12 +2456,12 @@ static int ocfs2_rotate_tree_right(struct inode *inode, goto out_ret_path; } - start = ocfs2_find_subtree_root(inode, left_path, right_path); + start = ocfs2_find_subtree_root(et, left_path, right_path); - mlog(0, "Subtree root at index %d (blk %llu, depth %d)\n", - start, - (unsigned long long) right_path->p_node[start].bh->b_blocknr, - right_path->p_tree_depth); + trace_ocfs2_rotate_subtree(start, + (unsigned long long) + right_path->p_node[start].bh->b_blocknr, + right_path->p_tree_depth); ret = ocfs2_extend_rotate_transaction(handle, start, orig_credits, right_path); @@ -2357,7 +2470,7 @@ static int ocfs2_rotate_tree_right(struct inode *inode, goto out; } - ret = ocfs2_rotate_subtree_right(inode, handle, left_path, + ret = ocfs2_rotate_subtree_right(handle, et, left_path, right_path, start); if (ret) { mlog_errno(ret); @@ -2389,8 +2502,7 @@ static int ocfs2_rotate_tree_right(struct inode *inode, */ ocfs2_mv_path(right_path, left_path); - ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, right_path, - &cpos); + ret = ocfs2_find_cpos_for_left_leaf(sb, right_path, &cpos); if (ret) { mlog_errno(ret); goto out; @@ -2404,15 +2516,37 @@ out_ret_path: return ret; } -static void ocfs2_update_edge_lengths(struct inode *inode, handle_t *handle, - struct ocfs2_path *path) +static int ocfs2_update_edge_lengths(handle_t *handle, + struct ocfs2_extent_tree *et, + int subtree_index, struct ocfs2_path *path) { - int i, idx; + int i, idx, ret; struct ocfs2_extent_rec *rec; struct ocfs2_extent_list *el; struct ocfs2_extent_block *eb; u32 range; + /* + * In normal tree rotation process, we will never touch the + * tree branch above subtree_index and ocfs2_extend_rotate_transaction + * doesn't reserve the credits for them either. + * + * But we do have a special case here which will update the rightmost + * records for all the bh in the path. + * So we have to allocate extra credits and access them. + */ + ret = ocfs2_extend_trans(handle, subtree_index); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_journal_access_path(et->et_ci, handle, path); + if (ret) { + mlog_errno(ret); + goto out; + } + /* Path should always be rightmost. */ eb = (struct ocfs2_extent_block *)path_leaf_bh(path)->b_data; BUG_ON(eb->h_next_leaf_blk != 0ULL); @@ -2433,9 +2567,12 @@ static void ocfs2_update_edge_lengths(struct inode *inode, handle_t *handle, ocfs2_journal_dirty(handle, path->p_node[i].bh); } +out: + return ret; } -static void ocfs2_unlink_path(struct inode *inode, handle_t *handle, +static void ocfs2_unlink_path(handle_t *handle, + struct ocfs2_extent_tree *et, struct ocfs2_cached_dealloc_ctxt *dealloc, struct ocfs2_path *path, int unlink_start) { @@ -2457,12 +2594,12 @@ static void ocfs2_unlink_path(struct inode *inode, handle_t *handle, mlog(ML_ERROR, "Inode %llu, attempted to remove extent block " "%llu with %u records\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci), (unsigned long long)le64_to_cpu(eb->h_blkno), le16_to_cpu(el->l_next_free_rec)); ocfs2_journal_dirty(handle, bh); - ocfs2_remove_from_cache(inode, bh); + ocfs2_remove_from_cache(et->et_ci, bh); continue; } @@ -2475,11 +2612,12 @@ static void ocfs2_unlink_path(struct inode *inode, handle_t *handle, if (ret) mlog_errno(ret); - ocfs2_remove_from_cache(inode, bh); + ocfs2_remove_from_cache(et->et_ci, bh); } } -static void ocfs2_unlink_subtree(struct inode *inode, handle_t *handle, +static void ocfs2_unlink_subtree(handle_t *handle, + struct ocfs2_extent_tree *et, struct ocfs2_path *left_path, struct ocfs2_path *right_path, int subtree_index, @@ -2510,17 +2648,17 @@ static void ocfs2_unlink_subtree(struct inode *inode, handle_t *handle, ocfs2_journal_dirty(handle, root_bh); ocfs2_journal_dirty(handle, path_leaf_bh(left_path)); - ocfs2_unlink_path(inode, handle, dealloc, right_path, + ocfs2_unlink_path(handle, et, dealloc, right_path, subtree_index + 1); } -static int ocfs2_rotate_subtree_left(struct inode *inode, handle_t *handle, +static int ocfs2_rotate_subtree_left(handle_t *handle, + struct ocfs2_extent_tree *et, struct ocfs2_path *left_path, struct ocfs2_path *right_path, int subtree_index, struct ocfs2_cached_dealloc_ctxt *dealloc, - int *deleted, - struct ocfs2_extent_tree *et) + int *deleted) { int ret, i, del_right_subtree = 0, right_has_empty = 0; struct buffer_head *root_bh, *et_root_bh = path_root_bh(right_path); @@ -2556,7 +2694,7 @@ static int ocfs2_rotate_subtree_left(struct inode *inode, handle_t *handle, return -EAGAIN; if (le16_to_cpu(right_leaf_el->l_next_free_rec) > 1) { - ret = ocfs2_journal_access_eb(handle, inode, + ret = ocfs2_journal_access_eb(handle, et->et_ci, path_leaf_bh(right_path), OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { @@ -2575,7 +2713,7 @@ static int ocfs2_rotate_subtree_left(struct inode *inode, handle_t *handle, * We have to update i_last_eb_blk during the meta * data delete. */ - ret = ocfs2_et_root_journal_access(handle, inode, et, + ret = ocfs2_et_root_journal_access(handle, et, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); @@ -2591,7 +2729,7 @@ static int ocfs2_rotate_subtree_left(struct inode *inode, handle_t *handle, */ BUG_ON(right_has_empty && !del_right_subtree); - ret = ocfs2_path_bh_journal_access(handle, inode, right_path, + ret = ocfs2_path_bh_journal_access(handle, et->et_ci, right_path, subtree_index); if (ret) { mlog_errno(ret); @@ -2599,14 +2737,14 @@ static int ocfs2_rotate_subtree_left(struct inode *inode, handle_t *handle, } for(i = subtree_index + 1; i < path_num_items(right_path); i++) { - ret = ocfs2_path_bh_journal_access(handle, inode, + ret = ocfs2_path_bh_journal_access(handle, et->et_ci, right_path, i); if (ret) { mlog_errno(ret); goto out; } - ret = ocfs2_path_bh_journal_access(handle, inode, + ret = ocfs2_path_bh_journal_access(handle, et->et_ci, left_path, i); if (ret) { mlog_errno(ret); @@ -2635,17 +2773,18 @@ static int ocfs2_rotate_subtree_left(struct inode *inode, handle_t *handle, ocfs2_remove_empty_extent(right_leaf_el); } - ret = ocfs2_journal_dirty(handle, path_leaf_bh(left_path)); - if (ret) - mlog_errno(ret); - ret = ocfs2_journal_dirty(handle, path_leaf_bh(right_path)); - if (ret) - mlog_errno(ret); + ocfs2_journal_dirty(handle, path_leaf_bh(left_path)); + ocfs2_journal_dirty(handle, path_leaf_bh(right_path)); if (del_right_subtree) { - ocfs2_unlink_subtree(inode, handle, left_path, right_path, + ocfs2_unlink_subtree(handle, et, left_path, right_path, subtree_index, dealloc); - ocfs2_update_edge_lengths(inode, handle, left_path); + ret = ocfs2_update_edge_lengths(handle, et, subtree_index, + left_path); + if (ret) { + mlog_errno(ret); + goto out; + } eb = (struct ocfs2_extent_block *)path_leaf_bh(left_path)->b_data; ocfs2_et_set_last_eb_blk(et, le64_to_cpu(eb->h_blkno)); @@ -2658,13 +2797,11 @@ static int ocfs2_rotate_subtree_left(struct inode *inode, handle_t *handle, if (right_has_empty) ocfs2_remove_empty_extent(left_leaf_el); - ret = ocfs2_journal_dirty(handle, et_root_bh); - if (ret) - mlog_errno(ret); + ocfs2_journal_dirty(handle, et_root_bh); *deleted = 1; } else - ocfs2_complete_edge_insert(inode, handle, left_path, right_path, + ocfs2_complete_edge_insert(handle, left_path, right_path, subtree_index); out: @@ -2680,8 +2817,8 @@ out: * This looks similar, but is subtly different to * ocfs2_find_cpos_for_left_leaf(). */ -static int ocfs2_find_cpos_for_right_leaf(struct super_block *sb, - struct ocfs2_path *path, u32 *cpos) +int ocfs2_find_cpos_for_right_leaf(struct super_block *sb, + struct ocfs2_path *path, u32 *cpos) { int i, j, ret = 0; u64 blkno; @@ -2750,8 +2887,8 @@ out: return ret; } -static int ocfs2_rotate_rightmost_leaf_left(struct inode *inode, - handle_t *handle, +static int ocfs2_rotate_rightmost_leaf_left(handle_t *handle, + struct ocfs2_extent_tree *et, struct ocfs2_path *path) { int ret; @@ -2761,7 +2898,7 @@ static int ocfs2_rotate_rightmost_leaf_left(struct inode *inode, if (!ocfs2_is_empty_extent(&el->l_recs[0])) return 0; - ret = ocfs2_path_bh_journal_access(handle, inode, path, + ret = ocfs2_path_bh_journal_access(handle, et->et_ci, path, path_num_items(path) - 1); if (ret) { mlog_errno(ret); @@ -2769,33 +2906,30 @@ static int ocfs2_rotate_rightmost_leaf_left(struct inode *inode, } ocfs2_remove_empty_extent(el); - - ret = ocfs2_journal_dirty(handle, bh); - if (ret) - mlog_errno(ret); + ocfs2_journal_dirty(handle, bh); out: return ret; } -static int __ocfs2_rotate_tree_left(struct inode *inode, - handle_t *handle, int orig_credits, +static int __ocfs2_rotate_tree_left(handle_t *handle, + struct ocfs2_extent_tree *et, + int orig_credits, struct ocfs2_path *path, struct ocfs2_cached_dealloc_ctxt *dealloc, - struct ocfs2_path **empty_extent_path, - struct ocfs2_extent_tree *et) + struct ocfs2_path **empty_extent_path) { int ret, subtree_root, deleted; u32 right_cpos; struct ocfs2_path *left_path = NULL; struct ocfs2_path *right_path = NULL; + struct super_block *sb = ocfs2_metadata_cache_get_super(et->et_ci); BUG_ON(!ocfs2_is_empty_extent(&(path_leaf_el(path)->l_recs[0]))); *empty_extent_path = NULL; - ret = ocfs2_find_cpos_for_right_leaf(inode->i_sb, path, - &right_cpos); + ret = ocfs2_find_cpos_for_right_leaf(sb, path, &right_cpos); if (ret) { mlog_errno(ret); goto out; @@ -2818,17 +2952,16 @@ static int __ocfs2_rotate_tree_left(struct inode *inode, } while (right_cpos) { - ret = ocfs2_find_path(inode, right_path, right_cpos); + ret = ocfs2_find_path(et->et_ci, right_path, right_cpos); if (ret) { mlog_errno(ret); goto out; } - subtree_root = ocfs2_find_subtree_root(inode, left_path, + subtree_root = ocfs2_find_subtree_root(et, left_path, right_path); - mlog(0, "Subtree root at index %d (blk %llu, depth %d)\n", - subtree_root, + trace_ocfs2_rotate_subtree(subtree_root, (unsigned long long) right_path->p_node[subtree_root].bh->b_blocknr, right_path->p_tree_depth); @@ -2844,16 +2977,16 @@ static int __ocfs2_rotate_tree_left(struct inode *inode, * Caller might still want to make changes to the * tree root, so re-add it to the journal here. */ - ret = ocfs2_path_bh_journal_access(handle, inode, + ret = ocfs2_path_bh_journal_access(handle, et->et_ci, left_path, 0); if (ret) { mlog_errno(ret); goto out; } - ret = ocfs2_rotate_subtree_left(inode, handle, left_path, + ret = ocfs2_rotate_subtree_left(handle, et, left_path, right_path, subtree_root, - dealloc, &deleted, et); + dealloc, &deleted); if (ret == -EAGAIN) { /* * The rotation has to temporarily stop due to @@ -2880,7 +3013,7 @@ static int __ocfs2_rotate_tree_left(struct inode *inode, ocfs2_mv_path(left_path, right_path); - ret = ocfs2_find_cpos_for_right_leaf(inode->i_sb, left_path, + ret = ocfs2_find_cpos_for_right_leaf(sb, left_path, &right_cpos); if (ret) { mlog_errno(ret); @@ -2895,10 +3028,10 @@ out: return ret; } -static int ocfs2_remove_rightmost_path(struct inode *inode, handle_t *handle, +static int ocfs2_remove_rightmost_path(handle_t *handle, + struct ocfs2_extent_tree *et, struct ocfs2_path *path, - struct ocfs2_cached_dealloc_ctxt *dealloc, - struct ocfs2_extent_tree *et) + struct ocfs2_cached_dealloc_ctxt *dealloc) { int ret, subtree_index; u32 cpos; @@ -2907,7 +3040,7 @@ static int ocfs2_remove_rightmost_path(struct inode *inode, handle_t *handle, struct ocfs2_extent_list *el; - ret = ocfs2_et_sanity_check(inode, et); + ret = ocfs2_et_sanity_check(et); if (ret) goto out; /* @@ -2922,13 +3055,14 @@ static int ocfs2_remove_rightmost_path(struct inode *inode, handle_t *handle, goto out; } - ret = ocfs2_journal_access_path(inode, handle, path); + ret = ocfs2_journal_access_path(et->et_ci, handle, path); if (ret) { mlog_errno(ret); goto out; } - ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, path, &cpos); + ret = ocfs2_find_cpos_for_left_leaf(ocfs2_metadata_cache_get_super(et->et_ci), + path, &cpos); if (ret) { mlog_errno(ret); goto out; @@ -2946,23 +3080,28 @@ static int ocfs2_remove_rightmost_path(struct inode *inode, handle_t *handle, goto out; } - ret = ocfs2_find_path(inode, left_path, cpos); + ret = ocfs2_find_path(et->et_ci, left_path, cpos); if (ret) { mlog_errno(ret); goto out; } - ret = ocfs2_journal_access_path(inode, handle, left_path); + ret = ocfs2_journal_access_path(et->et_ci, handle, left_path); if (ret) { mlog_errno(ret); goto out; } - subtree_index = ocfs2_find_subtree_root(inode, left_path, path); + subtree_index = ocfs2_find_subtree_root(et, left_path, path); - ocfs2_unlink_subtree(inode, handle, left_path, path, + ocfs2_unlink_subtree(handle, et, left_path, path, subtree_index, dealloc); - ocfs2_update_edge_lengths(inode, handle, left_path); + ret = ocfs2_update_edge_lengths(handle, et, subtree_index, + left_path); + if (ret) { + mlog_errno(ret); + goto out; + } eb = (struct ocfs2_extent_block *)path_leaf_bh(left_path)->b_data; ocfs2_et_set_last_eb_blk(et, le64_to_cpu(eb->h_blkno)); @@ -2971,10 +3110,10 @@ static int ocfs2_remove_rightmost_path(struct inode *inode, handle_t *handle, * 'path' is also the leftmost path which * means it must be the only one. This gets * handled differently because we want to - * revert the inode back to having extents + * revert the root back to having extents * in-line. */ - ocfs2_unlink_path(inode, handle, dealloc, path, 1); + ocfs2_unlink_path(handle, et, dealloc, path, 1); el = et->et_root_el; el->l_tree_depth = 0; @@ -3007,10 +3146,10 @@ out: * the rightmost tree leaf record is removed so the caller is * responsible for detecting and correcting that. */ -static int ocfs2_rotate_tree_left(struct inode *inode, handle_t *handle, +static int ocfs2_rotate_tree_left(handle_t *handle, + struct ocfs2_extent_tree *et, struct ocfs2_path *path, - struct ocfs2_cached_dealloc_ctxt *dealloc, - struct ocfs2_extent_tree *et) + struct ocfs2_cached_dealloc_ctxt *dealloc) { int ret, orig_credits = handle->h_buffer_credits; struct ocfs2_path *tmp_path = NULL, *restart_path = NULL; @@ -3027,8 +3166,7 @@ rightmost_no_delete: * Inline extents. This is trivially handled, so do * it up front. */ - ret = ocfs2_rotate_rightmost_leaf_left(inode, handle, - path); + ret = ocfs2_rotate_rightmost_leaf_left(handle, et, path); if (ret) mlog_errno(ret); goto out; @@ -3044,7 +3182,7 @@ rightmost_no_delete: * * 1) is handled via ocfs2_rotate_rightmost_leaf_left() * 2a) we need the left branch so that we can update it with the unlink - * 2b) we need to bring the inode back to inline extents. + * 2b) we need to bring the root back to inline extents. */ eb = (struct ocfs2_extent_block *)path_leaf_bh(path)->b_data; @@ -3060,9 +3198,9 @@ rightmost_no_delete: if (le16_to_cpu(el->l_next_free_rec) == 0) { ret = -EIO; - ocfs2_error(inode->i_sb, - "Inode %llu has empty extent block at %llu", - (unsigned long long)OCFS2_I(inode)->ip_blkno, + ocfs2_error(ocfs2_metadata_cache_get_super(et->et_ci), + "Owner %llu has empty extent block at %llu", + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci), (unsigned long long)le64_to_cpu(eb->h_blkno)); goto out; } @@ -3076,8 +3214,8 @@ rightmost_no_delete: * nonempty list. */ - ret = ocfs2_remove_rightmost_path(inode, handle, path, - dealloc, et); + ret = ocfs2_remove_rightmost_path(handle, et, path, + dealloc); if (ret) mlog_errno(ret); goto out; @@ -3088,8 +3226,8 @@ rightmost_no_delete: * and restarting from there. */ try_rotate: - ret = __ocfs2_rotate_tree_left(inode, handle, orig_credits, path, - dealloc, &restart_path, et); + ret = __ocfs2_rotate_tree_left(handle, et, orig_credits, path, + dealloc, &restart_path); if (ret && ret != -EAGAIN) { mlog_errno(ret); goto out; @@ -3099,9 +3237,9 @@ try_rotate: tmp_path = restart_path; restart_path = NULL; - ret = __ocfs2_rotate_tree_left(inode, handle, orig_credits, + ret = __ocfs2_rotate_tree_left(handle, et, orig_credits, tmp_path, dealloc, - &restart_path, et); + &restart_path); if (ret && ret != -EAGAIN) { mlog_errno(ret); goto out; @@ -3152,7 +3290,7 @@ static void ocfs2_cleanup_merge(struct ocfs2_extent_list *el, } } -static int ocfs2_get_right_path(struct inode *inode, +static int ocfs2_get_right_path(struct ocfs2_extent_tree *et, struct ocfs2_path *left_path, struct ocfs2_path **ret_right_path) { @@ -3169,8 +3307,8 @@ static int ocfs2_get_right_path(struct inode *inode, left_el = path_leaf_el(left_path); BUG_ON(left_el->l_next_free_rec != left_el->l_count); - ret = ocfs2_find_cpos_for_right_leaf(inode->i_sb, left_path, - &right_cpos); + ret = ocfs2_find_cpos_for_right_leaf(ocfs2_metadata_cache_get_super(et->et_ci), + left_path, &right_cpos); if (ret) { mlog_errno(ret); goto out; @@ -3186,7 +3324,7 @@ static int ocfs2_get_right_path(struct inode *inode, goto out; } - ret = ocfs2_find_path(inode, right_path, right_cpos); + ret = ocfs2_find_path(et->et_ci, right_path, right_cpos); if (ret) { mlog_errno(ret); goto out; @@ -3206,9 +3344,9 @@ out: * For index == l_count - 1, the "next" means the 1st extent rec of the * next extent block. */ -static int ocfs2_merge_rec_right(struct inode *inode, - struct ocfs2_path *left_path, +static int ocfs2_merge_rec_right(struct ocfs2_path *left_path, handle_t *handle, + struct ocfs2_extent_tree *et, struct ocfs2_extent_rec *split_rec, int index) { @@ -3229,7 +3367,7 @@ static int ocfs2_merge_rec_right(struct inode *inode, if (index == le16_to_cpu(el->l_next_free_rec) - 1 && le16_to_cpu(el->l_next_free_rec) == le16_to_cpu(el->l_count)) { /* we meet with a cross extent block merge. */ - ret = ocfs2_get_right_path(inode, left_path, &right_path); + ret = ocfs2_get_right_path(et, left_path, &right_path); if (ret) { mlog_errno(ret); goto out; @@ -3248,8 +3386,8 @@ static int ocfs2_merge_rec_right(struct inode *inode, le16_to_cpu(left_rec->e_leaf_clusters) != le32_to_cpu(right_rec->e_cpos)); - subtree_index = ocfs2_find_subtree_root(inode, - left_path, right_path); + subtree_index = ocfs2_find_subtree_root(et, left_path, + right_path); ret = ocfs2_extend_rotate_transaction(handle, subtree_index, handle->h_buffer_credits, @@ -3262,7 +3400,7 @@ static int ocfs2_merge_rec_right(struct inode *inode, root_bh = left_path->p_node[subtree_index].bh; BUG_ON(root_bh != right_path->p_node[subtree_index].bh); - ret = ocfs2_path_bh_journal_access(handle, inode, right_path, + ret = ocfs2_path_bh_journal_access(handle, et->et_ci, right_path, subtree_index); if (ret) { mlog_errno(ret); @@ -3271,14 +3409,14 @@ static int ocfs2_merge_rec_right(struct inode *inode, for (i = subtree_index + 1; i < path_num_items(right_path); i++) { - ret = ocfs2_path_bh_journal_access(handle, inode, + ret = ocfs2_path_bh_journal_access(handle, et->et_ci, right_path, i); if (ret) { mlog_errno(ret); goto out; } - ret = ocfs2_path_bh_journal_access(handle, inode, + ret = ocfs2_path_bh_journal_access(handle, et->et_ci, left_path, i); if (ret) { mlog_errno(ret); @@ -3291,7 +3429,7 @@ static int ocfs2_merge_rec_right(struct inode *inode, right_rec = &el->l_recs[index + 1]; } - ret = ocfs2_path_bh_journal_access(handle, inode, left_path, + ret = ocfs2_path_bh_journal_access(handle, et->et_ci, left_path, path_num_items(left_path) - 1); if (ret) { mlog_errno(ret); @@ -3302,22 +3440,17 @@ static int ocfs2_merge_rec_right(struct inode *inode, le32_add_cpu(&right_rec->e_cpos, -split_clusters); le64_add_cpu(&right_rec->e_blkno, - -ocfs2_clusters_to_blocks(inode->i_sb, split_clusters)); + -ocfs2_clusters_to_blocks(ocfs2_metadata_cache_get_super(et->et_ci), + split_clusters)); le16_add_cpu(&right_rec->e_leaf_clusters, split_clusters); ocfs2_cleanup_merge(el, index); - ret = ocfs2_journal_dirty(handle, bh); - if (ret) - mlog_errno(ret); - + ocfs2_journal_dirty(handle, bh); if (right_path) { - ret = ocfs2_journal_dirty(handle, path_leaf_bh(right_path)); - if (ret) - mlog_errno(ret); - - ocfs2_complete_edge_insert(inode, handle, left_path, - right_path, subtree_index); + ocfs2_journal_dirty(handle, path_leaf_bh(right_path)); + ocfs2_complete_edge_insert(handle, left_path, right_path, + subtree_index); } out: if (right_path) @@ -3325,7 +3458,7 @@ out: return ret; } -static int ocfs2_get_left_path(struct inode *inode, +static int ocfs2_get_left_path(struct ocfs2_extent_tree *et, struct ocfs2_path *right_path, struct ocfs2_path **ret_left_path) { @@ -3338,7 +3471,7 @@ static int ocfs2_get_left_path(struct inode *inode, /* This function shouldn't be called for non-trees. */ BUG_ON(right_path->p_tree_depth == 0); - ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, + ret = ocfs2_find_cpos_for_left_leaf(ocfs2_metadata_cache_get_super(et->et_ci), right_path, &left_cpos); if (ret) { mlog_errno(ret); @@ -3355,7 +3488,7 @@ static int ocfs2_get_left_path(struct inode *inode, goto out; } - ret = ocfs2_find_path(inode, left_path, left_cpos); + ret = ocfs2_find_path(et->et_ci, left_path, left_cpos); if (ret) { mlog_errno(ret); goto out; @@ -3378,12 +3511,11 @@ out: * remove the rightmost leaf extent block in the right_path and change * the right path to indicate the new rightmost path. */ -static int ocfs2_merge_rec_left(struct inode *inode, - struct ocfs2_path *right_path, +static int ocfs2_merge_rec_left(struct ocfs2_path *right_path, handle_t *handle, + struct ocfs2_extent_tree *et, struct ocfs2_extent_rec *split_rec, struct ocfs2_cached_dealloc_ctxt *dealloc, - struct ocfs2_extent_tree *et, int index) { int ret, i, subtree_index = 0, has_empty_extent = 0; @@ -3401,7 +3533,7 @@ static int ocfs2_merge_rec_left(struct inode *inode, right_rec = &el->l_recs[index]; if (index == 0) { /* we meet with a cross extent block merge. */ - ret = ocfs2_get_left_path(inode, right_path, &left_path); + ret = ocfs2_get_left_path(et, right_path, &left_path); if (ret) { mlog_errno(ret); goto out; @@ -3417,8 +3549,8 @@ static int ocfs2_merge_rec_left(struct inode *inode, le16_to_cpu(left_rec->e_leaf_clusters) != le32_to_cpu(split_rec->e_cpos)); - subtree_index = ocfs2_find_subtree_root(inode, - left_path, right_path); + subtree_index = ocfs2_find_subtree_root(et, left_path, + right_path); ret = ocfs2_extend_rotate_transaction(handle, subtree_index, handle->h_buffer_credits, @@ -3431,7 +3563,7 @@ static int ocfs2_merge_rec_left(struct inode *inode, root_bh = left_path->p_node[subtree_index].bh; BUG_ON(root_bh != right_path->p_node[subtree_index].bh); - ret = ocfs2_path_bh_journal_access(handle, inode, right_path, + ret = ocfs2_path_bh_journal_access(handle, et->et_ci, right_path, subtree_index); if (ret) { mlog_errno(ret); @@ -3440,14 +3572,14 @@ static int ocfs2_merge_rec_left(struct inode *inode, for (i = subtree_index + 1; i < path_num_items(right_path); i++) { - ret = ocfs2_path_bh_journal_access(handle, inode, + ret = ocfs2_path_bh_journal_access(handle, et->et_ci, right_path, i); if (ret) { mlog_errno(ret); goto out; } - ret = ocfs2_path_bh_journal_access(handle, inode, + ret = ocfs2_path_bh_journal_access(handle, et->et_ci, left_path, i); if (ret) { mlog_errno(ret); @@ -3460,7 +3592,7 @@ static int ocfs2_merge_rec_left(struct inode *inode, has_empty_extent = 1; } - ret = ocfs2_path_bh_journal_access(handle, inode, right_path, + ret = ocfs2_path_bh_journal_access(handle, et->et_ci, right_path, path_num_items(right_path) - 1); if (ret) { mlog_errno(ret); @@ -3479,19 +3611,15 @@ static int ocfs2_merge_rec_left(struct inode *inode, le32_add_cpu(&right_rec->e_cpos, split_clusters); le64_add_cpu(&right_rec->e_blkno, - ocfs2_clusters_to_blocks(inode->i_sb, split_clusters)); + ocfs2_clusters_to_blocks(ocfs2_metadata_cache_get_super(et->et_ci), + split_clusters)); le16_add_cpu(&right_rec->e_leaf_clusters, -split_clusters); ocfs2_cleanup_merge(el, index); - ret = ocfs2_journal_dirty(handle, bh); - if (ret) - mlog_errno(ret); - + ocfs2_journal_dirty(handle, bh); if (left_path) { - ret = ocfs2_journal_dirty(handle, path_leaf_bh(left_path)); - if (ret) - mlog_errno(ret); + ocfs2_journal_dirty(handle, path_leaf_bh(left_path)); /* * In the situation that the right_rec is empty and the extent @@ -3501,9 +3629,9 @@ static int ocfs2_merge_rec_left(struct inode *inode, if (le16_to_cpu(right_rec->e_leaf_clusters) == 0 && le16_to_cpu(el->l_next_free_rec) == 1) { - ret = ocfs2_remove_rightmost_path(inode, handle, + ret = ocfs2_remove_rightmost_path(handle, et, right_path, - dealloc, et); + dealloc); if (ret) { mlog_errno(ret); goto out; @@ -3515,7 +3643,7 @@ static int ocfs2_merge_rec_left(struct inode *inode, ocfs2_mv_path(right_path, left_path); left_path = NULL; } else - ocfs2_complete_edge_insert(inode, handle, left_path, + ocfs2_complete_edge_insert(handle, left_path, right_path, subtree_index); } out: @@ -3524,15 +3652,13 @@ out: return ret; } -static int ocfs2_try_to_merge_extent(struct inode *inode, - handle_t *handle, +static int ocfs2_try_to_merge_extent(handle_t *handle, + struct ocfs2_extent_tree *et, struct ocfs2_path *path, int split_index, struct ocfs2_extent_rec *split_rec, struct ocfs2_cached_dealloc_ctxt *dealloc, - struct ocfs2_merge_ctxt *ctxt, - struct ocfs2_extent_tree *et) - + struct ocfs2_merge_ctxt *ctxt) { int ret = 0; struct ocfs2_extent_list *el = path_leaf_el(path); @@ -3548,8 +3674,7 @@ static int ocfs2_try_to_merge_extent(struct inode *inode, * extents - having more than one in a leaf is * illegal. */ - ret = ocfs2_rotate_tree_left(inode, handle, path, - dealloc, et); + ret = ocfs2_rotate_tree_left(handle, et, path, dealloc); if (ret) { mlog_errno(ret); goto out; @@ -3578,8 +3703,7 @@ static int ocfs2_try_to_merge_extent(struct inode *inode, * prevoius extent block. It is more efficient and easier * if we do merge_right first and merge_left later. */ - ret = ocfs2_merge_rec_right(inode, path, - handle, split_rec, + ret = ocfs2_merge_rec_right(path, handle, et, split_rec, split_index); if (ret) { mlog_errno(ret); @@ -3592,8 +3716,7 @@ static int ocfs2_try_to_merge_extent(struct inode *inode, BUG_ON(!ocfs2_is_empty_extent(&el->l_recs[0])); /* The merge left us with an empty extent, remove it. */ - ret = ocfs2_rotate_tree_left(inode, handle, path, - dealloc, et); + ret = ocfs2_rotate_tree_left(handle, et, path, dealloc); if (ret) { mlog_errno(ret); goto out; @@ -3605,18 +3728,15 @@ static int ocfs2_try_to_merge_extent(struct inode *inode, * Note that we don't pass split_rec here on purpose - * we've merged it into the rec already. */ - ret = ocfs2_merge_rec_left(inode, path, - handle, rec, - dealloc, et, - split_index); + ret = ocfs2_merge_rec_left(path, handle, et, rec, + dealloc, split_index); if (ret) { mlog_errno(ret); goto out; } - ret = ocfs2_rotate_tree_left(inode, handle, path, - dealloc, et); + ret = ocfs2_rotate_tree_left(handle, et, path, dealloc); /* * Error from this last rotate is not critical, so * print but don't bubble it up. @@ -3633,19 +3753,16 @@ static int ocfs2_try_to_merge_extent(struct inode *inode, * the record on the left (hence the left merge). */ if (ctxt->c_contig_type == CONTIG_RIGHT) { - ret = ocfs2_merge_rec_left(inode, - path, - handle, split_rec, - dealloc, et, + ret = ocfs2_merge_rec_left(path, handle, et, + split_rec, dealloc, split_index); if (ret) { mlog_errno(ret); goto out; } } else { - ret = ocfs2_merge_rec_right(inode, - path, - handle, split_rec, + ret = ocfs2_merge_rec_right(path, handle, + et, split_rec, split_index); if (ret) { mlog_errno(ret); @@ -3658,8 +3775,8 @@ static int ocfs2_try_to_merge_extent(struct inode *inode, * The merge may have left an empty extent in * our leaf. Try to rotate it away. */ - ret = ocfs2_rotate_tree_left(inode, handle, path, - dealloc, et); + ret = ocfs2_rotate_tree_left(handle, et, path, + dealloc); if (ret) mlog_errno(ret); ret = 0; @@ -3705,10 +3822,10 @@ static void ocfs2_subtract_from_rec(struct super_block *sb, * list. If this leaf is part of an allocation tree, it is assumed * that the tree above has been prepared. */ -static void ocfs2_insert_at_leaf(struct ocfs2_extent_rec *insert_rec, +static void ocfs2_insert_at_leaf(struct ocfs2_extent_tree *et, + struct ocfs2_extent_rec *insert_rec, struct ocfs2_extent_list *el, - struct ocfs2_insert_type *insert, - struct inode *inode) + struct ocfs2_insert_type *insert) { int i = insert->ins_contig_index; unsigned int range; @@ -3720,7 +3837,8 @@ static void ocfs2_insert_at_leaf(struct ocfs2_extent_rec *insert_rec, i = ocfs2_search_extent_list(el, le32_to_cpu(insert_rec->e_cpos)); BUG_ON(i == -1); rec = &el->l_recs[i]; - ocfs2_subtract_from_rec(inode->i_sb, insert->ins_split, rec, + ocfs2_subtract_from_rec(ocfs2_metadata_cache_get_super(et->et_ci), + insert->ins_split, rec, insert_rec); goto rotate; } @@ -3762,10 +3880,10 @@ static void ocfs2_insert_at_leaf(struct ocfs2_extent_rec *insert_rec, mlog_bug_on_msg(le16_to_cpu(el->l_next_free_rec) >= le16_to_cpu(el->l_count), - "inode %lu, depth %u, count %u, next free %u, " + "owner %llu, depth %u, count %u, next free %u, " "rec.cpos %u, rec.clusters %u, " "insert.cpos %u, insert.clusters %u\n", - inode->i_ino, + ocfs2_metadata_cache_owner(et->et_ci), le16_to_cpu(el->l_tree_depth), le16_to_cpu(el->l_count), le16_to_cpu(el->l_next_free_rec), @@ -3793,8 +3911,8 @@ rotate: ocfs2_rotate_leaf(el, insert_rec); } -static void ocfs2_adjust_rightmost_records(struct inode *inode, - handle_t *handle, +static void ocfs2_adjust_rightmost_records(handle_t *handle, + struct ocfs2_extent_tree *et, struct ocfs2_path *path, struct ocfs2_extent_rec *insert_rec) { @@ -3812,9 +3930,9 @@ static void ocfs2_adjust_rightmost_records(struct inode *inode, next_free = le16_to_cpu(el->l_next_free_rec); if (next_free == 0) { - ocfs2_error(inode->i_sb, - "Dinode %llu has a bad extent list", - (unsigned long long)OCFS2_I(inode)->ip_blkno); + ocfs2_error(ocfs2_metadata_cache_get_super(et->et_ci), + "Owner %llu has a bad extent list", + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci)); ret = -EIO; return; } @@ -3827,14 +3945,12 @@ static void ocfs2_adjust_rightmost_records(struct inode *inode, le32_add_cpu(&rec->e_int_clusters, -le32_to_cpu(rec->e_cpos)); - ret = ocfs2_journal_dirty(handle, bh); - if (ret) - mlog_errno(ret); - + ocfs2_journal_dirty(handle, bh); } } -static int ocfs2_append_rec_to_path(struct inode *inode, handle_t *handle, +static int ocfs2_append_rec_to_path(handle_t *handle, + struct ocfs2_extent_tree *et, struct ocfs2_extent_rec *insert_rec, struct ocfs2_path *right_path, struct ocfs2_path **ret_left_path) @@ -3862,16 +3978,18 @@ static int ocfs2_append_rec_to_path(struct inode *inode, handle_t *handle, (next_free == 1 && ocfs2_is_empty_extent(&el->l_recs[0]))) { u32 left_cpos; - ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, right_path, - &left_cpos); + ret = ocfs2_find_cpos_for_left_leaf(ocfs2_metadata_cache_get_super(et->et_ci), + right_path, &left_cpos); if (ret) { mlog_errno(ret); goto out; } - mlog(0, "Append may need a left path update. cpos: %u, " - "left_cpos: %u\n", le32_to_cpu(insert_rec->e_cpos), - left_cpos); + trace_ocfs2_append_rec_to_path( + (unsigned long long) + ocfs2_metadata_cache_owner(et->et_ci), + le32_to_cpu(insert_rec->e_cpos), + left_cpos); /* * No need to worry if the append is already in the @@ -3885,7 +4003,8 @@ static int ocfs2_append_rec_to_path(struct inode *inode, handle_t *handle, goto out; } - ret = ocfs2_find_path(inode, left_path, left_cpos); + ret = ocfs2_find_path(et->et_ci, left_path, + left_cpos); if (ret) { mlog_errno(ret); goto out; @@ -3898,13 +4017,13 @@ static int ocfs2_append_rec_to_path(struct inode *inode, handle_t *handle, } } - ret = ocfs2_journal_access_path(inode, handle, right_path); + ret = ocfs2_journal_access_path(et->et_ci, handle, right_path); if (ret) { mlog_errno(ret); goto out; } - ocfs2_adjust_rightmost_records(inode, handle, right_path, insert_rec); + ocfs2_adjust_rightmost_records(handle, et, right_path, insert_rec); *ret_left_path = left_path; ret = 0; @@ -3915,7 +4034,7 @@ out: return ret; } -static void ocfs2_split_record(struct inode *inode, +static void ocfs2_split_record(struct ocfs2_extent_tree *et, struct ocfs2_path *left_path, struct ocfs2_path *right_path, struct ocfs2_extent_rec *split_rec, @@ -3988,7 +4107,8 @@ static void ocfs2_split_record(struct inode *inode, } rec = &el->l_recs[index]; - ocfs2_subtract_from_rec(inode->i_sb, split, rec, split_rec); + ocfs2_subtract_from_rec(ocfs2_metadata_cache_get_super(et->et_ci), + split, rec, split_rec); ocfs2_rotate_leaf(insert_el, split_rec); } @@ -4000,8 +4120,8 @@ static void ocfs2_split_record(struct inode *inode, * in. left_path should only be passed in if we need to update that * portion of the tree after an edge insert. */ -static int ocfs2_insert_path(struct inode *inode, - handle_t *handle, +static int ocfs2_insert_path(handle_t *handle, + struct ocfs2_extent_tree *et, struct ocfs2_path *left_path, struct ocfs2_path *right_path, struct ocfs2_extent_rec *insert_rec, @@ -4011,23 +4131,19 @@ static int ocfs2_insert_path(struct inode *inode, struct buffer_head *leaf_bh = path_leaf_bh(right_path); if (left_path) { - int credits = handle->h_buffer_credits; - /* * There's a chance that left_path got passed back to * us without being accounted for in the * journal. Extend our transaction here to be sure we * can change those blocks. */ - credits += left_path->p_tree_depth; - - ret = ocfs2_extend_trans(handle, credits); + ret = ocfs2_extend_trans(handle, left_path->p_tree_depth); if (ret < 0) { mlog_errno(ret); goto out; } - ret = ocfs2_journal_access_path(inode, handle, left_path); + ret = ocfs2_journal_access_path(et->et_ci, handle, left_path); if (ret < 0) { mlog_errno(ret); goto out; @@ -4038,7 +4154,7 @@ static int ocfs2_insert_path(struct inode *inode, * Pass both paths to the journal. The majority of inserts * will be touching all components anyway. */ - ret = ocfs2_journal_access_path(inode, handle, right_path); + ret = ocfs2_journal_access_path(et->et_ci, handle, right_path); if (ret < 0) { mlog_errno(ret); goto out; @@ -4050,7 +4166,7 @@ static int ocfs2_insert_path(struct inode *inode, * of splits, but it's easier to just let one separate * function sort it all out. */ - ocfs2_split_record(inode, left_path, right_path, + ocfs2_split_record(et, left_path, right_path, insert_rec, insert->ins_split); /* @@ -4059,17 +4175,13 @@ static int ocfs2_insert_path(struct inode *inode, * dirty this for us. */ if (left_path) - ret = ocfs2_journal_dirty(handle, - path_leaf_bh(left_path)); - if (ret) - mlog_errno(ret); + ocfs2_journal_dirty(handle, + path_leaf_bh(left_path)); } else - ocfs2_insert_at_leaf(insert_rec, path_leaf_el(right_path), - insert, inode); + ocfs2_insert_at_leaf(et, insert_rec, path_leaf_el(right_path), + insert); - ret = ocfs2_journal_dirty(handle, leaf_bh); - if (ret) - mlog_errno(ret); + ocfs2_journal_dirty(handle, leaf_bh); if (left_path) { /* @@ -4078,10 +4190,10 @@ static int ocfs2_insert_path(struct inode *inode, * * XXX: Should we extend the transaction here? */ - subtree_index = ocfs2_find_subtree_root(inode, left_path, + subtree_index = ocfs2_find_subtree_root(et, left_path, right_path); - ocfs2_complete_edge_insert(inode, handle, left_path, - right_path, subtree_index); + ocfs2_complete_edge_insert(handle, left_path, right_path, + subtree_index); } ret = 0; @@ -4089,8 +4201,7 @@ out: return ret; } -static int ocfs2_do_insert_extent(struct inode *inode, - handle_t *handle, +static int ocfs2_do_insert_extent(handle_t *handle, struct ocfs2_extent_tree *et, struct ocfs2_extent_rec *insert_rec, struct ocfs2_insert_type *type) @@ -4103,7 +4214,7 @@ static int ocfs2_do_insert_extent(struct inode *inode, el = et->et_root_el; - ret = ocfs2_et_root_journal_access(handle, inode, et, + ret = ocfs2_et_root_journal_access(handle, et, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); @@ -4111,7 +4222,7 @@ static int ocfs2_do_insert_extent(struct inode *inode, } if (le16_to_cpu(el->l_tree_depth) == 0) { - ocfs2_insert_at_leaf(insert_rec, el, type, inode); + ocfs2_insert_at_leaf(et, insert_rec, el, type); goto out_update_clusters; } @@ -4134,7 +4245,7 @@ static int ocfs2_do_insert_extent(struct inode *inode, cpos = UINT_MAX; } - ret = ocfs2_find_path(inode, right_path, cpos); + ret = ocfs2_find_path(et->et_ci, right_path, cpos); if (ret) { mlog_errno(ret); goto out; @@ -4153,7 +4264,7 @@ static int ocfs2_do_insert_extent(struct inode *inode, * can wind up skipping both of these two special cases... */ if (rotate) { - ret = ocfs2_rotate_tree_right(inode, handle, type->ins_split, + ret = ocfs2_rotate_tree_right(handle, et, type->ins_split, le32_to_cpu(insert_rec->e_cpos), right_path, &left_path); if (ret) { @@ -4165,7 +4276,7 @@ static int ocfs2_do_insert_extent(struct inode *inode, * ocfs2_rotate_tree_right() might have extended the * transaction without re-journaling our tree root. */ - ret = ocfs2_et_root_journal_access(handle, inode, et, + ret = ocfs2_et_root_journal_access(handle, et, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); @@ -4173,7 +4284,7 @@ static int ocfs2_do_insert_extent(struct inode *inode, } } else if (type->ins_appending == APPEND_TAIL && type->ins_contig != CONTIG_LEFT) { - ret = ocfs2_append_rec_to_path(inode, handle, insert_rec, + ret = ocfs2_append_rec_to_path(handle, et, insert_rec, right_path, &left_path); if (ret) { mlog_errno(ret); @@ -4181,7 +4292,7 @@ static int ocfs2_do_insert_extent(struct inode *inode, } } - ret = ocfs2_insert_path(inode, handle, left_path, right_path, + ret = ocfs2_insert_path(handle, et, left_path, right_path, insert_rec, type); if (ret) { mlog_errno(ret); @@ -4190,12 +4301,10 @@ static int ocfs2_do_insert_extent(struct inode *inode, out_update_clusters: if (type->ins_split == SPLIT_NONE) - ocfs2_et_update_clusters(inode, et, + ocfs2_et_update_clusters(et, le16_to_cpu(insert_rec->e_leaf_clusters)); - ret = ocfs2_journal_dirty(handle, et->et_root_bh); - if (ret) - mlog_errno(ret); + ocfs2_journal_dirty(handle, et->et_root_bh); out: ocfs2_free_path(left_path); @@ -4205,7 +4314,8 @@ out: } static enum ocfs2_contig_type -ocfs2_figure_merge_contig_type(struct inode *inode, struct ocfs2_path *path, +ocfs2_figure_merge_contig_type(struct ocfs2_extent_tree *et, + struct ocfs2_path *path, struct ocfs2_extent_list *el, int index, struct ocfs2_extent_rec *split_rec) { @@ -4217,12 +4327,12 @@ ocfs2_figure_merge_contig_type(struct inode *inode, struct ocfs2_path *path, struct ocfs2_path *left_path = NULL, *right_path = NULL; struct buffer_head *bh; struct ocfs2_extent_block *eb; + struct super_block *sb = ocfs2_metadata_cache_get_super(et->et_ci); if (index > 0) { rec = &el->l_recs[index - 1]; } else if (path->p_tree_depth > 0) { - status = ocfs2_find_cpos_for_left_leaf(inode->i_sb, - path, &left_cpos); + status = ocfs2_find_cpos_for_left_leaf(sb, path, &left_cpos); if (status) goto out; @@ -4231,7 +4341,8 @@ ocfs2_figure_merge_contig_type(struct inode *inode, struct ocfs2_path *path, if (!left_path) goto out; - status = ocfs2_find_path(inode, left_path, left_cpos); + status = ocfs2_find_path(et->et_ci, left_path, + left_cpos); if (status) goto out; @@ -4241,7 +4352,7 @@ ocfs2_figure_merge_contig_type(struct inode *inode, struct ocfs2_path *path, le16_to_cpu(new_el->l_count)) { bh = path_leaf_bh(left_path); eb = (struct ocfs2_extent_block *)bh->b_data; - ocfs2_error(inode->i_sb, + ocfs2_error(sb, "Extent block #%llu has an " "invalid l_next_free_rec of " "%d. It should have " @@ -4266,7 +4377,7 @@ ocfs2_figure_merge_contig_type(struct inode *inode, struct ocfs2_path *path, if (split_rec->e_cpos == el->l_recs[index].e_cpos) ret = CONTIG_RIGHT; } else { - ret = ocfs2_extent_contig(inode, rec, split_rec); + ret = ocfs2_et_extent_contig(et, rec, split_rec); } } @@ -4275,8 +4386,7 @@ ocfs2_figure_merge_contig_type(struct inode *inode, struct ocfs2_path *path, rec = &el->l_recs[index + 1]; else if (le16_to_cpu(el->l_next_free_rec) == le16_to_cpu(el->l_count) && path->p_tree_depth > 0) { - status = ocfs2_find_cpos_for_right_leaf(inode->i_sb, - path, &right_cpos); + status = ocfs2_find_cpos_for_right_leaf(sb, path, &right_cpos); if (status) goto out; @@ -4287,7 +4397,7 @@ ocfs2_figure_merge_contig_type(struct inode *inode, struct ocfs2_path *path, if (!right_path) goto out; - status = ocfs2_find_path(inode, right_path, right_cpos); + status = ocfs2_find_path(et->et_ci, right_path, right_cpos); if (status) goto out; @@ -4297,7 +4407,7 @@ ocfs2_figure_merge_contig_type(struct inode *inode, struct ocfs2_path *path, if (le16_to_cpu(new_el->l_next_free_rec) <= 1) { bh = path_leaf_bh(right_path); eb = (struct ocfs2_extent_block *)bh->b_data; - ocfs2_error(inode->i_sb, + ocfs2_error(sb, "Extent block #%llu has an " "invalid l_next_free_rec of %d", (unsigned long long)le64_to_cpu(eb->h_blkno), @@ -4312,7 +4422,7 @@ ocfs2_figure_merge_contig_type(struct inode *inode, struct ocfs2_path *path, if (rec) { enum ocfs2_contig_type contig_type; - contig_type = ocfs2_extent_contig(inode, rec, split_rec); + contig_type = ocfs2_et_extent_contig(et, rec, split_rec); if (contig_type == CONTIG_LEFT && ret == CONTIG_RIGHT) ret = CONTIG_LEFTRIGHT; @@ -4329,11 +4439,10 @@ out: return ret; } -static void ocfs2_figure_contig_type(struct inode *inode, +static void ocfs2_figure_contig_type(struct ocfs2_extent_tree *et, struct ocfs2_insert_type *insert, struct ocfs2_extent_list *el, - struct ocfs2_extent_rec *insert_rec, - struct ocfs2_extent_tree *et) + struct ocfs2_extent_rec *insert_rec) { int i; enum ocfs2_contig_type contig_type = CONTIG_NONE; @@ -4341,8 +4450,8 @@ static void ocfs2_figure_contig_type(struct inode *inode, BUG_ON(le16_to_cpu(el->l_tree_depth) != 0); for(i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) { - contig_type = ocfs2_extent_contig(inode, &el->l_recs[i], - insert_rec); + contig_type = ocfs2_et_extent_contig(et, &el->l_recs[i], + insert_rec); if (contig_type != CONTIG_NONE) { insert->ins_contig_index = i; break; @@ -4411,7 +4520,7 @@ set_tail_append: } /* - * Helper function called at the begining of an insert. + * Helper function called at the beginning of an insert. * * This computes a few things that are commonly used in the process of * inserting into the btree: @@ -4423,8 +4532,7 @@ set_tail_append: * All of the information is stored on the ocfs2_insert_type * structure. */ -static int ocfs2_figure_insert_type(struct inode *inode, - struct ocfs2_extent_tree *et, +static int ocfs2_figure_insert_type(struct ocfs2_extent_tree *et, struct buffer_head **last_eb_bh, struct ocfs2_extent_rec *insert_rec, int *free_records, @@ -4448,11 +4556,11 @@ static int ocfs2_figure_insert_type(struct inode *inode, * ocfs2_figure_insert_type() and ocfs2_add_branch() * may want it later. */ - ret = ocfs2_read_extent_block(inode, + ret = ocfs2_read_extent_block(et->et_ci, ocfs2_et_get_last_eb_blk(et), &bh); if (ret) { - mlog_exit(ret); + mlog_errno(ret); goto out; } eb = (struct ocfs2_extent_block *) bh->b_data; @@ -4471,7 +4579,7 @@ static int ocfs2_figure_insert_type(struct inode *inode, le16_to_cpu(el->l_next_free_rec); if (!insert->ins_tree_depth) { - ocfs2_figure_contig_type(inode, insert, el, insert_rec, et); + ocfs2_figure_contig_type(et, insert, el, insert_rec); ocfs2_figure_appending_type(insert, el, insert_rec); return 0; } @@ -4489,7 +4597,7 @@ static int ocfs2_figure_insert_type(struct inode *inode, * us the rightmost tree path. This is accounted for below in * the appending code. */ - ret = ocfs2_find_path(inode, path, le32_to_cpu(insert_rec->e_cpos)); + ret = ocfs2_find_path(et->et_ci, path, le32_to_cpu(insert_rec->e_cpos)); if (ret) { mlog_errno(ret); goto out; @@ -4505,7 +4613,7 @@ static int ocfs2_figure_insert_type(struct inode *inode, * into two types of appends: simple record append, or a * rotate inside the tail leaf. */ - ocfs2_figure_contig_type(inode, insert, el, insert_rec, et); + ocfs2_figure_contig_type(et, insert, el, insert_rec); /* * The insert code isn't quite ready to deal with all cases of @@ -4550,13 +4658,11 @@ out: } /* - * Insert an extent into an inode btree. + * Insert an extent into a btree. * - * The caller needs to update fe->i_clusters + * The caller needs to update the owning btree's cluster count. */ -int ocfs2_insert_extent(struct ocfs2_super *osb, - handle_t *handle, - struct inode *inode, +int ocfs2_insert_extent(handle_t *handle, struct ocfs2_extent_tree *et, u32 cpos, u64 start_blk, @@ -4570,35 +4676,34 @@ int ocfs2_insert_extent(struct ocfs2_super *osb, struct ocfs2_insert_type insert = {0, }; struct ocfs2_extent_rec rec; - mlog(0, "add %u clusters at position %u to inode %llu\n", - new_clusters, cpos, (unsigned long long)OCFS2_I(inode)->ip_blkno); + trace_ocfs2_insert_extent_start( + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci), + cpos, new_clusters); memset(&rec, 0, sizeof(rec)); rec.e_cpos = cpu_to_le32(cpos); rec.e_blkno = cpu_to_le64(start_blk); rec.e_leaf_clusters = cpu_to_le16(new_clusters); rec.e_flags = flags; - status = ocfs2_et_insert_check(inode, et, &rec); + status = ocfs2_et_insert_check(et, &rec); if (status) { mlog_errno(status); goto bail; } - status = ocfs2_figure_insert_type(inode, et, &last_eb_bh, &rec, + status = ocfs2_figure_insert_type(et, &last_eb_bh, &rec, &free_records, &insert); if (status < 0) { mlog_errno(status); goto bail; } - mlog(0, "Insert.appending: %u, Insert.Contig: %u, " - "Insert.contig_index: %d, Insert.free_records: %d, " - "Insert.tree_depth: %d\n", - insert.ins_appending, insert.ins_contig, insert.ins_contig_index, - free_records, insert.ins_tree_depth); + trace_ocfs2_insert_extent(insert.ins_appending, insert.ins_contig, + insert.ins_contig_index, free_records, + insert.ins_tree_depth); if (insert.ins_contig == CONTIG_NONE && free_records == 0) { - status = ocfs2_grow_tree(inode, handle, et, + status = ocfs2_grow_tree(handle, et, &insert.ins_tree_depth, &last_eb_bh, meta_ac); if (status) { @@ -4608,16 +4713,15 @@ int ocfs2_insert_extent(struct ocfs2_super *osb, } /* Finally, we can add clusters. This might rotate the tree for us. */ - status = ocfs2_do_insert_extent(inode, handle, et, &rec, &insert); + status = ocfs2_do_insert_extent(handle, et, &rec, &insert); if (status < 0) mlog_errno(status); - else if (et->et_ops == &ocfs2_dinode_et_ops) - ocfs2_extent_map_insert_rec(inode, &rec); + else + ocfs2_et_extent_map_insert(et, &rec); bail: brelse(last_eb_bh); - mlog_exit(status); return status; } @@ -4628,30 +4732,31 @@ bail: * it is not limited to the file storage. Any extent tree can use this * function if it implements the proper ocfs2_extent_tree. */ -int ocfs2_add_clusters_in_btree(struct ocfs2_super *osb, - struct inode *inode, +int ocfs2_add_clusters_in_btree(handle_t *handle, + struct ocfs2_extent_tree *et, u32 *logical_offset, u32 clusters_to_add, int mark_unwritten, - struct ocfs2_extent_tree *et, - handle_t *handle, struct ocfs2_alloc_context *data_ac, struct ocfs2_alloc_context *meta_ac, enum ocfs2_alloc_restarted *reason_ret) { - int status = 0; + int status = 0, err = 0; + int need_free = 0; int free_extents; enum ocfs2_alloc_restarted reason = RESTART_NONE; u32 bit_off, num_bits; u64 block; u8 flags = 0; + struct ocfs2_super *osb = + OCFS2_SB(ocfs2_metadata_cache_get_super(et->et_ci)); BUG_ON(!clusters_to_add); if (mark_unwritten) flags = OCFS2_EXT_UNWRITTEN; - free_extents = ocfs2_num_free_extents(osb, inode, et); + free_extents = ocfs2_num_free_extents(osb, et); if (free_extents < 0) { status = free_extents; mlog_errno(status); @@ -4664,20 +4769,20 @@ int ocfs2_add_clusters_in_btree(struct ocfs2_super *osb, * 2) we are so fragmented, we've needed to add metadata too * many times. */ if (!free_extents && !meta_ac) { - mlog(0, "we haven't reserved any metadata!\n"); + err = -1; status = -EAGAIN; reason = RESTART_META; goto leave; } else if ((!free_extents) && (ocfs2_alloc_context_bits_left(meta_ac) < ocfs2_extend_meta_needed(et->et_root_el))) { - mlog(0, "filesystem is really fragmented...\n"); + err = -2; status = -EAGAIN; reason = RESTART_META; goto leave; } - status = __ocfs2_claim_clusters(osb, handle, data_ac, 1, + status = __ocfs2_claim_clusters(handle, data_ac, 1, clusters_to_add, &bit_off, &num_bits); if (status < 0) { if (status != -ENOSPC) @@ -4688,44 +4793,54 @@ int ocfs2_add_clusters_in_btree(struct ocfs2_super *osb, BUG_ON(num_bits > clusters_to_add); /* reserve our write early -- insert_extent may update the tree root */ - status = ocfs2_et_root_journal_access(handle, inode, et, + status = ocfs2_et_root_journal_access(handle, et, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); - goto leave; + need_free = 1; + goto bail; } block = ocfs2_clusters_to_blocks(osb->sb, bit_off); - mlog(0, "Allocating %u clusters at block %u for inode %llu\n", - num_bits, bit_off, (unsigned long long)OCFS2_I(inode)->ip_blkno); - status = ocfs2_insert_extent(osb, handle, inode, et, - *logical_offset, block, + trace_ocfs2_add_clusters_in_btree( + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci), + bit_off, num_bits); + status = ocfs2_insert_extent(handle, et, *logical_offset, block, num_bits, flags, meta_ac); if (status < 0) { mlog_errno(status); - goto leave; + need_free = 1; + goto bail; } - status = ocfs2_journal_dirty(handle, et->et_root_bh); - if (status < 0) { - mlog_errno(status); - goto leave; - } + ocfs2_journal_dirty(handle, et->et_root_bh); clusters_to_add -= num_bits; *logical_offset += num_bits; if (clusters_to_add) { - mlog(0, "need to alloc once more, wanted = %u\n", - clusters_to_add); + err = clusters_to_add; status = -EAGAIN; reason = RESTART_TRANS; } +bail: + if (need_free) { + if (data_ac->ac_which == OCFS2_AC_USE_LOCAL) + ocfs2_free_local_alloc_bits(osb, handle, data_ac, + bit_off, num_bits); + else + ocfs2_free_clusters(handle, + data_ac->ac_inode, + data_ac->ac_bh, + ocfs2_clusters_to_blocks(osb->sb, bit_off), + num_bits); + } + leave: - mlog_exit(status); if (reason_ret) *reason_ret = reason; + trace_ocfs2_add_clusters_in_btree_ret(status, reason, err); return status; } @@ -4749,10 +4864,9 @@ static void ocfs2_make_right_split_rec(struct super_block *sb, split_rec->e_flags = rec->e_flags; } -static int ocfs2_split_and_insert(struct inode *inode, - handle_t *handle, - struct ocfs2_path *path, +static int ocfs2_split_and_insert(handle_t *handle, struct ocfs2_extent_tree *et, + struct ocfs2_path *path, struct buffer_head **last_eb_bh, int split_index, struct ocfs2_extent_rec *orig_split_rec, @@ -4785,7 +4899,7 @@ leftright: if (le16_to_cpu(rightmost_el->l_next_free_rec) == le16_to_cpu(rightmost_el->l_count)) { - ret = ocfs2_grow_tree(inode, handle, et, + ret = ocfs2_grow_tree(handle, et, &depth, last_eb_bh, meta_ac); if (ret) { mlog_errno(ret); @@ -4814,8 +4928,8 @@ leftright: */ insert.ins_split = SPLIT_RIGHT; - ocfs2_make_right_split_rec(inode->i_sb, &tmprec, insert_range, - &rec); + ocfs2_make_right_split_rec(ocfs2_metadata_cache_get_super(et->et_ci), + &tmprec, insert_range, &rec); split_rec = tmprec; @@ -4823,7 +4937,7 @@ leftright: do_leftright = 1; } - ret = ocfs2_do_insert_extent(inode, handle, et, &split_rec, &insert); + ret = ocfs2_do_insert_extent(handle, et, &split_rec, &insert); if (ret) { mlog_errno(ret); goto out; @@ -4839,7 +4953,7 @@ leftright: ocfs2_reinit_path(path, 1); cpos = le32_to_cpu(split_rec.e_cpos); - ret = ocfs2_find_path(inode, path, cpos); + ret = ocfs2_find_path(et->et_ci, path, cpos); if (ret) { mlog_errno(ret); goto out; @@ -4854,8 +4968,8 @@ out: return ret; } -static int ocfs2_replace_extent_rec(struct inode *inode, - handle_t *handle, +static int ocfs2_replace_extent_rec(handle_t *handle, + struct ocfs2_extent_tree *et, struct ocfs2_path *path, struct ocfs2_extent_list *el, int split_index, @@ -4863,7 +4977,7 @@ static int ocfs2_replace_extent_rec(struct inode *inode, { int ret; - ret = ocfs2_path_bh_journal_access(handle, inode, path, + ret = ocfs2_path_bh_journal_access(handle, et->et_ci, path, path_num_items(path) - 1); if (ret) { mlog_errno(ret); @@ -4878,9 +4992,8 @@ out: } /* - * Mark part or all of the extent record at split_index in the leaf - * pointed to by path as written. This removes the unwritten - * extent flag. + * Split part or all of the extent record at split_index in the leaf + * pointed to by path. Merge with the contiguous extent record if needed. * * Care is taken to handle contiguousness so as to not grow the tree. * @@ -4897,14 +5010,13 @@ out: * have been brought into cache (and pinned via the journal), so the * extra overhead is not expressed in terms of disk reads. */ -static int __ocfs2_mark_extent_written(struct inode *inode, - struct ocfs2_extent_tree *et, - handle_t *handle, - struct ocfs2_path *path, - int split_index, - struct ocfs2_extent_rec *split_rec, - struct ocfs2_alloc_context *meta_ac, - struct ocfs2_cached_dealloc_ctxt *dealloc) +int ocfs2_split_extent(handle_t *handle, + struct ocfs2_extent_tree *et, + struct ocfs2_path *path, + int split_index, + struct ocfs2_extent_rec *split_rec, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_cached_dealloc_ctxt *dealloc) { int ret = 0; struct ocfs2_extent_list *el = path_leaf_el(path); @@ -4913,12 +5025,6 @@ static int __ocfs2_mark_extent_written(struct inode *inode, struct ocfs2_merge_ctxt ctxt; struct ocfs2_extent_list *rightmost_el; - if (!(rec->e_flags & OCFS2_EXT_UNWRITTEN)) { - ret = -EIO; - mlog_errno(ret); - goto out; - } - if (le32_to_cpu(rec->e_cpos) > le32_to_cpu(split_rec->e_cpos) || ((le32_to_cpu(rec->e_cpos) + le16_to_cpu(rec->e_leaf_clusters)) < (le32_to_cpu(split_rec->e_cpos) + le16_to_cpu(split_rec->e_leaf_clusters)))) { @@ -4927,23 +5033,23 @@ static int __ocfs2_mark_extent_written(struct inode *inode, goto out; } - ctxt.c_contig_type = ocfs2_figure_merge_contig_type(inode, path, el, + ctxt.c_contig_type = ocfs2_figure_merge_contig_type(et, path, el, split_index, split_rec); /* * The core merge / split code wants to know how much room is - * left in this inodes allocation tree, so we pass the + * left in this allocation tree, so we pass the * rightmost extent list. */ if (path->p_tree_depth) { struct ocfs2_extent_block *eb; - ret = ocfs2_read_extent_block(inode, + ret = ocfs2_read_extent_block(et->et_ci, ocfs2_et_get_last_eb_blk(et), &last_eb_bh); if (ret) { - mlog_exit(ret); + mlog_errno(ret); goto out; } @@ -4960,25 +5066,24 @@ static int __ocfs2_mark_extent_written(struct inode *inode, ctxt.c_has_empty_extent = ocfs2_is_empty_extent(&el->l_recs[0]); - mlog(0, "index: %d, contig: %u, has_empty: %u, split_covers: %u\n", - split_index, ctxt.c_contig_type, ctxt.c_has_empty_extent, - ctxt.c_split_covers_rec); + trace_ocfs2_split_extent(split_index, ctxt.c_contig_type, + ctxt.c_has_empty_extent, + ctxt.c_split_covers_rec); if (ctxt.c_contig_type == CONTIG_NONE) { if (ctxt.c_split_covers_rec) - ret = ocfs2_replace_extent_rec(inode, handle, - path, el, + ret = ocfs2_replace_extent_rec(handle, et, path, el, split_index, split_rec); else - ret = ocfs2_split_and_insert(inode, handle, path, et, + ret = ocfs2_split_and_insert(handle, et, path, &last_eb_bh, split_index, split_rec, meta_ac); if (ret) mlog_errno(ret); } else { - ret = ocfs2_try_to_merge_extent(inode, handle, path, + ret = ocfs2_try_to_merge_extent(handle, et, path, split_index, split_rec, - dealloc, &ctxt, et); + dealloc, &ctxt); if (ret) mlog_errno(ret); } @@ -4989,46 +5094,31 @@ out: } /* - * Mark the already-existing extent at cpos as written for len clusters. + * Change the flags of the already-existing extent at cpos for len clusters. + * + * new_flags: the flags we want to set. + * clear_flags: the flags we want to clear. + * phys: the new physical offset we want this new extent starts from. * * If the existing extent is larger than the request, initiate a * split. An attempt will be made at merging with adjacent extents. * * The caller is responsible for passing down meta_ac if we'll need it. */ -int ocfs2_mark_extent_written(struct inode *inode, - struct ocfs2_extent_tree *et, - handle_t *handle, u32 cpos, u32 len, u32 phys, - struct ocfs2_alloc_context *meta_ac, - struct ocfs2_cached_dealloc_ctxt *dealloc) +int ocfs2_change_extent_flag(handle_t *handle, + struct ocfs2_extent_tree *et, + u32 cpos, u32 len, u32 phys, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_cached_dealloc_ctxt *dealloc, + int new_flags, int clear_flags) { int ret, index; - u64 start_blkno = ocfs2_clusters_to_blocks(inode->i_sb, phys); + struct super_block *sb = ocfs2_metadata_cache_get_super(et->et_ci); + u64 start_blkno = ocfs2_clusters_to_blocks(sb, phys); struct ocfs2_extent_rec split_rec; struct ocfs2_path *left_path = NULL; struct ocfs2_extent_list *el; - - mlog(0, "Inode %lu cpos %u, len %u, phys %u (%llu)\n", - inode->i_ino, cpos, len, phys, (unsigned long long)start_blkno); - - if (!ocfs2_writes_unwritten_extents(OCFS2_SB(inode->i_sb))) { - ocfs2_error(inode->i_sb, "Inode %llu has unwritten extents " - "that are being written to, but the feature bit " - "is not set in the super block.", - (unsigned long long)OCFS2_I(inode)->ip_blkno); - ret = -EROFS; - goto out; - } - - /* - * XXX: This should be fixed up so that we just re-insert the - * next extent records. - * - * XXX: This is a hack on the extent tree, maybe it should be - * an op? - */ - if (et->et_ops == &ocfs2_dinode_et_ops) - ocfs2_extent_map_trunc(inode, 0); + struct ocfs2_extent_rec *rec; left_path = ocfs2_new_path_from_et(et); if (!left_path) { @@ -5037,7 +5127,7 @@ int ocfs2_mark_extent_written(struct inode *inode, goto out; } - ret = ocfs2_find_path(inode, left_path, cpos); + ret = ocfs2_find_path(et->et_ci, left_path, cpos); if (ret) { mlog_errno(ret); goto out; @@ -5046,38 +5136,107 @@ int ocfs2_mark_extent_written(struct inode *inode, index = ocfs2_search_extent_list(el, cpos); if (index == -1 || index >= le16_to_cpu(el->l_next_free_rec)) { - ocfs2_error(inode->i_sb, - "Inode %llu has an extent at cpos %u which can no " + ocfs2_error(sb, + "Owner %llu has an extent at cpos %u which can no " "longer be found.\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, cpos); + (unsigned long long) + ocfs2_metadata_cache_owner(et->et_ci), cpos); ret = -EROFS; goto out; } + ret = -EIO; + rec = &el->l_recs[index]; + if (new_flags && (rec->e_flags & new_flags)) { + mlog(ML_ERROR, "Owner %llu tried to set %d flags on an " + "extent that already had them", + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci), + new_flags); + goto out; + } + + if (clear_flags && !(rec->e_flags & clear_flags)) { + mlog(ML_ERROR, "Owner %llu tried to clear %d flags on an " + "extent that didn't have them", + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci), + clear_flags); + goto out; + } + memset(&split_rec, 0, sizeof(struct ocfs2_extent_rec)); split_rec.e_cpos = cpu_to_le32(cpos); split_rec.e_leaf_clusters = cpu_to_le16(len); split_rec.e_blkno = cpu_to_le64(start_blkno); - split_rec.e_flags = path_leaf_el(left_path)->l_recs[index].e_flags; - split_rec.e_flags &= ~OCFS2_EXT_UNWRITTEN; - - ret = __ocfs2_mark_extent_written(inode, et, handle, left_path, - index, &split_rec, meta_ac, - dealloc); + split_rec.e_flags = rec->e_flags; + if (new_flags) + split_rec.e_flags |= new_flags; + if (clear_flags) + split_rec.e_flags &= ~clear_flags; + + ret = ocfs2_split_extent(handle, et, left_path, + index, &split_rec, meta_ac, + dealloc); if (ret) mlog_errno(ret); out: ocfs2_free_path(left_path); return ret; + +} + +/* + * Mark the already-existing extent at cpos as written for len clusters. + * This removes the unwritten extent flag. + * + * If the existing extent is larger than the request, initiate a + * split. An attempt will be made at merging with adjacent extents. + * + * The caller is responsible for passing down meta_ac if we'll need it. + */ +int ocfs2_mark_extent_written(struct inode *inode, + struct ocfs2_extent_tree *et, + handle_t *handle, u32 cpos, u32 len, u32 phys, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_cached_dealloc_ctxt *dealloc) +{ + int ret; + + trace_ocfs2_mark_extent_written( + (unsigned long long)OCFS2_I(inode)->ip_blkno, + cpos, len, phys); + + if (!ocfs2_writes_unwritten_extents(OCFS2_SB(inode->i_sb))) { + ocfs2_error(inode->i_sb, "Inode %llu has unwritten extents " + "that are being written to, but the feature bit " + "is not set in the super block.", + (unsigned long long)OCFS2_I(inode)->ip_blkno); + ret = -EROFS; + goto out; + } + + /* + * XXX: This should be fixed up so that we just re-insert the + * next extent records. + */ + ocfs2_et_extent_map_truncate(et, 0); + + ret = ocfs2_change_extent_flag(handle, et, cpos, + len, phys, meta_ac, dealloc, + 0, OCFS2_EXT_UNWRITTEN); + if (ret) + mlog_errno(ret); + +out: + return ret; } -static int ocfs2_split_tree(struct inode *inode, struct ocfs2_extent_tree *et, - handle_t *handle, struct ocfs2_path *path, +static int ocfs2_split_tree(handle_t *handle, struct ocfs2_extent_tree *et, + struct ocfs2_path *path, int index, u32 new_range, struct ocfs2_alloc_context *meta_ac) { - int ret, depth, credits = handle->h_buffer_credits; + int ret, depth, credits; struct buffer_head *last_eb_bh = NULL; struct ocfs2_extent_block *eb; struct ocfs2_extent_list *rightmost_el, *el; @@ -5090,11 +5249,12 @@ static int ocfs2_split_tree(struct inode *inode, struct ocfs2_extent_tree *et, */ el = path_leaf_el(path); rec = &el->l_recs[index]; - ocfs2_make_right_split_rec(inode->i_sb, &split_rec, new_range, rec); + ocfs2_make_right_split_rec(ocfs2_metadata_cache_get_super(et->et_ci), + &split_rec, new_range, rec); depth = path->p_tree_depth; if (depth > 0) { - ret = ocfs2_read_extent_block(inode, + ret = ocfs2_read_extent_block(et->et_ci, ocfs2_et_get_last_eb_blk(et), &last_eb_bh); if (ret < 0) { @@ -5107,8 +5267,8 @@ static int ocfs2_split_tree(struct inode *inode, struct ocfs2_extent_tree *et, } else rightmost_el = path_leaf_el(path); - credits += path->p_tree_depth + - ocfs2_extend_meta_needed(et->et_root_el); + credits = path->p_tree_depth + + ocfs2_extend_meta_needed(et->et_root_el); ret = ocfs2_extend_trans(handle, credits); if (ret) { mlog_errno(ret); @@ -5117,7 +5277,7 @@ static int ocfs2_split_tree(struct inode *inode, struct ocfs2_extent_tree *et, if (le16_to_cpu(rightmost_el->l_next_free_rec) == le16_to_cpu(rightmost_el->l_count)) { - ret = ocfs2_grow_tree(inode, handle, et, &depth, &last_eb_bh, + ret = ocfs2_grow_tree(handle, et, &depth, &last_eb_bh, meta_ac); if (ret) { mlog_errno(ret); @@ -5131,7 +5291,7 @@ static int ocfs2_split_tree(struct inode *inode, struct ocfs2_extent_tree *et, insert.ins_split = SPLIT_RIGHT; insert.ins_tree_depth = depth; - ret = ocfs2_do_insert_extent(inode, handle, et, &split_rec, &insert); + ret = ocfs2_do_insert_extent(handle, et, &split_rec, &insert); if (ret) mlog_errno(ret); @@ -5140,23 +5300,23 @@ out: return ret; } -static int ocfs2_truncate_rec(struct inode *inode, handle_t *handle, +static int ocfs2_truncate_rec(handle_t *handle, + struct ocfs2_extent_tree *et, struct ocfs2_path *path, int index, struct ocfs2_cached_dealloc_ctxt *dealloc, - u32 cpos, u32 len, - struct ocfs2_extent_tree *et) + u32 cpos, u32 len) { int ret; u32 left_cpos, rec_range, trunc_range; int wants_rotate = 0, is_rightmost_tree_rec = 0; - struct super_block *sb = inode->i_sb; + struct super_block *sb = ocfs2_metadata_cache_get_super(et->et_ci); struct ocfs2_path *left_path = NULL; struct ocfs2_extent_list *el = path_leaf_el(path); struct ocfs2_extent_rec *rec; struct ocfs2_extent_block *eb; if (ocfs2_is_empty_extent(&el->l_recs[0]) && index > 0) { - ret = ocfs2_rotate_tree_left(inode, handle, path, dealloc, et); + ret = ocfs2_rotate_tree_left(handle, et, path, dealloc); if (ret) { mlog_errno(ret); goto out; @@ -5188,14 +5348,13 @@ static int ocfs2_truncate_rec(struct inode *inode, handle_t *handle, * by this leaf and the one to it's left. * * There are two cases we can skip: - * 1) Path is the leftmost one in our inode tree. + * 1) Path is the leftmost one in our btree. * 2) The leaf is rightmost and will be empty after * we remove the extent record - the rotate code * knows how to update the newly formed edge. */ - ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, path, - &left_cpos); + ret = ocfs2_find_cpos_for_left_leaf(sb, path, &left_cpos); if (ret) { mlog_errno(ret); goto out; @@ -5209,7 +5368,8 @@ static int ocfs2_truncate_rec(struct inode *inode, handle_t *handle, goto out; } - ret = ocfs2_find_path(inode, left_path, left_cpos); + ret = ocfs2_find_path(et->et_ci, left_path, + left_cpos); if (ret) { mlog_errno(ret); goto out; @@ -5225,13 +5385,13 @@ static int ocfs2_truncate_rec(struct inode *inode, handle_t *handle, goto out; } - ret = ocfs2_journal_access_path(inode, handle, path); + ret = ocfs2_journal_access_path(et->et_ci, handle, path); if (ret) { mlog_errno(ret); goto out; } - ret = ocfs2_journal_access_path(inode, handle, left_path); + ret = ocfs2_journal_access_path(et->et_ci, handle, left_path); if (ret) { mlog_errno(ret); goto out; @@ -5254,7 +5414,7 @@ static int ocfs2_truncate_rec(struct inode *inode, handle_t *handle, * be deleted by the rotate code. */ rec = &el->l_recs[next_free - 1]; - ocfs2_adjust_rightmost_records(inode, handle, path, + ocfs2_adjust_rightmost_records(handle, et, path, rec); } } else if (le32_to_cpu(rec->e_cpos) == cpos) { @@ -5266,11 +5426,12 @@ static int ocfs2_truncate_rec(struct inode *inode, handle_t *handle, /* Remove rightmost portion of the record */ le16_add_cpu(&rec->e_leaf_clusters, -len); if (is_rightmost_tree_rec) - ocfs2_adjust_rightmost_records(inode, handle, path, rec); + ocfs2_adjust_rightmost_records(handle, et, path, rec); } else { /* Caller should have trapped this. */ - mlog(ML_ERROR, "Inode %llu: Invalid record truncate: (%u, %u) " - "(%u, %u)\n", (unsigned long long)OCFS2_I(inode)->ip_blkno, + mlog(ML_ERROR, "Owner %llu: Invalid record truncate: (%u, %u) " + "(%u, %u)\n", + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci), le32_to_cpu(rec->e_cpos), le16_to_cpu(rec->e_leaf_clusters), cpos, len); BUG(); @@ -5279,14 +5440,14 @@ static int ocfs2_truncate_rec(struct inode *inode, handle_t *handle, if (left_path) { int subtree_index; - subtree_index = ocfs2_find_subtree_root(inode, left_path, path); - ocfs2_complete_edge_insert(inode, handle, left_path, path, + subtree_index = ocfs2_find_subtree_root(et, left_path, path); + ocfs2_complete_edge_insert(handle, left_path, path, subtree_index); } ocfs2_journal_dirty(handle, path_leaf_bh(path)); - ret = ocfs2_rotate_tree_left(inode, handle, path, dealloc, et); + ret = ocfs2_rotate_tree_left(handle, et, path, dealloc); if (ret) { mlog_errno(ret); goto out; @@ -5297,9 +5458,9 @@ out: return ret; } -int ocfs2_remove_extent(struct inode *inode, +int ocfs2_remove_extent(handle_t *handle, struct ocfs2_extent_tree *et, - u32 cpos, u32 len, handle_t *handle, + u32 cpos, u32 len, struct ocfs2_alloc_context *meta_ac, struct ocfs2_cached_dealloc_ctxt *dealloc) { @@ -5309,7 +5470,11 @@ int ocfs2_remove_extent(struct inode *inode, struct ocfs2_extent_list *el; struct ocfs2_path *path = NULL; - ocfs2_extent_map_trunc(inode, 0); + /* + * XXX: Why are we truncating to 0 instead of wherever this + * affects us? + */ + ocfs2_et_extent_map_truncate(et, 0); path = ocfs2_new_path_from_et(et); if (!path) { @@ -5318,7 +5483,7 @@ int ocfs2_remove_extent(struct inode *inode, goto out; } - ret = ocfs2_find_path(inode, path, cpos); + ret = ocfs2_find_path(et->et_ci, path, cpos); if (ret) { mlog_errno(ret); goto out; @@ -5327,10 +5492,11 @@ int ocfs2_remove_extent(struct inode *inode, el = path_leaf_el(path); index = ocfs2_search_extent_list(el, cpos); if (index == -1 || index >= le16_to_cpu(el->l_next_free_rec)) { - ocfs2_error(inode->i_sb, - "Inode %llu has an extent at cpos %u which can no " + ocfs2_error(ocfs2_metadata_cache_get_super(et->et_ci), + "Owner %llu has an extent at cpos %u which can no " "longer be found.\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, cpos); + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci), + cpos); ret = -EROFS; goto out; } @@ -5357,20 +5523,20 @@ int ocfs2_remove_extent(struct inode *inode, BUG_ON(cpos < le32_to_cpu(rec->e_cpos) || trunc_range > rec_range); - mlog(0, "Inode %llu, remove (cpos %u, len %u). Existing index %d " - "(cpos %u, len %u)\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, cpos, len, index, - le32_to_cpu(rec->e_cpos), ocfs2_rec_clusters(el, rec)); + trace_ocfs2_remove_extent( + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci), + cpos, len, index, le32_to_cpu(rec->e_cpos), + ocfs2_rec_clusters(el, rec)); if (le32_to_cpu(rec->e_cpos) == cpos || rec_range == trunc_range) { - ret = ocfs2_truncate_rec(inode, handle, path, index, dealloc, - cpos, len, et); + ret = ocfs2_truncate_rec(handle, et, path, index, dealloc, + cpos, len); if (ret) { mlog_errno(ret); goto out; } } else { - ret = ocfs2_split_tree(inode, et, handle, path, index, + ret = ocfs2_split_tree(handle, et, path, index, trunc_range, meta_ac); if (ret) { mlog_errno(ret); @@ -5383,7 +5549,7 @@ int ocfs2_remove_extent(struct inode *inode, */ ocfs2_reinit_path(path, 1); - ret = ocfs2_find_path(inode, path, cpos); + ret = ocfs2_find_path(et->et_ci, path, cpos); if (ret) { mlog_errno(ret); goto out; @@ -5392,9 +5558,9 @@ int ocfs2_remove_extent(struct inode *inode, el = path_leaf_el(path); index = ocfs2_search_extent_list(el, cpos); if (index == -1 || index >= le16_to_cpu(el->l_next_free_rec)) { - ocfs2_error(inode->i_sb, - "Inode %llu: split at cpos %u lost record.", - (unsigned long long)OCFS2_I(inode)->ip_blkno, + ocfs2_error(ocfs2_metadata_cache_get_super(et->et_ci), + "Owner %llu: split at cpos %u lost record.", + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci), cpos); ret = -EROFS; goto out; @@ -5408,18 +5574,18 @@ int ocfs2_remove_extent(struct inode *inode, rec_range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec); if (rec_range != trunc_range) { - ocfs2_error(inode->i_sb, - "Inode %llu: error after split at cpos %u" + ocfs2_error(ocfs2_metadata_cache_get_super(et->et_ci), + "Owner %llu: error after split at cpos %u" "trunc len %u, existing record is (%u,%u)", - (unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci), cpos, len, le32_to_cpu(rec->e_cpos), ocfs2_rec_clusters(el, rec)); ret = -EROFS; goto out; } - ret = ocfs2_truncate_rec(inode, handle, path, index, dealloc, - cpos, len, et); + ret = ocfs2_truncate_rec(handle, et, path, index, dealloc, + cpos, len); if (ret) { mlog_errno(ret); goto out; @@ -5431,22 +5597,100 @@ out: return ret; } +/* + * ocfs2_reserve_blocks_for_rec_trunc() would look basically the + * same as ocfs2_lock_alloctors(), except for it accepts a blocks + * number to reserve some extra blocks, and it only handles meta + * data allocations. + * + * Currently, only ocfs2_remove_btree_range() uses it for truncating + * and punching holes. + */ +static int ocfs2_reserve_blocks_for_rec_trunc(struct inode *inode, + struct ocfs2_extent_tree *et, + u32 extents_to_split, + struct ocfs2_alloc_context **ac, + int extra_blocks) +{ + int ret = 0, num_free_extents; + unsigned int max_recs_needed = 2 * extents_to_split; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + + *ac = NULL; + + num_free_extents = ocfs2_num_free_extents(osb, et); + if (num_free_extents < 0) { + ret = num_free_extents; + mlog_errno(ret); + goto out; + } + + if (!num_free_extents || + (ocfs2_sparse_alloc(osb) && num_free_extents < max_recs_needed)) + extra_blocks += ocfs2_extend_meta_needed(et->et_root_el); + + if (extra_blocks) { + ret = ocfs2_reserve_new_metadata_blocks(osb, extra_blocks, ac); + if (ret < 0) { + if (ret != -ENOSPC) + mlog_errno(ret); + goto out; + } + } + +out: + if (ret) { + if (*ac) { + ocfs2_free_alloc_context(*ac); + *ac = NULL; + } + } + + return ret; +} + int ocfs2_remove_btree_range(struct inode *inode, struct ocfs2_extent_tree *et, - u32 cpos, u32 phys_cpos, u32 len, - struct ocfs2_cached_dealloc_ctxt *dealloc) + u32 cpos, u32 phys_cpos, u32 len, int flags, + struct ocfs2_cached_dealloc_ctxt *dealloc, + u64 refcount_loc) { - int ret; + int ret, credits = 0, extra_blocks = 0; u64 phys_blkno = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos); struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); struct inode *tl_inode = osb->osb_tl_inode; handle_t *handle; struct ocfs2_alloc_context *meta_ac = NULL; + struct ocfs2_refcount_tree *ref_tree = NULL; + + if ((flags & OCFS2_EXT_REFCOUNTED) && len) { + BUG_ON(!(OCFS2_I(inode)->ip_dyn_features & + OCFS2_HAS_REFCOUNT_FL)); - ret = ocfs2_lock_allocators(inode, et, 0, 1, NULL, &meta_ac); + ret = ocfs2_lock_refcount_tree(osb, refcount_loc, 1, + &ref_tree, NULL); + if (ret) { + mlog_errno(ret); + goto bail; + } + + ret = ocfs2_prepare_refcount_change_for_del(inode, + refcount_loc, + phys_blkno, + len, + &credits, + &extra_blocks); + if (ret < 0) { + mlog_errno(ret); + goto bail; + } + } + + ret = ocfs2_reserve_blocks_for_rec_trunc(inode, et, 1, &meta_ac, + extra_blocks); if (ret) { mlog_errno(ret); - return ret; + goto bail; } mutex_lock(&tl_inode->i_mutex); @@ -5459,50 +5703,61 @@ int ocfs2_remove_btree_range(struct inode *inode, } } - handle = ocfs2_start_trans(osb, ocfs2_remove_extent_credits(osb->sb)); + handle = ocfs2_start_trans(osb, + ocfs2_remove_extent_credits(osb->sb) + credits); if (IS_ERR(handle)) { ret = PTR_ERR(handle); mlog_errno(ret); goto out; } - ret = ocfs2_et_root_journal_access(handle, inode, et, + ret = ocfs2_et_root_journal_access(handle, et, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); - goto out; + goto out_commit; } - vfs_dq_free_space_nodirty(inode, + dquot_free_space_nodirty(inode, ocfs2_clusters_to_bytes(inode->i_sb, len)); - ret = ocfs2_remove_extent(inode, et, cpos, len, handle, meta_ac, - dealloc); + ret = ocfs2_remove_extent(handle, et, cpos, len, meta_ac, dealloc); if (ret) { mlog_errno(ret); goto out_commit; } - ocfs2_et_update_clusters(inode, et, -len); + ocfs2_et_update_clusters(et, -len); + ocfs2_update_inode_fsync_trans(handle, inode, 1); - ret = ocfs2_journal_dirty(handle, et->et_root_bh); - if (ret) { - mlog_errno(ret); - goto out_commit; - } + ocfs2_journal_dirty(handle, et->et_root_bh); - ret = ocfs2_truncate_log_append(osb, handle, phys_blkno, len); - if (ret) - mlog_errno(ret); + if (phys_blkno) { + if (flags & OCFS2_EXT_REFCOUNTED) + ret = ocfs2_decrease_refcount(inode, handle, + ocfs2_blocks_to_clusters(osb->sb, + phys_blkno), + len, meta_ac, + dealloc, 1); + else + ret = ocfs2_truncate_log_append(osb, handle, + phys_blkno, len); + if (ret) + mlog_errno(ret); + + } out_commit: ocfs2_commit_trans(osb, handle); out: mutex_unlock(&tl_inode->i_mutex); - +bail: if (meta_ac) ocfs2_free_alloc_context(meta_ac); + if (ref_tree) + ocfs2_unlock_refcount_tree(osb, ref_tree, 1); + return ret; } @@ -5551,9 +5806,6 @@ int ocfs2_truncate_log_append(struct ocfs2_super *osb, struct ocfs2_dinode *di; struct ocfs2_truncate_log *tl; - mlog_entry("start_blk = %llu, num_clusters = %u\n", - (unsigned long long)start_blk, num_clusters); - BUG_ON(mutex_trylock(&tl_inode->i_mutex)); start_cluster = ocfs2_blocks_to_clusters(osb->sb, start_blk); @@ -5583,17 +5835,16 @@ int ocfs2_truncate_log_append(struct ocfs2_super *osb, goto bail; } - status = ocfs2_journal_access_di(handle, tl_inode, tl_bh, + status = ocfs2_journal_access_di(handle, INODE_CACHE(tl_inode), tl_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); goto bail; } - mlog(0, "Log truncate of %u clusters starting at cluster %u to " - "%llu (index = %d)\n", num_clusters, start_cluster, - (unsigned long long)OCFS2_I(tl_inode)->ip_blkno, index); - + trace_ocfs2_truncate_log_append( + (unsigned long long)OCFS2_I(tl_inode)->ip_blkno, index, + start_cluster, num_clusters); if (ocfs2_truncate_log_can_coalesce(tl, start_cluster)) { /* * Move index back to the record we are coalescing with. @@ -5602,23 +5853,20 @@ int ocfs2_truncate_log_append(struct ocfs2_super *osb, index--; num_clusters += le32_to_cpu(tl->tl_recs[index].t_clusters); - mlog(0, "Coalesce with index %u (start = %u, clusters = %u)\n", - index, le32_to_cpu(tl->tl_recs[index].t_start), - num_clusters); + trace_ocfs2_truncate_log_append( + (unsigned long long)OCFS2_I(tl_inode)->ip_blkno, + index, le32_to_cpu(tl->tl_recs[index].t_start), + num_clusters); } else { tl->tl_recs[index].t_start = cpu_to_le32(start_cluster); tl->tl_used = cpu_to_le16(index + 1); } tl->tl_recs[index].t_clusters = cpu_to_le32(num_clusters); - status = ocfs2_journal_dirty(handle, tl_bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } + ocfs2_journal_dirty(handle, tl_bh); + osb->truncated_clusters += num_clusters; bail: - mlog_exit(status); return status; } @@ -5637,15 +5885,13 @@ static int ocfs2_replay_truncate_records(struct ocfs2_super *osb, struct inode *tl_inode = osb->osb_tl_inode; struct buffer_head *tl_bh = osb->osb_tl_bh; - mlog_entry_void(); - di = (struct ocfs2_dinode *) tl_bh->b_data; tl = &di->id2.i_dealloc; i = le16_to_cpu(tl->tl_used) - 1; while (i >= 0) { /* Caller has given us at least enough credits to * update the truncate log dinode */ - status = ocfs2_journal_access_di(handle, tl_inode, tl_bh, + status = ocfs2_journal_access_di(handle, INODE_CACHE(tl_inode), tl_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); @@ -5654,11 +5900,7 @@ static int ocfs2_replay_truncate_records(struct ocfs2_super *osb, tl->tl_used = cpu_to_le16(i); - status = ocfs2_journal_dirty(handle, tl_bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } + ocfs2_journal_dirty(handle, tl_bh); /* TODO: Perhaps we can calculate the bulk of the * credits up front rather than extending like @@ -5678,8 +5920,9 @@ static int ocfs2_replay_truncate_records(struct ocfs2_super *osb, /* if start_blk is not set, we ignore the record as * invalid. */ if (start_blk) { - mlog(0, "free record %d, start = %u, clusters = %u\n", - i, le32_to_cpu(rec.t_start), num_clusters); + trace_ocfs2_replay_truncate_records( + (unsigned long long)OCFS2_I(tl_inode)->ip_blkno, + i, le32_to_cpu(rec.t_start), num_clusters); status = ocfs2_free_clusters(handle, data_alloc_inode, data_alloc_bh, start_blk, @@ -5692,8 +5935,9 @@ static int ocfs2_replay_truncate_records(struct ocfs2_super *osb, i--; } + osb->truncated_clusters = 0; + bail: - mlog_exit(status); return status; } @@ -5710,8 +5954,6 @@ int __ocfs2_flush_truncate_log(struct ocfs2_super *osb) struct ocfs2_dinode *di; struct ocfs2_truncate_log *tl; - mlog_entry_void(); - BUG_ON(mutex_trylock(&tl_inode->i_mutex)); di = (struct ocfs2_dinode *) tl_bh->b_data; @@ -5723,8 +5965,9 @@ int __ocfs2_flush_truncate_log(struct ocfs2_super *osb) tl = &di->id2.i_dealloc; num_to_flush = le16_to_cpu(tl->tl_used); - mlog(0, "Flush %u records from truncate log #%llu\n", - num_to_flush, (unsigned long long)OCFS2_I(tl_inode)->ip_blkno); + trace_ocfs2_flush_truncate_log( + (unsigned long long)OCFS2_I(tl_inode)->ip_blkno, + num_to_flush); if (!num_to_flush) { status = 0; goto out; @@ -5770,7 +6013,6 @@ out_mutex: iput(data_alloc_inode); out: - mlog_exit(status); return status; } @@ -5793,22 +6035,19 @@ static void ocfs2_truncate_log_worker(struct work_struct *work) container_of(work, struct ocfs2_super, osb_truncate_log_wq.work); - mlog_entry_void(); - status = ocfs2_flush_truncate_log(osb); if (status < 0) mlog_errno(status); else - ocfs2_init_inode_steal_slot(osb); - - mlog_exit(status); + ocfs2_init_steal_slots(osb); } #define OCFS2_TRUNCATE_LOG_FLUSH_INTERVAL (2 * HZ) void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb, int cancel) { - if (osb->osb_tl_inode) { + if (osb->osb_tl_inode && + atomic_read(&osb->osb_tl_disable) == 0) { /* We want to push off log flushes while truncates are * still running. */ if (cancel) @@ -5847,7 +6086,6 @@ static int ocfs2_get_truncate_log_info(struct ocfs2_super *osb, *tl_inode = inode; *tl_bh = bh; bail: - mlog_exit(status); return status; } @@ -5867,7 +6105,7 @@ int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb, *tl_copy = NULL; - mlog(0, "recover truncate log from slot %d\n", slot_num); + trace_ocfs2_begin_truncate_log_recovery(slot_num); status = ocfs2_get_truncate_log_info(osb, slot_num, &tl_inode, &tl_bh); if (status < 0) { @@ -5884,8 +6122,7 @@ int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb, tl = &di->id2.i_dealloc; if (le16_to_cpu(tl->tl_used)) { - mlog(0, "We'll have %u logs to recover\n", - le16_to_cpu(tl->tl_used)); + trace_ocfs2_truncate_log_recovery_num(le16_to_cpu(tl->tl_used)); *tl_copy = kmalloc(tl_bh->b_size, GFP_KERNEL); if (!(*tl_copy)) { @@ -5903,7 +6140,7 @@ int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb, tl->tl_used = 0; ocfs2_compute_meta_ecc(osb->sb, tl_bh->b_data, &di->i_check); - status = ocfs2_write_block(osb, tl_bh, tl_inode); + status = ocfs2_write_block(osb, tl_bh, INODE_CACHE(tl_inode)); if (status < 0) { mlog_errno(status); goto bail; @@ -5918,9 +6155,9 @@ bail: if (status < 0 && (*tl_copy)) { kfree(*tl_copy); *tl_copy = NULL; + mlog_errno(status); } - mlog_exit(status); return status; } @@ -5935,8 +6172,6 @@ int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb, struct inode *tl_inode = osb->osb_tl_inode; struct ocfs2_truncate_log *tl; - mlog_entry_void(); - if (OCFS2_I(tl_inode)->ip_blkno == le64_to_cpu(tl_copy->i_blkno)) { mlog(ML_ERROR, "Asked to recover my own truncate log!\n"); return -EINVAL; @@ -5944,8 +6179,9 @@ int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb, tl = &tl_copy->id2.i_dealloc; num_recs = le16_to_cpu(tl->tl_used); - mlog(0, "cleanup %u records from %llu\n", num_recs, - (unsigned long long)le64_to_cpu(tl_copy->i_blkno)); + trace_ocfs2_complete_truncate_log_recovery( + (unsigned long long)le64_to_cpu(tl_copy->i_blkno), + num_recs); mutex_lock(&tl_inode->i_mutex); for(i = 0; i < num_recs; i++) { @@ -5980,7 +6216,6 @@ int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb, bail_up: mutex_unlock(&tl_inode->i_mutex); - mlog_exit(status); return status; } @@ -5989,7 +6224,7 @@ void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb) int status; struct inode *tl_inode = osb->osb_tl_inode; - mlog_entry_void(); + atomic_set(&osb->osb_tl_disable, 1); if (tl_inode) { cancel_delayed_work(&osb->osb_truncate_log_wq); @@ -6002,8 +6237,6 @@ void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb) brelse(osb->osb_tl_bh); iput(osb->osb_tl_inode); } - - mlog_exit_void(); } int ocfs2_truncate_log_init(struct ocfs2_super *osb) @@ -6012,8 +6245,6 @@ int ocfs2_truncate_log_init(struct ocfs2_super *osb) struct inode *tl_inode = NULL; struct buffer_head *tl_bh = NULL; - mlog_entry_void(); - status = ocfs2_get_truncate_log_info(osb, osb->slot_num, &tl_inode, @@ -6026,10 +6257,10 @@ int ocfs2_truncate_log_init(struct ocfs2_super *osb) * until we're sure all is well. */ INIT_DELAYED_WORK(&osb->osb_truncate_log_wq, ocfs2_truncate_log_worker); + atomic_set(&osb->osb_tl_disable, 0); osb->osb_tl_bh = tl_bh; osb->osb_tl_inode = tl_inode; - mlog_exit(status); return status; } @@ -6059,6 +6290,7 @@ int ocfs2_truncate_log_init(struct ocfs2_super *osb) */ struct ocfs2_cached_block_free { struct ocfs2_cached_block_free *free_next; + u64 free_bg; u64 free_blk; unsigned int free_bit; }; @@ -6105,10 +6337,13 @@ static int ocfs2_free_cached_blocks(struct ocfs2_super *osb, } while (head) { - bg_blkno = ocfs2_which_suballoc_group(head->free_blk, - head->free_bit); - mlog(0, "Free bit: (bit %u, blkno %llu)\n", - head->free_bit, (unsigned long long)head->free_blk); + if (head->free_bg) + bg_blkno = head->free_bg; + else + bg_blkno = ocfs2_which_suballoc_group(head->free_blk, + head->free_bit); + trace_ocfs2_free_cached_blocks( + (unsigned long long)head->free_blk, head->free_bit); ret = ocfs2_free_suballoc_bits(handle, inode, di_bh, head->free_bit, bg_blkno, 1); @@ -6154,15 +6389,14 @@ int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt, int ret = 0; struct ocfs2_cached_block_free *item; - item = kmalloc(sizeof(*item), GFP_NOFS); + item = kzalloc(sizeof(*item), GFP_NOFS); if (item == NULL) { ret = -ENOMEM; mlog_errno(ret); return ret; } - mlog(0, "Insert clusters: (bit %u, blk %llu)\n", - bit, (unsigned long long)blkno); + trace_ocfs2_cache_cluster_dealloc((unsigned long long)blkno, bit); item->free_blk = blkno; item->free_bit = bit; @@ -6237,8 +6471,8 @@ int ocfs2_run_deallocs(struct ocfs2_super *osb, fl = ctxt->c_first_suballocator; if (fl->f_first) { - mlog(0, "Free items: (type %u, slot %d)\n", - fl->f_inode_type, fl->f_slot); + trace_ocfs2_run_deallocs(fl->f_inode_type, + fl->f_slot); ret2 = ocfs2_free_cached_blocks(osb, fl->f_inode_type, fl->f_slot, @@ -6293,9 +6527,9 @@ ocfs2_find_per_slot_free_list(int type, return fl; } -static int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt, - int type, int slot, u64 blkno, - unsigned int bit) +int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt, + int type, int slot, u64 suballoc, + u64 blkno, unsigned int bit) { int ret; struct ocfs2_per_slot_free_list *fl; @@ -6308,16 +6542,18 @@ static int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt, goto out; } - item = kmalloc(sizeof(*item), GFP_NOFS); + item = kzalloc(sizeof(*item), GFP_NOFS); if (item == NULL) { ret = -ENOMEM; mlog_errno(ret); goto out; } - mlog(0, "Insert: (type %d, slot %u, bit %u, blk %llu)\n", - type, slot, bit, (unsigned long long)blkno); + trace_ocfs2_cache_block_dealloc(type, slot, + (unsigned long long)suballoc, + (unsigned long long)blkno, bit); + item->free_bg = suballoc; item->free_blk = blkno; item->free_bit = bit; item->free_next = fl->f_first; @@ -6334,421 +6570,11 @@ static int ocfs2_cache_extent_block_free(struct ocfs2_cached_dealloc_ctxt *ctxt, { return ocfs2_cache_block_dealloc(ctxt, EXTENT_ALLOC_SYSTEM_INODE, le16_to_cpu(eb->h_suballoc_slot), + le64_to_cpu(eb->h_suballoc_loc), le64_to_cpu(eb->h_blkno), le16_to_cpu(eb->h_suballoc_bit)); } -/* This function will figure out whether the currently last extent - * block will be deleted, and if it will, what the new last extent - * block will be so we can update his h_next_leaf_blk field, as well - * as the dinodes i_last_eb_blk */ -static int ocfs2_find_new_last_ext_blk(struct inode *inode, - unsigned int clusters_to_del, - struct ocfs2_path *path, - struct buffer_head **new_last_eb) -{ - int next_free, ret = 0; - u32 cpos; - struct ocfs2_extent_rec *rec; - struct ocfs2_extent_block *eb; - struct ocfs2_extent_list *el; - struct buffer_head *bh = NULL; - - *new_last_eb = NULL; - - /* we have no tree, so of course, no last_eb. */ - if (!path->p_tree_depth) - goto out; - - /* trunc to zero special case - this makes tree_depth = 0 - * regardless of what it is. */ - if (OCFS2_I(inode)->ip_clusters == clusters_to_del) - goto out; - - el = path_leaf_el(path); - BUG_ON(!el->l_next_free_rec); - - /* - * Make sure that this extent list will actually be empty - * after we clear away the data. We can shortcut out if - * there's more than one non-empty extent in the - * list. Otherwise, a check of the remaining extent is - * necessary. - */ - next_free = le16_to_cpu(el->l_next_free_rec); - rec = NULL; - if (ocfs2_is_empty_extent(&el->l_recs[0])) { - if (next_free > 2) - goto out; - - /* We may have a valid extent in index 1, check it. */ - if (next_free == 2) - rec = &el->l_recs[1]; - - /* - * Fall through - no more nonempty extents, so we want - * to delete this leaf. - */ - } else { - if (next_free > 1) - goto out; - - rec = &el->l_recs[0]; - } - - if (rec) { - /* - * Check it we'll only be trimming off the end of this - * cluster. - */ - if (le16_to_cpu(rec->e_leaf_clusters) > clusters_to_del) - goto out; - } - - ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, path, &cpos); - if (ret) { - mlog_errno(ret); - goto out; - } - - ret = ocfs2_find_leaf(inode, path_root_el(path), cpos, &bh); - if (ret) { - mlog_errno(ret); - goto out; - } - - eb = (struct ocfs2_extent_block *) bh->b_data; - el = &eb->h_list; - - /* ocfs2_find_leaf() gets the eb from ocfs2_read_extent_block(). - * Any corruption is a code bug. */ - BUG_ON(!OCFS2_IS_VALID_EXTENT_BLOCK(eb)); - - *new_last_eb = bh; - get_bh(*new_last_eb); - mlog(0, "returning block %llu, (cpos: %u)\n", - (unsigned long long)le64_to_cpu(eb->h_blkno), cpos); -out: - brelse(bh); - - return ret; -} - -/* - * Trim some clusters off the rightmost edge of a tree. Only called - * during truncate. - * - * The caller needs to: - * - start journaling of each path component. - * - compute and fully set up any new last ext block - */ -static int ocfs2_trim_tree(struct inode *inode, struct ocfs2_path *path, - handle_t *handle, struct ocfs2_truncate_context *tc, - u32 clusters_to_del, u64 *delete_start) -{ - int ret, i, index = path->p_tree_depth; - u32 new_edge = 0; - u64 deleted_eb = 0; - struct buffer_head *bh; - struct ocfs2_extent_list *el; - struct ocfs2_extent_rec *rec; - - *delete_start = 0; - - while (index >= 0) { - bh = path->p_node[index].bh; - el = path->p_node[index].el; - - mlog(0, "traveling tree (index = %d, block = %llu)\n", - index, (unsigned long long)bh->b_blocknr); - - BUG_ON(le16_to_cpu(el->l_next_free_rec) == 0); - - if (index != - (path->p_tree_depth - le16_to_cpu(el->l_tree_depth))) { - ocfs2_error(inode->i_sb, - "Inode %lu has invalid ext. block %llu", - inode->i_ino, - (unsigned long long)bh->b_blocknr); - ret = -EROFS; - goto out; - } - -find_tail_record: - i = le16_to_cpu(el->l_next_free_rec) - 1; - rec = &el->l_recs[i]; - - mlog(0, "Extent list before: record %d: (%u, %u, %llu), " - "next = %u\n", i, le32_to_cpu(rec->e_cpos), - ocfs2_rec_clusters(el, rec), - (unsigned long long)le64_to_cpu(rec->e_blkno), - le16_to_cpu(el->l_next_free_rec)); - - BUG_ON(ocfs2_rec_clusters(el, rec) < clusters_to_del); - - if (le16_to_cpu(el->l_tree_depth) == 0) { - /* - * If the leaf block contains a single empty - * extent and no records, we can just remove - * the block. - */ - if (i == 0 && ocfs2_is_empty_extent(rec)) { - memset(rec, 0, - sizeof(struct ocfs2_extent_rec)); - el->l_next_free_rec = cpu_to_le16(0); - - goto delete; - } - - /* - * Remove any empty extents by shifting things - * left. That should make life much easier on - * the code below. This condition is rare - * enough that we shouldn't see a performance - * hit. - */ - if (ocfs2_is_empty_extent(&el->l_recs[0])) { - le16_add_cpu(&el->l_next_free_rec, -1); - - for(i = 0; - i < le16_to_cpu(el->l_next_free_rec); i++) - el->l_recs[i] = el->l_recs[i + 1]; - - memset(&el->l_recs[i], 0, - sizeof(struct ocfs2_extent_rec)); - - /* - * We've modified our extent list. The - * simplest way to handle this change - * is to being the search from the - * start again. - */ - goto find_tail_record; - } - - le16_add_cpu(&rec->e_leaf_clusters, -clusters_to_del); - - /* - * We'll use "new_edge" on our way back up the - * tree to know what our rightmost cpos is. - */ - new_edge = le16_to_cpu(rec->e_leaf_clusters); - new_edge += le32_to_cpu(rec->e_cpos); - - /* - * The caller will use this to delete data blocks. - */ - *delete_start = le64_to_cpu(rec->e_blkno) - + ocfs2_clusters_to_blocks(inode->i_sb, - le16_to_cpu(rec->e_leaf_clusters)); - - /* - * If it's now empty, remove this record. - */ - if (le16_to_cpu(rec->e_leaf_clusters) == 0) { - memset(rec, 0, - sizeof(struct ocfs2_extent_rec)); - le16_add_cpu(&el->l_next_free_rec, -1); - } - } else { - if (le64_to_cpu(rec->e_blkno) == deleted_eb) { - memset(rec, 0, - sizeof(struct ocfs2_extent_rec)); - le16_add_cpu(&el->l_next_free_rec, -1); - - goto delete; - } - - /* Can this actually happen? */ - if (le16_to_cpu(el->l_next_free_rec) == 0) - goto delete; - - /* - * We never actually deleted any clusters - * because our leaf was empty. There's no - * reason to adjust the rightmost edge then. - */ - if (new_edge == 0) - goto delete; - - rec->e_int_clusters = cpu_to_le32(new_edge); - le32_add_cpu(&rec->e_int_clusters, - -le32_to_cpu(rec->e_cpos)); - - /* - * A deleted child record should have been - * caught above. - */ - BUG_ON(le32_to_cpu(rec->e_int_clusters) == 0); - } - -delete: - ret = ocfs2_journal_dirty(handle, bh); - if (ret) { - mlog_errno(ret); - goto out; - } - - mlog(0, "extent list container %llu, after: record %d: " - "(%u, %u, %llu), next = %u.\n", - (unsigned long long)bh->b_blocknr, i, - le32_to_cpu(rec->e_cpos), ocfs2_rec_clusters(el, rec), - (unsigned long long)le64_to_cpu(rec->e_blkno), - le16_to_cpu(el->l_next_free_rec)); - - /* - * We must be careful to only attempt delete of an - * extent block (and not the root inode block). - */ - if (index > 0 && le16_to_cpu(el->l_next_free_rec) == 0) { - struct ocfs2_extent_block *eb = - (struct ocfs2_extent_block *)bh->b_data; - - /* - * Save this for use when processing the - * parent block. - */ - deleted_eb = le64_to_cpu(eb->h_blkno); - - mlog(0, "deleting this extent block.\n"); - - ocfs2_remove_from_cache(inode, bh); - - BUG_ON(ocfs2_rec_clusters(el, &el->l_recs[0])); - BUG_ON(le32_to_cpu(el->l_recs[0].e_cpos)); - BUG_ON(le64_to_cpu(el->l_recs[0].e_blkno)); - - ret = ocfs2_cache_extent_block_free(&tc->tc_dealloc, eb); - /* An error here is not fatal. */ - if (ret < 0) - mlog_errno(ret); - } else { - deleted_eb = 0; - } - - index--; - } - - ret = 0; -out: - return ret; -} - -static int ocfs2_do_truncate(struct ocfs2_super *osb, - unsigned int clusters_to_del, - struct inode *inode, - struct buffer_head *fe_bh, - handle_t *handle, - struct ocfs2_truncate_context *tc, - struct ocfs2_path *path) -{ - int status; - struct ocfs2_dinode *fe; - struct ocfs2_extent_block *last_eb = NULL; - struct ocfs2_extent_list *el; - struct buffer_head *last_eb_bh = NULL; - u64 delete_blk = 0; - - fe = (struct ocfs2_dinode *) fe_bh->b_data; - - status = ocfs2_find_new_last_ext_blk(inode, clusters_to_del, - path, &last_eb_bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } - - /* - * Each component will be touched, so we might as well journal - * here to avoid having to handle errors later. - */ - status = ocfs2_journal_access_path(inode, handle, path); - if (status < 0) { - mlog_errno(status); - goto bail; - } - - if (last_eb_bh) { - status = ocfs2_journal_access_eb(handle, inode, last_eb_bh, - OCFS2_JOURNAL_ACCESS_WRITE); - if (status < 0) { - mlog_errno(status); - goto bail; - } - - last_eb = (struct ocfs2_extent_block *) last_eb_bh->b_data; - } - - el = &(fe->id2.i_list); - - /* - * Lower levels depend on this never happening, but it's best - * to check it up here before changing the tree. - */ - if (el->l_tree_depth && el->l_recs[0].e_int_clusters == 0) { - ocfs2_error(inode->i_sb, - "Inode %lu has an empty extent record, depth %u\n", - inode->i_ino, le16_to_cpu(el->l_tree_depth)); - status = -EROFS; - goto bail; - } - - vfs_dq_free_space_nodirty(inode, - ocfs2_clusters_to_bytes(osb->sb, clusters_to_del)); - spin_lock(&OCFS2_I(inode)->ip_lock); - OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters) - - clusters_to_del; - spin_unlock(&OCFS2_I(inode)->ip_lock); - le32_add_cpu(&fe->i_clusters, -clusters_to_del); - inode->i_blocks = ocfs2_inode_sector_count(inode); - - status = ocfs2_trim_tree(inode, path, handle, tc, - clusters_to_del, &delete_blk); - if (status) { - mlog_errno(status); - goto bail; - } - - if (le32_to_cpu(fe->i_clusters) == 0) { - /* trunc to zero is a special case. */ - el->l_tree_depth = 0; - fe->i_last_eb_blk = 0; - } else if (last_eb) - fe->i_last_eb_blk = last_eb->h_blkno; - - status = ocfs2_journal_dirty(handle, fe_bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } - - if (last_eb) { - /* If there will be a new last extent block, then by - * definition, there cannot be any leaves to the right of - * him. */ - last_eb->h_next_leaf_blk = 0; - status = ocfs2_journal_dirty(handle, last_eb_bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } - } - - if (delete_blk) { - status = ocfs2_truncate_log_append(osb, handle, delete_blk, - clusters_to_del); - if (status < 0) { - mlog_errno(status); - goto bail; - } - } - status = 0; -bail: - - mlog_exit(status); - return status; -} - static int ocfs2_zero_func(handle_t *handle, struct buffer_head *bh) { set_buffer_uptodate(bh); @@ -6756,9 +6582,9 @@ static int ocfs2_zero_func(handle_t *handle, struct buffer_head *bh) return 0; } -static void ocfs2_map_and_dirty_page(struct inode *inode, handle_t *handle, - unsigned int from, unsigned int to, - struct page *page, int zero, u64 *phys) +void ocfs2_map_and_dirty_page(struct inode *inode, handle_t *handle, + unsigned int from, unsigned int to, + struct page *page, int zero, u64 *phys) { int ret, partial = 0; @@ -6826,25 +6652,21 @@ out: ocfs2_unlock_and_free_pages(pages, numpages); } -static int ocfs2_grab_eof_pages(struct inode *inode, loff_t start, loff_t end, - struct page **pages, int *num) +int ocfs2_grab_pages(struct inode *inode, loff_t start, loff_t end, + struct page **pages, int *num) { int numpages, ret = 0; - struct super_block *sb = inode->i_sb; struct address_space *mapping = inode->i_mapping; unsigned long index; loff_t last_page_bytes; BUG_ON(start > end); - BUG_ON(start >> OCFS2_SB(sb)->s_clustersize_bits != - (end - 1) >> OCFS2_SB(sb)->s_clustersize_bits); - numpages = 0; last_page_bytes = PAGE_ALIGN(end); index = start >> PAGE_CACHE_SHIFT; do { - pages[numpages] = grab_cache_page(mapping, index); + pages[numpages] = find_or_create_page(mapping, index, GFP_NOFS); if (!pages[numpages]) { ret = -ENOMEM; mlog_errno(ret); @@ -6867,6 +6689,17 @@ out: return ret; } +static int ocfs2_grab_eof_pages(struct inode *inode, loff_t start, loff_t end, + struct page **pages, int *num) +{ + struct super_block *sb = inode->i_sb; + + BUG_ON(start >> OCFS2_SB(sb)->s_clustersize_bits != + (end - 1) >> OCFS2_SB(sb)->s_clustersize_bits); + + return ocfs2_grab_pages(inode, start, end, pages, num); +} + /* * Zero the area past i_size but still within an allocated * cluster. This avoids exposing nonzero data on subsequent file @@ -6933,14 +6766,13 @@ int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle, * wait on them - the truncate_inode_pages() call later will * do that for us. */ - ret = do_sync_mapping_range(inode->i_mapping, range_start, - range_end - 1, SYNC_FILE_RANGE_WRITE); + ret = filemap_fdatawrite_range(inode->i_mapping, range_start, + range_end - 1); if (ret) mlog_errno(ret); out: - if (pages) - kfree(pages); + kfree(pages); return ret; } @@ -6994,6 +6826,8 @@ int ocfs2_convert_inline_data_to_extents(struct inode *inode, struct buffer_head *di_bh) { int ret, i, has_data, num_pages = 0; + int need_free = 0; + u32 bit_off, num; handle_t *handle; u64 uninitialized_var(block); struct ocfs2_inode_info *oi = OCFS2_I(inode); @@ -7031,7 +6865,7 @@ int ocfs2_convert_inline_data_to_extents(struct inode *inode, goto out_unlock; } - ret = ocfs2_journal_access_di(handle, inode, di_bh, + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); @@ -7039,18 +6873,18 @@ int ocfs2_convert_inline_data_to_extents(struct inode *inode, } if (has_data) { - u32 bit_off, num; unsigned int page_end; u64 phys; - if (vfs_dq_alloc_space_nodirty(inode, - ocfs2_clusters_to_bytes(osb->sb, 1))) { - ret = -EDQUOT; + ret = dquot_alloc_space_nodirty(inode, + ocfs2_clusters_to_bytes(osb->sb, 1)); + if (ret) goto out_commit; - } did_quota = 1; - ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off, + data_ac->ac_resv = &OCFS2_I(inode)->ip_la_data_resv; + + ret = ocfs2_claim_clusters(handle, data_ac, 1, &bit_off, &num); if (ret) { mlog_errno(ret); @@ -7074,6 +6908,7 @@ int ocfs2_convert_inline_data_to_extents(struct inode *inode, ret = ocfs2_grab_eof_pages(inode, 0, end, pages, &num_pages); if (ret) { mlog_errno(ret); + need_free = 1; goto out_commit; } @@ -7084,6 +6919,7 @@ int ocfs2_convert_inline_data_to_extents(struct inode *inode, ret = ocfs2_read_inline_data(inode, pages[0], di_bh); if (ret) { mlog_errno(ret); + need_free = 1; goto out_commit; } @@ -7101,6 +6937,7 @@ int ocfs2_convert_inline_data_to_extents(struct inode *inode, di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features); spin_unlock(&oi->ip_lock); + ocfs2_update_inode_fsync_trans(handle, inode, 1); ocfs2_dinode_new_extent_list(inode, di); ocfs2_journal_dirty(handle, di_bh); @@ -7111,11 +6948,11 @@ int ocfs2_convert_inline_data_to_extents(struct inode *inode, * this proves to be false, we could always re-build * the in-inode data from our pages. */ - ocfs2_init_dinode_extent_tree(&et, inode, di_bh); - ret = ocfs2_insert_extent(osb, handle, inode, &et, - 0, block, 1, 0, NULL); + ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), di_bh); + ret = ocfs2_insert_extent(handle, &et, 0, block, 1, 0, NULL); if (ret) { mlog_errno(ret); + need_free = 1; goto out_commit; } @@ -7124,9 +6961,21 @@ int ocfs2_convert_inline_data_to_extents(struct inode *inode, out_commit: if (ret < 0 && did_quota) - vfs_dq_free_space_nodirty(inode, + dquot_free_space_nodirty(inode, ocfs2_clusters_to_bytes(osb->sb, 1)); + if (need_free) { + if (data_ac->ac_which == OCFS2_AC_USE_LOCAL) + ocfs2_free_local_alloc_bits(osb, handle, data_ac, + bit_off, num); + else + ocfs2_free_clusters(handle, + data_ac->ac_inode, + data_ac->ac_bh, + ocfs2_clusters_to_blocks(osb->sb, bit_off), + num); + } + ocfs2_commit_trans(osb, handle); out_unlock: @@ -7150,23 +6999,27 @@ out: */ int ocfs2_commit_truncate(struct ocfs2_super *osb, struct inode *inode, - struct buffer_head *fe_bh, - struct ocfs2_truncate_context *tc) + struct buffer_head *di_bh) { - int status, i, credits, tl_sem = 0; - u32 clusters_to_del, new_highest_cpos, range; + int status = 0, i, flags = 0; + u32 new_highest_cpos, range, trunc_cpos, trunc_len, phys_cpos, coff; + u64 blkno = 0; struct ocfs2_extent_list *el; - handle_t *handle = NULL; - struct inode *tl_inode = osb->osb_tl_inode; + struct ocfs2_extent_rec *rec; struct ocfs2_path *path = NULL; - struct ocfs2_dinode *di = (struct ocfs2_dinode *)fe_bh->b_data; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + struct ocfs2_extent_list *root_el = &(di->id2.i_list); + u64 refcount_loc = le64_to_cpu(di->i_refcount_loc); + struct ocfs2_extent_tree et; + struct ocfs2_cached_dealloc_ctxt dealloc; - mlog_entry_void(); + ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), di_bh); + ocfs2_init_dealloc_ctxt(&dealloc); new_highest_cpos = ocfs2_clusters_for_bytes(osb->sb, i_size_read(inode)); - path = ocfs2_new_path(fe_bh, &di->id2.i_list, + path = ocfs2_new_path(di_bh, &di->id2.i_list, ocfs2_journal_access_di); if (!path) { status = -ENOMEM; @@ -7188,14 +7041,17 @@ start: /* * Truncate always works against the rightmost tree branch. */ - status = ocfs2_find_path(inode, path, UINT_MAX); + status = ocfs2_find_path(INODE_CACHE(inode), path, UINT_MAX); if (status) { mlog_errno(status); goto bail; } - mlog(0, "inode->ip_clusters = %u, tree_depth = %u\n", - OCFS2_I(inode)->ip_clusters, path->p_tree_depth); + trace_ocfs2_commit_truncate( + (unsigned long long)OCFS2_I(inode)->ip_blkno, + new_highest_cpos, + OCFS2_I(inode)->ip_clusters, + path->p_tree_depth); /* * By now, el will point to the extent list on the bottom most @@ -7219,61 +7075,60 @@ start: } i = le16_to_cpu(el->l_next_free_rec) - 1; - range = le32_to_cpu(el->l_recs[i].e_cpos) + - ocfs2_rec_clusters(el, &el->l_recs[i]); - if (i == 0 && ocfs2_is_empty_extent(&el->l_recs[i])) { - clusters_to_del = 0; - } else if (le32_to_cpu(el->l_recs[i].e_cpos) >= new_highest_cpos) { - clusters_to_del = ocfs2_rec_clusters(el, &el->l_recs[i]); + rec = &el->l_recs[i]; + flags = rec->e_flags; + range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec); + + if (i == 0 && ocfs2_is_empty_extent(rec)) { + /* + * Lower levels depend on this never happening, but it's best + * to check it up here before changing the tree. + */ + if (root_el->l_tree_depth && rec->e_int_clusters == 0) { + ocfs2_error(inode->i_sb, "Inode %lu has an empty " + "extent record, depth %u\n", inode->i_ino, + le16_to_cpu(root_el->l_tree_depth)); + status = -EROFS; + goto bail; + } + trunc_cpos = le32_to_cpu(rec->e_cpos); + trunc_len = 0; + blkno = 0; + } else if (le32_to_cpu(rec->e_cpos) >= new_highest_cpos) { + /* + * Truncate entire record. + */ + trunc_cpos = le32_to_cpu(rec->e_cpos); + trunc_len = ocfs2_rec_clusters(el, rec); + blkno = le64_to_cpu(rec->e_blkno); } else if (range > new_highest_cpos) { - clusters_to_del = (ocfs2_rec_clusters(el, &el->l_recs[i]) + - le32_to_cpu(el->l_recs[i].e_cpos)) - - new_highest_cpos; + /* + * Partial truncate. it also should be + * the last truncate we're doing. + */ + trunc_cpos = new_highest_cpos; + trunc_len = range - new_highest_cpos; + coff = new_highest_cpos - le32_to_cpu(rec->e_cpos); + blkno = le64_to_cpu(rec->e_blkno) + + ocfs2_clusters_to_blocks(inode->i_sb, coff); } else { + /* + * Truncate completed, leave happily. + */ status = 0; goto bail; } - mlog(0, "clusters_to_del = %u in this pass, tail blk=%llu\n", - clusters_to_del, (unsigned long long)path_leaf_bh(path)->b_blocknr); - - mutex_lock(&tl_inode->i_mutex); - tl_sem = 1; - /* ocfs2_truncate_log_needs_flush guarantees us at least one - * record is free for use. If there isn't any, we flush to get - * an empty truncate log. */ - if (ocfs2_truncate_log_needs_flush(osb)) { - status = __ocfs2_flush_truncate_log(osb); - if (status < 0) { - mlog_errno(status); - goto bail; - } - } - - credits = ocfs2_calc_tree_trunc_credits(osb->sb, clusters_to_del, - (struct ocfs2_dinode *)fe_bh->b_data, - el); - handle = ocfs2_start_trans(osb, credits); - if (IS_ERR(handle)) { - status = PTR_ERR(handle); - handle = NULL; - mlog_errno(status); - goto bail; - } + phys_cpos = ocfs2_blocks_to_clusters(inode->i_sb, blkno); - status = ocfs2_do_truncate(osb, clusters_to_del, inode, fe_bh, handle, - tc, path); + status = ocfs2_remove_btree_range(inode, &et, trunc_cpos, + phys_cpos, trunc_len, flags, &dealloc, + refcount_loc); if (status < 0) { mlog_errno(status); goto bail; } - mutex_unlock(&tl_inode->i_mutex); - tl_sem = 0; - - ocfs2_commit_trans(osb, handle); - handle = NULL; - ocfs2_reinit_path(path, 1); /* @@ -7286,78 +7141,10 @@ bail: ocfs2_schedule_truncate_log_flush(osb, 1); - if (tl_sem) - mutex_unlock(&tl_inode->i_mutex); - - if (handle) - ocfs2_commit_trans(osb, handle); - - ocfs2_run_deallocs(osb, &tc->tc_dealloc); + ocfs2_run_deallocs(osb, &dealloc); ocfs2_free_path(path); - /* This will drop the ext_alloc cluster lock for us */ - ocfs2_free_truncate_context(tc); - - mlog_exit(status); - return status; -} - -/* - * Expects the inode to already be locked. - */ -int ocfs2_prepare_truncate(struct ocfs2_super *osb, - struct inode *inode, - struct buffer_head *fe_bh, - struct ocfs2_truncate_context **tc) -{ - int status; - unsigned int new_i_clusters; - struct ocfs2_dinode *fe; - struct ocfs2_extent_block *eb; - struct buffer_head *last_eb_bh = NULL; - - mlog_entry_void(); - - *tc = NULL; - - new_i_clusters = ocfs2_clusters_for_bytes(osb->sb, - i_size_read(inode)); - fe = (struct ocfs2_dinode *) fe_bh->b_data; - - mlog(0, "fe->i_clusters = %u, new_i_clusters = %u, fe->i_size =" - "%llu\n", le32_to_cpu(fe->i_clusters), new_i_clusters, - (unsigned long long)le64_to_cpu(fe->i_size)); - - *tc = kzalloc(sizeof(struct ocfs2_truncate_context), GFP_KERNEL); - if (!(*tc)) { - status = -ENOMEM; - mlog_errno(status); - goto bail; - } - ocfs2_init_dealloc_ctxt(&(*tc)->tc_dealloc); - - if (fe->id2.i_list.l_tree_depth) { - status = ocfs2_read_extent_block(inode, - le64_to_cpu(fe->i_last_eb_blk), - &last_eb_bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } - eb = (struct ocfs2_extent_block *) last_eb_bh->b_data; - } - - (*tc)->tc_last_eb_bh = last_eb_bh; - - status = 0; -bail: - if (status < 0) { - if (*tc) - ocfs2_free_truncate_context(*tc); - *tc = NULL; - } - mlog_exit_void(); return status; } @@ -7377,7 +7164,7 @@ int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh, if (end > i_size_read(inode)) end = i_size_read(inode); - BUG_ON(start >= end); + BUG_ON(start > end); if (!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) || !(le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_DATA_FL) || @@ -7400,7 +7187,7 @@ int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh, goto out; } - ret = ocfs2_journal_access_di(handle, inode, di_bh, + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); @@ -7427,6 +7214,7 @@ int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh, di->i_ctime = di->i_mtime = cpu_to_le64(inode->i_ctime.tv_sec); di->i_ctime_nsec = di->i_mtime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec); + ocfs2_update_inode_fsync_trans(handle, inode, 1); ocfs2_journal_dirty(handle, di_bh); out_commit: @@ -7436,17 +7224,163 @@ out: return ret; } -static void ocfs2_free_truncate_context(struct ocfs2_truncate_context *tc) +static int ocfs2_trim_extent(struct super_block *sb, + struct ocfs2_group_desc *gd, + u32 start, u32 count) { - /* - * The caller is responsible for completing deallocation - * before freeing the context. - */ - if (tc->tc_dealloc.c_first_suballocator != NULL) - mlog(ML_NOTICE, - "Truncate completion has non-empty dealloc context\n"); + u64 discard, bcount; + + bcount = ocfs2_clusters_to_blocks(sb, count); + discard = le64_to_cpu(gd->bg_blkno) + + ocfs2_clusters_to_blocks(sb, start); + + trace_ocfs2_trim_extent(sb, (unsigned long long)discard, bcount); - brelse(tc->tc_last_eb_bh); + return sb_issue_discard(sb, discard, bcount, GFP_NOFS, 0); +} + +static int ocfs2_trim_group(struct super_block *sb, + struct ocfs2_group_desc *gd, + u32 start, u32 max, u32 minbits) +{ + int ret = 0, count = 0, next; + void *bitmap = gd->bg_bitmap; - kfree(tc); + if (le16_to_cpu(gd->bg_free_bits_count) < minbits) + return 0; + + trace_ocfs2_trim_group((unsigned long long)le64_to_cpu(gd->bg_blkno), + start, max, minbits); + + while (start < max) { + start = ocfs2_find_next_zero_bit(bitmap, max, start); + if (start >= max) + break; + next = ocfs2_find_next_bit(bitmap, max, start); + + if ((next - start) >= minbits) { + ret = ocfs2_trim_extent(sb, gd, + start, next - start); + if (ret < 0) { + mlog_errno(ret); + break; + } + count += next - start; + } + start = next + 1; + + if (fatal_signal_pending(current)) { + count = -ERESTARTSYS; + break; + } + + if ((le16_to_cpu(gd->bg_free_bits_count) - count) < minbits) + break; + } + + if (ret < 0) + count = ret; + + return count; +} + +int ocfs2_trim_fs(struct super_block *sb, struct fstrim_range *range) +{ + struct ocfs2_super *osb = OCFS2_SB(sb); + u64 start, len, trimmed, first_group, last_group, group; + int ret, cnt; + u32 first_bit, last_bit, minlen; + struct buffer_head *main_bm_bh = NULL; + struct inode *main_bm_inode = NULL; + struct buffer_head *gd_bh = NULL; + struct ocfs2_dinode *main_bm; + struct ocfs2_group_desc *gd = NULL; + + start = range->start >> osb->s_clustersize_bits; + len = range->len >> osb->s_clustersize_bits; + minlen = range->minlen >> osb->s_clustersize_bits; + + if (minlen >= osb->bitmap_cpg || range->len < sb->s_blocksize) + return -EINVAL; + + main_bm_inode = ocfs2_get_system_file_inode(osb, + GLOBAL_BITMAP_SYSTEM_INODE, + OCFS2_INVALID_SLOT); + if (!main_bm_inode) { + ret = -EIO; + mlog_errno(ret); + goto out; + } + + mutex_lock(&main_bm_inode->i_mutex); + + ret = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 0); + if (ret < 0) { + mlog_errno(ret); + goto out_mutex; + } + main_bm = (struct ocfs2_dinode *)main_bm_bh->b_data; + + if (start >= le32_to_cpu(main_bm->i_clusters)) { + ret = -EINVAL; + goto out_unlock; + } + + len = range->len >> osb->s_clustersize_bits; + if (start + len > le32_to_cpu(main_bm->i_clusters)) + len = le32_to_cpu(main_bm->i_clusters) - start; + + trace_ocfs2_trim_fs(start, len, minlen); + + /* Determine first and last group to examine based on start and len */ + first_group = ocfs2_which_cluster_group(main_bm_inode, start); + if (first_group == osb->first_cluster_group_blkno) + first_bit = start; + else + first_bit = start - ocfs2_blocks_to_clusters(sb, first_group); + last_group = ocfs2_which_cluster_group(main_bm_inode, start + len - 1); + last_bit = osb->bitmap_cpg; + + trimmed = 0; + for (group = first_group; group <= last_group;) { + if (first_bit + len >= osb->bitmap_cpg) + last_bit = osb->bitmap_cpg; + else + last_bit = first_bit + len; + + ret = ocfs2_read_group_descriptor(main_bm_inode, + main_bm, group, + &gd_bh); + if (ret < 0) { + mlog_errno(ret); + break; + } + + gd = (struct ocfs2_group_desc *)gd_bh->b_data; + cnt = ocfs2_trim_group(sb, gd, first_bit, last_bit, minlen); + brelse(gd_bh); + gd_bh = NULL; + if (cnt < 0) { + ret = cnt; + mlog_errno(ret); + break; + } + + trimmed += cnt; + len -= osb->bitmap_cpg - first_bit; + first_bit = 0; + if (group == osb->first_cluster_group_blkno) + group = ocfs2_clusters_to_blocks(sb, osb->bitmap_cpg); + else + group += ocfs2_clusters_to_blocks(sb, osb->bitmap_cpg); + } + range->len = trimmed * sb->s_blocksize; +out_unlock: + ocfs2_inode_unlock(main_bm_inode, 0); + brelse(main_bm_bh); +out_mutex: + mutex_unlock(&main_bm_inode->i_mutex); + iput(main_bm_inode); +out: + return ret; } diff --git a/fs/ocfs2/alloc.h b/fs/ocfs2/alloc.h index 353254ba29e..ca381c58412 100644 --- a/fs/ocfs2/alloc.h +++ b/fs/ocfs2/alloc.h @@ -45,7 +45,8 @@ * * ocfs2_extent_tree contains info for the root of the b-tree, it must have a * root ocfs2_extent_list and a root_bh so that they can be used in the b-tree - * functions. With metadata ecc, we now call different journal_access + * functions. It needs the ocfs2_caching_info structure associated with + * I/O on the tree. With metadata ecc, we now call different journal_access * functions for each type of metadata, so it must have the * root_journal_access function. * ocfs2_extent_tree_operations abstract the normal operations we do for @@ -56,6 +57,7 @@ struct ocfs2_extent_tree { struct ocfs2_extent_tree_operations *et_ops; struct buffer_head *et_root_bh; struct ocfs2_extent_list *et_root_el; + struct ocfs2_caching_info *et_ci; ocfs2_journal_access_func et_root_journal_access; void *et_object; unsigned int et_max_leaf_clusters; @@ -66,31 +68,32 @@ struct ocfs2_extent_tree { * specified object buffer. */ void ocfs2_init_dinode_extent_tree(struct ocfs2_extent_tree *et, - struct inode *inode, + struct ocfs2_caching_info *ci, struct buffer_head *bh); void ocfs2_init_xattr_tree_extent_tree(struct ocfs2_extent_tree *et, - struct inode *inode, + struct ocfs2_caching_info *ci, struct buffer_head *bh); struct ocfs2_xattr_value_buf; void ocfs2_init_xattr_value_extent_tree(struct ocfs2_extent_tree *et, - struct inode *inode, + struct ocfs2_caching_info *ci, struct ocfs2_xattr_value_buf *vb); void ocfs2_init_dx_root_extent_tree(struct ocfs2_extent_tree *et, - struct inode *inode, + struct ocfs2_caching_info *ci, struct buffer_head *bh); +void ocfs2_init_refcount_extent_tree(struct ocfs2_extent_tree *et, + struct ocfs2_caching_info *ci, + struct buffer_head *bh); /* * Read an extent block into *bh. If *bh is NULL, a bh will be * allocated. This is a cached read. The extent block will be validated * with ocfs2_validate_extent_block(). */ -int ocfs2_read_extent_block(struct inode *inode, u64 eb_blkno, +int ocfs2_read_extent_block(struct ocfs2_caching_info *ci, u64 eb_blkno, struct buffer_head **bh); struct ocfs2_alloc_context; -int ocfs2_insert_extent(struct ocfs2_super *osb, - handle_t *handle, - struct inode *inode, +int ocfs2_insert_extent(handle_t *handle, struct ocfs2_extent_tree *et, u32 cpos, u64 start_blk, @@ -103,34 +106,45 @@ enum ocfs2_alloc_restarted { RESTART_TRANS, RESTART_META }; -int ocfs2_add_clusters_in_btree(struct ocfs2_super *osb, - struct inode *inode, +int ocfs2_add_clusters_in_btree(handle_t *handle, + struct ocfs2_extent_tree *et, u32 *logical_offset, u32 clusters_to_add, int mark_unwritten, - struct ocfs2_extent_tree *et, - handle_t *handle, struct ocfs2_alloc_context *data_ac, struct ocfs2_alloc_context *meta_ac, enum ocfs2_alloc_restarted *reason_ret); struct ocfs2_cached_dealloc_ctxt; +struct ocfs2_path; +int ocfs2_split_extent(handle_t *handle, + struct ocfs2_extent_tree *et, + struct ocfs2_path *path, + int split_index, + struct ocfs2_extent_rec *split_rec, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_cached_dealloc_ctxt *dealloc); int ocfs2_mark_extent_written(struct inode *inode, struct ocfs2_extent_tree *et, handle_t *handle, u32 cpos, u32 len, u32 phys, struct ocfs2_alloc_context *meta_ac, struct ocfs2_cached_dealloc_ctxt *dealloc); -int ocfs2_remove_extent(struct inode *inode, - struct ocfs2_extent_tree *et, - u32 cpos, u32 len, handle_t *handle, +int ocfs2_change_extent_flag(handle_t *handle, + struct ocfs2_extent_tree *et, + u32 cpos, u32 len, u32 phys, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_cached_dealloc_ctxt *dealloc, + int new_flags, int clear_flags); +int ocfs2_remove_extent(handle_t *handle, struct ocfs2_extent_tree *et, + u32 cpos, u32 len, struct ocfs2_alloc_context *meta_ac, struct ocfs2_cached_dealloc_ctxt *dealloc); int ocfs2_remove_btree_range(struct inode *inode, struct ocfs2_extent_tree *et, - u32 cpos, u32 phys_cpos, u32 len, - struct ocfs2_cached_dealloc_ctxt *dealloc); + u32 cpos, u32 phys_cpos, u32 len, int flags, + struct ocfs2_cached_dealloc_ctxt *dealloc, + u64 refcount_loc); int ocfs2_num_free_extents(struct ocfs2_super *osb, - struct inode *inode, struct ocfs2_extent_tree *et); /* @@ -195,6 +209,9 @@ static inline void ocfs2_init_dealloc_ctxt(struct ocfs2_cached_dealloc_ctxt *c) } int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt, u64 blkno, unsigned int bit); +int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt, + int type, int slot, u64 suballoc, u64 blkno, + unsigned int bit); static inline int ocfs2_dealloc_has_cluster(struct ocfs2_cached_dealloc_ctxt *c) { return c->c_global_allocator != NULL; @@ -211,21 +228,18 @@ struct ocfs2_truncate_context { int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle, u64 range_start, u64 range_end); -int ocfs2_prepare_truncate(struct ocfs2_super *osb, - struct inode *inode, - struct buffer_head *fe_bh, - struct ocfs2_truncate_context **tc); int ocfs2_commit_truncate(struct ocfs2_super *osb, struct inode *inode, - struct buffer_head *fe_bh, - struct ocfs2_truncate_context *tc); + struct buffer_head *di_bh); int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh, unsigned int start, unsigned int end, int trunc); -int ocfs2_find_leaf(struct inode *inode, struct ocfs2_extent_list *root_el, - u32 cpos, struct buffer_head **leaf_bh); +int ocfs2_find_leaf(struct ocfs2_caching_info *ci, + struct ocfs2_extent_list *root_el, u32 cpos, + struct buffer_head **leaf_bh); int ocfs2_search_extent_list(struct ocfs2_extent_list *el, u32 v_cluster); +int ocfs2_trim_fs(struct super_block *sb, struct fstrim_range *range); /* * Helper function to look at the # of clusters in an extent record. */ @@ -254,4 +268,57 @@ static inline int ocfs2_is_empty_extent(struct ocfs2_extent_rec *rec) return !rec->e_leaf_clusters; } +int ocfs2_grab_pages(struct inode *inode, loff_t start, loff_t end, + struct page **pages, int *num); +void ocfs2_map_and_dirty_page(struct inode *inode, handle_t *handle, + unsigned int from, unsigned int to, + struct page *page, int zero, u64 *phys); +/* + * Structures which describe a path through a btree, and functions to + * manipulate them. + * + * The idea here is to be as generic as possible with the tree + * manipulation code. + */ +struct ocfs2_path_item { + struct buffer_head *bh; + struct ocfs2_extent_list *el; +}; + +#define OCFS2_MAX_PATH_DEPTH 5 + +struct ocfs2_path { + int p_tree_depth; + ocfs2_journal_access_func p_root_access; + struct ocfs2_path_item p_node[OCFS2_MAX_PATH_DEPTH]; +}; + +#define path_root_bh(_path) ((_path)->p_node[0].bh) +#define path_root_el(_path) ((_path)->p_node[0].el) +#define path_root_access(_path)((_path)->p_root_access) +#define path_leaf_bh(_path) ((_path)->p_node[(_path)->p_tree_depth].bh) +#define path_leaf_el(_path) ((_path)->p_node[(_path)->p_tree_depth].el) +#define path_num_items(_path) ((_path)->p_tree_depth + 1) + +void ocfs2_reinit_path(struct ocfs2_path *path, int keep_root); +void ocfs2_free_path(struct ocfs2_path *path); +int ocfs2_find_path(struct ocfs2_caching_info *ci, + struct ocfs2_path *path, + u32 cpos); +struct ocfs2_path *ocfs2_new_path_from_path(struct ocfs2_path *path); +struct ocfs2_path *ocfs2_new_path_from_et(struct ocfs2_extent_tree *et); +int ocfs2_path_bh_journal_access(handle_t *handle, + struct ocfs2_caching_info *ci, + struct ocfs2_path *path, + int idx); +int ocfs2_journal_access_path(struct ocfs2_caching_info *ci, + handle_t *handle, + struct ocfs2_path *path); +int ocfs2_find_cpos_for_right_leaf(struct super_block *sb, + struct ocfs2_path *path, u32 *cpos); +int ocfs2_find_cpos_for_left_leaf(struct super_block *sb, + struct ocfs2_path *path, u32 *cpos); +int ocfs2_find_subtree_root(struct ocfs2_extent_tree *et, + struct ocfs2_path *left, + struct ocfs2_path *right); #endif /* OCFS2_ALLOC_H */ diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index b2c52b3a148..4a231a166cf 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -29,7 +29,6 @@ #include <linux/mpage.h> #include <linux/quotaops.h> -#define MLOG_MASK_PREFIX ML_FILE_IO #include <cluster/masklog.h> #include "ocfs2.h" @@ -44,6 +43,8 @@ #include "suballoc.h" #include "super.h" #include "symlink.h" +#include "refcounttree.h" +#include "ocfs2_trace.h" #include "buffer_head_io.h" @@ -58,8 +59,9 @@ static int ocfs2_symlink_get_block(struct inode *inode, sector_t iblock, struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); void *kaddr; - mlog_entry("(0x%p, %llu, 0x%p, %d)\n", inode, - (unsigned long long)iblock, bh_result, create); + trace_ocfs2_symlink_get_block( + (unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)iblock, bh_result, create); BUG_ON(ocfs2_inode_is_fast_symlink(inode)); @@ -78,6 +80,7 @@ static int ocfs2_symlink_get_block(struct inode *inode, sector_t iblock, if ((u64)iblock >= ocfs2_clusters_to_blocks(inode->i_sb, le32_to_cpu(fe->i_clusters))) { + err = -ENOMEM; mlog(ML_ERROR, "block offset is outside the allocated size: " "%llu\n", (unsigned long long)iblock); goto bail; @@ -90,6 +93,7 @@ static int ocfs2_symlink_get_block(struct inode *inode, sector_t iblock, iblock; buffer_cache_bh = sb_getblk(osb->sb, blkno); if (!buffer_cache_bh) { + err = -ENOMEM; mlog(ML_ERROR, "couldn't getblock for symlink!\n"); goto bail; } @@ -100,7 +104,7 @@ static int ocfs2_symlink_get_block(struct inode *inode, sector_t iblock, * copy, the data is still good. */ if (buffer_jbd(buffer_cache_bh) && ocfs2_inode_is_new(inode)) { - kaddr = kmap_atomic(bh_result->b_page, KM_USER0); + kaddr = kmap_atomic(bh_result->b_page); if (!kaddr) { mlog(ML_ERROR, "couldn't kmap!\n"); goto bail; @@ -108,7 +112,7 @@ static int ocfs2_symlink_get_block(struct inode *inode, sector_t iblock, memcpy(kaddr + (bh_result->b_size * iblock), buffer_cache_bh->b_data, bh_result->b_size); - kunmap_atomic(kaddr, KM_USER0); + kunmap_atomic(kaddr); set_buffer_uptodate(bh_result); } brelse(buffer_cache_bh); @@ -122,12 +126,11 @@ static int ocfs2_symlink_get_block(struct inode *inode, sector_t iblock, bail: brelse(bh); - mlog_exit(err); return err; } -static int ocfs2_get_block(struct inode *inode, sector_t iblock, - struct buffer_head *bh_result, int create) +int ocfs2_get_block(struct inode *inode, sector_t iblock, + struct buffer_head *bh_result, int create) { int err = 0; unsigned int ext_flags; @@ -135,8 +138,8 @@ static int ocfs2_get_block(struct inode *inode, sector_t iblock, u64 p_blkno, count, past_eof; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - mlog_entry("(0x%p, %llu, 0x%p, %d)\n", inode, - (unsigned long long)iblock, bh_result, create); + trace_ocfs2_get_block((unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)iblock, bh_result, create); if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SYSTEM_FILE) mlog(ML_NOTICE, "get_block on system inode 0x%p (%lu)\n", @@ -164,7 +167,7 @@ static int ocfs2_get_block(struct inode *inode, sector_t iblock, * ocfs2 never allocates in this function - the only time we * need to use BH_New is when we're extending i_size on a file * system which doesn't support holes, in which case BH_New - * allows block_prepare_write() to zero. + * allows __block_write_begin() to zero. * * If we see this on a sparse file system, then a truncate has * raced us and removed the cluster. In this case, we clear @@ -193,21 +196,21 @@ static int ocfs2_get_block(struct inode *inode, sector_t iblock, (unsigned long long)OCFS2_I(inode)->ip_blkno); mlog(ML_ERROR, "Size %llu, clusters %u\n", (unsigned long long)i_size_read(inode), OCFS2_I(inode)->ip_clusters); dump_stack(); + goto bail; } + } - past_eof = ocfs2_blocks_for_bytes(inode->i_sb, i_size_read(inode)); - mlog(0, "Inode %lu, past_eof = %llu\n", inode->i_ino, - (unsigned long long)past_eof); + past_eof = ocfs2_blocks_for_bytes(inode->i_sb, i_size_read(inode)); - if (create && (iblock >= past_eof)) - set_buffer_new(bh_result); - } + trace_ocfs2_get_block_end((unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)past_eof); + if (create && (iblock >= past_eof)) + set_buffer_new(bh_result); bail: if (err < 0) err = -EIO; - mlog_exit(err); return err; } @@ -235,13 +238,13 @@ int ocfs2_read_inline_data(struct inode *inode, struct page *page, return -EROFS; } - kaddr = kmap_atomic(page, KM_USER0); + kaddr = kmap_atomic(page); if (size) memcpy(kaddr, di->id2.i_data.id_data, size); /* Clear the remaining part of the page */ memset(kaddr + size, 0, PAGE_CACHE_SIZE - size); flush_dcache_page(page); - kunmap_atomic(kaddr, KM_USER0); + kunmap_atomic(kaddr); SetPageUptodate(page); @@ -277,7 +280,8 @@ static int ocfs2_readpage(struct file *file, struct page *page) loff_t start = (loff_t)page->index << PAGE_CACHE_SHIFT; int ret, unlock = 1; - mlog_entry("(0x%p, %lu)\n", file, (page ? page->index : 0)); + trace_ocfs2_readpage((unsigned long long)oi->ip_blkno, + (page ? page->index : 0)); ret = ocfs2_inode_lock_with_page(inode, NULL, 0, page); if (ret != 0) { @@ -288,7 +292,15 @@ static int ocfs2_readpage(struct file *file, struct page *page) } if (down_read_trylock(&oi->ip_alloc_sem) == 0) { + /* + * Unlock the page and cycle ip_alloc_sem so that we don't + * busyloop waiting for ip_alloc_sem to unlock + */ ret = AOP_TRUNCATED_PAGE; + unlock_page(page); + unlock = 0; + down_read(&oi->ip_alloc_sem); + up_read(&oi->ip_alloc_sem); goto out_inode_unlock; } @@ -322,7 +334,6 @@ out_inode_unlock: out: if (unlock) unlock_page(page); - mlog_exit(ret); return ret; } @@ -395,30 +406,11 @@ out_unlock: */ static int ocfs2_writepage(struct page *page, struct writeback_control *wbc) { - int ret; - - mlog_entry("(0x%p)\n", page); - - ret = block_write_full_page(page, ocfs2_get_block, wbc); - - mlog_exit(ret); - - return ret; -} - -/* - * This is called from ocfs2_write_zero_page() which has handled it's - * own cluster locking and has ensured allocation exists for those - * blocks to be written. - */ -int ocfs2_prepare_write_nolock(struct inode *inode, struct page *page, - unsigned from, unsigned to) -{ - int ret; - - ret = block_prepare_write(page, from, to, ocfs2_get_block); + trace_ocfs2_writepage( + (unsigned long long)OCFS2_I(page->mapping->host)->ip_blkno, + page->index); - return ret; + return block_write_full_page(page, ocfs2_get_block, wbc); } /* Taken from ext3. We don't necessarily need the full blown @@ -457,36 +449,6 @@ int walk_page_buffers( handle_t *handle, return ret; } -handle_t *ocfs2_start_walk_page_trans(struct inode *inode, - struct page *page, - unsigned from, - unsigned to) -{ - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - handle_t *handle; - int ret = 0; - - handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS); - if (IS_ERR(handle)) { - ret = -ENOMEM; - mlog_errno(ret); - goto out; - } - - if (ocfs2_should_order_data(inode)) { - ret = ocfs2_jbd2_file_inode(handle, inode); - if (ret < 0) - mlog_errno(ret); - } -out: - if (ret) { - if (!IS_ERR(handle)) - ocfs2_commit_trans(osb, handle); - handle = ERR_PTR(ret); - } - return handle; -} - static sector_t ocfs2_bmap(struct address_space *mapping, sector_t block) { sector_t status; @@ -494,7 +456,8 @@ static sector_t ocfs2_bmap(struct address_space *mapping, sector_t block) int err = 0; struct inode *inode = mapping->host; - mlog_entry("(block = %llu)\n", (unsigned long long)block); + trace_ocfs2_bmap((unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)block); /* We don't need to lock journal system files, since they aren't * accessed concurrently from multiple nodes. @@ -528,8 +491,6 @@ static sector_t ocfs2_bmap(struct address_space *mapping, sector_t block) bail: status = err ? 0 : p_blkno; - mlog_exit((int)status); - return status; } @@ -545,6 +506,9 @@ bail: * * called like this: dio->get_blocks(dio->inode, fs_startblk, * fs_count, map_bh, dio->rw == WRITE); + * + * Note that we never bother to allocate blocks here, and thus ignore the + * create argument. */ static int ocfs2_direct_IO_get_blocks(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create) @@ -561,14 +525,6 @@ static int ocfs2_direct_IO_get_blocks(struct inode *inode, sector_t iblock, inode_blocks = ocfs2_blocks_for_bytes(inode->i_sb, i_size_read(inode)); - /* - * Any write past EOF is not allowed because we'd be extending. - */ - if (create && (iblock + max_blocks) > inode_blocks) { - ret = -EIO; - goto bail; - } - /* This figures out the size of the next contiguous block, and * our logical offset */ ret = ocfs2_extent_map_get_blocks(inode, iblock, &p_blkno, @@ -580,14 +536,8 @@ static int ocfs2_direct_IO_get_blocks(struct inode *inode, sector_t iblock, goto bail; } - if (!ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb)) && !p_blkno && create) { - ocfs2_error(inode->i_sb, - "Inode %llu has a hole at block %llu\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, - (unsigned long long)iblock); - ret = -EROFS; - goto bail; - } + /* We should already CoW the refcounted extent in case of create. */ + BUG_ON(create && (ext_flags & OCFS2_EXT_REFCOUNTED)); /* * get_more_blocks() expects us to describe a hole by clearing @@ -597,20 +547,8 @@ static int ocfs2_direct_IO_get_blocks(struct inode *inode, sector_t iblock, */ if (p_blkno && !(ext_flags & OCFS2_EXT_UNWRITTEN)) map_bh(bh_result, inode->i_sb, p_blkno); - else { - /* - * ocfs2_prepare_inode_for_write() should have caught - * the case where we'd be filling a hole and triggered - * a buffered write instead. - */ - if (create) { - ret = -EIO; - mlog_errno(ret); - goto bail; - } - + else clear_buffer_mapped(bh_result); - } /* make sure we don't map more than max_blocks blocks here as that's all the kernel will handle at this point. */ @@ -621,63 +559,51 @@ bail: return ret; } -/* +/* * ocfs2_dio_end_io is called by the dio core when a dio is finished. We're - * particularly interested in the aio/dio case. Like the core uses - * i_alloc_sem, we use the rw_lock DLM lock to protect io on one node from - * truncation on another. + * particularly interested in the aio/dio case. We use the rw_lock DLM lock + * to protect io on one node from truncation on another. */ static void ocfs2_dio_end_io(struct kiocb *iocb, loff_t offset, ssize_t bytes, void *private) { - struct inode *inode = iocb->ki_filp->f_path.dentry->d_inode; + struct inode *inode = file_inode(iocb->ki_filp); int level; /* this io's submitter should not have unlocked this before we could */ BUG_ON(!ocfs2_iocb_is_rw_locked(iocb)); + if (ocfs2_iocb_is_sem_locked(iocb)) + ocfs2_iocb_clear_sem_locked(iocb); + + if (ocfs2_iocb_is_unaligned_aio(iocb)) { + ocfs2_iocb_clear_unaligned_aio(iocb); + + mutex_unlock(&OCFS2_I(inode)->ip_unaligned_aio); + } + ocfs2_iocb_clear_rw_locked(iocb); level = ocfs2_iocb_rw_locked_level(iocb); - if (!level) - up_read(&inode->i_alloc_sem); ocfs2_rw_unlock(inode, level); } -/* - * ocfs2_invalidatepage() and ocfs2_releasepage() are shamelessly stolen - * from ext3. PageChecked() bits have been removed as OCFS2 does not - * do journalled data. - */ -static void ocfs2_invalidatepage(struct page *page, unsigned long offset) -{ - journal_t *journal = OCFS2_SB(page->mapping->host->i_sb)->journal->j_journal; - - jbd2_journal_invalidatepage(journal, page, offset); -} - static int ocfs2_releasepage(struct page *page, gfp_t wait) { - journal_t *journal = OCFS2_SB(page->mapping->host->i_sb)->journal->j_journal; - if (!page_has_buffers(page)) return 0; - return jbd2_journal_try_to_free_buffers(journal, page, wait); + return try_to_free_buffers(page); } static ssize_t ocfs2_direct_IO(int rw, struct kiocb *iocb, - const struct iovec *iov, - loff_t offset, - unsigned long nr_segs) + struct iov_iter *iter, + loff_t offset) { struct file *file = iocb->ki_filp; - struct inode *inode = file->f_path.dentry->d_inode->i_mapping->host; - int ret; - - mlog_entry_void(); + struct inode *inode = file_inode(file)->i_mapping->host; /* * Fallback to buffered I/O if we see an inode without @@ -686,14 +612,14 @@ static ssize_t ocfs2_direct_IO(int rw, if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) return 0; - ret = blockdev_direct_IO_no_locking(rw, iocb, inode, - inode->i_sb->s_bdev, iov, offset, - nr_segs, - ocfs2_direct_IO_get_blocks, - ocfs2_dio_end_io); + /* Fallback to buffered I/O if we are appending. */ + if (i_size_read(inode) <= offset) + return 0; - mlog_exit(ret); - return ret; + return __blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, + iter, offset, + ocfs2_direct_IO_get_blocks, + ocfs2_dio_end_io, NULL, 0); } static void ocfs2_figure_cluster_boundaries(struct ocfs2_super *osb, @@ -740,7 +666,7 @@ static void ocfs2_clear_page_regions(struct page *page, ocfs2_figure_cluster_boundaries(osb, cpos, &cluster_start, &cluster_end); - kaddr = kmap_atomic(page, KM_USER0); + kaddr = kmap_atomic(page); if (from || to) { if (from > cluster_start) @@ -751,7 +677,7 @@ static void ocfs2_clear_page_regions(struct page *page, memset(kaddr + cluster_start, 0, cluster_end - cluster_start); } - kunmap_atomic(kaddr, KM_USER0); + kunmap_atomic(kaddr); } /* @@ -776,7 +702,7 @@ static int ocfs2_should_read_blk(struct inode *inode, struct page *page, } /* - * Some of this taken from block_prepare_write(). We already have our + * Some of this taken from __block_write_begin(). We already have our * mapping by now though, and the entire write will be allocating or * it won't, so not much need to use BH_New. * @@ -894,18 +820,17 @@ struct ocfs2_write_cluster_desc { */ unsigned c_new; unsigned c_unwritten; + unsigned c_needs_zero; }; -static inline int ocfs2_should_zero_cluster(struct ocfs2_write_cluster_desc *d) -{ - return d->c_new || d->c_unwritten; -} - struct ocfs2_write_ctxt { /* Logical cluster position / len of write */ u32 w_cpos; u32 w_clen; + /* First cluster allocated in a nonsparse extend */ + u32 w_first_new_cpos; + struct ocfs2_write_cluster_desc w_desc[OCFS2_MAX_CLUSTERS_PER_PAGE]; /* @@ -928,11 +853,17 @@ struct ocfs2_write_ctxt { * out in so that future reads from that region will get * zero's. */ - struct page *w_pages[OCFS2_MAX_CTXT_PAGES]; unsigned int w_num_pages; + struct page *w_pages[OCFS2_MAX_CTXT_PAGES]; struct page *w_target_page; /* + * w_target_locked is used for page_mkwrite path indicating no unlocking + * against w_target_page in ocfs2_write_end_nolock. + */ + unsigned int w_target_locked:1; + + /* * ocfs2_write_end() uses this to know what the real range to * write in the target should be. */ @@ -965,6 +896,24 @@ void ocfs2_unlock_and_free_pages(struct page **pages, int num_pages) static void ocfs2_free_write_ctxt(struct ocfs2_write_ctxt *wc) { + int i; + + /* + * w_target_locked is only set to true in the page_mkwrite() case. + * The intent is to allow us to lock the target page from write_begin() + * to write_end(). The caller must hold a ref on w_target_page. + */ + if (wc->w_target_locked) { + BUG_ON(!wc->w_target_page); + for (i = 0; i < wc->w_num_pages; i++) { + if (wc->w_target_page == wc->w_pages[i]) { + wc->w_pages[i] = NULL; + break; + } + } + mark_page_accessed(wc->w_target_page); + page_cache_release(wc->w_target_page); + } ocfs2_unlock_and_free_pages(wc->w_pages, wc->w_num_pages); brelse(wc->w_di_bh); @@ -983,6 +932,7 @@ static int ocfs2_alloc_write_ctxt(struct ocfs2_write_ctxt **wcp, return -ENOMEM; wc->w_cpos = pos >> osb->s_clustersize_bits; + wc->w_first_new_cpos = UINT_MAX; cend = (pos + len - 1) >> osb->s_clustersize_bits; wc->w_clen = cend - wc->w_cpos + 1; get_bh(di_bh); @@ -1082,6 +1032,12 @@ static int ocfs2_prepare_page_for_write(struct inode *inode, u64 *p_blkno, ocfs2_figure_cluster_boundaries(OCFS2_SB(inode->i_sb), cpos, &cluster_start, &cluster_end); + /* treat the write as new if the a hole/lseek spanned across + * the page boundary. + */ + new = new | ((i_size_read(inode) <= page_offset(page)) && + (page_offset(page) <= user_pos)); + if (page == wc->w_target_page) { map_from = user_pos & (PAGE_CACHE_SIZE - 1); map_to = map_from + user_len; @@ -1148,23 +1104,37 @@ out: */ static int ocfs2_grab_pages_for_write(struct address_space *mapping, struct ocfs2_write_ctxt *wc, - u32 cpos, loff_t user_pos, int new, + u32 cpos, loff_t user_pos, + unsigned user_len, int new, struct page *mmap_page) { int ret = 0, i; - unsigned long start, target_index, index; + unsigned long start, target_index, end_index, index; struct inode *inode = mapping->host; + loff_t last_byte; target_index = user_pos >> PAGE_CACHE_SHIFT; /* * Figure out how many pages we'll be manipulating here. For * non allocating write, we just change the one - * page. Otherwise, we'll need a whole clusters worth. + * page. Otherwise, we'll need a whole clusters worth. If we're + * writing past i_size, we only need enough pages to cover the + * last page of the write. */ if (new) { wc->w_num_pages = ocfs2_pages_per_cluster(inode->i_sb); start = ocfs2_align_clusters_to_page_index(inode->i_sb, cpos); + /* + * We need the index *past* the last page we could possibly + * touch. This is the page past the end of the write or + * i_size, whichever is greater. + */ + last_byte = max(user_pos + user_len, i_size_read(inode)); + BUG_ON(last_byte < 1); + end_index = ((last_byte - 1) >> PAGE_CACHE_SHIFT) + 1; + if ((start + wc->w_num_pages) > end_index) + wc->w_num_pages = end_index - start; } else { wc->w_num_pages = 1; start = target_index; @@ -1181,20 +1151,17 @@ static int ocfs2_grab_pages_for_write(struct address_space *mapping, */ lock_page(mmap_page); + /* Exit and let the caller retry */ if (mmap_page->mapping != mapping) { + WARN_ON(mmap_page->mapping); unlock_page(mmap_page); - /* - * Sanity check - the locking in - * ocfs2_pagemkwrite() should ensure - * that this code doesn't trigger. - */ - ret = -EINVAL; - mlog_errno(ret); + ret = -EAGAIN; goto out; } page_cache_get(mmap_page); wc->w_pages[i] = mmap_page; + wc->w_target_locked = true; } else { wc->w_pages[i] = find_or_create_page(mapping, index, GFP_NOFS); @@ -1204,11 +1171,14 @@ static int ocfs2_grab_pages_for_write(struct address_space *mapping, goto out; } } + wait_for_stable_page(wc->w_pages[i]); if (index == target_index) wc->w_target_page = wc->w_pages[i]; } out: + if (ret) + wc->w_target_locked = false; return ret; } @@ -1217,20 +1187,18 @@ out: */ static int ocfs2_write_cluster(struct address_space *mapping, u32 phys, unsigned int unwritten, + unsigned int should_zero, struct ocfs2_alloc_context *data_ac, struct ocfs2_alloc_context *meta_ac, struct ocfs2_write_ctxt *wc, u32 cpos, loff_t user_pos, unsigned user_len) { - int ret, i, new, should_zero = 0; + int ret, i, new; u64 v_blkno, p_blkno; struct inode *inode = mapping->host; struct ocfs2_extent_tree et; new = phys == 0 ? 1 : 0; - if (new || unwritten) - should_zero = 1; - if (new) { u32 tmp_pos; @@ -1260,7 +1228,8 @@ static int ocfs2_write_cluster(struct address_space *mapping, goto out; } } else if (unwritten) { - ocfs2_init_dinode_extent_tree(&et, inode, wc->w_di_bh); + ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), + wc->w_di_bh); ret = ocfs2_mark_extent_written(inode, &et, wc->w_handle, cpos, 1, phys, meta_ac, &wc->w_dealloc); @@ -1301,7 +1270,7 @@ static int ocfs2_write_cluster(struct address_space *mapping, if (tmpret) { mlog_errno(tmpret); if (ret == 0) - tmpret = ret; + ret = tmpret; } } @@ -1341,7 +1310,9 @@ static int ocfs2_write_cluster_by_desc(struct address_space *mapping, local_len = osb->s_clustersize - cluster_off; ret = ocfs2_write_cluster(mapping, desc->c_phys, - desc->c_unwritten, data_ac, meta_ac, + desc->c_unwritten, + desc->c_needs_zero, + data_ac, meta_ac, wc, desc->c_cpos, pos, local_len); if (ret) { mlog_errno(ret); @@ -1391,14 +1362,14 @@ static void ocfs2_set_target_boundaries(struct ocfs2_super *osb, * newly allocated cluster. */ desc = &wc->w_desc[0]; - if (ocfs2_should_zero_cluster(desc)) + if (desc->c_needs_zero) ocfs2_figure_cluster_boundaries(osb, desc->c_cpos, &wc->w_target_from, NULL); desc = &wc->w_desc[wc->w_clen - 1]; - if (ocfs2_should_zero_cluster(desc)) + if (desc->c_needs_zero) ocfs2_figure_cluster_boundaries(osb, desc->c_cpos, NULL, @@ -1447,6 +1418,9 @@ static int ocfs2_populate_write_desc(struct inode *inode, goto out; } + /* We should already CoW the refcountd extent. */ + BUG_ON(ext_flags & OCFS2_EXT_REFCOUNTED); + /* * Assume worst case - that we're writing in * the middle of the extent. @@ -1466,13 +1440,28 @@ static int ocfs2_populate_write_desc(struct inode *inode, phys++; } + /* + * If w_first_new_cpos is < UINT_MAX, we have a non-sparse + * file that got extended. w_first_new_cpos tells us + * where the newly allocated clusters are so we can + * zero them. + */ + if (desc->c_cpos >= wc->w_first_new_cpos) { + BUG_ON(phys == 0); + desc->c_needs_zero = 1; + } + desc->c_phys = phys; if (phys == 0) { desc->c_new = 1; + desc->c_needs_zero = 1; *clusters_to_alloc = *clusters_to_alloc + 1; } - if (ext_flags & OCFS2_EXT_UNWRITTEN) + + if (ext_flags & OCFS2_EXT_UNWRITTEN) { desc->c_unwritten = 1; + desc->c_needs_zero = 1; + } num_clusters--; } @@ -1512,7 +1501,7 @@ static int ocfs2_write_begin_inline(struct address_space *mapping, goto out; } - ret = ocfs2_journal_access_di(handle, inode, wc->w_di_bh, + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), wc->w_di_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { ocfs2_commit_trans(osb, handle); @@ -1557,9 +1546,9 @@ static int ocfs2_try_to_write_inline_data(struct address_space *mapping, struct ocfs2_inode_info *oi = OCFS2_I(inode); struct ocfs2_dinode *di = NULL; - mlog(0, "Inode %llu, write of %u bytes at off %llu. features: 0x%x\n", - (unsigned long long)oi->ip_blkno, len, (unsigned long long)pos, - oi->ip_dyn_features); + trace_ocfs2_try_to_write_inline_data((unsigned long long)oi->ip_blkno, + len, (unsigned long long)pos, + oi->ip_dyn_features); /* * Handle inodes which already have inline data 1st. @@ -1618,34 +1607,86 @@ out: * write path can treat it as an non-allocating write, which has no * special case code for sparse/nonsparse files. */ -static int ocfs2_expand_nonsparse_inode(struct inode *inode, loff_t pos, - unsigned len, +static int ocfs2_expand_nonsparse_inode(struct inode *inode, + struct buffer_head *di_bh, + loff_t pos, unsigned len, struct ocfs2_write_ctxt *wc) { int ret; - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); loff_t newsize = pos + len; - if (ocfs2_sparse_alloc(osb)) - return 0; + BUG_ON(ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))); if (newsize <= i_size_read(inode)) return 0; - ret = ocfs2_extend_no_holes(inode, newsize, newsize - len); + ret = ocfs2_extend_no_holes(inode, di_bh, newsize, pos); if (ret) mlog_errno(ret); + wc->w_first_new_cpos = + ocfs2_clusters_for_bytes(inode->i_sb, i_size_read(inode)); + return ret; } -int ocfs2_write_begin_nolock(struct address_space *mapping, +static int ocfs2_zero_tail(struct inode *inode, struct buffer_head *di_bh, + loff_t pos) +{ + int ret = 0; + + BUG_ON(!ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))); + if (pos > i_size_read(inode)) + ret = ocfs2_zero_extend(inode, di_bh, pos); + + return ret; +} + +/* + * Try to flush truncate logs if we can free enough clusters from it. + * As for return value, "< 0" means error, "0" no space and "1" means + * we have freed enough spaces and let the caller try to allocate again. + */ +static int ocfs2_try_to_free_truncate_log(struct ocfs2_super *osb, + unsigned int needed) +{ + tid_t target; + int ret = 0; + unsigned int truncated_clusters; + + mutex_lock(&osb->osb_tl_inode->i_mutex); + truncated_clusters = osb->truncated_clusters; + mutex_unlock(&osb->osb_tl_inode->i_mutex); + + /* + * Check whether we can succeed in allocating if we free + * the truncate log. + */ + if (truncated_clusters < needed) + goto out; + + ret = ocfs2_flush_truncate_log(osb); + if (ret) { + mlog_errno(ret); + goto out; + } + + if (jbd2_journal_start_commit(osb->journal->j_journal, &target)) { + jbd2_log_wait_commit(osb->journal->j_journal, target); + ret = 1; + } +out: + return ret; +} + +int ocfs2_write_begin_nolock(struct file *filp, + struct address_space *mapping, loff_t pos, unsigned len, unsigned flags, struct page **pagep, void **fsdata, struct buffer_head *di_bh, struct page *mmap_page) { - int ret, credits = OCFS2_INODE_UPDATE_CREDITS; - unsigned int clusters_to_alloc, extents_to_split; + int ret, cluster_of_pages, credits = OCFS2_INODE_UPDATE_CREDITS; + unsigned int clusters_to_alloc, extents_to_split, clusters_need = 0; struct ocfs2_write_ctxt *wc; struct inode *inode = mapping->host; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); @@ -1654,7 +1695,9 @@ int ocfs2_write_begin_nolock(struct address_space *mapping, struct ocfs2_alloc_context *meta_ac = NULL; handle_t *handle; struct ocfs2_extent_tree et; + int try_free = 1, ret1; +try_again: ret = ocfs2_alloc_write_ctxt(&wc, osb, pos, len, di_bh); if (ret) { mlog_errno(ret); @@ -1674,21 +1717,47 @@ int ocfs2_write_begin_nolock(struct address_space *mapping, } } - ret = ocfs2_expand_nonsparse_inode(inode, pos, len, wc); + if (ocfs2_sparse_alloc(osb)) + ret = ocfs2_zero_tail(inode, di_bh, pos); + else + ret = ocfs2_expand_nonsparse_inode(inode, di_bh, pos, len, + wc); if (ret) { mlog_errno(ret); goto out; } + ret = ocfs2_check_range_for_refcount(inode, pos, len); + if (ret < 0) { + mlog_errno(ret); + goto out; + } else if (ret == 1) { + clusters_need = wc->w_clen; + ret = ocfs2_refcount_cow(inode, di_bh, + wc->w_cpos, wc->w_clen, UINT_MAX); + if (ret) { + mlog_errno(ret); + goto out; + } + } + ret = ocfs2_populate_write_desc(inode, wc, &clusters_to_alloc, &extents_to_split); if (ret) { mlog_errno(ret); goto out; } + clusters_need += clusters_to_alloc; di = (struct ocfs2_dinode *)wc->w_di_bh->b_data; + trace_ocfs2_write_begin_nolock( + (unsigned long long)OCFS2_I(inode)->ip_blkno, + (long long)i_size_read(inode), + le32_to_cpu(di->i_clusters), + pos, len, flags, mmap_page, + clusters_to_alloc, extents_to_split); + /* * We set w_target_from, w_target_to here so that * ocfs2_write_end() knows which range in the target page to @@ -1701,13 +1770,8 @@ int ocfs2_write_begin_nolock(struct address_space *mapping, * ocfs2_lock_allocators(). It greatly over-estimates * the work to be done. */ - mlog(0, "extend inode %llu, i_size = %lld, di->i_clusters = %u," - " clusters_to_add = %u, extents_to_split = %u\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, - (long long)i_size_read(inode), le32_to_cpu(di->i_clusters), - clusters_to_alloc, extents_to_split); - - ocfs2_init_dinode_extent_tree(&et, inode, wc->w_di_bh); + ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), + wc->w_di_bh); ret = ocfs2_lock_allocators(inode, &et, clusters_to_alloc, extents_to_split, &data_ac, &meta_ac); @@ -1716,14 +1780,27 @@ int ocfs2_write_begin_nolock(struct address_space *mapping, goto out; } + if (data_ac) + data_ac->ac_resv = &OCFS2_I(inode)->ip_la_data_resv; + credits = ocfs2_calc_extend_credits(inode->i_sb, - &di->id2.i_list, - clusters_to_alloc); + &di->id2.i_list); } - ocfs2_set_target_boundaries(osb, wc, pos, len, - clusters_to_alloc + extents_to_split); + /* + * We have to zero sparse allocated clusters, unwritten extent clusters, + * and non-sparse clusters we just extended. For non-sparse writes, + * we know zeros will only be needed in the first and/or last cluster. + */ + if (clusters_to_alloc || extents_to_split || + (wc->w_clen && (wc->w_desc[0].c_needs_zero || + wc->w_desc[wc->w_clen - 1].c_needs_zero))) + cluster_of_pages = 1; + else + cluster_of_pages = 0; + + ocfs2_set_target_boundaries(osb, wc, pos, len, cluster_of_pages); handle = ocfs2_start_trans(osb, credits); if (IS_ERR(handle)) { @@ -1734,16 +1811,17 @@ int ocfs2_write_begin_nolock(struct address_space *mapping, wc->w_handle = handle; - if (clusters_to_alloc && vfs_dq_alloc_space_nodirty(inode, - ocfs2_clusters_to_bytes(osb->sb, clusters_to_alloc))) { - ret = -EDQUOT; - goto out_commit; + if (clusters_to_alloc) { + ret = dquot_alloc_space_nodirty(inode, + ocfs2_clusters_to_bytes(osb->sb, clusters_to_alloc)); + if (ret) + goto out_commit; } /* * We don't want this to fail in ocfs2_write_end(), so do it * here. */ - ret = ocfs2_journal_access_di(handle, inode, wc->w_di_bh, + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), wc->w_di_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); @@ -1755,14 +1833,25 @@ int ocfs2_write_begin_nolock(struct address_space *mapping, * that we can zero and flush if we error after adding the * extent. */ - ret = ocfs2_grab_pages_for_write(mapping, wc, wc->w_cpos, pos, - clusters_to_alloc + extents_to_split, - mmap_page); - if (ret) { + ret = ocfs2_grab_pages_for_write(mapping, wc, wc->w_cpos, pos, len, + cluster_of_pages, mmap_page); + if (ret && ret != -EAGAIN) { mlog_errno(ret); goto out_quota; } + /* + * ocfs2_grab_pages_for_write() returns -EAGAIN if it could not lock + * the target page. In this case, we exit with no error and no target + * page. This will trigger the caller, page_mkwrite(), to re-try + * the operation. + */ + if (ret == -EAGAIN) { + BUG_ON(wc->w_target_page); + ret = 0; + goto out_quota; + } + ret = ocfs2_write_cluster_by_desc(mapping, data_ac, meta_ac, wc, pos, len); if (ret) { @@ -1781,7 +1870,7 @@ success: return 0; out_quota: if (clusters_to_alloc) - vfs_dq_free_space(inode, + dquot_free_space(inode, ocfs2_clusters_to_bytes(osb->sb, clusters_to_alloc)); out_commit: ocfs2_commit_trans(osb, handle); @@ -1789,10 +1878,30 @@ out_commit: out: ocfs2_free_write_ctxt(wc); - if (data_ac) + if (data_ac) { ocfs2_free_alloc_context(data_ac); - if (meta_ac) + data_ac = NULL; + } + if (meta_ac) { ocfs2_free_alloc_context(meta_ac); + meta_ac = NULL; + } + + if (ret == -ENOSPC && try_free) { + /* + * Try to free some truncate log so that we can have enough + * clusters to allocate. + */ + try_free = 0; + + ret1 = ocfs2_try_to_free_truncate_log(osb, clusters_need); + if (ret1 == 1) + goto try_again; + + if (ret1 < 0) + mlog_errno(ret1); + } + return ret; } @@ -1819,7 +1928,7 @@ static int ocfs2_write_begin(struct file *file, struct address_space *mapping, */ down_write(&OCFS2_I(inode)->ip_alloc_sem); - ret = ocfs2_write_begin_nolock(mapping, pos, len, flags, pagep, + ret = ocfs2_write_begin_nolock(file, mapping, pos, len, flags, pagep, fsdata, di_bh, NULL); if (ret) { mlog_errno(ret); @@ -1853,12 +1962,12 @@ static void ocfs2_write_end_inline(struct inode *inode, loff_t pos, } } - kaddr = kmap_atomic(wc->w_target_page, KM_USER0); + kaddr = kmap_atomic(wc->w_target_page); memcpy(di->id2.i_data.id_data + pos, kaddr + pos, *copied); - kunmap_atomic(kaddr, KM_USER0); + kunmap_atomic(kaddr); - mlog(0, "Data written to inode at offset %llu. " - "id_count = %u, copied = %u, i_dyn_features = 0x%x\n", + trace_ocfs2_write_end_inline( + (unsigned long long)OCFS2_I(inode)->ip_blkno, (unsigned long long)pos, *copied, le16_to_cpu(di->id2.i_data.id_count), le16_to_cpu(di->i_dyn_features)); @@ -1920,7 +2029,7 @@ int ocfs2_write_end_nolock(struct address_space *mapping, out_write_size: pos += copied; - if (pos > inode->i_size) { + if (pos > i_size_read(inode)) { i_size_write(inode, pos); mark_inode_dirty(inode); } @@ -1929,6 +2038,7 @@ out_write_size: inode->i_mtime = inode->i_ctime = CURRENT_TIME; di->i_mtime = di->i_ctime = cpu_to_le64(inode->i_mtime.tv_sec); di->i_mtime_nsec = di->i_ctime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec); + ocfs2_update_inode_fsync_trans(handle, inode, 1); ocfs2_journal_dirty(handle, wc->w_di_bh); ocfs2_commit_trans(osb, handle); @@ -1962,10 +2072,10 @@ const struct address_space_operations ocfs2_aops = { .write_begin = ocfs2_write_begin, .write_end = ocfs2_write_end, .bmap = ocfs2_bmap, - .sync_page = block_sync_page, .direct_IO = ocfs2_direct_IO, - .invalidatepage = ocfs2_invalidatepage, + .invalidatepage = block_invalidatepage, .releasepage = ocfs2_releasepage, .migratepage = buffer_migrate_page, .is_partially_uptodate = block_is_partially_uptodate, + .error_remove_page = generic_error_remove_page, }; diff --git a/fs/ocfs2/aops.h b/fs/ocfs2/aops.h index 503e49232e1..6cae155d54d 100644 --- a/fs/ocfs2/aops.h +++ b/fs/ocfs2/aops.h @@ -22,8 +22,7 @@ #ifndef OCFS2_AOPS_H #define OCFS2_AOPS_H -int ocfs2_prepare_write_nolock(struct inode *inode, struct page *page, - unsigned from, unsigned to); +#include <linux/aio.h> handle_t *ocfs2_start_walk_page_trans(struct inode *inode, struct page *page, @@ -48,7 +47,8 @@ int ocfs2_write_end_nolock(struct address_space *mapping, loff_t pos, unsigned len, unsigned copied, struct page *page, void *fsdata); -int ocfs2_write_begin_nolock(struct address_space *mapping, +int ocfs2_write_begin_nolock(struct file *filp, + struct address_space *mapping, loff_t pos, unsigned len, unsigned flags, struct page **pagep, void **fsdata, struct buffer_head *di_bh, struct page *mmap_page); @@ -57,6 +57,8 @@ int ocfs2_read_inline_data(struct inode *inode, struct page *page, struct buffer_head *di_bh); int ocfs2_size_fits_inline_data(struct buffer_head *di_bh, u64 new_size); +int ocfs2_get_block(struct inode *inode, sector_t iblock, + struct buffer_head *bh_result, int create); /* all ocfs2_dio_end_io()'s fault */ #define ocfs2_iocb_is_rw_locked(iocb) \ test_bit(0, (unsigned long *)&iocb->private) @@ -68,8 +70,36 @@ static inline void ocfs2_iocb_set_rw_locked(struct kiocb *iocb, int level) else clear_bit(1, (unsigned long *)&iocb->private); } + +/* + * Using a named enum representing lock types in terms of #N bit stored in + * iocb->private, which is going to be used for communication between + * ocfs2_dio_end_io() and ocfs2_file_aio_write/read(). + */ +enum ocfs2_iocb_lock_bits { + OCFS2_IOCB_RW_LOCK = 0, + OCFS2_IOCB_RW_LOCK_LEVEL, + OCFS2_IOCB_SEM, + OCFS2_IOCB_UNALIGNED_IO, + OCFS2_IOCB_NUM_LOCKS +}; + #define ocfs2_iocb_clear_rw_locked(iocb) \ - clear_bit(0, (unsigned long *)&iocb->private) + clear_bit(OCFS2_IOCB_RW_LOCK, (unsigned long *)&iocb->private) #define ocfs2_iocb_rw_locked_level(iocb) \ - test_bit(1, (unsigned long *)&iocb->private) + test_bit(OCFS2_IOCB_RW_LOCK_LEVEL, (unsigned long *)&iocb->private) +#define ocfs2_iocb_set_sem_locked(iocb) \ + set_bit(OCFS2_IOCB_SEM, (unsigned long *)&iocb->private) +#define ocfs2_iocb_clear_sem_locked(iocb) \ + clear_bit(OCFS2_IOCB_SEM, (unsigned long *)&iocb->private) +#define ocfs2_iocb_is_sem_locked(iocb) \ + test_bit(OCFS2_IOCB_SEM, (unsigned long *)&iocb->private) + +#define ocfs2_iocb_set_unaligned_aio(iocb) \ + set_bit(OCFS2_IOCB_UNALIGNED_IO, (unsigned long *)&iocb->private) +#define ocfs2_iocb_clear_unaligned_aio(iocb) \ + clear_bit(OCFS2_IOCB_UNALIGNED_IO, (unsigned long *)&iocb->private) +#define ocfs2_iocb_is_unaligned_aio(iocb) \ + test_bit(OCFS2_IOCB_UNALIGNED_IO, (unsigned long *)&iocb->private) + #endif /* OCFS2_FILE_H */ diff --git a/fs/ocfs2/blockcheck.c b/fs/ocfs2/blockcheck.c index 2a947c44e59..0725e605465 100644 --- a/fs/ocfs2/blockcheck.c +++ b/fs/ocfs2/blockcheck.c @@ -22,6 +22,9 @@ #include <linux/crc32.h> #include <linux/buffer_head.h> #include <linux/bitops.h> +#include <linux/debugfs.h> +#include <linux/module.h> +#include <linux/fs.h> #include <asm/byteorder.h> #include <cluster/masklog.h> @@ -44,7 +47,7 @@ * Calculate the bit offset in the hamming code buffer based on the bit's * offset in the data buffer. Since the hamming code reserves all * power-of-two bits for parity, the data bit number and the code bit - * number are offest by all the parity bits beforehand. + * number are offset by all the parity bits beforehand. * * Recall that bit numbers in hamming code are 1-based. This function * takes the 0-based data bit from the caller. @@ -222,6 +225,155 @@ void ocfs2_hamming_fix_block(void *data, unsigned int blocksize, ocfs2_hamming_fix(data, blocksize * 8, 0, fix); } + +/* + * Debugfs handling. + */ + +#ifdef CONFIG_DEBUG_FS + +static int blockcheck_u64_get(void *data, u64 *val) +{ + *val = *(u64 *)data; + return 0; +} +DEFINE_SIMPLE_ATTRIBUTE(blockcheck_fops, blockcheck_u64_get, NULL, "%llu\n"); + +static struct dentry *blockcheck_debugfs_create(const char *name, + struct dentry *parent, + u64 *value) +{ + return debugfs_create_file(name, S_IFREG | S_IRUSR, parent, value, + &blockcheck_fops); +} + +static void ocfs2_blockcheck_debug_remove(struct ocfs2_blockcheck_stats *stats) +{ + if (stats) { + debugfs_remove(stats->b_debug_check); + stats->b_debug_check = NULL; + debugfs_remove(stats->b_debug_failure); + stats->b_debug_failure = NULL; + debugfs_remove(stats->b_debug_recover); + stats->b_debug_recover = NULL; + debugfs_remove(stats->b_debug_dir); + stats->b_debug_dir = NULL; + } +} + +static int ocfs2_blockcheck_debug_install(struct ocfs2_blockcheck_stats *stats, + struct dentry *parent) +{ + int rc = -EINVAL; + + if (!stats) + goto out; + + stats->b_debug_dir = debugfs_create_dir("blockcheck", parent); + if (!stats->b_debug_dir) + goto out; + + stats->b_debug_check = + blockcheck_debugfs_create("blocks_checked", + stats->b_debug_dir, + &stats->b_check_count); + + stats->b_debug_failure = + blockcheck_debugfs_create("checksums_failed", + stats->b_debug_dir, + &stats->b_failure_count); + + stats->b_debug_recover = + blockcheck_debugfs_create("ecc_recoveries", + stats->b_debug_dir, + &stats->b_recover_count); + if (stats->b_debug_check && stats->b_debug_failure && + stats->b_debug_recover) + rc = 0; + +out: + if (rc) + ocfs2_blockcheck_debug_remove(stats); + return rc; +} +#else +static inline int ocfs2_blockcheck_debug_install(struct ocfs2_blockcheck_stats *stats, + struct dentry *parent) +{ + return 0; +} + +static inline void ocfs2_blockcheck_debug_remove(struct ocfs2_blockcheck_stats *stats) +{ +} +#endif /* CONFIG_DEBUG_FS */ + +/* Always-called wrappers for starting and stopping the debugfs files */ +int ocfs2_blockcheck_stats_debugfs_install(struct ocfs2_blockcheck_stats *stats, + struct dentry *parent) +{ + return ocfs2_blockcheck_debug_install(stats, parent); +} + +void ocfs2_blockcheck_stats_debugfs_remove(struct ocfs2_blockcheck_stats *stats) +{ + ocfs2_blockcheck_debug_remove(stats); +} + +static void ocfs2_blockcheck_inc_check(struct ocfs2_blockcheck_stats *stats) +{ + u64 new_count; + + if (!stats) + return; + + spin_lock(&stats->b_lock); + stats->b_check_count++; + new_count = stats->b_check_count; + spin_unlock(&stats->b_lock); + + if (!new_count) + mlog(ML_NOTICE, "Block check count has wrapped\n"); +} + +static void ocfs2_blockcheck_inc_failure(struct ocfs2_blockcheck_stats *stats) +{ + u64 new_count; + + if (!stats) + return; + + spin_lock(&stats->b_lock); + stats->b_failure_count++; + new_count = stats->b_failure_count; + spin_unlock(&stats->b_lock); + + if (!new_count) + mlog(ML_NOTICE, "Checksum failure count has wrapped\n"); +} + +static void ocfs2_blockcheck_inc_recover(struct ocfs2_blockcheck_stats *stats) +{ + u64 new_count; + + if (!stats) + return; + + spin_lock(&stats->b_lock); + stats->b_recover_count++; + new_count = stats->b_recover_count; + spin_unlock(&stats->b_lock); + + if (!new_count) + mlog(ML_NOTICE, "ECC recovery count has wrapped\n"); +} + + + +/* + * These are the low-level APIs for using the ocfs2_block_check structure. + */ + /* * This function generates check information for a block. * data is the block to be checked. bc is a pointer to the @@ -251,7 +403,7 @@ void ocfs2_block_check_compute(void *data, size_t blocksize, * No ecc'd ocfs2 structure is larger than 4K, so ecc will be no * larger than 16 bits. */ - BUG_ON(ecc > USHORT_MAX); + BUG_ON(ecc > USHRT_MAX); bc->bc_crc32e = cpu_to_le32(crc); bc->bc_ecc = cpu_to_le16((u16)ecc); @@ -266,43 +418,50 @@ void ocfs2_block_check_compute(void *data, size_t blocksize, * Again, the data passed in should be the on-disk endian. */ int ocfs2_block_check_validate(void *data, size_t blocksize, - struct ocfs2_block_check *bc) + struct ocfs2_block_check *bc, + struct ocfs2_blockcheck_stats *stats) { int rc = 0; - struct ocfs2_block_check check; + u32 bc_crc32e; + u16 bc_ecc; u32 crc, ecc; - check.bc_crc32e = le32_to_cpu(bc->bc_crc32e); - check.bc_ecc = le16_to_cpu(bc->bc_ecc); + ocfs2_blockcheck_inc_check(stats); + + bc_crc32e = le32_to_cpu(bc->bc_crc32e); + bc_ecc = le16_to_cpu(bc->bc_ecc); memset(bc, 0, sizeof(struct ocfs2_block_check)); /* Fast path - if the crc32 validates, we're good to go */ crc = crc32_le(~0, data, blocksize); - if (crc == check.bc_crc32e) + if (crc == bc_crc32e) goto out; + ocfs2_blockcheck_inc_failure(stats); mlog(ML_ERROR, - "CRC32 failed: stored: %u, computed %u. Applying ECC.\n", - (unsigned int)check.bc_crc32e, (unsigned int)crc); + "CRC32 failed: stored: 0x%x, computed 0x%x. Applying ECC.\n", + (unsigned int)bc_crc32e, (unsigned int)crc); /* Ok, try ECC fixups */ ecc = ocfs2_hamming_encode_block(data, blocksize); - ocfs2_hamming_fix_block(data, blocksize, ecc ^ check.bc_ecc); + ocfs2_hamming_fix_block(data, blocksize, ecc ^ bc_ecc); /* And check the crc32 again */ crc = crc32_le(~0, data, blocksize); - if (crc == check.bc_crc32e) + if (crc == bc_crc32e) { + ocfs2_blockcheck_inc_recover(stats); goto out; + } - mlog(ML_ERROR, "Fixed CRC32 failed: stored: %u, computed %u\n", - (unsigned int)check.bc_crc32e, (unsigned int)crc); + mlog(ML_ERROR, "Fixed CRC32 failed: stored: 0x%x, computed 0x%x\n", + (unsigned int)bc_crc32e, (unsigned int)crc); rc = -EIO; out: - bc->bc_crc32e = cpu_to_le32(check.bc_crc32e); - bc->bc_ecc = cpu_to_le16(check.bc_ecc); + bc->bc_crc32e = cpu_to_le32(bc_crc32e); + bc->bc_ecc = cpu_to_le16(bc_ecc); return rc; } @@ -350,7 +509,7 @@ void ocfs2_block_check_compute_bhs(struct buffer_head **bhs, int nr, * No ecc'd ocfs2 structure is larger than 4K, so ecc will be no * larger than 16 bits. */ - BUG_ON(ecc > USHORT_MAX); + BUG_ON(ecc > USHRT_MAX); bc->bc_crc32e = cpu_to_le32(crc); bc->bc_ecc = cpu_to_le16((u16)ecc); @@ -366,10 +525,12 @@ void ocfs2_block_check_compute_bhs(struct buffer_head **bhs, int nr, * Again, the data passed in should be the on-disk endian. */ int ocfs2_block_check_validate_bhs(struct buffer_head **bhs, int nr, - struct ocfs2_block_check *bc) + struct ocfs2_block_check *bc, + struct ocfs2_blockcheck_stats *stats) { int i, rc = 0; - struct ocfs2_block_check check; + u32 bc_crc32e; + u16 bc_ecc; u32 crc, ecc, fix; BUG_ON(nr < 0); @@ -377,20 +538,23 @@ int ocfs2_block_check_validate_bhs(struct buffer_head **bhs, int nr, if (!nr) return 0; - check.bc_crc32e = le32_to_cpu(bc->bc_crc32e); - check.bc_ecc = le16_to_cpu(bc->bc_ecc); + ocfs2_blockcheck_inc_check(stats); + + bc_crc32e = le32_to_cpu(bc->bc_crc32e); + bc_ecc = le16_to_cpu(bc->bc_ecc); memset(bc, 0, sizeof(struct ocfs2_block_check)); /* Fast path - if the crc32 validates, we're good to go */ for (i = 0, crc = ~0; i < nr; i++) crc = crc32_le(crc, bhs[i]->b_data, bhs[i]->b_size); - if (crc == check.bc_crc32e) + if (crc == bc_crc32e) goto out; + ocfs2_blockcheck_inc_failure(stats); mlog(ML_ERROR, "CRC32 failed: stored: %u, computed %u. Applying ECC.\n", - (unsigned int)check.bc_crc32e, (unsigned int)crc); + (unsigned int)bc_crc32e, (unsigned int)crc); /* Ok, try ECC fixups */ for (i = 0, ecc = 0; i < nr; i++) { @@ -403,7 +567,7 @@ int ocfs2_block_check_validate_bhs(struct buffer_head **bhs, int nr, bhs[i]->b_size * 8, bhs[i]->b_size * 8 * i); } - fix = ecc ^ check.bc_ecc; + fix = ecc ^ bc_ecc; for (i = 0; i < nr; i++) { /* * Try the fix against each buffer. It will only affect @@ -416,17 +580,19 @@ int ocfs2_block_check_validate_bhs(struct buffer_head **bhs, int nr, /* And check the crc32 again */ for (i = 0, crc = ~0; i < nr; i++) crc = crc32_le(crc, bhs[i]->b_data, bhs[i]->b_size); - if (crc == check.bc_crc32e) + if (crc == bc_crc32e) { + ocfs2_blockcheck_inc_recover(stats); goto out; + } mlog(ML_ERROR, "Fixed CRC32 failed: stored: %u, computed %u\n", - (unsigned int)check.bc_crc32e, (unsigned int)crc); + (unsigned int)bc_crc32e, (unsigned int)crc); rc = -EIO; out: - bc->bc_crc32e = cpu_to_le32(check.bc_crc32e); - bc->bc_ecc = cpu_to_le16(check.bc_ecc); + bc->bc_crc32e = cpu_to_le32(bc_crc32e); + bc->bc_ecc = cpu_to_le16(bc_ecc); return rc; } @@ -448,9 +614,11 @@ int ocfs2_validate_meta_ecc(struct super_block *sb, void *data, struct ocfs2_block_check *bc) { int rc = 0; + struct ocfs2_super *osb = OCFS2_SB(sb); - if (ocfs2_meta_ecc(OCFS2_SB(sb))) - rc = ocfs2_block_check_validate(data, sb->s_blocksize, bc); + if (ocfs2_meta_ecc(osb)) + rc = ocfs2_block_check_validate(data, sb->s_blocksize, bc, + &osb->osb_ecc_stats); return rc; } @@ -468,9 +636,11 @@ int ocfs2_validate_meta_ecc_bhs(struct super_block *sb, struct ocfs2_block_check *bc) { int rc = 0; + struct ocfs2_super *osb = OCFS2_SB(sb); - if (ocfs2_meta_ecc(OCFS2_SB(sb))) - rc = ocfs2_block_check_validate_bhs(bhs, nr, bc); + if (ocfs2_meta_ecc(osb)) + rc = ocfs2_block_check_validate_bhs(bhs, nr, bc, + &osb->osb_ecc_stats); return rc; } diff --git a/fs/ocfs2/blockcheck.h b/fs/ocfs2/blockcheck.h index 70ec3feda32..d4b69febf70 100644 --- a/fs/ocfs2/blockcheck.h +++ b/fs/ocfs2/blockcheck.h @@ -21,6 +21,24 @@ #define OCFS2_BLOCKCHECK_H +/* Count errors and error correction from blockcheck.c */ +struct ocfs2_blockcheck_stats { + spinlock_t b_lock; + u64 b_check_count; /* Number of blocks we've checked */ + u64 b_failure_count; /* Number of failed checksums */ + u64 b_recover_count; /* Number of blocks fixed by ecc */ + + /* + * debugfs entries, used if this is passed to + * ocfs2_blockcheck_stats_debugfs_install() + */ + struct dentry *b_debug_dir; /* Parent of the debugfs files */ + struct dentry *b_debug_check; /* Exposes b_check_count */ + struct dentry *b_debug_failure; /* Exposes b_failure_count */ + struct dentry *b_debug_recover; /* Exposes b_recover_count */ +}; + + /* High level block API */ void ocfs2_compute_meta_ecc(struct super_block *sb, void *data, struct ocfs2_block_check *bc); @@ -37,11 +55,18 @@ int ocfs2_validate_meta_ecc_bhs(struct super_block *sb, void ocfs2_block_check_compute(void *data, size_t blocksize, struct ocfs2_block_check *bc); int ocfs2_block_check_validate(void *data, size_t blocksize, - struct ocfs2_block_check *bc); + struct ocfs2_block_check *bc, + struct ocfs2_blockcheck_stats *stats); void ocfs2_block_check_compute_bhs(struct buffer_head **bhs, int nr, struct ocfs2_block_check *bc); int ocfs2_block_check_validate_bhs(struct buffer_head **bhs, int nr, - struct ocfs2_block_check *bc); + struct ocfs2_block_check *bc, + struct ocfs2_blockcheck_stats *stats); + +/* Debug Initialization */ +int ocfs2_blockcheck_stats_debugfs_install(struct ocfs2_blockcheck_stats *stats, + struct dentry *parent); +void ocfs2_blockcheck_stats_debugfs_remove(struct ocfs2_blockcheck_stats *stats); /* * Hamming code functions diff --git a/fs/ocfs2/buffer_head_io.c b/fs/ocfs2/buffer_head_io.c index 15c8e6deee2..1edcb141f63 100644 --- a/fs/ocfs2/buffer_head_io.c +++ b/fs/ocfs2/buffer_head_io.c @@ -25,7 +25,6 @@ #include <linux/fs.h> #include <linux/types.h> -#include <linux/slab.h> #include <linux/highmem.h> #include <cluster/masklog.h> @@ -36,8 +35,8 @@ #include "inode.h" #include "journal.h" #include "uptodate.h" - #include "buffer_head_io.h" +#include "ocfs2_trace.h" /* * Bits on bh->b_state used by ocfs2. @@ -52,12 +51,11 @@ enum ocfs2_state_bits { BUFFER_FNS(NeedsValidate, needs_validate); int ocfs2_write_block(struct ocfs2_super *osb, struct buffer_head *bh, - struct inode *inode) + struct ocfs2_caching_info *ci) { int ret = 0; - mlog_entry("(bh->b_blocknr = %llu, inode=%p)\n", - (unsigned long long)bh->b_blocknr, inode); + trace_ocfs2_write_block((unsigned long long)bh->b_blocknr, ci); BUG_ON(bh->b_blocknr < OCFS2_SUPER_BLOCK_BLKNO); BUG_ON(buffer_jbd(bh)); @@ -67,10 +65,11 @@ int ocfs2_write_block(struct ocfs2_super *osb, struct buffer_head *bh, * can get modified during recovery even if read-only. */ if (ocfs2_is_hard_readonly(osb)) { ret = -EROFS; + mlog_errno(ret); goto out; } - mutex_lock(&OCFS2_I(inode)->ip_io_mutex); + ocfs2_metadata_cache_io_lock(ci); lock_buffer(bh); set_buffer_uptodate(bh); @@ -85,18 +84,17 @@ int ocfs2_write_block(struct ocfs2_super *osb, struct buffer_head *bh, wait_on_buffer(bh); if (buffer_uptodate(bh)) { - ocfs2_set_buffer_uptodate(inode, bh); + ocfs2_set_buffer_uptodate(ci, bh); } else { /* We don't need to remove the clustered uptodate * information for this bh as it's not marked locally * uptodate. */ ret = -EIO; - put_bh(bh); + mlog_errno(ret); } - mutex_unlock(&OCFS2_I(inode)->ip_io_mutex); + ocfs2_metadata_cache_io_unlock(ci); out: - mlog_exit(ret); return ret; } @@ -107,16 +105,16 @@ int ocfs2_read_blocks_sync(struct ocfs2_super *osb, u64 block, unsigned int i; struct buffer_head *bh; - if (!nr) { - mlog(ML_BH_IO, "No buffers will be read!\n"); + trace_ocfs2_read_blocks_sync((unsigned long long)block, nr); + + if (!nr) goto bail; - } for (i = 0 ; i < nr ; i++) { if (bhs[i] == NULL) { bhs[i] = sb_getblk(osb->sb, block++); if (bhs[i] == NULL) { - status = -EIO; + status = -ENOMEM; mlog_errno(status); goto bail; } @@ -124,10 +122,8 @@ int ocfs2_read_blocks_sync(struct ocfs2_super *osb, u64 block, bh = bhs[i]; if (buffer_jbd(bh)) { - mlog(ML_BH_IO, - "trying to sync read a jbd " - "managed bh (blocknr = %llu), skipping\n", - (unsigned long long)bh->b_blocknr); + trace_ocfs2_read_blocks_sync_jbd( + (unsigned long long)bh->b_blocknr); continue; } @@ -177,7 +173,7 @@ bail: return status; } -int ocfs2_read_blocks(struct inode *inode, u64 block, int nr, +int ocfs2_read_blocks(struct ocfs2_caching_info *ci, u64 block, int nr, struct buffer_head *bhs[], int flags, int (*validate)(struct super_block *sb, struct buffer_head *bh)) @@ -185,11 +181,11 @@ int ocfs2_read_blocks(struct inode *inode, u64 block, int nr, int status = 0; int i, ignore_cache = 0; struct buffer_head *bh; + struct super_block *sb = ocfs2_metadata_cache_get_super(ci); - mlog_entry("(inode=%p, block=(%llu), nr=(%d), flags=%d)\n", - inode, (unsigned long long)block, nr, flags); + trace_ocfs2_read_blocks_begin(ci, (unsigned long long)block, nr, flags); - BUG_ON(!inode); + BUG_ON(!ci); BUG_ON((flags & OCFS2_BH_READAHEAD) && (flags & OCFS2_BH_IGNORE_CACHE)); @@ -207,18 +203,17 @@ int ocfs2_read_blocks(struct inode *inode, u64 block, int nr, } if (nr == 0) { - mlog(ML_BH_IO, "No buffers will be read!\n"); status = 0; goto bail; } - mutex_lock(&OCFS2_I(inode)->ip_io_mutex); + ocfs2_metadata_cache_io_lock(ci); for (i = 0 ; i < nr ; i++) { if (bhs[i] == NULL) { - bhs[i] = sb_getblk(inode->i_sb, block++); + bhs[i] = sb_getblk(sb, block++); if (bhs[i] == NULL) { - mutex_unlock(&OCFS2_I(inode)->ip_io_mutex); - status = -EIO; + ocfs2_metadata_cache_io_unlock(ci); + status = -ENOMEM; mlog_errno(status); goto bail; } @@ -250,21 +245,19 @@ int ocfs2_read_blocks(struct inode *inode, u64 block, int nr, * before our is-it-in-flight check. */ - if (!ignore_cache && !ocfs2_buffer_uptodate(inode, bh)) { - mlog(ML_UPTODATE, - "bh (%llu), inode %llu not uptodate\n", + if (!ignore_cache && !ocfs2_buffer_uptodate(ci, bh)) { + trace_ocfs2_read_blocks_from_disk( (unsigned long long)bh->b_blocknr, - (unsigned long long)OCFS2_I(inode)->ip_blkno); + (unsigned long long)ocfs2_metadata_cache_owner(ci)); /* We're using ignore_cache here to say * "go to disk" */ ignore_cache = 1; } + trace_ocfs2_read_blocks_bh((unsigned long long)bh->b_blocknr, + ignore_cache, buffer_jbd(bh), buffer_dirty(bh)); + if (buffer_jbd(bh)) { - if (ignore_cache) - mlog(ML_BH_IO, "trying to sync read a jbd " - "managed bh (blocknr = %llu)\n", - (unsigned long long)bh->b_blocknr); continue; } @@ -272,9 +265,6 @@ int ocfs2_read_blocks(struct inode *inode, u64 block, int nr, if (buffer_dirty(bh)) { /* This should probably be a BUG, or * at least return an error. */ - mlog(ML_BH_IO, "asking me to sync read a dirty " - "buffer! (blocknr = %llu)\n", - (unsigned long long)bh->b_blocknr); continue; } @@ -283,7 +273,7 @@ int ocfs2_read_blocks(struct inode *inode, u64 block, int nr, * previously submitted request than we are * done here. */ if ((flags & OCFS2_BH_READAHEAD) - && ocfs2_buffer_read_ahead(inode, bh)) + && ocfs2_buffer_read_ahead(ci, bh)) continue; lock_buffer(bh); @@ -305,7 +295,7 @@ int ocfs2_read_blocks(struct inode *inode, u64 block, int nr, * buffer lock. */ if (!(flags & OCFS2_BH_IGNORE_CACHE) && !(flags & OCFS2_BH_READAHEAD) - && ocfs2_buffer_uptodate(inode, bh)) { + && ocfs2_buffer_uptodate(ci, bh)) { unlock_buffer(bh); continue; } @@ -327,7 +317,7 @@ int ocfs2_read_blocks(struct inode *inode, u64 block, int nr, if (!(flags & OCFS2_BH_READAHEAD)) { /* We know this can't have changed as we hold the - * inode sem. Avoid doing any work on the bh if the + * owner sem. Avoid doing any work on the bh if the * journal has it. */ if (!buffer_jbd(bh)) wait_on_buffer(bh); @@ -351,7 +341,7 @@ int ocfs2_read_blocks(struct inode *inode, u64 block, int nr, * that better not have changed */ BUG_ON(buffer_jbd(bh)); clear_buffer_needs_validate(bh); - status = validate(inode->i_sb, bh); + status = validate(sb, bh); if (status) { put_bh(bh); bhs[i] = NULL; @@ -363,18 +353,15 @@ int ocfs2_read_blocks(struct inode *inode, u64 block, int nr, /* Always set the buffer in the cache, even if it was * a forced read, or read-ahead which hasn't yet * completed. */ - ocfs2_set_buffer_uptodate(inode, bh); + ocfs2_set_buffer_uptodate(ci, bh); } - mutex_unlock(&OCFS2_I(inode)->ip_io_mutex); + ocfs2_metadata_cache_io_unlock(ci); - mlog(ML_BH_IO, "block=(%llu), nr=(%d), cached=%s, flags=0x%x\n", - (unsigned long long)block, nr, - ((flags & OCFS2_BH_IGNORE_CACHE) || ignore_cache) ? "no" : "yes", - flags); + trace_ocfs2_read_blocks_end((unsigned long long)block, nr, + flags, ignore_cache); bail: - mlog_exit(status); return status; } @@ -399,21 +386,21 @@ static void ocfs2_check_super_or_backup(struct super_block *sb, /* * Write super block and backups doesn't need to collaborate with journal, - * so we don't need to lock ip_io_mutex and inode doesn't need to bea passed + * so we don't need to lock ip_io_mutex and ci doesn't need to bea passed * into this function. */ int ocfs2_write_super_or_backup(struct ocfs2_super *osb, struct buffer_head *bh) { int ret = 0; - - mlog_entry_void(); + struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; BUG_ON(buffer_jbd(bh)); ocfs2_check_super_or_backup(osb->sb, bh->b_blocknr); if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb)) { ret = -EROFS; + mlog_errno(ret); goto out; } @@ -425,16 +412,16 @@ int ocfs2_write_super_or_backup(struct ocfs2_super *osb, get_bh(bh); /* for end_buffer_write_sync() */ bh->b_end_io = end_buffer_write_sync; + ocfs2_compute_meta_ecc(osb->sb, bh->b_data, &di->i_check); submit_bh(WRITE, bh); wait_on_buffer(bh); if (!buffer_uptodate(bh)) { ret = -EIO; - put_bh(bh); + mlog_errno(ret); } out: - mlog_exit(ret); return ret; } diff --git a/fs/ocfs2/buffer_head_io.h b/fs/ocfs2/buffer_head_io.h index c75d682dadd..b97bcc6dde7 100644 --- a/fs/ocfs2/buffer_head_io.h +++ b/fs/ocfs2/buffer_head_io.h @@ -33,7 +33,7 @@ void ocfs2_end_buffer_io_sync(struct buffer_head *bh, int ocfs2_write_block(struct ocfs2_super *osb, struct buffer_head *bh, - struct inode *inode); + struct ocfs2_caching_info *ci); int ocfs2_read_blocks_sync(struct ocfs2_super *osb, u64 block, unsigned int nr, struct buffer_head *bhs[]); @@ -44,7 +44,7 @@ int ocfs2_read_blocks_sync(struct ocfs2_super *osb, u64 block, * be set even for a READAHEAD call, as it marks the buffer for later * validation. */ -int ocfs2_read_blocks(struct inode *inode, u64 block, int nr, +int ocfs2_read_blocks(struct ocfs2_caching_info *ci, u64 block, int nr, struct buffer_head *bhs[], int flags, int (*validate)(struct super_block *sb, struct buffer_head *bh)); @@ -55,7 +55,7 @@ int ocfs2_write_super_or_backup(struct ocfs2_super *osb, #define OCFS2_BH_IGNORE_CACHE 1 #define OCFS2_BH_READAHEAD 8 -static inline int ocfs2_read_block(struct inode *inode, u64 off, +static inline int ocfs2_read_block(struct ocfs2_caching_info *ci, u64 off, struct buffer_head **bh, int (*validate)(struct super_block *sb, struct buffer_head *bh)) @@ -68,7 +68,7 @@ static inline int ocfs2_read_block(struct inode *inode, u64 off, goto bail; } - status = ocfs2_read_blocks(inode, off, 1, bh, 0, validate); + status = ocfs2_read_blocks(ci, off, 1, bh, 0, validate); bail: return status; diff --git a/fs/ocfs2/cluster/Makefile b/fs/ocfs2/cluster/Makefile index bc8c5e7d860..1aefc0350ec 100644 --- a/fs/ocfs2/cluster/Makefile +++ b/fs/ocfs2/cluster/Makefile @@ -1,4 +1,4 @@ obj-$(CONFIG_OCFS2_FS) += ocfs2_nodemanager.o ocfs2_nodemanager-objs := heartbeat.o masklog.o sys.o nodemanager.o \ - quorum.o tcp.o netdebug.o ver.o + quorum.o tcp.o netdebug.o diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c index 09cc25d0461..73039295d0d 100644 --- a/fs/ocfs2/cluster/heartbeat.c +++ b/fs/ocfs2/cluster/heartbeat.c @@ -34,6 +34,8 @@ #include <linux/crc32.h> #include <linux/time.h> #include <linux/debugfs.h> +#include <linux/slab.h> +#include <linux/bitmap.h> #include "heartbeat.h" #include "tcp.h" @@ -61,10 +63,53 @@ static unsigned long o2hb_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)]; static LIST_HEAD(o2hb_node_events); static DECLARE_WAIT_QUEUE_HEAD(o2hb_steady_queue); +/* + * In global heartbeat, we maintain a series of region bitmaps. + * - o2hb_region_bitmap allows us to limit the region number to max region. + * - o2hb_live_region_bitmap tracks live regions (seen steady iterations). + * - o2hb_quorum_region_bitmap tracks live regions that have seen all nodes + * heartbeat on it. + * - o2hb_failed_region_bitmap tracks the regions that have seen io timeouts. + */ +static unsigned long o2hb_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)]; +static unsigned long o2hb_live_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)]; +static unsigned long o2hb_quorum_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)]; +static unsigned long o2hb_failed_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)]; + +#define O2HB_DB_TYPE_LIVENODES 0 +#define O2HB_DB_TYPE_LIVEREGIONS 1 +#define O2HB_DB_TYPE_QUORUMREGIONS 2 +#define O2HB_DB_TYPE_FAILEDREGIONS 3 +#define O2HB_DB_TYPE_REGION_LIVENODES 4 +#define O2HB_DB_TYPE_REGION_NUMBER 5 +#define O2HB_DB_TYPE_REGION_ELAPSED_TIME 6 +#define O2HB_DB_TYPE_REGION_PINNED 7 +struct o2hb_debug_buf { + int db_type; + int db_size; + int db_len; + void *db_data; +}; + +static struct o2hb_debug_buf *o2hb_db_livenodes; +static struct o2hb_debug_buf *o2hb_db_liveregions; +static struct o2hb_debug_buf *o2hb_db_quorumregions; +static struct o2hb_debug_buf *o2hb_db_failedregions; + #define O2HB_DEBUG_DIR "o2hb" #define O2HB_DEBUG_LIVENODES "livenodes" +#define O2HB_DEBUG_LIVEREGIONS "live_regions" +#define O2HB_DEBUG_QUORUMREGIONS "quorum_regions" +#define O2HB_DEBUG_FAILEDREGIONS "failed_regions" +#define O2HB_DEBUG_REGION_NUMBER "num" +#define O2HB_DEBUG_REGION_ELAPSED_TIME "elapsed_time_in_ms" +#define O2HB_DEBUG_REGION_PINNED "pinned" + static struct dentry *o2hb_debug_dir; static struct dentry *o2hb_debug_livenodes; +static struct dentry *o2hb_debug_liveregions; +static struct dentry *o2hb_debug_quorumregions; +static struct dentry *o2hb_debug_failedregions; static LIST_HEAD(o2hb_all_regions); @@ -76,9 +121,48 @@ static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type); #define O2HB_DEFAULT_BLOCK_BITS 9 +enum o2hb_heartbeat_modes { + O2HB_HEARTBEAT_LOCAL = 0, + O2HB_HEARTBEAT_GLOBAL, + O2HB_HEARTBEAT_NUM_MODES, +}; + +char *o2hb_heartbeat_mode_desc[O2HB_HEARTBEAT_NUM_MODES] = { + "local", /* O2HB_HEARTBEAT_LOCAL */ + "global", /* O2HB_HEARTBEAT_GLOBAL */ +}; + unsigned int o2hb_dead_threshold = O2HB_DEFAULT_DEAD_THRESHOLD; +unsigned int o2hb_heartbeat_mode = O2HB_HEARTBEAT_LOCAL; + +/* + * o2hb_dependent_users tracks the number of registered callbacks that depend + * on heartbeat. o2net and o2dlm are two entities that register this callback. + * However only o2dlm depends on the heartbeat. It does not want the heartbeat + * to stop while a dlm domain is still active. + */ +unsigned int o2hb_dependent_users; + +/* + * In global heartbeat mode, all regions are pinned if there are one or more + * dependent users and the quorum region count is <= O2HB_PIN_CUT_OFF. All + * regions are unpinned if the region count exceeds the cut off or the number + * of dependent users falls to zero. + */ +#define O2HB_PIN_CUT_OFF 3 -/* Only sets a new threshold if there are no active regions. +/* + * In local heartbeat mode, we assume the dlm domain name to be the same as + * region uuid. This is true for domains created for the file system but not + * necessarily true for userdlm domains. This is a known limitation. + * + * In global heartbeat mode, we pin/unpin all o2hb regions. This solution + * works for both file system and userdlm domains. + */ +static int o2hb_region_pin(const char *region_uuid); +static void o2hb_region_unpin(const char *region_uuid); + +/* Only sets a new threshold if there are no active regions. * * No locking or otherwise interesting code is required for reading * o2hb_dead_threshold as it can't change once regions are active and @@ -93,6 +177,22 @@ static void o2hb_dead_threshold_set(unsigned int threshold) } } +static int o2hb_global_heartbeat_mode_set(unsigned int hb_mode) +{ + int ret = -1; + + if (hb_mode < O2HB_HEARTBEAT_NUM_MODES) { + spin_lock(&o2hb_live_lock); + if (list_empty(&o2hb_all_regions)) { + o2hb_heartbeat_mode = hb_mode; + ret = 0; + } + spin_unlock(&o2hb_live_lock); + } + + return ret; +} + struct o2hb_node_event { struct list_head hn_item; enum o2hb_callback_type hn_event_type; @@ -116,7 +216,10 @@ struct o2hb_region { struct config_item hr_item; struct list_head hr_all_item; - unsigned hr_unclean_stop:1; + unsigned hr_unclean_stop:1, + hr_aborted_start:1, + hr_item_pinned:1, + hr_item_dropped:1; /* protected by the hr_callback_sem */ struct task_struct *hr_task; @@ -134,11 +237,29 @@ struct o2hb_region { struct block_device *hr_bdev; struct o2hb_disk_slot *hr_slots; + /* live node map of this region */ + unsigned long hr_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)]; + unsigned int hr_region_num; + + struct dentry *hr_debug_dir; + struct dentry *hr_debug_livenodes; + struct dentry *hr_debug_regnum; + struct dentry *hr_debug_elapsed_time; + struct dentry *hr_debug_pinned; + struct o2hb_debug_buf *hr_db_livenodes; + struct o2hb_debug_buf *hr_db_regnum; + struct o2hb_debug_buf *hr_db_elapsed_time; + struct o2hb_debug_buf *hr_db_pinned; + /* let the person setting up hb wait for it to return until it * has reached a 'steady' state. This will be fixed when we have * a more complete api that doesn't lead to this sort of fragility. */ atomic_t hr_steady_iterations; + /* terminate o2hb thread if it does not reach steady state + * (hr_steady_iterations == 0) within hr_unsteady_iterations */ + atomic_t hr_unsteady_iterations; + char hr_dev_name[BDEVNAME_SIZE]; unsigned int hr_timeout_ms; @@ -164,20 +285,54 @@ struct o2hb_bio_wait_ctxt { static void o2hb_write_timeout(struct work_struct *work) { + int failed, quorum; + unsigned long flags; struct o2hb_region *reg = container_of(work, struct o2hb_region, hr_write_timeout_work.work); mlog(ML_ERROR, "Heartbeat write timeout to device %s after %u " "milliseconds\n", reg->hr_dev_name, - jiffies_to_msecs(jiffies - reg->hr_last_timeout_start)); + jiffies_to_msecs(jiffies - reg->hr_last_timeout_start)); + + if (o2hb_global_heartbeat_active()) { + spin_lock_irqsave(&o2hb_live_lock, flags); + if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap)) + set_bit(reg->hr_region_num, o2hb_failed_region_bitmap); + failed = bitmap_weight(o2hb_failed_region_bitmap, + O2NM_MAX_REGIONS); + quorum = bitmap_weight(o2hb_quorum_region_bitmap, + O2NM_MAX_REGIONS); + spin_unlock_irqrestore(&o2hb_live_lock, flags); + + mlog(ML_HEARTBEAT, "Number of regions %d, failed regions %d\n", + quorum, failed); + + /* + * Fence if the number of failed regions >= half the number + * of quorum regions + */ + if ((failed << 1) < quorum) + return; + } + o2quo_disk_timeout(); } static void o2hb_arm_write_timeout(struct o2hb_region *reg) { - mlog(0, "Queue write timeout for %u ms\n", O2HB_MAX_WRITE_TIMEOUT_MS); + /* Arm writeout only after thread reaches steady state */ + if (atomic_read(®->hr_steady_iterations) != 0) + return; + + mlog(ML_HEARTBEAT, "Queue write timeout for %u ms\n", + O2HB_MAX_WRITE_TIMEOUT_MS); + if (o2hb_global_heartbeat_active()) { + spin_lock(&o2hb_live_lock); + clear_bit(reg->hr_region_num, o2hb_failed_region_bitmap); + spin_unlock(&o2hb_live_lock); + } cancel_delayed_work(®->hr_write_timeout_work); reg->hr_last_timeout_start = jiffies; schedule_delayed_work(®->hr_write_timeout_work, @@ -186,8 +341,7 @@ static void o2hb_arm_write_timeout(struct o2hb_region *reg) static void o2hb_disarm_write_timeout(struct o2hb_region *reg) { - cancel_delayed_work(®->hr_write_timeout_work); - flush_scheduled_work(); + cancel_delayed_work_sync(®->hr_write_timeout_work); } static inline void o2hb_bio_wait_init(struct o2hb_bio_wait_ctxt *wc) @@ -214,11 +368,7 @@ static inline void o2hb_bio_wait_dec(struct o2hb_bio_wait_ctxt *wc, static void o2hb_wait_on_io(struct o2hb_region *reg, struct o2hb_bio_wait_ctxt *wc) { - struct address_space *mapping = reg->hr_bdev->bd_inode->i_mapping; - - blk_run_address_space(mapping); o2hb_bio_wait_dec(wc, 1); - wait_for_completion(&wc->wc_io_complete); } @@ -263,7 +413,7 @@ static struct bio *o2hb_setup_one_bio(struct o2hb_region *reg, } /* Must put everything in 512 byte sectors for the bio... */ - bio->bi_sector = (reg->hr_start_block + cs) << (bits - 9); + bio->bi_iter.bi_sector = (reg->hr_start_block + cs) << (bits - 9); bio->bi_bdev = reg->hr_bdev; bio->bi_private = wc; bio->bi_end_io = o2hb_bio_end_io; @@ -342,7 +492,7 @@ static int o2hb_issue_node_write(struct o2hb_region *reg, } atomic_inc(&write_wc->wc_num_reqs); - submit_bio(WRITE, bio); + submit_bio(WRITE_SYNC, bio); status = 0; bail: @@ -388,27 +538,50 @@ static int o2hb_verify_crc(struct o2hb_region *reg, return read == computed; } -/* We want to make sure that nobody is heartbeating on top of us -- - * this will help detect an invalid configuration. */ -static int o2hb_check_last_timestamp(struct o2hb_region *reg) +/* + * Compare the slot data with what we wrote in the last iteration. + * If the match fails, print an appropriate error message. This is to + * detect errors like... another node hearting on the same slot, + * flaky device that is losing writes, etc. + * Returns 1 if check succeeds, 0 otherwise. + */ +static int o2hb_check_own_slot(struct o2hb_region *reg) { - int node_num, ret; struct o2hb_disk_slot *slot; struct o2hb_disk_heartbeat_block *hb_block; + char *errstr; - node_num = o2nm_this_node(); - - ret = 1; - slot = ®->hr_slots[node_num]; + slot = ®->hr_slots[o2nm_this_node()]; /* Don't check on our 1st timestamp */ - if (slot->ds_last_time) { - hb_block = slot->ds_raw_block; + if (!slot->ds_last_time) + return 0; - if (le64_to_cpu(hb_block->hb_seq) != slot->ds_last_time) - ret = 0; - } + hb_block = slot->ds_raw_block; + if (le64_to_cpu(hb_block->hb_seq) == slot->ds_last_time && + le64_to_cpu(hb_block->hb_generation) == slot->ds_last_generation && + hb_block->hb_node == slot->ds_node_num) + return 1; + +#define ERRSTR1 "Another node is heartbeating on device" +#define ERRSTR2 "Heartbeat generation mismatch on device" +#define ERRSTR3 "Heartbeat sequence mismatch on device" + + if (hb_block->hb_node != slot->ds_node_num) + errstr = ERRSTR1; + else if (le64_to_cpu(hb_block->hb_generation) != + slot->ds_last_generation) + errstr = ERRSTR2; + else + errstr = ERRSTR3; - return ret; + mlog(ML_ERROR, "%s (%s): expected(%u:0x%llx, 0x%llx), " + "ondisk(%u:0x%llx, 0x%llx)\n", errstr, reg->hr_dev_name, + slot->ds_node_num, (unsigned long long)slot->ds_last_generation, + (unsigned long long)slot->ds_last_time, hb_block->hb_node, + (unsigned long long)le64_to_cpu(hb_block->hb_generation), + (unsigned long long)le64_to_cpu(hb_block->hb_seq)); + + return 0; } static inline void o2hb_prepare_block(struct o2hb_region *reg, @@ -447,11 +620,9 @@ static void o2hb_fire_callbacks(struct o2hb_callback *hbcall, struct o2nm_node *node, int idx) { - struct list_head *iter; struct o2hb_callback_func *f; - list_for_each(iter, &hbcall->list) { - f = list_entry(iter, struct o2hb_callback_func, hc_item); + list_for_each_entry(f, &hbcall->list, hc_item) { mlog(ML_HEARTBEAT, "calling funcs %p\n", f); (f->hc_func)(node, idx, f->hc_data); } @@ -460,16 +631,9 @@ static void o2hb_fire_callbacks(struct o2hb_callback *hbcall, /* Will run the list in order until we process the passed event */ static void o2hb_run_event_list(struct o2hb_node_event *queued_event) { - int empty; struct o2hb_callback *hbcall; struct o2hb_node_event *event; - spin_lock(&o2hb_live_lock); - empty = list_empty(&queued_event->hn_item); - spin_unlock(&o2hb_live_lock); - if (empty) - return; - /* Holding callback sem assures we don't alter the callback * lists when doing this, and serializes ourselves with other * processes wanting callbacks. */ @@ -511,6 +675,8 @@ static void o2hb_queue_node_event(struct o2hb_node_event *event, { assert_spin_locked(&o2hb_live_lock); + BUG_ON((!node) && (type != O2HB_NODE_DOWN_CB)); + event->hn_event_type = type; event->hn_node = node; event->hn_node_num = node_num; @@ -526,6 +692,7 @@ static void o2hb_shutdown_slot(struct o2hb_disk_slot *slot) struct o2hb_node_event event = { .hn_item = LIST_HEAD_INIT(event.hn_item), }; struct o2nm_node *node; + int queued = 0; node = o2nm_get_node_by_num(slot->ds_node_num); if (!node) @@ -543,15 +710,60 @@ static void o2hb_shutdown_slot(struct o2hb_disk_slot *slot) o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, node, slot->ds_node_num); + queued = 1; } } spin_unlock(&o2hb_live_lock); - o2hb_run_event_list(&event); + if (queued) + o2hb_run_event_list(&event); o2nm_node_put(node); } +static void o2hb_set_quorum_device(struct o2hb_region *reg) +{ + if (!o2hb_global_heartbeat_active()) + return; + + /* Prevent race with o2hb_heartbeat_group_drop_item() */ + if (kthread_should_stop()) + return; + + /* Tag region as quorum only after thread reaches steady state */ + if (atomic_read(®->hr_steady_iterations) != 0) + return; + + spin_lock(&o2hb_live_lock); + + if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap)) + goto unlock; + + /* + * A region can be added to the quorum only when it sees all + * live nodes heartbeat on it. In other words, the region has been + * added to all nodes. + */ + if (memcmp(reg->hr_live_node_bitmap, o2hb_live_node_bitmap, + sizeof(o2hb_live_node_bitmap))) + goto unlock; + + printk(KERN_NOTICE "o2hb: Region %s (%s) is now a quorum device\n", + config_item_name(®->hr_item), reg->hr_dev_name); + + set_bit(reg->hr_region_num, o2hb_quorum_region_bitmap); + + /* + * If global heartbeat active, unpin all regions if the + * region count > CUT_OFF + */ + if (bitmap_weight(o2hb_quorum_region_bitmap, + O2NM_MAX_REGIONS) > O2HB_PIN_CUT_OFF) + o2hb_region_unpin(NULL); +unlock: + spin_unlock(&o2hb_live_lock); +} + static int o2hb_check_slot(struct o2hb_region *reg, struct o2hb_disk_slot *slot) { @@ -563,14 +775,23 @@ static int o2hb_check_slot(struct o2hb_region *reg, u64 cputime; unsigned int dead_ms = o2hb_dead_threshold * O2HB_REGION_TIMEOUT_MS; unsigned int slot_dead_ms; + int tmp; + int queued = 0; memcpy(hb_block, slot->ds_raw_block, reg->hr_block_bytes); - /* Is this correct? Do we assume that the node doesn't exist - * if we're not configured for him? */ + /* + * If a node is no longer configured but is still in the livemap, we + * may need to clear that bit from the livemap. + */ node = o2nm_get_node_by_num(slot->ds_node_num); - if (!node) - return 0; + if (!node) { + spin_lock(&o2hb_live_lock); + tmp = test_bit(slot->ds_node_num, o2hb_live_node_bitmap); + spin_unlock(&o2hb_live_lock); + if (!tmp) + return 0; + } if (!o2hb_verify_crc(reg, hb_block)) { /* all paths from here will drop o2hb_live_lock for @@ -623,7 +844,7 @@ static int o2hb_check_slot(struct o2hb_region *reg, "seq %llu last %llu changed %u equal %u\n", slot->ds_node_num, (long long)slot->ds_last_generation, le32_to_cpu(hb_block->hb_cksum), - (unsigned long long)le64_to_cpu(hb_block->hb_seq), + (unsigned long long)le64_to_cpu(hb_block->hb_seq), (unsigned long long)slot->ds_last_time, slot->ds_changed_samples, slot->ds_equal_samples); @@ -637,14 +858,19 @@ fire_callbacks: mlog(ML_HEARTBEAT, "Node %d (id 0x%llx) joined my region\n", slot->ds_node_num, (long long)slot->ds_last_generation); + set_bit(slot->ds_node_num, reg->hr_live_node_bitmap); + /* first on the list generates a callback */ if (list_empty(&o2hb_live_slots[slot->ds_node_num])) { + mlog(ML_HEARTBEAT, "o2hb: Add node %d to live nodes " + "bitmap\n", slot->ds_node_num); set_bit(slot->ds_node_num, o2hb_live_node_bitmap); o2hb_queue_node_event(&event, O2HB_NODE_UP_CB, node, slot->ds_node_num); changed = 1; + queued = 1; } list_add_tail(&slot->ds_live_item, @@ -682,15 +908,21 @@ fire_callbacks: mlog(ML_HEARTBEAT, "Node %d left my region\n", slot->ds_node_num); + clear_bit(slot->ds_node_num, reg->hr_live_node_bitmap); + /* last off the live_slot generates a callback */ list_del_init(&slot->ds_live_item); if (list_empty(&o2hb_live_slots[slot->ds_node_num])) { + mlog(ML_HEARTBEAT, "o2hb: Remove node %d from live " + "nodes bitmap\n", slot->ds_node_num); clear_bit(slot->ds_node_num, o2hb_live_node_bitmap); - o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, node, - slot->ds_node_num); + /* node can be null */ + o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, + node, slot->ds_node_num); changed = 1; + queued = 1; } /* We don't clear this because the node is still @@ -706,48 +938,50 @@ fire_callbacks: out: spin_unlock(&o2hb_live_lock); - o2hb_run_event_list(&event); + if (queued) + o2hb_run_event_list(&event); - o2nm_node_put(node); + if (node) + o2nm_node_put(node); return changed; } -/* This could be faster if we just implmented a find_last_bit, but I - * don't think the circumstances warrant it. */ -static int o2hb_highest_node(unsigned long *nodes, - int numbits) +static int o2hb_highest_node(unsigned long *nodes, int numbits) { - int highest, node; - - highest = numbits; - node = -1; - while ((node = find_next_bit(nodes, numbits, node + 1)) != -1) { - if (node >= numbits) - break; - - highest = node; - } - - return highest; + return find_last_bit(nodes, numbits); } static int o2hb_do_disk_heartbeat(struct o2hb_region *reg) { - int i, ret, highest_node, change = 0; + int i, ret, highest_node; + int membership_change = 0, own_slot_ok = 0; unsigned long configured_nodes[BITS_TO_LONGS(O2NM_MAX_NODES)]; + unsigned long live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)]; struct o2hb_bio_wait_ctxt write_wc; ret = o2nm_configured_node_map(configured_nodes, sizeof(configured_nodes)); if (ret) { mlog_errno(ret); - return ret; + goto bail; + } + + /* + * If a node is not configured but is in the livemap, we still need + * to read the slot so as to be able to remove it from the livemap. + */ + o2hb_fill_node_map(live_node_bitmap, sizeof(live_node_bitmap)); + i = -1; + while ((i = find_next_bit(live_node_bitmap, + O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) { + set_bit(i, configured_nodes); } highest_node = o2hb_highest_node(configured_nodes, O2NM_MAX_NODES); if (highest_node >= O2NM_MAX_NODES) { - mlog(ML_NOTICE, "ocfs2_heartbeat: no configured nodes found!\n"); - return -EINVAL; + mlog(ML_NOTICE, "o2hb: No configured nodes found!\n"); + ret = -EINVAL; + goto bail; } /* No sense in reading the slots of nodes that don't exist @@ -757,31 +991,27 @@ static int o2hb_do_disk_heartbeat(struct o2hb_region *reg) ret = o2hb_read_slots(reg, highest_node + 1); if (ret < 0) { mlog_errno(ret); - return ret; + goto bail; } /* With an up to date view of the slots, we can check that no * other node has been improperly configured to heartbeat in * our slot. */ - if (!o2hb_check_last_timestamp(reg)) - mlog(ML_ERROR, "Device \"%s\": another node is heartbeating " - "in our slot!\n", reg->hr_dev_name); + own_slot_ok = o2hb_check_own_slot(reg); /* fill in the proper info for our next heartbeat */ o2hb_prepare_block(reg, reg->hr_generation); - /* And fire off the write. Note that we don't wait on this I/O - * until later. */ ret = o2hb_issue_node_write(reg, &write_wc); if (ret < 0) { mlog_errno(ret); - return ret; + goto bail; } i = -1; - while((i = find_next_bit(configured_nodes, O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) { - - change |= o2hb_check_slot(reg, ®->hr_slots[i]); + while((i = find_next_bit(configured_nodes, + O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) { + membership_change |= o2hb_check_slot(reg, ®->hr_slots[i]); } /* @@ -796,18 +1026,39 @@ static int o2hb_do_disk_heartbeat(struct o2hb_region *reg) * disk */ mlog(ML_ERROR, "Write error %d on device \"%s\"\n", write_wc.wc_error, reg->hr_dev_name); - return write_wc.wc_error; + ret = write_wc.wc_error; + goto bail; } - o2hb_arm_write_timeout(reg); + /* Skip disarming the timeout if own slot has stale/bad data */ + if (own_slot_ok) { + o2hb_set_quorum_device(reg); + o2hb_arm_write_timeout(reg); + } +bail: /* let the person who launched us know when things are steady */ - if (!change && (atomic_read(®->hr_steady_iterations) != 0)) { - if (atomic_dec_and_test(®->hr_steady_iterations)) + if (atomic_read(®->hr_steady_iterations) != 0) { + if (!ret && own_slot_ok && !membership_change) { + if (atomic_dec_and_test(®->hr_steady_iterations)) + wake_up(&o2hb_steady_queue); + } + } + + if (atomic_read(®->hr_steady_iterations) != 0) { + if (atomic_dec_and_test(®->hr_unsteady_iterations)) { + printk(KERN_NOTICE "o2hb: Unable to stabilize " + "heartbeart on region %s (%s)\n", + config_item_name(®->hr_item), + reg->hr_dev_name); + atomic_set(®->hr_steady_iterations, 0); + reg->hr_aborted_start = 1; wake_up(&o2hb_steady_queue); + ret = -EIO; + } } - return 0; + return ret; } /* Subtract b from a, storing the result in a. a *must* have a larger @@ -856,9 +1107,13 @@ static int o2hb_thread(void *data) mlog(ML_HEARTBEAT|ML_KTHREAD, "hb thread running\n"); - set_user_nice(current, -20); + set_user_nice(current, MIN_NICE); - while (!kthread_should_stop() && !reg->hr_unclean_stop) { + /* Pin node */ + o2nm_depend_this_node(); + + while (!kthread_should_stop() && + !reg->hr_unclean_stop && !reg->hr_aborted_start) { /* We track the time spent inside * o2hb_do_disk_heartbeat so that we avoid more than * hr_timeout_ms between disk writes. On busy systems @@ -866,20 +1121,19 @@ static int o2hb_thread(void *data) * likely to time itself out. */ do_gettimeofday(&before_hb); - i = 0; - do { - ret = o2hb_do_disk_heartbeat(reg); - } while (ret && ++i < 2); + ret = o2hb_do_disk_heartbeat(reg); do_gettimeofday(&after_hb); elapsed_msec = o2hb_elapsed_msecs(&before_hb, &after_hb); - mlog(0, "start = %lu.%lu, end = %lu.%lu, msec = %u\n", + mlog(ML_HEARTBEAT, + "start = %lu.%lu, end = %lu.%lu, msec = %u\n", before_hb.tv_sec, (unsigned long) before_hb.tv_usec, after_hb.tv_sec, (unsigned long) after_hb.tv_usec, elapsed_msec); - if (elapsed_msec < reg->hr_timeout_ms) { + if (!kthread_should_stop() && + elapsed_msec < reg->hr_timeout_ms) { /* the kthread api has blocked signals for us so no * need to record the return value. */ msleep_interruptible(reg->hr_timeout_ms - elapsed_msec); @@ -896,17 +1150,20 @@ static int o2hb_thread(void *data) * to timeout on this region when we could just as easily * write a clear generation - thus indicating to them that * this node has left this region. - * - * XXX: Should we skip this on unclean_stop? */ - o2hb_prepare_block(reg, 0); - ret = o2hb_issue_node_write(reg, &write_wc); - if (ret == 0) { - o2hb_wait_on_io(reg, &write_wc); - } else { - mlog_errno(ret); + */ + if (!reg->hr_unclean_stop && !reg->hr_aborted_start) { + o2hb_prepare_block(reg, 0); + ret = o2hb_issue_node_write(reg, &write_wc); + if (ret == 0) + o2hb_wait_on_io(reg, &write_wc); + else + mlog_errno(ret); } - mlog(ML_HEARTBEAT|ML_KTHREAD, "hb thread exiting\n"); + /* Unpin node */ + o2nm_undepend_this_node(); + + mlog(ML_HEARTBEAT|ML_KTHREAD, "o2hb thread exiting\n"); return 0; } @@ -914,21 +1171,68 @@ static int o2hb_thread(void *data) #ifdef CONFIG_DEBUG_FS static int o2hb_debug_open(struct inode *inode, struct file *file) { + struct o2hb_debug_buf *db = inode->i_private; + struct o2hb_region *reg; unsigned long map[BITS_TO_LONGS(O2NM_MAX_NODES)]; + unsigned long lts; char *buf = NULL; int i = -1; int out = 0; + /* max_nodes should be the largest bitmap we pass here */ + BUG_ON(sizeof(map) < db->db_size); + buf = kmalloc(PAGE_SIZE, GFP_KERNEL); if (!buf) goto bail; - o2hb_fill_node_map(map, sizeof(map)); + switch (db->db_type) { + case O2HB_DB_TYPE_LIVENODES: + case O2HB_DB_TYPE_LIVEREGIONS: + case O2HB_DB_TYPE_QUORUMREGIONS: + case O2HB_DB_TYPE_FAILEDREGIONS: + spin_lock(&o2hb_live_lock); + memcpy(map, db->db_data, db->db_size); + spin_unlock(&o2hb_live_lock); + break; - while ((i = find_next_bit(map, O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) + case O2HB_DB_TYPE_REGION_LIVENODES: + spin_lock(&o2hb_live_lock); + reg = (struct o2hb_region *)db->db_data; + memcpy(map, reg->hr_live_node_bitmap, db->db_size); + spin_unlock(&o2hb_live_lock); + break; + + case O2HB_DB_TYPE_REGION_NUMBER: + reg = (struct o2hb_region *)db->db_data; + out += snprintf(buf + out, PAGE_SIZE - out, "%d\n", + reg->hr_region_num); + goto done; + + case O2HB_DB_TYPE_REGION_ELAPSED_TIME: + reg = (struct o2hb_region *)db->db_data; + lts = reg->hr_last_timeout_start; + /* If 0, it has never been set before */ + if (lts) + lts = jiffies_to_msecs(jiffies - lts); + out += snprintf(buf + out, PAGE_SIZE - out, "%lu\n", lts); + goto done; + + case O2HB_DB_TYPE_REGION_PINNED: + reg = (struct o2hb_region *)db->db_data; + out += snprintf(buf + out, PAGE_SIZE - out, "%u\n", + !!reg->hr_item_pinned); + goto done; + + default: + goto done; + } + + while ((i = find_next_bit(map, db->db_len, i + 1)) < db->db_len) out += snprintf(buf + out, PAGE_SIZE - out, "%d ", i); out += snprintf(buf + out, PAGE_SIZE - out, "\n"); +done: i_size_write(inode, out); file->private_data = buf; @@ -966,7 +1270,7 @@ static ssize_t o2hb_debug_read(struct file *file, char __user *buf, } #endif /* CONFIG_DEBUG_FS */ -static struct file_operations o2hb_debug_fops = { +static const struct file_operations o2hb_debug_fops = { .open = o2hb_debug_open, .release = o2hb_debug_release, .read = o2hb_debug_read, @@ -975,10 +1279,104 @@ static struct file_operations o2hb_debug_fops = { void o2hb_exit(void) { - if (o2hb_debug_livenodes) - debugfs_remove(o2hb_debug_livenodes); - if (o2hb_debug_dir) - debugfs_remove(o2hb_debug_dir); + kfree(o2hb_db_livenodes); + kfree(o2hb_db_liveregions); + kfree(o2hb_db_quorumregions); + kfree(o2hb_db_failedregions); + debugfs_remove(o2hb_debug_failedregions); + debugfs_remove(o2hb_debug_quorumregions); + debugfs_remove(o2hb_debug_liveregions); + debugfs_remove(o2hb_debug_livenodes); + debugfs_remove(o2hb_debug_dir); +} + +static struct dentry *o2hb_debug_create(const char *name, struct dentry *dir, + struct o2hb_debug_buf **db, int db_len, + int type, int size, int len, void *data) +{ + *db = kmalloc(db_len, GFP_KERNEL); + if (!*db) + return NULL; + + (*db)->db_type = type; + (*db)->db_size = size; + (*db)->db_len = len; + (*db)->db_data = data; + + return debugfs_create_file(name, S_IFREG|S_IRUSR, dir, *db, + &o2hb_debug_fops); +} + +static int o2hb_debug_init(void) +{ + int ret = -ENOMEM; + + o2hb_debug_dir = debugfs_create_dir(O2HB_DEBUG_DIR, NULL); + if (!o2hb_debug_dir) { + mlog_errno(ret); + goto bail; + } + + o2hb_debug_livenodes = o2hb_debug_create(O2HB_DEBUG_LIVENODES, + o2hb_debug_dir, + &o2hb_db_livenodes, + sizeof(*o2hb_db_livenodes), + O2HB_DB_TYPE_LIVENODES, + sizeof(o2hb_live_node_bitmap), + O2NM_MAX_NODES, + o2hb_live_node_bitmap); + if (!o2hb_debug_livenodes) { + mlog_errno(ret); + goto bail; + } + + o2hb_debug_liveregions = o2hb_debug_create(O2HB_DEBUG_LIVEREGIONS, + o2hb_debug_dir, + &o2hb_db_liveregions, + sizeof(*o2hb_db_liveregions), + O2HB_DB_TYPE_LIVEREGIONS, + sizeof(o2hb_live_region_bitmap), + O2NM_MAX_REGIONS, + o2hb_live_region_bitmap); + if (!o2hb_debug_liveregions) { + mlog_errno(ret); + goto bail; + } + + o2hb_debug_quorumregions = + o2hb_debug_create(O2HB_DEBUG_QUORUMREGIONS, + o2hb_debug_dir, + &o2hb_db_quorumregions, + sizeof(*o2hb_db_quorumregions), + O2HB_DB_TYPE_QUORUMREGIONS, + sizeof(o2hb_quorum_region_bitmap), + O2NM_MAX_REGIONS, + o2hb_quorum_region_bitmap); + if (!o2hb_debug_quorumregions) { + mlog_errno(ret); + goto bail; + } + + o2hb_debug_failedregions = + o2hb_debug_create(O2HB_DEBUG_FAILEDREGIONS, + o2hb_debug_dir, + &o2hb_db_failedregions, + sizeof(*o2hb_db_failedregions), + O2HB_DB_TYPE_FAILEDREGIONS, + sizeof(o2hb_failed_region_bitmap), + O2NM_MAX_REGIONS, + o2hb_failed_region_bitmap); + if (!o2hb_debug_failedregions) { + mlog_errno(ret); + goto bail; + } + + ret = 0; +bail: + if (ret) + o2hb_exit(); + + return ret; } int o2hb_init(void) @@ -994,24 +1392,14 @@ int o2hb_init(void) INIT_LIST_HEAD(&o2hb_node_events); memset(o2hb_live_node_bitmap, 0, sizeof(o2hb_live_node_bitmap)); + memset(o2hb_region_bitmap, 0, sizeof(o2hb_region_bitmap)); + memset(o2hb_live_region_bitmap, 0, sizeof(o2hb_live_region_bitmap)); + memset(o2hb_quorum_region_bitmap, 0, sizeof(o2hb_quorum_region_bitmap)); + memset(o2hb_failed_region_bitmap, 0, sizeof(o2hb_failed_region_bitmap)); - o2hb_debug_dir = debugfs_create_dir(O2HB_DEBUG_DIR, NULL); - if (!o2hb_debug_dir) { - mlog_errno(-ENOMEM); - return -ENOMEM; - } - - o2hb_debug_livenodes = debugfs_create_file(O2HB_DEBUG_LIVENODES, - S_IFREG|S_IRUSR, - o2hb_debug_dir, NULL, - &o2hb_debug_fops); - if (!o2hb_debug_livenodes) { - mlog_errno(-ENOMEM); - debugfs_remove(o2hb_debug_dir); - return -ENOMEM; - } + o2hb_dependent_users = 0; - return 0; + return o2hb_debug_init(); } /* if we're already in a callback then we're already serialized by the sem */ @@ -1057,8 +1445,9 @@ static void o2hb_region_release(struct config_item *item) struct page *page; struct o2hb_region *reg = to_o2hb_region(item); - if (reg->hr_tmp_block) - kfree(reg->hr_tmp_block); + mlog(ML_HEARTBEAT, "hb region release (%s)\n", reg->hr_dev_name); + + kfree(reg->hr_tmp_block); if (reg->hr_slot_data) { for (i = 0; i < reg->hr_num_pages; i++) { @@ -1072,8 +1461,15 @@ static void o2hb_region_release(struct config_item *item) if (reg->hr_bdev) blkdev_put(reg->hr_bdev, FMODE_READ|FMODE_WRITE); - if (reg->hr_slots) - kfree(reg->hr_slots); + kfree(reg->hr_slots); + + kfree(reg->hr_db_regnum); + kfree(reg->hr_db_livenodes); + debugfs_remove(reg->hr_debug_livenodes); + debugfs_remove(reg->hr_debug_regnum); + debugfs_remove(reg->hr_debug_elapsed_time); + debugfs_remove(reg->hr_debug_pinned); + debugfs_remove(reg->hr_debug_dir); spin_lock(&o2hb_live_lock); list_del(®->hr_all_item); @@ -1291,8 +1687,6 @@ static int o2hb_populate_slot_data(struct o2hb_region *reg) struct o2hb_disk_slot *slot; struct o2hb_disk_heartbeat_block *hb_block; - mlog_entry_void(); - ret = o2hb_read_slots(reg, reg->hr_blocks); if (ret) { mlog_errno(ret); @@ -1314,7 +1708,6 @@ static int o2hb_populate_slot_data(struct o2hb_region *reg) } out: - mlog_exit(ret); return ret; } @@ -1327,9 +1720,10 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg, long fd; int sectsize; char *p = (char *)page; - struct file *filp = NULL; - struct inode *inode = NULL; + struct fd f; + struct inode *inode; ssize_t ret = -EINVAL; + int live_threshold; if (reg->hr_bdev) goto out; @@ -1346,26 +1740,26 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg, if (fd < 0 || fd >= INT_MAX) goto out; - filp = fget(fd); - if (filp == NULL) + f = fdget(fd); + if (f.file == NULL) goto out; if (reg->hr_blocks == 0 || reg->hr_start_block == 0 || reg->hr_block_bytes == 0) - goto out; + goto out2; - inode = igrab(filp->f_mapping->host); + inode = igrab(f.file->f_mapping->host); if (inode == NULL) - goto out; + goto out2; if (!S_ISBLK(inode->i_mode)) - goto out; + goto out3; - reg->hr_bdev = I_BDEV(filp->f_mapping->host); - ret = blkdev_get(reg->hr_bdev, FMODE_WRITE | FMODE_READ); + reg->hr_bdev = I_BDEV(f.file->f_mapping->host); + ret = blkdev_get(reg->hr_bdev, FMODE_WRITE | FMODE_READ, NULL); if (ret) { reg->hr_bdev = NULL; - goto out; + goto out3; } inode = NULL; @@ -1377,7 +1771,7 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg, "blocksize %u incorrect for device, expected %d", reg->hr_block_bytes, sectsize); ret = -EINVAL; - goto out; + goto out3; } o2hb_init_region_params(reg); @@ -1391,13 +1785,13 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg, ret = o2hb_map_slot_data(reg); if (ret) { mlog_errno(ret); - goto out; + goto out3; } ret = o2hb_populate_slot_data(reg); if (ret) { mlog_errno(ret); - goto out; + goto out3; } INIT_DELAYED_WORK(®->hr_write_timeout_work, o2hb_write_timeout); @@ -1406,15 +1800,28 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg, * A node is considered live after it has beat LIVE_THRESHOLD * times. We're not steady until we've given them a chance * _after_ our first read. + * The default threshold is bare minimum so as to limit the delay + * during mounts. For global heartbeat, the threshold doubled for the + * first region. */ - atomic_set(®->hr_steady_iterations, O2HB_LIVE_THRESHOLD + 1); + live_threshold = O2HB_LIVE_THRESHOLD; + if (o2hb_global_heartbeat_active()) { + spin_lock(&o2hb_live_lock); + if (bitmap_weight(o2hb_region_bitmap, O2NM_MAX_REGIONS) == 1) + live_threshold <<= 1; + spin_unlock(&o2hb_live_lock); + } + ++live_threshold; + atomic_set(®->hr_steady_iterations, live_threshold); + /* unsteady_iterations is double the steady_iterations */ + atomic_set(®->hr_unsteady_iterations, (live_threshold << 1)); hb_task = kthread_run(o2hb_thread, reg, "o2hb-%s", reg->hr_item.ci_name); if (IS_ERR(hb_task)) { ret = PTR_ERR(hb_task); mlog_errno(ret); - goto out; + goto out3; } spin_lock(&o2hb_live_lock); @@ -1424,20 +1831,20 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg, ret = wait_event_interruptible(o2hb_steady_queue, atomic_read(®->hr_steady_iterations) == 0); if (ret) { - /* We got interrupted (hello ptrace!). Clean up */ - spin_lock(&o2hb_live_lock); - hb_task = reg->hr_task; - reg->hr_task = NULL; - spin_unlock(&o2hb_live_lock); + atomic_set(®->hr_steady_iterations, 0); + reg->hr_aborted_start = 1; + } - if (hb_task) - kthread_stop(hb_task); - goto out; + if (reg->hr_aborted_start) { + ret = -EIO; + goto out3; } /* Ok, we were woken. Make sure it wasn't by drop_item() */ spin_lock(&o2hb_live_lock); hb_task = reg->hr_task; + if (o2hb_global_heartbeat_active()) + set_bit(reg->hr_region_num, o2hb_live_region_bitmap); spin_unlock(&o2hb_live_lock); if (hb_task) @@ -1445,11 +1852,15 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg, else ret = -EIO; + if (hb_task && o2hb_global_heartbeat_active()) + printk(KERN_NOTICE "o2hb: Heartbeat started on region %s (%s)\n", + config_item_name(®->hr_item), reg->hr_dev_name); + +out3: + iput(inode); +out2: + fdput(f); out: - if (filp) - fput(filp); - if (inode) - iput(inode); if (ret < 0) { if (reg->hr_bdev) { blkdev_put(reg->hr_bdev, FMODE_READ|FMODE_WRITE); @@ -1583,22 +1994,113 @@ static struct o2hb_heartbeat_group *to_o2hb_heartbeat_group(struct config_group : NULL; } +static int o2hb_debug_region_init(struct o2hb_region *reg, struct dentry *dir) +{ + int ret = -ENOMEM; + + reg->hr_debug_dir = + debugfs_create_dir(config_item_name(®->hr_item), dir); + if (!reg->hr_debug_dir) { + mlog_errno(ret); + goto bail; + } + + reg->hr_debug_livenodes = + o2hb_debug_create(O2HB_DEBUG_LIVENODES, + reg->hr_debug_dir, + &(reg->hr_db_livenodes), + sizeof(*(reg->hr_db_livenodes)), + O2HB_DB_TYPE_REGION_LIVENODES, + sizeof(reg->hr_live_node_bitmap), + O2NM_MAX_NODES, reg); + if (!reg->hr_debug_livenodes) { + mlog_errno(ret); + goto bail; + } + + reg->hr_debug_regnum = + o2hb_debug_create(O2HB_DEBUG_REGION_NUMBER, + reg->hr_debug_dir, + &(reg->hr_db_regnum), + sizeof(*(reg->hr_db_regnum)), + O2HB_DB_TYPE_REGION_NUMBER, + 0, O2NM_MAX_NODES, reg); + if (!reg->hr_debug_regnum) { + mlog_errno(ret); + goto bail; + } + + reg->hr_debug_elapsed_time = + o2hb_debug_create(O2HB_DEBUG_REGION_ELAPSED_TIME, + reg->hr_debug_dir, + &(reg->hr_db_elapsed_time), + sizeof(*(reg->hr_db_elapsed_time)), + O2HB_DB_TYPE_REGION_ELAPSED_TIME, + 0, 0, reg); + if (!reg->hr_debug_elapsed_time) { + mlog_errno(ret); + goto bail; + } + + reg->hr_debug_pinned = + o2hb_debug_create(O2HB_DEBUG_REGION_PINNED, + reg->hr_debug_dir, + &(reg->hr_db_pinned), + sizeof(*(reg->hr_db_pinned)), + O2HB_DB_TYPE_REGION_PINNED, + 0, 0, reg); + if (!reg->hr_debug_pinned) { + mlog_errno(ret); + goto bail; + } + + ret = 0; +bail: + return ret; +} + static struct config_item *o2hb_heartbeat_group_make_item(struct config_group *group, const char *name) { struct o2hb_region *reg = NULL; + int ret; reg = kzalloc(sizeof(struct o2hb_region), GFP_KERNEL); if (reg == NULL) return ERR_PTR(-ENOMEM); - config_item_init_type_name(®->hr_item, name, &o2hb_region_type); + if (strlen(name) > O2HB_MAX_REGION_NAME_LEN) { + ret = -ENAMETOOLONG; + goto free; + } spin_lock(&o2hb_live_lock); + reg->hr_region_num = 0; + if (o2hb_global_heartbeat_active()) { + reg->hr_region_num = find_first_zero_bit(o2hb_region_bitmap, + O2NM_MAX_REGIONS); + if (reg->hr_region_num >= O2NM_MAX_REGIONS) { + spin_unlock(&o2hb_live_lock); + ret = -EFBIG; + goto free; + } + set_bit(reg->hr_region_num, o2hb_region_bitmap); + } list_add_tail(®->hr_all_item, &o2hb_all_regions); spin_unlock(&o2hb_live_lock); + config_item_init_type_name(®->hr_item, name, &o2hb_region_type); + + ret = o2hb_debug_region_init(reg, o2hb_debug_dir); + if (ret) { + config_item_put(®->hr_item); + goto free; + } + return ®->hr_item; +free: + kfree(reg); + return ERR_PTR(ret); } static void o2hb_heartbeat_group_drop_item(struct config_group *group, @@ -1606,26 +2108,62 @@ static void o2hb_heartbeat_group_drop_item(struct config_group *group, { struct task_struct *hb_task; struct o2hb_region *reg = to_o2hb_region(item); + int quorum_region = 0; /* stop the thread when the user removes the region dir */ spin_lock(&o2hb_live_lock); hb_task = reg->hr_task; reg->hr_task = NULL; + reg->hr_item_dropped = 1; spin_unlock(&o2hb_live_lock); if (hb_task) kthread_stop(hb_task); + if (o2hb_global_heartbeat_active()) { + spin_lock(&o2hb_live_lock); + clear_bit(reg->hr_region_num, o2hb_region_bitmap); + clear_bit(reg->hr_region_num, o2hb_live_region_bitmap); + if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap)) + quorum_region = 1; + clear_bit(reg->hr_region_num, o2hb_quorum_region_bitmap); + spin_unlock(&o2hb_live_lock); + printk(KERN_NOTICE "o2hb: Heartbeat %s on region %s (%s)\n", + ((atomic_read(®->hr_steady_iterations) == 0) ? + "stopped" : "start aborted"), config_item_name(item), + reg->hr_dev_name); + } + /* * If we're racing a dev_write(), we need to wake them. They will * check reg->hr_task */ if (atomic_read(®->hr_steady_iterations) != 0) { + reg->hr_aborted_start = 1; atomic_set(®->hr_steady_iterations, 0); wake_up(&o2hb_steady_queue); } config_item_put(item); + + if (!o2hb_global_heartbeat_active() || !quorum_region) + return; + + /* + * If global heartbeat active and there are dependent users, + * pin all regions if quorum region count <= CUT_OFF + */ + spin_lock(&o2hb_live_lock); + + if (!o2hb_dependent_users) + goto unlock; + + if (bitmap_weight(o2hb_quorum_region_bitmap, + O2NM_MAX_REGIONS) <= O2HB_PIN_CUT_OFF) + o2hb_region_pin(NULL); + +unlock: + spin_unlock(&o2hb_live_lock); } struct o2hb_heartbeat_group_attribute { @@ -1685,6 +2223,41 @@ static ssize_t o2hb_heartbeat_group_threshold_store(struct o2hb_heartbeat_group return count; } +static +ssize_t o2hb_heartbeat_group_mode_show(struct o2hb_heartbeat_group *group, + char *page) +{ + return sprintf(page, "%s\n", + o2hb_heartbeat_mode_desc[o2hb_heartbeat_mode]); +} + +static +ssize_t o2hb_heartbeat_group_mode_store(struct o2hb_heartbeat_group *group, + const char *page, size_t count) +{ + unsigned int i; + int ret; + size_t len; + + len = (page[count - 1] == '\n') ? count - 1 : count; + if (!len) + return -EINVAL; + + for (i = 0; i < O2HB_HEARTBEAT_NUM_MODES; ++i) { + if (strnicmp(page, o2hb_heartbeat_mode_desc[i], len)) + continue; + + ret = o2hb_global_heartbeat_mode_set(i); + if (!ret) + printk(KERN_NOTICE "o2hb: Heartbeat mode set to %s\n", + o2hb_heartbeat_mode_desc[i]); + return count; + } + + return -EINVAL; + +} + static struct o2hb_heartbeat_group_attribute o2hb_heartbeat_group_attr_threshold = { .attr = { .ca_owner = THIS_MODULE, .ca_name = "dead_threshold", @@ -1693,12 +2266,21 @@ static struct o2hb_heartbeat_group_attribute o2hb_heartbeat_group_attr_threshold .store = o2hb_heartbeat_group_threshold_store, }; +static struct o2hb_heartbeat_group_attribute o2hb_heartbeat_group_attr_mode = { + .attr = { .ca_owner = THIS_MODULE, + .ca_name = "mode", + .ca_mode = S_IRUGO | S_IWUSR }, + .show = o2hb_heartbeat_group_mode_show, + .store = o2hb_heartbeat_group_mode_store, +}; + static struct configfs_attribute *o2hb_heartbeat_group_attrs[] = { &o2hb_heartbeat_group_attr_threshold.attr, + &o2hb_heartbeat_group_attr_mode.attr, NULL, }; -static struct configfs_item_operations o2hb_hearbeat_group_item_ops = { +static struct configfs_item_operations o2hb_heartbeat_group_item_ops = { .show_attribute = o2hb_heartbeat_group_show, .store_attribute = o2hb_heartbeat_group_store, }; @@ -1710,7 +2292,7 @@ static struct configfs_group_operations o2hb_heartbeat_group_group_ops = { static struct config_item_type o2hb_heartbeat_group_type = { .ct_group_ops = &o2hb_heartbeat_group_group_ops, - .ct_item_ops = &o2hb_hearbeat_group_item_ops, + .ct_item_ops = &o2hb_heartbeat_group_item_ops, .ct_attrs = o2hb_heartbeat_group_attrs, .ct_owner = THIS_MODULE, }; @@ -1742,7 +2324,7 @@ void o2hb_free_hb_set(struct config_group *group) kfree(hs); } -/* hb callback registration and issueing */ +/* hb callback registration and issuing */ static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type) { @@ -1767,70 +2349,150 @@ void o2hb_setup_callback(struct o2hb_callback_func *hc, } EXPORT_SYMBOL_GPL(o2hb_setup_callback); -static struct o2hb_region *o2hb_find_region(const char *region_uuid) +/* + * In local heartbeat mode, region_uuid passed matches the dlm domain name. + * In global heartbeat mode, region_uuid passed is NULL. + * + * In local, we only pin the matching region. In global we pin all the active + * regions. + */ +static int o2hb_region_pin(const char *region_uuid) { - struct o2hb_region *p, *reg = NULL; + int ret = 0, found = 0; + struct o2hb_region *reg; + char *uuid; assert_spin_locked(&o2hb_live_lock); - list_for_each_entry(p, &o2hb_all_regions, hr_all_item) { - if (!strcmp(region_uuid, config_item_name(&p->hr_item))) { - reg = p; - break; + list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) { + if (reg->hr_item_dropped) + continue; + + uuid = config_item_name(®->hr_item); + + /* local heartbeat */ + if (region_uuid) { + if (strcmp(region_uuid, uuid)) + continue; + found = 1; + } + + if (reg->hr_item_pinned || reg->hr_item_dropped) + goto skip_pin; + + /* Ignore ENOENT only for local hb (userdlm domain) */ + ret = o2nm_depend_item(®->hr_item); + if (!ret) { + mlog(ML_CLUSTER, "Pin region %s\n", uuid); + reg->hr_item_pinned = 1; + } else { + if (ret == -ENOENT && found) + ret = 0; + else { + mlog(ML_ERROR, "Pin region %s fails with %d\n", + uuid, ret); + break; + } } +skip_pin: + if (found) + break; } - return reg; + return ret; } -static int o2hb_region_get(const char *region_uuid) +/* + * In local heartbeat mode, region_uuid passed matches the dlm domain name. + * In global heartbeat mode, region_uuid passed is NULL. + * + * In local, we only unpin the matching region. In global we unpin all the + * active regions. + */ +static void o2hb_region_unpin(const char *region_uuid) { - int ret = 0; struct o2hb_region *reg; + char *uuid; + int found = 0; - spin_lock(&o2hb_live_lock); - - reg = o2hb_find_region(region_uuid); - if (!reg) - ret = -ENOENT; - spin_unlock(&o2hb_live_lock); - - if (ret) - goto out; + assert_spin_locked(&o2hb_live_lock); - ret = o2nm_depend_this_node(); - if (ret) - goto out; + list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) { + if (reg->hr_item_dropped) + continue; - ret = o2nm_depend_item(®->hr_item); - if (ret) - o2nm_undepend_this_node(); + uuid = config_item_name(®->hr_item); + if (region_uuid) { + if (strcmp(region_uuid, uuid)) + continue; + found = 1; + } -out: - return ret; + if (reg->hr_item_pinned) { + mlog(ML_CLUSTER, "Unpin region %s\n", uuid); + o2nm_undepend_item(®->hr_item); + reg->hr_item_pinned = 0; + } + if (found) + break; + } } -static void o2hb_region_put(const char *region_uuid) +static int o2hb_region_inc_user(const char *region_uuid) { - struct o2hb_region *reg; + int ret = 0; spin_lock(&o2hb_live_lock); - reg = o2hb_find_region(region_uuid); + /* local heartbeat */ + if (!o2hb_global_heartbeat_active()) { + ret = o2hb_region_pin(region_uuid); + goto unlock; + } + /* + * if global heartbeat active and this is the first dependent user, + * pin all regions if quorum region count <= CUT_OFF + */ + o2hb_dependent_users++; + if (o2hb_dependent_users > 1) + goto unlock; + + if (bitmap_weight(o2hb_quorum_region_bitmap, + O2NM_MAX_REGIONS) <= O2HB_PIN_CUT_OFF) + ret = o2hb_region_pin(NULL); + +unlock: spin_unlock(&o2hb_live_lock); + return ret; +} - if (reg) { - o2nm_undepend_item(®->hr_item); - o2nm_undepend_this_node(); +void o2hb_region_dec_user(const char *region_uuid) +{ + spin_lock(&o2hb_live_lock); + + /* local heartbeat */ + if (!o2hb_global_heartbeat_active()) { + o2hb_region_unpin(region_uuid); + goto unlock; } + + /* + * if global heartbeat active and there are no dependent users, + * unpin all quorum regions + */ + o2hb_dependent_users--; + if (!o2hb_dependent_users) + o2hb_region_unpin(NULL); + +unlock: + spin_unlock(&o2hb_live_lock); } int o2hb_register_callback(const char *region_uuid, struct o2hb_callback_func *hc) { - struct o2hb_callback_func *tmp; - struct list_head *iter; + struct o2hb_callback_func *f; struct o2hb_callback *hbcall; int ret; @@ -1844,17 +2506,18 @@ int o2hb_register_callback(const char *region_uuid, } if (region_uuid) { - ret = o2hb_region_get(region_uuid); - if (ret) + ret = o2hb_region_inc_user(region_uuid); + if (ret) { + mlog_errno(ret); goto out; + } } down_write(&o2hb_callback_sem); - list_for_each(iter, &hbcall->list) { - tmp = list_entry(iter, struct o2hb_callback_func, hc_item); - if (hc->hc_priority < tmp->hc_priority) { - list_add_tail(&hc->hc_item, iter); + list_for_each_entry(f, &hbcall->list, hc_item) { + if (hc->hc_priority < f->hc_priority) { + list_add_tail(&hc->hc_item, &f->hc_item); break; } } @@ -1864,7 +2527,7 @@ int o2hb_register_callback(const char *region_uuid, up_write(&o2hb_callback_sem); ret = 0; out: - mlog(ML_HEARTBEAT, "returning %d on behalf of %p for funcs %p\n", + mlog(ML_CLUSTER, "returning %d on behalf of %p for funcs %p\n", ret, __builtin_return_address(0), hc); return ret; } @@ -1875,7 +2538,7 @@ void o2hb_unregister_callback(const char *region_uuid, { BUG_ON(hc->hc_magic != O2HB_CB_MAGIC); - mlog(ML_HEARTBEAT, "on behalf of %p for funcs %p\n", + mlog(ML_CLUSTER, "on behalf of %p for funcs %p\n", __builtin_return_address(0), hc); /* XXX Can this happen _with_ a region reference? */ @@ -1883,7 +2546,7 @@ void o2hb_unregister_callback(const char *region_uuid, return; if (region_uuid) - o2hb_region_put(region_uuid); + o2hb_region_dec_user(region_uuid); down_write(&o2hb_callback_sem); @@ -1960,3 +2623,37 @@ void o2hb_stop_all_regions(void) spin_unlock(&o2hb_live_lock); } EXPORT_SYMBOL_GPL(o2hb_stop_all_regions); + +int o2hb_get_all_regions(char *region_uuids, u8 max_regions) +{ + struct o2hb_region *reg; + int numregs = 0; + char *p; + + spin_lock(&o2hb_live_lock); + + p = region_uuids; + list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) { + if (reg->hr_item_dropped) + continue; + + mlog(0, "Region: %s\n", config_item_name(®->hr_item)); + if (numregs < max_regions) { + memcpy(p, config_item_name(®->hr_item), + O2HB_MAX_REGION_NAME_LEN); + p += O2HB_MAX_REGION_NAME_LEN; + } + numregs++; + } + + spin_unlock(&o2hb_live_lock); + + return numregs; +} +EXPORT_SYMBOL_GPL(o2hb_get_all_regions); + +int o2hb_global_heartbeat_active(void) +{ + return (o2hb_heartbeat_mode == O2HB_HEARTBEAT_GLOBAL); +} +EXPORT_SYMBOL(o2hb_global_heartbeat_active); diff --git a/fs/ocfs2/cluster/heartbeat.h b/fs/ocfs2/cluster/heartbeat.h index 2f1649253b4..00ad8e8fea5 100644 --- a/fs/ocfs2/cluster/heartbeat.h +++ b/fs/ocfs2/cluster/heartbeat.h @@ -31,6 +31,8 @@ #define O2HB_REGION_TIMEOUT_MS 2000 +#define O2HB_MAX_REGION_NAME_LEN 32 + /* number of changes to be seen as live */ #define O2HB_LIVE_THRESHOLD 2 /* number of equal samples to be seen as dead */ @@ -81,5 +83,7 @@ int o2hb_check_node_heartbeating(u8 node_num); int o2hb_check_node_heartbeating_from_callback(u8 node_num); int o2hb_check_local_node_heartbeating(void); void o2hb_stop_all_regions(void); +int o2hb_get_all_regions(char *region_uuids, u8 numregions); +int o2hb_global_heartbeat_active(void); #endif /* O2CLUSTER_HEARTBEAT_H */ diff --git a/fs/ocfs2/cluster/masklog.c b/fs/ocfs2/cluster/masklog.c index 96df5416993..07ac24fd925 100644 --- a/fs/ocfs2/cluster/masklog.c +++ b/fs/ocfs2/cluster/masklog.c @@ -30,7 +30,7 @@ struct mlog_bits mlog_and_bits = MLOG_BITS_RHS(MLOG_INITIAL_AND_MASK); EXPORT_SYMBOL_GPL(mlog_and_bits); -struct mlog_bits mlog_not_bits = MLOG_BITS_RHS(MLOG_INITIAL_NOT_MASK); +struct mlog_bits mlog_not_bits = MLOG_BITS_RHS(0); EXPORT_SYMBOL_GPL(mlog_not_bits); static ssize_t mlog_mask_show(u64 mask, char *buf) @@ -80,8 +80,6 @@ struct mlog_attribute { } static struct mlog_attribute mlog_attrs[MLOG_MAX_BITS] = { - define_mask(ENTRY), - define_mask(EXIT), define_mask(TCP), define_mask(MSG), define_mask(SOCKET), @@ -93,24 +91,12 @@ static struct mlog_attribute mlog_attrs[MLOG_MAX_BITS] = { define_mask(DLM_THREAD), define_mask(DLM_MASTER), define_mask(DLM_RECOVERY), - define_mask(AIO), - define_mask(JOURNAL), - define_mask(DISK_ALLOC), - define_mask(SUPER), - define_mask(FILE_IO), - define_mask(EXTENT_MAP), define_mask(DLM_GLUE), - define_mask(BH_IO), - define_mask(UPTODATE), - define_mask(NAMEI), - define_mask(INODE), define_mask(VOTE), - define_mask(DCACHE), define_mask(CONN), define_mask(QUORUM), - define_mask(EXPORT), - define_mask(XATTR), - define_mask(QUOTA), + define_mask(BASTS), + define_mask(CLUSTER), define_mask(ERROR), define_mask(NOTICE), define_mask(KTHREAD), @@ -134,7 +120,7 @@ static ssize_t mlog_store(struct kobject *obj, struct attribute *attr, return mlog_mask_store(mlog_attr->mask, buf, count); } -static struct sysfs_ops mlog_attr_ops = { +static const struct sysfs_ops mlog_attr_ops = { .show = mlog_show, .store = mlog_store, }; diff --git a/fs/ocfs2/cluster/masklog.h b/fs/ocfs2/cluster/masklog.h index 7e72a81bc2d..2260fb9e650 100644 --- a/fs/ocfs2/cluster/masklog.h +++ b/fs/ocfs2/cluster/masklog.h @@ -48,79 +48,64 @@ * only emit the appropriage printk() when the caller passes in a constant * mask, as is almost always the case. * - * All this bitmask nonsense is hidden from the /proc interface so that Joel - * doesn't have an aneurism. Reading the file gives a straight forward - * indication of which bits are on or off: - * ENTRY off - * EXIT off + * All this bitmask nonsense is managed from the files under + * /sys/fs/o2cb/logmask/. Reading the files gives a straightforward + * indication of which bits are allowed (allow) or denied (off/deny). + * ENTRY deny + * EXIT deny * TCP off * MSG off * SOCKET off - * ERROR off - * NOTICE on + * ERROR allow + * NOTICE allow * * Writing changes the state of a given bit and requires a strictly formatted * single write() call: * - * write(fd, "ENTRY on", 8); + * write(fd, "allow", 5); * - * would turn the entry bit on. "1" is also accepted in the place of "on", and - * "off" and "0" behave as expected. + * Echoing allow/deny/off string into the logmask files can flip the bits + * on or off as expected; here is the bash script for example: * - * Some trivial shell can flip all the bits on or off: + * log_mask="/sys/fs/o2cb/log_mask" + * for node in ENTRY EXIT TCP MSG SOCKET ERROR NOTICE; do + * echo allow >"$log_mask"/"$node" + * done * - * log_mask="/proc/fs/ocfs2_nodemanager/log_mask" - * cat $log_mask | ( - * while read bit status; do - * # $1 is "on" or "off", say - * echo "$bit $1" > $log_mask - * done - * ) + * The debugfs.ocfs2 tool can also flip the bits with the -l option: + * + * debugfs.ocfs2 -l TCP allow */ /* for task_struct */ #include <linux/sched.h> /* bits that are frequently given and infrequently matched in the low word */ -/* NOTE: If you add a flag, you need to also update mlog.c! */ -#define ML_ENTRY 0x0000000000000001ULL /* func call entry */ -#define ML_EXIT 0x0000000000000002ULL /* func call exit */ -#define ML_TCP 0x0000000000000004ULL /* net cluster/tcp.c */ -#define ML_MSG 0x0000000000000008ULL /* net network messages */ -#define ML_SOCKET 0x0000000000000010ULL /* net socket lifetime */ -#define ML_HEARTBEAT 0x0000000000000020ULL /* hb all heartbeat tracking */ -#define ML_HB_BIO 0x0000000000000040ULL /* hb io tracing */ -#define ML_DLMFS 0x0000000000000080ULL /* dlm user dlmfs */ -#define ML_DLM 0x0000000000000100ULL /* dlm general debugging */ -#define ML_DLM_DOMAIN 0x0000000000000200ULL /* dlm domain debugging */ -#define ML_DLM_THREAD 0x0000000000000400ULL /* dlm domain thread */ -#define ML_DLM_MASTER 0x0000000000000800ULL /* dlm master functions */ -#define ML_DLM_RECOVERY 0x0000000000001000ULL /* dlm master functions */ -#define ML_AIO 0x0000000000002000ULL /* ocfs2 aio read and write */ -#define ML_JOURNAL 0x0000000000004000ULL /* ocfs2 journalling functions */ -#define ML_DISK_ALLOC 0x0000000000008000ULL /* ocfs2 disk allocation */ -#define ML_SUPER 0x0000000000010000ULL /* ocfs2 mount / umount */ -#define ML_FILE_IO 0x0000000000020000ULL /* ocfs2 file I/O */ -#define ML_EXTENT_MAP 0x0000000000040000ULL /* ocfs2 extent map caching */ -#define ML_DLM_GLUE 0x0000000000080000ULL /* ocfs2 dlm glue layer */ -#define ML_BH_IO 0x0000000000100000ULL /* ocfs2 buffer I/O */ -#define ML_UPTODATE 0x0000000000200000ULL /* ocfs2 caching sequence #'s */ -#define ML_NAMEI 0x0000000000400000ULL /* ocfs2 directory / namespace */ -#define ML_INODE 0x0000000000800000ULL /* ocfs2 inode manipulation */ -#define ML_VOTE 0x0000000001000000ULL /* ocfs2 node messaging */ -#define ML_DCACHE 0x0000000002000000ULL /* ocfs2 dcache operations */ -#define ML_CONN 0x0000000004000000ULL /* net connection management */ -#define ML_QUORUM 0x0000000008000000ULL /* net connection quorum */ -#define ML_EXPORT 0x0000000010000000ULL /* ocfs2 export operations */ -#define ML_XATTR 0x0000000020000000ULL /* ocfs2 extended attributes */ -#define ML_QUOTA 0x0000000040000000ULL /* ocfs2 quota operations */ +/* NOTE: If you add a flag, you need to also update masklog.c! */ +#define ML_TCP 0x0000000000000001ULL /* net cluster/tcp.c */ +#define ML_MSG 0x0000000000000002ULL /* net network messages */ +#define ML_SOCKET 0x0000000000000004ULL /* net socket lifetime */ +#define ML_HEARTBEAT 0x0000000000000008ULL /* hb all heartbeat tracking */ +#define ML_HB_BIO 0x0000000000000010ULL /* hb io tracing */ +#define ML_DLMFS 0x0000000000000020ULL /* dlm user dlmfs */ +#define ML_DLM 0x0000000000000040ULL /* dlm general debugging */ +#define ML_DLM_DOMAIN 0x0000000000000080ULL /* dlm domain debugging */ +#define ML_DLM_THREAD 0x0000000000000100ULL /* dlm domain thread */ +#define ML_DLM_MASTER 0x0000000000000200ULL /* dlm master functions */ +#define ML_DLM_RECOVERY 0x0000000000000400ULL /* dlm master functions */ +#define ML_DLM_GLUE 0x0000000000000800ULL /* ocfs2 dlm glue layer */ +#define ML_VOTE 0x0000000000001000ULL /* ocfs2 node messaging */ +#define ML_CONN 0x0000000000002000ULL /* net connection management */ +#define ML_QUORUM 0x0000000000004000ULL /* net connection quorum */ +#define ML_BASTS 0x0000000000008000ULL /* dlmglue asts and basts */ +#define ML_CLUSTER 0x0000000000010000ULL /* cluster stack */ + /* bits that are infrequently given and frequently matched in the high word */ -#define ML_ERROR 0x0000000100000000ULL /* sent to KERN_ERR */ -#define ML_NOTICE 0x0000000200000000ULL /* setn to KERN_NOTICE */ -#define ML_KTHREAD 0x0000000400000000ULL /* kernel thread activity */ +#define ML_ERROR 0x1000000000000000ULL /* sent to KERN_ERR */ +#define ML_NOTICE 0x2000000000000000ULL /* setn to KERN_NOTICE */ +#define ML_KTHREAD 0x4000000000000000ULL /* kernel thread activity */ #define MLOG_INITIAL_AND_MASK (ML_ERROR|ML_NOTICE) -#define MLOG_INITIAL_NOT_MASK (ML_ENTRY|ML_EXIT) #ifndef MLOG_MASK_PREFIX #define MLOG_MASK_PREFIX 0 #endif @@ -194,9 +179,9 @@ extern struct mlog_bits mlog_and_bits, mlog_not_bits; * previous token if args expands to nothing. */ #define __mlog_printk(level, fmt, args...) \ - printk(level "(%u,%lu):%s:%d " fmt, task_pid_nr(current), \ - __mlog_cpu_guess, __PRETTY_FUNCTION__, __LINE__ , \ - ##args) + printk(level "(%s,%u,%lu):%s:%d " fmt, current->comm, \ + task_pid_nr(current), __mlog_cpu_guess, \ + __PRETTY_FUNCTION__, __LINE__ , ##args) #define mlog(mask, fmt, args...) do { \ u64 __m = MLOG_MASK_PREFIX | (mask); \ @@ -214,62 +199,11 @@ extern struct mlog_bits mlog_and_bits, mlog_not_bits; #define mlog_errno(st) do { \ int _st = (st); \ if (_st != -ERESTARTSYS && _st != -EINTR && \ - _st != AOP_TRUNCATED_PAGE && _st != -ENOSPC) \ + _st != AOP_TRUNCATED_PAGE && _st != -ENOSPC && \ + _st != -EDQUOT) \ mlog(ML_ERROR, "status = %lld\n", (long long)_st); \ } while (0) -#if defined(CONFIG_OCFS2_DEBUG_MASKLOG) -#define mlog_entry(fmt, args...) do { \ - mlog(ML_ENTRY, "ENTRY:" fmt , ##args); \ -} while (0) - -#define mlog_entry_void() do { \ - mlog(ML_ENTRY, "ENTRY:\n"); \ -} while (0) - -/* - * We disable this for sparse. - */ -#if !defined(__CHECKER__) -#define mlog_exit(st) do { \ - if (__builtin_types_compatible_p(typeof(st), unsigned long)) \ - mlog(ML_EXIT, "EXIT: %lu\n", (unsigned long) (st)); \ - else if (__builtin_types_compatible_p(typeof(st), signed long)) \ - mlog(ML_EXIT, "EXIT: %ld\n", (signed long) (st)); \ - else if (__builtin_types_compatible_p(typeof(st), unsigned int) \ - || __builtin_types_compatible_p(typeof(st), unsigned short) \ - || __builtin_types_compatible_p(typeof(st), unsigned char)) \ - mlog(ML_EXIT, "EXIT: %u\n", (unsigned int) (st)); \ - else if (__builtin_types_compatible_p(typeof(st), signed int) \ - || __builtin_types_compatible_p(typeof(st), signed short) \ - || __builtin_types_compatible_p(typeof(st), signed char)) \ - mlog(ML_EXIT, "EXIT: %d\n", (signed int) (st)); \ - else if (__builtin_types_compatible_p(typeof(st), long long)) \ - mlog(ML_EXIT, "EXIT: %lld\n", (long long) (st)); \ - else \ - mlog(ML_EXIT, "EXIT: %llu\n", (unsigned long long) (st)); \ -} while (0) -#else -#define mlog_exit(st) do { \ - mlog(ML_EXIT, "EXIT: %lld\n", (long long) (st)); \ -} while (0) -#endif - -#define mlog_exit_ptr(ptr) do { \ - mlog(ML_EXIT, "EXIT: %p\n", ptr); \ -} while (0) - -#define mlog_exit_void() do { \ - mlog(ML_EXIT, "EXIT\n"); \ -} while (0) -#else -#define mlog_entry(...) do { } while (0) -#define mlog_entry_void(...) do { } while (0) -#define mlog_exit(...) do { } while (0) -#define mlog_exit_ptr(...) do { } while (0) -#define mlog_exit_void(...) do { } while (0) -#endif /* defined(CONFIG_OCFS2_DEBUG_MASKLOG) */ - #define mlog_bug_on_msg(cond, fmt, args...) do { \ if (cond) { \ mlog(ML_ERROR, "bug expression: " #cond "\n"); \ diff --git a/fs/ocfs2/cluster/netdebug.c b/fs/ocfs2/cluster/netdebug.c index f8424874fa0..73ba81928bc 100644 --- a/fs/ocfs2/cluster/netdebug.c +++ b/fs/ocfs2/cluster/netdebug.c @@ -46,10 +46,17 @@ #define O2NET_DEBUG_DIR "o2net" #define SC_DEBUG_NAME "sock_containers" #define NST_DEBUG_NAME "send_tracking" +#define STATS_DEBUG_NAME "stats" +#define NODES_DEBUG_NAME "connected_nodes" + +#define SHOW_SOCK_CONTAINERS 0 +#define SHOW_SOCK_STATS 1 static struct dentry *o2net_dentry; static struct dentry *sc_dentry; static struct dentry *nst_dentry; +static struct dentry *stats_dentry; +static struct dentry *nodes_dentry; static DEFINE_SPINLOCK(o2net_debug_lock); @@ -123,37 +130,42 @@ static void *nst_seq_next(struct seq_file *seq, void *v, loff_t *pos) static int nst_seq_show(struct seq_file *seq, void *v) { struct o2net_send_tracking *nst, *dummy_nst = seq->private; + ktime_t now; + s64 sock, send, status; spin_lock(&o2net_debug_lock); nst = next_nst(dummy_nst); + if (!nst) + goto out; - if (nst != NULL) { - /* get_task_comm isn't exported. oh well. */ - seq_printf(seq, "%p:\n" - " pid: %lu\n" - " tgid: %lu\n" - " process name: %s\n" - " node: %u\n" - " sc: %p\n" - " message id: %d\n" - " message type: %u\n" - " message key: 0x%08x\n" - " sock acquiry: %lu.%ld\n" - " send start: %lu.%ld\n" - " wait start: %lu.%ld\n", - nst, (unsigned long)nst->st_task->pid, - (unsigned long)nst->st_task->tgid, - nst->st_task->comm, nst->st_node, - nst->st_sc, nst->st_id, nst->st_msg_type, - nst->st_msg_key, - nst->st_sock_time.tv_sec, - (long)nst->st_sock_time.tv_usec, - nst->st_send_time.tv_sec, - (long)nst->st_send_time.tv_usec, - nst->st_status_time.tv_sec, - (long)nst->st_status_time.tv_usec); - } + now = ktime_get(); + sock = ktime_to_us(ktime_sub(now, nst->st_sock_time)); + send = ktime_to_us(ktime_sub(now, nst->st_send_time)); + status = ktime_to_us(ktime_sub(now, nst->st_status_time)); + + /* get_task_comm isn't exported. oh well. */ + seq_printf(seq, "%p:\n" + " pid: %lu\n" + " tgid: %lu\n" + " process name: %s\n" + " node: %u\n" + " sc: %p\n" + " message id: %d\n" + " message type: %u\n" + " message key: 0x%08x\n" + " sock acquiry: %lld usecs ago\n" + " send start: %lld usecs ago\n" + " wait start: %lld usecs ago\n", + nst, (unsigned long)task_pid_nr(nst->st_task), + (unsigned long)nst->st_task->tgid, + nst->st_task->comm, nst->st_node, + nst->st_sc, nst->st_id, nst->st_msg_type, + nst->st_msg_key, + (long long)sock, + (long long)send, + (long long)status); +out: spin_unlock(&o2net_debug_lock); return 0; @@ -163,7 +175,7 @@ static void nst_seq_stop(struct seq_file *seq, void *v) { } -static struct seq_operations nst_seq_ops = { +static const struct seq_operations nst_seq_ops = { .start = nst_seq_start, .next = nst_seq_next, .stop = nst_seq_stop, @@ -207,7 +219,7 @@ static int nst_fop_release(struct inode *inode, struct file *file) return seq_release_private(inode, file); } -static struct file_operations nst_seq_fops = { +static const struct file_operations nst_seq_fops = { .open = nst_fop_open, .read = seq_read, .llseek = seq_lseek, @@ -228,6 +240,11 @@ void o2net_debug_del_sc(struct o2net_sock_container *sc) spin_unlock(&o2net_debug_lock); } +struct o2net_sock_debug { + int dbg_ctxt; + struct o2net_sock_container *dbg_sock; +}; + static struct o2net_sock_container *next_sc(struct o2net_sock_container *sc_start) { @@ -253,7 +270,8 @@ static struct o2net_sock_container static void *sc_seq_start(struct seq_file *seq, loff_t *pos) { - struct o2net_sock_container *sc, *dummy_sc = seq->private; + struct o2net_sock_debug *sd = seq->private; + struct o2net_sock_container *sc, *dummy_sc = sd->dbg_sock; spin_lock(&o2net_debug_lock); sc = next_sc(dummy_sc); @@ -264,7 +282,8 @@ static void *sc_seq_start(struct seq_file *seq, loff_t *pos) static void *sc_seq_next(struct seq_file *seq, void *v, loff_t *pos) { - struct o2net_sock_container *sc, *dummy_sc = seq->private; + struct o2net_sock_debug *sd = seq->private; + struct o2net_sock_container *sc, *dummy_sc = sd->dbg_sock; spin_lock(&o2net_debug_lock); sc = next_sc(dummy_sc); @@ -276,65 +295,107 @@ static void *sc_seq_next(struct seq_file *seq, void *v, loff_t *pos) return sc; /* unused, just needs to be null when done */ } -#define TV_SEC_USEC(TV) TV.tv_sec, (long)TV.tv_usec +#ifdef CONFIG_OCFS2_FS_STATS +# define sc_send_count(_s) ((_s)->sc_send_count) +# define sc_recv_count(_s) ((_s)->sc_recv_count) +# define sc_tv_acquiry_total_ns(_s) (ktime_to_ns((_s)->sc_tv_acquiry_total)) +# define sc_tv_send_total_ns(_s) (ktime_to_ns((_s)->sc_tv_send_total)) +# define sc_tv_status_total_ns(_s) (ktime_to_ns((_s)->sc_tv_status_total)) +# define sc_tv_process_total_ns(_s) (ktime_to_ns((_s)->sc_tv_process_total)) +#else +# define sc_send_count(_s) (0U) +# define sc_recv_count(_s) (0U) +# define sc_tv_acquiry_total_ns(_s) (0LL) +# define sc_tv_send_total_ns(_s) (0LL) +# define sc_tv_status_total_ns(_s) (0LL) +# define sc_tv_process_total_ns(_s) (0LL) +#endif + +/* So that debugfs.ocfs2 can determine which format is being used */ +#define O2NET_STATS_STR_VERSION 1 +static void sc_show_sock_stats(struct seq_file *seq, + struct o2net_sock_container *sc) +{ + if (!sc) + return; + + seq_printf(seq, "%d,%u,%lu,%lld,%lld,%lld,%lu,%lld\n", O2NET_STATS_STR_VERSION, + sc->sc_node->nd_num, (unsigned long)sc_send_count(sc), + (long long)sc_tv_acquiry_total_ns(sc), + (long long)sc_tv_send_total_ns(sc), + (long long)sc_tv_status_total_ns(sc), + (unsigned long)sc_recv_count(sc), + (long long)sc_tv_process_total_ns(sc)); +} + +static void sc_show_sock_container(struct seq_file *seq, + struct o2net_sock_container *sc) +{ + struct inet_sock *inet = NULL; + __be32 saddr = 0, daddr = 0; + __be16 sport = 0, dport = 0; + + if (!sc) + return; + + if (sc->sc_sock) { + inet = inet_sk(sc->sc_sock->sk); + /* the stack's structs aren't sparse endian clean */ + saddr = (__force __be32)inet->inet_saddr; + daddr = (__force __be32)inet->inet_daddr; + sport = (__force __be16)inet->inet_sport; + dport = (__force __be16)inet->inet_dport; + } + + /* XXX sigh, inet-> doesn't have sparse annotation so any + * use of it here generates a warning with -Wbitwise */ + seq_printf(seq, "%p:\n" + " krefs: %d\n" + " sock: %pI4:%u -> " + "%pI4:%u\n" + " remote node: %s\n" + " page off: %zu\n" + " handshake ok: %u\n" + " timer: %lld usecs\n" + " data ready: %lld usecs\n" + " advance start: %lld usecs\n" + " advance stop: %lld usecs\n" + " func start: %lld usecs\n" + " func stop: %lld usecs\n" + " func key: 0x%08x\n" + " func type: %u\n", + sc, + atomic_read(&sc->sc_kref.refcount), + &saddr, inet ? ntohs(sport) : 0, + &daddr, inet ? ntohs(dport) : 0, + sc->sc_node->nd_name, + sc->sc_page_off, + sc->sc_handshake_ok, + (long long)ktime_to_us(sc->sc_tv_timer), + (long long)ktime_to_us(sc->sc_tv_data_ready), + (long long)ktime_to_us(sc->sc_tv_advance_start), + (long long)ktime_to_us(sc->sc_tv_advance_stop), + (long long)ktime_to_us(sc->sc_tv_func_start), + (long long)ktime_to_us(sc->sc_tv_func_stop), + sc->sc_msg_key, + sc->sc_msg_type); +} static int sc_seq_show(struct seq_file *seq, void *v) { - struct o2net_sock_container *sc, *dummy_sc = seq->private; + struct o2net_sock_debug *sd = seq->private; + struct o2net_sock_container *sc, *dummy_sc = sd->dbg_sock; spin_lock(&o2net_debug_lock); sc = next_sc(dummy_sc); - if (sc != NULL) { - struct inet_sock *inet = NULL; - - __be32 saddr = 0, daddr = 0; - __be16 sport = 0, dport = 0; - - if (sc->sc_sock) { - inet = inet_sk(sc->sc_sock->sk); - /* the stack's structs aren't sparse endian clean */ - saddr = (__force __be32)inet->saddr; - daddr = (__force __be32)inet->daddr; - sport = (__force __be16)inet->sport; - dport = (__force __be16)inet->dport; - } - - /* XXX sigh, inet-> doesn't have sparse annotation so any - * use of it here generates a warning with -Wbitwise */ - seq_printf(seq, "%p:\n" - " krefs: %d\n" - " sock: %pI4:%u -> " - "%pI4:%u\n" - " remote node: %s\n" - " page off: %zu\n" - " handshake ok: %u\n" - " timer: %lu.%ld\n" - " data ready: %lu.%ld\n" - " advance start: %lu.%ld\n" - " advance stop: %lu.%ld\n" - " func start: %lu.%ld\n" - " func stop: %lu.%ld\n" - " func key: %u\n" - " func type: %u\n", - sc, - atomic_read(&sc->sc_kref.refcount), - &saddr, inet ? ntohs(sport) : 0, - &daddr, inet ? ntohs(dport) : 0, - sc->sc_node->nd_name, - sc->sc_page_off, - sc->sc_handshake_ok, - TV_SEC_USEC(sc->sc_tv_timer), - TV_SEC_USEC(sc->sc_tv_data_ready), - TV_SEC_USEC(sc->sc_tv_advance_start), - TV_SEC_USEC(sc->sc_tv_advance_stop), - TV_SEC_USEC(sc->sc_tv_func_start), - TV_SEC_USEC(sc->sc_tv_func_stop), - sc->sc_msg_key, - sc->sc_msg_type); + if (sc) { + if (sd->dbg_ctxt == SHOW_SOCK_CONTAINERS) + sc_show_sock_container(seq, sc); + else + sc_show_sock_stats(seq, sc); } - spin_unlock(&o2net_debug_lock); return 0; @@ -344,14 +405,14 @@ static void sc_seq_stop(struct seq_file *seq, void *v) { } -static struct seq_operations sc_seq_ops = { +static const struct seq_operations sc_seq_ops = { .start = sc_seq_start, .next = sc_seq_next, .stop = sc_seq_stop, .show = sc_seq_show, }; -static int sc_fop_open(struct inode *inode, struct file *file) +static int sc_common_open(struct file *file, struct o2net_sock_debug *sd) { struct o2net_sock_container *dummy_sc; struct seq_file *seq; @@ -369,7 +430,8 @@ static int sc_fop_open(struct inode *inode, struct file *file) goto out; seq = file->private_data; - seq->private = dummy_sc; + seq->private = sd; + sd->dbg_sock = dummy_sc; o2net_debug_add_sc(dummy_sc); dummy_sc = NULL; @@ -382,62 +444,136 @@ out: static int sc_fop_release(struct inode *inode, struct file *file) { struct seq_file *seq = file->private_data; - struct o2net_sock_container *dummy_sc = seq->private; + struct o2net_sock_debug *sd = seq->private; + struct o2net_sock_container *dummy_sc = sd->dbg_sock; o2net_debug_del_sc(dummy_sc); return seq_release_private(inode, file); } -static struct file_operations sc_seq_fops = { +static int stats_fop_open(struct inode *inode, struct file *file) +{ + struct o2net_sock_debug *sd; + + sd = kmalloc(sizeof(struct o2net_sock_debug), GFP_KERNEL); + if (sd == NULL) + return -ENOMEM; + + sd->dbg_ctxt = SHOW_SOCK_STATS; + sd->dbg_sock = NULL; + + return sc_common_open(file, sd); +} + +static const struct file_operations stats_seq_fops = { + .open = stats_fop_open, + .read = seq_read, + .llseek = seq_lseek, + .release = sc_fop_release, +}; + +static int sc_fop_open(struct inode *inode, struct file *file) +{ + struct o2net_sock_debug *sd; + + sd = kmalloc(sizeof(struct o2net_sock_debug), GFP_KERNEL); + if (sd == NULL) + return -ENOMEM; + + sd->dbg_ctxt = SHOW_SOCK_CONTAINERS; + sd->dbg_sock = NULL; + + return sc_common_open(file, sd); +} + +static const struct file_operations sc_seq_fops = { .open = sc_fop_open, .read = seq_read, .llseek = seq_lseek, .release = sc_fop_release, }; -int o2net_debugfs_init(void) +static int o2net_fill_bitmap(char *buf, int len) { - o2net_dentry = debugfs_create_dir(O2NET_DEBUG_DIR, NULL); - if (!o2net_dentry) { - mlog_errno(-ENOMEM); - goto bail; - } + unsigned long map[BITS_TO_LONGS(O2NM_MAX_NODES)]; + int i = -1, out = 0; - nst_dentry = debugfs_create_file(NST_DEBUG_NAME, S_IFREG|S_IRUSR, - o2net_dentry, NULL, - &nst_seq_fops); - if (!nst_dentry) { - mlog_errno(-ENOMEM); - goto bail; - } + o2net_fill_node_map(map, sizeof(map)); - sc_dentry = debugfs_create_file(SC_DEBUG_NAME, S_IFREG|S_IRUSR, - o2net_dentry, NULL, - &sc_seq_fops); - if (!sc_dentry) { - mlog_errno(-ENOMEM); - goto bail; - } + while ((i = find_next_bit(map, O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) + out += snprintf(buf + out, PAGE_SIZE - out, "%d ", i); + out += snprintf(buf + out, PAGE_SIZE - out, "\n"); + + return out; +} + +static int nodes_fop_open(struct inode *inode, struct file *file) +{ + char *buf; + + buf = kmalloc(PAGE_SIZE, GFP_KERNEL); + if (!buf) + return -ENOMEM; + + i_size_write(inode, o2net_fill_bitmap(buf, PAGE_SIZE)); + + file->private_data = buf; return 0; -bail: - if (sc_dentry) - debugfs_remove(sc_dentry); - if (nst_dentry) - debugfs_remove(nst_dentry); - if (o2net_dentry) - debugfs_remove(o2net_dentry); - return -ENOMEM; } +static int o2net_debug_release(struct inode *inode, struct file *file) +{ + kfree(file->private_data); + return 0; +} + +static ssize_t o2net_debug_read(struct file *file, char __user *buf, + size_t nbytes, loff_t *ppos) +{ + return simple_read_from_buffer(buf, nbytes, ppos, file->private_data, + i_size_read(file->f_mapping->host)); +} + +static const struct file_operations nodes_fops = { + .open = nodes_fop_open, + .release = o2net_debug_release, + .read = o2net_debug_read, + .llseek = generic_file_llseek, +}; + void o2net_debugfs_exit(void) { - if (sc_dentry) - debugfs_remove(sc_dentry); - if (nst_dentry) - debugfs_remove(nst_dentry); + debugfs_remove(nodes_dentry); + debugfs_remove(stats_dentry); + debugfs_remove(sc_dentry); + debugfs_remove(nst_dentry); + debugfs_remove(o2net_dentry); +} + +int o2net_debugfs_init(void) +{ + umode_t mode = S_IFREG|S_IRUSR; + + o2net_dentry = debugfs_create_dir(O2NET_DEBUG_DIR, NULL); if (o2net_dentry) - debugfs_remove(o2net_dentry); + nst_dentry = debugfs_create_file(NST_DEBUG_NAME, mode, + o2net_dentry, NULL, &nst_seq_fops); + if (nst_dentry) + sc_dentry = debugfs_create_file(SC_DEBUG_NAME, mode, + o2net_dentry, NULL, &sc_seq_fops); + if (sc_dentry) + stats_dentry = debugfs_create_file(STATS_DEBUG_NAME, mode, + o2net_dentry, NULL, &stats_seq_fops); + if (stats_dentry) + nodes_dentry = debugfs_create_file(NODES_DEBUG_NAME, mode, + o2net_dentry, NULL, &nodes_fops); + if (nodes_dentry) + return 0; + + o2net_debugfs_exit(); + mlog_errno(-ENOMEM); + return -ENOMEM; } #endif /* CONFIG_DEBUG_FS */ diff --git a/fs/ocfs2/cluster/nodemanager.c b/fs/ocfs2/cluster/nodemanager.c index 7ee6188bc79..441c84e169e 100644 --- a/fs/ocfs2/cluster/nodemanager.c +++ b/fs/ocfs2/cluster/nodemanager.c @@ -19,6 +19,7 @@ * Boston, MA 021110-1307, USA. */ +#include <linux/slab.h> #include <linux/kernel.h> #include <linux/module.h> #include <linux/configfs.h> @@ -28,13 +29,16 @@ #include "heartbeat.h" #include "masklog.h" #include "sys.h" -#include "ver.h" /* for now we operate under the assertion that there can be only one * cluster active at a time. Changing this will require trickling * cluster references throughout where nodes are looked up */ struct o2nm_cluster *o2nm_single_cluster = NULL; +char *o2nm_fence_method_desc[O2NM_FENCE_METHODS] = { + "reset", /* O2NM_FENCE_RESET */ + "panic", /* O2NM_FENCE_PANIC */ +}; struct o2nm_node *o2nm_get_node_by_num(u8 node_num) { @@ -579,6 +583,43 @@ static ssize_t o2nm_cluster_attr_reconnect_delay_ms_write( return o2nm_cluster_attr_write(page, count, &cluster->cl_reconnect_delay_ms); } + +static ssize_t o2nm_cluster_attr_fence_method_read( + struct o2nm_cluster *cluster, char *page) +{ + ssize_t ret = 0; + + if (cluster) + ret = sprintf(page, "%s\n", + o2nm_fence_method_desc[cluster->cl_fence_method]); + return ret; +} + +static ssize_t o2nm_cluster_attr_fence_method_write( + struct o2nm_cluster *cluster, const char *page, size_t count) +{ + unsigned int i; + + if (page[count - 1] != '\n') + goto bail; + + for (i = 0; i < O2NM_FENCE_METHODS; ++i) { + if (count != strlen(o2nm_fence_method_desc[i]) + 1) + continue; + if (strncasecmp(page, o2nm_fence_method_desc[i], count - 1)) + continue; + if (cluster->cl_fence_method != i) { + printk(KERN_INFO "ocfs2: Changing fence method to %s\n", + o2nm_fence_method_desc[i]); + cluster->cl_fence_method = i; + } + return count; + } + +bail: + return -EINVAL; +} + static struct o2nm_cluster_attribute o2nm_cluster_attr_idle_timeout_ms = { .attr = { .ca_owner = THIS_MODULE, .ca_name = "idle_timeout_ms", @@ -603,10 +644,19 @@ static struct o2nm_cluster_attribute o2nm_cluster_attr_reconnect_delay_ms = { .store = o2nm_cluster_attr_reconnect_delay_ms_write, }; +static struct o2nm_cluster_attribute o2nm_cluster_attr_fence_method = { + .attr = { .ca_owner = THIS_MODULE, + .ca_name = "fence_method", + .ca_mode = S_IRUGO | S_IWUSR }, + .show = o2nm_cluster_attr_fence_method_read, + .store = o2nm_cluster_attr_fence_method_write, +}; + static struct configfs_attribute *o2nm_cluster_attrs[] = { &o2nm_cluster_attr_idle_timeout_ms.attr, &o2nm_cluster_attr_keepalive_delay_ms.attr, &o2nm_cluster_attr_reconnect_delay_ms.attr, + &o2nm_cluster_attr_fence_method.attr, NULL, }; static ssize_t o2nm_cluster_show(struct config_item *item, @@ -660,6 +710,8 @@ static struct config_item *o2nm_node_group_make_item(struct config_group *group, config_item_init_type_name(&node->nd_item, name, &o2nm_node_type); spin_lock_init(&node->nd_lock); + mlog(ML_CLUSTER, "o2nm: Registering node %s\n", name); + return &node->nd_item; } @@ -693,6 +745,9 @@ static void o2nm_node_group_drop_item(struct config_group *group, } write_unlock(&cluster->cl_nodes_lock); + mlog(ML_CLUSTER, "o2nm: Unregistered node %s\n", + config_item_name(&node->nd_item)); + config_item_put(item); } @@ -778,6 +833,7 @@ static struct config_group *o2nm_cluster_group_make_group(struct config_group *g cluster->cl_reconnect_delay_ms = O2NET_RECONNECT_DELAY_MS_DEFAULT; cluster->cl_idle_timeout_ms = O2NET_IDLE_TIMEOUT_MS_DEFAULT; cluster->cl_keepalive_delay_ms = O2NET_KEEPALIVE_DELAY_MS_DEFAULT; + cluster->cl_fence_method = O2NM_FENCE_RESET; ret = &cluster->cl_group; o2nm_single_cluster = cluster; @@ -888,8 +944,6 @@ static int __init init_o2nm(void) { int ret = -1; - cluster_print_version(); - ret = o2hb_init(); if (ret) goto out; @@ -927,6 +981,7 @@ out: MODULE_AUTHOR("Oracle"); MODULE_LICENSE("GPL"); +MODULE_DESCRIPTION("OCFS2 cluster management"); module_init(init_o2nm) module_exit(exit_o2nm) diff --git a/fs/ocfs2/cluster/nodemanager.h b/fs/ocfs2/cluster/nodemanager.h index c992ea0da4a..09ea2d388bb 100644 --- a/fs/ocfs2/cluster/nodemanager.h +++ b/fs/ocfs2/cluster/nodemanager.h @@ -33,6 +33,12 @@ #include <linux/configfs.h> #include <linux/rbtree.h> +enum o2nm_fence_method { + O2NM_FENCE_RESET = 0, + O2NM_FENCE_PANIC, + O2NM_FENCE_METHODS, /* Number of fence methods */ +}; + struct o2nm_node { spinlock_t nd_lock; struct config_item nd_item; @@ -58,6 +64,7 @@ struct o2nm_cluster { unsigned int cl_idle_timeout_ms; unsigned int cl_keepalive_delay_ms; unsigned int cl_reconnect_delay_ms; + enum o2nm_fence_method cl_fence_method; /* this bitmap is part of a hack for disk bitmap.. will go eventually. - zab */ unsigned long cl_nodes_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)]; diff --git a/fs/ocfs2/cluster/ocfs2_nodemanager.h b/fs/ocfs2/cluster/ocfs2_nodemanager.h index 5b9854bad57..49b594325be 100644 --- a/fs/ocfs2/cluster/ocfs2_nodemanager.h +++ b/fs/ocfs2/cluster/ocfs2_nodemanager.h @@ -36,4 +36,10 @@ /* host name, group name, cluster name all 64 bytes */ #define O2NM_MAX_NAME_LEN 64 // __NEW_UTS_LEN +/* + * Maximum number of global heartbeat regions allowed. + * **CAUTION** Changing this number will break dlm compatibility. + */ +#define O2NM_MAX_REGIONS 32 + #endif /* _OCFS2_NODEMANAGER_H */ diff --git a/fs/ocfs2/cluster/quorum.c b/fs/ocfs2/cluster/quorum.c index bbacf7da48a..1ec141e758d 100644 --- a/fs/ocfs2/cluster/quorum.c +++ b/fs/ocfs2/cluster/quorum.c @@ -44,7 +44,6 @@ * and if they're the last, they fire off the decision. */ #include <linux/kernel.h> -#include <linux/slab.h> #include <linux/workqueue.h> #include <linux/reboot.h> @@ -74,11 +73,23 @@ static void o2quo_fence_self(void) * threads can still schedule, etc, etc */ o2hb_stop_all_regions(); - printk("ocfs2 is very sorry to be fencing this system by restarting\n"); - emergency_restart(); + switch (o2nm_single_cluster->cl_fence_method) { + case O2NM_FENCE_PANIC: + panic("*** ocfs2 is very sorry to be fencing this system by " + "panicing ***\n"); + break; + default: + WARN_ON(o2nm_single_cluster->cl_fence_method >= + O2NM_FENCE_METHODS); + case O2NM_FENCE_RESET: + printk(KERN_ERR "*** ocfs2 is very sorry to be fencing this " + "system by restarting ***\n"); + emergency_restart(); + break; + }; } -/* Indicate that a timeout occured on a hearbeat region write. The +/* Indicate that a timeout occurred on a hearbeat region write. The * other nodes in the cluster may consider us dead at that time so we * want to "fence" ourselves so that we don't scribble on the disk * after they think they've recovered us. This can't solve all @@ -250,10 +261,10 @@ void o2quo_hb_still_up(u8 node) spin_unlock(&qs->qs_lock); } -/* This is analagous to hb_up. as a node's connection comes up we delay the +/* This is analogous to hb_up. as a node's connection comes up we delay the * quorum decision until we see it heartbeating. the hold will be droped in * hb_up or hb_down. it might be perpetuated by con_err until hb_down. if - * it's already heartbeating we we might be dropping a hold that conn_up got. + * it's already heartbeating we might be dropping a hold that conn_up got. * */ void o2quo_conn_up(u8 node) { @@ -314,5 +325,7 @@ void o2quo_init(void) void o2quo_exit(void) { - flush_scheduled_work(); + struct o2quo_state *qs = &o2quo_state; + + flush_work(&qs->qs_work); } diff --git a/fs/ocfs2/cluster/sys.c b/fs/ocfs2/cluster/sys.c index bc702dab5d1..b7f57271d49 100644 --- a/fs/ocfs2/cluster/sys.c +++ b/fs/ocfs2/cluster/sys.c @@ -41,7 +41,7 @@ static ssize_t version_show(struct kobject *kobj, struct kobj_attribute *attr, return snprintf(buf, PAGE_SIZE, "%u\n", O2NM_API_VERSION); } static struct kobj_attribute attr_version = - __ATTR(interface_revision, S_IFREG | S_IRUGO, version_show, NULL); + __ATTR(interface_revision, S_IRUGO, version_show, NULL); static struct attribute *o2cb_attrs[] = { &attr_version.attr, @@ -57,7 +57,6 @@ static struct kset *o2cb_kset; void o2cb_sys_shutdown(void) { mlog_sys_shutdown(); - sysfs_remove_link(NULL, "o2cb"); kset_unregister(o2cb_kset); } @@ -69,14 +68,6 @@ int o2cb_sys_init(void) if (!o2cb_kset) return -ENOMEM; - /* - * Create this symlink for backwards compatibility with old - * versions of ocfs2-tools which look for things in /sys/o2cb. - */ - ret = sysfs_create_link(NULL, &o2cb_kset->kobj, "o2cb"); - if (ret) - goto error; - ret = sysfs_create_group(&o2cb_kset->kobj, &o2cb_attr_group); if (ret) goto error; diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c index 9fbe849f634..681691bc233 100644 --- a/fs/ocfs2/cluster/tcp.c +++ b/fs/ocfs2/cluster/tcp.c @@ -59,6 +59,7 @@ #include <linux/idr.h> #include <linux/kref.h> #include <linux/net.h> +#include <linux/export.h> #include <net/tcp.h> #include <asm/uaccess.h> @@ -72,9 +73,9 @@ #include "tcp_internal.h" -#define SC_NODEF_FMT "node %s (num %u) at %u.%u.%u.%u:%u" +#define SC_NODEF_FMT "node %s (num %u) at %pI4:%u" #define SC_NODEF_ARGS(sc) sc->sc_node->nd_name, sc->sc_node->nd_num, \ - NIPQUAD(sc->sc_node->nd_ipv4_address), \ + &sc->sc_node->nd_ipv4_address, \ ntohs(sc->sc_node->nd_ipv4_port) /* @@ -107,7 +108,7 @@ static struct rb_root o2net_handler_tree = RB_ROOT; static struct o2net_node o2net_nodes[O2NM_MAX_NODES]; /* XXX someday we'll need better accounting */ -static struct socket *o2net_listen_sock = NULL; +static struct socket *o2net_listen_sock; /* * listen work is only queued by the listening socket callbacks on the @@ -136,7 +137,7 @@ static int o2net_sys_err_translations[O2NET_ERR_MAX] = static void o2net_sc_connect_completed(struct work_struct *work); static void o2net_rx_until_empty(struct work_struct *work); static void o2net_shutdown_sc(struct work_struct *work); -static void o2net_listen_data_ready(struct sock *sk, int bytes); +static void o2net_listen_data_ready(struct sock *sk); static void o2net_sc_send_keep_req(struct work_struct *work); static void o2net_idle_timer(unsigned long data); static void o2net_sc_postpone_idle(struct o2net_sock_container *sc); @@ -153,74 +154,125 @@ static void o2net_init_nst(struct o2net_send_tracking *nst, u32 msgtype, nst->st_node = node; } -static void o2net_set_nst_sock_time(struct o2net_send_tracking *nst) +static inline void o2net_set_nst_sock_time(struct o2net_send_tracking *nst) { - do_gettimeofday(&nst->st_sock_time); + nst->st_sock_time = ktime_get(); } -static void o2net_set_nst_send_time(struct o2net_send_tracking *nst) +static inline void o2net_set_nst_send_time(struct o2net_send_tracking *nst) { - do_gettimeofday(&nst->st_send_time); + nst->st_send_time = ktime_get(); } -static void o2net_set_nst_status_time(struct o2net_send_tracking *nst) +static inline void o2net_set_nst_status_time(struct o2net_send_tracking *nst) { - do_gettimeofday(&nst->st_status_time); + nst->st_status_time = ktime_get(); } -static void o2net_set_nst_sock_container(struct o2net_send_tracking *nst, - struct o2net_sock_container *sc) +static inline void o2net_set_nst_sock_container(struct o2net_send_tracking *nst, + struct o2net_sock_container *sc) { nst->st_sc = sc; } -static void o2net_set_nst_msg_id(struct o2net_send_tracking *nst, u32 msg_id) +static inline void o2net_set_nst_msg_id(struct o2net_send_tracking *nst, + u32 msg_id) { nst->st_id = msg_id; } -#else /* CONFIG_DEBUG_FS */ - -static inline void o2net_init_nst(struct o2net_send_tracking *nst, u32 msgtype, - u32 msgkey, struct task_struct *task, u8 node) +static inline void o2net_set_sock_timer(struct o2net_sock_container *sc) { + sc->sc_tv_timer = ktime_get(); } -static inline void o2net_set_nst_sock_time(struct o2net_send_tracking *nst) +static inline void o2net_set_data_ready_time(struct o2net_sock_container *sc) { + sc->sc_tv_data_ready = ktime_get(); } -static inline void o2net_set_nst_send_time(struct o2net_send_tracking *nst) +static inline void o2net_set_advance_start_time(struct o2net_sock_container *sc) { + sc->sc_tv_advance_start = ktime_get(); } -static inline void o2net_set_nst_status_time(struct o2net_send_tracking *nst) +static inline void o2net_set_advance_stop_time(struct o2net_sock_container *sc) { + sc->sc_tv_advance_stop = ktime_get(); } -static inline void o2net_set_nst_sock_container(struct o2net_send_tracking *nst, - struct o2net_sock_container *sc) +static inline void o2net_set_func_start_time(struct o2net_sock_container *sc) { + sc->sc_tv_func_start = ktime_get(); } -static inline void o2net_set_nst_msg_id(struct o2net_send_tracking *nst, - u32 msg_id) +static inline void o2net_set_func_stop_time(struct o2net_sock_container *sc) { + sc->sc_tv_func_stop = ktime_get(); } +#else /* CONFIG_DEBUG_FS */ +# define o2net_init_nst(a, b, c, d, e) +# define o2net_set_nst_sock_time(a) +# define o2net_set_nst_send_time(a) +# define o2net_set_nst_status_time(a) +# define o2net_set_nst_sock_container(a, b) +# define o2net_set_nst_msg_id(a, b) +# define o2net_set_sock_timer(a) +# define o2net_set_data_ready_time(a) +# define o2net_set_advance_start_time(a) +# define o2net_set_advance_stop_time(a) +# define o2net_set_func_start_time(a) +# define o2net_set_func_stop_time(a) #endif /* CONFIG_DEBUG_FS */ -static inline int o2net_reconnect_delay(void) +#ifdef CONFIG_OCFS2_FS_STATS +static ktime_t o2net_get_func_run_time(struct o2net_sock_container *sc) +{ + return ktime_sub(sc->sc_tv_func_stop, sc->sc_tv_func_start); +} + +static void o2net_update_send_stats(struct o2net_send_tracking *nst, + struct o2net_sock_container *sc) +{ + sc->sc_tv_status_total = ktime_add(sc->sc_tv_status_total, + ktime_sub(ktime_get(), + nst->st_status_time)); + sc->sc_tv_send_total = ktime_add(sc->sc_tv_send_total, + ktime_sub(nst->st_status_time, + nst->st_send_time)); + sc->sc_tv_acquiry_total = ktime_add(sc->sc_tv_acquiry_total, + ktime_sub(nst->st_send_time, + nst->st_sock_time)); + sc->sc_send_count++; +} + +static void o2net_update_recv_stats(struct o2net_sock_container *sc) +{ + sc->sc_tv_process_total = ktime_add(sc->sc_tv_process_total, + o2net_get_func_run_time(sc)); + sc->sc_recv_count++; +} + +#else + +# define o2net_update_send_stats(a, b) + +# define o2net_update_recv_stats(sc) + +#endif /* CONFIG_OCFS2_FS_STATS */ + +static inline unsigned int o2net_reconnect_delay(void) { return o2nm_single_cluster->cl_reconnect_delay_ms; } -static inline int o2net_keepalive_delay(void) +static inline unsigned int o2net_keepalive_delay(void) { return o2nm_single_cluster->cl_keepalive_delay_ms; } -static inline int o2net_idle_timeout(void) +static inline unsigned int o2net_idle_timeout(void) { return o2nm_single_cluster->cl_idle_timeout_ms; } @@ -252,28 +304,22 @@ static u8 o2net_num_from_nn(struct o2net_node *nn) static int o2net_prep_nsw(struct o2net_node *nn, struct o2net_status_wait *nsw) { - int ret = 0; - - do { - if (!idr_pre_get(&nn->nn_status_idr, GFP_ATOMIC)) { - ret = -EAGAIN; - break; - } - spin_lock(&nn->nn_lock); - ret = idr_get_new(&nn->nn_status_idr, nsw, &nsw->ns_id); - if (ret == 0) - list_add_tail(&nsw->ns_node_item, - &nn->nn_status_list); - spin_unlock(&nn->nn_lock); - } while (ret == -EAGAIN); + int ret; - if (ret == 0) { - init_waitqueue_head(&nsw->ns_wq); - nsw->ns_sys_status = O2NET_ERR_NONE; - nsw->ns_status = 0; + spin_lock(&nn->nn_lock); + ret = idr_alloc(&nn->nn_status_idr, nsw, 0, 0, GFP_ATOMIC); + if (ret >= 0) { + nsw->ns_id = ret; + list_add_tail(&nsw->ns_node_item, &nn->nn_status_list); } + spin_unlock(&nn->nn_lock); + if (ret < 0) + return ret; - return ret; + init_waitqueue_head(&nsw->ns_wq); + nsw->ns_sys_status = O2NET_ERR_NONE; + nsw->ns_status = 0; + return 0; } static void o2net_complete_nsw_locked(struct o2net_node *nn, @@ -355,10 +401,14 @@ static void sc_kref_release(struct kref *kref) sc->sc_sock = NULL; } + o2nm_undepend_item(&sc->sc_node->nd_item); o2nm_node_put(sc->sc_node); sc->sc_node = NULL; o2net_debug_del_sc(sc); + + if (sc->sc_page) + __free_page(sc->sc_page); kfree(sc); } @@ -376,6 +426,7 @@ static struct o2net_sock_container *sc_alloc(struct o2nm_node *node) { struct o2net_sock_container *sc, *ret = NULL; struct page *page = NULL; + int status = 0; page = alloc_page(GFP_NOFS); sc = kzalloc(sizeof(*sc), GFP_NOFS); @@ -386,6 +437,13 @@ static struct o2net_sock_container *sc_alloc(struct o2nm_node *node) o2nm_node_get(node); sc->sc_node = node; + /* pin the node item of the remote node */ + status = o2nm_depend_item(&node->nd_item); + if (status) { + mlog_errno(status); + o2nm_node_put(node); + goto out; + } INIT_WORK(&sc->sc_connect_work, o2net_sc_connect_completed); INIT_WORK(&sc->sc_rx_work, o2net_rx_until_empty); INIT_WORK(&sc->sc_shutdown_work, o2net_shutdown_sc); @@ -485,17 +543,18 @@ static void o2net_set_nn_state(struct o2net_node *nn, } if (was_valid && !valid) { - printk(KERN_INFO "o2net: no longer connected to " - SC_NODEF_FMT "\n", SC_NODEF_ARGS(old_sc)); + if (old_sc) + printk(KERN_NOTICE "o2net: No longer connected to " + SC_NODEF_FMT "\n", SC_NODEF_ARGS(old_sc)); o2net_complete_nodes_nsw(nn); } if (!was_valid && valid) { o2quo_conn_up(o2net_num_from_nn(nn)); cancel_delayed_work(&nn->nn_connect_expired); - printk(KERN_INFO "o2net: %s " SC_NODEF_FMT "\n", + printk(KERN_NOTICE "o2net: %s " SC_NODEF_FMT "\n", o2nm_this_node() > sc->sc_node->nd_num ? - "connected to" : "accepted connection from", + "Connected to" : "Accepted connection from", SC_NODEF_ARGS(sc)); } @@ -505,7 +564,7 @@ static void o2net_set_nn_state(struct o2net_node *nn, * the work queue actually being up. */ if (!valid && o2net_wq) { unsigned long delay; - /* delay if we're withing a RECONNECT_DELAY of the + /* delay if we're within a RECONNECT_DELAY of the * last attempt */ delay = (nn->nn_last_connect_attempt + msecs_to_jiffies(o2net_reconnect_delay())) @@ -538,15 +597,15 @@ static void o2net_set_nn_state(struct o2net_node *nn, } /* see o2net_register_callbacks() */ -static void o2net_data_ready(struct sock *sk, int bytes) +static void o2net_data_ready(struct sock *sk) { - void (*ready)(struct sock *sk, int bytes); + void (*ready)(struct sock *sk); read_lock(&sk->sk_callback_lock); if (sk->sk_user_data) { struct o2net_sock_container *sc = sk->sk_user_data; sclog(sc, "data_ready hit\n"); - do_gettimeofday(&sc->sc_tv_data_ready); + o2net_set_data_ready_time(sc); o2net_sc_queue_work(sc, &sc->sc_rx_work); ready = sc->sc_data_ready; } else { @@ -554,7 +613,7 @@ static void o2net_data_ready(struct sock *sk, int bytes) } read_unlock(&sk->sk_callback_lock); - ready(sk, bytes); + ready(sk); } /* see o2net_register_callbacks() */ @@ -575,16 +634,19 @@ static void o2net_state_change(struct sock *sk) state_change = sc->sc_state_change; switch(sk->sk_state) { - /* ignore connecting sockets as they make progress */ - case TCP_SYN_SENT: - case TCP_SYN_RECV: - break; - case TCP_ESTABLISHED: - o2net_sc_queue_work(sc, &sc->sc_connect_work); - break; - default: - o2net_sc_queue_work(sc, &sc->sc_shutdown_work); - break; + /* ignore connecting sockets as they make progress */ + case TCP_SYN_SENT: + case TCP_SYN_RECV: + break; + case TCP_ESTABLISHED: + o2net_sc_queue_work(sc, &sc->sc_connect_work); + break; + default: + printk(KERN_INFO "o2net: Connection to " SC_NODEF_FMT + " shutdown, state %d\n", + SC_NODEF_ARGS(sc), sk->sk_state); + o2net_sc_queue_work(sc, &sc->sc_shutdown_work); + break; } out: read_unlock(&sk->sk_callback_lock); @@ -704,32 +766,32 @@ static struct o2net_msg_handler * o2net_handler_tree_lookup(u32 msg_type, u32 key, struct rb_node ***ret_p, struct rb_node **ret_parent) { - struct rb_node **p = &o2net_handler_tree.rb_node; - struct rb_node *parent = NULL; + struct rb_node **p = &o2net_handler_tree.rb_node; + struct rb_node *parent = NULL; struct o2net_msg_handler *nmh, *ret = NULL; int cmp; - while (*p) { - parent = *p; - nmh = rb_entry(parent, struct o2net_msg_handler, nh_node); + while (*p) { + parent = *p; + nmh = rb_entry(parent, struct o2net_msg_handler, nh_node); cmp = o2net_handler_cmp(nmh, msg_type, key); - if (cmp < 0) - p = &(*p)->rb_left; - else if (cmp > 0) - p = &(*p)->rb_right; - else { + if (cmp < 0) + p = &(*p)->rb_left; + else if (cmp > 0) + p = &(*p)->rb_right; + else { ret = nmh; - break; + break; } - } + } - if (ret_p != NULL) - *ret_p = p; - if (ret_parent != NULL) - *ret_parent = parent; + if (ret_p != NULL) + *ret_p = p; + if (ret_parent != NULL) + *ret_parent = parent; - return ret; + return ret; } static void o2net_handler_kref_release(struct kref *kref) @@ -806,7 +868,7 @@ int o2net_register_handler(u32 msg_type, u32 key, u32 max_len, /* we've had some trouble with handlers seemingly vanishing. */ mlog_bug_on_msg(o2net_handler_tree_lookup(msg_type, key, &p, &parent) == NULL, - "couldn't find handler we *just* registerd " + "couldn't find handler we *just* registered " "for type %u key %08x\n", msg_type, key); } write_unlock(&o2net_handler_lock); @@ -854,57 +916,30 @@ static struct o2net_msg_handler *o2net_handler_get(u32 msg_type, u32 key) static int o2net_recv_tcp_msg(struct socket *sock, void *data, size_t len) { - int ret; - mm_segment_t oldfs; - struct kvec vec = { - .iov_len = len, - .iov_base = data, - }; - struct msghdr msg = { - .msg_iovlen = 1, - .msg_iov = (struct iovec *)&vec, - .msg_flags = MSG_DONTWAIT, - }; - - oldfs = get_fs(); - set_fs(get_ds()); - ret = sock_recvmsg(sock, &msg, len, msg.msg_flags); - set_fs(oldfs); - - return ret; + struct kvec vec = { .iov_len = len, .iov_base = data, }; + struct msghdr msg = { .msg_flags = MSG_DONTWAIT, }; + return kernel_recvmsg(sock, &msg, &vec, 1, len, msg.msg_flags); } static int o2net_send_tcp_msg(struct socket *sock, struct kvec *vec, size_t veclen, size_t total) { int ret; - mm_segment_t oldfs; - struct msghdr msg = { - .msg_iov = (struct iovec *)vec, - .msg_iovlen = veclen, - }; + struct msghdr msg; if (sock == NULL) { ret = -EINVAL; goto out; } - oldfs = get_fs(); - set_fs(get_ds()); - ret = sock_sendmsg(sock, &msg, total); - set_fs(oldfs); - if (ret != total) { - mlog(ML_ERROR, "sendmsg returned %d instead of %zu\n", ret, - total); - if (ret >= 0) - ret = -EPIPE; /* should be smarter, I bet */ - goto out; - } - - ret = 0; + ret = kernel_sendmsg(sock, &msg, vec, veclen, total); + if (likely(ret == total)) + return 0; + mlog(ML_ERROR, "sendmsg returned %d instead of %zu\n", ret, total); + if (ret >= 0) + ret = -EPIPE; /* should be smarter, I bet */ out: - if (ret < 0) - mlog(0, "returning error: %d\n", ret); + mlog(0, "returning error: %d\n", ret); return ret; } @@ -930,7 +965,7 @@ static void o2net_sendpage(struct o2net_sock_container *sc, cond_resched(); continue; } - mlog(ML_ERROR, "sendpage of size %zu to " SC_NODEF_FMT + mlog(ML_ERROR, "sendpage of size %zu to " SC_NODEF_FMT " failed with %zd\n", size, SC_NODEF_ARGS(sc), ret); o2net_ensure_shutdown(nn, sc, 0); break; @@ -971,10 +1006,29 @@ static int o2net_tx_can_proceed(struct o2net_node *nn, return ret; } +/* Get a map of all nodes to which this node is currently connected to */ +void o2net_fill_node_map(unsigned long *map, unsigned bytes) +{ + struct o2net_sock_container *sc; + int node, ret; + + BUG_ON(bytes < (BITS_TO_LONGS(O2NM_MAX_NODES) * sizeof(unsigned long))); + + memset(map, 0, bytes); + for (node = 0; node < O2NM_MAX_NODES; ++node) { + o2net_tx_can_proceed(o2net_nn_from_num(node), &sc, &ret); + if (!ret) { + set_bit(node, map); + sc_put(sc); + } + } +} +EXPORT_SYMBOL_GPL(o2net_fill_node_map); + int o2net_send_message_vec(u32 msg_type, u32 key, struct kvec *caller_vec, size_t caller_veclen, u8 target_node, int *status) { - int ret, error = 0; + int ret = 0; struct o2net_msg *msg = NULL; size_t veclen, caller_bytes = 0; struct kvec *vec = NULL; @@ -1015,10 +1069,7 @@ int o2net_send_message_vec(u32 msg_type, u32 key, struct kvec *caller_vec, o2net_set_nst_sock_time(&nst); - ret = wait_event_interruptible(nn->nn_sc_wq, - o2net_tx_can_proceed(nn, &sc, &error)); - if (!ret && error) - ret = error; + wait_event(nn->nn_sc_wq, o2net_tx_can_proceed(nn, &sc, &ret)); if (ret) goto out; @@ -1070,6 +1121,8 @@ int o2net_send_message_vec(u32 msg_type, u32 key, struct kvec *caller_vec, o2net_set_nst_status_time(&nst); wait_event(nsw.ns_wq, o2net_nsw_completed(nn, &nsw)); + o2net_update_send_stats(&nst, sc); + /* Note that we avoid overwriting the callers status return * variable if a system error was reported on the other * side. Callers beware. */ @@ -1083,10 +1136,8 @@ out: o2net_debug_del_nst(&nst); /* must be before dropping sc and node */ if (sc) sc_put(sc); - if (vec) - kfree(vec); - if (msg) - kfree(msg); + kfree(vec); + kfree(msg); o2net_complete_nsw(nn, &nsw, 0, 0, 0); return ret; } @@ -1183,13 +1234,15 @@ static int o2net_process_message(struct o2net_sock_container *sc, if (syserr != O2NET_ERR_NONE) goto out_respond; - do_gettimeofday(&sc->sc_tv_func_start); + o2net_set_func_start_time(sc); sc->sc_msg_key = be32_to_cpu(hdr->key); sc->sc_msg_type = be16_to_cpu(hdr->msg_type); handler_status = (nmh->nh_func)(hdr, sizeof(struct o2net_msg) + be16_to_cpu(hdr->data_len), nmh->nh_func_data, &ret_data); - do_gettimeofday(&sc->sc_tv_func_stop); + o2net_set_func_stop_time(sc); + + o2net_update_recv_stats(sc); out_respond: /* this destroys the hdr, so don't use it after this */ @@ -1220,11 +1273,11 @@ static int o2net_check_handshake(struct o2net_sock_container *sc) struct o2net_node *nn = o2net_nn_from_num(sc->sc_node->nd_num); if (hand->protocol_version != cpu_to_be64(O2NET_PROTOCOL_VERSION)) { - mlog(ML_NOTICE, SC_NODEF_FMT " advertised net protocol " - "version %llu but %llu is required, disconnecting\n", - SC_NODEF_ARGS(sc), - (unsigned long long)be64_to_cpu(hand->protocol_version), - O2NET_PROTOCOL_VERSION); + printk(KERN_NOTICE "o2net: " SC_NODEF_FMT " Advertised net " + "protocol version %llu but %llu is required. " + "Disconnecting.\n", SC_NODEF_ARGS(sc), + (unsigned long long)be64_to_cpu(hand->protocol_version), + O2NET_PROTOCOL_VERSION); /* don't bother reconnecting if its the wrong version. */ o2net_ensure_shutdown(nn, sc, -ENOTCONN); @@ -1238,33 +1291,33 @@ static int o2net_check_handshake(struct o2net_sock_container *sc) */ if (be32_to_cpu(hand->o2net_idle_timeout_ms) != o2net_idle_timeout()) { - mlog(ML_NOTICE, SC_NODEF_FMT " uses a network idle timeout of " - "%u ms, but we use %u ms locally. disconnecting\n", - SC_NODEF_ARGS(sc), - be32_to_cpu(hand->o2net_idle_timeout_ms), - o2net_idle_timeout()); + printk(KERN_NOTICE "o2net: " SC_NODEF_FMT " uses a network " + "idle timeout of %u ms, but we use %u ms locally. " + "Disconnecting.\n", SC_NODEF_ARGS(sc), + be32_to_cpu(hand->o2net_idle_timeout_ms), + o2net_idle_timeout()); o2net_ensure_shutdown(nn, sc, -ENOTCONN); return -1; } if (be32_to_cpu(hand->o2net_keepalive_delay_ms) != o2net_keepalive_delay()) { - mlog(ML_NOTICE, SC_NODEF_FMT " uses a keepalive delay of " - "%u ms, but we use %u ms locally. disconnecting\n", - SC_NODEF_ARGS(sc), - be32_to_cpu(hand->o2net_keepalive_delay_ms), - o2net_keepalive_delay()); + printk(KERN_NOTICE "o2net: " SC_NODEF_FMT " uses a keepalive " + "delay of %u ms, but we use %u ms locally. " + "Disconnecting.\n", SC_NODEF_ARGS(sc), + be32_to_cpu(hand->o2net_keepalive_delay_ms), + o2net_keepalive_delay()); o2net_ensure_shutdown(nn, sc, -ENOTCONN); return -1; } if (be32_to_cpu(hand->o2hb_heartbeat_timeout_ms) != O2HB_MAX_WRITE_TIMEOUT_MS) { - mlog(ML_NOTICE, SC_NODEF_FMT " uses a heartbeat timeout of " - "%u ms, but we use %u ms locally. disconnecting\n", - SC_NODEF_ARGS(sc), - be32_to_cpu(hand->o2hb_heartbeat_timeout_ms), - O2HB_MAX_WRITE_TIMEOUT_MS); + printk(KERN_NOTICE "o2net: " SC_NODEF_FMT " uses a heartbeat " + "timeout of %u ms, but we use %u ms locally. " + "Disconnecting.\n", SC_NODEF_ARGS(sc), + be32_to_cpu(hand->o2hb_heartbeat_timeout_ms), + O2HB_MAX_WRITE_TIMEOUT_MS); o2net_ensure_shutdown(nn, sc, -ENOTCONN); return -1; } @@ -1300,7 +1353,7 @@ static int o2net_advance_rx(struct o2net_sock_container *sc) size_t datalen; sclog(sc, "receiving\n"); - do_gettimeofday(&sc->sc_tv_advance_start); + o2net_set_advance_start_time(sc); if (unlikely(sc->sc_handshake_ok == 0)) { if(sc->sc_page_off < sizeof(struct o2net_handshake)) { @@ -1375,7 +1428,7 @@ static int o2net_advance_rx(struct o2net_sock_container *sc) out: sclog(sc, "ret = %d\n", ret); - do_gettimeofday(&sc->sc_tv_advance_stop); + o2net_set_advance_stop_time(sc); return ret; } @@ -1475,27 +1528,16 @@ static void o2net_idle_timer(unsigned long data) { struct o2net_sock_container *sc = (struct o2net_sock_container *)data; struct o2net_node *nn = o2net_nn_from_num(sc->sc_node->nd_num); - struct timeval now; +#ifdef CONFIG_DEBUG_FS + unsigned long msecs = ktime_to_ms(ktime_get()) - + ktime_to_ms(sc->sc_tv_timer); +#else + unsigned long msecs = o2net_idle_timeout(); +#endif - do_gettimeofday(&now); - - printk(KERN_INFO "o2net: connection to " SC_NODEF_FMT " has been idle for %u.%u " - "seconds, shutting it down.\n", SC_NODEF_ARGS(sc), - o2net_idle_timeout() / 1000, - o2net_idle_timeout() % 1000); - mlog(ML_NOTICE, "here are some times that might help debug the " - "situation: (tmr %ld.%ld now %ld.%ld dr %ld.%ld adv " - "%ld.%ld:%ld.%ld func (%08x:%u) %ld.%ld:%ld.%ld)\n", - sc->sc_tv_timer.tv_sec, (long) sc->sc_tv_timer.tv_usec, - now.tv_sec, (long) now.tv_usec, - sc->sc_tv_data_ready.tv_sec, (long) sc->sc_tv_data_ready.tv_usec, - sc->sc_tv_advance_start.tv_sec, - (long) sc->sc_tv_advance_start.tv_usec, - sc->sc_tv_advance_stop.tv_sec, - (long) sc->sc_tv_advance_stop.tv_usec, - sc->sc_msg_key, sc->sc_msg_type, - sc->sc_tv_func_start.tv_sec, (long) sc->sc_tv_func_start.tv_usec, - sc->sc_tv_func_stop.tv_sec, (long) sc->sc_tv_func_stop.tv_usec); + printk(KERN_NOTICE "o2net: Connection to " SC_NODEF_FMT " has been " + "idle for %lu.%lu secs, shutting it down.\n", SC_NODEF_ARGS(sc), + msecs / 1000, msecs % 1000); /* * Initialize the nn_timeout so that the next connection attempt @@ -1511,7 +1553,7 @@ static void o2net_sc_reset_idle_timer(struct o2net_sock_container *sc) o2net_sc_cancel_delayed_work(sc, &sc->sc_keepalive_work); o2net_sc_queue_delayed_work(sc, &sc->sc_keepalive_work, msecs_to_jiffies(o2net_keepalive_delay())); - do_gettimeofday(&sc->sc_tv_timer); + o2net_set_sock_timer(sc); mod_timer(&sc->sc_idle_timeout, jiffies + msecs_to_jiffies(o2net_idle_timeout())); } @@ -1627,13 +1669,12 @@ static void o2net_start_connect(struct work_struct *work) ret = 0; out: - if (ret) { - mlog(ML_NOTICE, "connect attempt to " SC_NODEF_FMT " failed " - "with errno %d\n", SC_NODEF_ARGS(sc), ret); + if (ret && sc) { + printk(KERN_NOTICE "o2net: Connect attempt to " SC_NODEF_FMT + " failed with errno %d\n", SC_NODEF_ARGS(sc), ret); /* 0 err so that another will be queued and attempted * from set_nn_state */ - if (sc) - o2net_ensure_shutdown(nn, sc, 0); + o2net_ensure_shutdown(nn, sc, 0); } if (sc) sc_put(sc); @@ -1652,8 +1693,8 @@ static void o2net_connect_expired(struct work_struct *work) spin_lock(&nn->nn_lock); if (!nn->nn_sc_valid) { - mlog(ML_ERROR, "no connection established with node %u after " - "%u.%u seconds, giving up and returning errors.\n", + printk(KERN_NOTICE "o2net: No connection established with " + "node %u after %u.%u seconds, giving up.\n", o2net_num_from_nn(nn), o2net_idle_timeout() / 1000, o2net_idle_timeout() % 1000); @@ -1696,6 +1737,9 @@ static void o2net_hb_node_down_cb(struct o2nm_node *node, int node_num, { o2quo_hb_down(node_num); + if (!node) + return; + if (node_num != o2nm_this_node()) o2net_disconnect_node(node); @@ -1709,6 +1753,8 @@ static void o2net_hb_node_up_cb(struct o2nm_node *node, int node_num, o2quo_hb_up(node_num); + BUG_ON(!node); + /* ensure an immediate connect attempt */ nn->nn_last_connect_attempt = jiffies - (msecs_to_jiffies(o2net_reconnect_delay()) + 1); @@ -1753,16 +1799,18 @@ int o2net_register_hb_callbacks(void) /* ------------------------------------------------------------ */ -static int o2net_accept_one(struct socket *sock) +static int o2net_accept_one(struct socket *sock, int *more) { int ret, slen; struct sockaddr_in sin; struct socket *new_sock = NULL; struct o2nm_node *node = NULL; + struct o2nm_node *local_node = NULL; struct o2net_sock_container *sc = NULL; struct o2net_node *nn; BUG_ON(sock == NULL); + *more = 0; ret = sock_create_lite(sock->sk->sk_family, sock->sk->sk_type, sock->sk->sk_protocol, &new_sock); if (ret) @@ -1774,6 +1822,7 @@ static int o2net_accept_one(struct socket *sock) if (ret < 0) goto out; + *more = 1; new_sock->sk->sk_allocation = GFP_ATOMIC; ret = o2net_set_nodelay(new_sock); @@ -1790,17 +1839,25 @@ static int o2net_accept_one(struct socket *sock) node = o2nm_get_node_by_ip(sin.sin_addr.s_addr); if (node == NULL) { - mlog(ML_NOTICE, "attempt to connect from unknown node at %pI4:%d\n", - &sin.sin_addr.s_addr, ntohs(sin.sin_port)); + printk(KERN_NOTICE "o2net: Attempt to connect from unknown " + "node at %pI4:%d\n", &sin.sin_addr.s_addr, + ntohs(sin.sin_port)); ret = -EINVAL; goto out; } - if (o2nm_this_node() > node->nd_num) { - mlog(ML_NOTICE, "unexpected connect attempted from a lower " - "numbered node '%s' at " "%pI4:%d with num %u\n", - node->nd_name, &sin.sin_addr.s_addr, - ntohs(sin.sin_port), node->nd_num); + if (o2nm_this_node() >= node->nd_num) { + local_node = o2nm_get_node_by_num(o2nm_this_node()); + if (local_node) + printk(KERN_NOTICE "o2net: Unexpected connect attempt " + "seen at node '%s' (%u, %pI4:%d) from " + "node '%s' (%u, %pI4:%d)\n", + local_node->nd_name, local_node->nd_num, + &(local_node->nd_ipv4_address), + ntohs(local_node->nd_ipv4_port), + node->nd_name, + node->nd_num, &sin.sin_addr.s_addr, + ntohs(sin.sin_port)); ret = -EINVAL; goto out; } @@ -1825,10 +1882,10 @@ static int o2net_accept_one(struct socket *sock) ret = 0; spin_unlock(&nn->nn_lock); if (ret) { - mlog(ML_NOTICE, "attempt to connect from node '%s' at " - "%pI4:%d but it already has an open connection\n", - node->nd_name, &sin.sin_addr.s_addr, - ntohs(sin.sin_port)); + printk(KERN_NOTICE "o2net: Attempt to connect from node '%s' " + "at %pI4:%d but it already has an open connection\n", + node->nd_name, &sin.sin_addr.s_addr, + ntohs(sin.sin_port)); goto out; } @@ -1857,21 +1914,48 @@ out: sock_release(new_sock); if (node) o2nm_node_put(node); + if (local_node) + o2nm_node_put(local_node); if (sc) sc_put(sc); return ret; } +/* + * This function is invoked in response to one or more + * pending accepts at softIRQ level. We must drain the + * entire que before returning. + */ + static void o2net_accept_many(struct work_struct *work) { struct socket *sock = o2net_listen_sock; - while (o2net_accept_one(sock) == 0) + int more; + int err; + + /* + * It is critical to note that due to interrupt moderation + * at the network driver level, we can't assume to get a + * softIRQ for every single conn since tcp SYN packets + * can arrive back-to-back, and therefore many pending + * accepts may result in just 1 softIRQ. If we terminate + * the o2net_accept_one() loop upon seeing an err, what happens + * to the rest of the conns in the queue? If no new SYN + * arrives for hours, no softIRQ will be delivered, + * and the connections will just sit in the queue. + */ + + for (;;) { + err = o2net_accept_one(sock, &more); + if (!more) + break; cond_resched(); + } } -static void o2net_listen_data_ready(struct sock *sk, int bytes) +static void o2net_listen_data_ready(struct sock *sk) { - void (*ready)(struct sock *sk, int bytes); + void (*ready)(struct sock *sk); read_lock(&sk->sk_callback_lock); ready = sk->sk_user_data; @@ -1880,18 +1964,29 @@ static void o2net_listen_data_ready(struct sock *sk, int bytes) goto out; } - /* ->sk_data_ready is also called for a newly established child socket - * before it has been accepted and the acceptor has set up their - * data_ready.. we only want to queue listen work for our listening - * socket */ + /* This callback may called twice when a new connection + * is being established as a child socket inherits everything + * from a parent LISTEN socket, including the data_ready cb of + * the parent. This leads to a hazard. In o2net_accept_one() + * we are still initializing the child socket but have not + * changed the inherited data_ready callback yet when + * data starts arriving. + * We avoid this hazard by checking the state. + * For the listening socket, the state will be TCP_LISTEN; for the new + * socket, will be TCP_ESTABLISHED. Also, in this case, + * sk->sk_user_data is not a valid function pointer. + */ + if (sk->sk_state == TCP_LISTEN) { - mlog(ML_TCP, "bytes: %d\n", bytes); queue_work(o2net_wq, &o2net_listen_work); + } else { + ready = NULL; } out: read_unlock(&sk->sk_callback_lock); - ready(sk, bytes); + if (ready != NULL) + ready(sk); } static int o2net_open_listening_sock(__be32 addr, __be16 port) @@ -1906,7 +2001,7 @@ static int o2net_open_listening_sock(__be32 addr, __be16 port) ret = sock_create(PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock); if (ret < 0) { - mlog(ML_ERROR, "unable to create socket, ret=%d\n", ret); + printk(KERN_ERR "o2net: Error %d while creating socket\n", ret); goto out; } @@ -1920,19 +2015,18 @@ static int o2net_open_listening_sock(__be32 addr, __be16 port) o2net_listen_sock = sock; INIT_WORK(&o2net_listen_work, o2net_accept_many); - sock->sk->sk_reuse = 1; + sock->sk->sk_reuse = SK_CAN_REUSE; ret = sock->ops->bind(sock, (struct sockaddr *)&sin, sizeof(sin)); if (ret < 0) { - mlog(ML_ERROR, "unable to bind socket at %pI4:%u, " - "ret=%d\n", &addr, ntohs(port), ret); + printk(KERN_ERR "o2net: Error %d while binding socket at " + "%pI4:%u\n", ret, &addr, ntohs(port)); goto out; } ret = sock->ops->listen(sock, 64); - if (ret < 0) { - mlog(ML_ERROR, "unable to listen on %pI4:%u, ret=%d\n", - &addr, ntohs(port), ret); - } + if (ret < 0) + printk(KERN_ERR "o2net: Error %d while listening on %pI4:%u\n", + ret, &addr, ntohs(port)); out: if (ret) { diff --git a/fs/ocfs2/cluster/tcp.h b/fs/ocfs2/cluster/tcp.h index fd6179eb26d..5bada2a69b5 100644 --- a/fs/ocfs2/cluster/tcp.h +++ b/fs/ocfs2/cluster/tcp.h @@ -106,6 +106,8 @@ int o2net_register_handler(u32 msg_type, u32 key, u32 max_len, struct list_head *unreg_list); void o2net_unregister_handler_list(struct list_head *list); +void o2net_fill_node_map(unsigned long *map, unsigned bytes); + struct o2nm_node; int o2net_register_hb_callbacks(void); void o2net_unregister_hb_callbacks(void); diff --git a/fs/ocfs2/cluster/tcp_internal.h b/fs/ocfs2/cluster/tcp_internal.h index 8d58cfe410b..dc024367110 100644 --- a/fs/ocfs2/cluster/tcp_internal.h +++ b/fs/ocfs2/cluster/tcp_internal.h @@ -32,10 +32,10 @@ * on their number */ #define O2NET_QUORUM_DELAY_MS ((o2hb_dead_threshold + 2) * O2HB_REGION_TIMEOUT_MS) -/* +/* * This version number represents quite a lot, unfortunately. It not * only represents the raw network message protocol on the wire but also - * locking semantics of the file system using the protocol. It should + * locking semantics of the file system using the protocol. It should * be somewhere else, I'm sure, but right now it isn't. * * With version 11, we separate out the filesystem locking portion. The @@ -129,7 +129,7 @@ struct o2net_node { struct o2net_sock_container { struct kref sc_kref; - /* the next two are vaild for the life time of the sc */ + /* the next two are valid for the life time of the sc */ struct socket *sc_sock; struct o2nm_node *sc_node; @@ -165,19 +165,28 @@ struct o2net_sock_container { /* original handlers for the sockets */ void (*sc_state_change)(struct sock *sk); - void (*sc_data_ready)(struct sock *sk, int bytes); -#ifdef CONFIG_DEBUG_FS - struct list_head sc_net_debug_item; -#endif - struct timeval sc_tv_timer; - struct timeval sc_tv_data_ready; - struct timeval sc_tv_advance_start; - struct timeval sc_tv_advance_stop; - struct timeval sc_tv_func_start; - struct timeval sc_tv_func_stop; + void (*sc_data_ready)(struct sock *sk); + u32 sc_msg_key; u16 sc_msg_type; +#ifdef CONFIG_DEBUG_FS + struct list_head sc_net_debug_item; + ktime_t sc_tv_timer; + ktime_t sc_tv_data_ready; + ktime_t sc_tv_advance_start; + ktime_t sc_tv_advance_stop; + ktime_t sc_tv_func_start; + ktime_t sc_tv_func_stop; +#endif +#ifdef CONFIG_OCFS2_FS_STATS + ktime_t sc_tv_acquiry_total; + ktime_t sc_tv_send_total; + ktime_t sc_tv_status_total; + u32 sc_send_count; + u32 sc_recv_count; + ktime_t sc_tv_process_total; +#endif struct mutex sc_send_lock; }; @@ -220,9 +229,9 @@ struct o2net_send_tracking { u32 st_msg_type; u32 st_msg_key; u8 st_node; - struct timeval st_sock_time; - struct timeval st_send_time; - struct timeval st_status_time; + ktime_t st_sock_time; + ktime_t st_send_time; + ktime_t st_status_time; }; #else struct o2net_send_tracking { diff --git a/fs/ocfs2/cluster/ver.c b/fs/ocfs2/cluster/ver.c deleted file mode 100644 index a56eee6abad..00000000000 --- a/fs/ocfs2/cluster/ver.c +++ /dev/null @@ -1,42 +0,0 @@ -/* -*- mode: c; c-basic-offset: 8; -*- - * vim: noexpandtab sw=8 ts=8 sts=0: - * - * ver.c - * - * version string - * - * Copyright (C) 2002, 2005 Oracle. All rights reserved. - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public - * License as published by the Free Software Foundation; either - * version 2 of the License, or (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - * General Public License for more details. - * - * You should have received a copy of the GNU General Public - * License along with this program; if not, write to the - * Free Software Foundation, Inc., 59 Temple Place - Suite 330, - * Boston, MA 021110-1307, USA. - */ - -#include <linux/module.h> -#include <linux/kernel.h> - -#include "ver.h" - -#define CLUSTER_BUILD_VERSION "1.5.0" - -#define VERSION_STR "OCFS2 Node Manager " CLUSTER_BUILD_VERSION - -void cluster_print_version(void) -{ - printk(KERN_INFO "%s\n", VERSION_STR); -} - -MODULE_DESCRIPTION(VERSION_STR); - -MODULE_VERSION(CLUSTER_BUILD_VERSION); diff --git a/fs/ocfs2/cluster/ver.h b/fs/ocfs2/cluster/ver.h deleted file mode 100644 index 32554c3382c..00000000000 --- a/fs/ocfs2/cluster/ver.h +++ /dev/null @@ -1,31 +0,0 @@ -/* -*- mode: c; c-basic-offset: 8; -*- - * vim: noexpandtab sw=8 ts=8 sts=0: - * - * ver.h - * - * Function prototypes - * - * Copyright (C) 2005 Oracle. All rights reserved. - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public - * License as published by the Free Software Foundation; either - * version 2 of the License, or (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - * General Public License for more details. - * - * You should have received a copy of the GNU General Public - * License along with this program; if not, write to the - * Free Software Foundation, Inc., 59 Temple Place - Suite 330, - * Boston, MA 021110-1307, USA. - */ - -#ifndef O2CLUSTER_VER_H -#define O2CLUSTER_VER_H - -void cluster_print_version(void); - -#endif /* O2CLUSTER_VER_H */ diff --git a/fs/ocfs2/dcache.c b/fs/ocfs2/dcache.c index b574431a031..e2e05a106be 100644 --- a/fs/ocfs2/dcache.c +++ b/fs/ocfs2/dcache.c @@ -28,7 +28,6 @@ #include <linux/slab.h> #include <linux/namei.h> -#define MLOG_MASK_PREFIX ML_DCACHE #include <cluster/masklog.h> #include "ocfs2.h" @@ -38,24 +37,48 @@ #include "dlmglue.h" #include "file.h" #include "inode.h" -#include "super.h" +#include "ocfs2_trace.h" + +void ocfs2_dentry_attach_gen(struct dentry *dentry) +{ + unsigned long gen = + OCFS2_I(dentry->d_parent->d_inode)->ip_dir_lock_gen; + BUG_ON(dentry->d_inode); + dentry->d_fsdata = (void *)gen; +} -static int ocfs2_dentry_revalidate(struct dentry *dentry, - struct nameidata *nd) +static int ocfs2_dentry_revalidate(struct dentry *dentry, unsigned int flags) { - struct inode *inode = dentry->d_inode; + struct inode *inode; int ret = 0; /* if all else fails, just return false */ - struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); + struct ocfs2_super *osb; + + if (flags & LOOKUP_RCU) + return -ECHILD; - mlog_entry("(0x%p, '%.*s')\n", dentry, - dentry->d_name.len, dentry->d_name.name); + inode = dentry->d_inode; + osb = OCFS2_SB(dentry->d_sb); - /* Never trust a negative dentry - force a new lookup. */ + trace_ocfs2_dentry_revalidate(dentry, dentry->d_name.len, + dentry->d_name.name); + + /* For a negative dentry - + * check the generation number of the parent and compare with the + * one stored in the inode. + */ if (inode == NULL) { - mlog(0, "negative dentry: %.*s\n", dentry->d_name.len, - dentry->d_name.name); - goto bail; + unsigned long gen = (unsigned long) dentry->d_fsdata; + unsigned long pgen; + spin_lock(&dentry->d_lock); + pgen = OCFS2_I(dentry->d_parent->d_inode)->ip_dir_lock_gen; + spin_unlock(&dentry->d_lock); + trace_ocfs2_dentry_revalidate_negative(dentry->d_name.len, + dentry->d_name.name, + pgen, gen); + if (gen != pgen) + goto bail; + goto valid; } BUG_ON(!osb); @@ -67,8 +90,8 @@ static int ocfs2_dentry_revalidate(struct dentry *dentry, /* did we or someone else delete this inode? */ if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_DELETED) { spin_unlock(&OCFS2_I(inode)->ip_lock); - mlog(0, "inode (%llu) deleted, returning false\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno); + trace_ocfs2_dentry_revalidate_delete( + (unsigned long long)OCFS2_I(inode)->ip_blkno); goto bail; } spin_unlock(&OCFS2_I(inode)->ip_lock); @@ -78,18 +101,27 @@ static int ocfs2_dentry_revalidate(struct dentry *dentry, * inode nlink hits zero, it never goes back. */ if (inode->i_nlink == 0) { - mlog(0, "Inode %llu orphaned, returning false " - "dir = %d\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, - S_ISDIR(inode->i_mode)); + trace_ocfs2_dentry_revalidate_orphaned( + (unsigned long long)OCFS2_I(inode)->ip_blkno, + S_ISDIR(inode->i_mode)); goto bail; } + /* + * If the last lookup failed to create dentry lock, let us + * redo it. + */ + if (!dentry->d_fsdata) { + trace_ocfs2_dentry_revalidate_nofsdata( + (unsigned long long)OCFS2_I(inode)->ip_blkno); + goto bail; + } + +valid: ret = 1; bail: - mlog_exit(ret); - + trace_ocfs2_dentry_revalidate_ret(ret); return ret; } @@ -137,28 +169,24 @@ struct dentry *ocfs2_find_local_alias(struct inode *inode, u64 parent_blkno, int skip_unhashed) { - struct list_head *p; - struct dentry *dentry = NULL; - - spin_lock(&dcache_lock); - - list_for_each(p, &inode->i_dentry) { - dentry = list_entry(p, struct dentry, d_alias); + struct dentry *dentry; + spin_lock(&inode->i_lock); + hlist_for_each_entry(dentry, &inode->i_dentry, d_alias) { + spin_lock(&dentry->d_lock); if (ocfs2_match_dentry(dentry, parent_blkno, skip_unhashed)) { - mlog(0, "dentry found: %.*s\n", - dentry->d_name.len, dentry->d_name.name); + trace_ocfs2_find_local_alias(dentry->d_name.len, + dentry->d_name.name); - dget_locked(dentry); - break; + dget_dlock(dentry); + spin_unlock(&dentry->d_lock); + spin_unlock(&inode->i_lock); + return dentry; } - - dentry = NULL; + spin_unlock(&dentry->d_lock); } - - spin_unlock(&dcache_lock); - - return dentry; + spin_unlock(&inode->i_lock); + return NULL; } DEFINE_SPINLOCK(dentry_attach_lock); @@ -203,9 +231,8 @@ int ocfs2_dentry_attach_lock(struct dentry *dentry, struct dentry *alias; struct ocfs2_dentry_lock *dl = dentry->d_fsdata; - mlog(0, "Attach \"%.*s\", parent %llu, fsdata: %p\n", - dentry->d_name.len, dentry->d_name.name, - (unsigned long long)parent_blkno, dl); + trace_ocfs2_dentry_attach_lock(dentry->d_name.len, dentry->d_name.name, + (unsigned long long)parent_blkno, dl); /* * Negative dentry. We ignore these for now. @@ -216,6 +243,12 @@ int ocfs2_dentry_attach_lock(struct dentry *dentry, if (!inode) return 0; + if (!dentry->d_inode && dentry->d_fsdata) { + /* Converting a negative dentry to positive + Clear dentry->d_fsdata */ + dentry->d_fsdata = dl = NULL; + } + if (dl) { mlog_bug_on_msg(dl->dl_parent_blkno != parent_blkno, " \"%.*s\": old parent: %llu, new: %llu\n", @@ -249,7 +282,9 @@ int ocfs2_dentry_attach_lock(struct dentry *dentry, (unsigned long long)parent_blkno, (unsigned long long)dl->dl_parent_blkno); - mlog(0, "Found: %s\n", dl->dl_lockres.l_name); + trace_ocfs2_dentry_attach_lock_found(dl->dl_lockres.l_name, + (unsigned long long)parent_blkno, + (unsigned long long)OCFS2_I(inode)->ip_blkno); goto out_attach; } @@ -310,34 +345,6 @@ out_attach: return ret; } -static DEFINE_SPINLOCK(dentry_list_lock); - -/* We limit the number of dentry locks to drop in one go. We have - * this limit so that we don't starve other users of ocfs2_wq. */ -#define DL_INODE_DROP_COUNT 64 - -/* Drop inode references from dentry locks */ -void ocfs2_drop_dl_inodes(struct work_struct *work) -{ - struct ocfs2_super *osb = container_of(work, struct ocfs2_super, - dentry_lock_work); - struct ocfs2_dentry_lock *dl; - int drop_count = DL_INODE_DROP_COUNT; - - spin_lock(&dentry_list_lock); - while (osb->dentry_lock_list && drop_count--) { - dl = osb->dentry_lock_list; - osb->dentry_lock_list = dl->dl_next; - spin_unlock(&dentry_list_lock); - iput(dl->dl_inode); - kfree(dl); - spin_lock(&dentry_list_lock); - } - if (osb->dentry_lock_list) - queue_work(ocfs2_wq, &osb->dentry_lock_work); - spin_unlock(&dentry_list_lock); -} - /* * ocfs2_dentry_iput() and friends. * @@ -362,23 +369,16 @@ void ocfs2_drop_dl_inodes(struct work_struct *work) static void ocfs2_drop_dentry_lock(struct ocfs2_super *osb, struct ocfs2_dentry_lock *dl) { + iput(dl->dl_inode); ocfs2_simple_drop_lockres(osb, &dl->dl_lockres); ocfs2_lock_res_free(&dl->dl_lockres); - - /* We leave dropping of inode reference to ocfs2_wq as that can - * possibly lead to inode deletion which gets tricky */ - spin_lock(&dentry_list_lock); - if (!osb->dentry_lock_list) - queue_work(ocfs2_wq, &osb->dentry_lock_work); - dl->dl_next = osb->dentry_lock_list; - osb->dentry_lock_list = dl; - spin_unlock(&dentry_list_lock); + kfree(dl); } void ocfs2_dentry_lock_put(struct ocfs2_super *osb, struct ocfs2_dentry_lock *dl) { - int unlock; + int unlock = 0; BUG_ON(dl->dl_count == 0); diff --git a/fs/ocfs2/dcache.h b/fs/ocfs2/dcache.h index faa12e75f98..55f58892b15 100644 --- a/fs/ocfs2/dcache.h +++ b/fs/ocfs2/dcache.h @@ -29,13 +29,8 @@ extern const struct dentry_operations ocfs2_dentry_ops; struct ocfs2_dentry_lock { - /* Use count of dentry lock */ unsigned int dl_count; - union { - /* Linked list of dentry locks to release */ - struct ocfs2_dentry_lock *dl_next; - u64 dl_parent_blkno; - }; + u64 dl_parent_blkno; /* * The ocfs2_dentry_lock keeps an inode reference until @@ -52,8 +47,6 @@ int ocfs2_dentry_attach_lock(struct dentry *dentry, struct inode *inode, void ocfs2_dentry_lock_put(struct ocfs2_super *osb, struct ocfs2_dentry_lock *dl); -void ocfs2_drop_dl_inodes(struct work_struct *work); - struct dentry *ocfs2_find_local_alias(struct inode *inode, u64 parent_blkno, int skip_unhashed); @@ -61,5 +54,6 @@ void ocfs2_dentry_move(struct dentry *dentry, struct dentry *target, struct inode *old_dir, struct inode *new_dir); extern spinlock_t dentry_attach_lock; +void ocfs2_dentry_attach_gen(struct dentry *dentry); #endif /* OCFS2_DCACHE_H */ diff --git a/fs/ocfs2/dir.c b/fs/ocfs2/dir.c index c5752305627..0717662b4ae 100644 --- a/fs/ocfs2/dir.c +++ b/fs/ocfs2/dir.c @@ -43,7 +43,6 @@ #include <linux/quotaops.h> #include <linux/sort.h> -#define MLOG_MASK_PREFIX ML_NAMEI #include <cluster/masklog.h> #include "ocfs2.h" @@ -61,13 +60,13 @@ #include "super.h" #include "sysfile.h" #include "uptodate.h" +#include "ocfs2_trace.h" #include "buffer_head_io.h" #define NAMEI_RA_CHUNKS 2 #define NAMEI_RA_BLOCKS 4 #define NAMEI_RA_SIZE (NAMEI_RA_CHUNKS * NAMEI_RA_BLOCKS) -#define NAMEI_RA_INDEX(c,b) (((c) * NAMEI_RA_BLOCKS) + (b)) static unsigned char ocfs2_filetype_table[] = { DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK @@ -176,7 +175,7 @@ static int ocfs2_dx_dir_link_trailer(struct inode *dir, handle_t *handle, struct ocfs2_dx_root_block *dx_root; struct ocfs2_dir_block_trailer *trailer; - ret = ocfs2_journal_access_dr(handle, dir, dx_root_bh, + ret = ocfs2_journal_access_dr(handle, INODE_CACHE(dir), dx_root_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); @@ -322,21 +321,23 @@ static int ocfs2_check_dir_entry(struct inode * dir, const char *error_msg = NULL; const int rlen = le16_to_cpu(de->rec_len); - if (rlen < OCFS2_DIR_REC_LEN(1)) + if (unlikely(rlen < OCFS2_DIR_REC_LEN(1))) error_msg = "rec_len is smaller than minimal"; - else if (rlen % 4 != 0) + else if (unlikely(rlen % 4 != 0)) error_msg = "rec_len % 4 != 0"; - else if (rlen < OCFS2_DIR_REC_LEN(de->name_len)) + else if (unlikely(rlen < OCFS2_DIR_REC_LEN(de->name_len))) error_msg = "rec_len is too small for name_len"; - else if (((char *) de - bh->b_data) + rlen > dir->i_sb->s_blocksize) + else if (unlikely( + ((char *) de - bh->b_data) + rlen > dir->i_sb->s_blocksize)) error_msg = "directory entry across blocks"; - if (error_msg != NULL) + if (unlikely(error_msg != NULL)) mlog(ML_ERROR, "bad entry in directory #%llu: %s - " "offset=%lu, inode=%llu, rec_len=%d, name_len=%d\n", (unsigned long long)OCFS2_I(dir)->ip_blkno, error_msg, offset, (unsigned long long)le64_to_cpu(de->inode), rlen, de->name_len); + return error_msg == NULL ? 1 : 0; } @@ -354,7 +355,7 @@ static inline int ocfs2_match(int len, /* * Returns 0 if not found, -1 on failure, and 1 on success */ -static int inline ocfs2_search_dirblock(struct buffer_head *bh, +static inline int ocfs2_search_dirblock(struct buffer_head *bh, struct inode *dir, const char *name, int namelen, unsigned long offset, @@ -367,8 +368,6 @@ static int inline ocfs2_search_dirblock(struct buffer_head *bh, int de_len; int ret = 0; - mlog_entry_void(); - de_buf = first_de; dlimit = de_buf + bytes; @@ -402,7 +401,7 @@ static int inline ocfs2_search_dirblock(struct buffer_head *bh, } bail: - mlog_exit(ret); + trace_ocfs2_search_dirblock(ret); return ret; } @@ -447,8 +446,7 @@ static int ocfs2_validate_dir_block(struct super_block *sb, * We don't validate dirents here, that's handled * in-place when the code walks them. */ - mlog(0, "Validating dirblock %llu\n", - (unsigned long long)bh->b_blocknr); + trace_ocfs2_validate_dir_block((unsigned long long)bh->b_blocknr); BUG_ON(!buffer_uptodate(bh)); @@ -564,7 +562,8 @@ static int ocfs2_read_dir_block_direct(struct inode *dir, u64 phys, int ret; struct buffer_head *tmp = *bh; - ret = ocfs2_read_block(dir, phys, &tmp, ocfs2_validate_dir_block); + ret = ocfs2_read_block(INODE_CACHE(dir), phys, &tmp, + ocfs2_validate_dir_block); if (ret) { mlog_errno(ret); goto out; @@ -622,7 +621,8 @@ static int ocfs2_read_dx_root(struct inode *dir, struct ocfs2_dinode *di, u64 blkno = le64_to_cpu(di->i_dx_root); struct buffer_head *tmp = *dx_root_bh; - ret = ocfs2_read_block(dir, blkno, &tmp, ocfs2_validate_dx_root); + ret = ocfs2_read_block(INODE_CACHE(dir), blkno, &tmp, + ocfs2_validate_dx_root); /* If ocfs2_read_block() got us a new bh, pass it up. */ if (!ret && !*dx_root_bh) @@ -662,7 +662,8 @@ static int ocfs2_read_dx_leaf(struct inode *dir, u64 blkno, int ret; struct buffer_head *tmp = *dx_leaf_bh; - ret = ocfs2_read_block(dir, blkno, &tmp, ocfs2_validate_dx_leaf); + ret = ocfs2_read_block(INODE_CACHE(dir), blkno, &tmp, + ocfs2_validate_dx_leaf); /* If ocfs2_read_block() got us a new bh, pass it up. */ if (!ret && !*dx_leaf_bh) @@ -680,7 +681,7 @@ static int ocfs2_read_dx_leaves(struct inode *dir, u64 start, int num, { int ret; - ret = ocfs2_read_blocks(dir, start, num, dx_leaf_bhs, 0, + ret = ocfs2_read_blocks(INODE_CACHE(dir), start, num, dx_leaf_bhs, 0, ocfs2_validate_dx_leaf); if (ret) mlog_errno(ret); @@ -703,8 +704,6 @@ static struct buffer_head *ocfs2_find_entry_el(const char *name, int namelen, int num = 0; int nblocks, i, err; - mlog_entry_void(); - sb = dir->i_sb; nblocks = i_size_read(dir) >> sb->s_blocksize_bits; @@ -785,7 +784,7 @@ cleanup_and_exit: for (; ra_ptr < ra_max; ra_ptr++) brelse(bh_use[ra_ptr]); - mlog_exit_ptr(ret); + trace_ocfs2_find_entry_el(ret); return ret; } @@ -802,7 +801,8 @@ static int ocfs2_dx_dir_lookup_rec(struct inode *inode, struct ocfs2_extent_rec *rec = NULL; if (el->l_tree_depth) { - ret = ocfs2_find_leaf(inode, el, major_hash, &eb_bh); + ret = ocfs2_find_leaf(INODE_CACHE(inode), el, major_hash, + &eb_bh); if (ret) { mlog_errno(ret); goto out; @@ -946,11 +946,9 @@ static int ocfs2_dx_dir_search(const char *name, int namelen, goto out; } - mlog(0, "Dir %llu: name: \"%.*s\", lookup of hash: %u.0x%x " - "returns: %llu\n", - (unsigned long long)OCFS2_I(dir)->ip_blkno, - namelen, name, hinfo->major_hash, hinfo->minor_hash, - (unsigned long long)phys); + trace_ocfs2_dx_dir_search((unsigned long long)OCFS2_I(dir)->ip_blkno, + namelen, name, hinfo->major_hash, + hinfo->minor_hash, (unsigned long long)phys); ret = ocfs2_read_dx_leaf(dir, phys, &dx_leaf_bh); if (ret) { @@ -960,9 +958,9 @@ static int ocfs2_dx_dir_search(const char *name, int namelen, dx_leaf = (struct ocfs2_dx_leaf *) dx_leaf_bh->b_data; - mlog(0, "leaf info: num_used: %d, count: %d\n", - le16_to_cpu(dx_leaf->dl_list.de_num_used), - le16_to_cpu(dx_leaf->dl_list.de_count)); + trace_ocfs2_dx_dir_search_leaf_info( + le16_to_cpu(dx_leaf->dl_list.de_num_used), + le16_to_cpu(dx_leaf->dl_list.de_count)); entry_list = &dx_leaf->dl_list; @@ -1133,7 +1131,8 @@ int ocfs2_update_entry(struct inode *dir, handle_t *handle, if (OCFS2_I(dir)->ip_dyn_features & OCFS2_INLINE_DATA_FL) access = ocfs2_journal_access_di; - ret = access(handle, dir, de_bh, OCFS2_JOURNAL_ACCESS_WRITE); + ret = access(handle, INODE_CACHE(dir), de_bh, + OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); goto out; @@ -1161,8 +1160,6 @@ static int __ocfs2_delete_entry(handle_t *handle, struct inode *dir, int i, status = -ENOENT; ocfs2_journal_access_func access = ocfs2_journal_access_db; - mlog_entry("(0x%p, 0x%p, 0x%p, 0x%p)\n", handle, dir, de_del, bh); - if (OCFS2_I(dir)->ip_dyn_features & OCFS2_INLINE_DATA_FL) access = ocfs2_journal_access_di; @@ -1176,7 +1173,7 @@ static int __ocfs2_delete_entry(handle_t *handle, struct inode *dir, goto bail; } if (de == de_del) { - status = access(handle, dir, bh, + status = access(handle, INODE_CACHE(dir), bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { status = -EIO; @@ -1186,10 +1183,9 @@ static int __ocfs2_delete_entry(handle_t *handle, struct inode *dir, if (pde) le16_add_cpu(&pde->rec_len, le16_to_cpu(de->rec_len)); - else - de->inode = 0; + de->inode = 0; dir->i_version++; - status = ocfs2_journal_dirty(handle, bh); + ocfs2_journal_dirty(handle, bh); goto bail; } i += le16_to_cpu(de->rec_len); @@ -1197,7 +1193,6 @@ static int __ocfs2_delete_entry(handle_t *handle, struct inode *dir, de = (struct ocfs2_dir_entry *)((char *)de + le16_to_cpu(de->rec_len)); } bail: - mlog_exit(status); return status; } @@ -1326,7 +1321,7 @@ static int ocfs2_delete_entry_dx(handle_t *handle, struct inode *dir, * the entry count needs to be updated. Also, we might be * adding to the start of the free list. */ - ret = ocfs2_journal_access_dr(handle, dir, dx_root_bh, + ret = ocfs2_journal_access_dr(handle, INODE_CACHE(dir), dx_root_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); @@ -1334,7 +1329,7 @@ static int ocfs2_delete_entry_dx(handle_t *handle, struct inode *dir, } if (!ocfs2_dx_root_inline(dx_root)) { - ret = ocfs2_journal_access_dl(handle, dir, + ret = ocfs2_journal_access_dl(handle, INODE_CACHE(dir), lookup->dl_dx_leaf_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { @@ -1343,8 +1338,8 @@ static int ocfs2_delete_entry_dx(handle_t *handle, struct inode *dir, } } - mlog(0, "Dir %llu: delete entry at index: %d\n", - (unsigned long long)OCFS2_I(dir)->ip_blkno, index); + trace_ocfs2_delete_entry_dx((unsigned long long)OCFS2_I(dir)->ip_blkno, + index); ret = __ocfs2_delete_entry(handle, dir, lookup->dl_entry, leaf_bh, leaf_bh->b_data, leaf_bh->b_size); @@ -1493,7 +1488,7 @@ static int __ocfs2_dx_dir_leaf_insert(struct inode *dir, handle_t *handle, int ret; struct ocfs2_dx_leaf *dx_leaf; - ret = ocfs2_journal_access_dl(handle, dir, dx_leaf_bh, + ret = ocfs2_journal_access_dl(handle, INODE_CACHE(dir), dx_leaf_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); @@ -1523,7 +1518,7 @@ static int ocfs2_dx_dir_insert(struct inode *dir, handle_t *handle, struct ocfs2_dx_root_block *dx_root; struct buffer_head *dx_root_bh = lookup->dl_dx_root_bh; - ret = ocfs2_journal_access_dr(handle, dir, dx_root_bh, + ret = ocfs2_journal_access_dr(handle, INODE_CACHE(dir), dx_root_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); @@ -1627,8 +1622,6 @@ int __ocfs2_add_entry(handle_t *handle, struct buffer_head *insert_bh = lookup->dl_leaf_bh; char *data_start = insert_bh->b_data; - mlog_entry_void(); - if (!namelen) return -EINVAL; @@ -1645,11 +1638,13 @@ int __ocfs2_add_entry(handle_t *handle, */ if (ocfs2_free_list_at_root(lookup)) { bh = lookup->dl_dx_root_bh; - retval = ocfs2_journal_access_dr(handle, dir, bh, + retval = ocfs2_journal_access_dr(handle, + INODE_CACHE(dir), bh, OCFS2_JOURNAL_ACCESS_WRITE); } else { bh = lookup->dl_prev_leaf_bh; - retval = ocfs2_journal_access_db(handle, dir, bh, + retval = ocfs2_journal_access_db(handle, + INODE_CACHE(dir), bh, OCFS2_JOURNAL_ACCESS_WRITE); } if (retval) { @@ -1700,11 +1695,13 @@ int __ocfs2_add_entry(handle_t *handle, } if (insert_bh == parent_fe_bh) - status = ocfs2_journal_access_di(handle, dir, + status = ocfs2_journal_access_di(handle, + INODE_CACHE(dir), insert_bh, OCFS2_JOURNAL_ACCESS_WRITE); else { - status = ocfs2_journal_access_db(handle, dir, + status = ocfs2_journal_access_db(handle, + INODE_CACHE(dir), insert_bh, OCFS2_JOURNAL_ACCESS_WRITE); @@ -1743,7 +1740,7 @@ int __ocfs2_add_entry(handle_t *handle, ocfs2_recalc_free_list(dir, handle, lookup); dir->i_version++; - status = ocfs2_journal_dirty(handle, insert_bh); + ocfs2_journal_dirty(handle, insert_bh); retval = 0; goto bail; } @@ -1756,18 +1753,18 @@ int __ocfs2_add_entry(handle_t *handle, * from ever getting here. */ retval = -ENOSPC; bail: + if (retval) + mlog_errno(retval); - mlog_exit(retval); return retval; } static int ocfs2_dir_foreach_blk_id(struct inode *inode, u64 *f_version, - loff_t *f_pos, void *priv, - filldir_t filldir, int *filldir_err) + struct dir_context *ctx) { - int ret, i, filldir_ret; - unsigned long offset = *f_pos; + int ret, i; + unsigned long offset = ctx->pos; struct buffer_head *di_bh = NULL; struct ocfs2_dinode *di; struct ocfs2_inline_data *data; @@ -1783,8 +1780,7 @@ static int ocfs2_dir_foreach_blk_id(struct inode *inode, di = (struct ocfs2_dinode *)di_bh->b_data; data = &di->id2.i_data; - while (*f_pos < i_size_read(inode)) { -revalidate: + while (ctx->pos < i_size_read(inode)) { /* If the dir block has changed since the last call to * readdir(2), then we might be pointing to an invalid * dirent right now. Scan from the start of the block @@ -1804,50 +1800,31 @@ revalidate: break; i += le16_to_cpu(de->rec_len); } - *f_pos = offset = i; + ctx->pos = offset = i; *f_version = inode->i_version; } - de = (struct ocfs2_dir_entry *) (data->id_data + *f_pos); - if (!ocfs2_check_dir_entry(inode, de, di_bh, *f_pos)) { + de = (struct ocfs2_dir_entry *) (data->id_data + ctx->pos); + if (!ocfs2_check_dir_entry(inode, de, di_bh, ctx->pos)) { /* On error, skip the f_pos to the end. */ - *f_pos = i_size_read(inode); - goto out; + ctx->pos = i_size_read(inode); + break; } offset += le16_to_cpu(de->rec_len); if (le64_to_cpu(de->inode)) { - /* We might block in the next section - * if the data destination is - * currently swapped out. So, use a - * version stamp to detect whether or - * not the directory has been modified - * during the copy operation. - */ - u64 version = *f_version; unsigned char d_type = DT_UNKNOWN; if (de->file_type < OCFS2_FT_MAX) d_type = ocfs2_filetype_table[de->file_type]; - filldir_ret = filldir(priv, de->name, - de->name_len, - *f_pos, - le64_to_cpu(de->inode), - d_type); - if (filldir_ret) { - if (filldir_err) - *filldir_err = filldir_ret; - break; - } - if (version != *f_version) - goto revalidate; + if (!dir_emit(ctx, de->name, de->name_len, + le64_to_cpu(de->inode), d_type)) + goto out; } - *f_pos += le16_to_cpu(de->rec_len); + ctx->pos += le16_to_cpu(de->rec_len); } - out: brelse(di_bh); - return 0; } @@ -1857,27 +1834,26 @@ out: */ static int ocfs2_dir_foreach_blk_el(struct inode *inode, u64 *f_version, - loff_t *f_pos, void *priv, - filldir_t filldir, int *filldir_err) + struct dir_context *ctx, + bool persist) { - int error = 0; unsigned long offset, blk, last_ra_blk = 0; - int i, stored; + int i; struct buffer_head * bh, * tmp; struct ocfs2_dir_entry * de; struct super_block * sb = inode->i_sb; unsigned int ra_sectors = 16; + int stored = 0; - stored = 0; bh = NULL; - offset = (*f_pos) & (sb->s_blocksize - 1); + offset = ctx->pos & (sb->s_blocksize - 1); - while (!error && !stored && *f_pos < i_size_read(inode)) { - blk = (*f_pos) >> sb->s_blocksize_bits; + while (ctx->pos < i_size_read(inode)) { + blk = ctx->pos >> sb->s_blocksize_bits; if (ocfs2_read_dir_block(inode, blk, &bh, 0)) { /* Skip the corrupt dirblock and keep trying */ - *f_pos += sb->s_blocksize - offset; + ctx->pos += sb->s_blocksize - offset; continue; } @@ -1899,7 +1875,6 @@ static int ocfs2_dir_foreach_blk_el(struct inode *inode, ra_sectors = 8; } -revalidate: /* If the dir block has changed since the last call to * readdir(2), then we might be pointing to an invalid * dirent right now. Scan from the start of the block @@ -1919,93 +1894,64 @@ revalidate: i += le16_to_cpu(de->rec_len); } offset = i; - *f_pos = ((*f_pos) & ~(sb->s_blocksize - 1)) + ctx->pos = (ctx->pos & ~(sb->s_blocksize - 1)) | offset; *f_version = inode->i_version; } - while (!error && *f_pos < i_size_read(inode) + while (ctx->pos < i_size_read(inode) && offset < sb->s_blocksize) { de = (struct ocfs2_dir_entry *) (bh->b_data + offset); if (!ocfs2_check_dir_entry(inode, de, bh, offset)) { /* On error, skip the f_pos to the next block. */ - *f_pos = ((*f_pos) | (sb->s_blocksize - 1)) + 1; + ctx->pos = (ctx->pos | (sb->s_blocksize - 1)) + 1; brelse(bh); - goto out; + continue; } - offset += le16_to_cpu(de->rec_len); if (le64_to_cpu(de->inode)) { - /* We might block in the next section - * if the data destination is - * currently swapped out. So, use a - * version stamp to detect whether or - * not the directory has been modified - * during the copy operation. - */ - unsigned long version = *f_version; unsigned char d_type = DT_UNKNOWN; if (de->file_type < OCFS2_FT_MAX) d_type = ocfs2_filetype_table[de->file_type]; - error = filldir(priv, de->name, + if (!dir_emit(ctx, de->name, de->name_len, - *f_pos, le64_to_cpu(de->inode), - d_type); - if (error) { - if (filldir_err) - *filldir_err = error; - break; + d_type)) { + brelse(bh); + return 0; } - if (version != *f_version) - goto revalidate; - stored ++; + stored++; } - *f_pos += le16_to_cpu(de->rec_len); + offset += le16_to_cpu(de->rec_len); + ctx->pos += le16_to_cpu(de->rec_len); } offset = 0; brelse(bh); bh = NULL; + if (!persist && stored) + break; } - - stored = 0; -out: - return stored; + return 0; } static int ocfs2_dir_foreach_blk(struct inode *inode, u64 *f_version, - loff_t *f_pos, void *priv, filldir_t filldir, - int *filldir_err) + struct dir_context *ctx, + bool persist) { if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) - return ocfs2_dir_foreach_blk_id(inode, f_version, f_pos, priv, - filldir, filldir_err); - - return ocfs2_dir_foreach_blk_el(inode, f_version, f_pos, priv, filldir, - filldir_err); + return ocfs2_dir_foreach_blk_id(inode, f_version, ctx); + return ocfs2_dir_foreach_blk_el(inode, f_version, ctx, persist); } /* * This is intended to be called from inside other kernel functions, * so we fake some arguments. */ -int ocfs2_dir_foreach(struct inode *inode, loff_t *f_pos, void *priv, - filldir_t filldir) +int ocfs2_dir_foreach(struct inode *inode, struct dir_context *ctx) { - int ret = 0, filldir_err = 0; u64 version = inode->i_version; - - while (*f_pos < i_size_read(inode)) { - ret = ocfs2_dir_foreach_blk(inode, &version, f_pos, priv, - filldir, &filldir_err); - if (ret || filldir_err) - break; - } - - if (ret > 0) - ret = -EIO; - + ocfs2_dir_foreach_blk(inode, &version, ctx, true); return 0; } @@ -2013,16 +1959,15 @@ int ocfs2_dir_foreach(struct inode *inode, loff_t *f_pos, void *priv, * ocfs2_readdir() * */ -int ocfs2_readdir(struct file * filp, void * dirent, filldir_t filldir) +int ocfs2_readdir(struct file *file, struct dir_context *ctx) { int error = 0; - struct inode *inode = filp->f_path.dentry->d_inode; + struct inode *inode = file_inode(file); int lock_level = 0; - mlog_entry("dirino=%llu\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno); + trace_ocfs2_readdir((unsigned long long)OCFS2_I(inode)->ip_blkno); - error = ocfs2_inode_lock_atime(inode, filp->f_vfsmnt, &lock_level); + error = ocfs2_inode_lock_atime(inode, file->f_path.mnt, &lock_level); if (lock_level && error >= 0) { /* We release EX lock which used to update atime * and get PR lock again to reduce contention @@ -2038,13 +1983,13 @@ int ocfs2_readdir(struct file * filp, void * dirent, filldir_t filldir) goto bail_nolock; } - error = ocfs2_dir_foreach_blk(inode, &filp->f_version, &filp->f_pos, - dirent, filldir, NULL); + error = ocfs2_dir_foreach_blk(inode, &file->f_version, ctx, false); ocfs2_inode_unlock(inode, lock_level); + if (error) + mlog_errno(error); bail_nolock: - mlog_exit(error); return error; } @@ -2060,8 +2005,8 @@ int ocfs2_find_files_on_disk(const char *name, { int status = -ENOENT; - mlog(0, "name=%.*s, blkno=%p, inode=%llu\n", namelen, name, blkno, - (unsigned long long)OCFS2_I(inode)->ip_blkno); + trace_ocfs2_find_files_on_disk(namelen, name, blkno, + (unsigned long long)OCFS2_I(inode)->ip_blkno); status = ocfs2_find_entry(name, namelen, inode, lookup); if (status) @@ -2105,8 +2050,8 @@ int ocfs2_check_dir_for_entry(struct inode *dir, int ret; struct ocfs2_dir_lookup_result lookup = { NULL, }; - mlog_entry("dir %llu, name '%.*s'\n", - (unsigned long long)OCFS2_I(dir)->ip_blkno, namelen, name); + trace_ocfs2_check_dir_for_entry( + (unsigned long long)OCFS2_I(dir)->ip_blkno, namelen, name); ret = -EEXIST; if (ocfs2_find_entry(name, namelen, dir, &lookup) == 0) @@ -2116,11 +2061,13 @@ int ocfs2_check_dir_for_entry(struct inode *dir, bail: ocfs2_free_dir_lookup_result(&lookup); - mlog_exit(ret); + if (ret) + mlog_errno(ret); return ret; } struct ocfs2_empty_dir_priv { + struct dir_context ctx; unsigned seen_dot; unsigned seen_dot_dot; unsigned seen_other; @@ -2205,10 +2152,9 @@ out: int ocfs2_empty_dir(struct inode *inode) { int ret; - loff_t start = 0; - struct ocfs2_empty_dir_priv priv; - - memset(&priv, 0, sizeof(priv)); + struct ocfs2_empty_dir_priv priv = { + .ctx.actor = ocfs2_empty_dir_filldir, + }; if (ocfs2_dir_indexed(inode)) { ret = ocfs2_empty_dir_dx(inode, &priv); @@ -2220,7 +2166,7 @@ int ocfs2_empty_dir(struct inode *inode) */ } - ret = ocfs2_dir_foreach(inode, &start, &priv, ocfs2_empty_dir_filldir); + ret = ocfs2_dir_foreach(inode, &priv.ctx); if (ret) mlog_errno(ret); @@ -2280,7 +2226,7 @@ static int ocfs2_fill_new_dir_id(struct ocfs2_super *osb, struct ocfs2_inline_data *data = &di->id2.i_data; unsigned int size = le16_to_cpu(data->id_count); - ret = ocfs2_journal_access_di(handle, inode, di_bh, + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); @@ -2288,15 +2234,10 @@ static int ocfs2_fill_new_dir_id(struct ocfs2_super *osb, } ocfs2_fill_initial_dirents(inode, parent, data->id_data, size); - ocfs2_journal_dirty(handle, di_bh); - if (ret) { - mlog_errno(ret); - goto out; - } i_size_write(inode, size); - inode->i_nlink = 2; + set_nlink(inode, 2); inode->i_blocks = ocfs2_inode_sector_count(inode); ret = ocfs2_mark_inode_dirty(handle, inode, di_bh); @@ -2320,8 +2261,6 @@ static int ocfs2_fill_new_dir_el(struct ocfs2_super *osb, struct buffer_head *new_bh = NULL; struct ocfs2_dir_entry *de; - mlog_entry_void(); - if (ocfs2_new_dir_wants_trailer(inode)) size = ocfs2_dir_trailer_blk_off(parent->i_sb); @@ -2332,9 +2271,9 @@ static int ocfs2_fill_new_dir_el(struct ocfs2_super *osb, goto bail; } - ocfs2_set_new_buffer_uptodate(inode, new_bh); + ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode), new_bh); - status = ocfs2_journal_access_db(handle, inode, new_bh, + status = ocfs2_journal_access_db(handle, INODE_CACHE(inode), new_bh, OCFS2_JOURNAL_ACCESS_CREATE); if (status < 0) { mlog_errno(status); @@ -2357,14 +2296,10 @@ static int ocfs2_fill_new_dir_el(struct ocfs2_super *osb, ocfs2_init_dir_trailer(inode, new_bh, size); } - status = ocfs2_journal_dirty(handle, new_bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } + ocfs2_journal_dirty(handle, new_bh); i_size_write(inode, inode->i_sb->s_blocksize); - inode->i_nlink = 2; + set_nlink(inode, 2); inode->i_blocks = ocfs2_inode_sector_count(inode); status = ocfs2_mark_inode_dirty(handle, inode, fe_bh); if (status < 0) { @@ -2380,7 +2315,6 @@ static int ocfs2_fill_new_dir_el(struct ocfs2_super *osb, bail: brelse(new_bh); - mlog_exit(status); return status; } @@ -2395,32 +2329,32 @@ static int ocfs2_dx_dir_attach_index(struct ocfs2_super *osb, int ret; struct ocfs2_dinode *di = (struct ocfs2_dinode *) di_bh->b_data; u16 dr_suballoc_bit; - u64 dr_blkno; + u64 suballoc_loc, dr_blkno; unsigned int num_bits; struct buffer_head *dx_root_bh = NULL; struct ocfs2_dx_root_block *dx_root; struct ocfs2_dir_block_trailer *trailer = ocfs2_trailer_from_bh(dirdata_bh, dir->i_sb); - ret = ocfs2_claim_metadata(osb, handle, meta_ac, 1, &dr_suballoc_bit, - &num_bits, &dr_blkno); + ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc, + &dr_suballoc_bit, &num_bits, &dr_blkno); if (ret) { mlog_errno(ret); goto out; } - mlog(0, "Dir %llu, attach new index block: %llu\n", - (unsigned long long)OCFS2_I(dir)->ip_blkno, - (unsigned long long)dr_blkno); + trace_ocfs2_dx_dir_attach_index( + (unsigned long long)OCFS2_I(dir)->ip_blkno, + (unsigned long long)dr_blkno); dx_root_bh = sb_getblk(osb->sb, dr_blkno); if (dx_root_bh == NULL) { - ret = -EIO; + ret = -ENOMEM; goto out; } - ocfs2_set_new_buffer_uptodate(dir, dx_root_bh); + ocfs2_set_new_buffer_uptodate(INODE_CACHE(dir), dx_root_bh); - ret = ocfs2_journal_access_dr(handle, dir, dx_root_bh, + ret = ocfs2_journal_access_dr(handle, INODE_CACHE(dir), dx_root_bh, OCFS2_JOURNAL_ACCESS_CREATE); if (ret < 0) { mlog_errno(ret); @@ -2430,7 +2364,8 @@ static int ocfs2_dx_dir_attach_index(struct ocfs2_super *osb, dx_root = (struct ocfs2_dx_root_block *)dx_root_bh->b_data; memset(dx_root, 0, osb->sb->s_blocksize); strcpy(dx_root->dr_signature, OCFS2_DX_ROOT_SIGNATURE); - dx_root->dr_suballoc_slot = cpu_to_le16(osb->slot_num); + dx_root->dr_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot); + dx_root->dr_suballoc_loc = cpu_to_le64(suballoc_loc); dx_root->dr_suballoc_bit = cpu_to_le16(dr_suballoc_bit); dx_root->dr_fs_generation = cpu_to_le32(osb->fs_generation); dx_root->dr_blkno = cpu_to_le64(dr_blkno); @@ -2449,12 +2384,9 @@ static int ocfs2_dx_dir_attach_index(struct ocfs2_super *osb, dx_root->dr_list.l_count = cpu_to_le16(ocfs2_extent_recs_per_dx_root(osb->sb)); } + ocfs2_journal_dirty(handle, dx_root_bh); - ret = ocfs2_journal_dirty(handle, dx_root_bh); - if (ret) - mlog_errno(ret); - - ret = ocfs2_journal_access_di(handle, dir, di_bh, + ret = ocfs2_journal_access_di(handle, INODE_CACHE(dir), di_bh, OCFS2_JOURNAL_ACCESS_CREATE); if (ret) { mlog_errno(ret); @@ -2463,12 +2395,12 @@ static int ocfs2_dx_dir_attach_index(struct ocfs2_super *osb, di->i_dx_root = cpu_to_le64(dr_blkno); + spin_lock(&OCFS2_I(dir)->ip_lock); OCFS2_I(dir)->ip_dyn_features |= OCFS2_INDEXED_DIR_FL; di->i_dyn_features = cpu_to_le16(OCFS2_I(dir)->ip_dyn_features); + spin_unlock(&OCFS2_I(dir)->ip_lock); - ret = ocfs2_journal_dirty(handle, di_bh); - if (ret) - mlog_errno(ret); + ocfs2_journal_dirty(handle, di_bh); *ret_dx_root_bh = dx_root_bh; dx_root_bh = NULL; @@ -2490,14 +2422,14 @@ static int ocfs2_dx_dir_format_cluster(struct ocfs2_super *osb, for (i = 0; i < num_dx_leaves; i++) { bh = sb_getblk(osb->sb, start_blk + i); if (bh == NULL) { - ret = -EIO; + ret = -ENOMEM; goto out; } dx_leaves[i] = bh; - ocfs2_set_new_buffer_uptodate(dir, bh); + ocfs2_set_new_buffer_uptodate(INODE_CACHE(dir), bh); - ret = ocfs2_journal_access_dl(handle, dir, bh, + ret = ocfs2_journal_access_dl(handle, INODE_CACHE(dir), bh, OCFS2_JOURNAL_ACCESS_CREATE); if (ret < 0) { mlog_errno(ret); @@ -2513,11 +2445,10 @@ static int ocfs2_dx_dir_format_cluster(struct ocfs2_super *osb, dx_leaf->dl_list.de_count = cpu_to_le16(ocfs2_dx_entries_per_leaf(osb->sb)); - mlog(0, - "Dir %llu, format dx_leaf: %llu, entry count: %u\n", - (unsigned long long)OCFS2_I(dir)->ip_blkno, - (unsigned long long)bh->b_blocknr, - le16_to_cpu(dx_leaf->dl_list.de_count)); + trace_ocfs2_dx_dir_format_cluster( + (unsigned long long)OCFS2_I(dir)->ip_blkno, + (unsigned long long)bh->b_blocknr, + le16_to_cpu(dx_leaf->dl_list.de_count)); ocfs2_journal_dirty(handle, bh); } @@ -2549,7 +2480,7 @@ static int __ocfs2_dx_dir_new_cluster(struct inode *dir, * chance of contiguousness as the directory grows in number * of entries. */ - ret = __ocfs2_claim_clusters(osb, handle, data_ac, 1, 1, &phys, &num); + ret = __ocfs2_claim_clusters(handle, data_ac, 1, 1, &phys, &num); if (ret) { mlog_errno(ret); goto out; @@ -2582,7 +2513,6 @@ static int ocfs2_dx_dir_new_cluster(struct inode *dir, { int ret; u64 phys_blkno; - struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); ret = __ocfs2_dx_dir_new_cluster(dir, cpos, handle, data_ac, dx_leaves, num_dx_leaves, &phys_blkno); @@ -2591,7 +2521,7 @@ static int ocfs2_dx_dir_new_cluster(struct inode *dir, goto out; } - ret = ocfs2_insert_extent(osb, handle, dir, et, cpos, phys_blkno, 1, 0, + ret = ocfs2_insert_extent(handle, et, cpos, phys_blkno, 1, 0, meta_ac); if (ret) mlog_errno(ret); @@ -2762,12 +2692,11 @@ static void ocfs2_dx_dir_index_root_block(struct inode *dir, ocfs2_dx_dir_name_hash(dir, de->name, de->name_len, &hinfo); - mlog(0, - "dir: %llu, major: 0x%x minor: 0x%x, index: %u, name: %.*s\n", - (unsigned long long)dir->i_ino, hinfo.major_hash, - hinfo.minor_hash, - le16_to_cpu(dx_root->dr_entries.de_num_used), - de->name_len, de->name); + trace_ocfs2_dx_dir_index_root_block( + (unsigned long long)dir->i_ino, + hinfo.major_hash, hinfo.minor_hash, + de->name_len, de->name, + le16_to_cpu(dx_root->dr_entries.de_num_used)); ocfs2_dx_entry_list_insert(&dx_root->dr_entries, &hinfo, dirent_blk); @@ -2884,7 +2813,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh, bytes = blocks_wanted << sb->s_blocksize_bits; struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); struct ocfs2_inode_info *oi = OCFS2_I(dir); - struct ocfs2_alloc_context *data_ac; + struct ocfs2_alloc_context *data_ac = NULL; struct ocfs2_alloc_context *meta_ac = NULL; struct buffer_head *dirdata_bh = NULL; struct buffer_head *dx_root_bh = NULL; @@ -2895,11 +2824,13 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh, struct ocfs2_extent_tree dx_et; int did_quota = 0, bytes_allocated = 0; - ocfs2_init_dinode_extent_tree(&et, dir, di_bh); + ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(dir), di_bh); alloc = ocfs2_clusters_for_bytes(sb, bytes); dx_alloc = 0; + down_write(&oi->ip_alloc_sem); + if (ocfs2_supports_indexed_dirs(osb)) { credits += ocfs2_add_dir_index_credits(sb); @@ -2940,8 +2871,6 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh, goto out; } - down_write(&oi->ip_alloc_sem); - /* * Prepare for worst case allocation scenario of two separate * extents in the unindexed tree. @@ -2953,15 +2882,13 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh, if (IS_ERR(handle)) { ret = PTR_ERR(handle); mlog_errno(ret); - goto out_sem; + goto out; } - if (vfs_dq_alloc_space_nodirty(dir, - ocfs2_clusters_to_bytes(osb->sb, - alloc + dx_alloc))) { - ret = -EDQUOT; + ret = dquot_alloc_space_nodirty(dir, + ocfs2_clusters_to_bytes(osb->sb, alloc + dx_alloc)); + if (ret) goto out_commit; - } did_quota = 1; if (ocfs2_supports_indexed_dirs(osb) && !dx_inline) { @@ -2985,7 +2912,9 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh, * if we only get one now, that's enough to continue. The rest * will be claimed after the conversion to extents. */ - ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off, &len); + if (ocfs2_dir_resv_allowed(osb)) + data_ac->ac_resv = &oi->ip_la_data_resv; + ret = ocfs2_claim_clusters(handle, data_ac, 1, &bit_off, &len); if (ret) { mlog_errno(ret); goto out_commit; @@ -3000,14 +2929,14 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh, blkno = ocfs2_clusters_to_blocks(dir->i_sb, bit_off); dirdata_bh = sb_getblk(sb, blkno); if (!dirdata_bh) { - ret = -EIO; + ret = -ENOMEM; mlog_errno(ret); goto out_commit; } - ocfs2_set_new_buffer_uptodate(dir, dirdata_bh); + ocfs2_set_new_buffer_uptodate(INODE_CACHE(dir), dirdata_bh); - ret = ocfs2_journal_access_db(handle, dir, dirdata_bh, + ret = ocfs2_journal_access_db(handle, INODE_CACHE(dir), dirdata_bh, OCFS2_JOURNAL_ACCESS_CREATE); if (ret) { mlog_errno(ret); @@ -3028,11 +2957,8 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh, ocfs2_init_dir_trailer(dir, dirdata_bh, i); } - ret = ocfs2_journal_dirty(handle, dirdata_bh); - if (ret) { - mlog_errno(ret); - goto out_commit; - } + ocfs2_update_inode_fsync_trans(handle, dir, 1); + ocfs2_journal_dirty(handle, dirdata_bh); if (ocfs2_supports_indexed_dirs(osb) && !dx_inline) { /* @@ -3060,7 +2986,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh, * We let the later dirent insert modify c/mtime - to the user * the data hasn't changed. */ - ret = ocfs2_journal_access_di(handle, dir, di_bh, + ret = ocfs2_journal_access_di(handle, INODE_CACHE(dir), di_bh, OCFS2_JOURNAL_ACCESS_CREATE); if (ret) { mlog_errno(ret); @@ -3080,12 +3006,13 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh, di->i_size = cpu_to_le64(sb->s_blocksize); di->i_ctime = di->i_mtime = cpu_to_le64(dir->i_ctime.tv_sec); di->i_ctime_nsec = di->i_mtime_nsec = cpu_to_le32(dir->i_ctime.tv_nsec); + ocfs2_update_inode_fsync_trans(handle, dir, 1); /* * This should never fail as our extent list is empty and all * related blocks have been journaled already. */ - ret = ocfs2_insert_extent(osb, handle, dir, &et, 0, blkno, len, + ret = ocfs2_insert_extent(handle, &et, 0, blkno, len, 0, NULL); if (ret) { mlog_errno(ret); @@ -3098,11 +3025,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh, */ dir->i_blocks = ocfs2_inode_sector_count(dir); - ret = ocfs2_journal_dirty(handle, di_bh); - if (ret) { - mlog_errno(ret); - goto out_commit; - } + ocfs2_journal_dirty(handle, di_bh); if (ocfs2_supports_indexed_dirs(osb)) { ret = ocfs2_dx_dir_attach_index(osb, handle, dir, di_bh, @@ -3117,8 +3040,10 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh, ocfs2_dx_dir_index_root_block(dir, dx_root_bh, dirdata_bh); } else { - ocfs2_init_dx_root_extent_tree(&dx_et, dir, dx_root_bh); - ret = ocfs2_insert_extent(osb, handle, dir, &dx_et, 0, + ocfs2_init_dx_root_extent_tree(&dx_et, + INODE_CACHE(dir), + dx_root_bh); + ret = ocfs2_insert_extent(handle, &dx_et, 0, dx_insert_blkno, 1, 0, NULL); if (ret) mlog_errno(ret); @@ -3130,7 +3055,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh, * pass. Claim the 2nd cluster as a separate extent. */ if (alloc > len) { - ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off, + ret = ocfs2_claim_clusters(handle, data_ac, 1, &bit_off, &len); if (ret) { mlog_errno(ret); @@ -3138,7 +3063,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh, } blkno = ocfs2_clusters_to_blocks(dir->i_sb, bit_off); - ret = ocfs2_insert_extent(osb, handle, dir, &et, 1, + ret = ocfs2_insert_extent(handle, &et, 1, blkno, len, 0, NULL); if (ret) { mlog_errno(ret); @@ -3168,14 +3093,12 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh, out_commit: if (ret < 0 && did_quota) - vfs_dq_free_space_nodirty(dir, bytes_allocated); + dquot_free_space_nodirty(dir, bytes_allocated); ocfs2_commit_trans(osb, handle); -out_sem: - up_write(&oi->ip_alloc_sem); - out: + up_write(&oi->ip_alloc_sem); if (data_ac) ocfs2_free_alloc_context(data_ac); if (meta_ac) @@ -3213,11 +3136,10 @@ static int ocfs2_do_extend_dir(struct super_block *sb, if (extend) { u32 offset = OCFS2_I(dir)->ip_clusters; - if (vfs_dq_alloc_space_nodirty(dir, - ocfs2_clusters_to_bytes(sb, 1))) { - status = -EDQUOT; + status = dquot_alloc_space_nodirty(dir, + ocfs2_clusters_to_bytes(sb, 1)); + if (status) goto bail; - } did_quota = 1; status = ocfs2_add_inode_data(OCFS2_SB(sb), dir, &offset, @@ -3239,15 +3161,14 @@ static int ocfs2_do_extend_dir(struct super_block *sb, *new_bh = sb_getblk(sb, p_blkno); if (!*new_bh) { - status = -EIO; + status = -ENOMEM; mlog_errno(status); goto bail; } status = 0; bail: if (did_quota && status < 0) - vfs_dq_free_space_nodirty(dir, ocfs2_clusters_to_bytes(sb, 1)); - mlog_exit(status); + dquot_free_space_nodirty(dir, ocfs2_clusters_to_bytes(sb, 1)); return status; } @@ -3282,8 +3203,6 @@ static int ocfs2_extend_dir(struct ocfs2_super *osb, struct ocfs2_extent_tree et; struct buffer_head *dx_root_bh = lookup->dl_dx_root_bh; - mlog_entry_void(); - if (OCFS2_I(dir)->ip_dyn_features & OCFS2_INLINE_DATA_FL) { /* * This would be a code error as an inline directory should @@ -3322,21 +3241,26 @@ static int ocfs2_extend_dir(struct ocfs2_super *osb, brelse(new_bh); new_bh = NULL; + down_write(&OCFS2_I(dir)->ip_alloc_sem); + drop_alloc_sem = 1; dir_i_size = i_size_read(dir); credits = OCFS2_SIMPLE_DIR_EXTEND_CREDITS; goto do_extend; } + down_write(&OCFS2_I(dir)->ip_alloc_sem); + drop_alloc_sem = 1; dir_i_size = i_size_read(dir); - mlog(0, "extending dir %llu (i_size = %lld)\n", - (unsigned long long)OCFS2_I(dir)->ip_blkno, dir_i_size); + trace_ocfs2_extend_dir((unsigned long long)OCFS2_I(dir)->ip_blkno, + dir_i_size); /* dir->i_size is always block aligned. */ spin_lock(&OCFS2_I(dir)->ip_lock); if (dir_i_size == ocfs2_clusters_to_bytes(sb, OCFS2_I(dir)->ip_clusters)) { spin_unlock(&OCFS2_I(dir)->ip_lock); - ocfs2_init_dinode_extent_tree(&et, dir, parent_fe_bh); - num_free_extents = ocfs2_num_free_extents(osb, dir, &et); + ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(dir), + parent_fe_bh); + num_free_extents = ocfs2_num_free_extents(osb, &et); if (num_free_extents < 0) { status = num_free_extents; mlog_errno(status); @@ -3359,7 +3283,10 @@ static int ocfs2_extend_dir(struct ocfs2_super *osb, goto bail; } - credits = ocfs2_calc_extend_credits(sb, el, 1); + if (ocfs2_dir_resv_allowed(osb)) + data_ac->ac_resv = &OCFS2_I(dir)->ip_la_data_resv; + + credits = ocfs2_calc_extend_credits(sb, el); } else { spin_unlock(&OCFS2_I(dir)->ip_lock); credits = OCFS2_SIMPLE_DIR_EXTEND_CREDITS; @@ -3370,9 +3297,6 @@ do_extend: credits++; /* For attaching the new dirent block to the * dx_root */ - down_write(&OCFS2_I(dir)->ip_alloc_sem); - drop_alloc_sem = 1; - handle = ocfs2_start_trans(osb, credits); if (IS_ERR(handle)) { status = PTR_ERR(handle); @@ -3388,9 +3312,9 @@ do_extend: goto bail; } - ocfs2_set_new_buffer_uptodate(dir, new_bh); + ocfs2_set_new_buffer_uptodate(INODE_CACHE(dir), new_bh); - status = ocfs2_journal_access_db(handle, dir, new_bh, + status = ocfs2_journal_access_db(handle, INODE_CACHE(dir), new_bh, OCFS2_JOURNAL_ACCESS_CREATE); if (status < 0) { mlog_errno(status); @@ -3416,11 +3340,8 @@ do_extend: } else { de->rec_len = cpu_to_le16(sb->s_blocksize); } - status = ocfs2_journal_dirty(handle, new_bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } + ocfs2_update_inode_fsync_trans(handle, dir, 1); + ocfs2_journal_dirty(handle, new_bh); dir_i_size += dir->i_sb->s_blocksize; i_size_write(dir, dir_i_size); @@ -3435,10 +3356,10 @@ bail_bh: *new_de_bh = new_bh; get_bh(*new_de_bh); bail: - if (drop_alloc_sem) - up_write(&OCFS2_I(dir)->ip_alloc_sem); if (handle) ocfs2_commit_trans(osb, handle); + if (drop_alloc_sem) + up_write(&OCFS2_I(dir)->ip_alloc_sem); if (data_ac) ocfs2_free_alloc_context(data_ac); @@ -3447,7 +3368,6 @@ bail: brelse(new_bh); - mlog_exit(status); return status; } @@ -3594,8 +3514,9 @@ next: status = 0; bail: brelse(bh); + if (status) + mlog_errno(status); - mlog_exit(status); return status; } @@ -3798,7 +3719,7 @@ static int ocfs2_dx_dir_rebalance_credits(struct ocfs2_super *osb, { int credits = ocfs2_clusters_to_blocks(osb->sb, 2); - credits += ocfs2_calc_extend_credits(osb->sb, &dx_root->dr_list, 1); + credits += ocfs2_calc_extend_credits(osb->sb, &dx_root->dr_list); credits += ocfs2_quota_trans_credits(osb->sb); return credits; } @@ -3826,11 +3747,11 @@ static int ocfs2_dx_dir_rebalance(struct ocfs2_super *osb, struct inode *dir, struct ocfs2_dx_root_block *dx_root; struct ocfs2_dx_leaf *tmp_dx_leaf = NULL; - mlog(0, "DX Dir: %llu, rebalance leaf leaf_blkno: %llu insert: %u\n", - (unsigned long long)OCFS2_I(dir)->ip_blkno, - (unsigned long long)leaf_blkno, insert_hash); + trace_ocfs2_dx_dir_rebalance((unsigned long long)OCFS2_I(dir)->ip_blkno, + (unsigned long long)leaf_blkno, + insert_hash); - ocfs2_init_dx_root_extent_tree(&et, dir, dx_root_bh); + ocfs2_init_dx_root_extent_tree(&et, INODE_CACHE(dir), dx_root_bh); dx_root = (struct ocfs2_dx_root_block *)dx_root_bh->b_data; /* @@ -3879,14 +3800,13 @@ static int ocfs2_dx_dir_rebalance(struct ocfs2_super *osb, struct inode *dir, goto out; } - if (vfs_dq_alloc_space_nodirty(dir, - ocfs2_clusters_to_bytes(dir->i_sb, 1))) { - ret = -EDQUOT; + ret = dquot_alloc_space_nodirty(dir, + ocfs2_clusters_to_bytes(dir->i_sb, 1)); + if (ret) goto out_commit; - } did_quota = 1; - ret = ocfs2_journal_access_dl(handle, dir, dx_leaf_bh, + ret = ocfs2_journal_access_dl(handle, INODE_CACHE(dir), dx_leaf_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); @@ -3900,11 +3820,7 @@ static int ocfs2_dx_dir_rebalance(struct ocfs2_super *osb, struct inode *dir, sizeof(struct ocfs2_dx_entry), dx_leaf_sort_cmp, dx_leaf_sort_swap); - ret = ocfs2_journal_dirty(handle, dx_leaf_bh); - if (ret) { - mlog_errno(ret); - goto out_commit; - } + ocfs2_journal_dirty(handle, dx_leaf_bh); ret = ocfs2_dx_dir_find_leaf_split(dx_leaf, leaf_cpos, insert_hash, &split_hash); @@ -3913,8 +3829,7 @@ static int ocfs2_dx_dir_rebalance(struct ocfs2_super *osb, struct inode *dir, goto out_commit; } - mlog(0, "Split leaf (%u) at %u, insert major hash is %u\n", - leaf_cpos, split_hash, insert_hash); + trace_ocfs2_dx_dir_rebalance_split(leaf_cpos, split_hash, insert_hash); /* * We have to carefully order operations here. There are items @@ -3949,15 +3864,6 @@ static int ocfs2_dx_dir_rebalance(struct ocfs2_super *osb, struct inode *dir, goto out_commit; } - for (i = 0; i < num_dx_leaves; i++) { - ret = ocfs2_journal_access_dl(handle, dir, orig_dx_leaves[i], - OCFS2_JOURNAL_ACCESS_WRITE); - if (ret) { - mlog_errno(ret); - goto out_commit; - } - } - cpos = split_hash; ret = ocfs2_dx_dir_new_cluster(dir, &et, cpos, handle, data_ac, meta_ac, new_dx_leaves, @@ -3967,14 +3873,33 @@ static int ocfs2_dx_dir_rebalance(struct ocfs2_super *osb, struct inode *dir, goto out_commit; } + for (i = 0; i < num_dx_leaves; i++) { + ret = ocfs2_journal_access_dl(handle, INODE_CACHE(dir), + orig_dx_leaves[i], + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + ret = ocfs2_journal_access_dl(handle, INODE_CACHE(dir), + new_dx_leaves[i], + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + } + ocfs2_dx_dir_transfer_leaf(dir, split_hash, handle, tmp_dx_leaf, orig_dx_leaves, new_dx_leaves, num_dx_leaves); out_commit: if (ret < 0 && did_quota) - vfs_dq_free_space_nodirty(dir, + dquot_free_space_nodirty(dir, ocfs2_clusters_to_bytes(dir->i_sb, 1)); + ocfs2_update_inode_fsync_trans(handle, dir, 1); ocfs2_commit_trans(osb, handle); out: @@ -4154,11 +4079,10 @@ static int ocfs2_expand_inline_dx_root(struct inode *dir, goto out; } - if (vfs_dq_alloc_space_nodirty(dir, - ocfs2_clusters_to_bytes(osb->sb, 1))) { - ret = -EDQUOT; + ret = dquot_alloc_space_nodirty(dir, + ocfs2_clusters_to_bytes(osb->sb, 1)); + if (ret) goto out_commit; - } did_quota = 1; /* @@ -4166,7 +4090,7 @@ static int ocfs2_expand_inline_dx_root(struct inode *dir, * failure to add the dx_root_bh to the journal won't result * us losing clusters. */ - ret = ocfs2_journal_access_dr(handle, dir, dx_root_bh, + ret = ocfs2_journal_access_dr(handle, INODE_CACHE(dir), dx_root_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); @@ -4208,18 +4132,18 @@ static int ocfs2_expand_inline_dx_root(struct inode *dir, /* This should never fail considering we start with an empty * dx_root. */ - ocfs2_init_dx_root_extent_tree(&et, dir, dx_root_bh); - ret = ocfs2_insert_extent(osb, handle, dir, &et, 0, - insert_blkno, 1, 0, NULL); + ocfs2_init_dx_root_extent_tree(&et, INODE_CACHE(dir), dx_root_bh); + ret = ocfs2_insert_extent(handle, &et, 0, insert_blkno, 1, 0, NULL); if (ret) mlog_errno(ret); did_quota = 0; + ocfs2_update_inode_fsync_trans(handle, dir, 1); ocfs2_journal_dirty(handle, dx_root_bh); out_commit: if (ret < 0 && did_quota) - vfs_dq_free_space_nodirty(dir, + dquot_free_space_nodirty(dir, ocfs2_clusters_to_bytes(dir->i_sb, 1)); ocfs2_commit_trans(osb, handle); @@ -4364,8 +4288,8 @@ int ocfs2_prepare_dir_for_insert(struct ocfs2_super *osb, unsigned int blocks_wanted = 1; struct buffer_head *bh = NULL; - mlog(0, "getting ready to insert namelen %d into dir %llu\n", - namelen, (unsigned long long)OCFS2_I(dir)->ip_blkno); + trace_ocfs2_prepare_dir_for_insert( + (unsigned long long)OCFS2_I(dir)->ip_blkno, namelen); if (!namelen) { ret = -EINVAL; @@ -4470,22 +4394,28 @@ static int ocfs2_dx_dir_remove_index(struct inode *dir, goto out_unlock; } - ret = ocfs2_journal_access_di(handle, dir, di_bh, + ret = ocfs2_journal_access_di(handle, INODE_CACHE(dir), di_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); goto out_commit; } + spin_lock(&OCFS2_I(dir)->ip_lock); OCFS2_I(dir)->ip_dyn_features &= ~OCFS2_INDEXED_DIR_FL; di->i_dyn_features = cpu_to_le16(OCFS2_I(dir)->ip_dyn_features); + spin_unlock(&OCFS2_I(dir)->ip_lock); di->i_dx_root = cpu_to_le64(0ULL); + ocfs2_update_inode_fsync_trans(handle, dir, 1); ocfs2_journal_dirty(handle, di_bh); blk = le64_to_cpu(dx_root->dr_blkno); bit = le16_to_cpu(dx_root->dr_suballoc_bit); - bg_blkno = ocfs2_which_suballoc_group(blk, bit); + if (dx_root->dr_suballoc_loc) + bg_blkno = le64_to_cpu(dx_root->dr_suballoc_loc); + else + bg_blkno = ocfs2_which_suballoc_group(blk, bit); ret = ocfs2_free_suballoc_bits(handle, dx_alloc_inode, dx_alloc_bh, bit, bg_blkno, 1); if (ret) @@ -4533,7 +4463,7 @@ int ocfs2_dx_dir_truncate(struct inode *dir, struct buffer_head *di_bh) if (ocfs2_dx_root_inline(dx_root)) goto remove_index; - ocfs2_init_dx_root_extent_tree(&et, dir, dx_root_bh); + ocfs2_init_dx_root_extent_tree(&et, INODE_CACHE(dir), dx_root_bh); /* XXX: What if dr_clusters is too large? */ while (le32_to_cpu(dx_root->dr_clusters)) { @@ -4546,8 +4476,8 @@ int ocfs2_dx_dir_truncate(struct inode *dir, struct buffer_head *di_bh) p_cpos = ocfs2_blocks_to_clusters(dir->i_sb, blkno); - ret = ocfs2_remove_btree_range(dir, &et, cpos, p_cpos, clen, - &dealloc); + ret = ocfs2_remove_btree_range(dir, &et, cpos, p_cpos, clen, 0, + &dealloc, 0); if (ret) { mlog_errno(ret); goto out; @@ -4566,7 +4496,7 @@ remove_index: goto out; } - ocfs2_remove_from_cache(dir, dx_root_bh); + ocfs2_remove_from_cache(INODE_CACHE(dir), dx_root_bh); out: ocfs2_schedule_truncate_log_flush(osb, 1); ocfs2_run_deallocs(osb, &dealloc); diff --git a/fs/ocfs2/dir.h b/fs/ocfs2/dir.h index e683f3deb64..f0344b75b14 100644 --- a/fs/ocfs2/dir.h +++ b/fs/ocfs2/dir.h @@ -92,9 +92,8 @@ int ocfs2_find_files_on_disk(const char *name, struct ocfs2_dir_lookup_result *res); int ocfs2_lookup_ino_from_name(struct inode *dir, const char *name, int namelen, u64 *blkno); -int ocfs2_readdir(struct file *filp, void *dirent, filldir_t filldir); -int ocfs2_dir_foreach(struct inode *inode, loff_t *f_pos, void *priv, - filldir_t filldir); +int ocfs2_readdir(struct file *file, struct dir_context *ctx); +int ocfs2_dir_foreach(struct inode *inode, struct dir_context *ctx); int ocfs2_prepare_dir_for_insert(struct ocfs2_super *osb, struct inode *dir, struct buffer_head *parent_fe_bh, diff --git a/fs/ocfs2/dlm/Makefile b/fs/ocfs2/dlm/Makefile index 19036137570..bd1aab1f49a 100644 --- a/fs/ocfs2/dlm/Makefile +++ b/fs/ocfs2/dlm/Makefile @@ -1,8 +1,7 @@ -EXTRA_CFLAGS += -Ifs/ocfs2 +ccflags-y := -Ifs/ocfs2 -obj-$(CONFIG_OCFS2_FS_O2CB) += ocfs2_dlm.o ocfs2_dlmfs.o +obj-$(CONFIG_OCFS2_FS_O2CB) += ocfs2_dlm.o ocfs2_dlm-objs := dlmdomain.o dlmdebug.o dlmthread.o dlmrecovery.o \ - dlmmaster.o dlmast.o dlmconvert.o dlmlock.o dlmunlock.o dlmver.o + dlmmaster.o dlmast.o dlmconvert.o dlmlock.o dlmunlock.o -ocfs2_dlmfs-objs := userdlm.o dlmfs.o dlmfsver.o diff --git a/fs/ocfs2/dlm/dlmapi.h b/fs/ocfs2/dlm/dlmapi.h index b5786a787fa..3cfa114aa39 100644 --- a/fs/ocfs2/dlm/dlmapi.h +++ b/fs/ocfs2/dlm/dlmapi.h @@ -95,7 +95,7 @@ const char *dlm_errname(enum dlm_status err); mlog(ML_ERROR, "dlm status = %s\n", dlm_errname((st))); \ } while (0) -#define DLM_LKSB_UNUSED1 0x01 +#define DLM_LKSB_UNUSED1 0x01 #define DLM_LKSB_PUT_LVB 0x02 #define DLM_LKSB_GET_LVB 0x04 #define DLM_LKSB_UNUSED2 0x08 diff --git a/fs/ocfs2/dlm/dlmast.c b/fs/ocfs2/dlm/dlmast.c index d07ddbe4b28..b46278f9ae4 100644 --- a/fs/ocfs2/dlm/dlmast.c +++ b/fs/ocfs2/dlm/dlmast.c @@ -28,9 +28,7 @@ #include <linux/module.h> #include <linux/fs.h> #include <linux/types.h> -#include <linux/slab.h> #include <linux/highmem.h> -#include <linux/utsname.h> #include <linux/init.h> #include <linux/sysctl.h> #include <linux/random.h> @@ -90,22 +88,31 @@ static int dlm_should_cancel_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock) return 0; } -static void __dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock) +void __dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock) { - mlog_entry_void(); + struct dlm_lock_resource *res; BUG_ON(!dlm); BUG_ON(!lock); + res = lock->lockres; + assert_spin_locked(&dlm->ast_lock); + if (!list_empty(&lock->ast_list)) { - mlog(ML_ERROR, "ast list not empty!! pending=%d, newlevel=%d\n", + mlog(ML_ERROR, "%s: res %.*s, lock %u:%llu, " + "AST list not empty, pending %d, newlevel %d\n", + dlm->name, res->lockname.len, res->lockname.name, + dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)), + dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)), lock->ast_pending, lock->ml.type); BUG(); } - BUG_ON(!list_empty(&lock->ast_list)); if (lock->ast_pending) - mlog(0, "lock has an ast getting flushed right now\n"); + mlog(0, "%s: res %.*s, lock %u:%llu, AST getting flushed\n", + dlm->name, res->lockname.len, res->lockname.name, + dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)), + dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie))); /* putting lock on list, add a ref */ dlm_lock_get(lock); @@ -113,9 +120,10 @@ static void __dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock) /* check to see if this ast obsoletes the bast */ if (dlm_should_cancel_bast(dlm, lock)) { - struct dlm_lock_resource *res = lock->lockres; - mlog(0, "%s: cancelling bast for %.*s\n", - dlm->name, res->lockname.len, res->lockname.name); + mlog(0, "%s: res %.*s, lock %u:%llu, Cancelling BAST\n", + dlm->name, res->lockname.len, res->lockname.name, + dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)), + dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie))); lock->bast_pending = 0; list_del_init(&lock->bast_list); lock->ml.highest_blocked = LKM_IVMODE; @@ -125,7 +133,7 @@ static void __dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock) dlm_lock_put(lock); /* free up the reserved bast that we are cancelling. * guaranteed that this will not be the last reserved - * ast because *both* an ast and a bast were reserved + * ast because *both* an ast and a bast were reserved * to get to this point. the res->spinlock will not be * taken here */ dlm_lockres_release_ast(dlm, res); @@ -137,8 +145,6 @@ static void __dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock) void dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock) { - mlog_entry_void(); - BUG_ON(!dlm); BUG_ON(!lock); @@ -148,17 +154,23 @@ void dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock) } -static void __dlm_queue_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock) +void __dlm_queue_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock) { - mlog_entry_void(); + struct dlm_lock_resource *res; BUG_ON(!dlm); BUG_ON(!lock); + assert_spin_locked(&dlm->ast_lock); + res = lock->lockres; + BUG_ON(!list_empty(&lock->bast_list)); if (lock->bast_pending) - mlog(0, "lock has a bast getting flushed right now\n"); + mlog(0, "%s: res %.*s, lock %u:%llu, BAST getting flushed\n", + dlm->name, res->lockname.len, res->lockname.name, + dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)), + dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie))); /* putting lock on list, add a ref */ dlm_lock_get(lock); @@ -170,8 +182,6 @@ static void __dlm_queue_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock) void dlm_queue_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock) { - mlog_entry_void(); - BUG_ON(!dlm); BUG_ON(!lock); @@ -187,9 +197,8 @@ static void dlm_update_lvb(struct dlm_ctxt *dlm, struct dlm_lock_resource *res, BUG_ON(!lksb); /* only updates if this node masters the lockres */ + spin_lock(&res->spinlock); if (res->owner == dlm->node_num) { - - spin_lock(&res->spinlock); /* check the lksb flags for the direction */ if (lksb->flags & DLM_LKSB_GET_LVB) { mlog(0, "getting lvb from lockres for %s node\n", @@ -204,8 +213,8 @@ static void dlm_update_lvb(struct dlm_ctxt *dlm, struct dlm_lock_resource *res, * here. In the future we might want to clear it at the time * the put is actually done. */ - spin_unlock(&res->spinlock); } + spin_unlock(&res->spinlock); /* reset any lvb flags on the lksb */ lksb->flags &= ~(DLM_LKSB_PUT_LVB|DLM_LKSB_GET_LVB); @@ -217,7 +226,10 @@ void dlm_do_local_ast(struct dlm_ctxt *dlm, struct dlm_lock_resource *res, dlm_astlockfunc_t *fn; struct dlm_lockstatus *lksb; - mlog_entry_void(); + mlog(0, "%s: res %.*s, lock %u:%llu, Local AST\n", dlm->name, + res->lockname.len, res->lockname.name, + dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)), + dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie))); lksb = lock->lksb; fn = lock->ast; @@ -235,7 +247,10 @@ int dlm_do_remote_ast(struct dlm_ctxt *dlm, struct dlm_lock_resource *res, struct dlm_lockstatus *lksb; int lksbflags; - mlog_entry_void(); + mlog(0, "%s: res %.*s, lock %u:%llu, Remote AST\n", dlm->name, + res->lockname.len, res->lockname.name, + dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)), + dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie))); lksb = lock->lksb; BUG_ON(lock->ml.node == dlm->node_num); @@ -254,9 +269,14 @@ void dlm_do_local_bast(struct dlm_ctxt *dlm, struct dlm_lock_resource *res, { dlm_bastlockfunc_t *fn = lock->bast; - mlog_entry_void(); BUG_ON(lock->ml.node != dlm->node_num); + mlog(0, "%s: res %.*s, lock %u:%llu, Local BAST, blocked %d\n", + dlm->name, res->lockname.len, res->lockname.name, + dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)), + dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)), + blocked_type); + (*fn)(lock->astdata, blocked_type); } @@ -272,8 +292,8 @@ int dlm_proxy_ast_handler(struct o2net_msg *msg, u32 len, void *data, struct dlm_lock *lock = NULL; struct dlm_proxy_ast *past = (struct dlm_proxy_ast *) msg->buf; char *name; - struct list_head *iter, *head=NULL; - u64 cookie; + struct list_head *head = NULL; + __be64 cookie; u32 flags; u8 node; @@ -336,7 +356,8 @@ int dlm_proxy_ast_handler(struct o2net_msg *msg, u32 len, void *data, /* cannot get a proxy ast message if this node owns it */ BUG_ON(res->owner == dlm->node_num); - mlog(0, "lockres %.*s\n", res->lockname.len, res->lockname.name); + mlog(0, "%s: res %.*s\n", dlm->name, res->lockname.len, + res->lockname.name); spin_lock(&res->spinlock); if (res->state & DLM_LOCK_RES_RECOVERING) { @@ -352,8 +373,7 @@ int dlm_proxy_ast_handler(struct o2net_msg *msg, u32 len, void *data, /* try convert queue for both ast/bast */ head = &res->converting; lock = NULL; - list_for_each(iter, head) { - lock = list_entry (iter, struct dlm_lock, list); + list_for_each_entry(lock, head, list) { if (lock->ml.cookie == cookie) goto do_ast; } @@ -364,8 +384,7 @@ int dlm_proxy_ast_handler(struct o2net_msg *msg, u32 len, void *data, else head = &res->granted; - list_for_each(iter, head) { - lock = list_entry (iter, struct dlm_lock, list); + list_for_each_entry(lock, head, list) { if (lock->ml.cookie == cookie) goto do_ast; } @@ -386,8 +405,12 @@ do_ast: if (past->type == DLM_AST) { /* do not alter lock refcount. switching lists. */ list_move_tail(&lock->list, &res->granted); - mlog(0, "ast: Adding to granted list... type=%d, " - "convert_type=%d\n", lock->ml.type, lock->ml.convert_type); + mlog(0, "%s: res %.*s, lock %u:%llu, Granted type %d => %d\n", + dlm->name, res->lockname.len, res->lockname.name, + dlm_get_lock_cookie_node(be64_to_cpu(cookie)), + dlm_get_lock_cookie_seq(be64_to_cpu(cookie)), + lock->ml.type, lock->ml.convert_type); + if (lock->ml.convert_type != LKM_IVMODE) { lock->ml.type = lock->ml.convert_type; lock->ml.convert_type = LKM_IVMODE; @@ -430,9 +453,9 @@ int dlm_send_proxy_ast_msg(struct dlm_ctxt *dlm, struct dlm_lock_resource *res, size_t veclen = 1; int status; - mlog_entry("res %.*s, to=%u, type=%d, blocked_type=%d\n", - res->lockname.len, res->lockname.name, lock->ml.node, - msg_type, blocked_type); + mlog(0, "%s: res %.*s, to %u, type %d, blocked_type %d\n", dlm->name, + res->lockname.len, res->lockname.name, lock->ml.node, msg_type, + blocked_type); memset(&past, 0, sizeof(struct dlm_proxy_ast)); past.node_idx = dlm->node_num; @@ -445,7 +468,6 @@ int dlm_send_proxy_ast_msg(struct dlm_ctxt *dlm, struct dlm_lock_resource *res, vec[0].iov_len = sizeof(struct dlm_proxy_ast); vec[0].iov_base = &past; if (flags & DLM_LKSB_GET_LVB) { - mlog(0, "returning requested LVB data\n"); be32_add_cpu(&past.flags, LKM_GET_LVB); vec[1].iov_len = DLM_LVB_LEN; vec[1].iov_base = lock->lksb->lvb; @@ -455,7 +477,9 @@ int dlm_send_proxy_ast_msg(struct dlm_ctxt *dlm, struct dlm_lock_resource *res, ret = o2net_send_message_vec(DLM_PROXY_AST_MSG, dlm->key, vec, veclen, lock->ml.node, &status); if (ret < 0) - mlog_errno(ret); + mlog(ML_ERROR, "%s: res %.*s, error %d send AST to node %u\n", + dlm->name, res->lockname.len, res->lockname.name, ret, + lock->ml.node); else { if (status == DLM_RECOVERING) { mlog(ML_ERROR, "sent AST to node %u, it thinks this " diff --git a/fs/ocfs2/dlm/dlmcommon.h b/fs/ocfs2/dlm/dlmcommon.h index 0102be35980..fae17c640df 100644 --- a/fs/ocfs2/dlm/dlmcommon.h +++ b/fs/ocfs2/dlm/dlmcommon.h @@ -37,7 +37,7 @@ #define DLM_THREAD_SHUFFLE_INTERVAL 5 // flush everything every 5 passes #define DLM_THREAD_MS 200 // flush at least every 200 ms -#define DLM_HASH_SIZE_DEFAULT (1 << 14) +#define DLM_HASH_SIZE_DEFAULT (1 << 17) #if DLM_HASH_SIZE_DEFAULT < PAGE_SIZE # define DLM_HASH_PAGES 1 #else @@ -50,10 +50,10 @@ #define dlm_lockid_hash(_n, _l) full_name_hash(_n, _l) enum dlm_mle_type { - DLM_MLE_BLOCK, - DLM_MLE_MASTER, - DLM_MLE_MIGRATION, - DLM_MLE_NUM_TYPES + DLM_MLE_BLOCK = 0, + DLM_MLE_MASTER = 1, + DLM_MLE_MIGRATION = 2, + DLM_MLE_NUM_TYPES = 3, }; struct dlm_master_list_entry { @@ -82,8 +82,8 @@ struct dlm_master_list_entry { enum dlm_ast_type { DLM_AST = 0, - DLM_BAST, - DLM_ASTUNLOCK + DLM_BAST = 1, + DLM_ASTUNLOCK = 2, }; @@ -108,7 +108,6 @@ static inline int dlm_is_recovery_lock(const char *lock_name, int name_len) struct dlm_recovery_ctxt { struct list_head resources; - struct list_head received; struct list_head node_data; u8 new_master; u8 dead_node; @@ -119,9 +118,9 @@ struct dlm_recovery_ctxt enum dlm_ctxt_state { DLM_CTXT_NEW = 0, - DLM_CTXT_JOINED, - DLM_CTXT_IN_SHUTDOWN, - DLM_CTXT_LEAVING, + DLM_CTXT_JOINED = 1, + DLM_CTXT_IN_SHUTDOWN = 2, + DLM_CTXT_LEAVING = 3, }; struct dlm_ctxt @@ -144,6 +143,7 @@ struct dlm_ctxt wait_queue_head_t dlm_join_events; unsigned long live_nodes_map[BITS_TO_LONGS(O2NM_MAX_NODES)]; unsigned long domain_map[BITS_TO_LONGS(O2NM_MAX_NODES)]; + unsigned long exit_domain_map[BITS_TO_LONGS(O2NM_MAX_NODES)]; unsigned long recovery_map[BITS_TO_LONGS(O2NM_MAX_NODES)]; struct dlm_recovery_ctxt reco; spinlock_t master_lock; @@ -331,6 +331,7 @@ struct dlm_lock_resource u16 state; char lvb[DLM_LVB_LEN]; unsigned int inflight_locks; + unsigned int inflight_assert_workers; unsigned long refmap[BITS_TO_LONGS(O2NM_MAX_NODES)]; }; @@ -388,8 +389,8 @@ struct dlm_lock enum dlm_lockres_list { DLM_GRANTED_LIST = 0, - DLM_CONVERTING_LIST, - DLM_BLOCKED_LIST + DLM_CONVERTING_LIST = 1, + DLM_BLOCKED_LIST = 2, }; static inline int dlm_lvb_is_empty(char *lvb) @@ -401,6 +402,18 @@ static inline int dlm_lvb_is_empty(char *lvb) return 1; } +static inline char *dlm_list_in_text(enum dlm_lockres_list idx) +{ + if (idx == DLM_GRANTED_LIST) + return "granted"; + else if (idx == DLM_CONVERTING_LIST) + return "converting"; + else if (idx == DLM_BLOCKED_LIST) + return "blocked"; + else + return "unknown"; +} + static inline struct list_head * dlm_list_idx_to_ptr(struct dlm_lock_resource *res, enum dlm_lockres_list idx) { @@ -427,25 +440,28 @@ struct dlm_node_iter enum { - DLM_MASTER_REQUEST_MSG = 500, - DLM_UNUSED_MSG1, /* 501 */ - DLM_ASSERT_MASTER_MSG, /* 502 */ - DLM_CREATE_LOCK_MSG, /* 503 */ - DLM_CONVERT_LOCK_MSG, /* 504 */ - DLM_PROXY_AST_MSG, /* 505 */ - DLM_UNLOCK_LOCK_MSG, /* 506 */ - DLM_DEREF_LOCKRES_MSG, /* 507 */ - DLM_MIGRATE_REQUEST_MSG, /* 508 */ - DLM_MIG_LOCKRES_MSG, /* 509 */ - DLM_QUERY_JOIN_MSG, /* 510 */ - DLM_ASSERT_JOINED_MSG, /* 511 */ - DLM_CANCEL_JOIN_MSG, /* 512 */ - DLM_EXIT_DOMAIN_MSG, /* 513 */ - DLM_MASTER_REQUERY_MSG, /* 514 */ - DLM_LOCK_REQUEST_MSG, /* 515 */ - DLM_RECO_DATA_DONE_MSG, /* 516 */ - DLM_BEGIN_RECO_MSG, /* 517 */ - DLM_FINALIZE_RECO_MSG /* 518 */ + DLM_MASTER_REQUEST_MSG = 500, + DLM_UNUSED_MSG1 = 501, + DLM_ASSERT_MASTER_MSG = 502, + DLM_CREATE_LOCK_MSG = 503, + DLM_CONVERT_LOCK_MSG = 504, + DLM_PROXY_AST_MSG = 505, + DLM_UNLOCK_LOCK_MSG = 506, + DLM_DEREF_LOCKRES_MSG = 507, + DLM_MIGRATE_REQUEST_MSG = 508, + DLM_MIG_LOCKRES_MSG = 509, + DLM_QUERY_JOIN_MSG = 510, + DLM_ASSERT_JOINED_MSG = 511, + DLM_CANCEL_JOIN_MSG = 512, + DLM_EXIT_DOMAIN_MSG = 513, + DLM_MASTER_REQUERY_MSG = 514, + DLM_LOCK_REQUEST_MSG = 515, + DLM_RECO_DATA_DONE_MSG = 516, + DLM_BEGIN_RECO_MSG = 517, + DLM_FINALIZE_RECO_MSG = 518, + DLM_QUERY_REGION = 519, + DLM_QUERY_NODEINFO = 520, + DLM_BEGIN_EXIT_DOMAIN_MSG = 521, }; struct dlm_reco_node_data @@ -458,19 +474,19 @@ struct dlm_reco_node_data enum { DLM_RECO_NODE_DATA_DEAD = -1, DLM_RECO_NODE_DATA_INIT = 0, - DLM_RECO_NODE_DATA_REQUESTING, - DLM_RECO_NODE_DATA_REQUESTED, - DLM_RECO_NODE_DATA_RECEIVING, - DLM_RECO_NODE_DATA_DONE, - DLM_RECO_NODE_DATA_FINALIZE_SENT, + DLM_RECO_NODE_DATA_REQUESTING = 1, + DLM_RECO_NODE_DATA_REQUESTED = 2, + DLM_RECO_NODE_DATA_RECEIVING = 3, + DLM_RECO_NODE_DATA_DONE = 4, + DLM_RECO_NODE_DATA_FINALIZE_SENT = 5, }; enum { DLM_MASTER_RESP_NO = 0, - DLM_MASTER_RESP_YES, - DLM_MASTER_RESP_MAYBE, - DLM_MASTER_RESP_ERROR + DLM_MASTER_RESP_YES = 1, + DLM_MASTER_RESP_MAYBE = 2, + DLM_MASTER_RESP_ERROR = 3, }; @@ -647,9 +663,9 @@ struct dlm_proxy_ast #define DLM_MOD_KEY (0x666c6172) enum dlm_query_join_response_code { JOIN_DISALLOW = 0, - JOIN_OK, - JOIN_OK_NO_MAP, - JOIN_PROTOCOL_MISMATCH, + JOIN_OK = 1, + JOIN_OK_NO_MAP = 2, + JOIN_PROTOCOL_MISMATCH = 3, }; struct dlm_query_join_packet { @@ -663,7 +679,7 @@ struct dlm_query_join_packet { }; union dlm_query_join_response { - u32 intval; + __be32 intval; struct dlm_query_join_packet packet; }; @@ -727,6 +743,31 @@ struct dlm_cancel_join u8 domain[O2NM_MAX_NAME_LEN]; }; +struct dlm_query_region { + u8 qr_node; + u8 qr_numregions; + u8 qr_namelen; + u8 pad1; + u8 qr_domain[O2NM_MAX_NAME_LEN]; + u8 qr_regions[O2HB_MAX_REGION_NAME_LEN * O2NM_MAX_REGIONS]; +}; + +struct dlm_node_info { + u8 ni_nodenum; + u8 pad1; + __be16 ni_ipv4_port; + __be32 ni_ipv4_address; +}; + +struct dlm_query_nodeinfo { + u8 qn_nodenum; + u8 qn_numnodes; + u8 qn_namelen; + u8 pad1; + u8 qn_domain[O2NM_MAX_NAME_LEN]; + struct dlm_node_info qn_nodes[O2NM_MAX_NODES]; +}; + struct dlm_exit_domain { u8 node_idx; @@ -818,8 +859,8 @@ void dlm_complete_recovery_thread(struct dlm_ctxt *dlm); void dlm_wait_for_recovery(struct dlm_ctxt *dlm); void dlm_kick_recovery_thread(struct dlm_ctxt *dlm); int dlm_is_node_dead(struct dlm_ctxt *dlm, u8 node); -int dlm_wait_for_node_death(struct dlm_ctxt *dlm, u8 node, int timeout); -int dlm_wait_for_node_recovery(struct dlm_ctxt *dlm, u8 node, int timeout); +void dlm_wait_for_node_death(struct dlm_ctxt *dlm, u8 node, int timeout); +void dlm_wait_for_node_recovery(struct dlm_ctxt *dlm, u8 node, int timeout); void dlm_put(struct dlm_ctxt *dlm); struct dlm_ctxt *dlm_grab(struct dlm_ctxt *dlm); @@ -836,9 +877,8 @@ static inline void dlm_lockres_get(struct dlm_lock_resource *res) kref_get(&res->refs); } void dlm_lockres_put(struct dlm_lock_resource *res); -void __dlm_unhash_lockres(struct dlm_lock_resource *res); -void __dlm_insert_lockres(struct dlm_ctxt *dlm, - struct dlm_lock_resource *res); +void __dlm_unhash_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res); +void __dlm_insert_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res); struct dlm_lock_resource * __dlm_lookup_lockres_full(struct dlm_ctxt *dlm, const char *name, unsigned int len, @@ -861,49 +901,23 @@ struct dlm_lock_resource *dlm_new_lockres(struct dlm_ctxt *dlm, const char *name, unsigned int namelen); -#define dlm_lockres_set_refmap_bit(bit,res) \ - __dlm_lockres_set_refmap_bit(bit,res,__FILE__,__LINE__) -#define dlm_lockres_clear_refmap_bit(bit,res) \ - __dlm_lockres_clear_refmap_bit(bit,res,__FILE__,__LINE__) +void dlm_lockres_set_refmap_bit(struct dlm_ctxt *dlm, + struct dlm_lock_resource *res, int bit); +void dlm_lockres_clear_refmap_bit(struct dlm_ctxt *dlm, + struct dlm_lock_resource *res, int bit); -static inline void __dlm_lockres_set_refmap_bit(int bit, - struct dlm_lock_resource *res, - const char *file, - int line) -{ - //printk("%s:%d:%.*s: setting bit %d\n", file, line, - // res->lockname.len, res->lockname.name, bit); - set_bit(bit, res->refmap); -} - -static inline void __dlm_lockres_clear_refmap_bit(int bit, - struct dlm_lock_resource *res, - const char *file, - int line) -{ - //printk("%s:%d:%.*s: clearing bit %d\n", file, line, - // res->lockname.len, res->lockname.name, bit); - clear_bit(bit, res->refmap); -} +void dlm_lockres_drop_inflight_ref(struct dlm_ctxt *dlm, + struct dlm_lock_resource *res); +void dlm_lockres_grab_inflight_ref(struct dlm_ctxt *dlm, + struct dlm_lock_resource *res); -void __dlm_lockres_drop_inflight_ref(struct dlm_ctxt *dlm, - struct dlm_lock_resource *res, - const char *file, - int line); -void __dlm_lockres_grab_inflight_ref(struct dlm_ctxt *dlm, - struct dlm_lock_resource *res, - int new_lockres, - const char *file, - int line); -#define dlm_lockres_drop_inflight_ref(d,r) \ - __dlm_lockres_drop_inflight_ref(d,r,__FILE__,__LINE__) -#define dlm_lockres_grab_inflight_ref(d,r) \ - __dlm_lockres_grab_inflight_ref(d,r,0,__FILE__,__LINE__) -#define dlm_lockres_grab_inflight_ref_new(d,r) \ - __dlm_lockres_grab_inflight_ref(d,r,1,__FILE__,__LINE__) +void __dlm_lockres_grab_inflight_worker(struct dlm_ctxt *dlm, + struct dlm_lock_resource *res); void dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock); void dlm_queue_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock); +void __dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock); +void __dlm_queue_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock); void dlm_do_local_ast(struct dlm_ctxt *dlm, struct dlm_lock_resource *res, struct dlm_lock *lock); @@ -1028,6 +1042,7 @@ int dlm_drop_lockres_ref(struct dlm_ctxt *dlm, struct dlm_lock_resource *res); void dlm_clean_master_list(struct dlm_ctxt *dlm, u8 dead_node); +void dlm_force_free_mles(struct dlm_ctxt *dlm); int dlm_lock_basts_flushed(struct dlm_ctxt *dlm, struct dlm_lock *lock); int __dlm_lockres_has_locks(struct dlm_lock_resource *res); int __dlm_lockres_unused(struct dlm_lock_resource *res); @@ -1067,11 +1082,9 @@ static inline int dlm_lock_compatible(int existing, int request) static inline int dlm_lock_on_list(struct list_head *head, struct dlm_lock *lock) { - struct list_head *iter; struct dlm_lock *tmplock; - list_for_each(iter, head) { - tmplock = list_entry(iter, struct dlm_lock, list); + list_for_each_entry(tmplock, head, list) { if (tmplock == lock) return 1; } diff --git a/fs/ocfs2/dlm/dlmconvert.c b/fs/ocfs2/dlm/dlmconvert.c index 75997b4deaf..e36d63ff178 100644 --- a/fs/ocfs2/dlm/dlmconvert.c +++ b/fs/ocfs2/dlm/dlmconvert.c @@ -28,9 +28,7 @@ #include <linux/module.h> #include <linux/fs.h> #include <linux/types.h> -#include <linux/slab.h> #include <linux/highmem.h> -#include <linux/utsname.h> #include <linux/init.h> #include <linux/sysctl.h> #include <linux/random.h> @@ -125,13 +123,12 @@ static enum dlm_status __dlmconvert_master(struct dlm_ctxt *dlm, int *kick_thread) { enum dlm_status status = DLM_NORMAL; - struct list_head *iter; struct dlm_lock *tmplock=NULL; assert_spin_locked(&res->spinlock); - mlog_entry("type=%d, convert_type=%d, new convert_type=%d\n", - lock->ml.type, lock->ml.convert_type, type); + mlog(0, "type=%d, convert_type=%d, new convert_type=%d\n", + lock->ml.type, lock->ml.convert_type, type); spin_lock(&lock->spinlock); @@ -187,16 +184,14 @@ static enum dlm_status __dlmconvert_master(struct dlm_ctxt *dlm, /* upconvert from here on */ status = DLM_NORMAL; - list_for_each(iter, &res->granted) { - tmplock = list_entry(iter, struct dlm_lock, list); + list_for_each_entry(tmplock, &res->granted, list) { if (tmplock == lock) continue; if (!dlm_lock_compatible(tmplock->ml.type, type)) goto switch_queues; } - list_for_each(iter, &res->converting) { - tmplock = list_entry(iter, struct dlm_lock, list); + list_for_each_entry(tmplock, &res->converting, list) { if (!dlm_lock_compatible(tmplock->ml.type, type)) goto switch_queues; /* existing conversion requests take precedence */ @@ -355,7 +350,7 @@ static enum dlm_status dlm_send_remote_convert_request(struct dlm_ctxt *dlm, struct kvec vec[2]; size_t veclen = 1; - mlog_entry("%.*s\n", res->lockname.len, res->lockname.name); + mlog(0, "%.*s\n", res->lockname.len, res->lockname.name); memset(&convert, 0, sizeof(struct dlm_convert_lock)); convert.node_idx = dlm->node_num; @@ -392,12 +387,14 @@ static enum dlm_status dlm_send_remote_convert_request(struct dlm_ctxt *dlm, } else if (ret != DLM_NORMAL && ret != DLM_NOTQUEUED) dlm_error(ret); } else { - mlog_errno(tmpret); + mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to " + "node %u\n", tmpret, DLM_CONVERT_LOCK_MSG, dlm->key, + res->owner); if (dlm_is_host_down(tmpret)) { /* instead of logging the same network error over * and over, sleep here and wait for the heartbeat * to notice the node is dead. times out after 5s. */ - dlm_wait_for_node_death(dlm, res->owner, + dlm_wait_for_node_death(dlm, res->owner, DLM_NODE_DEATH_WAIT_MAX); ret = DLM_RECOVERING; mlog(0, "node %u died so returning DLM_RECOVERING " @@ -424,8 +421,8 @@ int dlm_convert_lock_handler(struct o2net_msg *msg, u32 len, void *data, struct dlm_ctxt *dlm = data; struct dlm_convert_lock *cnv = (struct dlm_convert_lock *)msg->buf; struct dlm_lock_resource *res = NULL; - struct list_head *iter; struct dlm_lock *lock = NULL; + struct dlm_lock *tmp_lock; struct dlm_lockstatus *lksb; enum dlm_status status = DLM_NORMAL; u32 flags; @@ -471,14 +468,13 @@ int dlm_convert_lock_handler(struct o2net_msg *msg, u32 len, void *data, dlm_error(status); goto leave; } - list_for_each(iter, &res->granted) { - lock = list_entry(iter, struct dlm_lock, list); - if (lock->ml.cookie == cnv->cookie && - lock->ml.node == cnv->node_idx) { + list_for_each_entry(tmp_lock, &res->granted, list) { + if (tmp_lock->ml.cookie == cnv->cookie && + tmp_lock->ml.node == cnv->node_idx) { + lock = tmp_lock; dlm_lock_get(lock); break; } - lock = NULL; } spin_unlock(&res->spinlock); if (!lock) { diff --git a/fs/ocfs2/dlm/dlmdebug.c b/fs/ocfs2/dlm/dlmdebug.c index df52f706f66..18f13c2e4a1 100644 --- a/fs/ocfs2/dlm/dlmdebug.c +++ b/fs/ocfs2/dlm/dlmdebug.c @@ -27,10 +27,10 @@ #include <linux/types.h> #include <linux/slab.h> #include <linux/highmem.h> -#include <linux/utsname.h> #include <linux/sysctl.h> #include <linux/spinlock.h> #include <linux/debugfs.h> +#include <linux/export.h> #include "cluster/heartbeat.h" #include "cluster/nodemanager.h" @@ -96,14 +96,13 @@ static void __dlm_print_lock(struct dlm_lock *lock) void __dlm_print_one_lock_resource(struct dlm_lock_resource *res) { - struct list_head *iter2; struct dlm_lock *lock; char buf[DLM_LOCKID_NAME_MAX]; assert_spin_locked(&res->spinlock); stringify_lockname(res->lockname.name, res->lockname.len, - buf, sizeof(buf) - 1); + buf, sizeof(buf)); printk("lockres: %s, owner=%u, state=%u\n", buf, res->owner, res->state); printk(" last used: %lu, refcnt: %u, on purge list: %s\n", @@ -118,18 +117,15 @@ void __dlm_print_one_lock_resource(struct dlm_lock_resource *res) res->inflight_locks, atomic_read(&res->asts_reserved)); dlm_print_lockres_refmap(res); printk(" granted queue:\n"); - list_for_each(iter2, &res->granted) { - lock = list_entry(iter2, struct dlm_lock, list); + list_for_each_entry(lock, &res->granted, list) { __dlm_print_lock(lock); } printk(" converting queue:\n"); - list_for_each(iter2, &res->converting) { - lock = list_entry(iter2, struct dlm_lock, list); + list_for_each_entry(lock, &res->converting, list) { __dlm_print_lock(lock); } printk(" blocked queue:\n"); - list_for_each(iter2, &res->blocked) { - lock = list_entry(iter2, struct dlm_lock, list); + list_for_each_entry(lock, &res->blocked, list) { __dlm_print_lock(lock); } } @@ -342,7 +338,7 @@ void dlm_print_one_mle(struct dlm_master_list_entry *mle) #ifdef CONFIG_DEBUG_FS -static struct dentry *dlm_debugfs_root = NULL; +static struct dentry *dlm_debugfs_root; #define DLM_DEBUGFS_DIR "o2dlm" #define DLM_DEBUGFS_DLM_STATE "dlm_state" @@ -371,92 +367,46 @@ static void dlm_debug_get(struct dlm_debug_ctxt *dc) kref_get(&dc->debug_refcnt); } -static struct debug_buffer *debug_buffer_allocate(void) +static int debug_release(struct inode *inode, struct file *file) { - struct debug_buffer *db = NULL; - - db = kzalloc(sizeof(struct debug_buffer), GFP_KERNEL); - if (!db) - goto bail; - - db->len = PAGE_SIZE; - db->buf = kmalloc(db->len, GFP_KERNEL); - if (!db->buf) - goto bail; - - return db; -bail: - kfree(db); - return NULL; -} - -static ssize_t debug_buffer_read(struct file *file, char __user *buf, - size_t nbytes, loff_t *ppos) -{ - struct debug_buffer *db = file->private_data; - - return simple_read_from_buffer(buf, nbytes, ppos, db->buf, db->len); -} - -static loff_t debug_buffer_llseek(struct file *file, loff_t off, int whence) -{ - struct debug_buffer *db = file->private_data; - loff_t new = -1; - - switch (whence) { - case 0: - new = off; - break; - case 1: - new = file->f_pos + off; - break; - } - - if (new < 0 || new > db->len) - return -EINVAL; - - return (file->f_pos = new); + free_page((unsigned long)file->private_data); + return 0; } -static int debug_buffer_release(struct inode *inode, struct file *file) +static ssize_t debug_read(struct file *file, char __user *buf, + size_t nbytes, loff_t *ppos) { - struct debug_buffer *db = (struct debug_buffer *)file->private_data; - - if (db) - kfree(db->buf); - kfree(db); - - return 0; + return simple_read_from_buffer(buf, nbytes, ppos, file->private_data, + i_size_read(file->f_mapping->host)); } /* end - util funcs */ /* begin - purge list funcs */ -static int debug_purgelist_print(struct dlm_ctxt *dlm, struct debug_buffer *db) +static int debug_purgelist_print(struct dlm_ctxt *dlm, char *buf, int len) { struct dlm_lock_resource *res; int out = 0; unsigned long total = 0; - out += snprintf(db->buf + out, db->len - out, + out += snprintf(buf + out, len - out, "Dumping Purgelist for Domain: %s\n", dlm->name); spin_lock(&dlm->spinlock); list_for_each_entry(res, &dlm->purge_list, purge) { ++total; - if (db->len - out < 100) + if (len - out < 100) continue; spin_lock(&res->spinlock); out += stringify_lockname(res->lockname.name, res->lockname.len, - db->buf + out, db->len - out); - out += snprintf(db->buf + out, db->len - out, "\t%ld\n", + buf + out, len - out); + out += snprintf(buf + out, len - out, "\t%ld\n", (jiffies - res->last_used)/HZ); spin_unlock(&res->spinlock); } spin_unlock(&dlm->spinlock); - out += snprintf(db->buf + out, db->len - out, - "Total on list: %ld\n", total); + out += snprintf(buf + out, len - out, "Total on list: %ld\n", total); return out; } @@ -464,59 +414,56 @@ static int debug_purgelist_print(struct dlm_ctxt *dlm, struct debug_buffer *db) static int debug_purgelist_open(struct inode *inode, struct file *file) { struct dlm_ctxt *dlm = inode->i_private; - struct debug_buffer *db; + char *buf = NULL; - db = debug_buffer_allocate(); - if (!db) + buf = (char *) get_zeroed_page(GFP_NOFS); + if (!buf) goto bail; - db->len = debug_purgelist_print(dlm, db); + i_size_write(inode, debug_purgelist_print(dlm, buf, PAGE_SIZE - 1)); - file->private_data = db; + file->private_data = buf; return 0; bail: return -ENOMEM; } -static struct file_operations debug_purgelist_fops = { +static const struct file_operations debug_purgelist_fops = { .open = debug_purgelist_open, - .release = debug_buffer_release, - .read = debug_buffer_read, - .llseek = debug_buffer_llseek, + .release = debug_release, + .read = debug_read, + .llseek = generic_file_llseek, }; /* end - purge list funcs */ /* begin - debug mle funcs */ -static int debug_mle_print(struct dlm_ctxt *dlm, struct debug_buffer *db) +static int debug_mle_print(struct dlm_ctxt *dlm, char *buf, int len) { struct dlm_master_list_entry *mle; struct hlist_head *bucket; - struct hlist_node *list; int i, out = 0; - unsigned long total = 0, longest = 0, bktcnt; + unsigned long total = 0, longest = 0, bucket_count = 0; - out += snprintf(db->buf + out, db->len - out, + out += snprintf(buf + out, len - out, "Dumping MLEs for Domain: %s\n", dlm->name); spin_lock(&dlm->master_lock); for (i = 0; i < DLM_HASH_BUCKETS; i++) { bucket = dlm_master_hash(dlm, i); - hlist_for_each(list, bucket) { - mle = hlist_entry(list, struct dlm_master_list_entry, - master_hash_node); + hlist_for_each_entry(mle, bucket, master_hash_node) { ++total; - ++bktcnt; - if (db->len - out < 200) + ++bucket_count; + if (len - out < 200) continue; - out += dump_mle(mle, db->buf + out, db->len - out); + out += dump_mle(mle, buf + out, len - out); } - longest = max(longest, bktcnt); - bktcnt = 0; + longest = max(longest, bucket_count); + bucket_count = 0; } spin_unlock(&dlm->master_lock); - out += snprintf(db->buf + out, db->len - out, + out += snprintf(buf + out, len - out, "Total: %ld, Longest: %ld\n", total, longest); return out; } @@ -524,26 +471,26 @@ static int debug_mle_print(struct dlm_ctxt *dlm, struct debug_buffer *db) static int debug_mle_open(struct inode *inode, struct file *file) { struct dlm_ctxt *dlm = inode->i_private; - struct debug_buffer *db; + char *buf = NULL; - db = debug_buffer_allocate(); - if (!db) + buf = (char *) get_zeroed_page(GFP_NOFS); + if (!buf) goto bail; - db->len = debug_mle_print(dlm, db); + i_size_write(inode, debug_mle_print(dlm, buf, PAGE_SIZE - 1)); - file->private_data = db; + file->private_data = buf; return 0; bail: return -ENOMEM; } -static struct file_operations debug_mle_fops = { +static const struct file_operations debug_mle_fops = { .open = debug_mle_open, - .release = debug_buffer_release, - .read = debug_buffer_read, - .llseek = debug_buffer_llseek, + .release = debug_release, + .read = debug_read, + .llseek = generic_file_llseek, }; /* end - debug mle funcs */ @@ -637,8 +584,14 @@ static void *lockres_seq_start(struct seq_file *m, loff_t *pos) spin_lock(&dlm->track_lock); if (oldres) track_list = &oldres->tracking; - else + else { track_list = &dlm->tracking_list; + if (list_empty(track_list)) { + dl = NULL; + spin_unlock(&dlm->track_lock); + goto bail; + } + } list_for_each_entry(res, track_list, tracking) { if (&res->tracking == &dlm->tracking_list) @@ -661,6 +614,7 @@ static void *lockres_seq_start(struct seq_file *m, loff_t *pos) } else dl = NULL; +bail: /* passed to seq_show */ return dl; } @@ -683,7 +637,7 @@ static int lockres_seq_show(struct seq_file *s, void *v) return 0; } -static struct seq_operations debug_lockres_ops = { +static const struct seq_operations debug_lockres_ops = { .start = lockres_seq_start, .stop = lockres_seq_stop, .next = lockres_seq_next, @@ -716,7 +670,7 @@ static int debug_lockres_open(struct inode *inode, struct file *file) goto bail; } - seq = (struct seq_file *) file->private_data; + seq = file->private_data; seq->private = dl; dlm_grab(dlm); @@ -732,7 +686,7 @@ bail: static int debug_lockres_release(struct inode *inode, struct file *file) { - struct seq_file *seq = (struct seq_file *)file->private_data; + struct seq_file *seq = file->private_data; struct debug_lockres *dl = (struct debug_lockres *)seq->private; if (dl->dl_res) @@ -742,7 +696,7 @@ static int debug_lockres_release(struct inode *inode, struct file *file) return seq_release_private(inode, file); } -static struct file_operations debug_lockres_fops = { +static const struct file_operations debug_lockres_fops = { .open = debug_lockres_open, .release = debug_lockres_release, .read = seq_read, @@ -751,7 +705,7 @@ static struct file_operations debug_lockres_fops = { /* end - debug lockres funcs */ /* begin - debug state funcs */ -static int debug_state_print(struct dlm_ctxt *dlm, struct debug_buffer *db) +static int debug_state_print(struct dlm_ctxt *dlm, char *buf, int len) { int out = 0; struct dlm_reco_node_data *node; @@ -775,33 +729,41 @@ static int debug_state_print(struct dlm_ctxt *dlm, struct debug_buffer *db) } /* Domain: xxxxxxxxxx Key: 0xdfbac769 */ - out += snprintf(db->buf + out, db->len - out, - "Domain: %s Key: 0x%08x\n", dlm->name, dlm->key); + out += snprintf(buf + out, len - out, + "Domain: %s Key: 0x%08x Protocol: %d.%d\n", + dlm->name, dlm->key, dlm->dlm_locking_proto.pv_major, + dlm->dlm_locking_proto.pv_minor); /* Thread Pid: xxx Node: xxx State: xxxxx */ - out += snprintf(db->buf + out, db->len - out, + out += snprintf(buf + out, len - out, "Thread Pid: %d Node: %d State: %s\n", - dlm->dlm_thread_task->pid, dlm->node_num, state); + task_pid_nr(dlm->dlm_thread_task), dlm->node_num, state); /* Number of Joins: xxx Joining Node: xxx */ - out += snprintf(db->buf + out, db->len - out, + out += snprintf(buf + out, len - out, "Number of Joins: %d Joining Node: %d\n", dlm->num_joins, dlm->joining_node); /* Domain Map: xx xx xx */ - out += snprintf(db->buf + out, db->len - out, "Domain Map: "); + out += snprintf(buf + out, len - out, "Domain Map: "); out += stringify_nodemap(dlm->domain_map, O2NM_MAX_NODES, - db->buf + out, db->len - out); - out += snprintf(db->buf + out, db->len - out, "\n"); + buf + out, len - out); + out += snprintf(buf + out, len - out, "\n"); + + /* Exit Domain Map: xx xx xx */ + out += snprintf(buf + out, len - out, "Exit Domain Map: "); + out += stringify_nodemap(dlm->exit_domain_map, O2NM_MAX_NODES, + buf + out, len - out); + out += snprintf(buf + out, len - out, "\n"); /* Live Map: xx xx xx */ - out += snprintf(db->buf + out, db->len - out, "Live Map: "); + out += snprintf(buf + out, len - out, "Live Map: "); out += stringify_nodemap(dlm->live_nodes_map, O2NM_MAX_NODES, - db->buf + out, db->len - out); - out += snprintf(db->buf + out, db->len - out, "\n"); + buf + out, len - out); + out += snprintf(buf + out, len - out, "\n"); /* Lock Resources: xxx (xxx) */ - out += snprintf(db->buf + out, db->len - out, + out += snprintf(buf + out, len - out, "Lock Resources: %d (%d)\n", atomic_read(&dlm->res_cur_count), atomic_read(&dlm->res_tot_count)); @@ -813,29 +775,29 @@ static int debug_state_print(struct dlm_ctxt *dlm, struct debug_buffer *db) cur_mles += atomic_read(&dlm->mle_cur_count[i]); /* MLEs: xxx (xxx) */ - out += snprintf(db->buf + out, db->len - out, + out += snprintf(buf + out, len - out, "MLEs: %d (%d)\n", cur_mles, tot_mles); /* Blocking: xxx (xxx) */ - out += snprintf(db->buf + out, db->len - out, + out += snprintf(buf + out, len - out, " Blocking: %d (%d)\n", atomic_read(&dlm->mle_cur_count[DLM_MLE_BLOCK]), atomic_read(&dlm->mle_tot_count[DLM_MLE_BLOCK])); /* Mastery: xxx (xxx) */ - out += snprintf(db->buf + out, db->len - out, + out += snprintf(buf + out, len - out, " Mastery: %d (%d)\n", atomic_read(&dlm->mle_cur_count[DLM_MLE_MASTER]), atomic_read(&dlm->mle_tot_count[DLM_MLE_MASTER])); /* Migration: xxx (xxx) */ - out += snprintf(db->buf + out, db->len - out, + out += snprintf(buf + out, len - out, " Migration: %d (%d)\n", atomic_read(&dlm->mle_cur_count[DLM_MLE_MIGRATION]), atomic_read(&dlm->mle_tot_count[DLM_MLE_MIGRATION])); /* Lists: Dirty=Empty Purge=InUse PendingASTs=Empty ... */ - out += snprintf(db->buf + out, db->len - out, + out += snprintf(buf + out, len - out, "Lists: Dirty=%s Purge=%s PendingASTs=%s " "PendingBASTs=%s\n", (list_empty(&dlm->dirty_list) ? "Empty" : "InUse"), @@ -844,12 +806,12 @@ static int debug_state_print(struct dlm_ctxt *dlm, struct debug_buffer *db) (list_empty(&dlm->pending_basts) ? "Empty" : "InUse")); /* Purge Count: xxx Refs: xxx */ - out += snprintf(db->buf + out, db->len - out, + out += snprintf(buf + out, len - out, "Purge Count: %d Refs: %d\n", dlm->purge_count, atomic_read(&dlm->dlm_refs.refcount)); /* Dead Node: xxx */ - out += snprintf(db->buf + out, db->len - out, + out += snprintf(buf + out, len - out, "Dead Node: %d\n", dlm->reco.dead_node); /* What about DLM_RECO_STATE_FINALIZE? */ @@ -859,19 +821,19 @@ static int debug_state_print(struct dlm_ctxt *dlm, struct debug_buffer *db) state = "INACTIVE"; /* Recovery Pid: xxxx Master: xxx State: xxxx */ - out += snprintf(db->buf + out, db->len - out, + out += snprintf(buf + out, len - out, "Recovery Pid: %d Master: %d State: %s\n", - dlm->dlm_reco_thread_task->pid, + task_pid_nr(dlm->dlm_reco_thread_task), dlm->reco.new_master, state); /* Recovery Map: xx xx */ - out += snprintf(db->buf + out, db->len - out, "Recovery Map: "); + out += snprintf(buf + out, len - out, "Recovery Map: "); out += stringify_nodemap(dlm->recovery_map, O2NM_MAX_NODES, - db->buf + out, db->len - out); - out += snprintf(db->buf + out, db->len - out, "\n"); + buf + out, len - out); + out += snprintf(buf + out, len - out, "\n"); /* Recovery Node State: */ - out += snprintf(db->buf + out, db->len - out, "Recovery Node State:\n"); + out += snprintf(buf + out, len - out, "Recovery Node State:\n"); list_for_each_entry(node, &dlm->reco.node_data, list) { switch (node->state) { case DLM_RECO_NODE_DATA_INIT: @@ -899,7 +861,7 @@ static int debug_state_print(struct dlm_ctxt *dlm, struct debug_buffer *db) state = "BAD"; break; } - out += snprintf(db->buf + out, db->len - out, "\t%u - %s\n", + out += snprintf(buf + out, len - out, "\t%u - %s\n", node->node_num, state); } @@ -911,26 +873,26 @@ static int debug_state_print(struct dlm_ctxt *dlm, struct debug_buffer *db) static int debug_state_open(struct inode *inode, struct file *file) { struct dlm_ctxt *dlm = inode->i_private; - struct debug_buffer *db = NULL; + char *buf = NULL; - db = debug_buffer_allocate(); - if (!db) + buf = (char *) get_zeroed_page(GFP_NOFS); + if (!buf) goto bail; - db->len = debug_state_print(dlm, db); + i_size_write(inode, debug_state_print(dlm, buf, PAGE_SIZE - 1)); - file->private_data = db; + file->private_data = buf; return 0; bail: return -ENOMEM; } -static struct file_operations debug_state_fops = { +static const struct file_operations debug_state_fops = { .open = debug_state_open, - .release = debug_buffer_release, - .read = debug_buffer_read, - .llseek = debug_buffer_llseek, + .release = debug_release, + .read = debug_read, + .llseek = generic_file_llseek, }; /* end - debug state funcs */ @@ -994,14 +956,10 @@ void dlm_debug_shutdown(struct dlm_ctxt *dlm) struct dlm_debug_ctxt *dc = dlm->dlm_debug_ctxt; if (dc) { - if (dc->debug_purgelist_dentry) - debugfs_remove(dc->debug_purgelist_dentry); - if (dc->debug_mle_dentry) - debugfs_remove(dc->debug_mle_dentry); - if (dc->debug_lockres_dentry) - debugfs_remove(dc->debug_lockres_dentry); - if (dc->debug_state_dentry) - debugfs_remove(dc->debug_state_dentry); + debugfs_remove(dc->debug_purgelist_dentry); + debugfs_remove(dc->debug_mle_dentry); + debugfs_remove(dc->debug_lockres_dentry); + debugfs_remove(dc->debug_state_dentry); dlm_debug_put(dc); } } @@ -1032,8 +990,7 @@ bail: void dlm_destroy_debugfs_subroot(struct dlm_ctxt *dlm) { - if (dlm->dlm_debugfs_subroot) - debugfs_remove(dlm->dlm_debugfs_subroot); + debugfs_remove(dlm->dlm_debugfs_subroot); } /* debugfs root */ @@ -1049,7 +1006,6 @@ int dlm_create_debugfs_root(void) void dlm_destroy_debugfs_root(void) { - if (dlm_debugfs_root) - debugfs_remove(dlm_debugfs_root); + debugfs_remove(dlm_debugfs_root); } #endif /* CONFIG_DEBUG_FS */ diff --git a/fs/ocfs2/dlm/dlmdebug.h b/fs/ocfs2/dlm/dlmdebug.h index 8c686d22f9c..1f27c4812d1 100644 --- a/fs/ocfs2/dlm/dlmdebug.h +++ b/fs/ocfs2/dlm/dlmdebug.h @@ -37,11 +37,6 @@ struct dlm_debug_ctxt { struct dentry *debug_purgelist_dentry; }; -struct debug_buffer { - int len; - char *buf; -}; - struct debug_lockres { int dl_len; char *dl_buf; diff --git a/fs/ocfs2/dlm/dlmdomain.c b/fs/ocfs2/dlm/dlmdomain.c index 4d9e6b288dd..39efc5057a3 100644 --- a/fs/ocfs2/dlm/dlmdomain.c +++ b/fs/ocfs2/dlm/dlmdomain.c @@ -28,7 +28,6 @@ #include <linux/types.h> #include <linux/slab.h> #include <linux/highmem.h> -#include <linux/utsname.h> #include <linux/init.h> #include <linux/spinlock.h> #include <linux/delay.h> @@ -44,8 +43,6 @@ #include "dlmdomain.h" #include "dlmdebug.h" -#include "dlmver.h" - #define MLOG_MASK_PREFIX (ML_DLM|ML_DLM_DOMAIN) #include "cluster/masklog.h" @@ -129,10 +126,16 @@ static DECLARE_WAIT_QUEUE_HEAD(dlm_domain_events); * will have a negotiated version with the same major number and a minor * number equal or smaller. The dlm_ctxt->dlm_locking_proto field should * be used to determine what a running domain is actually using. + * + * New in version 1.1: + * - Message DLM_QUERY_REGION added to support global heartbeat + * - Message DLM_QUERY_NODEINFO added to allow online node removes + * New in version 1.2: + * - Message DLM_BEGIN_EXIT_DOMAIN_MSG added to mark start of exit domain */ static const struct dlm_protocol_version dlm_protocol = { .pv_major = 1, - .pv_minor = 0, + .pv_minor = 2, }; #define DLM_DOMAIN_BACKOFF_MS 200 @@ -143,6 +146,8 @@ static int dlm_assert_joined_handler(struct o2net_msg *msg, u32 len, void *data, void **ret_data); static int dlm_cancel_join_handler(struct o2net_msg *msg, u32 len, void *data, void **ret_data); +static int dlm_query_region_handler(struct o2net_msg *msg, u32 len, + void *data, void **ret_data); static int dlm_exit_domain_handler(struct o2net_msg *msg, u32 len, void *data, void **ret_data); static int dlm_protocol_compare(struct dlm_protocol_version *existing, @@ -150,16 +155,18 @@ static int dlm_protocol_compare(struct dlm_protocol_version *existing, static void dlm_unregister_domain_handlers(struct dlm_ctxt *dlm); -void __dlm_unhash_lockres(struct dlm_lock_resource *lockres) +void __dlm_unhash_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res) { - if (!hlist_unhashed(&lockres->hash_node)) { - hlist_del_init(&lockres->hash_node); - dlm_lockres_put(lockres); - } + if (hlist_unhashed(&res->hash_node)) + return; + + mlog(0, "%s: Unhash res %.*s\n", dlm->name, res->lockname.len, + res->lockname.name); + hlist_del_init(&res->hash_node); + dlm_lockres_put(res); } -void __dlm_insert_lockres(struct dlm_ctxt *dlm, - struct dlm_lock_resource *res) +void __dlm_insert_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res) { struct hlist_head *bucket; struct qstr *q; @@ -173,6 +180,9 @@ void __dlm_insert_lockres(struct dlm_ctxt *dlm, dlm_lockres_get(res); hlist_add_head(&res->hash_node, bucket); + + mlog(0, "%s: Hash res %.*s\n", dlm->name, res->lockname.len, + res->lockname.name); } struct dlm_lock_resource * __dlm_lookup_lockres_full(struct dlm_ctxt *dlm, @@ -181,17 +191,15 @@ struct dlm_lock_resource * __dlm_lookup_lockres_full(struct dlm_ctxt *dlm, unsigned int hash) { struct hlist_head *bucket; - struct hlist_node *list; + struct dlm_lock_resource *res; - mlog_entry("%.*s\n", len, name); + mlog(0, "%.*s\n", len, name); assert_spin_locked(&dlm->spinlock); bucket = dlm_lockres_hash(dlm, hash); - hlist_for_each(list, bucket) { - struct dlm_lock_resource *res = hlist_entry(list, - struct dlm_lock_resource, hash_node); + hlist_for_each_entry(res, bucket, hash_node) { if (res->lockname.name[0] != name[0]) continue; if (unlikely(res->lockname.len != len)) @@ -217,7 +225,7 @@ struct dlm_lock_resource * __dlm_lookup_lockres(struct dlm_ctxt *dlm, { struct dlm_lock_resource *res = NULL; - mlog_entry("%.*s\n", len, name); + mlog(0, "%.*s\n", len, name); assert_spin_locked(&dlm->spinlock); @@ -250,22 +258,19 @@ struct dlm_lock_resource * dlm_lookup_lockres(struct dlm_ctxt *dlm, static struct dlm_ctxt * __dlm_lookup_domain_full(const char *domain, int len) { - struct dlm_ctxt *tmp = NULL; - struct list_head *iter; + struct dlm_ctxt *tmp; assert_spin_locked(&dlm_domain_lock); /* tmp->name here is always NULL terminated, * but domain may not be! */ - list_for_each(iter, &dlm_domains) { - tmp = list_entry (iter, struct dlm_ctxt, list); + list_for_each_entry(tmp, &dlm_domains, list) { if (strlen(tmp->name) == len && memcmp(tmp->name, domain, len)==0) - break; - tmp = NULL; + return tmp; } - return tmp; + return NULL; } /* For null terminated domain strings ONLY */ @@ -307,9 +312,7 @@ static void dlm_free_ctxt_mem(struct dlm_ctxt *dlm) if (dlm->master_hash) dlm_free_pagevec((void **)dlm->master_hash, DLM_HASH_PAGES); - if (dlm->name) - kfree(dlm->name); - + kfree(dlm->name); kfree(dlm); } @@ -356,25 +359,22 @@ static void __dlm_get(struct dlm_ctxt *dlm) * you shouldn't trust your pointer. */ struct dlm_ctxt *dlm_grab(struct dlm_ctxt *dlm) { - struct list_head *iter; - struct dlm_ctxt *target = NULL; + struct dlm_ctxt *target; + struct dlm_ctxt *ret = NULL; spin_lock(&dlm_domain_lock); - list_for_each(iter, &dlm_domains) { - target = list_entry (iter, struct dlm_ctxt, list); - + list_for_each_entry(target, &dlm_domains, list) { if (target == dlm) { __dlm_get(target); + ret = target; break; } - - target = NULL; } spin_unlock(&dlm_domain_lock); - return target; + return ret; } int dlm_domain_fully_joined(struct dlm_ctxt *dlm) @@ -444,19 +444,21 @@ redo_bucket: dropped = dlm_empty_lockres(dlm, res); spin_lock(&res->spinlock); - __dlm_lockres_calc_usage(dlm, res); - iter = res->hash_node.next; + if (dropped) + __dlm_lockres_calc_usage(dlm, res); + else + iter = res->hash_node.next; spin_unlock(&res->spinlock); dlm_lockres_put(res); - if (dropped) + if (dropped) { + cond_resched_lock(&dlm->spinlock); goto redo_bucket; + } } cond_resched_lock(&dlm->spinlock); num += n; - mlog(0, "%s: touched %d lockreses in bucket %d " - "(tot=%d)\n", dlm->name, n, i, num); } spin_unlock(&dlm->spinlock); wake_up(&dlm->dlm_thread_wq); @@ -483,6 +485,28 @@ static int dlm_no_joining_node(struct dlm_ctxt *dlm) return ret; } +static int dlm_begin_exit_domain_handler(struct o2net_msg *msg, u32 len, + void *data, void **ret_data) +{ + struct dlm_ctxt *dlm = data; + unsigned int node; + struct dlm_exit_domain *exit_msg = (struct dlm_exit_domain *) msg->buf; + + if (!dlm_grab(dlm)) + return 0; + + node = exit_msg->node_idx; + mlog(0, "%s: Node %u sent a begin exit domain message\n", dlm->name, node); + + spin_lock(&dlm->spinlock); + set_bit(node, dlm->exit_domain_map); + spin_unlock(&dlm->spinlock); + + dlm_put(dlm); + + return 0; +} + static void dlm_mark_domain_leaving(struct dlm_ctxt *dlm) { /* Yikes, a double spinlock! I need domain_lock for the dlm @@ -508,17 +532,17 @@ again: static void __dlm_print_nodes(struct dlm_ctxt *dlm) { - int node = -1; + int node = -1, num = 0; assert_spin_locked(&dlm->spinlock); - printk(KERN_INFO "ocfs2_dlm: Nodes in domain (\"%s\"): ", dlm->name); - + printk("( "); while ((node = find_next_bit(dlm->domain_map, O2NM_MAX_NODES, node + 1)) < O2NM_MAX_NODES) { printk("%d ", node); + ++num; } - printk("\n"); + printk(") %u nodes\n", num); } static int dlm_exit_domain_handler(struct o2net_msg *msg, u32 len, void *data, @@ -528,17 +552,17 @@ static int dlm_exit_domain_handler(struct o2net_msg *msg, u32 len, void *data, unsigned int node; struct dlm_exit_domain *exit_msg = (struct dlm_exit_domain *) msg->buf; - mlog_entry("%p %u %p", msg, len, data); + mlog(0, "%p %u %p", msg, len, data); if (!dlm_grab(dlm)) return 0; node = exit_msg->node_idx; - printk(KERN_INFO "ocfs2_dlm: Node %u leaves domain %s\n", node, dlm->name); - spin_lock(&dlm->spinlock); clear_bit(node, dlm->domain_map); + clear_bit(node, dlm->exit_domain_map); + printk(KERN_NOTICE "o2dlm: Node %u leaves domain %s ", node, dlm->name); __dlm_print_nodes(dlm); /* notify anything attached to the heartbeat events */ @@ -551,27 +575,56 @@ static int dlm_exit_domain_handler(struct o2net_msg *msg, u32 len, void *data, return 0; } -static int dlm_send_one_domain_exit(struct dlm_ctxt *dlm, +static int dlm_send_one_domain_exit(struct dlm_ctxt *dlm, u32 msg_type, unsigned int node) { int status; struct dlm_exit_domain leave_msg; - mlog(0, "Asking node %u if we can leave the domain %s me = %u\n", - node, dlm->name, dlm->node_num); + mlog(0, "%s: Sending domain exit message %u to node %u\n", dlm->name, + msg_type, node); memset(&leave_msg, 0, sizeof(leave_msg)); leave_msg.node_idx = dlm->node_num; - status = o2net_send_message(DLM_EXIT_DOMAIN_MSG, dlm->key, - &leave_msg, sizeof(leave_msg), node, - NULL); - - mlog(0, "status return %d from o2net_send_message\n", status); + status = o2net_send_message(msg_type, dlm->key, &leave_msg, + sizeof(leave_msg), node, NULL); + if (status < 0) + mlog(ML_ERROR, "Error %d sending domain exit message %u " + "to node %u on domain %s\n", status, msg_type, node, + dlm->name); return status; } +static void dlm_begin_exit_domain(struct dlm_ctxt *dlm) +{ + int node = -1; + + /* Support for begin exit domain was added in 1.2 */ + if (dlm->dlm_locking_proto.pv_major == 1 && + dlm->dlm_locking_proto.pv_minor < 2) + return; + + /* + * Unlike DLM_EXIT_DOMAIN_MSG, DLM_BEGIN_EXIT_DOMAIN_MSG is purely + * informational. Meaning if a node does not receive the message, + * so be it. + */ + spin_lock(&dlm->spinlock); + while (1) { + node = find_next_bit(dlm->domain_map, O2NM_MAX_NODES, node + 1); + if (node >= O2NM_MAX_NODES) + break; + if (node == dlm->node_num) + continue; + + spin_unlock(&dlm->spinlock); + dlm_send_one_domain_exit(dlm, DLM_BEGIN_EXIT_DOMAIN_MSG, node); + spin_lock(&dlm->spinlock); + } + spin_unlock(&dlm->spinlock); +} static void dlm_leave_domain(struct dlm_ctxt *dlm) { @@ -597,7 +650,8 @@ static void dlm_leave_domain(struct dlm_ctxt *dlm) clear_node = 1; - status = dlm_send_one_domain_exit(dlm, node); + status = dlm_send_one_domain_exit(dlm, DLM_EXIT_DOMAIN_MSG, + node); if (status < 0 && status != -ENOPROTOOPT && status != -ENOTCONN) { @@ -672,6 +726,7 @@ void dlm_unregister_domain(struct dlm_ctxt *dlm) if (leave) { mlog(0, "shutting down domain %s\n", dlm->name); + dlm_begin_exit_domain(dlm); /* We changed dlm state, notify the thread */ dlm_kick_thread(dlm, NULL); @@ -692,6 +747,8 @@ void dlm_unregister_domain(struct dlm_ctxt *dlm) dlm_mark_domain_leaving(dlm); dlm_leave_domain(dlm); + printk(KERN_NOTICE "o2dlm: Leaving domain %s\n", dlm->name); + dlm_force_free_mles(dlm); dlm_complete_dlm_shutdown(dlm); } dlm_put(dlm); @@ -749,7 +806,7 @@ static void dlm_query_join_packet_to_wire(struct dlm_query_join_packet *packet, union dlm_query_join_response response; response.packet = *packet; - *wire = cpu_to_be32(response.intval); + *wire = be32_to_cpu(response.intval); } static void dlm_query_join_wire_to_packet(u32 wire, @@ -817,7 +874,7 @@ static int dlm_query_join_handler(struct o2net_msg *msg, u32 len, void *data, } /* Once the dlm ctxt is marked as leaving then we don't want - * to be put in someone's domain map. + * to be put in someone's domain map. * Also, explicitly disallow joining at certain troublesome * times (ie. during recovery). */ if (dlm && dlm->dlm_state != DLM_CTXT_LEAVING) { @@ -902,10 +959,19 @@ static int dlm_assert_joined_handler(struct o2net_msg *msg, u32 len, void *data, * domain. Set him in the map and clean up our * leftover join state. */ BUG_ON(dlm->joining_node != assert->node_idx); + + if (dlm->reco.state & DLM_RECO_STATE_ACTIVE) { + mlog(0, "dlm recovery is ongoing, disallow join\n"); + spin_unlock(&dlm->spinlock); + spin_unlock(&dlm_domain_lock); + return -EAGAIN; + } + set_bit(assert->node_idx, dlm->domain_map); + clear_bit(assert->node_idx, dlm->exit_domain_map); __dlm_set_joining_node(dlm, DLM_LOCK_RES_OWNER_UNKNOWN); - printk(KERN_INFO "ocfs2_dlm: Node %u joins domain %s\n", + printk(KERN_NOTICE "o2dlm: Node %u joins domain %s ", assert->node_idx, dlm->name); __dlm_print_nodes(dlm); @@ -919,6 +985,371 @@ static int dlm_assert_joined_handler(struct o2net_msg *msg, u32 len, void *data, return 0; } +static int dlm_match_regions(struct dlm_ctxt *dlm, + struct dlm_query_region *qr, + char *local, int locallen) +{ + char *remote = qr->qr_regions; + char *l, *r; + int localnr, i, j, foundit; + int status = 0; + + if (!o2hb_global_heartbeat_active()) { + if (qr->qr_numregions) { + mlog(ML_ERROR, "Domain %s: Joining node %d has global " + "heartbeat enabled but local node %d does not\n", + qr->qr_domain, qr->qr_node, dlm->node_num); + status = -EINVAL; + } + goto bail; + } + + if (o2hb_global_heartbeat_active() && !qr->qr_numregions) { + mlog(ML_ERROR, "Domain %s: Local node %d has global " + "heartbeat enabled but joining node %d does not\n", + qr->qr_domain, dlm->node_num, qr->qr_node); + status = -EINVAL; + goto bail; + } + + r = remote; + for (i = 0; i < qr->qr_numregions; ++i) { + mlog(0, "Region %.*s\n", O2HB_MAX_REGION_NAME_LEN, r); + r += O2HB_MAX_REGION_NAME_LEN; + } + + localnr = min(O2NM_MAX_REGIONS, locallen/O2HB_MAX_REGION_NAME_LEN); + localnr = o2hb_get_all_regions(local, (u8)localnr); + + /* compare local regions with remote */ + l = local; + for (i = 0; i < localnr; ++i) { + foundit = 0; + r = remote; + for (j = 0; j <= qr->qr_numregions; ++j) { + if (!memcmp(l, r, O2HB_MAX_REGION_NAME_LEN)) { + foundit = 1; + break; + } + r += O2HB_MAX_REGION_NAME_LEN; + } + if (!foundit) { + status = -EINVAL; + mlog(ML_ERROR, "Domain %s: Region '%.*s' registered " + "in local node %d but not in joining node %d\n", + qr->qr_domain, O2HB_MAX_REGION_NAME_LEN, l, + dlm->node_num, qr->qr_node); + goto bail; + } + l += O2HB_MAX_REGION_NAME_LEN; + } + + /* compare remote with local regions */ + r = remote; + for (i = 0; i < qr->qr_numregions; ++i) { + foundit = 0; + l = local; + for (j = 0; j < localnr; ++j) { + if (!memcmp(r, l, O2HB_MAX_REGION_NAME_LEN)) { + foundit = 1; + break; + } + l += O2HB_MAX_REGION_NAME_LEN; + } + if (!foundit) { + status = -EINVAL; + mlog(ML_ERROR, "Domain %s: Region '%.*s' registered " + "in joining node %d but not in local node %d\n", + qr->qr_domain, O2HB_MAX_REGION_NAME_LEN, r, + qr->qr_node, dlm->node_num); + goto bail; + } + r += O2HB_MAX_REGION_NAME_LEN; + } + +bail: + return status; +} + +static int dlm_send_regions(struct dlm_ctxt *dlm, unsigned long *node_map) +{ + struct dlm_query_region *qr = NULL; + int status, ret = 0, i; + char *p; + + if (find_next_bit(node_map, O2NM_MAX_NODES, 0) >= O2NM_MAX_NODES) + goto bail; + + qr = kzalloc(sizeof(struct dlm_query_region), GFP_KERNEL); + if (!qr) { + ret = -ENOMEM; + mlog_errno(ret); + goto bail; + } + + qr->qr_node = dlm->node_num; + qr->qr_namelen = strlen(dlm->name); + memcpy(qr->qr_domain, dlm->name, qr->qr_namelen); + /* if local hb, the numregions will be zero */ + if (o2hb_global_heartbeat_active()) + qr->qr_numregions = o2hb_get_all_regions(qr->qr_regions, + O2NM_MAX_REGIONS); + + p = qr->qr_regions; + for (i = 0; i < qr->qr_numregions; ++i, p += O2HB_MAX_REGION_NAME_LEN) + mlog(0, "Region %.*s\n", O2HB_MAX_REGION_NAME_LEN, p); + + i = -1; + while ((i = find_next_bit(node_map, O2NM_MAX_NODES, + i + 1)) < O2NM_MAX_NODES) { + if (i == dlm->node_num) + continue; + + mlog(0, "Sending regions to node %d\n", i); + + ret = o2net_send_message(DLM_QUERY_REGION, DLM_MOD_KEY, qr, + sizeof(struct dlm_query_region), + i, &status); + if (ret >= 0) + ret = status; + if (ret) { + mlog(ML_ERROR, "Region mismatch %d, node %d\n", + ret, i); + break; + } + } + +bail: + kfree(qr); + return ret; +} + +static int dlm_query_region_handler(struct o2net_msg *msg, u32 len, + void *data, void **ret_data) +{ + struct dlm_query_region *qr; + struct dlm_ctxt *dlm = NULL; + char *local = NULL; + int status = 0; + + qr = (struct dlm_query_region *) msg->buf; + + mlog(0, "Node %u queries hb regions on domain %s\n", qr->qr_node, + qr->qr_domain); + + /* buffer used in dlm_mast_regions() */ + local = kmalloc(sizeof(qr->qr_regions), GFP_KERNEL); + if (!local) + return -ENOMEM; + + status = -EINVAL; + + spin_lock(&dlm_domain_lock); + dlm = __dlm_lookup_domain_full(qr->qr_domain, qr->qr_namelen); + if (!dlm) { + mlog(ML_ERROR, "Node %d queried hb regions on domain %s " + "before join domain\n", qr->qr_node, qr->qr_domain); + goto out_domain_lock; + } + + spin_lock(&dlm->spinlock); + if (dlm->joining_node != qr->qr_node) { + mlog(ML_ERROR, "Node %d queried hb regions on domain %s " + "but joining node is %d\n", qr->qr_node, qr->qr_domain, + dlm->joining_node); + goto out_dlm_lock; + } + + /* Support for global heartbeat was added in 1.1 */ + if (dlm->dlm_locking_proto.pv_major == 1 && + dlm->dlm_locking_proto.pv_minor == 0) { + mlog(ML_ERROR, "Node %d queried hb regions on domain %s " + "but active dlm protocol is %d.%d\n", qr->qr_node, + qr->qr_domain, dlm->dlm_locking_proto.pv_major, + dlm->dlm_locking_proto.pv_minor); + goto out_dlm_lock; + } + + status = dlm_match_regions(dlm, qr, local, sizeof(qr->qr_regions)); + +out_dlm_lock: + spin_unlock(&dlm->spinlock); + +out_domain_lock: + spin_unlock(&dlm_domain_lock); + + kfree(local); + + return status; +} + +static int dlm_match_nodes(struct dlm_ctxt *dlm, struct dlm_query_nodeinfo *qn) +{ + struct o2nm_node *local; + struct dlm_node_info *remote; + int i, j; + int status = 0; + + for (j = 0; j < qn->qn_numnodes; ++j) + mlog(0, "Node %3d, %pI4:%u\n", qn->qn_nodes[j].ni_nodenum, + &(qn->qn_nodes[j].ni_ipv4_address), + ntohs(qn->qn_nodes[j].ni_ipv4_port)); + + for (i = 0; i < O2NM_MAX_NODES && !status; ++i) { + local = o2nm_get_node_by_num(i); + remote = NULL; + for (j = 0; j < qn->qn_numnodes; ++j) { + if (qn->qn_nodes[j].ni_nodenum == i) { + remote = &(qn->qn_nodes[j]); + break; + } + } + + if (!local && !remote) + continue; + + if ((local && !remote) || (!local && remote)) + status = -EINVAL; + + if (!status && + ((remote->ni_nodenum != local->nd_num) || + (remote->ni_ipv4_port != local->nd_ipv4_port) || + (remote->ni_ipv4_address != local->nd_ipv4_address))) + status = -EINVAL; + + if (status) { + if (remote && !local) + mlog(ML_ERROR, "Domain %s: Node %d (%pI4:%u) " + "registered in joining node %d but not in " + "local node %d\n", qn->qn_domain, + remote->ni_nodenum, + &(remote->ni_ipv4_address), + ntohs(remote->ni_ipv4_port), + qn->qn_nodenum, dlm->node_num); + if (local && !remote) + mlog(ML_ERROR, "Domain %s: Node %d (%pI4:%u) " + "registered in local node %d but not in " + "joining node %d\n", qn->qn_domain, + local->nd_num, &(local->nd_ipv4_address), + ntohs(local->nd_ipv4_port), + dlm->node_num, qn->qn_nodenum); + BUG_ON((!local && !remote)); + } + + if (local) + o2nm_node_put(local); + } + + return status; +} + +static int dlm_send_nodeinfo(struct dlm_ctxt *dlm, unsigned long *node_map) +{ + struct dlm_query_nodeinfo *qn = NULL; + struct o2nm_node *node; + int ret = 0, status, count, i; + + if (find_next_bit(node_map, O2NM_MAX_NODES, 0) >= O2NM_MAX_NODES) + goto bail; + + qn = kzalloc(sizeof(struct dlm_query_nodeinfo), GFP_KERNEL); + if (!qn) { + ret = -ENOMEM; + mlog_errno(ret); + goto bail; + } + + for (i = 0, count = 0; i < O2NM_MAX_NODES; ++i) { + node = o2nm_get_node_by_num(i); + if (!node) + continue; + qn->qn_nodes[count].ni_nodenum = node->nd_num; + qn->qn_nodes[count].ni_ipv4_port = node->nd_ipv4_port; + qn->qn_nodes[count].ni_ipv4_address = node->nd_ipv4_address; + mlog(0, "Node %3d, %pI4:%u\n", node->nd_num, + &(node->nd_ipv4_address), ntohs(node->nd_ipv4_port)); + ++count; + o2nm_node_put(node); + } + + qn->qn_nodenum = dlm->node_num; + qn->qn_numnodes = count; + qn->qn_namelen = strlen(dlm->name); + memcpy(qn->qn_domain, dlm->name, qn->qn_namelen); + + i = -1; + while ((i = find_next_bit(node_map, O2NM_MAX_NODES, + i + 1)) < O2NM_MAX_NODES) { + if (i == dlm->node_num) + continue; + + mlog(0, "Sending nodeinfo to node %d\n", i); + + ret = o2net_send_message(DLM_QUERY_NODEINFO, DLM_MOD_KEY, + qn, sizeof(struct dlm_query_nodeinfo), + i, &status); + if (ret >= 0) + ret = status; + if (ret) { + mlog(ML_ERROR, "node mismatch %d, node %d\n", ret, i); + break; + } + } + +bail: + kfree(qn); + return ret; +} + +static int dlm_query_nodeinfo_handler(struct o2net_msg *msg, u32 len, + void *data, void **ret_data) +{ + struct dlm_query_nodeinfo *qn; + struct dlm_ctxt *dlm = NULL; + int locked = 0, status = -EINVAL; + + qn = (struct dlm_query_nodeinfo *) msg->buf; + + mlog(0, "Node %u queries nodes on domain %s\n", qn->qn_nodenum, + qn->qn_domain); + + spin_lock(&dlm_domain_lock); + dlm = __dlm_lookup_domain_full(qn->qn_domain, qn->qn_namelen); + if (!dlm) { + mlog(ML_ERROR, "Node %d queried nodes on domain %s before " + "join domain\n", qn->qn_nodenum, qn->qn_domain); + goto bail; + } + + spin_lock(&dlm->spinlock); + locked = 1; + if (dlm->joining_node != qn->qn_nodenum) { + mlog(ML_ERROR, "Node %d queried nodes on domain %s but " + "joining node is %d\n", qn->qn_nodenum, qn->qn_domain, + dlm->joining_node); + goto bail; + } + + /* Support for node query was added in 1.1 */ + if (dlm->dlm_locking_proto.pv_major == 1 && + dlm->dlm_locking_proto.pv_minor == 0) { + mlog(ML_ERROR, "Node %d queried nodes on domain %s " + "but active dlm protocol is %d.%d\n", qn->qn_nodenum, + qn->qn_domain, dlm->dlm_locking_proto.pv_major, + dlm->dlm_locking_proto.pv_minor); + goto bail; + } + + status = dlm_match_nodes(dlm, qn); + +bail: + if (locked) + spin_unlock(&dlm->spinlock); + spin_unlock(&dlm_domain_lock); + + return status; +} + static int dlm_cancel_join_handler(struct o2net_msg *msg, u32 len, void *data, void **ret_data) { @@ -963,7 +1394,9 @@ static int dlm_send_one_join_cancel(struct dlm_ctxt *dlm, &cancel_msg, sizeof(cancel_msg), node, NULL); if (status < 0) { - mlog_errno(status); + mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to " + "node %u\n", status, DLM_CANCEL_JOIN_MSG, DLM_MOD_KEY, + node); goto bail; } @@ -1030,10 +1463,11 @@ static int dlm_request_join(struct dlm_ctxt *dlm, byte_copymap(join_msg.node_map, dlm->live_nodes_map, O2NM_MAX_NODES); status = o2net_send_message(DLM_QUERY_JOIN_MSG, DLM_MOD_KEY, &join_msg, - sizeof(join_msg), node, - &join_resp); + sizeof(join_msg), node, &join_resp); if (status < 0 && status != -ENOPROTOOPT) { - mlog_errno(status); + mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to " + "node %u\n", status, DLM_QUERY_JOIN_MSG, DLM_MOD_KEY, + node); goto bail; } dlm_query_join_wire_to_packet(join_resp, &packet); @@ -1091,6 +1525,7 @@ static int dlm_send_one_join_assert(struct dlm_ctxt *dlm, unsigned int node) { int status; + int ret; struct dlm_assert_joined assert_msg; mlog(0, "Sending join assert to node %u\n", node); @@ -1102,9 +1537,13 @@ static int dlm_send_one_join_assert(struct dlm_ctxt *dlm, status = o2net_send_message(DLM_ASSERT_JOINED_MSG, DLM_MOD_KEY, &assert_msg, sizeof(assert_msg), node, - NULL); + &ret); if (status < 0) - mlog_errno(status); + mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to " + "node %u\n", status, DLM_ASSERT_JOINED_MSG, DLM_MOD_KEY, + node); + else + status = ret; return status; } @@ -1178,7 +1617,7 @@ static int dlm_try_to_join_domain(struct dlm_ctxt *dlm) struct domain_join_ctxt *ctxt; enum dlm_query_join_response_code response = JOIN_DISALLOW; - mlog_entry("%p", dlm); + mlog(0, "%p", dlm); ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL); if (!ctxt) { @@ -1234,6 +1673,21 @@ static int dlm_try_to_join_domain(struct dlm_ctxt *dlm) set_bit(dlm->node_num, dlm->domain_map); spin_unlock(&dlm->spinlock); + /* Support for global heartbeat and node info was added in 1.1 */ + if (dlm->dlm_locking_proto.pv_major > 1 || + dlm->dlm_locking_proto.pv_minor > 0) { + status = dlm_send_nodeinfo(dlm, ctxt->yes_resp_map); + if (status) { + mlog_errno(status); + goto bail; + } + status = dlm_send_regions(dlm, ctxt->yes_resp_map); + if (status) { + mlog_errno(status); + goto bail; + } + } + dlm_send_join_asserts(dlm, ctxt->yes_resp_map); /* Joined state *must* be set before the joining node @@ -1248,8 +1702,10 @@ static int dlm_try_to_join_domain(struct dlm_ctxt *dlm) bail: spin_lock(&dlm->spinlock); __dlm_set_joining_node(dlm, DLM_LOCK_RES_OWNER_UNKNOWN); - if (!status) + if (!status) { + printk(KERN_NOTICE "o2dlm: Joining domain %s ", dlm->name); __dlm_print_nodes(dlm); + } spin_unlock(&dlm->spinlock); if (ctxt) { @@ -1270,8 +1726,8 @@ bail: static void dlm_unregister_domain_handlers(struct dlm_ctxt *dlm) { - o2hb_unregister_callback(NULL, &dlm->dlm_hb_up); - o2hb_unregister_callback(NULL, &dlm->dlm_hb_down); + o2hb_unregister_callback(dlm->name, &dlm->dlm_hb_up); + o2hb_unregister_callback(dlm->name, &dlm->dlm_hb_down); o2net_unregister_handler_list(&dlm->dlm_domain_handlers); } @@ -1283,13 +1739,13 @@ static int dlm_register_domain_handlers(struct dlm_ctxt *dlm) o2hb_setup_callback(&dlm->dlm_hb_down, O2HB_NODE_DOWN_CB, dlm_hb_node_down_cb, dlm, DLM_HB_NODE_DOWN_PRI); - status = o2hb_register_callback(NULL, &dlm->dlm_hb_down); + status = o2hb_register_callback(dlm->name, &dlm->dlm_hb_down); if (status) goto bail; o2hb_setup_callback(&dlm->dlm_hb_up, O2HB_NODE_UP_CB, dlm_hb_node_up_cb, dlm, DLM_HB_NODE_UP_PRI); - status = o2hb_register_callback(NULL, &dlm->dlm_hb_up); + status = o2hb_register_callback(dlm->name, &dlm->dlm_hb_up); if (status) goto bail; @@ -1399,6 +1855,13 @@ static int dlm_register_domain_handlers(struct dlm_ctxt *dlm) if (status) goto bail; + status = o2net_register_handler(DLM_BEGIN_EXIT_DOMAIN_MSG, dlm->key, + sizeof(struct dlm_exit_domain), + dlm_begin_exit_domain_handler, + dlm, NULL, &dlm->dlm_domain_handlers); + if (status) + goto bail; + bail: if (status) dlm_unregister_domain_handlers(dlm); @@ -1422,19 +1885,19 @@ static int dlm_join_domain(struct dlm_ctxt *dlm) goto bail; } - status = dlm_debug_init(dlm); + status = dlm_launch_thread(dlm); if (status < 0) { mlog_errno(status); goto bail; } - status = dlm_launch_thread(dlm); + status = dlm_launch_recovery_thread(dlm); if (status < 0) { mlog_errno(status); goto bail; } - status = dlm_launch_recovery_thread(dlm); + status = dlm_debug_init(dlm); if (status < 0) { mlog_errno(status); goto bail; @@ -1517,7 +1980,7 @@ static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain, goto leave; } - dlm->name = kmalloc(strlen(domain) + 1, GFP_KERNEL); + dlm->name = kstrdup(domain, GFP_KERNEL); if (dlm->name == NULL) { mlog_errno(-ENOMEM); kfree(dlm); @@ -1551,7 +2014,6 @@ static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain, for (i = 0; i < DLM_HASH_BUCKETS; i++) INIT_HLIST_HEAD(dlm_master_hash(dlm, i)); - strcpy(dlm->name, domain); dlm->key = key; dlm->node_num = o2nm_this_node(); @@ -1572,7 +2034,6 @@ static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain, INIT_LIST_HEAD(&dlm->list); INIT_LIST_HEAD(&dlm->dirty_list); INIT_LIST_HEAD(&dlm->reco.resources); - INIT_LIST_HEAD(&dlm->reco.received); INIT_LIST_HEAD(&dlm->reco.node_data); INIT_LIST_HEAD(&dlm->purge_list); INIT_LIST_HEAD(&dlm->dlm_domain_handlers); @@ -1666,19 +2127,12 @@ struct dlm_ctxt * dlm_register_domain(const char *domain, struct dlm_ctxt *dlm = NULL; struct dlm_ctxt *new_ctxt = NULL; - if (strlen(domain) > O2NM_MAX_NAME_LEN) { + if (strlen(domain) >= O2NM_MAX_NAME_LEN) { ret = -ENAMETOOLONG; mlog(ML_ERROR, "domain name length too long\n"); goto leave; } - if (!o2hb_check_local_node_heartbeating()) { - mlog(ML_ERROR, "the local node has not been configured, or is " - "not heartbeating\n"); - ret = -EPROTO; - goto leave; - } - mlog(0, "register called for domain \"%s\"\n", domain); retry: @@ -1704,6 +2158,7 @@ retry: } if (dlm_protocol_compare(&dlm->fs_locking_proto, fs_proto)) { + spin_unlock(&dlm_domain_lock); mlog(ML_ERROR, "Requested locking protocol version is not " "compatible with already registered domain " @@ -1800,7 +2255,21 @@ static int dlm_register_net_handlers(void) sizeof(struct dlm_cancel_join), dlm_cancel_join_handler, NULL, NULL, &dlm_join_handlers); + if (status) + goto bail; + status = o2net_register_handler(DLM_QUERY_REGION, DLM_MOD_KEY, + sizeof(struct dlm_query_region), + dlm_query_region_handler, + NULL, NULL, &dlm_join_handlers); + + if (status) + goto bail; + + status = o2net_register_handler(DLM_QUERY_NODEINFO, DLM_MOD_KEY, + sizeof(struct dlm_query_nodeinfo), + dlm_query_nodeinfo_handler, + NULL, NULL, &dlm_join_handlers); bail: if (status < 0) dlm_unregister_net_handlers(); @@ -1824,13 +2293,10 @@ static DECLARE_RWSEM(dlm_callback_sem); void dlm_fire_domain_eviction_callbacks(struct dlm_ctxt *dlm, int node_num) { - struct list_head *iter; struct dlm_eviction_cb *cb; down_read(&dlm_callback_sem); - list_for_each(iter, &dlm->dlm_eviction_callbacks) { - cb = list_entry(iter, struct dlm_eviction_cb, ec_item); - + list_for_each_entry(cb, &dlm->dlm_eviction_callbacks, ec_item) { cb->ec_func(node_num, cb->ec_data); } up_read(&dlm_callback_sem); @@ -1867,8 +2333,6 @@ static int __init dlm_init(void) { int status; - dlm_print_version(); - status = dlm_init_mle_cache(); if (status) { mlog(ML_ERROR, "Could not create o2dlm_mle slabcache\n"); @@ -1918,6 +2382,7 @@ static void __exit dlm_exit (void) MODULE_AUTHOR("Oracle"); MODULE_LICENSE("GPL"); +MODULE_DESCRIPTION("OCFS2 Distributed Lock Management"); module_init(dlm_init); module_exit(dlm_exit); diff --git a/fs/ocfs2/dlm/dlmfsver.c b/fs/ocfs2/dlm/dlmfsver.c deleted file mode 100644 index a733b3321f8..00000000000 --- a/fs/ocfs2/dlm/dlmfsver.c +++ /dev/null @@ -1,42 +0,0 @@ -/* -*- mode: c; c-basic-offset: 8; -*- - * vim: noexpandtab sw=8 ts=8 sts=0: - * - * dlmfsver.c - * - * version string - * - * Copyright (C) 2002, 2005 Oracle. All rights reserved. - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public - * License as published by the Free Software Foundation; either - * version 2 of the License, or (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - * General Public License for more details. - * - * You should have received a copy of the GNU General Public - * License along with this program; if not, write to the - * Free Software Foundation, Inc., 59 Temple Place - Suite 330, - * Boston, MA 021110-1307, USA. - */ - -#include <linux/module.h> -#include <linux/kernel.h> - -#include "dlmfsver.h" - -#define DLM_BUILD_VERSION "1.5.0" - -#define VERSION_STR "OCFS2 DLMFS " DLM_BUILD_VERSION - -void dlmfs_print_version(void) -{ - printk(KERN_INFO "%s\n", VERSION_STR); -} - -MODULE_DESCRIPTION(VERSION_STR); - -MODULE_VERSION(DLM_BUILD_VERSION); diff --git a/fs/ocfs2/dlm/dlmfsver.h b/fs/ocfs2/dlm/dlmfsver.h deleted file mode 100644 index f35eadbed25..00000000000 --- a/fs/ocfs2/dlm/dlmfsver.h +++ /dev/null @@ -1,31 +0,0 @@ -/* -*- mode: c; c-basic-offset: 8; -*- - * vim: noexpandtab sw=8 ts=8 sts=0: - * - * dlmver.h - * - * Function prototypes - * - * Copyright (C) 2005 Oracle. All rights reserved. - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public - * License as published by the Free Software Foundation; either - * version 2 of the License, or (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - * General Public License for more details. - * - * You should have received a copy of the GNU General Public - * License along with this program; if not, write to the - * Free Software Foundation, Inc., 59 Temple Place - Suite 330, - * Boston, MA 021110-1307, USA. - */ - -#ifndef DLMFS_VER_H -#define DLMFS_VER_H - -void dlmfs_print_version(void); - -#endif /* DLMFS_VER_H */ diff --git a/fs/ocfs2/dlm/dlmlock.c b/fs/ocfs2/dlm/dlmlock.c index 83a9f2972ac..66c2a491f68 100644 --- a/fs/ocfs2/dlm/dlmlock.c +++ b/fs/ocfs2/dlm/dlmlock.c @@ -30,7 +30,6 @@ #include <linux/types.h> #include <linux/slab.h> #include <linux/highmem.h> -#include <linux/utsname.h> #include <linux/init.h> #include <linux/sysctl.h> #include <linux/random.h> @@ -53,7 +52,7 @@ #define MLOG_MASK_PREFIX ML_DLM #include "cluster/masklog.h" -static struct kmem_cache *dlm_lock_cache = NULL; +static struct kmem_cache *dlm_lock_cache; static DEFINE_SPINLOCK(dlm_cookie_lock); static u64 dlm_next_cookie = 1; @@ -92,21 +91,19 @@ void dlm_destroy_lock_cache(void) static int dlm_can_grant_new_lock(struct dlm_lock_resource *res, struct dlm_lock *lock) { - struct list_head *iter; struct dlm_lock *tmplock; - list_for_each(iter, &res->granted) { - tmplock = list_entry(iter, struct dlm_lock, list); - + list_for_each_entry(tmplock, &res->granted, list) { if (!dlm_lock_compatible(tmplock->ml.type, lock->ml.type)) return 0; } - list_for_each(iter, &res->converting) { - tmplock = list_entry(iter, struct dlm_lock, list); - + list_for_each_entry(tmplock, &res->converting, list) { if (!dlm_lock_compatible(tmplock->ml.type, lock->ml.type)) return 0; + if (!dlm_lock_compatible(tmplock->ml.convert_type, + lock->ml.type)) + return 0; } return 1; @@ -126,7 +123,7 @@ static enum dlm_status dlmlock_master(struct dlm_ctxt *dlm, int call_ast = 0, kick_thread = 0; enum dlm_status status = DLM_NORMAL; - mlog_entry("type=%d\n", lock->ml.type); + mlog(0, "type=%d\n", lock->ml.type); spin_lock(&res->spinlock); /* if called from dlm_create_lock_handler, need to @@ -176,15 +173,12 @@ static enum dlm_status dlmlock_master(struct dlm_ctxt *dlm, lock->ml.node); } } else { + status = DLM_NORMAL; dlm_lock_get(lock); list_add_tail(&lock->list, &res->blocked); kick_thread = 1; } } - /* reduce the inflight count, this may result in the lockres - * being purged below during calc_usage */ - if (lock->ml.node == dlm->node_num) - dlm_lockres_drop_inflight_ref(dlm, res); spin_unlock(&res->spinlock); wake_up(&res->wq); @@ -225,14 +219,20 @@ static enum dlm_status dlmlock_remote(struct dlm_ctxt *dlm, enum dlm_status status = DLM_DENIED; int lockres_changed = 1; - mlog_entry("type=%d\n", lock->ml.type); - mlog(0, "lockres %.*s, flags = 0x%x\n", res->lockname.len, + mlog(0, "type=%d, lockres %.*s, flags = 0x%x\n", + lock->ml.type, res->lockname.len, res->lockname.name, flags); + /* + * Wait if resource is getting recovered, remastered, etc. + * If the resource was remastered and new owner is self, then exit. + */ spin_lock(&res->spinlock); - - /* will exit this call with spinlock held */ __dlm_wait_on_lockres(res); + if (res->owner == dlm->node_num) { + spin_unlock(&res->spinlock); + return DLM_RECOVERING; + } res->state |= DLM_LOCK_RES_IN_PROGRESS; /* add lock to local (secondary) queue */ @@ -270,7 +270,7 @@ static enum dlm_status dlmlock_remote(struct dlm_ctxt *dlm, } dlm_revert_pending_lock(res, lock); dlm_lock_put(lock); - } else if (dlm_is_recovery_lock(res->lockname.name, + } else if (dlm_is_recovery_lock(res->lockname.name, res->lockname.len)) { /* special case for the $RECOVERY lock. * there will never be an AST delivered to put @@ -306,8 +306,6 @@ static enum dlm_status dlm_send_remote_lock_request(struct dlm_ctxt *dlm, int tmpret, status = 0; enum dlm_status ret; - mlog_entry_void(); - memset(&create, 0, sizeof(create)); create.node_idx = dlm->node_num; create.requested_type = lock->ml.type; @@ -319,25 +317,23 @@ static enum dlm_status dlm_send_remote_lock_request(struct dlm_ctxt *dlm, tmpret = o2net_send_message(DLM_CREATE_LOCK_MSG, dlm->key, &create, sizeof(create), res->owner, &status); if (tmpret >= 0) { - // successfully sent and received - ret = status; // this is already a dlm_status + ret = status; if (ret == DLM_REJECTED) { - mlog(ML_ERROR, "%s:%.*s: BUG. this is a stale lockres " - "no longer owned by %u. that node is coming back " - "up currently.\n", dlm->name, create.namelen, + mlog(ML_ERROR, "%s: res %.*s, Stale lockres no longer " + "owned by node %u. That node is coming back up " + "currently.\n", dlm->name, create.namelen, create.name, res->owner); dlm_print_one_lock_resource(res); BUG(); } } else { - mlog_errno(tmpret); - if (dlm_is_host_down(tmpret)) { + mlog(ML_ERROR, "%s: res %.*s, Error %d send CREATE LOCK to " + "node %u\n", dlm->name, create.namelen, create.name, + tmpret, res->owner); + if (dlm_is_host_down(tmpret)) ret = DLM_RECOVERING; - mlog(0, "node %u died so returning DLM_RECOVERING " - "from lock message!\n", res->owner); - } else { + else ret = dlm_err_to_dlm_status(tmpret); - } } return ret; @@ -430,7 +426,7 @@ struct dlm_lock * dlm_new_lock(int type, u8 node, u64 cookie, struct dlm_lock *lock; int kernel_allocated = 0; - lock = (struct dlm_lock *) kmem_cache_zalloc(dlm_lock_cache, GFP_NOFS); + lock = kmem_cache_zalloc(dlm_lock_cache, GFP_NOFS); if (!lock) return NULL; @@ -438,7 +434,7 @@ struct dlm_lock * dlm_new_lock(int type, u8 node, u64 cookie, /* zero memory only if kernel-allocated */ lksb = kzalloc(sizeof(*lksb), GFP_NOFS); if (!lksb) { - kfree(lock); + kmem_cache_free(dlm_lock_cache, lock); return NULL; } kernel_allocated = 1; @@ -473,8 +469,6 @@ int dlm_create_lock_handler(struct o2net_msg *msg, u32 len, void *data, BUG_ON(!dlm); - mlog_entry_void(); - if (!dlm_grab(dlm)) return DLM_REJECTED; @@ -718,18 +712,10 @@ retry_lock: if (status == DLM_RECOVERING || status == DLM_MIGRATING || status == DLM_FORWARD) { - mlog(0, "retrying lock with migration/" - "recovery/in progress\n"); msleep(100); - /* no waiting for dlm_reco_thread */ if (recovery) { if (status != DLM_RECOVERING) goto retry_lock; - - mlog(0, "%s: got RECOVERING " - "for $RECOVERY lock, master " - "was %u\n", dlm->name, - res->owner); /* wait to see the node go down, then * drop down and allow the lockres to * get cleaned up. need to remaster. */ @@ -741,6 +727,14 @@ retry_lock: } } + /* Inflight taken in dlm_get_lock_resource() is dropped here */ + spin_lock(&res->spinlock); + dlm_lockres_drop_inflight_ref(dlm, res); + spin_unlock(&res->spinlock); + + dlm_lockres_calc_usage(dlm, res); + dlm_kick_thread(dlm, res); + if (status != DLM_NORMAL) { lock->lksb->flags &= ~DLM_LKSB_GET_LVB; if (status != DLM_NOTQUEUED) diff --git a/fs/ocfs2/dlm/dlmmaster.c b/fs/ocfs2/dlm/dlmmaster.c index f8b653fcd4d..82abf0cc9a1 100644 --- a/fs/ocfs2/dlm/dlmmaster.c +++ b/fs/ocfs2/dlm/dlmmaster.c @@ -30,7 +30,6 @@ #include <linux/types.h> #include <linux/slab.h> #include <linux/highmem.h> -#include <linux/utsname.h> #include <linux/init.h> #include <linux/sysctl.h> #include <linux/random.h> @@ -83,9 +82,9 @@ static inline int dlm_mle_equal(struct dlm_ctxt *dlm, return 1; } -static struct kmem_cache *dlm_lockres_cache = NULL; -static struct kmem_cache *dlm_lockname_cache = NULL; -static struct kmem_cache *dlm_mle_cache = NULL; +static struct kmem_cache *dlm_lockres_cache; +static struct kmem_cache *dlm_lockname_cache; +static struct kmem_cache *dlm_mle_cache; static void dlm_mle_release(struct kref *kref); static void dlm_init_mle(struct dlm_master_list_entry *mle, @@ -343,16 +342,13 @@ static int dlm_find_mle(struct dlm_ctxt *dlm, { struct dlm_master_list_entry *tmpmle; struct hlist_head *bucket; - struct hlist_node *list; unsigned int hash; assert_spin_locked(&dlm->master_lock); hash = dlm_lockid_hash(name, namelen); bucket = dlm_master_hash(dlm, hash); - hlist_for_each(list, bucket) { - tmpmle = hlist_entry(list, struct dlm_master_list_entry, - master_hash_node); + hlist_for_each_entry(tmpmle, bucket, master_hash_node) { if (!dlm_mle_equal(dlm, tmpmle, name, namelen)) continue; dlm_get_mle(tmpmle); @@ -367,7 +363,7 @@ void dlm_hb_event_notify_attached(struct dlm_ctxt *dlm, int idx, int node_up) struct dlm_master_list_entry *mle; assert_spin_locked(&dlm->spinlock); - + list_for_each_entry(mle, &dlm->mle_hb_events, hb_events) { if (node_up) dlm_mle_node_up(dlm, mle, NULL, idx); @@ -427,8 +423,6 @@ static void dlm_mle_release(struct kref *kref) struct dlm_master_list_entry *mle; struct dlm_ctxt *dlm; - mlog_entry_void(); - mle = container_of(kref, struct dlm_master_list_entry, mle_refs); dlm = mle->dlm; @@ -478,11 +472,15 @@ bail: void dlm_destroy_master_caches(void) { - if (dlm_lockname_cache) + if (dlm_lockname_cache) { kmem_cache_destroy(dlm_lockname_cache); + dlm_lockname_cache = NULL; + } - if (dlm_lockres_cache) + if (dlm_lockres_cache) { kmem_cache_destroy(dlm_lockres_cache); + dlm_lockres_cache = NULL; + } } static void dlm_lockres_release(struct kref *kref) @@ -512,8 +510,6 @@ static void dlm_lockres_release(struct kref *kref) atomic_dec(&dlm->res_cur_count); - dlm_put(dlm); - if (!hlist_unhashed(&res->hash_node) || !list_empty(&res->granted) || !list_empty(&res->converting) || @@ -585,9 +581,8 @@ static void dlm_init_lockres(struct dlm_ctxt *dlm, atomic_set(&res->asts_reserved, 0); res->migration_pending = 0; res->inflight_locks = 0; + res->inflight_assert_workers = 0; - /* put in dlm_lockres_release */ - dlm_grab(dlm); res->dlm = dlm; kref_init(&res->refs); @@ -618,13 +613,11 @@ struct dlm_lock_resource *dlm_new_lockres(struct dlm_ctxt *dlm, { struct dlm_lock_resource *res = NULL; - res = (struct dlm_lock_resource *) - kmem_cache_zalloc(dlm_lockres_cache, GFP_NOFS); + res = kmem_cache_zalloc(dlm_lockres_cache, GFP_NOFS); if (!res) goto error; - res->lockname.name = (char *) - kmem_cache_zalloc(dlm_lockname_cache, GFP_NOFS); + res->lockname.name = kmem_cache_zalloc(dlm_lockname_cache, GFP_NOFS); if (!res->lockname.name) goto error; @@ -640,42 +633,94 @@ error: return NULL; } -void __dlm_lockres_grab_inflight_ref(struct dlm_ctxt *dlm, - struct dlm_lock_resource *res, - int new_lockres, - const char *file, - int line) +void dlm_lockres_set_refmap_bit(struct dlm_ctxt *dlm, + struct dlm_lock_resource *res, int bit) +{ + assert_spin_locked(&res->spinlock); + + mlog(0, "res %.*s, set node %u, %ps()\n", res->lockname.len, + res->lockname.name, bit, __builtin_return_address(0)); + + set_bit(bit, res->refmap); +} + +void dlm_lockres_clear_refmap_bit(struct dlm_ctxt *dlm, + struct dlm_lock_resource *res, int bit) { - if (!new_lockres) - assert_spin_locked(&res->spinlock); + assert_spin_locked(&res->spinlock); + + mlog(0, "res %.*s, clr node %u, %ps()\n", res->lockname.len, + res->lockname.name, bit, __builtin_return_address(0)); + + clear_bit(bit, res->refmap); +} + + +void dlm_lockres_grab_inflight_ref(struct dlm_ctxt *dlm, + struct dlm_lock_resource *res) +{ + assert_spin_locked(&res->spinlock); - if (!test_bit(dlm->node_num, res->refmap)) { - BUG_ON(res->inflight_locks != 0); - dlm_lockres_set_refmap_bit(dlm->node_num, res); - } res->inflight_locks++; - mlog(0, "%s:%.*s: inflight++: now %u\n", - dlm->name, res->lockname.len, res->lockname.name, - res->inflight_locks); + + mlog(0, "%s: res %.*s, inflight++: now %u, %ps()\n", dlm->name, + res->lockname.len, res->lockname.name, res->inflight_locks, + __builtin_return_address(0)); } -void __dlm_lockres_drop_inflight_ref(struct dlm_ctxt *dlm, - struct dlm_lock_resource *res, - const char *file, - int line) +void dlm_lockres_drop_inflight_ref(struct dlm_ctxt *dlm, + struct dlm_lock_resource *res) { assert_spin_locked(&res->spinlock); BUG_ON(res->inflight_locks == 0); + res->inflight_locks--; - mlog(0, "%s:%.*s: inflight--: now %u\n", - dlm->name, res->lockname.len, res->lockname.name, - res->inflight_locks); - if (res->inflight_locks == 0) - dlm_lockres_clear_refmap_bit(dlm->node_num, res); + + mlog(0, "%s: res %.*s, inflight--: now %u, %ps()\n", dlm->name, + res->lockname.len, res->lockname.name, res->inflight_locks, + __builtin_return_address(0)); + wake_up(&res->wq); } +void __dlm_lockres_grab_inflight_worker(struct dlm_ctxt *dlm, + struct dlm_lock_resource *res) +{ + assert_spin_locked(&res->spinlock); + res->inflight_assert_workers++; + mlog(0, "%s:%.*s: inflight assert worker++: now %u\n", + dlm->name, res->lockname.len, res->lockname.name, + res->inflight_assert_workers); +} + +static void dlm_lockres_grab_inflight_worker(struct dlm_ctxt *dlm, + struct dlm_lock_resource *res) +{ + spin_lock(&res->spinlock); + __dlm_lockres_grab_inflight_worker(dlm, res); + spin_unlock(&res->spinlock); +} + +static void __dlm_lockres_drop_inflight_worker(struct dlm_ctxt *dlm, + struct dlm_lock_resource *res) +{ + assert_spin_locked(&res->spinlock); + BUG_ON(res->inflight_assert_workers == 0); + res->inflight_assert_workers--; + mlog(0, "%s:%.*s: inflight assert worker--: now %u\n", + dlm->name, res->lockname.len, res->lockname.name, + res->inflight_assert_workers); +} + +static void dlm_lockres_drop_inflight_worker(struct dlm_ctxt *dlm, + struct dlm_lock_resource *res) +{ + spin_lock(&res->spinlock); + __dlm_lockres_drop_inflight_worker(dlm, res); + spin_unlock(&res->spinlock); +} + /* * lookup a lock resource by name. * may already exist in the hashtable. @@ -706,7 +751,6 @@ struct dlm_lock_resource * dlm_get_lock_resource(struct dlm_ctxt *dlm, unsigned int hash; int tries = 0; int bit, wait_on_recovery = 0; - int drop_inflight_if_nonlocal = 0; BUG_ON(!lockid); @@ -718,36 +762,33 @@ lookup: spin_lock(&dlm->spinlock); tmpres = __dlm_lookup_lockres_full(dlm, lockid, namelen, hash); if (tmpres) { - int dropping_ref = 0; - spin_unlock(&dlm->spinlock); - spin_lock(&tmpres->spinlock); - /* We wait for the other thread that is mastering the resource */ + /* Wait on the thread that is mastering the resource */ if (tmpres->owner == DLM_LOCK_RES_OWNER_UNKNOWN) { __dlm_wait_on_lockres(tmpres); BUG_ON(tmpres->owner == DLM_LOCK_RES_OWNER_UNKNOWN); + spin_unlock(&tmpres->spinlock); + dlm_lockres_put(tmpres); + tmpres = NULL; + goto lookup; } - if (tmpres->owner == dlm->node_num) { - BUG_ON(tmpres->state & DLM_LOCK_RES_DROPPING_REF); - dlm_lockres_grab_inflight_ref(dlm, tmpres); - } else if (tmpres->state & DLM_LOCK_RES_DROPPING_REF) - dropping_ref = 1; - spin_unlock(&tmpres->spinlock); - - /* wait until done messaging the master, drop our ref to allow - * the lockres to be purged, start over. */ - if (dropping_ref) { - spin_lock(&tmpres->spinlock); - __dlm_wait_on_lockres_flags(tmpres, DLM_LOCK_RES_DROPPING_REF); + /* Wait on the resource purge to complete before continuing */ + if (tmpres->state & DLM_LOCK_RES_DROPPING_REF) { + BUG_ON(tmpres->owner == dlm->node_num); + __dlm_wait_on_lockres_flags(tmpres, + DLM_LOCK_RES_DROPPING_REF); spin_unlock(&tmpres->spinlock); dlm_lockres_put(tmpres); tmpres = NULL; goto lookup; } - mlog(0, "found in hash!\n"); + /* Grab inflight ref to pin the resource */ + dlm_lockres_grab_inflight_ref(dlm, tmpres); + + spin_unlock(&tmpres->spinlock); if (res) dlm_lockres_put(res); res = tmpres; @@ -758,8 +799,7 @@ lookup: spin_unlock(&dlm->spinlock); mlog(0, "allocating a new resource\n"); /* nothing found and we need to allocate one. */ - alloc_mle = (struct dlm_master_list_entry *) - kmem_cache_alloc(dlm_mle_cache, GFP_NOFS); + alloc_mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS); if (!alloc_mle) goto leave; res = dlm_new_lockres(dlm, lockid, namelen); @@ -818,7 +858,7 @@ lookup: dlm_mle_detach_hb_events(dlm, mle); dlm_put_mle(mle); mle = NULL; - /* this is lame, but we cant wait on either + /* this is lame, but we can't wait on either * the mle or lockres waitqueue here */ if (mig) msleep(100); @@ -834,13 +874,13 @@ lookup: __dlm_insert_mle(dlm, mle); /* still holding the dlm spinlock, check the recovery map - * to see if there are any nodes that still need to be + * to see if there are any nodes that still need to be * considered. these will not appear in the mle nodemap * but they might own this lockres. wait on them. */ bit = find_next_bit(dlm->recovery_map, O2NM_MAX_NODES, 0); if (bit < O2NM_MAX_NODES) { - mlog(ML_NOTICE, "%s:%.*s: at least one node (%d) to " - "recover before lock mastery can begin\n", + mlog(0, "%s: res %.*s, At least one node (%d) " + "to recover before lock mastery can begin\n", dlm->name, namelen, (char *)lockid, bit); wait_on_recovery = 1; } @@ -853,12 +893,11 @@ lookup: /* finally add the lockres to its hash bucket */ __dlm_insert_lockres(dlm, res); - /* since this lockres is new it doesnt not require the spinlock */ - dlm_lockres_grab_inflight_ref_new(dlm, res); - /* if this node does not become the master make sure to drop - * this inflight reference below */ - drop_inflight_if_nonlocal = 1; + /* Grab inflight ref to pin the resource */ + spin_lock(&res->spinlock); + dlm_lockres_grab_inflight_ref(dlm, res); + spin_unlock(&res->spinlock); /* get an extra ref on the mle in case this is a BLOCK * if so, the creator of the BLOCK may try to put the last @@ -874,8 +913,8 @@ redo_request: * dlm spinlock would be detectable be a change on the mle, * so we only need to clear out the recovery map once. */ if (dlm_is_recovery_lock(lockid, namelen)) { - mlog(ML_NOTICE, "%s: recovery map is not empty, but " - "must master $RECOVERY lock now\n", dlm->name); + mlog(0, "%s: Recovery map is not empty, but must " + "master $RECOVERY lock now\n", dlm->name); if (!dlm_pre_master_reco_lockres(dlm, res)) wait_on_recovery = 0; else { @@ -884,7 +923,7 @@ redo_request: msleep(500); } continue; - } + } dlm_kick_recovery_thread(dlm); msleep(1000); @@ -893,8 +932,8 @@ redo_request: spin_lock(&dlm->spinlock); bit = find_next_bit(dlm->recovery_map, O2NM_MAX_NODES, 0); if (bit < O2NM_MAX_NODES) { - mlog(ML_NOTICE, "%s:%.*s: at least one node (%d) to " - "recover before lock mastery can begin\n", + mlog(0, "%s: res %.*s, At least one node (%d) " + "to recover before lock mastery can begin\n", dlm->name, namelen, (char *)lockid, bit); wait_on_recovery = 1; } else @@ -923,8 +962,8 @@ redo_request: * yet, keep going until it does. this is how the * master will know that asserts are needed back to * the lower nodes. */ - mlog(0, "%s:%.*s: requests only up to %u but master " - "is %u, keep going\n", dlm->name, namelen, + mlog(0, "%s: res %.*s, Requests only up to %u but " + "master is %u, keep going\n", dlm->name, namelen, lockid, nodenum, mle->master); } } @@ -934,14 +973,13 @@ wait: ret = dlm_wait_for_lock_mastery(dlm, res, mle, &blocked); if (ret < 0) { wait_on_recovery = 1; - mlog(0, "%s:%.*s: node map changed, redo the " - "master request now, blocked=%d\n", - dlm->name, res->lockname.len, + mlog(0, "%s: res %.*s, Node map changed, redo the master " + "request now, blocked=%d\n", dlm->name, res->lockname.len, res->lockname.name, blocked); if (++tries > 20) { - mlog(ML_ERROR, "%s:%.*s: spinning on " - "dlm_wait_for_lock_mastery, blocked=%d\n", - dlm->name, res->lockname.len, + mlog(ML_ERROR, "%s: res %.*s, Spinning on " + "dlm_wait_for_lock_mastery, blocked = %d\n", + dlm->name, res->lockname.len, res->lockname.name, blocked); dlm_print_one_lock_resource(res); dlm_print_one_mle(mle); @@ -950,7 +988,8 @@ wait: goto redo_request; } - mlog(0, "lockres mastered by %u\n", res->owner); + mlog(0, "%s: res %.*s, Mastered by %u\n", dlm->name, res->lockname.len, + res->lockname.name, res->owner); /* make sure we never continue without this */ BUG_ON(res->owner == O2NM_MAX_NODES); @@ -962,8 +1001,6 @@ wait: wake_waiters: spin_lock(&res->spinlock); - if (res->owner != dlm->node_num && drop_inflight_if_nonlocal) - dlm_lockres_drop_inflight_ref(dlm, res); res->state &= ~DLM_LOCK_RES_IN_PROGRESS; spin_unlock(&res->spinlock); wake_up(&res->wq); @@ -1030,7 +1067,7 @@ recheck: ret = dlm_restart_lock_mastery(dlm, res, mle, *blocked); b = (mle->type == DLM_MLE_BLOCK); if ((*blocked && !b) || (!*blocked && b)) { - mlog(0, "%s:%.*s: status change: old=%d new=%d\n", + mlog(0, "%s:%.*s: status change: old=%d new=%d\n", dlm->name, res->lockname.len, res->lockname.name, *blocked, b); *blocked = b; @@ -1436,9 +1473,7 @@ way_up_top: } if (res->owner == dlm->node_num) { - mlog(0, "%s:%.*s: setting bit %u in refmap\n", - dlm->name, namelen, name, request->node_idx); - dlm_lockres_set_refmap_bit(request->node_idx, res); + dlm_lockres_set_refmap_bit(dlm, res, request->node_idx); spin_unlock(&res->spinlock); response = DLM_MASTER_RESP_YES; if (mle) @@ -1503,10 +1538,8 @@ way_up_top: * go back and clean the mles on any * other nodes */ dispatch_assert = 1; - dlm_lockres_set_refmap_bit(request->node_idx, res); - mlog(0, "%s:%.*s: setting bit %u in refmap\n", - dlm->name, namelen, name, - request->node_idx); + dlm_lockres_set_refmap_bit(dlm, res, + request->node_idx); } else response = DLM_MASTER_RESP_NO; } else { @@ -1543,8 +1576,7 @@ way_up_top: spin_unlock(&dlm->master_lock); spin_unlock(&dlm->spinlock); - mle = (struct dlm_master_list_entry *) - kmem_cache_alloc(dlm_mle_cache, GFP_NOFS); + mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS); if (!mle) { response = DLM_MASTER_RESP_ERROR; mlog_errno(-ENOMEM); @@ -1603,13 +1635,14 @@ send_response: } mlog(0, "%u is the owner of %.*s, cleaning everyone else\n", dlm->node_num, res->lockname.len, res->lockname.name); - ret = dlm_dispatch_assert_master(dlm, res, 0, request->node_idx, + ret = dlm_dispatch_assert_master(dlm, res, 0, request->node_idx, DLM_ASSERT_MASTER_MLE_CLEANUP); if (ret < 0) { mlog(ML_ERROR, "failed to dispatch assert master work\n"); response = DLM_MASTER_RESP_ERROR; dlm_lockres_put(res); - } + } else + dlm_lockres_grab_inflight_worker(dlm, res); } else { if (res) dlm_lockres_put(res); @@ -1667,7 +1700,9 @@ again: tmpret = o2net_send_message(DLM_ASSERT_MASTER_MSG, dlm->key, &assert, sizeof(assert), to, &r); if (tmpret < 0) { - mlog(0, "assert_master returned %d!\n", tmpret); + mlog(ML_ERROR, "Error %d when sending message %u (key " + "0x%x) to node %u\n", tmpret, + DLM_ASSERT_MASTER_MSG, dlm->key, to); if (!dlm_is_host_down(tmpret)) { mlog(ML_ERROR, "unhandled error=%d!\n", tmpret); BUG(); @@ -1702,7 +1737,7 @@ again: if (r & DLM_ASSERT_RESPONSE_REASSERT) { mlog(0, "%.*s: node %u create mles on other " - "nodes and requests a re-assert\n", + "nodes and requests a re-assert\n", namelen, lockname, to); reassert = 1; } @@ -1711,7 +1746,7 @@ again: "lockres, set the bit in the refmap\n", namelen, lockname, to); spin_lock(&res->spinlock); - dlm_lockres_set_refmap_bit(to, res); + dlm_lockres_set_refmap_bit(dlm, res, to); spin_unlock(&res->spinlock); } } @@ -1813,7 +1848,7 @@ int dlm_assert_master_handler(struct o2net_msg *msg, u32 len, void *data, spin_unlock(&dlm->master_lock); spin_unlock(&dlm->spinlock); goto done; - } + } } } spin_unlock(&dlm->master_lock); @@ -1876,7 +1911,6 @@ int dlm_assert_master_handler(struct o2net_msg *msg, u32 len, void *data, ok: spin_unlock(&res->spinlock); } - spin_unlock(&dlm->spinlock); // mlog(0, "woo! got an assert_master from node %u!\n", // assert->node_idx); @@ -1884,7 +1918,7 @@ ok: int extra_ref = 0; int nn = -1; int rr, err = 0; - + spin_lock(&mle->spinlock); if (mle->type == DLM_MLE_BLOCK || mle->type == DLM_MLE_MIGRATION) extra_ref = 1; @@ -1892,10 +1926,12 @@ ok: /* MASTER mle: if any bits set in the response map * then the calling node needs to re-assert to clear * up nodes that this node contacted */ - while ((nn = find_next_bit (mle->response_map, O2NM_MAX_NODES, + while ((nn = find_next_bit (mle->response_map, O2NM_MAX_NODES, nn+1)) < O2NM_MAX_NODES) { - if (nn != dlm->node_num && nn != assert->node_idx) + if (nn != dlm->node_num && nn != assert->node_idx) { master_request = 1; + break; + } } } mle->master = assert->node_idx; @@ -1927,7 +1963,6 @@ ok: /* master is known, detach if not already detached. * ensures that only one assert_master call will happen * on this mle. */ - spin_lock(&dlm->spinlock); spin_lock(&dlm->master_lock); rr = atomic_read(&mle->mle_refs.refcount); @@ -1960,7 +1995,6 @@ ok: __dlm_put_mle(mle); } spin_unlock(&dlm->master_lock); - spin_unlock(&dlm->spinlock); } else if (res) { if (res->owner != assert->node_idx) { mlog(0, "assert_master from %u, but current " @@ -1968,6 +2002,7 @@ ok: res->owner, namelen, name); } } + spin_unlock(&dlm->spinlock); done: ret = 0; @@ -2003,7 +2038,7 @@ kill: __dlm_print_one_lock_resource(res); spin_unlock(&res->spinlock); spin_unlock(&dlm->spinlock); - *ret_data = (void *)res; + *ret_data = (void *)res; dlm_put(dlm); return -EINVAL; } @@ -2027,7 +2062,7 @@ int dlm_dispatch_assert_master(struct dlm_ctxt *dlm, int ignore_higher, u8 request_from, u32 flags) { struct dlm_work_item *item; - item = kzalloc(sizeof(*item), GFP_NOFS); + item = kzalloc(sizeof(*item), GFP_ATOMIC); if (!item) return -ENOMEM; @@ -2041,10 +2076,10 @@ int dlm_dispatch_assert_master(struct dlm_ctxt *dlm, item->u.am.request_from = request_from; item->u.am.flags = flags; - if (ignore_higher) - mlog(0, "IGNORE HIGHER: %.*s\n", res->lockname.len, + if (ignore_higher) + mlog(0, "IGNORE HIGHER: %.*s\n", res->lockname.len, res->lockname.name); - + spin_lock(&dlm->work_lock); list_add_tail(&item->list, &dlm->work_list); spin_unlock(&dlm->work_lock); @@ -2122,6 +2157,8 @@ static void dlm_assert_master_worker(struct dlm_work_item *item, void *data) dlm_lockres_release_ast(dlm, res); put: + dlm_lockres_drop_inflight_worker(dlm, res); + dlm_lockres_put(res); mlog(0, "finished with dlm_assert_master_worker\n"); @@ -2134,7 +2171,7 @@ put: * think that $RECOVERY is currently mastered by a dead node. If so, * we wait a short time to allow that node to get notified by its own * heartbeat stack, then check again. All $RECOVERY lock resources - * mastered by dead nodes are purged when the hearbeat callback is + * mastered by dead nodes are purged when the hearbeat callback is * fired, so we can know for sure that it is safe to continue once * the node returns a live node or no node. */ static int dlm_pre_master_reco_lockres(struct dlm_ctxt *dlm, @@ -2175,7 +2212,7 @@ static int dlm_pre_master_reco_lockres(struct dlm_ctxt *dlm, ret = -EAGAIN; } spin_unlock(&dlm->spinlock); - mlog(0, "%s: reco lock master is %u\n", dlm->name, + mlog(0, "%s: reco lock master is %u\n", dlm->name, master); break; } @@ -2198,8 +2235,6 @@ int dlm_drop_lockres_ref(struct dlm_ctxt *dlm, struct dlm_lock_resource *res) namelen = res->lockname.len; BUG_ON(namelen > O2NM_MAX_NAME_LEN); - mlog(0, "%s:%.*s: sending deref to %d\n", - dlm->name, namelen, lockname, res->owner); memset(&deref, 0, sizeof(deref)); deref.node_idx = dlm->node_num; deref.namelen = namelen; @@ -2208,12 +2243,12 @@ int dlm_drop_lockres_ref(struct dlm_ctxt *dlm, struct dlm_lock_resource *res) ret = o2net_send_message(DLM_DEREF_LOCKRES_MSG, dlm->key, &deref, sizeof(deref), res->owner, &r); if (ret < 0) - mlog_errno(ret); + mlog(ML_ERROR, "%s: res %.*s, error %d send DEREF to node %u\n", + dlm->name, namelen, lockname, ret, res->owner); else if (r < 0) { /* BAD. other node says I did not have a ref. */ - mlog(ML_ERROR,"while dropping ref on %s:%.*s " - "(master=%u) got %d.\n", dlm->name, namelen, - lockname, res->owner, r); + mlog(ML_ERROR, "%s: res %.*s, DEREF to node %u got %d\n", + dlm->name, namelen, lockname, res->owner, r); dlm_print_one_lock_resource(res); BUG(); } @@ -2269,7 +2304,7 @@ int dlm_deref_lockres_handler(struct o2net_msg *msg, u32 len, void *data, else { BUG_ON(res->state & DLM_LOCK_RES_DROPPING_REF); if (test_bit(node, res->refmap)) { - dlm_lockres_clear_refmap_bit(node, res); + dlm_lockres_clear_refmap_bit(dlm, res, node); cleared = 1; } } @@ -2329,7 +2364,7 @@ static void dlm_deref_lockres_worker(struct dlm_work_item *item, void *data) BUG_ON(res->state & DLM_LOCK_RES_DROPPING_REF); if (test_bit(node, res->refmap)) { __dlm_wait_on_lockres_flags(res, DLM_LOCK_RES_SETREF_INPROG); - dlm_lockres_clear_refmap_bit(node, res); + dlm_lockres_clear_refmap_bit(dlm, res, node); cleared = 1; } spin_unlock(&res->spinlock); @@ -2348,55 +2383,59 @@ static void dlm_deref_lockres_worker(struct dlm_work_item *item, void *data) dlm_lockres_put(res); } -/* Checks whether the lockres can be migrated. Returns 0 if yes, < 0 - * if not. If 0, numlocks is set to the number of locks in the lockres. +/* + * A migrateable resource is one that is : + * 1. locally mastered, and, + * 2. zero local locks, and, + * 3. one or more non-local locks, or, one or more references + * Returns 1 if yes, 0 if not. */ static int dlm_is_lockres_migrateable(struct dlm_ctxt *dlm, - struct dlm_lock_resource *res, - int *numlocks) + struct dlm_lock_resource *res) { - int ret; - int i; - int count = 0; + enum dlm_lockres_list idx; + int nonlocal = 0, node_ref; struct list_head *queue; struct dlm_lock *lock; + u64 cookie; assert_spin_locked(&res->spinlock); - ret = -EINVAL; - if (res->owner == DLM_LOCK_RES_OWNER_UNKNOWN) { - mlog(0, "cannot migrate lockres with unknown owner!\n"); - goto leave; - } + /* delay migration when the lockres is in MIGRATING state */ + if (res->state & DLM_LOCK_RES_MIGRATING) + return 0; - if (res->owner != dlm->node_num) { - mlog(0, "cannot migrate lockres this node doesn't own!\n"); - goto leave; - } + if (res->owner != dlm->node_num) + return 0; - ret = 0; - queue = &res->granted; - for (i = 0; i < 3; i++) { + for (idx = DLM_GRANTED_LIST; idx <= DLM_BLOCKED_LIST; idx++) { + queue = dlm_list_idx_to_ptr(res, idx); list_for_each_entry(lock, queue, list) { - ++count; - if (lock->ml.node == dlm->node_num) { - mlog(0, "found a lock owned by this node still " - "on the %s queue! will not migrate this " - "lockres\n", (i == 0 ? "granted" : - (i == 1 ? "converting" : - "blocked"))); - ret = -ENOTEMPTY; - goto leave; + if (lock->ml.node != dlm->node_num) { + nonlocal++; + continue; } + cookie = be64_to_cpu(lock->ml.cookie); + mlog(0, "%s: Not migrateable res %.*s, lock %u:%llu on " + "%s list\n", dlm->name, res->lockname.len, + res->lockname.name, + dlm_get_lock_cookie_node(cookie), + dlm_get_lock_cookie_seq(cookie), + dlm_list_in_text(idx)); + return 0; } - queue++; } - *numlocks = count; - mlog(0, "migrateable lockres having %d locks\n", *numlocks); + if (!nonlocal) { + node_ref = find_next_bit(res->refmap, O2NM_MAX_NODES, 0); + if (node_ref >= O2NM_MAX_NODES) + return 0; + } -leave: - return ret; + mlog(0, "%s: res %.*s, Migrateable\n", dlm->name, res->lockname.len, + res->lockname.name); + + return 1; } /* @@ -2405,8 +2444,7 @@ leave: static int dlm_migrate_lockres(struct dlm_ctxt *dlm, - struct dlm_lock_resource *res, - u8 target) + struct dlm_lock_resource *res, u8 target) { struct dlm_master_list_entry *mle = NULL; struct dlm_master_list_entry *oldmle = NULL; @@ -2415,39 +2453,20 @@ static int dlm_migrate_lockres(struct dlm_ctxt *dlm, const char *name; unsigned int namelen; int mle_added = 0; - int numlocks; int wake = 0; if (!dlm_grab(dlm)) return -EINVAL; + BUG_ON(target == O2NM_MAX_NODES); + name = res->lockname.name; namelen = res->lockname.len; - mlog(0, "migrating %.*s to %u\n", namelen, name, target); - - /* - * ensure this lockres is a proper candidate for migration - */ - spin_lock(&res->spinlock); - ret = dlm_is_lockres_migrateable(dlm, res, &numlocks); - if (ret < 0) { - spin_unlock(&res->spinlock); - goto leave; - } - spin_unlock(&res->spinlock); - - /* no work to do */ - if (numlocks == 0) { - mlog(0, "no locks were found on this lockres! done!\n"); - goto leave; - } - - /* - * preallocate up front - * if this fails, abort - */ + mlog(0, "%s: Migrating %.*s to node %u\n", dlm->name, namelen, name, + target); + /* preallocate up front. if this fails, abort */ ret = -ENOMEM; mres = (struct dlm_migratable_lockres *) __get_free_page(GFP_NOFS); if (!mres) { @@ -2455,8 +2474,7 @@ static int dlm_migrate_lockres(struct dlm_ctxt *dlm, goto leave; } - mle = (struct dlm_master_list_entry *) kmem_cache_alloc(dlm_mle_cache, - GFP_NOFS); + mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS); if (!mle) { mlog_errno(ret); goto leave; @@ -2464,35 +2482,10 @@ static int dlm_migrate_lockres(struct dlm_ctxt *dlm, ret = 0; /* - * find a node to migrate the lockres to - */ - - mlog(0, "picking a migration node\n"); - spin_lock(&dlm->spinlock); - /* pick a new node */ - if (!test_bit(target, dlm->domain_map) || - target >= O2NM_MAX_NODES) { - target = dlm_pick_migration_target(dlm, res); - } - mlog(0, "node %u chosen for migration\n", target); - - if (target >= O2NM_MAX_NODES || - !test_bit(target, dlm->domain_map)) { - /* target chosen is not alive */ - ret = -EINVAL; - } - - if (ret) { - spin_unlock(&dlm->spinlock); - goto fail; - } - - mlog(0, "continuing with target = %u\n", target); - - /* * clear any existing master requests and * add the migration mle to the list */ + spin_lock(&dlm->spinlock); spin_lock(&dlm->master_lock); ret = dlm_add_migration_mle(dlm, res, mle, &oldmle, name, namelen, target, dlm->node_num); @@ -2533,6 +2526,7 @@ fail: dlm_put_mle(mle); } else if (mle) { kmem_cache_free(dlm_mle_cache, mle); + mle = NULL; } goto leave; } @@ -2576,6 +2570,9 @@ fail: res->state &= ~DLM_LOCK_RES_MIGRATING; wake = 1; spin_unlock(&res->spinlock); + if (dlm_is_host_down(ret)) + dlm_wait_for_node_death(dlm, target, + DLM_NODE_DEATH_WAIT_MAX); goto leave; } @@ -2587,7 +2584,7 @@ fail: * is complete everywhere. if the target dies while this is * going on, some nodes could potentially see the target as the * master, so it is important that my recovery finds the migration - * mle and sets the master to UNKNONWN. */ + * mle and sets the master to UNKNOWN. */ /* wait for new node to assert master */ @@ -2603,7 +2600,7 @@ fail: mlog(0, "%s:%.*s: timed out during migration\n", dlm->name, res->lockname.len, res->lockname.name); - /* avoid hang during shutdown when migrating lockres + /* avoid hang during shutdown when migrating lockres * to a node which also goes down */ if (dlm_is_node_dead(dlm, target)) { mlog(0, "%s:%.*s: expected migration " @@ -2651,69 +2648,52 @@ leave: if (wake) wake_up(&res->wq); - /* TODO: cleanup */ if (mres) free_page((unsigned long)mres); dlm_put(dlm); - mlog(0, "returning %d\n", ret); + mlog(0, "%s: Migrating %.*s to %u, returns %d\n", dlm->name, namelen, + name, target, ret); return ret; } #define DLM_MIGRATION_RETRY_MS 100 -/* Should be called only after beginning the domain leave process. +/* + * Should be called only after beginning the domain leave process. * There should not be any remaining locks on nonlocal lock resources, * and there should be no local locks left on locally mastered resources. * * Called with the dlm spinlock held, may drop it to do migration, but * will re-acquire before exit. * - * Returns: 1 if dlm->spinlock was dropped/retaken, 0 if never dropped */ + * Returns: 1 if dlm->spinlock was dropped/retaken, 0 if never dropped + */ int dlm_empty_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res) { int ret; int lock_dropped = 0; - int numlocks; + u8 target = O2NM_MAX_NODES; + + assert_spin_locked(&dlm->spinlock); spin_lock(&res->spinlock); - if (res->owner != dlm->node_num) { - if (!__dlm_lockres_unused(res)) { - mlog(ML_ERROR, "%s:%.*s: this node is not master, " - "trying to free this but locks remain\n", - dlm->name, res->lockname.len, res->lockname.name); - } - spin_unlock(&res->spinlock); - goto leave; - } + if (dlm_is_lockres_migrateable(dlm, res)) + target = dlm_pick_migration_target(dlm, res); + spin_unlock(&res->spinlock); - /* No need to migrate a lockres having no locks */ - ret = dlm_is_lockres_migrateable(dlm, res, &numlocks); - if (ret >= 0 && numlocks == 0) { - spin_unlock(&res->spinlock); + if (target == O2NM_MAX_NODES) goto leave; - } - spin_unlock(&res->spinlock); /* Wheee! Migrate lockres here! Will sleep so drop spinlock. */ spin_unlock(&dlm->spinlock); lock_dropped = 1; - while (1) { - ret = dlm_migrate_lockres(dlm, res, O2NM_MAX_NODES); - if (ret >= 0) - break; - if (ret == -ENOTEMPTY) { - mlog(ML_ERROR, "lockres %.*s still has local locks!\n", - res->lockname.len, res->lockname.name); - BUG(); - } - - mlog(0, "lockres %.*s: migrate failed, " - "retrying\n", res->lockname.len, - res->lockname.name); - msleep(DLM_MIGRATION_RETRY_MS); - } + ret = dlm_migrate_lockres(dlm, res, target); + if (ret) + mlog(0, "%s: res %.*s, Migrate to node %u failed with %d\n", + dlm->name, res->lockname.len, res->lockname.name, + target, ret); spin_lock(&dlm->spinlock); leave: return lock_dropped; @@ -2739,7 +2719,7 @@ static int dlm_migration_can_proceed(struct dlm_ctxt *dlm, can_proceed = !!(res->state & DLM_LOCK_RES_MIGRATING); spin_unlock(&res->spinlock); - /* target has died, so make the caller break out of the + /* target has died, so make the caller break out of the * wait_event, but caller must recheck the domain_map */ spin_lock(&dlm->spinlock); if (!test_bit(mig_target, dlm->domain_map)) @@ -2812,14 +2792,8 @@ again: mlog(0, "trying again...\n"); goto again; } - /* now that we are sure the MIGRATING state is there, drop - * the unneded state which blocked threads trying to DIRTY */ - spin_lock(&res->spinlock); - BUG_ON(!(res->state & DLM_LOCK_RES_BLOCK_DIRTY)); - BUG_ON(!(res->state & DLM_LOCK_RES_MIGRATING)); - res->state &= ~DLM_LOCK_RES_BLOCK_DIRTY; - spin_unlock(&res->spinlock); + ret = 0; /* did the target go down or die? */ spin_lock(&dlm->spinlock); if (!test_bit(target, dlm->domain_map)) { @@ -2830,9 +2804,21 @@ again: spin_unlock(&dlm->spinlock); /* + * if target is down, we need to clear DLM_LOCK_RES_BLOCK_DIRTY for + * another try; otherwise, we are sure the MIGRATING state is there, + * drop the unneded state which blocked threads trying to DIRTY + */ + spin_lock(&res->spinlock); + BUG_ON(!(res->state & DLM_LOCK_RES_BLOCK_DIRTY)); + res->state &= ~DLM_LOCK_RES_BLOCK_DIRTY; + if (!ret) + BUG_ON(!(res->state & DLM_LOCK_RES_MIGRATING)); + spin_unlock(&res->spinlock); + + /* * at this point: * - * o the DLM_LOCK_RES_MIGRATING flag is set + * o the DLM_LOCK_RES_MIGRATING flag is set if target not down * o there are no pending asts on this lockres * o all processes trying to reserve an ast on this * lockres must wait for the MIGRATING flag to clear @@ -2864,7 +2850,8 @@ static void dlm_remove_nonlocal_locks(struct dlm_ctxt *dlm, BUG_ON(!list_empty(&lock->bast_list)); BUG_ON(lock->ast_pending); BUG_ON(lock->bast_pending); - dlm_lockres_clear_refmap_bit(lock->ml.node, res); + dlm_lockres_clear_refmap_bit(dlm, res, + lock->ml.node); list_del_init(&lock->list); dlm_lock_put(lock); /* In a normal unlock, we would have added a @@ -2885,61 +2872,61 @@ static void dlm_remove_nonlocal_locks(struct dlm_ctxt *dlm, mlog(0, "%s:%.*s: node %u had a ref to this " "migrating lockres, clearing\n", dlm->name, res->lockname.len, res->lockname.name, bit); - dlm_lockres_clear_refmap_bit(bit, res); + dlm_lockres_clear_refmap_bit(dlm, res, bit); } bit++; } } -/* for now this is not too intelligent. we will - * need stats to make this do the right thing. - * this just finds the first lock on one of the - * queues and uses that node as the target. */ +/* + * Pick a node to migrate the lock resource to. This function selects a + * potential target based first on the locks and then on refmap. It skips + * nodes that are in the process of exiting the domain. + */ static u8 dlm_pick_migration_target(struct dlm_ctxt *dlm, struct dlm_lock_resource *res) { - int i; + enum dlm_lockres_list idx; struct list_head *queue = &res->granted; struct dlm_lock *lock; - int nodenum; + int noderef; + u8 nodenum = O2NM_MAX_NODES; assert_spin_locked(&dlm->spinlock); + assert_spin_locked(&res->spinlock); - spin_lock(&res->spinlock); - for (i=0; i<3; i++) { + /* Go through all the locks */ + for (idx = DLM_GRANTED_LIST; idx <= DLM_BLOCKED_LIST; idx++) { + queue = dlm_list_idx_to_ptr(res, idx); list_for_each_entry(lock, queue, list) { - /* up to the caller to make sure this node - * is alive */ - if (lock->ml.node != dlm->node_num) { - spin_unlock(&res->spinlock); - return lock->ml.node; - } + if (lock->ml.node == dlm->node_num) + continue; + if (test_bit(lock->ml.node, dlm->exit_domain_map)) + continue; + nodenum = lock->ml.node; + goto bail; } - queue++; } - spin_unlock(&res->spinlock); - mlog(0, "have not found a suitable target yet! checking domain map\n"); - /* ok now we're getting desperate. pick anyone alive. */ - nodenum = -1; + /* Go thru the refmap */ + noderef = -1; while (1) { - nodenum = find_next_bit(dlm->domain_map, - O2NM_MAX_NODES, nodenum+1); - mlog(0, "found %d in domain map\n", nodenum); - if (nodenum >= O2NM_MAX_NODES) + noderef = find_next_bit(res->refmap, O2NM_MAX_NODES, + noderef + 1); + if (noderef >= O2NM_MAX_NODES) break; - if (nodenum != dlm->node_num) { - mlog(0, "picking %d\n", nodenum); - return nodenum; - } + if (noderef == dlm->node_num) + continue; + if (test_bit(noderef, dlm->exit_domain_map)) + continue; + nodenum = noderef; + goto bail; } - mlog(0, "giving up. no master to migrate to\n"); - return DLM_LOCK_RES_OWNER_UNKNOWN; +bail: + return nodenum; } - - /* this is called by the new master once all lockres * data has been received */ static int dlm_do_migrate_request(struct dlm_ctxt *dlm, @@ -2978,7 +2965,9 @@ static int dlm_do_migrate_request(struct dlm_ctxt *dlm, &migrate, sizeof(migrate), nodenum, &status); if (ret < 0) { - mlog(0, "migrate_request returned %d!\n", ret); + mlog(ML_ERROR, "%s: res %.*s, Error %d send " + "MIGRATE_REQUEST to node %u\n", dlm->name, + migrate.namelen, migrate.name, ret, nodenum); if (!dlm_is_host_down(ret)) { mlog(ML_ERROR, "unhandled error=%d!\n", ret); BUG(); @@ -2997,7 +2986,7 @@ static int dlm_do_migrate_request(struct dlm_ctxt *dlm, dlm->name, res->lockname.len, res->lockname.name, nodenum); spin_lock(&res->spinlock); - dlm_lockres_set_refmap_bit(nodenum, res); + dlm_lockres_set_refmap_bit(dlm, res, nodenum); spin_unlock(&res->spinlock); } } @@ -3036,8 +3025,7 @@ int dlm_migrate_request_handler(struct o2net_msg *msg, u32 len, void *data, hash = dlm_lockid_hash(name, namelen); /* preallocate.. if this fails, abort */ - mle = (struct dlm_master_list_entry *) kmem_cache_alloc(dlm_mle_cache, - GFP_NOFS); + mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS); if (!mle) { ret = -ENOMEM; @@ -3047,8 +3035,6 @@ int dlm_migrate_request_handler(struct o2net_msg *msg, u32 len, void *data, /* check for pre-existing lock */ spin_lock(&dlm->spinlock); res = __dlm_lookup_lockres(dlm, name, namelen, hash); - spin_lock(&dlm->master_lock); - if (res) { spin_lock(&res->spinlock); if (res->state & DLM_LOCK_RES_RECOVERING) { @@ -3066,14 +3052,15 @@ int dlm_migrate_request_handler(struct o2net_msg *msg, u32 len, void *data, spin_unlock(&res->spinlock); } + spin_lock(&dlm->master_lock); /* ignore status. only nonzero status would BUG. */ ret = dlm_add_migration_mle(dlm, res, mle, &oldmle, name, namelen, migrate->new_master, migrate->master); -unlock: spin_unlock(&dlm->master_lock); +unlock: spin_unlock(&dlm->spinlock); if (oldmle) { @@ -3108,8 +3095,6 @@ static int dlm_add_migration_mle(struct dlm_ctxt *dlm, *oldmle = NULL; - mlog_entry_void(); - assert_spin_locked(&dlm->spinlock); assert_spin_locked(&dlm->master_lock); @@ -3144,11 +3129,15 @@ static int dlm_add_migration_mle(struct dlm_ctxt *dlm, /* remove it so that only one mle will be found */ __dlm_unlink_mle(dlm, tmp); __dlm_mle_detach_hb_events(dlm, tmp); - ret = DLM_MIGRATE_RESPONSE_MASTERY_REF; - mlog(0, "%s:%.*s: master=%u, newmaster=%u, " - "telling master to get ref for cleared out mle " - "during migration\n", dlm->name, namelen, name, - master, new_master); + if (tmp->type == DLM_MLE_MASTER) { + ret = DLM_MIGRATE_RESPONSE_MASTERY_REF; + mlog(0, "%s:%.*s: master=%u, newmaster=%u, " + "telling master to get ref " + "for cleared out mle during " + "migration\n", dlm->name, + namelen, name, master, + new_master); + } } spin_unlock(&tmp->spinlock); } @@ -3246,10 +3235,10 @@ void dlm_clean_master_list(struct dlm_ctxt *dlm, u8 dead_node) struct dlm_master_list_entry *mle; struct dlm_lock_resource *res; struct hlist_head *bucket; - struct hlist_node *list; + struct hlist_node *tmp; unsigned int i; - mlog_entry("dlm=%s, dead node=%u\n", dlm->name, dead_node); + mlog(0, "dlm=%s, dead node=%u\n", dlm->name, dead_node); top: assert_spin_locked(&dlm->spinlock); @@ -3257,10 +3246,7 @@ top: spin_lock(&dlm->master_lock); for (i = 0; i < DLM_HASH_BUCKETS; i++) { bucket = dlm_master_hash(dlm, i); - hlist_for_each(list, bucket) { - mle = hlist_entry(list, struct dlm_master_list_entry, - master_hash_node); - + hlist_for_each_entry_safe(mle, tmp, bucket, master_hash_node) { BUG_ON(mle->type != DLM_MLE_BLOCK && mle->type != DLM_MLE_MASTER && mle->type != DLM_MLE_MIGRATION); @@ -3335,7 +3321,7 @@ int dlm_finish_migration(struct dlm_ctxt *dlm, struct dlm_lock_resource *res, * mastery reference here since old_master will briefly have * a reference after the migration completes */ spin_lock(&res->spinlock); - dlm_lockres_set_refmap_bit(old_master, res); + dlm_lockres_set_refmap_bit(dlm, res, old_master); spin_unlock(&res->spinlock); mlog(0, "now time to do a migrate request to other nodes\n"); @@ -3435,3 +3421,41 @@ void dlm_lockres_release_ast(struct dlm_ctxt *dlm, wake_up(&res->wq); wake_up(&dlm->migration_wq); } + +void dlm_force_free_mles(struct dlm_ctxt *dlm) +{ + int i; + struct hlist_head *bucket; + struct dlm_master_list_entry *mle; + struct hlist_node *tmp; + + /* + * We notified all other nodes that we are exiting the domain and + * marked the dlm state to DLM_CTXT_LEAVING. If any mles are still + * around we force free them and wake any processes that are waiting + * on the mles + */ + spin_lock(&dlm->spinlock); + spin_lock(&dlm->master_lock); + + BUG_ON(dlm->dlm_state != DLM_CTXT_LEAVING); + BUG_ON((find_next_bit(dlm->domain_map, O2NM_MAX_NODES, 0) < O2NM_MAX_NODES)); + + for (i = 0; i < DLM_HASH_BUCKETS; i++) { + bucket = dlm_master_hash(dlm, i); + hlist_for_each_entry_safe(mle, tmp, bucket, master_hash_node) { + if (mle->type != DLM_MLE_BLOCK) { + mlog(ML_ERROR, "bad mle: %p\n", mle); + dlm_print_one_mle(mle); + } + atomic_set(&mle->woken, 1); + wake_up(&mle->wq); + + __dlm_unlink_mle(dlm, mle); + __dlm_mle_detach_hb_events(dlm, mle); + __dlm_put_mle(mle); + } + } + spin_unlock(&dlm->master_lock); + spin_unlock(&dlm->spinlock); +} diff --git a/fs/ocfs2/dlm/dlmrecovery.c b/fs/ocfs2/dlm/dlmrecovery.c index bcb9260c373..45067faf569 100644 --- a/fs/ocfs2/dlm/dlmrecovery.c +++ b/fs/ocfs2/dlm/dlmrecovery.c @@ -30,7 +30,6 @@ #include <linux/types.h> #include <linux/slab.h> #include <linux/highmem.h> -#include <linux/utsname.h> #include <linux/init.h> #include <linux/sysctl.h> #include <linux/random.h> @@ -56,9 +55,6 @@ static void dlm_do_local_recovery_cleanup(struct dlm_ctxt *dlm, u8 dead_node); static int dlm_recovery_thread(void *data); -void dlm_complete_recovery_thread(struct dlm_ctxt *dlm); -int dlm_launch_recovery_thread(struct dlm_ctxt *dlm); -void dlm_kick_recovery_thread(struct dlm_ctxt *dlm); static int dlm_do_recovery(struct dlm_ctxt *dlm); static int dlm_pick_recovery_master(struct dlm_ctxt *dlm); @@ -311,7 +307,7 @@ static int dlm_recovery_thread(void *data) mlog(0, "dlm thread running for %s...\n", dlm->name); while (!kthread_should_stop()) { - if (dlm_joined(dlm)) { + if (dlm_domain_fully_joined(dlm)) { status = dlm_do_recovery(dlm); if (status == -EAGAIN) { /* do not sleep, recheck immediately. */ @@ -363,40 +359,38 @@ static int dlm_is_node_recovered(struct dlm_ctxt *dlm, u8 node) } -int dlm_wait_for_node_death(struct dlm_ctxt *dlm, u8 node, int timeout) +void dlm_wait_for_node_death(struct dlm_ctxt *dlm, u8 node, int timeout) { - if (timeout) { - mlog(ML_NOTICE, "%s: waiting %dms for notification of " - "death of node %u\n", dlm->name, timeout, node); + if (dlm_is_node_dead(dlm, node)) + return; + + printk(KERN_NOTICE "o2dlm: Waiting on the death of node %u in " + "domain %s\n", node, dlm->name); + + if (timeout) wait_event_timeout(dlm->dlm_reco_thread_wq, - dlm_is_node_dead(dlm, node), - msecs_to_jiffies(timeout)); - } else { - mlog(ML_NOTICE, "%s: waiting indefinitely for notification " - "of death of node %u\n", dlm->name, node); + dlm_is_node_dead(dlm, node), + msecs_to_jiffies(timeout)); + else wait_event(dlm->dlm_reco_thread_wq, dlm_is_node_dead(dlm, node)); - } - /* for now, return 0 */ - return 0; } -int dlm_wait_for_node_recovery(struct dlm_ctxt *dlm, u8 node, int timeout) +void dlm_wait_for_node_recovery(struct dlm_ctxt *dlm, u8 node, int timeout) { - if (timeout) { - mlog(0, "%s: waiting %dms for notification of " - "recovery of node %u\n", dlm->name, timeout, node); + if (dlm_is_node_recovered(dlm, node)) + return; + + printk(KERN_NOTICE "o2dlm: Waiting on the recovery of node %u in " + "domain %s\n", node, dlm->name); + + if (timeout) wait_event_timeout(dlm->dlm_reco_thread_wq, - dlm_is_node_recovered(dlm, node), - msecs_to_jiffies(timeout)); - } else { - mlog(0, "%s: waiting indefinitely for notification " - "of recovery of node %u\n", dlm->name, node); + dlm_is_node_recovered(dlm, node), + msecs_to_jiffies(timeout)); + else wait_event(dlm->dlm_reco_thread_wq, dlm_is_node_recovered(dlm, node)); - } - /* for now, return 0 */ - return 0; } /* callers of the top-level api calls (dlmlock/dlmunlock) should @@ -431,6 +425,8 @@ static void dlm_begin_recovery(struct dlm_ctxt *dlm) { spin_lock(&dlm->spinlock); BUG_ON(dlm->reco.state & DLM_RECO_STATE_ACTIVE); + printk(KERN_NOTICE "o2dlm: Begin recovery on domain %s for node %u\n", + dlm->name, dlm->reco.dead_node); dlm->reco.state |= DLM_RECO_STATE_ACTIVE; spin_unlock(&dlm->spinlock); } @@ -441,9 +437,18 @@ static void dlm_end_recovery(struct dlm_ctxt *dlm) BUG_ON(!(dlm->reco.state & DLM_RECO_STATE_ACTIVE)); dlm->reco.state &= ~DLM_RECO_STATE_ACTIVE; spin_unlock(&dlm->spinlock); + printk(KERN_NOTICE "o2dlm: End recovery on domain %s\n", dlm->name); wake_up(&dlm->reco.event); } +static void dlm_print_recovery_master(struct dlm_ctxt *dlm) +{ + printk(KERN_NOTICE "o2dlm: Node %u (%s) is the Recovery Master for the " + "dead node %u in domain %s\n", dlm->reco.new_master, + (dlm->node_num == dlm->reco.new_master ? "me" : "he"), + dlm->reco.dead_node, dlm->name); +} + static int dlm_do_recovery(struct dlm_ctxt *dlm) { int status = 0; @@ -464,7 +469,7 @@ static int dlm_do_recovery(struct dlm_ctxt *dlm) if (dlm->reco.dead_node == O2NM_INVALID_NODE_NUM) { int bit; - bit = find_next_bit (dlm->recovery_map, O2NM_MAX_NODES+1, 0); + bit = find_next_bit (dlm->recovery_map, O2NM_MAX_NODES, 0); if (bit >= O2NM_MAX_NODES || bit < 0) dlm_set_reco_dead_node(dlm, O2NM_INVALID_NODE_NUM); else @@ -506,9 +511,8 @@ static int dlm_do_recovery(struct dlm_ctxt *dlm) } mlog(0, "another node will master this recovery session.\n"); } - mlog(0, "dlm=%s (%d), new_master=%u, this node=%u, dead_node=%u\n", - dlm->name, task_pid_nr(dlm->dlm_reco_thread_task), dlm->reco.new_master, - dlm->node_num, dlm->reco.dead_node); + + dlm_print_recovery_master(dlm); /* it is safe to start everything back up here * because all of the dead node's lock resources @@ -519,15 +523,13 @@ static int dlm_do_recovery(struct dlm_ctxt *dlm) return 0; master_here: - mlog(ML_NOTICE, "(%d) Node %u is the Recovery Master for the Dead Node " - "%u for Domain %s\n", task_pid_nr(dlm->dlm_reco_thread_task), - dlm->node_num, dlm->reco.dead_node, dlm->name); + dlm_print_recovery_master(dlm); status = dlm_remaster_locks(dlm, dlm->reco.dead_node); if (status < 0) { /* we should never hit this anymore */ - mlog(ML_ERROR, "error %d remastering locks for node %u, " - "retrying.\n", status, dlm->reco.dead_node); + mlog(ML_ERROR, "%s: Error %d remastering locks for node %u, " + "retrying.\n", dlm->name, status, dlm->reco.dead_node); /* yield a bit to allow any final network messages * to get handled on remaining nodes */ msleep(100); @@ -535,7 +537,10 @@ master_here: /* success! see if any other nodes need recovery */ mlog(0, "DONE mastering recovery of %s:%u here(this=%u)!\n", dlm->name, dlm->reco.dead_node, dlm->node_num); - dlm_reset_recovery(dlm); + spin_lock(&dlm->spinlock); + __dlm_reset_recovery(dlm); + dlm->reco.state &= ~DLM_RECO_STATE_FINALIZE; + spin_unlock(&dlm->spinlock); } dlm_end_recovery(dlm); @@ -568,7 +573,7 @@ static int dlm_remaster_locks(struct dlm_ctxt *dlm, u8 dead_node) BUG_ON(ndata->state != DLM_RECO_NODE_DATA_INIT); ndata->state = DLM_RECO_NODE_DATA_REQUESTING; - mlog(0, "requesting lock info from node %u\n", + mlog(0, "%s: Requesting lock info from node %u\n", dlm->name, ndata->node_num); if (ndata->node_num == dlm->node_num) { @@ -641,7 +646,7 @@ static int dlm_remaster_locks(struct dlm_ctxt *dlm, u8 dead_node) spin_unlock(&dlm_reco_state_lock); } - mlog(0, "done requesting all lock info\n"); + mlog(0, "%s: Done requesting all lock info\n", dlm->name); /* nodes should be sending reco data now * just need to wait */ @@ -693,6 +698,14 @@ static int dlm_remaster_locks(struct dlm_ctxt *dlm, u8 dead_node) if (all_nodes_done) { int ret; + /* Set this flag on recovery master to avoid + * a new recovery for another dead node start + * before the recovery is not done. That may + * cause recovery hung.*/ + spin_lock(&dlm->spinlock); + dlm->reco.state |= DLM_RECO_STATE_FINALIZE; + spin_unlock(&dlm->spinlock); + /* all nodes are now in DLM_RECO_NODE_DATA_DONE state * just send a finalize message to everyone and * clean up */ @@ -728,7 +741,6 @@ static int dlm_remaster_locks(struct dlm_ctxt *dlm, u8 dead_node) if (destroy) dlm_destroy_recovery_area(dlm, dead_node); - mlog_exit(status); return status; } @@ -785,7 +797,8 @@ static int dlm_request_all_locks(struct dlm_ctxt *dlm, u8 request_from, u8 dead_node) { struct dlm_lock_request lr; - enum dlm_status ret; + int ret; + int status; mlog(0, "\n"); @@ -798,14 +811,16 @@ static int dlm_request_all_locks(struct dlm_ctxt *dlm, u8 request_from, lr.dead_node = dead_node; // send message - ret = DLM_NOLOCKMGR; ret = o2net_send_message(DLM_LOCK_REQUEST_MSG, dlm->key, - &lr, sizeof(lr), request_from, NULL); + &lr, sizeof(lr), request_from, &status); /* negative status is handled by caller */ if (ret < 0) - mlog_errno(ret); - + mlog(ML_ERROR, "%s: Error %d send LOCK_REQUEST to node %u " + "to recover dead node %u\n", dlm->name, ret, + request_from, dead_node); + else + ret = status; // return from here, then // sleep until all received or error return ret; @@ -956,10 +971,10 @@ static int dlm_send_all_done_msg(struct dlm_ctxt *dlm, u8 dead_node, u8 send_to) ret = o2net_send_message(DLM_RECO_DATA_DONE_MSG, dlm->key, &done_msg, sizeof(done_msg), send_to, &tmpret); if (ret < 0) { + mlog(ML_ERROR, "%s: Error %d send RECO_DATA_DONE to node %u " + "to recover dead node %u\n", dlm->name, ret, send_to, + dead_node); if (!dlm_is_host_down(ret)) { - mlog_errno(ret); - mlog(ML_ERROR, "%s: unknown error sending data-done " - "to %u\n", dlm->name, send_to); BUG(); } } else @@ -1051,7 +1066,7 @@ static void dlm_move_reco_locks_to_list(struct dlm_ctxt *dlm, if (lock->ml.node == dead_node) { mlog(0, "AHA! there was " "a $RECOVERY lock for dead " - "node %u (%s)!\n", + "node %u (%s)!\n", dead_node, dlm->name); list_del_init(&lock->list); dlm_lock_put(lock); @@ -1118,7 +1133,7 @@ static int dlm_send_mig_lockres_msg(struct dlm_ctxt *dlm, mlog(0, "%s:%.*s: sending mig lockres (%s) to %u\n", dlm->name, res->lockname.len, res->lockname.name, - orig_flags & DLM_MRES_MIGRATION ? "migrate" : "recovery", + orig_flags & DLM_MRES_MIGRATION ? "migration" : "recovery", send_to); /* send it */ @@ -1127,7 +1142,11 @@ static int dlm_send_mig_lockres_msg(struct dlm_ctxt *dlm, if (ret < 0) { /* XXX: negative status is not handled. * this will end up killing this node. */ - mlog_errno(ret); + mlog(ML_ERROR, "%s: res %.*s, Error %d send MIG_LOCKRES to " + "node %u (%s)\n", dlm->name, mres->lockname_len, + mres->lockname, ret, send_to, + (orig_flags & DLM_MRES_MIGRATION ? + "migration" : "recovery")); } else { /* might get an -ENOMEM back here */ ret = status; @@ -1165,6 +1184,39 @@ static void dlm_init_migratable_lockres(struct dlm_migratable_lockres *mres, mres->master = master; } +static void dlm_prepare_lvb_for_migration(struct dlm_lock *lock, + struct dlm_migratable_lockres *mres, + int queue) +{ + if (!lock->lksb) + return; + + /* Ignore lvb in all locks in the blocked list */ + if (queue == DLM_BLOCKED_LIST) + return; + + /* Only consider lvbs in locks with granted EX or PR lock levels */ + if (lock->ml.type != LKM_EXMODE && lock->ml.type != LKM_PRMODE) + return; + + if (dlm_lvb_is_empty(mres->lvb)) { + memcpy(mres->lvb, lock->lksb->lvb, DLM_LVB_LEN); + return; + } + + /* Ensure the lvb copied for migration matches in other valid locks */ + if (!memcmp(mres->lvb, lock->lksb->lvb, DLM_LVB_LEN)) + return; + + mlog(ML_ERROR, "Mismatched lvb in lock cookie=%u:%llu, name=%.*s, " + "node=%u\n", + dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)), + dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)), + lock->lockres->lockname.len, lock->lockres->lockname.name, + lock->ml.node); + dlm_print_one_lock_resource(lock->lockres); + BUG(); +} /* returns 1 if this lock fills the network structure, * 0 otherwise */ @@ -1182,20 +1234,7 @@ static int dlm_add_lock_to_array(struct dlm_lock *lock, ml->list = queue; if (lock->lksb) { ml->flags = lock->lksb->flags; - /* send our current lvb */ - if (ml->type == LKM_EXMODE || - ml->type == LKM_PRMODE) { - /* if it is already set, this had better be a PR - * and it has to match */ - if (!dlm_lvb_is_empty(mres->lvb) && - (ml->type == LKM_EXMODE || - memcmp(mres->lvb, lock->lksb->lvb, DLM_LVB_LEN))) { - mlog(ML_ERROR, "mismatched lvbs!\n"); - dlm_print_one_lock_resource(lock->lockres); - BUG(); - } - memcpy(mres->lvb, lock->lksb->lvb, DLM_LVB_LEN); - } + dlm_prepare_lvb_for_migration(lock, mres, queue); } ml->node = lock->ml.node; mres->num_locks++; @@ -1379,6 +1418,7 @@ int dlm_mig_lockres_handler(struct o2net_msg *msg, u32 len, void *data, mres->lockname_len, mres->lockname); ret = -EFAULT; spin_unlock(&res->spinlock); + dlm_lockres_put(res); goto leave; } res->state |= DLM_LOCK_RES_MIGRATING; @@ -1469,13 +1509,11 @@ leave: dlm_put(dlm); if (ret < 0) { - if (buf) - kfree(buf); - if (item) - kfree(item); + kfree(buf); + kfree(item); + mlog_errno(ret); } - mlog_exit(ret); return ret; } @@ -1544,7 +1582,6 @@ leave: dlm_lockres_put(res); } kfree(data); - mlog_exit(ret); } @@ -1623,7 +1660,9 @@ int dlm_do_master_requery(struct dlm_ctxt *dlm, struct dlm_lock_resource *res, &req, sizeof(req), nodenum, &status); /* XXX: negative status not handled properly here. */ if (ret < 0) - mlog_errno(ret); + mlog(ML_ERROR, "Error %d when sending message %u (key " + "0x%x) to node %u\n", ret, DLM_MASTER_REQUERY_MSG, + dlm->key, nodenum); else { BUG_ON(status < 0); BUG_ON(status > DLM_LOCK_RES_OWNER_UNKNOWN); @@ -1669,7 +1708,8 @@ int dlm_master_requery_handler(struct o2net_msg *msg, u32 len, void *data, mlog_errno(-ENOMEM); /* retry!? */ BUG(); - } + } else + __dlm_lockres_grab_inflight_worker(dlm, res); } else /* put.. incase we are not the master */ dlm_lockres_put(res); spin_unlock(&res->spinlock); @@ -1722,15 +1762,16 @@ static int dlm_process_recovery_data(struct dlm_ctxt *dlm, struct dlm_migratable_lockres *mres) { struct dlm_migratable_lock *ml; - struct list_head *queue; + struct list_head *queue, *iter; struct list_head *tmpq = NULL; struct dlm_lock *newlock = NULL; struct dlm_lockstatus *lksb = NULL; int ret = 0; int i, j, bad; - struct dlm_lock *lock = NULL; + struct dlm_lock *lock; u8 from = O2NM_MAX_NODES; unsigned int added = 0; + __be64 c; mlog(0, "running %d locks for this lockres\n", mres->num_locks); for (i=0; i<mres->num_locks; i++) { @@ -1743,7 +1784,7 @@ static int dlm_process_recovery_data(struct dlm_ctxt *dlm, dlm->name, mres->lockname_len, mres->lockname, from); spin_lock(&res->spinlock); - dlm_lockres_set_refmap_bit(from, res); + dlm_lockres_set_refmap_bit(dlm, res, from); spin_unlock(&res->spinlock); added++; break; @@ -1762,14 +1803,16 @@ static int dlm_process_recovery_data(struct dlm_ctxt *dlm, /* MIGRATION ONLY! */ BUG_ON(!(mres->flags & DLM_MRES_MIGRATION)); + lock = NULL; spin_lock(&res->spinlock); for (j = DLM_GRANTED_LIST; j <= DLM_BLOCKED_LIST; j++) { tmpq = dlm_list_idx_to_ptr(res, j); - list_for_each_entry(lock, tmpq, list) { - if (lock->ml.cookie != ml->cookie) - lock = NULL; - else + list_for_each(iter, tmpq) { + lock = list_entry(iter, + struct dlm_lock, list); + if (lock->ml.cookie == ml->cookie) break; + lock = NULL; } if (lock) break; @@ -1778,19 +1821,48 @@ static int dlm_process_recovery_data(struct dlm_ctxt *dlm, /* lock is always created locally first, and * destroyed locally last. it must be on the list */ if (!lock) { - __be64 c = ml->cookie; - mlog(ML_ERROR, "could not find local lock " - "with cookie %u:%llu!\n", + c = ml->cookie; + mlog(ML_ERROR, "Could not find local lock " + "with cookie %u:%llu, node %u, " + "list %u, flags 0x%x, type %d, " + "conv %d, highest blocked %d\n", dlm_get_lock_cookie_node(be64_to_cpu(c)), - dlm_get_lock_cookie_seq(be64_to_cpu(c))); + dlm_get_lock_cookie_seq(be64_to_cpu(c)), + ml->node, ml->list, ml->flags, ml->type, + ml->convert_type, ml->highest_blocked); + __dlm_print_one_lock_resource(res); + BUG(); + } + + if (lock->ml.node != ml->node) { + c = lock->ml.cookie; + mlog(ML_ERROR, "Mismatched node# in lock " + "cookie %u:%llu, name %.*s, node %u\n", + dlm_get_lock_cookie_node(be64_to_cpu(c)), + dlm_get_lock_cookie_seq(be64_to_cpu(c)), + res->lockname.len, res->lockname.name, + lock->ml.node); + c = ml->cookie; + mlog(ML_ERROR, "Migrate lock cookie %u:%llu, " + "node %u, list %u, flags 0x%x, type %d, " + "conv %d, highest blocked %d\n", + dlm_get_lock_cookie_node(be64_to_cpu(c)), + dlm_get_lock_cookie_seq(be64_to_cpu(c)), + ml->node, ml->list, ml->flags, ml->type, + ml->convert_type, ml->highest_blocked); __dlm_print_one_lock_resource(res); BUG(); } - BUG_ON(lock->ml.node != ml->node); if (tmpq != queue) { - mlog(0, "lock was on %u instead of %u for %.*s\n", - j, ml->list, res->lockname.len, res->lockname.name); + c = ml->cookie; + mlog(0, "Lock cookie %u:%llu was on list %u " + "instead of list %u for %.*s\n", + dlm_get_lock_cookie_node(be64_to_cpu(c)), + dlm_get_lock_cookie_seq(be64_to_cpu(c)), + j, ml->list, res->lockname.len, + res->lockname.name); + __dlm_print_one_lock_resource(res); spin_unlock(&res->spinlock); continue; } @@ -1828,6 +1900,13 @@ static int dlm_process_recovery_data(struct dlm_ctxt *dlm, if (ml->type == LKM_NLMODE) goto skip_lvb; + /* + * If the lock is in the blocked list it can't have a valid lvb, + * so skip it + */ + if (ml->list == DLM_BLOCKED_LIST) + goto skip_lvb; + if (!dlm_lvb_is_empty(mres->lvb)) { if (lksb->flags & DLM_LKSB_PUT_LVB) { /* other node was trying to update @@ -1840,7 +1919,7 @@ static int dlm_process_recovery_data(struct dlm_ctxt *dlm, * the lvb. */ memcpy(res->lvb, mres->lvb, DLM_LVB_LEN); } else { - /* otherwise, the node is sending its + /* otherwise, the node is sending its * most recent valid lvb info */ BUG_ON(ml->type != LKM_EXMODE && ml->type != LKM_PRMODE); @@ -1887,7 +1966,7 @@ skip_lvb: spin_lock(&res->spinlock); list_for_each_entry(lock, queue, list) { if (lock->ml.cookie == ml->cookie) { - __be64 c = lock->ml.cookie; + c = lock->ml.cookie; mlog(ML_ERROR, "%s:%.*s: %u:%llu: lock already " "exists on this lockres!\n", dlm->name, res->lockname.len, res->lockname.name, @@ -1908,11 +1987,19 @@ skip_lvb: } if (!bad) { dlm_lock_get(newlock); - list_add_tail(&newlock->list, queue); + if (mres->flags & DLM_MRES_RECOVERY && + ml->list == DLM_CONVERTING_LIST && + newlock->ml.type > + newlock->ml.convert_type) { + /* newlock is doing downconvert, add it to the + * head of converting list */ + list_add(&newlock->list, queue); + } else + list_add_tail(&newlock->list, queue); mlog(0, "%s:%.*s: added lock for node %u, " "setting refmap bit\n", dlm->name, res->lockname.len, res->lockname.name, ml->node); - dlm_lockres_set_refmap_bit(ml->node, res); + dlm_lockres_set_refmap_bit(dlm, res, ml->node); added++; } spin_unlock(&res->spinlock); @@ -1931,7 +2018,6 @@ leave: dlm_lock_put(newlock); } - mlog_exit(ret); return ret; } @@ -1942,6 +2028,8 @@ void dlm_move_lockres_to_recovery_list(struct dlm_ctxt *dlm, struct list_head *queue; struct dlm_lock *lock, *next; + assert_spin_locked(&dlm->spinlock); + assert_spin_locked(&res->spinlock); res->state |= DLM_LOCK_RES_RECOVERING; if (!list_empty(&res->recovering)) { mlog(0, @@ -2022,16 +2110,16 @@ static void dlm_finish_local_lockres_recovery(struct dlm_ctxt *dlm, u8 dead_node, u8 new_master) { int i; - struct hlist_node *hash_iter; struct hlist_head *bucket; struct dlm_lock_resource *res, *next; - mlog_entry_void(); - assert_spin_locked(&dlm->spinlock); list_for_each_entry_safe(res, next, &dlm->reco.resources, recovering) { if (res->owner == dead_node) { + mlog(0, "%s: res %.*s, Changing owner from %u to %u\n", + dlm->name, res->lockname.len, res->lockname.name, + res->owner, new_master); list_del_init(&res->recovering); spin_lock(&res->spinlock); /* new_master has our reference from @@ -2052,41 +2140,31 @@ static void dlm_finish_local_lockres_recovery(struct dlm_ctxt *dlm, * if necessary */ for (i = 0; i < DLM_HASH_BUCKETS; i++) { bucket = dlm_lockres_hash(dlm, i); - hlist_for_each_entry(res, hash_iter, bucket, hash_node) { - if (res->state & DLM_LOCK_RES_RECOVERING) { - if (res->owner == dead_node) { - mlog(0, "(this=%u) res %.*s owner=%u " - "was not on recovering list, but " - "clearing state anyway\n", - dlm->node_num, res->lockname.len, - res->lockname.name, new_master); - } else if (res->owner == dlm->node_num) { - mlog(0, "(this=%u) res %.*s owner=%u " - "was not on recovering list, " - "owner is THIS node, clearing\n", - dlm->node_num, res->lockname.len, - res->lockname.name, new_master); - } else - continue; + hlist_for_each_entry(res, bucket, hash_node) { + if (!(res->state & DLM_LOCK_RES_RECOVERING)) + continue; - if (!list_empty(&res->recovering)) { - mlog(0, "%s:%.*s: lockres was " - "marked RECOVERING, owner=%u\n", - dlm->name, res->lockname.len, - res->lockname.name, res->owner); - list_del_init(&res->recovering); - dlm_lockres_put(res); - } - spin_lock(&res->spinlock); - /* new_master has our reference from - * the lock state sent during recovery */ - dlm_change_lockres_owner(dlm, res, new_master); - res->state &= ~DLM_LOCK_RES_RECOVERING; - if (__dlm_lockres_has_locks(res)) - __dlm_dirty_lockres(dlm, res); - spin_unlock(&res->spinlock); - wake_up(&res->wq); + if (res->owner != dead_node && + res->owner != dlm->node_num) + continue; + + if (!list_empty(&res->recovering)) { + list_del_init(&res->recovering); + dlm_lockres_put(res); } + + /* new_master has our reference from + * the lock state sent during recovery */ + mlog(0, "%s: res %.*s, Changing owner from %u to %u\n", + dlm->name, res->lockname.len, res->lockname.name, + res->owner, new_master); + spin_lock(&res->spinlock); + dlm_change_lockres_owner(dlm, res, new_master); + res->state &= ~DLM_LOCK_RES_RECOVERING; + if (__dlm_lockres_has_locks(res)) + __dlm_dirty_lockres(dlm, res); + spin_unlock(&res->spinlock); + wake_up(&res->wq); } } } @@ -2115,7 +2193,7 @@ static void dlm_revalidate_lvb(struct dlm_ctxt *dlm, assert_spin_locked(&res->spinlock); if (res->owner == dlm->node_num) - /* if this node owned the lockres, and if the dead node + /* if this node owned the lockres, and if the dead node * had an EX when he died, blank out the lvb */ search_node = dead_node; else { @@ -2153,7 +2231,7 @@ static void dlm_free_dead_locks(struct dlm_ctxt *dlm, /* this node is the lockres master: * 1) remove any stale locks for the dead node - * 2) if the dead node had an EX when he died, blank out the lvb + * 2) if the dead node had an EX when he died, blank out the lvb */ assert_spin_locked(&dlm->spinlock); assert_spin_locked(&res->spinlock); @@ -2194,13 +2272,18 @@ static void dlm_free_dead_locks(struct dlm_ctxt *dlm, mlog(0, "%s:%.*s: freed %u locks for dead node %u, " "dropping ref from lockres\n", dlm->name, res->lockname.len, res->lockname.name, freed, dead_node); - BUG_ON(!test_bit(dead_node, res->refmap)); - dlm_lockres_clear_refmap_bit(dead_node, res); + if(!test_bit(dead_node, res->refmap)) { + mlog(ML_ERROR, "%s:%.*s: freed %u locks for dead node %u, " + "but ref was not set\n", dlm->name, + res->lockname.len, res->lockname.name, freed, dead_node); + __dlm_print_one_lock_resource(res); + } + dlm_lockres_clear_refmap_bit(dlm, res, dead_node); } else if (test_bit(dead_node, res->refmap)) { mlog(0, "%s:%.*s: dead node %u had a ref, but had " "no locks and had not purged before dying\n", dlm->name, res->lockname.len, res->lockname.name, dead_node); - dlm_lockres_clear_refmap_bit(dead_node, res); + dlm_lockres_clear_refmap_bit(dlm, res, dead_node); } /* do not kick thread yet */ @@ -2216,7 +2299,6 @@ static void dlm_free_dead_locks(struct dlm_ctxt *dlm, static void dlm_do_local_recovery_cleanup(struct dlm_ctxt *dlm, u8 dead_node) { - struct hlist_node *iter; struct dlm_lock_resource *res; int i; struct hlist_head *bucket; @@ -2242,7 +2324,7 @@ static void dlm_do_local_recovery_cleanup(struct dlm_ctxt *dlm, u8 dead_node) */ for (i = 0; i < DLM_HASH_BUCKETS; i++) { bucket = dlm_lockres_hash(dlm, i); - hlist_for_each_entry(res, iter, bucket, hash_node) { + hlist_for_each_entry(res, bucket, hash_node) { /* always prune any $RECOVERY entries for dead nodes, * otherwise hangs can occur during later recovery */ if (dlm_is_recovery_lock(res->lockname.name, @@ -2261,27 +2343,31 @@ static void dlm_do_local_recovery_cleanup(struct dlm_ctxt *dlm, u8 dead_node) } spin_unlock(&res->spinlock); continue; - } + } spin_lock(&res->spinlock); /* zero the lvb if necessary */ dlm_revalidate_lvb(dlm, res, dead_node); if (res->owner == dead_node) { - if (res->state & DLM_LOCK_RES_DROPPING_REF) - mlog(0, "%s:%.*s: owned by " - "dead node %u, this node was " - "dropping its ref when it died. " - "continue, dropping the flag.\n", + if (res->state & DLM_LOCK_RES_DROPPING_REF) { + mlog(ML_NOTICE, "%s: res %.*s, Skip " + "recovery as it is being freed\n", dlm->name, res->lockname.len, - res->lockname.name, dead_node); - - /* the wake_up for this will happen when the - * RECOVERING flag is dropped later */ - res->state &= ~DLM_LOCK_RES_DROPPING_REF; + res->lockname.name); + } else + dlm_move_lockres_to_recovery_list(dlm, + res); - dlm_move_lockres_to_recovery_list(dlm, res); } else if (res->owner == dlm->node_num) { dlm_free_dead_locks(dlm, res, dead_node); __dlm_lockres_calc_usage(dlm, res); + } else if (res->owner == DLM_LOCK_RES_OWNER_UNKNOWN) { + if (test_bit(dead_node, res->refmap)) { + mlog(0, "%s:%.*s: dead node %u had a ref, but had " + "no locks and had not purged before dying\n", + dlm->name, res->lockname.len, + res->lockname.name, dead_node); + dlm_lockres_clear_refmap_bit(dlm, res, dead_node); + } } spin_unlock(&res->spinlock); } @@ -2340,6 +2426,7 @@ static void __dlm_hb_node_down(struct dlm_ctxt *dlm, int idx) mlog(0, "node %u being removed from domain map!\n", idx); clear_bit(idx, dlm->domain_map); + clear_bit(idx, dlm->exit_domain_map); /* wake up migration waiters if a node goes down. * perhaps later we can genericize this for other waiters. */ wake_up(&dlm->migration_wq); @@ -2412,7 +2499,7 @@ static void dlm_reco_unlock_ast(void *astdata, enum dlm_status st) * this function on each node racing to become the recovery * master will not stop attempting this until either: * a) this node gets the EX (and becomes the recovery master), - * or b) dlm->reco.new_master gets set to some nodenum + * or b) dlm->reco.new_master gets set to some nodenum * != O2NM_INVALID_NODE_NUM (another node will do the reco). * so each time a recovery master is needed, the entire cluster * will sync at this point. if the new master dies, that will @@ -2425,7 +2512,7 @@ static int dlm_pick_recovery_master(struct dlm_ctxt *dlm) mlog(0, "starting recovery of %s at %lu, dead=%u, this=%u\n", dlm->name, jiffies, dlm->reco.dead_node, dlm->node_num); -again: +again: memset(&lksb, 0, sizeof(lksb)); ret = dlmlock(dlm, LKM_EXMODE, &lksb, LKM_NOQUEUE|LKM_RECOVERY, @@ -2438,8 +2525,8 @@ again: if (ret == DLM_NORMAL) { mlog(0, "dlm=%s dlmlock says I got it (this=%u)\n", dlm->name, dlm->node_num); - - /* got the EX lock. check to see if another node + + /* got the EX lock. check to see if another node * just became the reco master */ if (dlm_reco_master_ready(dlm)) { mlog(0, "%s: got reco EX lock, but %u will " @@ -2452,12 +2539,12 @@ again: /* see if recovery was already finished elsewhere */ spin_lock(&dlm->spinlock); if (dlm->reco.dead_node == O2NM_INVALID_NODE_NUM) { - status = -EINVAL; + status = -EINVAL; mlog(0, "%s: got reco EX lock, but " "node got recovered already\n", dlm->name); if (dlm->reco.new_master != O2NM_INVALID_NODE_NUM) { mlog(ML_ERROR, "%s: new master is %u " - "but no dead node!\n", + "but no dead node!\n", dlm->name, dlm->reco.new_master); BUG(); } @@ -2469,7 +2556,7 @@ again: * set the master and send the messages to begin recovery */ if (!status) { mlog(0, "%s: dead=%u, this=%u, sending " - "begin_reco now\n", dlm->name, + "begin_reco now\n", dlm->name, dlm->reco.dead_node, dlm->node_num); status = dlm_send_begin_reco_message(dlm, dlm->reco.dead_node); @@ -2502,7 +2589,7 @@ again: mlog(0, "dlm=%s dlmlock says another node got it (this=%u)\n", dlm->name, dlm->node_num); /* another node is master. wait on - * reco.new_master != O2NM_INVALID_NODE_NUM + * reco.new_master != O2NM_INVALID_NODE_NUM * for at most one second */ wait_event_timeout(dlm->dlm_reco_thread_wq, dlm_reco_master_ready(dlm), @@ -2549,8 +2636,6 @@ static int dlm_send_begin_reco_message(struct dlm_ctxt *dlm, u8 dead_node) int nodenum; int status; - mlog_entry("%u\n", dead_node); - mlog(0, "%s: dead node is %u\n", dlm->name, dead_node); spin_lock(&dlm->spinlock); @@ -2586,17 +2671,32 @@ retry: if (dlm_is_host_down(ret)) { /* node is down. not involved in recovery * so just keep going */ - mlog(0, "%s: node %u was down when sending " + mlog(ML_NOTICE, "%s: node %u was down when sending " "begin reco msg (%d)\n", dlm->name, nodenum, ret); ret = 0; } + + /* + * Prior to commit aad1b15310b9bcd59fa81ab8f2b1513b59553ea8, + * dlm_begin_reco_handler() returned EAGAIN and not -EAGAIN. + * We are handling both for compatibility reasons. + */ + if (ret == -EAGAIN || ret == EAGAIN) { + mlog(0, "%s: trying to start recovery of node " + "%u, but node %u is waiting for last recovery " + "to complete, backoff for a bit\n", dlm->name, + dead_node, nodenum); + msleep(100); + goto retry; + } if (ret < 0) { struct dlm_lock_resource *res; - /* this is now a serious problem, possibly ENOMEM + + /* this is now a serious problem, possibly ENOMEM * in the network stack. must retry */ mlog_errno(ret); mlog(ML_ERROR, "begin reco of dlm %s to node %u " - " returned %d\n", dlm->name, nodenum, ret); + "returned %d\n", dlm->name, nodenum, ret); res = dlm_lookup_lockres(dlm, DLM_RECOVERY_LOCK_NAME, DLM_RECOVERY_LOCK_NAME_LEN); if (res) { @@ -2605,18 +2705,10 @@ retry: } else { mlog(ML_ERROR, "recovery lock not found\n"); } - /* sleep for a bit in hopes that we can avoid + /* sleep for a bit in hopes that we can avoid * another ENOMEM */ msleep(100); goto retry; - } else if (ret == EAGAIN) { - mlog(0, "%s: trying to start recovery of node " - "%u, but node %u is waiting for last recovery " - "to complete, backoff for a bit\n", dlm->name, - dead_node, nodenum); - /* TODO Look into replacing msleep with cond_resched() */ - msleep(100); - goto retry; } } @@ -2640,7 +2732,8 @@ int dlm_begin_reco_handler(struct o2net_msg *msg, u32 len, void *data, dlm->name, br->node_idx, br->dead_node, dlm->reco.dead_node, dlm->reco.new_master); spin_unlock(&dlm->spinlock); - return EAGAIN; + dlm_put(dlm); + return -EAGAIN; } spin_unlock(&dlm->spinlock); @@ -2665,7 +2758,7 @@ int dlm_begin_reco_handler(struct o2net_msg *msg, u32 len, void *data, } if (dlm->reco.dead_node != O2NM_INVALID_NODE_NUM) { mlog(ML_NOTICE, "%s: dead_node previously set to %u, " - "node %u changing it to %u\n", dlm->name, + "node %u changing it to %u\n", dlm->name, dlm->reco.dead_node, br->node_idx, br->dead_node); } dlm_set_reco_master(dlm, br->node_idx); @@ -2729,10 +2822,12 @@ stage2: if (ret >= 0) ret = status; if (ret < 0) { - mlog_errno(ret); + mlog(ML_ERROR, "Error %d when sending message %u (key " + "0x%x) to node %u\n", ret, DLM_FINALIZE_RECO_MSG, + dlm->key, nodenum); if (dlm_is_host_down(ret)) { - /* this has no effect on this recovery - * session, so set the status to zero to + /* this has no effect on this recovery + * session, so set the status to zero to * finish out the last recovery */ mlog(ML_ERROR, "node %u went down after this " "node finished recovery.\n", nodenum); @@ -2769,7 +2864,7 @@ int dlm_finalize_reco_handler(struct o2net_msg *msg, u32 len, void *data, mlog(0, "%s: node %u finalizing recovery stage%d of " "node %u (%u:%u)\n", dlm->name, fr->node_idx, stage, fr->dead_node, dlm->reco.dead_node, dlm->reco.new_master); - + spin_lock(&dlm->spinlock); if (dlm->reco.new_master != fr->node_idx) { @@ -2809,8 +2904,8 @@ int dlm_finalize_reco_handler(struct o2net_msg *msg, u32 len, void *data, BUG(); } dlm->reco.state &= ~DLM_RECO_STATE_FINALIZE; + __dlm_reset_recovery(dlm); spin_unlock(&dlm->spinlock); - dlm_reset_recovery(dlm); dlm_kick_recovery_thread(dlm); break; default: diff --git a/fs/ocfs2/dlm/dlmthread.c b/fs/ocfs2/dlm/dlmthread.c index d490b66ad9d..69aac6f088a 100644 --- a/fs/ocfs2/dlm/dlmthread.c +++ b/fs/ocfs2/dlm/dlmthread.c @@ -28,9 +28,7 @@ #include <linux/module.h> #include <linux/fs.h> #include <linux/types.h> -#include <linux/slab.h> #include <linux/highmem.h> -#include <linux/utsname.h> #include <linux/init.h> #include <linux/sysctl.h> #include <linux/random.h> @@ -94,19 +92,29 @@ int __dlm_lockres_has_locks(struct dlm_lock_resource *res) * truly ready to be freed. */ int __dlm_lockres_unused(struct dlm_lock_resource *res) { - if (!__dlm_lockres_has_locks(res) && - (list_empty(&res->dirty) && !(res->state & DLM_LOCK_RES_DIRTY))) { - /* try not to scan the bitmap unless the first two - * conditions are already true */ - int bit = find_next_bit(res->refmap, O2NM_MAX_NODES, 0); - if (bit >= O2NM_MAX_NODES) { - /* since the bit for dlm->node_num is not - * set, inflight_locks better be zero */ - BUG_ON(res->inflight_locks != 0); - return 1; - } - } - return 0; + int bit; + + assert_spin_locked(&res->spinlock); + + if (__dlm_lockres_has_locks(res)) + return 0; + + /* Locks are in the process of being created */ + if (res->inflight_locks) + return 0; + + if (!list_empty(&res->dirty) || res->state & DLM_LOCK_RES_DIRTY) + return 0; + + if (res->state & DLM_LOCK_RES_RECOVERING) + return 0; + + /* Another node has this resource with this node as the master */ + bit = find_next_bit(res->refmap, O2NM_MAX_NODES, 0); + if (bit < O2NM_MAX_NODES) + return 0; + + return 1; } @@ -116,15 +124,13 @@ int __dlm_lockres_unused(struct dlm_lock_resource *res) void __dlm_lockres_calc_usage(struct dlm_ctxt *dlm, struct dlm_lock_resource *res) { - mlog_entry("%.*s\n", res->lockname.len, res->lockname.name); - assert_spin_locked(&dlm->spinlock); assert_spin_locked(&res->spinlock); if (__dlm_lockres_unused(res)){ if (list_empty(&res->purge)) { - mlog(0, "putting lockres %.*s:%p onto purge list\n", - res->lockname.len, res->lockname.name, res); + mlog(0, "%s: Adding res %.*s to purge list\n", + dlm->name, res->lockname.len, res->lockname.name); res->last_used = jiffies; dlm_lockres_get(res); @@ -132,8 +138,8 @@ void __dlm_lockres_calc_usage(struct dlm_ctxt *dlm, dlm->purge_count++; } } else if (!list_empty(&res->purge)) { - mlog(0, "removing lockres %.*s:%p from purge list, owner=%u\n", - res->lockname.len, res->lockname.name, res, res->owner); + mlog(0, "%s: Removing res %.*s from purge list\n", + dlm->name, res->lockname.len, res->lockname.name); list_del_init(&res->purge); dlm_lockres_put(res); @@ -144,7 +150,6 @@ void __dlm_lockres_calc_usage(struct dlm_ctxt *dlm, void dlm_lockres_calc_usage(struct dlm_ctxt *dlm, struct dlm_lock_resource *res) { - mlog_entry("%.*s\n", res->lockname.len, res->lockname.name); spin_lock(&dlm->spinlock); spin_lock(&res->spinlock); @@ -154,45 +159,24 @@ void dlm_lockres_calc_usage(struct dlm_ctxt *dlm, spin_unlock(&dlm->spinlock); } -static int dlm_purge_lockres(struct dlm_ctxt *dlm, +static void dlm_purge_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res) { int master; int ret = 0; - spin_lock(&res->spinlock); - if (!__dlm_lockres_unused(res)) { - mlog(0, "%s:%.*s: tried to purge but not unused\n", - dlm->name, res->lockname.len, res->lockname.name); - __dlm_print_one_lock_resource(res); - spin_unlock(&res->spinlock); - BUG(); - } - - if (res->state & DLM_LOCK_RES_MIGRATING) { - mlog(0, "%s:%.*s: Delay dropref as this lockres is " - "being remastered\n", dlm->name, res->lockname.len, - res->lockname.name); - /* Re-add the lockres to the end of the purge list */ - if (!list_empty(&res->purge)) { - list_del_init(&res->purge); - list_add_tail(&res->purge, &dlm->purge_list); - } - spin_unlock(&res->spinlock); - return 0; - } + assert_spin_locked(&dlm->spinlock); + assert_spin_locked(&res->spinlock); master = (res->owner == dlm->node_num); - if (!master) - res->state |= DLM_LOCK_RES_DROPPING_REF; - spin_unlock(&res->spinlock); - - mlog(0, "purging lockres %.*s, master = %d\n", res->lockname.len, - res->lockname.name, master); + mlog(0, "%s: Purging res %.*s, master %d\n", dlm->name, + res->lockname.len, res->lockname.name, master); if (!master) { + res->state |= DLM_LOCK_RES_DROPPING_REF; /* drop spinlock... retake below */ + spin_unlock(&res->spinlock); spin_unlock(&dlm->spinlock); spin_lock(&res->spinlock); @@ -203,34 +187,38 @@ static int dlm_purge_lockres(struct dlm_ctxt *dlm, /* clear our bit from the master's refmap, ignore errors */ ret = dlm_drop_lockres_ref(dlm, res); if (ret < 0) { - mlog_errno(ret); if (!dlm_is_host_down(ret)) BUG(); } - mlog(0, "%s:%.*s: dlm_deref_lockres returned %d\n", - dlm->name, res->lockname.len, res->lockname.name, ret); spin_lock(&dlm->spinlock); + spin_lock(&res->spinlock); } if (!list_empty(&res->purge)) { - mlog(0, "removing lockres %.*s:%p from purgelist, " - "master = %d\n", res->lockname.len, res->lockname.name, - res, master); + mlog(0, "%s: Removing res %.*s from purgelist, master %d\n", + dlm->name, res->lockname.len, res->lockname.name, master); list_del_init(&res->purge); dlm_lockres_put(res); dlm->purge_count--; } - __dlm_unhash_lockres(res); + + if (!__dlm_lockres_unused(res)) { + mlog(ML_ERROR, "%s: res %.*s in use after deref\n", + dlm->name, res->lockname.len, res->lockname.name); + __dlm_print_one_lock_resource(res); + BUG(); + } + + __dlm_unhash_lockres(dlm, res); /* lockres is not in the hash now. drop the flag and wake up * any processes waiting in dlm_get_lock_resource. */ if (!master) { - spin_lock(&res->spinlock); res->state &= ~DLM_LOCK_RES_DROPPING_REF; spin_unlock(&res->spinlock); wake_up(&res->wq); - } - return 0; + } else + spin_unlock(&res->spinlock); } static void dlm_run_purge_list(struct dlm_ctxt *dlm, @@ -249,17 +237,7 @@ static void dlm_run_purge_list(struct dlm_ctxt *dlm, lockres = list_entry(dlm->purge_list.next, struct dlm_lock_resource, purge); - /* Status of the lockres *might* change so double - * check. If the lockres is unused, holding the dlm - * spinlock will prevent people from getting and more - * refs on it -- there's no need to keep the lockres - * spinlock. */ spin_lock(&lockres->spinlock); - unused = __dlm_lockres_unused(lockres); - spin_unlock(&lockres->spinlock); - - if (!unused) - continue; purge_jiffies = lockres->last_used + msecs_to_jiffies(DLM_PURGE_INTERVAL_MS); @@ -271,15 +249,32 @@ static void dlm_run_purge_list(struct dlm_ctxt *dlm, * in tail order, we can stop at the first * unpurgable resource -- anyone added after * him will have a greater last_used value */ + spin_unlock(&lockres->spinlock); break; } + /* Status of the lockres *might* change so double + * check. If the lockres is unused, holding the dlm + * spinlock will prevent people from getting and more + * refs on it. */ + unused = __dlm_lockres_unused(lockres); + if (!unused || + (lockres->state & DLM_LOCK_RES_MIGRATING) || + (lockres->inflight_assert_workers != 0)) { + mlog(0, "%s: res %.*s is in use or being remastered, " + "used %d, state %d, assert master workers %u\n", + dlm->name, lockres->lockname.len, + lockres->lockname.name, + !unused, lockres->state, + lockres->inflight_assert_workers); + list_move_tail(&lockres->purge, &dlm->purge_list); + spin_unlock(&lockres->spinlock); + continue; + } + dlm_lockres_get(lockres); - /* This may drop and reacquire the dlm spinlock if it - * has to do migration. */ - if (dlm_purge_lockres(dlm, lockres)) - BUG(); + dlm_purge_lockres(dlm, lockres); dlm_lockres_put(lockres); @@ -294,19 +289,15 @@ static void dlm_shuffle_lists(struct dlm_ctxt *dlm, struct dlm_lock_resource *res) { struct dlm_lock *lock, *target; - struct list_head *iter; - struct list_head *head; int can_grant = 1; - //mlog(0, "res->lockname.len=%d\n", res->lockname.len); - //mlog(0, "res->lockname.name=%p\n", res->lockname.name); - //mlog(0, "shuffle res %.*s\n", res->lockname.len, - // res->lockname.name); - - /* because this function is called with the lockres + /* + * Because this function is called with the lockres * spinlock, and because we know that it is not migrating/ * recovering/in-progress, it is fine to reserve asts and - * basts right before queueing them all throughout */ + * basts right before queueing them all throughout + */ + assert_spin_locked(&dlm->ast_lock); assert_spin_locked(&res->spinlock); BUG_ON((res->state & (DLM_LOCK_RES_MIGRATING| DLM_LOCK_RES_RECOVERING| @@ -315,18 +306,16 @@ static void dlm_shuffle_lists(struct dlm_ctxt *dlm, converting: if (list_empty(&res->converting)) goto blocked; - mlog(0, "res %.*s has locks on a convert queue\n", res->lockname.len, - res->lockname.name); + mlog(0, "%s: res %.*s has locks on the convert queue\n", dlm->name, + res->lockname.len, res->lockname.name); target = list_entry(res->converting.next, struct dlm_lock, list); if (target->ml.convert_type == LKM_IVMODE) { - mlog(ML_ERROR, "%.*s: converting a lock with no " - "convert_type!\n", res->lockname.len, res->lockname.name); + mlog(ML_ERROR, "%s: res %.*s converting lock to invalid mode\n", + dlm->name, res->lockname.len, res->lockname.name); BUG(); } - head = &res->granted; - list_for_each(iter, head) { - lock = list_entry(iter, struct dlm_lock, list); + list_for_each_entry(lock, &res->granted, list) { if (lock==target) continue; if (!dlm_lock_compatible(lock->ml.type, @@ -335,7 +324,7 @@ converting: /* queue the BAST if not already */ if (lock->ml.highest_blocked == LKM_IVMODE) { __dlm_lockres_reserve_ast(res); - dlm_queue_bast(dlm, lock); + __dlm_queue_bast(dlm, lock); } /* update the highest_blocked if needed */ if (lock->ml.highest_blocked < target->ml.convert_type) @@ -343,9 +332,8 @@ converting: target->ml.convert_type; } } - head = &res->converting; - list_for_each(iter, head) { - lock = list_entry(iter, struct dlm_lock, list); + + list_for_each_entry(lock, &res->converting, list) { if (lock==target) continue; if (!dlm_lock_compatible(lock->ml.type, @@ -353,7 +341,7 @@ converting: can_grant = 0; if (lock->ml.highest_blocked == LKM_IVMODE) { __dlm_lockres_reserve_ast(res); - dlm_queue_bast(dlm, lock); + __dlm_queue_bast(dlm, lock); } if (lock->ml.highest_blocked < target->ml.convert_type) lock->ml.highest_blocked = @@ -366,9 +354,12 @@ converting: spin_lock(&target->spinlock); BUG_ON(target->ml.highest_blocked != LKM_IVMODE); - mlog(0, "calling ast for converting lock: %.*s, have: %d, " - "granting: %d, node: %u\n", res->lockname.len, - res->lockname.name, target->ml.type, + mlog(0, "%s: res %.*s, AST for Converting lock %u:%llu, type " + "%d => %d, node %u\n", dlm->name, res->lockname.len, + res->lockname.name, + dlm_get_lock_cookie_node(be64_to_cpu(target->ml.cookie)), + dlm_get_lock_cookie_seq(be64_to_cpu(target->ml.cookie)), + target->ml.type, target->ml.convert_type, target->ml.node); target->ml.type = target->ml.convert_type; @@ -381,7 +372,7 @@ converting: spin_unlock(&target->spinlock); __dlm_lockres_reserve_ast(res); - dlm_queue_ast(dlm, target); + __dlm_queue_ast(dlm, target); /* go back and check for more */ goto converting; } @@ -391,32 +382,28 @@ blocked: goto leave; target = list_entry(res->blocked.next, struct dlm_lock, list); - head = &res->granted; - list_for_each(iter, head) { - lock = list_entry(iter, struct dlm_lock, list); + list_for_each_entry(lock, &res->granted, list) { if (lock==target) continue; if (!dlm_lock_compatible(lock->ml.type, target->ml.type)) { can_grant = 0; if (lock->ml.highest_blocked == LKM_IVMODE) { __dlm_lockres_reserve_ast(res); - dlm_queue_bast(dlm, lock); + __dlm_queue_bast(dlm, lock); } if (lock->ml.highest_blocked < target->ml.type) lock->ml.highest_blocked = target->ml.type; } } - head = &res->converting; - list_for_each(iter, head) { - lock = list_entry(iter, struct dlm_lock, list); + list_for_each_entry(lock, &res->converting, list) { if (lock==target) continue; if (!dlm_lock_compatible(lock->ml.type, target->ml.type)) { can_grant = 0; if (lock->ml.highest_blocked == LKM_IVMODE) { __dlm_lockres_reserve_ast(res); - dlm_queue_bast(dlm, lock); + __dlm_queue_bast(dlm, lock); } if (lock->ml.highest_blocked < target->ml.type) lock->ml.highest_blocked = target->ml.type; @@ -429,11 +416,14 @@ blocked: spin_lock(&target->spinlock); BUG_ON(target->ml.highest_blocked != LKM_IVMODE); - mlog(0, "calling ast for blocked lock: %.*s, granting: %d, " - "node: %u\n", res->lockname.len, res->lockname.name, + mlog(0, "%s: res %.*s, AST for Blocked lock %u:%llu, type %d, " + "node %u\n", dlm->name, res->lockname.len, + res->lockname.name, + dlm_get_lock_cookie_node(be64_to_cpu(target->ml.cookie)), + dlm_get_lock_cookie_seq(be64_to_cpu(target->ml.cookie)), target->ml.type, target->ml.node); - // target->ml.type is already correct + /* target->ml.type is already correct */ list_move_tail(&target->list, &res->granted); BUG_ON(!target->lksb); @@ -442,7 +432,7 @@ blocked: spin_unlock(&target->spinlock); __dlm_lockres_reserve_ast(res); - dlm_queue_ast(dlm, target); + __dlm_queue_ast(dlm, target); /* go back and check for more */ goto converting; } @@ -454,7 +444,6 @@ leave: /* must have NO locks when calling this with res !=NULL * */ void dlm_kick_thread(struct dlm_ctxt *dlm, struct dlm_lock_resource *res) { - mlog_entry("dlm=%p, res=%p\n", dlm, res); if (res) { spin_lock(&dlm->spinlock); spin_lock(&res->spinlock); @@ -467,8 +456,6 @@ void dlm_kick_thread(struct dlm_ctxt *dlm, struct dlm_lock_resource *res) void __dlm_dirty_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res) { - mlog_entry("dlm=%p, res=%p\n", dlm, res); - assert_spin_locked(&dlm->spinlock); assert_spin_locked(&res->spinlock); @@ -485,13 +472,16 @@ void __dlm_dirty_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res) res->state |= DLM_LOCK_RES_DIRTY; } } + + mlog(0, "%s: res %.*s\n", dlm->name, res->lockname.len, + res->lockname.name); } /* Launch the NM thread for the mounted volume */ int dlm_launch_thread(struct dlm_ctxt *dlm) { - mlog(0, "starting dlm thread...\n"); + mlog(0, "Starting dlm_thread...\n"); dlm->dlm_thread_task = kthread_run(dlm_thread, dlm, "dlm_thread"); if (IS_ERR(dlm->dlm_thread_task)) { @@ -506,7 +496,7 @@ int dlm_launch_thread(struct dlm_ctxt *dlm) void dlm_complete_thread(struct dlm_ctxt *dlm) { if (dlm->dlm_thread_task) { - mlog(ML_KTHREAD, "waiting for dlm thread to exit\n"); + mlog(ML_KTHREAD, "Waiting for dlm thread to exit\n"); kthread_stop(dlm->dlm_thread_task); dlm->dlm_thread_task = NULL; } @@ -537,7 +527,12 @@ static void dlm_flush_asts(struct dlm_ctxt *dlm) /* get an extra ref on lock */ dlm_lock_get(lock); res = lock->lockres; - mlog(0, "delivering an ast for this lockres\n"); + mlog(0, "%s: res %.*s, Flush AST for lock %u:%llu, type %d, " + "node %u\n", dlm->name, res->lockname.len, + res->lockname.name, + dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)), + dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)), + lock->ml.type, lock->ml.node); BUG_ON(!lock->ast_pending); @@ -558,9 +553,9 @@ static void dlm_flush_asts(struct dlm_ctxt *dlm) /* possible that another ast was queued while * we were delivering the last one */ if (!list_empty(&lock->ast_list)) { - mlog(0, "aha another ast got queued while " - "we were finishing the last one. will " - "keep the ast_pending flag set.\n"); + mlog(0, "%s: res %.*s, AST queued while flushing last " + "one\n", dlm->name, res->lockname.len, + res->lockname.name); } else lock->ast_pending = 0; @@ -591,8 +586,12 @@ static void dlm_flush_asts(struct dlm_ctxt *dlm) dlm_lock_put(lock); spin_unlock(&dlm->ast_lock); - mlog(0, "delivering a bast for this lockres " - "(blocked = %d\n", hi); + mlog(0, "%s: res %.*s, Flush BAST for lock %u:%llu, " + "blocked %d, node %u\n", + dlm->name, res->lockname.len, res->lockname.name, + dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)), + dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)), + hi, lock->ml.node); if (lock->ml.node != dlm->node_num) { ret = dlm_send_proxy_bast(dlm, res, lock, hi); @@ -606,9 +605,9 @@ static void dlm_flush_asts(struct dlm_ctxt *dlm) /* possible that another bast was queued while * we were delivering the last one */ if (!list_empty(&lock->bast_list)) { - mlog(0, "aha another bast got queued while " - "we were finishing the last one. will " - "keep the bast_pending flag set.\n"); + mlog(0, "%s: res %.*s, BAST queued while flushing last " + "one\n", dlm->name, res->lockname.len, + res->lockname.name); } else lock->bast_pending = 0; @@ -672,14 +671,16 @@ static int dlm_thread(void *data) /* lockres can be re-dirtied/re-added to the * dirty_list in this gap, but that is ok */ + spin_lock(&dlm->ast_lock); spin_lock(&res->spinlock); if (res->owner != dlm->node_num) { __dlm_print_one_lock_resource(res); - mlog(ML_ERROR, "inprog:%s, mig:%s, reco:%s, dirty:%s\n", - res->state & DLM_LOCK_RES_IN_PROGRESS ? "yes" : "no", - res->state & DLM_LOCK_RES_MIGRATING ? "yes" : "no", - res->state & DLM_LOCK_RES_RECOVERING ? "yes" : "no", - res->state & DLM_LOCK_RES_DIRTY ? "yes" : "no"); + mlog(ML_ERROR, "%s: inprog %d, mig %d, reco %d," + " dirty %d\n", dlm->name, + !!(res->state & DLM_LOCK_RES_IN_PROGRESS), + !!(res->state & DLM_LOCK_RES_MIGRATING), + !!(res->state & DLM_LOCK_RES_RECOVERING), + !!(res->state & DLM_LOCK_RES_DIRTY)); } BUG_ON(res->owner != dlm->node_num); @@ -692,8 +693,9 @@ static int dlm_thread(void *data) /* move it to the tail and keep going */ res->state &= ~DLM_LOCK_RES_DIRTY; spin_unlock(&res->spinlock); - mlog(0, "delaying list shuffling for in-" - "progress lockres %.*s, state=%d\n", + spin_unlock(&dlm->ast_lock); + mlog(0, "%s: res %.*s, inprogress, delay list " + "shuffle, state %d\n", dlm->name, res->lockname.len, res->lockname.name, res->state); delay = 1; @@ -705,14 +707,11 @@ static int dlm_thread(void *data) * spinlock and do NOT have the dlm lock. * safe to reserve/queue asts and run the lists. */ - mlog(0, "calling dlm_shuffle_lists with dlm=%s, " - "res=%.*s\n", dlm->name, - res->lockname.len, res->lockname.name); - /* called while holding lockres lock */ dlm_shuffle_lists(dlm, res); res->state &= ~DLM_LOCK_RES_DIRTY; spin_unlock(&res->spinlock); + spin_unlock(&dlm->ast_lock); dlm_lockres_calc_usage(dlm, res); @@ -731,7 +730,8 @@ in_progress: /* unlikely, but we may need to give time to * other tasks */ if (!--n) { - mlog(0, "throttling dlm_thread\n"); + mlog(0, "%s: Throttling dlm thread\n", + dlm->name); break; } } diff --git a/fs/ocfs2/dlm/dlmunlock.c b/fs/ocfs2/dlm/dlmunlock.c index fcf879ed693..2e3c9dbab68 100644 --- a/fs/ocfs2/dlm/dlmunlock.c +++ b/fs/ocfs2/dlm/dlmunlock.c @@ -28,9 +28,7 @@ #include <linux/module.h> #include <linux/fs.h> #include <linux/types.h> -#include <linux/slab.h> #include <linux/highmem.h> -#include <linux/utsname.h> #include <linux/init.h> #include <linux/sysctl.h> #include <linux/random.h> @@ -122,7 +120,7 @@ static enum dlm_status dlmunlock_common(struct dlm_ctxt *dlm, * that still has AST's pending... */ in_use = !list_empty(&lock->ast_list); spin_unlock(&dlm->ast_lock); - if (in_use) { + if (in_use && !(flags & LKM_CANCEL)) { mlog(ML_ERROR, "lockres %.*s: Someone is calling dlmunlock " "while waiting for an ast!", res->lockname.len, res->lockname.name); @@ -131,7 +129,7 @@ static enum dlm_status dlmunlock_common(struct dlm_ctxt *dlm, spin_lock(&res->spinlock); if (res->state & DLM_LOCK_RES_IN_PROGRESS) { - if (master_node) { + if (master_node && !(flags & LKM_CANCEL)) { mlog(ML_ERROR, "lockres in progress!\n"); spin_unlock(&res->spinlock); return DLM_FORWARD; @@ -191,9 +189,11 @@ static enum dlm_status dlmunlock_common(struct dlm_ctxt *dlm, actions &= ~(DLM_UNLOCK_REMOVE_LOCK| DLM_UNLOCK_REGRANT_LOCK| DLM_UNLOCK_CLEAR_CONVERT_TYPE); - } else if (status == DLM_RECOVERING || - status == DLM_MIGRATING || - status == DLM_FORWARD) { + } else if (status == DLM_RECOVERING || + status == DLM_MIGRATING || + status == DLM_FORWARD || + status == DLM_NOLOCKMGR + ) { /* must clear the actions because this unlock * is about to be retried. cannot free or do * any list manipulation. */ @@ -202,7 +202,8 @@ static enum dlm_status dlmunlock_common(struct dlm_ctxt *dlm, res->lockname.name, status==DLM_RECOVERING?"recovering": (status==DLM_MIGRATING?"migrating": - "forward")); + (status == DLM_FORWARD ? "forward" : + "nolockmanager"))); actions = 0; } if (flags & LKM_CANCEL) @@ -319,7 +320,7 @@ static enum dlm_status dlm_send_remote_unlock_request(struct dlm_ctxt *dlm, struct kvec vec[2]; size_t veclen = 1; - mlog_entry("%.*s\n", res->lockname.len, res->lockname.name); + mlog(0, "%.*s\n", res->lockname.len, res->lockname.name); if (owner == dlm->node_num) { /* ended up trying to contact ourself. this means @@ -356,7 +357,8 @@ static enum dlm_status dlm_send_remote_unlock_request(struct dlm_ctxt *dlm, mlog(0, "master was in-progress. retry\n"); ret = status; } else { - mlog_errno(tmpret); + mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to " + "node %u\n", tmpret, DLM_UNLOCK_LOCK_MSG, dlm->key, owner); if (dlm_is_host_down(tmpret)) { /* NOTE: this seems strange, but it is what we want. * when the master goes down during a cancel or @@ -365,7 +367,10 @@ static enum dlm_status dlm_send_remote_unlock_request(struct dlm_ctxt *dlm, * updated state to the recovery master. this thread * just needs to finish out the operation and call * the unlockast. */ - ret = DLM_NORMAL; + if (dlm_is_node_dead(dlm, owner)) + ret = DLM_NORMAL; + else + ret = DLM_NOLOCKMGR; } else { /* something bad. this will BUG in ocfs2 */ ret = dlm_err_to_dlm_status(tmpret); @@ -389,7 +394,6 @@ int dlm_unlock_lock_handler(struct o2net_msg *msg, u32 len, void *data, struct dlm_ctxt *dlm = data; struct dlm_unlock_lock *unlock = (struct dlm_unlock_lock *)msg->buf; struct dlm_lock_resource *res = NULL; - struct list_head *iter; struct dlm_lock *lock = NULL; enum dlm_status status = DLM_NORMAL; int found = 0, i; @@ -459,8 +463,7 @@ int dlm_unlock_lock_handler(struct o2net_msg *msg, u32 len, void *data, } for (i=0; i<3; i++) { - list_for_each(iter, queue) { - lock = list_entry(iter, struct dlm_lock, list); + list_for_each_entry(lock, queue, list) { if (lock->ml.cookie == unlock->cookie && lock->ml.node == unlock->node_idx) { dlm_lock_get(lock); @@ -589,8 +592,6 @@ enum dlm_status dlmunlock(struct dlm_ctxt *dlm, struct dlm_lockstatus *lksb, struct dlm_lock *lock = NULL; int call_ast, is_master; - mlog_entry_void(); - if (!lksb) { dlm_error(DLM_BADARGS); return DLM_BADARGS; @@ -643,7 +644,9 @@ retry: if (status == DLM_RECOVERING || status == DLM_MIGRATING || - status == DLM_FORWARD) { + status == DLM_FORWARD || + status == DLM_NOLOCKMGR) { + /* We want to go away for a tiny bit to allow recovery * / migration to complete on this resource. I don't * know of any wait queue we could sleep on as this @@ -655,21 +658,21 @@ retry: msleep(50); mlog(0, "retrying unlock due to pending recovery/" - "migration/in-progress\n"); + "migration/in-progress/reconnect\n"); goto retry; } if (call_ast) { mlog(0, "calling unlockast(%p, %d)\n", data, status); if (is_master) { - /* it is possible that there is one last bast + /* it is possible that there is one last bast * pending. make sure it is flushed, then * call the unlockast. * not an issue if this is a mastered remotely, * since this lock has been removed from the * lockres queues and cannot be found. */ dlm_kick_thread(dlm, NULL); - wait_event(dlm->ast_wq, + wait_event(dlm->ast_wq, dlm_lock_basts_flushed(dlm, lock)); } (*unlockast)(data, status); diff --git a/fs/ocfs2/dlm/dlmver.c b/fs/ocfs2/dlm/dlmver.c deleted file mode 100644 index dfc0da4d158..00000000000 --- a/fs/ocfs2/dlm/dlmver.c +++ /dev/null @@ -1,42 +0,0 @@ -/* -*- mode: c; c-basic-offset: 8; -*- - * vim: noexpandtab sw=8 ts=8 sts=0: - * - * dlmver.c - * - * version string - * - * Copyright (C) 2002, 2005 Oracle. All rights reserved. - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public - * License as published by the Free Software Foundation; either - * version 2 of the License, or (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - * General Public License for more details. - * - * You should have received a copy of the GNU General Public - * License along with this program; if not, write to the - * Free Software Foundation, Inc., 59 Temple Place - Suite 330, - * Boston, MA 021110-1307, USA. - */ - -#include <linux/module.h> -#include <linux/kernel.h> - -#include "dlmver.h" - -#define DLM_BUILD_VERSION "1.5.0" - -#define VERSION_STR "OCFS2 DLM " DLM_BUILD_VERSION - -void dlm_print_version(void) -{ - printk(KERN_INFO "%s\n", VERSION_STR); -} - -MODULE_DESCRIPTION(VERSION_STR); - -MODULE_VERSION(DLM_BUILD_VERSION); diff --git a/fs/ocfs2/dlm/dlmver.h b/fs/ocfs2/dlm/dlmver.h deleted file mode 100644 index f674aee77a1..00000000000 --- a/fs/ocfs2/dlm/dlmver.h +++ /dev/null @@ -1,31 +0,0 @@ -/* -*- mode: c; c-basic-offset: 8; -*- - * vim: noexpandtab sw=8 ts=8 sts=0: - * - * dlmfsver.h - * - * Function prototypes - * - * Copyright (C) 2005 Oracle. All rights reserved. - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public - * License as published by the Free Software Foundation; either - * version 2 of the License, or (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - * General Public License for more details. - * - * You should have received a copy of the GNU General Public - * License along with this program; if not, write to the - * Free Software Foundation, Inc., 59 Temple Place - Suite 330, - * Boston, MA 021110-1307, USA. - */ - -#ifndef DLM_VER_H -#define DLM_VER_H - -void dlm_print_version(void); - -#endif /* DLM_VER_H */ diff --git a/fs/ocfs2/dlmfs/Makefile b/fs/ocfs2/dlmfs/Makefile new file mode 100644 index 00000000000..eed3db8c5b4 --- /dev/null +++ b/fs/ocfs2/dlmfs/Makefile @@ -0,0 +1,5 @@ +ccflags-y := -Ifs/ocfs2 + +obj-$(CONFIG_OCFS2_FS) += ocfs2_dlmfs.o + +ocfs2_dlmfs-objs := userdlm.o dlmfs.o diff --git a/fs/ocfs2/dlm/dlmfs.c b/fs/ocfs2/dlmfs/dlmfs.c index 1c9efb406a9..09b7d9dac71 100644 --- a/fs/ocfs2/dlm/dlmfs.c +++ b/fs/ocfs2/dlmfs/dlmfs.c @@ -43,24 +43,16 @@ #include <linux/init.h> #include <linux/string.h> #include <linux/backing-dev.h> +#include <linux/poll.h> #include <asm/uaccess.h> - -#include "cluster/nodemanager.h" -#include "cluster/heartbeat.h" -#include "cluster/tcp.h" - -#include "dlmapi.h" - +#include "stackglue.h" #include "userdlm.h" -#include "dlmfsver.h" - #define MLOG_MASK_PREFIX ML_DLMFS #include "cluster/masklog.h" -#include "ocfs2_lockingver.h" static const struct super_operations dlmfs_ops; static const struct file_operations dlmfs_file_operations; @@ -71,15 +63,46 @@ static struct kmem_cache *dlmfs_inode_cache; struct workqueue_struct *user_dlm_worker; + + /* - * This is the userdlmfs locking protocol version. + * These are the ABI capabilities of dlmfs. + * + * Over time, dlmfs has added some features that were not part of the + * initial ABI. Unfortunately, some of these features are not detectable + * via standard usage. For example, Linux's default poll always returns + * POLLIN, so there is no way for a caller of poll(2) to know when dlmfs + * added poll support. Instead, we provide this list of new capabilities. * - * See fs/ocfs2/dlmglue.c for more details on locking versions. + * Capabilities is a read-only attribute. We do it as a module parameter + * so we can discover it whether dlmfs is built in, loaded, or even not + * loaded. + * + * The ABI features are local to this machine's dlmfs mount. This is + * distinct from the locking protocol, which is concerned with inter-node + * interaction. + * + * Capabilities: + * - bast : POLLIN against the file descriptor of a held lock + * signifies a bast fired on the lock. */ -static const struct dlm_protocol_version user_locking_protocol = { - .pv_major = OCFS2_LOCKING_PROTOCOL_MAJOR, - .pv_minor = OCFS2_LOCKING_PROTOCOL_MINOR, -}; +#define DLMFS_CAPABILITIES "bast stackglue" +static int param_set_dlmfs_capabilities(const char *val, + struct kernel_param *kp) +{ + printk(KERN_ERR "%s: readonly parameter\n", kp->name); + return -EINVAL; +} +static int param_get_dlmfs_capabilities(char *buffer, + struct kernel_param *kp) +{ + return strlcpy(buffer, DLMFS_CAPABILITIES, + strlen(DLMFS_CAPABILITIES) + 1); +} +module_param_call(capabilities, param_set_dlmfs_capabilities, + param_get_dlmfs_capabilities, NULL, 0444); +MODULE_PARM_DESC(capabilities, DLMFS_CAPABILITIES); + /* * decodes a set of open flags into a valid lock level and a set of flags. @@ -88,20 +111,20 @@ static const struct dlm_protocol_version user_locking_protocol = { * O_RDONLY -> PRMODE level * O_WRONLY -> EXMODE level * - * O_NONBLOCK -> LKM_NOQUEUE + * O_NONBLOCK -> NOQUEUE */ static int dlmfs_decode_open_flags(int open_flags, int *level, int *flags) { if (open_flags & (O_WRONLY|O_RDWR)) - *level = LKM_EXMODE; + *level = DLM_LOCK_EX; else - *level = LKM_PRMODE; + *level = DLM_LOCK_PR; *flags = 0; if (open_flags & O_NONBLOCK) - *flags |= LKM_NOQUEUE; + *flags |= DLM_LKF_NOQUEUE; return 0; } @@ -142,7 +165,7 @@ static int dlmfs_file_open(struct inode *inode, * to be able userspace to be able to distinguish a * valid lock request from one that simply couldn't be * granted. */ - if (flags & LKM_NOQUEUE && status == -EAGAIN) + if (flags & DLM_LKF_NOQUEUE && status == -EAGAIN) status = -ETXTBSY; kfree(fp); goto bail; @@ -158,8 +181,7 @@ static int dlmfs_file_release(struct inode *inode, { int level, status; struct dlmfs_inode_private *ip = DLMFS_I(inode); - struct dlmfs_filp_private *fp = - (struct dlmfs_filp_private *) file->private_data; + struct dlmfs_filp_private *fp = file->private_data; if (S_ISDIR(inode->i_mode)) BUG(); @@ -169,7 +191,7 @@ static int dlmfs_file_release(struct inode *inode, status = 0; if (fp) { level = fp->fp_lock_level; - if (level != LKM_IVMODE) + if (level != DLM_LOCK_IV) user_dlm_cluster_unlock(&ip->ip_lockres, level); kfree(fp); @@ -179,15 +201,50 @@ static int dlmfs_file_release(struct inode *inode, return 0; } +/* + * We do ->setattr() just to override size changes. Our size is the size + * of the LVB and nothing else. + */ +static int dlmfs_file_setattr(struct dentry *dentry, struct iattr *attr) +{ + int error; + struct inode *inode = dentry->d_inode; + + attr->ia_valid &= ~ATTR_SIZE; + error = inode_change_ok(inode, attr); + if (error) + return error; + + setattr_copy(inode, attr); + mark_inode_dirty(inode); + return 0; +} + +static unsigned int dlmfs_file_poll(struct file *file, poll_table *wait) +{ + int event = 0; + struct inode *inode = file_inode(file); + struct dlmfs_inode_private *ip = DLMFS_I(inode); + + poll_wait(file, &ip->ip_lockres.l_event, wait); + + spin_lock(&ip->ip_lockres.l_lock); + if (ip->ip_lockres.l_flags & USER_LOCK_BLOCKED) + event = POLLIN | POLLRDNORM; + spin_unlock(&ip->ip_lockres.l_lock); + + return event; +} + static ssize_t dlmfs_file_read(struct file *filp, char __user *buf, size_t count, loff_t *ppos) { int bytes_left; - ssize_t readlen; + ssize_t readlen, got; char *lvb_buf; - struct inode *inode = filp->f_path.dentry->d_inode; + struct inode *inode = file_inode(filp); mlog(0, "inode %lu, count = %zu, *ppos = %llu\n", inode->i_ino, count, *ppos); @@ -205,15 +262,19 @@ static ssize_t dlmfs_file_read(struct file *filp, if ((count + *ppos) > i_size_read(inode)) readlen = i_size_read(inode) - *ppos; else - readlen = count - *ppos; + readlen = count; lvb_buf = kmalloc(readlen, GFP_NOFS); if (!lvb_buf) return -ENOMEM; - user_dlm_read_lvb(inode, lvb_buf, readlen); - bytes_left = __copy_to_user(buf, lvb_buf, readlen); - readlen -= bytes_left; + got = user_dlm_read_lvb(inode, lvb_buf, readlen); + if (got) { + BUG_ON(got != readlen); + bytes_left = __copy_to_user(buf, lvb_buf, readlen); + readlen -= bytes_left; + } else + readlen = 0; kfree(lvb_buf); @@ -231,7 +292,7 @@ static ssize_t dlmfs_file_write(struct file *filp, int bytes_left; ssize_t writelen; char *lvb_buf; - struct inode *inode = filp->f_path.dentry->d_inode; + struct inode *inode = file_inode(filp); mlog(0, "inode %lu, count = %zu, *ppos = %llu\n", inode->i_ino, count, *ppos); @@ -272,7 +333,7 @@ static void dlmfs_init_once(void *foo) struct dlmfs_inode_private *ip = (struct dlmfs_inode_private *) foo; - ip->ip_dlm = NULL; + ip->ip_conn = NULL; ip->ip_parent = NULL; inode_init_once(&ip->ip_vfs_inode); @@ -289,18 +350,23 @@ static struct inode *dlmfs_alloc_inode(struct super_block *sb) return &ip->ip_vfs_inode; } -static void dlmfs_destroy_inode(struct inode *inode) +static void dlmfs_i_callback(struct rcu_head *head) { + struct inode *inode = container_of(head, struct inode, i_rcu); kmem_cache_free(dlmfs_inode_cache, DLMFS_I(inode)); } -static void dlmfs_clear_inode(struct inode *inode) +static void dlmfs_destroy_inode(struct inode *inode) +{ + call_rcu(&inode->i_rcu, dlmfs_i_callback); +} + +static void dlmfs_evict_inode(struct inode *inode) { int status; struct dlmfs_inode_private *ip; - if (!inode) - return; + clear_inode(inode); mlog(0, "inode %lu\n", inode->i_ino); @@ -314,17 +380,18 @@ static void dlmfs_clear_inode(struct inode *inode) goto clear_fields; } - mlog(0, "we're a directory, ip->ip_dlm = 0x%p\n", ip->ip_dlm); + mlog(0, "we're a directory, ip->ip_conn = 0x%p\n", ip->ip_conn); /* we must be a directory. If required, lets unregister the * dlm context now. */ - if (ip->ip_dlm) - user_dlm_unregister_context(ip->ip_dlm); + if (ip->ip_conn) + user_dlm_unregister(ip->ip_conn); clear_fields: ip->ip_parent = NULL; - ip->ip_dlm = NULL; + ip->ip_conn = NULL; } static struct backing_dev_info dlmfs_backing_dev_info = { + .name = "ocfs2-dlmfs", .ra_pages = 0, /* No readahead */ .capabilities = BDI_CAP_NO_ACCT_AND_WRITEBACK, }; @@ -332,15 +399,11 @@ static struct backing_dev_info dlmfs_backing_dev_info = { static struct inode *dlmfs_get_root_inode(struct super_block *sb) { struct inode *inode = new_inode(sb); - int mode = S_IFDIR | 0755; - struct dlmfs_inode_private *ip; + umode_t mode = S_IFDIR | 0755; if (inode) { - ip = DLMFS_I(inode); - - inode->i_mode = mode; - inode->i_uid = current_fsuid(); - inode->i_gid = current_fsgid(); + inode->i_ino = get_next_ino(); + inode_init_owner(inode, NULL, mode); inode->i_mapping->backing_dev_info = &dlmfs_backing_dev_info; inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME; inc_nlink(inode); @@ -354,7 +417,7 @@ static struct inode *dlmfs_get_root_inode(struct super_block *sb) static struct inode *dlmfs_get_inode(struct inode *parent, struct dentry *dentry, - int mode) + umode_t mode) { struct super_block *sb = parent->i_sb; struct inode * inode = new_inode(sb); @@ -363,14 +426,13 @@ static struct inode *dlmfs_get_inode(struct inode *parent, if (!inode) return NULL; - inode->i_mode = mode; - inode->i_uid = current_fsuid(); - inode->i_gid = current_fsgid(); + inode->i_ino = get_next_ino(); + inode_init_owner(inode, parent, mode); inode->i_mapping->backing_dev_info = &dlmfs_backing_dev_info; inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME; ip = DLMFS_I(inode); - ip->ip_dlm = DLMFS_I(parent)->ip_dlm; + ip->ip_conn = DLMFS_I(parent)->ip_conn; switch (mode & S_IFMT) { default: @@ -402,13 +464,6 @@ static struct inode *dlmfs_get_inode(struct inode *parent, inc_nlink(inode); break; } - - if (parent->i_mode & S_ISGID) { - inode->i_gid = parent->i_gid; - if (S_ISDIR(mode)) - inode->i_mode |= S_ISGID; - } - return inode; } @@ -418,19 +473,18 @@ static struct inode *dlmfs_get_inode(struct inode *parent, /* SMP-safe */ static int dlmfs_mkdir(struct inode * dir, struct dentry * dentry, - int mode) + umode_t mode) { int status; struct inode *inode = NULL; struct qstr *domain = &dentry->d_name; struct dlmfs_inode_private *ip; - struct dlm_ctxt *dlm; - struct dlm_protocol_version proto = user_locking_protocol; + struct ocfs2_cluster_connection *conn; mlog(0, "mkdir %.*s\n", domain->len, domain->name); /* verify that we have a proper domain */ - if (domain->len >= O2NM_MAX_NAME_LEN) { + if (domain->len >= GROUP_NAME_MAX) { status = -EINVAL; mlog(ML_ERROR, "invalid domain name for directory.\n"); goto bail; @@ -445,14 +499,14 @@ static int dlmfs_mkdir(struct inode * dir, ip = DLMFS_I(inode); - dlm = user_dlm_register_context(domain, &proto); - if (IS_ERR(dlm)) { - status = PTR_ERR(dlm); + conn = user_dlm_register(domain); + if (IS_ERR(conn)) { + status = PTR_ERR(conn); mlog(ML_ERROR, "Error %d could not register domain \"%.*s\"\n", status, domain->len, domain->name); goto bail; } - ip->ip_dlm = dlm; + ip->ip_conn = conn; inc_nlink(dir); d_instantiate(dentry, inode); @@ -467,8 +521,8 @@ bail: static int dlmfs_create(struct inode *dir, struct dentry *dentry, - int mode, - struct nameidata *nd) + umode_t mode, + bool excl) { int status = 0; struct inode *inode; @@ -524,32 +578,24 @@ static int dlmfs_fill_super(struct super_block * sb, void * data, int silent) { - struct inode * inode; - struct dentry * root; - sb->s_maxbytes = MAX_LFS_FILESIZE; sb->s_blocksize = PAGE_CACHE_SIZE; sb->s_blocksize_bits = PAGE_CACHE_SHIFT; sb->s_magic = DLMFS_MAGIC; sb->s_op = &dlmfs_ops; - inode = dlmfs_get_root_inode(sb); - if (!inode) - return -ENOMEM; - - root = d_alloc_root(inode); - if (!root) { - iput(inode); + sb->s_root = d_make_root(dlmfs_get_root_inode(sb)); + if (!sb->s_root) return -ENOMEM; - } - sb->s_root = root; return 0; } static const struct file_operations dlmfs_file_operations = { .open = dlmfs_file_open, .release = dlmfs_file_release, + .poll = dlmfs_file_poll, .read = dlmfs_file_read, .write = dlmfs_file_write, + .llseek = default_llseek, }; static const struct inode_operations dlmfs_dir_inode_operations = { @@ -569,34 +615,34 @@ static const struct super_operations dlmfs_ops = { .statfs = simple_statfs, .alloc_inode = dlmfs_alloc_inode, .destroy_inode = dlmfs_destroy_inode, - .clear_inode = dlmfs_clear_inode, + .evict_inode = dlmfs_evict_inode, .drop_inode = generic_delete_inode, }; static const struct inode_operations dlmfs_file_inode_operations = { .getattr = simple_getattr, + .setattr = dlmfs_file_setattr, }; -static int dlmfs_get_sb(struct file_system_type *fs_type, - int flags, const char *dev_name, void *data, struct vfsmount *mnt) +static struct dentry *dlmfs_mount(struct file_system_type *fs_type, + int flags, const char *dev_name, void *data) { - return get_sb_nodev(fs_type, flags, data, dlmfs_fill_super, mnt); + return mount_nodev(fs_type, flags, data, dlmfs_fill_super); } static struct file_system_type dlmfs_fs_type = { .owner = THIS_MODULE, .name = "ocfs2_dlmfs", - .get_sb = dlmfs_get_sb, + .mount = dlmfs_mount, .kill_sb = kill_litter_super, }; +MODULE_ALIAS_FS("ocfs2_dlmfs"); static int __init init_dlmfs_fs(void) { int status; int cleanup_inode = 0, cleanup_worker = 0; - dlmfs_print_version(); - status = bdi_init(&dlmfs_backing_dev_info); if (status) return status; @@ -619,6 +665,7 @@ static int __init init_dlmfs_fs(void) } cleanup_worker = 1; + user_dlm_set_locking_protocol(); status = register_filesystem(&dlmfs_fs_type); bail: if (status) { @@ -639,6 +686,11 @@ static void __exit exit_dlmfs_fs(void) flush_workqueue(user_dlm_worker); destroy_workqueue(user_dlm_worker); + /* + * Make sure all delayed rcu free inodes are flushed before we + * destroy cache. + */ + rcu_barrier(); kmem_cache_destroy(dlmfs_inode_cache); bdi_destroy(&dlmfs_backing_dev_info); @@ -646,6 +698,7 @@ static void __exit exit_dlmfs_fs(void) MODULE_AUTHOR("Oracle"); MODULE_LICENSE("GPL"); +MODULE_DESCRIPTION("OCFS2 DLM-Filesystem"); module_init(init_dlmfs_fs) module_exit(exit_dlmfs_fs) diff --git a/fs/ocfs2/dlm/userdlm.c b/fs/ocfs2/dlmfs/userdlm.c index 4cb1d3dae25..0499e3fb7bd 100644 --- a/fs/ocfs2/dlm/userdlm.c +++ b/fs/ocfs2/dlmfs/userdlm.c @@ -34,18 +34,19 @@ #include <linux/types.h> #include <linux/crc32.h> - -#include "cluster/nodemanager.h" -#include "cluster/heartbeat.h" -#include "cluster/tcp.h" - -#include "dlmapi.h" - +#include "ocfs2_lockingver.h" +#include "stackglue.h" #include "userdlm.h" #define MLOG_MASK_PREFIX ML_DLMFS #include "cluster/masklog.h" + +static inline struct user_lock_res *user_lksb_to_lock_res(struct ocfs2_dlm_lksb *lksb) +{ + return container_of(lksb, struct user_lock_res, l_lksb); +} + static inline int user_check_wait_flag(struct user_lock_res *lockres, int flag) { @@ -73,15 +74,15 @@ static inline void user_wait_on_blocked_lock(struct user_lock_res *lockres) } /* I heart container_of... */ -static inline struct dlm_ctxt * -dlm_ctxt_from_user_lockres(struct user_lock_res *lockres) +static inline struct ocfs2_cluster_connection * +cluster_connection_from_user_lockres(struct user_lock_res *lockres) { struct dlmfs_inode_private *ip; ip = container_of(lockres, struct dlmfs_inode_private, ip_lockres); - return ip->ip_dlm; + return ip->ip_conn; } static struct inode * @@ -103,9 +104,9 @@ static inline void user_recover_from_dlm_error(struct user_lock_res *lockres) } #define user_log_dlm_error(_func, _stat, _lockres) do { \ - mlog(ML_ERROR, "Dlm error \"%s\" while calling %s on " \ - "resource %.*s: %s\n", dlm_errname(_stat), _func, \ - _lockres->l_namelen, _lockres->l_name, dlm_errmsg(_stat)); \ + mlog(ML_ERROR, "Dlm error %d while calling %s on " \ + "resource %.*s\n", _stat, _func, \ + _lockres->l_namelen, _lockres->l_name); \ } while (0) /* WARNING: This function lives in a world where the only three lock @@ -113,34 +114,35 @@ static inline void user_recover_from_dlm_error(struct user_lock_res *lockres) * lock types are added. */ static inline int user_highest_compat_lock_level(int level) { - int new_level = LKM_EXMODE; + int new_level = DLM_LOCK_EX; - if (level == LKM_EXMODE) - new_level = LKM_NLMODE; - else if (level == LKM_PRMODE) - new_level = LKM_PRMODE; + if (level == DLM_LOCK_EX) + new_level = DLM_LOCK_NL; + else if (level == DLM_LOCK_PR) + new_level = DLM_LOCK_PR; return new_level; } -static void user_ast(void *opaque) +static void user_ast(struct ocfs2_dlm_lksb *lksb) { - struct user_lock_res *lockres = opaque; - struct dlm_lockstatus *lksb; + struct user_lock_res *lockres = user_lksb_to_lock_res(lksb); + int status; - mlog(0, "AST fired for lockres %.*s\n", lockres->l_namelen, - lockres->l_name); + mlog(ML_BASTS, "AST fired for lockres %.*s, level %d => %d\n", + lockres->l_namelen, lockres->l_name, lockres->l_level, + lockres->l_requested); spin_lock(&lockres->l_lock); - lksb = &(lockres->l_lksb); - if (lksb->status != DLM_NORMAL) { + status = ocfs2_dlm_lock_status(&lockres->l_lksb); + if (status) { mlog(ML_ERROR, "lksb status value of %u on lockres %.*s\n", - lksb->status, lockres->l_namelen, lockres->l_name); + status, lockres->l_namelen, lockres->l_name); spin_unlock(&lockres->l_lock); return; } - mlog_bug_on_msg(lockres->l_requested == LKM_IVMODE, + mlog_bug_on_msg(lockres->l_requested == DLM_LOCK_IV, "Lockres %.*s, requested ivmode. flags 0x%x\n", lockres->l_namelen, lockres->l_name, lockres->l_flags); @@ -148,13 +150,13 @@ static void user_ast(void *opaque) if (lockres->l_requested < lockres->l_level) { if (lockres->l_requested <= user_highest_compat_lock_level(lockres->l_blocking)) { - lockres->l_blocking = LKM_NLMODE; + lockres->l_blocking = DLM_LOCK_NL; lockres->l_flags &= ~USER_LOCK_BLOCKED; } } lockres->l_level = lockres->l_requested; - lockres->l_requested = LKM_IVMODE; + lockres->l_requested = DLM_LOCK_IV; lockres->l_flags |= USER_LOCK_ATTACHED; lockres->l_flags &= ~USER_LOCK_BUSY; @@ -193,11 +195,11 @@ static void __user_dlm_cond_queue_lockres(struct user_lock_res *lockres) return; switch (lockres->l_blocking) { - case LKM_EXMODE: + case DLM_LOCK_EX: if (!lockres->l_ex_holders && !lockres->l_ro_holders) queue = 1; break; - case LKM_PRMODE: + case DLM_LOCK_PR: if (!lockres->l_ex_holders) queue = 1; break; @@ -209,12 +211,12 @@ static void __user_dlm_cond_queue_lockres(struct user_lock_res *lockres) __user_dlm_queue_lockres(lockres); } -static void user_bast(void *opaque, int level) +static void user_bast(struct ocfs2_dlm_lksb *lksb, int level) { - struct user_lock_res *lockres = opaque; + struct user_lock_res *lockres = user_lksb_to_lock_res(lksb); - mlog(0, "Blocking AST fired for lockres %.*s. Blocking level %d\n", - lockres->l_namelen, lockres->l_name, level); + mlog(ML_BASTS, "BAST fired for lockres %.*s, blocking %d, level %d\n", + lockres->l_namelen, lockres->l_name, level, lockres->l_level); spin_lock(&lockres->l_lock); lockres->l_flags |= USER_LOCK_BLOCKED; @@ -227,15 +229,15 @@ static void user_bast(void *opaque, int level) wake_up(&lockres->l_event); } -static void user_unlock_ast(void *opaque, enum dlm_status status) +static void user_unlock_ast(struct ocfs2_dlm_lksb *lksb, int status) { - struct user_lock_res *lockres = opaque; + struct user_lock_res *lockres = user_lksb_to_lock_res(lksb); - mlog(0, "UNLOCK AST called on lock %.*s\n", lockres->l_namelen, - lockres->l_name); + mlog(ML_BASTS, "UNLOCK AST fired for lockres %.*s, flags 0x%x\n", + lockres->l_namelen, lockres->l_name, lockres->l_flags); - if (status != DLM_NORMAL && status != DLM_CANCELGRANT) - mlog(ML_ERROR, "Dlm returns status %d\n", status); + if (status) + mlog(ML_ERROR, "dlm returns status %d\n", status); spin_lock(&lockres->l_lock); /* The teardown flag gets set early during the unlock process, @@ -243,7 +245,7 @@ static void user_unlock_ast(void *opaque, enum dlm_status status) * for a concurrent cancel. */ if (lockres->l_flags & USER_LOCK_IN_TEARDOWN && !(lockres->l_flags & USER_LOCK_IN_CANCEL)) { - lockres->l_level = LKM_IVMODE; + lockres->l_level = DLM_LOCK_IV; } else if (status == DLM_CANCELGRANT) { /* We tried to cancel a convert request, but it was * already granted. Don't clear the busy flag - the @@ -254,7 +256,7 @@ static void user_unlock_ast(void *opaque, enum dlm_status status) } else { BUG_ON(!(lockres->l_flags & USER_LOCK_IN_CANCEL)); /* Cancel succeeded, we want to re-queue */ - lockres->l_requested = LKM_IVMODE; /* cancel an + lockres->l_requested = DLM_LOCK_IV; /* cancel an * upconvert * request. */ lockres->l_flags &= ~USER_LOCK_IN_CANCEL; @@ -271,6 +273,21 @@ out_noclear: wake_up(&lockres->l_event); } +/* + * This is the userdlmfs locking protocol version. + * + * See fs/ocfs2/dlmglue.c for more details on locking versions. + */ +static struct ocfs2_locking_protocol user_dlm_lproto = { + .lp_max_version = { + .pv_major = OCFS2_LOCKING_PROTOCOL_MAJOR, + .pv_minor = OCFS2_LOCKING_PROTOCOL_MINOR, + }, + .lp_lock_ast = user_ast, + .lp_blocking_ast = user_bast, + .lp_unlock_ast = user_unlock_ast, +}; + static inline void user_dlm_drop_inode_ref(struct user_lock_res *lockres) { struct inode *inode; @@ -283,10 +300,10 @@ static void user_dlm_unblock_lock(struct work_struct *work) int new_level, status; struct user_lock_res *lockres = container_of(work, struct user_lock_res, l_work); - struct dlm_ctxt *dlm = dlm_ctxt_from_user_lockres(lockres); + struct ocfs2_cluster_connection *conn = + cluster_connection_from_user_lockres(lockres); - mlog(0, "processing lockres %.*s\n", lockres->l_namelen, - lockres->l_name); + mlog(0, "lockres %.*s\n", lockres->l_namelen, lockres->l_name); spin_lock(&lockres->l_lock); @@ -304,17 +321,23 @@ static void user_dlm_unblock_lock(struct work_struct *work) * flag, and finally we might get another bast which re-queues * us before our ast for the downconvert is called. */ if (!(lockres->l_flags & USER_LOCK_BLOCKED)) { + mlog(ML_BASTS, "lockres %.*s USER_LOCK_BLOCKED\n", + lockres->l_namelen, lockres->l_name); spin_unlock(&lockres->l_lock); goto drop_ref; } if (lockres->l_flags & USER_LOCK_IN_TEARDOWN) { + mlog(ML_BASTS, "lockres %.*s USER_LOCK_IN_TEARDOWN\n", + lockres->l_namelen, lockres->l_name); spin_unlock(&lockres->l_lock); goto drop_ref; } if (lockres->l_flags & USER_LOCK_BUSY) { if (lockres->l_flags & USER_LOCK_IN_CANCEL) { + mlog(ML_BASTS, "lockres %.*s USER_LOCK_IN_CANCEL\n", + lockres->l_namelen, lockres->l_name); spin_unlock(&lockres->l_lock); goto drop_ref; } @@ -322,32 +345,31 @@ static void user_dlm_unblock_lock(struct work_struct *work) lockres->l_flags |= USER_LOCK_IN_CANCEL; spin_unlock(&lockres->l_lock); - status = dlmunlock(dlm, - &lockres->l_lksb, - LKM_CANCEL, - user_unlock_ast, - lockres); - if (status != DLM_NORMAL) - user_log_dlm_error("dlmunlock", status, lockres); + status = ocfs2_dlm_unlock(conn, &lockres->l_lksb, + DLM_LKF_CANCEL); + if (status) + user_log_dlm_error("ocfs2_dlm_unlock", status, lockres); goto drop_ref; } /* If there are still incompat holders, we can exit safely * without worrying about re-queueing this lock as that will * happen on the last call to user_cluster_unlock. */ - if ((lockres->l_blocking == LKM_EXMODE) + if ((lockres->l_blocking == DLM_LOCK_EX) && (lockres->l_ex_holders || lockres->l_ro_holders)) { spin_unlock(&lockres->l_lock); - mlog(0, "can't downconvert for ex: ro = %u, ex = %u\n", - lockres->l_ro_holders, lockres->l_ex_holders); + mlog(ML_BASTS, "lockres %.*s, EX/PR Holders %u,%u\n", + lockres->l_namelen, lockres->l_name, + lockres->l_ex_holders, lockres->l_ro_holders); goto drop_ref; } - if ((lockres->l_blocking == LKM_PRMODE) + if ((lockres->l_blocking == DLM_LOCK_PR) && lockres->l_ex_holders) { spin_unlock(&lockres->l_lock); - mlog(0, "can't downconvert for pr: ex = %u\n", - lockres->l_ex_holders); + mlog(ML_BASTS, "lockres %.*s, EX Holders %u\n", + lockres->l_namelen, lockres->l_name, + lockres->l_ex_holders); goto drop_ref; } @@ -355,22 +377,17 @@ static void user_dlm_unblock_lock(struct work_struct *work) new_level = user_highest_compat_lock_level(lockres->l_blocking); lockres->l_requested = new_level; lockres->l_flags |= USER_LOCK_BUSY; - mlog(0, "Downconvert lock from %d to %d\n", - lockres->l_level, new_level); + mlog(ML_BASTS, "lockres %.*s, downconvert %d => %d\n", + lockres->l_namelen, lockres->l_name, lockres->l_level, new_level); spin_unlock(&lockres->l_lock); /* need lock downconvert request now... */ - status = dlmlock(dlm, - new_level, - &lockres->l_lksb, - LKM_CONVERT|LKM_VALBLK, - lockres->l_name, - lockres->l_namelen, - user_ast, - lockres, - user_bast); - if (status != DLM_NORMAL) { - user_log_dlm_error("dlmlock", status, lockres); + status = ocfs2_dlm_lock(conn, new_level, &lockres->l_lksb, + DLM_LKF_CONVERT|DLM_LKF_VALBLK, + lockres->l_name, + lockres->l_namelen); + if (status) { + user_log_dlm_error("ocfs2_dlm_lock", status, lockres); user_recover_from_dlm_error(lockres); } @@ -382,10 +399,10 @@ static inline void user_dlm_inc_holders(struct user_lock_res *lockres, int level) { switch(level) { - case LKM_EXMODE: + case DLM_LOCK_EX: lockres->l_ex_holders++; break; - case LKM_PRMODE: + case DLM_LOCK_PR: lockres->l_ro_holders++; break; default: @@ -410,20 +427,19 @@ int user_dlm_cluster_lock(struct user_lock_res *lockres, int lkm_flags) { int status, local_flags; - struct dlm_ctxt *dlm = dlm_ctxt_from_user_lockres(lockres); + struct ocfs2_cluster_connection *conn = + cluster_connection_from_user_lockres(lockres); - if (level != LKM_EXMODE && - level != LKM_PRMODE) { + if (level != DLM_LOCK_EX && + level != DLM_LOCK_PR) { mlog(ML_ERROR, "lockres %.*s: invalid request!\n", lockres->l_namelen, lockres->l_name); status = -EINVAL; goto bail; } - mlog(0, "lockres %.*s: asking for %s lock, passed flags = 0x%x\n", - lockres->l_namelen, lockres->l_name, - (level == LKM_EXMODE) ? "LKM_EXMODE" : "LKM_PRMODE", - lkm_flags); + mlog(ML_BASTS, "lockres %.*s, level %d, flags = 0x%x\n", + lockres->l_namelen, lockres->l_name, level, lkm_flags); again: if (signal_pending(current)) { @@ -457,35 +473,26 @@ again: } if (level > lockres->l_level) { - local_flags = lkm_flags | LKM_VALBLK; - if (lockres->l_level != LKM_IVMODE) - local_flags |= LKM_CONVERT; + local_flags = lkm_flags | DLM_LKF_VALBLK; + if (lockres->l_level != DLM_LOCK_IV) + local_flags |= DLM_LKF_CONVERT; lockres->l_requested = level; lockres->l_flags |= USER_LOCK_BUSY; spin_unlock(&lockres->l_lock); - BUG_ON(level == LKM_IVMODE); - BUG_ON(level == LKM_NLMODE); + BUG_ON(level == DLM_LOCK_IV); + BUG_ON(level == DLM_LOCK_NL); /* call dlm_lock to upgrade lock now */ - status = dlmlock(dlm, - level, - &lockres->l_lksb, - local_flags, - lockres->l_name, - lockres->l_namelen, - user_ast, - lockres, - user_bast); - if (status != DLM_NORMAL) { - if ((lkm_flags & LKM_NOQUEUE) && - (status == DLM_NOTQUEUED)) - status = -EAGAIN; - else { - user_log_dlm_error("dlmlock", status, lockres); - status = -EINVAL; - } + status = ocfs2_dlm_lock(conn, level, &lockres->l_lksb, + local_flags, lockres->l_name, + lockres->l_namelen); + if (status) { + if ((lkm_flags & DLM_LKF_NOQUEUE) && + (status != -EAGAIN)) + user_log_dlm_error("ocfs2_dlm_lock", + status, lockres); user_recover_from_dlm_error(lockres); goto bail; } @@ -506,11 +513,11 @@ static inline void user_dlm_dec_holders(struct user_lock_res *lockres, int level) { switch(level) { - case LKM_EXMODE: + case DLM_LOCK_EX: BUG_ON(!lockres->l_ex_holders); lockres->l_ex_holders--; break; - case LKM_PRMODE: + case DLM_LOCK_PR: BUG_ON(!lockres->l_ro_holders); lockres->l_ro_holders--; break; @@ -522,8 +529,8 @@ static inline void user_dlm_dec_holders(struct user_lock_res *lockres, void user_dlm_cluster_unlock(struct user_lock_res *lockres, int level) { - if (level != LKM_EXMODE && - level != LKM_PRMODE) { + if (level != DLM_LOCK_EX && + level != DLM_LOCK_PR) { mlog(ML_ERROR, "lockres %.*s: invalid request!\n", lockres->l_namelen, lockres->l_name); return; @@ -540,33 +547,40 @@ void user_dlm_write_lvb(struct inode *inode, unsigned int len) { struct user_lock_res *lockres = &DLMFS_I(inode)->ip_lockres; - char *lvb = lockres->l_lksb.lvb; + char *lvb; BUG_ON(len > DLM_LVB_LEN); spin_lock(&lockres->l_lock); - BUG_ON(lockres->l_level < LKM_EXMODE); + BUG_ON(lockres->l_level < DLM_LOCK_EX); + lvb = ocfs2_dlm_lvb(&lockres->l_lksb); memcpy(lvb, val, len); spin_unlock(&lockres->l_lock); } -void user_dlm_read_lvb(struct inode *inode, - char *val, - unsigned int len) +ssize_t user_dlm_read_lvb(struct inode *inode, + char *val, + unsigned int len) { struct user_lock_res *lockres = &DLMFS_I(inode)->ip_lockres; - char *lvb = lockres->l_lksb.lvb; + char *lvb; + ssize_t ret = len; BUG_ON(len > DLM_LVB_LEN); spin_lock(&lockres->l_lock); - BUG_ON(lockres->l_level < LKM_PRMODE); - memcpy(val, lvb, len); + BUG_ON(lockres->l_level < DLM_LOCK_PR); + if (ocfs2_dlm_lvb_valid(&lockres->l_lksb)) { + lvb = ocfs2_dlm_lvb(&lockres->l_lksb); + memcpy(val, lvb, len); + } else + ret = 0; spin_unlock(&lockres->l_lock); + return ret; } void user_dlm_lock_res_init(struct user_lock_res *lockres, @@ -576,9 +590,9 @@ void user_dlm_lock_res_init(struct user_lock_res *lockres, spin_lock_init(&lockres->l_lock); init_waitqueue_head(&lockres->l_event); - lockres->l_level = LKM_IVMODE; - lockres->l_requested = LKM_IVMODE; - lockres->l_blocking = LKM_IVMODE; + lockres->l_level = DLM_LOCK_IV; + lockres->l_requested = DLM_LOCK_IV; + lockres->l_blocking = DLM_LOCK_IV; /* should have been checked before getting here. */ BUG_ON(dentry->d_name.len >= USER_DLM_LOCK_ID_MAX_LEN); @@ -592,9 +606,10 @@ void user_dlm_lock_res_init(struct user_lock_res *lockres, int user_dlm_destroy_lock(struct user_lock_res *lockres) { int status = -EBUSY; - struct dlm_ctxt *dlm = dlm_ctxt_from_user_lockres(lockres); + struct ocfs2_cluster_connection *conn = + cluster_connection_from_user_lockres(lockres); - mlog(0, "asked to destroy %.*s\n", lockres->l_namelen, lockres->l_name); + mlog(ML_BASTS, "lockres %.*s\n", lockres->l_namelen, lockres->l_name); spin_lock(&lockres->l_lock); if (lockres->l_flags & USER_LOCK_IN_TEARDOWN) { @@ -627,14 +642,9 @@ int user_dlm_destroy_lock(struct user_lock_res *lockres) lockres->l_flags |= USER_LOCK_BUSY; spin_unlock(&lockres->l_lock); - status = dlmunlock(dlm, - &lockres->l_lksb, - LKM_VALBLK, - user_unlock_ast, - lockres); - if (status != DLM_NORMAL) { - user_log_dlm_error("dlmunlock", status, lockres); - status = -EINVAL; + status = ocfs2_dlm_unlock(conn, &lockres->l_lksb, DLM_LKF_VALBLK); + if (status) { + user_log_dlm_error("ocfs2_dlm_unlock", status, lockres); goto bail; } @@ -645,32 +655,34 @@ bail: return status; } -struct dlm_ctxt *user_dlm_register_context(struct qstr *name, - struct dlm_protocol_version *proto) +static void user_dlm_recovery_handler_noop(int node_num, + void *recovery_data) { - struct dlm_ctxt *dlm; - u32 dlm_key; - char *domain; - - domain = kmalloc(name->len + 1, GFP_NOFS); - if (!domain) { - mlog_errno(-ENOMEM); - return ERR_PTR(-ENOMEM); - } + /* We ignore recovery events */ + return; +} - dlm_key = crc32_le(0, name->name, name->len); +void user_dlm_set_locking_protocol(void) +{ + ocfs2_stack_glue_set_max_proto_version(&user_dlm_lproto.lp_max_version); +} - snprintf(domain, name->len + 1, "%.*s", name->len, name->name); +struct ocfs2_cluster_connection *user_dlm_register(struct qstr *name) +{ + int rc; + struct ocfs2_cluster_connection *conn; - dlm = dlm_register_domain(domain, dlm_key, proto); - if (IS_ERR(dlm)) - mlog_errno(PTR_ERR(dlm)); + rc = ocfs2_cluster_connect_agnostic(name->name, name->len, + &user_dlm_lproto, + user_dlm_recovery_handler_noop, + NULL, &conn); + if (rc) + mlog_errno(rc); - kfree(domain); - return dlm; + return rc ? ERR_PTR(rc) : conn; } -void user_dlm_unregister_context(struct dlm_ctxt *dlm) +void user_dlm_unregister(struct ocfs2_cluster_connection *conn) { - dlm_unregister_domain(dlm); + ocfs2_cluster_disconnect(conn, 0); } diff --git a/fs/ocfs2/dlm/userdlm.h b/fs/ocfs2/dlmfs/userdlm.h index 0c3cc03c61f..3b42d79531d 100644 --- a/fs/ocfs2/dlm/userdlm.h +++ b/fs/ocfs2/dlmfs/userdlm.h @@ -57,7 +57,7 @@ struct user_lock_res { int l_level; unsigned int l_ro_holders; unsigned int l_ex_holders; - struct dlm_lockstatus l_lksb; + struct ocfs2_dlm_lksb l_lksb; int l_requested; int l_blocking; @@ -80,15 +80,15 @@ void user_dlm_cluster_unlock(struct user_lock_res *lockres, void user_dlm_write_lvb(struct inode *inode, const char *val, unsigned int len); -void user_dlm_read_lvb(struct inode *inode, - char *val, - unsigned int len); -struct dlm_ctxt *user_dlm_register_context(struct qstr *name, - struct dlm_protocol_version *proto); -void user_dlm_unregister_context(struct dlm_ctxt *dlm); +ssize_t user_dlm_read_lvb(struct inode *inode, + char *val, + unsigned int len); +struct ocfs2_cluster_connection *user_dlm_register(struct qstr *name); +void user_dlm_unregister(struct ocfs2_cluster_connection *conn); +void user_dlm_set_locking_protocol(void); struct dlmfs_inode_private { - struct dlm_ctxt *ip_dlm; + struct ocfs2_cluster_connection *ip_conn; struct user_lock_res ip_lockres; /* unused for directories. */ struct inode *ip_parent; diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c index e15fc7d5082..52cfe99ae05 100644 --- a/fs/ocfs2/dlmglue.c +++ b/fs/ocfs2/dlmglue.c @@ -53,6 +53,7 @@ #include "super.h" #include "uptodate.h" #include "quota.h" +#include "refcounttree.h" #include "buffer_head_io.h" @@ -63,7 +64,7 @@ struct ocfs2_mask_waiter { unsigned long mw_mask; unsigned long mw_goal; #ifdef CONFIG_OCFS2_FS_STATS - unsigned long long mw_lock_start; + ktime_t mw_lock_start; #endif }; @@ -92,6 +93,9 @@ struct ocfs2_unblock_ctl { enum ocfs2_unblock_action unblock_action; }; +/* Lockdep class keys */ +struct lock_class_key lockdep_keys[OCFS2_NUM_LOCK_TYPES]; + static int ocfs2_check_meta_downconvert(struct ocfs2_lock_res *lockres, int new_level); static void ocfs2_set_meta_lvb(struct ocfs2_lock_res *lockres); @@ -107,6 +111,11 @@ static void ocfs2_dentry_post_unlock(struct ocfs2_super *osb, static void ocfs2_set_qinfo_lvb(struct ocfs2_lock_res *lockres); +static int ocfs2_check_refcount_downconvert(struct ocfs2_lock_res *lockres, + int new_level); +static int ocfs2_refcount_convert_worker(struct ocfs2_lock_res *lockres, + int blocking); + #define mlog_meta_lvb(__level, __lockres) ocfs2_dump_meta_lvb_info(__level, __PRETTY_FUNCTION__, __LINE__, __lockres) /* This aids in debugging situations where a bad LVB might be involved. */ @@ -248,6 +257,10 @@ static struct ocfs2_lock_res_ops ocfs2_nfs_sync_lops = { .flags = 0, }; +static struct ocfs2_lock_res_ops ocfs2_orphan_scan_lops = { + .flags = LOCK_TYPE_REQUIRES_REFRESH|LOCK_TYPE_USES_LVB, +}; + static struct ocfs2_lock_res_ops ocfs2_dentry_lops = { .get_osb = ocfs2_get_dentry_osb, .post_unlock = ocfs2_dentry_post_unlock, @@ -271,6 +284,12 @@ static struct ocfs2_lock_res_ops ocfs2_qinfo_lops = { .flags = LOCK_TYPE_REQUIRES_REFRESH | LOCK_TYPE_USES_LVB, }; +static struct ocfs2_lock_res_ops ocfs2_refcount_block_lops = { + .check_downconvert = ocfs2_check_refcount_downconvert, + .downconvert_worker = ocfs2_refcount_convert_worker, + .flags = 0, +}; + static inline int ocfs2_is_inode_lock(struct ocfs2_lock_res *lockres) { return lockres->l_type == OCFS2_LOCK_TYPE_META || @@ -278,6 +297,11 @@ static inline int ocfs2_is_inode_lock(struct ocfs2_lock_res *lockres) lockres->l_type == OCFS2_LOCK_TYPE_OPEN; } +static inline struct ocfs2_lock_res *ocfs2_lksb_to_lock_res(struct ocfs2_dlm_lksb *lksb) +{ + return container_of(lksb, struct ocfs2_lock_res, l_lksb); +} + static inline struct inode *ocfs2_lock_res_inode(struct ocfs2_lock_res *lockres) { BUG_ON(!ocfs2_is_inode_lock(lockres)); @@ -299,6 +323,12 @@ static inline struct ocfs2_mem_dqinfo *ocfs2_lock_res_qinfo(struct ocfs2_lock_re return (struct ocfs2_mem_dqinfo *)lockres->l_priv; } +static inline struct ocfs2_refcount_tree * +ocfs2_lock_res_refcount_tree(struct ocfs2_lock_res *res) +{ + return container_of(res, struct ocfs2_refcount_tree, rf_lockres); +} + static inline struct ocfs2_super *ocfs2_get_lockres_osb(struct ocfs2_lock_res *lockres) { if (lockres->l_ops->get_osb) @@ -313,9 +343,16 @@ static int ocfs2_lock_create(struct ocfs2_super *osb, u32 dlm_flags); static inline int ocfs2_may_continue_on_blocked_lock(struct ocfs2_lock_res *lockres, int wanted); -static void ocfs2_cluster_unlock(struct ocfs2_super *osb, - struct ocfs2_lock_res *lockres, - int level); +static void __ocfs2_cluster_unlock(struct ocfs2_super *osb, + struct ocfs2_lock_res *lockres, + int level, unsigned long caller_ip); +static inline void ocfs2_cluster_unlock(struct ocfs2_super *osb, + struct ocfs2_lock_res *lockres, + int level) +{ + __ocfs2_cluster_unlock(osb, lockres, level, _RET_IP_); +} + static inline void ocfs2_generic_handle_downconvert_action(struct ocfs2_lock_res *lockres); static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lockres); static inline void ocfs2_generic_handle_attach_action(struct ocfs2_lock_res *lockres); @@ -360,8 +397,6 @@ static void ocfs2_build_lock_name(enum ocfs2_lock_type type, { int len; - mlog_entry_void(); - BUG_ON(type >= OCFS2_NUM_LOCK_TYPES); len = snprintf(name, OCFS2_LOCK_ID_MAX_LEN, "%c%s%016llx%08x", @@ -371,8 +406,6 @@ static void ocfs2_build_lock_name(enum ocfs2_lock_type type, BUG_ON(len != (OCFS2_LOCK_ID_MAX_LEN - 1)); mlog(0, "built lock resource with name: %s\n", name); - - mlog_exit_void(); } static DEFINE_SPINLOCK(ocfs2_dlm_tracking_lock); @@ -398,44 +431,41 @@ static void ocfs2_remove_lockres_tracking(struct ocfs2_lock_res *res) #ifdef CONFIG_OCFS2_FS_STATS static void ocfs2_init_lock_stats(struct ocfs2_lock_res *res) { - res->l_lock_num_prmode = 0; - res->l_lock_num_prmode_failed = 0; - res->l_lock_total_prmode = 0; - res->l_lock_max_prmode = 0; - res->l_lock_num_exmode = 0; - res->l_lock_num_exmode_failed = 0; - res->l_lock_total_exmode = 0; - res->l_lock_max_exmode = 0; res->l_lock_refresh = 0; + memset(&res->l_lock_prmode, 0, sizeof(struct ocfs2_lock_stats)); + memset(&res->l_lock_exmode, 0, sizeof(struct ocfs2_lock_stats)); } static void ocfs2_update_lock_stats(struct ocfs2_lock_res *res, int level, struct ocfs2_mask_waiter *mw, int ret) { - unsigned long long *num, *sum; - unsigned int *max, *failed; - struct timespec ts = current_kernel_time(); - unsigned long long time = timespec_to_ns(&ts) - mw->mw_lock_start; - - if (level == LKM_PRMODE) { - num = &res->l_lock_num_prmode; - sum = &res->l_lock_total_prmode; - max = &res->l_lock_max_prmode; - failed = &res->l_lock_num_prmode_failed; - } else if (level == LKM_EXMODE) { - num = &res->l_lock_num_exmode; - sum = &res->l_lock_total_exmode; - max = &res->l_lock_max_exmode; - failed = &res->l_lock_num_exmode_failed; - } else + u32 usec; + ktime_t kt; + struct ocfs2_lock_stats *stats; + + if (level == LKM_PRMODE) + stats = &res->l_lock_prmode; + else if (level == LKM_EXMODE) + stats = &res->l_lock_exmode; + else return; - (*num)++; - (*sum) += time; - if (time > *max) - *max = time; + kt = ktime_sub(ktime_get(), mw->mw_lock_start); + usec = ktime_to_us(kt); + + stats->ls_gets++; + stats->ls_total += ktime_to_ns(kt); + /* overflow */ + if (unlikely(stats->ls_gets == 0)) { + stats->ls_gets++; + stats->ls_total = ktime_to_ns(kt); + } + + if (stats->ls_max < usec) + stats->ls_max = usec; + if (ret) - (*failed)++; + stats->ls_fail++; } static inline void ocfs2_track_lock_refresh(struct ocfs2_lock_res *lockres) @@ -445,8 +475,7 @@ static inline void ocfs2_track_lock_refresh(struct ocfs2_lock_res *lockres) static inline void ocfs2_init_start_time(struct ocfs2_mask_waiter *mw) { - struct timespec ts = current_kernel_time(); - mw->mw_lock_start = timespec_to_ns(&ts); + mw->mw_lock_start = ktime_get(); } #else static inline void ocfs2_init_lock_stats(struct ocfs2_lock_res *res) @@ -485,6 +514,13 @@ static void ocfs2_lock_res_init_common(struct ocfs2_super *osb, ocfs2_add_lockres_tracking(res, osb->osb_dlm_debug); ocfs2_init_lock_stats(res); +#ifdef CONFIG_DEBUG_LOCK_ALLOC + if (type != OCFS2_LOCK_TYPE_OPEN) + lockdep_init_map(&res->l_lockdep_map, ocfs2_lock_type_strings[type], + &lockdep_keys[type], 0); + else + res->l_lockdep_map.key = NULL; +#endif } void ocfs2_lock_res_init_once(struct ocfs2_lock_res *res) @@ -637,6 +673,15 @@ static void ocfs2_nfs_sync_lock_res_init(struct ocfs2_lock_res *res, &ocfs2_nfs_sync_lops, osb); } +static void ocfs2_orphan_scan_lock_res_init(struct ocfs2_lock_res *res, + struct ocfs2_super *osb) +{ + ocfs2_lock_res_init_once(res); + ocfs2_build_lock_name(OCFS2_LOCK_TYPE_ORPHAN_SCAN, 0, 0, res->l_name); + ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_ORPHAN_SCAN, + &ocfs2_orphan_scan_lops, osb); +} + void ocfs2_file_lock_res_init(struct ocfs2_lock_res *lockres, struct ocfs2_file_private *fp) { @@ -663,10 +708,19 @@ void ocfs2_qinfo_lock_res_init(struct ocfs2_lock_res *lockres, info); } -void ocfs2_lock_res_free(struct ocfs2_lock_res *res) +void ocfs2_refcount_lock_res_init(struct ocfs2_lock_res *lockres, + struct ocfs2_super *osb, u64 ref_blkno, + unsigned int generation) { - mlog_entry_void(); + ocfs2_lock_res_init_once(lockres); + ocfs2_build_lock_name(OCFS2_LOCK_TYPE_REFCOUNT, ref_blkno, + generation, lockres->l_name); + ocfs2_lock_res_init_common(osb, lockres, OCFS2_LOCK_TYPE_REFCOUNT, + &ocfs2_refcount_block_lops, osb); +} +void ocfs2_lock_res_free(struct ocfs2_lock_res *res) +{ if (!(res->l_flags & OCFS2_LOCK_INITIALIZED)) return; @@ -692,14 +746,11 @@ void ocfs2_lock_res_free(struct ocfs2_lock_res *res) memset(&res->l_lksb, 0, sizeof(res->l_lksb)); res->l_flags = 0UL; - mlog_exit_void(); } static inline void ocfs2_inc_holders(struct ocfs2_lock_res *lockres, int level) { - mlog_entry_void(); - BUG_ON(!lockres); switch(level) { @@ -712,15 +763,11 @@ static inline void ocfs2_inc_holders(struct ocfs2_lock_res *lockres, default: BUG(); } - - mlog_exit_void(); } static inline void ocfs2_dec_holders(struct ocfs2_lock_res *lockres, int level) { - mlog_entry_void(); - BUG_ON(!lockres); switch(level) { @@ -735,7 +782,6 @@ static inline void ocfs2_dec_holders(struct ocfs2_lock_res *lockres, default: BUG(); } - mlog_exit_void(); } /* WARNING: This function lives in a world where the only three lock @@ -782,8 +828,6 @@ static void lockres_clear_flags(struct ocfs2_lock_res *lockres, static inline void ocfs2_generic_handle_downconvert_action(struct ocfs2_lock_res *lockres) { - mlog_entry_void(); - BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY)); BUG_ON(!(lockres->l_flags & OCFS2_LOCK_ATTACHED)); BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED)); @@ -796,14 +840,10 @@ static inline void ocfs2_generic_handle_downconvert_action(struct ocfs2_lock_res lockres_clear_flags(lockres, OCFS2_LOCK_BLOCKED); } lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); - - mlog_exit_void(); } static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lockres) { - mlog_entry_void(); - BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY)); BUG_ON(!(lockres->l_flags & OCFS2_LOCK_ATTACHED)); @@ -816,15 +856,19 @@ static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lo lockres_or_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH); lockres->l_level = lockres->l_requested; - lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); - mlog_exit_void(); + /* + * We set the OCFS2_LOCK_UPCONVERT_FINISHING flag before clearing + * the OCFS2_LOCK_BUSY flag to prevent the dc thread from + * downconverting the lock before the upconvert has fully completed. + */ + lockres_or_flags(lockres, OCFS2_LOCK_UPCONVERT_FINISHING); + + lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); } static inline void ocfs2_generic_handle_attach_action(struct ocfs2_lock_res *lockres) { - mlog_entry_void(); - BUG_ON((!(lockres->l_flags & OCFS2_LOCK_BUSY))); BUG_ON(lockres->l_flags & OCFS2_LOCK_ATTACHED); @@ -836,20 +880,15 @@ static inline void ocfs2_generic_handle_attach_action(struct ocfs2_lock_res *loc lockres->l_level = lockres->l_requested; lockres_or_flags(lockres, OCFS2_LOCK_ATTACHED); lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); - - mlog_exit_void(); } static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres, int level) { int needs_downconvert = 0; - mlog_entry_void(); assert_spin_locked(&lockres->l_lock); - lockres_or_flags(lockres, OCFS2_LOCK_BLOCKED); - if (level > lockres->l_blocking) { /* only schedule a downconvert if we haven't already scheduled * one that goes low enough to satisfy the level we're @@ -862,7 +901,13 @@ static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres, lockres->l_blocking = level; } - mlog_exit(needs_downconvert); + mlog(ML_BASTS, "lockres %s, block %d, level %d, l_block %d, dwn %d\n", + lockres->l_name, level, lockres->l_level, lockres->l_blocking, + needs_downconvert); + + if (needs_downconvert) + lockres_or_flags(lockres, OCFS2_LOCK_BLOCKED); + mlog(0, "needs_downconvert = %d\n", needs_downconvert); return needs_downconvert; } @@ -972,18 +1017,17 @@ static unsigned int lockres_set_pending(struct ocfs2_lock_res *lockres) return lockres->l_pending_gen; } - -static void ocfs2_blocking_ast(void *opaque, int level) +static void ocfs2_blocking_ast(struct ocfs2_dlm_lksb *lksb, int level) { - struct ocfs2_lock_res *lockres = opaque; + struct ocfs2_lock_res *lockres = ocfs2_lksb_to_lock_res(lksb); struct ocfs2_super *osb = ocfs2_get_lockres_osb(lockres); int needs_downconvert; unsigned long flags; BUG_ON(level <= DLM_LOCK_NL); - mlog(0, "BAST fired for lockres %s, blocking %d, level %d type %s\n", - lockres->l_name, level, lockres->l_level, + mlog(ML_BASTS, "BAST fired for lockres %s, blocking %d, level %d, " + "type %s\n", lockres->l_name, level, lockres->l_level, ocfs2_lock_type_string(lockres->l_type)); /* @@ -1004,9 +1048,9 @@ static void ocfs2_blocking_ast(void *opaque, int level) ocfs2_wake_downconvert_thread(osb); } -static void ocfs2_locking_ast(void *opaque) +static void ocfs2_locking_ast(struct ocfs2_dlm_lksb *lksb) { - struct ocfs2_lock_res *lockres = opaque; + struct ocfs2_lock_res *lockres = ocfs2_lksb_to_lock_res(lksb); struct ocfs2_super *osb = ocfs2_get_lockres_osb(lockres); unsigned long flags; int status; @@ -1027,6 +1071,10 @@ static void ocfs2_locking_ast(void *opaque) return; } + mlog(ML_BASTS, "AST fired for lockres %s, action %d, unlock %d, " + "level %d => %d\n", lockres->l_name, lockres->l_action, + lockres->l_unlock_action, lockres->l_level, lockres->l_requested); + switch(lockres->l_action) { case OCFS2_AST_ATTACH: ocfs2_generic_handle_attach_action(lockres); @@ -1039,8 +1087,8 @@ static void ocfs2_locking_ast(void *opaque) ocfs2_generic_handle_downconvert_action(lockres); break; default: - mlog(ML_ERROR, "lockres %s: ast fired with invalid action: %u " - "lockres flags = 0x%lx, unlock action: %u\n", + mlog(ML_ERROR, "lockres %s: AST fired with invalid action: %u, " + "flags 0x%lx, unlock: %u\n", lockres->l_name, lockres->l_action, lockres->l_flags, lockres->l_unlock_action); BUG(); @@ -1066,14 +1114,91 @@ out: spin_unlock_irqrestore(&lockres->l_lock, flags); } +static void ocfs2_unlock_ast(struct ocfs2_dlm_lksb *lksb, int error) +{ + struct ocfs2_lock_res *lockres = ocfs2_lksb_to_lock_res(lksb); + unsigned long flags; + + mlog(ML_BASTS, "UNLOCK AST fired for lockres %s, action = %d\n", + lockres->l_name, lockres->l_unlock_action); + + spin_lock_irqsave(&lockres->l_lock, flags); + if (error) { + mlog(ML_ERROR, "Dlm passes error %d for lock %s, " + "unlock_action %d\n", error, lockres->l_name, + lockres->l_unlock_action); + spin_unlock_irqrestore(&lockres->l_lock, flags); + return; + } + + switch(lockres->l_unlock_action) { + case OCFS2_UNLOCK_CANCEL_CONVERT: + mlog(0, "Cancel convert success for %s\n", lockres->l_name); + lockres->l_action = OCFS2_AST_INVALID; + /* Downconvert thread may have requeued this lock, we + * need to wake it. */ + if (lockres->l_flags & OCFS2_LOCK_BLOCKED) + ocfs2_wake_downconvert_thread(ocfs2_get_lockres_osb(lockres)); + break; + case OCFS2_UNLOCK_DROP_LOCK: + lockres->l_level = DLM_LOCK_IV; + break; + default: + BUG(); + } + + lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); + lockres->l_unlock_action = OCFS2_UNLOCK_INVALID; + wake_up(&lockres->l_event); + spin_unlock_irqrestore(&lockres->l_lock, flags); +} + +/* + * This is the filesystem locking protocol. It provides the lock handling + * hooks for the underlying DLM. It has a maximum version number. + * The version number allows interoperability with systems running at + * the same major number and an equal or smaller minor number. + * + * Whenever the filesystem does new things with locks (adds or removes a + * lock, orders them differently, does different things underneath a lock), + * the version must be changed. The protocol is negotiated when joining + * the dlm domain. A node may join the domain if its major version is + * identical to all other nodes and its minor version is greater than + * or equal to all other nodes. When its minor version is greater than + * the other nodes, it will run at the minor version specified by the + * other nodes. + * + * If a locking change is made that will not be compatible with older + * versions, the major number must be increased and the minor version set + * to zero. If a change merely adds a behavior that can be disabled when + * speaking to older versions, the minor version must be increased. If a + * change adds a fully backwards compatible change (eg, LVB changes that + * are just ignored by older versions), the version does not need to be + * updated. + */ +static struct ocfs2_locking_protocol lproto = { + .lp_max_version = { + .pv_major = OCFS2_LOCKING_PROTOCOL_MAJOR, + .pv_minor = OCFS2_LOCKING_PROTOCOL_MINOR, + }, + .lp_lock_ast = ocfs2_locking_ast, + .lp_blocking_ast = ocfs2_blocking_ast, + .lp_unlock_ast = ocfs2_unlock_ast, +}; + +void ocfs2_set_locking_protocol(void) +{ + ocfs2_stack_glue_set_max_proto_version(&lproto.lp_max_version); +} + static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres, int convert) { unsigned long flags; - mlog_entry_void(); spin_lock_irqsave(&lockres->l_lock, flags); lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); + lockres_clear_flags(lockres, OCFS2_LOCK_UPCONVERT_FINISHING); if (convert) lockres->l_action = OCFS2_AST_INVALID; else @@ -1081,7 +1206,6 @@ static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres, spin_unlock_irqrestore(&lockres->l_lock, flags); wake_up(&lockres->l_event); - mlog_exit_void(); } /* Note: If we detect another process working on the lock (i.e., @@ -1097,8 +1221,6 @@ static int ocfs2_lock_create(struct ocfs2_super *osb, unsigned long flags; unsigned int gen; - mlog_entry_void(); - mlog(0, "lock %s, level = %d, flags = %u\n", lockres->l_name, level, dlm_flags); @@ -1120,8 +1242,7 @@ static int ocfs2_lock_create(struct ocfs2_super *osb, &lockres->l_lksb, dlm_flags, lockres->l_name, - OCFS2_LOCK_ID_MAX_LEN - 1, - lockres); + OCFS2_LOCK_ID_MAX_LEN - 1); lockres_clear_pending(lockres, gen, osb); if (ret) { ocfs2_log_dlm_error("ocfs2_dlm_lock", ret, lockres); @@ -1131,7 +1252,6 @@ static int ocfs2_lock_create(struct ocfs2_super *osb, mlog(0, "lock %s, return from ocfs2_dlm_lock\n", lockres->l_name); bail: - mlog_exit(ret); return ret; } @@ -1184,7 +1304,7 @@ static int ocfs2_wait_for_mask(struct ocfs2_mask_waiter *mw) { wait_for_completion(&mw->mw_complete); /* Re-arm the completion in case we want to wait on it again */ - INIT_COMPLETION(mw->mw_complete); + reinit_completion(&mw->mw_complete); return mw->mw_status; } @@ -1235,15 +1355,17 @@ static int ocfs2_wait_for_mask_interruptible(struct ocfs2_mask_waiter *mw, else ret = mw->mw_status; /* Re-arm the completion in case we want to wait on it again */ - INIT_COMPLETION(mw->mw_complete); + reinit_completion(&mw->mw_complete); return ret; } -static int ocfs2_cluster_lock(struct ocfs2_super *osb, - struct ocfs2_lock_res *lockres, - int level, - u32 lkm_flags, - int arg_flags) +static int __ocfs2_cluster_lock(struct ocfs2_super *osb, + struct ocfs2_lock_res *lockres, + int level, + u32 lkm_flags, + int arg_flags, + int l_subclass, + unsigned long caller_ip) { struct ocfs2_mask_waiter mw; int wait, catch_signals = !(osb->s_mount_opt & OCFS2_MOUNT_NOINTR); @@ -1252,8 +1374,6 @@ static int ocfs2_cluster_lock(struct ocfs2_super *osb, unsigned int gen; int noqueue_attempted = 0; - mlog_entry_void(); - ocfs2_init_mask_waiter(&mw); if (lockres->l_ops->flags & LOCK_TYPE_USES_LVB) @@ -1262,13 +1382,13 @@ static int ocfs2_cluster_lock(struct ocfs2_super *osb, again: wait = 0; + spin_lock_irqsave(&lockres->l_lock, flags); + if (catch_signals && signal_pending(current)) { ret = -ERESTARTSYS; - goto out; + goto unlock; } - spin_lock_irqsave(&lockres->l_lock, flags); - mlog_bug_on_msg(lockres->l_flags & OCFS2_LOCK_FREEING, "Cluster lock called on freeing lockres %s! flags " "0x%lx\n", lockres->l_name, lockres->l_flags); @@ -1285,6 +1405,25 @@ again: goto unlock; } + if (lockres->l_flags & OCFS2_LOCK_UPCONVERT_FINISHING) { + /* + * We've upconverted. If the lock now has a level we can + * work with, we take it. If, however, the lock is not at the + * required level, we go thru the full cycle. One way this could + * happen is if a process requesting an upconvert to PR is + * closely followed by another requesting upconvert to an EX. + * If the process requesting EX lands here, we want it to + * continue attempting to upconvert and let the process + * requesting PR take the lock. + * If multiple processes request upconvert to PR, the first one + * here will take the lock. The others will have to go thru the + * OCFS2_LOCK_BLOCKED check to ensure that there is no pending + * downconvert request. + */ + if (level <= lockres->l_level) + goto update_holders; + } + if (lockres->l_flags & OCFS2_LOCK_BLOCKED && !ocfs2_may_continue_on_blocked_lock(lockres, level)) { /* is the lock is currently blocked on behalf of @@ -1322,7 +1461,7 @@ again: BUG_ON(level == DLM_LOCK_IV); BUG_ON(level == DLM_LOCK_NL); - mlog(0, "lock %s, convert from %d to level = %d\n", + mlog(ML_BASTS, "lockres %s, convert from %d to %d\n", lockres->l_name, lockres->l_level, level); /* call dlm_lock to upgrade lock now */ @@ -1331,8 +1470,7 @@ again: &lockres->l_lksb, lkm_flags, lockres->l_name, - OCFS2_LOCK_ID_MAX_LEN - 1, - lockres); + OCFS2_LOCK_ID_MAX_LEN - 1); lockres_clear_pending(lockres, gen, osb); if (ret) { if (!(lkm_flags & DLM_LKF_NOQUEUE) || @@ -1355,11 +1493,14 @@ again: goto again; } +update_holders: /* Ok, if we get here then we're good to go. */ ocfs2_inc_holders(lockres, level); ret = 0; unlock: + lockres_clear_flags(lockres, OCFS2_LOCK_UPCONVERT_FINISHING); + spin_unlock_irqrestore(&lockres->l_lock, flags); out: /* @@ -1386,22 +1527,47 @@ out: } ocfs2_update_lock_stats(lockres, level, &mw, ret); - mlog_exit(ret); +#ifdef CONFIG_DEBUG_LOCK_ALLOC + if (!ret && lockres->l_lockdep_map.key != NULL) { + if (level == DLM_LOCK_PR) + rwsem_acquire_read(&lockres->l_lockdep_map, l_subclass, + !!(arg_flags & OCFS2_META_LOCK_NOQUEUE), + caller_ip); + else + rwsem_acquire(&lockres->l_lockdep_map, l_subclass, + !!(arg_flags & OCFS2_META_LOCK_NOQUEUE), + caller_ip); + } +#endif return ret; } -static void ocfs2_cluster_unlock(struct ocfs2_super *osb, - struct ocfs2_lock_res *lockres, - int level) +static inline int ocfs2_cluster_lock(struct ocfs2_super *osb, + struct ocfs2_lock_res *lockres, + int level, + u32 lkm_flags, + int arg_flags) +{ + return __ocfs2_cluster_lock(osb, lockres, level, lkm_flags, arg_flags, + 0, _RET_IP_); +} + + +static void __ocfs2_cluster_unlock(struct ocfs2_super *osb, + struct ocfs2_lock_res *lockres, + int level, + unsigned long caller_ip) { unsigned long flags; - mlog_entry_void(); spin_lock_irqsave(&lockres->l_lock, flags); ocfs2_dec_holders(lockres, level); ocfs2_downconvert_on_unlock(osb, lockres); spin_unlock_irqrestore(&lockres->l_lock, flags); - mlog_exit_void(); +#ifdef CONFIG_DEBUG_LOCK_ALLOC + if (lockres->l_lockdep_map.key != NULL) + rwsem_release(&lockres->l_lockdep_map, 1, caller_ip); +#endif } static int ocfs2_create_new_lock(struct ocfs2_super *osb, @@ -1435,8 +1601,6 @@ int ocfs2_create_new_inode_locks(struct inode *inode) BUG_ON(!inode); BUG_ON(!ocfs2_inode_is_new(inode)); - mlog_entry_void(); - mlog(0, "Inode %llu\n", (unsigned long long)OCFS2_I(inode)->ip_blkno); /* NOTE: That we don't increment any of the holder counts, nor @@ -1470,7 +1634,6 @@ int ocfs2_create_new_inode_locks(struct inode *inode) } bail: - mlog_exit(ret); return ret; } @@ -1482,8 +1645,6 @@ int ocfs2_rw_lock(struct inode *inode, int write) BUG_ON(!inode); - mlog_entry_void(); - mlog(0, "inode %llu take %s RW lock\n", (unsigned long long)OCFS2_I(inode)->ip_blkno, write ? "EXMODE" : "PRMODE"); @@ -1500,7 +1661,6 @@ int ocfs2_rw_lock(struct inode *inode, int write) if (status < 0) mlog_errno(status); - mlog_exit(status); return status; } @@ -1510,16 +1670,12 @@ void ocfs2_rw_unlock(struct inode *inode, int write) struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_rw_lockres; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - mlog_entry_void(); - mlog(0, "inode %llu drop %s RW lock\n", (unsigned long long)OCFS2_I(inode)->ip_blkno, write ? "EXMODE" : "PRMODE"); if (!ocfs2_mount_local(osb)) ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level); - - mlog_exit_void(); } /* @@ -1533,12 +1689,10 @@ int ocfs2_open_lock(struct inode *inode) BUG_ON(!inode); - mlog_entry_void(); - mlog(0, "inode %llu take PRMODE open lock\n", (unsigned long long)OCFS2_I(inode)->ip_blkno); - if (ocfs2_mount_local(osb)) + if (ocfs2_is_hard_readonly(osb) || ocfs2_mount_local(osb)) goto out; lockres = &OCFS2_I(inode)->ip_open_lockres; @@ -1549,7 +1703,6 @@ int ocfs2_open_lock(struct inode *inode) mlog_errno(status); out: - mlog_exit(status); return status; } @@ -1561,12 +1714,16 @@ int ocfs2_try_open_lock(struct inode *inode, int write) BUG_ON(!inode); - mlog_entry_void(); - mlog(0, "inode %llu try to take %s open lock\n", (unsigned long long)OCFS2_I(inode)->ip_blkno, write ? "EXMODE" : "PRMODE"); + if (ocfs2_is_hard_readonly(osb)) { + if (write) + status = -EROFS; + goto out; + } + if (ocfs2_mount_local(osb)) goto out; @@ -1584,7 +1741,6 @@ int ocfs2_try_open_lock(struct inode *inode, int write) level, DLM_LKF_NOQUEUE, 0); out: - mlog_exit(status); return status; } @@ -1596,8 +1752,6 @@ void ocfs2_open_unlock(struct inode *inode) struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_open_lockres; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - mlog_entry_void(); - mlog(0, "inode %llu drop open lock\n", (unsigned long long)OCFS2_I(inode)->ip_blkno); @@ -1612,7 +1766,7 @@ void ocfs2_open_unlock(struct inode *inode) DLM_LOCK_EX); out: - mlog_exit_void(); + return; } static int ocfs2_flock_handle_signal(struct ocfs2_lock_res *lockres, @@ -1666,7 +1820,7 @@ out: * ocfs2_file_lock() and ocfs2_file_unlock() map to a single pair of * flock() calls. The locking approach this requires is sufficiently * different from all other cluster lock types that we implement a - * seperate path to the "low-level" dlm calls. In particular: + * separate path to the "low-level" dlm calls. In particular: * * - No optimization of lock levels is done - we take at exactly * what's been requested. @@ -1736,8 +1890,7 @@ int ocfs2_file_lock(struct file *file, int ex, int trylock) spin_unlock_irqrestore(&lockres->l_lock, flags); ret = ocfs2_dlm_lock(osb->cconn, level, &lockres->l_lksb, lkm_flags, - lockres->l_name, OCFS2_LOCK_ID_MAX_LEN - 1, - lockres); + lockres->l_name, OCFS2_LOCK_ID_MAX_LEN - 1); if (ret) { if (!trylock || (ret != -EAGAIN)) { ocfs2_log_dlm_error("ocfs2_dlm_lock", ret, lockres); @@ -1764,7 +1917,7 @@ int ocfs2_file_lock(struct file *file, int ex, int trylock) * outstanding lock request, so a cancel convert is * required. We intentionally overwrite 'ret' - if the * cancel fails and the lock was granted, it's easier - * to just bubble sucess back up to the user. + * to just bubble success back up to the user. */ ret = ocfs2_flock_handle_signal(lockres, level); } else if (!ret && (level > lockres->l_level)) { @@ -1829,8 +1982,6 @@ static void ocfs2_downconvert_on_unlock(struct ocfs2_super *osb, { int kick = 0; - mlog_entry_void(); - /* If we know that another node is waiting on our lock, kick * the downconvert thread * pre-emptively when we reach a release * condition. */ @@ -1851,8 +2002,6 @@ static void ocfs2_downconvert_on_unlock(struct ocfs2_super *osb, if (kick) ocfs2_wake_downconvert_thread(osb); - - mlog_exit_void(); } #define OCFS2_SEC_BITS 34 @@ -1881,8 +2030,6 @@ static void __ocfs2_stuff_meta_lvb(struct inode *inode) struct ocfs2_lock_res *lockres = &oi->ip_inode_lockres; struct ocfs2_meta_lvb *lvb; - mlog_entry_void(); - lvb = ocfs2_dlm_lvb(&lockres->l_lksb); /* @@ -1898,8 +2045,8 @@ static void __ocfs2_stuff_meta_lvb(struct inode *inode) lvb->lvb_version = OCFS2_LVB_VERSION; lvb->lvb_isize = cpu_to_be64(i_size_read(inode)); lvb->lvb_iclusters = cpu_to_be32(oi->ip_clusters); - lvb->lvb_iuid = cpu_to_be32(inode->i_uid); - lvb->lvb_igid = cpu_to_be32(inode->i_gid); + lvb->lvb_iuid = cpu_to_be32(i_uid_read(inode)); + lvb->lvb_igid = cpu_to_be32(i_gid_read(inode)); lvb->lvb_imode = cpu_to_be16(inode->i_mode); lvb->lvb_inlink = cpu_to_be16(inode->i_nlink); lvb->lvb_iatime_packed = @@ -1914,8 +2061,6 @@ static void __ocfs2_stuff_meta_lvb(struct inode *inode) out: mlog_meta_lvb(0, lockres); - - mlog_exit_void(); } static void ocfs2_unpack_timespec(struct timespec *spec, @@ -1931,8 +2076,6 @@ static void ocfs2_refresh_inode_from_lvb(struct inode *inode) struct ocfs2_lock_res *lockres = &oi->ip_inode_lockres; struct ocfs2_meta_lvb *lvb; - mlog_entry_void(); - mlog_meta_lvb(0, lockres); lvb = ocfs2_dlm_lvb(&lockres->l_lksb); @@ -1952,10 +2095,10 @@ static void ocfs2_refresh_inode_from_lvb(struct inode *inode) else inode->i_blocks = ocfs2_inode_sector_count(inode); - inode->i_uid = be32_to_cpu(lvb->lvb_iuid); - inode->i_gid = be32_to_cpu(lvb->lvb_igid); + i_uid_write(inode, be32_to_cpu(lvb->lvb_iuid)); + i_gid_write(inode, be32_to_cpu(lvb->lvb_igid)); inode->i_mode = be16_to_cpu(lvb->lvb_imode); - inode->i_nlink = be16_to_cpu(lvb->lvb_inlink); + set_nlink(inode, be16_to_cpu(lvb->lvb_inlink)); ocfs2_unpack_timespec(&inode->i_atime, be64_to_cpu(lvb->lvb_iatime_packed)); ocfs2_unpack_timespec(&inode->i_mtime, @@ -1963,8 +2106,6 @@ static void ocfs2_refresh_inode_from_lvb(struct inode *inode) ocfs2_unpack_timespec(&inode->i_ctime, be64_to_cpu(lvb->lvb_ictime_packed)); spin_unlock(&oi->ip_lock); - - mlog_exit_void(); } static inline int ocfs2_meta_lvb_is_trustable(struct inode *inode, @@ -1972,7 +2113,8 @@ static inline int ocfs2_meta_lvb_is_trustable(struct inode *inode, { struct ocfs2_meta_lvb *lvb = ocfs2_dlm_lvb(&lockres->l_lksb); - if (lvb->lvb_version == OCFS2_LVB_VERSION + if (ocfs2_dlm_lvb_valid(&lockres->l_lksb) + && lvb->lvb_version == OCFS2_LVB_VERSION && be32_to_cpu(lvb->lvb_igeneration) == inode->i_generation) return 1; return 0; @@ -1990,8 +2132,6 @@ static int ocfs2_should_refresh_lock_res(struct ocfs2_lock_res *lockres) unsigned long flags; int status = 0; - mlog_entry_void(); - refresh_check: spin_lock_irqsave(&lockres->l_lock, flags); if (!(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH)) { @@ -2012,7 +2152,7 @@ refresh_check: status = 1; bail: - mlog_exit(status); + mlog(0, "status %d\n", status); return status; } @@ -2022,7 +2162,6 @@ static inline void ocfs2_complete_lock_res_refresh(struct ocfs2_lock_res *lockre int status) { unsigned long flags; - mlog_entry_void(); spin_lock_irqsave(&lockres->l_lock, flags); lockres_clear_flags(lockres, OCFS2_LOCK_REFRESHING); @@ -2031,8 +2170,6 @@ static inline void ocfs2_complete_lock_res_refresh(struct ocfs2_lock_res *lockre spin_unlock_irqrestore(&lockres->l_lock, flags); wake_up(&lockres->l_event); - - mlog_exit_void(); } /* may or may not return a bh if it went to disk. */ @@ -2045,8 +2182,6 @@ static int ocfs2_inode_lock_update(struct inode *inode, struct ocfs2_dinode *fe; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - mlog_entry_void(); - if (ocfs2_mount_local(osb)) goto bail; @@ -2066,7 +2201,7 @@ static int ocfs2_inode_lock_update(struct inode *inode, /* This will discard any caching information we might have had * for the inode metadata. */ - ocfs2_metadata_cache_purge(inode); + ocfs2_metadata_cache_purge(INODE_CACHE(inode)); ocfs2_extent_map_trunc(inode, 0); @@ -2115,7 +2250,6 @@ static int ocfs2_inode_lock_update(struct inode *inode, bail_refresh: ocfs2_complete_lock_res_refresh(lockres, status); bail: - mlog_exit(status); return status; } @@ -2145,10 +2279,11 @@ static int ocfs2_assign_bh(struct inode *inode, * returns < 0 error if the callback will never be called, otherwise * the result of the lock will be communicated via the callback. */ -int ocfs2_inode_lock_full(struct inode *inode, - struct buffer_head **ret_bh, - int ex, - int arg_flags) +int ocfs2_inode_lock_full_nested(struct inode *inode, + struct buffer_head **ret_bh, + int ex, + int arg_flags, + int subclass) { int status, level, acquired; u32 dlm_flags; @@ -2158,8 +2293,6 @@ int ocfs2_inode_lock_full(struct inode *inode, BUG_ON(!inode); - mlog_entry_void(); - mlog(0, "inode %llu, take %s META lock\n", (unsigned long long)OCFS2_I(inode)->ip_blkno, ex ? "EXMODE" : "PRMODE"); @@ -2171,7 +2304,7 @@ int ocfs2_inode_lock_full(struct inode *inode, if (ocfs2_is_hard_readonly(osb)) { if (ex) status = -EROFS; - goto bail; + goto getbh; } if (ocfs2_mount_local(osb)) @@ -2186,9 +2319,10 @@ int ocfs2_inode_lock_full(struct inode *inode, if (arg_flags & OCFS2_META_LOCK_NOQUEUE) dlm_flags |= DLM_LKF_NOQUEUE; - status = ocfs2_cluster_lock(osb, lockres, level, dlm_flags, arg_flags); + status = __ocfs2_cluster_lock(osb, lockres, level, dlm_flags, + arg_flags, subclass, _RET_IP_); if (status < 0) { - if (status != -EAGAIN && status != -EIOCBRETRY) + if (status != -EAGAIN) mlog_errno(status); goto bail; } @@ -2228,7 +2362,7 @@ local: mlog_errno(status); goto bail; } - +getbh: if (ret_bh) { status = ocfs2_assign_bh(inode, ret_bh, local_bh); if (status < 0) { @@ -2250,7 +2384,6 @@ bail: if (local_bh) brelse(local_bh); - mlog_exit(status); return status; } @@ -2300,7 +2433,6 @@ int ocfs2_inode_lock_atime(struct inode *inode, { int ret; - mlog_entry_void(); ret = ocfs2_inode_lock(inode, NULL, 0); if (ret < 0) { mlog_errno(ret); @@ -2328,7 +2460,6 @@ int ocfs2_inode_lock_atime(struct inode *inode, } else *level = 0; - mlog_exit(ret); return ret; } @@ -2339,8 +2470,6 @@ void ocfs2_inode_unlock(struct inode *inode, struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_inode_lockres; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - mlog_entry_void(); - mlog(0, "inode %llu drop %s META lock\n", (unsigned long long)OCFS2_I(inode)->ip_blkno, ex ? "EXMODE" : "PRMODE"); @@ -2348,8 +2477,47 @@ void ocfs2_inode_unlock(struct inode *inode, if (!ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb)) && !ocfs2_mount_local(osb)) ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level); +} - mlog_exit_void(); +int ocfs2_orphan_scan_lock(struct ocfs2_super *osb, u32 *seqno) +{ + struct ocfs2_lock_res *lockres; + struct ocfs2_orphan_scan_lvb *lvb; + int status = 0; + + if (ocfs2_is_hard_readonly(osb)) + return -EROFS; + + if (ocfs2_mount_local(osb)) + return 0; + + lockres = &osb->osb_orphan_scan.os_lockres; + status = ocfs2_cluster_lock(osb, lockres, DLM_LOCK_EX, 0, 0); + if (status < 0) + return status; + + lvb = ocfs2_dlm_lvb(&lockres->l_lksb); + if (ocfs2_dlm_lvb_valid(&lockres->l_lksb) && + lvb->lvb_version == OCFS2_ORPHAN_LVB_VERSION) + *seqno = be32_to_cpu(lvb->lvb_os_seqno); + else + *seqno = osb->osb_orphan_scan.os_seqno + 1; + + return status; +} + +void ocfs2_orphan_scan_unlock(struct ocfs2_super *osb, u32 seqno) +{ + struct ocfs2_lock_res *lockres; + struct ocfs2_orphan_scan_lvb *lvb; + + if (!ocfs2_is_hard_readonly(osb) && !ocfs2_mount_local(osb)) { + lockres = &osb->osb_orphan_scan.os_lockres; + lvb = ocfs2_dlm_lvb(&lockres->l_lksb); + lvb->lvb_version = OCFS2_ORPHAN_LVB_VERSION; + lvb->lvb_os_seqno = cpu_to_be32(seqno); + ocfs2_cluster_unlock(osb, lockres, DLM_LOCK_EX); + } } int ocfs2_super_lock(struct ocfs2_super *osb, @@ -2359,8 +2527,6 @@ int ocfs2_super_lock(struct ocfs2_super *osb, int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR; struct ocfs2_lock_res *lockres = &osb->osb_super_lockres; - mlog_entry_void(); - if (ocfs2_is_hard_readonly(osb)) return -EROFS; @@ -2378,21 +2544,18 @@ int ocfs2_super_lock(struct ocfs2_super *osb, * refreshed, so we do it here. Of course, making sense of * everything is up to the caller :) */ status = ocfs2_should_refresh_lock_res(lockres); - if (status < 0) { - mlog_errno(status); - goto bail; - } if (status) { status = ocfs2_refresh_slot_info(osb); ocfs2_complete_lock_res_refresh(lockres, status); - if (status < 0) + if (status < 0) { + ocfs2_cluster_unlock(osb, lockres, level); mlog_errno(status); + } ocfs2_track_lock_refresh(lockres); } bail: - mlog_exit(status); return status; } @@ -2469,8 +2632,11 @@ int ocfs2_dentry_lock(struct dentry *dentry, int ex) BUG_ON(!dl); - if (ocfs2_is_hard_readonly(osb)) - return -EROFS; + if (ocfs2_is_hard_readonly(osb)) { + if (ex) + return -EROFS; + return 0; + } if (ocfs2_mount_local(osb)) return 0; @@ -2488,7 +2654,7 @@ void ocfs2_dentry_unlock(struct dentry *dentry, int ex) struct ocfs2_dentry_lock *dl = dentry->d_fsdata; struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); - if (!ocfs2_mount_local(osb)) + if (!ocfs2_is_hard_readonly(osb) && !ocfs2_mount_local(osb)) ocfs2_cluster_unlock(osb, &dl->dl_lockres, level); } @@ -2611,8 +2777,15 @@ static void *ocfs2_dlm_seq_next(struct seq_file *m, void *v, loff_t *pos) return iter; } -/* So that debugfs.ocfs2 can determine which format is being used */ -#define OCFS2_DLM_DEBUG_STR_VERSION 2 +/* + * Version is used by debugfs.ocfs2 to determine the format being used + * + * New in version 2 + * - Lock stats printed + * New in version 3 + * - Max time in lock stats is in usecs (instead of nsecs) + */ +#define OCFS2_DLM_DEBUG_STR_VERSION 3 static int ocfs2_dlm_seq_show(struct seq_file *m, void *v) { int i; @@ -2654,18 +2827,18 @@ static int ocfs2_dlm_seq_show(struct seq_file *m, void *v) seq_printf(m, "0x%x\t", lvb[i]); #ifdef CONFIG_OCFS2_FS_STATS -# define lock_num_prmode(_l) (_l)->l_lock_num_prmode -# define lock_num_exmode(_l) (_l)->l_lock_num_exmode -# define lock_num_prmode_failed(_l) (_l)->l_lock_num_prmode_failed -# define lock_num_exmode_failed(_l) (_l)->l_lock_num_exmode_failed -# define lock_total_prmode(_l) (_l)->l_lock_total_prmode -# define lock_total_exmode(_l) (_l)->l_lock_total_exmode -# define lock_max_prmode(_l) (_l)->l_lock_max_prmode -# define lock_max_exmode(_l) (_l)->l_lock_max_exmode -# define lock_refresh(_l) (_l)->l_lock_refresh +# define lock_num_prmode(_l) ((_l)->l_lock_prmode.ls_gets) +# define lock_num_exmode(_l) ((_l)->l_lock_exmode.ls_gets) +# define lock_num_prmode_failed(_l) ((_l)->l_lock_prmode.ls_fail) +# define lock_num_exmode_failed(_l) ((_l)->l_lock_exmode.ls_fail) +# define lock_total_prmode(_l) ((_l)->l_lock_prmode.ls_total) +# define lock_total_exmode(_l) ((_l)->l_lock_exmode.ls_total) +# define lock_max_prmode(_l) ((_l)->l_lock_prmode.ls_max) +# define lock_max_exmode(_l) ((_l)->l_lock_exmode.ls_max) +# define lock_refresh(_l) ((_l)->l_lock_refresh) #else -# define lock_num_prmode(_l) (0ULL) -# define lock_num_exmode(_l) (0ULL) +# define lock_num_prmode(_l) (0) +# define lock_num_exmode(_l) (0) # define lock_num_prmode_failed(_l) (0) # define lock_num_exmode_failed(_l) (0) # define lock_total_prmode(_l) (0ULL) @@ -2675,8 +2848,8 @@ static int ocfs2_dlm_seq_show(struct seq_file *m, void *v) # define lock_refresh(_l) (0) #endif /* The following seq_print was added in version 2 of this output */ - seq_printf(m, "%llu\t" - "%llu\t" + seq_printf(m, "%u\t" + "%u\t" "%u\t" "%u\t" "%llu\t" @@ -2708,7 +2881,7 @@ static const struct seq_operations ocfs2_dlm_seq_ops = { static int ocfs2_dlm_debug_release(struct inode *inode, struct file *file) { - struct seq_file *seq = (struct seq_file *) file->private_data; + struct seq_file *seq = file->private_data; struct ocfs2_dlm_seq_priv *priv = seq->private; struct ocfs2_lock_res *res = &priv->p_iter_res; @@ -2742,7 +2915,7 @@ static int ocfs2_dlm_debug_open(struct inode *inode, struct file *file) goto out; } - seq = (struct seq_file *) file->private_data; + seq = file->private_data; seq->private = priv; ocfs2_add_lockres_tracking(&priv->p_iter_res, @@ -2796,8 +2969,6 @@ int ocfs2_dlm_init(struct ocfs2_super *osb) int status = 0; struct ocfs2_cluster_connection *conn = NULL; - mlog_entry_void(); - if (ocfs2_mount_local(osb)) { osb->node_num = 0; goto local; @@ -2820,16 +2991,18 @@ int ocfs2_dlm_init(struct ocfs2_super *osb) /* for now, uuid == domain */ status = ocfs2_cluster_connect(osb->osb_cluster_stack, + osb->osb_cluster_name, + strlen(osb->osb_cluster_name), osb->uuid_str, strlen(osb->uuid_str), - ocfs2_do_node_down, osb, + &lproto, ocfs2_do_node_down, osb, &conn); if (status) { mlog_errno(status); goto bail; } - status = ocfs2_cluster_this_node(&osb->node_num); + status = ocfs2_cluster_this_node(conn, &osb->node_num); if (status < 0) { mlog_errno(status); mlog(ML_ERROR, @@ -2842,6 +3015,7 @@ local: ocfs2_super_lock_res_init(&osb->osb_super_lockres, osb); ocfs2_rename_lock_res_init(&osb->osb_rename_lockres, osb); ocfs2_nfs_sync_lock_res_init(&osb->osb_nfs_sync_lockres, osb); + ocfs2_orphan_scan_lock_res_init(&osb->osb_orphan_scan.os_lockres, osb); osb->cconn = conn; @@ -2853,15 +3027,12 @@ bail: kthread_stop(osb->dc_task); } - mlog_exit(status); return status; } void ocfs2_dlm_shutdown(struct ocfs2_super *osb, int hangup_pending) { - mlog_entry_void(); - ocfs2_drop_osb_locks(osb); /* @@ -2878,56 +3049,12 @@ void ocfs2_dlm_shutdown(struct ocfs2_super *osb, ocfs2_lock_res_free(&osb->osb_super_lockres); ocfs2_lock_res_free(&osb->osb_rename_lockres); ocfs2_lock_res_free(&osb->osb_nfs_sync_lockres); + ocfs2_lock_res_free(&osb->osb_orphan_scan.os_lockres); ocfs2_cluster_disconnect(osb->cconn, hangup_pending); osb->cconn = NULL; ocfs2_dlm_shutdown_debug(osb); - - mlog_exit_void(); -} - -static void ocfs2_unlock_ast(void *opaque, int error) -{ - struct ocfs2_lock_res *lockres = opaque; - unsigned long flags; - - mlog_entry_void(); - - mlog(0, "UNLOCK AST called on lock %s, action = %d\n", lockres->l_name, - lockres->l_unlock_action); - - spin_lock_irqsave(&lockres->l_lock, flags); - if (error) { - mlog(ML_ERROR, "Dlm passes error %d for lock %s, " - "unlock_action %d\n", error, lockres->l_name, - lockres->l_unlock_action); - spin_unlock_irqrestore(&lockres->l_lock, flags); - return; - } - - switch(lockres->l_unlock_action) { - case OCFS2_UNLOCK_CANCEL_CONVERT: - mlog(0, "Cancel convert success for %s\n", lockres->l_name); - lockres->l_action = OCFS2_AST_INVALID; - /* Downconvert thread may have requeued this lock, we - * need to wake it. */ - if (lockres->l_flags & OCFS2_LOCK_BLOCKED) - ocfs2_wake_downconvert_thread(ocfs2_get_lockres_osb(lockres)); - break; - case OCFS2_UNLOCK_DROP_LOCK: - lockres->l_level = DLM_LOCK_IV; - break; - default: - BUG(); - } - - lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); - lockres->l_unlock_action = OCFS2_UNLOCK_INVALID; - wake_up(&lockres->l_event); - spin_unlock_irqrestore(&lockres->l_lock, flags); - - mlog_exit_void(); } static int ocfs2_drop_lock(struct ocfs2_super *osb, @@ -2997,8 +3124,7 @@ static int ocfs2_drop_lock(struct ocfs2_super *osb, mlog(0, "lock %s\n", lockres->l_name); - ret = ocfs2_dlm_unlock(osb->cconn, &lockres->l_lksb, lkm_flags, - lockres); + ret = ocfs2_dlm_unlock(osb->cconn, &lockres->l_lksb, lkm_flags); if (ret) { ocfs2_log_dlm_error("ocfs2_dlm_unlock", ret, lockres); mlog(ML_ERROR, "lockres flags: %lu\n", lockres->l_flags); @@ -3010,26 +3136,63 @@ static int ocfs2_drop_lock(struct ocfs2_super *osb, ocfs2_wait_on_busy_lock(lockres); out: - mlog_exit(0); return 0; } +static void ocfs2_process_blocked_lock(struct ocfs2_super *osb, + struct ocfs2_lock_res *lockres); + /* Mark the lockres as being dropped. It will no longer be * queued if blocking, but we still may have to wait on it * being dequeued from the downconvert thread before we can consider - * it safe to drop. + * it safe to drop. * * You can *not* attempt to call cluster_lock on this lockres anymore. */ -void ocfs2_mark_lockres_freeing(struct ocfs2_lock_res *lockres) +void ocfs2_mark_lockres_freeing(struct ocfs2_super *osb, + struct ocfs2_lock_res *lockres) { int status; struct ocfs2_mask_waiter mw; - unsigned long flags; + unsigned long flags, flags2; ocfs2_init_mask_waiter(&mw); spin_lock_irqsave(&lockres->l_lock, flags); lockres->l_flags |= OCFS2_LOCK_FREEING; + if (lockres->l_flags & OCFS2_LOCK_QUEUED && current == osb->dc_task) { + /* + * We know the downconvert is queued but not in progress + * because we are the downconvert thread and processing + * different lock. So we can just remove the lock from the + * queue. This is not only an optimization but also a way + * to avoid the following deadlock: + * ocfs2_dentry_post_unlock() + * ocfs2_dentry_lock_put() + * ocfs2_drop_dentry_lock() + * iput() + * ocfs2_evict_inode() + * ocfs2_clear_inode() + * ocfs2_mark_lockres_freeing() + * ... blocks waiting for OCFS2_LOCK_QUEUED + * since we are the downconvert thread which + * should clear the flag. + */ + spin_unlock_irqrestore(&lockres->l_lock, flags); + spin_lock_irqsave(&osb->dc_task_lock, flags2); + list_del_init(&lockres->l_blocked_list); + osb->blocked_lock_count--; + spin_unlock_irqrestore(&osb->dc_task_lock, flags2); + /* + * Warn if we recurse into another post_unlock call. Strictly + * speaking it isn't a problem but we need to be careful if + * that happens (stack overflow, deadlocks, ...) so warn if + * ocfs2 grows a path for which this can happen. + */ + WARN_ON_ONCE(lockres->l_ops->post_unlock); + /* Since the lock is freeing we don't do much in the fn below */ + ocfs2_process_blocked_lock(osb, lockres); + return; + } while (lockres->l_flags & OCFS2_LOCK_QUEUED) { lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_QUEUED, 0); spin_unlock_irqrestore(&lockres->l_lock, flags); @@ -3050,7 +3213,7 @@ void ocfs2_simple_drop_lockres(struct ocfs2_super *osb, { int ret; - ocfs2_mark_lockres_freeing(lockres); + ocfs2_mark_lockres_freeing(osb, lockres); ret = ocfs2_drop_lock(osb, lockres); if (ret) mlog_errno(ret); @@ -3061,14 +3224,13 @@ static void ocfs2_drop_osb_locks(struct ocfs2_super *osb) ocfs2_simple_drop_lockres(osb, &osb->osb_super_lockres); ocfs2_simple_drop_lockres(osb, &osb->osb_rename_lockres); ocfs2_simple_drop_lockres(osb, &osb->osb_nfs_sync_lockres); + ocfs2_simple_drop_lockres(osb, &osb->osb_orphan_scan.os_lockres); } int ocfs2_drop_inode_locks(struct inode *inode) { int status, err; - mlog_entry_void(); - /* No need to call ocfs2_mark_lockres_freeing here - * ocfs2_clear_inode has done it for us. */ @@ -3093,7 +3255,6 @@ int ocfs2_drop_inode_locks(struct inode *inode) if (err < 0 && !status) status = err; - mlog_exit(status); return status; } @@ -3105,13 +3266,20 @@ static unsigned int ocfs2_prepare_downconvert(struct ocfs2_lock_res *lockres, BUG_ON(lockres->l_blocking <= DLM_LOCK_NL); if (lockres->l_level <= new_level) { - mlog(ML_ERROR, "lockres->l_level (%d) <= new_level (%d)\n", - lockres->l_level, new_level); + mlog(ML_ERROR, "lockres %s, lvl %d <= %d, blcklst %d, mask %d, " + "type %d, flags 0x%lx, hold %d %d, act %d %d, req %d, " + "block %d, pgen %d\n", lockres->l_name, lockres->l_level, + new_level, list_empty(&lockres->l_blocked_list), + list_empty(&lockres->l_mask_waiters), lockres->l_type, + lockres->l_flags, lockres->l_ro_holders, + lockres->l_ex_holders, lockres->l_action, + lockres->l_unlock_action, lockres->l_requested, + lockres->l_blocking, lockres->l_pending_gen); BUG(); } - mlog(0, "lock %s, new_level = %d, l_blocking = %d\n", - lockres->l_name, new_level, lockres->l_blocking); + mlog(ML_BASTS, "lockres %s, level %d => %d, blocking %d\n", + lockres->l_name, lockres->l_level, new_level, lockres->l_blocking); lockres->l_action = OCFS2_AST_DOWNCONVERT; lockres->l_requested = new_level; @@ -3128,7 +3296,8 @@ static int ocfs2_downconvert_lock(struct ocfs2_super *osb, int ret; u32 dlm_flags = DLM_LKF_CONVERT; - mlog_entry_void(); + mlog(ML_BASTS, "lockres %s, level %d => %d\n", lockres->l_name, + lockres->l_level, new_level); if (lvb) dlm_flags |= DLM_LKF_VALBLK; @@ -3138,8 +3307,7 @@ static int ocfs2_downconvert_lock(struct ocfs2_super *osb, &lockres->l_lksb, dlm_flags, lockres->l_name, - OCFS2_LOCK_ID_MAX_LEN - 1, - lockres); + OCFS2_LOCK_ID_MAX_LEN - 1); lockres_clear_pending(lockres, generation, osb); if (ret) { ocfs2_log_dlm_error("ocfs2_dlm_lock", ret, lockres); @@ -3149,7 +3317,6 @@ static int ocfs2_downconvert_lock(struct ocfs2_super *osb, ret = 0; bail: - mlog_exit(ret); return ret; } @@ -3159,15 +3326,11 @@ static int ocfs2_prepare_cancel_convert(struct ocfs2_super *osb, { assert_spin_locked(&lockres->l_lock); - mlog_entry_void(); - mlog(0, "lock %s\n", lockres->l_name); - if (lockres->l_unlock_action == OCFS2_UNLOCK_CANCEL_CONVERT) { /* If we're already trying to cancel a lock conversion * then just drop the spinlock and allow the caller to * requeue this lock. */ - - mlog(0, "Lockres %s, skip convert\n", lockres->l_name); + mlog(ML_BASTS, "lockres %s, skip convert\n", lockres->l_name); return 0; } @@ -3182,6 +3345,8 @@ static int ocfs2_prepare_cancel_convert(struct ocfs2_super *osb, "lock %s, invalid flags: 0x%lx\n", lockres->l_name, lockres->l_flags); + mlog(ML_BASTS, "lockres %s\n", lockres->l_name); + return 1; } @@ -3190,19 +3355,15 @@ static int ocfs2_cancel_convert(struct ocfs2_super *osb, { int ret; - mlog_entry_void(); - mlog(0, "lock %s\n", lockres->l_name); - ret = ocfs2_dlm_unlock(osb->cconn, &lockres->l_lksb, - DLM_LKF_CANCEL, lockres); + DLM_LKF_CANCEL); if (ret) { ocfs2_log_dlm_error("ocfs2_dlm_unlock", ret, lockres); ocfs2_recover_from_dlm_error(lockres, 0); } - mlog(0, "lock %s return from ocfs2_dlm_unlock\n", lockres->l_name); + mlog(ML_BASTS, "lockres %s\n", lockres->l_name); - mlog_exit(ret); return ret; } @@ -3213,17 +3374,24 @@ static int ocfs2_unblock_lock(struct ocfs2_super *osb, unsigned long flags; int blocking; int new_level; + int level; int ret = 0; int set_lvb = 0; unsigned int gen; - mlog_entry_void(); - spin_lock_irqsave(&lockres->l_lock, flags); - BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED)); - recheck: + /* + * Is it still blocking? If not, we have no more work to do. + */ + if (!(lockres->l_flags & OCFS2_LOCK_BLOCKED)) { + BUG_ON(lockres->l_blocking != DLM_LOCK_NL); + spin_unlock_irqrestore(&lockres->l_lock, flags); + ret = 0; + goto leave; + } + if (lockres->l_flags & OCFS2_LOCK_BUSY) { /* XXX * This is a *big* race. The OCFS2_LOCK_PENDING flag @@ -3248,8 +3416,11 @@ recheck: * at the same time they set OCFS2_DLM_BUSY. They must * clear OCFS2_DLM_PENDING after dlm_lock() returns. */ - if (lockres->l_flags & OCFS2_LOCK_PENDING) + if (lockres->l_flags & OCFS2_LOCK_PENDING) { + mlog(ML_BASTS, "lockres %s, ReQ: Pending\n", + lockres->l_name); goto leave_requeue; + } ctl->requeue = 1; ret = ocfs2_prepare_cancel_convert(osb, lockres); @@ -3262,31 +3433,70 @@ recheck: goto leave; } + /* + * This prevents livelocks. OCFS2_LOCK_UPCONVERT_FINISHING flag is + * set when the ast is received for an upconvert just before the + * OCFS2_LOCK_BUSY flag is cleared. Now if the fs received a bast + * on the heels of the ast, we want to delay the downconvert just + * enough to allow the up requestor to do its task. Because this + * lock is in the blocked queue, the lock will be downconverted + * as soon as the requestor is done with the lock. + */ + if (lockres->l_flags & OCFS2_LOCK_UPCONVERT_FINISHING) + goto leave_requeue; + + /* + * How can we block and yet be at NL? We were trying to upconvert + * from NL and got canceled. The code comes back here, and now + * we notice and clear BLOCKING. + */ + if (lockres->l_level == DLM_LOCK_NL) { + BUG_ON(lockres->l_ex_holders || lockres->l_ro_holders); + mlog(ML_BASTS, "lockres %s, Aborting dc\n", lockres->l_name); + lockres->l_blocking = DLM_LOCK_NL; + lockres_clear_flags(lockres, OCFS2_LOCK_BLOCKED); + spin_unlock_irqrestore(&lockres->l_lock, flags); + goto leave; + } + /* if we're blocking an exclusive and we have *any* holders, * then requeue. */ if ((lockres->l_blocking == DLM_LOCK_EX) - && (lockres->l_ex_holders || lockres->l_ro_holders)) + && (lockres->l_ex_holders || lockres->l_ro_holders)) { + mlog(ML_BASTS, "lockres %s, ReQ: EX/PR Holders %u,%u\n", + lockres->l_name, lockres->l_ex_holders, + lockres->l_ro_holders); goto leave_requeue; + } /* If it's a PR we're blocking, then only * requeue if we've got any EX holders */ if (lockres->l_blocking == DLM_LOCK_PR && - lockres->l_ex_holders) + lockres->l_ex_holders) { + mlog(ML_BASTS, "lockres %s, ReQ: EX Holders %u\n", + lockres->l_name, lockres->l_ex_holders); goto leave_requeue; + } /* * Can we get a lock in this state if the holder counts are * zero? The meta data unblock code used to check this. */ if ((lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH) - && (lockres->l_flags & OCFS2_LOCK_REFRESHING)) + && (lockres->l_flags & OCFS2_LOCK_REFRESHING)) { + mlog(ML_BASTS, "lockres %s, ReQ: Lock Refreshing\n", + lockres->l_name); goto leave_requeue; + } new_level = ocfs2_highest_compat_lock_level(lockres->l_blocking); if (lockres->l_ops->check_downconvert - && !lockres->l_ops->check_downconvert(lockres, new_level)) + && !lockres->l_ops->check_downconvert(lockres, new_level)) { + mlog(ML_BASTS, "lockres %s, ReQ: Checkpointing\n", + lockres->l_name); goto leave_requeue; + } /* If we get here, then we know that there are no more * incompatible holders (and anyone asking for an incompatible @@ -3299,17 +3509,24 @@ recheck: * may sleep, so we save off a copy of what we're blocking as * it may change while we're not holding the spin lock. */ blocking = lockres->l_blocking; + level = lockres->l_level; spin_unlock_irqrestore(&lockres->l_lock, flags); ctl->unblock_action = lockres->l_ops->downconvert_worker(lockres, blocking); - if (ctl->unblock_action == UNBLOCK_STOP_POST) + if (ctl->unblock_action == UNBLOCK_STOP_POST) { + mlog(ML_BASTS, "lockres %s, UNBLOCK_STOP_POST\n", + lockres->l_name); goto leave; + } spin_lock_irqsave(&lockres->l_lock, flags); - if (blocking != lockres->l_blocking) { + if ((blocking != lockres->l_blocking) || (level != lockres->l_level)) { /* If this changed underneath us, then we can't drop * it just yet. */ + mlog(ML_BASTS, "lockres %s, block=%d:%d, level=%d:%d, " + "Recheck\n", lockres->l_name, blocking, + lockres->l_blocking, level, lockres->l_level); goto recheck; } @@ -3336,14 +3553,14 @@ downconvert: gen); leave: - mlog_exit(ret); + if (ret) + mlog_errno(ret); return ret; leave_requeue: spin_unlock_irqrestore(&lockres->l_lock, flags); ctl->requeue = 1; - mlog_exit(0); return 0; } @@ -3352,10 +3569,18 @@ static int ocfs2_data_convert_worker(struct ocfs2_lock_res *lockres, { struct inode *inode; struct address_space *mapping; + struct ocfs2_inode_info *oi; inode = ocfs2_lock_res_inode(lockres); mapping = inode->i_mapping; + if (S_ISDIR(inode->i_mode)) { + oi = OCFS2_I(inode); + oi->ip_dir_lock_gen++; + mlog(0, "generation: %u\n", oi->ip_dir_lock_gen); + goto out; + } + if (!S_ISREG(inode->i_mode)) goto out; @@ -3388,11 +3613,11 @@ out: return UNBLOCK_CONTINUE; } -static int ocfs2_check_meta_downconvert(struct ocfs2_lock_res *lockres, - int new_level) +static int ocfs2_ci_checkpointed(struct ocfs2_caching_info *ci, + struct ocfs2_lock_res *lockres, + int new_level) { - struct inode *inode = ocfs2_lock_res_inode(lockres); - int checkpointed = ocfs2_inode_fully_checkpointed(inode); + int checkpointed = ocfs2_ci_fully_checkpointed(ci); BUG_ON(new_level != DLM_LOCK_NL && new_level != DLM_LOCK_PR); BUG_ON(lockres->l_level != DLM_LOCK_EX && !checkpointed); @@ -3400,10 +3625,18 @@ static int ocfs2_check_meta_downconvert(struct ocfs2_lock_res *lockres, if (checkpointed) return 1; - ocfs2_start_checkpoint(OCFS2_SB(inode->i_sb)); + ocfs2_start_checkpoint(OCFS2_SB(ocfs2_metadata_cache_get_super(ci))); return 0; } +static int ocfs2_check_meta_downconvert(struct ocfs2_lock_res *lockres, + int new_level) +{ + struct inode *inode = ocfs2_lock_res_inode(lockres); + + return ocfs2_ci_checkpointed(INODE_CACHE(inode), lockres, new_level); +} + static void ocfs2_set_meta_lvb(struct ocfs2_lock_res *lockres) { struct inode *inode = ocfs2_lock_res_inode(lockres); @@ -3533,6 +3766,26 @@ static int ocfs2_dentry_convert_worker(struct ocfs2_lock_res *lockres, return UNBLOCK_CONTINUE_POST; } +static int ocfs2_check_refcount_downconvert(struct ocfs2_lock_res *lockres, + int new_level) +{ + struct ocfs2_refcount_tree *tree = + ocfs2_lock_res_refcount_tree(lockres); + + return ocfs2_ci_checkpointed(&tree->rf_ci, lockres, new_level); +} + +static int ocfs2_refcount_convert_worker(struct ocfs2_lock_res *lockres, + int blocking) +{ + struct ocfs2_refcount_tree *tree = + ocfs2_lock_res_refcount_tree(lockres); + + ocfs2_metadata_cache_purge(&tree->rf_ci); + + return UNBLOCK_CONTINUE; +} + static void ocfs2_set_qinfo_lvb(struct ocfs2_lock_res *lockres) { struct ocfs2_qinfo_lvb *lvb; @@ -3540,8 +3793,6 @@ static void ocfs2_set_qinfo_lvb(struct ocfs2_lock_res *lockres) struct mem_dqinfo *info = sb_dqinfo(oinfo->dqi_gi.dqi_sb, oinfo->dqi_gi.dqi_type); - mlog_entry_void(); - lvb = ocfs2_dlm_lvb(&lockres->l_lksb); lvb->lvb_version = OCFS2_QINFO_LVB_VERSION; lvb->lvb_bgrace = cpu_to_be32(info->dqi_bgrace); @@ -3550,8 +3801,6 @@ static void ocfs2_set_qinfo_lvb(struct ocfs2_lock_res *lockres) lvb->lvb_blocks = cpu_to_be32(oinfo->dqi_gi.dqi_blocks); lvb->lvb_free_blk = cpu_to_be32(oinfo->dqi_gi.dqi_free_blk); lvb->lvb_free_entry = cpu_to_be32(oinfo->dqi_gi.dqi_free_entry); - - mlog_exit_void(); } void ocfs2_qinfo_unlock(struct ocfs2_mem_dqinfo *oinfo, int ex) @@ -3560,10 +3809,8 @@ void ocfs2_qinfo_unlock(struct ocfs2_mem_dqinfo *oinfo, int ex) struct ocfs2_super *osb = OCFS2_SB(oinfo->dqi_gi.dqi_sb); int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR; - mlog_entry_void(); if (!ocfs2_is_hard_readonly(osb) && !ocfs2_mount_local(osb)) ocfs2_cluster_unlock(osb, lockres, level); - mlog_exit_void(); } static int ocfs2_refresh_qinfo(struct ocfs2_mem_dqinfo *oinfo) @@ -3576,7 +3823,8 @@ static int ocfs2_refresh_qinfo(struct ocfs2_mem_dqinfo *oinfo) struct ocfs2_global_disk_dqinfo *gdinfo; int status = 0; - if (lvb->lvb_version == OCFS2_QINFO_LVB_VERSION) { + if (ocfs2_dlm_lvb_valid(&lockres->l_lksb) && + lvb->lvb_version == OCFS2_QINFO_LVB_VERSION) { info->dqi_bgrace = be32_to_cpu(lvb->lvb_bgrace); info->dqi_igrace = be32_to_cpu(lvb->lvb_igrace); oinfo->dqi_syncms = be32_to_cpu(lvb->lvb_syncms); @@ -3585,7 +3833,8 @@ static int ocfs2_refresh_qinfo(struct ocfs2_mem_dqinfo *oinfo) oinfo->dqi_gi.dqi_free_entry = be32_to_cpu(lvb->lvb_free_entry); } else { - status = ocfs2_read_quota_block(oinfo->dqi_gqinode, 0, &bh); + status = ocfs2_read_quota_phys_block(oinfo->dqi_gqinode, + oinfo->dqi_giblk, &bh); if (status) { mlog_errno(status); goto bail; @@ -3616,8 +3865,6 @@ int ocfs2_qinfo_lock(struct ocfs2_mem_dqinfo *oinfo, int ex) int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR; int status = 0; - mlog_entry_void(); - /* On RO devices, locking really isn't needed... */ if (ocfs2_is_hard_readonly(osb)) { if (ex) @@ -3640,48 +3887,39 @@ int ocfs2_qinfo_lock(struct ocfs2_mem_dqinfo *oinfo, int ex) ocfs2_qinfo_unlock(oinfo, ex); ocfs2_complete_lock_res_refresh(lockres, status); bail: - mlog_exit(status); return status; } -/* - * This is the filesystem locking protocol. It provides the lock handling - * hooks for the underlying DLM. It has a maximum version number. - * The version number allows interoperability with systems running at - * the same major number and an equal or smaller minor number. - * - * Whenever the filesystem does new things with locks (adds or removes a - * lock, orders them differently, does different things underneath a lock), - * the version must be changed. The protocol is negotiated when joining - * the dlm domain. A node may join the domain if its major version is - * identical to all other nodes and its minor version is greater than - * or equal to all other nodes. When its minor version is greater than - * the other nodes, it will run at the minor version specified by the - * other nodes. - * - * If a locking change is made that will not be compatible with older - * versions, the major number must be increased and the minor version set - * to zero. If a change merely adds a behavior that can be disabled when - * speaking to older versions, the minor version must be increased. If a - * change adds a fully backwards compatible change (eg, LVB changes that - * are just ignored by older versions), the version does not need to be - * updated. - */ -static struct ocfs2_locking_protocol lproto = { - .lp_max_version = { - .pv_major = OCFS2_LOCKING_PROTOCOL_MAJOR, - .pv_minor = OCFS2_LOCKING_PROTOCOL_MINOR, - }, - .lp_lock_ast = ocfs2_locking_ast, - .lp_blocking_ast = ocfs2_blocking_ast, - .lp_unlock_ast = ocfs2_unlock_ast, -}; - -void ocfs2_set_locking_protocol(void) +int ocfs2_refcount_lock(struct ocfs2_refcount_tree *ref_tree, int ex) { - ocfs2_stack_glue_set_locking_protocol(&lproto); + int status; + int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR; + struct ocfs2_lock_res *lockres = &ref_tree->rf_lockres; + struct ocfs2_super *osb = lockres->l_priv; + + + if (ocfs2_is_hard_readonly(osb)) + return -EROFS; + + if (ocfs2_mount_local(osb)) + return 0; + + status = ocfs2_cluster_lock(osb, lockres, level, 0, 0); + if (status < 0) + mlog_errno(status); + + return status; } +void ocfs2_refcount_unlock(struct ocfs2_refcount_tree *ref_tree, int ex) +{ + int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR; + struct ocfs2_lock_res *lockres = &ref_tree->rf_lockres; + struct ocfs2_super *osb = lockres->l_priv; + + if (!ocfs2_mount_local(osb)) + ocfs2_cluster_unlock(osb, lockres, level); +} static void ocfs2_process_blocked_lock(struct ocfs2_super *osb, struct ocfs2_lock_res *lockres) @@ -3694,12 +3932,10 @@ static void ocfs2_process_blocked_lock(struct ocfs2_super *osb, * considered valid until we remove the OCFS2_LOCK_QUEUED * flag. */ - mlog_entry_void(); - BUG_ON(!lockres); BUG_ON(!lockres->l_ops); - mlog(0, "lockres %s blocked.\n", lockres->l_name); + mlog(ML_BASTS, "lockres %s blocked\n", lockres->l_name); /* Detect whether a lock has been marked as going away while * the downconvert thread was processing other things. A lock can @@ -3722,21 +3958,19 @@ unqueue: } else ocfs2_schedule_blocked_lock(osb, lockres); - mlog(0, "lockres %s, requeue = %s.\n", lockres->l_name, + mlog(ML_BASTS, "lockres %s, requeue = %s.\n", lockres->l_name, ctl.requeue ? "yes" : "no"); spin_unlock_irqrestore(&lockres->l_lock, flags); if (ctl.unblock_action != UNBLOCK_CONTINUE && lockres->l_ops->post_unlock) lockres->l_ops->post_unlock(osb, lockres); - - mlog_exit_void(); } static void ocfs2_schedule_blocked_lock(struct ocfs2_super *osb, struct ocfs2_lock_res *lockres) { - mlog_entry_void(); + unsigned long flags; assert_spin_locked(&lockres->l_lock); @@ -3744,32 +3978,29 @@ static void ocfs2_schedule_blocked_lock(struct ocfs2_super *osb, /* Do not schedule a lock for downconvert when it's on * the way to destruction - any nodes wanting access * to the resource will get it soon. */ - mlog(0, "Lockres %s won't be scheduled: flags 0x%lx\n", + mlog(ML_BASTS, "lockres %s won't be scheduled: flags 0x%lx\n", lockres->l_name, lockres->l_flags); return; } lockres_or_flags(lockres, OCFS2_LOCK_QUEUED); - spin_lock(&osb->dc_task_lock); + spin_lock_irqsave(&osb->dc_task_lock, flags); if (list_empty(&lockres->l_blocked_list)) { list_add_tail(&lockres->l_blocked_list, &osb->blocked_lock_list); osb->blocked_lock_count++; } - spin_unlock(&osb->dc_task_lock); - - mlog_exit_void(); + spin_unlock_irqrestore(&osb->dc_task_lock, flags); } static void ocfs2_downconvert_thread_do_work(struct ocfs2_super *osb) { unsigned long processed; + unsigned long flags; struct ocfs2_lock_res *lockres; - mlog_entry_void(); - - spin_lock(&osb->dc_task_lock); + spin_lock_irqsave(&osb->dc_task_lock, flags); /* grab this early so we know to try again if a state change and * wake happens part-way through our work */ osb->dc_work_sequence = osb->dc_wake_sequence; @@ -3782,40 +4013,40 @@ static void ocfs2_downconvert_thread_do_work(struct ocfs2_super *osb) struct ocfs2_lock_res, l_blocked_list); list_del_init(&lockres->l_blocked_list); osb->blocked_lock_count--; - spin_unlock(&osb->dc_task_lock); + spin_unlock_irqrestore(&osb->dc_task_lock, flags); BUG_ON(!processed); processed--; ocfs2_process_blocked_lock(osb, lockres); - spin_lock(&osb->dc_task_lock); + spin_lock_irqsave(&osb->dc_task_lock, flags); } - spin_unlock(&osb->dc_task_lock); - - mlog_exit_void(); + spin_unlock_irqrestore(&osb->dc_task_lock, flags); } static int ocfs2_downconvert_thread_lists_empty(struct ocfs2_super *osb) { int empty = 0; + unsigned long flags; - spin_lock(&osb->dc_task_lock); + spin_lock_irqsave(&osb->dc_task_lock, flags); if (list_empty(&osb->blocked_lock_list)) empty = 1; - spin_unlock(&osb->dc_task_lock); + spin_unlock_irqrestore(&osb->dc_task_lock, flags); return empty; } static int ocfs2_downconvert_thread_should_wake(struct ocfs2_super *osb) { int should_wake = 0; + unsigned long flags; - spin_lock(&osb->dc_task_lock); + spin_lock_irqsave(&osb->dc_task_lock, flags); if (osb->dc_work_sequence != osb->dc_wake_sequence) should_wake = 1; - spin_unlock(&osb->dc_task_lock); + spin_unlock_irqrestore(&osb->dc_task_lock, flags); return should_wake; } @@ -3845,10 +4076,12 @@ static int ocfs2_downconvert_thread(void *arg) void ocfs2_wake_downconvert_thread(struct ocfs2_super *osb) { - spin_lock(&osb->dc_task_lock); + unsigned long flags; + + spin_lock_irqsave(&osb->dc_task_lock, flags); /* make sure the voting thread gets a swipe at whatever changes * the caller may have made to the voting state */ osb->dc_wake_sequence++; - spin_unlock(&osb->dc_task_lock); + spin_unlock_irqrestore(&osb->dc_task_lock, flags); wake_up(&osb->dc_event); } diff --git a/fs/ocfs2/dlmglue.h b/fs/ocfs2/dlmglue.h index e1fd5721cd7..d293a22c32c 100644 --- a/fs/ocfs2/dlmglue.h +++ b/fs/ocfs2/dlmglue.h @@ -62,6 +62,14 @@ struct ocfs2_qinfo_lvb { __be32 lvb_free_entry; }; +#define OCFS2_ORPHAN_LVB_VERSION 1 + +struct ocfs2_orphan_scan_lvb { + __u8 lvb_version; + __u8 lvb_reserved[3]; + __be32 lvb_os_seqno; +}; + /* ocfs2_inode_lock_full() 'arg_flags' flags */ /* don't wait on recovery. */ #define OCFS2_META_LOCK_RECOVERY (0x01) @@ -70,6 +78,15 @@ struct ocfs2_qinfo_lvb { /* don't block waiting for the downconvert thread, instead return -EAGAIN */ #define OCFS2_LOCK_NONBLOCK (0x04) +/* Locking subclasses of inode cluster lock */ +enum { + OI_LS_NORMAL = 0, + OI_LS_PARENT, + OI_LS_RENAME1, + OI_LS_RENAME2, + OI_LS_REFLINK_TARGET, +}; + int ocfs2_dlm_init(struct ocfs2_super *osb); void ocfs2_dlm_shutdown(struct ocfs2_super *osb, int hangup_pending); void ocfs2_lock_res_init_once(struct ocfs2_lock_res *res); @@ -85,6 +102,9 @@ void ocfs2_file_lock_res_init(struct ocfs2_lock_res *lockres, struct ocfs2_mem_dqinfo; void ocfs2_qinfo_lock_res_init(struct ocfs2_lock_res *lockres, struct ocfs2_mem_dqinfo *info); +void ocfs2_refcount_lock_res_init(struct ocfs2_lock_res *lockres, + struct ocfs2_super *osb, u64 ref_blkno, + unsigned int generation); void ocfs2_lock_res_free(struct ocfs2_lock_res *res); int ocfs2_create_new_inode_locks(struct inode *inode); int ocfs2_drop_inode_locks(struct inode *inode); @@ -96,23 +116,32 @@ void ocfs2_open_unlock(struct inode *inode); int ocfs2_inode_lock_atime(struct inode *inode, struct vfsmount *vfsmnt, int *level); -int ocfs2_inode_lock_full(struct inode *inode, +int ocfs2_inode_lock_full_nested(struct inode *inode, struct buffer_head **ret_bh, int ex, - int arg_flags); + int arg_flags, + int subclass); int ocfs2_inode_lock_with_page(struct inode *inode, struct buffer_head **ret_bh, int ex, struct page *page); +/* Variants without special locking class or flags */ +#define ocfs2_inode_lock_full(i, r, e, f)\ + ocfs2_inode_lock_full_nested(i, r, e, f, OI_LS_NORMAL) +#define ocfs2_inode_lock_nested(i, b, e, s)\ + ocfs2_inode_lock_full_nested(i, b, e, 0, s) /* 99% of the time we don't want to supply any additional flags -- * those are for very specific cases only. */ -#define ocfs2_inode_lock(i, b, e) ocfs2_inode_lock_full(i, b, e, 0) +#define ocfs2_inode_lock(i, b, e) ocfs2_inode_lock_full_nested(i, b, e, 0, OI_LS_NORMAL) void ocfs2_inode_unlock(struct inode *inode, int ex); int ocfs2_super_lock(struct ocfs2_super *osb, int ex); void ocfs2_super_unlock(struct ocfs2_super *osb, int ex); +int ocfs2_orphan_scan_lock(struct ocfs2_super *osb, u32 *seqno); +void ocfs2_orphan_scan_unlock(struct ocfs2_super *osb, u32 seqno); + int ocfs2_rename_lock(struct ocfs2_super *osb); void ocfs2_rename_unlock(struct ocfs2_super *osb); int ocfs2_nfs_sync_lock(struct ocfs2_super *osb, int ex); @@ -123,9 +152,13 @@ int ocfs2_file_lock(struct file *file, int ex, int trylock); void ocfs2_file_unlock(struct file *file); int ocfs2_qinfo_lock(struct ocfs2_mem_dqinfo *oinfo, int ex); void ocfs2_qinfo_unlock(struct ocfs2_mem_dqinfo *oinfo, int ex); +struct ocfs2_refcount_tree; +int ocfs2_refcount_lock(struct ocfs2_refcount_tree *ref_tree, int ex); +void ocfs2_refcount_unlock(struct ocfs2_refcount_tree *ref_tree, int ex); -void ocfs2_mark_lockres_freeing(struct ocfs2_lock_res *lockres); +void ocfs2_mark_lockres_freeing(struct ocfs2_super *osb, + struct ocfs2_lock_res *lockres); void ocfs2_simple_drop_lockres(struct ocfs2_super *osb, struct ocfs2_lock_res *lockres); diff --git a/fs/ocfs2/export.c b/fs/ocfs2/export.c index 15713cbb865..29651167190 100644 --- a/fs/ocfs2/export.c +++ b/fs/ocfs2/export.c @@ -26,7 +26,6 @@ #include <linux/fs.h> #include <linux/types.h> -#define MLOG_MASK_PREFIX ML_EXPORT #include <cluster/masklog.h> #include "ocfs2.h" @@ -40,6 +39,7 @@ #include "buffer_head_io.h" #include "suballoc.h" +#include "ocfs2_trace.h" struct ocfs2_inode_handle { @@ -56,10 +56,9 @@ static struct dentry *ocfs2_get_dentry(struct super_block *sb, int status, set; struct dentry *result; - mlog_entry("(0x%p, 0x%p)\n", sb, handle); + trace_ocfs2_get_dentry_begin(sb, handle, (unsigned long long)blkno); if (blkno == 0) { - mlog(0, "nfs wants inode with blkno: 0\n"); result = ERR_PTR(-ESTALE); goto bail; } @@ -83,6 +82,7 @@ static struct dentry *ocfs2_get_dentry(struct super_block *sb, } status = ocfs2_test_inode_bit(osb, blkno, &set); + trace_ocfs2_get_dentry_test_bit(status, set); if (status < 0) { if (status == -EINVAL) { /* @@ -90,18 +90,14 @@ static struct dentry *ocfs2_get_dentry(struct super_block *sb, * as an inode, we return -ESTALE to be * nice */ - mlog(0, "test inode bit failed %d\n", status); status = -ESTALE; - } else { + } else mlog(ML_ERROR, "test inode bit failed %d\n", status); - } goto unlock_nfs_sync; } /* If the inode allocator bit is clear, this inode must be stale */ if (!set) { - mlog(0, "inode %llu suballoc bit is clear\n", - (unsigned long long)blkno); status = -ESTALE; goto unlock_nfs_sync; } @@ -114,8 +110,8 @@ unlock_nfs_sync: check_err: if (status < 0) { if (status == -ESTALE) { - mlog(0, "stale inode ino: %llu generation: %u\n", - (unsigned long long)blkno, handle->ih_generation); + trace_ocfs2_get_dentry_stale((unsigned long long)blkno, + handle->ih_generation); } result = ERR_PTR(status); goto bail; @@ -130,20 +126,19 @@ check_err: check_gen: if (handle->ih_generation != inode->i_generation) { iput(inode); - mlog(0, "stale inode ino: %llu generation: %u\n", - (unsigned long long)blkno, handle->ih_generation); + trace_ocfs2_get_dentry_generation((unsigned long long)blkno, + handle->ih_generation, + inode->i_generation); result = ERR_PTR(-ESTALE); goto bail; } result = d_obtain_alias(inode); - if (!IS_ERR(result)) - result->d_op = &ocfs2_dentry_ops; - else + if (IS_ERR(result)) mlog_errno(PTR_ERR(result)); bail: - mlog_exit_ptr(result); + trace_ocfs2_get_dentry_end(result); return result; } @@ -154,11 +149,8 @@ static struct dentry *ocfs2_get_parent(struct dentry *child) struct dentry *parent; struct inode *dir = child->d_inode; - mlog_entry("(0x%p, '%.*s')\n", child, - child->d_name.len, child->d_name.name); - - mlog(0, "find parent of directory %llu\n", - (unsigned long long)OCFS2_I(dir)->ip_blkno); + trace_ocfs2_get_parent(child, child->d_name.len, child->d_name.name, + (unsigned long long)OCFS2_I(dir)->ip_blkno); status = ocfs2_inode_lock(dir, NULL, 0); if (status < 0) { @@ -175,55 +167,53 @@ static struct dentry *ocfs2_get_parent(struct dentry *child) } parent = d_obtain_alias(ocfs2_iget(OCFS2_SB(dir->i_sb), blkno, 0, 0)); - if (!IS_ERR(parent)) - parent->d_op = &ocfs2_dentry_ops; bail_unlock: ocfs2_inode_unlock(dir, 0); bail: - mlog_exit_ptr(parent); + trace_ocfs2_get_parent_end(parent); return parent; } -static int ocfs2_encode_fh(struct dentry *dentry, u32 *fh_in, int *max_len, - int connectable) +static int ocfs2_encode_fh(struct inode *inode, u32 *fh_in, int *max_len, + struct inode *parent) { - struct inode *inode = dentry->d_inode; int len = *max_len; int type = 1; u64 blkno; u32 generation; __le32 *fh = (__force __le32 *) fh_in; - mlog_entry("(0x%p, '%.*s', 0x%p, %d, %d)\n", dentry, - dentry->d_name.len, dentry->d_name.name, - fh, len, connectable); +#ifdef TRACE_HOOKS_ARE_NOT_BRAINDEAD_IN_YOUR_OPINION +#error "You go ahead and fix that mess, then. Somehow" + trace_ocfs2_encode_fh_begin(dentry, dentry->d_name.len, + dentry->d_name.name, + fh, len, connectable); +#endif - if (len < 3 || (connectable && len < 6)) { - mlog(ML_ERROR, "fh buffer is too small for encoding\n"); - type = 255; + if (parent && (len < 6)) { + *max_len = 6; + type = FILEID_INVALID; + goto bail; + } else if (len < 3) { + *max_len = 3; + type = FILEID_INVALID; goto bail; } blkno = OCFS2_I(inode)->ip_blkno; generation = inode->i_generation; - mlog(0, "Encoding fh: blkno: %llu, generation: %u\n", - (unsigned long long)blkno, generation); + trace_ocfs2_encode_fh_self((unsigned long long)blkno, generation); len = 3; fh[0] = cpu_to_le32((u32)(blkno >> 32)); fh[1] = cpu_to_le32((u32)(blkno & 0xffffffff)); fh[2] = cpu_to_le32(generation); - if (connectable && !S_ISDIR(inode->i_mode)) { - struct inode *parent; - - spin_lock(&dentry->d_lock); - - parent = dentry->d_parent->d_inode; + if (parent) { blkno = OCFS2_I(parent)->ip_blkno; generation = parent->i_generation; @@ -231,19 +221,17 @@ static int ocfs2_encode_fh(struct dentry *dentry, u32 *fh_in, int *max_len, fh[4] = cpu_to_le32((u32)(blkno & 0xffffffff)); fh[5] = cpu_to_le32(generation); - spin_unlock(&dentry->d_lock); - len = 6; type = 2; - mlog(0, "Encoding parent: blkno: %llu, generation: %u\n", - (unsigned long long)blkno, generation); + trace_ocfs2_encode_fh_parent((unsigned long long)blkno, + generation); } - + *max_len = len; bail: - mlog_exit(type); + trace_ocfs2_encode_fh_type(type); return type; } diff --git a/fs/ocfs2/extent_map.c b/fs/ocfs2/extent_map.c index f2bb1a04d25..767370b656c 100644 --- a/fs/ocfs2/extent_map.c +++ b/fs/ocfs2/extent_map.c @@ -24,10 +24,10 @@ #include <linux/fs.h> #include <linux/init.h> +#include <linux/slab.h> #include <linux/types.h> #include <linux/fiemap.h> -#define MLOG_MASK_PREFIX ML_EXTENT_MAP #include <cluster/masklog.h> #include "ocfs2.h" @@ -37,6 +37,8 @@ #include "extent_map.h" #include "inode.h" #include "super.h" +#include "symlink.h" +#include "ocfs2_trace.h" #include "buffer_head_io.h" @@ -191,7 +193,7 @@ static int ocfs2_try_to_merge_extent_map(struct ocfs2_extent_map_item *emi, emi->ei_clusters += ins->ei_clusters; return 1; } else if ((ins->ei_phys + ins->ei_clusters) == emi->ei_phys && - (ins->ei_cpos + ins->ei_clusters) == emi->ei_phys && + (ins->ei_cpos + ins->ei_clusters) == emi->ei_cpos && ins->ei_flags == emi->ei_flags) { emi->ei_phys = ins->ei_phys; emi->ei_cpos = ins->ei_cpos; @@ -280,8 +282,7 @@ search: spin_unlock(&oi->ip_lock); out: - if (new_emi) - kfree(new_emi); + kfree(new_emi); } static int ocfs2_last_eb_is_empty(struct inode *inode, @@ -293,7 +294,7 @@ static int ocfs2_last_eb_is_empty(struct inode *inode, struct ocfs2_extent_block *eb; struct ocfs2_extent_list *el; - ret = ocfs2_read_extent_block(inode, last_eb_blk, &eb_bh); + ret = ocfs2_read_extent_block(INODE_CACHE(inode), last_eb_blk, &eb_bh); if (ret) { mlog_errno(ret); goto out; @@ -353,11 +354,11 @@ static int ocfs2_search_for_hole_index(struct ocfs2_extent_list *el, * eb_bh is NULL. Otherwise, eb_bh should point to the extent block * containing el. */ -static int ocfs2_figure_hole_clusters(struct inode *inode, - struct ocfs2_extent_list *el, - struct buffer_head *eb_bh, - u32 v_cluster, - u32 *num_clusters) +int ocfs2_figure_hole_clusters(struct ocfs2_caching_info *ci, + struct ocfs2_extent_list *el, + struct buffer_head *eb_bh, + u32 v_cluster, + u32 *num_clusters) { int ret, i; struct buffer_head *next_eb_bh = NULL; @@ -375,7 +376,7 @@ static int ocfs2_figure_hole_clusters(struct inode *inode, if (le64_to_cpu(eb->h_next_leaf_blk) == 0ULL) goto no_more_extents; - ret = ocfs2_read_extent_block(inode, + ret = ocfs2_read_extent_block(ci, le64_to_cpu(eb->h_next_leaf_blk), &next_eb_bh); if (ret) { @@ -428,7 +429,8 @@ static int ocfs2_get_clusters_nocache(struct inode *inode, tree_height = le16_to_cpu(el->l_tree_depth); if (tree_height > 0) { - ret = ocfs2_find_leaf(inode, el, v_cluster, &eb_bh); + ret = ocfs2_find_leaf(INODE_CACHE(inode), el, v_cluster, + &eb_bh); if (ret) { mlog_errno(ret); goto out; @@ -451,11 +453,12 @@ static int ocfs2_get_clusters_nocache(struct inode *inode, if (i == -1) { /* * Holes can be larger than the maximum size of an - * extent, so we return their lengths in a seperate + * extent, so we return their lengths in a separate * field. */ if (hole_len) { - ret = ocfs2_figure_hole_clusters(inode, el, eb_bh, + ret = ocfs2_figure_hole_clusters(INODE_CACHE(inode), + el, eb_bh, v_cluster, &len); if (ret) { mlog_errno(ret); @@ -539,7 +542,8 @@ static void ocfs2_relative_extent_offsets(struct super_block *sb, int ocfs2_xattr_get_clusters(struct inode *inode, u32 v_cluster, u32 *p_cluster, u32 *num_clusters, - struct ocfs2_extent_list *el) + struct ocfs2_extent_list *el, + unsigned int *extent_flags) { int ret = 0, i; struct buffer_head *eb_bh = NULL; @@ -548,7 +552,8 @@ int ocfs2_xattr_get_clusters(struct inode *inode, u32 v_cluster, u32 coff; if (el->l_tree_depth) { - ret = ocfs2_find_leaf(inode, el, v_cluster, &eb_bh); + ret = ocfs2_find_leaf(INODE_CACHE(inode), el, v_cluster, + &eb_bh); if (ret) { mlog_errno(ret); goto out; @@ -590,6 +595,9 @@ int ocfs2_xattr_get_clusters(struct inode *inode, u32 v_cluster, *p_cluster = *p_cluster + coff; if (num_clusters) *num_clusters = ocfs2_rec_clusters(el, rec) - coff; + + if (extent_flags) + *extent_flags = rec->e_flags; } out: if (eb_bh) @@ -696,6 +704,12 @@ out: return ret; } +/* + * The ocfs2_fiemap_inline() may be a little bit misleading, since + * it not only handles the fiemap for inlined files, but also deals + * with the fast symlink, cause they have no difference for extent + * mapping per se. + */ static int ocfs2_fiemap_inline(struct inode *inode, struct buffer_head *di_bh, struct fiemap_extent_info *fieinfo, u64 map_start) @@ -708,11 +722,18 @@ static int ocfs2_fiemap_inline(struct inode *inode, struct buffer_head *di_bh, struct ocfs2_inode_info *oi = OCFS2_I(inode); di = (struct ocfs2_dinode *)di_bh->b_data; - id_count = le16_to_cpu(di->id2.i_data.id_count); + if (ocfs2_inode_is_fast_symlink(inode)) + id_count = ocfs2_fast_symlink_chars(inode->i_sb); + else + id_count = le16_to_cpu(di->id2.i_data.id_count); if (map_start < id_count) { phys = oi->ip_blkno << inode->i_sb->s_blocksize_bits; - phys += offsetof(struct ocfs2_dinode, id2.i_data.id_data); + if (ocfs2_inode_is_fast_symlink(inode)) + phys += offsetof(struct ocfs2_dinode, id2.i_symlink); + else + phys += offsetof(struct ocfs2_dinode, + id2.i_data.id_data); ret = fiemap_fill_next_extent(fieinfo, 0, phys, id_count, flags); @@ -749,9 +770,10 @@ int ocfs2_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, down_read(&OCFS2_I(inode)->ip_alloc_sem); /* - * Handle inline-data separately. + * Handle inline-data and fast symlink separately. */ - if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) { + if ((OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) || + ocfs2_inode_is_fast_symlink(inode)) { ret = ocfs2_fiemap_inline(inode, di_bh, fieinfo, map_start); goto out_unlock; } @@ -759,7 +781,6 @@ int ocfs2_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, cpos = map_start >> osb->s_clustersize_bits; mapping_end = ocfs2_clusters_for_bytes(inode->i_sb, map_start + map_len); - mapping_end -= cpos; is_last = 0; while (cpos < mapping_end && !is_last) { u32 fe_flags; @@ -768,7 +789,7 @@ int ocfs2_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, &hole_size, &rec, &is_last); if (ret) { mlog_errno(ret); - goto out; + goto out_unlock; } if (rec.e_blkno == 0ULL) { @@ -779,6 +800,8 @@ int ocfs2_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, fe_flags = 0; if (rec.e_flags & OCFS2_EXT_UNWRITTEN) fe_flags |= FIEMAP_EXTENT_UNWRITTEN; + if (rec.e_flags & OCFS2_EXT_REFCOUNTED) + fe_flags |= FIEMAP_EXTENT_SHARED; if (is_last) fe_flags |= FIEMAP_EXTENT_LAST; len_bytes = (u64)le16_to_cpu(rec.e_leaf_clusters) << osb->s_clustersize_bits; @@ -807,6 +830,100 @@ out: return ret; } +int ocfs2_seek_data_hole_offset(struct file *file, loff_t *offset, int whence) +{ + struct inode *inode = file->f_mapping->host; + int ret; + unsigned int is_last = 0, is_data = 0; + u16 cs_bits = OCFS2_SB(inode->i_sb)->s_clustersize_bits; + u32 cpos, cend, clen, hole_size; + u64 extoff, extlen; + struct buffer_head *di_bh = NULL; + struct ocfs2_extent_rec rec; + + BUG_ON(whence != SEEK_DATA && whence != SEEK_HOLE); + + ret = ocfs2_inode_lock(inode, &di_bh, 0); + if (ret) { + mlog_errno(ret); + goto out; + } + + down_read(&OCFS2_I(inode)->ip_alloc_sem); + + if (*offset >= i_size_read(inode)) { + ret = -ENXIO; + goto out_unlock; + } + + if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) { + if (whence == SEEK_HOLE) + *offset = i_size_read(inode); + goto out_unlock; + } + + clen = 0; + cpos = *offset >> cs_bits; + cend = ocfs2_clusters_for_bytes(inode->i_sb, i_size_read(inode)); + + while (cpos < cend && !is_last) { + ret = ocfs2_get_clusters_nocache(inode, di_bh, cpos, &hole_size, + &rec, &is_last); + if (ret) { + mlog_errno(ret); + goto out_unlock; + } + + extoff = cpos; + extoff <<= cs_bits; + + if (rec.e_blkno == 0ULL) { + clen = hole_size; + is_data = 0; + } else { + clen = le16_to_cpu(rec.e_leaf_clusters) - + (cpos - le32_to_cpu(rec.e_cpos)); + is_data = (rec.e_flags & OCFS2_EXT_UNWRITTEN) ? 0 : 1; + } + + if ((!is_data && whence == SEEK_HOLE) || + (is_data && whence == SEEK_DATA)) { + if (extoff > *offset) + *offset = extoff; + goto out_unlock; + } + + if (!is_last) + cpos += clen; + } + + if (whence == SEEK_HOLE) { + extoff = cpos; + extoff <<= cs_bits; + extlen = clen; + extlen <<= cs_bits; + + if ((extoff + extlen) > i_size_read(inode)) + extlen = i_size_read(inode) - extoff; + extoff += extlen; + if (extoff > *offset) + *offset = extoff; + goto out_unlock; + } + + ret = -ENXIO; + +out_unlock: + + brelse(di_bh); + + up_read(&OCFS2_I(inode)->ip_alloc_sem); + + ocfs2_inode_unlock(inode, 0); +out: + return ret; +} + int ocfs2_read_virt_blocks(struct inode *inode, u64 v_block, int nr, struct buffer_head *bhs[], int flags, int (*validate)(struct super_block *sb, @@ -816,10 +933,9 @@ int ocfs2_read_virt_blocks(struct inode *inode, u64 v_block, int nr, u64 p_block, p_count; int i, count, done = 0; - mlog_entry("(inode = %p, v_block = %llu, nr = %d, bhs = %p, " - "flags = %x, validate = %p)\n", - inode, (unsigned long long)v_block, nr, bhs, flags, - validate); + trace_ocfs2_read_virt_blocks( + inode, (unsigned long long)v_block, nr, bhs, flags, + validate); if (((v_block + nr - 1) << inode->i_sb->s_blocksize_bits) >= i_size_read(inode)) { @@ -862,8 +978,8 @@ int ocfs2_read_virt_blocks(struct inode *inode, u64 v_block, int nr, BUG_ON(bhs[done + i]->b_blocknr != (p_block + i)); } - rc = ocfs2_read_blocks(inode, p_block, count, bhs + done, - flags, validate); + rc = ocfs2_read_blocks(INODE_CACHE(inode), p_block, count, + bhs + done, flags, validate); if (rc) { mlog_errno(rc); break; @@ -872,7 +988,6 @@ int ocfs2_read_virt_blocks(struct inode *inode, u64 v_block, int nr, } out: - mlog_exit(rc); return rc; } diff --git a/fs/ocfs2/extent_map.h b/fs/ocfs2/extent_map.h index b7dd9731b46..67ea57d2fd5 100644 --- a/fs/ocfs2/extent_map.h +++ b/fs/ocfs2/extent_map.h @@ -53,14 +53,22 @@ int ocfs2_extent_map_get_blocks(struct inode *inode, u64 v_blkno, u64 *p_blkno, int ocfs2_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, u64 map_start, u64 map_len); +int ocfs2_seek_data_hole_offset(struct file *file, loff_t *offset, int origin); + int ocfs2_xattr_get_clusters(struct inode *inode, u32 v_cluster, u32 *p_cluster, u32 *num_clusters, - struct ocfs2_extent_list *el); + struct ocfs2_extent_list *el, + unsigned int *extent_flags); int ocfs2_read_virt_blocks(struct inode *inode, u64 v_block, int nr, struct buffer_head *bhs[], int flags, int (*validate)(struct super_block *sb, struct buffer_head *bh)); +int ocfs2_figure_hole_clusters(struct ocfs2_caching_info *ci, + struct ocfs2_extent_list *el, + struct buffer_head *eb_bh, + u32 v_cluster, + u32 *num_clusters); static inline int ocfs2_read_virt_block(struct inode *inode, u64 v_block, struct buffer_head **bh, int (*validate)(struct super_block *sb, diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c index c2a87c885b7..2930e231f3f 100644 --- a/fs/ocfs2/file.c +++ b/fs/ocfs2/file.c @@ -36,8 +36,8 @@ #include <linux/writeback.h> #include <linux/falloc.h> #include <linux/quotaops.h> +#include <linux/blkdev.h> -#define MLOG_MASK_PREFIX ML_INODE #include <cluster/masklog.h> #include "ocfs2.h" @@ -59,15 +59,11 @@ #include "xattr.h" #include "acl.h" #include "quota.h" +#include "refcounttree.h" +#include "ocfs2_trace.h" #include "buffer_head_io.h" -static int ocfs2_sync_inode(struct inode *inode) -{ - filemap_fdatawrite(inode->i_mapping); - return sync_mapping_buffers(inode->i_mapping); -} - static int ocfs2_init_file_private(struct inode *inode, struct file *file) { struct ocfs2_file_private *fp; @@ -103,8 +99,13 @@ static int ocfs2_file_open(struct inode *inode, struct file *file) int mode = file->f_flags; struct ocfs2_inode_info *oi = OCFS2_I(inode); - mlog_entry("(0x%p, 0x%p, '%.*s')\n", inode, file, - file->f_path.dentry->d_name.len, file->f_path.dentry->d_name.name); + trace_ocfs2_file_open(inode, file, file->f_path.dentry, + (unsigned long long)OCFS2_I(inode)->ip_blkno, + file->f_path.dentry->d_name.len, + file->f_path.dentry->d_name.name, mode); + + if (file->f_mode & FMODE_WRITE) + dquot_initialize(inode); spin_lock(&oi->ip_lock); @@ -136,7 +137,6 @@ static int ocfs2_file_open(struct inode *inode, struct file *file) } leave: - mlog_exit(status); return status; } @@ -144,19 +144,19 @@ static int ocfs2_file_release(struct inode *inode, struct file *file) { struct ocfs2_inode_info *oi = OCFS2_I(inode); - mlog_entry("(0x%p, 0x%p, '%.*s')\n", inode, file, - file->f_path.dentry->d_name.len, - file->f_path.dentry->d_name.name); - spin_lock(&oi->ip_lock); if (!--oi->ip_open_count) oi->ip_flags &= ~OCFS2_INODE_OPEN_DIRECT; + + trace_ocfs2_file_release(inode, file, file->f_path.dentry, + oi->ip_blkno, + file->f_path.dentry->d_name.len, + file->f_path.dentry->d_name.name, + oi->ip_open_count); spin_unlock(&oi->ip_lock); ocfs2_free_file_private(inode, file); - mlog_exit(0); - return 0; } @@ -171,27 +171,44 @@ static int ocfs2_dir_release(struct inode *inode, struct file *file) return 0; } -static int ocfs2_sync_file(struct file *file, - struct dentry *dentry, +static int ocfs2_sync_file(struct file *file, loff_t start, loff_t end, int datasync) { int err = 0; - journal_t *journal; - struct inode *inode = dentry->d_inode; + struct inode *inode = file->f_mapping->host; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_inode_info *oi = OCFS2_I(inode); + journal_t *journal = osb->journal->j_journal; + int ret; + tid_t commit_tid; + bool needs_barrier = false; - mlog_entry("(0x%p, 0x%p, %d, '%.*s')\n", file, dentry, datasync, - dentry->d_name.len, dentry->d_name.name); + trace_ocfs2_sync_file(inode, file, file->f_path.dentry, + OCFS2_I(inode)->ip_blkno, + file->f_path.dentry->d_name.len, + file->f_path.dentry->d_name.name, + (unsigned long long)datasync); + + if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb)) + return -EROFS; - err = ocfs2_sync_inode(dentry->d_inode); + err = filemap_write_and_wait_range(inode->i_mapping, start, end); if (err) - goto bail; + return err; - journal = osb->journal->j_journal; - err = jbd2_journal_force_commit(journal); + commit_tid = datasync ? oi->i_datasync_tid : oi->i_sync_tid; + if (journal->j_flags & JBD2_BARRIER && + !jbd2_trans_will_send_data_barrier(journal, commit_tid)) + needs_barrier = true; + err = jbd2_complete_transaction(journal, commit_tid); + if (needs_barrier) { + ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL); + if (!err) + err = ret; + } -bail: - mlog_exit(err); + if (err) + mlog_errno(err); return (err < 0) ? -EIO : 0; } @@ -247,8 +264,6 @@ int ocfs2_update_inode_atime(struct inode *inode, handle_t *handle; struct ocfs2_dinode *di = (struct ocfs2_dinode *) bh->b_data; - mlog_entry_void(); - handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS); if (IS_ERR(handle)) { ret = PTR_ERR(handle); @@ -256,7 +271,7 @@ int ocfs2_update_inode_atime(struct inode *inode, goto out; } - ret = ocfs2_journal_access_di(handle, inode, bh, + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); @@ -271,15 +286,12 @@ int ocfs2_update_inode_atime(struct inode *inode, inode->i_atime = CURRENT_TIME; di->i_atime = cpu_to_le64(inode->i_atime.tv_sec); di->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec); - - ret = ocfs2_journal_dirty(handle, bh); - if (ret < 0) - mlog_errno(ret); + ocfs2_update_inode_fsync_trans(handle, inode, 0); + ocfs2_journal_dirty(handle, bh); out_commit: ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle); out: - mlog_exit(ret); return ret; } @@ -290,7 +302,6 @@ static int ocfs2_set_inode_size(handle_t *handle, { int status; - mlog_entry_void(); i_size_write(inode, new_i_size); inode->i_blocks = ocfs2_inode_sector_count(inode); inode->i_ctime = inode->i_mtime = CURRENT_TIME; @@ -302,7 +313,6 @@ static int ocfs2_set_inode_size(handle_t *handle, } bail: - mlog_exit(status); return status; } @@ -326,11 +336,45 @@ int ocfs2_simple_size_update(struct inode *inode, if (ret < 0) mlog_errno(ret); + ocfs2_update_inode_fsync_trans(handle, inode, 0); ocfs2_commit_trans(osb, handle); out: return ret; } +static int ocfs2_cow_file_pos(struct inode *inode, + struct buffer_head *fe_bh, + u64 offset) +{ + int status; + u32 phys, cpos = offset >> OCFS2_SB(inode->i_sb)->s_clustersize_bits; + unsigned int num_clusters = 0; + unsigned int ext_flags = 0; + + /* + * If the new offset is aligned to the range of the cluster, there is + * no space for ocfs2_zero_range_for_truncate to fill, so no need to + * CoW either. + */ + if ((offset & (OCFS2_SB(inode->i_sb)->s_clustersize - 1)) == 0) + return 0; + + status = ocfs2_get_clusters(inode, cpos, &phys, + &num_clusters, &ext_flags); + if (status) { + mlog_errno(status); + goto out; + } + + if (!(ext_flags & OCFS2_EXT_REFCOUNTED)) + goto out; + + return ocfs2_refcount_cow(inode, fe_bh, cpos, 1, cpos+1); + +out: + return status; +} + static int ocfs2_orphan_for_truncate(struct ocfs2_super *osb, struct inode *inode, struct buffer_head *fe_bh, @@ -341,7 +385,16 @@ static int ocfs2_orphan_for_truncate(struct ocfs2_super *osb, struct ocfs2_dinode *di; u64 cluster_bytes; - mlog_entry_void(); + /* + * We need to CoW the cluster contains the offset if it is reflinked + * since we will call ocfs2_zero_range_for_truncate later which will + * write "0" from offset to the end of the cluster. + */ + status = ocfs2_cow_file_pos(inode, fe_bh, new_i_size); + if (status) { + mlog_errno(status); + return status; + } /* TODO: This needs to actually orphan the inode in this * transaction. */ @@ -353,7 +406,7 @@ static int ocfs2_orphan_for_truncate(struct ocfs2_super *osb, goto out; } - status = ocfs2_journal_access_di(handle, inode, fe_bh, + status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), fe_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); @@ -378,16 +431,13 @@ static int ocfs2_orphan_for_truncate(struct ocfs2_super *osb, di->i_size = cpu_to_le64(new_i_size); di->i_ctime = di->i_mtime = cpu_to_le64(inode->i_ctime.tv_sec); di->i_ctime_nsec = di->i_mtime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec); + ocfs2_update_inode_fsync_trans(handle, inode, 0); - status = ocfs2_journal_dirty(handle, fe_bh); - if (status < 0) - mlog_errno(status); + ocfs2_journal_dirty(handle, fe_bh); out_commit: ocfs2_commit_trans(osb, handle); out: - - mlog_exit(status); return status; } @@ -398,16 +448,15 @@ static int ocfs2_truncate_file(struct inode *inode, int status = 0; struct ocfs2_dinode *fe = NULL; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - struct ocfs2_truncate_context *tc = NULL; - - mlog_entry("(inode = %llu, new_i_size = %llu\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, - (unsigned long long)new_i_size); /* We trust di_bh because it comes from ocfs2_inode_lock(), which * already validated it */ fe = (struct ocfs2_dinode *) di_bh->b_data; + trace_ocfs2_truncate_file((unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)le64_to_cpu(fe->i_size), + (unsigned long long)new_i_size); + mlog_bug_on_msg(le64_to_cpu(fe->i_size) != i_size_read(inode), "Inode %llu, inode i_size = %lld != di " "i_size = %llu, i_flags = 0x%x\n", @@ -417,26 +466,19 @@ static int ocfs2_truncate_file(struct inode *inode, le32_to_cpu(fe->i_flags)); if (new_i_size > le64_to_cpu(fe->i_size)) { - mlog(0, "asked to truncate file with size (%llu) to size (%llu)!\n", - (unsigned long long)le64_to_cpu(fe->i_size), - (unsigned long long)new_i_size); + trace_ocfs2_truncate_file_error( + (unsigned long long)le64_to_cpu(fe->i_size), + (unsigned long long)new_i_size); status = -EINVAL; mlog_errno(status); goto bail; } - mlog(0, "inode %llu, i_size = %llu, new_i_size = %llu\n", - (unsigned long long)le64_to_cpu(fe->i_blkno), - (unsigned long long)le64_to_cpu(fe->i_size), - (unsigned long long)new_i_size); - - /* lets handle the simple truncate cases before doing any more - * cluster locking. */ - if (new_i_size == le64_to_cpu(fe->i_size)) - goto bail; - down_write(&OCFS2_I(inode)->ip_alloc_sem); + ocfs2_resv_discard(&osb->osb_la_resmap, + &OCFS2_I(inode)->ip_la_data_resv); + /* * The inode lock forced other nodes to sync and drop their * pages, which (correctly) happens even if we have a truncate @@ -466,13 +508,7 @@ static int ocfs2_truncate_file(struct inode *inode, goto bail_unlock_sem; } - status = ocfs2_prepare_truncate(osb, inode, di_bh, &tc); - if (status < 0) { - mlog_errno(status); - goto bail_unlock_sem; - } - - status = ocfs2_commit_truncate(osb, inode, di_bh, tc); + status = ocfs2_commit_truncate(osb, inode, di_bh); if (status < 0) { mlog_errno(status); goto bail_unlock_sem; @@ -483,8 +519,9 @@ bail_unlock_sem: up_write(&OCFS2_I(inode)->ip_alloc_sem); bail: + if (!status && OCFS2_I(inode)->ip_clusters == 0) + status = ocfs2_try_remove_refcount_tree(inode, di_bh); - mlog_exit(status); return status; } @@ -512,11 +549,10 @@ int ocfs2_add_inode_data(struct ocfs2_super *osb, int ret; struct ocfs2_extent_tree et; - ocfs2_init_dinode_extent_tree(&et, inode, fe_bh); - ret = ocfs2_add_clusters_in_btree(osb, inode, logical_offset, - clusters_to_add, mark_unwritten, - &et, handle, - data_ac, meta_ac, reason_ret); + ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), fe_bh); + ret = ocfs2_add_clusters_in_btree(handle, &et, logical_offset, + clusters_to_add, mark_unwritten, + data_ac, meta_ac, reason_ret); return ret; } @@ -538,10 +574,8 @@ static int __ocfs2_extend_allocation(struct inode *inode, u32 logical_start, struct ocfs2_extent_tree et; int did_quota = 0; - mlog_entry("(clusters_to_add = %u)\n", clusters_to_add); - /* - * This function only exists for file systems which don't + * Unwritten extent only exists for file systems which * support holes. */ BUG_ON(mark_unwritten && !ocfs2_sparse_alloc(osb)); @@ -556,12 +590,7 @@ static int __ocfs2_extend_allocation(struct inode *inode, u32 logical_start, restart_all: BUG_ON(le32_to_cpu(fe->i_clusters) != OCFS2_I(inode)->ip_clusters); - mlog(0, "extend inode %llu, i_size = %lld, di->i_clusters = %u, " - "clusters_to_add = %u\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, - (long long)i_size_read(inode), le32_to_cpu(fe->i_clusters), - clusters_to_add); - ocfs2_init_dinode_extent_tree(&et, inode, bh); + ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), bh); status = ocfs2_lock_allocators(inode, &et, clusters_to_add, 0, &data_ac, &meta_ac); if (status) { @@ -569,8 +598,7 @@ restart_all: goto leave; } - credits = ocfs2_calc_extend_credits(osb->sb, &fe->id2.i_list, - clusters_to_add); + credits = ocfs2_calc_extend_credits(osb->sb, &fe->id2.i_list); handle = ocfs2_start_trans(osb, credits); if (IS_ERR(handle)) { status = PTR_ERR(handle); @@ -580,17 +608,22 @@ restart_all: } restarted_transaction: - if (vfs_dq_alloc_space_nodirty(inode, ocfs2_clusters_to_bytes(osb->sb, - clusters_to_add))) { - status = -EDQUOT; + trace_ocfs2_extend_allocation( + (unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)i_size_read(inode), + le32_to_cpu(fe->i_clusters), clusters_to_add, + why, restart_func); + + status = dquot_alloc_space_nodirty(inode, + ocfs2_clusters_to_bytes(osb->sb, clusters_to_add)); + if (status) goto leave; - } did_quota = 1; /* reserve a write to the file entry early on - that we if we * run out of credits in the allocation path, we can still * update i_size. */ - status = ocfs2_journal_access_di(handle, inode, bh, + status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); @@ -614,34 +647,25 @@ restarted_transaction: mlog_errno(status); goto leave; } - - status = ocfs2_journal_dirty(handle, bh); - if (status < 0) { - mlog_errno(status); - goto leave; - } + ocfs2_update_inode_fsync_trans(handle, inode, 1); + ocfs2_journal_dirty(handle, bh); spin_lock(&OCFS2_I(inode)->ip_lock); clusters_to_add -= (OCFS2_I(inode)->ip_clusters - prev_clusters); spin_unlock(&OCFS2_I(inode)->ip_lock); /* Release unused quota reservation */ - vfs_dq_free_space(inode, + dquot_free_space(inode, ocfs2_clusters_to_bytes(osb->sb, clusters_to_add)); did_quota = 0; if (why != RESTART_NONE && clusters_to_add) { if (why == RESTART_META) { - mlog(0, "restarting function.\n"); restart_func = 1; + status = 0; } else { BUG_ON(why != RESTART_TRANS); - mlog(0, "restarting transaction.\n"); - /* TODO: This can be more intelligent. */ - credits = ocfs2_calc_extend_credits(osb->sb, - &fe->id2.i_list, - clusters_to_add); - status = ocfs2_extend_trans(handle, credits); + status = ocfs2_allocate_extend_trans(handle, 1); if (status < 0) { /* handle still has to be committed at * this point. */ @@ -653,15 +677,15 @@ restarted_transaction: } } - mlog(0, "fe: i_clusters = %u, i_size=%llu\n", + trace_ocfs2_extend_allocation_end(OCFS2_I(inode)->ip_blkno, le32_to_cpu(fe->i_clusters), - (unsigned long long)le64_to_cpu(fe->i_size)); - mlog(0, "inode: ip_clusters=%u, i_size=%lld\n", - OCFS2_I(inode)->ip_clusters, (long long)i_size_read(inode)); + (unsigned long long)le64_to_cpu(fe->i_size), + OCFS2_I(inode)->ip_clusters, + (unsigned long long)i_size_read(inode)); leave: if (status < 0 && did_quota) - vfs_dq_free_space(inode, + dquot_free_space(inode, ocfs2_clusters_to_bytes(osb->sb, clusters_to_add)); if (handle) { ocfs2_commit_trans(osb, handle); @@ -682,65 +706,142 @@ leave: brelse(bh); bh = NULL; - mlog_exit(status); return status; } +/* + * While a write will already be ordering the data, a truncate will not. + * Thus, we need to explicitly order the zeroed pages. + */ +static handle_t *ocfs2_zero_start_ordered_transaction(struct inode *inode, + struct buffer_head *di_bh) +{ + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + handle_t *handle = NULL; + int ret = 0; + + if (!ocfs2_should_order_data(inode)) + goto out; + + handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS); + if (IS_ERR(handle)) { + ret = -ENOMEM; + mlog_errno(ret); + goto out; + } + + ret = ocfs2_jbd2_file_inode(handle, inode); + if (ret < 0) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) + mlog_errno(ret); + ocfs2_update_inode_fsync_trans(handle, inode, 1); + +out: + if (ret) { + if (!IS_ERR(handle)) + ocfs2_commit_trans(osb, handle); + handle = ERR_PTR(ret); + } + return handle; +} + /* Some parts of this taken from generic_cont_expand, which turned out * to be too fragile to do exactly what we need without us having to * worry about recursive locking in ->write_begin() and ->write_end(). */ -static int ocfs2_write_zero_page(struct inode *inode, - u64 size) +static int ocfs2_write_zero_page(struct inode *inode, u64 abs_from, + u64 abs_to, struct buffer_head *di_bh) { struct address_space *mapping = inode->i_mapping; struct page *page; - unsigned long index; - unsigned int offset; + unsigned long index = abs_from >> PAGE_CACHE_SHIFT; handle_t *handle = NULL; - int ret; + int ret = 0; + unsigned zero_from, zero_to, block_start, block_end; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; - offset = (size & (PAGE_CACHE_SIZE-1)); /* Within page */ - /* ugh. in prepare/commit_write, if from==to==start of block, we - ** skip the prepare. make sure we never send an offset for the start - ** of a block - */ - if ((offset & (inode->i_sb->s_blocksize - 1)) == 0) { - offset++; - } - index = size >> PAGE_CACHE_SHIFT; + BUG_ON(abs_from >= abs_to); + BUG_ON(abs_to > (((u64)index + 1) << PAGE_CACHE_SHIFT)); + BUG_ON(abs_from & (inode->i_blkbits - 1)); - page = grab_cache_page(mapping, index); + page = find_or_create_page(mapping, index, GFP_NOFS); if (!page) { ret = -ENOMEM; mlog_errno(ret); goto out; } - ret = ocfs2_prepare_write_nolock(inode, page, offset, offset); - if (ret < 0) { - mlog_errno(ret); - goto out_unlock; - } + /* Get the offsets within the page that we want to zero */ + zero_from = abs_from & (PAGE_CACHE_SIZE - 1); + zero_to = abs_to & (PAGE_CACHE_SIZE - 1); + if (!zero_to) + zero_to = PAGE_CACHE_SIZE; - if (ocfs2_should_order_data(inode)) { - handle = ocfs2_start_walk_page_trans(inode, page, offset, - offset); - if (IS_ERR(handle)) { - ret = PTR_ERR(handle); - handle = NULL; + trace_ocfs2_write_zero_page( + (unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)abs_from, + (unsigned long long)abs_to, + index, zero_from, zero_to); + + /* We know that zero_from is block aligned */ + for (block_start = zero_from; block_start < zero_to; + block_start = block_end) { + block_end = block_start + (1 << inode->i_blkbits); + + /* + * block_start is block-aligned. Bump it by one to force + * __block_write_begin and block_commit_write to zero the + * whole block. + */ + ret = __block_write_begin(page, block_start + 1, 0, + ocfs2_get_block); + if (ret < 0) { + mlog_errno(ret); goto out_unlock; } - } - /* must not update i_size! */ - ret = block_commit_write(page, offset, offset); - if (ret < 0) - mlog_errno(ret); - else - ret = 0; + if (!handle) { + handle = ocfs2_zero_start_ordered_transaction(inode, + di_bh); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + handle = NULL; + break; + } + } + + /* must not update i_size! */ + ret = block_commit_write(page, block_start + 1, + block_start + 1); + if (ret < 0) + mlog_errno(ret); + else + ret = 0; + } - if (handle) + if (handle) { + /* + * fs-writeback will release the dirty pages without page lock + * whose offset are over inode size, the release happens at + * block_write_full_page(). + */ + i_size_write(inode, abs_to); + inode->i_blocks = ocfs2_inode_sector_count(inode); + di->i_size = cpu_to_le64((u64)i_size_read(inode)); + inode->i_mtime = inode->i_ctime = CURRENT_TIME; + di->i_mtime = di->i_ctime = cpu_to_le64(inode->i_mtime.tv_sec); + di->i_ctime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec); + di->i_mtime_nsec = di->i_ctime_nsec; + ocfs2_journal_dirty(handle, di_bh); + ocfs2_update_inode_fsync_trans(handle, inode, 1); ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle); + } + out_unlock: unlock_page(page); page_cache_release(page); @@ -748,22 +849,115 @@ out: return ret; } -static int ocfs2_zero_extend(struct inode *inode, - u64 zero_to_size) +/* + * Find the next range to zero. We do this in terms of bytes because + * that's what ocfs2_zero_extend() wants, and it is dealing with the + * pagecache. We may return multiple extents. + * + * zero_start and zero_end are ocfs2_zero_extend()s current idea of what + * needs to be zeroed. range_start and range_end return the next zeroing + * range. A subsequent call should pass the previous range_end as its + * zero_start. If range_end is 0, there's nothing to do. + * + * Unwritten extents are skipped over. Refcounted extents are CoWd. + */ +static int ocfs2_zero_extend_get_range(struct inode *inode, + struct buffer_head *di_bh, + u64 zero_start, u64 zero_end, + u64 *range_start, u64 *range_end) { - int ret = 0; - u64 start_off; - struct super_block *sb = inode->i_sb; + int rc = 0, needs_cow = 0; + u32 p_cpos, zero_clusters = 0; + u32 zero_cpos = + zero_start >> OCFS2_SB(inode->i_sb)->s_clustersize_bits; + u32 last_cpos = ocfs2_clusters_for_bytes(inode->i_sb, zero_end); + unsigned int num_clusters = 0; + unsigned int ext_flags = 0; + + while (zero_cpos < last_cpos) { + rc = ocfs2_get_clusters(inode, zero_cpos, &p_cpos, + &num_clusters, &ext_flags); + if (rc) { + mlog_errno(rc); + goto out; + } - start_off = ocfs2_align_bytes_to_blocks(sb, i_size_read(inode)); - while (start_off < zero_to_size) { - ret = ocfs2_write_zero_page(inode, start_off); - if (ret < 0) { - mlog_errno(ret); + if (p_cpos && !(ext_flags & OCFS2_EXT_UNWRITTEN)) { + zero_clusters = num_clusters; + if (ext_flags & OCFS2_EXT_REFCOUNTED) + needs_cow = 1; + break; + } + + zero_cpos += num_clusters; + } + if (!zero_clusters) { + *range_end = 0; + goto out; + } + + while ((zero_cpos + zero_clusters) < last_cpos) { + rc = ocfs2_get_clusters(inode, zero_cpos + zero_clusters, + &p_cpos, &num_clusters, + &ext_flags); + if (rc) { + mlog_errno(rc); goto out; } - start_off += sb->s_blocksize; + if (!p_cpos || (ext_flags & OCFS2_EXT_UNWRITTEN)) + break; + if (ext_flags & OCFS2_EXT_REFCOUNTED) + needs_cow = 1; + zero_clusters += num_clusters; + } + if ((zero_cpos + zero_clusters) > last_cpos) + zero_clusters = last_cpos - zero_cpos; + + if (needs_cow) { + rc = ocfs2_refcount_cow(inode, di_bh, zero_cpos, + zero_clusters, UINT_MAX); + if (rc) { + mlog_errno(rc); + goto out; + } + } + + *range_start = ocfs2_clusters_to_bytes(inode->i_sb, zero_cpos); + *range_end = ocfs2_clusters_to_bytes(inode->i_sb, + zero_cpos + zero_clusters); + +out: + return rc; +} + +/* + * Zero one range returned from ocfs2_zero_extend_get_range(). The caller + * has made sure that the entire range needs zeroing. + */ +static int ocfs2_zero_extend_range(struct inode *inode, u64 range_start, + u64 range_end, struct buffer_head *di_bh) +{ + int rc = 0; + u64 next_pos; + u64 zero_pos = range_start; + + trace_ocfs2_zero_extend_range( + (unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)range_start, + (unsigned long long)range_end); + BUG_ON(range_start >= range_end); + + while (zero_pos < range_end) { + next_pos = (zero_pos & PAGE_CACHE_MASK) + PAGE_CACHE_SIZE; + if (next_pos > range_end) + next_pos = range_end; + rc = ocfs2_write_zero_page(inode, zero_pos, next_pos, di_bh); + if (rc < 0) { + mlog_errno(rc); + break; + } + zero_pos = next_pos; /* * Very large extends have the potential to lock up @@ -772,16 +966,63 @@ static int ocfs2_zero_extend(struct inode *inode, cond_resched(); } -out: + return rc; +} + +int ocfs2_zero_extend(struct inode *inode, struct buffer_head *di_bh, + loff_t zero_to_size) +{ + int ret = 0; + u64 zero_start, range_start = 0, range_end = 0; + struct super_block *sb = inode->i_sb; + + zero_start = ocfs2_align_bytes_to_blocks(sb, i_size_read(inode)); + trace_ocfs2_zero_extend((unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)zero_start, + (unsigned long long)i_size_read(inode)); + while (zero_start < zero_to_size) { + ret = ocfs2_zero_extend_get_range(inode, di_bh, zero_start, + zero_to_size, + &range_start, + &range_end); + if (ret) { + mlog_errno(ret); + break; + } + if (!range_end) + break; + /* Trim the ends */ + if (range_start < zero_start) + range_start = zero_start; + if (range_end > zero_to_size) + range_end = zero_to_size; + + ret = ocfs2_zero_extend_range(inode, range_start, + range_end, di_bh); + if (ret) { + mlog_errno(ret); + break; + } + zero_start = range_end; + } + return ret; } -int ocfs2_extend_no_holes(struct inode *inode, u64 new_i_size, u64 zero_to) +int ocfs2_extend_no_holes(struct inode *inode, struct buffer_head *di_bh, + u64 new_i_size, u64 zero_to) { int ret; u32 clusters_to_add; struct ocfs2_inode_info *oi = OCFS2_I(inode); + /* + * Only quota files call this without a bh, and they can't be + * refcounted. + */ + BUG_ON(!di_bh && (oi->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL)); + BUG_ON(!di_bh && !(oi->ip_flags & OCFS2_INODE_SYSTEM_FILE)); + clusters_to_add = ocfs2_clusters_for_bytes(inode->i_sb, new_i_size); if (clusters_to_add < oi->ip_clusters) clusters_to_add = 0; @@ -802,7 +1043,7 @@ int ocfs2_extend_no_holes(struct inode *inode, u64 new_i_size, u64 zero_to) * still need to zero the area between the old i_size and the * new i_size. */ - ret = ocfs2_zero_extend(inode, zero_to); + ret = ocfs2_zero_extend(inode, di_bh, zero_to); if (ret < 0) mlog_errno(ret); @@ -824,27 +1065,15 @@ static int ocfs2_extend_file(struct inode *inode, goto out; if (i_size_read(inode) == new_i_size) - goto out; + goto out; BUG_ON(new_i_size < i_size_read(inode)); /* - * Fall through for converting inline data, even if the fs - * supports sparse files. - * - * The check for inline data here is legal - nobody can add - * the feature since we have i_mutex. We must check it again - * after acquiring ip_alloc_sem though, as paths like mmap - * might have raced us to converting the inode to extents. - */ - if (!(oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) - && ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) - goto out_update_size; - - /* * The alloc sem blocks people in read/write from reading our * allocation until we're done changing it. We depend on * i_mutex to block other extend/truncate calls while we're - * here. + * here. We even have to hold it for sparse files because there + * might be some tail zeroing. */ down_write(&oi->ip_alloc_sem); @@ -861,14 +1090,16 @@ static int ocfs2_extend_file(struct inode *inode, ret = ocfs2_convert_inline_data_to_extents(inode, di_bh); if (ret) { up_write(&oi->ip_alloc_sem); - mlog_errno(ret); goto out; } } - if (!ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) - ret = ocfs2_extend_no_holes(inode, new_i_size, new_i_size); + if (ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) + ret = ocfs2_zero_extend(inode, di_bh, new_i_size); + else + ret = ocfs2_extend_no_holes(inode, di_bh, new_i_size, + new_i_size); up_write(&oi->ip_alloc_sem); @@ -894,39 +1125,31 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr) struct ocfs2_super *osb = OCFS2_SB(sb); struct buffer_head *bh = NULL; handle_t *handle = NULL; - int locked[MAXQUOTAS] = {0, 0}; - int credits, qtype; - struct ocfs2_mem_dqinfo *oinfo; + struct dquot *transfer_to[MAXQUOTAS] = { }; + int qtype; - mlog_entry("(0x%p, '%.*s')\n", dentry, - dentry->d_name.len, dentry->d_name.name); + trace_ocfs2_setattr(inode, dentry, + (unsigned long long)OCFS2_I(inode)->ip_blkno, + dentry->d_name.len, dentry->d_name.name, + attr->ia_valid, attr->ia_mode, + from_kuid(&init_user_ns, attr->ia_uid), + from_kgid(&init_user_ns, attr->ia_gid)); /* ensuring we don't even attempt to truncate a symlink */ if (S_ISLNK(inode->i_mode)) attr->ia_valid &= ~ATTR_SIZE; - if (attr->ia_valid & ATTR_MODE) - mlog(0, "mode change: %d\n", attr->ia_mode); - if (attr->ia_valid & ATTR_UID) - mlog(0, "uid change: %d\n", attr->ia_uid); - if (attr->ia_valid & ATTR_GID) - mlog(0, "gid change: %d\n", attr->ia_gid); - if (attr->ia_valid & ATTR_SIZE) - mlog(0, "size change...\n"); - if (attr->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_CTIME)) - mlog(0, "time change...\n"); - #define OCFS2_VALID_ATTRS (ATTR_ATIME | ATTR_MTIME | ATTR_CTIME | ATTR_SIZE \ | ATTR_GID | ATTR_UID | ATTR_MODE) - if (!(attr->ia_valid & OCFS2_VALID_ATTRS)) { - mlog(0, "can't handle attrs: 0x%x\n", attr->ia_valid); + if (!(attr->ia_valid & OCFS2_VALID_ATTRS)) return 0; - } status = inode_change_ok(inode, attr); if (status) return status; + if (is_quota_modification(inode, attr)) + dquot_initialize(inode); size_change = S_ISREG(inode->i_mode) && attr->ia_valid & ATTR_SIZE; if (size_change) { status = ocfs2_rw_lock(inode, 1); @@ -943,13 +1166,14 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr) goto bail_unlock_rw; } - if (size_change && attr->ia_size != i_size_read(inode)) { - if (attr->ia_size > sb->s_maxbytes) { - status = -EFBIG; + if (size_change) { + status = inode_newsize_ok(inode, attr->ia_size); + if (status) goto bail_unlock; - } - if (i_size_read(inode) > attr->ia_size) { + inode_dio_wait(inode); + + if (i_size_read(inode) >= attr->ia_size) { if (ocfs2_should_order_data(inode)) { status = ocfs2_begin_ordered_truncate(inode, attr->ia_size); @@ -967,38 +1191,39 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr) } } - if ((attr->ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid) || - (attr->ia_valid & ATTR_GID && attr->ia_gid != inode->i_gid)) { - credits = OCFS2_INODE_UPDATE_CREDITS; - if (attr->ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid + if ((attr->ia_valid & ATTR_UID && !uid_eq(attr->ia_uid, inode->i_uid)) || + (attr->ia_valid & ATTR_GID && !gid_eq(attr->ia_gid, inode->i_gid))) { + /* + * Gather pointers to quota structures so that allocation / + * freeing of quota structures happens here and not inside + * dquot_transfer() where we have problems with lock ordering + */ + if (attr->ia_valid & ATTR_UID && !uid_eq(attr->ia_uid, inode->i_uid) && OCFS2_HAS_RO_COMPAT_FEATURE(sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) { - oinfo = sb_dqinfo(sb, USRQUOTA)->dqi_priv; - status = ocfs2_lock_global_qf(oinfo, 1); - if (status < 0) + transfer_to[USRQUOTA] = dqget(sb, make_kqid_uid(attr->ia_uid)); + if (!transfer_to[USRQUOTA]) { + status = -ESRCH; goto bail_unlock; - credits += ocfs2_calc_qinit_credits(sb, USRQUOTA) + - ocfs2_calc_qdel_credits(sb, USRQUOTA); - locked[USRQUOTA] = 1; + } } - if (attr->ia_valid & ATTR_GID && attr->ia_gid != inode->i_gid + if (attr->ia_valid & ATTR_GID && !gid_eq(attr->ia_gid, inode->i_gid) && OCFS2_HAS_RO_COMPAT_FEATURE(sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) { - oinfo = sb_dqinfo(sb, GRPQUOTA)->dqi_priv; - status = ocfs2_lock_global_qf(oinfo, 1); - if (status < 0) + transfer_to[GRPQUOTA] = dqget(sb, make_kqid_gid(attr->ia_gid)); + if (!transfer_to[GRPQUOTA]) { + status = -ESRCH; goto bail_unlock; - credits += ocfs2_calc_qinit_credits(sb, GRPQUOTA) + - ocfs2_calc_qdel_credits(sb, GRPQUOTA); - locked[GRPQUOTA] = 1; + } } - handle = ocfs2_start_trans(osb, credits); + handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS + + 2 * ocfs2_quota_trans_credits(sb)); if (IS_ERR(handle)) { status = PTR_ERR(handle); mlog_errno(status); goto bail_unlock; } - status = vfs_dq_transfer(inode, attr) ? -EDQUOT : 0; + status = __dquot_transfer(inode, transfer_to); if (status < 0) goto bail_commit; } else { @@ -1010,18 +1235,8 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr) } } - /* - * This will intentionally not wind up calling vmtruncate(), - * since all the work for a size change has been done above. - * Otherwise, we could get into problems with truncate as - * ip_alloc_sem is used there to protect against i_size - * changes. - */ - status = inode_setattr(inode, attr); - if (status < 0) { - mlog_errno(status); - goto bail_commit; - } + setattr_copy(inode, attr); + mark_inode_dirty(inode); status = ocfs2_mark_inode_dirty(handle, inode, bh); if (status < 0) @@ -1030,12 +1245,6 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr) bail_commit: ocfs2_commit_trans(osb, handle); bail_unlock: - for (qtype = 0; qtype < MAXQUOTAS; qtype++) { - if (!locked[qtype]) - continue; - oinfo = sb_dqinfo(sb, qtype)->dqi_priv; - ocfs2_unlock_global_qf(oinfo, 1); - } ocfs2_inode_unlock(inode, 1); bail_unlock_rw: if (size_change) @@ -1043,13 +1252,16 @@ bail_unlock_rw: bail: brelse(bh); + /* Release quota pointers in case we acquired them */ + for (qtype = 0; qtype < MAXQUOTAS; qtype++) + dqput(transfer_to[qtype]); + if (!status && attr->ia_valid & ATTR_MODE) { - status = ocfs2_acl_chmod(inode); + status = posix_acl_chmod(inode, inode->i_mode); if (status < 0) mlog_errno(status); } - mlog_exit(status); return status; } @@ -1062,8 +1274,6 @@ int ocfs2_getattr(struct vfsmount *mnt, struct ocfs2_super *osb = sb->s_fs_info; int err; - mlog_entry_void(); - err = ocfs2_inode_revalidate(dentry); if (err) { if (err != -ENOENT) @@ -1077,8 +1287,6 @@ int ocfs2_getattr(struct vfsmount *mnt, stat->blksize = osb->s_clustersize; bail: - mlog_exit(err); - return err; } @@ -1086,7 +1294,8 @@ int ocfs2_permission(struct inode *inode, int mask) { int ret; - mlog_entry_void(); + if (mask & MAY_NOT_BLOCK) + return -ECHILD; ret = ocfs2_inode_lock(inode, NULL, 0); if (ret) { @@ -1095,11 +1304,10 @@ int ocfs2_permission(struct inode *inode, int mask) goto out; } - ret = generic_permission(inode, mask, ocfs2_check_acl); + ret = generic_permission(inode, mask); ocfs2_inode_unlock(inode, 0); out: - mlog_exit(ret); return ret; } @@ -1111,8 +1319,9 @@ static int __ocfs2_write_remove_suid(struct inode *inode, struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); struct ocfs2_dinode *di; - mlog_entry("(Inode %llu, mode 0%o)\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, inode->i_mode); + trace_ocfs2_write_remove_suid( + (unsigned long long)OCFS2_I(inode)->ip_blkno, + inode->i_mode); handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS); if (IS_ERR(handle)) { @@ -1121,7 +1330,7 @@ static int __ocfs2_write_remove_suid(struct inode *inode, goto out; } - ret = ocfs2_journal_access_di(handle, inode, bh, + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret < 0) { mlog_errno(ret); @@ -1134,15 +1343,13 @@ static int __ocfs2_write_remove_suid(struct inode *inode, di = (struct ocfs2_dinode *) bh->b_data; di->i_mode = cpu_to_le16(inode->i_mode); + ocfs2_update_inode_fsync_trans(handle, inode, 0); - ret = ocfs2_journal_dirty(handle, bh); - if (ret < 0) - mlog_errno(ret); + ocfs2_journal_dirty(handle, bh); out_trans: ocfs2_commit_trans(osb, handle); out: - mlog_exit(ret); return ret; } @@ -1321,8 +1528,9 @@ static int ocfs2_zero_partial_clusters(struct inode *inode, * partial clusters here. There's no need to worry about * physical allocation - the zeroing code knows to skip holes. */ - mlog(0, "byte start: %llu, end: %llu\n", - (unsigned long long)start, (unsigned long long)end); + trace_ocfs2_zero_partial_clusters( + (unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)start, (unsigned long long)end); /* * If both edges are on a cluster boundary then there's no @@ -1346,8 +1554,8 @@ static int ocfs2_zero_partial_clusters(struct inode *inode, if (tmpend > end) tmpend = end; - mlog(0, "1st range: start: %llu, tmpend: %llu\n", - (unsigned long long)start, (unsigned long long)tmpend); + trace_ocfs2_zero_partial_clusters_range1((unsigned long long)start, + (unsigned long long)tmpend); ret = ocfs2_zero_range_for_truncate(inode, handle, start, tmpend); if (ret) @@ -1361,33 +1569,125 @@ static int ocfs2_zero_partial_clusters(struct inode *inode, */ start = end & ~(osb->s_clustersize - 1); - mlog(0, "2nd range: start: %llu, end: %llu\n", - (unsigned long long)start, (unsigned long long)end); + trace_ocfs2_zero_partial_clusters_range2( + (unsigned long long)start, (unsigned long long)end); ret = ocfs2_zero_range_for_truncate(inode, handle, start, end); if (ret) mlog_errno(ret); } + ocfs2_update_inode_fsync_trans(handle, inode, 1); ocfs2_commit_trans(osb, handle); out: return ret; } +static int ocfs2_find_rec(struct ocfs2_extent_list *el, u32 pos) +{ + int i; + struct ocfs2_extent_rec *rec = NULL; + + for (i = le16_to_cpu(el->l_next_free_rec) - 1; i >= 0; i--) { + + rec = &el->l_recs[i]; + + if (le32_to_cpu(rec->e_cpos) < pos) + break; + } + + return i; +} + +/* + * Helper to calculate the punching pos and length in one run, we handle the + * following three cases in order: + * + * - remove the entire record + * - remove a partial record + * - no record needs to be removed (hole-punching completed) +*/ +static void ocfs2_calc_trunc_pos(struct inode *inode, + struct ocfs2_extent_list *el, + struct ocfs2_extent_rec *rec, + u32 trunc_start, u32 *trunc_cpos, + u32 *trunc_len, u32 *trunc_end, + u64 *blkno, int *done) +{ + int ret = 0; + u32 coff, range; + + range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec); + + if (le32_to_cpu(rec->e_cpos) >= trunc_start) { + /* + * remove an entire extent record. + */ + *trunc_cpos = le32_to_cpu(rec->e_cpos); + /* + * Skip holes if any. + */ + if (range < *trunc_end) + *trunc_end = range; + *trunc_len = *trunc_end - le32_to_cpu(rec->e_cpos); + *blkno = le64_to_cpu(rec->e_blkno); + *trunc_end = le32_to_cpu(rec->e_cpos); + } else if (range > trunc_start) { + /* + * remove a partial extent record, which means we're + * removing the last extent record. + */ + *trunc_cpos = trunc_start; + /* + * skip hole if any. + */ + if (range < *trunc_end) + *trunc_end = range; + *trunc_len = *trunc_end - trunc_start; + coff = trunc_start - le32_to_cpu(rec->e_cpos); + *blkno = le64_to_cpu(rec->e_blkno) + + ocfs2_clusters_to_blocks(inode->i_sb, coff); + *trunc_end = trunc_start; + } else { + /* + * It may have two following possibilities: + * + * - last record has been removed + * - trunc_start was within a hole + * + * both two cases mean the completion of hole punching. + */ + ret = 1; + } + + *done = ret; +} + static int ocfs2_remove_inode_range(struct inode *inode, struct buffer_head *di_bh, u64 byte_start, u64 byte_len) { - int ret = 0; - u32 trunc_start, trunc_len, cpos, phys_cpos, alloc_size; + int ret = 0, flags = 0, done = 0, i; + u32 trunc_start, trunc_len, trunc_end, trunc_cpos, phys_cpos; + u32 cluster_in_el; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); struct ocfs2_cached_dealloc_ctxt dealloc; struct address_space *mapping = inode->i_mapping; struct ocfs2_extent_tree et; + struct ocfs2_path *path = NULL; + struct ocfs2_extent_list *el = NULL; + struct ocfs2_extent_rec *rec = NULL; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + u64 blkno, refcount_loc = le64_to_cpu(di->i_refcount_loc); - ocfs2_init_dinode_extent_tree(&et, inode, di_bh); + ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), di_bh); ocfs2_init_dealloc_ctxt(&dealloc); + trace_ocfs2_remove_inode_range( + (unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)byte_start, + (unsigned long long)byte_len); + if (byte_len == 0) return 0; @@ -1409,17 +1709,30 @@ static int ocfs2_remove_inode_range(struct inode *inode, goto out; } - trunc_start = ocfs2_clusters_for_bytes(osb->sb, byte_start); - trunc_len = (byte_start + byte_len) >> osb->s_clustersize_bits; - if (trunc_len >= trunc_start) - trunc_len -= trunc_start; - else - trunc_len = 0; + /* + * For reflinks, we may need to CoW 2 clusters which might be + * partially zero'd later, if hole's start and end offset were + * within one cluster(means is not exactly aligned to clustersize). + */ - mlog(0, "Inode: %llu, start: %llu, len: %llu, cstart: %u, clen: %u\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, - (unsigned long long)byte_start, - (unsigned long long)byte_len, trunc_start, trunc_len); + if (OCFS2_I(inode)->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL) { + + ret = ocfs2_cow_file_pos(inode, di_bh, byte_start); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_cow_file_pos(inode, di_bh, byte_start + byte_len); + if (ret) { + mlog_errno(ret); + goto out; + } + } + + trunc_start = ocfs2_clusters_for_bytes(osb->sb, byte_start); + trunc_end = (byte_start + byte_len) >> osb->s_clustersize_bits; + cluster_in_el = trunc_end; ret = ocfs2_zero_partial_clusters(inode, byte_start, byte_len); if (ret) { @@ -1427,36 +1740,85 @@ static int ocfs2_remove_inode_range(struct inode *inode, goto out; } - cpos = trunc_start; - while (trunc_len) { - ret = ocfs2_get_clusters(inode, cpos, &phys_cpos, - &alloc_size, NULL); + path = ocfs2_new_path_from_et(&et); + if (!path) { + ret = -ENOMEM; + mlog_errno(ret); + goto out; + } + + while (trunc_end > trunc_start) { + + ret = ocfs2_find_path(INODE_CACHE(inode), path, + cluster_in_el); if (ret) { mlog_errno(ret); goto out; } - if (alloc_size > trunc_len) - alloc_size = trunc_len; + el = path_leaf_el(path); - /* Only do work for non-holes */ - if (phys_cpos != 0) { - ret = ocfs2_remove_btree_range(inode, &et, cpos, - phys_cpos, alloc_size, - &dealloc); + i = ocfs2_find_rec(el, trunc_end); + /* + * Need to go to previous extent block. + */ + if (i < 0) { + if (path->p_tree_depth == 0) + break; + + ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, + path, + &cluster_in_el); if (ret) { mlog_errno(ret); goto out; } + + /* + * We've reached the leftmost extent block, + * it's safe to leave. + */ + if (cluster_in_el == 0) + break; + + /* + * The 'pos' searched for previous extent block is + * always one cluster less than actual trunc_end. + */ + trunc_end = cluster_in_el + 1; + + ocfs2_reinit_path(path, 1); + + continue; + + } else + rec = &el->l_recs[i]; + + ocfs2_calc_trunc_pos(inode, el, rec, trunc_start, &trunc_cpos, + &trunc_len, &trunc_end, &blkno, &done); + if (done) + break; + + flags = rec->e_flags; + phys_cpos = ocfs2_blocks_to_clusters(inode->i_sb, blkno); + + ret = ocfs2_remove_btree_range(inode, &et, trunc_cpos, + phys_cpos, trunc_len, flags, + &dealloc, refcount_loc); + if (ret < 0) { + mlog_errno(ret); + goto out; } - cpos += alloc_size; - trunc_len -= alloc_size; + cluster_in_el = trunc_end; + + ocfs2_reinit_path(path, 1); } ocfs2_truncate_cluster_pages(inode, byte_start, byte_len); out: + ocfs2_free_path(path); ocfs2_schedule_truncate_log_flush(osb, 1); ocfs2_run_deallocs(osb, &dealloc); @@ -1530,7 +1892,8 @@ static int __ocfs2_change_file_space(struct file *file, struct inode *inode, } size = sr->l_start + sr->l_len; - if (cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64) { + if (cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64 || + cmd == OCFS2_IOC_UNRESVSP || cmd == OCFS2_IOC_UNRESVSP64) { if (sr->l_len <= 0) { ret = -EINVAL; goto out_inode_unlock; @@ -1588,6 +1951,9 @@ static int __ocfs2_change_file_space(struct file *file, struct inode *inode, if (ret < 0) mlog_errno(ret); + if (file && (file->f_flags & O_SYNC)) + handle->h_sync = 1; + ocfs2_commit_trans(osb, handle); out_inode_unlock: @@ -1604,8 +1970,9 @@ out: int ocfs2_change_file_space(struct file *file, unsigned int cmd, struct ocfs2_space_resv *sr) { - struct inode *inode = file->f_path.dentry->d_inode; + struct inode *inode = file_inode(file); struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + int ret; if ((cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64) && !ocfs2_writes_unwritten_extents(osb)) @@ -1620,44 +1987,131 @@ int ocfs2_change_file_space(struct file *file, unsigned int cmd, if (!(file->f_mode & FMODE_WRITE)) return -EBADF; - return __ocfs2_change_file_space(file, inode, file->f_pos, cmd, sr, 0); + ret = mnt_want_write_file(file); + if (ret) + return ret; + ret = __ocfs2_change_file_space(file, inode, file->f_pos, cmd, sr, 0); + mnt_drop_write_file(file); + return ret; } -static long ocfs2_fallocate(struct inode *inode, int mode, loff_t offset, +static long ocfs2_fallocate(struct file *file, int mode, loff_t offset, loff_t len) { + struct inode *inode = file_inode(file); struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); struct ocfs2_space_resv sr; int change_size = 1; + int cmd = OCFS2_IOC_RESVSP64; + if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE)) + return -EOPNOTSUPP; if (!ocfs2_writes_unwritten_extents(osb)) return -EOPNOTSUPP; - if (S_ISDIR(inode->i_mode)) - return -ENODEV; - if (mode & FALLOC_FL_KEEP_SIZE) change_size = 0; + if (mode & FALLOC_FL_PUNCH_HOLE) + cmd = OCFS2_IOC_UNRESVSP64; + sr.l_whence = 0; sr.l_start = (s64)offset; sr.l_len = (s64)len; - return __ocfs2_change_file_space(NULL, inode, offset, - OCFS2_IOC_RESVSP64, &sr, change_size); + return __ocfs2_change_file_space(NULL, inode, offset, cmd, &sr, + change_size); +} + +int ocfs2_check_range_for_refcount(struct inode *inode, loff_t pos, + size_t count) +{ + int ret = 0; + unsigned int extent_flags; + u32 cpos, clusters, extent_len, phys_cpos; + struct super_block *sb = inode->i_sb; + + if (!ocfs2_refcount_tree(OCFS2_SB(inode->i_sb)) || + !(OCFS2_I(inode)->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL) || + OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) + return 0; + + cpos = pos >> OCFS2_SB(sb)->s_clustersize_bits; + clusters = ocfs2_clusters_for_bytes(sb, pos + count) - cpos; + + while (clusters) { + ret = ocfs2_get_clusters(inode, cpos, &phys_cpos, &extent_len, + &extent_flags); + if (ret < 0) { + mlog_errno(ret); + goto out; + } + + if (phys_cpos && (extent_flags & OCFS2_EXT_REFCOUNTED)) { + ret = 1; + break; + } + + if (extent_len > clusters) + extent_len = clusters; + + clusters -= extent_len; + cpos += extent_len; + } +out: + return ret; +} + +static int ocfs2_is_io_unaligned(struct inode *inode, size_t count, loff_t pos) +{ + int blockmask = inode->i_sb->s_blocksize - 1; + loff_t final_size = pos + count; + + if ((pos & blockmask) || (final_size & blockmask)) + return 1; + return 0; } -static int ocfs2_prepare_inode_for_write(struct dentry *dentry, +static int ocfs2_prepare_inode_for_refcount(struct inode *inode, + struct file *file, + loff_t pos, size_t count, + int *meta_level) +{ + int ret; + struct buffer_head *di_bh = NULL; + u32 cpos = pos >> OCFS2_SB(inode->i_sb)->s_clustersize_bits; + u32 clusters = + ocfs2_clusters_for_bytes(inode->i_sb, pos + count) - cpos; + + ret = ocfs2_inode_lock(inode, &di_bh, 1); + if (ret) { + mlog_errno(ret); + goto out; + } + + *meta_level = 1; + + ret = ocfs2_refcount_cow(inode, di_bh, cpos, clusters, UINT_MAX); + if (ret) + mlog_errno(ret); +out: + brelse(di_bh); + return ret; +} + +static int ocfs2_prepare_inode_for_write(struct file *file, loff_t *ppos, size_t count, int appending, - int *direct_io) + int *direct_io, + int *has_refcount) { int ret = 0, meta_level = 0; + struct dentry *dentry = file->f_path.dentry; struct inode *inode = dentry->d_inode; - loff_t saved_pos, end; + loff_t saved_pos = 0, end; - /* + /* * We start with a read level meta lock and only jump to an ex * if we need to make modifications here. */ @@ -1674,7 +2128,7 @@ static int ocfs2_prepare_inode_for_write(struct dentry *dentry, * remove_suid() calls ->setattr without any hint that * we may have already done our cluster locking. Since * ocfs2_setattr() *must* take cluster locks to - * proceeed, this will lead us to recursively lock the + * proceed, this will lead us to recursively lock the * inode. There's also the dinode i_size state which * can be lost via setattr during extending writes (we * set inode->i_size at the end of a write. */ @@ -1694,15 +2148,34 @@ static int ocfs2_prepare_inode_for_write(struct dentry *dentry, /* work on a copy of ppos until we're sure that we won't have * to recalculate it due to relocking. */ - if (appending) { + if (appending) saved_pos = i_size_read(inode); - mlog(0, "O_APPEND: inode->i_size=%llu\n", saved_pos); - } else { + else saved_pos = *ppos; - } end = saved_pos + count; + ret = ocfs2_check_range_for_refcount(inode, saved_pos, count); + if (ret == 1) { + ocfs2_inode_unlock(inode, meta_level); + meta_level = -1; + + ret = ocfs2_prepare_inode_for_refcount(inode, + file, + saved_pos, + count, + &meta_level); + if (has_refcount) + *has_refcount = 1; + if (direct_io) + *direct_io = 0; + } + + if (ret < 0) { + mlog_errno(ret); + goto out_unlock; + } + /* * Skip the O_DIRECT checks if we don't need * them. @@ -1749,74 +2222,107 @@ static int ocfs2_prepare_inode_for_write(struct dentry *dentry, *ppos = saved_pos; out_unlock: - ocfs2_inode_unlock(inode, meta_level); + trace_ocfs2_prepare_inode_for_write(OCFS2_I(inode)->ip_blkno, + saved_pos, appending, count, + direct_io, has_refcount); + + if (meta_level >= 0) + ocfs2_inode_unlock(inode, meta_level); out: return ret; } -static ssize_t ocfs2_file_aio_write(struct kiocb *iocb, - const struct iovec *iov, - unsigned long nr_segs, - loff_t pos) +static ssize_t ocfs2_file_write_iter(struct kiocb *iocb, + struct iov_iter *from) { int ret, direct_io, appending, rw_level, have_alloc_sem = 0; - int can_do_direct; + int can_do_direct, has_refcount = 0; ssize_t written = 0; - size_t ocount; /* original count */ - size_t count; /* after file limit checks */ + size_t count = iov_iter_count(from); loff_t old_size, *ppos = &iocb->ki_pos; u32 old_clusters; struct file *file = iocb->ki_filp; - struct inode *inode = file->f_path.dentry->d_inode; + struct inode *inode = file_inode(file); struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + int full_coherency = !(osb->s_mount_opt & + OCFS2_MOUNT_COHERENCY_BUFFERED); + int unaligned_dio = 0; - mlog_entry("(0x%p, %u, '%.*s')\n", file, - (unsigned int)nr_segs, - file->f_path.dentry->d_name.len, - file->f_path.dentry->d_name.name); + trace_ocfs2_file_aio_write(inode, file, file->f_path.dentry, + (unsigned long long)OCFS2_I(inode)->ip_blkno, + file->f_path.dentry->d_name.len, + file->f_path.dentry->d_name.name, + (unsigned int)from->nr_segs); /* GRRRRR */ - if (iocb->ki_left == 0) + if (iocb->ki_nbytes == 0) return 0; - vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE); - appending = file->f_flags & O_APPEND ? 1 : 0; direct_io = file->f_flags & O_DIRECT ? 1 : 0; mutex_lock(&inode->i_mutex); + ocfs2_iocb_clear_sem_locked(iocb); + relock: - /* to match setattr's i_mutex -> i_alloc_sem -> rw_lock ordering */ + /* to match setattr's i_mutex -> rw_lock ordering */ if (direct_io) { - down_read(&inode->i_alloc_sem); have_alloc_sem = 1; + /* communicate with ocfs2_dio_end_io */ + ocfs2_iocb_set_sem_locked(iocb); } - /* concurrent O_DIRECT writes are allowed */ - rw_level = !direct_io; + /* + * Concurrent O_DIRECT writes are allowed with + * mount_option "coherency=buffered". + */ + rw_level = (!direct_io || full_coherency); + ret = ocfs2_rw_lock(inode, rw_level); if (ret < 0) { mlog_errno(ret); goto out_sems; } + /* + * O_DIRECT writes with "coherency=full" need to take EX cluster + * inode_lock to guarantee coherency. + */ + if (direct_io && full_coherency) { + /* + * We need to take and drop the inode lock to force + * other nodes to drop their caches. Buffered I/O + * already does this in write_begin(). + */ + ret = ocfs2_inode_lock(inode, NULL, 1); + if (ret < 0) { + mlog_errno(ret); + goto out; + } + + ocfs2_inode_unlock(inode, 1); + } + can_do_direct = direct_io; - ret = ocfs2_prepare_inode_for_write(file->f_path.dentry, ppos, - iocb->ki_left, appending, - &can_do_direct); + ret = ocfs2_prepare_inode_for_write(file, ppos, + iocb->ki_nbytes, appending, + &can_do_direct, &has_refcount); if (ret < 0) { mlog_errno(ret); goto out; } + if (direct_io && !is_sync_kiocb(iocb)) + unaligned_dio = ocfs2_is_io_unaligned(inode, iocb->ki_nbytes, + *ppos); + /* * We can't complete the direct I/O as requested, fall back to * buffered I/O. */ if (direct_io && !can_do_direct) { ocfs2_rw_unlock(inode, rw_level); - up_read(&inode->i_alloc_sem); have_alloc_sem = 0; rw_level = -1; @@ -1825,6 +2331,16 @@ relock: goto relock; } + if (unaligned_dio) { + /* + * Wait on previous unaligned aio to complete before + * proceeding. + */ + mutex_lock(&OCFS2_I(inode)->ip_unaligned_aio); + /* Mark the iocb as needing an unlock in ocfs2_dio_end_io */ + ocfs2_iocb_set_unaligned_aio(iocb); + } + /* * To later detect whether a journal commit for sync writes is * necessary, we sample i_size, and cluster count here. @@ -1835,67 +2351,68 @@ relock: /* communicate with ocfs2_dio_end_io */ ocfs2_iocb_set_rw_locked(iocb, rw_level); - if (direct_io) { - ret = generic_segment_checks(iov, &nr_segs, &ocount, - VERIFY_READ); - if (ret) - goto out_dio; - - ret = generic_write_checks(file, ppos, &count, - S_ISBLK(inode->i_mode)); - if (ret) - goto out_dio; + ret = generic_write_checks(file, ppos, &count, + S_ISBLK(inode->i_mode)); + if (ret) + goto out_dio; - written = generic_file_direct_write(iocb, iov, &nr_segs, *ppos, - ppos, count, ocount); + iov_iter_truncate(from, count); + if (direct_io) { + written = generic_file_direct_write(iocb, from, *ppos); if (written < 0) { - /* - * direct write may have instantiated a few - * blocks outside i_size. Trim these off again. - * Don't need i_size_read because we hold i_mutex. - */ - if (*ppos + count > inode->i_size) - vmtruncate(inode, inode->i_size); ret = written; goto out_dio; } } else { - written = generic_file_aio_write_nolock(iocb, iov, nr_segs, - *ppos); + current->backing_dev_info = file->f_mapping->backing_dev_info; + written = generic_perform_write(file, from, *ppos); + if (likely(written >= 0)) + iocb->ki_pos = *ppos + written; + current->backing_dev_info = NULL; } out_dio: /* buffered aio wouldn't have proper lock coverage today */ BUG_ON(ret == -EIOCBQUEUED && !(file->f_flags & O_DIRECT)); - if ((file->f_flags & O_SYNC && !direct_io) || IS_SYNC(inode)) { - /* - * The generic write paths have handled getting data - * to disk, but since we don't make use of the dirty - * inode list, a manual journal commit is necessary - * here. - */ - if (old_size != i_size_read(inode) || - old_clusters != OCFS2_I(inode)->ip_clusters) { + if (((file->f_flags & O_DSYNC) && !direct_io) || IS_SYNC(inode) || + ((file->f_flags & O_DIRECT) && !direct_io)) { + ret = filemap_fdatawrite_range(file->f_mapping, *ppos, + *ppos + count - 1); + if (ret < 0) + written = ret; + + if (!ret && ((old_size != i_size_read(inode)) || + (old_clusters != OCFS2_I(inode)->ip_clusters) || + has_refcount)) { ret = jbd2_journal_force_commit(osb->journal->j_journal); if (ret < 0) written = ret; } + + if (!ret) + ret = filemap_fdatawait_range(file->f_mapping, *ppos, + *ppos + count - 1); } - /* + /* * deep in g_f_a_w_n()->ocfs2_direct_IO we pass in a ocfs2_dio_end_io * function pointer which is called when o_direct io completes so that - * it can unlock our rw lock. (it's the clustered equivalent of - * i_alloc_sem; protects truncate from racing with pending ios). + * it can unlock our rw lock. * Unfortunately there are error cases which call end_io and others * that don't. so we don't have to unlock the rw_lock if either an * async dio is going to do it in the future or an end_io after an * error has already done it. */ - if (ret == -EIOCBQUEUED || !ocfs2_iocb_is_rw_locked(iocb)) { + if ((ret == -EIOCBQUEUED) || (!ocfs2_iocb_is_rw_locked(iocb))) { rw_level = -1; have_alloc_sem = 0; + unaligned_dio = 0; + } + + if (unaligned_dio) { + ocfs2_iocb_clear_unaligned_aio(iocb); + mutex_unlock(&OCFS2_I(inode)->ip_unaligned_aio); } out: @@ -1904,109 +2421,12 @@ out: out_sems: if (have_alloc_sem) - up_read(&inode->i_alloc_sem); + ocfs2_iocb_clear_sem_locked(iocb); mutex_unlock(&inode->i_mutex); - mlog_exit(ret); - return written ? written : ret; -} - -static int ocfs2_splice_to_file(struct pipe_inode_info *pipe, - struct file *out, - struct splice_desc *sd) -{ - int ret; - - ret = ocfs2_prepare_inode_for_write(out->f_path.dentry, &sd->pos, - sd->total_len, 0, NULL); - if (ret < 0) { - mlog_errno(ret); - return ret; - } - - return splice_from_pipe_feed(pipe, sd, pipe_to_file); -} - -static ssize_t ocfs2_file_splice_write(struct pipe_inode_info *pipe, - struct file *out, - loff_t *ppos, - size_t len, - unsigned int flags) -{ - int ret; - struct address_space *mapping = out->f_mapping; - struct inode *inode = mapping->host; - struct splice_desc sd = { - .total_len = len, - .flags = flags, - .pos = *ppos, - .u.file = out, - }; - - mlog_entry("(0x%p, 0x%p, %u, '%.*s')\n", out, pipe, - (unsigned int)len, - out->f_path.dentry->d_name.len, - out->f_path.dentry->d_name.name); - - if (pipe->inode) - mutex_lock_nested(&pipe->inode->i_mutex, I_MUTEX_PARENT); - - splice_from_pipe_begin(&sd); - do { - ret = splice_from_pipe_next(pipe, &sd); - if (ret <= 0) - break; - - mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD); - ret = ocfs2_rw_lock(inode, 1); - if (ret < 0) - mlog_errno(ret); - else { - ret = ocfs2_splice_to_file(pipe, out, &sd); - ocfs2_rw_unlock(inode, 1); - } - mutex_unlock(&inode->i_mutex); - } while (ret > 0); - splice_from_pipe_end(pipe, &sd); - - if (pipe->inode) - mutex_unlock(&pipe->inode->i_mutex); - - if (sd.num_spliced) - ret = sd.num_spliced; - - if (ret > 0) { - unsigned long nr_pages; - - *ppos += ret; - nr_pages = (ret + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; - - /* - * If file or inode is SYNC and we actually wrote some data, - * sync it. - */ - if (unlikely((out->f_flags & O_SYNC) || IS_SYNC(inode))) { - int err; - - mutex_lock(&inode->i_mutex); - err = ocfs2_rw_lock(inode, 1); - if (err < 0) { - mlog_errno(err); - } else { - err = generic_osync_inode(inode, mapping, - OSYNC_METADATA|OSYNC_DATA); - ocfs2_rw_unlock(inode, 1); - } - mutex_unlock(&inode->i_mutex); - - if (err) - ret = err; - } - balance_dirty_pages_ratelimited_nr(mapping, nr_pages); - } - - mlog_exit(ret); + if (written) + ret = written; return ret; } @@ -2016,44 +2436,43 @@ static ssize_t ocfs2_file_splice_read(struct file *in, size_t len, unsigned int flags) { - int ret = 0; - struct inode *inode = in->f_path.dentry->d_inode; + int ret = 0, lock_level = 0; + struct inode *inode = file_inode(in); - mlog_entry("(0x%p, 0x%p, %u, '%.*s')\n", in, pipe, - (unsigned int)len, - in->f_path.dentry->d_name.len, - in->f_path.dentry->d_name.name); + trace_ocfs2_file_splice_read(inode, in, in->f_path.dentry, + (unsigned long long)OCFS2_I(inode)->ip_blkno, + in->f_path.dentry->d_name.len, + in->f_path.dentry->d_name.name, len); /* - * See the comment in ocfs2_file_aio_read() + * See the comment in ocfs2_file_read_iter() */ - ret = ocfs2_inode_lock(inode, NULL, 0); + ret = ocfs2_inode_lock_atime(inode, in->f_path.mnt, &lock_level); if (ret < 0) { mlog_errno(ret); goto bail; } - ocfs2_inode_unlock(inode, 0); + ocfs2_inode_unlock(inode, lock_level); ret = generic_file_splice_read(in, ppos, pipe, len, flags); bail: - mlog_exit(ret); return ret; } -static ssize_t ocfs2_file_aio_read(struct kiocb *iocb, - const struct iovec *iov, - unsigned long nr_segs, - loff_t pos) +static ssize_t ocfs2_file_read_iter(struct kiocb *iocb, + struct iov_iter *to) { int ret = 0, rw_level = -1, have_alloc_sem = 0, lock_level = 0; struct file *filp = iocb->ki_filp; - struct inode *inode = filp->f_path.dentry->d_inode; + struct inode *inode = file_inode(filp); + + trace_ocfs2_file_aio_read(inode, filp, filp->f_path.dentry, + (unsigned long long)OCFS2_I(inode)->ip_blkno, + filp->f_path.dentry->d_name.len, + filp->f_path.dentry->d_name.name, + to->nr_segs); /* GRRRRR */ - mlog_entry("(0x%p, %u, '%.*s')\n", filp, - (unsigned int)nr_segs, - filp->f_path.dentry->d_name.len, - filp->f_path.dentry->d_name.name); if (!inode) { ret = -EINVAL; @@ -2061,13 +2480,15 @@ static ssize_t ocfs2_file_aio_read(struct kiocb *iocb, goto bail; } - /* + ocfs2_iocb_clear_sem_locked(iocb); + + /* * buffered reads protect themselves in ->readpage(). O_DIRECT reads * need locks to protect pending reads from racing with truncate. */ if (filp->f_flags & O_DIRECT) { - down_read(&inode->i_alloc_sem); have_alloc_sem = 1; + ocfs2_iocb_set_sem_locked(iocb); ret = ocfs2_rw_lock(inode, 0); if (ret < 0) { @@ -2083,26 +2504,25 @@ static ssize_t ocfs2_file_aio_read(struct kiocb *iocb, * We're fine letting folks race truncates and extending * writes with read across the cluster, just like they can * locally. Hence no rw_lock during read. - * + * * Take and drop the meta data lock to update inode fields * like i_size. This allows the checks down below - * generic_file_aio_read() a chance of actually working. + * generic_file_aio_read() a chance of actually working. */ - ret = ocfs2_inode_lock_atime(inode, filp->f_vfsmnt, &lock_level); + ret = ocfs2_inode_lock_atime(inode, filp->f_path.mnt, &lock_level); if (ret < 0) { mlog_errno(ret); goto bail; } ocfs2_inode_unlock(inode, lock_level); - ret = generic_file_aio_read(iocb, iov, nr_segs, iocb->ki_pos); - if (ret == -EINVAL) - mlog(0, "generic_file_aio_read returned -EINVAL\n"); + ret = generic_file_read_iter(iocb, to); + trace_generic_file_aio_read_ret(ret); /* buffered aio wouldn't have proper lock coverage today */ BUG_ON(ret == -EIOCBQUEUED && !(filp->f_flags & O_DIRECT)); - /* see ocfs2_file_aio_write */ + /* see ocfs2_file_write_iter */ if (ret == -EIOCBQUEUED || !ocfs2_iocb_is_rw_locked(iocb)) { rw_level = -1; have_alloc_sem = 0; @@ -2110,14 +2530,64 @@ static ssize_t ocfs2_file_aio_read(struct kiocb *iocb, bail: if (have_alloc_sem) - up_read(&inode->i_alloc_sem); - if (rw_level != -1) + ocfs2_iocb_clear_sem_locked(iocb); + + if (rw_level != -1) ocfs2_rw_unlock(inode, rw_level); - mlog_exit(ret); return ret; } +/* Refer generic_file_llseek_unlocked() */ +static loff_t ocfs2_file_llseek(struct file *file, loff_t offset, int whence) +{ + struct inode *inode = file->f_mapping->host; + int ret = 0; + + mutex_lock(&inode->i_mutex); + + switch (whence) { + case SEEK_SET: + break; + case SEEK_END: + /* SEEK_END requires the OCFS2 inode lock for the file + * because it references the file's size. + */ + ret = ocfs2_inode_lock(inode, NULL, 0); + if (ret < 0) { + mlog_errno(ret); + goto out; + } + offset += i_size_read(inode); + ocfs2_inode_unlock(inode, 0); + break; + case SEEK_CUR: + if (offset == 0) { + offset = file->f_pos; + goto out; + } + offset += file->f_pos; + break; + case SEEK_DATA: + case SEEK_HOLE: + ret = ocfs2_seek_data_hole_offset(file, &offset, whence); + if (ret) + goto out; + break; + default: + ret = -EINVAL; + goto out; + } + + offset = vfs_setpos(file, offset, inode->i_sb->s_maxbytes); + +out: + mutex_unlock(&inode->i_mutex); + if (ret) + return ret; + return offset; +} + const struct inode_operations ocfs2_file_iops = { .setattr = ocfs2_setattr, .getattr = ocfs2_getattr, @@ -2126,14 +2596,17 @@ const struct inode_operations ocfs2_file_iops = { .getxattr = generic_getxattr, .listxattr = ocfs2_listxattr, .removexattr = generic_removexattr, - .fallocate = ocfs2_fallocate, .fiemap = ocfs2_fiemap, + .get_acl = ocfs2_iop_get_acl, + .set_acl = ocfs2_iop_set_acl, }; const struct inode_operations ocfs2_special_file_iops = { .setattr = ocfs2_setattr, .getattr = ocfs2_getattr, .permission = ocfs2_permission, + .get_acl = ocfs2_iop_get_acl, + .set_acl = ocfs2_iop_set_acl, }; /* @@ -2141,15 +2614,15 @@ const struct inode_operations ocfs2_special_file_iops = { * ocfs2_fops_no_plocks and ocfs2_dops_no_plocks! */ const struct file_operations ocfs2_fops = { - .llseek = generic_file_llseek, - .read = do_sync_read, - .write = do_sync_write, + .llseek = ocfs2_file_llseek, + .read = new_sync_read, + .write = new_sync_write, .mmap = ocfs2_mmap, .fsync = ocfs2_sync_file, .release = ocfs2_file_release, .open = ocfs2_file_open, - .aio_read = ocfs2_file_aio_read, - .aio_write = ocfs2_file_aio_write, + .read_iter = ocfs2_file_read_iter, + .write_iter = ocfs2_file_write_iter, .unlocked_ioctl = ocfs2_ioctl, #ifdef CONFIG_COMPAT .compat_ioctl = ocfs2_compat_ioctl, @@ -2157,13 +2630,14 @@ const struct file_operations ocfs2_fops = { .lock = ocfs2_lock, .flock = ocfs2_flock, .splice_read = ocfs2_file_splice_read, - .splice_write = ocfs2_file_splice_write, + .splice_write = iter_file_splice_write, + .fallocate = ocfs2_fallocate, }; const struct file_operations ocfs2_dops = { .llseek = generic_file_llseek, .read = generic_read_dir, - .readdir = ocfs2_readdir, + .iterate = ocfs2_readdir, .fsync = ocfs2_sync_file, .release = ocfs2_dir_release, .open = ocfs2_dir_open, @@ -2188,28 +2662,29 @@ const struct file_operations ocfs2_dops = { * the cluster. */ const struct file_operations ocfs2_fops_no_plocks = { - .llseek = generic_file_llseek, - .read = do_sync_read, - .write = do_sync_write, + .llseek = ocfs2_file_llseek, + .read = new_sync_read, + .write = new_sync_write, .mmap = ocfs2_mmap, .fsync = ocfs2_sync_file, .release = ocfs2_file_release, .open = ocfs2_file_open, - .aio_read = ocfs2_file_aio_read, - .aio_write = ocfs2_file_aio_write, + .read_iter = ocfs2_file_read_iter, + .write_iter = ocfs2_file_write_iter, .unlocked_ioctl = ocfs2_ioctl, #ifdef CONFIG_COMPAT .compat_ioctl = ocfs2_compat_ioctl, #endif .flock = ocfs2_flock, .splice_read = ocfs2_file_splice_read, - .splice_write = ocfs2_file_splice_write, + .splice_write = iter_file_splice_write, + .fallocate = ocfs2_fallocate, }; const struct file_operations ocfs2_dops_no_plocks = { .llseek = generic_file_llseek, .read = generic_read_dir, - .readdir = ocfs2_readdir, + .iterate = ocfs2_readdir, .fsync = ocfs2_sync_file, .release = ocfs2_dir_release, .open = ocfs2_dir_open, diff --git a/fs/ocfs2/file.h b/fs/ocfs2/file.h index 172f9fbc9fc..97bf761c9e7 100644 --- a/fs/ocfs2/file.h +++ b/fs/ocfs2/file.h @@ -54,8 +54,10 @@ int ocfs2_add_inode_data(struct ocfs2_super *osb, int ocfs2_simple_size_update(struct inode *inode, struct buffer_head *di_bh, u64 new_i_size); -int ocfs2_extend_no_holes(struct inode *inode, u64 new_i_size, - u64 zero_to); +int ocfs2_extend_no_holes(struct inode *inode, struct buffer_head *di_bh, + u64 new_i_size, u64 zero_to); +int ocfs2_zero_extend(struct inode *inode, struct buffer_head *di_bh, + loff_t zero_to); int ocfs2_setattr(struct dentry *dentry, struct iattr *attr); int ocfs2_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat); @@ -69,4 +71,6 @@ int ocfs2_update_inode_atime(struct inode *inode, int ocfs2_change_file_space(struct file *file, unsigned int cmd, struct ocfs2_space_resv *sr); +int ocfs2_check_range_for_refcount(struct inode *inode, loff_t pos, + size_t count); #endif /* OCFS2_FILE_H */ diff --git a/fs/ocfs2/heartbeat.c b/fs/ocfs2/heartbeat.c index c6e7213db86..d8208b20dc5 100644 --- a/fs/ocfs2/heartbeat.c +++ b/fs/ocfs2/heartbeat.c @@ -26,10 +26,8 @@ #include <linux/fs.h> #include <linux/types.h> -#include <linux/slab.h> #include <linux/highmem.h> -#define MLOG_MASK_PREFIX ML_SUPER #include <cluster/masklog.h> #include "ocfs2.h" @@ -38,6 +36,7 @@ #include "heartbeat.h" #include "inode.h" #include "journal.h" +#include "ocfs2_trace.h" #include "buffer_head_io.h" @@ -67,7 +66,7 @@ void ocfs2_do_node_down(int node_num, void *data) BUG_ON(osb->node_num == node_num); - mlog(0, "ocfs2: node down event for %d\n", node_num); + trace_ocfs2_do_node_down(node_num); if (!osb->cconn) { /* diff --git a/fs/ocfs2/inode.c b/fs/ocfs2/inode.c index 10e1fa87396..437de7f768c 100644 --- a/fs/ocfs2/inode.c +++ b/fs/ocfs2/inode.c @@ -25,14 +25,12 @@ #include <linux/fs.h> #include <linux/types.h> -#include <linux/slab.h> #include <linux/highmem.h> #include <linux/pagemap.h> #include <linux/quotaops.h> #include <asm/byteorder.h> -#define MLOG_MASK_PREFIX ML_INODE #include <cluster/masklog.h> #include "ocfs2.h" @@ -53,6 +51,8 @@ #include "sysfile.h" #include "uptodate.h" #include "xattr.h" +#include "refcounttree.h" +#include "ocfs2_trace.h" #include "buffer_head_io.h" @@ -130,8 +130,10 @@ struct inode *ocfs2_iget(struct ocfs2_super *osb, u64 blkno, unsigned flags, struct inode *inode = NULL; struct super_block *sb = osb->sb; struct ocfs2_find_inode_args args; + journal_t *journal = OCFS2_SB(sb)->journal->j_journal; - mlog_entry("(blkno = %llu)\n", (unsigned long long)blkno); + trace_ocfs2_iget_begin((unsigned long long)blkno, flags, + sysfile_type); /* Ok. By now we've either got the offsets passed to us by the * caller, or we just pulled them off the bh. Lets do some @@ -152,27 +154,52 @@ struct inode *ocfs2_iget(struct ocfs2_super *osb, u64 blkno, unsigned flags, /* inode was *not* in the inode cache. 2.6.x requires * us to do our own read_inode call and unlock it * afterwards. */ - if (inode && inode->i_state & I_NEW) { - mlog(0, "Inode was not in inode cache, reading it.\n"); - ocfs2_read_locked_inode(inode, &args); - unlock_new_inode(inode); - } if (inode == NULL) { inode = ERR_PTR(-ENOMEM); mlog_errno(PTR_ERR(inode)); goto bail; } + trace_ocfs2_iget5_locked(inode->i_state); + if (inode->i_state & I_NEW) { + ocfs2_read_locked_inode(inode, &args); + unlock_new_inode(inode); + } if (is_bad_inode(inode)) { iput(inode); inode = ERR_PTR(-ESTALE); goto bail; } + /* + * Set transaction id's of transactions that have to be committed + * to finish f[data]sync. We set them to currently running transaction + * as we cannot be sure that the inode or some of its metadata isn't + * part of the transaction - the inode could have been reclaimed and + * now it is reread from disk. + */ + if (journal) { + transaction_t *transaction; + tid_t tid; + struct ocfs2_inode_info *oi = OCFS2_I(inode); + + read_lock(&journal->j_state_lock); + if (journal->j_running_transaction) + transaction = journal->j_running_transaction; + else + transaction = journal->j_committing_transaction; + if (transaction) + tid = transaction->t_tid; + else + tid = journal->j_commit_sequence; + read_unlock(&journal->j_state_lock); + oi->i_sync_tid = tid; + oi->i_datasync_tid = tid; + } + bail: if (!IS_ERR(inode)) { - mlog(0, "returning inode with number %llu\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno); - mlog_exit_ptr(inode); + trace_ocfs2_iget_end(inode, + (unsigned long long)OCFS2_I(inode)->ip_blkno); } return inode; @@ -192,18 +219,17 @@ static int ocfs2_find_actor(struct inode *inode, void *opaque) struct ocfs2_inode_info *oi = OCFS2_I(inode); int ret = 0; - mlog_entry("(0x%p, %lu, 0x%p)\n", inode, inode->i_ino, opaque); - args = opaque; mlog_bug_on_msg(!inode, "No inode in find actor!\n"); + trace_ocfs2_find_actor(inode, inode->i_ino, opaque, args->fi_blkno); + if (oi->ip_blkno != args->fi_blkno) goto bail; ret = 1; bail: - mlog_exit(ret); return ret; } @@ -215,16 +241,24 @@ bail: static int ocfs2_init_locked_inode(struct inode *inode, void *opaque) { struct ocfs2_find_inode_args *args = opaque; - - mlog_entry("inode = %p, opaque = %p\n", inode, opaque); + static struct lock_class_key ocfs2_quota_ip_alloc_sem_key, + ocfs2_file_ip_alloc_sem_key; inode->i_ino = args->fi_ino; OCFS2_I(inode)->ip_blkno = args->fi_blkno; if (args->fi_sysfile_type != 0) lockdep_set_class(&inode->i_mutex, &ocfs2_sysfile_lock_key[args->fi_sysfile_type]); + if (args->fi_sysfile_type == USER_QUOTA_SYSTEM_INODE || + args->fi_sysfile_type == GROUP_QUOTA_SYSTEM_INODE || + args->fi_sysfile_type == LOCAL_USER_QUOTA_SYSTEM_INODE || + args->fi_sysfile_type == LOCAL_GROUP_QUOTA_SYSTEM_INODE) + lockdep_set_class(&OCFS2_I(inode)->ip_alloc_sem, + &ocfs2_quota_ip_alloc_sem_key); + else + lockdep_set_class(&OCFS2_I(inode)->ip_alloc_sem, + &ocfs2_file_ip_alloc_sem_key); - mlog_exit(0); return 0; } @@ -235,9 +269,6 @@ void ocfs2_populate_inode(struct inode *inode, struct ocfs2_dinode *fe, struct ocfs2_super *osb; int use_plocks = 1; - mlog_entry("(0x%p, size:%llu)\n", inode, - (unsigned long long)le64_to_cpu(fe->i_size)); - sb = inode->i_sb; osb = OCFS2_SB(sb); @@ -265,15 +296,17 @@ void ocfs2_populate_inode(struct inode *inode, struct ocfs2_dinode *fe, inode->i_generation = le32_to_cpu(fe->i_generation); inode->i_rdev = huge_decode_dev(le64_to_cpu(fe->id1.dev1.i_rdev)); inode->i_mode = le16_to_cpu(fe->i_mode); - inode->i_uid = le32_to_cpu(fe->i_uid); - inode->i_gid = le32_to_cpu(fe->i_gid); + i_uid_write(inode, le32_to_cpu(fe->i_uid)); + i_gid_write(inode, le32_to_cpu(fe->i_gid)); /* Fast symlinks will have i_size but no allocated clusters. */ - if (S_ISLNK(inode->i_mode) && !fe->i_clusters) + if (S_ISLNK(inode->i_mode) && !fe->i_clusters) { inode->i_blocks = 0; - else + inode->i_mapping->a_ops = &ocfs2_fast_symlink_aops; + } else { inode->i_blocks = ocfs2_inode_sector_count(inode); - inode->i_mapping->a_ops = &ocfs2_aops; + inode->i_mapping->a_ops = &ocfs2_aops; + } inode->i_atime.tv_sec = le64_to_cpu(fe->i_atime); inode->i_atime.tv_nsec = le32_to_cpu(fe->i_atime_nsec); inode->i_mtime.tv_sec = le64_to_cpu(fe->i_mtime); @@ -287,22 +320,22 @@ void ocfs2_populate_inode(struct inode *inode, struct ocfs2_dinode *fe, (unsigned long long)OCFS2_I(inode)->ip_blkno, (unsigned long long)le64_to_cpu(fe->i_blkno)); - inode->i_nlink = ocfs2_read_links_count(fe); + set_nlink(inode, ocfs2_read_links_count(fe)); + trace_ocfs2_populate_inode(OCFS2_I(inode)->ip_blkno, + le32_to_cpu(fe->i_flags)); if (fe->i_flags & cpu_to_le32(OCFS2_SYSTEM_FL)) { OCFS2_I(inode)->ip_flags |= OCFS2_INODE_SYSTEM_FILE; inode->i_flags |= S_NOQUOTA; } - + if (fe->i_flags & cpu_to_le32(OCFS2_LOCAL_ALLOC_FL)) { OCFS2_I(inode)->ip_flags |= OCFS2_INODE_BITMAP; - mlog(0, "local alloc inode: i_ino=%lu\n", inode->i_ino); } else if (fe->i_flags & cpu_to_le32(OCFS2_BITMAP_FL)) { OCFS2_I(inode)->ip_flags |= OCFS2_INODE_BITMAP; } else if (fe->i_flags & cpu_to_le32(OCFS2_QUOTA_FL)) { inode->i_flags |= S_NOQUOTA; } else if (fe->i_flags & cpu_to_le32(OCFS2_SUPER_BLOCK_FL)) { - mlog(0, "superblock inode: i_ino=%lu\n", inode->i_ino); /* we can't actually hit this as read_inode can't * handle superblocks today ;-) */ BUG(); @@ -324,12 +357,10 @@ void ocfs2_populate_inode(struct inode *inode, struct ocfs2_dinode *fe, else inode->i_fop = &ocfs2_dops_no_plocks; i_size_write(inode, le64_to_cpu(fe->i_size)); + OCFS2_I(inode)->ip_dir_lock_gen = 1; break; case S_IFLNK: - if (ocfs2_inode_is_fast_symlink(inode)) - inode->i_op = &ocfs2_fast_symlink_inode_operations; - else - inode->i_op = &ocfs2_symlink_inode_operations; + inode->i_op = &ocfs2_symlink_inode_operations; i_size_write(inode, le64_to_cpu(fe->i_size)); break; default: @@ -365,7 +396,10 @@ void ocfs2_populate_inode(struct inode *inode, struct ocfs2_dinode *fe, OCFS2_I(inode)->ip_last_used_slot = 0; OCFS2_I(inode)->ip_last_used_group = 0; - mlog_exit_void(); + + if (S_ISDIR(inode->i_mode)) + ocfs2_resv_set_type(&OCFS2_I(inode)->ip_la_data_resv, + OCFS2_RESV_FLAG_DIR); } static int ocfs2_read_locked_inode(struct inode *inode, @@ -378,22 +412,10 @@ static int ocfs2_read_locked_inode(struct inode *inode, int status, can_lock; u32 generation = 0; - mlog_entry("(0x%p, 0x%p)\n", inode, args); - status = -EINVAL; - if (inode == NULL || inode->i_sb == NULL) { - mlog(ML_ERROR, "bad inode\n"); - return status; - } sb = inode->i_sb; osb = OCFS2_SB(sb); - if (!args) { - mlog(ML_ERROR, "bad inode args\n"); - make_bad_inode(inode); - return status; - } - /* * To improve performance of cold-cache inode stats, we take * the cluster lock here if possible. @@ -418,7 +440,7 @@ static int ocfs2_read_locked_inode(struct inode *inode, * #1 and #2 can be simply solved by never taking the lock * here for system files (which are the only type we read * during mount). It's a heavier approach, but our main - * concern is user-accesible files anyway. + * concern is user-accessible files anyway. * * #3 works itself out because we'll eventually take the * cluster lock before trusting anything anyway. @@ -427,6 +449,9 @@ static int ocfs2_read_locked_inode(struct inode *inode, && !(args->fi_flags & OCFS2_FI_FLAG_ORPHAN_RECOVERY) && !ocfs2_mount_local(osb); + trace_ocfs2_read_locked_inode( + (unsigned long long)OCFS2_I(inode)->ip_blkno, can_lock); + /* * To maintain backwards compatibility with older versions of * ocfs2-tools, we still store the generation value for system @@ -463,7 +488,7 @@ static int ocfs2_read_locked_inode(struct inode *inode, if (args->fi_flags & OCFS2_FI_FLAG_ORPHAN_RECOVERY) { status = ocfs2_try_open_lock(inode, 0); if (status) { - make_bad_inode(inode); + make_bad_inode(inode); return status; } } @@ -473,7 +498,11 @@ static int ocfs2_read_locked_inode(struct inode *inode, OCFS2_BH_IGNORE_CACHE); } else { status = ocfs2_read_blocks_sync(osb, args->fi_blkno, 1, &bh); - if (!status) + /* + * If buffer is in jbd, then its checksum may not have been + * computed as yet. + */ + if (!status && !buffer_jbd(bh)) status = ocfs2_validate_inode_block(osb->sb, bh); } if (status < 0) { @@ -514,7 +543,6 @@ bail: if (args && bh) brelse(bh); - mlog_exit(status); return status; } @@ -528,12 +556,9 @@ static int ocfs2_truncate_for_delete(struct ocfs2_super *osb, struct buffer_head *fe_bh) { int status = 0; - struct ocfs2_truncate_context *tc = NULL; struct ocfs2_dinode *fe; handle_t *handle = NULL; - mlog_entry_void(); - fe = (struct ocfs2_dinode *) fe_bh->b_data; /* @@ -547,11 +572,13 @@ static int ocfs2_truncate_for_delete(struct ocfs2_super *osb, handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS); if (IS_ERR(handle)) { status = PTR_ERR(handle); + handle = NULL; mlog_errno(status); goto out; } - status = ocfs2_journal_access_di(handle, inode, fe_bh, + status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), + fe_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); @@ -569,13 +596,7 @@ static int ocfs2_truncate_for_delete(struct ocfs2_super *osb, ocfs2_commit_trans(osb, handle); handle = NULL; - status = ocfs2_prepare_truncate(osb, inode, fe_bh, &tc); - if (status < 0) { - mlog_errno(status); - goto out; - } - - status = ocfs2_commit_truncate(osb, inode, fe_bh, tc); + status = ocfs2_commit_truncate(osb, inode, fe_bh); if (status < 0) { mlog_errno(status); goto out; @@ -585,7 +606,6 @@ static int ocfs2_truncate_for_delete(struct ocfs2_super *osb, out: if (handle) ocfs2_commit_trans(osb, handle); - mlog_exit(status); return status; } @@ -627,15 +647,17 @@ static int ocfs2_remove_inode(struct inode *inode, goto bail_unlock; } - status = ocfs2_orphan_del(osb, handle, orphan_dir_inode, inode, - orphan_dir_bh); - if (status < 0) { - mlog_errno(status); - goto bail_commit; + if (!(OCFS2_I(inode)->ip_flags & OCFS2_INODE_SKIP_ORPHAN_DIR)) { + status = ocfs2_orphan_del(osb, handle, orphan_dir_inode, inode, + orphan_dir_bh); + if (status < 0) { + mlog_errno(status); + goto bail_commit; + } } /* set the inodes dtime */ - status = ocfs2_journal_access_di(handle, inode, di_bh, + status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); @@ -644,15 +666,10 @@ static int ocfs2_remove_inode(struct inode *inode, di->i_dtime = cpu_to_le64(CURRENT_TIME.tv_sec); di->i_flags &= cpu_to_le32(~(OCFS2_VALID_FL | OCFS2_ORPHANED_FL)); + ocfs2_journal_dirty(handle, di_bh); - status = ocfs2_journal_dirty(handle, di_bh); - if (status < 0) { - mlog_errno(status); - goto bail_commit; - } - - ocfs2_remove_from_cache(inode, di_bh); - vfs_dq_free_inode(inode); + ocfs2_remove_from_cache(INODE_CACHE(inode), di_bh); + dquot_free_inode(inode); status = ocfs2_free_dinode(handle, inode_alloc_inode, inode_alloc_bh, di); @@ -671,7 +688,7 @@ bail: return status; } -/* +/* * Serialize with orphan dir recovery. If the process doing * recovery on this orphan dir does an iget() with the dir * i_mutex held, we'll deadlock here. Instead we detect this @@ -684,8 +701,6 @@ static int ocfs2_check_orphan_recovery_state(struct ocfs2_super *osb, spin_lock(&osb->osb_lock); if (ocfs2_node_map_test_bit(osb, &osb->osb_recovering_orphan_dirs, slot)) { - mlog(0, "Recovery is happening on orphan dir %d, will skip " - "this inode\n", slot); ret = -EDEADLK; goto out; } @@ -694,6 +709,7 @@ static int ocfs2_check_orphan_recovery_state(struct ocfs2_super *osb, osb->osb_orphan_wipes[slot]++; out: spin_unlock(&osb->osb_lock); + trace_ocfs2_check_orphan_recovery_state(slot, ret); return ret; } @@ -710,38 +726,39 @@ static void ocfs2_signal_wipe_completion(struct ocfs2_super *osb, static int ocfs2_wipe_inode(struct inode *inode, struct buffer_head *di_bh) { - int status, orphaned_slot; + int status, orphaned_slot = -1; struct inode *orphan_dir_inode = NULL; struct buffer_head *orphan_dir_bh = NULL; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - struct ocfs2_dinode *di; + struct ocfs2_dinode *di = (struct ocfs2_dinode *) di_bh->b_data; - di = (struct ocfs2_dinode *) di_bh->b_data; - orphaned_slot = le16_to_cpu(di->i_orphaned_slot); + if (!(OCFS2_I(inode)->ip_flags & OCFS2_INODE_SKIP_ORPHAN_DIR)) { + orphaned_slot = le16_to_cpu(di->i_orphaned_slot); - status = ocfs2_check_orphan_recovery_state(osb, orphaned_slot); - if (status) - return status; + status = ocfs2_check_orphan_recovery_state(osb, orphaned_slot); + if (status) + return status; - orphan_dir_inode = ocfs2_get_system_file_inode(osb, - ORPHAN_DIR_SYSTEM_INODE, - orphaned_slot); - if (!orphan_dir_inode) { - status = -EEXIST; - mlog_errno(status); - goto bail; - } + orphan_dir_inode = ocfs2_get_system_file_inode(osb, + ORPHAN_DIR_SYSTEM_INODE, + orphaned_slot); + if (!orphan_dir_inode) { + status = -EEXIST; + mlog_errno(status); + goto bail; + } - /* Lock the orphan dir. The lock will be held for the entire - * delete_inode operation. We do this now to avoid races with - * recovery completion on other nodes. */ - mutex_lock(&orphan_dir_inode->i_mutex); - status = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1); - if (status < 0) { - mutex_unlock(&orphan_dir_inode->i_mutex); + /* Lock the orphan dir. The lock will be held for the entire + * delete_inode operation. We do this now to avoid races with + * recovery completion on other nodes. */ + mutex_lock(&orphan_dir_inode->i_mutex); + status = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1); + if (status < 0) { + mutex_unlock(&orphan_dir_inode->i_mutex); - mlog_errno(status); - goto bail; + mlog_errno(status); + goto bail; + } } /* we do this while holding the orphan dir lock because we @@ -770,12 +787,21 @@ static int ocfs2_wipe_inode(struct inode *inode, goto bail_unlock_dir; } + status = ocfs2_remove_refcount_tree(inode, di_bh); + if (status < 0) { + mlog_errno(status); + goto bail_unlock_dir; + } + status = ocfs2_remove_inode(inode, di_bh, orphan_dir_inode, orphan_dir_bh); if (status < 0) mlog_errno(status); bail_unlock_dir: + if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SKIP_ORPHAN_DIR) + return status; + ocfs2_inode_unlock(orphan_dir_inode, 1); mutex_unlock(&orphan_dir_inode->i_mutex); brelse(orphan_dir_bh); @@ -794,6 +820,10 @@ static int ocfs2_inode_is_valid_to_delete(struct inode *inode) struct ocfs2_inode_info *oi = OCFS2_I(inode); struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + trace_ocfs2_inode_is_valid_to_delete(current, osb->dc_task, + (unsigned long long)oi->ip_blkno, + oi->ip_flags); + /* We shouldn't be getting here for the root directory * inode.. */ if (inode == osb->root_inode) { @@ -801,16 +831,15 @@ static int ocfs2_inode_is_valid_to_delete(struct inode *inode) goto bail; } - /* If we're coming from downconvert_thread we can't go into our own - * voting [hello, deadlock city!], so unforuntately we just - * have to skip deleting this guy. That's OK though because - * the node who's doing the actual deleting should handle it - * anyway. */ - if (current == osb->dc_task) { - mlog(0, "Skipping delete of %lu because we're currently " - "in downconvert\n", inode->i_ino); + /* + * If we're coming from downconvert_thread we can't go into our own + * voting [hello, deadlock city!] so we cannot delete the inode. But + * since we dropped last inode ref when downconverting dentry lock, + * we cannot have the file open and thus the node doing unlink will + * take care of deleting the inode. + */ + if (current == osb->dc_task) goto bail; - } spin_lock(&oi->ip_lock); /* OCFS2 *never* deletes system files. This should technically @@ -822,15 +851,6 @@ static int ocfs2_inode_is_valid_to_delete(struct inode *inode) goto bail_unlock; } - /* If we have allowd wipe of this inode for another node, it - * will be marked here so we can safely skip it. Recovery will - * cleanup any inodes we might inadvertantly skip here. */ - if (oi->ip_flags & OCFS2_INODE_SKIP_DELETE) { - mlog(0, "Skipping delete of %lu because another node " - "has done this for us.\n", inode->i_ino); - goto bail_unlock; - } - ret = 1; bail_unlock: spin_unlock(&oi->ip_lock); @@ -846,32 +866,45 @@ static int ocfs2_query_inode_wipe(struct inode *inode, struct buffer_head *di_bh, int *wipe) { - int status = 0; + int status = 0, reason = 0; struct ocfs2_inode_info *oi = OCFS2_I(inode); struct ocfs2_dinode *di; *wipe = 0; + trace_ocfs2_query_inode_wipe_begin((unsigned long long)oi->ip_blkno, + inode->i_nlink); + /* While we were waiting for the cluster lock in * ocfs2_delete_inode, another node might have asked to delete * the inode. Recheck our flags to catch this. */ if (!ocfs2_inode_is_valid_to_delete(inode)) { - mlog(0, "Skipping delete of %llu because flags changed\n", - (unsigned long long)oi->ip_blkno); + reason = 1; goto bail; } /* Now that we have an up to date inode, we can double check * the link count. */ - if (inode->i_nlink) { - mlog(0, "Skipping delete of %llu because nlink = %u\n", - (unsigned long long)oi->ip_blkno, inode->i_nlink); + if (inode->i_nlink) goto bail; - } /* Do some basic inode verification... */ di = (struct ocfs2_dinode *) di_bh->b_data; - if (!(di->i_flags & cpu_to_le32(OCFS2_ORPHANED_FL))) { + if (!(di->i_flags & cpu_to_le32(OCFS2_ORPHANED_FL)) && + !(oi->ip_flags & OCFS2_INODE_SKIP_ORPHAN_DIR)) { + /* + * Inodes in the orphan dir must have ORPHANED_FL. The only + * inodes that come back out of the orphan dir are reflink + * targets. A reflink target may be moved out of the orphan + * dir between the time we scan the directory and the time we + * process it. This would lead to HAS_REFCOUNT_FL being set but + * ORPHANED_FL not. + */ + if (di->i_dyn_features & cpu_to_le16(OCFS2_HAS_REFCOUNT_FL)) { + reason = 2; + goto bail; + } + /* for lack of a better error? */ status = -EEXIST; mlog(ML_ERROR, @@ -896,7 +929,7 @@ static int ocfs2_query_inode_wipe(struct inode *inode, * the inode open lock in ocfs2_read_locked_inode(). When we * get to ->delete_inode(), each node tries to convert it's * lock to an exclusive. Trylocks are serialized by the inode - * meta data lock. If the upconvert suceeds, we know the inode + * meta data lock. If the upconvert succeeds, we know the inode * is no longer live and can be deleted. * * Though we call this with the meta data lock held, the @@ -905,8 +938,7 @@ static int ocfs2_query_inode_wipe(struct inode *inode, status = ocfs2_try_open_lock(inode, 1); if (status == -EAGAIN) { status = 0; - mlog(0, "Skipping delete of %llu because it is in use on " - "other nodes\n", (unsigned long long)oi->ip_blkno); + reason = 3; goto bail; } if (status < 0) { @@ -915,11 +947,10 @@ static int ocfs2_query_inode_wipe(struct inode *inode, } *wipe = 1; - mlog(0, "Inode %llu is ok to wipe from orphan dir %u\n", - (unsigned long long)oi->ip_blkno, - le16_to_cpu(di->i_orphaned_slot)); + trace_ocfs2_query_inode_wipe_succ(le16_to_cpu(di->i_orphaned_slot)); bail: + trace_ocfs2_query_inode_wipe_end(status, reason); return status; } @@ -929,28 +960,28 @@ bail: static void ocfs2_cleanup_delete_inode(struct inode *inode, int sync_data) { - mlog(0, "Cleanup inode %llu, sync = %d\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, sync_data); + trace_ocfs2_cleanup_delete_inode( + (unsigned long long)OCFS2_I(inode)->ip_blkno, sync_data); if (sync_data) - write_inode_now(inode, 1); - truncate_inode_pages(&inode->i_data, 0); + filemap_write_and_wait(inode->i_mapping); + truncate_inode_pages_final(&inode->i_data); } -void ocfs2_delete_inode(struct inode *inode) +static void ocfs2_delete_inode(struct inode *inode) { int wipe, status; - sigset_t blocked, oldset; + sigset_t oldset; struct buffer_head *di_bh = NULL; - mlog_entry("(inode->i_ino = %lu)\n", inode->i_ino); + trace_ocfs2_delete_inode(inode->i_ino, + (unsigned long long)OCFS2_I(inode)->ip_blkno, + is_bad_inode(inode)); /* When we fail in read_inode() we mark inode as bad. The second test * catches the case when inode allocation fails before allocating * a block for inode. */ - if (is_bad_inode(inode) || !OCFS2_I(inode)->ip_blkno) { - mlog(0, "Skipping delete of bad inode\n"); + if (is_bad_inode(inode) || !OCFS2_I(inode)->ip_blkno) goto bail; - } if (!ocfs2_inode_is_valid_to_delete(inode)) { /* It's probably not necessary to truncate_inode_pages @@ -960,17 +991,13 @@ void ocfs2_delete_inode(struct inode *inode) goto bail; } + dquot_initialize(inode); + /* We want to block signals in delete_inode as the lock and * messaging paths may return us -ERESTARTSYS. Which would * cause us to exit early, resulting in inodes being orphaned * forever. */ - sigfillset(&blocked); - status = sigprocmask(SIG_BLOCK, &blocked, &oldset); - if (status < 0) { - mlog_errno(status); - ocfs2_cleanup_delete_inode(inode, 1); - goto bail; - } + ocfs2_block_signals(&oldset); /* * Synchronize us against ocfs2_get_dentry. We take this in @@ -1044,39 +1071,39 @@ bail_unlock_nfs_sync: ocfs2_nfs_sync_unlock(OCFS2_SB(inode->i_sb), 0); bail_unblock: - status = sigprocmask(SIG_SETMASK, &oldset, NULL); - if (status < 0) - mlog_errno(status); + ocfs2_unblock_signals(&oldset); bail: - clear_inode(inode); - mlog_exit_void(); + return; } -void ocfs2_clear_inode(struct inode *inode) +static void ocfs2_clear_inode(struct inode *inode) { int status; struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - mlog_entry_void(); - - if (!inode) - goto bail; - - mlog(0, "Clearing inode: %llu, nlink = %u\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, inode->i_nlink); + clear_inode(inode); + trace_ocfs2_clear_inode((unsigned long long)oi->ip_blkno, + inode->i_nlink); mlog_bug_on_msg(OCFS2_SB(inode->i_sb) == NULL, "Inode=%lu\n", inode->i_ino); + dquot_drop(inode); + /* To preven remote deletes we hold open lock before, now it * is time to unlock PR and EX open locks. */ ocfs2_open_unlock(inode); /* Do these before all the other work so that we don't bounce * the downconvert thread while waiting to destroy the locks. */ - ocfs2_mark_lockres_freeing(&oi->ip_rw_lockres); - ocfs2_mark_lockres_freeing(&oi->ip_inode_lockres); - ocfs2_mark_lockres_freeing(&oi->ip_open_lockres); + ocfs2_mark_lockres_freeing(osb, &oi->ip_rw_lockres); + ocfs2_mark_lockres_freeing(osb, &oi->ip_inode_lockres); + ocfs2_mark_lockres_freeing(osb, &oi->ip_open_lockres); + + ocfs2_resv_discard(&OCFS2_SB(inode->i_sb)->osb_la_resmap, + &oi->ip_la_data_resv); + ocfs2_resv_init_once(&oi->ip_la_data_resv); /* We very well may get a clear_inode before all an inodes * metadata has hit disk. Of course, we can't drop any cluster @@ -1101,13 +1128,14 @@ void ocfs2_clear_inode(struct inode *inode) ocfs2_lock_res_free(&oi->ip_inode_lockres); ocfs2_lock_res_free(&oi->ip_open_lockres); - ocfs2_metadata_cache_purge(inode); + ocfs2_metadata_cache_exit(INODE_CACHE(inode)); - mlog_bug_on_msg(oi->ip_metadata_cache.ci_num_cached, + mlog_bug_on_msg(INODE_CACHE(inode)->ci_num_cached, "Clear inode of %llu, inode has %u cache items\n", - (unsigned long long)oi->ip_blkno, oi->ip_metadata_cache.ci_num_cached); + (unsigned long long)oi->ip_blkno, + INODE_CACHE(inode)->ci_num_cached); - mlog_bug_on_msg(!(oi->ip_flags & OCFS2_INODE_CACHE_INLINE), + mlog_bug_on_msg(!(INODE_CACHE(inode)->ci_flags & OCFS2_CACHE_FL_INLINE), "Clear inode of %llu, inode has a bad flag\n", (unsigned long long)oi->ip_blkno); @@ -1134,9 +1162,7 @@ void ocfs2_clear_inode(struct inode *inode) (unsigned long long)oi->ip_blkno, oi->ip_open_count); /* Clear all other flags. */ - oi->ip_flags = OCFS2_INODE_CACHE_INLINE; - oi->ip_created_trans = 0; - oi->ip_last_trans = 0; + oi->ip_flags = 0; oi->ip_dir_start_lookup = 0; oi->ip_blkno = 0ULL; @@ -1147,29 +1173,36 @@ void ocfs2_clear_inode(struct inode *inode) */ jbd2_journal_release_jbd_inode(OCFS2_SB(inode->i_sb)->journal->j_journal, &oi->ip_jinode); +} -bail: - mlog_exit_void(); +void ocfs2_evict_inode(struct inode *inode) +{ + if (!inode->i_nlink || + (OCFS2_I(inode)->ip_flags & OCFS2_INODE_MAYBE_ORPHANED)) { + ocfs2_delete_inode(inode); + } else { + truncate_inode_pages_final(&inode->i_data); + } + ocfs2_clear_inode(inode); } /* Called under inode_lock, with no more references on the * struct inode, so it's safe here to check the flags field * and to manipulate i_nlink without any other locks. */ -void ocfs2_drop_inode(struct inode *inode) +int ocfs2_drop_inode(struct inode *inode) { struct ocfs2_inode_info *oi = OCFS2_I(inode); + int res; - mlog_entry_void(); - - mlog(0, "Drop inode %llu, nlink = %u, ip_flags = 0x%x\n", - (unsigned long long)oi->ip_blkno, inode->i_nlink, oi->ip_flags); + trace_ocfs2_drop_inode((unsigned long long)oi->ip_blkno, + inode->i_nlink, oi->ip_flags); if (oi->ip_flags & OCFS2_INODE_MAYBE_ORPHANED) - generic_delete_inode(inode); + res = 1; else - generic_drop_inode(inode); + res = generic_drop_inode(inode); - mlog_exit_void(); + return res; } /* @@ -1180,11 +1213,11 @@ int ocfs2_inode_revalidate(struct dentry *dentry) struct inode *inode = dentry->d_inode; int status = 0; - mlog_entry("(inode = 0x%p, ino = %llu)\n", inode, - inode ? (unsigned long long)OCFS2_I(inode)->ip_blkno : 0ULL); + trace_ocfs2_inode_revalidate(inode, + inode ? (unsigned long long)OCFS2_I(inode)->ip_blkno : 0ULL, + inode ? (unsigned long long)OCFS2_I(inode)->ip_flags : 0); if (!inode) { - mlog(0, "eep, no inode!\n"); status = -ENOENT; goto bail; } @@ -1192,7 +1225,6 @@ int ocfs2_inode_revalidate(struct dentry *dentry) spin_lock(&OCFS2_I(inode)->ip_lock); if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_DELETED) { spin_unlock(&OCFS2_I(inode)->ip_lock); - mlog(0, "inode deleted!\n"); status = -ENOENT; goto bail; } @@ -1208,8 +1240,6 @@ int ocfs2_inode_revalidate(struct dentry *dentry) } ocfs2_inode_unlock(inode, 0); bail: - mlog_exit(status); - return status; } @@ -1225,10 +1255,9 @@ int ocfs2_mark_inode_dirty(handle_t *handle, int status; struct ocfs2_dinode *fe = (struct ocfs2_dinode *) bh->b_data; - mlog_entry("(inode %llu)\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno); + trace_ocfs2_mark_inode_dirty((unsigned long long)OCFS2_I(inode)->ip_blkno); - status = ocfs2_journal_access_di(handle, inode, bh, + status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); @@ -1244,8 +1273,8 @@ int ocfs2_mark_inode_dirty(handle_t *handle, fe->i_size = cpu_to_le64(i_size_read(inode)); ocfs2_set_links_count(fe, inode->i_nlink); - fe->i_uid = cpu_to_le32(inode->i_uid); - fe->i_gid = cpu_to_le32(inode->i_gid); + fe->i_uid = cpu_to_le32(i_uid_read(inode)); + fe->i_gid = cpu_to_le32(i_gid_read(inode)); fe->i_mode = cpu_to_le16(inode->i_mode); fe->i_atime = cpu_to_le64(inode->i_atime.tv_sec); fe->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec); @@ -1254,14 +1283,9 @@ int ocfs2_mark_inode_dirty(handle_t *handle, fe->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec); fe->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec); - status = ocfs2_journal_dirty(handle, bh); - if (status < 0) - mlog_errno(status); - - status = 0; + ocfs2_journal_dirty(handle, bh); + ocfs2_update_inode_fsync_trans(handle, inode, 1); leave: - - mlog_exit(status); return status; } @@ -1280,9 +1304,9 @@ void ocfs2_refresh_inode(struct inode *inode, OCFS2_I(inode)->ip_dyn_features = le16_to_cpu(fe->i_dyn_features); ocfs2_set_inode_flags(inode); i_size_write(inode, le64_to_cpu(fe->i_size)); - inode->i_nlink = ocfs2_read_links_count(fe); - inode->i_uid = le32_to_cpu(fe->i_uid); - inode->i_gid = le32_to_cpu(fe->i_gid); + set_nlink(inode, ocfs2_read_links_count(fe)); + i_uid_write(inode, le32_to_cpu(fe->i_uid)); + i_gid_write(inode, le32_to_cpu(fe->i_gid)); inode->i_mode = le16_to_cpu(fe->i_mode); if (S_ISLNK(inode->i_mode) && le32_to_cpu(fe->i_clusters) == 0) inode->i_blocks = 0; @@ -1304,8 +1328,7 @@ int ocfs2_validate_inode_block(struct super_block *sb, int rc; struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; - mlog(0, "Validating dinode %llu\n", - (unsigned long long)bh->b_blocknr); + trace_ocfs2_validate_inode_block((unsigned long long)bh->b_blocknr); BUG_ON(!buffer_uptodate(bh)); @@ -1369,8 +1392,8 @@ int ocfs2_read_inode_block_full(struct inode *inode, struct buffer_head **bh, int rc; struct buffer_head *tmp = *bh; - rc = ocfs2_read_blocks(inode, OCFS2_I(inode)->ip_blkno, 1, &tmp, - flags, ocfs2_validate_inode_block); + rc = ocfs2_read_blocks(INODE_CACHE(inode), OCFS2_I(inode)->ip_blkno, + 1, &tmp, flags, ocfs2_validate_inode_block); /* If ocfs2_read_blocks() got us a new bh, pass it up. */ if (!rc && !*bh) @@ -1383,3 +1406,56 @@ int ocfs2_read_inode_block(struct inode *inode, struct buffer_head **bh) { return ocfs2_read_inode_block_full(inode, bh, 0); } + + +static u64 ocfs2_inode_cache_owner(struct ocfs2_caching_info *ci) +{ + struct ocfs2_inode_info *oi = cache_info_to_inode(ci); + + return oi->ip_blkno; +} + +static struct super_block *ocfs2_inode_cache_get_super(struct ocfs2_caching_info *ci) +{ + struct ocfs2_inode_info *oi = cache_info_to_inode(ci); + + return oi->vfs_inode.i_sb; +} + +static void ocfs2_inode_cache_lock(struct ocfs2_caching_info *ci) +{ + struct ocfs2_inode_info *oi = cache_info_to_inode(ci); + + spin_lock(&oi->ip_lock); +} + +static void ocfs2_inode_cache_unlock(struct ocfs2_caching_info *ci) +{ + struct ocfs2_inode_info *oi = cache_info_to_inode(ci); + + spin_unlock(&oi->ip_lock); +} + +static void ocfs2_inode_cache_io_lock(struct ocfs2_caching_info *ci) +{ + struct ocfs2_inode_info *oi = cache_info_to_inode(ci); + + mutex_lock(&oi->ip_io_mutex); +} + +static void ocfs2_inode_cache_io_unlock(struct ocfs2_caching_info *ci) +{ + struct ocfs2_inode_info *oi = cache_info_to_inode(ci); + + mutex_unlock(&oi->ip_io_mutex); +} + +const struct ocfs2_caching_operations ocfs2_inode_caching_ops = { + .co_owner = ocfs2_inode_cache_owner, + .co_get_super = ocfs2_inode_cache_get_super, + .co_cache_lock = ocfs2_inode_cache_lock, + .co_cache_unlock = ocfs2_inode_cache_unlock, + .co_io_lock = ocfs2_inode_cache_io_lock, + .co_io_unlock = ocfs2_inode_cache_io_unlock, +}; + diff --git a/fs/ocfs2/inode.h b/fs/ocfs2/inode.h index ea71525aad4..a6c991c0fc9 100644 --- a/fs/ocfs2/inode.h +++ b/fs/ocfs2/inode.h @@ -43,39 +43,43 @@ struct ocfs2_inode_info /* protects extended attribute changes on this inode */ struct rw_semaphore ip_xattr_sem; + /* Number of outstanding AIO's which are not page aligned */ + struct mutex ip_unaligned_aio; + /* These fields are protected by ip_lock */ spinlock_t ip_lock; u32 ip_open_count; - u32 ip_clusters; struct list_head ip_io_markers; + u32 ip_clusters; + u16 ip_dyn_features; struct mutex ip_io_mutex; - u32 ip_flags; /* see below */ u32 ip_attr; /* inode attributes */ - u16 ip_dyn_features; /* protected by recovery_lock. */ struct inode *ip_next_orphan; - u32 ip_dir_start_lookup; - - /* next two are protected by trans_inc_lock */ - /* which transaction were we created on? Zero if none. */ - unsigned long ip_created_trans; - /* last transaction we were a part of. */ - unsigned long ip_last_trans; - struct ocfs2_caching_info ip_metadata_cache; - struct ocfs2_extent_map ip_extent_map; - struct inode vfs_inode; struct jbd2_inode ip_jinode; + u32 ip_dir_start_lookup; + /* Only valid if the inode is the dir. */ u32 ip_last_used_slot; u64 ip_last_used_group; + u32 ip_dir_lock_gen; + + struct ocfs2_alloc_reservation ip_la_data_resv; + + /* + * Transactions that contain inode's metadata needed to complete + * fsync and fdatasync, respectively. + */ + tid_t i_sync_tid; + tid_t i_datasync_tid; }; /* @@ -87,8 +91,6 @@ struct ocfs2_inode_info #define OCFS2_INODE_BITMAP 0x00000004 /* This inode has been wiped from disk */ #define OCFS2_INODE_DELETED 0x00000008 -/* Another node is deleting, so our delete is a nop */ -#define OCFS2_INODE_SKIP_DELETE 0x00000010 /* Has the inode been orphaned on another node? * * This hints to ocfs2_drop_inode that it should clear i_nlink before @@ -103,11 +105,11 @@ struct ocfs2_inode_info * rely on ocfs2_delete_inode to sort things out under the proper * cluster locks. */ -#define OCFS2_INODE_MAYBE_ORPHANED 0x00000020 +#define OCFS2_INODE_MAYBE_ORPHANED 0x00000010 /* Does someone have the file open O_DIRECT */ -#define OCFS2_INODE_OPEN_DIRECT 0x00000040 -/* Indicates that the metadata cache should be used as an array. */ -#define OCFS2_INODE_CACHE_INLINE 0x00000080 +#define OCFS2_INODE_OPEN_DIRECT 0x00000020 +/* Tell the inode wipe code it's not in orphan dir */ +#define OCFS2_INODE_SKIP_ORPHAN_DIR 0x00000040 static inline struct ocfs2_inode_info *OCFS2_I(struct inode *inode) { @@ -120,10 +122,15 @@ static inline struct ocfs2_inode_info *OCFS2_I(struct inode *inode) extern struct kmem_cache *ocfs2_inode_cache; extern const struct address_space_operations ocfs2_aops; +extern const struct ocfs2_caching_operations ocfs2_inode_caching_ops; -void ocfs2_clear_inode(struct inode *inode); -void ocfs2_delete_inode(struct inode *inode); -void ocfs2_drop_inode(struct inode *inode); +static inline struct ocfs2_caching_info *INODE_CACHE(struct inode *inode) +{ + return &OCFS2_I(inode)->ip_metadata_cache; +} + +void ocfs2_evict_inode(struct inode *inode); +int ocfs2_drop_inode(struct inode *inode); /* Flags for ocfs2_iget() */ #define OCFS2_FI_FLAG_SYSFILE 0x1 @@ -145,8 +152,6 @@ void ocfs2_refresh_inode(struct inode *inode, int ocfs2_mark_inode_dirty(handle_t *handle, struct inode *inode, struct buffer_head *bh); -int ocfs2_aio_read(struct file *file, struct kiocb *req, struct iocb *iocb); -int ocfs2_aio_write(struct file *file, struct kiocb *req, struct iocb *iocb); struct buffer_head *ocfs2_bread(struct inode *inode, int block, int *err, int reada); @@ -172,4 +177,10 @@ int ocfs2_read_inode_block(struct inode *inode, struct buffer_head **bh); /* The same, but can be passed OCFS2_BH_* flags */ int ocfs2_read_inode_block_full(struct inode *inode, struct buffer_head **bh, int flags); + +static inline struct ocfs2_inode_info *cache_info_to_inode(struct ocfs2_caching_info *ci) +{ + return container_of(ci, struct ocfs2_inode_info, ip_metadata_cache); +} + #endif /* OCFS2_INODE_H */ diff --git a/fs/ocfs2/ioctl.c b/fs/ocfs2/ioctl.c index 9fcd36dcc9a..6f66b3751ac 100644 --- a/fs/ocfs2/ioctl.c +++ b/fs/ocfs2/ioctl.c @@ -7,9 +7,9 @@ #include <linux/fs.h> #include <linux/mount.h> -#include <linux/smp_lock.h> +#include <linux/blkdev.h> +#include <linux/compat.h> -#define MLOG_MASK_PREFIX ML_INODE #include <cluster/masklog.h> #include "ocfs2.h" @@ -22,8 +22,44 @@ #include "ocfs2_fs.h" #include "ioctl.h" #include "resize.h" +#include "refcounttree.h" +#include "sysfile.h" +#include "dir.h" +#include "buffer_head_io.h" +#include "suballoc.h" +#include "move_extents.h" -#include <linux/ext2_fs.h> +#define o2info_from_user(a, b) \ + copy_from_user(&(a), (b), sizeof(a)) +#define o2info_to_user(a, b) \ + copy_to_user((typeof(a) __user *)b, &(a), sizeof(a)) + +/* + * This call is void because we are already reporting an error that may + * be -EFAULT. The error will be returned from the ioctl(2) call. It's + * just a best-effort to tell userspace that this request caused the error. + */ +static inline void o2info_set_request_error(struct ocfs2_info_request *kreq, + struct ocfs2_info_request __user *req) +{ + kreq->ir_flags |= OCFS2_INFO_FL_ERROR; + (void)put_user(kreq->ir_flags, (__u32 __user *)&(req->ir_flags)); +} + +static inline void o2info_set_request_filled(struct ocfs2_info_request *req) +{ + req->ir_flags |= OCFS2_INFO_FL_FILLED; +} + +static inline void o2info_clear_request_filled(struct ocfs2_info_request *req) +{ + req->ir_flags &= ~OCFS2_INFO_FL_FILLED; +} + +static inline int o2info_coherent(struct ocfs2_info_request *req) +{ + return (!(req->ir_flags & OCFS2_INFO_FL_NON_COHERENT)); +} static int ocfs2_get_inode_attr(struct inode *inode, unsigned *flags) { @@ -38,7 +74,6 @@ static int ocfs2_get_inode_attr(struct inode *inode, unsigned *flags) *flags = OCFS2_I(inode)->ip_attr; ocfs2_inode_unlock(inode, 0); - mlog_exit(status); return status; } @@ -61,19 +96,12 @@ static int ocfs2_set_inode_attr(struct inode *inode, unsigned flags, } status = -EACCES; - if (!is_owner_or_cap(inode)) + if (!inode_owner_or_capable(inode)) goto bail_unlock; if (!S_ISDIR(inode->i_mode)) flags &= ~OCFS2_DIRSYNC_FL; - handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS); - if (IS_ERR(handle)) { - status = PTR_ERR(handle); - mlog_errno(status); - goto bail_unlock; - } - oldflags = ocfs2_inode->ip_attr; flags = flags & mask; flags |= oldflags & ~mask; @@ -89,6 +117,13 @@ static int ocfs2_set_inode_attr(struct inode *inode, unsigned flags, goto bail_unlock; } + handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS); + if (IS_ERR(handle)) { + status = PTR_ERR(handle); + mlog_errno(status); + goto bail_unlock; + } + ocfs2_inode->ip_attr = flags; ocfs2_set_inode_flags(inode); @@ -97,6 +132,7 @@ static int ocfs2_set_inode_attr(struct inode *inode, unsigned flags, mlog_errno(status); ocfs2_commit_trans(osb, handle); + bail_unlock: ocfs2_inode_unlock(inode, 1); bail: @@ -104,18 +140,764 @@ bail: brelse(bh); - mlog_exit(status); + return status; +} + +static int ocfs2_info_handle_blocksize(struct inode *inode, + struct ocfs2_info_request __user *req) +{ + int status = -EFAULT; + struct ocfs2_info_blocksize oib; + + if (o2info_from_user(oib, req)) + goto bail; + + oib.ib_blocksize = inode->i_sb->s_blocksize; + + o2info_set_request_filled(&oib.ib_req); + + if (o2info_to_user(oib, req)) + goto bail; + + status = 0; +bail: + if (status) + o2info_set_request_error(&oib.ib_req, req); + + return status; +} + +static int ocfs2_info_handle_clustersize(struct inode *inode, + struct ocfs2_info_request __user *req) +{ + int status = -EFAULT; + struct ocfs2_info_clustersize oic; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + + if (o2info_from_user(oic, req)) + goto bail; + + oic.ic_clustersize = osb->s_clustersize; + + o2info_set_request_filled(&oic.ic_req); + + if (o2info_to_user(oic, req)) + goto bail; + + status = 0; +bail: + if (status) + o2info_set_request_error(&oic.ic_req, req); + + return status; +} + +static int ocfs2_info_handle_maxslots(struct inode *inode, + struct ocfs2_info_request __user *req) +{ + int status = -EFAULT; + struct ocfs2_info_maxslots oim; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + + if (o2info_from_user(oim, req)) + goto bail; + + oim.im_max_slots = osb->max_slots; + + o2info_set_request_filled(&oim.im_req); + + if (o2info_to_user(oim, req)) + goto bail; + + status = 0; +bail: + if (status) + o2info_set_request_error(&oim.im_req, req); + + return status; +} + +static int ocfs2_info_handle_label(struct inode *inode, + struct ocfs2_info_request __user *req) +{ + int status = -EFAULT; + struct ocfs2_info_label oil; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + + if (o2info_from_user(oil, req)) + goto bail; + + memcpy(oil.il_label, osb->vol_label, OCFS2_MAX_VOL_LABEL_LEN); + + o2info_set_request_filled(&oil.il_req); + + if (o2info_to_user(oil, req)) + goto bail; + + status = 0; +bail: + if (status) + o2info_set_request_error(&oil.il_req, req); + + return status; +} + +static int ocfs2_info_handle_uuid(struct inode *inode, + struct ocfs2_info_request __user *req) +{ + int status = -EFAULT; + struct ocfs2_info_uuid oiu; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + + if (o2info_from_user(oiu, req)) + goto bail; + + memcpy(oiu.iu_uuid_str, osb->uuid_str, OCFS2_TEXT_UUID_LEN + 1); + + o2info_set_request_filled(&oiu.iu_req); + + if (o2info_to_user(oiu, req)) + goto bail; + + status = 0; +bail: + if (status) + o2info_set_request_error(&oiu.iu_req, req); + + return status; +} + +static int ocfs2_info_handle_fs_features(struct inode *inode, + struct ocfs2_info_request __user *req) +{ + int status = -EFAULT; + struct ocfs2_info_fs_features oif; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + + if (o2info_from_user(oif, req)) + goto bail; + + oif.if_compat_features = osb->s_feature_compat; + oif.if_incompat_features = osb->s_feature_incompat; + oif.if_ro_compat_features = osb->s_feature_ro_compat; + + o2info_set_request_filled(&oif.if_req); + + if (o2info_to_user(oif, req)) + goto bail; + + status = 0; +bail: + if (status) + o2info_set_request_error(&oif.if_req, req); + + return status; +} + +static int ocfs2_info_handle_journal_size(struct inode *inode, + struct ocfs2_info_request __user *req) +{ + int status = -EFAULT; + struct ocfs2_info_journal_size oij; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + + if (o2info_from_user(oij, req)) + goto bail; + + oij.ij_journal_size = i_size_read(osb->journal->j_inode); + + o2info_set_request_filled(&oij.ij_req); + + if (o2info_to_user(oij, req)) + goto bail; + + status = 0; +bail: + if (status) + o2info_set_request_error(&oij.ij_req, req); + + return status; +} + +static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb, + struct inode *inode_alloc, u64 blkno, + struct ocfs2_info_freeinode *fi, + u32 slot) +{ + int status = 0, unlock = 0; + + struct buffer_head *bh = NULL; + struct ocfs2_dinode *dinode_alloc = NULL; + + if (inode_alloc) + mutex_lock(&inode_alloc->i_mutex); + + if (o2info_coherent(&fi->ifi_req)) { + status = ocfs2_inode_lock(inode_alloc, &bh, 0); + if (status < 0) { + mlog_errno(status); + goto bail; + } + unlock = 1; + } else { + status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh); + if (status < 0) { + mlog_errno(status); + goto bail; + } + } + + dinode_alloc = (struct ocfs2_dinode *)bh->b_data; + + fi->ifi_stat[slot].lfi_total = + le32_to_cpu(dinode_alloc->id1.bitmap1.i_total); + fi->ifi_stat[slot].lfi_free = + le32_to_cpu(dinode_alloc->id1.bitmap1.i_total) - + le32_to_cpu(dinode_alloc->id1.bitmap1.i_used); + +bail: + if (unlock) + ocfs2_inode_unlock(inode_alloc, 0); + + if (inode_alloc) + mutex_unlock(&inode_alloc->i_mutex); + + brelse(bh); + + return status; +} + +static int ocfs2_info_handle_freeinode(struct inode *inode, + struct ocfs2_info_request __user *req) +{ + u32 i; + u64 blkno = -1; + char namebuf[40]; + int status = -EFAULT, type = INODE_ALLOC_SYSTEM_INODE; + struct ocfs2_info_freeinode *oifi = NULL; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct inode *inode_alloc = NULL; + + oifi = kzalloc(sizeof(struct ocfs2_info_freeinode), GFP_KERNEL); + if (!oifi) { + status = -ENOMEM; + mlog_errno(status); + goto out_err; + } + + if (o2info_from_user(*oifi, req)) + goto bail; + + oifi->ifi_slotnum = osb->max_slots; + + for (i = 0; i < oifi->ifi_slotnum; i++) { + if (o2info_coherent(&oifi->ifi_req)) { + inode_alloc = ocfs2_get_system_file_inode(osb, type, i); + if (!inode_alloc) { + mlog(ML_ERROR, "unable to get alloc inode in " + "slot %u\n", i); + status = -EIO; + goto bail; + } + } else { + ocfs2_sprintf_system_inode_name(namebuf, + sizeof(namebuf), + type, i); + status = ocfs2_lookup_ino_from_name(osb->sys_root_inode, + namebuf, + strlen(namebuf), + &blkno); + if (status < 0) { + status = -ENOENT; + goto bail; + } + } + + status = ocfs2_info_scan_inode_alloc(osb, inode_alloc, blkno, oifi, i); + + iput(inode_alloc); + inode_alloc = NULL; + + if (status < 0) + goto bail; + } + + o2info_set_request_filled(&oifi->ifi_req); + + if (o2info_to_user(*oifi, req)) + goto bail; + + status = 0; +bail: + if (status) + o2info_set_request_error(&oifi->ifi_req, req); + + kfree(oifi); +out_err: + return status; +} + +static void o2ffg_update_histogram(struct ocfs2_info_free_chunk_list *hist, + unsigned int chunksize) +{ + int index; + + index = __ilog2_u32(chunksize); + if (index >= OCFS2_INFO_MAX_HIST) + index = OCFS2_INFO_MAX_HIST - 1; + + hist->fc_chunks[index]++; + hist->fc_clusters[index] += chunksize; +} + +static void o2ffg_update_stats(struct ocfs2_info_freefrag_stats *stats, + unsigned int chunksize) +{ + if (chunksize > stats->ffs_max) + stats->ffs_max = chunksize; + + if (chunksize < stats->ffs_min) + stats->ffs_min = chunksize; + + stats->ffs_avg += chunksize; + stats->ffs_free_chunks_real++; +} + +static void ocfs2_info_update_ffg(struct ocfs2_info_freefrag *ffg, + unsigned int chunksize) +{ + o2ffg_update_histogram(&(ffg->iff_ffs.ffs_fc_hist), chunksize); + o2ffg_update_stats(&(ffg->iff_ffs), chunksize); +} + +static int ocfs2_info_freefrag_scan_chain(struct ocfs2_super *osb, + struct inode *gb_inode, + struct ocfs2_dinode *gb_dinode, + struct ocfs2_chain_rec *rec, + struct ocfs2_info_freefrag *ffg, + u32 chunks_in_group) +{ + int status = 0, used; + u64 blkno; + + struct buffer_head *bh = NULL; + struct ocfs2_group_desc *bg = NULL; + + unsigned int max_bits, num_clusters; + unsigned int offset = 0, cluster, chunk; + unsigned int chunk_free, last_chunksize = 0; + + if (!le32_to_cpu(rec->c_free)) + goto bail; + + do { + if (!bg) + blkno = le64_to_cpu(rec->c_blkno); + else + blkno = le64_to_cpu(bg->bg_next_group); + + if (bh) { + brelse(bh); + bh = NULL; + } + + if (o2info_coherent(&ffg->iff_req)) + status = ocfs2_read_group_descriptor(gb_inode, + gb_dinode, + blkno, &bh); + else + status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh); + + if (status < 0) { + mlog(ML_ERROR, "Can't read the group descriptor # " + "%llu from device.", (unsigned long long)blkno); + status = -EIO; + goto bail; + } + + bg = (struct ocfs2_group_desc *)bh->b_data; + + if (!le16_to_cpu(bg->bg_free_bits_count)) + continue; + + max_bits = le16_to_cpu(bg->bg_bits); + offset = 0; + + for (chunk = 0; chunk < chunks_in_group; chunk++) { + /* + * last chunk may be not an entire one. + */ + if ((offset + ffg->iff_chunksize) > max_bits) + num_clusters = max_bits - offset; + else + num_clusters = ffg->iff_chunksize; + + chunk_free = 0; + for (cluster = 0; cluster < num_clusters; cluster++) { + used = ocfs2_test_bit(offset, + (unsigned long *)bg->bg_bitmap); + /* + * - chunk_free counts free clusters in #N chunk. + * - last_chunksize records the size(in) clusters + * for the last real free chunk being counted. + */ + if (!used) { + last_chunksize++; + chunk_free++; + } + + if (used && last_chunksize) { + ocfs2_info_update_ffg(ffg, + last_chunksize); + last_chunksize = 0; + } + + offset++; + } + + if (chunk_free == ffg->iff_chunksize) + ffg->iff_ffs.ffs_free_chunks++; + } + + /* + * need to update the info for last free chunk. + */ + if (last_chunksize) + ocfs2_info_update_ffg(ffg, last_chunksize); + + } while (le64_to_cpu(bg->bg_next_group)); + +bail: + brelse(bh); + + return status; +} + +static int ocfs2_info_freefrag_scan_bitmap(struct ocfs2_super *osb, + struct inode *gb_inode, u64 blkno, + struct ocfs2_info_freefrag *ffg) +{ + u32 chunks_in_group; + int status = 0, unlock = 0, i; + + struct buffer_head *bh = NULL; + struct ocfs2_chain_list *cl = NULL; + struct ocfs2_chain_rec *rec = NULL; + struct ocfs2_dinode *gb_dinode = NULL; + + if (gb_inode) + mutex_lock(&gb_inode->i_mutex); + + if (o2info_coherent(&ffg->iff_req)) { + status = ocfs2_inode_lock(gb_inode, &bh, 0); + if (status < 0) { + mlog_errno(status); + goto bail; + } + unlock = 1; + } else { + status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh); + if (status < 0) { + mlog_errno(status); + goto bail; + } + } + + gb_dinode = (struct ocfs2_dinode *)bh->b_data; + cl = &(gb_dinode->id2.i_chain); + + /* + * Chunksize(in) clusters from userspace should be + * less than clusters in a group. + */ + if (ffg->iff_chunksize > le16_to_cpu(cl->cl_cpg)) { + status = -EINVAL; + goto bail; + } + + memset(&ffg->iff_ffs, 0, sizeof(struct ocfs2_info_freefrag_stats)); + + ffg->iff_ffs.ffs_min = ~0U; + ffg->iff_ffs.ffs_clusters = + le32_to_cpu(gb_dinode->id1.bitmap1.i_total); + ffg->iff_ffs.ffs_free_clusters = ffg->iff_ffs.ffs_clusters - + le32_to_cpu(gb_dinode->id1.bitmap1.i_used); + + chunks_in_group = le16_to_cpu(cl->cl_cpg) / ffg->iff_chunksize + 1; + + for (i = 0; i < le16_to_cpu(cl->cl_next_free_rec); i++) { + rec = &(cl->cl_recs[i]); + status = ocfs2_info_freefrag_scan_chain(osb, gb_inode, + gb_dinode, + rec, ffg, + chunks_in_group); + if (status) + goto bail; + } + + if (ffg->iff_ffs.ffs_free_chunks_real) + ffg->iff_ffs.ffs_avg = (ffg->iff_ffs.ffs_avg / + ffg->iff_ffs.ffs_free_chunks_real); +bail: + if (unlock) + ocfs2_inode_unlock(gb_inode, 0); + + if (gb_inode) + mutex_unlock(&gb_inode->i_mutex); + + if (gb_inode) + iput(gb_inode); + + brelse(bh); + + return status; +} + +static int ocfs2_info_handle_freefrag(struct inode *inode, + struct ocfs2_info_request __user *req) +{ + u64 blkno = -1; + char namebuf[40]; + int status = -EFAULT, type = GLOBAL_BITMAP_SYSTEM_INODE; + + struct ocfs2_info_freefrag *oiff; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct inode *gb_inode = NULL; + + oiff = kzalloc(sizeof(struct ocfs2_info_freefrag), GFP_KERNEL); + if (!oiff) { + status = -ENOMEM; + mlog_errno(status); + goto out_err; + } + + if (o2info_from_user(*oiff, req)) + goto bail; + /* + * chunksize from userspace should be power of 2. + */ + if ((oiff->iff_chunksize & (oiff->iff_chunksize - 1)) || + (!oiff->iff_chunksize)) { + status = -EINVAL; + goto bail; + } + + if (o2info_coherent(&oiff->iff_req)) { + gb_inode = ocfs2_get_system_file_inode(osb, type, + OCFS2_INVALID_SLOT); + if (!gb_inode) { + mlog(ML_ERROR, "unable to get global_bitmap inode\n"); + status = -EIO; + goto bail; + } + } else { + ocfs2_sprintf_system_inode_name(namebuf, sizeof(namebuf), type, + OCFS2_INVALID_SLOT); + status = ocfs2_lookup_ino_from_name(osb->sys_root_inode, + namebuf, + strlen(namebuf), + &blkno); + if (status < 0) { + status = -ENOENT; + goto bail; + } + } + + status = ocfs2_info_freefrag_scan_bitmap(osb, gb_inode, blkno, oiff); + if (status < 0) + goto bail; + + o2info_set_request_filled(&oiff->iff_req); + + if (o2info_to_user(*oiff, req)) { + status = -EFAULT; + goto bail; + } + + status = 0; +bail: + if (status) + o2info_set_request_error(&oiff->iff_req, req); + + kfree(oiff); +out_err: + return status; +} + +static int ocfs2_info_handle_unknown(struct inode *inode, + struct ocfs2_info_request __user *req) +{ + int status = -EFAULT; + struct ocfs2_info_request oir; + + if (o2info_from_user(oir, req)) + goto bail; + + o2info_clear_request_filled(&oir); + + if (o2info_to_user(oir, req)) + goto bail; + + status = 0; +bail: + if (status) + o2info_set_request_error(&oir, req); + + return status; +} + +/* + * Validate and distinguish OCFS2_IOC_INFO requests. + * + * - validate the magic number. + * - distinguish different requests. + * - validate size of different requests. + */ +static int ocfs2_info_handle_request(struct inode *inode, + struct ocfs2_info_request __user *req) +{ + int status = -EFAULT; + struct ocfs2_info_request oir; + + if (o2info_from_user(oir, req)) + goto bail; + + status = -EINVAL; + if (oir.ir_magic != OCFS2_INFO_MAGIC) + goto bail; + + switch (oir.ir_code) { + case OCFS2_INFO_BLOCKSIZE: + if (oir.ir_size == sizeof(struct ocfs2_info_blocksize)) + status = ocfs2_info_handle_blocksize(inode, req); + break; + case OCFS2_INFO_CLUSTERSIZE: + if (oir.ir_size == sizeof(struct ocfs2_info_clustersize)) + status = ocfs2_info_handle_clustersize(inode, req); + break; + case OCFS2_INFO_MAXSLOTS: + if (oir.ir_size == sizeof(struct ocfs2_info_maxslots)) + status = ocfs2_info_handle_maxslots(inode, req); + break; + case OCFS2_INFO_LABEL: + if (oir.ir_size == sizeof(struct ocfs2_info_label)) + status = ocfs2_info_handle_label(inode, req); + break; + case OCFS2_INFO_UUID: + if (oir.ir_size == sizeof(struct ocfs2_info_uuid)) + status = ocfs2_info_handle_uuid(inode, req); + break; + case OCFS2_INFO_FS_FEATURES: + if (oir.ir_size == sizeof(struct ocfs2_info_fs_features)) + status = ocfs2_info_handle_fs_features(inode, req); + break; + case OCFS2_INFO_JOURNAL_SIZE: + if (oir.ir_size == sizeof(struct ocfs2_info_journal_size)) + status = ocfs2_info_handle_journal_size(inode, req); + break; + case OCFS2_INFO_FREEINODE: + if (oir.ir_size == sizeof(struct ocfs2_info_freeinode)) + status = ocfs2_info_handle_freeinode(inode, req); + break; + case OCFS2_INFO_FREEFRAG: + if (oir.ir_size == sizeof(struct ocfs2_info_freefrag)) + status = ocfs2_info_handle_freefrag(inode, req); + break; + default: + status = ocfs2_info_handle_unknown(inode, req); + break; + } + +bail: + return status; +} + +static int ocfs2_get_request_ptr(struct ocfs2_info *info, int idx, + u64 *req_addr, int compat_flag) +{ + int status = -EFAULT; + u64 __user *bp = NULL; + + if (compat_flag) { +#ifdef CONFIG_COMPAT + /* + * pointer bp stores the base address of a pointers array, + * which collects all addresses of separate request. + */ + bp = (u64 __user *)(unsigned long)compat_ptr(info->oi_requests); +#else + BUG(); +#endif + } else + bp = (u64 __user *)(unsigned long)(info->oi_requests); + + if (o2info_from_user(*req_addr, bp + idx)) + goto bail; + + status = 0; +bail: + return status; +} + +/* + * OCFS2_IOC_INFO handles an array of requests passed from userspace. + * + * ocfs2_info_handle() recevies a large info aggregation, grab and + * validate the request count from header, then break it into small + * pieces, later specific handlers can handle them one by one. + * + * Idea here is to make each separate request small enough to ensure + * a better backward&forward compatibility, since a small piece of + * request will be less likely to be broken if disk layout get changed. + */ +static int ocfs2_info_handle(struct inode *inode, struct ocfs2_info *info, + int compat_flag) +{ + int i, status = 0; + u64 req_addr; + struct ocfs2_info_request __user *reqp; + + if ((info->oi_count > OCFS2_INFO_MAX_REQUEST) || + (!info->oi_requests)) { + status = -EINVAL; + goto bail; + } + + for (i = 0; i < info->oi_count; i++) { + + status = ocfs2_get_request_ptr(info, i, &req_addr, compat_flag); + if (status) + break; + + reqp = (struct ocfs2_info_request __user *)(unsigned long)req_addr; + if (!reqp) { + status = -EINVAL; + goto bail; + } + + status = ocfs2_info_handle_request(inode, reqp); + if (status) + break; + } + +bail: return status; } long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) { - struct inode *inode = filp->f_path.dentry->d_inode; + struct inode *inode = file_inode(filp); unsigned int flags; int new_clusters; int status; struct ocfs2_space_resv sr; struct ocfs2_new_group_input input; + struct reflink_arguments args; + const char __user *old_path; + const char __user *new_path; + bool preserve; + struct ocfs2_info info; + void __user *argp = (void __user *)arg; switch (cmd) { case OCFS2_IOC_GETFLAGS: @@ -129,12 +911,12 @@ long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) if (get_user(flags, (int __user *) arg)) return -EFAULT; - status = mnt_want_write(filp->f_path.mnt); + status = mnt_want_write_file(filp); if (status) return status; status = ocfs2_set_inode_attr(inode, flags, OCFS2_FL_MODIFIABLE); - mnt_drop_write(filp->f_path.mnt); + mnt_drop_write_file(filp); return status; case OCFS2_IOC_RESVSP: case OCFS2_IOC_RESVSP64: @@ -151,7 +933,12 @@ long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) if (get_user(new_clusters, (int __user *)arg)) return -EFAULT; - return ocfs2_group_extend(inode, new_clusters); + status = mnt_want_write_file(filp); + if (status) + return status; + status = ocfs2_group_extend(inode, new_clusters); + mnt_drop_write_file(filp); + return status; case OCFS2_IOC_GROUP_ADD: case OCFS2_IOC_GROUP_ADD64: if (!capable(CAP_SYS_RESOURCE)) @@ -160,7 +947,54 @@ long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) if (copy_from_user(&input, (int __user *) arg, sizeof(input))) return -EFAULT; - return ocfs2_group_add(inode, &input); + status = mnt_want_write_file(filp); + if (status) + return status; + status = ocfs2_group_add(inode, &input); + mnt_drop_write_file(filp); + return status; + case OCFS2_IOC_REFLINK: + if (copy_from_user(&args, argp, sizeof(args))) + return -EFAULT; + old_path = (const char __user *)(unsigned long)args.old_path; + new_path = (const char __user *)(unsigned long)args.new_path; + preserve = (args.preserve != 0); + + return ocfs2_reflink_ioctl(inode, old_path, new_path, preserve); + case OCFS2_IOC_INFO: + if (copy_from_user(&info, argp, sizeof(struct ocfs2_info))) + return -EFAULT; + + return ocfs2_info_handle(inode, &info, 0); + case FITRIM: + { + struct super_block *sb = inode->i_sb; + struct request_queue *q = bdev_get_queue(sb->s_bdev); + struct fstrim_range range; + int ret = 0; + + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + + if (!blk_queue_discard(q)) + return -EOPNOTSUPP; + + if (copy_from_user(&range, argp, sizeof(range))) + return -EFAULT; + + range.minlen = max_t(u64, q->limits.discard_granularity, + range.minlen); + ret = ocfs2_trim_fs(sb, &range); + if (ret < 0) + return ret; + + if (copy_to_user(argp, &range, sizeof(range))) + return -EFAULT; + + return 0; + } + case OCFS2_IOC_MOVE_EXT: + return ocfs2_ioctl_move_extents(filp, argp); default: return -ENOTTY; } @@ -169,6 +1003,12 @@ long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) #ifdef CONFIG_COMPAT long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg) { + bool preserve; + struct reflink_arguments args; + struct inode *inode = file_inode(file); + struct ocfs2_info info; + void __user *argp = (void __user *)arg; + switch (cmd) { case OCFS2_IOC32_GETFLAGS: cmd = OCFS2_IOC_GETFLAGS; @@ -183,6 +1023,21 @@ long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg) case OCFS2_IOC_GROUP_EXTEND: case OCFS2_IOC_GROUP_ADD: case OCFS2_IOC_GROUP_ADD64: + case FITRIM: + break; + case OCFS2_IOC_REFLINK: + if (copy_from_user(&args, argp, sizeof(args))) + return -EFAULT; + preserve = (args.preserve != 0); + + return ocfs2_reflink_ioctl(inode, compat_ptr(args.old_path), + compat_ptr(args.new_path), preserve); + case OCFS2_IOC_INFO: + if (copy_from_user(&info, argp, sizeof(struct ocfs2_info))) + return -EFAULT; + + return ocfs2_info_handle(inode, &info, 1); + case OCFS2_IOC_MOVE_EXT: break; default: return -ENOIOCTLCMD; diff --git a/fs/ocfs2/ioctl.h b/fs/ocfs2/ioctl.h index cf9a5ee30fe..0cd5323bd3f 100644 --- a/fs/ocfs2/ioctl.h +++ b/fs/ocfs2/ioctl.h @@ -7,10 +7,10 @@ * */ -#ifndef OCFS2_IOCTL_H -#define OCFS2_IOCTL_H +#ifndef OCFS2_IOCTL_PROTO_H +#define OCFS2_IOCTL_PROTO_H long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg); long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg); -#endif /* OCFS2_IOCTL_H */ +#endif /* OCFS2_IOCTL_PROTO_H */ diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c index a20a0f1e37f..4b0c68849b3 100644 --- a/fs/ocfs2/journal.c +++ b/fs/ocfs2/journal.c @@ -28,8 +28,10 @@ #include <linux/slab.h> #include <linux/highmem.h> #include <linux/kthread.h> +#include <linux/time.h> +#include <linux/random.h> +#include <linux/delay.h> -#define MLOG_MASK_PREFIX ML_JOURNAL #include <cluster/masklog.h> #include "ocfs2.h" @@ -46,12 +48,16 @@ #include "slot_map.h" #include "super.h" #include "sysfile.h" +#include "uptodate.h" #include "quota.h" #include "buffer_head_io.h" +#include "ocfs2_trace.h" DEFINE_SPINLOCK(trans_inc_lock); +#define ORPHAN_SCAN_SCHEDULE_TIMEOUT 300000 + static int ocfs2_force_read_journal(struct inode *inode); static int ocfs2_recover_node(struct ocfs2_super *osb, int node_num, int slot_num); @@ -296,19 +302,17 @@ static int ocfs2_commit_cache(struct ocfs2_super *osb) { int status = 0; unsigned int flushed; - unsigned long old_id; struct ocfs2_journal *journal = NULL; - mlog_entry_void(); - journal = osb->journal; /* Flush all pending commits and checkpoint the journal. */ down_write(&journal->j_trans_barrier); - if (atomic_read(&journal->j_num_trans) == 0) { + flushed = atomic_read(&journal->j_num_trans); + trace_ocfs2_commit_cache_begin(flushed); + if (flushed == 0) { up_write(&journal->j_trans_barrier); - mlog(0, "No transactions for me to flush!\n"); goto finally; } @@ -321,25 +325,20 @@ static int ocfs2_commit_cache(struct ocfs2_super *osb) goto finally; } - old_id = ocfs2_inc_trans_id(journal); + ocfs2_inc_trans_id(journal); flushed = atomic_read(&journal->j_num_trans); atomic_set(&journal->j_num_trans, 0); up_write(&journal->j_trans_barrier); - mlog(0, "commit_thread: flushed transaction %lu (%u handles)\n", - journal->j_trans_id, flushed); + trace_ocfs2_commit_cache_end(journal->j_trans_id, flushed); ocfs2_wake_downconvert_thread(osb); wake_up(&journal->j_checkpointed); finally: - mlog_exit(status); return status; } -/* pass it NULL and it will allocate a new handle object for you. If - * you pass it a handle however, it may still return error, in which - * case it has free'd the passed handle for you. */ handle_t *ocfs2_start_trans(struct ocfs2_super *osb, int max_buffs) { journal_t *journal = osb->journal->j_journal; @@ -357,11 +356,14 @@ handle_t *ocfs2_start_trans(struct ocfs2_super *osb, int max_buffs) if (journal_current_handle()) return jbd2_journal_start(journal, max_buffs); + sb_start_intwrite(osb->sb); + down_read(&osb->journal->j_trans_barrier); handle = jbd2_journal_start(journal, max_buffs); if (IS_ERR(handle)) { up_read(&osb->journal->j_trans_barrier); + sb_end_intwrite(osb->sb); mlog_errno(PTR_ERR(handle)); @@ -390,16 +392,16 @@ int ocfs2_commit_trans(struct ocfs2_super *osb, if (ret < 0) mlog_errno(ret); - if (!nested) + if (!nested) { up_read(&journal->j_trans_barrier); + sb_end_intwrite(osb->sb); + } return ret; } /* - * 'nblocks' is what you want to add to the current - * transaction. extend_trans will either extend the current handle by - * nblocks, or commit it and start a new one with nblocks credits. + * 'nblocks' is what you want to add to the current transaction. * * This might call jbd2_journal_restart() which will commit dirty buffers * and then restart the transaction. Before calling @@ -417,14 +419,17 @@ int ocfs2_commit_trans(struct ocfs2_super *osb, */ int ocfs2_extend_trans(handle_t *handle, int nblocks) { - int status; + int status, old_nblocks; BUG_ON(!handle); - BUG_ON(!nblocks); + BUG_ON(nblocks < 0); + + if (!nblocks) + return 0; - mlog_entry_void(); + old_nblocks = handle->h_buffer_credits; - mlog(0, "Trying to extend transaction by %d blocks\n", nblocks); + trace_ocfs2_extend_trans(old_nblocks, nblocks); #ifdef CONFIG_OCFS2_DEBUG_FS status = 1; @@ -437,10 +442,9 @@ int ocfs2_extend_trans(handle_t *handle, int nblocks) #endif if (status > 0) { - mlog(0, - "jbd2_journal_extend failed, trying " - "jbd2_journal_restart\n"); - status = jbd2_journal_restart(handle, nblocks); + trace_ocfs2_extend_trans_restart(old_nblocks + nblocks); + status = jbd2_journal_restart(handle, + old_nblocks + nblocks); if (status < 0) { mlog_errno(status); goto bail; @@ -449,11 +453,44 @@ int ocfs2_extend_trans(handle_t *handle, int nblocks) status = 0; bail: + return status; +} + +/* + * If we have fewer than thresh credits, extend by OCFS2_MAX_TRANS_DATA. + * If that fails, restart the transaction & regain write access for the + * buffer head which is used for metadata modifications. + * Taken from Ext4: extend_or_restart_transaction() + */ +int ocfs2_allocate_extend_trans(handle_t *handle, int thresh) +{ + int status, old_nblks; + + BUG_ON(!handle); + + old_nblks = handle->h_buffer_credits; + trace_ocfs2_allocate_extend_trans(old_nblks, thresh); + + if (old_nblks < thresh) + return 0; + + status = jbd2_journal_extend(handle, OCFS2_MAX_TRANS_DATA); + if (status < 0) { + mlog_errno(status); + goto bail; + } + + if (status > 0) { + status = jbd2_journal_restart(handle, OCFS2_MAX_TRANS_DATA); + if (status < 0) + mlog_errno(status); + } - mlog_exit(status); +bail: return status; } + struct ocfs2_triggers { struct jbd2_buffer_trigger_type ot_triggers; int ot_offset; @@ -464,7 +501,7 @@ static inline struct ocfs2_triggers *to_ocfs2_trigger(struct jbd2_buffer_trigger return container_of(triggers, struct ocfs2_triggers, ot_triggers); } -static void ocfs2_commit_trigger(struct jbd2_buffer_trigger_type *triggers, +static void ocfs2_frozen_trigger(struct jbd2_buffer_trigger_type *triggers, struct buffer_head *bh, void *data, size_t size) { @@ -483,7 +520,7 @@ static void ocfs2_commit_trigger(struct jbd2_buffer_trigger_type *triggers, * Quota blocks have their own trigger because the struct ocfs2_block_check * offset depends on the blocksize. */ -static void ocfs2_dq_commit_trigger(struct jbd2_buffer_trigger_type *triggers, +static void ocfs2_dq_frozen_trigger(struct jbd2_buffer_trigger_type *triggers, struct buffer_head *bh, void *data, size_t size) { @@ -503,7 +540,7 @@ static void ocfs2_dq_commit_trigger(struct jbd2_buffer_trigger_type *triggers, * Directory blocks also have their own trigger because the * struct ocfs2_block_check offset depends on the blocksize. */ -static void ocfs2_db_commit_trigger(struct jbd2_buffer_trigger_type *triggers, +static void ocfs2_db_frozen_trigger(struct jbd2_buffer_trigger_type *triggers, struct buffer_head *bh, void *data, size_t size) { @@ -536,7 +573,7 @@ static void ocfs2_abort_trigger(struct jbd2_buffer_trigger_type *triggers, static struct ocfs2_triggers di_triggers = { .ot_triggers = { - .t_commit = ocfs2_commit_trigger, + .t_frozen = ocfs2_frozen_trigger, .t_abort = ocfs2_abort_trigger, }, .ot_offset = offsetof(struct ocfs2_dinode, i_check), @@ -544,15 +581,23 @@ static struct ocfs2_triggers di_triggers = { static struct ocfs2_triggers eb_triggers = { .ot_triggers = { - .t_commit = ocfs2_commit_trigger, + .t_frozen = ocfs2_frozen_trigger, .t_abort = ocfs2_abort_trigger, }, .ot_offset = offsetof(struct ocfs2_extent_block, h_check), }; +static struct ocfs2_triggers rb_triggers = { + .ot_triggers = { + .t_frozen = ocfs2_frozen_trigger, + .t_abort = ocfs2_abort_trigger, + }, + .ot_offset = offsetof(struct ocfs2_refcount_block, rf_check), +}; + static struct ocfs2_triggers gd_triggers = { .ot_triggers = { - .t_commit = ocfs2_commit_trigger, + .t_frozen = ocfs2_frozen_trigger, .t_abort = ocfs2_abort_trigger, }, .ot_offset = offsetof(struct ocfs2_group_desc, bg_check), @@ -560,14 +605,14 @@ static struct ocfs2_triggers gd_triggers = { static struct ocfs2_triggers db_triggers = { .ot_triggers = { - .t_commit = ocfs2_db_commit_trigger, + .t_frozen = ocfs2_db_frozen_trigger, .t_abort = ocfs2_abort_trigger, }, }; static struct ocfs2_triggers xb_triggers = { .ot_triggers = { - .t_commit = ocfs2_commit_trigger, + .t_frozen = ocfs2_frozen_trigger, .t_abort = ocfs2_abort_trigger, }, .ot_offset = offsetof(struct ocfs2_xattr_block, xb_check), @@ -575,14 +620,14 @@ static struct ocfs2_triggers xb_triggers = { static struct ocfs2_triggers dq_triggers = { .ot_triggers = { - .t_commit = ocfs2_dq_commit_trigger, + .t_frozen = ocfs2_dq_frozen_trigger, .t_abort = ocfs2_abort_trigger, }, }; static struct ocfs2_triggers dr_triggers = { .ot_triggers = { - .t_commit = ocfs2_commit_trigger, + .t_frozen = ocfs2_frozen_trigger, .t_abort = ocfs2_abort_trigger, }, .ot_offset = offsetof(struct ocfs2_dx_root_block, dr_check), @@ -590,30 +635,29 @@ static struct ocfs2_triggers dr_triggers = { static struct ocfs2_triggers dl_triggers = { .ot_triggers = { - .t_commit = ocfs2_commit_trigger, + .t_frozen = ocfs2_frozen_trigger, .t_abort = ocfs2_abort_trigger, }, .ot_offset = offsetof(struct ocfs2_dx_leaf, dl_check), }; static int __ocfs2_journal_access(handle_t *handle, - struct inode *inode, + struct ocfs2_caching_info *ci, struct buffer_head *bh, struct ocfs2_triggers *triggers, int type) { int status; + struct ocfs2_super *osb = + OCFS2_SB(ocfs2_metadata_cache_get_super(ci)); - BUG_ON(!inode); + BUG_ON(!ci || !ci->ci_ops); BUG_ON(!handle); BUG_ON(!bh); - mlog_entry("bh->b_blocknr=%llu, type=%d (\"%s\"), bh->b_size = %zu\n", - (unsigned long long)bh->b_blocknr, type, - (type == OCFS2_JOURNAL_ACCESS_CREATE) ? - "OCFS2_JOURNAL_ACCESS_CREATE" : - "OCFS2_JOURNAL_ACCESS_WRITE", - bh->b_size); + trace_ocfs2_journal_access( + (unsigned long long)ocfs2_metadata_cache_owner(ci), + (unsigned long long)bh->b_blocknr, type, bh->b_size); /* we can safely remove this assertion after testing. */ if (!buffer_uptodate(bh)) { @@ -623,15 +667,15 @@ static int __ocfs2_journal_access(handle_t *handle, BUG(); } - /* Set the current transaction information on the inode so + /* Set the current transaction information on the ci so * that the locking code knows whether it can drop it's locks - * on this inode or not. We're protected from the commit + * on this ci or not. We're protected from the commit * thread updating the current transaction id until * ocfs2_commit_trans() because ocfs2_start_trans() took * j_trans_barrier for us. */ - ocfs2_set_inode_lock_trans(OCFS2_SB(inode->i_sb)->journal, inode); + ocfs2_set_ci_lock_trans(osb->journal, ci); - mutex_lock(&OCFS2_I(inode)->ip_io_mutex); + ocfs2_metadata_cache_io_lock(ci); switch (type) { case OCFS2_JOURNAL_ACCESS_CREATE: case OCFS2_JOURNAL_ACCESS_WRITE: @@ -644,98 +688,88 @@ static int __ocfs2_journal_access(handle_t *handle, default: status = -EINVAL; - mlog(ML_ERROR, "Uknown access type!\n"); + mlog(ML_ERROR, "Unknown access type!\n"); } - if (!status && ocfs2_meta_ecc(OCFS2_SB(inode->i_sb)) && triggers) + if (!status && ocfs2_meta_ecc(osb) && triggers) jbd2_journal_set_triggers(bh, &triggers->ot_triggers); - mutex_unlock(&OCFS2_I(inode)->ip_io_mutex); + ocfs2_metadata_cache_io_unlock(ci); if (status < 0) mlog(ML_ERROR, "Error %d getting %d access to buffer!\n", status, type); - mlog_exit(status); return status; } -int ocfs2_journal_access_di(handle_t *handle, struct inode *inode, - struct buffer_head *bh, int type) +int ocfs2_journal_access_di(handle_t *handle, struct ocfs2_caching_info *ci, + struct buffer_head *bh, int type) { - return __ocfs2_journal_access(handle, inode, bh, &di_triggers, - type); + return __ocfs2_journal_access(handle, ci, bh, &di_triggers, type); } -int ocfs2_journal_access_eb(handle_t *handle, struct inode *inode, +int ocfs2_journal_access_eb(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type) { - return __ocfs2_journal_access(handle, inode, bh, &eb_triggers, - type); + return __ocfs2_journal_access(handle, ci, bh, &eb_triggers, type); } -int ocfs2_journal_access_gd(handle_t *handle, struct inode *inode, +int ocfs2_journal_access_rb(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type) { - return __ocfs2_journal_access(handle, inode, bh, &gd_triggers, + return __ocfs2_journal_access(handle, ci, bh, &rb_triggers, type); } -int ocfs2_journal_access_db(handle_t *handle, struct inode *inode, +int ocfs2_journal_access_gd(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type) { - return __ocfs2_journal_access(handle, inode, bh, &db_triggers, - type); + return __ocfs2_journal_access(handle, ci, bh, &gd_triggers, type); } -int ocfs2_journal_access_xb(handle_t *handle, struct inode *inode, +int ocfs2_journal_access_db(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type) { - return __ocfs2_journal_access(handle, inode, bh, &xb_triggers, - type); + return __ocfs2_journal_access(handle, ci, bh, &db_triggers, type); } -int ocfs2_journal_access_dq(handle_t *handle, struct inode *inode, +int ocfs2_journal_access_xb(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type) { - return __ocfs2_journal_access(handle, inode, bh, &dq_triggers, - type); + return __ocfs2_journal_access(handle, ci, bh, &xb_triggers, type); } -int ocfs2_journal_access_dr(handle_t *handle, struct inode *inode, +int ocfs2_journal_access_dq(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type) { - return __ocfs2_journal_access(handle, inode, bh, &dr_triggers, - type); + return __ocfs2_journal_access(handle, ci, bh, &dq_triggers, type); } -int ocfs2_journal_access_dl(handle_t *handle, struct inode *inode, +int ocfs2_journal_access_dr(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type) { - return __ocfs2_journal_access(handle, inode, bh, &dl_triggers, - type); + return __ocfs2_journal_access(handle, ci, bh, &dr_triggers, type); +} + +int ocfs2_journal_access_dl(handle_t *handle, struct ocfs2_caching_info *ci, + struct buffer_head *bh, int type) +{ + return __ocfs2_journal_access(handle, ci, bh, &dl_triggers, type); } -int ocfs2_journal_access(handle_t *handle, struct inode *inode, +int ocfs2_journal_access(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type) { - return __ocfs2_journal_access(handle, inode, bh, NULL, type); + return __ocfs2_journal_access(handle, ci, bh, NULL, type); } -int ocfs2_journal_dirty(handle_t *handle, - struct buffer_head *bh) +void ocfs2_journal_dirty(handle_t *handle, struct buffer_head *bh) { int status; - mlog_entry("(bh->b_blocknr=%llu)\n", - (unsigned long long)bh->b_blocknr); + trace_ocfs2_journal_dirty((unsigned long long)bh->b_blocknr); status = jbd2_journal_dirty_metadata(handle, bh); - if (status < 0) - mlog(ML_ERROR, "Could not dirty metadata buffer. " - "(bh->b_blocknr=%llu)\n", - (unsigned long long)bh->b_blocknr); - - mlog_exit(status); - return status; + BUG_ON(status); } #define OCFS2_DEFAULT_COMMIT_INTERVAL (HZ * JBD2_DEFAULT_MAX_COMMIT_AGE) @@ -748,13 +782,13 @@ void ocfs2_set_journal_params(struct ocfs2_super *osb) if (osb->osb_commit_interval) commit_interval = osb->osb_commit_interval; - spin_lock(&journal->j_state_lock); + write_lock(&journal->j_state_lock); journal->j_commit_interval = commit_interval; if (osb->s_mount_opt & OCFS2_MOUNT_BARRIER) journal->j_flags |= JBD2_BARRIER; else journal->j_flags &= ~JBD2_BARRIER; - spin_unlock(&journal->j_state_lock); + write_unlock(&journal->j_state_lock); } int ocfs2_journal_init(struct ocfs2_journal *journal, int *dirty) @@ -767,8 +801,6 @@ int ocfs2_journal_init(struct ocfs2_journal *journal, int *dirty) struct ocfs2_super *osb; int inode_lock = 0; - mlog_entry_void(); - BUG_ON(!journal); osb = journal->j_osb; @@ -805,17 +837,16 @@ int ocfs2_journal_init(struct ocfs2_journal *journal, int *dirty) inode_lock = 1; di = (struct ocfs2_dinode *)bh->b_data; - if (inode->i_size < OCFS2_MIN_JOURNAL_SIZE) { + if (i_size_read(inode) < OCFS2_MIN_JOURNAL_SIZE) { mlog(ML_ERROR, "Journal file size (%lld) is too small!\n", - inode->i_size); + i_size_read(inode)); status = -EINVAL; goto done; } - mlog(0, "inode->i_size = %lld\n", inode->i_size); - mlog(0, "inode->i_blocks = %llu\n", - (unsigned long long)inode->i_blocks); - mlog(0, "inode->ip_clusters = %u\n", OCFS2_I(inode)->ip_clusters); + trace_ocfs2_journal_init(i_size_read(inode), + (unsigned long long)inode->i_blocks, + OCFS2_I(inode)->ip_clusters); /* call the kernels journal init function now */ j_journal = jbd2_journal_init_inode(inode); @@ -825,8 +856,7 @@ int ocfs2_journal_init(struct ocfs2_journal *journal, int *dirty) goto done; } - mlog(0, "Returned from jbd2_journal_init_inode\n"); - mlog(0, "j_journal->j_maxlen = %u\n", j_journal->j_maxlen); + trace_ocfs2_journal_init_maxlen(j_journal->j_maxlen); *dirty = (le32_to_cpu(di->id1.journal1.ij_flags) & OCFS2_JOURNAL_DIRTY_FL); @@ -851,7 +881,6 @@ done: } } - mlog_exit(status); return status; } @@ -874,8 +903,6 @@ static int ocfs2_journal_toggle_dirty(struct ocfs2_super *osb, struct buffer_head *bh = journal->j_bh; struct ocfs2_dinode *fe; - mlog_entry_void(); - fe = (struct ocfs2_dinode *)bh->b_data; /* The journal bh on the osb always comes from ocfs2_journal_init() @@ -894,11 +921,10 @@ static int ocfs2_journal_toggle_dirty(struct ocfs2_super *osb, ocfs2_bump_recovery_generation(fe); ocfs2_compute_meta_ecc(osb->sb, bh->b_data, &fe->i_check); - status = ocfs2_write_block(osb, bh, journal->j_inode); + status = ocfs2_write_block(osb, bh, INODE_CACHE(journal->j_inode)); if (status < 0) mlog_errno(status); - mlog_exit(status); return status; } @@ -913,8 +939,6 @@ void ocfs2_journal_shutdown(struct ocfs2_super *osb) struct inode *inode = NULL; int num_running_trans = 0; - mlog_entry_void(); - BUG_ON(!osb); journal = osb->journal; @@ -931,10 +955,7 @@ void ocfs2_journal_shutdown(struct ocfs2_super *osb) BUG(); num_running_trans = atomic_read(&(osb->journal->j_num_trans)); - if (num_running_trans > 0) - mlog(0, "Shutting down journal: must wait on %d " - "running transactions!\n", - num_running_trans); + trace_ocfs2_journal_shutdown(num_running_trans); /* Do a commit_cache here. It will flush our journal, *and* * release any locks that are still held. @@ -947,7 +968,7 @@ void ocfs2_journal_shutdown(struct ocfs2_super *osb) * completely destroy the journal. */ if (osb->commit_task) { /* Wait for the commit thread */ - mlog(0, "Waiting for ocfs2commit to exit....\n"); + trace_ocfs2_journal_shutdown_wait(osb->commit_task); kthread_stop(osb->commit_task); osb->commit_task = NULL; } @@ -990,7 +1011,6 @@ void ocfs2_journal_shutdown(struct ocfs2_super *osb) done: if (inode) iput(inode); - mlog_exit_void(); } static void ocfs2_clear_journal_error(struct super_block *sb, @@ -1016,8 +1036,6 @@ int ocfs2_journal_load(struct ocfs2_journal *journal, int local, int replayed) int status = 0; struct ocfs2_super *osb; - mlog_entry_void(); - BUG_ON(!journal); osb = journal->j_osb; @@ -1051,7 +1069,6 @@ int ocfs2_journal_load(struct ocfs2_journal *journal, int local, int replayed) osb->commit_task = NULL; done: - mlog_exit(status); return status; } @@ -1062,8 +1079,6 @@ int ocfs2_journal_wipe(struct ocfs2_journal *journal, int full) { int status; - mlog_entry_void(); - BUG_ON(!journal); status = jbd2_journal_wipe(journal->j_journal, full); @@ -1077,7 +1092,6 @@ int ocfs2_journal_wipe(struct ocfs2_journal *journal, int full) mlog_errno(status); bail: - mlog_exit(status); return status; } @@ -1116,11 +1130,9 @@ static int ocfs2_force_read_journal(struct inode *inode) #define CONCURRENT_JOURNAL_FILL 32ULL struct buffer_head *bhs[CONCURRENT_JOURNAL_FILL]; - mlog_entry_void(); - memset(bhs, 0, sizeof(struct buffer_head *) * CONCURRENT_JOURNAL_FILL); - num_blocks = ocfs2_blocks_for_bytes(inode->i_sb, inode->i_size); + num_blocks = ocfs2_blocks_for_bytes(inode->i_sb, i_size_read(inode)); v_blkno = 0; while (v_blkno < num_blocks) { status = ocfs2_extent_map_get_blocks(inode, v_blkno, @@ -1153,7 +1165,6 @@ static int ocfs2_force_read_journal(struct inode *inode) bail: for(i = 0; i < CONCURRENT_JOURNAL_FILL; i++) brelse(bhs[i]); - mlog_exit(status); return status; } @@ -1177,7 +1188,7 @@ struct ocfs2_la_recovery_item { */ void ocfs2_complete_recovery(struct work_struct *work) { - int ret; + int ret = 0; struct ocfs2_journal *journal = container_of(work, struct ocfs2_journal, j_recovery_work); struct ocfs2_super *osb = journal->j_osb; @@ -1186,9 +1197,8 @@ void ocfs2_complete_recovery(struct work_struct *work) struct ocfs2_quota_recovery *qrec; LIST_HEAD(tmp_la_list); - mlog_entry_void(); - - mlog(0, "completing recovery from keventd\n"); + trace_ocfs2_complete_recovery( + (unsigned long long)OCFS2_I(journal->j_inode)->ip_blkno); spin_lock(&journal->j_lock); list_splice_init(&journal->j_la_cleanups, &tmp_la_list); @@ -1197,15 +1207,18 @@ void ocfs2_complete_recovery(struct work_struct *work) list_for_each_entry_safe(item, n, &tmp_la_list, lri_list) { list_del_init(&item->lri_list); - mlog(0, "Complete recovery for slot %d\n", item->lri_slot); - ocfs2_wait_on_quotas(osb); la_dinode = item->lri_la_dinode; - if (la_dinode) { - mlog(0, "Clean up local alloc %llu\n", - (unsigned long long)le64_to_cpu(la_dinode->i_blkno)); + tl_dinode = item->lri_tl_dinode; + qrec = item->lri_qrec; + trace_ocfs2_complete_recovery_slot(item->lri_slot, + la_dinode ? le64_to_cpu(la_dinode->i_blkno) : 0, + tl_dinode ? le64_to_cpu(tl_dinode->i_blkno) : 0, + qrec); + + if (la_dinode) { ret = ocfs2_complete_local_alloc_recovery(osb, la_dinode); if (ret < 0) @@ -1214,11 +1227,7 @@ void ocfs2_complete_recovery(struct work_struct *work) kfree(la_dinode); } - tl_dinode = item->lri_tl_dinode; if (tl_dinode) { - mlog(0, "Clean up truncate log %llu\n", - (unsigned long long)le64_to_cpu(tl_dinode->i_blkno)); - ret = ocfs2_complete_truncate_log_recovery(osb, tl_dinode); if (ret < 0) @@ -1231,9 +1240,7 @@ void ocfs2_complete_recovery(struct work_struct *work) if (ret < 0) mlog_errno(ret); - qrec = item->lri_qrec; if (qrec) { - mlog(0, "Recovering quota files"); ret = ocfs2_finish_quota_recovery(osb, qrec, item->lri_slot); if (ret < 0) @@ -1244,8 +1251,7 @@ void ocfs2_complete_recovery(struct work_struct *work) kfree(item); } - mlog(0, "Recovery completion\n"); - mlog_exit_void(); + trace_ocfs2_complete_recovery_end(ret); } /* NOTE: This function always eats your references to la_dinode and @@ -1264,11 +1270,8 @@ static void ocfs2_queue_recovery_completion(struct ocfs2_journal *journal, /* Though we wish to avoid it, we are in fact safe in * skipping local alloc cleanup as fsck.ocfs2 is more * than capable of reclaiming unused space. */ - if (la_dinode) - kfree(la_dinode); - - if (tl_dinode) - kfree(tl_dinode); + kfree(la_dinode); + kfree(tl_dinode); if (qrec) ocfs2_free_quota_recovery(qrec); @@ -1295,6 +1298,9 @@ void ocfs2_complete_mount_recovery(struct ocfs2_super *osb) { struct ocfs2_journal *journal = osb->journal; + if (ocfs2_is_hard_readonly(osb)) + return; + /* No need to queue up our truncate_log as regular cleanup will catch * that */ ocfs2_queue_recovery_completion(journal, osb->slot_num, @@ -1331,8 +1337,6 @@ static int __ocfs2_recovery_thread(void *arg) int rm_quota_used = 0, i; struct ocfs2_quota_recovery *qrec; - mlog_entry_void(); - status = ocfs2_wait_on_mount(osb); if (status < 0) { goto bail; @@ -1364,15 +1368,12 @@ restart: * clear it until ocfs2_recover_node() has succeeded. */ node_num = rm->rm_entries[0]; spin_unlock(&osb->osb_lock); - mlog(0, "checking node %d\n", node_num); slot_num = ocfs2_node_num_to_slot(osb, node_num); + trace_ocfs2_recovery_thread_node(node_num, slot_num); if (slot_num == -ENOENT) { status = 0; - mlog(0, "no slot for this node, so no recovery" - "required.\n"); goto skip_recovery; } - mlog(0, "node %d was using slot %d\n", node_num, slot_num); /* It is a bit subtle with quota recovery. We cannot do it * immediately because we have to obtain cluster locks from @@ -1399,7 +1400,7 @@ skip_recovery: spin_lock(&osb->osb_lock); } spin_unlock(&osb->osb_lock); - mlog(0, "All nodes recovered\n"); + trace_ocfs2_recovery_thread_end(status); /* Refresh all journal recovery generations from disk */ status = ocfs2_check_journals_nolocks(osb); @@ -1408,7 +1409,7 @@ skip_recovery: mlog_errno(status); /* Now it is right time to recover quotas... We have to do this under - * superblock lock so that noone can start using the slot (and crash) + * superblock lock so that no one can start using the slot (and crash) * before we recover it */ for (i = 0; i < rm_quota_used; i++) { qrec = ocfs2_begin_quota_recovery(osb, rm_quota[i]); @@ -1440,10 +1441,8 @@ bail: mutex_unlock(&osb->recovery_lock); - if (rm_quota) - kfree(rm_quota); + kfree(rm_quota); - mlog_exit(status); /* no one is callint kthread_stop() for us so the kthread() api * requires that we call do_exit(). And it isn't exported, but * complete_and_exit() seems to be a minimal wrapper around it. */ @@ -1453,19 +1452,15 @@ bail: void ocfs2_recovery_thread(struct ocfs2_super *osb, int node_num) { - mlog_entry("(node_num=%d, osb->node_num = %d)\n", - node_num, osb->node_num); - mutex_lock(&osb->recovery_lock); - if (osb->disable_recovery) - goto out; - /* People waiting on recovery will wait on - * the recovery map to empty. */ - if (ocfs2_recovery_map_set(osb, node_num)) - mlog(0, "node %d already in recovery map.\n", node_num); + trace_ocfs2_recovery_thread(node_num, osb->node_num, + osb->disable_recovery, osb->recovery_thread_task, + osb->disable_recovery ? + -1 : ocfs2_recovery_map_set(osb, node_num)); - mlog(0, "starting recovery thread...\n"); + if (osb->disable_recovery) + goto out; if (osb->recovery_thread_task) goto out; @@ -1480,8 +1475,6 @@ void ocfs2_recovery_thread(struct ocfs2_super *osb, int node_num) out: mutex_unlock(&osb->recovery_lock); wake_up(&osb->recovery_event); - - mlog_exit_void(); } static int ocfs2_read_journal_inode(struct ocfs2_super *osb, @@ -1555,7 +1548,7 @@ static int ocfs2_replay_journal(struct ocfs2_super *osb, * If not, it needs recovery. */ if (osb->slot_recovery_generations[slot_num] != slot_reco_gen) { - mlog(0, "Slot %u already recovered (old/new=%u/%u)\n", slot_num, + trace_ocfs2_replay_journal_recovered(slot_num, osb->slot_recovery_generations[slot_num], slot_reco_gen); osb->slot_recovery_generations[slot_num] = slot_reco_gen; status = -EBUSY; @@ -1566,7 +1559,7 @@ static int ocfs2_replay_journal(struct ocfs2_super *osb, status = ocfs2_inode_lock_full(inode, &bh, 1, OCFS2_META_LOCK_RECOVERY); if (status < 0) { - mlog(0, "status returned from ocfs2_inode_lock=%d\n", status); + trace_ocfs2_replay_journal_lock_err(status); if (status != -ERESTARTSYS) mlog(ML_ERROR, "Could not lock journal!\n"); goto done; @@ -1579,7 +1572,7 @@ static int ocfs2_replay_journal(struct ocfs2_super *osb, slot_reco_gen = ocfs2_get_recovery_generation(fe); if (!(flags & OCFS2_JOURNAL_DIRTY_FL)) { - mlog(0, "No recovery required for node %d\n", node_num); + trace_ocfs2_replay_journal_skip(node_num); /* Refresh recovery generation for the slot */ osb->slot_recovery_generations[slot_num] = slot_reco_gen; goto done; @@ -1588,9 +1581,9 @@ static int ocfs2_replay_journal(struct ocfs2_super *osb, /* we need to run complete recovery for offline orphan slots */ ocfs2_replay_map_set_state(osb, REPLAY_NEEDED); - mlog(ML_NOTICE, "Recovering node %d from slot %d on device (%u,%u)\n", - node_num, slot_num, - MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev)); + printk(KERN_NOTICE "ocfs2: Begin replay journal (node %d, slot %d) on "\ + "device (%u,%u)\n", node_num, slot_num, MAJOR(osb->sb->s_dev), + MINOR(osb->sb->s_dev)); OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters); @@ -1600,7 +1593,6 @@ static int ocfs2_replay_journal(struct ocfs2_super *osb, goto done; } - mlog(0, "calling journal_init_inode\n"); journal = jbd2_journal_init_inode(inode); if (journal == NULL) { mlog(ML_ERROR, "Linux journal layer error\n"); @@ -1620,7 +1612,6 @@ static int ocfs2_replay_journal(struct ocfs2_super *osb, ocfs2_clear_journal_error(osb->sb, journal, slot_num); /* wipe the journal */ - mlog(0, "flushing the journal.\n"); jbd2_journal_lock_updates(journal); status = jbd2_journal_flush(journal); jbd2_journal_unlock_updates(journal); @@ -1638,7 +1629,7 @@ static int ocfs2_replay_journal(struct ocfs2_super *osb, ocfs2_get_recovery_generation(fe); ocfs2_compute_meta_ecc(osb->sb, bh->b_data, &fe->i_check); - status = ocfs2_write_block(osb, bh, inode); + status = ocfs2_write_block(osb, bh, INODE_CACHE(inode)); if (status < 0) mlog_errno(status); @@ -1647,6 +1638,9 @@ static int ocfs2_replay_journal(struct ocfs2_super *osb, jbd2_journal_destroy(journal); + printk(KERN_NOTICE "ocfs2: End replay journal (node %d, slot %d) on "\ + "device (%u,%u)\n", node_num, slot_num, MAJOR(osb->sb->s_dev), + MINOR(osb->sb->s_dev)); done: /* drop the lock on this nodes journal */ if (got_lock) @@ -1657,7 +1651,6 @@ done: brelse(bh); - mlog_exit(status); return status; } @@ -1680,8 +1673,7 @@ static int ocfs2_recover_node(struct ocfs2_super *osb, struct ocfs2_dinode *la_copy = NULL; struct ocfs2_dinode *tl_copy = NULL; - mlog_entry("(node_num=%d, slot_num=%d, osb->node_num = %d)\n", - node_num, slot_num, osb->node_num); + trace_ocfs2_recover_node(node_num, slot_num, osb->node_num); /* Should not ever be called to recover ourselves -- in that * case we should've called ocfs2_journal_load instead. */ @@ -1690,9 +1682,7 @@ static int ocfs2_recover_node(struct ocfs2_super *osb, status = ocfs2_replay_journal(osb, node_num, slot_num); if (status < 0) { if (status == -EBUSY) { - mlog(0, "Skipping recovery for slot %u (node %u) " - "as another node has recovered it\n", slot_num, - node_num); + trace_ocfs2_recover_node_skip(slot_num, node_num); status = 0; goto done; } @@ -1727,7 +1717,6 @@ static int ocfs2_recover_node(struct ocfs2_super *osb, status = 0; done: - mlog_exit(status); return status; } @@ -1800,8 +1789,8 @@ int ocfs2_mark_dead_nodes(struct ocfs2_super *osb) spin_lock(&osb->osb_lock); osb->slot_recovery_generations[i] = gen; - mlog(0, "Slot %u recovery generation is %u\n", i, - osb->slot_recovery_generations[i]); + trace_ocfs2_mark_dead_nodes(i, + osb->slot_recovery_generations[i]); if (i == osb->slot_num) { spin_unlock(&osb->osb_lock); @@ -1837,11 +1826,158 @@ int ocfs2_mark_dead_nodes(struct ocfs2_super *osb) status = 0; bail: - mlog_exit(status); return status; } +/* + * Scan timer should get fired every ORPHAN_SCAN_SCHEDULE_TIMEOUT. Add some + * randomness to the timeout to minimize multple nodes firing the timer at the + * same time. + */ +static inline unsigned long ocfs2_orphan_scan_timeout(void) +{ + unsigned long time; + + get_random_bytes(&time, sizeof(time)); + time = ORPHAN_SCAN_SCHEDULE_TIMEOUT + (time % 5000); + return msecs_to_jiffies(time); +} + +/* + * ocfs2_queue_orphan_scan calls ocfs2_queue_recovery_completion for + * every slot, queuing a recovery of the slot on the ocfs2_wq thread. This + * is done to catch any orphans that are left over in orphan directories. + * + * It scans all slots, even ones that are in use. It does so to handle the + * case described below: + * + * Node 1 has an inode it was using. The dentry went away due to memory + * pressure. Node 1 closes the inode, but it's on the free list. The node + * has the open lock. + * Node 2 unlinks the inode. It grabs the dentry lock to notify others, + * but node 1 has no dentry and doesn't get the message. It trylocks the + * open lock, sees that another node has a PR, and does nothing. + * Later node 2 runs its orphan dir. It igets the inode, trylocks the + * open lock, sees the PR still, and does nothing. + * Basically, we have to trigger an orphan iput on node 1. The only way + * for this to happen is if node 1 runs node 2's orphan dir. + * + * ocfs2_queue_orphan_scan gets called every ORPHAN_SCAN_SCHEDULE_TIMEOUT + * seconds. It gets an EX lock on os_lockres and checks sequence number + * stored in LVB. If the sequence number has changed, it means some other + * node has done the scan. This node skips the scan and tracks the + * sequence number. If the sequence number didn't change, it means a scan + * hasn't happened. The node queues a scan and increments the + * sequence number in the LVB. + */ +void ocfs2_queue_orphan_scan(struct ocfs2_super *osb) +{ + struct ocfs2_orphan_scan *os; + int status, i; + u32 seqno = 0; + + os = &osb->osb_orphan_scan; + + if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE) + goto out; + + trace_ocfs2_queue_orphan_scan_begin(os->os_count, os->os_seqno, + atomic_read(&os->os_state)); + + status = ocfs2_orphan_scan_lock(osb, &seqno); + if (status < 0) { + if (status != -EAGAIN) + mlog_errno(status); + goto out; + } + + /* Do no queue the tasks if the volume is being umounted */ + if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE) + goto unlock; + + if (os->os_seqno != seqno) { + os->os_seqno = seqno; + goto unlock; + } + + for (i = 0; i < osb->max_slots; i++) + ocfs2_queue_recovery_completion(osb->journal, i, NULL, NULL, + NULL); + /* + * We queued a recovery on orphan slots, increment the sequence + * number and update LVB so other node will skip the scan for a while + */ + seqno++; + os->os_count++; + os->os_scantime = CURRENT_TIME; +unlock: + ocfs2_orphan_scan_unlock(osb, seqno); +out: + trace_ocfs2_queue_orphan_scan_end(os->os_count, os->os_seqno, + atomic_read(&os->os_state)); + return; +} + +/* Worker task that gets fired every ORPHAN_SCAN_SCHEDULE_TIMEOUT millsec */ +void ocfs2_orphan_scan_work(struct work_struct *work) +{ + struct ocfs2_orphan_scan *os; + struct ocfs2_super *osb; + + os = container_of(work, struct ocfs2_orphan_scan, + os_orphan_scan_work.work); + osb = os->os_osb; + + mutex_lock(&os->os_lock); + ocfs2_queue_orphan_scan(osb); + if (atomic_read(&os->os_state) == ORPHAN_SCAN_ACTIVE) + queue_delayed_work(ocfs2_wq, &os->os_orphan_scan_work, + ocfs2_orphan_scan_timeout()); + mutex_unlock(&os->os_lock); +} + +void ocfs2_orphan_scan_stop(struct ocfs2_super *osb) +{ + struct ocfs2_orphan_scan *os; + + os = &osb->osb_orphan_scan; + if (atomic_read(&os->os_state) == ORPHAN_SCAN_ACTIVE) { + atomic_set(&os->os_state, ORPHAN_SCAN_INACTIVE); + mutex_lock(&os->os_lock); + cancel_delayed_work(&os->os_orphan_scan_work); + mutex_unlock(&os->os_lock); + } +} + +void ocfs2_orphan_scan_init(struct ocfs2_super *osb) +{ + struct ocfs2_orphan_scan *os; + + os = &osb->osb_orphan_scan; + os->os_osb = osb; + os->os_count = 0; + os->os_seqno = 0; + mutex_init(&os->os_lock); + INIT_DELAYED_WORK(&os->os_orphan_scan_work, ocfs2_orphan_scan_work); +} + +void ocfs2_orphan_scan_start(struct ocfs2_super *osb) +{ + struct ocfs2_orphan_scan *os; + + os = &osb->osb_orphan_scan; + os->os_scantime = CURRENT_TIME; + if (ocfs2_is_hard_readonly(osb) || ocfs2_mount_local(osb)) + atomic_set(&os->os_state, ORPHAN_SCAN_INACTIVE); + else { + atomic_set(&os->os_state, ORPHAN_SCAN_ACTIVE); + queue_delayed_work(ocfs2_wq, &os->os_orphan_scan_work, + ocfs2_orphan_scan_timeout()); + } +} + struct ocfs2_orphan_filldir_priv { + struct dir_context ctx; struct inode *head; struct ocfs2_super *osb; }; @@ -1863,8 +1999,7 @@ static int ocfs2_orphan_filldir(void *priv, const char *name, int name_len, if (IS_ERR(iter)) return 0; - mlog(0, "queue orphan %llu\n", - (unsigned long long)OCFS2_I(iter)->ip_blkno); + trace_ocfs2_orphan_filldir((unsigned long long)OCFS2_I(iter)->ip_blkno); /* No locking is required for the next_orphan queue as there * is only ever a single process doing orphan recovery. */ OCFS2_I(iter)->ip_next_orphan = p->head; @@ -1879,11 +2014,11 @@ static int ocfs2_queue_orphans(struct ocfs2_super *osb, { int status; struct inode *orphan_dir_inode = NULL; - struct ocfs2_orphan_filldir_priv priv; - loff_t pos = 0; - - priv.osb = osb; - priv.head = *head; + struct ocfs2_orphan_filldir_priv priv = { + .ctx.actor = ocfs2_orphan_filldir, + .osb = osb, + .head = *head + }; orphan_dir_inode = ocfs2_get_system_file_inode(osb, ORPHAN_DIR_SYSTEM_INODE, @@ -1892,7 +2027,7 @@ static int ocfs2_queue_orphans(struct ocfs2_super *osb, status = -ENOENT; mlog_errno(status); return status; - } + } mutex_lock(&orphan_dir_inode->i_mutex); status = ocfs2_inode_lock(orphan_dir_inode, NULL, 0); @@ -1901,8 +2036,7 @@ static int ocfs2_queue_orphans(struct ocfs2_super *osb, goto out; } - status = ocfs2_dir_foreach(orphan_dir_inode, &pos, &priv, - ocfs2_orphan_filldir); + status = ocfs2_dir_foreach(orphan_dir_inode, &priv.ctx); if (status) { mlog_errno(status); goto out_cluster; @@ -1980,7 +2114,7 @@ static int ocfs2_recover_orphans(struct ocfs2_super *osb, struct inode *iter; struct ocfs2_inode_info *oi; - mlog(0, "Recover inodes from orphan dir in slot %d\n", slot); + trace_ocfs2_recover_orphans(slot); ocfs2_mark_recovering_orphan_dir(osb, slot); ret = ocfs2_queue_orphans(osb, slot, &inode); @@ -1993,17 +2127,12 @@ static int ocfs2_recover_orphans(struct ocfs2_super *osb, while (inode) { oi = OCFS2_I(inode); - mlog(0, "iput orphan %llu\n", (unsigned long long)oi->ip_blkno); + trace_ocfs2_recover_orphans_iput( + (unsigned long long)oi->ip_blkno); iter = oi->ip_next_orphan; spin_lock(&oi->ip_lock); - /* The remote delete code may have set these on the - * assumption that the other node would wipe them - * successfully. If they are still in the node's - * orphan dir, we need to reset that state. */ - oi->ip_flags &= ~(OCFS2_INODE_DELETED|OCFS2_INODE_SKIP_DELETE); - /* Set the proper information to get us going into * ocfs2_delete_inode. */ oi->ip_flags |= OCFS2_INODE_MAYBE_ORPHANED; @@ -2031,6 +2160,7 @@ static int __ocfs2_wait_on_mount(struct ocfs2_super *osb, int quota) * MOUNTED flag, but this is set right before * dismount_volume() so we can trust it. */ if (atomic_read(&osb->vol_state) == VOLUME_DISABLED) { + trace_ocfs2_wait_on_mount(VOLUME_DISABLED); mlog(0, "mount error, exiting!\n"); return -EBUSY; } @@ -2056,8 +2186,20 @@ static int ocfs2_commit_thread(void *arg) || kthread_should_stop()); status = ocfs2_commit_cache(osb); - if (status < 0) - mlog_errno(status); + if (status < 0) { + static unsigned long abort_warn_time; + + /* Warn about this once per minute */ + if (printk_timed_ratelimit(&abort_warn_time, 60*HZ)) + mlog(ML_ERROR, "status = %d, journal is " + "already aborted.\n", status); + /* + * After ocfs2_commit_cache() fails, j_num_trans has a + * non-zero value. Sleep here to avoid a busy-wait + * loop. + */ + msleep_interruptible(1000); + } if (kthread_should_stop() && atomic_read(&journal->j_num_trans)){ mlog(ML_KTHREAD, diff --git a/fs/ocfs2/journal.h b/fs/ocfs2/journal.h index eb7b76331eb..7f8cde94abf 100644 --- a/fs/ocfs2/journal.h +++ b/fs/ocfs2/journal.h @@ -67,11 +67,12 @@ struct ocfs2_journal { struct buffer_head *j_bh; /* Journal disk inode block */ atomic_t j_num_trans; /* Number of transactions * currently in the system. */ + spinlock_t j_lock; unsigned long j_trans_id; struct rw_semaphore j_trans_barrier; wait_queue_head_t j_checkpointed; - spinlock_t j_lock; + /* both fields protected by j_lock*/ struct list_head j_la_cleanups; struct work_struct j_recovery_work; }; @@ -90,60 +91,75 @@ static inline unsigned long ocfs2_inc_trans_id(struct ocfs2_journal *j) return old_id; } -static inline void ocfs2_set_inode_lock_trans(struct ocfs2_journal *journal, - struct inode *inode) +static inline void ocfs2_set_ci_lock_trans(struct ocfs2_journal *journal, + struct ocfs2_caching_info *ci) { spin_lock(&trans_inc_lock); - OCFS2_I(inode)->ip_last_trans = journal->j_trans_id; + ci->ci_last_trans = journal->j_trans_id; spin_unlock(&trans_inc_lock); } /* Used to figure out whether it's safe to drop a metadata lock on an - * inode. Returns true if all the inodes changes have been + * cached object. Returns true if all the object's changes have been * checkpointed to disk. You should be holding the spinlock on the * metadata lock while calling this to be sure that nobody can take * the lock and put it on another transaction. */ -static inline int ocfs2_inode_fully_checkpointed(struct inode *inode) +static inline int ocfs2_ci_fully_checkpointed(struct ocfs2_caching_info *ci) { int ret; - struct ocfs2_journal *journal = OCFS2_SB(inode->i_sb)->journal; + struct ocfs2_journal *journal = + OCFS2_SB(ocfs2_metadata_cache_get_super(ci))->journal; spin_lock(&trans_inc_lock); - ret = time_after(journal->j_trans_id, OCFS2_I(inode)->ip_last_trans); + ret = time_after(journal->j_trans_id, ci->ci_last_trans); spin_unlock(&trans_inc_lock); return ret; } -/* convenience function to check if an inode is still new (has never - * hit disk) Will do you a favor and set created_trans = 0 when you've - * been checkpointed. returns '1' if the inode is still new. */ -static inline int ocfs2_inode_is_new(struct inode *inode) +/* convenience function to check if an object backed by struct + * ocfs2_caching_info is still new (has never hit disk) Will do you a + * favor and set created_trans = 0 when you've + * been checkpointed. returns '1' if the ci is still new. */ +static inline int ocfs2_ci_is_new(struct ocfs2_caching_info *ci) { int ret; + struct ocfs2_journal *journal = + OCFS2_SB(ocfs2_metadata_cache_get_super(ci))->journal; + + spin_lock(&trans_inc_lock); + ret = !(time_after(journal->j_trans_id, ci->ci_created_trans)); + if (!ret) + ci->ci_created_trans = 0; + spin_unlock(&trans_inc_lock); + return ret; +} +/* Wrapper for inodes so we can check system files */ +static inline int ocfs2_inode_is_new(struct inode *inode) +{ /* System files are never "new" as they're written out by * mkfs. This helps us early during mount, before we have the * journal open and j_trans_id could be junk. */ if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SYSTEM_FILE) return 0; - spin_lock(&trans_inc_lock); - ret = !(time_after(OCFS2_SB(inode->i_sb)->journal->j_trans_id, - OCFS2_I(inode)->ip_created_trans)); - if (!ret) - OCFS2_I(inode)->ip_created_trans = 0; - spin_unlock(&trans_inc_lock); - return ret; + + return ocfs2_ci_is_new(INODE_CACHE(inode)); } -static inline void ocfs2_inode_set_new(struct ocfs2_super *osb, - struct inode *inode) +static inline void ocfs2_ci_set_new(struct ocfs2_super *osb, + struct ocfs2_caching_info *ci) { spin_lock(&trans_inc_lock); - OCFS2_I(inode)->ip_created_trans = osb->journal->j_trans_id; + ci->ci_created_trans = osb->journal->j_trans_id; spin_unlock(&trans_inc_lock); } /* Exported only for the journal struct init code in super.c. Do not call. */ +void ocfs2_orphan_scan_init(struct ocfs2_super *osb); +void ocfs2_orphan_scan_start(struct ocfs2_super *osb); +void ocfs2_orphan_scan_stop(struct ocfs2_super *osb); +void ocfs2_orphan_scan_exit(struct ocfs2_super *osb); + void ocfs2_complete_recovery(struct work_struct *work); void ocfs2_wait_for_recovery(struct ocfs2_super *osb); @@ -184,7 +200,6 @@ void ocfs2_complete_quota_recovery(struct ocfs2_super *osb); static inline void ocfs2_start_checkpoint(struct ocfs2_super *osb) { - atomic_set(&osb->needs_checkpoint, 1); wake_up(&osb->checkpoint_event); } @@ -195,17 +210,17 @@ static inline void ocfs2_checkpoint_inode(struct inode *inode) if (ocfs2_mount_local(osb)) return; - if (!ocfs2_inode_fully_checkpointed(inode)) { + if (!ocfs2_ci_fully_checkpointed(INODE_CACHE(inode))) { /* WARNING: This only kicks off a single * checkpoint. If someone races you and adds more * metadata to the journal, you won't know, and will - * wind up waiting *alot* longer than necessary. Right + * wind up waiting *a lot* longer than necessary. Right * now we only use this in clear_inode so that's * OK. */ ocfs2_start_checkpoint(osb); wait_event(osb->journal->j_checkpointed, - ocfs2_inode_fully_checkpointed(inode)); + ocfs2_ci_fully_checkpointed(INODE_CACHE(inode))); } } @@ -243,6 +258,17 @@ handle_t *ocfs2_start_trans(struct ocfs2_super *osb, int ocfs2_commit_trans(struct ocfs2_super *osb, handle_t *handle); int ocfs2_extend_trans(handle_t *handle, int nblocks); +int ocfs2_allocate_extend_trans(handle_t *handle, + int thresh); + +/* + * Define an arbitrary limit for the amount of data we will anticipate + * writing to any given transaction. For unbounded transactions such as + * fallocate(2) we can write more than this, but we always + * start off at the maximum transaction size and grow the transaction + * optimistically as we go. + */ +#define OCFS2_MAX_TRANS_DATA 64U /* * Create access is for when we get a newly created buffer and we're @@ -261,31 +287,34 @@ int ocfs2_extend_trans(handle_t *handle, int nblocks); /* ocfs2_inode */ -int ocfs2_journal_access_di(handle_t *handle, struct inode *inode, +int ocfs2_journal_access_di(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type); /* ocfs2_extent_block */ -int ocfs2_journal_access_eb(handle_t *handle, struct inode *inode, +int ocfs2_journal_access_eb(handle_t *handle, struct ocfs2_caching_info *ci, + struct buffer_head *bh, int type); +/* ocfs2_refcount_block */ +int ocfs2_journal_access_rb(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type); /* ocfs2_group_desc */ -int ocfs2_journal_access_gd(handle_t *handle, struct inode *inode, +int ocfs2_journal_access_gd(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type); /* ocfs2_xattr_block */ -int ocfs2_journal_access_xb(handle_t *handle, struct inode *inode, +int ocfs2_journal_access_xb(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type); /* quota blocks */ -int ocfs2_journal_access_dq(handle_t *handle, struct inode *inode, +int ocfs2_journal_access_dq(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type); /* dirblock */ -int ocfs2_journal_access_db(handle_t *handle, struct inode *inode, +int ocfs2_journal_access_db(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type); /* ocfs2_dx_root_block */ -int ocfs2_journal_access_dr(handle_t *handle, struct inode *inode, +int ocfs2_journal_access_dr(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type); /* ocfs2_dx_leaf */ -int ocfs2_journal_access_dl(handle_t *handle, struct inode *inode, +int ocfs2_journal_access_dl(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type); /* Anything that has no ecc */ -int ocfs2_journal_access(handle_t *handle, struct inode *inode, +int ocfs2_journal_access(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type); /* @@ -307,8 +336,7 @@ int ocfs2_journal_access(handle_t *handle, struct inode *inode, * <modify the bh> * ocfs2_journal_dirty(handle, bh); */ -int ocfs2_journal_dirty(handle_t *handle, - struct buffer_head *bh); +void ocfs2_journal_dirty(handle_t *handle, struct buffer_head *bh); /* * Credit Macros: @@ -325,20 +353,27 @@ int ocfs2_journal_dirty(handle_t *handle, /* extended attribute block update */ #define OCFS2_XATTR_BLOCK_UPDATE_CREDITS 1 +/* Update of a single quota block */ +#define OCFS2_QUOTA_BLOCK_UPDATE_CREDITS 1 + /* global quotafile inode update, data block */ -#define OCFS2_QINFO_WRITE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1) +#define OCFS2_QINFO_WRITE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + \ + OCFS2_QUOTA_BLOCK_UPDATE_CREDITS) +#define OCFS2_LOCAL_QINFO_WRITE_CREDITS OCFS2_QUOTA_BLOCK_UPDATE_CREDITS /* * The two writes below can accidentally see global info dirty due * to set_info() quotactl so make them prepared for the writes. */ /* quota data block, global info */ /* Write to local quota file */ -#define OCFS2_QWRITE_CREDITS (OCFS2_QINFO_WRITE_CREDITS + 1) +#define OCFS2_QWRITE_CREDITS (OCFS2_QINFO_WRITE_CREDITS + \ + OCFS2_QUOTA_BLOCK_UPDATE_CREDITS) /* global quota data block, local quota data block, global quota inode, * global quota info */ -#define OCFS2_QSYNC_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 3) +#define OCFS2_QSYNC_CREDITS (OCFS2_QINFO_WRITE_CREDITS + \ + 2 * OCFS2_QUOTA_BLOCK_UPDATE_CREDITS) static inline int ocfs2_quota_trans_credits(struct super_block *sb) { @@ -351,11 +386,6 @@ static inline int ocfs2_quota_trans_credits(struct super_block *sb) return credits; } -/* Number of credits needed for removing quota structure from file */ -int ocfs2_calc_qdel_credits(struct super_block *sb, int type); -/* Number of credits needed for initialization of new quota structure */ -int ocfs2_calc_qinit_credits(struct super_block *sb, int type); - /* group extend. inode update and last group update. */ #define OCFS2_GROUP_EXTEND_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1) @@ -385,9 +415,9 @@ static inline int ocfs2_remove_extent_credits(struct super_block *sb) ocfs2_quota_trans_credits(sb); } -/* data block for new dir/symlink, 2 for bitmap updates (bitmap fe + - * bitmap block for the new bit) dx_root update for free list */ -#define OCFS2_DIR_LINK_ADDITIONAL_CREDITS (1 + 2 + 1) +/* data block for new dir/symlink, allocation of directory block, dx_root + * update for free list */ +#define OCFS2_DIR_LINK_ADDITIONAL_CREDITS (1 + OCFS2_SUBALLOC_ALLOC + 1) static inline int ocfs2_add_dir_index_credits(struct super_block *sb) { @@ -421,10 +451,11 @@ static inline int ocfs2_mknod_credits(struct super_block *sb, int is_dir, #define OCFS2_SIMPLE_DIR_EXTEND_CREDITS (2) /* file update (nlink, etc) + directory mtime/ctime + dir entry block + quota - * update on dir + index leaf + dx root update for free list */ + * update on dir + index leaf + dx root update for free list + + * previous dirblock update in the free list */ static inline int ocfs2_link_credits(struct super_block *sb) { - return 2*OCFS2_INODE_UPDATE_CREDITS + 3 + + return 2*OCFS2_INODE_UPDATE_CREDITS + 4 + ocfs2_quota_trans_credits(sb); } @@ -470,14 +501,30 @@ static inline int ocfs2_calc_dxi_expand_credits(struct super_block *sb) return credits; } +/* inode update, new refcount block and its allocation credits. */ +#define OCFS2_REFCOUNT_TREE_CREATE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1 \ + + OCFS2_SUBALLOC_ALLOC) + +/* inode and the refcount block update. */ +#define OCFS2_REFCOUNT_TREE_SET_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1) + +/* + * inode and the refcount block update. + * It doesn't include the credits for sub alloc change. + * So if we need to free the bit, OCFS2_SUBALLOC_FREE needs to be added. + */ +#define OCFS2_REFCOUNT_TREE_REMOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1) + +/* 2 metadata alloc, 2 new blocks and root refcount block */ +#define OCFS2_EXPAND_REFCOUNT_TREE_CREDITS (OCFS2_SUBALLOC_ALLOC * 2 + 3) + /* * Please note that the caller must make sure that root_el is the root * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise * the result may be wrong. */ static inline int ocfs2_calc_extend_credits(struct super_block *sb, - struct ocfs2_extent_list *root_el, - u32 bits_wanted) + struct ocfs2_extent_list *root_el) { int bitmap_blocks, sysfile_bitmap_blocks, extent_blocks; @@ -525,6 +572,18 @@ static inline int ocfs2_calc_group_alloc_credits(struct super_block *sb, return blocks; } +/* + * Allocating a discontiguous block group requires the credits from + * ocfs2_calc_group_alloc_credits() as well as enough credits to fill + * the group descriptor's extent list. The caller already has started + * the transaction with ocfs2_calc_group_alloc_credits(). They extend + * it with these credits. + */ +static inline int ocfs2_calc_bg_discontig_credits(struct super_block *sb) +{ + return ocfs2_extent_recs_per_gd(sb); +} + static inline int ocfs2_calc_tree_trunc_credits(struct super_block *sb, unsigned int clusters_to_del, struct ocfs2_dinode *fe, @@ -567,4 +626,15 @@ static inline int ocfs2_begin_ordered_truncate(struct inode *inode, new_size); } +static inline void ocfs2_update_inode_fsync_trans(handle_t *handle, + struct inode *inode, + int datasync) +{ + struct ocfs2_inode_info *oi = OCFS2_I(inode); + + oi->i_sync_tid = handle->h_transaction->t_tid; + if (datasync) + oi->i_datasync_tid = handle->h_transaction->t_tid; +} + #endif /* OCFS2_JOURNAL_H */ diff --git a/fs/ocfs2/localalloc.c b/fs/ocfs2/localalloc.c index bac7e6abaf4..04401345562 100644 --- a/fs/ocfs2/localalloc.c +++ b/fs/ocfs2/localalloc.c @@ -29,7 +29,6 @@ #include <linux/highmem.h> #include <linux/bitops.h> -#define MLOG_MASK_PREFIX ML_DISK_ALLOC #include <cluster/masklog.h> #include "ocfs2.h" @@ -43,6 +42,7 @@ #include "suballoc.h" #include "super.h" #include "sysfile.h" +#include "ocfs2_trace.h" #include "buffer_head_io.h" @@ -52,7 +52,8 @@ static u32 ocfs2_local_alloc_count_bits(struct ocfs2_dinode *alloc); static int ocfs2_local_alloc_find_clear_bits(struct ocfs2_super *osb, struct ocfs2_dinode *alloc, - u32 numbits); + u32 *numbits, + struct ocfs2_alloc_reservation *resv); static void ocfs2_clear_local_alloc(struct ocfs2_dinode *alloc); @@ -74,6 +75,150 @@ static int ocfs2_local_alloc_new_window(struct ocfs2_super *osb, static int ocfs2_local_alloc_slide_window(struct ocfs2_super *osb, struct inode *local_alloc_inode); +/* + * ocfs2_la_default_mb() - determine a default size, in megabytes of + * the local alloc. + * + * Generally, we'd like to pick as large a local alloc as + * possible. Performance on large workloads tends to scale + * proportionally to la size. In addition to that, the reservations + * code functions more efficiently as it can reserve more windows for + * write. + * + * Some things work against us when trying to choose a large local alloc: + * + * - We need to ensure our sizing is picked to leave enough space in + * group descriptors for other allocations (such as block groups, + * etc). Picking default sizes which are a multiple of 4 could help + * - block groups are allocated in 2mb and 4mb chunks. + * + * - Likewise, we don't want to starve other nodes of bits on small + * file systems. This can easily be taken care of by limiting our + * default to a reasonable size (256M) on larger cluster sizes. + * + * - Some file systems can't support very large sizes - 4k and 8k in + * particular are limited to less than 128 and 256 megabytes respectively. + * + * The following reference table shows group descriptor and local + * alloc maximums at various cluster sizes (4k blocksize) + * + * csize: 4K group: 126M la: 121M + * csize: 8K group: 252M la: 243M + * csize: 16K group: 504M la: 486M + * csize: 32K group: 1008M la: 972M + * csize: 64K group: 2016M la: 1944M + * csize: 128K group: 4032M la: 3888M + * csize: 256K group: 8064M la: 7776M + * csize: 512K group: 16128M la: 15552M + * csize: 1024K group: 32256M la: 31104M + */ +#define OCFS2_LA_MAX_DEFAULT_MB 256 +#define OCFS2_LA_OLD_DEFAULT 8 +unsigned int ocfs2_la_default_mb(struct ocfs2_super *osb) +{ + unsigned int la_mb; + unsigned int gd_mb; + unsigned int la_max_mb; + unsigned int megs_per_slot; + struct super_block *sb = osb->sb; + + gd_mb = ocfs2_clusters_to_megabytes(osb->sb, + 8 * ocfs2_group_bitmap_size(sb, 0, osb->s_feature_incompat)); + + /* + * This takes care of files systems with very small group + * descriptors - 512 byte blocksize at cluster sizes lower + * than 16K and also 1k blocksize with 4k cluster size. + */ + if ((sb->s_blocksize == 512 && osb->s_clustersize <= 8192) + || (sb->s_blocksize == 1024 && osb->s_clustersize == 4096)) + return OCFS2_LA_OLD_DEFAULT; + + /* + * Leave enough room for some block groups and make the final + * value we work from a multiple of 4. + */ + gd_mb -= 16; + gd_mb &= 0xFFFFFFFB; + + la_mb = gd_mb; + + /* + * Keep window sizes down to a reasonable default + */ + if (la_mb > OCFS2_LA_MAX_DEFAULT_MB) { + /* + * Some clustersize / blocksize combinations will have + * given us a larger than OCFS2_LA_MAX_DEFAULT_MB + * default size, but get poor distribution when + * limited to exactly 256 megabytes. + * + * As an example, 16K clustersize at 4K blocksize + * gives us a cluster group size of 504M. Paring the + * local alloc size down to 256 however, would give us + * only one window and around 200MB left in the + * cluster group. Instead, find the first size below + * 256 which would give us an even distribution. + * + * Larger cluster group sizes actually work out pretty + * well when pared to 256, so we don't have to do this + * for any group that fits more than two + * OCFS2_LA_MAX_DEFAULT_MB windows. + */ + if (gd_mb > (2 * OCFS2_LA_MAX_DEFAULT_MB)) + la_mb = 256; + else { + unsigned int gd_mult = gd_mb; + + while (gd_mult > 256) + gd_mult = gd_mult >> 1; + + la_mb = gd_mult; + } + } + + megs_per_slot = osb->osb_clusters_at_boot / osb->max_slots; + megs_per_slot = ocfs2_clusters_to_megabytes(osb->sb, megs_per_slot); + /* Too many nodes, too few disk clusters. */ + if (megs_per_slot < la_mb) + la_mb = megs_per_slot; + + /* We can't store more bits than we can in a block. */ + la_max_mb = ocfs2_clusters_to_megabytes(osb->sb, + ocfs2_local_alloc_size(sb) * 8); + if (la_mb > la_max_mb) + la_mb = la_max_mb; + + return la_mb; +} + +void ocfs2_la_set_sizes(struct ocfs2_super *osb, int requested_mb) +{ + struct super_block *sb = osb->sb; + unsigned int la_default_mb = ocfs2_la_default_mb(osb); + unsigned int la_max_mb; + + la_max_mb = ocfs2_clusters_to_megabytes(sb, + ocfs2_local_alloc_size(sb) * 8); + + trace_ocfs2_la_set_sizes(requested_mb, la_max_mb, la_default_mb); + + if (requested_mb == -1) { + /* No user request - use defaults */ + osb->local_alloc_default_bits = + ocfs2_megabytes_to_clusters(sb, la_default_mb); + } else if (requested_mb > la_max_mb) { + /* Request is too big, we give the maximum available */ + osb->local_alloc_default_bits = + ocfs2_megabytes_to_clusters(sb, la_max_mb); + } else { + osb->local_alloc_default_bits = + ocfs2_megabytes_to_clusters(sb, requested_mb); + } + + osb->local_alloc_bits = osb->local_alloc_default_bits; +} + static inline int ocfs2_la_state_enabled(struct ocfs2_super *osb) { return (osb->local_alloc_state == OCFS2_LA_THROTTLED || @@ -130,8 +275,8 @@ int ocfs2_alloc_should_use_local(struct ocfs2_super *osb, u64 bits) ret = 1; bail: - mlog(0, "state=%d, bits=%llu, la_bits=%d, ret=%d\n", - osb->local_alloc_state, (unsigned long long)bits, la_bits, ret); + trace_ocfs2_alloc_should_use_local( + (unsigned long long)bits, osb->local_alloc_state, la_bits, ret); spin_unlock(&osb->osb_lock); return ret; } @@ -145,8 +290,6 @@ int ocfs2_load_local_alloc(struct ocfs2_super *osb) struct inode *inode = NULL; struct ocfs2_local_alloc *la; - mlog_entry_void(); - if (osb->local_alloc_bits == 0) goto bail; @@ -156,7 +299,7 @@ int ocfs2_load_local_alloc(struct ocfs2_super *osb) osb->local_alloc_bits, (osb->bitmap_cpg - 1)); osb->local_alloc_bits = ocfs2_megabytes_to_clusters(osb->sb, - OCFS2_DEFAULT_LOCAL_ALLOC_SIZE); + ocfs2_la_default_mb(osb)); } /* read the alloc off disk */ @@ -218,9 +361,10 @@ bail: if (inode) iput(inode); - mlog(0, "Local alloc window bits = %d\n", osb->local_alloc_bits); + trace_ocfs2_load_local_alloc(osb->local_alloc_bits); - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -242,8 +386,6 @@ void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb) struct ocfs2_dinode *alloc_copy = NULL; struct ocfs2_dinode *alloc = NULL; - mlog_entry_void(); - cancel_delayed_work(&osb->la_enable_wq); flush_workqueue(ocfs2_wq); @@ -262,6 +404,8 @@ void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb) osb->local_alloc_state = OCFS2_LA_DISABLED; + ocfs2_resmap_uninit(&osb->osb_la_resmap); + main_bm_inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE, OCFS2_INVALID_SLOT); @@ -297,20 +441,15 @@ void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb) } memcpy(alloc_copy, alloc, bh->b_size); - status = ocfs2_journal_access_di(handle, local_alloc_inode, bh, - OCFS2_JOURNAL_ACCESS_WRITE); + status = ocfs2_journal_access_di(handle, INODE_CACHE(local_alloc_inode), + bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); goto out_commit; } ocfs2_clear_local_alloc(alloc); - - status = ocfs2_journal_dirty(handle, bh); - if (status < 0) { - mlog_errno(status); - goto out_commit; - } + ocfs2_journal_dirty(handle, bh); brelse(bh); osb->local_alloc_bh = NULL; @@ -337,10 +476,7 @@ out: if (local_alloc_inode) iput(local_alloc_inode); - if (alloc_copy) - kfree(alloc_copy); - - mlog_exit_void(); + kfree(alloc_copy); } /* @@ -359,7 +495,7 @@ int ocfs2_begin_local_alloc_recovery(struct ocfs2_super *osb, struct inode *inode = NULL; struct ocfs2_dinode *alloc; - mlog_entry("(slot_num = %d)\n", slot_num); + trace_ocfs2_begin_local_alloc_recovery(slot_num); *alloc_copy = NULL; @@ -392,12 +528,12 @@ int ocfs2_begin_local_alloc_recovery(struct ocfs2_super *osb, ocfs2_clear_local_alloc(alloc); ocfs2_compute_meta_ecc(osb->sb, alloc_bh->b_data, &alloc->i_check); - status = ocfs2_write_block(osb, alloc_bh, inode); + status = ocfs2_write_block(osb, alloc_bh, INODE_CACHE(inode)); if (status < 0) mlog_errno(status); bail: - if ((status < 0) && (*alloc_copy)) { + if (status < 0) { kfree(*alloc_copy); *alloc_copy = NULL; } @@ -409,7 +545,8 @@ bail: iput(inode); } - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -427,8 +564,6 @@ int ocfs2_complete_local_alloc_recovery(struct ocfs2_super *osb, struct buffer_head *main_bm_bh = NULL; struct inode *main_bm_inode; - mlog_entry_void(); - main_bm_inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE, OCFS2_INVALID_SLOT); @@ -476,51 +611,12 @@ out_mutex: out: if (!status) - ocfs2_init_inode_steal_slot(osb); - mlog_exit(status); + ocfs2_init_steal_slots(osb); + if (status) + mlog_errno(status); return status; } -/* Check to see if the local alloc window is within ac->ac_max_block */ -static int ocfs2_local_alloc_in_range(struct inode *inode, - struct ocfs2_alloc_context *ac, - u32 bits_wanted) -{ - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - struct ocfs2_dinode *alloc; - struct ocfs2_local_alloc *la; - int start; - u64 block_off; - - if (!ac->ac_max_block) - return 1; - - alloc = (struct ocfs2_dinode *) osb->local_alloc_bh->b_data; - la = OCFS2_LOCAL_ALLOC(alloc); - - start = ocfs2_local_alloc_find_clear_bits(osb, alloc, bits_wanted); - if (start == -1) { - mlog_errno(-ENOSPC); - return 0; - } - - /* - * Converting (bm_off + start + bits_wanted) to blocks gives us - * the blkno just past our actual allocation. This is perfect - * to compare with ac_max_block. - */ - block_off = ocfs2_clusters_to_blocks(inode->i_sb, - le32_to_cpu(la->la_bm_off) + - start + bits_wanted); - mlog(0, "Checking %llu against %llu\n", - (unsigned long long)block_off, - (unsigned long long)ac->ac_max_block); - if (block_off > ac->ac_max_block) - return 0; - - return 1; -} - /* * make sure we've got at least bits_wanted contiguous bits in the * local alloc. You lose them when you drop i_mutex. @@ -537,8 +633,6 @@ int ocfs2_reserve_local_alloc_bits(struct ocfs2_super *osb, struct inode *local_alloc_inode; unsigned int free_bits; - mlog_entry_void(); - BUG_ON(!ac); local_alloc_inode = @@ -609,21 +703,6 @@ int ocfs2_reserve_local_alloc_bits(struct ocfs2_super *osb, goto bail; } - if (ac->ac_max_block) - mlog(0, "Calling in_range for max block %llu\n", - (unsigned long long)ac->ac_max_block); - - if (!ocfs2_local_alloc_in_range(local_alloc_inode, ac, - bits_wanted)) { - /* - * The window is outside ac->ac_max_block. - * This errno tells the caller to keep localalloc enabled - * but to get the allocation from the main bitmap. - */ - status = -EFBIG; - goto bail; - } - ac->ac_inode = local_alloc_inode; /* We should never use localalloc from another slot */ ac->ac_alloc_slot = osb->slot_num; @@ -637,10 +716,12 @@ bail: iput(local_alloc_inode); } - mlog(0, "bits=%d, slot=%d, ret=%d\n", bits_wanted, osb->slot_num, - status); + trace_ocfs2_reserve_local_alloc_bits( + (unsigned long long)ac->ac_max_block, + bits_wanted, osb->slot_num, status); - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -657,14 +738,14 @@ int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb, struct ocfs2_dinode *alloc; struct ocfs2_local_alloc *la; - mlog_entry_void(); BUG_ON(ac->ac_which != OCFS2_AC_USE_LOCAL); local_alloc_inode = ac->ac_inode; alloc = (struct ocfs2_dinode *) osb->local_alloc_bh->b_data; la = OCFS2_LOCAL_ALLOC(alloc); - start = ocfs2_local_alloc_find_clear_bits(osb, alloc, bits_wanted); + start = ocfs2_local_alloc_find_clear_bits(osb, alloc, &bits_wanted, + ac->ac_resv); if (start == -1) { /* TODO: Shouldn't we just BUG here? */ status = -ENOSPC; @@ -674,11 +755,10 @@ int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb, bitmap = la->la_bitmap; *bit_off = le32_to_cpu(la->la_bm_off) + start; - /* local alloc is always contiguous by nature -- we never - * delete bits from it! */ *num_bits = bits_wanted; - status = ocfs2_journal_access_di(handle, local_alloc_inode, + status = ocfs2_journal_access_di(handle, + INODE_CACHE(local_alloc_inode), osb->local_alloc_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { @@ -686,55 +766,114 @@ int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb, goto bail; } + ocfs2_resmap_claimed_bits(&osb->osb_la_resmap, ac->ac_resv, start, + bits_wanted); + while(bits_wanted--) ocfs2_set_bit(start++, bitmap); le32_add_cpu(&alloc->id1.bitmap1.i_used, *num_bits); + ocfs2_journal_dirty(handle, osb->local_alloc_bh); + +bail: + if (status) + mlog_errno(status); + return status; +} - status = ocfs2_journal_dirty(handle, osb->local_alloc_bh); +int ocfs2_free_local_alloc_bits(struct ocfs2_super *osb, + handle_t *handle, + struct ocfs2_alloc_context *ac, + u32 bit_off, + u32 num_bits) +{ + int status, start; + u32 clear_bits; + struct inode *local_alloc_inode; + void *bitmap; + struct ocfs2_dinode *alloc; + struct ocfs2_local_alloc *la; + + BUG_ON(ac->ac_which != OCFS2_AC_USE_LOCAL); + + local_alloc_inode = ac->ac_inode; + alloc = (struct ocfs2_dinode *) osb->local_alloc_bh->b_data; + la = OCFS2_LOCAL_ALLOC(alloc); + + bitmap = la->la_bitmap; + start = bit_off - le32_to_cpu(la->la_bm_off); + clear_bits = num_bits; + + status = ocfs2_journal_access_di(handle, + INODE_CACHE(local_alloc_inode), + osb->local_alloc_bh, + OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); goto bail; } - status = 0; + while (clear_bits--) + ocfs2_clear_bit(start++, bitmap); + + le32_add_cpu(&alloc->id1.bitmap1.i_used, -num_bits); + ocfs2_journal_dirty(handle, osb->local_alloc_bh); + bail: - mlog_exit(status); return status; } static u32 ocfs2_local_alloc_count_bits(struct ocfs2_dinode *alloc) { - int i; - u8 *buffer; - u32 count = 0; + u32 count; struct ocfs2_local_alloc *la = OCFS2_LOCAL_ALLOC(alloc); - mlog_entry_void(); - - buffer = la->la_bitmap; - for (i = 0; i < le16_to_cpu(la->la_size); i++) - count += hweight8(buffer[i]); + count = memweight(la->la_bitmap, le16_to_cpu(la->la_size)); - mlog_exit(count); + trace_ocfs2_local_alloc_count_bits(count); return count; } static int ocfs2_local_alloc_find_clear_bits(struct ocfs2_super *osb, - struct ocfs2_dinode *alloc, - u32 numbits) + struct ocfs2_dinode *alloc, + u32 *numbits, + struct ocfs2_alloc_reservation *resv) { int numfound, bitoff, left, startoff, lastzero; + int local_resv = 0; + struct ocfs2_alloc_reservation r; void *bitmap = NULL; - - mlog_entry("(numbits wanted = %u)\n", numbits); + struct ocfs2_reservation_map *resmap = &osb->osb_la_resmap; if (!alloc->id1.bitmap1.i_total) { - mlog(0, "No bits in my window!\n"); bitoff = -1; goto bail; } + if (!resv) { + local_resv = 1; + ocfs2_resv_init_once(&r); + ocfs2_resv_set_type(&r, OCFS2_RESV_FLAG_TMP); + resv = &r; + } + + numfound = *numbits; + if (ocfs2_resmap_resv_bits(resmap, resv, &bitoff, &numfound) == 0) { + if (numfound < *numbits) + *numbits = numfound; + goto bail; + } + + /* + * Code error. While reservations are enabled, local + * allocation should _always_ go through them. + */ + BUG_ON(osb->osb_resv_level != 0); + + /* + * Reservations are disabled. Handle this the old way. + */ + bitmap = OCFS2_LOCAL_ALLOC(alloc)->la_bitmap; numfound = bitoff = startoff = 0; @@ -760,22 +899,27 @@ static int ocfs2_local_alloc_find_clear_bits(struct ocfs2_super *osb, startoff = bitoff+1; } /* we got everything we needed */ - if (numfound == numbits) { + if (numfound == *numbits) { /* mlog(0, "Found it all!\n"); */ break; } } - mlog(0, "Exiting loop, bitoff = %d, numfound = %d\n", bitoff, - numfound); + trace_ocfs2_local_alloc_find_clear_bits_search_bitmap(bitoff, numfound); - if (numfound == numbits) + if (numfound == *numbits) bitoff = startoff - numfound; else bitoff = -1; bail: - mlog_exit(bitoff); + if (local_resv) + ocfs2_resv_discard(resmap, resv); + + trace_ocfs2_local_alloc_find_clear_bits(*numbits, + le32_to_cpu(alloc->id1.bitmap1.i_total), + bitoff, numfound); + return bitoff; } @@ -783,15 +927,12 @@ static void ocfs2_clear_local_alloc(struct ocfs2_dinode *alloc) { struct ocfs2_local_alloc *la = OCFS2_LOCAL_ALLOC(alloc); int i; - mlog_entry_void(); alloc->id1.bitmap1.i_total = 0; alloc->id1.bitmap1.i_used = 0; la->la_bm_off = 0; for(i = 0; i < le16_to_cpu(la->la_size); i++) la->la_bitmap[i] = 0; - - mlog_exit_void(); } #if 0 @@ -832,18 +973,16 @@ static int ocfs2_sync_local_to_main(struct ocfs2_super *osb, void *bitmap; struct ocfs2_local_alloc *la = OCFS2_LOCAL_ALLOC(alloc); - mlog_entry("total = %u, used = %u\n", - le32_to_cpu(alloc->id1.bitmap1.i_total), - le32_to_cpu(alloc->id1.bitmap1.i_used)); + trace_ocfs2_sync_local_to_main( + le32_to_cpu(alloc->id1.bitmap1.i_total), + le32_to_cpu(alloc->id1.bitmap1.i_used)); if (!alloc->id1.bitmap1.i_total) { - mlog(0, "nothing to sync!\n"); goto bail; } if (le32_to_cpu(alloc->id1.bitmap1.i_used) == le32_to_cpu(alloc->id1.bitmap1.i_total)) { - mlog(0, "all bits were taken!\n"); goto bail; } @@ -865,14 +1004,15 @@ static int ocfs2_sync_local_to_main(struct ocfs2_super *osb, ocfs2_clusters_to_blocks(osb->sb, start - count); - mlog(0, "freeing %u bits starting at local alloc bit " - "%u (la_start_blk = %llu, blkno = %llu)\n", + trace_ocfs2_sync_local_to_main_free( count, start - count, (unsigned long long)la_start_blk, (unsigned long long)blkno); - status = ocfs2_free_clusters(handle, main_bm_inode, - main_bm_bh, blkno, count); + status = ocfs2_release_clusters(handle, + main_bm_inode, + main_bm_bh, blkno, + count); if (status < 0) { mlog_errno(status); goto bail; @@ -885,7 +1025,8 @@ static int ocfs2_sync_local_to_main(struct ocfs2_super *osb, } bail: - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -984,7 +1125,6 @@ static int ocfs2_local_alloc_reserve_for_window(struct ocfs2_super *osb, retry_enospc: (*ac)->ac_bits_wanted = osb->local_alloc_bits; - status = ocfs2_reserve_cluster_bitmap_bits(osb, *ac); if (status == -ENOSPC) { if (ocfs2_recalc_la_window(osb, OCFS2_LA_EVENT_ENOSPC) == @@ -1011,7 +1151,8 @@ bail: *ac = NULL; } - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -1027,17 +1168,12 @@ static int ocfs2_local_alloc_new_window(struct ocfs2_super *osb, struct ocfs2_dinode *alloc = NULL; struct ocfs2_local_alloc *la; - mlog_entry_void(); - alloc = (struct ocfs2_dinode *) osb->local_alloc_bh->b_data; la = OCFS2_LOCAL_ALLOC(alloc); - if (alloc->id1.bitmap1.i_total) - mlog(0, "asking me to alloc a new window over a non-empty " - "one\n"); - - mlog(0, "Allocating %u clusters for a new window.\n", - osb->local_alloc_bits); + trace_ocfs2_local_alloc_new_window( + le32_to_cpu(alloc->id1.bitmap1.i_total), + osb->local_alloc_bits); /* Instruct the allocation code to try the most recently used * cluster group. We'll re-record the group used this pass @@ -1047,7 +1183,7 @@ static int ocfs2_local_alloc_new_window(struct ocfs2_super *osb, /* we used the generic suballoc reserve function, but we set * everything up nicely, so there's no reason why we can't use * the more specific cluster api to claim bits. */ - status = ocfs2_claim_clusters(osb, handle, ac, osb->local_alloc_bits, + status = ocfs2_claim_clusters(handle, ac, osb->local_alloc_bits, &cluster_off, &cluster_count); if (status == -ENOSPC) { retry_enospc: @@ -1060,7 +1196,8 @@ retry_enospc: OCFS2_LA_DISABLED) goto bail; - status = ocfs2_claim_clusters(osb, handle, ac, + ac->ac_bits_wanted = osb->local_alloc_bits; + status = ocfs2_claim_clusters(handle, ac, osb->local_alloc_bits, &cluster_off, &cluster_count); @@ -1095,13 +1232,16 @@ retry_enospc: memset(OCFS2_LOCAL_ALLOC(alloc)->la_bitmap, 0, le16_to_cpu(la->la_size)); - mlog(0, "New window allocated:\n"); - mlog(0, "window la_bm_off = %u\n", - OCFS2_LOCAL_ALLOC(alloc)->la_bm_off); - mlog(0, "window bits = %u\n", le32_to_cpu(alloc->id1.bitmap1.i_total)); + ocfs2_resmap_restart(&osb->osb_la_resmap, cluster_count, + OCFS2_LOCAL_ALLOC(alloc)->la_bitmap); + + trace_ocfs2_local_alloc_new_window_result( + OCFS2_LOCAL_ALLOC(alloc)->la_bm_off, + le32_to_cpu(alloc->id1.bitmap1.i_total)); bail: - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -1118,8 +1258,6 @@ static int ocfs2_local_alloc_slide_window(struct ocfs2_super *osb, struct ocfs2_dinode *alloc_copy = NULL; struct ocfs2_alloc_context *ac = NULL; - mlog_entry_void(); - ocfs2_recalc_la_window(osb, OCFS2_LA_EVENT_SLIDE); /* This will lock the main bitmap for us. */ @@ -1156,7 +1294,8 @@ static int ocfs2_local_alloc_slide_window(struct ocfs2_super *osb, } memcpy(alloc_copy, alloc, osb->local_alloc_bh->b_size); - status = ocfs2_journal_access_di(handle, local_alloc_inode, + status = ocfs2_journal_access_di(handle, + INODE_CACHE(local_alloc_inode), osb->local_alloc_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { @@ -1165,12 +1304,7 @@ static int ocfs2_local_alloc_slide_window(struct ocfs2_super *osb, } ocfs2_clear_local_alloc(alloc); - - status = ocfs2_journal_dirty(handle, osb->local_alloc_bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } + ocfs2_journal_dirty(handle, osb->local_alloc_bh); status = ocfs2_sync_local_to_main(osb, handle, alloc_copy, main_bm_inode, main_bm_bh); @@ -1188,7 +1322,6 @@ static int ocfs2_local_alloc_slide_window(struct ocfs2_super *osb, atomic_inc(&osb->alloc_stats.moves); - status = 0; bail: if (handle) ocfs2_commit_trans(osb, handle); @@ -1198,13 +1331,13 @@ bail: if (main_bm_inode) iput(main_bm_inode); - if (alloc_copy) - kfree(alloc_copy); + kfree(alloc_copy); if (ac) ocfs2_free_alloc_context(ac); - mlog_exit(status); + if (status) + mlog_errno(status); return status; } diff --git a/fs/ocfs2/localalloc.h b/fs/ocfs2/localalloc.h index ac5ea9f8665..44a7d1fb2de 100644 --- a/fs/ocfs2/localalloc.h +++ b/fs/ocfs2/localalloc.h @@ -30,6 +30,9 @@ int ocfs2_load_local_alloc(struct ocfs2_super *osb); void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb); +void ocfs2_la_set_sizes(struct ocfs2_super *osb, int requested_mb); +unsigned int ocfs2_la_default_mb(struct ocfs2_super *osb); + int ocfs2_begin_local_alloc_recovery(struct ocfs2_super *osb, int node_num, struct ocfs2_dinode **alloc_copy); @@ -52,6 +55,12 @@ int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb, u32 *bit_off, u32 *num_bits); +int ocfs2_free_local_alloc_bits(struct ocfs2_super *osb, + handle_t *handle, + struct ocfs2_alloc_context *ac, + u32 bit_off, + u32 num_bits); + void ocfs2_local_alloc_seen_free_bits(struct ocfs2_super *osb, unsigned int num_clusters); void ocfs2_la_enable_worker(struct work_struct *work); diff --git a/fs/ocfs2/locks.c b/fs/ocfs2/locks.c index 544ac624517..6b6d092b099 100644 --- a/fs/ocfs2/locks.c +++ b/fs/ocfs2/locks.c @@ -26,7 +26,6 @@ #include <linux/fs.h> #include <linux/fcntl.h> -#define MLOG_MASK_PREFIX ML_INODE #include <cluster/masklog.h> #include "ocfs2.h" @@ -83,6 +82,8 @@ static int ocfs2_do_flock(struct file *file, struct inode *inode, } ret = flock_lock_file_wait(file, fl); + if (ret) + ocfs2_file_unlock(file); out: mutex_unlock(&fp->fp_mutex); @@ -133,7 +134,7 @@ int ocfs2_lock(struct file *file, int cmd, struct file_lock *fl) if (!(fl->fl_flags & FL_POSIX)) return -ENOLCK; - if (__mandatory_lock(inode)) + if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK) return -ENOLCK; return ocfs2_plock(osb->cconn, OCFS2_I(inode)->ip_blkno, file, cmd, fl); diff --git a/fs/ocfs2/mmap.c b/fs/ocfs2/mmap.c index b606496b72e..10d66c75cec 100644 --- a/fs/ocfs2/mmap.c +++ b/fs/ocfs2/mmap.c @@ -25,14 +25,12 @@ #include <linux/fs.h> #include <linux/types.h> -#include <linux/slab.h> #include <linux/highmem.h> #include <linux/pagemap.h> #include <linux/uio.h> #include <linux/signal.h> #include <linux/rbtree.h> -#define MLOG_MASK_PREFIX ML_FILE_IO #include <cluster/masklog.h> #include "ocfs2.h" @@ -42,52 +40,29 @@ #include "file.h" #include "inode.h" #include "mmap.h" +#include "super.h" +#include "ocfs2_trace.h" -static inline int ocfs2_vm_op_block_sigs(sigset_t *blocked, sigset_t *oldset) -{ - /* The best way to deal with signals in the vm path is - * to block them upfront, rather than allowing the - * locking paths to return -ERESTARTSYS. */ - sigfillset(blocked); - - /* We should technically never get a bad return value - * from sigprocmask */ - return sigprocmask(SIG_BLOCK, blocked, oldset); -} - -static inline int ocfs2_vm_op_unblock_sigs(sigset_t *oldset) -{ - return sigprocmask(SIG_SETMASK, oldset, NULL); -} static int ocfs2_fault(struct vm_area_struct *area, struct vm_fault *vmf) { - sigset_t blocked, oldset; - int error, ret; - - mlog_entry("(area=%p, page offset=%lu)\n", area, vmf->pgoff); - - error = ocfs2_vm_op_block_sigs(&blocked, &oldset); - if (error < 0) { - mlog_errno(error); - ret = VM_FAULT_SIGBUS; - goto out; - } + sigset_t oldset; + int ret; + ocfs2_block_signals(&oldset); ret = filemap_fault(area, vmf); + ocfs2_unblock_signals(&oldset); - error = ocfs2_vm_op_unblock_sigs(&oldset); - if (error < 0) - mlog_errno(error); -out: - mlog_exit_ptr(vmf->page); + trace_ocfs2_fault(OCFS2_I(area->vm_file->f_mapping->host)->ip_blkno, + area, vmf->page, vmf->pgoff); return ret; } -static int __ocfs2_page_mkwrite(struct inode *inode, struct buffer_head *di_bh, +static int __ocfs2_page_mkwrite(struct file *file, struct buffer_head *di_bh, struct page *page) { - int ret; + int ret = VM_FAULT_NOPAGE; + struct inode *inode = file_inode(file); struct address_space *mapping = inode->i_mapping; loff_t pos = page_offset(page); unsigned int len = PAGE_CACHE_SIZE; @@ -96,30 +71,25 @@ static int __ocfs2_page_mkwrite(struct inode *inode, struct buffer_head *di_bh, void *fsdata; loff_t size = i_size_read(inode); - /* - * Another node might have truncated while we were waiting on - * cluster locks. - */ - last_index = size >> PAGE_CACHE_SHIFT; - if (page->index > last_index) { - ret = -EINVAL; - goto out; - } + last_index = (size - 1) >> PAGE_CACHE_SHIFT; /* - * The i_size check above doesn't catch the case where nodes - * truncated and then re-extended the file. We'll re-check the - * page mapping after taking the page lock inside of - * ocfs2_write_begin_nolock(). + * There are cases that lead to the page no longer bebongs to the + * mapping. + * 1) pagecache truncates locally due to memory pressure. + * 2) pagecache truncates when another is taking EX lock against + * inode lock. see ocfs2_data_convert_worker. + * + * The i_size check doesn't catch the case where nodes truncated and + * then re-extended the file. We'll re-check the page mapping after + * taking the page lock inside of ocfs2_write_begin_nolock(). + * + * Let VM retry with these cases. */ - if (!PageUptodate(page) || page->mapping != inode->i_mapping) { - /* - * the page has been umapped in ocfs2_data_downconvert_worker. - * So return 0 here and let VFS retry. - */ - ret = 0; + if ((page->mapping != inode->i_mapping) || + (!PageUptodate(page)) || + (page_offset(page) >= size)) goto out; - } /* * Call ocfs2_write_begin() and ocfs2_write_end() to take @@ -132,24 +102,28 @@ static int __ocfs2_page_mkwrite(struct inode *inode, struct buffer_head *di_bh, * because the "write" would invalidate their data. */ if (page->index == last_index) - len = size & ~PAGE_CACHE_MASK; + len = ((size - 1) & ~PAGE_CACHE_MASK) + 1; - ret = ocfs2_write_begin_nolock(mapping, pos, len, 0, &locked_page, + ret = ocfs2_write_begin_nolock(file, mapping, pos, len, 0, &locked_page, &fsdata, di_bh, page); if (ret) { if (ret != -ENOSPC) mlog_errno(ret); + if (ret == -ENOMEM) + ret = VM_FAULT_OOM; + else + ret = VM_FAULT_SIGBUS; goto out; } - ret = ocfs2_write_end_nolock(mapping, pos, len, len, locked_page, - fsdata); - if (ret < 0) { - mlog_errno(ret); + if (!locked_page) { + ret = VM_FAULT_NOPAGE; goto out; } + ret = ocfs2_write_end_nolock(mapping, pos, len, len, locked_page, + fsdata); BUG_ON(ret != len); - ret = 0; + ret = VM_FAULT_LOCKED; out: return ret; } @@ -157,16 +131,13 @@ out: static int ocfs2_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) { struct page *page = vmf->page; - struct inode *inode = vma->vm_file->f_path.dentry->d_inode; + struct inode *inode = file_inode(vma->vm_file); struct buffer_head *di_bh = NULL; - sigset_t blocked, oldset; - int ret, ret2; + sigset_t oldset; + int ret; - ret = ocfs2_vm_op_block_sigs(&blocked, &oldset); - if (ret < 0) { - mlog_errno(ret); - return ret; - } + sb_start_pagefault(inode->i_sb); + ocfs2_block_signals(&oldset); /* * The cluster locks taken will block a truncate from another @@ -186,7 +157,7 @@ static int ocfs2_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) */ down_write(&OCFS2_I(inode)->ip_alloc_sem); - ret = __ocfs2_page_mkwrite(inode, di_bh, page); + ret = __ocfs2_page_mkwrite(vma->vm_file, di_bh, page); up_write(&OCFS2_I(inode)->ip_alloc_sem); @@ -194,33 +165,30 @@ static int ocfs2_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) ocfs2_inode_unlock(inode, 1); out: - ret2 = ocfs2_vm_op_unblock_sigs(&oldset); - if (ret2 < 0) - mlog_errno(ret2); - if (ret) - ret = VM_FAULT_SIGBUS; + ocfs2_unblock_signals(&oldset); + sb_end_pagefault(inode->i_sb); return ret; } -static struct vm_operations_struct ocfs2_file_vm_ops = { +static const struct vm_operations_struct ocfs2_file_vm_ops = { .fault = ocfs2_fault, .page_mkwrite = ocfs2_page_mkwrite, + .remap_pages = generic_file_remap_pages, }; int ocfs2_mmap(struct file *file, struct vm_area_struct *vma) { int ret = 0, lock_level = 0; - ret = ocfs2_inode_lock_atime(file->f_dentry->d_inode, - file->f_vfsmnt, &lock_level); + ret = ocfs2_inode_lock_atime(file_inode(file), + file->f_path.mnt, &lock_level); if (ret < 0) { mlog_errno(ret); goto out; } - ocfs2_inode_unlock(file->f_dentry->d_inode, lock_level); + ocfs2_inode_unlock(file_inode(file), lock_level); out: vma->vm_ops = &ocfs2_file_vm_ops; - vma->vm_flags |= VM_CAN_NONLINEAR; return 0; } diff --git a/fs/ocfs2/move_extents.c b/fs/ocfs2/move_extents.c new file mode 100644 index 00000000000..599eb4c4c8b --- /dev/null +++ b/fs/ocfs2/move_extents.c @@ -0,0 +1,1078 @@ +/* -*- mode: c; c-basic-offset: 8; -*- + * vim: noexpandtab sw=8 ts=8 sts=0: + * + * move_extents.c + * + * Copyright (C) 2011 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License version 2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + */ +#include <linux/fs.h> +#include <linux/types.h> +#include <linux/mount.h> +#include <linux/swap.h> + +#include <cluster/masklog.h> + +#include "ocfs2.h" +#include "ocfs2_ioctl.h" + +#include "alloc.h" +#include "aops.h" +#include "dlmglue.h" +#include "extent_map.h" +#include "inode.h" +#include "journal.h" +#include "suballoc.h" +#include "uptodate.h" +#include "super.h" +#include "dir.h" +#include "buffer_head_io.h" +#include "sysfile.h" +#include "refcounttree.h" +#include "move_extents.h" + +struct ocfs2_move_extents_context { + struct inode *inode; + struct file *file; + int auto_defrag; + int partial; + int credits; + u32 new_phys_cpos; + u32 clusters_moved; + u64 refcount_loc; + struct ocfs2_move_extents *range; + struct ocfs2_extent_tree et; + struct ocfs2_alloc_context *meta_ac; + struct ocfs2_alloc_context *data_ac; + struct ocfs2_cached_dealloc_ctxt dealloc; +}; + +static int __ocfs2_move_extent(handle_t *handle, + struct ocfs2_move_extents_context *context, + u32 cpos, u32 len, u32 p_cpos, u32 new_p_cpos, + int ext_flags) +{ + int ret = 0, index; + struct inode *inode = context->inode; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_extent_rec *rec, replace_rec; + struct ocfs2_path *path = NULL; + struct ocfs2_extent_list *el; + u64 ino = ocfs2_metadata_cache_owner(context->et.et_ci); + u64 old_blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cpos); + + ret = ocfs2_duplicate_clusters_by_page(handle, inode, cpos, + p_cpos, new_p_cpos, len); + if (ret) { + mlog_errno(ret); + goto out; + } + + memset(&replace_rec, 0, sizeof(replace_rec)); + replace_rec.e_cpos = cpu_to_le32(cpos); + replace_rec.e_leaf_clusters = cpu_to_le16(len); + replace_rec.e_blkno = cpu_to_le64(ocfs2_clusters_to_blocks(inode->i_sb, + new_p_cpos)); + + path = ocfs2_new_path_from_et(&context->et); + if (!path) { + ret = -ENOMEM; + mlog_errno(ret); + goto out; + } + + ret = ocfs2_find_path(INODE_CACHE(inode), path, cpos); + if (ret) { + mlog_errno(ret); + goto out; + } + + el = path_leaf_el(path); + + index = ocfs2_search_extent_list(el, cpos); + if (index == -1 || index >= le16_to_cpu(el->l_next_free_rec)) { + ocfs2_error(inode->i_sb, + "Inode %llu has an extent at cpos %u which can no " + "longer be found.\n", + (unsigned long long)ino, cpos); + ret = -EROFS; + goto out; + } + + rec = &el->l_recs[index]; + + BUG_ON(ext_flags != rec->e_flags); + /* + * after moving/defraging to new location, the extent is not going + * to be refcounted anymore. + */ + replace_rec.e_flags = ext_flags & ~OCFS2_EXT_REFCOUNTED; + + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), + context->et.et_root_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_split_extent(handle, &context->et, path, index, + &replace_rec, context->meta_ac, + &context->dealloc); + if (ret) { + mlog_errno(ret); + goto out; + } + + ocfs2_journal_dirty(handle, context->et.et_root_bh); + + context->new_phys_cpos = new_p_cpos; + + /* + * need I to append truncate log for old clusters? + */ + if (old_blkno) { + if (ext_flags & OCFS2_EXT_REFCOUNTED) + ret = ocfs2_decrease_refcount(inode, handle, + ocfs2_blocks_to_clusters(osb->sb, + old_blkno), + len, context->meta_ac, + &context->dealloc, 1); + else + ret = ocfs2_truncate_log_append(osb, handle, + old_blkno, len); + } + + ocfs2_update_inode_fsync_trans(handle, inode, 0); +out: + ocfs2_free_path(path); + return ret; +} + +/* + * lock allocators, and reserving appropriate number of bits for + * meta blocks and data clusters. + * + * in some cases, we don't need to reserve clusters, just let data_ac + * be NULL. + */ +static int ocfs2_lock_allocators_move_extents(struct inode *inode, + struct ocfs2_extent_tree *et, + u32 clusters_to_move, + u32 extents_to_split, + struct ocfs2_alloc_context **meta_ac, + struct ocfs2_alloc_context **data_ac, + int extra_blocks, + int *credits) +{ + int ret, num_free_extents; + unsigned int max_recs_needed = 2 * extents_to_split + clusters_to_move; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + + num_free_extents = ocfs2_num_free_extents(osb, et); + if (num_free_extents < 0) { + ret = num_free_extents; + mlog_errno(ret); + goto out; + } + + if (!num_free_extents || + (ocfs2_sparse_alloc(osb) && num_free_extents < max_recs_needed)) + extra_blocks += ocfs2_extend_meta_needed(et->et_root_el); + + ret = ocfs2_reserve_new_metadata_blocks(osb, extra_blocks, meta_ac); + if (ret) { + mlog_errno(ret); + goto out; + } + + if (data_ac) { + ret = ocfs2_reserve_clusters(osb, clusters_to_move, data_ac); + if (ret) { + mlog_errno(ret); + goto out; + } + } + + *credits += ocfs2_calc_extend_credits(osb->sb, et->et_root_el); + + mlog(0, "reserve metadata_blocks: %d, data_clusters: %u, credits: %d\n", + extra_blocks, clusters_to_move, *credits); +out: + if (ret) { + if (*meta_ac) { + ocfs2_free_alloc_context(*meta_ac); + *meta_ac = NULL; + } + } + + return ret; +} + +/* + * Using one journal handle to guarantee the data consistency in case + * crash happens anywhere. + * + * XXX: defrag can end up with finishing partial extent as requested, + * due to not enough contiguous clusters can be found in allocator. + */ +static int ocfs2_defrag_extent(struct ocfs2_move_extents_context *context, + u32 cpos, u32 phys_cpos, u32 *len, int ext_flags) +{ + int ret, credits = 0, extra_blocks = 0, partial = context->partial; + handle_t *handle; + struct inode *inode = context->inode; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct inode *tl_inode = osb->osb_tl_inode; + struct ocfs2_refcount_tree *ref_tree = NULL; + u32 new_phys_cpos, new_len; + u64 phys_blkno = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos); + + if ((ext_flags & OCFS2_EXT_REFCOUNTED) && *len) { + + BUG_ON(!(OCFS2_I(inode)->ip_dyn_features & + OCFS2_HAS_REFCOUNT_FL)); + + BUG_ON(!context->refcount_loc); + + ret = ocfs2_lock_refcount_tree(osb, context->refcount_loc, 1, + &ref_tree, NULL); + if (ret) { + mlog_errno(ret); + return ret; + } + + ret = ocfs2_prepare_refcount_change_for_del(inode, + context->refcount_loc, + phys_blkno, + *len, + &credits, + &extra_blocks); + if (ret) { + mlog_errno(ret); + goto out; + } + } + + ret = ocfs2_lock_allocators_move_extents(inode, &context->et, *len, 1, + &context->meta_ac, + &context->data_ac, + extra_blocks, &credits); + if (ret) { + mlog_errno(ret); + goto out; + } + + /* + * should be using allocation reservation strategy there? + * + * if (context->data_ac) + * context->data_ac->ac_resv = &OCFS2_I(inode)->ip_la_data_resv; + */ + + mutex_lock(&tl_inode->i_mutex); + + if (ocfs2_truncate_log_needs_flush(osb)) { + ret = __ocfs2_flush_truncate_log(osb); + if (ret < 0) { + mlog_errno(ret); + goto out_unlock_mutex; + } + } + + handle = ocfs2_start_trans(osb, credits); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto out_unlock_mutex; + } + + ret = __ocfs2_claim_clusters(handle, context->data_ac, 1, *len, + &new_phys_cpos, &new_len); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + /* + * allowing partial extent moving is kind of 'pros and cons', it makes + * whole defragmentation less likely to fail, on the contrary, the bad + * thing is it may make the fs even more fragmented after moving, let + * userspace make a good decision here. + */ + if (new_len != *len) { + mlog(0, "len_claimed: %u, len: %u\n", new_len, *len); + if (!partial) { + context->range->me_flags &= ~OCFS2_MOVE_EXT_FL_COMPLETE; + ret = -ENOSPC; + goto out_commit; + } + } + + mlog(0, "cpos: %u, phys_cpos: %u, new_phys_cpos: %u\n", cpos, + phys_cpos, new_phys_cpos); + + ret = __ocfs2_move_extent(handle, context, cpos, new_len, phys_cpos, + new_phys_cpos, ext_flags); + if (ret) + mlog_errno(ret); + + if (partial && (new_len != *len)) + *len = new_len; + + /* + * Here we should write the new page out first if we are + * in write-back mode. + */ + ret = ocfs2_cow_sync_writeback(inode->i_sb, context->inode, cpos, *len); + if (ret) + mlog_errno(ret); + +out_commit: + ocfs2_commit_trans(osb, handle); + +out_unlock_mutex: + mutex_unlock(&tl_inode->i_mutex); + + if (context->data_ac) { + ocfs2_free_alloc_context(context->data_ac); + context->data_ac = NULL; + } + + if (context->meta_ac) { + ocfs2_free_alloc_context(context->meta_ac); + context->meta_ac = NULL; + } + +out: + if (ref_tree) + ocfs2_unlock_refcount_tree(osb, ref_tree, 1); + + return ret; +} + +/* + * find the victim alloc group, where #blkno fits. + */ +static int ocfs2_find_victim_alloc_group(struct inode *inode, + u64 vict_blkno, + int type, int slot, + int *vict_bit, + struct buffer_head **ret_bh) +{ + int ret, i, bits_per_unit = 0; + u64 blkno; + char namebuf[40]; + + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct buffer_head *ac_bh = NULL, *gd_bh = NULL; + struct ocfs2_chain_list *cl; + struct ocfs2_chain_rec *rec; + struct ocfs2_dinode *ac_dinode; + struct ocfs2_group_desc *bg; + + ocfs2_sprintf_system_inode_name(namebuf, sizeof(namebuf), type, slot); + ret = ocfs2_lookup_ino_from_name(osb->sys_root_inode, namebuf, + strlen(namebuf), &blkno); + if (ret) { + ret = -ENOENT; + goto out; + } + + ret = ocfs2_read_blocks_sync(osb, blkno, 1, &ac_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + ac_dinode = (struct ocfs2_dinode *)ac_bh->b_data; + cl = &(ac_dinode->id2.i_chain); + rec = &(cl->cl_recs[0]); + + if (type == GLOBAL_BITMAP_SYSTEM_INODE) + bits_per_unit = osb->s_clustersize_bits - + inode->i_sb->s_blocksize_bits; + /* + * 'vict_blkno' was out of the valid range. + */ + if ((vict_blkno < le64_to_cpu(rec->c_blkno)) || + (vict_blkno >= (le32_to_cpu(ac_dinode->id1.bitmap1.i_total) << + bits_per_unit))) { + ret = -EINVAL; + goto out; + } + + for (i = 0; i < le16_to_cpu(cl->cl_next_free_rec); i++) { + + rec = &(cl->cl_recs[i]); + if (!rec) + continue; + + bg = NULL; + + do { + if (!bg) + blkno = le64_to_cpu(rec->c_blkno); + else + blkno = le64_to_cpu(bg->bg_next_group); + + if (gd_bh) { + brelse(gd_bh); + gd_bh = NULL; + } + + ret = ocfs2_read_blocks_sync(osb, blkno, 1, &gd_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + bg = (struct ocfs2_group_desc *)gd_bh->b_data; + + if (vict_blkno < (le64_to_cpu(bg->bg_blkno) + + le16_to_cpu(bg->bg_bits))) { + + *ret_bh = gd_bh; + *vict_bit = (vict_blkno - blkno) >> + bits_per_unit; + mlog(0, "find the victim group: #%llu, " + "total_bits: %u, vict_bit: %u\n", + blkno, le16_to_cpu(bg->bg_bits), + *vict_bit); + goto out; + } + + } while (le64_to_cpu(bg->bg_next_group)); + } + + ret = -EINVAL; +out: + brelse(ac_bh); + + /* + * caller has to release the gd_bh properly. + */ + return ret; +} + +/* + * XXX: helper to validate and adjust moving goal. + */ +static int ocfs2_validate_and_adjust_move_goal(struct inode *inode, + struct ocfs2_move_extents *range) +{ + int ret, goal_bit = 0; + + struct buffer_head *gd_bh = NULL; + struct ocfs2_group_desc *bg; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + int c_to_b = 1 << (osb->s_clustersize_bits - + inode->i_sb->s_blocksize_bits); + + /* + * make goal become cluster aligned. + */ + range->me_goal = ocfs2_block_to_cluster_start(inode->i_sb, + range->me_goal); + /* + * validate goal sits within global_bitmap, and return the victim + * group desc + */ + ret = ocfs2_find_victim_alloc_group(inode, range->me_goal, + GLOBAL_BITMAP_SYSTEM_INODE, + OCFS2_INVALID_SLOT, + &goal_bit, &gd_bh); + if (ret) + goto out; + + bg = (struct ocfs2_group_desc *)gd_bh->b_data; + + /* + * moving goal is not allowd to start with a group desc blok(#0 blk) + * let's compromise to the latter cluster. + */ + if (range->me_goal == le64_to_cpu(bg->bg_blkno)) + range->me_goal += c_to_b; + + /* + * movement is not gonna cross two groups. + */ + if ((le16_to_cpu(bg->bg_bits) - goal_bit) * osb->s_clustersize < + range->me_len) { + ret = -EINVAL; + goto out; + } + /* + * more exact validations/adjustments will be performed later during + * moving operation for each extent range. + */ + mlog(0, "extents get ready to be moved to #%llu block\n", + range->me_goal); + +out: + brelse(gd_bh); + + return ret; +} + +static void ocfs2_probe_alloc_group(struct inode *inode, struct buffer_head *bh, + int *goal_bit, u32 move_len, u32 max_hop, + u32 *phys_cpos) +{ + int i, used, last_free_bits = 0, base_bit = *goal_bit; + struct ocfs2_group_desc *gd = (struct ocfs2_group_desc *)bh->b_data; + u32 base_cpos = ocfs2_blocks_to_clusters(inode->i_sb, + le64_to_cpu(gd->bg_blkno)); + + for (i = base_bit; i < le16_to_cpu(gd->bg_bits); i++) { + + used = ocfs2_test_bit(i, (unsigned long *)gd->bg_bitmap); + if (used) { + /* + * we even tried searching the free chunk by jumping + * a 'max_hop' distance, but still failed. + */ + if ((i - base_bit) > max_hop) { + *phys_cpos = 0; + break; + } + + if (last_free_bits) + last_free_bits = 0; + + continue; + } else + last_free_bits++; + + if (last_free_bits == move_len) { + *goal_bit = i; + *phys_cpos = base_cpos + i; + break; + } + } + + mlog(0, "found phys_cpos: %u to fit the wanted moving.\n", *phys_cpos); +} + +static int ocfs2_move_extent(struct ocfs2_move_extents_context *context, + u32 cpos, u32 phys_cpos, u32 *new_phys_cpos, + u32 len, int ext_flags) +{ + int ret, credits = 0, extra_blocks = 0, goal_bit = 0; + handle_t *handle; + struct inode *inode = context->inode; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct inode *tl_inode = osb->osb_tl_inode; + struct inode *gb_inode = NULL; + struct buffer_head *gb_bh = NULL; + struct buffer_head *gd_bh = NULL; + struct ocfs2_group_desc *gd; + struct ocfs2_refcount_tree *ref_tree = NULL; + u32 move_max_hop = ocfs2_blocks_to_clusters(inode->i_sb, + context->range->me_threshold); + u64 phys_blkno, new_phys_blkno; + + phys_blkno = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos); + + if ((ext_flags & OCFS2_EXT_REFCOUNTED) && len) { + + BUG_ON(!(OCFS2_I(inode)->ip_dyn_features & + OCFS2_HAS_REFCOUNT_FL)); + + BUG_ON(!context->refcount_loc); + + ret = ocfs2_lock_refcount_tree(osb, context->refcount_loc, 1, + &ref_tree, NULL); + if (ret) { + mlog_errno(ret); + return ret; + } + + ret = ocfs2_prepare_refcount_change_for_del(inode, + context->refcount_loc, + phys_blkno, + len, + &credits, + &extra_blocks); + if (ret) { + mlog_errno(ret); + goto out; + } + } + + ret = ocfs2_lock_allocators_move_extents(inode, &context->et, len, 1, + &context->meta_ac, + NULL, extra_blocks, &credits); + if (ret) { + mlog_errno(ret); + goto out; + } + + /* + * need to count 2 extra credits for global_bitmap inode and + * group descriptor. + */ + credits += OCFS2_INODE_UPDATE_CREDITS + 1; + + /* + * ocfs2_move_extent() didn't reserve any clusters in lock_allocators() + * logic, while we still need to lock the global_bitmap. + */ + gb_inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE, + OCFS2_INVALID_SLOT); + if (!gb_inode) { + mlog(ML_ERROR, "unable to get global_bitmap inode\n"); + ret = -EIO; + goto out; + } + + mutex_lock(&gb_inode->i_mutex); + + ret = ocfs2_inode_lock(gb_inode, &gb_bh, 1); + if (ret) { + mlog_errno(ret); + goto out_unlock_gb_mutex; + } + + mutex_lock(&tl_inode->i_mutex); + + handle = ocfs2_start_trans(osb, credits); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto out_unlock_tl_inode; + } + + new_phys_blkno = ocfs2_clusters_to_blocks(inode->i_sb, *new_phys_cpos); + ret = ocfs2_find_victim_alloc_group(inode, new_phys_blkno, + GLOBAL_BITMAP_SYSTEM_INODE, + OCFS2_INVALID_SLOT, + &goal_bit, &gd_bh); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + /* + * probe the victim cluster group to find a proper + * region to fit wanted movement, it even will perfrom + * a best-effort attempt by compromising to a threshold + * around the goal. + */ + ocfs2_probe_alloc_group(inode, gd_bh, &goal_bit, len, move_max_hop, + new_phys_cpos); + if (!*new_phys_cpos) { + ret = -ENOSPC; + goto out_commit; + } + + ret = __ocfs2_move_extent(handle, context, cpos, len, phys_cpos, + *new_phys_cpos, ext_flags); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + gd = (struct ocfs2_group_desc *)gd_bh->b_data; + ret = ocfs2_alloc_dinode_update_counts(gb_inode, handle, gb_bh, len, + le16_to_cpu(gd->bg_chain)); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + ret = ocfs2_block_group_set_bits(handle, gb_inode, gd, gd_bh, + goal_bit, len); + if (ret) { + ocfs2_rollback_alloc_dinode_counts(gb_inode, gb_bh, len, + le16_to_cpu(gd->bg_chain)); + mlog_errno(ret); + } + + /* + * Here we should write the new page out first if we are + * in write-back mode. + */ + ret = ocfs2_cow_sync_writeback(inode->i_sb, context->inode, cpos, len); + if (ret) + mlog_errno(ret); + +out_commit: + ocfs2_commit_trans(osb, handle); + brelse(gd_bh); + +out_unlock_tl_inode: + mutex_unlock(&tl_inode->i_mutex); + + ocfs2_inode_unlock(gb_inode, 1); +out_unlock_gb_mutex: + mutex_unlock(&gb_inode->i_mutex); + brelse(gb_bh); + iput(gb_inode); + +out: + if (context->meta_ac) { + ocfs2_free_alloc_context(context->meta_ac); + context->meta_ac = NULL; + } + + if (ref_tree) + ocfs2_unlock_refcount_tree(osb, ref_tree, 1); + + return ret; +} + +/* + * Helper to calculate the defraging length in one run according to threshold. + */ +static void ocfs2_calc_extent_defrag_len(u32 *alloc_size, u32 *len_defraged, + u32 threshold, int *skip) +{ + if ((*alloc_size + *len_defraged) < threshold) { + /* + * proceed defragmentation until we meet the thresh + */ + *len_defraged += *alloc_size; + } else if (*len_defraged == 0) { + /* + * XXX: skip a large extent. + */ + *skip = 1; + } else { + /* + * split this extent to coalesce with former pieces as + * to reach the threshold. + * + * we're done here with one cycle of defragmentation + * in a size of 'thresh', resetting 'len_defraged' + * forces a new defragmentation. + */ + *alloc_size = threshold - *len_defraged; + *len_defraged = 0; + } +} + +static int __ocfs2_move_extents_range(struct buffer_head *di_bh, + struct ocfs2_move_extents_context *context) +{ + int ret = 0, flags, do_defrag, skip = 0; + u32 cpos, phys_cpos, move_start, len_to_move, alloc_size; + u32 len_defraged = 0, defrag_thresh = 0, new_phys_cpos = 0; + + struct inode *inode = context->inode; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + struct ocfs2_move_extents *range = context->range; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + + if ((i_size_read(inode) == 0) || (range->me_len == 0)) + return 0; + + if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) + return 0; + + context->refcount_loc = le64_to_cpu(di->i_refcount_loc); + + ocfs2_init_dinode_extent_tree(&context->et, INODE_CACHE(inode), di_bh); + ocfs2_init_dealloc_ctxt(&context->dealloc); + + /* + * TO-DO XXX: + * + * - xattr extents. + */ + + do_defrag = context->auto_defrag; + + /* + * extents moving happens in unit of clusters, for the sake + * of simplicity, we may ignore two clusters where 'byte_start' + * and 'byte_start + len' were within. + */ + move_start = ocfs2_clusters_for_bytes(osb->sb, range->me_start); + len_to_move = (range->me_start + range->me_len) >> + osb->s_clustersize_bits; + if (len_to_move >= move_start) + len_to_move -= move_start; + else + len_to_move = 0; + + if (do_defrag) { + defrag_thresh = range->me_threshold >> osb->s_clustersize_bits; + if (defrag_thresh <= 1) + goto done; + } else + new_phys_cpos = ocfs2_blocks_to_clusters(inode->i_sb, + range->me_goal); + + mlog(0, "Inode: %llu, start: %llu, len: %llu, cstart: %u, clen: %u, " + "thresh: %u\n", + (unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)range->me_start, + (unsigned long long)range->me_len, + move_start, len_to_move, defrag_thresh); + + cpos = move_start; + while (len_to_move) { + ret = ocfs2_get_clusters(inode, cpos, &phys_cpos, &alloc_size, + &flags); + if (ret) { + mlog_errno(ret); + goto out; + } + + if (alloc_size > len_to_move) + alloc_size = len_to_move; + + /* + * XXX: how to deal with a hole: + * + * - skip the hole of course + * - force a new defragmentation + */ + if (!phys_cpos) { + if (do_defrag) + len_defraged = 0; + + goto next; + } + + if (do_defrag) { + ocfs2_calc_extent_defrag_len(&alloc_size, &len_defraged, + defrag_thresh, &skip); + /* + * skip large extents + */ + if (skip) { + skip = 0; + goto next; + } + + mlog(0, "#Defrag: cpos: %u, phys_cpos: %u, " + "alloc_size: %u, len_defraged: %u\n", + cpos, phys_cpos, alloc_size, len_defraged); + + ret = ocfs2_defrag_extent(context, cpos, phys_cpos, + &alloc_size, flags); + } else { + ret = ocfs2_move_extent(context, cpos, phys_cpos, + &new_phys_cpos, alloc_size, + flags); + + new_phys_cpos += alloc_size; + } + + if (ret < 0) { + mlog_errno(ret); + goto out; + } + + context->clusters_moved += alloc_size; +next: + cpos += alloc_size; + len_to_move -= alloc_size; + } + +done: + range->me_flags |= OCFS2_MOVE_EXT_FL_COMPLETE; + +out: + range->me_moved_len = ocfs2_clusters_to_bytes(osb->sb, + context->clusters_moved); + range->me_new_offset = ocfs2_clusters_to_bytes(osb->sb, + context->new_phys_cpos); + + ocfs2_schedule_truncate_log_flush(osb, 1); + ocfs2_run_deallocs(osb, &context->dealloc); + + return ret; +} + +static int ocfs2_move_extents(struct ocfs2_move_extents_context *context) +{ + int status; + handle_t *handle; + struct inode *inode = context->inode; + struct ocfs2_dinode *di; + struct buffer_head *di_bh = NULL; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + + if (!inode) + return -ENOENT; + + if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb)) + return -EROFS; + + mutex_lock(&inode->i_mutex); + + /* + * This prevents concurrent writes from other nodes + */ + status = ocfs2_rw_lock(inode, 1); + if (status) { + mlog_errno(status); + goto out; + } + + status = ocfs2_inode_lock(inode, &di_bh, 1); + if (status) { + mlog_errno(status); + goto out_rw_unlock; + } + + /* + * rememer ip_xattr_sem also needs to be held if necessary + */ + down_write(&OCFS2_I(inode)->ip_alloc_sem); + + status = __ocfs2_move_extents_range(di_bh, context); + + up_write(&OCFS2_I(inode)->ip_alloc_sem); + if (status) { + mlog_errno(status); + goto out_inode_unlock; + } + + /* + * We update ctime for these changes + */ + handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS); + if (IS_ERR(handle)) { + status = PTR_ERR(handle); + mlog_errno(status); + goto out_inode_unlock; + } + + status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (status) { + mlog_errno(status); + goto out_commit; + } + + di = (struct ocfs2_dinode *)di_bh->b_data; + inode->i_ctime = CURRENT_TIME; + di->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec); + di->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec); + ocfs2_update_inode_fsync_trans(handle, inode, 0); + + ocfs2_journal_dirty(handle, di_bh); + +out_commit: + ocfs2_commit_trans(osb, handle); + +out_inode_unlock: + brelse(di_bh); + ocfs2_inode_unlock(inode, 1); +out_rw_unlock: + ocfs2_rw_unlock(inode, 1); +out: + mutex_unlock(&inode->i_mutex); + + return status; +} + +int ocfs2_ioctl_move_extents(struct file *filp, void __user *argp) +{ + int status; + + struct inode *inode = file_inode(filp); + struct ocfs2_move_extents range; + struct ocfs2_move_extents_context *context; + + if (!argp) + return -EINVAL; + + status = mnt_want_write_file(filp); + if (status) + return status; + + if ((!S_ISREG(inode->i_mode)) || !(filp->f_mode & FMODE_WRITE)) { + status = -EPERM; + goto out_drop; + } + + if (inode->i_flags & (S_IMMUTABLE|S_APPEND)) { + status = -EPERM; + goto out_drop; + } + + context = kzalloc(sizeof(struct ocfs2_move_extents_context), GFP_NOFS); + if (!context) { + status = -ENOMEM; + mlog_errno(status); + goto out_drop; + } + + context->inode = inode; + context->file = filp; + + if (copy_from_user(&range, argp, sizeof(range))) { + status = -EFAULT; + goto out_free; + } + + if (range.me_start > i_size_read(inode)) { + status = -EINVAL; + goto out_free; + } + + if (range.me_start + range.me_len > i_size_read(inode)) + range.me_len = i_size_read(inode) - range.me_start; + + context->range = ⦥ + + if (range.me_flags & OCFS2_MOVE_EXT_FL_AUTO_DEFRAG) { + context->auto_defrag = 1; + /* + * ok, the default theshold for the defragmentation + * is 1M, since our maximum clustersize was 1M also. + * any thought? + */ + if (!range.me_threshold) + range.me_threshold = 1024 * 1024; + + if (range.me_threshold > i_size_read(inode)) + range.me_threshold = i_size_read(inode); + + if (range.me_flags & OCFS2_MOVE_EXT_FL_PART_DEFRAG) + context->partial = 1; + } else { + /* + * first best-effort attempt to validate and adjust the goal + * (physical address in block), while it can't guarantee later + * operation can succeed all the time since global_bitmap may + * change a bit over time. + */ + + status = ocfs2_validate_and_adjust_move_goal(inode, &range); + if (status) + goto out_copy; + } + + status = ocfs2_move_extents(context); + if (status) + mlog_errno(status); +out_copy: + /* + * movement/defragmentation may end up being partially completed, + * that's the reason why we need to return userspace the finished + * length and new_offset even if failure happens somewhere. + */ + if (copy_to_user(argp, &range, sizeof(range))) + status = -EFAULT; + +out_free: + kfree(context); +out_drop: + mnt_drop_write_file(filp); + + return status; +} diff --git a/fs/ocfs2/move_extents.h b/fs/ocfs2/move_extents.h new file mode 100644 index 00000000000..4e143e81144 --- /dev/null +++ b/fs/ocfs2/move_extents.h @@ -0,0 +1,22 @@ +/* -*- mode: c; c-basic-offset: 8; -*- + * vim: noexpandtab sw=8 ts=8 sts=0: + * + * move_extents.h + * + * Copyright (C) 2011 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License version 2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + */ +#ifndef OCFS2_MOVE_EXTENTS_H +#define OCFS2_MOVE_EXTENTS_H + +int ocfs2_ioctl_move_extents(struct file *filp, void __user *argp); + +#endif /* OCFS2_MOVE_EXTENTS_H */ diff --git a/fs/ocfs2/namei.c b/fs/ocfs2/namei.c index 33464c6b60a..8add6f1030d 100644 --- a/fs/ocfs2/namei.c +++ b/fs/ocfs2/namei.c @@ -42,7 +42,6 @@ #include <linux/highmem.h> #include <linux/quotaops.h> -#define MLOG_MASK_PREFIX ML_NAMEI #include <cluster/masklog.h> #include "ocfs2.h" @@ -63,13 +62,13 @@ #include "uptodate.h" #include "xattr.h" #include "acl.h" +#include "ocfs2_trace.h" #include "buffer_head_io.h" static int ocfs2_mknod_locked(struct ocfs2_super *osb, struct inode *dir, struct inode *inode, - struct dentry *dentry, dev_t dev, struct buffer_head **new_fe_bh, struct buffer_head *parent_fe_bh, @@ -78,14 +77,14 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb, static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb, struct inode **ret_orphan_dir, - struct inode *inode, + u64 blkno, char *name, struct ocfs2_dir_lookup_result *lookup); static int ocfs2_orphan_add(struct ocfs2_super *osb, handle_t *handle, struct inode *inode, - struct ocfs2_dinode *fe, + struct buffer_head *fe_bh, char *name, struct ocfs2_dir_lookup_result *lookup, struct inode *orphan_dir_inode); @@ -99,7 +98,7 @@ static int ocfs2_create_symlink_data(struct ocfs2_super *osb, #define OCFS2_ORPHAN_NAMELEN ((int)(2 * sizeof(u64))) static struct dentry *ocfs2_lookup(struct inode *dir, struct dentry *dentry, - struct nameidata *nd) + unsigned int flags) { int status; u64 blkno; @@ -107,18 +106,16 @@ static struct dentry *ocfs2_lookup(struct inode *dir, struct dentry *dentry, struct dentry *ret; struct ocfs2_inode_info *oi; - mlog_entry("(0x%p, 0x%p, '%.*s')\n", dir, dentry, - dentry->d_name.len, dentry->d_name.name); + trace_ocfs2_lookup(dir, dentry, dentry->d_name.len, + dentry->d_name.name, + (unsigned long long)OCFS2_I(dir)->ip_blkno, 0); if (dentry->d_name.len > OCFS2_MAX_FILENAME_LEN) { ret = ERR_PTR(-ENAMETOOLONG); goto bail; } - mlog(0, "find name %.*s in directory %llu\n", dentry->d_name.len, - dentry->d_name.name, (unsigned long long)OCFS2_I(dir)->ip_blkno); - - status = ocfs2_inode_lock(dir, NULL, 0); + status = ocfs2_inode_lock_nested(dir, NULL, 0, OI_LS_PARENT); if (status < 0) { if (status != -ENOENT) mlog_errno(status); @@ -148,7 +145,6 @@ static struct dentry *ocfs2_lookup(struct inode *dir, struct dentry *dentry, spin_unlock(&oi->ip_lock); bail_add: - dentry->d_op = &ocfs2_dentry_ops; ret = d_splice_alias(inode, dentry); if (inode) { @@ -172,7 +168,8 @@ bail_add: ret = ERR_PTR(status); goto bail_unlock; } - } + } else + ocfs2_dentry_attach_gen(dentry); bail_unlock: /* Don't drop the cluster lock until *after* the d_add -- @@ -183,12 +180,12 @@ bail_unlock: bail: - mlog_exit_ptr(ret); + trace_ocfs2_lookup_ret(ret); return ret; } -static struct inode *ocfs2_get_init_inode(struct inode *dir, int mode) +static struct inode *ocfs2_get_init_inode(struct inode *dir, umode_t mode) { struct inode *inode; @@ -202,24 +199,30 @@ static struct inode *ocfs2_get_init_inode(struct inode *dir, int mode) * these are used by the support functions here and in * callers. */ if (S_ISDIR(mode)) - inode->i_nlink = 2; - else - inode->i_nlink = 1; - inode->i_uid = current_fsuid(); - if (dir->i_mode & S_ISGID) { - inode->i_gid = dir->i_gid; - if (S_ISDIR(mode)) - mode |= S_ISGID; - } else - inode->i_gid = current_fsgid(); - inode->i_mode = mode; - vfs_dq_init(inode); + set_nlink(inode, 2); + inode_init_owner(inode, dir, mode); + dquot_initialize(inode); return inode; } +static void ocfs2_cleanup_add_entry_failure(struct ocfs2_super *osb, + struct dentry *dentry, struct inode *inode) +{ + struct ocfs2_dentry_lock *dl = dentry->d_fsdata; + + ocfs2_simple_drop_lockres(osb, &dl->dl_lockres); + ocfs2_lock_res_free(&dl->dl_lockres); + BUG_ON(dl->dl_count != 1); + spin_lock(&dentry_attach_lock); + dentry->d_fsdata = NULL; + spin_unlock(&dentry_attach_lock); + kfree(dl); + iput(inode); +} + static int ocfs2_mknod(struct inode *dir, struct dentry *dentry, - int mode, + umode_t mode, dev_t dev) { int status = 0; @@ -240,10 +243,16 @@ static int ocfs2_mknod(struct inode *dir, }; int did_quota_inode = 0; struct ocfs2_dir_lookup_result lookup = { NULL, }; + sigset_t oldset; + int did_block_signals = 0; + struct posix_acl *default_acl = NULL, *acl = NULL; + struct ocfs2_dentry_lock *dl = NULL; + + trace_ocfs2_mknod(dir, dentry, dentry->d_name.len, dentry->d_name.name, + (unsigned long long)OCFS2_I(dir)->ip_blkno, + (unsigned long)dev, mode); - mlog_entry("(0x%p, 0x%p, %d, %lu, '%.*s')\n", dir, dentry, mode, - (unsigned long)dev, dentry->d_name.len, - dentry->d_name.name); + dquot_initialize(dir); /* get our super block */ osb = OCFS2_SB(dir->i_sb); @@ -297,7 +306,7 @@ static int ocfs2_mknod(struct inode *dir, } /* get security xattr */ - status = ocfs2_init_security_get(inode, dir, &si); + status = ocfs2_init_security_get(inode, dir, &dentry->d_name, &si); if (status) { if (status == -EOPNOTSUPP) si.enable = 0; @@ -339,6 +348,12 @@ static int ocfs2_mknod(struct inode *dir, goto leave; } + status = posix_acl_create(dir, &mode, &default_acl, &acl); + if (status) { + mlog_errno(status); + goto leave; + } + handle = ocfs2_start_trans(osb, ocfs2_mknod_credits(osb->sb, S_ISDIR(mode), xattr_credits)); @@ -349,17 +364,17 @@ static int ocfs2_mknod(struct inode *dir, goto leave; } - /* We don't use standard VFS wrapper because we don't want vfs_dq_init - * to be called. */ - if (sb_any_quota_active(osb->sb) && - osb->sb->dq_op->alloc_inode(inode, 1) == NO_QUOTA) { - status = -EDQUOT; + /* Starting to change things, restart is no longer possible. */ + ocfs2_block_signals(&oldset); + did_block_signals = 1; + + status = dquot_alloc_inode(inode); + if (status) goto leave; - } did_quota_inode = 1; /* do the real work now. */ - status = ocfs2_mknod_locked(osb, dir, inode, dentry, dev, + status = ocfs2_mknod_locked(osb, dir, inode, dev, &new_fe_bh, parent_fe_bh, handle, inode_ac); if (status < 0) { @@ -375,23 +390,29 @@ static int ocfs2_mknod(struct inode *dir, goto leave; } - status = ocfs2_journal_access_di(handle, dir, parent_fe_bh, + status = ocfs2_journal_access_di(handle, INODE_CACHE(dir), + parent_fe_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); goto leave; } ocfs2_add_links_count(dirfe, 1); - status = ocfs2_journal_dirty(handle, parent_fe_bh); - if (status < 0) { - mlog_errno(status); - goto leave; - } + ocfs2_journal_dirty(handle, parent_fe_bh); inc_nlink(dir); } - status = ocfs2_init_acl(handle, inode, dir, new_fe_bh, parent_fe_bh, - meta_ac, data_ac); + if (default_acl) { + status = ocfs2_set_acl(handle, inode, new_fe_bh, + ACL_TYPE_DEFAULT, default_acl, + meta_ac, data_ac); + } + if (!status && acl) { + status = ocfs2_set_acl(handle, inode, new_fe_bh, + ACL_TYPE_ACCESS, acl, + meta_ac, data_ac); + } + if (status < 0) { mlog_errno(status); goto leave; @@ -406,48 +427,51 @@ static int ocfs2_mknod(struct inode *dir, } } - status = ocfs2_add_entry(handle, dentry, inode, - OCFS2_I(inode)->ip_blkno, parent_fe_bh, - &lookup); - if (status < 0) { + /* + * Do this before adding the entry to the directory. We add + * also set d_op after success so that ->d_iput() will cleanup + * the dentry lock even if ocfs2_add_entry() fails below. + */ + status = ocfs2_dentry_attach_lock(dentry, inode, + OCFS2_I(dir)->ip_blkno); + if (status) { mlog_errno(status); goto leave; } - status = ocfs2_dentry_attach_lock(dentry, inode, - OCFS2_I(dir)->ip_blkno); - if (status) { + dl = dentry->d_fsdata; + + status = ocfs2_add_entry(handle, dentry, inode, + OCFS2_I(inode)->ip_blkno, parent_fe_bh, + &lookup); + if (status < 0) { mlog_errno(status); goto leave; } insert_inode_hash(inode); - dentry->d_op = &ocfs2_dentry_ops; d_instantiate(dentry, inode); status = 0; leave: + if (default_acl) + posix_acl_release(default_acl); + if (acl) + posix_acl_release(acl); if (status < 0 && did_quota_inode) - vfs_dq_free_inode(inode); + dquot_free_inode(inode); if (handle) ocfs2_commit_trans(osb, handle); ocfs2_inode_unlock(dir, 1); - - if (status == -ENOSPC) - mlog(0, "Disk is full\n"); + if (did_block_signals) + ocfs2_unblock_signals(&oldset); brelse(new_fe_bh); brelse(parent_fe_bh); - kfree(si.name); kfree(si.value); ocfs2_free_dir_lookup_result(&lookup); - if ((status < 0) && inode) { - clear_nlink(inode); - iput(inode); - } - if (inode_ac) ocfs2_free_alloc_context(inode_ac); @@ -457,41 +481,44 @@ leave: if (meta_ac) ocfs2_free_alloc_context(meta_ac); - mlog_exit(status); + /* + * We should call iput after the i_mutex of the bitmap been + * unlocked in ocfs2_free_alloc_context, or the + * ocfs2_delete_inode will mutex_lock again. + */ + if ((status < 0) && inode) { + if (dl) + ocfs2_cleanup_add_entry_failure(osb, dentry, inode); + + OCFS2_I(inode)->ip_flags |= OCFS2_INODE_SKIP_ORPHAN_DIR; + clear_nlink(inode); + iput(inode); + } + + if (status) + mlog_errno(status); return status; } -static int ocfs2_mknod_locked(struct ocfs2_super *osb, - struct inode *dir, - struct inode *inode, - struct dentry *dentry, - dev_t dev, - struct buffer_head **new_fe_bh, - struct buffer_head *parent_fe_bh, - handle_t *handle, - struct ocfs2_alloc_context *inode_ac) +static int __ocfs2_mknod_locked(struct inode *dir, + struct inode *inode, + dev_t dev, + struct buffer_head **new_fe_bh, + struct buffer_head *parent_fe_bh, + handle_t *handle, + struct ocfs2_alloc_context *inode_ac, + u64 fe_blkno, u64 suballoc_loc, u16 suballoc_bit) { int status = 0; + struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); struct ocfs2_dinode *fe = NULL; struct ocfs2_extent_list *fel; - u64 fe_blkno = 0; - u16 suballoc_bit; u16 feat; - - mlog_entry("(0x%p, 0x%p, %d, %lu, '%.*s')\n", dir, dentry, - inode->i_mode, (unsigned long)dev, dentry->d_name.len, - dentry->d_name.name); + struct ocfs2_inode_info *oi = OCFS2_I(inode); *new_fe_bh = NULL; - status = ocfs2_claim_new_inode(osb, handle, dir, parent_fe_bh, - inode_ac, &suballoc_bit, &fe_blkno); - if (status < 0) { - mlog_errno(status); - goto leave; - } - /* populate as many fields early on as possible - many of * these are used by the support functions here and in * callers. */ @@ -503,13 +530,14 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb, *new_fe_bh = sb_getblk(osb->sb, fe_blkno); if (!*new_fe_bh) { - status = -EIO; + status = -ENOMEM; mlog_errno(status); goto leave; } - ocfs2_set_new_buffer_uptodate(inode, *new_fe_bh); + ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode), *new_fe_bh); - status = ocfs2_journal_access_di(handle, inode, *new_fe_bh, + status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), + *new_fe_bh, OCFS2_JOURNAL_ACCESS_CREATE); if (status < 0) { mlog_errno(status); @@ -522,10 +550,11 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb, fe->i_generation = cpu_to_le32(inode->i_generation); fe->i_fs_generation = cpu_to_le32(osb->fs_generation); fe->i_blkno = cpu_to_le64(fe_blkno); + fe->i_suballoc_loc = cpu_to_le64(suballoc_loc); fe->i_suballoc_bit = cpu_to_le16(suballoc_bit); fe->i_suballoc_slot = cpu_to_le16(inode_ac->ac_alloc_slot); - fe->i_uid = cpu_to_le32(inode->i_uid); - fe->i_gid = cpu_to_le32(inode->i_gid); + fe->i_uid = cpu_to_le32(i_uid_read(inode)); + fe->i_gid = cpu_to_le32(i_gid_read(inode)); fe->i_mode = cpu_to_le16(inode->i_mode); if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) fe->id1.dev1.i_rdev = cpu_to_le64(huge_encode_dev(dev)); @@ -534,7 +563,7 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb, fe->i_last_eb_blk = 0; strcpy(fe->i_signature, OCFS2_INODE_SIGNATURE); - le32_add_cpu(&fe->i_flags, OCFS2_VALID_FL); + fe->i_flags |= cpu_to_le32(OCFS2_VALID_FL); fe->i_atime = fe->i_ctime = fe->i_mtime = cpu_to_le64(CURRENT_TIME.tv_sec); fe->i_mtime_nsec = fe->i_ctime_nsec = fe->i_atime_nsec = @@ -558,22 +587,18 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb, fel->l_count = cpu_to_le16(ocfs2_extent_recs_per_inode(osb->sb)); } - status = ocfs2_journal_dirty(handle, *new_fe_bh); - if (status < 0) { - mlog_errno(status); - goto leave; - } + ocfs2_journal_dirty(handle, *new_fe_bh); ocfs2_populate_inode(inode, fe, 1); - ocfs2_inode_set_new(osb, inode); + ocfs2_ci_set_new(osb, INODE_CACHE(inode)); if (!ocfs2_mount_local(osb)) { status = ocfs2_create_new_inode_locks(inode); if (status < 0) mlog_errno(status); } - status = 0; /* error in ocfs2_create_new_inode_locks is not - * critical */ + oi->i_sync_tid = handle->h_transaction->t_tid; + oi->i_datasync_tid = handle->h_transaction->t_tid; leave: if (status < 0) { @@ -583,35 +608,66 @@ leave: } } - mlog_exit(status); + if (status) + mlog_errno(status); return status; } +static int ocfs2_mknod_locked(struct ocfs2_super *osb, + struct inode *dir, + struct inode *inode, + dev_t dev, + struct buffer_head **new_fe_bh, + struct buffer_head *parent_fe_bh, + handle_t *handle, + struct ocfs2_alloc_context *inode_ac) +{ + int status = 0; + u64 suballoc_loc, fe_blkno = 0; + u16 suballoc_bit; + + *new_fe_bh = NULL; + + status = ocfs2_claim_new_inode(handle, dir, parent_fe_bh, + inode_ac, &suballoc_loc, + &suballoc_bit, &fe_blkno); + if (status < 0) { + mlog_errno(status); + return status; + } + + return __ocfs2_mknod_locked(dir, inode, dev, new_fe_bh, + parent_fe_bh, handle, inode_ac, + fe_blkno, suballoc_loc, suballoc_bit); +} + static int ocfs2_mkdir(struct inode *dir, struct dentry *dentry, - int mode) + umode_t mode) { int ret; - mlog_entry("(0x%p, 0x%p, %d, '%.*s')\n", dir, dentry, mode, - dentry->d_name.len, dentry->d_name.name); + trace_ocfs2_mkdir(dir, dentry, dentry->d_name.len, dentry->d_name.name, + OCFS2_I(dir)->ip_blkno, mode); ret = ocfs2_mknod(dir, dentry, mode | S_IFDIR, 0); - mlog_exit(ret); + if (ret) + mlog_errno(ret); return ret; } static int ocfs2_create(struct inode *dir, struct dentry *dentry, - int mode, - struct nameidata *nd) + umode_t mode, + bool excl) { int ret; - mlog_entry("(0x%p, 0x%p, %d, '%.*s')\n", dir, dentry, mode, - dentry->d_name.len, dentry->d_name.name); + trace_ocfs2_create(dir, dentry, dentry->d_name.len, dentry->d_name.name, + (unsigned long long)OCFS2_I(dir)->ip_blkno, mode); ret = ocfs2_mknod(dir, dentry, mode | S_IFREG, 0); - mlog_exit(ret); + if (ret) + mlog_errno(ret); return ret; } @@ -628,15 +684,19 @@ static int ocfs2_link(struct dentry *old_dentry, struct ocfs2_dinode *fe = NULL; struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); struct ocfs2_dir_lookup_result lookup = { NULL, }; + sigset_t oldset; + u64 old_de_ino; - mlog_entry("(inode=%lu, old='%.*s' new='%.*s')\n", inode->i_ino, - old_dentry->d_name.len, old_dentry->d_name.name, - dentry->d_name.len, dentry->d_name.name); + trace_ocfs2_link((unsigned long long)OCFS2_I(inode)->ip_blkno, + old_dentry->d_name.len, old_dentry->d_name.name, + dentry->d_name.len, dentry->d_name.name); if (S_ISDIR(inode->i_mode)) return -EPERM; - err = ocfs2_inode_lock(dir, &parent_fe_bh, 1); + dquot_initialize(dir); + + err = ocfs2_inode_lock_nested(dir, &parent_fe_bh, 1, OI_LS_PARENT); if (err < 0) { if (err != -ENOENT) mlog_errno(err); @@ -648,6 +708,22 @@ static int ocfs2_link(struct dentry *old_dentry, goto out; } + err = ocfs2_lookup_ino_from_name(dir, old_dentry->d_name.name, + old_dentry->d_name.len, &old_de_ino); + if (err) { + err = -ENOENT; + goto out; + } + + /* + * Check whether another node removed the source inode while we + * were in the vfs. + */ + if (old_de_ino != OCFS2_I(inode)->ip_blkno) { + err = -ENOENT; + goto out; + } + err = ocfs2_check_dir_for_entry(dir, dentry->d_name.name, dentry->d_name.len); if (err) @@ -682,7 +758,10 @@ static int ocfs2_link(struct dentry *old_dentry, goto out_unlock_inode; } - err = ocfs2_journal_access_di(handle, inode, fe_bh, + /* Starting to change things, restart is no longer possible. */ + ocfs2_block_signals(&oldset); + + err = ocfs2_journal_access_di(handle, INODE_CACHE(inode), fe_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (err < 0) { mlog_errno(err); @@ -694,14 +773,7 @@ static int ocfs2_link(struct dentry *old_dentry, ocfs2_set_links_count(fe, inode->i_nlink); fe->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec); fe->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec); - - err = ocfs2_journal_dirty(handle, fe_bh); - if (err < 0) { - ocfs2_add_links_count(fe, -1); - drop_nlink(inode); - mlog_errno(err); - goto out_commit; - } + ocfs2_journal_dirty(handle, fe_bh); err = ocfs2_add_entry(handle, dentry, inode, OCFS2_I(inode)->ip_blkno, @@ -719,12 +791,12 @@ static int ocfs2_link(struct dentry *old_dentry, goto out_commit; } - atomic_inc(&inode->i_count); - dentry->d_op = &ocfs2_dentry_ops; + ihold(inode); d_instantiate(dentry, inode); out_commit: ocfs2_commit_trans(osb, handle); + ocfs2_unblock_signals(&oldset); out_unlock_inode: ocfs2_inode_unlock(inode, 1); @@ -736,7 +808,8 @@ out: ocfs2_free_dir_lookup_result(&lookup); - mlog_exit(err); + if (err) + mlog_errno(err); return err; } @@ -758,7 +831,7 @@ static int ocfs2_remote_dentry_delete(struct dentry *dentry) return ret; } -static inline int inode_is_unlinkable(struct inode *inode) +static inline int ocfs2_inode_is_unlinkable(struct inode *inode) { if (S_ISDIR(inode->i_mode)) { if (inode->i_nlink == 2) @@ -776,6 +849,7 @@ static int ocfs2_unlink(struct inode *dir, { int status; int child_locked = 0; + bool is_unlinkable = false; struct inode *inode = dentry->d_inode; struct inode *orphan_dir = NULL; struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); @@ -788,19 +862,20 @@ static int ocfs2_unlink(struct inode *dir, struct ocfs2_dir_lookup_result lookup = { NULL, }; struct ocfs2_dir_lookup_result orphan_insert = { NULL, }; - mlog_entry("(0x%p, 0x%p, '%.*s')\n", dir, dentry, - dentry->d_name.len, dentry->d_name.name); + trace_ocfs2_unlink(dir, dentry, dentry->d_name.len, + dentry->d_name.name, + (unsigned long long)OCFS2_I(dir)->ip_blkno, + (unsigned long long)OCFS2_I(inode)->ip_blkno); - BUG_ON(dentry->d_parent->d_inode != dir); + dquot_initialize(dir); - mlog(0, "ino = %llu\n", (unsigned long long)OCFS2_I(inode)->ip_blkno); + BUG_ON(dentry->d_parent->d_inode != dir); - if (inode == osb->root_inode) { - mlog(0, "Cannot delete the root directory\n"); + if (inode == osb->root_inode) return -EPERM; - } - status = ocfs2_inode_lock(dir, &parent_node_bh, 1); + status = ocfs2_inode_lock_nested(dir, &parent_node_bh, 1, + OI_LS_PARENT); if (status < 0) { if (status != -ENOENT) mlog_errno(status); @@ -819,9 +894,10 @@ static int ocfs2_unlink(struct inode *dir, if (OCFS2_I(inode)->ip_blkno != blkno) { status = -ENOENT; - mlog(0, "ip_blkno %llu != dirent blkno %llu ip_flags = %x\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, - (unsigned long long)blkno, OCFS2_I(inode)->ip_flags); + trace_ocfs2_unlink_noent( + (unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)blkno, + OCFS2_I(inode)->ip_flags); goto leave; } @@ -848,13 +924,15 @@ static int ocfs2_unlink(struct inode *dir, goto leave; } - if (inode_is_unlinkable(inode)) { - status = ocfs2_prepare_orphan_dir(osb, &orphan_dir, inode, + if (ocfs2_inode_is_unlinkable(inode)) { + status = ocfs2_prepare_orphan_dir(osb, &orphan_dir, + OCFS2_I(inode)->ip_blkno, orphan_name, &orphan_insert); if (status < 0) { mlog_errno(status); goto leave; } + is_unlinkable = true; } handle = ocfs2_start_trans(osb, ocfs2_unlink_credits(osb->sb)); @@ -865,7 +943,7 @@ static int ocfs2_unlink(struct inode *dir, goto leave; } - status = ocfs2_journal_access_di(handle, inode, fe_bh, + status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), fe_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); @@ -874,15 +952,6 @@ static int ocfs2_unlink(struct inode *dir, fe = (struct ocfs2_dinode *) fe_bh->b_data; - if (inode_is_unlinkable(inode)) { - status = ocfs2_orphan_add(osb, handle, inode, fe, orphan_name, - &orphan_insert, orphan_dir); - if (status < 0) { - mlog_errno(status); - goto leave; - } - } - /* delete the name from the parent dir */ status = ocfs2_delete_entry(handle, dir, &lookup); if (status < 0) { @@ -894,12 +963,7 @@ static int ocfs2_unlink(struct inode *dir, drop_nlink(inode); drop_nlink(inode); ocfs2_set_links_count(fe, inode->i_nlink); - - status = ocfs2_journal_dirty(handle, fe_bh); - if (status < 0) { - mlog_errno(status); - goto leave; - } + ocfs2_journal_dirty(handle, fe_bh); dir->i_ctime = dir->i_mtime = CURRENT_TIME; if (S_ISDIR(inode->i_mode)) @@ -910,6 +974,14 @@ static int ocfs2_unlink(struct inode *dir, mlog_errno(status); if (S_ISDIR(inode->i_mode)) inc_nlink(dir); + goto leave; + } + + if (is_unlinkable) { + status = ocfs2_orphan_add(osb, handle, inode, fe_bh, + orphan_name, &orphan_insert, orphan_dir); + if (status < 0) + mlog_errno(status); } leave: @@ -934,11 +1006,71 @@ leave: ocfs2_free_dir_lookup_result(&orphan_insert); ocfs2_free_dir_lookup_result(&lookup); - mlog_exit(status); + if (status && (status != -ENOTEMPTY) && (status != -ENOENT)) + mlog_errno(status); return status; } +static int ocfs2_check_if_ancestor(struct ocfs2_super *osb, + u64 src_inode_no, u64 dest_inode_no) +{ + int ret = 0, i = 0; + u64 parent_inode_no = 0; + u64 child_inode_no = src_inode_no; + struct inode *child_inode; + +#define MAX_LOOKUP_TIMES 32 + while (1) { + child_inode = ocfs2_iget(osb, child_inode_no, 0, 0); + if (IS_ERR(child_inode)) { + ret = PTR_ERR(child_inode); + break; + } + + ret = ocfs2_inode_lock(child_inode, NULL, 0); + if (ret < 0) { + iput(child_inode); + if (ret != -ENOENT) + mlog_errno(ret); + break; + } + + ret = ocfs2_lookup_ino_from_name(child_inode, "..", 2, + &parent_inode_no); + ocfs2_inode_unlock(child_inode, 0); + iput(child_inode); + if (ret < 0) { + ret = -ENOENT; + break; + } + + if (parent_inode_no == dest_inode_no) { + ret = 1; + break; + } + + if (parent_inode_no == osb->root_inode->i_ino) { + ret = 0; + break; + } + + child_inode_no = parent_inode_no; + + if (++i >= MAX_LOOKUP_TIMES) { + mlog(ML_NOTICE, "max lookup times reached, filesystem " + "may have nested directories, " + "src inode: %llu, dest inode: %llu.\n", + (unsigned long long)src_inode_no, + (unsigned long long)dest_inode_no); + ret = 0; + break; + } + } + + return ret; +} + /* * The only place this should be used is rename! * if they have the same id, then the 1st one is the only one locked. @@ -950,25 +1082,41 @@ static int ocfs2_double_lock(struct ocfs2_super *osb, struct inode *inode2) { int status; + int inode1_is_ancestor, inode2_is_ancestor; struct ocfs2_inode_info *oi1 = OCFS2_I(inode1); struct ocfs2_inode_info *oi2 = OCFS2_I(inode2); struct buffer_head **tmpbh; struct inode *tmpinode; - mlog_entry("(inode1 = %llu, inode2 = %llu)\n", - (unsigned long long)oi1->ip_blkno, - (unsigned long long)oi2->ip_blkno); + trace_ocfs2_double_lock((unsigned long long)oi1->ip_blkno, + (unsigned long long)oi2->ip_blkno); if (*bh1) *bh1 = NULL; if (*bh2) *bh2 = NULL; - /* we always want to lock the one with the lower lockid first. */ + /* we always want to lock the one with the lower lockid first. + * and if they are nested, we lock ancestor first */ if (oi1->ip_blkno != oi2->ip_blkno) { - if (oi1->ip_blkno < oi2->ip_blkno) { + inode1_is_ancestor = ocfs2_check_if_ancestor(osb, oi2->ip_blkno, + oi1->ip_blkno); + if (inode1_is_ancestor < 0) { + status = inode1_is_ancestor; + goto bail; + } + + inode2_is_ancestor = ocfs2_check_if_ancestor(osb, oi1->ip_blkno, + oi2->ip_blkno); + if (inode2_is_ancestor < 0) { + status = inode2_is_ancestor; + goto bail; + } + + if ((inode1_is_ancestor == 1) || + (oi1->ip_blkno < oi2->ip_blkno && + inode2_is_ancestor == 0)) { /* switch id1 and id2 around */ - mlog(0, "switching them around...\n"); tmpbh = bh2; bh2 = bh1; bh1 = tmpbh; @@ -978,7 +1126,8 @@ static int ocfs2_double_lock(struct ocfs2_super *osb, inode1 = tmpinode; } /* lock id2 */ - status = ocfs2_inode_lock(inode2, bh2, 1); + status = ocfs2_inode_lock_nested(inode2, bh2, 1, + OI_LS_RENAME1); if (status < 0) { if (status != -ENOENT) mlog_errno(status); @@ -987,21 +1136,29 @@ static int ocfs2_double_lock(struct ocfs2_super *osb, } /* lock id1 */ - status = ocfs2_inode_lock(inode1, bh1, 1); + status = ocfs2_inode_lock_nested(inode1, bh1, 1, OI_LS_RENAME2); if (status < 0) { /* * An error return must mean that no cluster locks * were held on function exit. */ - if (oi1->ip_blkno != oi2->ip_blkno) + if (oi1->ip_blkno != oi2->ip_blkno) { ocfs2_inode_unlock(inode2, 1); + brelse(*bh2); + *bh2 = NULL; + } if (status != -ENOENT) mlog_errno(status); } + trace_ocfs2_double_lock_end( + (unsigned long long)OCFS2_I(inode1)->ip_blkno, + (unsigned long long)OCFS2_I(inode2)->ip_blkno); + bail: - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -1032,21 +1189,24 @@ static int ocfs2_rename(struct inode *old_dir, handle_t *handle = NULL; struct buffer_head *old_dir_bh = NULL; struct buffer_head *new_dir_bh = NULL; - nlink_t old_dir_nlink = old_dir->i_nlink; + u32 old_dir_nlink = old_dir->i_nlink; struct ocfs2_dinode *old_di; struct ocfs2_dir_lookup_result old_inode_dot_dot_res = { NULL, }; struct ocfs2_dir_lookup_result target_lookup_res = { NULL, }; struct ocfs2_dir_lookup_result old_entry_lookup = { NULL, }; struct ocfs2_dir_lookup_result orphan_insert = { NULL, }; struct ocfs2_dir_lookup_result target_insert = { NULL, }; + bool should_add_orphan = false; /* At some point it might be nice to break this function up a * bit. */ - mlog_entry("(0x%p, 0x%p, 0x%p, 0x%p, from='%.*s' to='%.*s')\n", - old_dir, old_dentry, new_dir, new_dentry, - old_dentry->d_name.len, old_dentry->d_name.name, - new_dentry->d_name.len, new_dentry->d_name.name); + trace_ocfs2_rename(old_dir, old_dentry, new_dir, new_dentry, + old_dentry->d_name.len, old_dentry->d_name.name, + new_dentry->d_name.len, new_dentry->d_name.name); + + dquot_initialize(old_dir); + dquot_initialize(new_dir); osb = OCFS2_SB(old_dir->i_sb); @@ -1073,6 +1233,21 @@ static int ocfs2_rename(struct inode *old_dir, goto bail; } rename_lock = 1; + + /* here we cannot guarantee the inodes haven't just been + * changed, so check if they are nested again */ + status = ocfs2_check_if_ancestor(osb, new_dir->i_ino, + old_inode->i_ino); + if (status < 0) { + mlog_errno(status); + goto bail; + } else if (status == 1) { + status = -EPERM; + trace_ocfs2_rename_not_permitted( + (unsigned long long)old_inode->i_ino, + (unsigned long long)new_dir->i_ino); + goto bail; + } } /* if old and new are the same, this'll just do one lock. */ @@ -1103,7 +1278,8 @@ static int ocfs2_rename(struct inode *old_dir, * won't have to concurrently downconvert the inode and the * dentry locks. */ - status = ocfs2_inode_lock(old_inode, &old_inode_bh, 1); + status = ocfs2_inode_lock_nested(old_inode, &old_inode_bh, 1, + OI_LS_PARENT); if (status < 0) { if (status != -ENOENT) mlog_errno(status); @@ -1199,16 +1375,15 @@ static int ocfs2_rename(struct inode *old_dir, if (!new_inode) { status = -EACCES; - mlog(0, "We found an inode for name %.*s but VFS " - "didn't give us one.\n", new_dentry->d_name.len, - new_dentry->d_name.name); + trace_ocfs2_rename_target_exists(new_dentry->d_name.len, + new_dentry->d_name.name); goto bail; } if (OCFS2_I(new_inode)->ip_blkno != newfe_blkno) { status = -EACCES; - mlog(0, "Inode %llu and dir %llu disagree. flags = %x\n", + trace_ocfs2_rename_disagree( (unsigned long long)OCFS2_I(new_inode)->ip_blkno, (unsigned long long)newfe_blkno, OCFS2_I(new_inode)->ip_flags); @@ -1231,20 +1406,19 @@ static int ocfs2_rename(struct inode *old_dir, newfe = (struct ocfs2_dinode *) newfe_bh->b_data; - mlog(0, "aha rename over existing... new_blkno=%llu " - "newfebh=%p bhblocknr=%llu\n", + trace_ocfs2_rename_over_existing( (unsigned long long)newfe_blkno, newfe_bh, newfe_bh ? (unsigned long long)newfe_bh->b_blocknr : 0ULL); if (S_ISDIR(new_inode->i_mode) || (new_inode->i_nlink == 1)) { status = ocfs2_prepare_orphan_dir(osb, &orphan_dir, - new_inode, - orphan_name, - &orphan_insert); + OCFS2_I(new_inode)->ip_blkno, + orphan_name, &orphan_insert); if (status < 0) { mlog_errno(status); goto bail; } + should_add_orphan = true; } } else { BUG_ON(new_dentry->d_parent->d_inode != new_dir); @@ -1281,24 +1455,14 @@ static int ocfs2_rename(struct inode *old_dir, goto bail; } } - status = ocfs2_journal_access_di(handle, new_inode, newfe_bh, + status = ocfs2_journal_access_di(handle, INODE_CACHE(new_inode), + newfe_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); goto bail; } - if (S_ISDIR(new_inode->i_mode) || - (ocfs2_read_links_count(newfe) == 1)) { - status = ocfs2_orphan_add(osb, handle, new_inode, - newfe, orphan_name, - &orphan_insert, orphan_dir); - if (status < 0) { - mlog_errno(status); - goto bail; - } - } - /* change the dirent to point to the correct inode */ status = ocfs2_update_entry(new_dir, handle, &target_lookup_res, old_inode); @@ -1312,11 +1476,15 @@ static int ocfs2_rename(struct inode *old_dir, ocfs2_set_links_count(newfe, 0); else ocfs2_add_links_count(newfe, -1); - - status = ocfs2_journal_dirty(handle, newfe_bh); - if (status < 0) { - mlog_errno(status); - goto bail; + ocfs2_journal_dirty(handle, newfe_bh); + if (should_add_orphan) { + status = ocfs2_orphan_add(osb, handle, new_inode, + newfe_bh, orphan_name, + &orphan_insert, orphan_dir); + if (status < 0) { + mlog_errno(status); + goto bail; + } } } else { /* if the name was not found in new_dir, add it now */ @@ -1328,17 +1496,15 @@ static int ocfs2_rename(struct inode *old_dir, old_inode->i_ctime = CURRENT_TIME; mark_inode_dirty(old_inode); - status = ocfs2_journal_access_di(handle, old_inode, old_inode_bh, + status = ocfs2_journal_access_di(handle, INODE_CACHE(old_inode), + old_inode_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status >= 0) { old_di = (struct ocfs2_dinode *) old_inode_bh->b_data; old_di->i_ctime = cpu_to_le64(old_inode->i_ctime.tv_sec); old_di->i_ctime_nsec = cpu_to_le32(old_inode->i_ctime.tv_nsec); - - status = ocfs2_journal_dirty(handle, old_inode_bh); - if (status < 0) - mlog_errno(status); + ocfs2_journal_dirty(handle, old_inode_bh); } else mlog_errno(status); @@ -1362,7 +1528,7 @@ static int ocfs2_rename(struct inode *old_dir, } if (new_inode) { - new_inode->i_nlink--; + drop_nlink(new_inode); new_inode->i_ctime = CURRENT_TIME; } old_dir->i_ctime = old_dir->i_mtime = CURRENT_TIME; @@ -1370,9 +1536,9 @@ static int ocfs2_rename(struct inode *old_dir, if (update_dot_dot) { status = ocfs2_update_entry(old_inode, handle, &old_inode_dot_dot_res, new_dir); - old_dir->i_nlink--; + drop_nlink(old_dir); if (new_inode) { - new_inode->i_nlink--; + drop_nlink(new_inode); } else { inc_nlink(new_dir); mark_inode_dirty(new_dir); @@ -1404,12 +1570,13 @@ static int ocfs2_rename(struct inode *old_dir, (int)old_dir_nlink, old_dir->i_nlink); } else { struct ocfs2_dinode *fe; - status = ocfs2_journal_access_di(handle, old_dir, - old_dir_bh, - OCFS2_JOURNAL_ACCESS_WRITE); + status = ocfs2_journal_access_di(handle, + INODE_CACHE(old_dir), + old_dir_bh, + OCFS2_JOURNAL_ACCESS_WRITE); fe = (struct ocfs2_dinode *) old_dir_bh->b_data; ocfs2_set_links_count(fe, old_dir->i_nlink); - status = ocfs2_journal_dirty(handle, old_dir_bh); + ocfs2_journal_dirty(handle, old_dir_bh); } } ocfs2_dentry_move(old_dentry, new_dentry, old_dir, new_dir); @@ -1454,7 +1621,8 @@ bail: brelse(old_dir_bh); brelse(new_dir_bh); - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -1479,9 +1647,8 @@ static int ocfs2_create_symlink_data(struct ocfs2_super *osb, * write i_size + 1 bytes. */ blocks = (bytes_left + sb->s_blocksize - 1) >> sb->s_blocksize_bits; - mlog_entry("i_blocks = %llu, i_size = %llu, blocks = %d\n", - (unsigned long long)inode->i_blocks, - i_size_read(inode), blocks); + trace_ocfs2_create_symlink_data((unsigned long long)inode->i_blocks, + i_size_read(inode), blocks); /* Sanity check -- make sure we're going to fit. */ if (bytes_left > @@ -1524,9 +1691,11 @@ static int ocfs2_create_symlink_data(struct ocfs2_super *osb, mlog_errno(status); goto bail; } - ocfs2_set_new_buffer_uptodate(inode, bhs[virtual]); + ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode), + bhs[virtual]); - status = ocfs2_journal_access(handle, inode, bhs[virtual], + status = ocfs2_journal_access(handle, INODE_CACHE(inode), + bhs[virtual], OCFS2_JOURNAL_ACCESS_CREATE); if (status < 0) { mlog_errno(status); @@ -1539,11 +1708,7 @@ static int ocfs2_create_symlink_data(struct ocfs2_super *osb, (bytes_left > sb->s_blocksize) ? sb->s_blocksize : bytes_left); - status = ocfs2_journal_dirty(handle, bhs[virtual]); - if (status < 0) { - mlog_errno(status); - goto bail; - } + ocfs2_journal_dirty(handle, bhs[virtual]); virtual++; p_blkno++; @@ -1559,7 +1724,8 @@ bail: kfree(bhs); } - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -1587,9 +1753,14 @@ static int ocfs2_symlink(struct inode *dir, }; int did_quota = 0, did_quota_inode = 0; struct ocfs2_dir_lookup_result lookup = { NULL, }; + sigset_t oldset; + int did_block_signals = 0; + struct ocfs2_dentry_lock *dl = NULL; - mlog_entry("(0x%p, 0x%p, symname='%s' actual='%.*s')\n", dir, - dentry, symname, dentry->d_name.len, dentry->d_name.name); + trace_ocfs2_symlink_begin(dir, dentry, symname, + dentry->d_name.len, dentry->d_name.name); + + dquot_initialize(dir); sb = dir->i_sb; osb = OCFS2_SB(sb); @@ -1641,7 +1812,7 @@ static int ocfs2_symlink(struct inode *dir, } /* get security xattr */ - status = ocfs2_init_security_get(inode, dir, &si); + status = ocfs2_init_security_get(inode, dir, &dentry->d_name, &si); if (status) { if (status == -EOPNOTSUPP) si.enable = 0; @@ -1680,16 +1851,21 @@ static int ocfs2_symlink(struct inode *dir, goto bail; } - /* We don't use standard VFS wrapper because we don't want vfs_dq_init - * to be called. */ - if (sb_any_quota_active(osb->sb) && - osb->sb->dq_op->alloc_inode(inode, 1) == NO_QUOTA) { - status = -EDQUOT; + /* Starting to change things, restart is no longer possible. */ + ocfs2_block_signals(&oldset); + did_block_signals = 1; + + status = dquot_alloc_inode(inode); + if (status) goto bail; - } did_quota_inode = 1; - status = ocfs2_mknod_locked(osb, dir, inode, dentry, + trace_ocfs2_symlink_create(dir, dentry, dentry->d_name.len, + dentry->d_name.name, + (unsigned long long)OCFS2_I(dir)->ip_blkno, + inode->i_mode); + + status = ocfs2_mknod_locked(osb, dir, inode, 0, &new_fe_bh, parent_fe_bh, handle, inode_ac); if (status < 0) { @@ -1700,16 +1876,16 @@ static int ocfs2_symlink(struct inode *dir, fe = (struct ocfs2_dinode *) new_fe_bh->b_data; inode->i_rdev = 0; newsize = l - 1; + inode->i_op = &ocfs2_symlink_inode_operations; if (l > ocfs2_fast_symlink_chars(sb)) { u32 offset = 0; - inode->i_op = &ocfs2_symlink_inode_operations; - if (vfs_dq_alloc_space_nodirty(inode, - ocfs2_clusters_to_bytes(osb->sb, 1))) { - status = -EDQUOT; + status = dquot_alloc_space_nodirty(inode, + ocfs2_clusters_to_bytes(osb->sb, 1)); + if (status) goto bail; - } did_quota = 1; + inode->i_mapping->a_ops = &ocfs2_aops; status = ocfs2_add_inode_data(osb, inode, &offset, 1, 0, new_fe_bh, handle, data_ac, NULL, @@ -1727,7 +1903,7 @@ static int ocfs2_symlink(struct inode *dir, i_size_write(inode, newsize); inode->i_blocks = ocfs2_inode_sector_count(inode); } else { - inode->i_op = &ocfs2_fast_symlink_inode_operations; + inode->i_mapping->a_ops = &ocfs2_fast_symlink_aops; memcpy((char *) fe->id2.i_symlink, symname, l); i_size_write(inode, newsize); inode->i_blocks = 0; @@ -1757,37 +1933,44 @@ static int ocfs2_symlink(struct inode *dir, } } - status = ocfs2_add_entry(handle, dentry, inode, - le64_to_cpu(fe->i_blkno), parent_fe_bh, - &lookup); - if (status < 0) { + /* + * Do this before adding the entry to the directory. We add + * also set d_op after success so that ->d_iput() will cleanup + * the dentry lock even if ocfs2_add_entry() fails below. + */ + status = ocfs2_dentry_attach_lock(dentry, inode, OCFS2_I(dir)->ip_blkno); + if (status) { mlog_errno(status); goto bail; } - status = ocfs2_dentry_attach_lock(dentry, inode, OCFS2_I(dir)->ip_blkno); - if (status) { + dl = dentry->d_fsdata; + + status = ocfs2_add_entry(handle, dentry, inode, + le64_to_cpu(fe->i_blkno), parent_fe_bh, + &lookup); + if (status < 0) { mlog_errno(status); goto bail; } insert_inode_hash(inode); - dentry->d_op = &ocfs2_dentry_ops; d_instantiate(dentry, inode); bail: if (status < 0 && did_quota) - vfs_dq_free_space_nodirty(inode, + dquot_free_space_nodirty(inode, ocfs2_clusters_to_bytes(osb->sb, 1)); if (status < 0 && did_quota_inode) - vfs_dq_free_inode(inode); + dquot_free_inode(inode); if (handle) ocfs2_commit_trans(osb, handle); ocfs2_inode_unlock(dir, 1); + if (did_block_signals) + ocfs2_unblock_signals(&oldset); brelse(new_fe_bh); brelse(parent_fe_bh); - kfree(si.name); kfree(si.value); ocfs2_free_dir_lookup_result(&lookup); if (inode_ac) @@ -1797,11 +1980,16 @@ bail: if (xattr_ac) ocfs2_free_alloc_context(xattr_ac); if ((status < 0) && inode) { + if (dl) + ocfs2_cleanup_add_entry_failure(osb, dentry, inode); + + OCFS2_I(inode)->ip_flags |= OCFS2_INODE_SKIP_ORPHAN_DIR; clear_nlink(inode); iput(inode); } - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -1810,8 +1998,6 @@ static int ocfs2_blkno_stringify(u64 blkno, char *name) { int status, namelen; - mlog_entry_void(); - namelen = snprintf(name, OCFS2_ORPHAN_NAMELEN + 1, "%016llx", (long long)blkno); if (namelen <= 0) { @@ -1828,76 +2014,133 @@ static int ocfs2_blkno_stringify(u64 blkno, char *name) goto bail; } - mlog(0, "built filename '%s' for orphan dir (len=%d)\n", name, - namelen); + trace_ocfs2_blkno_stringify(blkno, name, namelen); status = 0; bail: - mlog_exit(status); + if (status < 0) + mlog_errno(status); return status; } -static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb, - struct inode **ret_orphan_dir, - struct inode *inode, - char *name, - struct ocfs2_dir_lookup_result *lookup) +static int ocfs2_lookup_lock_orphan_dir(struct ocfs2_super *osb, + struct inode **ret_orphan_dir, + struct buffer_head **ret_orphan_dir_bh) { struct inode *orphan_dir_inode; struct buffer_head *orphan_dir_bh = NULL; - int status = 0; - - status = ocfs2_blkno_stringify(OCFS2_I(inode)->ip_blkno, name); - if (status < 0) { - mlog_errno(status); - return status; - } + int ret = 0; orphan_dir_inode = ocfs2_get_system_file_inode(osb, ORPHAN_DIR_SYSTEM_INODE, osb->slot_num); if (!orphan_dir_inode) { - status = -ENOENT; - mlog_errno(status); - return status; + ret = -ENOENT; + mlog_errno(ret); + return ret; } mutex_lock(&orphan_dir_inode->i_mutex); - status = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1); - if (status < 0) { - mlog_errno(status); - goto leave; + ret = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1); + if (ret < 0) { + mutex_unlock(&orphan_dir_inode->i_mutex); + iput(orphan_dir_inode); + + mlog_errno(ret); + return ret; } - status = ocfs2_prepare_dir_for_insert(osb, orphan_dir_inode, - orphan_dir_bh, name, - OCFS2_ORPHAN_NAMELEN, lookup); - if (status < 0) { - ocfs2_inode_unlock(orphan_dir_inode, 1); + *ret_orphan_dir = orphan_dir_inode; + *ret_orphan_dir_bh = orphan_dir_bh; - mlog_errno(status); - goto leave; + return 0; +} + +static int __ocfs2_prepare_orphan_dir(struct inode *orphan_dir_inode, + struct buffer_head *orphan_dir_bh, + u64 blkno, + char *name, + struct ocfs2_dir_lookup_result *lookup) +{ + int ret; + struct ocfs2_super *osb = OCFS2_SB(orphan_dir_inode->i_sb); + + ret = ocfs2_blkno_stringify(blkno, name); + if (ret < 0) { + mlog_errno(ret); + return ret; + } + + ret = ocfs2_prepare_dir_for_insert(osb, orphan_dir_inode, + orphan_dir_bh, name, + OCFS2_ORPHAN_NAMELEN, lookup); + if (ret < 0) { + mlog_errno(ret); + return ret; + } + + return 0; +} + +/** + * ocfs2_prepare_orphan_dir() - Prepare an orphan directory for + * insertion of an orphan. + * @osb: ocfs2 file system + * @ret_orphan_dir: Orphan dir inode - returned locked! + * @blkno: Actual block number of the inode to be inserted into orphan dir. + * @lookup: dir lookup result, to be passed back into functions like + * ocfs2_orphan_add + * + * Returns zero on success and the ret_orphan_dir, name and lookup + * fields will be populated. + * + * Returns non-zero on failure. + */ +static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb, + struct inode **ret_orphan_dir, + u64 blkno, + char *name, + struct ocfs2_dir_lookup_result *lookup) +{ + struct inode *orphan_dir_inode = NULL; + struct buffer_head *orphan_dir_bh = NULL; + int ret = 0; + + ret = ocfs2_lookup_lock_orphan_dir(osb, &orphan_dir_inode, + &orphan_dir_bh); + if (ret < 0) { + mlog_errno(ret); + return ret; + } + + ret = __ocfs2_prepare_orphan_dir(orphan_dir_inode, orphan_dir_bh, + blkno, name, lookup); + if (ret < 0) { + mlog_errno(ret); + goto out; } *ret_orphan_dir = orphan_dir_inode; -leave: - if (status) { +out: + brelse(orphan_dir_bh); + + if (ret) { + ocfs2_inode_unlock(orphan_dir_inode, 1); mutex_unlock(&orphan_dir_inode->i_mutex); iput(orphan_dir_inode); } - brelse(orphan_dir_bh); - - mlog_exit(status); - return status; + if (ret) + mlog_errno(ret); + return ret; } static int ocfs2_orphan_add(struct ocfs2_super *osb, handle_t *handle, struct inode *inode, - struct ocfs2_dinode *fe, + struct buffer_head *fe_bh, char *name, struct ocfs2_dir_lookup_result *lookup, struct inode *orphan_dir_inode) @@ -1905,8 +2148,10 @@ static int ocfs2_orphan_add(struct ocfs2_super *osb, struct buffer_head *orphan_dir_bh = NULL; int status = 0; struct ocfs2_dinode *orphan_fe; + struct ocfs2_dinode *fe = (struct ocfs2_dinode *) fe_bh->b_data; - mlog_entry("(inode->i_ino = %lu)\n", inode->i_ino); + trace_ocfs2_orphan_add_begin( + (unsigned long long)OCFS2_I(inode)->ip_blkno); status = ocfs2_read_inode_block(orphan_dir_inode, &orphan_dir_bh); if (status < 0) { @@ -1914,7 +2159,24 @@ static int ocfs2_orphan_add(struct ocfs2_super *osb, goto leave; } - status = ocfs2_journal_access_di(handle, orphan_dir_inode, orphan_dir_bh, + status = ocfs2_journal_access_di(handle, + INODE_CACHE(orphan_dir_inode), + orphan_dir_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (status < 0) { + mlog_errno(status); + goto leave; + } + + /* + * We're going to journal the change of i_flags and i_orphaned_slot. + * It's safe anyway, though some callers may duplicate the journaling. + * Journaling within the func just make the logic look more + * straightforward. + */ + status = ocfs2_journal_access_di(handle, + INODE_CACHE(inode), + fe_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); @@ -1926,13 +2188,8 @@ static int ocfs2_orphan_add(struct ocfs2_super *osb, orphan_fe = (struct ocfs2_dinode *) orphan_dir_bh->b_data; if (S_ISDIR(inode->i_mode)) ocfs2_add_links_count(orphan_fe, 1); - orphan_dir_inode->i_nlink = ocfs2_read_links_count(orphan_fe); - - status = ocfs2_journal_dirty(handle, orphan_dir_bh); - if (status < 0) { - mlog_errno(status); - goto leave; - } + set_nlink(orphan_dir_inode, ocfs2_read_links_count(orphan_fe)); + ocfs2_journal_dirty(handle, orphan_dir_bh); status = __ocfs2_add_entry(handle, orphan_dir_inode, name, OCFS2_ORPHAN_NAMELEN, inode, @@ -1940,23 +2197,32 @@ static int ocfs2_orphan_add(struct ocfs2_super *osb, orphan_dir_bh, lookup); if (status < 0) { mlog_errno(status); - goto leave; + goto rollback; } - le32_add_cpu(&fe->i_flags, OCFS2_ORPHANED_FL); + fe->i_flags |= cpu_to_le32(OCFS2_ORPHANED_FL); + OCFS2_I(inode)->ip_flags &= ~OCFS2_INODE_SKIP_ORPHAN_DIR; /* Record which orphan dir our inode now resides * in. delete_inode will use this to determine which orphan * dir to lock. */ fe->i_orphaned_slot = cpu_to_le16(osb->slot_num); - mlog(0, "Inode %llu orphaned in slot %d\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, osb->slot_num); + ocfs2_journal_dirty(handle, fe_bh); + + trace_ocfs2_orphan_add_end((unsigned long long)OCFS2_I(inode)->ip_blkno, + osb->slot_num); + +rollback: + if (status < 0) { + if (S_ISDIR(inode->i_mode)) + ocfs2_add_links_count(orphan_fe, -1); + set_nlink(orphan_dir_inode, ocfs2_read_links_count(orphan_fe)); + } leave: brelse(orphan_dir_bh); - mlog_exit(status); return status; } @@ -1972,17 +2238,15 @@ int ocfs2_orphan_del(struct ocfs2_super *osb, int status = 0; struct ocfs2_dir_lookup_result lookup = { NULL, }; - mlog_entry_void(); - status = ocfs2_blkno_stringify(OCFS2_I(inode)->ip_blkno, name); if (status < 0) { mlog_errno(status); goto leave; } - mlog(0, "removing '%s' from orphan dir %llu (namelen=%d)\n", - name, (unsigned long long)OCFS2_I(orphan_dir_inode)->ip_blkno, - OCFS2_ORPHAN_NAMELEN); + trace_ocfs2_orphan_del( + (unsigned long long)OCFS2_I(orphan_dir_inode)->ip_blkno, + name, OCFS2_ORPHAN_NAMELEN); /* find it's spot in the orphan directory */ status = ocfs2_find_entry(name, OCFS2_ORPHAN_NAMELEN, orphan_dir_inode, @@ -1999,7 +2263,9 @@ int ocfs2_orphan_del(struct ocfs2_super *osb, goto leave; } - status = ocfs2_journal_access_di(handle,orphan_dir_inode, orphan_dir_bh, + status = ocfs2_journal_access_di(handle, + INODE_CACHE(orphan_dir_inode), + orphan_dir_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); @@ -2010,18 +2276,368 @@ int ocfs2_orphan_del(struct ocfs2_super *osb, orphan_fe = (struct ocfs2_dinode *) orphan_dir_bh->b_data; if (S_ISDIR(inode->i_mode)) ocfs2_add_links_count(orphan_fe, -1); - orphan_dir_inode->i_nlink = ocfs2_read_links_count(orphan_fe); + set_nlink(orphan_dir_inode, ocfs2_read_links_count(orphan_fe)); + ocfs2_journal_dirty(handle, orphan_dir_bh); + +leave: + ocfs2_free_dir_lookup_result(&lookup); + + if (status) + mlog_errno(status); + return status; +} + +/** + * ocfs2_prep_new_orphaned_file() - Prepare the orphan dir to receive a newly + * allocated file. This is different from the typical 'add to orphan dir' + * operation in that the inode does not yet exist. This is a problem because + * the orphan dir stringifies the inode block number to come up with it's + * dirent. Obviously if the inode does not yet exist we have a chicken and egg + * problem. This function works around it by calling deeper into the orphan + * and suballoc code than other callers. Use this only by necessity. + * @dir: The directory which this inode will ultimately wind up under - not the + * orphan dir! + * @dir_bh: buffer_head the @dir inode block + * @orphan_name: string of length (CFS2_ORPHAN_NAMELEN + 1). Will be filled + * with the string to be used for orphan dirent. Pass back to the orphan dir + * code. + * @ret_orphan_dir: orphan dir inode returned to be passed back into orphan + * dir code. + * @ret_di_blkno: block number where the new inode will be allocated. + * @orphan_insert: Dir insert context to be passed back into orphan dir code. + * @ret_inode_ac: Inode alloc context to be passed back to the allocator. + * + * Returns zero on success and the ret_orphan_dir, name and lookup + * fields will be populated. + * + * Returns non-zero on failure. + */ +static int ocfs2_prep_new_orphaned_file(struct inode *dir, + struct buffer_head *dir_bh, + char *orphan_name, + struct inode **ret_orphan_dir, + u64 *ret_di_blkno, + struct ocfs2_dir_lookup_result *orphan_insert, + struct ocfs2_alloc_context **ret_inode_ac) +{ + int ret; + u64 di_blkno; + struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); + struct inode *orphan_dir = NULL; + struct buffer_head *orphan_dir_bh = NULL; + struct ocfs2_alloc_context *inode_ac = NULL; + + ret = ocfs2_lookup_lock_orphan_dir(osb, &orphan_dir, &orphan_dir_bh); + if (ret < 0) { + mlog_errno(ret); + return ret; + } + + /* reserve an inode spot */ + ret = ocfs2_reserve_new_inode(osb, &inode_ac); + if (ret < 0) { + if (ret != -ENOSPC) + mlog_errno(ret); + goto out; + } + + ret = ocfs2_find_new_inode_loc(dir, dir_bh, inode_ac, + &di_blkno); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = __ocfs2_prepare_orphan_dir(orphan_dir, orphan_dir_bh, + di_blkno, orphan_name, orphan_insert); + if (ret < 0) { + mlog_errno(ret); + goto out; + } + +out: + if (ret == 0) { + *ret_orphan_dir = orphan_dir; + *ret_di_blkno = di_blkno; + *ret_inode_ac = inode_ac; + /* + * orphan_name and orphan_insert are already up to + * date via prepare_orphan_dir + */ + } else { + /* Unroll reserve_new_inode* */ + if (inode_ac) + ocfs2_free_alloc_context(inode_ac); + + /* Unroll orphan dir locking */ + mutex_unlock(&orphan_dir->i_mutex); + ocfs2_inode_unlock(orphan_dir, 1); + iput(orphan_dir); + } + + brelse(orphan_dir_bh); + + return ret; +} + +int ocfs2_create_inode_in_orphan(struct inode *dir, + int mode, + struct inode **new_inode) +{ + int status, did_quota_inode = 0; + struct inode *inode = NULL; + struct inode *orphan_dir = NULL; + struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); + struct ocfs2_dinode *di = NULL; + handle_t *handle = NULL; + char orphan_name[OCFS2_ORPHAN_NAMELEN + 1]; + struct buffer_head *parent_di_bh = NULL; + struct buffer_head *new_di_bh = NULL; + struct ocfs2_alloc_context *inode_ac = NULL; + struct ocfs2_dir_lookup_result orphan_insert = { NULL, }; + u64 uninitialized_var(di_blkno), suballoc_loc; + u16 suballoc_bit; + + status = ocfs2_inode_lock(dir, &parent_di_bh, 1); + if (status < 0) { + if (status != -ENOENT) + mlog_errno(status); + return status; + } + + status = ocfs2_prep_new_orphaned_file(dir, parent_di_bh, + orphan_name, &orphan_dir, + &di_blkno, &orphan_insert, &inode_ac); + if (status < 0) { + if (status != -ENOSPC) + mlog_errno(status); + goto leave; + } + + inode = ocfs2_get_init_inode(dir, mode); + if (!inode) { + status = -ENOMEM; + mlog_errno(status); + goto leave; + } + + handle = ocfs2_start_trans(osb, ocfs2_mknod_credits(osb->sb, 0, 0)); + if (IS_ERR(handle)) { + status = PTR_ERR(handle); + handle = NULL; + mlog_errno(status); + goto leave; + } + + status = dquot_alloc_inode(inode); + if (status) + goto leave; + did_quota_inode = 1; + + status = ocfs2_claim_new_inode_at_loc(handle, dir, inode_ac, + &suballoc_loc, + &suballoc_bit, di_blkno); + if (status < 0) { + mlog_errno(status); + goto leave; + } + + clear_nlink(inode); + /* do the real work now. */ + status = __ocfs2_mknod_locked(dir, inode, + 0, &new_di_bh, parent_di_bh, handle, + inode_ac, di_blkno, suballoc_loc, + suballoc_bit); + if (status < 0) { + mlog_errno(status); + goto leave; + } - status = ocfs2_journal_dirty(handle, orphan_dir_bh); + di = (struct ocfs2_dinode *)new_di_bh->b_data; + status = ocfs2_orphan_add(osb, handle, inode, new_di_bh, orphan_name, + &orphan_insert, orphan_dir); if (status < 0) { mlog_errno(status); goto leave; } + /* get open lock so that only nodes can't remove it from orphan dir. */ + status = ocfs2_open_lock(inode); + if (status < 0) + mlog_errno(status); + + insert_inode_hash(inode); leave: + if (status < 0 && did_quota_inode) + dquot_free_inode(inode); + if (handle) + ocfs2_commit_trans(osb, handle); + + if (orphan_dir) { + /* This was locked for us in ocfs2_prepare_orphan_dir() */ + ocfs2_inode_unlock(orphan_dir, 1); + mutex_unlock(&orphan_dir->i_mutex); + iput(orphan_dir); + } + + if ((status < 0) && inode) { + clear_nlink(inode); + iput(inode); + } + + if (inode_ac) + ocfs2_free_alloc_context(inode_ac); + + brelse(new_di_bh); + + if (!status) + *new_inode = inode; + + ocfs2_free_dir_lookup_result(&orphan_insert); + + ocfs2_inode_unlock(dir, 1); + brelse(parent_di_bh); + return status; +} + +int ocfs2_mv_orphaned_inode_to_new(struct inode *dir, + struct inode *inode, + struct dentry *dentry) +{ + int status = 0; + struct buffer_head *parent_di_bh = NULL; + handle_t *handle = NULL; + struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); + struct ocfs2_dinode *dir_di, *di; + struct inode *orphan_dir_inode = NULL; + struct buffer_head *orphan_dir_bh = NULL; + struct buffer_head *di_bh = NULL; + struct ocfs2_dir_lookup_result lookup = { NULL, }; + + trace_ocfs2_mv_orphaned_inode_to_new(dir, dentry, + dentry->d_name.len, dentry->d_name.name, + (unsigned long long)OCFS2_I(dir)->ip_blkno, + (unsigned long long)OCFS2_I(inode)->ip_blkno); + + status = ocfs2_inode_lock(dir, &parent_di_bh, 1); + if (status < 0) { + if (status != -ENOENT) + mlog_errno(status); + return status; + } + + dir_di = (struct ocfs2_dinode *) parent_di_bh->b_data; + if (!dir_di->i_links_count) { + /* can't make a file in a deleted directory. */ + status = -ENOENT; + goto leave; + } + + status = ocfs2_check_dir_for_entry(dir, dentry->d_name.name, + dentry->d_name.len); + if (status) + goto leave; + + /* get a spot inside the dir. */ + status = ocfs2_prepare_dir_for_insert(osb, dir, parent_di_bh, + dentry->d_name.name, + dentry->d_name.len, &lookup); + if (status < 0) { + mlog_errno(status); + goto leave; + } + + orphan_dir_inode = ocfs2_get_system_file_inode(osb, + ORPHAN_DIR_SYSTEM_INODE, + osb->slot_num); + if (!orphan_dir_inode) { + status = -EEXIST; + mlog_errno(status); + goto leave; + } + + mutex_lock(&orphan_dir_inode->i_mutex); + + status = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1); + if (status < 0) { + mlog_errno(status); + mutex_unlock(&orphan_dir_inode->i_mutex); + iput(orphan_dir_inode); + goto leave; + } + + status = ocfs2_read_inode_block(inode, &di_bh); + if (status < 0) { + mlog_errno(status); + goto orphan_unlock; + } + + handle = ocfs2_start_trans(osb, ocfs2_rename_credits(osb->sb)); + if (IS_ERR(handle)) { + status = PTR_ERR(handle); + handle = NULL; + mlog_errno(status); + goto orphan_unlock; + } + + status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), + di_bh, OCFS2_JOURNAL_ACCESS_WRITE); + if (status < 0) { + mlog_errno(status); + goto out_commit; + } + + status = ocfs2_orphan_del(osb, handle, orphan_dir_inode, inode, + orphan_dir_bh); + if (status < 0) { + mlog_errno(status); + goto out_commit; + } + + di = (struct ocfs2_dinode *)di_bh->b_data; + di->i_flags &= ~cpu_to_le32(OCFS2_ORPHANED_FL); + di->i_orphaned_slot = 0; + set_nlink(inode, 1); + ocfs2_set_links_count(di, inode->i_nlink); + ocfs2_update_inode_fsync_trans(handle, inode, 1); + ocfs2_journal_dirty(handle, di_bh); + + status = ocfs2_add_entry(handle, dentry, inode, + OCFS2_I(inode)->ip_blkno, parent_di_bh, + &lookup); + if (status < 0) { + mlog_errno(status); + goto out_commit; + } + + status = ocfs2_dentry_attach_lock(dentry, inode, + OCFS2_I(dir)->ip_blkno); + if (status) { + mlog_errno(status); + goto out_commit; + } + + d_instantiate(dentry, inode); + status = 0; +out_commit: + ocfs2_commit_trans(osb, handle); +orphan_unlock: + ocfs2_inode_unlock(orphan_dir_inode, 1); + mutex_unlock(&orphan_dir_inode->i_mutex); + iput(orphan_dir_inode); +leave: + + ocfs2_inode_unlock(dir, 1); + + brelse(di_bh); + brelse(parent_di_bh); + brelse(orphan_dir_bh); + ocfs2_free_dir_lookup_result(&lookup); - mlog_exit(status); + if (status) + mlog_errno(status); + return status; } @@ -2042,4 +2658,7 @@ const struct inode_operations ocfs2_dir_iops = { .getxattr = generic_getxattr, .listxattr = ocfs2_listxattr, .removexattr = generic_removexattr, + .fiemap = ocfs2_fiemap, + .get_acl = ocfs2_iop_get_acl, + .set_acl = ocfs2_iop_set_acl, }; diff --git a/fs/ocfs2/namei.h b/fs/ocfs2/namei.h index 688aef64c87..e5d059d4f11 100644 --- a/fs/ocfs2/namei.h +++ b/fs/ocfs2/namei.h @@ -35,5 +35,11 @@ int ocfs2_orphan_del(struct ocfs2_super *osb, struct inode *orphan_dir_inode, struct inode *inode, struct buffer_head *orphan_dir_bh); +int ocfs2_create_inode_in_orphan(struct inode *dir, + int mode, + struct inode **new_inode); +int ocfs2_mv_orphaned_inode_to_new(struct inode *dir, + struct inode *new_inode, + struct dentry *new_dentry); #endif /* OCFS2_NAMEI_H */ diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h index 1386281950d..bbec539230f 100644 --- a/fs/ocfs2/ocfs2.h +++ b/fs/ocfs2/ocfs2.h @@ -30,37 +30,70 @@ #include <linux/sched.h> #include <linux/wait.h> #include <linux/list.h> +#include <linux/llist.h> #include <linux/rbtree.h> #include <linux/workqueue.h> #include <linux/kref.h> #include <linux/mutex.h> -#ifndef CONFIG_OCFS2_COMPAT_JBD -# include <linux/jbd2.h> -#else -# include <linux/jbd.h> -# include "ocfs2_jbd_compat.h" -#endif +#include <linux/lockdep.h> +#include <linux/jbd2.h> /* For union ocfs2_dlm_lksb */ #include "stackglue.h" #include "ocfs2_fs.h" #include "ocfs2_lockid.h" +#include "ocfs2_ioctl.h" + +/* For struct ocfs2_blockcheck_stats */ +#include "blockcheck.h" + +#include "reservations.h" + +/* Caching of metadata buffers */ /* Most user visible OCFS2 inodes will have very few pieces of * metadata, but larger files (including bitmaps, etc) must be taken * into account when designing an access scheme. We allow a small * amount of inlined blocks to be stored on an array and grow the * structure into a rb tree when necessary. */ -#define OCFS2_INODE_MAX_CACHE_ARRAY 2 +#define OCFS2_CACHE_INFO_MAX_ARRAY 2 + +/* Flags for ocfs2_caching_info */ +enum ocfs2_caching_info_flags { + /* Indicates that the metadata cache is using the inline array */ + OCFS2_CACHE_FL_INLINE = 1<<1, +}; + +struct ocfs2_caching_operations; struct ocfs2_caching_info { + /* + * The parent structure provides the locks, but because the + * parent structure can differ, it provides locking operations + * to struct ocfs2_caching_info. + */ + const struct ocfs2_caching_operations *ci_ops; + + /* next two are protected by trans_inc_lock */ + /* which transaction were we created on? Zero if none. */ + unsigned long ci_created_trans; + /* last transaction we were a part of. */ + unsigned long ci_last_trans; + + /* Cache structures */ + unsigned int ci_flags; unsigned int ci_num_cached; union { - sector_t ci_array[OCFS2_INODE_MAX_CACHE_ARRAY]; + sector_t ci_array[OCFS2_CACHE_INFO_MAX_ARRAY]; struct rb_root ci_tree; } ci_cache; }; +/* + * Need this prototype here instead of in uptodate.h because journal.h + * uses it. + */ +struct super_block *ocfs2_metadata_cache_get_super(struct ocfs2_caching_info *ci); /* this limits us to 256 nodes * if we need more, we can do a kmalloc for the map */ @@ -106,49 +139,84 @@ enum ocfs2_unlock_action { #define OCFS2_LOCK_PENDING (0x00000400) /* This lockres is pending a call to dlm_lock. Only exists with BUSY set. */ +#define OCFS2_LOCK_UPCONVERT_FINISHING (0x00000800) /* blocks the dc thread + * from downconverting + * before the upconvert + * has completed */ struct ocfs2_lock_res_ops; typedef void (*ocfs2_lock_callback)(int status, unsigned long data); +#ifdef CONFIG_OCFS2_FS_STATS +struct ocfs2_lock_stats { + u64 ls_total; /* Total wait in NSEC */ + u32 ls_gets; /* Num acquires */ + u32 ls_fail; /* Num failed acquires */ + + /* Storing max wait in usecs saves 24 bytes per inode */ + u32 ls_max; /* Max wait in USEC */ +}; +#endif + struct ocfs2_lock_res { void *l_priv; struct ocfs2_lock_res_ops *l_ops; - spinlock_t l_lock; + struct list_head l_blocked_list; struct list_head l_mask_waiters; - enum ocfs2_lock_type l_type; unsigned long l_flags; char l_name[OCFS2_LOCK_ID_MAX_LEN]; - int l_level; unsigned int l_ro_holders; unsigned int l_ex_holders; - union ocfs2_dlm_lksb l_lksb; + signed char l_level; + signed char l_requested; + signed char l_blocking; + + /* Data packed - type enum ocfs2_lock_type */ + unsigned char l_type; /* used from AST/BAST funcs. */ - enum ocfs2_ast_action l_action; - enum ocfs2_unlock_action l_unlock_action; - int l_requested; - int l_blocking; + /* Data packed - enum type ocfs2_ast_action */ + unsigned char l_action; + /* Data packed - enum type ocfs2_unlock_action */ + unsigned char l_unlock_action; unsigned int l_pending_gen; + spinlock_t l_lock; + + struct ocfs2_dlm_lksb l_lksb; + wait_queue_head_t l_event; struct list_head l_debug_list; #ifdef CONFIG_OCFS2_FS_STATS - unsigned long long l_lock_num_prmode; /* PR acquires */ - unsigned long long l_lock_num_exmode; /* EX acquires */ - unsigned int l_lock_num_prmode_failed; /* Failed PR gets */ - unsigned int l_lock_num_exmode_failed; /* Failed EX gets */ - unsigned long long l_lock_total_prmode; /* Tot wait for PR */ - unsigned long long l_lock_total_exmode; /* Tot wait for EX */ - unsigned int l_lock_max_prmode; /* Max wait for PR */ - unsigned int l_lock_max_exmode; /* Max wait for EX */ - unsigned int l_lock_refresh; /* Disk refreshes */ + struct ocfs2_lock_stats l_lock_prmode; /* PR mode stats */ + u32 l_lock_refresh; /* Disk refreshes */ + struct ocfs2_lock_stats l_lock_exmode; /* EX mode stats */ #endif +#ifdef CONFIG_DEBUG_LOCK_ALLOC + struct lockdep_map l_lockdep_map; +#endif +}; + +enum ocfs2_orphan_scan_state { + ORPHAN_SCAN_ACTIVE, + ORPHAN_SCAN_INACTIVE +}; + +struct ocfs2_orphan_scan { + struct mutex os_lock; + struct ocfs2_super *os_osb; + struct ocfs2_lock_res os_lockres; /* lock to synchronize scans */ + struct delayed_work os_orphan_scan_work; + struct timespec os_scantime; /* time this node ran the scan */ + u32 os_count; /* tracks node specific scans */ + u32 os_seqno; /* tracks cluster wide scans */ + atomic_t os_state; /* ACTIVE or INACTIVE */ }; struct ocfs2_dlm_debug { @@ -188,7 +256,7 @@ enum ocfs2_local_alloc_state enum ocfs2_mount_options { - OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Heartbeat started in local mode */ + OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Local heartbeat */ OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */ OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */ OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */ @@ -196,9 +264,15 @@ enum ocfs2_mount_options OCFS2_MOUNT_LOCALFLOCKS = 1 << 5, /* No cluster aware user file locks */ OCFS2_MOUNT_NOUSERXATTR = 1 << 6, /* No user xattr */ OCFS2_MOUNT_INODE64 = 1 << 7, /* Allow inode numbers > 2^32 */ - OCFS2_MOUNT_POSIX_ACL = 1 << 8, /* POSIX access control lists */ - OCFS2_MOUNT_USRQUOTA = 1 << 9, /* We support user quotas */ - OCFS2_MOUNT_GRPQUOTA = 1 << 10, /* We support group quotas */ + OCFS2_MOUNT_POSIX_ACL = 1 << 8, /* Force POSIX access control lists */ + OCFS2_MOUNT_NO_POSIX_ACL = 1 << 9, /* Disable POSIX access + control lists */ + OCFS2_MOUNT_USRQUOTA = 1 << 10, /* We support user quotas */ + OCFS2_MOUNT_GRPQUOTA = 1 << 11, /* We support group quotas */ + OCFS2_MOUNT_COHERENCY_BUFFERED = 1 << 12, /* Allow concurrent O_DIRECT + writes */ + OCFS2_MOUNT_HB_NONE = 1 << 13, /* No heartbeat */ + OCFS2_MOUNT_HB_GLOBAL = 1 << 14, /* Global heartbeat */ }; #define OCFS2_OSB_SOFT_RO 0x0001 @@ -211,14 +285,14 @@ struct ocfs2_slot_info; struct ocfs2_recovery_map; struct ocfs2_replay_map; struct ocfs2_quota_recovery; -struct ocfs2_dentry_lock; struct ocfs2_super { struct task_struct *commit_task; struct super_block *sb; struct inode *root_inode; struct inode *sys_root_inode; - struct inode *system_inodes[NUM_SYSTEM_INODES]; + struct inode *global_system_inodes[NUM_GLOBAL_SYSTEM_INODES]; + struct inode **local_system_inodes; struct ocfs2_slot_info *slot_info; @@ -248,7 +322,9 @@ struct ocfs2_super u32 s_next_generation; unsigned long osb_flags; s16 s_inode_steal_slot; + s16 s_meta_steal_slot; atomic_t s_num_inodes_stolen; + atomic_t s_num_meta_stolen; unsigned long s_mount_opt; unsigned int s_atime_quantum; @@ -269,7 +345,6 @@ struct ocfs2_super struct task_struct *recovery_thread_task; int disable_recovery; wait_queue_head_t checkpoint_event; - atomic_t needs_checkpoint; struct ocfs2_journal *journal; unsigned long osb_commit_interval; @@ -281,6 +356,9 @@ struct ocfs2_super */ unsigned int local_alloc_bits; unsigned int local_alloc_default_bits; + /* osb_clusters_at_boot can become stale! Do not trust it to + * be up to date. */ + unsigned int osb_clusters_at_boot; enum ocfs2_local_alloc_state local_alloc_state; /* protected * by osb_lock */ @@ -289,16 +367,25 @@ struct ocfs2_super u64 la_last_gd; + struct ocfs2_reservation_map osb_la_resmap; + + unsigned int osb_resv_level; + unsigned int osb_dir_resv_level; + /* Next three fields are for local node slot recovery during * mount. */ int dirty; struct ocfs2_dinode *local_alloc_copy; struct ocfs2_quota_recovery *quota_rec; + struct ocfs2_blockcheck_stats osb_ecc_stats; struct ocfs2_alloc_stats alloc_stats; char dev_str[20]; /* "major,minor" of the device */ + u8 osb_stackflags; + char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1]; + char osb_cluster_name[OCFS2_CLUSTER_NAME_LEN + 1]; struct ocfs2_cluster_connection *cconn; struct ocfs2_lock_res osb_super_lockres; struct ocfs2_lock_res osb_rename_lockres; @@ -325,10 +412,9 @@ struct ocfs2_super struct list_head blocked_lock_list; unsigned long blocked_lock_count; - /* List of dentry locks to release. Anyone can add locks to - * the list, ocfs2_wq processes the list */ - struct ocfs2_dentry_lock *dentry_lock_list; - struct work_struct dentry_lock_work; + /* List of dquot structures to drop last reference to */ + struct llist_head dquot_drop_list; + struct work_struct dquot_drop_work; wait_queue_head_t osb_mount_event; @@ -336,11 +422,19 @@ struct ocfs2_super struct inode *osb_tl_inode; struct buffer_head *osb_tl_bh; struct delayed_work osb_truncate_log_wq; + atomic_t osb_tl_disable; + /* + * How many clusters in our truncate log. + * It must be protected by osb_tl_inode->i_mutex. + */ + unsigned int truncated_clusters; struct ocfs2_node_map osb_recovering_orphan_dirs; unsigned int *osb_orphan_wipes; wait_queue_head_t osb_wipe_event; + struct ocfs2_orphan_scan osb_orphan_scan; + /* used to protect metaecc calculation check of xattr. */ spinlock_t osb_xattr_lock; @@ -349,12 +443,19 @@ struct ocfs2_super /* the group we used to allocate inodes. */ u64 osb_inode_alloc_group; + + /* rb tree root for refcount lock. */ + struct rb_root osb_rf_lock_tree; + struct ocfs2_refcount_tree *osb_ref_tree_lru; + + struct mutex system_file_mutex; }; #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info) /* Useful typedef for passing around journal access functions */ -typedef int (*ocfs2_journal_access_func)(handle_t *handle, struct inode *inode, +typedef int (*ocfs2_journal_access_func)(handle_t *handle, + struct ocfs2_caching_info *ci, struct buffer_head *bh, int type); static inline int ocfs2_should_order_data(struct inode *inode) @@ -414,6 +515,13 @@ static inline int ocfs2_supports_indexed_dirs(struct ocfs2_super *osb) return 0; } +static inline int ocfs2_supports_discontig_bg(struct ocfs2_super *osb) +{ + if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG) + return 1; + return 0; +} + static inline unsigned int ocfs2_link_max(struct ocfs2_super *osb) { if (ocfs2_supports_indexed_dirs(osb)) @@ -452,6 +560,13 @@ static inline void ocfs2_add_links_count(struct ocfs2_dinode *di, int n) ocfs2_set_links_count(di, links); } +static inline int ocfs2_refcount_tree(struct ocfs2_super *osb) +{ + if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE) + return 1; + return 0; +} + /* set / clear functions because cluster events can make these happen * in parallel so we want the transitions to be atomic. this also * means that any future flags osb_flags must be protected by spinlock @@ -498,10 +613,35 @@ static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb) return ret; } -static inline int ocfs2_userspace_stack(struct ocfs2_super *osb) +static inline int ocfs2_clusterinfo_valid(struct ocfs2_super *osb) { return (osb->s_feature_incompat & - OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK); + (OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK | + OCFS2_FEATURE_INCOMPAT_CLUSTERINFO)); +} + +static inline int ocfs2_userspace_stack(struct ocfs2_super *osb) +{ + if (ocfs2_clusterinfo_valid(osb) && + memcmp(osb->osb_cluster_stack, OCFS2_CLASSIC_CLUSTER_STACK, + OCFS2_STACK_LABEL_LEN)) + return 1; + return 0; +} + +static inline int ocfs2_o2cb_stack(struct ocfs2_super *osb) +{ + if (ocfs2_clusterinfo_valid(osb) && + !memcmp(osb->osb_cluster_stack, OCFS2_CLASSIC_CLUSTER_STACK, + OCFS2_STACK_LABEL_LEN)) + return 1; + return 0; +} + +static inline int ocfs2_cluster_o2cb_global_heartbeat(struct ocfs2_super *osb) +{ + return ocfs2_o2cb_stack(osb) && + (osb->osb_stackflags & OCFS2_CLUSTER_O2CB_GLOBAL_HEARTBEAT); } static inline int ocfs2_mount_local(struct ocfs2_super *osb) @@ -538,6 +678,9 @@ static inline int ocfs2_uses_extended_slot_map(struct ocfs2_super *osb) #define OCFS2_IS_VALID_DX_LEAF(ptr) \ (!strcmp((ptr)->dl_signature, OCFS2_DX_LEAF_SIGNATURE)) +#define OCFS2_IS_VALID_REFCOUNT_BLOCK(ptr) \ + (!strcmp((ptr)->rf_signature, OCFS2_REFCOUNT_BLOCK_SIGNATURE)) + static inline unsigned long ino_from_blkno(struct super_block *sb, u64 blkno) { @@ -673,37 +816,73 @@ static inline unsigned int ocfs2_megabytes_to_clusters(struct super_block *sb, return megs << (20 - OCFS2_SB(sb)->s_clustersize_bits); } -static inline void ocfs2_init_inode_steal_slot(struct ocfs2_super *osb) +static inline unsigned int ocfs2_clusters_to_megabytes(struct super_block *sb, + unsigned int clusters) { - spin_lock(&osb->osb_lock); - osb->s_inode_steal_slot = OCFS2_INVALID_SLOT; - spin_unlock(&osb->osb_lock); - atomic_set(&osb->s_num_inodes_stolen, 0); + return clusters >> (20 - OCFS2_SB(sb)->s_clustersize_bits); } -static inline void ocfs2_set_inode_steal_slot(struct ocfs2_super *osb, - s16 slot) +static inline void _ocfs2_set_bit(unsigned int bit, unsigned long *bitmap) { - spin_lock(&osb->osb_lock); - osb->s_inode_steal_slot = slot; - spin_unlock(&osb->osb_lock); + __set_bit_le(bit, bitmap); } +#define ocfs2_set_bit(bit, addr) _ocfs2_set_bit((bit), (unsigned long *)(addr)) -static inline s16 ocfs2_get_inode_steal_slot(struct ocfs2_super *osb) +static inline void _ocfs2_clear_bit(unsigned int bit, unsigned long *bitmap) { - s16 slot; + __clear_bit_le(bit, bitmap); +} +#define ocfs2_clear_bit(bit, addr) _ocfs2_clear_bit((bit), (unsigned long *)(addr)) - spin_lock(&osb->osb_lock); - slot = osb->s_inode_steal_slot; - spin_unlock(&osb->osb_lock); +#define ocfs2_test_bit test_bit_le +#define ocfs2_find_next_zero_bit find_next_zero_bit_le +#define ocfs2_find_next_bit find_next_bit_le + +static inline void *correct_addr_and_bit_unaligned(int *bit, void *addr) +{ +#if BITS_PER_LONG == 64 + *bit += ((unsigned long) addr & 7UL) << 3; + addr = (void *) ((unsigned long) addr & ~7UL); +#elif BITS_PER_LONG == 32 + *bit += ((unsigned long) addr & 3UL) << 3; + addr = (void *) ((unsigned long) addr & ~3UL); +#else +#error "how many bits you are?!" +#endif + return addr; +} + +static inline void ocfs2_set_bit_unaligned(int bit, void *bitmap) +{ + bitmap = correct_addr_and_bit_unaligned(&bit, bitmap); + ocfs2_set_bit(bit, bitmap); +} + +static inline void ocfs2_clear_bit_unaligned(int bit, void *bitmap) +{ + bitmap = correct_addr_and_bit_unaligned(&bit, bitmap); + ocfs2_clear_bit(bit, bitmap); +} - return slot; +static inline int ocfs2_test_bit_unaligned(int bit, void *bitmap) +{ + bitmap = correct_addr_and_bit_unaligned(&bit, bitmap); + return ocfs2_test_bit(bit, bitmap); +} + +static inline int ocfs2_find_next_zero_bit_unaligned(void *bitmap, int max, + int start) +{ + int fix = 0, ret, tmpmax; + bitmap = correct_addr_and_bit_unaligned(&fix, bitmap); + tmpmax = max + fix; + start += fix; + + ret = ocfs2_find_next_zero_bit(bitmap, tmpmax, start) - fix; + if (ret > max) + return max; + return ret; } -#define ocfs2_set_bit ext2_set_bit -#define ocfs2_clear_bit ext2_clear_bit -#define ocfs2_test_bit ext2_test_bit -#define ocfs2_find_next_zero_bit ext2_find_next_zero_bit -#define ocfs2_find_next_bit ext2_find_next_bit #endif /* OCFS2_H */ diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h index 7ab6e9e5e77..938387a10d5 100644 --- a/fs/ocfs2/ocfs2_fs.h +++ b/fs/ocfs2/ocfs2_fs.h @@ -68,6 +68,7 @@ #define OCFS2_DIR_TRAILER_SIGNATURE "DIRTRL1" #define OCFS2_DX_ROOT_SIGNATURE "DXDIR01" #define OCFS2_DX_LEAF_SIGNATURE "DXLEAF1" +#define OCFS2_REFCOUNT_BLOCK_SIGNATURE "REFCNT1" /* Compatibility flags */ #define OCFS2_HAS_COMPAT_FEATURE(sb,mask) \ @@ -98,7 +99,10 @@ | OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK \ | OCFS2_FEATURE_INCOMPAT_XATTR \ | OCFS2_FEATURE_INCOMPAT_META_ECC \ - | OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS) + | OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS \ + | OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE \ + | OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG \ + | OCFS2_FEATURE_INCOMPAT_CLUSTERINFO) #define OCFS2_FEATURE_RO_COMPAT_SUPP (OCFS2_FEATURE_RO_COMPAT_UNWRITTEN \ | OCFS2_FEATURE_RO_COMPAT_USRQUOTA \ | OCFS2_FEATURE_RO_COMPAT_GRPQUOTA) @@ -160,6 +164,19 @@ /* Metadata checksum and error correction */ #define OCFS2_FEATURE_INCOMPAT_META_ECC 0x0800 +/* Refcount tree support */ +#define OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE 0x1000 + +/* Discontigous block groups */ +#define OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG 0x2000 + +/* + * Incompat bit to indicate useable clusterinfo with stackflags for all + * cluster stacks (userspace adnd o2cb). If this bit is set, + * INCOMPAT_USERSPACE_STACK becomes superfluous and thus should not be set. + */ +#define OCFS2_FEATURE_INCOMPAT_CLUSTERINFO 0x4000 + /* * backup superblock flag is used to indicate that this volume * has backup superblocks. @@ -223,74 +240,43 @@ #define OCFS2_HAS_XATTR_FL (0x0002) #define OCFS2_INLINE_XATTR_FL (0x0004) #define OCFS2_INDEXED_DIR_FL (0x0008) +#define OCFS2_HAS_REFCOUNT_FL (0x0010) /* Inode attributes, keep in sync with EXT2 */ -#define OCFS2_SECRM_FL (0x00000001) /* Secure deletion */ -#define OCFS2_UNRM_FL (0x00000002) /* Undelete */ -#define OCFS2_COMPR_FL (0x00000004) /* Compress file */ -#define OCFS2_SYNC_FL (0x00000008) /* Synchronous updates */ -#define OCFS2_IMMUTABLE_FL (0x00000010) /* Immutable file */ -#define OCFS2_APPEND_FL (0x00000020) /* writes to file may only append */ -#define OCFS2_NODUMP_FL (0x00000040) /* do not dump file */ -#define OCFS2_NOATIME_FL (0x00000080) /* do not update atime */ -#define OCFS2_DIRSYNC_FL (0x00010000) /* dirsync behaviour (directories only) */ - -#define OCFS2_FL_VISIBLE (0x000100FF) /* User visible flags */ -#define OCFS2_FL_MODIFIABLE (0x000100FF) /* User modifiable flags */ +#define OCFS2_SECRM_FL FS_SECRM_FL /* Secure deletion */ +#define OCFS2_UNRM_FL FS_UNRM_FL /* Undelete */ +#define OCFS2_COMPR_FL FS_COMPR_FL /* Compress file */ +#define OCFS2_SYNC_FL FS_SYNC_FL /* Synchronous updates */ +#define OCFS2_IMMUTABLE_FL FS_IMMUTABLE_FL /* Immutable file */ +#define OCFS2_APPEND_FL FS_APPEND_FL /* writes to file may only append */ +#define OCFS2_NODUMP_FL FS_NODUMP_FL /* do not dump file */ +#define OCFS2_NOATIME_FL FS_NOATIME_FL /* do not update atime */ +/* Reserved for compression usage... */ +#define OCFS2_DIRTY_FL FS_DIRTY_FL +#define OCFS2_COMPRBLK_FL FS_COMPRBLK_FL /* One or more compressed clusters */ +#define OCFS2_NOCOMP_FL FS_NOCOMP_FL /* Don't compress */ +#define OCFS2_ECOMPR_FL FS_ECOMPR_FL /* Compression error */ +/* End compression flags --- maybe not all used */ +#define OCFS2_BTREE_FL FS_BTREE_FL /* btree format dir */ +#define OCFS2_INDEX_FL FS_INDEX_FL /* hash-indexed directory */ +#define OCFS2_IMAGIC_FL FS_IMAGIC_FL /* AFS directory */ +#define OCFS2_JOURNAL_DATA_FL FS_JOURNAL_DATA_FL /* Reserved for ext3 */ +#define OCFS2_NOTAIL_FL FS_NOTAIL_FL /* file tail should not be merged */ +#define OCFS2_DIRSYNC_FL FS_DIRSYNC_FL /* dirsync behaviour (directories only) */ +#define OCFS2_TOPDIR_FL FS_TOPDIR_FL /* Top of directory hierarchies*/ +#define OCFS2_RESERVED_FL FS_RESERVED_FL /* reserved for ext2 lib */ + +#define OCFS2_FL_VISIBLE FS_FL_USER_VISIBLE /* User visible flags */ +#define OCFS2_FL_MODIFIABLE FS_FL_USER_MODIFIABLE /* User modifiable flags */ /* * Extent record flags (e_node.leaf.flags) */ -#define OCFS2_EXT_UNWRITTEN (0x01) /* Extent is allocated but - * unwritten */ - -/* - * ioctl commands - */ -#define OCFS2_IOC_GETFLAGS _IOR('f', 1, long) -#define OCFS2_IOC_SETFLAGS _IOW('f', 2, long) -#define OCFS2_IOC32_GETFLAGS _IOR('f', 1, int) -#define OCFS2_IOC32_SETFLAGS _IOW('f', 2, int) - -/* - * Space reservation / allocation / free ioctls and argument structure - * are designed to be compatible with XFS. - * - * ALLOCSP* and FREESP* are not and will never be supported, but are - * included here for completeness. - */ -struct ocfs2_space_resv { - __s16 l_type; - __s16 l_whence; - __s64 l_start; - __s64 l_len; /* len == 0 means until end of file */ - __s32 l_sysid; - __u32 l_pid; - __s32 l_pad[4]; /* reserve area */ -}; - -#define OCFS2_IOC_ALLOCSP _IOW ('X', 10, struct ocfs2_space_resv) -#define OCFS2_IOC_FREESP _IOW ('X', 11, struct ocfs2_space_resv) -#define OCFS2_IOC_RESVSP _IOW ('X', 40, struct ocfs2_space_resv) -#define OCFS2_IOC_UNRESVSP _IOW ('X', 41, struct ocfs2_space_resv) -#define OCFS2_IOC_ALLOCSP64 _IOW ('X', 36, struct ocfs2_space_resv) -#define OCFS2_IOC_FREESP64 _IOW ('X', 37, struct ocfs2_space_resv) -#define OCFS2_IOC_RESVSP64 _IOW ('X', 42, struct ocfs2_space_resv) -#define OCFS2_IOC_UNRESVSP64 _IOW ('X', 43, struct ocfs2_space_resv) - -/* Used to pass group descriptor data when online resize is done */ -struct ocfs2_new_group_input { - __u64 group; /* Group descriptor's blkno. */ - __u32 clusters; /* Total number of clusters in this group */ - __u32 frees; /* Total free clusters in this group */ - __u16 chain; /* Chain for this group */ - __u16 reserved1; - __u32 reserved2; -}; - -#define OCFS2_IOC_GROUP_EXTEND _IOW('o', 1, int) -#define OCFS2_IOC_GROUP_ADD _IOW('o', 2,struct ocfs2_new_group_input) -#define OCFS2_IOC_GROUP_ADD64 _IOW('o', 3,struct ocfs2_new_group_input) +#define OCFS2_EXT_UNWRITTEN (0x01) /* Extent is allocated but + * unwritten */ +#define OCFS2_EXT_REFCOUNTED (0x02) /* Extent is reference + * counted in an associated + * refcount tree */ /* * Journal Flags (ocfs2_dinode.id1.journal1.i_flags) @@ -314,27 +300,27 @@ struct ocfs2_new_group_input { #define OCFS2_VOL_UUID_LEN 16 #define OCFS2_MAX_VOL_LABEL_LEN 64 -/* The alternate, userspace stack fields */ +/* The cluster stack fields */ #define OCFS2_STACK_LABEL_LEN 4 #define OCFS2_CLUSTER_NAME_LEN 16 +/* Classic (historically speaking) cluster stack */ +#define OCFS2_CLASSIC_CLUSTER_STACK "o2cb" + /* Journal limits (in bytes) */ #define OCFS2_MIN_JOURNAL_SIZE (4 * 1024 * 1024) /* - * Default local alloc size (in megabytes) - * - * The value chosen should be such that most allocations, including new - * block groups, use local alloc. - */ -#define OCFS2_DEFAULT_LOCAL_ALLOC_SIZE 8 - -/* * Inline extended attribute size (in bytes) * The value chosen should be aligned to 16 byte boundaries. */ #define OCFS2_MIN_XATTR_INLINE_SIZE 256 +/* + * Cluster info flags (ocfs2_cluster_info.ci_stackflags) + */ +#define OCFS2_CLUSTER_O2CB_GLOBAL_HEARTBEAT (0x01) + struct ocfs2_system_inode_info { char *si_name; int si_iflags; @@ -352,6 +338,7 @@ enum { USER_QUOTA_SYSTEM_INODE, GROUP_QUOTA_SYSTEM_INODE, #define OCFS2_LAST_GLOBAL_SYSTEM_INODE GROUP_QUOTA_SYSTEM_INODE +#define OCFS2_FIRST_LOCAL_SYSTEM_INODE ORPHAN_DIR_SYSTEM_INODE ORPHAN_DIR_SYSTEM_INODE, EXTENT_ALLOC_SYSTEM_INODE, INODE_ALLOC_SYSTEM_INODE, @@ -360,8 +347,12 @@ enum { TRUNCATE_LOG_SYSTEM_INODE, LOCAL_USER_QUOTA_SYSTEM_INODE, LOCAL_GROUP_QUOTA_SYSTEM_INODE, +#define OCFS2_LAST_LOCAL_SYSTEM_INODE LOCAL_GROUP_QUOTA_SYSTEM_INODE NUM_SYSTEM_INODES }; +#define NUM_GLOBAL_SYSTEM_INODES OCFS2_FIRST_LOCAL_SYSTEM_INODE +#define NUM_LOCAL_SYSTEM_INODES \ + (NUM_SYSTEM_INODES - OCFS2_FIRST_LOCAL_SYSTEM_INODE) static struct ocfs2_system_inode_info ocfs2_system_inodes[NUM_SYSTEM_INODES] = { /* Global system inodes (single copy) */ @@ -390,6 +381,7 @@ static struct ocfs2_system_inode_info ocfs2_system_inodes[NUM_SYSTEM_INODES] = { /* Parameter passed from mount.ocfs2 to module */ #define OCFS2_HB_NONE "heartbeat=none" #define OCFS2_HB_LOCAL "heartbeat=local" +#define OCFS2_HB_GLOBAL "heartbeat=global" /* * OCFS2 directory file types. Only the low 3 bits are used. The @@ -449,7 +441,7 @@ static unsigned char ocfs2_type_by_mode[S_IFMT >> S_SHIFT] = { struct ocfs2_block_check { /*00*/ __le32 bc_crc32e; /* 802.3 Ethernet II CRC32 */ __le16 bc_ecc; /* Single-error-correction parity vector. - This is a simple Hamming code dependant + This is a simple Hamming code dependent on the blocksize. OCFS2's maximum blocksize, 4K, requires 16 parity bits, so we fit in __le16. */ @@ -551,7 +543,10 @@ struct ocfs2_extent_block block group */ __le32 h_fs_generation; /* Must match super block */ __le64 h_blkno; /* Offset on disk, in blocks */ -/*20*/ __le64 h_reserved3; +/*20*/ __le64 h_suballoc_loc; /* Suballocator block group this + eb belongs to. Only valid + if allocated from a + discontiguous block group */ __le64 h_next_leaf_blk; /* Offset on disk, in blocks, of next leaf header pointing to data */ @@ -593,9 +588,21 @@ struct ocfs2_slot_map_extended { */ }; +/* + * ci_stackflags is only valid if the incompat bit + * OCFS2_FEATURE_INCOMPAT_CLUSTERINFO is set. + */ struct ocfs2_cluster_info { /*00*/ __u8 ci_stack[OCFS2_STACK_LABEL_LEN]; - __le32 ci_reserved; + union { + __le32 ci_reserved; + struct { + __u8 ci_stackflags; + __u8 ci_reserved1; + __u8 ci_reserved2; + __u8 ci_reserved3; + }; + }; /*08*/ __u8 ci_cluster[OCFS2_CLUSTER_NAME_LEN]; /*18*/ }; @@ -632,9 +639,9 @@ struct ocfs2_super_block { * group header */ /*50*/ __u8 s_label[OCFS2_MAX_VOL_LABEL_LEN]; /* Label for mounting, etc. */ /*90*/ __u8 s_uuid[OCFS2_VOL_UUID_LEN]; /* 128-bit uuid */ -/*A0*/ struct ocfs2_cluster_info s_cluster_info; /* Selected userspace - stack. Only valid - with INCOMPAT flag. */ +/*A0*/ struct ocfs2_cluster_info s_cluster_info; /* Only valid if either + userspace or clusterinfo + INCOMPAT flag set. */ /*B8*/ __le16 s_xattr_inline_size; /* extended attribute inline size for this fs*/ __le16 s_reserved0; @@ -717,7 +724,12 @@ struct ocfs2_dinode { __le64 i_xattr_loc; /*80*/ struct ocfs2_block_check i_check; /* Error checking */ /*88*/ __le64 i_dx_root; /* Pointer to dir index root block */ - __le64 i_reserved2[5]; +/*90*/ __le64 i_refcount_loc; + __le64 i_suballoc_loc; /* Suballocator block group this + inode belongs to. Only valid + if allocated from a + discontiguous block group */ +/*A0*/ __le64 i_reserved2[3]; /*B8*/ union { __le64 i_pad1; /* Generic way to refer to this 64bit union */ @@ -738,7 +750,7 @@ struct ocfs2_dinode { after an unclean shutdown */ } journal1; - } id1; /* Inode type dependant 1 */ + } id1; /* Inode type dependent 1 */ /*C0*/ union { struct ocfs2_super_block i_super; struct ocfs2_local_alloc i_lab; @@ -852,7 +864,12 @@ struct ocfs2_dx_root_block { __le32 dr_reserved2; __le64 dr_free_blk; /* Pointer to head of free * unindexed block list. */ - __le64 dr_reserved3[15]; + __le64 dr_suballoc_loc; /* Suballocator block group + this root belongs to. + Only valid if allocated + from a discontiguous + block group */ + __le64 dr_reserved3[14]; union { struct ocfs2_extent_list dr_list; /* Keep this aligned to 128 * bits for maximum space @@ -878,6 +895,13 @@ struct ocfs2_dx_leaf { }; /* + * Largest bitmap for a block (suballocator) group in bytes. This limit + * does not affect cluster groups (global allocator). Cluster group + * bitmaps run to the end of the block. + */ +#define OCFS2_MAX_BG_BITMAP_SIZE 256 + +/* * On disk allocator group structure for OCFS2 */ struct ocfs2_group_desc @@ -898,7 +922,87 @@ struct ocfs2_group_desc __le64 bg_blkno; /* Offset on disk, in blocks */ /*30*/ struct ocfs2_block_check bg_check; /* Error checking */ __le64 bg_reserved2; -/*40*/ __u8 bg_bitmap[0]; +/*40*/ union { + __u8 bg_bitmap[0]; + struct { + /* + * Block groups may be discontiguous when + * OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG is set. + * The extents of a discontigous block group are + * stored in bg_list. It is a flat list. + * l_tree_depth must always be zero. A + * discontiguous group is signified by a non-zero + * bg_list->l_next_free_rec. Only block groups + * can be discontiguous; Cluster groups cannot. + * We've never made a block group with more than + * 2048 blocks (256 bytes of bg_bitmap). This + * codifies that limit so that we can fit bg_list. + * bg_size of a discontiguous block group will + * be 256 to match bg_bitmap_filler. + */ + __u8 bg_bitmap_filler[OCFS2_MAX_BG_BITMAP_SIZE]; +/*140*/ struct ocfs2_extent_list bg_list; + }; + }; +/* Actual on-disk size is one block */ +}; + +struct ocfs2_refcount_rec { +/*00*/ __le64 r_cpos; /* Physical offset, in clusters */ + __le32 r_clusters; /* Clusters covered by this extent */ + __le32 r_refcount; /* Reference count of this extent */ +/*10*/ +}; +#define OCFS2_32BIT_POS_MASK (0xffffffffULL) + +#define OCFS2_REFCOUNT_LEAF_FL (0x00000001) +#define OCFS2_REFCOUNT_TREE_FL (0x00000002) + +struct ocfs2_refcount_list { +/*00*/ __le16 rl_count; /* Maximum number of entries possible + in rl_records */ + __le16 rl_used; /* Current number of used records */ + __le32 rl_reserved2; + __le64 rl_reserved1; /* Pad to sizeof(ocfs2_refcount_record) */ +/*10*/ struct ocfs2_refcount_rec rl_recs[0]; /* Refcount records */ +}; + + +struct ocfs2_refcount_block { +/*00*/ __u8 rf_signature[8]; /* Signature for verification */ + __le16 rf_suballoc_slot; /* Slot suballocator this block + belongs to */ + __le16 rf_suballoc_bit; /* Bit offset in suballocator + block group */ + __le32 rf_fs_generation; /* Must match superblock */ +/*10*/ __le64 rf_blkno; /* Offset on disk, in blocks */ + __le64 rf_parent; /* Parent block, only valid if + OCFS2_REFCOUNT_LEAF_FL is set in + rf_flags */ +/*20*/ struct ocfs2_block_check rf_check; /* Error checking */ + __le64 rf_last_eb_blk; /* Pointer to last extent block */ +/*30*/ __le32 rf_count; /* Number of inodes sharing this + refcount tree */ + __le32 rf_flags; /* See the flags above */ + __le32 rf_clusters; /* clusters covered by refcount tree. */ + __le32 rf_cpos; /* cluster offset in refcount tree.*/ +/*40*/ __le32 rf_generation; /* generation number. all be the same + * for the same refcount tree. */ + __le32 rf_reserved0; + __le64 rf_suballoc_loc; /* Suballocator block group this + refcount block belongs to. Only + valid if allocated from a + discontiguous block group */ +/*50*/ __le64 rf_reserved1[6]; +/*80*/ union { + struct ocfs2_refcount_list rf_records; /* List of refcount + records */ + struct ocfs2_extent_list rf_list; /* Extent record list, + only valid if + OCFS2_REFCOUNT_TREE_FL + is set in rf_flags */ + }; +/* Actual on-disk size is one block */ }; /* @@ -915,7 +1019,7 @@ struct ocfs2_xattr_entry { __le16 xe_name_offset; /* byte offset from the 1st entry in the local xattr storage(inode, xattr block or xattr bucket). */ - __u8 xe_name_len; /* xattr name len, does't include prefix. */ + __u8 xe_name_len; /* xattr name len, doesn't include prefix. */ __u8 xe_type; /* the low 7 bits indicate the name prefix * type and the highest bit indicates whether * the EA is stored in the local storage. */ @@ -1001,7 +1105,10 @@ struct ocfs2_xattr_block { real xattr or a xattr tree. */ __le16 xb_reserved0; __le32 xb_reserved1; - __le64 xb_reserved2; + __le64 xb_suballoc_loc; /* Suballocator block group this + xattr block belongs to. Only + valid if allocated from a + discontiguous block group */ /*30*/ union { struct ocfs2_xattr_header xb_header; /* xattr header if this block contains xattr */ @@ -1129,7 +1236,7 @@ struct ocfs2_local_disk_dqinfo { /* Header of one chunk of a quota file */ struct ocfs2_local_disk_chunk { __le32 dqc_free; /* Number of free entries in the bitmap */ - u8 dqc_bitmap[0]; /* Bitmap of entries in the corresponding + __u8 dqc_bitmap[0]; /* Bitmap of entries in the corresponding * chunk of quota file */ }; @@ -1238,6 +1345,16 @@ static inline u16 ocfs2_extent_recs_per_eb(struct super_block *sb) return size / sizeof(struct ocfs2_extent_rec); } +static inline u16 ocfs2_extent_recs_per_gd(struct super_block *sb) +{ + int size; + + size = sb->s_blocksize - + offsetof(struct ocfs2_group_desc, bg_list.l_recs); + + return size / sizeof(struct ocfs2_extent_rec); +} + static inline int ocfs2_dx_entries_per_leaf(struct super_block *sb) { int size; @@ -1268,13 +1385,23 @@ static inline u16 ocfs2_local_alloc_size(struct super_block *sb) return size; } -static inline int ocfs2_group_bitmap_size(struct super_block *sb) +static inline int ocfs2_group_bitmap_size(struct super_block *sb, + int suballocator, + u32 feature_incompat) { - int size; - - size = sb->s_blocksize - + int size = sb->s_blocksize - offsetof(struct ocfs2_group_desc, bg_bitmap); + /* + * The cluster allocator uses the entire block. Suballocators have + * never used more than OCFS2_MAX_BG_BITMAP_SIZE. Unfortunately, older + * code expects bg_size set to the maximum. Thus we must keep + * bg_size as-is unless discontig_bg is enabled. + */ + if (suballocator && + (feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG)) + size = OCFS2_MAX_BG_BITMAP_SIZE; + return size; } @@ -1312,15 +1439,48 @@ static inline u16 ocfs2_xattr_recs_per_xb(struct super_block *sb) return size / sizeof(struct ocfs2_extent_rec); } + +static inline u16 ocfs2_extent_recs_per_rb(struct super_block *sb) +{ + int size; + + size = sb->s_blocksize - + offsetof(struct ocfs2_refcount_block, rf_list.l_recs); + + return size / sizeof(struct ocfs2_extent_rec); +} + +static inline u16 ocfs2_refcount_recs_per_rb(struct super_block *sb) +{ + int size; + + size = sb->s_blocksize - + offsetof(struct ocfs2_refcount_block, rf_records.rl_recs); + + return size / sizeof(struct ocfs2_refcount_rec); +} + +static inline u32 +ocfs2_get_ref_rec_low_cpos(const struct ocfs2_refcount_rec *rec) +{ + return le64_to_cpu(rec->r_cpos) & OCFS2_32BIT_POS_MASK; +} #else static inline int ocfs2_fast_symlink_chars(int blocksize) { return blocksize - offsetof(struct ocfs2_dinode, id2.i_symlink); } -static inline int ocfs2_max_inline_data(int blocksize) +static inline int ocfs2_max_inline_data_with_xattr(int blocksize, + struct ocfs2_dinode *di) { - return blocksize - offsetof(struct ocfs2_dinode, id2.i_data.id_data); + if (di && (di->i_dyn_features & OCFS2_INLINE_XATTR_FL)) + return blocksize - + offsetof(struct ocfs2_dinode, id2.i_data.id_data) - + di->i_xattr_inline_size; + else + return blocksize - + offsetof(struct ocfs2_dinode, id2.i_data.id_data); } static inline int ocfs2_extent_recs_per_inode(int blocksize) @@ -1353,23 +1513,43 @@ static inline int ocfs2_extent_recs_per_eb(int blocksize) return size / sizeof(struct ocfs2_extent_rec); } -static inline int ocfs2_local_alloc_size(int blocksize) +static inline int ocfs2_extent_recs_per_gd(int blocksize) { int size; size = blocksize - - offsetof(struct ocfs2_dinode, id2.i_lab.la_bitmap); + offsetof(struct ocfs2_group_desc, bg_list.l_recs); - return size; + return size / sizeof(struct ocfs2_extent_rec); } -static inline int ocfs2_group_bitmap_size(int blocksize) +static inline int ocfs2_local_alloc_size(int blocksize) { int size; size = blocksize - + offsetof(struct ocfs2_dinode, id2.i_lab.la_bitmap); + + return size; +} + +static inline int ocfs2_group_bitmap_size(int blocksize, + int suballocator, + uint32_t feature_incompat) +{ + int size = sb->s_blocksize - offsetof(struct ocfs2_group_desc, bg_bitmap); + /* + * The cluster allocator uses the entire block. Suballocators have + * never used more than OCFS2_MAX_BG_BITMAP_SIZE. Unfortunately, older + * code expects bg_size set to the maximum. Thus we must keep + * bg_size as-is unless discontig_bg is enabled. + */ + if (suballocator && + (feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG)) + size = OCFS2_MAX_BG_BITMAP_SIZE; + return size; } @@ -1442,5 +1622,19 @@ static inline void ocfs2_set_de_type(struct ocfs2_dir_entry *de, de->file_type = ocfs2_type_by_mode[(mode & S_IFMT)>>S_SHIFT]; } +static inline int ocfs2_gd_is_discontig(struct ocfs2_group_desc *gd) +{ + if ((offsetof(struct ocfs2_group_desc, bg_bitmap) + + le16_to_cpu(gd->bg_size)) != + offsetof(struct ocfs2_group_desc, bg_list)) + return 0; + /* + * Only valid to check l_next_free_rec if + * bg_bitmap + bg_size == bg_list. + */ + if (!gd->bg_list.l_next_free_rec) + return 0; + return 1; +} #endif /* _OCFS2_FS_H */ diff --git a/fs/ocfs2/ocfs2_ioctl.h b/fs/ocfs2/ocfs2_ioctl.h new file mode 100644 index 00000000000..5b27ff1fa57 --- /dev/null +++ b/fs/ocfs2/ocfs2_ioctl.h @@ -0,0 +1,242 @@ +/* -*- mode: c; c-basic-offset: 8; -*- + * vim: noexpandtab sw=8 ts=8 sts=0: + * + * ocfs2_ioctl.h + * + * Defines OCFS2 ioctls. + * + * Copyright (C) 2010 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License, version 2, as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + */ + +#ifndef OCFS2_IOCTL_H +#define OCFS2_IOCTL_H + +/* + * ioctl commands + */ +#define OCFS2_IOC_GETFLAGS FS_IOC_GETFLAGS +#define OCFS2_IOC_SETFLAGS FS_IOC_SETFLAGS +#define OCFS2_IOC32_GETFLAGS FS_IOC32_GETFLAGS +#define OCFS2_IOC32_SETFLAGS FS_IOC32_SETFLAGS + +/* + * Space reservation / allocation / free ioctls and argument structure + * are designed to be compatible with XFS. + * + * ALLOCSP* and FREESP* are not and will never be supported, but are + * included here for completeness. + */ +struct ocfs2_space_resv { + __s16 l_type; + __s16 l_whence; + __s64 l_start; + __s64 l_len; /* len == 0 means until end of file */ + __s32 l_sysid; + __u32 l_pid; + __s32 l_pad[4]; /* reserve area */ +}; + +#define OCFS2_IOC_ALLOCSP _IOW ('X', 10, struct ocfs2_space_resv) +#define OCFS2_IOC_FREESP _IOW ('X', 11, struct ocfs2_space_resv) +#define OCFS2_IOC_RESVSP _IOW ('X', 40, struct ocfs2_space_resv) +#define OCFS2_IOC_UNRESVSP _IOW ('X', 41, struct ocfs2_space_resv) +#define OCFS2_IOC_ALLOCSP64 _IOW ('X', 36, struct ocfs2_space_resv) +#define OCFS2_IOC_FREESP64 _IOW ('X', 37, struct ocfs2_space_resv) +#define OCFS2_IOC_RESVSP64 _IOW ('X', 42, struct ocfs2_space_resv) +#define OCFS2_IOC_UNRESVSP64 _IOW ('X', 43, struct ocfs2_space_resv) + +/* Used to pass group descriptor data when online resize is done */ +struct ocfs2_new_group_input { + __u64 group; /* Group descriptor's blkno. */ + __u32 clusters; /* Total number of clusters in this group */ + __u32 frees; /* Total free clusters in this group */ + __u16 chain; /* Chain for this group */ + __u16 reserved1; + __u32 reserved2; +}; + +#define OCFS2_IOC_GROUP_EXTEND _IOW('o', 1, int) +#define OCFS2_IOC_GROUP_ADD _IOW('o', 2,struct ocfs2_new_group_input) +#define OCFS2_IOC_GROUP_ADD64 _IOW('o', 3,struct ocfs2_new_group_input) + +/* Used to pass 2 file names to reflink. */ +struct reflink_arguments { + __u64 old_path; + __u64 new_path; + __u64 preserve; +}; +#define OCFS2_IOC_REFLINK _IOW('o', 4, struct reflink_arguments) + +/* Following definitions dedicated for ocfs2_info_request ioctls. */ +#define OCFS2_INFO_MAX_REQUEST (50) +#define OCFS2_TEXT_UUID_LEN (OCFS2_VOL_UUID_LEN * 2) + +/* Magic number of all requests */ +#define OCFS2_INFO_MAGIC (0x4F32494E) + +/* + * Always try to separate info request into small pieces to + * guarantee the backward&forward compatibility. + */ +struct ocfs2_info { + __u64 oi_requests; /* Array of __u64 pointers to requests */ + __u32 oi_count; /* Number of requests in info_requests */ + __u32 oi_pad; +}; + +struct ocfs2_info_request { +/*00*/ __u32 ir_magic; /* Magic number */ + __u32 ir_code; /* Info request code */ + __u32 ir_size; /* Size of request */ + __u32 ir_flags; /* Request flags */ +/*10*/ /* Request specific fields */ +}; + +struct ocfs2_info_clustersize { + struct ocfs2_info_request ic_req; + __u32 ic_clustersize; + __u32 ic_pad; +}; + +struct ocfs2_info_blocksize { + struct ocfs2_info_request ib_req; + __u32 ib_blocksize; + __u32 ib_pad; +}; + +struct ocfs2_info_maxslots { + struct ocfs2_info_request im_req; + __u32 im_max_slots; + __u32 im_pad; +}; + +struct ocfs2_info_label { + struct ocfs2_info_request il_req; + __u8 il_label[OCFS2_MAX_VOL_LABEL_LEN]; +} __attribute__ ((packed)); + +struct ocfs2_info_uuid { + struct ocfs2_info_request iu_req; + __u8 iu_uuid_str[OCFS2_TEXT_UUID_LEN + 1]; +} __attribute__ ((packed)); + +struct ocfs2_info_fs_features { + struct ocfs2_info_request if_req; + __u32 if_compat_features; + __u32 if_incompat_features; + __u32 if_ro_compat_features; + __u32 if_pad; +}; + +struct ocfs2_info_journal_size { + struct ocfs2_info_request ij_req; + __u64 ij_journal_size; +}; + +struct ocfs2_info_freeinode { + struct ocfs2_info_request ifi_req; + struct ocfs2_info_local_freeinode { + __u64 lfi_total; + __u64 lfi_free; + } ifi_stat[OCFS2_MAX_SLOTS]; + __u32 ifi_slotnum; /* out */ + __u32 ifi_pad; +}; + +#define OCFS2_INFO_MAX_HIST (32) + +struct ocfs2_info_freefrag { + struct ocfs2_info_request iff_req; + struct ocfs2_info_freefrag_stats { /* (out) */ + struct ocfs2_info_free_chunk_list { + __u32 fc_chunks[OCFS2_INFO_MAX_HIST]; + __u32 fc_clusters[OCFS2_INFO_MAX_HIST]; + } ffs_fc_hist; + __u32 ffs_clusters; + __u32 ffs_free_clusters; + __u32 ffs_free_chunks; + __u32 ffs_free_chunks_real; + __u32 ffs_min; /* Minimum free chunksize in clusters */ + __u32 ffs_max; + __u32 ffs_avg; + __u32 ffs_pad; + } iff_ffs; + __u32 iff_chunksize; /* chunksize in clusters(in) */ + __u32 iff_pad; +}; + +/* Codes for ocfs2_info_request */ +enum ocfs2_info_type { + OCFS2_INFO_CLUSTERSIZE = 1, + OCFS2_INFO_BLOCKSIZE, + OCFS2_INFO_MAXSLOTS, + OCFS2_INFO_LABEL, + OCFS2_INFO_UUID, + OCFS2_INFO_FS_FEATURES, + OCFS2_INFO_JOURNAL_SIZE, + OCFS2_INFO_FREEINODE, + OCFS2_INFO_FREEFRAG, + OCFS2_INFO_NUM_TYPES +}; + +/* Flags for struct ocfs2_info_request */ +/* Filled by the caller */ +#define OCFS2_INFO_FL_NON_COHERENT (0x00000001) /* Cluster coherency not + required. This is a hint. + It is up to ocfs2 whether + the request can be fulfilled + without locking. */ +/* Filled by ocfs2 */ +#define OCFS2_INFO_FL_FILLED (0x40000000) /* Filesystem understood + this request and + filled in the answer */ + +#define OCFS2_INFO_FL_ERROR (0x80000000) /* Error happened during + request handling. */ + +#define OCFS2_IOC_INFO _IOR('o', 5, struct ocfs2_info) + +struct ocfs2_move_extents { +/* All values are in bytes */ + /* in */ + __u64 me_start; /* Virtual start in the file to move */ + __u64 me_len; /* Length of the extents to be moved */ + __u64 me_goal; /* Physical offset of the goal, + it's in block unit */ + __u64 me_threshold; /* Maximum distance from goal or threshold + for auto defragmentation */ + __u64 me_flags; /* Flags for the operation: + * - auto defragmentation. + * - refcount,xattr cases. + */ + /* out */ + __u64 me_moved_len; /* Moved/defraged length */ + __u64 me_new_offset; /* Resulting physical location */ + __u32 me_reserved[2]; /* Reserved for futhure */ +}; + +#define OCFS2_MOVE_EXT_FL_AUTO_DEFRAG (0x00000001) /* Kernel manages to + claim new clusters + as the goal place + for extents moving */ +#define OCFS2_MOVE_EXT_FL_PART_DEFRAG (0x00000002) /* Allow partial extent + moving, is to make + movement less likely + to fail, may make fs + even more fragmented */ +#define OCFS2_MOVE_EXT_FL_COMPLETE (0x00000004) /* Move or defragmenation + completely gets done. + */ + +#define OCFS2_IOC_MOVE_EXT _IOW('o', 6, struct ocfs2_move_extents) + +#endif /* OCFS2_IOCTL_H */ diff --git a/fs/ocfs2/ocfs2_lockid.h b/fs/ocfs2/ocfs2_lockid.h index a53ce87481b..d277aabf5df 100644 --- a/fs/ocfs2/ocfs2_lockid.h +++ b/fs/ocfs2/ocfs2_lockid.h @@ -48,6 +48,8 @@ enum ocfs2_lock_type { OCFS2_LOCK_TYPE_FLOCK, OCFS2_LOCK_TYPE_QINFO, OCFS2_LOCK_TYPE_NFS_SYNC, + OCFS2_LOCK_TYPE_ORPHAN_SCAN, + OCFS2_LOCK_TYPE_REFCOUNT, OCFS2_NUM_LOCK_TYPES }; @@ -85,6 +87,12 @@ static inline char ocfs2_lock_type_char(enum ocfs2_lock_type type) case OCFS2_LOCK_TYPE_NFS_SYNC: c = 'Y'; break; + case OCFS2_LOCK_TYPE_ORPHAN_SCAN: + c = 'P'; + break; + case OCFS2_LOCK_TYPE_REFCOUNT: + c = 'T'; + break; default: c = '\0'; } @@ -104,6 +112,9 @@ static char *ocfs2_lock_type_strings[] = { [OCFS2_LOCK_TYPE_OPEN] = "Open", [OCFS2_LOCK_TYPE_FLOCK] = "Flock", [OCFS2_LOCK_TYPE_QINFO] = "Quota", + [OCFS2_LOCK_TYPE_NFS_SYNC] = "NFSSync", + [OCFS2_LOCK_TYPE_ORPHAN_SCAN] = "OrphanScan", + [OCFS2_LOCK_TYPE_REFCOUNT] = "Refcount", }; static inline const char *ocfs2_lock_type_string(enum ocfs2_lock_type type) diff --git a/fs/ocfs2/ocfs2_lockingver.h b/fs/ocfs2/ocfs2_lockingver.h index 82d5eeac0ff..2e45c8d2ea7 100644 --- a/fs/ocfs2/ocfs2_lockingver.h +++ b/fs/ocfs2/ocfs2_lockingver.h @@ -23,6 +23,8 @@ /* * The protocol version for ocfs2 cluster locking. See dlmglue.c for * more details. + * + * 1.0 - Initial locking version from ocfs2 1.4. */ #define OCFS2_LOCKING_PROTOCOL_MAJOR 1 #define OCFS2_LOCKING_PROTOCOL_MINOR 0 diff --git a/fs/ocfs2/ocfs2_trace.h b/fs/ocfs2/ocfs2_trace.h new file mode 100644 index 00000000000..6cb019b7c6a --- /dev/null +++ b/fs/ocfs2/ocfs2_trace.h @@ -0,0 +1,2768 @@ +#undef TRACE_SYSTEM +#define TRACE_SYSTEM ocfs2 + +#if !defined(_TRACE_OCFS2_H) || defined(TRACE_HEADER_MULTI_READ) +#define _TRACE_OCFS2_H + +#include <linux/tracepoint.h> + +DECLARE_EVENT_CLASS(ocfs2__int, + TP_PROTO(int num), + TP_ARGS(num), + TP_STRUCT__entry( + __field(int, num) + ), + TP_fast_assign( + __entry->num = num; + ), + TP_printk("%d", __entry->num) +); + +#define DEFINE_OCFS2_INT_EVENT(name) \ +DEFINE_EVENT(ocfs2__int, name, \ + TP_PROTO(int num), \ + TP_ARGS(num)) + +DECLARE_EVENT_CLASS(ocfs2__uint, + TP_PROTO(unsigned int num), + TP_ARGS(num), + TP_STRUCT__entry( + __field( unsigned int, num ) + ), + TP_fast_assign( + __entry->num = num; + ), + TP_printk("%u", __entry->num) +); + +#define DEFINE_OCFS2_UINT_EVENT(name) \ +DEFINE_EVENT(ocfs2__uint, name, \ + TP_PROTO(unsigned int num), \ + TP_ARGS(num)) + +DECLARE_EVENT_CLASS(ocfs2__ull, + TP_PROTO(unsigned long long blkno), + TP_ARGS(blkno), + TP_STRUCT__entry( + __field(unsigned long long, blkno) + ), + TP_fast_assign( + __entry->blkno = blkno; + ), + TP_printk("%llu", __entry->blkno) +); + +#define DEFINE_OCFS2_ULL_EVENT(name) \ +DEFINE_EVENT(ocfs2__ull, name, \ + TP_PROTO(unsigned long long num), \ + TP_ARGS(num)) + +DECLARE_EVENT_CLASS(ocfs2__pointer, + TP_PROTO(void *pointer), + TP_ARGS(pointer), + TP_STRUCT__entry( + __field(void *, pointer) + ), + TP_fast_assign( + __entry->pointer = pointer; + ), + TP_printk("%p", __entry->pointer) +); + +#define DEFINE_OCFS2_POINTER_EVENT(name) \ +DEFINE_EVENT(ocfs2__pointer, name, \ + TP_PROTO(void *pointer), \ + TP_ARGS(pointer)) + +DECLARE_EVENT_CLASS(ocfs2__string, + TP_PROTO(const char *name), + TP_ARGS(name), + TP_STRUCT__entry( + __string(name,name) + ), + TP_fast_assign( + __assign_str(name, name); + ), + TP_printk("%s", __get_str(name)) +); + +#define DEFINE_OCFS2_STRING_EVENT(name) \ +DEFINE_EVENT(ocfs2__string, name, \ + TP_PROTO(const char *name), \ + TP_ARGS(name)) + +DECLARE_EVENT_CLASS(ocfs2__int_int, + TP_PROTO(int value1, int value2), + TP_ARGS(value1, value2), + TP_STRUCT__entry( + __field(int, value1) + __field(int, value2) + ), + TP_fast_assign( + __entry->value1 = value1; + __entry->value2 = value2; + ), + TP_printk("%d %d", __entry->value1, __entry->value2) +); + +#define DEFINE_OCFS2_INT_INT_EVENT(name) \ +DEFINE_EVENT(ocfs2__int_int, name, \ + TP_PROTO(int val1, int val2), \ + TP_ARGS(val1, val2)) + +DECLARE_EVENT_CLASS(ocfs2__uint_int, + TP_PROTO(unsigned int value1, int value2), + TP_ARGS(value1, value2), + TP_STRUCT__entry( + __field(unsigned int, value1) + __field(int, value2) + ), + TP_fast_assign( + __entry->value1 = value1; + __entry->value2 = value2; + ), + TP_printk("%u %d", __entry->value1, __entry->value2) +); + +#define DEFINE_OCFS2_UINT_INT_EVENT(name) \ +DEFINE_EVENT(ocfs2__uint_int, name, \ + TP_PROTO(unsigned int val1, int val2), \ + TP_ARGS(val1, val2)) + +DECLARE_EVENT_CLASS(ocfs2__uint_uint, + TP_PROTO(unsigned int value1, unsigned int value2), + TP_ARGS(value1, value2), + TP_STRUCT__entry( + __field(unsigned int, value1) + __field(unsigned int, value2) + ), + TP_fast_assign( + __entry->value1 = value1; + __entry->value2 = value2; + ), + TP_printk("%u %u", __entry->value1, __entry->value2) +); + +#define DEFINE_OCFS2_UINT_UINT_EVENT(name) \ +DEFINE_EVENT(ocfs2__uint_uint, name, \ + TP_PROTO(unsigned int val1, unsigned int val2), \ + TP_ARGS(val1, val2)) + +DECLARE_EVENT_CLASS(ocfs2__ull_uint, + TP_PROTO(unsigned long long value1, unsigned int value2), + TP_ARGS(value1, value2), + TP_STRUCT__entry( + __field(unsigned long long, value1) + __field(unsigned int, value2) + ), + TP_fast_assign( + __entry->value1 = value1; + __entry->value2 = value2; + ), + TP_printk("%llu %u", __entry->value1, __entry->value2) +); + +#define DEFINE_OCFS2_ULL_UINT_EVENT(name) \ +DEFINE_EVENT(ocfs2__ull_uint, name, \ + TP_PROTO(unsigned long long val1, unsigned int val2), \ + TP_ARGS(val1, val2)) + +DECLARE_EVENT_CLASS(ocfs2__ull_int, + TP_PROTO(unsigned long long value1, int value2), + TP_ARGS(value1, value2), + TP_STRUCT__entry( + __field(unsigned long long, value1) + __field(int, value2) + ), + TP_fast_assign( + __entry->value1 = value1; + __entry->value2 = value2; + ), + TP_printk("%llu %d", __entry->value1, __entry->value2) +); + +#define DEFINE_OCFS2_ULL_INT_EVENT(name) \ +DEFINE_EVENT(ocfs2__ull_int, name, \ + TP_PROTO(unsigned long long val1, int val2), \ + TP_ARGS(val1, val2)) + +DECLARE_EVENT_CLASS(ocfs2__ull_ull, + TP_PROTO(unsigned long long value1, unsigned long long value2), + TP_ARGS(value1, value2), + TP_STRUCT__entry( + __field(unsigned long long, value1) + __field(unsigned long long, value2) + ), + TP_fast_assign( + __entry->value1 = value1; + __entry->value2 = value2; + ), + TP_printk("%llu %llu", __entry->value1, __entry->value2) +); + +#define DEFINE_OCFS2_ULL_ULL_EVENT(name) \ +DEFINE_EVENT(ocfs2__ull_ull, name, \ + TP_PROTO(unsigned long long val1, unsigned long long val2), \ + TP_ARGS(val1, val2)) + +DECLARE_EVENT_CLASS(ocfs2__ull_ull_uint, + TP_PROTO(unsigned long long value1, + unsigned long long value2, unsigned int value3), + TP_ARGS(value1, value2, value3), + TP_STRUCT__entry( + __field(unsigned long long, value1) + __field(unsigned long long, value2) + __field(unsigned int, value3) + ), + TP_fast_assign( + __entry->value1 = value1; + __entry->value2 = value2; + __entry->value3 = value3; + ), + TP_printk("%llu %llu %u", + __entry->value1, __entry->value2, __entry->value3) +); + +#define DEFINE_OCFS2_ULL_ULL_UINT_EVENT(name) \ +DEFINE_EVENT(ocfs2__ull_ull_uint, name, \ + TP_PROTO(unsigned long long val1, \ + unsigned long long val2, unsigned int val3), \ + TP_ARGS(val1, val2, val3)) + +DECLARE_EVENT_CLASS(ocfs2__ull_uint_uint, + TP_PROTO(unsigned long long value1, + unsigned int value2, unsigned int value3), + TP_ARGS(value1, value2, value3), + TP_STRUCT__entry( + __field(unsigned long long, value1) + __field(unsigned int, value2) + __field(unsigned int, value3) + ), + TP_fast_assign( + __entry->value1 = value1; + __entry->value2 = value2; + __entry->value3 = value3; + ), + TP_printk("%llu %u %u", __entry->value1, + __entry->value2, __entry->value3) +); + +#define DEFINE_OCFS2_ULL_UINT_UINT_EVENT(name) \ +DEFINE_EVENT(ocfs2__ull_uint_uint, name, \ + TP_PROTO(unsigned long long val1, \ + unsigned int val2, unsigned int val3), \ + TP_ARGS(val1, val2, val3)) + +DECLARE_EVENT_CLASS(ocfs2__uint_uint_uint, + TP_PROTO(unsigned int value1, unsigned int value2, + unsigned int value3), + TP_ARGS(value1, value2, value3), + TP_STRUCT__entry( + __field( unsigned int, value1 ) + __field( unsigned int, value2 ) + __field( unsigned int, value3 ) + ), + TP_fast_assign( + __entry->value1 = value1; + __entry->value2 = value2; + __entry->value3 = value3; + ), + TP_printk("%u %u %u", __entry->value1, __entry->value2, __entry->value3) +); + +#define DEFINE_OCFS2_UINT_UINT_UINT_EVENT(name) \ +DEFINE_EVENT(ocfs2__uint_uint_uint, name, \ + TP_PROTO(unsigned int value1, unsigned int value2, \ + unsigned int value3), \ + TP_ARGS(value1, value2, value3)) + +DECLARE_EVENT_CLASS(ocfs2__ull_ull_ull, + TP_PROTO(unsigned long long value1, + unsigned long long value2, unsigned long long value3), + TP_ARGS(value1, value2, value3), + TP_STRUCT__entry( + __field(unsigned long long, value1) + __field(unsigned long long, value2) + __field(unsigned long long, value3) + ), + TP_fast_assign( + __entry->value1 = value1; + __entry->value2 = value2; + __entry->value3 = value3; + ), + TP_printk("%llu %llu %llu", + __entry->value1, __entry->value2, __entry->value3) +); + +#define DEFINE_OCFS2_ULL_ULL_ULL_EVENT(name) \ +DEFINE_EVENT(ocfs2__ull_ull_ull, name, \ + TP_PROTO(unsigned long long value1, unsigned long long value2, \ + unsigned long long value3), \ + TP_ARGS(value1, value2, value3)) + +DECLARE_EVENT_CLASS(ocfs2__ull_int_int_int, + TP_PROTO(unsigned long long ull, int value1, int value2, int value3), + TP_ARGS(ull, value1, value2, value3), + TP_STRUCT__entry( + __field( unsigned long long, ull ) + __field( int, value1 ) + __field( int, value2 ) + __field( int, value3 ) + ), + TP_fast_assign( + __entry->ull = ull; + __entry->value1 = value1; + __entry->value2 = value2; + __entry->value3 = value3; + ), + TP_printk("%llu %d %d %d", + __entry->ull, __entry->value1, + __entry->value2, __entry->value3) +); + +#define DEFINE_OCFS2_ULL_INT_INT_INT_EVENT(name) \ +DEFINE_EVENT(ocfs2__ull_int_int_int, name, \ + TP_PROTO(unsigned long long ull, int value1, \ + int value2, int value3), \ + TP_ARGS(ull, value1, value2, value3)) + +DECLARE_EVENT_CLASS(ocfs2__ull_uint_uint_uint, + TP_PROTO(unsigned long long ull, unsigned int value1, + unsigned int value2, unsigned int value3), + TP_ARGS(ull, value1, value2, value3), + TP_STRUCT__entry( + __field(unsigned long long, ull) + __field(unsigned int, value1) + __field(unsigned int, value2) + __field(unsigned int, value3) + ), + TP_fast_assign( + __entry->ull = ull; + __entry->value1 = value1; + __entry->value2 = value2; + __entry->value3 = value3; + ), + TP_printk("%llu %u %u %u", + __entry->ull, __entry->value1, + __entry->value2, __entry->value3) +); + +#define DEFINE_OCFS2_ULL_UINT_UINT_UINT_EVENT(name) \ +DEFINE_EVENT(ocfs2__ull_uint_uint_uint, name, \ + TP_PROTO(unsigned long long ull, unsigned int value1, \ + unsigned int value2, unsigned int value3), \ + TP_ARGS(ull, value1, value2, value3)) + +DECLARE_EVENT_CLASS(ocfs2__ull_ull_uint_uint, + TP_PROTO(unsigned long long value1, unsigned long long value2, + unsigned int value3, unsigned int value4), + TP_ARGS(value1, value2, value3, value4), + TP_STRUCT__entry( + __field(unsigned long long, value1) + __field(unsigned long long, value2) + __field(unsigned int, value3) + __field(unsigned int, value4) + ), + TP_fast_assign( + __entry->value1 = value1; + __entry->value2 = value2; + __entry->value3 = value3; + __entry->value4 = value4; + ), + TP_printk("%llu %llu %u %u", + __entry->value1, __entry->value2, + __entry->value3, __entry->value4) +); + +#define DEFINE_OCFS2_ULL_ULL_UINT_UINT_EVENT(name) \ +DEFINE_EVENT(ocfs2__ull_ull_uint_uint, name, \ + TP_PROTO(unsigned long long ull, unsigned long long ull1, \ + unsigned int value2, unsigned int value3), \ + TP_ARGS(ull, ull1, value2, value3)) + +/* Trace events for fs/ocfs2/alloc.c. */ +DECLARE_EVENT_CLASS(ocfs2__btree_ops, + TP_PROTO(unsigned long long owner,\ + unsigned int value1, unsigned int value2), + TP_ARGS(owner, value1, value2), + TP_STRUCT__entry( + __field(unsigned long long, owner) + __field(unsigned int, value1) + __field(unsigned int, value2) + ), + TP_fast_assign( + __entry->owner = owner; + __entry->value1 = value1; + __entry->value2 = value2; + ), + TP_printk("%llu %u %u", + __entry->owner, __entry->value1, __entry->value2) +); + +#define DEFINE_OCFS2_BTREE_EVENT(name) \ +DEFINE_EVENT(ocfs2__btree_ops, name, \ + TP_PROTO(unsigned long long owner, \ + unsigned int value1, unsigned int value2), \ + TP_ARGS(owner, value1, value2)) + +DEFINE_OCFS2_BTREE_EVENT(ocfs2_adjust_rightmost_branch); + +DEFINE_OCFS2_BTREE_EVENT(ocfs2_rotate_tree_right); + +DEFINE_OCFS2_BTREE_EVENT(ocfs2_append_rec_to_path); + +DEFINE_OCFS2_BTREE_EVENT(ocfs2_insert_extent_start); + +DEFINE_OCFS2_BTREE_EVENT(ocfs2_add_clusters_in_btree); + +DEFINE_OCFS2_INT_EVENT(ocfs2_num_free_extents); + +DEFINE_OCFS2_INT_EVENT(ocfs2_complete_edge_insert); + +TRACE_EVENT(ocfs2_grow_tree, + TP_PROTO(unsigned long long owner, int depth), + TP_ARGS(owner, depth), + TP_STRUCT__entry( + __field(unsigned long long, owner) + __field(int, depth) + ), + TP_fast_assign( + __entry->owner = owner; + __entry->depth = depth; + ), + TP_printk("%llu %d", __entry->owner, __entry->depth) +); + +TRACE_EVENT(ocfs2_rotate_subtree, + TP_PROTO(int subtree_root, unsigned long long blkno, + int depth), + TP_ARGS(subtree_root, blkno, depth), + TP_STRUCT__entry( + __field(int, subtree_root) + __field(unsigned long long, blkno) + __field(int, depth) + ), + TP_fast_assign( + __entry->subtree_root = subtree_root; + __entry->blkno = blkno; + __entry->depth = depth; + ), + TP_printk("%d %llu %d", __entry->subtree_root, + __entry->blkno, __entry->depth) +); + +TRACE_EVENT(ocfs2_insert_extent, + TP_PROTO(unsigned int ins_appending, unsigned int ins_contig, + int ins_contig_index, int free_records, int ins_tree_depth), + TP_ARGS(ins_appending, ins_contig, ins_contig_index, free_records, + ins_tree_depth), + TP_STRUCT__entry( + __field(unsigned int, ins_appending) + __field(unsigned int, ins_contig) + __field(int, ins_contig_index) + __field(int, free_records) + __field(int, ins_tree_depth) + ), + TP_fast_assign( + __entry->ins_appending = ins_appending; + __entry->ins_contig = ins_contig; + __entry->ins_contig_index = ins_contig_index; + __entry->free_records = free_records; + __entry->ins_tree_depth = ins_tree_depth; + ), + TP_printk("%u %u %d %d %d", + __entry->ins_appending, __entry->ins_contig, + __entry->ins_contig_index, __entry->free_records, + __entry->ins_tree_depth) +); + +TRACE_EVENT(ocfs2_split_extent, + TP_PROTO(int split_index, unsigned int c_contig_type, + unsigned int c_has_empty_extent, + unsigned int c_split_covers_rec), + TP_ARGS(split_index, c_contig_type, + c_has_empty_extent, c_split_covers_rec), + TP_STRUCT__entry( + __field(int, split_index) + __field(unsigned int, c_contig_type) + __field(unsigned int, c_has_empty_extent) + __field(unsigned int, c_split_covers_rec) + ), + TP_fast_assign( + __entry->split_index = split_index; + __entry->c_contig_type = c_contig_type; + __entry->c_has_empty_extent = c_has_empty_extent; + __entry->c_split_covers_rec = c_split_covers_rec; + ), + TP_printk("%d %u %u %u", __entry->split_index, __entry->c_contig_type, + __entry->c_has_empty_extent, __entry->c_split_covers_rec) +); + +TRACE_EVENT(ocfs2_remove_extent, + TP_PROTO(unsigned long long owner, unsigned int cpos, + unsigned int len, int index, + unsigned int e_cpos, unsigned int clusters), + TP_ARGS(owner, cpos, len, index, e_cpos, clusters), + TP_STRUCT__entry( + __field(unsigned long long, owner) + __field(unsigned int, cpos) + __field(unsigned int, len) + __field(int, index) + __field(unsigned int, e_cpos) + __field(unsigned int, clusters) + ), + TP_fast_assign( + __entry->owner = owner; + __entry->cpos = cpos; + __entry->len = len; + __entry->index = index; + __entry->e_cpos = e_cpos; + __entry->clusters = clusters; + ), + TP_printk("%llu %u %u %d %u %u", + __entry->owner, __entry->cpos, __entry->len, __entry->index, + __entry->e_cpos, __entry->clusters) +); + +TRACE_EVENT(ocfs2_commit_truncate, + TP_PROTO(unsigned long long ino, unsigned int new_cpos, + unsigned int clusters, unsigned int depth), + TP_ARGS(ino, new_cpos, clusters, depth), + TP_STRUCT__entry( + __field(unsigned long long, ino) + __field(unsigned int, new_cpos) + __field(unsigned int, clusters) + __field(unsigned int, depth) + ), + TP_fast_assign( + __entry->ino = ino; + __entry->new_cpos = new_cpos; + __entry->clusters = clusters; + __entry->depth = depth; + ), + TP_printk("%llu %u %u %u", + __entry->ino, __entry->new_cpos, + __entry->clusters, __entry->depth) +); + +TRACE_EVENT(ocfs2_validate_extent_block, + TP_PROTO(unsigned long long blkno), + TP_ARGS(blkno), + TP_STRUCT__entry( + __field(unsigned long long, blkno) + ), + TP_fast_assign( + __entry->blkno = blkno; + ), + TP_printk("%llu ", __entry->blkno) +); + +TRACE_EVENT(ocfs2_rotate_leaf, + TP_PROTO(unsigned int insert_cpos, int insert_index, + int has_empty, int next_free, + unsigned int l_count), + TP_ARGS(insert_cpos, insert_index, has_empty, + next_free, l_count), + TP_STRUCT__entry( + __field(unsigned int, insert_cpos) + __field(int, insert_index) + __field(int, has_empty) + __field(int, next_free) + __field(unsigned int, l_count) + ), + TP_fast_assign( + __entry->insert_cpos = insert_cpos; + __entry->insert_index = insert_index; + __entry->has_empty = has_empty; + __entry->next_free = next_free; + __entry->l_count = l_count; + ), + TP_printk("%u %d %d %d %u", __entry->insert_cpos, + __entry->insert_index, __entry->has_empty, + __entry->next_free, __entry->l_count) +); + +TRACE_EVENT(ocfs2_add_clusters_in_btree_ret, + TP_PROTO(int status, int reason, int err), + TP_ARGS(status, reason, err), + TP_STRUCT__entry( + __field(int, status) + __field(int, reason) + __field(int, err) + ), + TP_fast_assign( + __entry->status = status; + __entry->reason = reason; + __entry->err = err; + ), + TP_printk("%d %d %d", __entry->status, + __entry->reason, __entry->err) +); + +TRACE_EVENT(ocfs2_mark_extent_written, + TP_PROTO(unsigned long long owner, unsigned int cpos, + unsigned int len, unsigned int phys), + TP_ARGS(owner, cpos, len, phys), + TP_STRUCT__entry( + __field(unsigned long long, owner) + __field(unsigned int, cpos) + __field(unsigned int, len) + __field(unsigned int, phys) + ), + TP_fast_assign( + __entry->owner = owner; + __entry->cpos = cpos; + __entry->len = len; + __entry->phys = phys; + ), + TP_printk("%llu %u %u %u", + __entry->owner, __entry->cpos, + __entry->len, __entry->phys) +); + +DECLARE_EVENT_CLASS(ocfs2__truncate_log_ops, + TP_PROTO(unsigned long long blkno, int index, + unsigned int start, unsigned int num), + TP_ARGS(blkno, index, start, num), + TP_STRUCT__entry( + __field(unsigned long long, blkno) + __field(int, index) + __field(unsigned int, start) + __field(unsigned int, num) + ), + TP_fast_assign( + __entry->blkno = blkno; + __entry->index = index; + __entry->start = start; + __entry->num = num; + ), + TP_printk("%llu %d %u %u", + __entry->blkno, __entry->index, + __entry->start, __entry->num) +); + +#define DEFINE_OCFS2_TRUNCATE_LOG_OPS_EVENT(name) \ +DEFINE_EVENT(ocfs2__truncate_log_ops, name, \ + TP_PROTO(unsigned long long blkno, int index, \ + unsigned int start, unsigned int num), \ + TP_ARGS(blkno, index, start, num)) + +DEFINE_OCFS2_TRUNCATE_LOG_OPS_EVENT(ocfs2_truncate_log_append); + +DEFINE_OCFS2_TRUNCATE_LOG_OPS_EVENT(ocfs2_replay_truncate_records); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_flush_truncate_log); + +DEFINE_OCFS2_INT_EVENT(ocfs2_begin_truncate_log_recovery); + +DEFINE_OCFS2_INT_EVENT(ocfs2_truncate_log_recovery_num); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_complete_truncate_log_recovery); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_free_cached_blocks); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_cache_cluster_dealloc); + +DEFINE_OCFS2_INT_INT_EVENT(ocfs2_run_deallocs); + +TRACE_EVENT(ocfs2_cache_block_dealloc, + TP_PROTO(int type, int slot, unsigned long long suballoc, + unsigned long long blkno, unsigned int bit), + TP_ARGS(type, slot, suballoc, blkno, bit), + TP_STRUCT__entry( + __field(int, type) + __field(int, slot) + __field(unsigned long long, suballoc) + __field(unsigned long long, blkno) + __field(unsigned int, bit) + ), + TP_fast_assign( + __entry->type = type; + __entry->slot = slot; + __entry->suballoc = suballoc; + __entry->blkno = blkno; + __entry->bit = bit; + ), + TP_printk("%d %d %llu %llu %u", + __entry->type, __entry->slot, __entry->suballoc, + __entry->blkno, __entry->bit) +); + +TRACE_EVENT(ocfs2_trim_extent, + TP_PROTO(struct super_block *sb, unsigned long long blk, + unsigned long long count), + TP_ARGS(sb, blk, count), + TP_STRUCT__entry( + __field(int, dev_major) + __field(int, dev_minor) + __field(unsigned long long, blk) + __field(__u64, count) + ), + TP_fast_assign( + __entry->dev_major = MAJOR(sb->s_dev); + __entry->dev_minor = MINOR(sb->s_dev); + __entry->blk = blk; + __entry->count = count; + ), + TP_printk("%d %d %llu %llu", + __entry->dev_major, __entry->dev_minor, + __entry->blk, __entry->count) +); + +DEFINE_OCFS2_ULL_UINT_UINT_UINT_EVENT(ocfs2_trim_group); + +DEFINE_OCFS2_ULL_ULL_ULL_EVENT(ocfs2_trim_fs); + +/* End of trace events for fs/ocfs2/alloc.c. */ + +/* Trace events for fs/ocfs2/localalloc.c. */ + +DEFINE_OCFS2_UINT_UINT_UINT_EVENT(ocfs2_la_set_sizes); + +DEFINE_OCFS2_ULL_INT_INT_INT_EVENT(ocfs2_alloc_should_use_local); + +DEFINE_OCFS2_INT_EVENT(ocfs2_load_local_alloc); + +DEFINE_OCFS2_INT_EVENT(ocfs2_begin_local_alloc_recovery); + +DEFINE_OCFS2_ULL_INT_INT_INT_EVENT(ocfs2_reserve_local_alloc_bits); + +DEFINE_OCFS2_UINT_EVENT(ocfs2_local_alloc_count_bits); + +DEFINE_OCFS2_INT_INT_EVENT(ocfs2_local_alloc_find_clear_bits_search_bitmap); + +DEFINE_OCFS2_ULL_INT_INT_INT_EVENT(ocfs2_local_alloc_find_clear_bits); + +DEFINE_OCFS2_INT_INT_EVENT(ocfs2_sync_local_to_main); + +TRACE_EVENT(ocfs2_sync_local_to_main_free, + TP_PROTO(int count, int bit, unsigned long long start_blk, + unsigned long long blkno), + TP_ARGS(count, bit, start_blk, blkno), + TP_STRUCT__entry( + __field(int, count) + __field(int, bit) + __field(unsigned long long, start_blk) + __field(unsigned long long, blkno) + ), + TP_fast_assign( + __entry->count = count; + __entry->bit = bit; + __entry->start_blk = start_blk; + __entry->blkno = blkno; + ), + TP_printk("%d %d %llu %llu", + __entry->count, __entry->bit, __entry->start_blk, + __entry->blkno) +); + +DEFINE_OCFS2_INT_INT_EVENT(ocfs2_local_alloc_new_window); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_local_alloc_new_window_result); + +/* End of trace events for fs/ocfs2/localalloc.c. */ + +/* Trace events for fs/ocfs2/resize.c. */ + +DEFINE_OCFS2_UINT_UINT_EVENT(ocfs2_update_last_group_and_inode); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_group_extend); + +DEFINE_OCFS2_ULL_UINT_UINT_UINT_EVENT(ocfs2_group_add); + +/* End of trace events for fs/ocfs2/resize.c. */ + +/* Trace events for fs/ocfs2/suballoc.c. */ + +DEFINE_OCFS2_ULL_EVENT(ocfs2_validate_group_descriptor); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_block_group_alloc_contig); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_block_group_alloc_discontig); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_block_group_alloc); + +DEFINE_OCFS2_UINT_UINT_EVENT(ocfs2_reserve_suballoc_bits_nospc); + +DEFINE_OCFS2_UINT_UINT_UINT_EVENT(ocfs2_reserve_suballoc_bits_no_new_group); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_reserve_new_inode_new_group); + +DEFINE_OCFS2_UINT_UINT_EVENT(ocfs2_block_group_set_bits); + +TRACE_EVENT(ocfs2_relink_block_group, + TP_PROTO(unsigned long long i_blkno, unsigned int chain, + unsigned long long bg_blkno, + unsigned long long prev_blkno), + TP_ARGS(i_blkno, chain, bg_blkno, prev_blkno), + TP_STRUCT__entry( + __field(unsigned long long, i_blkno) + __field(unsigned int, chain) + __field(unsigned long long, bg_blkno) + __field(unsigned long long, prev_blkno) + ), + TP_fast_assign( + __entry->i_blkno = i_blkno; + __entry->chain = chain; + __entry->bg_blkno = bg_blkno; + __entry->prev_blkno = prev_blkno; + ), + TP_printk("%llu %u %llu %llu", + __entry->i_blkno, __entry->chain, __entry->bg_blkno, + __entry->prev_blkno) +); + +DEFINE_OCFS2_ULL_UINT_UINT_UINT_EVENT(ocfs2_cluster_group_search_wrong_max_bits); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_cluster_group_search_max_block); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_block_group_search_max_block); + +DEFINE_OCFS2_ULL_UINT_UINT_EVENT(ocfs2_search_chain_begin); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_search_chain_succ); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_search_chain_end); + +DEFINE_OCFS2_UINT_EVENT(ocfs2_claim_suballoc_bits); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_claim_new_inode_at_loc); + +DEFINE_OCFS2_UINT_UINT_EVENT(ocfs2_block_group_clear_bits); + +TRACE_EVENT(ocfs2_free_suballoc_bits, + TP_PROTO(unsigned long long inode, unsigned long long group, + unsigned int start_bit, unsigned int count), + TP_ARGS(inode, group, start_bit, count), + TP_STRUCT__entry( + __field(unsigned long long, inode) + __field(unsigned long long, group) + __field(unsigned int, start_bit) + __field(unsigned int, count) + ), + TP_fast_assign( + __entry->inode = inode; + __entry->group = group; + __entry->start_bit = start_bit; + __entry->count = count; + ), + TP_printk("%llu %llu %u %u", __entry->inode, __entry->group, + __entry->start_bit, __entry->count) +); + +TRACE_EVENT(ocfs2_free_clusters, + TP_PROTO(unsigned long long bg_blkno, unsigned long long start_blk, + unsigned int start_bit, unsigned int count), + TP_ARGS(bg_blkno, start_blk, start_bit, count), + TP_STRUCT__entry( + __field(unsigned long long, bg_blkno) + __field(unsigned long long, start_blk) + __field(unsigned int, start_bit) + __field(unsigned int, count) + ), + TP_fast_assign( + __entry->bg_blkno = bg_blkno; + __entry->start_blk = start_blk; + __entry->start_bit = start_bit; + __entry->count = count; + ), + TP_printk("%llu %llu %u %u", __entry->bg_blkno, __entry->start_blk, + __entry->start_bit, __entry->count) +); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_get_suballoc_slot_bit); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_test_suballoc_bit); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_test_inode_bit); + +/* End of trace events for fs/ocfs2/suballoc.c. */ + +/* Trace events for fs/ocfs2/refcounttree.c. */ + +DEFINE_OCFS2_ULL_EVENT(ocfs2_validate_refcount_block); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_purge_refcount_trees); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_create_refcount_tree); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_create_refcount_tree_blkno); + +DEFINE_OCFS2_ULL_INT_INT_INT_EVENT(ocfs2_change_refcount_rec); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_expand_inline_ref_root); + +DEFINE_OCFS2_ULL_UINT_UINT_EVENT(ocfs2_divide_leaf_refcount_block); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_new_leaf_refcount_block); + +DECLARE_EVENT_CLASS(ocfs2__refcount_tree_ops, + TP_PROTO(unsigned long long blkno, int index, + unsigned long long cpos, + unsigned int clusters, unsigned int refcount), + TP_ARGS(blkno, index, cpos, clusters, refcount), + TP_STRUCT__entry( + __field(unsigned long long, blkno) + __field(int, index) + __field(unsigned long long, cpos) + __field(unsigned int, clusters) + __field(unsigned int, refcount) + ), + TP_fast_assign( + __entry->blkno = blkno; + __entry->index = index; + __entry->cpos = cpos; + __entry->clusters = clusters; + __entry->refcount = refcount; + ), + TP_printk("%llu %d %llu %u %u", __entry->blkno, __entry->index, + __entry->cpos, __entry->clusters, __entry->refcount) +); + +#define DEFINE_OCFS2_REFCOUNT_TREE_OPS_EVENT(name) \ +DEFINE_EVENT(ocfs2__refcount_tree_ops, name, \ + TP_PROTO(unsigned long long blkno, int index, \ + unsigned long long cpos, \ + unsigned int count, unsigned int refcount), \ + TP_ARGS(blkno, index, cpos, count, refcount)) + +DEFINE_OCFS2_REFCOUNT_TREE_OPS_EVENT(ocfs2_insert_refcount_rec); + +TRACE_EVENT(ocfs2_split_refcount_rec, + TP_PROTO(unsigned long long cpos, + unsigned int clusters, unsigned int refcount, + unsigned long long split_cpos, + unsigned int split_clusters, unsigned int split_refcount), + TP_ARGS(cpos, clusters, refcount, + split_cpos, split_clusters, split_refcount), + TP_STRUCT__entry( + __field(unsigned long long, cpos) + __field(unsigned int, clusters) + __field(unsigned int, refcount) + __field(unsigned long long, split_cpos) + __field(unsigned int, split_clusters) + __field(unsigned int, split_refcount) + ), + TP_fast_assign( + __entry->cpos = cpos; + __entry->clusters = clusters; + __entry->refcount = refcount; + __entry->split_cpos = split_cpos; + __entry->split_clusters = split_clusters; + __entry->split_refcount = split_refcount; + ), + TP_printk("%llu %u %u %llu %u %u", + __entry->cpos, __entry->clusters, __entry->refcount, + __entry->split_cpos, __entry->split_clusters, + __entry->split_refcount) +); + +DEFINE_OCFS2_REFCOUNT_TREE_OPS_EVENT(ocfs2_split_refcount_rec_insert); + +DEFINE_OCFS2_ULL_ULL_UINT_EVENT(ocfs2_increase_refcount_begin); + +DEFINE_OCFS2_ULL_UINT_UINT_EVENT(ocfs2_increase_refcount_change); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_increase_refcount_insert); + +DEFINE_OCFS2_ULL_UINT_UINT_EVENT(ocfs2_increase_refcount_split); + +DEFINE_OCFS2_ULL_ULL_UINT_EVENT(ocfs2_remove_refcount_extent); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_restore_refcount_block); + +DEFINE_OCFS2_ULL_ULL_UINT_EVENT(ocfs2_decrease_refcount_rec); + +TRACE_EVENT(ocfs2_decrease_refcount, + TP_PROTO(unsigned long long owner, + unsigned long long cpos, + unsigned int len, int delete), + TP_ARGS(owner, cpos, len, delete), + TP_STRUCT__entry( + __field(unsigned long long, owner) + __field(unsigned long long, cpos) + __field(unsigned int, len) + __field(int, delete) + ), + TP_fast_assign( + __entry->owner = owner; + __entry->cpos = cpos; + __entry->len = len; + __entry->delete = delete; + ), + TP_printk("%llu %llu %u %d", + __entry->owner, __entry->cpos, __entry->len, __entry->delete) +); + +DEFINE_OCFS2_ULL_UINT_UINT_UINT_EVENT(ocfs2_mark_extent_refcounted); + +DEFINE_OCFS2_ULL_UINT_UINT_UINT_EVENT(ocfs2_calc_refcount_meta_credits); + +TRACE_EVENT(ocfs2_calc_refcount_meta_credits_iterate, + TP_PROTO(int recs_add, unsigned long long cpos, + unsigned int clusters, unsigned long long r_cpos, + unsigned int r_clusters, unsigned int refcount, int index), + TP_ARGS(recs_add, cpos, clusters, r_cpos, r_clusters, refcount, index), + TP_STRUCT__entry( + __field(int, recs_add) + __field(unsigned long long, cpos) + __field(unsigned int, clusters) + __field(unsigned long long, r_cpos) + __field(unsigned int, r_clusters) + __field(unsigned int, refcount) + __field(int, index) + ), + TP_fast_assign( + __entry->recs_add = recs_add; + __entry->cpos = cpos; + __entry->clusters = clusters; + __entry->r_cpos = r_cpos; + __entry->r_clusters = r_clusters; + __entry->refcount = refcount; + __entry->index = index; + ), + TP_printk("%d %llu %u %llu %u %u %d", + __entry->recs_add, __entry->cpos, __entry->clusters, + __entry->r_cpos, __entry->r_clusters, + __entry->refcount, __entry->index) +); + +DEFINE_OCFS2_INT_INT_EVENT(ocfs2_add_refcount_flag); + +DEFINE_OCFS2_INT_INT_EVENT(ocfs2_prepare_refcount_change_for_del); + +DEFINE_OCFS2_INT_INT_EVENT(ocfs2_lock_refcount_allocators); + +DEFINE_OCFS2_ULL_UINT_UINT_UINT_EVENT(ocfs2_duplicate_clusters_by_page); + +DEFINE_OCFS2_ULL_UINT_UINT_UINT_EVENT(ocfs2_duplicate_clusters_by_jbd); + +TRACE_EVENT(ocfs2_clear_ext_refcount, + TP_PROTO(unsigned long long ino, unsigned int cpos, + unsigned int len, unsigned int p_cluster, + unsigned int ext_flags), + TP_ARGS(ino, cpos, len, p_cluster, ext_flags), + TP_STRUCT__entry( + __field(unsigned long long, ino) + __field(unsigned int, cpos) + __field(unsigned int, len) + __field(unsigned int, p_cluster) + __field(unsigned int, ext_flags) + ), + TP_fast_assign( + __entry->ino = ino; + __entry->cpos = cpos; + __entry->len = len; + __entry->p_cluster = p_cluster; + __entry->ext_flags = ext_flags; + ), + TP_printk("%llu %u %u %u %u", + __entry->ino, __entry->cpos, __entry->len, + __entry->p_cluster, __entry->ext_flags) +); + +TRACE_EVENT(ocfs2_replace_clusters, + TP_PROTO(unsigned long long ino, unsigned int cpos, + unsigned int old, unsigned int new, unsigned int len, + unsigned int ext_flags), + TP_ARGS(ino, cpos, old, new, len, ext_flags), + TP_STRUCT__entry( + __field(unsigned long long, ino) + __field(unsigned int, cpos) + __field(unsigned int, old) + __field(unsigned int, new) + __field(unsigned int, len) + __field(unsigned int, ext_flags) + ), + TP_fast_assign( + __entry->ino = ino; + __entry->cpos = cpos; + __entry->old = old; + __entry->new = new; + __entry->len = len; + __entry->ext_flags = ext_flags; + ), + TP_printk("%llu %u %u %u %u %u", + __entry->ino, __entry->cpos, __entry->old, __entry->new, + __entry->len, __entry->ext_flags) +); + +DEFINE_OCFS2_ULL_UINT_UINT_UINT_EVENT(ocfs2_make_clusters_writable); + +TRACE_EVENT(ocfs2_refcount_cow_hunk, + TP_PROTO(unsigned long long ino, unsigned int cpos, + unsigned int write_len, unsigned int max_cpos, + unsigned int cow_start, unsigned int cow_len), + TP_ARGS(ino, cpos, write_len, max_cpos, cow_start, cow_len), + TP_STRUCT__entry( + __field(unsigned long long, ino) + __field(unsigned int, cpos) + __field(unsigned int, write_len) + __field(unsigned int, max_cpos) + __field(unsigned int, cow_start) + __field(unsigned int, cow_len) + ), + TP_fast_assign( + __entry->ino = ino; + __entry->cpos = cpos; + __entry->write_len = write_len; + __entry->max_cpos = max_cpos; + __entry->cow_start = cow_start; + __entry->cow_len = cow_len; + ), + TP_printk("%llu %u %u %u %u %u", + __entry->ino, __entry->cpos, __entry->write_len, + __entry->max_cpos, __entry->cow_start, __entry->cow_len) +); + +/* End of trace events for fs/ocfs2/refcounttree.c. */ + +/* Trace events for fs/ocfs2/aops.c. */ + +DECLARE_EVENT_CLASS(ocfs2__get_block, + TP_PROTO(unsigned long long ino, unsigned long long iblock, + void *bh_result, int create), + TP_ARGS(ino, iblock, bh_result, create), + TP_STRUCT__entry( + __field(unsigned long long, ino) + __field(unsigned long long, iblock) + __field(void *, bh_result) + __field(int, create) + ), + TP_fast_assign( + __entry->ino = ino; + __entry->iblock = iblock; + __entry->bh_result = bh_result; + __entry->create = create; + ), + TP_printk("%llu %llu %p %d", + __entry->ino, __entry->iblock, + __entry->bh_result, __entry->create) +); + +#define DEFINE_OCFS2_GET_BLOCK_EVENT(name) \ +DEFINE_EVENT(ocfs2__get_block, name, \ + TP_PROTO(unsigned long long ino, unsigned long long iblock, \ + void *bh_result, int create), \ + TP_ARGS(ino, iblock, bh_result, create)) + +DEFINE_OCFS2_GET_BLOCK_EVENT(ocfs2_symlink_get_block); + +DEFINE_OCFS2_GET_BLOCK_EVENT(ocfs2_get_block); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_get_block_end); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_readpage); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_writepage); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_bmap); + +TRACE_EVENT(ocfs2_try_to_write_inline_data, + TP_PROTO(unsigned long long ino, unsigned int len, + unsigned long long pos, unsigned int flags), + TP_ARGS(ino, len, pos, flags), + TP_STRUCT__entry( + __field(unsigned long long, ino) + __field(unsigned int, len) + __field(unsigned long long, pos) + __field(unsigned int, flags) + ), + TP_fast_assign( + __entry->ino = ino; + __entry->len = len; + __entry->pos = pos; + __entry->flags = flags; + ), + TP_printk("%llu %u %llu 0x%x", + __entry->ino, __entry->len, __entry->pos, __entry->flags) +); + +TRACE_EVENT(ocfs2_write_begin_nolock, + TP_PROTO(unsigned long long ino, + long long i_size, unsigned int i_clusters, + unsigned long long pos, unsigned int len, + unsigned int flags, void *page, + unsigned int clusters, unsigned int extents_to_split), + TP_ARGS(ino, i_size, i_clusters, pos, len, flags, + page, clusters, extents_to_split), + TP_STRUCT__entry( + __field(unsigned long long, ino) + __field(long long, i_size) + __field(unsigned int, i_clusters) + __field(unsigned long long, pos) + __field(unsigned int, len) + __field(unsigned int, flags) + __field(void *, page) + __field(unsigned int, clusters) + __field(unsigned int, extents_to_split) + ), + TP_fast_assign( + __entry->ino = ino; + __entry->i_size = i_size; + __entry->i_clusters = i_clusters; + __entry->pos = pos; + __entry->len = len; + __entry->flags = flags; + __entry->page = page; + __entry->clusters = clusters; + __entry->extents_to_split = extents_to_split; + ), + TP_printk("%llu %lld %u %llu %u %u %p %u %u", + __entry->ino, __entry->i_size, __entry->i_clusters, + __entry->pos, __entry->len, + __entry->flags, __entry->page, __entry->clusters, + __entry->extents_to_split) +); + +TRACE_EVENT(ocfs2_write_end_inline, + TP_PROTO(unsigned long long ino, + unsigned long long pos, unsigned int copied, + unsigned int id_count, unsigned int features), + TP_ARGS(ino, pos, copied, id_count, features), + TP_STRUCT__entry( + __field(unsigned long long, ino) + __field(unsigned long long, pos) + __field(unsigned int, copied) + __field(unsigned int, id_count) + __field(unsigned int, features) + ), + TP_fast_assign( + __entry->ino = ino; + __entry->pos = pos; + __entry->copied = copied; + __entry->id_count = id_count; + __entry->features = features; + ), + TP_printk("%llu %llu %u %u %u", + __entry->ino, __entry->pos, __entry->copied, + __entry->id_count, __entry->features) +); + +/* End of trace events for fs/ocfs2/aops.c. */ + +/* Trace events for fs/ocfs2/mmap.c. */ + +TRACE_EVENT(ocfs2_fault, + TP_PROTO(unsigned long long ino, + void *area, void *page, unsigned long pgoff), + TP_ARGS(ino, area, page, pgoff), + TP_STRUCT__entry( + __field(unsigned long long, ino) + __field(void *, area) + __field(void *, page) + __field(unsigned long, pgoff) + ), + TP_fast_assign( + __entry->ino = ino; + __entry->area = area; + __entry->page = page; + __entry->pgoff = pgoff; + ), + TP_printk("%llu %p %p %lu", + __entry->ino, __entry->area, __entry->page, __entry->pgoff) +); + +/* End of trace events for fs/ocfs2/mmap.c. */ + +/* Trace events for fs/ocfs2/file.c. */ + +DECLARE_EVENT_CLASS(ocfs2__file_ops, + TP_PROTO(void *inode, void *file, void *dentry, + unsigned long long ino, + unsigned int d_len, const unsigned char *d_name, + unsigned long long para), + TP_ARGS(inode, file, dentry, ino, d_len, d_name, para), + TP_STRUCT__entry( + __field(void *, inode) + __field(void *, file) + __field(void *, dentry) + __field(unsigned long long, ino) + __field(unsigned int, d_len) + __string(d_name, d_name) + __field(unsigned long long, para) + ), + TP_fast_assign( + __entry->inode = inode; + __entry->file = file; + __entry->dentry = dentry; + __entry->ino = ino; + __entry->d_len = d_len; + __assign_str(d_name, d_name); + __entry->para = para; + ), + TP_printk("%p %p %p %llu %llu %.*s", __entry->inode, __entry->file, + __entry->dentry, __entry->ino, __entry->para, + __entry->d_len, __get_str(d_name)) +); + +#define DEFINE_OCFS2_FILE_OPS(name) \ +DEFINE_EVENT(ocfs2__file_ops, name, \ +TP_PROTO(void *inode, void *file, void *dentry, \ + unsigned long long ino, \ + unsigned int d_len, const unsigned char *d_name, \ + unsigned long long mode), \ + TP_ARGS(inode, file, dentry, ino, d_len, d_name, mode)) + +DEFINE_OCFS2_FILE_OPS(ocfs2_file_open); + +DEFINE_OCFS2_FILE_OPS(ocfs2_file_release); + +DEFINE_OCFS2_FILE_OPS(ocfs2_sync_file); + +DEFINE_OCFS2_FILE_OPS(ocfs2_file_aio_write); + +DEFINE_OCFS2_FILE_OPS(ocfs2_file_splice_write); + +DEFINE_OCFS2_FILE_OPS(ocfs2_file_splice_read); + +DEFINE_OCFS2_FILE_OPS(ocfs2_file_aio_read); + +DEFINE_OCFS2_ULL_ULL_ULL_EVENT(ocfs2_truncate_file); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_truncate_file_error); + +TRACE_EVENT(ocfs2_extend_allocation, + TP_PROTO(unsigned long long ip_blkno, unsigned long long size, + unsigned int clusters, unsigned int clusters_to_add, + int why, int restart_func), + TP_ARGS(ip_blkno, size, clusters, clusters_to_add, why, restart_func), + TP_STRUCT__entry( + __field(unsigned long long, ip_blkno) + __field(unsigned long long, size) + __field(unsigned int, clusters) + __field(unsigned int, clusters_to_add) + __field(int, why) + __field(int, restart_func) + ), + TP_fast_assign( + __entry->ip_blkno = ip_blkno; + __entry->size = size; + __entry->clusters = clusters; + __entry->clusters_to_add = clusters_to_add; + __entry->why = why; + __entry->restart_func = restart_func; + ), + TP_printk("%llu %llu %u %u %d %d", + __entry->ip_blkno, __entry->size, __entry->clusters, + __entry->clusters_to_add, __entry->why, __entry->restart_func) +); + +TRACE_EVENT(ocfs2_extend_allocation_end, + TP_PROTO(unsigned long long ino, + unsigned int di_clusters, unsigned long long di_size, + unsigned int ip_clusters, unsigned long long i_size), + TP_ARGS(ino, di_clusters, di_size, ip_clusters, i_size), + TP_STRUCT__entry( + __field(unsigned long long, ino) + __field(unsigned int, di_clusters) + __field(unsigned long long, di_size) + __field(unsigned int, ip_clusters) + __field(unsigned long long, i_size) + ), + TP_fast_assign( + __entry->ino = ino; + __entry->di_clusters = di_clusters; + __entry->di_size = di_size; + __entry->ip_clusters = ip_clusters; + __entry->i_size = i_size; + ), + TP_printk("%llu %u %llu %u %llu", __entry->ino, __entry->di_clusters, + __entry->di_size, __entry->ip_clusters, __entry->i_size) +); + +TRACE_EVENT(ocfs2_write_zero_page, + TP_PROTO(unsigned long long ino, + unsigned long long abs_from, unsigned long long abs_to, + unsigned long index, unsigned int zero_from, + unsigned int zero_to), + TP_ARGS(ino, abs_from, abs_to, index, zero_from, zero_to), + TP_STRUCT__entry( + __field(unsigned long long, ino) + __field(unsigned long long, abs_from) + __field(unsigned long long, abs_to) + __field(unsigned long, index) + __field(unsigned int, zero_from) + __field(unsigned int, zero_to) + ), + TP_fast_assign( + __entry->ino = ino; + __entry->abs_from = abs_from; + __entry->abs_to = abs_to; + __entry->index = index; + __entry->zero_from = zero_from; + __entry->zero_to = zero_to; + ), + TP_printk("%llu %llu %llu %lu %u %u", __entry->ino, + __entry->abs_from, __entry->abs_to, + __entry->index, __entry->zero_from, __entry->zero_to) +); + +DEFINE_OCFS2_ULL_ULL_ULL_EVENT(ocfs2_zero_extend_range); + +DEFINE_OCFS2_ULL_ULL_ULL_EVENT(ocfs2_zero_extend); + +TRACE_EVENT(ocfs2_setattr, + TP_PROTO(void *inode, void *dentry, + unsigned long long ino, + unsigned int d_len, const unsigned char *d_name, + unsigned int ia_valid, unsigned int ia_mode, + unsigned int ia_uid, unsigned int ia_gid), + TP_ARGS(inode, dentry, ino, d_len, d_name, + ia_valid, ia_mode, ia_uid, ia_gid), + TP_STRUCT__entry( + __field(void *, inode) + __field(void *, dentry) + __field(unsigned long long, ino) + __field(unsigned int, d_len) + __string(d_name, d_name) + __field(unsigned int, ia_valid) + __field(unsigned int, ia_mode) + __field(unsigned int, ia_uid) + __field(unsigned int, ia_gid) + ), + TP_fast_assign( + __entry->inode = inode; + __entry->dentry = dentry; + __entry->ino = ino; + __entry->d_len = d_len; + __assign_str(d_name, d_name); + __entry->ia_valid = ia_valid; + __entry->ia_mode = ia_mode; + __entry->ia_uid = ia_uid; + __entry->ia_gid = ia_gid; + ), + TP_printk("%p %p %llu %.*s %u %u %u %u", __entry->inode, + __entry->dentry, __entry->ino, __entry->d_len, + __get_str(d_name), __entry->ia_valid, __entry->ia_mode, + __entry->ia_uid, __entry->ia_gid) +); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_write_remove_suid); + +DEFINE_OCFS2_ULL_ULL_ULL_EVENT(ocfs2_zero_partial_clusters); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_zero_partial_clusters_range1); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_zero_partial_clusters_range2); + +DEFINE_OCFS2_ULL_ULL_ULL_EVENT(ocfs2_remove_inode_range); + +TRACE_EVENT(ocfs2_prepare_inode_for_write, + TP_PROTO(unsigned long long ino, unsigned long long saved_pos, + int appending, unsigned long count, + int *direct_io, int *has_refcount), + TP_ARGS(ino, saved_pos, appending, count, direct_io, has_refcount), + TP_STRUCT__entry( + __field(unsigned long long, ino) + __field(unsigned long long, saved_pos) + __field(int, appending) + __field(unsigned long, count) + __field(int, direct_io) + __field(int, has_refcount) + ), + TP_fast_assign( + __entry->ino = ino; + __entry->saved_pos = saved_pos; + __entry->appending = appending; + __entry->count = count; + __entry->direct_io = direct_io ? *direct_io : -1; + __entry->has_refcount = has_refcount ? *has_refcount : -1; + ), + TP_printk("%llu %llu %d %lu %d %d", __entry->ino, + __entry->saved_pos, __entry->appending, __entry->count, + __entry->direct_io, __entry->has_refcount) +); + +DEFINE_OCFS2_INT_EVENT(generic_file_aio_read_ret); + +/* End of trace events for fs/ocfs2/file.c. */ + +/* Trace events for fs/ocfs2/inode.c. */ + +TRACE_EVENT(ocfs2_iget_begin, + TP_PROTO(unsigned long long ino, unsigned int flags, int sysfile_type), + TP_ARGS(ino, flags, sysfile_type), + TP_STRUCT__entry( + __field(unsigned long long, ino) + __field(unsigned int, flags) + __field(int, sysfile_type) + ), + TP_fast_assign( + __entry->ino = ino; + __entry->flags = flags; + __entry->sysfile_type = sysfile_type; + ), + TP_printk("%llu %u %d", __entry->ino, + __entry->flags, __entry->sysfile_type) +); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_iget5_locked); + +TRACE_EVENT(ocfs2_iget_end, + TP_PROTO(void *inode, unsigned long long ino), + TP_ARGS(inode, ino), + TP_STRUCT__entry( + __field(void *, inode) + __field(unsigned long long, ino) + ), + TP_fast_assign( + __entry->inode = inode; + __entry->ino = ino; + ), + TP_printk("%p %llu", __entry->inode, __entry->ino) +); + +TRACE_EVENT(ocfs2_find_actor, + TP_PROTO(void *inode, unsigned long long ino, + void *args, unsigned long long fi_blkno), + TP_ARGS(inode, ino, args, fi_blkno), + TP_STRUCT__entry( + __field(void *, inode) + __field(unsigned long long, ino) + __field(void *, args) + __field(unsigned long long, fi_blkno) + ), + TP_fast_assign( + __entry->inode = inode; + __entry->ino = ino; + __entry->args = args; + __entry->fi_blkno = fi_blkno; + ), + TP_printk("%p %llu %p %llu", __entry->inode, __entry->ino, + __entry->args, __entry->fi_blkno) +); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_populate_inode); + +DEFINE_OCFS2_ULL_INT_EVENT(ocfs2_read_locked_inode); + +DEFINE_OCFS2_INT_INT_EVENT(ocfs2_check_orphan_recovery_state); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_validate_inode_block); + +TRACE_EVENT(ocfs2_inode_is_valid_to_delete, + TP_PROTO(void *task, void *dc_task, unsigned long long ino, + unsigned int flags), + TP_ARGS(task, dc_task, ino, flags), + TP_STRUCT__entry( + __field(void *, task) + __field(void *, dc_task) + __field(unsigned long long, ino) + __field(unsigned int, flags) + ), + TP_fast_assign( + __entry->task = task; + __entry->dc_task = dc_task; + __entry->ino = ino; + __entry->flags = flags; + ), + TP_printk("%p %p %llu %u", __entry->task, __entry->dc_task, + __entry->ino, __entry->flags) +); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_query_inode_wipe_begin); + +DEFINE_OCFS2_UINT_EVENT(ocfs2_query_inode_wipe_succ); + +DEFINE_OCFS2_INT_INT_EVENT(ocfs2_query_inode_wipe_end); + +DEFINE_OCFS2_ULL_INT_EVENT(ocfs2_cleanup_delete_inode); + +DEFINE_OCFS2_ULL_ULL_UINT_EVENT(ocfs2_delete_inode); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_clear_inode); + +DEFINE_OCFS2_ULL_UINT_UINT_EVENT(ocfs2_drop_inode); + +TRACE_EVENT(ocfs2_inode_revalidate, + TP_PROTO(void *inode, unsigned long long ino, + unsigned int flags), + TP_ARGS(inode, ino, flags), + TP_STRUCT__entry( + __field(void *, inode) + __field(unsigned long long, ino) + __field(unsigned int, flags) + ), + TP_fast_assign( + __entry->inode = inode; + __entry->ino = ino; + __entry->flags = flags; + ), + TP_printk("%p %llu %u", __entry->inode, __entry->ino, __entry->flags) +); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_mark_inode_dirty); + +/* End of trace events for fs/ocfs2/inode.c. */ + +/* Trace events for fs/ocfs2/extent_map.c. */ + +TRACE_EVENT(ocfs2_read_virt_blocks, + TP_PROTO(void *inode, unsigned long long vblock, int nr, + void *bhs, unsigned int flags, void *validate), + TP_ARGS(inode, vblock, nr, bhs, flags, validate), + TP_STRUCT__entry( + __field(void *, inode) + __field(unsigned long long, vblock) + __field(int, nr) + __field(void *, bhs) + __field(unsigned int, flags) + __field(void *, validate) + ), + TP_fast_assign( + __entry->inode = inode; + __entry->vblock = vblock; + __entry->nr = nr; + __entry->bhs = bhs; + __entry->flags = flags; + __entry->validate = validate; + ), + TP_printk("%p %llu %d %p %x %p", __entry->inode, __entry->vblock, + __entry->nr, __entry->bhs, __entry->flags, __entry->validate) +); + +/* End of trace events for fs/ocfs2/extent_map.c. */ + +/* Trace events for fs/ocfs2/slot_map.c. */ + +DEFINE_OCFS2_UINT_EVENT(ocfs2_refresh_slot_info); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_map_slot_buffers); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_map_slot_buffers_block); + +DEFINE_OCFS2_INT_EVENT(ocfs2_find_slot); + +/* End of trace events for fs/ocfs2/slot_map.c. */ + +/* Trace events for fs/ocfs2/heartbeat.c. */ + +DEFINE_OCFS2_INT_EVENT(ocfs2_do_node_down); + +/* End of trace events for fs/ocfs2/heartbeat.c. */ + +/* Trace events for fs/ocfs2/super.c. */ + +TRACE_EVENT(ocfs2_remount, + TP_PROTO(unsigned long s_flags, unsigned long osb_flags, int flags), + TP_ARGS(s_flags, osb_flags, flags), + TP_STRUCT__entry( + __field(unsigned long, s_flags) + __field(unsigned long, osb_flags) + __field(int, flags) + ), + TP_fast_assign( + __entry->s_flags = s_flags; + __entry->osb_flags = osb_flags; + __entry->flags = flags; + ), + TP_printk("%lu %lu %d", __entry->s_flags, + __entry->osb_flags, __entry->flags) +); + +TRACE_EVENT(ocfs2_fill_super, + TP_PROTO(void *sb, void *data, int silent), + TP_ARGS(sb, data, silent), + TP_STRUCT__entry( + __field(void *, sb) + __field(void *, data) + __field(int, silent) + ), + TP_fast_assign( + __entry->sb = sb; + __entry->data = data; + __entry->silent = silent; + ), + TP_printk("%p %p %d", __entry->sb, + __entry->data, __entry->silent) +); + +TRACE_EVENT(ocfs2_parse_options, + TP_PROTO(int is_remount, char *options), + TP_ARGS(is_remount, options), + TP_STRUCT__entry( + __field(int, is_remount) + __string(options, options) + ), + TP_fast_assign( + __entry->is_remount = is_remount; + __assign_str(options, options); + ), + TP_printk("%d %s", __entry->is_remount, __get_str(options)) +); + +DEFINE_OCFS2_POINTER_EVENT(ocfs2_put_super); + +TRACE_EVENT(ocfs2_statfs, + TP_PROTO(void *sb, void *buf), + TP_ARGS(sb, buf), + TP_STRUCT__entry( + __field(void *, sb) + __field(void *, buf) + ), + TP_fast_assign( + __entry->sb = sb; + __entry->buf = buf; + ), + TP_printk("%p %p", __entry->sb, __entry->buf) +); + +DEFINE_OCFS2_POINTER_EVENT(ocfs2_dismount_volume); + +TRACE_EVENT(ocfs2_initialize_super, + TP_PROTO(char *label, char *uuid_str, unsigned long long root_dir, + unsigned long long system_dir, int cluster_bits), + TP_ARGS(label, uuid_str, root_dir, system_dir, cluster_bits), + TP_STRUCT__entry( + __string(label, label) + __string(uuid_str, uuid_str) + __field(unsigned long long, root_dir) + __field(unsigned long long, system_dir) + __field(int, cluster_bits) + ), + TP_fast_assign( + __assign_str(label, label); + __assign_str(uuid_str, uuid_str); + __entry->root_dir = root_dir; + __entry->system_dir = system_dir; + __entry->cluster_bits = cluster_bits; + ), + TP_printk("%s %s %llu %llu %d", __get_str(label), __get_str(uuid_str), + __entry->root_dir, __entry->system_dir, __entry->cluster_bits) +); + +/* End of trace events for fs/ocfs2/super.c. */ + +/* Trace events for fs/ocfs2/xattr.c. */ + +DEFINE_OCFS2_ULL_EVENT(ocfs2_validate_xattr_block); + +DEFINE_OCFS2_UINT_EVENT(ocfs2_xattr_extend_allocation); + +TRACE_EVENT(ocfs2_init_xattr_set_ctxt, + TP_PROTO(const char *name, int meta, int clusters, int credits), + TP_ARGS(name, meta, clusters, credits), + TP_STRUCT__entry( + __string(name, name) + __field(int, meta) + __field(int, clusters) + __field(int, credits) + ), + TP_fast_assign( + __assign_str(name, name); + __entry->meta = meta; + __entry->clusters = clusters; + __entry->credits = credits; + ), + TP_printk("%s %d %d %d", __get_str(name), __entry->meta, + __entry->clusters, __entry->credits) +); + +DECLARE_EVENT_CLASS(ocfs2__xattr_find, + TP_PROTO(unsigned long long ino, const char *name, int name_index, + unsigned int hash, unsigned long long location, + int xe_index), + TP_ARGS(ino, name, name_index, hash, location, xe_index), + TP_STRUCT__entry( + __field(unsigned long long, ino) + __string(name, name) + __field(int, name_index) + __field(unsigned int, hash) + __field(unsigned long long, location) + __field(int, xe_index) + ), + TP_fast_assign( + __entry->ino = ino; + __assign_str(name, name); + __entry->name_index = name_index; + __entry->hash = hash; + __entry->location = location; + __entry->xe_index = xe_index; + ), + TP_printk("%llu %s %d %u %llu %d", __entry->ino, __get_str(name), + __entry->name_index, __entry->hash, __entry->location, + __entry->xe_index) +); + +#define DEFINE_OCFS2_XATTR_FIND_EVENT(name) \ +DEFINE_EVENT(ocfs2__xattr_find, name, \ +TP_PROTO(unsigned long long ino, const char *name, int name_index, \ + unsigned int hash, unsigned long long bucket, \ + int xe_index), \ + TP_ARGS(ino, name, name_index, hash, bucket, xe_index)) + +DEFINE_OCFS2_XATTR_FIND_EVENT(ocfs2_xattr_bucket_find); + +DEFINE_OCFS2_XATTR_FIND_EVENT(ocfs2_xattr_index_block_find); + +DEFINE_OCFS2_XATTR_FIND_EVENT(ocfs2_xattr_index_block_find_rec); + +DEFINE_OCFS2_ULL_ULL_UINT_EVENT(ocfs2_iterate_xattr_buckets); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_iterate_xattr_bucket); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_cp_xattr_block_to_bucket_begin); + +DEFINE_OCFS2_UINT_UINT_UINT_EVENT(ocfs2_cp_xattr_block_to_bucket_end); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_xattr_create_index_block_begin); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_xattr_create_index_block); + +DEFINE_OCFS2_ULL_UINT_UINT_UINT_EVENT(ocfs2_defrag_xattr_bucket); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_mv_xattr_bucket_cross_cluster); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_divide_xattr_bucket_begin); + +DEFINE_OCFS2_UINT_UINT_UINT_EVENT(ocfs2_divide_xattr_bucket_move); + +DEFINE_OCFS2_ULL_ULL_UINT_EVENT(ocfs2_cp_xattr_bucket); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_mv_xattr_buckets); + +DEFINE_OCFS2_ULL_ULL_UINT_EVENT(ocfs2_adjust_xattr_cross_cluster); + +DEFINE_OCFS2_ULL_ULL_UINT_UINT_EVENT(ocfs2_add_new_xattr_cluster_begin); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_add_new_xattr_cluster); + +DEFINE_OCFS2_ULL_UINT_UINT_EVENT(ocfs2_add_new_xattr_cluster_insert); + +DEFINE_OCFS2_ULL_ULL_UINT_UINT_EVENT(ocfs2_extend_xattr_bucket); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_add_new_xattr_bucket); + +DEFINE_OCFS2_ULL_UINT_UINT_EVENT(ocfs2_xattr_bucket_value_truncate); + +DEFINE_OCFS2_ULL_ULL_UINT_UINT_EVENT(ocfs2_rm_xattr_cluster); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_reflink_xattr_header); + +DEFINE_OCFS2_ULL_INT_EVENT(ocfs2_create_empty_xattr_block); + +DEFINE_OCFS2_STRING_EVENT(ocfs2_xattr_set_entry_bucket); + +DEFINE_OCFS2_STRING_EVENT(ocfs2_xattr_set_entry_index_block); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_xattr_bucket_value_refcount); + +DEFINE_OCFS2_ULL_UINT_UINT_EVENT(ocfs2_reflink_xattr_buckets); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_reflink_xattr_rec); + +/* End of trace events for fs/ocfs2/xattr.c. */ + +/* Trace events for fs/ocfs2/reservations.c. */ + +DEFINE_OCFS2_UINT_UINT_EVENT(ocfs2_resv_insert); + +DEFINE_OCFS2_ULL_UINT_UINT_UINT_EVENT(ocfs2_resmap_find_free_bits_begin); + +DEFINE_OCFS2_UINT_UINT_EVENT(ocfs2_resmap_find_free_bits_end); + +TRACE_EVENT(ocfs2_resv_find_window_begin, + TP_PROTO(unsigned int r_start, unsigned int r_end, unsigned int goal, + unsigned int wanted, int empty_root), + TP_ARGS(r_start, r_end, goal, wanted, empty_root), + TP_STRUCT__entry( + __field(unsigned int, r_start) + __field(unsigned int, r_end) + __field(unsigned int, goal) + __field(unsigned int, wanted) + __field(int, empty_root) + ), + TP_fast_assign( + __entry->r_start = r_start; + __entry->r_end = r_end; + __entry->goal = goal; + __entry->wanted = wanted; + __entry->empty_root = empty_root; + ), + TP_printk("%u %u %u %u %d", __entry->r_start, __entry->r_end, + __entry->goal, __entry->wanted, __entry->empty_root) +); + +DEFINE_OCFS2_UINT_UINT_EVENT(ocfs2_resv_find_window_prev); + +DEFINE_OCFS2_INT_INT_EVENT(ocfs2_resv_find_window_next); + +DEFINE_OCFS2_UINT_UINT_UINT_EVENT(ocfs2_cannibalize_resv_begin); + +TRACE_EVENT(ocfs2_cannibalize_resv_end, + TP_PROTO(unsigned int start, unsigned int end, unsigned int len, + unsigned int last_start, unsigned int last_len), + TP_ARGS(start, end, len, last_start, last_len), + TP_STRUCT__entry( + __field(unsigned int, start) + __field(unsigned int, end) + __field(unsigned int, len) + __field(unsigned int, last_start) + __field(unsigned int, last_len) + ), + TP_fast_assign( + __entry->start = start; + __entry->end = end; + __entry->len = len; + __entry->last_start = last_start; + __entry->last_len = last_len; + ), + TP_printk("%u %u %u %u %u", __entry->start, __entry->end, + __entry->len, __entry->last_start, __entry->last_len) +); + +DEFINE_OCFS2_UINT_UINT_EVENT(ocfs2_resmap_resv_bits); + +TRACE_EVENT(ocfs2_resmap_claimed_bits_begin, + TP_PROTO(unsigned int cstart, unsigned int cend, unsigned int clen, + unsigned int r_start, unsigned int r_end, unsigned int r_len, + unsigned int last_start, unsigned int last_len), + TP_ARGS(cstart, cend, clen, r_start, r_end, + r_len, last_start, last_len), + TP_STRUCT__entry( + __field(unsigned int, cstart) + __field(unsigned int, cend) + __field(unsigned int, clen) + __field(unsigned int, r_start) + __field(unsigned int, r_end) + __field(unsigned int, r_len) + __field(unsigned int, last_start) + __field(unsigned int, last_len) + ), + TP_fast_assign( + __entry->cstart = cstart; + __entry->cend = cend; + __entry->clen = clen; + __entry->r_start = r_start; + __entry->r_end = r_end; + __entry->r_len = r_len; + __entry->last_start = last_start; + __entry->last_len = last_len; + ), + TP_printk("%u %u %u %u %u %u %u %u", + __entry->cstart, __entry->cend, __entry->clen, + __entry->r_start, __entry->r_end, __entry->r_len, + __entry->last_start, __entry->last_len) +); + +TRACE_EVENT(ocfs2_resmap_claimed_bits_end, + TP_PROTO(unsigned int start, unsigned int end, unsigned int len, + unsigned int last_start, unsigned int last_len), + TP_ARGS(start, end, len, last_start, last_len), + TP_STRUCT__entry( + __field(unsigned int, start) + __field(unsigned int, end) + __field(unsigned int, len) + __field(unsigned int, last_start) + __field(unsigned int, last_len) + ), + TP_fast_assign( + __entry->start = start; + __entry->end = end; + __entry->len = len; + __entry->last_start = last_start; + __entry->last_len = last_len; + ), + TP_printk("%u %u %u %u %u", __entry->start, __entry->end, + __entry->len, __entry->last_start, __entry->last_len) +); + +/* End of trace events for fs/ocfs2/reservations.c. */ + +/* Trace events for fs/ocfs2/quota_local.c. */ + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_recover_local_quota_file); + +DEFINE_OCFS2_INT_EVENT(ocfs2_finish_quota_recovery); + +DEFINE_OCFS2_ULL_ULL_UINT_EVENT(olq_set_dquot); + +/* End of trace events for fs/ocfs2/quota_local.c. */ + +/* Trace events for fs/ocfs2/quota_global.c. */ + +DEFINE_OCFS2_ULL_EVENT(ocfs2_validate_quota_block); + +TRACE_EVENT(ocfs2_sync_dquot, + TP_PROTO(unsigned int dq_id, long long dqb_curspace, + long long spacechange, long long curinodes, + long long inodechange), + TP_ARGS(dq_id, dqb_curspace, spacechange, curinodes, inodechange), + TP_STRUCT__entry( + __field(unsigned int, dq_id) + __field(long long, dqb_curspace) + __field(long long, spacechange) + __field(long long, curinodes) + __field(long long, inodechange) + ), + TP_fast_assign( + __entry->dq_id = dq_id; + __entry->dqb_curspace = dqb_curspace; + __entry->spacechange = spacechange; + __entry->curinodes = curinodes; + __entry->inodechange = inodechange; + ), + TP_printk("%u %lld %lld %lld %lld", __entry->dq_id, + __entry->dqb_curspace, __entry->spacechange, + __entry->curinodes, __entry->inodechange) +); + +TRACE_EVENT(ocfs2_sync_dquot_helper, + TP_PROTO(unsigned int dq_id, unsigned int dq_type, unsigned long type, + const char *s_id), + TP_ARGS(dq_id, dq_type, type, s_id), + + TP_STRUCT__entry( + __field(unsigned int, dq_id) + __field(unsigned int, dq_type) + __field(unsigned long, type) + __string(s_id, s_id) + ), + TP_fast_assign( + __entry->dq_id = dq_id; + __entry->dq_type = dq_type; + __entry->type = type; + __assign_str(s_id, s_id); + ), + TP_printk("%u %u %lu %s", __entry->dq_id, __entry->dq_type, + __entry->type, __get_str(s_id)) +); + +DEFINE_OCFS2_UINT_INT_EVENT(ocfs2_write_dquot); + +DEFINE_OCFS2_UINT_INT_EVENT(ocfs2_release_dquot); + +DEFINE_OCFS2_UINT_INT_EVENT(ocfs2_acquire_dquot); + +DEFINE_OCFS2_UINT_INT_EVENT(ocfs2_mark_dquot_dirty); + +/* End of trace events for fs/ocfs2/quota_global.c. */ + +/* Trace events for fs/ocfs2/dir.c. */ +DEFINE_OCFS2_INT_EVENT(ocfs2_search_dirblock); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_validate_dir_block); + +DEFINE_OCFS2_POINTER_EVENT(ocfs2_find_entry_el); + +TRACE_EVENT(ocfs2_dx_dir_search, + TP_PROTO(unsigned long long ino, int namelen, const char *name, + unsigned int major_hash, unsigned int minor_hash, + unsigned long long blkno), + TP_ARGS(ino, namelen, name, major_hash, minor_hash, blkno), + TP_STRUCT__entry( + __field(unsigned long long, ino) + __field(int, namelen) + __string(name, name) + __field(unsigned int, major_hash) + __field(unsigned int,minor_hash) + __field(unsigned long long, blkno) + ), + TP_fast_assign( + __entry->ino = ino; + __entry->namelen = namelen; + __assign_str(name, name); + __entry->major_hash = major_hash; + __entry->minor_hash = minor_hash; + __entry->blkno = blkno; + ), + TP_printk("%llu %.*s %u %u %llu", __entry->ino, + __entry->namelen, __get_str(name), + __entry->major_hash, __entry->minor_hash, __entry->blkno) +); + +DEFINE_OCFS2_UINT_UINT_EVENT(ocfs2_dx_dir_search_leaf_info); + +DEFINE_OCFS2_ULL_INT_EVENT(ocfs2_delete_entry_dx); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_readdir); + +TRACE_EVENT(ocfs2_find_files_on_disk, + TP_PROTO(int namelen, const char *name, void *blkno, + unsigned long long dir), + TP_ARGS(namelen, name, blkno, dir), + TP_STRUCT__entry( + __field(int, namelen) + __string(name, name) + __field(void *, blkno) + __field(unsigned long long, dir) + ), + TP_fast_assign( + __entry->namelen = namelen; + __assign_str(name, name); + __entry->blkno = blkno; + __entry->dir = dir; + ), + TP_printk("%.*s %p %llu", __entry->namelen, __get_str(name), + __entry->blkno, __entry->dir) +); + +TRACE_EVENT(ocfs2_check_dir_for_entry, + TP_PROTO(unsigned long long dir, int namelen, const char *name), + TP_ARGS(dir, namelen, name), + TP_STRUCT__entry( + __field(unsigned long long, dir) + __field(int, namelen) + __string(name, name) + ), + TP_fast_assign( + __entry->dir = dir; + __entry->namelen = namelen; + __assign_str(name, name); + ), + TP_printk("%llu %.*s", __entry->dir, + __entry->namelen, __get_str(name)) +); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_dx_dir_attach_index); + +DEFINE_OCFS2_ULL_ULL_UINT_EVENT(ocfs2_dx_dir_format_cluster); + +TRACE_EVENT(ocfs2_dx_dir_index_root_block, + TP_PROTO(unsigned long long dir, + unsigned int major_hash, unsigned int minor_hash, + int namelen, const char *name, unsigned int num_used), + TP_ARGS(dir, major_hash, minor_hash, namelen, name, num_used), + TP_STRUCT__entry( + __field(unsigned long long, dir) + __field(unsigned int, major_hash) + __field(unsigned int, minor_hash) + __field(int, namelen) + __string(name, name) + __field(unsigned int, num_used) + ), + TP_fast_assign( + __entry->dir = dir; + __entry->major_hash = major_hash; + __entry->minor_hash = minor_hash; + __entry->namelen = namelen; + __assign_str(name, name); + __entry->num_used = num_used; + ), + TP_printk("%llu %x %x %.*s %u", __entry->dir, + __entry->major_hash, __entry->minor_hash, + __entry->namelen, __get_str(name), __entry->num_used) +); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_extend_dir); + +DEFINE_OCFS2_ULL_ULL_UINT_EVENT(ocfs2_dx_dir_rebalance); + +DEFINE_OCFS2_UINT_UINT_UINT_EVENT(ocfs2_dx_dir_rebalance_split); + +DEFINE_OCFS2_ULL_INT_EVENT(ocfs2_prepare_dir_for_insert); + +/* End of trace events for fs/ocfs2/dir.c. */ + +/* Trace events for fs/ocfs2/namei.c. */ + +DECLARE_EVENT_CLASS(ocfs2__dentry_ops, + TP_PROTO(void *dir, void *dentry, int name_len, const char *name, + unsigned long long dir_blkno, unsigned long long extra), + TP_ARGS(dir, dentry, name_len, name, dir_blkno, extra), + TP_STRUCT__entry( + __field(void *, dir) + __field(void *, dentry) + __field(int, name_len) + __string(name, name) + __field(unsigned long long, dir_blkno) + __field(unsigned long long, extra) + ), + TP_fast_assign( + __entry->dir = dir; + __entry->dentry = dentry; + __entry->name_len = name_len; + __assign_str(name, name); + __entry->dir_blkno = dir_blkno; + __entry->extra = extra; + ), + TP_printk("%p %p %.*s %llu %llu", __entry->dir, __entry->dentry, + __entry->name_len, __get_str(name), + __entry->dir_blkno, __entry->extra) +); + +#define DEFINE_OCFS2_DENTRY_OPS(name) \ +DEFINE_EVENT(ocfs2__dentry_ops, name, \ +TP_PROTO(void *dir, void *dentry, int name_len, const char *name, \ + unsigned long long dir_blkno, unsigned long long extra), \ + TP_ARGS(dir, dentry, name_len, name, dir_blkno, extra)) + +DEFINE_OCFS2_DENTRY_OPS(ocfs2_lookup); + +DEFINE_OCFS2_DENTRY_OPS(ocfs2_mkdir); + +DEFINE_OCFS2_DENTRY_OPS(ocfs2_create); + +DEFINE_OCFS2_DENTRY_OPS(ocfs2_unlink); + +DEFINE_OCFS2_DENTRY_OPS(ocfs2_symlink_create); + +DEFINE_OCFS2_DENTRY_OPS(ocfs2_mv_orphaned_inode_to_new); + +DEFINE_OCFS2_POINTER_EVENT(ocfs2_lookup_ret); + +TRACE_EVENT(ocfs2_mknod, + TP_PROTO(void *dir, void *dentry, int name_len, const char *name, + unsigned long long dir_blkno, unsigned long dev, int mode), + TP_ARGS(dir, dentry, name_len, name, dir_blkno, dev, mode), + TP_STRUCT__entry( + __field(void *, dir) + __field(void *, dentry) + __field(int, name_len) + __string(name, name) + __field(unsigned long long, dir_blkno) + __field(unsigned long, dev) + __field(int, mode) + ), + TP_fast_assign( + __entry->dir = dir; + __entry->dentry = dentry; + __entry->name_len = name_len; + __assign_str(name, name); + __entry->dir_blkno = dir_blkno; + __entry->dev = dev; + __entry->mode = mode; + ), + TP_printk("%p %p %.*s %llu %lu %d", __entry->dir, __entry->dentry, + __entry->name_len, __get_str(name), + __entry->dir_blkno, __entry->dev, __entry->mode) +); + +TRACE_EVENT(ocfs2_link, + TP_PROTO(unsigned long long ino, int old_len, const char *old_name, + int name_len, const char *name), + TP_ARGS(ino, old_len, old_name, name_len, name), + TP_STRUCT__entry( + __field(unsigned long long, ino) + __field(int, old_len) + __string(old_name, old_name) + __field(int, name_len) + __string(name, name) + ), + TP_fast_assign( + __entry->ino = ino; + __entry->old_len = old_len; + __assign_str(old_name, old_name); + __entry->name_len = name_len; + __assign_str(name, name); + ), + TP_printk("%llu %.*s %.*s", __entry->ino, + __entry->old_len, __get_str(old_name), + __entry->name_len, __get_str(name)) +); + +DEFINE_OCFS2_ULL_ULL_UINT_EVENT(ocfs2_unlink_noent); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_double_lock); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_double_lock_end); + +TRACE_EVENT(ocfs2_rename, + TP_PROTO(void *old_dir, void *old_dentry, + void *new_dir, void *new_dentry, + int old_len, const char *old_name, + int new_len, const char *new_name), + TP_ARGS(old_dir, old_dentry, new_dir, new_dentry, + old_len, old_name, new_len, new_name), + TP_STRUCT__entry( + __field(void *, old_dir) + __field(void *, old_dentry) + __field(void *, new_dir) + __field(void *, new_dentry) + __field(int, old_len) + __string(old_name, old_name) + __field(int, new_len) + __string(new_name, new_name) + ), + TP_fast_assign( + __entry->old_dir = old_dir; + __entry->old_dentry = old_dentry; + __entry->new_dir = new_dir; + __entry->new_dentry = new_dentry; + __entry->old_len = old_len; + __assign_str(old_name, old_name); + __entry->new_len = new_len; + __assign_str(new_name, new_name); + ), + TP_printk("%p %p %p %p %.*s %.*s", + __entry->old_dir, __entry->old_dentry, + __entry->new_dir, __entry->new_dentry, + __entry->old_len, __get_str(old_name), + __entry->new_len, __get_str(new_name)) +); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_rename_not_permitted); + +TRACE_EVENT(ocfs2_rename_target_exists, + TP_PROTO(int new_len, const char *new_name), + TP_ARGS(new_len, new_name), + TP_STRUCT__entry( + __field(int, new_len) + __string(new_name, new_name) + ), + TP_fast_assign( + __entry->new_len = new_len; + __assign_str(new_name, new_name); + ), + TP_printk("%.*s", __entry->new_len, __get_str(new_name)) +); + +DEFINE_OCFS2_ULL_ULL_UINT_EVENT(ocfs2_rename_disagree); + +TRACE_EVENT(ocfs2_rename_over_existing, + TP_PROTO(unsigned long long new_blkno, void *new_bh, + unsigned long long newdi_blkno), + TP_ARGS(new_blkno, new_bh, newdi_blkno), + TP_STRUCT__entry( + __field(unsigned long long, new_blkno) + __field(void *, new_bh) + __field(unsigned long long, newdi_blkno) + ), + TP_fast_assign( + __entry->new_blkno = new_blkno; + __entry->new_bh = new_bh; + __entry->newdi_blkno = newdi_blkno; + ), + TP_printk("%llu %p %llu", __entry->new_blkno, __entry->new_bh, + __entry->newdi_blkno) +); + +DEFINE_OCFS2_ULL_ULL_UINT_EVENT(ocfs2_create_symlink_data); + +TRACE_EVENT(ocfs2_symlink_begin, + TP_PROTO(void *dir, void *dentry, const char *symname, + int len, const char *name), + TP_ARGS(dir, dentry, symname, len, name), + TP_STRUCT__entry( + __field(void *, dir) + __field(void *, dentry) + __field(const char *, symname) + __field(int, len) + __string(name, name) + ), + TP_fast_assign( + __entry->dir = dir; + __entry->dentry = dentry; + __entry->symname = symname; + __entry->len = len; + __assign_str(name, name); + ), + TP_printk("%p %p %s %.*s", __entry->dir, __entry->dentry, + __entry->symname, __entry->len, __get_str(name)) +); + +TRACE_EVENT(ocfs2_blkno_stringify, + TP_PROTO(unsigned long long blkno, const char *name, int namelen), + TP_ARGS(blkno, name, namelen), + TP_STRUCT__entry( + __field(unsigned long long, blkno) + __string(name, name) + __field(int, namelen) + ), + TP_fast_assign( + __entry->blkno = blkno; + __assign_str(name, name); + __entry->namelen = namelen; + ), + TP_printk("%llu %s %d", __entry->blkno, __get_str(name), + __entry->namelen) +); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_orphan_add_begin); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_orphan_add_end); + +TRACE_EVENT(ocfs2_orphan_del, + TP_PROTO(unsigned long long dir, const char *name, int namelen), + TP_ARGS(dir, name, namelen), + TP_STRUCT__entry( + __field(unsigned long long, dir) + __string(name, name) + __field(int, namelen) + ), + TP_fast_assign( + __entry->dir = dir; + __assign_str(name, name); + __entry->namelen = namelen; + ), + TP_printk("%llu %s %d", __entry->dir, __get_str(name), + __entry->namelen) +); + +/* End of trace events for fs/ocfs2/namei.c. */ + +/* Trace events for fs/ocfs2/dcache.c. */ + +TRACE_EVENT(ocfs2_dentry_revalidate, + TP_PROTO(void *dentry, int len, const char *name), + TP_ARGS(dentry, len, name), + TP_STRUCT__entry( + __field(void *, dentry) + __field(int, len) + __string(name, name) + ), + TP_fast_assign( + __entry->dentry = dentry; + __entry->len = len; + __assign_str(name, name); + ), + TP_printk("%p %.*s", __entry->dentry, __entry->len, __get_str(name)) +); + +TRACE_EVENT(ocfs2_dentry_revalidate_negative, + TP_PROTO(int len, const char *name, unsigned long pgen, + unsigned long gen), + TP_ARGS(len, name, pgen, gen), + TP_STRUCT__entry( + __field(int, len) + __string(name, name) + __field(unsigned long, pgen) + __field(unsigned long, gen) + ), + TP_fast_assign( + __entry->len = len; + __assign_str(name, name); + __entry->pgen = pgen; + __entry->gen = gen; + ), + TP_printk("%.*s %lu %lu", __entry->len, __get_str(name), + __entry->pgen, __entry->gen) +); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_dentry_revalidate_delete); + +DEFINE_OCFS2_ULL_INT_EVENT(ocfs2_dentry_revalidate_orphaned); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_dentry_revalidate_nofsdata); + +DEFINE_OCFS2_INT_EVENT(ocfs2_dentry_revalidate_ret); + +TRACE_EVENT(ocfs2_find_local_alias, + TP_PROTO(int len, const char *name), + TP_ARGS(len, name), + TP_STRUCT__entry( + __field(int, len) + __string(name, name) + ), + TP_fast_assign( + __entry->len = len; + __assign_str(name, name); + ), + TP_printk("%.*s", __entry->len, __get_str(name)) +); + +TRACE_EVENT(ocfs2_dentry_attach_lock, + TP_PROTO(int len, const char *name, + unsigned long long parent, void *fsdata), + TP_ARGS(len, name, parent, fsdata), + TP_STRUCT__entry( + __field(int, len) + __string(name, name) + __field(unsigned long long, parent) + __field(void *, fsdata) + ), + TP_fast_assign( + __entry->len = len; + __assign_str(name, name); + __entry->parent = parent; + __entry->fsdata = fsdata; + ), + TP_printk("%.*s %llu %p", __entry->len, __get_str(name), + __entry->parent, __entry->fsdata) +); + +TRACE_EVENT(ocfs2_dentry_attach_lock_found, + TP_PROTO(const char *name, unsigned long long parent, + unsigned long long ino), + TP_ARGS(name, parent, ino), + TP_STRUCT__entry( + __string(name, name) + __field(unsigned long long, parent) + __field(unsigned long long, ino) + ), + TP_fast_assign( + __assign_str(name, name); + __entry->parent = parent; + __entry->ino = ino; + ), + TP_printk("%s %llu %llu", __get_str(name), __entry->parent, __entry->ino) +); +/* End of trace events for fs/ocfs2/dcache.c. */ + +/* Trace events for fs/ocfs2/export.c. */ + +TRACE_EVENT(ocfs2_get_dentry_begin, + TP_PROTO(void *sb, void *handle, unsigned long long blkno), + TP_ARGS(sb, handle, blkno), + TP_STRUCT__entry( + __field(void *, sb) + __field(void *, handle) + __field(unsigned long long, blkno) + ), + TP_fast_assign( + __entry->sb = sb; + __entry->handle = handle; + __entry->blkno = blkno; + ), + TP_printk("%p %p %llu", __entry->sb, __entry->handle, __entry->blkno) +); + +DEFINE_OCFS2_INT_INT_EVENT(ocfs2_get_dentry_test_bit); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_get_dentry_stale); + +DEFINE_OCFS2_ULL_UINT_UINT_EVENT(ocfs2_get_dentry_generation); + +DEFINE_OCFS2_POINTER_EVENT(ocfs2_get_dentry_end); + +TRACE_EVENT(ocfs2_get_parent, + TP_PROTO(void *child, int len, const char *name, + unsigned long long ino), + TP_ARGS(child, len, name, ino), + TP_STRUCT__entry( + __field(void *, child) + __field(int, len) + __string(name, name) + __field(unsigned long long, ino) + ), + TP_fast_assign( + __entry->child = child; + __entry->len = len; + __assign_str(name, name); + __entry->ino = ino; + ), + TP_printk("%p %.*s %llu", __entry->child, __entry->len, + __get_str(name), __entry->ino) +); + +DEFINE_OCFS2_POINTER_EVENT(ocfs2_get_parent_end); + +TRACE_EVENT(ocfs2_encode_fh_begin, + TP_PROTO(void *dentry, int name_len, const char *name, + void *fh, int len, int connectable), + TP_ARGS(dentry, name_len, name, fh, len, connectable), + TP_STRUCT__entry( + __field(void *, dentry) + __field(int, name_len) + __string(name, name) + __field(void *, fh) + __field(int, len) + __field(int, connectable) + ), + TP_fast_assign( + __entry->dentry = dentry; + __entry->name_len = name_len; + __assign_str(name, name); + __entry->fh = fh; + __entry->len = len; + __entry->connectable = connectable; + ), + TP_printk("%p %.*s %p %d %d", __entry->dentry, __entry->name_len, + __get_str(name), __entry->fh, __entry->len, + __entry->connectable) +); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_encode_fh_self); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_encode_fh_parent); + +DEFINE_OCFS2_INT_EVENT(ocfs2_encode_fh_type); + +/* End of trace events for fs/ocfs2/export.c. */ + +/* Trace events for fs/ocfs2/journal.c. */ + +DEFINE_OCFS2_UINT_EVENT(ocfs2_commit_cache_begin); + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_commit_cache_end); + +DEFINE_OCFS2_INT_INT_EVENT(ocfs2_extend_trans); + +DEFINE_OCFS2_INT_EVENT(ocfs2_extend_trans_restart); + +DEFINE_OCFS2_INT_INT_EVENT(ocfs2_allocate_extend_trans); + +DEFINE_OCFS2_ULL_ULL_UINT_UINT_EVENT(ocfs2_journal_access); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_journal_dirty); + +DEFINE_OCFS2_ULL_ULL_UINT_EVENT(ocfs2_journal_init); + +DEFINE_OCFS2_UINT_EVENT(ocfs2_journal_init_maxlen); + +DEFINE_OCFS2_INT_EVENT(ocfs2_journal_shutdown); + +DEFINE_OCFS2_POINTER_EVENT(ocfs2_journal_shutdown_wait); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_complete_recovery); + +DEFINE_OCFS2_INT_EVENT(ocfs2_complete_recovery_end); + +TRACE_EVENT(ocfs2_complete_recovery_slot, + TP_PROTO(int slot, unsigned long long la_ino, + unsigned long long tl_ino, void *qrec), + TP_ARGS(slot, la_ino, tl_ino, qrec), + TP_STRUCT__entry( + __field(int, slot) + __field(unsigned long long, la_ino) + __field(unsigned long long, tl_ino) + __field(void *, qrec) + ), + TP_fast_assign( + __entry->slot = slot; + __entry->la_ino = la_ino; + __entry->tl_ino = tl_ino; + __entry->qrec = qrec; + ), + TP_printk("%d %llu %llu %p", __entry->slot, __entry->la_ino, + __entry->tl_ino, __entry->qrec) +); + +DEFINE_OCFS2_INT_INT_EVENT(ocfs2_recovery_thread_node); + +DEFINE_OCFS2_INT_EVENT(ocfs2_recovery_thread_end); + +TRACE_EVENT(ocfs2_recovery_thread, + TP_PROTO(int node_num, int osb_node_num, int disable, + void *recovery_thread, int map_set), + TP_ARGS(node_num, osb_node_num, disable, recovery_thread, map_set), + TP_STRUCT__entry( + __field(int, node_num) + __field(int, osb_node_num) + __field(int,disable) + __field(void *, recovery_thread) + __field(int,map_set) + ), + TP_fast_assign( + __entry->node_num = node_num; + __entry->osb_node_num = osb_node_num; + __entry->disable = disable; + __entry->recovery_thread = recovery_thread; + __entry->map_set = map_set; + ), + TP_printk("%d %d %d %p %d", __entry->node_num, + __entry->osb_node_num, __entry->disable, + __entry->recovery_thread, __entry->map_set) +); + +DEFINE_OCFS2_UINT_UINT_UINT_EVENT(ocfs2_replay_journal_recovered); + +DEFINE_OCFS2_INT_EVENT(ocfs2_replay_journal_lock_err); + +DEFINE_OCFS2_INT_EVENT(ocfs2_replay_journal_skip); + +DEFINE_OCFS2_UINT_UINT_UINT_EVENT(ocfs2_recover_node); + +DEFINE_OCFS2_UINT_UINT_EVENT(ocfs2_recover_node_skip); + +DEFINE_OCFS2_UINT_UINT_EVENT(ocfs2_mark_dead_nodes); + +DEFINE_OCFS2_UINT_UINT_UINT_EVENT(ocfs2_queue_orphan_scan_begin); + +DEFINE_OCFS2_UINT_UINT_UINT_EVENT(ocfs2_queue_orphan_scan_end); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_orphan_filldir); + +DEFINE_OCFS2_INT_EVENT(ocfs2_recover_orphans); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_recover_orphans_iput); + +DEFINE_OCFS2_INT_EVENT(ocfs2_wait_on_mount); + +/* End of trace events for fs/ocfs2/journal.c. */ + +/* Trace events for fs/ocfs2/buffer_head_io.c. */ + +DEFINE_OCFS2_ULL_UINT_EVENT(ocfs2_read_blocks_sync); + +DEFINE_OCFS2_ULL_EVENT(ocfs2_read_blocks_sync_jbd); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_read_blocks_from_disk); + +DEFINE_OCFS2_ULL_INT_INT_INT_EVENT(ocfs2_read_blocks_bh); + +DEFINE_OCFS2_ULL_INT_INT_INT_EVENT(ocfs2_read_blocks_end); + +TRACE_EVENT(ocfs2_write_block, + TP_PROTO(unsigned long long block, void *ci), + TP_ARGS(block, ci), + TP_STRUCT__entry( + __field(unsigned long long, block) + __field(void *, ci) + ), + TP_fast_assign( + __entry->block = block; + __entry->ci = ci; + ), + TP_printk("%llu %p", __entry->block, __entry->ci) +); + +TRACE_EVENT(ocfs2_read_blocks_begin, + TP_PROTO(void *ci, unsigned long long block, + unsigned int nr, int flags), + TP_ARGS(ci, block, nr, flags), + TP_STRUCT__entry( + __field(void *, ci) + __field(unsigned long long, block) + __field(unsigned int, nr) + __field(int, flags) + ), + TP_fast_assign( + __entry->ci = ci; + __entry->block = block; + __entry->nr = nr; + __entry->flags = flags; + ), + TP_printk("%p %llu %u %d", __entry->ci, __entry->block, + __entry->nr, __entry->flags) +); + +/* End of trace events for fs/ocfs2/buffer_head_io.c. */ + +/* Trace events for fs/ocfs2/uptodate.c. */ + +DEFINE_OCFS2_ULL_EVENT(ocfs2_purge_copied_metadata_tree); + +DEFINE_OCFS2_ULL_UINT_UINT_EVENT(ocfs2_metadata_cache_purge); + +DEFINE_OCFS2_ULL_ULL_UINT_EVENT(ocfs2_buffer_cached_begin); + +TRACE_EVENT(ocfs2_buffer_cached_end, + TP_PROTO(int index, void *item), + TP_ARGS(index, item), + TP_STRUCT__entry( + __field(int, index) + __field(void *, item) + ), + TP_fast_assign( + __entry->index = index; + __entry->item = item; + ), + TP_printk("%d %p", __entry->index, __entry->item) +); + +DEFINE_OCFS2_ULL_ULL_UINT_EVENT(ocfs2_append_cache_array); + +DEFINE_OCFS2_ULL_ULL_UINT_EVENT(ocfs2_insert_cache_tree); + +DEFINE_OCFS2_ULL_UINT_UINT_EVENT(ocfs2_expand_cache); + +DEFINE_OCFS2_ULL_UINT_UINT_EVENT(ocfs2_set_buffer_uptodate); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_set_buffer_uptodate_begin); + +DEFINE_OCFS2_ULL_UINT_UINT_EVENT(ocfs2_remove_metadata_array); + +DEFINE_OCFS2_ULL_ULL_EVENT(ocfs2_remove_metadata_tree); + +DEFINE_OCFS2_ULL_ULL_UINT_UINT_EVENT(ocfs2_remove_block_from_cache); + +/* End of trace events for fs/ocfs2/uptodate.c. */ +#endif /* _TRACE_OCFS2_H */ + +/* This part must be outside protection */ +#undef TRACE_INCLUDE_PATH +#define TRACE_INCLUDE_PATH . +#define TRACE_INCLUDE_FILE ocfs2_trace +#include <trace/define_trace.h> diff --git a/fs/ocfs2/quota.h b/fs/ocfs2/quota.h index 7365e2e0870..f266d67df3c 100644 --- a/fs/ocfs2/quota.h +++ b/fs/ocfs2/quota.h @@ -17,20 +17,18 @@ #include "ocfs2.h" -/* Common stuff */ -/* id number of quota format */ -#define QFMT_OCFS2 3 - /* * In-memory structures */ struct ocfs2_dquot { struct dquot dq_dquot; /* Generic VFS dquot */ loff_t dq_local_off; /* Offset in the local quota file */ + u64 dq_local_phys_blk; /* Physical block carrying quota structure */ struct ocfs2_quota_chunk *dq_chunk; /* Chunk dquot is in */ unsigned int dq_use_count; /* Number of nodes having reference to this entry in global quota file */ s64 dq_origspace; /* Last globally synced space usage */ s64 dq_originodes; /* Last globally synced inode usage */ + struct llist_node list; /* Member of list of dquots to drop */ }; /* Description of one chunk to recover in memory */ @@ -50,14 +48,14 @@ struct ocfs2_mem_dqinfo { unsigned int dqi_chunks; /* Number of chunks in local quota file */ unsigned int dqi_blocks; /* Number of blocks allocated for local quota file */ unsigned int dqi_syncms; /* How often should we sync with other nodes */ - unsigned int dqi_syncjiff; /* Precomputed dqi_syncms in jiffies */ struct list_head dqi_chunk; /* List of chunks */ struct inode *dqi_gqinode; /* Global quota file inode */ struct ocfs2_lock_res dqi_gqlock; /* Lock protecting quota information structure */ struct buffer_head *dqi_gqi_bh; /* Buffer head with global quota file inode - set only if inode lock is obtained */ int dqi_gqi_count; /* Number of holders of dqi_gqi_bh */ + u64 dqi_giblk; /* Number of block with global information header */ struct buffer_head *dqi_lqi_bh; /* Buffer head with local quota file inode */ - struct buffer_head *dqi_ibh; /* Buffer with information header */ + struct buffer_head *dqi_libh; /* Buffer with local information header */ struct qtree_mem_dqinfo dqi_gi; /* Info about global file */ struct delayed_work dqi_sync_work; /* Work for syncing dquots */ struct ocfs2_quota_recovery *dqi_rec; /* Pointer to recovery @@ -107,13 +105,15 @@ static inline int ocfs2_global_release_dquot(struct dquot *dquot) int ocfs2_lock_global_qf(struct ocfs2_mem_dqinfo *oinfo, int ex); void ocfs2_unlock_global_qf(struct ocfs2_mem_dqinfo *oinfo, int ex); -int ocfs2_read_quota_block(struct inode *inode, u64 v_block, - struct buffer_head **bh); - -extern struct dquot_operations ocfs2_quota_operations; +int ocfs2_validate_quota_block(struct super_block *sb, struct buffer_head *bh); +int ocfs2_read_quota_phys_block(struct inode *inode, u64 p_block, + struct buffer_head **bh); +int ocfs2_create_local_dquot(struct dquot *dquot); +int ocfs2_local_release_dquot(handle_t *handle, struct dquot *dquot); +int ocfs2_local_write_dquot(struct dquot *dquot); +void ocfs2_drop_dquot_refs(struct work_struct *work); + +extern const struct dquot_operations ocfs2_quota_operations; extern struct quota_format_type ocfs2_quota_format; -int ocfs2_quota_setup(void); -void ocfs2_quota_shutdown(void); - #endif /* _OCFS2_QUOTA_H */ diff --git a/fs/ocfs2/quota_global.c b/fs/ocfs2/quota_global.c index 1ed0f7c8686..b990a62cff5 100644 --- a/fs/ocfs2/quota_global.c +++ b/fs/ocfs2/quota_global.c @@ -3,14 +3,15 @@ */ #include <linux/spinlock.h> #include <linux/fs.h> +#include <linux/slab.h> #include <linux/quota.h> #include <linux/quotaops.h> #include <linux/dqblk_qtree.h> #include <linux/jiffies.h> #include <linux/writeback.h> #include <linux/workqueue.h> +#include <linux/llist.h> -#define MLOG_MASK_PREFIX ML_QUOTA #include <cluster/masklog.h> #include "ocfs2_fs.h" @@ -23,9 +24,45 @@ #include "sysfile.h" #include "dlmglue.h" #include "uptodate.h" +#include "super.h" +#include "buffer_head_io.h" #include "quota.h" +#include "ocfs2_trace.h" -static struct workqueue_struct *ocfs2_quota_wq = NULL; +/* + * Locking of quotas with OCFS2 is rather complex. Here are rules that + * should be obeyed by all the functions: + * - any write of quota structure (either to local or global file) is protected + * by dqio_mutex or dquot->dq_lock. + * - any modification of global quota file holds inode cluster lock, i_mutex, + * and ip_alloc_sem of the global quota file (achieved by + * ocfs2_lock_global_qf). It also has to hold qinfo_lock. + * - an allocation of new blocks for local quota file is protected by + * its ip_alloc_sem + * + * A rough sketch of locking dependencies (lf = local file, gf = global file): + * Normal filesystem operation: + * start_trans -> dqio_mutex -> write to lf + * Syncing of local and global file: + * ocfs2_lock_global_qf -> start_trans -> dqio_mutex -> qinfo_lock -> + * write to gf + * -> write to lf + * Acquire dquot for the first time: + * dq_lock -> ocfs2_lock_global_qf -> qinfo_lock -> read from gf + * -> alloc space for gf + * -> start_trans -> qinfo_lock -> write to gf + * -> ip_alloc_sem of lf -> alloc space for lf + * -> write to lf + * Release last reference to dquot: + * dq_lock -> ocfs2_lock_global_qf -> start_trans -> qinfo_lock -> write to gf + * -> write to lf + * Note that all the above operations also hold the inode cluster lock of lf. + * Recovery: + * inode cluster lock of recovered lf + * -> read bitmaps -> ip_alloc_sem of lf + * -> ocfs2_lock_global_qf -> start_trans -> dqio_mutex -> qinfo_lock -> + * write to gf + */ static void qsync_work_fn(struct work_struct *work); @@ -59,7 +96,7 @@ static void ocfs2_global_mem2diskdqb(void *dp, struct dquot *dquot) struct ocfs2_global_disk_dqblk *d = dp; struct mem_dqblk *m = &dquot->dq_dqb; - d->dqb_id = cpu_to_le32(dquot->dq_id); + d->dqb_id = cpu_to_le32(from_kqid(&init_user_ns, dquot->dq_id)); d->dqb_use_count = cpu_to_le32(OCFS2_DQUOT(dquot)->dq_use_count); d->dqb_ihardlimit = cpu_to_le64(m->dqb_ihardlimit); d->dqb_isoftlimit = cpu_to_le64(m->dqb_isoftlimit); @@ -69,17 +106,21 @@ static void ocfs2_global_mem2diskdqb(void *dp, struct dquot *dquot) d->dqb_curspace = cpu_to_le64(m->dqb_curspace); d->dqb_btime = cpu_to_le64(m->dqb_btime); d->dqb_itime = cpu_to_le64(m->dqb_itime); + d->dqb_pad1 = d->dqb_pad2 = 0; } static int ocfs2_global_is_id(void *dp, struct dquot *dquot) { struct ocfs2_global_disk_dqblk *d = dp; struct ocfs2_mem_dqinfo *oinfo = - sb_dqinfo(dquot->dq_sb, dquot->dq_type)->dqi_priv; + sb_dqinfo(dquot->dq_sb, dquot->dq_id.type)->dqi_priv; if (qtree_entry_unused(&oinfo->dqi_gi, dp)) return 0; - return le32_to_cpu(d->dqb_id) == dquot->dq_id; + + return qid_eq(make_kqid(&init_user_ns, dquot->dq_id.type, + le32_to_cpu(d->dqb_id)), + dquot->dq_id); } struct qtree_fmt_operations ocfs2_global_ops = { @@ -88,14 +129,12 @@ struct qtree_fmt_operations ocfs2_global_ops = { .is_id = ocfs2_global_is_id, }; -static int ocfs2_validate_quota_block(struct super_block *sb, - struct buffer_head *bh) +int ocfs2_validate_quota_block(struct super_block *sb, struct buffer_head *bh) { struct ocfs2_disk_dqtrailer *dqt = ocfs2_block_dqtrailer(sb->s_blocksize, bh->b_data); - mlog(0, "Validating quota block %llu\n", - (unsigned long long)bh->b_blocknr); + trace_ocfs2_validate_quota_block((unsigned long long)bh->b_blocknr); BUG_ON(!buffer_uptodate(bh)); @@ -107,45 +146,19 @@ static int ocfs2_validate_quota_block(struct super_block *sb, return ocfs2_validate_meta_ecc(sb, bh->b_data, &dqt->dq_check); } -int ocfs2_read_quota_block(struct inode *inode, u64 v_block, - struct buffer_head **bh) +int ocfs2_read_quota_phys_block(struct inode *inode, u64 p_block, + struct buffer_head **bhp) { - int rc = 0; - struct buffer_head *tmp = *bh; + int rc; - rc = ocfs2_read_virt_blocks(inode, v_block, 1, &tmp, 0, - ocfs2_validate_quota_block); + *bhp = NULL; + rc = ocfs2_read_blocks(INODE_CACHE(inode), p_block, 1, bhp, 0, + ocfs2_validate_quota_block); if (rc) mlog_errno(rc); - - /* If ocfs2_read_virt_blocks() got us a new bh, pass it up. */ - if (!rc && !*bh) - *bh = tmp; - return rc; } -static int ocfs2_get_quota_block(struct inode *inode, int block, - struct buffer_head **bh) -{ - u64 pblock, pcount; - int err; - - down_read(&OCFS2_I(inode)->ip_alloc_sem); - err = ocfs2_extent_map_get_blocks(inode, block, &pblock, &pcount, NULL); - up_read(&OCFS2_I(inode)->ip_alloc_sem); - if (err) { - mlog_errno(err); - return err; - } - *bh = sb_getblk(inode->i_sb, pblock); - if (!*bh) { - err = -EIO; - mlog_errno(err); - } - return err;; -} - /* Read data from global quotafile - avoid pagecache and such because we cannot * afford acquiring the locks... We use quota cluster lock to serialize * operations. Caller is responsible for acquiring it. */ @@ -160,6 +173,7 @@ ssize_t ocfs2_quota_read(struct super_block *sb, int type, char *data, int err = 0; struct buffer_head *bh; size_t toread, tocopy; + u64 pblock = 0, pcount = 0; if (off > i_size) return 0; @@ -168,8 +182,19 @@ ssize_t ocfs2_quota_read(struct super_block *sb, int type, char *data, toread = len; while (toread > 0) { tocopy = min_t(size_t, (sb->s_blocksize - offset), toread); + if (!pcount) { + err = ocfs2_extent_map_get_blocks(gqinode, blk, &pblock, + &pcount, NULL); + if (err) { + mlog_errno(err); + return err; + } + } else { + pcount--; + pblock++; + } bh = NULL; - err = ocfs2_read_quota_block(gqinode, blk, &bh); + err = ocfs2_read_quota_phys_block(gqinode, pblock, &bh); if (err) { mlog_errno(err); return err; @@ -197,6 +222,7 @@ ssize_t ocfs2_quota_write(struct super_block *sb, int type, int err = 0, new = 0, ja_type; struct buffer_head *bh = NULL; handle_t *handle = journal_current_handle(); + u64 pblock, pcount; if (!handle) { mlog(ML_ERROR, "Quota write (off=%llu, len=%llu) cancelled " @@ -209,32 +235,37 @@ ssize_t ocfs2_quota_write(struct super_block *sb, int type, len = sb->s_blocksize - OCFS2_QBLK_RESERVED_SPACE - offset; } - mutex_lock_nested(&gqinode->i_mutex, I_MUTEX_QUOTA); - if (gqinode->i_size < off + len) { - down_write(&OCFS2_I(gqinode)->ip_alloc_sem); - err = ocfs2_extend_no_holes(gqinode, off + len, off); - up_write(&OCFS2_I(gqinode)->ip_alloc_sem); - if (err < 0) - goto out; + if (i_size_read(gqinode) < off + len) { + loff_t rounded_end = + ocfs2_align_bytes_to_blocks(sb, off + len); + + /* Space is already allocated in ocfs2_acquire_dquot() */ err = ocfs2_simple_size_update(gqinode, oinfo->dqi_gqi_bh, - off + len); + rounded_end); if (err < 0) goto out; new = 1; } + err = ocfs2_extent_map_get_blocks(gqinode, blk, &pblock, &pcount, NULL); + if (err) { + mlog_errno(err); + goto out; + } /* Not rewriting whole block? */ if ((offset || len < sb->s_blocksize - OCFS2_QBLK_RESERVED_SPACE) && !new) { - err = ocfs2_read_quota_block(gqinode, blk, &bh); + err = ocfs2_read_quota_phys_block(gqinode, pblock, &bh); ja_type = OCFS2_JOURNAL_ACCESS_WRITE; } else { - err = ocfs2_get_quota_block(gqinode, blk, &bh); + bh = sb_getblk(sb, pblock); + if (!bh) + err = -ENOMEM; ja_type = OCFS2_JOURNAL_ACCESS_CREATE; } if (err) { mlog_errno(err); - return err; + goto out; } lock_buffer(bh); if (new) @@ -243,25 +274,22 @@ ssize_t ocfs2_quota_write(struct super_block *sb, int type, flush_dcache_page(bh->b_page); set_buffer_uptodate(bh); unlock_buffer(bh); - ocfs2_set_buffer_uptodate(gqinode, bh); - err = ocfs2_journal_access_dq(handle, gqinode, bh, ja_type); + ocfs2_set_buffer_uptodate(INODE_CACHE(gqinode), bh); + err = ocfs2_journal_access_dq(handle, INODE_CACHE(gqinode), bh, + ja_type); if (err < 0) { brelse(bh); goto out; } - err = ocfs2_journal_dirty(handle, bh); + ocfs2_journal_dirty(handle, bh); brelse(bh); - if (err < 0) - goto out; out: if (err) { - mutex_unlock(&gqinode->i_mutex); mlog_errno(err); return err; } gqinode->i_version++; ocfs2_mark_inode_dirty(handle, gqinode, oinfo->dqi_gqi_bh); - mutex_unlock(&gqinode->i_mutex); return len; } @@ -279,11 +307,23 @@ int ocfs2_lock_global_qf(struct ocfs2_mem_dqinfo *oinfo, int ex) else WARN_ON(bh != oinfo->dqi_gqi_bh); spin_unlock(&dq_data_lock); + if (ex) { + mutex_lock(&oinfo->dqi_gqinode->i_mutex); + down_write(&OCFS2_I(oinfo->dqi_gqinode)->ip_alloc_sem); + } else { + down_read(&OCFS2_I(oinfo->dqi_gqinode)->ip_alloc_sem); + } return 0; } void ocfs2_unlock_global_qf(struct ocfs2_mem_dqinfo *oinfo, int ex) { + if (ex) { + up_write(&OCFS2_I(oinfo->dqi_gqinode)->ip_alloc_sem); + mutex_unlock(&oinfo->dqi_gqinode->i_mutex); + } else { + up_read(&OCFS2_I(oinfo->dqi_gqinode)->ip_alloc_sem); + } ocfs2_inode_unlock(oinfo->dqi_gqinode, ex); brelse(oinfo->dqi_gqi_bh); spin_lock(&dq_data_lock); @@ -301,10 +341,9 @@ int ocfs2_global_read_info(struct super_block *sb, int type) struct ocfs2_global_disk_dqinfo dinfo; struct mem_dqinfo *info = sb_dqinfo(sb, type); struct ocfs2_mem_dqinfo *oinfo = info->dqi_priv; + u64 pcount; int status; - mlog_entry_void(); - /* Read global header */ gqinode = ocfs2_get_system_file_inode(OCFS2_SB(sb), ino[type], OCFS2_INVALID_SLOT); @@ -327,9 +366,19 @@ int ocfs2_global_read_info(struct super_block *sb, int type) mlog_errno(status); goto out_err; } + + status = ocfs2_extent_map_get_blocks(gqinode, 0, &oinfo->dqi_giblk, + &pcount, NULL); + if (status < 0) + goto out_unlock; + + status = ocfs2_qinfo_lock(oinfo, 0); + if (status < 0) + goto out_unlock; status = sb->s_op->quota_read(sb, type, (char *)&dinfo, sizeof(struct ocfs2_global_disk_dqinfo), OCFS2_GLOBAL_INFO_OFF); + ocfs2_qinfo_unlock(oinfo, 0); ocfs2_unlock_global_qf(oinfo, 0); if (status != sizeof(struct ocfs2_global_disk_dqinfo)) { mlog(ML_ERROR, "Cannot read global quota info (%d).\n", @@ -342,7 +391,6 @@ int ocfs2_global_read_info(struct super_block *sb, int type) info->dqi_bgrace = le32_to_cpu(dinfo.dqi_bgrace); info->dqi_igrace = le32_to_cpu(dinfo.dqi_igrace); oinfo->dqi_syncms = le32_to_cpu(dinfo.dqi_syncms); - oinfo->dqi_syncjiff = msecs_to_jiffies(oinfo->dqi_syncms); oinfo->dqi_gi.dqi_blocks = le32_to_cpu(dinfo.dqi_blocks); oinfo->dqi_gi.dqi_free_blk = le32_to_cpu(dinfo.dqi_free_blk); oinfo->dqi_gi.dqi_free_entry = le32_to_cpu(dinfo.dqi_free_entry); @@ -351,12 +399,15 @@ int ocfs2_global_read_info(struct super_block *sb, int type) OCFS2_QBLK_RESERVED_SPACE; oinfo->dqi_gi.dqi_qtree_depth = qtree_depth(&oinfo->dqi_gi); INIT_DELAYED_WORK(&oinfo->dqi_sync_work, qsync_work_fn); - queue_delayed_work(ocfs2_quota_wq, &oinfo->dqi_sync_work, - oinfo->dqi_syncjiff); + schedule_delayed_work(&oinfo->dqi_sync_work, + msecs_to_jiffies(oinfo->dqi_syncms)); out_err: - mlog_exit(status); return status; +out_unlock: + ocfs2_unlock_global_qf(oinfo, 0); + mlog_errno(status); + goto out_err; } /* Write information to global quota file. Expects exlusive lock on quota @@ -402,44 +453,23 @@ int ocfs2_global_write_info(struct super_block *sb, int type) return err; } -/* Read in information from global quota file and acquire a reference to it. - * dquot_acquire() has already started the transaction and locked quota file */ -int ocfs2_global_read_dquot(struct dquot *dquot) +static int ocfs2_global_qinit_alloc(struct super_block *sb, int type) { - int err, err2, ex = 0; - struct ocfs2_mem_dqinfo *info = - sb_dqinfo(dquot->dq_sb, dquot->dq_type)->dqi_priv; + struct ocfs2_mem_dqinfo *oinfo = sb_dqinfo(sb, type)->dqi_priv; - err = ocfs2_qinfo_lock(info, 0); - if (err < 0) - goto out; - err = qtree_read_dquot(&info->dqi_gi, dquot); - if (err < 0) - goto out_qlock; - OCFS2_DQUOT(dquot)->dq_use_count++; - OCFS2_DQUOT(dquot)->dq_origspace = dquot->dq_dqb.dqb_curspace; - OCFS2_DQUOT(dquot)->dq_originodes = dquot->dq_dqb.dqb_curinodes; - if (!dquot->dq_off) { /* No real quota entry? */ - /* Upgrade to exclusive lock for allocation */ - err = ocfs2_qinfo_lock(info, 1); - if (err < 0) - goto out_qlock; - ex = 1; - } - err = qtree_write_dquot(&info->dqi_gi, dquot); - if (ex && info_dirty(sb_dqinfo(dquot->dq_sb, dquot->dq_type))) { - err2 = __ocfs2_global_write_info(dquot->dq_sb, dquot->dq_type); - if (!err) - err = err2; - } -out_qlock: - if (ex) - ocfs2_qinfo_unlock(info, 1); - ocfs2_qinfo_unlock(info, 0); -out: - if (err < 0) - mlog_errno(err); - return err; + /* + * We may need to allocate tree blocks and a leaf block but not the + * root block + */ + return oinfo->dqi_gi.dqi_qtree_depth; +} + +static int ocfs2_calc_global_qinit_credits(struct super_block *sb, int type) +{ + /* We modify all the allocated blocks, tree root, info block and + * the inode */ + return (ocfs2_global_qinit_alloc(sb, type) + 2) * + OCFS2_QUOTA_BLOCK_UPDATE_CREDITS + 1; } /* Sync local information about quota modifications with global quota file. @@ -449,7 +479,7 @@ int __ocfs2_sync_dquot(struct dquot *dquot, int freeing) { int err, err2; struct super_block *sb = dquot->dq_sb; - int type = dquot->dq_type; + int type = dquot->dq_id.type; struct ocfs2_mem_dqinfo *info = sb_dqinfo(sb, type)->dqi_priv; struct ocfs2_global_disk_dqblk dqblk; s64 spacechange, inodechange; @@ -478,9 +508,11 @@ int __ocfs2_sync_dquot(struct dquot *dquot, int freeing) olditime = dquot->dq_dqb.dqb_itime; oldbtime = dquot->dq_dqb.dqb_btime; ocfs2_global_disk2memdqb(dquot, &dqblk); - mlog(0, "Syncing global dquot %u space %lld+%lld, inodes %lld+%lld\n", - dquot->dq_id, dquot->dq_dqb.dqb_curspace, (long long)spacechange, - dquot->dq_dqb.dqb_curinodes, (long long)inodechange); + trace_ocfs2_sync_dquot(from_kqid(&init_user_ns, dquot->dq_id), + dquot->dq_dqb.dqb_curspace, + (long long)spacechange, + dquot->dq_dqb.dqb_curinodes, + (long long)inodechange); if (!test_bit(DQ_LASTSET_B + QIF_SPACE_B, &dquot->dq_flags)) dquot->dq_dqb.dqb_curspace += spacechange; if (!test_bit(DQ_LASTSET_B + QIF_INODES_B, &dquot->dq_flags)) @@ -527,9 +559,9 @@ int __ocfs2_sync_dquot(struct dquot *dquot, int freeing) spin_unlock(&dq_data_lock); err = ocfs2_qinfo_lock(info, freeing); if (err < 0) { - mlog(ML_ERROR, "Failed to lock quota info, loosing quota write" - " (type=%d, id=%u)\n", dquot->dq_type, - (unsigned)dquot->dq_id); + mlog(ML_ERROR, "Failed to lock quota info, losing quota write" + " (type=%d, id=%u)\n", dquot->dq_id.type, + (unsigned)from_kqid(&init_user_ns, dquot->dq_id)); goto out; } if (freeing) @@ -564,9 +596,10 @@ static int ocfs2_sync_dquot_helper(struct dquot *dquot, unsigned long type) struct ocfs2_super *osb = OCFS2_SB(sb); int status = 0; - mlog_entry("id=%u qtype=%u type=%lu device=%s\n", dquot->dq_id, - dquot->dq_type, type, sb->s_id); - if (type != dquot->dq_type) + trace_ocfs2_sync_dquot_helper(from_kqid(&init_user_ns, dquot->dq_id), + dquot->dq_id.type, + type, sb->s_id); + if (type != dquot->dq_id.type) goto out; status = ocfs2_lock_global_qf(oinfo, 1); if (status < 0) @@ -580,19 +613,17 @@ static int ocfs2_sync_dquot_helper(struct dquot *dquot, unsigned long type) } mutex_lock(&sb_dqopt(sb)->dqio_mutex); status = ocfs2_sync_dquot(dquot); - mutex_unlock(&sb_dqopt(sb)->dqio_mutex); if (status < 0) mlog_errno(status); /* We have to write local structure as well... */ - dquot_mark_dquot_dirty(dquot); - status = dquot_commit(dquot); + status = ocfs2_local_write_dquot(dquot); if (status < 0) mlog_errno(status); + mutex_unlock(&sb_dqopt(sb)->dqio_mutex); ocfs2_commit_trans(osb, handle); out_ilock: ocfs2_unlock_global_qf(oinfo, 1); out: - mlog_exit(status); return status; } @@ -604,8 +635,8 @@ static void qsync_work_fn(struct work_struct *work) struct super_block *sb = oinfo->dqi_gqinode->i_sb; dquot_scan_active(sb, ocfs2_sync_dquot_helper, oinfo->dqi_type); - queue_delayed_work(ocfs2_quota_wq, &oinfo->dqi_sync_work, - oinfo->dqi_syncjiff); + schedule_delayed_work(&oinfo->dqi_sync_work, + msecs_to_jiffies(oinfo->dqi_syncms)); } /* @@ -618,7 +649,8 @@ static int ocfs2_write_dquot(struct dquot *dquot) struct ocfs2_super *osb = OCFS2_SB(dquot->dq_sb); int status = 0; - mlog_entry("id=%u, type=%d", dquot->dq_id, dquot->dq_type); + trace_ocfs2_write_dquot(from_kqid(&init_user_ns, dquot->dq_id), + dquot->dq_id.type); handle = ocfs2_start_trans(osb, OCFS2_QWRITE_CREDITS); if (IS_ERR(handle)) { @@ -626,106 +658,205 @@ static int ocfs2_write_dquot(struct dquot *dquot) mlog_errno(status); goto out; } - status = dquot_commit(dquot); + mutex_lock(&sb_dqopt(dquot->dq_sb)->dqio_mutex); + status = ocfs2_local_write_dquot(dquot); + mutex_unlock(&sb_dqopt(dquot->dq_sb)->dqio_mutex); ocfs2_commit_trans(osb, handle); out: - mlog_exit(status); return status; } -int ocfs2_calc_qdel_credits(struct super_block *sb, int type) +static int ocfs2_calc_qdel_credits(struct super_block *sb, int type) { - struct ocfs2_mem_dqinfo *oinfo; - int features[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA, - OCFS2_FEATURE_RO_COMPAT_GRPQUOTA }; + struct ocfs2_mem_dqinfo *oinfo = sb_dqinfo(sb, type)->dqi_priv; + /* + * We modify tree, leaf block, global info, local chunk header, + * global and local inode; OCFS2_QINFO_WRITE_CREDITS already + * accounts for inode update + */ + return (oinfo->dqi_gi.dqi_qtree_depth + 2) * + OCFS2_QUOTA_BLOCK_UPDATE_CREDITS + + OCFS2_QINFO_WRITE_CREDITS + + OCFS2_INODE_UPDATE_CREDITS; +} - if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, features[type])) - return 0; +void ocfs2_drop_dquot_refs(struct work_struct *work) +{ + struct ocfs2_super *osb = container_of(work, struct ocfs2_super, + dquot_drop_work); + struct llist_node *list; + struct ocfs2_dquot *odquot, *next_odquot; - oinfo = sb_dqinfo(sb, type)->dqi_priv; - /* We modify tree, leaf block, global info, local chunk header, - * global and local inode */ - return oinfo->dqi_gi.dqi_qtree_depth + 2 + 1 + - 2 * OCFS2_INODE_UPDATE_CREDITS; + list = llist_del_all(&osb->dquot_drop_list); + llist_for_each_entry_safe(odquot, next_odquot, list, list) { + /* Drop the reference we acquired in ocfs2_dquot_release() */ + dqput(&odquot->dq_dquot); + } } +/* + * Called when the last reference to dquot is dropped. If we are called from + * downconvert thread, we cannot do all the handling here because grabbing + * quota lock could deadlock (the node holding the quota lock could need some + * other cluster lock to proceed but with blocked downconvert thread we cannot + * release any lock). + */ static int ocfs2_release_dquot(struct dquot *dquot) { handle_t *handle; struct ocfs2_mem_dqinfo *oinfo = - sb_dqinfo(dquot->dq_sb, dquot->dq_type)->dqi_priv; + sb_dqinfo(dquot->dq_sb, dquot->dq_id.type)->dqi_priv; struct ocfs2_super *osb = OCFS2_SB(dquot->dq_sb); int status = 0; - mlog_entry("id=%u, type=%d", dquot->dq_id, dquot->dq_type); + trace_ocfs2_release_dquot(from_kqid(&init_user_ns, dquot->dq_id), + dquot->dq_id.type); + mutex_lock(&dquot->dq_lock); + /* Check whether we are not racing with some other dqget() */ + if (atomic_read(&dquot->dq_count) > 1) + goto out; + /* Running from downconvert thread? Postpone quota processing to wq */ + if (current == osb->dc_task) { + /* + * Grab our own reference to dquot and queue it for delayed + * dropping. Quota code rechecks after calling + * ->release_dquot() and won't free dquot structure. + */ + dqgrab(dquot); + /* First entry on list -> queue work */ + if (llist_add(&OCFS2_DQUOT(dquot)->list, &osb->dquot_drop_list)) + queue_work(ocfs2_wq, &osb->dquot_drop_work); + goto out; + } status = ocfs2_lock_global_qf(oinfo, 1); if (status < 0) goto out; handle = ocfs2_start_trans(osb, - ocfs2_calc_qdel_credits(dquot->dq_sb, dquot->dq_type)); + ocfs2_calc_qdel_credits(dquot->dq_sb, dquot->dq_id.type)); if (IS_ERR(handle)) { status = PTR_ERR(handle); mlog_errno(status); goto out_ilock; } - status = dquot_release(dquot); + + status = ocfs2_global_release_dquot(dquot); + if (status < 0) { + mlog_errno(status); + goto out_trans; + } + status = ocfs2_local_release_dquot(handle, dquot); + /* + * If we fail here, we cannot do much as global structure is + * already released. So just complain... + */ + if (status < 0) + mlog_errno(status); + /* + * Clear dq_off so that we search for the structure in quota file next + * time we acquire it. The structure might be deleted and reallocated + * elsewhere by another node while our dquot structure is on freelist. + */ + dquot->dq_off = 0; + clear_bit(DQ_ACTIVE_B, &dquot->dq_flags); +out_trans: ocfs2_commit_trans(osb, handle); out_ilock: ocfs2_unlock_global_qf(oinfo, 1); out: - mlog_exit(status); + mutex_unlock(&dquot->dq_lock); + if (status) + mlog_errno(status); return status; } -int ocfs2_calc_qinit_credits(struct super_block *sb, int type) -{ - struct ocfs2_mem_dqinfo *oinfo; - int features[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA, - OCFS2_FEATURE_RO_COMPAT_GRPQUOTA }; - struct ocfs2_dinode *lfe, *gfe; - - if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, features[type])) - return 0; - - oinfo = sb_dqinfo(sb, type)->dqi_priv; - gfe = (struct ocfs2_dinode *)oinfo->dqi_gqi_bh->b_data; - lfe = (struct ocfs2_dinode *)oinfo->dqi_lqi_bh->b_data; - /* We can extend local file + global file. In local file we - * can modify info, chunk header block and dquot block. In - * global file we can modify info, tree and leaf block */ - return ocfs2_calc_extend_credits(sb, &lfe->id2.i_list, 0) + - ocfs2_calc_extend_credits(sb, &gfe->id2.i_list, 0) + - 3 + oinfo->dqi_gi.dqi_qtree_depth + 2; -} - +/* + * Read global dquot structure from disk or create it if it does + * not exist. Also update use count of the global structure and + * create structure in node-local quota file. + */ static int ocfs2_acquire_dquot(struct dquot *dquot) { + int status = 0, err; + int ex = 0; + struct super_block *sb = dquot->dq_sb; + struct ocfs2_super *osb = OCFS2_SB(sb); + int type = dquot->dq_id.type; + struct ocfs2_mem_dqinfo *info = sb_dqinfo(sb, type)->dqi_priv; + struct inode *gqinode = info->dqi_gqinode; + int need_alloc = ocfs2_global_qinit_alloc(sb, type); handle_t *handle; - struct ocfs2_mem_dqinfo *oinfo = - sb_dqinfo(dquot->dq_sb, dquot->dq_type)->dqi_priv; - struct ocfs2_super *osb = OCFS2_SB(dquot->dq_sb); - int status = 0; - mlog_entry("id=%u, type=%d", dquot->dq_id, dquot->dq_type); - /* We need an exclusive lock, because we're going to update use count - * and instantiate possibly new dquot structure */ - status = ocfs2_lock_global_qf(oinfo, 1); + trace_ocfs2_acquire_dquot(from_kqid(&init_user_ns, dquot->dq_id), + type); + mutex_lock(&dquot->dq_lock); + /* + * We need an exclusive lock, because we're going to update use count + * and instantiate possibly new dquot structure + */ + status = ocfs2_lock_global_qf(info, 1); if (status < 0) goto out; + status = ocfs2_qinfo_lock(info, 0); + if (status < 0) + goto out_dq; + /* + * We always want to read dquot structure from disk because we don't + * know what happened with it while it was on freelist. + */ + status = qtree_read_dquot(&info->dqi_gi, dquot); + ocfs2_qinfo_unlock(info, 0); + if (status < 0) + goto out_dq; + + OCFS2_DQUOT(dquot)->dq_use_count++; + OCFS2_DQUOT(dquot)->dq_origspace = dquot->dq_dqb.dqb_curspace; + OCFS2_DQUOT(dquot)->dq_originodes = dquot->dq_dqb.dqb_curinodes; + if (!dquot->dq_off) { /* No real quota entry? */ + ex = 1; + /* + * Add blocks to quota file before we start a transaction since + * locking allocators ranks above a transaction start + */ + WARN_ON(journal_current_handle()); + status = ocfs2_extend_no_holes(gqinode, NULL, + i_size_read(gqinode) + (need_alloc << sb->s_blocksize_bits), + i_size_read(gqinode)); + if (status < 0) + goto out_dq; + } + handle = ocfs2_start_trans(osb, - ocfs2_calc_qinit_credits(dquot->dq_sb, dquot->dq_type)); + ocfs2_calc_global_qinit_credits(sb, type)); if (IS_ERR(handle)) { status = PTR_ERR(handle); - mlog_errno(status); - goto out_ilock; + goto out_dq; } - status = dquot_acquire(dquot); + status = ocfs2_qinfo_lock(info, ex); + if (status < 0) + goto out_trans; + status = qtree_write_dquot(&info->dqi_gi, dquot); + if (ex && info_dirty(sb_dqinfo(sb, type))) { + err = __ocfs2_global_write_info(sb, type); + if (!status) + status = err; + } + ocfs2_qinfo_unlock(info, ex); +out_trans: ocfs2_commit_trans(osb, handle); -out_ilock: - ocfs2_unlock_global_qf(oinfo, 1); +out_dq: + ocfs2_unlock_global_qf(info, 1); + if (status < 0) + goto out; + + status = ocfs2_create_local_dquot(dquot); + if (status < 0) + goto out; + set_bit(DQ_ACTIVE_B, &dquot->dq_flags); out: - mlog_exit(status); + mutex_unlock(&dquot->dq_lock); + if (status) + mlog_errno(status); return status; } @@ -740,13 +871,13 @@ static int ocfs2_mark_dquot_dirty(struct dquot *dquot) int sync = 0; int status; struct super_block *sb = dquot->dq_sb; - int type = dquot->dq_type; + int type = dquot->dq_id.type; struct ocfs2_mem_dqinfo *oinfo = sb_dqinfo(sb, type)->dqi_priv; handle_t *handle; struct ocfs2_super *osb = OCFS2_SB(sb); - mlog_entry("id=%u, type=%d", dquot->dq_id, type); - dquot_mark_dquot_dirty(dquot); + trace_ocfs2_mark_dquot_dirty(from_kqid(&init_user_ns, dquot->dq_id), + type); /* In case user set some limits, sync dquot immediately to global * quota file so that information propagates quicker */ @@ -769,19 +900,22 @@ static int ocfs2_mark_dquot_dirty(struct dquot *dquot) mlog_errno(status); goto out_ilock; } + mutex_lock(&sb_dqopt(sb)->dqio_mutex); status = ocfs2_sync_dquot(dquot); if (status < 0) { mlog_errno(status); - goto out_trans; + goto out_dlock; } /* Now write updated local dquot structure */ - status = dquot_commit(dquot); -out_trans: + status = ocfs2_local_write_dquot(dquot); +out_dlock: + mutex_unlock(&sb_dqopt(sb)->dqio_mutex); ocfs2_commit_trans(osb, handle); out_ilock: ocfs2_unlock_global_qf(oinfo, 1); out: - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -792,8 +926,6 @@ static int ocfs2_write_info(struct super_block *sb, int type) int status = 0; struct ocfs2_mem_dqinfo *oinfo = sb_dqinfo(sb, type)->dqi_priv; - mlog_entry_void(); - status = ocfs2_lock_global_qf(oinfo, 1); if (status < 0) goto out; @@ -808,7 +940,8 @@ static int ocfs2_write_info(struct super_block *sb, int type) out_ilock: ocfs2_unlock_global_qf(oinfo, 1); out: - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -827,15 +960,8 @@ static void ocfs2_destroy_dquot(struct dquot *dquot) kmem_cache_free(ocfs2_dquot_cachep, dquot); } -struct dquot_operations ocfs2_quota_operations = { - .initialize = dquot_initialize, - .drop = dquot_drop, - .alloc_space = dquot_alloc_space, - .alloc_inode = dquot_alloc_inode, - .free_space = dquot_free_space, - .free_inode = dquot_free_inode, - .transfer = dquot_transfer, - .write_dquot = ocfs2_write_dquot, +const struct dquot_operations ocfs2_quota_operations = { + /* We never make dquot dirty so .write_dquot is never called */ .acquire_dquot = ocfs2_acquire_dquot, .release_dquot = ocfs2_release_dquot, .mark_dirty = ocfs2_mark_dquot_dirty, @@ -843,20 +969,3 @@ struct dquot_operations ocfs2_quota_operations = { .alloc_dquot = ocfs2_alloc_dquot, .destroy_dquot = ocfs2_destroy_dquot, }; - -int ocfs2_quota_setup(void) -{ - ocfs2_quota_wq = create_workqueue("o2quot"); - if (!ocfs2_quota_wq) - return -ENOMEM; - return 0; -} - -void ocfs2_quota_shutdown(void) -{ - if (ocfs2_quota_wq) { - flush_workqueue(ocfs2_quota_wq); - destroy_workqueue(ocfs2_quota_wq); - ocfs2_quota_wq = NULL; - } -} diff --git a/fs/ocfs2/quota_local.c b/fs/ocfs2/quota_local.c index 07deec5e972..2001862bf2b 100644 --- a/fs/ocfs2/quota_local.c +++ b/fs/ocfs2/quota_local.c @@ -3,11 +3,11 @@ */ #include <linux/fs.h> +#include <linux/slab.h> #include <linux/quota.h> #include <linux/quotaops.h> #include <linux/module.h> -#define MLOG_MASK_PREFIX ML_QUOTA #include <cluster/masklog.h> #include "ocfs2_fs.h" @@ -20,6 +20,9 @@ #include "sysfile.h" #include "dlmglue.h" #include "quota.h" +#include "uptodate.h" +#include "super.h" +#include "ocfs2_trace.h" /* Number of local quota structures per block */ static inline unsigned int ol_quota_entries_per_block(struct super_block *sb) @@ -100,13 +103,14 @@ static int ocfs2_modify_bh(struct inode *inode, struct buffer_head *bh, handle_t *handle; int status; - handle = ocfs2_start_trans(OCFS2_SB(sb), 1); + handle = ocfs2_start_trans(OCFS2_SB(sb), + OCFS2_QUOTA_BLOCK_UPDATE_CREDITS); if (IS_ERR(handle)) { status = PTR_ERR(handle); mlog_errno(status); return status; } - status = ocfs2_journal_access_dq(handle, inode, bh, + status = ocfs2_journal_access_dq(handle, INODE_CACHE(inode), bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); @@ -116,12 +120,8 @@ static int ocfs2_modify_bh(struct inode *inode, struct buffer_head *bh, lock_buffer(bh); modify(bh, private); unlock_buffer(bh); - status = ocfs2_journal_dirty(handle, bh); - if (status < 0) { - mlog_errno(status); - ocfs2_commit_trans(OCFS2_SB(sb), handle); - return status; - } + ocfs2_journal_dirty(handle, bh); + status = ocfs2_commit_trans(OCFS2_SB(sb), handle); if (status < 0) { mlog_errno(status); @@ -130,6 +130,39 @@ static int ocfs2_modify_bh(struct inode *inode, struct buffer_head *bh, return 0; } +/* + * Read quota block from a given logical offset. + * + * This function acquires ip_alloc_sem and thus it must not be called with a + * transaction started. + */ +static int ocfs2_read_quota_block(struct inode *inode, u64 v_block, + struct buffer_head **bh) +{ + int rc = 0; + struct buffer_head *tmp = *bh; + + if (i_size_read(inode) >> inode->i_sb->s_blocksize_bits <= v_block) { + ocfs2_error(inode->i_sb, + "Quota file %llu is probably corrupted! Requested " + "to read block %Lu but file has size only %Lu\n", + (unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)v_block, + (unsigned long long)i_size_read(inode)); + return -EIO; + } + rc = ocfs2_read_virt_blocks(inode, v_block, 1, &tmp, 0, + ocfs2_validate_quota_block); + if (rc) + mlog_errno(rc); + + /* If ocfs2_read_virt_blocks() got us a new bh, pass it up. */ + if (!rc && !*bh) + *bh = tmp; + + return rc; +} + /* Check whether we understand format of quota files */ static int ocfs2_local_check_quota_file(struct super_block *sb, int type) { @@ -371,7 +404,9 @@ struct ocfs2_quota_recovery *ocfs2_begin_quota_recovery( int status = 0; struct ocfs2_quota_recovery *rec; - mlog(ML_NOTICE, "Beginning quota recovery in slot %u\n", slot_num); + printk(KERN_NOTICE "ocfs2: Beginning quota recovery on device (%s) for " + "slot %u\n", osb->dev_str, slot_num); + rec = ocfs2_alloc_quota_recovery(); if (!rec) return ERR_PTR(-ENOMEM); @@ -442,11 +477,7 @@ static int ocfs2_recover_local_quota_file(struct inode *lqinode, struct ocfs2_recovery_chunk *rchunk, *next; qsize_t spacechange, inodechange; - mlog_entry("ino=%lu type=%u", (unsigned long)lqinode->i_ino, type); - - status = ocfs2_lock_global_qf(oinfo, 1); - if (status < 0) - goto out; + trace_ocfs2_recover_local_quota_file((unsigned long)lqinode->i_ino, type); list_for_each_entry_safe(rchunk, next, &(rec->r_list[type]), rc_list) { chunk = rchunk->rc_chunk; @@ -459,7 +490,7 @@ static int ocfs2_recover_local_quota_file(struct inode *lqinode, break; } dchunk = (struct ocfs2_local_disk_chunk *)hbh->b_data; - for_each_bit(bit, rchunk->rc_bitmap, ol_chunk_entries(sb)) { + for_each_set_bit(bit, rchunk->rc_bitmap, ol_chunk_entries(sb)) { qbh = NULL; status = ocfs2_read_quota_block(lqinode, ol_dqblk_block(sb, chunk, bit), @@ -470,7 +501,9 @@ static int ocfs2_recover_local_quota_file(struct inode *lqinode, } dqblk = (struct ocfs2_local_disk_dqblk *)(qbh->b_data + ol_dqblk_block_off(sb, chunk, bit)); - dquot = dqget(sb, le64_to_cpu(dqblk->dqb_id), type); + dquot = dqget(sb, + make_kqid(&init_user_ns, type, + le64_to_cpu(dqblk->dqb_id))); if (!dquot) { status = -EIO; mlog(ML_ERROR, "Failed to get quota structure " @@ -480,12 +513,18 @@ static int ocfs2_recover_local_quota_file(struct inode *lqinode, type); goto out_put_bh; } + status = ocfs2_lock_global_qf(oinfo, 1); + if (status < 0) { + mlog_errno(status); + goto out_put_dquot; + } + handle = ocfs2_start_trans(OCFS2_SB(sb), OCFS2_QSYNC_CREDITS); if (IS_ERR(handle)) { status = PTR_ERR(handle); mlog_errno(status); - goto out_put_dquot; + goto out_drop_lock; } mutex_lock(&sb_dqopt(sb)->dqio_mutex); spin_lock(&dq_data_lock); @@ -506,23 +545,24 @@ static int ocfs2_recover_local_quota_file(struct inode *lqinode, goto out_commit; } /* Release local quota file entry */ - status = ocfs2_journal_access_dq(handle, lqinode, + status = ocfs2_journal_access_dq(handle, + INODE_CACHE(lqinode), qbh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); goto out_commit; } lock_buffer(qbh); - WARN_ON(!ocfs2_test_bit(bit, dchunk->dqc_bitmap)); - ocfs2_clear_bit(bit, dchunk->dqc_bitmap); + WARN_ON(!ocfs2_test_bit_unaligned(bit, dchunk->dqc_bitmap)); + ocfs2_clear_bit_unaligned(bit, dchunk->dqc_bitmap); le32_add_cpu(&dchunk->dqc_free, 1); unlock_buffer(qbh); - status = ocfs2_journal_dirty(handle, qbh); - if (status < 0) - mlog_errno(status); + ocfs2_journal_dirty(handle, qbh); out_commit: mutex_unlock(&sb_dqopt(sb)->dqio_mutex); ocfs2_commit_trans(OCFS2_SB(sb), handle); +out_drop_lock: + ocfs2_unlock_global_qf(oinfo, 1); out_put_dquot: dqput(dquot); out_put_bh: @@ -537,11 +577,10 @@ out_put_bh: if (status < 0) break; } - ocfs2_unlock_global_qf(oinfo, 1); -out: if (status < 0) free_recovery_list(&(rec->r_list[type])); - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -561,12 +600,14 @@ int ocfs2_finish_quota_recovery(struct ocfs2_super *osb, struct inode *lqinode; unsigned int flags; - mlog(ML_NOTICE, "Finishing quota recovery in slot %u\n", slot_num); + printk(KERN_NOTICE "ocfs2: Finishing quota recovery on device (%s) for " + "slot %u\n", osb->dev_str, slot_num); + mutex_lock(&sb_dqopt(sb)->dqonoff_mutex); for (type = 0; type < MAXQUOTAS; type++) { if (list_empty(&(rec->r_list[type]))) continue; - mlog(0, "Recovering quota in slot %d\n", slot_num); + trace_ocfs2_finish_quota_recovery(slot_num); lqinode = ocfs2_get_system_file_inode(osb, ino[type], slot_num); if (!lqinode) { status = -ENOENT; @@ -577,8 +618,9 @@ int ocfs2_finish_quota_recovery(struct ocfs2_super *osb, /* Someone else is holding the lock? Then he must be * doing the recovery. Just skip the file... */ if (status == -EAGAIN) { - mlog(ML_NOTICE, "skipping quota recovery for slot %d " - "because quota file is locked.\n", slot_num); + printk(KERN_NOTICE "ocfs2: Skipping quota recovery on " + "device (%s) for slot %d because quota file is " + "locked.\n", osb->dev_str, slot_num); status = 0; goto out_put; } else if (status < 0) { @@ -608,13 +650,15 @@ int ocfs2_finish_quota_recovery(struct ocfs2_super *osb, goto out_bh; /* Mark quota file as clean if we are recovering quota file of * some other node. */ - handle = ocfs2_start_trans(osb, 1); + handle = ocfs2_start_trans(osb, + OCFS2_LOCAL_QINFO_WRITE_CREDITS); if (IS_ERR(handle)) { status = PTR_ERR(handle); mlog_errno(status); goto out_bh; } - status = ocfs2_journal_access_dq(handle, lqinode, bh, + status = ocfs2_journal_access_dq(handle, INODE_CACHE(lqinode), + bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); @@ -623,9 +667,7 @@ int ocfs2_finish_quota_recovery(struct ocfs2_super *osb, lock_buffer(bh); ldinfo->dqi_flags = cpu_to_le32(flags | OLQF_CLEAN); unlock_buffer(bh); - status = ocfs2_journal_dirty(handle, bh); - if (status < 0) - mlog_errno(status); + ocfs2_journal_dirty(handle, bh); out_trans: ocfs2_commit_trans(osb, handle); out_bh: @@ -655,6 +697,9 @@ static int ocfs2_local_read_info(struct super_block *sb, int type) struct ocfs2_quota_recovery *rec; int locked = 0; + /* We don't need the lock and we have to acquire quota file locks + * which will later depend on this lock */ + mutex_unlock(&sb_dqopt(sb)->dqio_mutex); info->dqi_maxblimit = 0x7fffffffffffffffLL; info->dqi_maxilimit = 0x7fffffffffffffffLL; oinfo = kmalloc(sizeof(struct ocfs2_mem_dqinfo), GFP_NOFS); @@ -668,7 +713,7 @@ static int ocfs2_local_read_info(struct super_block *sb, int type) INIT_LIST_HEAD(&oinfo->dqi_chunk); oinfo->dqi_rec = NULL; oinfo->dqi_lqi_bh = NULL; - oinfo->dqi_ibh = NULL; + oinfo->dqi_libh = NULL; status = ocfs2_global_read_info(sb, type); if (status < 0) @@ -694,7 +739,7 @@ static int ocfs2_local_read_info(struct super_block *sb, int type) info->dqi_flags = le32_to_cpu(ldinfo->dqi_flags); oinfo->dqi_chunks = le32_to_cpu(ldinfo->dqi_chunks); oinfo->dqi_blocks = le32_to_cpu(ldinfo->dqi_blocks); - oinfo->dqi_ibh = bh; + oinfo->dqi_libh = bh; /* We crashed when using local quota file? */ if (!(info->dqi_flags & OLQF_CLEAN)) { @@ -733,6 +778,7 @@ static int ocfs2_local_read_info(struct super_block *sb, int type) goto out_err; } + mutex_lock(&sb_dqopt(sb)->dqio_mutex); return 0; out_err: if (oinfo) { @@ -746,6 +792,7 @@ out_err: kfree(oinfo); } brelse(bh); + mutex_lock(&sb_dqopt(sb)->dqio_mutex); return -1; } @@ -754,7 +801,7 @@ static int ocfs2_local_write_info(struct super_block *sb, int type) { struct mem_dqinfo *info = sb_dqinfo(sb, type); struct buffer_head *bh = ((struct ocfs2_mem_dqinfo *)info->dqi_priv) - ->dqi_ibh; + ->dqi_libh; int status; status = ocfs2_modify_bh(sb_dqopt(sb)->files[type], bh, olq_update_info, @@ -777,10 +824,6 @@ static int ocfs2_local_free_info(struct super_block *sb, int type) int mark_clean = 1, len; int status; - /* At this point we know there are no more dquots and thus - * even if there's some sync in the pdflush queue, it won't - * find any dquots and return without doing anything */ - cancel_delayed_work_sync(&oinfo->dqi_sync_work); iput(oinfo->dqi_gqinode); ocfs2_simple_drop_lockres(OCFS2_SB(sb), &oinfo->dqi_gqlock); ocfs2_lock_res_free(&oinfo->dqi_gqlock); @@ -815,7 +858,7 @@ static int ocfs2_local_free_info(struct super_block *sb, int type) /* Mark local file as clean */ info->dqi_flags |= OLQF_CLEAN; status = ocfs2_modify_bh(sb_dqopt(sb)->files[type], - oinfo->dqi_ibh, + oinfo->dqi_libh, olq_update_info, info); if (status < 0) { @@ -825,7 +868,7 @@ static int ocfs2_local_free_info(struct super_block *sb, int type) out: ocfs2_inode_unlock(sb_dqopt(sb)->files[type], 1); - brelse(oinfo->dqi_ibh); + brelse(oinfo->dqi_libh); brelse(oinfo->dqi_lqi_bh); kfree(oinfo); return 0; @@ -840,35 +883,36 @@ static void olq_set_dquot(struct buffer_head *bh, void *private) dqblk = (struct ocfs2_local_disk_dqblk *)(bh->b_data + ol_dqblk_block_offset(sb, od->dq_local_off)); - dqblk->dqb_id = cpu_to_le64(od->dq_dquot.dq_id); + dqblk->dqb_id = cpu_to_le64(from_kqid(&init_user_ns, + od->dq_dquot.dq_id)); spin_lock(&dq_data_lock); dqblk->dqb_spacemod = cpu_to_le64(od->dq_dquot.dq_dqb.dqb_curspace - od->dq_origspace); dqblk->dqb_inodemod = cpu_to_le64(od->dq_dquot.dq_dqb.dqb_curinodes - od->dq_originodes); spin_unlock(&dq_data_lock); - mlog(0, "Writing local dquot %u space %lld inodes %lld\n", - od->dq_dquot.dq_id, (long long)le64_to_cpu(dqblk->dqb_spacemod), - (long long)le64_to_cpu(dqblk->dqb_inodemod)); + trace_olq_set_dquot( + (unsigned long long)le64_to_cpu(dqblk->dqb_spacemod), + (unsigned long long)le64_to_cpu(dqblk->dqb_inodemod), + from_kqid(&init_user_ns, od->dq_dquot.dq_id)); } /* Write dquot to local quota file */ -static int ocfs2_local_write_dquot(struct dquot *dquot) +int ocfs2_local_write_dquot(struct dquot *dquot) { struct super_block *sb = dquot->dq_sb; struct ocfs2_dquot *od = OCFS2_DQUOT(dquot); - struct buffer_head *bh = NULL; + struct buffer_head *bh; + struct inode *lqinode = sb_dqopt(sb)->files[dquot->dq_id.type]; int status; - status = ocfs2_read_quota_block(sb_dqopt(sb)->files[dquot->dq_type], - ol_dqblk_file_block(sb, od->dq_local_off), - &bh); + status = ocfs2_read_quota_phys_block(lqinode, od->dq_local_phys_blk, + &bh); if (status) { mlog_errno(status); goto out; } - status = ocfs2_modify_bh(sb_dqopt(sb)->files[dquot->dq_type], bh, - olq_set_dquot, od); + status = ocfs2_modify_bh(lqinode, bh, olq_set_dquot, od); if (status < 0) { mlog_errno(status); goto out; @@ -908,7 +952,7 @@ static struct ocfs2_quota_chunk *ocfs2_find_free_entry(struct super_block *sb, * ol_quota_entries_per_block(sb); } - found = ocfs2_find_next_zero_bit(dchunk->dqc_bitmap, len, 0); + found = ocfs2_find_next_zero_bit_unaligned(dchunk->dqc_bitmap, len, 0); /* We failed? */ if (found == len) { mlog(ML_ERROR, "Did not find empty entry in chunk %d with %u" @@ -933,19 +977,19 @@ static struct ocfs2_quota_chunk *ocfs2_local_quota_add_chunk( struct ocfs2_local_disk_chunk *dchunk; int status; handle_t *handle; - struct buffer_head *bh = NULL; + struct buffer_head *bh = NULL, *dbh = NULL; u64 p_blkno; /* We are protected by dqio_sem so no locking needed */ - status = ocfs2_extend_no_holes(lqinode, - lqinode->i_size + 2 * sb->s_blocksize, - lqinode->i_size); + status = ocfs2_extend_no_holes(lqinode, NULL, + i_size_read(lqinode) + 2 * sb->s_blocksize, + i_size_read(lqinode)); if (status < 0) { mlog_errno(status); goto out; } status = ocfs2_simple_size_update(lqinode, oinfo->dqi_lqi_bh, - lqinode->i_size + 2 * sb->s_blocksize); + i_size_read(lqinode) + 2 * sb->s_blocksize); if (status < 0) { mlog_errno(status); goto out; @@ -957,32 +1001,33 @@ static struct ocfs2_quota_chunk *ocfs2_local_quota_add_chunk( mlog_errno(status); goto out; } + /* Local quota info and two new blocks we initialize */ + handle = ocfs2_start_trans(OCFS2_SB(sb), + OCFS2_LOCAL_QINFO_WRITE_CREDITS + + 2 * OCFS2_QUOTA_BLOCK_UPDATE_CREDITS); + if (IS_ERR(handle)) { + status = PTR_ERR(handle); + mlog_errno(status); + goto out; + } - down_read(&OCFS2_I(lqinode)->ip_alloc_sem); + /* Initialize chunk header */ status = ocfs2_extent_map_get_blocks(lqinode, oinfo->dqi_blocks, &p_blkno, NULL, NULL); - up_read(&OCFS2_I(lqinode)->ip_alloc_sem); if (status < 0) { mlog_errno(status); - goto out; + goto out_trans; } bh = sb_getblk(sb, p_blkno); if (!bh) { status = -ENOMEM; mlog_errno(status); - goto out; + goto out_trans; } dchunk = (struct ocfs2_local_disk_chunk *)bh->b_data; - - handle = ocfs2_start_trans(OCFS2_SB(sb), 2); - if (IS_ERR(handle)) { - status = PTR_ERR(handle); - mlog_errno(status); - goto out; - } - - status = ocfs2_journal_access_dq(handle, lqinode, bh, - OCFS2_JOURNAL_ACCESS_WRITE); + ocfs2_set_new_buffer_uptodate(INODE_CACHE(lqinode), bh); + status = ocfs2_journal_access_dq(handle, INODE_CACHE(lqinode), bh, + OCFS2_JOURNAL_ACCESS_CREATE); if (status < 0) { mlog_errno(status); goto out_trans; @@ -992,14 +1037,35 @@ static struct ocfs2_quota_chunk *ocfs2_local_quota_add_chunk( memset(dchunk->dqc_bitmap, 0, sb->s_blocksize - sizeof(struct ocfs2_local_disk_chunk) - OCFS2_QBLK_RESERVED_SPACE); - set_buffer_uptodate(bh); unlock_buffer(bh); - status = ocfs2_journal_dirty(handle, bh); + ocfs2_journal_dirty(handle, bh); + + /* Initialize new block with structures */ + status = ocfs2_extent_map_get_blocks(lqinode, oinfo->dqi_blocks + 1, + &p_blkno, NULL, NULL); + if (status < 0) { + mlog_errno(status); + goto out_trans; + } + dbh = sb_getblk(sb, p_blkno); + if (!dbh) { + status = -ENOMEM; + mlog_errno(status); + goto out_trans; + } + ocfs2_set_new_buffer_uptodate(INODE_CACHE(lqinode), dbh); + status = ocfs2_journal_access_dq(handle, INODE_CACHE(lqinode), dbh, + OCFS2_JOURNAL_ACCESS_CREATE); if (status < 0) { mlog_errno(status); goto out_trans; } + lock_buffer(dbh); + memset(dbh->b_data, 0, sb->s_blocksize - OCFS2_QBLK_RESERVED_SPACE); + unlock_buffer(dbh); + ocfs2_journal_dirty(handle, dbh); + /* Update local quotafile info */ oinfo->dqi_blocks += 2; oinfo->dqi_chunks++; status = ocfs2_local_write_info(sb, type); @@ -1024,6 +1090,7 @@ out_trans: ocfs2_commit_trans(OCFS2_SB(sb), handle); out: brelse(bh); + brelse(dbh); kmem_cache_free(ocfs2_qf_chunk_cachep, chunk); return ERR_PTR(status); } @@ -1041,6 +1108,8 @@ static struct ocfs2_quota_chunk *ocfs2_extend_local_quota_file( struct ocfs2_local_disk_chunk *dchunk; int epb = ol_quota_entries_per_block(sb); unsigned int chunk_blocks; + struct buffer_head *bh; + u64 p_blkno; int status; handle_t *handle; @@ -1055,26 +1124,59 @@ static struct ocfs2_quota_chunk *ocfs2_extend_local_quota_file( return ocfs2_local_quota_add_chunk(sb, type, offset); /* We are protected by dqio_sem so no locking needed */ - status = ocfs2_extend_no_holes(lqinode, - lqinode->i_size + sb->s_blocksize, - lqinode->i_size); + status = ocfs2_extend_no_holes(lqinode, NULL, + i_size_read(lqinode) + sb->s_blocksize, + i_size_read(lqinode)); if (status < 0) { mlog_errno(status); goto out; } status = ocfs2_simple_size_update(lqinode, oinfo->dqi_lqi_bh, - lqinode->i_size + sb->s_blocksize); + i_size_read(lqinode) + sb->s_blocksize); + if (status < 0) { + mlog_errno(status); + goto out; + } + + /* Get buffer from the just added block */ + status = ocfs2_extent_map_get_blocks(lqinode, oinfo->dqi_blocks, + &p_blkno, NULL, NULL); if (status < 0) { mlog_errno(status); goto out; } - handle = ocfs2_start_trans(OCFS2_SB(sb), 2); + bh = sb_getblk(sb, p_blkno); + if (!bh) { + status = -ENOMEM; + mlog_errno(status); + goto out; + } + ocfs2_set_new_buffer_uptodate(INODE_CACHE(lqinode), bh); + + /* Local quota info, chunk header and the new block we initialize */ + handle = ocfs2_start_trans(OCFS2_SB(sb), + OCFS2_LOCAL_QINFO_WRITE_CREDITS + + 2 * OCFS2_QUOTA_BLOCK_UPDATE_CREDITS); if (IS_ERR(handle)) { status = PTR_ERR(handle); mlog_errno(status); goto out; } - status = ocfs2_journal_access_dq(handle, lqinode, chunk->qc_headerbh, + /* Zero created block */ + status = ocfs2_journal_access_dq(handle, INODE_CACHE(lqinode), bh, + OCFS2_JOURNAL_ACCESS_CREATE); + if (status < 0) { + mlog_errno(status); + goto out_trans; + } + lock_buffer(bh); + memset(bh->b_data, 0, sb->s_blocksize); + unlock_buffer(bh); + ocfs2_journal_dirty(handle, bh); + + /* Update chunk header */ + status = ocfs2_journal_access_dq(handle, INODE_CACHE(lqinode), + chunk->qc_headerbh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); @@ -1085,11 +1187,9 @@ static struct ocfs2_quota_chunk *ocfs2_extend_local_quota_file( lock_buffer(chunk->qc_headerbh); le32_add_cpu(&dchunk->dqc_free, ol_quota_entries_per_block(sb)); unlock_buffer(chunk->qc_headerbh); - status = ocfs2_journal_dirty(handle, chunk->qc_headerbh); - if (status < 0) { - mlog_errno(status); - goto out_trans; - } + ocfs2_journal_dirty(handle, chunk->qc_headerbh); + + /* Update file header */ oinfo->dqi_blocks++; status = ocfs2_local_write_info(sb, type); if (status < 0) { @@ -1116,31 +1216,41 @@ static void olq_alloc_dquot(struct buffer_head *bh, void *private) struct ocfs2_local_disk_chunk *dchunk; dchunk = (struct ocfs2_local_disk_chunk *)bh->b_data; - ocfs2_set_bit(*offset, dchunk->dqc_bitmap); + ocfs2_set_bit_unaligned(*offset, dchunk->dqc_bitmap); le32_add_cpu(&dchunk->dqc_free, -1); } /* Create dquot in the local file for given id */ -static int ocfs2_create_local_dquot(struct dquot *dquot) +int ocfs2_create_local_dquot(struct dquot *dquot) { struct super_block *sb = dquot->dq_sb; - int type = dquot->dq_type; + int type = dquot->dq_id.type; struct inode *lqinode = sb_dqopt(sb)->files[type]; struct ocfs2_quota_chunk *chunk; struct ocfs2_dquot *od = OCFS2_DQUOT(dquot); int offset; int status; + u64 pcount; + down_write(&OCFS2_I(lqinode)->ip_alloc_sem); chunk = ocfs2_find_free_entry(sb, type, &offset); if (!chunk) { chunk = ocfs2_extend_local_quota_file(sb, type, &offset); - if (IS_ERR(chunk)) - return PTR_ERR(chunk); + if (IS_ERR(chunk)) { + status = PTR_ERR(chunk); + goto out; + } } else if (IS_ERR(chunk)) { - return PTR_ERR(chunk); + status = PTR_ERR(chunk); + goto out; } od->dq_local_off = ol_dqblk_off(sb, chunk->qc_num, offset); od->dq_chunk = chunk; + status = ocfs2_extent_map_get_blocks(lqinode, + ol_dqblk_block(sb, chunk->qc_num, offset), + &od->dq_local_phys_blk, + &pcount, + NULL); /* Initialize dquot structure on disk */ status = ocfs2_local_write_dquot(dquot); @@ -1157,57 +1267,25 @@ static int ocfs2_create_local_dquot(struct dquot *dquot) goto out; } out: + up_write(&OCFS2_I(lqinode)->ip_alloc_sem); return status; } -/* Create entry in local file for dquot, load data from the global file */ -static int ocfs2_local_read_dquot(struct dquot *dquot) -{ - int status; - - mlog_entry("id=%u, type=%d\n", dquot->dq_id, dquot->dq_type); - - status = ocfs2_global_read_dquot(dquot); - if (status < 0) { - mlog_errno(status); - goto out_err; - } - - /* Now create entry in the local quota file */ - status = ocfs2_create_local_dquot(dquot); - if (status < 0) { - mlog_errno(status); - goto out_err; - } - mlog_exit(0); - return 0; -out_err: - mlog_exit(status); - return status; -} - -/* Release dquot structure from local quota file. ocfs2_release_dquot() has - * already started a transaction and obtained exclusive lock for global - * quota file. */ -static int ocfs2_local_release_dquot(struct dquot *dquot) +/* + * Release dquot structure from local quota file. ocfs2_release_dquot() has + * already started a transaction and written all changes to global quota file + */ +int ocfs2_local_release_dquot(handle_t *handle, struct dquot *dquot) { int status; - int type = dquot->dq_type; + int type = dquot->dq_id.type; struct ocfs2_dquot *od = OCFS2_DQUOT(dquot); struct super_block *sb = dquot->dq_sb; struct ocfs2_local_disk_chunk *dchunk; int offset; - handle_t *handle = journal_current_handle(); - BUG_ON(!handle); - /* First write all local changes to global file */ - status = ocfs2_global_release_dquot(dquot); - if (status < 0) { - mlog_errno(status); - goto out; - } - - status = ocfs2_journal_access_dq(handle, sb_dqopt(sb)->files[type], + status = ocfs2_journal_access_dq(handle, + INODE_CACHE(sb_dqopt(sb)->files[type]), od->dq_chunk->qc_headerbh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); @@ -1219,31 +1297,20 @@ static int ocfs2_local_release_dquot(struct dquot *dquot) (od->dq_chunk->qc_headerbh->b_data); /* Mark structure as freed */ lock_buffer(od->dq_chunk->qc_headerbh); - ocfs2_clear_bit(offset, dchunk->dqc_bitmap); + ocfs2_clear_bit_unaligned(offset, dchunk->dqc_bitmap); le32_add_cpu(&dchunk->dqc_free, 1); unlock_buffer(od->dq_chunk->qc_headerbh); - status = ocfs2_journal_dirty(handle, od->dq_chunk->qc_headerbh); - if (status < 0) { - mlog_errno(status); - goto out; - } - status = 0; + ocfs2_journal_dirty(handle, od->dq_chunk->qc_headerbh); + out: - /* Clear the read bit so that next time someone uses this - * dquot he reads fresh info from disk and allocates local - * dquot structure */ - clear_bit(DQ_READ_B, &dquot->dq_flags); return status; } -static struct quota_format_ops ocfs2_format_ops = { +static const struct quota_format_ops ocfs2_format_ops = { .check_quota_file = ocfs2_local_check_quota_file, .read_file_info = ocfs2_local_read_info, .write_file_info = ocfs2_global_write_info, .free_file_info = ocfs2_local_free_info, - .read_dqblk = ocfs2_local_read_dquot, - .commit_dqblk = ocfs2_local_write_dquot, - .release_dqblk = ocfs2_local_release_dquot, }; struct quota_format_type ocfs2_quota_format = { diff --git a/fs/ocfs2/refcounttree.c b/fs/ocfs2/refcounttree.c new file mode 100644 index 00000000000..636aab69ead --- /dev/null +++ b/fs/ocfs2/refcounttree.c @@ -0,0 +1,4476 @@ +/* -*- mode: c; c-basic-offset: 8; -*- + * vim: noexpandtab sw=8 ts=8 sts=0: + * + * refcounttree.c + * + * Copyright (C) 2009 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License version 2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + */ + +#include <linux/sort.h> +#include <cluster/masklog.h> +#include "ocfs2.h" +#include "inode.h" +#include "alloc.h" +#include "suballoc.h" +#include "journal.h" +#include "uptodate.h" +#include "super.h" +#include "buffer_head_io.h" +#include "blockcheck.h" +#include "refcounttree.h" +#include "sysfile.h" +#include "dlmglue.h" +#include "extent_map.h" +#include "aops.h" +#include "xattr.h" +#include "namei.h" +#include "ocfs2_trace.h" + +#include <linux/bio.h> +#include <linux/blkdev.h> +#include <linux/slab.h> +#include <linux/writeback.h> +#include <linux/pagevec.h> +#include <linux/swap.h> +#include <linux/security.h> +#include <linux/fsnotify.h> +#include <linux/quotaops.h> +#include <linux/namei.h> +#include <linux/mount.h> +#include <linux/posix_acl.h> + +struct ocfs2_cow_context { + struct inode *inode; + u32 cow_start; + u32 cow_len; + struct ocfs2_extent_tree data_et; + struct ocfs2_refcount_tree *ref_tree; + struct buffer_head *ref_root_bh; + struct ocfs2_alloc_context *meta_ac; + struct ocfs2_alloc_context *data_ac; + struct ocfs2_cached_dealloc_ctxt dealloc; + void *cow_object; + struct ocfs2_post_refcount *post_refcount; + int extra_credits; + int (*get_clusters)(struct ocfs2_cow_context *context, + u32 v_cluster, u32 *p_cluster, + u32 *num_clusters, + unsigned int *extent_flags); + int (*cow_duplicate_clusters)(handle_t *handle, + struct inode *inode, + u32 cpos, u32 old_cluster, + u32 new_cluster, u32 new_len); +}; + +static inline struct ocfs2_refcount_tree * +cache_info_to_refcount(struct ocfs2_caching_info *ci) +{ + return container_of(ci, struct ocfs2_refcount_tree, rf_ci); +} + +static int ocfs2_validate_refcount_block(struct super_block *sb, + struct buffer_head *bh) +{ + int rc; + struct ocfs2_refcount_block *rb = + (struct ocfs2_refcount_block *)bh->b_data; + + trace_ocfs2_validate_refcount_block((unsigned long long)bh->b_blocknr); + + BUG_ON(!buffer_uptodate(bh)); + + /* + * If the ecc fails, we return the error but otherwise + * leave the filesystem running. We know any error is + * local to this block. + */ + rc = ocfs2_validate_meta_ecc(sb, bh->b_data, &rb->rf_check); + if (rc) { + mlog(ML_ERROR, "Checksum failed for refcount block %llu\n", + (unsigned long long)bh->b_blocknr); + return rc; + } + + + if (!OCFS2_IS_VALID_REFCOUNT_BLOCK(rb)) { + ocfs2_error(sb, + "Refcount block #%llu has bad signature %.*s", + (unsigned long long)bh->b_blocknr, 7, + rb->rf_signature); + return -EINVAL; + } + + if (le64_to_cpu(rb->rf_blkno) != bh->b_blocknr) { + ocfs2_error(sb, + "Refcount block #%llu has an invalid rf_blkno " + "of %llu", + (unsigned long long)bh->b_blocknr, + (unsigned long long)le64_to_cpu(rb->rf_blkno)); + return -EINVAL; + } + + if (le32_to_cpu(rb->rf_fs_generation) != OCFS2_SB(sb)->fs_generation) { + ocfs2_error(sb, + "Refcount block #%llu has an invalid " + "rf_fs_generation of #%u", + (unsigned long long)bh->b_blocknr, + le32_to_cpu(rb->rf_fs_generation)); + return -EINVAL; + } + + return 0; +} + +static int ocfs2_read_refcount_block(struct ocfs2_caching_info *ci, + u64 rb_blkno, + struct buffer_head **bh) +{ + int rc; + struct buffer_head *tmp = *bh; + + rc = ocfs2_read_block(ci, rb_blkno, &tmp, + ocfs2_validate_refcount_block); + + /* If ocfs2_read_block() got us a new bh, pass it up. */ + if (!rc && !*bh) + *bh = tmp; + + return rc; +} + +static u64 ocfs2_refcount_cache_owner(struct ocfs2_caching_info *ci) +{ + struct ocfs2_refcount_tree *rf = cache_info_to_refcount(ci); + + return rf->rf_blkno; +} + +static struct super_block * +ocfs2_refcount_cache_get_super(struct ocfs2_caching_info *ci) +{ + struct ocfs2_refcount_tree *rf = cache_info_to_refcount(ci); + + return rf->rf_sb; +} + +static void ocfs2_refcount_cache_lock(struct ocfs2_caching_info *ci) +{ + struct ocfs2_refcount_tree *rf = cache_info_to_refcount(ci); + + spin_lock(&rf->rf_lock); +} + +static void ocfs2_refcount_cache_unlock(struct ocfs2_caching_info *ci) +{ + struct ocfs2_refcount_tree *rf = cache_info_to_refcount(ci); + + spin_unlock(&rf->rf_lock); +} + +static void ocfs2_refcount_cache_io_lock(struct ocfs2_caching_info *ci) +{ + struct ocfs2_refcount_tree *rf = cache_info_to_refcount(ci); + + mutex_lock(&rf->rf_io_mutex); +} + +static void ocfs2_refcount_cache_io_unlock(struct ocfs2_caching_info *ci) +{ + struct ocfs2_refcount_tree *rf = cache_info_to_refcount(ci); + + mutex_unlock(&rf->rf_io_mutex); +} + +static const struct ocfs2_caching_operations ocfs2_refcount_caching_ops = { + .co_owner = ocfs2_refcount_cache_owner, + .co_get_super = ocfs2_refcount_cache_get_super, + .co_cache_lock = ocfs2_refcount_cache_lock, + .co_cache_unlock = ocfs2_refcount_cache_unlock, + .co_io_lock = ocfs2_refcount_cache_io_lock, + .co_io_unlock = ocfs2_refcount_cache_io_unlock, +}; + +static struct ocfs2_refcount_tree * +ocfs2_find_refcount_tree(struct ocfs2_super *osb, u64 blkno) +{ + struct rb_node *n = osb->osb_rf_lock_tree.rb_node; + struct ocfs2_refcount_tree *tree = NULL; + + while (n) { + tree = rb_entry(n, struct ocfs2_refcount_tree, rf_node); + + if (blkno < tree->rf_blkno) + n = n->rb_left; + else if (blkno > tree->rf_blkno) + n = n->rb_right; + else + return tree; + } + + return NULL; +} + +/* osb_lock is already locked. */ +static void ocfs2_insert_refcount_tree(struct ocfs2_super *osb, + struct ocfs2_refcount_tree *new) +{ + u64 rf_blkno = new->rf_blkno; + struct rb_node *parent = NULL; + struct rb_node **p = &osb->osb_rf_lock_tree.rb_node; + struct ocfs2_refcount_tree *tmp; + + while (*p) { + parent = *p; + + tmp = rb_entry(parent, struct ocfs2_refcount_tree, + rf_node); + + if (rf_blkno < tmp->rf_blkno) + p = &(*p)->rb_left; + else if (rf_blkno > tmp->rf_blkno) + p = &(*p)->rb_right; + else { + /* This should never happen! */ + mlog(ML_ERROR, "Duplicate refcount block %llu found!\n", + (unsigned long long)rf_blkno); + BUG(); + } + } + + rb_link_node(&new->rf_node, parent, p); + rb_insert_color(&new->rf_node, &osb->osb_rf_lock_tree); +} + +static void ocfs2_free_refcount_tree(struct ocfs2_refcount_tree *tree) +{ + ocfs2_metadata_cache_exit(&tree->rf_ci); + ocfs2_simple_drop_lockres(OCFS2_SB(tree->rf_sb), &tree->rf_lockres); + ocfs2_lock_res_free(&tree->rf_lockres); + kfree(tree); +} + +static inline void +ocfs2_erase_refcount_tree_from_list_no_lock(struct ocfs2_super *osb, + struct ocfs2_refcount_tree *tree) +{ + rb_erase(&tree->rf_node, &osb->osb_rf_lock_tree); + if (osb->osb_ref_tree_lru && osb->osb_ref_tree_lru == tree) + osb->osb_ref_tree_lru = NULL; +} + +static void ocfs2_erase_refcount_tree_from_list(struct ocfs2_super *osb, + struct ocfs2_refcount_tree *tree) +{ + spin_lock(&osb->osb_lock); + ocfs2_erase_refcount_tree_from_list_no_lock(osb, tree); + spin_unlock(&osb->osb_lock); +} + +static void ocfs2_kref_remove_refcount_tree(struct kref *kref) +{ + struct ocfs2_refcount_tree *tree = + container_of(kref, struct ocfs2_refcount_tree, rf_getcnt); + + ocfs2_free_refcount_tree(tree); +} + +static inline void +ocfs2_refcount_tree_get(struct ocfs2_refcount_tree *tree) +{ + kref_get(&tree->rf_getcnt); +} + +static inline void +ocfs2_refcount_tree_put(struct ocfs2_refcount_tree *tree) +{ + kref_put(&tree->rf_getcnt, ocfs2_kref_remove_refcount_tree); +} + +static inline void ocfs2_init_refcount_tree_ci(struct ocfs2_refcount_tree *new, + struct super_block *sb) +{ + ocfs2_metadata_cache_init(&new->rf_ci, &ocfs2_refcount_caching_ops); + mutex_init(&new->rf_io_mutex); + new->rf_sb = sb; + spin_lock_init(&new->rf_lock); +} + +static inline void ocfs2_init_refcount_tree_lock(struct ocfs2_super *osb, + struct ocfs2_refcount_tree *new, + u64 rf_blkno, u32 generation) +{ + init_rwsem(&new->rf_sem); + ocfs2_refcount_lock_res_init(&new->rf_lockres, osb, + rf_blkno, generation); +} + +static struct ocfs2_refcount_tree* +ocfs2_allocate_refcount_tree(struct ocfs2_super *osb, u64 rf_blkno) +{ + struct ocfs2_refcount_tree *new; + + new = kzalloc(sizeof(struct ocfs2_refcount_tree), GFP_NOFS); + if (!new) + return NULL; + + new->rf_blkno = rf_blkno; + kref_init(&new->rf_getcnt); + ocfs2_init_refcount_tree_ci(new, osb->sb); + + return new; +} + +static int ocfs2_get_refcount_tree(struct ocfs2_super *osb, u64 rf_blkno, + struct ocfs2_refcount_tree **ret_tree) +{ + int ret = 0; + struct ocfs2_refcount_tree *tree, *new = NULL; + struct buffer_head *ref_root_bh = NULL; + struct ocfs2_refcount_block *ref_rb; + + spin_lock(&osb->osb_lock); + if (osb->osb_ref_tree_lru && + osb->osb_ref_tree_lru->rf_blkno == rf_blkno) + tree = osb->osb_ref_tree_lru; + else + tree = ocfs2_find_refcount_tree(osb, rf_blkno); + if (tree) + goto out; + + spin_unlock(&osb->osb_lock); + + new = ocfs2_allocate_refcount_tree(osb, rf_blkno); + if (!new) { + ret = -ENOMEM; + mlog_errno(ret); + return ret; + } + /* + * We need the generation to create the refcount tree lock and since + * it isn't changed during the tree modification, we are safe here to + * read without protection. + * We also have to purge the cache after we create the lock since the + * refcount block may have the stale data. It can only be trusted when + * we hold the refcount lock. + */ + ret = ocfs2_read_refcount_block(&new->rf_ci, rf_blkno, &ref_root_bh); + if (ret) { + mlog_errno(ret); + ocfs2_metadata_cache_exit(&new->rf_ci); + kfree(new); + return ret; + } + + ref_rb = (struct ocfs2_refcount_block *)ref_root_bh->b_data; + new->rf_generation = le32_to_cpu(ref_rb->rf_generation); + ocfs2_init_refcount_tree_lock(osb, new, rf_blkno, + new->rf_generation); + ocfs2_metadata_cache_purge(&new->rf_ci); + + spin_lock(&osb->osb_lock); + tree = ocfs2_find_refcount_tree(osb, rf_blkno); + if (tree) + goto out; + + ocfs2_insert_refcount_tree(osb, new); + + tree = new; + new = NULL; + +out: + *ret_tree = tree; + + osb->osb_ref_tree_lru = tree; + + spin_unlock(&osb->osb_lock); + + if (new) + ocfs2_free_refcount_tree(new); + + brelse(ref_root_bh); + return ret; +} + +static int ocfs2_get_refcount_block(struct inode *inode, u64 *ref_blkno) +{ + int ret; + struct buffer_head *di_bh = NULL; + struct ocfs2_dinode *di; + + ret = ocfs2_read_inode_block(inode, &di_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + BUG_ON(!(OCFS2_I(inode)->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL)); + + di = (struct ocfs2_dinode *)di_bh->b_data; + *ref_blkno = le64_to_cpu(di->i_refcount_loc); + brelse(di_bh); +out: + return ret; +} + +static int __ocfs2_lock_refcount_tree(struct ocfs2_super *osb, + struct ocfs2_refcount_tree *tree, int rw) +{ + int ret; + + ret = ocfs2_refcount_lock(tree, rw); + if (ret) { + mlog_errno(ret); + goto out; + } + + if (rw) + down_write(&tree->rf_sem); + else + down_read(&tree->rf_sem); + +out: + return ret; +} + +/* + * Lock the refcount tree pointed by ref_blkno and return the tree. + * In most case, we lock the tree and read the refcount block. + * So read it here if the caller really needs it. + * + * If the tree has been re-created by other node, it will free the + * old one and re-create it. + */ +int ocfs2_lock_refcount_tree(struct ocfs2_super *osb, + u64 ref_blkno, int rw, + struct ocfs2_refcount_tree **ret_tree, + struct buffer_head **ref_bh) +{ + int ret, delete_tree = 0; + struct ocfs2_refcount_tree *tree = NULL; + struct buffer_head *ref_root_bh = NULL; + struct ocfs2_refcount_block *rb; + +again: + ret = ocfs2_get_refcount_tree(osb, ref_blkno, &tree); + if (ret) { + mlog_errno(ret); + return ret; + } + + ocfs2_refcount_tree_get(tree); + + ret = __ocfs2_lock_refcount_tree(osb, tree, rw); + if (ret) { + mlog_errno(ret); + ocfs2_refcount_tree_put(tree); + goto out; + } + + ret = ocfs2_read_refcount_block(&tree->rf_ci, tree->rf_blkno, + &ref_root_bh); + if (ret) { + mlog_errno(ret); + ocfs2_unlock_refcount_tree(osb, tree, rw); + ocfs2_refcount_tree_put(tree); + goto out; + } + + rb = (struct ocfs2_refcount_block *)ref_root_bh->b_data; + /* + * If the refcount block has been freed and re-created, we may need + * to recreate the refcount tree also. + * + * Here we just remove the tree from the rb-tree, and the last + * kref holder will unlock and delete this refcount_tree. + * Then we goto "again" and ocfs2_get_refcount_tree will create + * the new refcount tree for us. + */ + if (tree->rf_generation != le32_to_cpu(rb->rf_generation)) { + if (!tree->rf_removed) { + ocfs2_erase_refcount_tree_from_list(osb, tree); + tree->rf_removed = 1; + delete_tree = 1; + } + + ocfs2_unlock_refcount_tree(osb, tree, rw); + /* + * We get an extra reference when we create the refcount + * tree, so another put will destroy it. + */ + if (delete_tree) + ocfs2_refcount_tree_put(tree); + brelse(ref_root_bh); + ref_root_bh = NULL; + goto again; + } + + *ret_tree = tree; + if (ref_bh) { + *ref_bh = ref_root_bh; + ref_root_bh = NULL; + } +out: + brelse(ref_root_bh); + return ret; +} + +void ocfs2_unlock_refcount_tree(struct ocfs2_super *osb, + struct ocfs2_refcount_tree *tree, int rw) +{ + if (rw) + up_write(&tree->rf_sem); + else + up_read(&tree->rf_sem); + + ocfs2_refcount_unlock(tree, rw); + ocfs2_refcount_tree_put(tree); +} + +void ocfs2_purge_refcount_trees(struct ocfs2_super *osb) +{ + struct rb_node *node; + struct ocfs2_refcount_tree *tree; + struct rb_root *root = &osb->osb_rf_lock_tree; + + while ((node = rb_last(root)) != NULL) { + tree = rb_entry(node, struct ocfs2_refcount_tree, rf_node); + + trace_ocfs2_purge_refcount_trees( + (unsigned long long) tree->rf_blkno); + + rb_erase(&tree->rf_node, root); + ocfs2_free_refcount_tree(tree); + } +} + +/* + * Create a refcount tree for an inode. + * We take for granted that the inode is already locked. + */ +static int ocfs2_create_refcount_tree(struct inode *inode, + struct buffer_head *di_bh) +{ + int ret; + handle_t *handle = NULL; + struct ocfs2_alloc_context *meta_ac = NULL; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct buffer_head *new_bh = NULL; + struct ocfs2_refcount_block *rb; + struct ocfs2_refcount_tree *new_tree = NULL, *tree = NULL; + u16 suballoc_bit_start; + u32 num_got; + u64 suballoc_loc, first_blkno; + + BUG_ON(oi->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL); + + trace_ocfs2_create_refcount_tree( + (unsigned long long)OCFS2_I(inode)->ip_blkno); + + ret = ocfs2_reserve_new_metadata_blocks(osb, 1, &meta_ac); + if (ret) { + mlog_errno(ret); + goto out; + } + + handle = ocfs2_start_trans(osb, OCFS2_REFCOUNT_TREE_CREATE_CREDITS); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto out; + } + + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc, + &suballoc_bit_start, &num_got, + &first_blkno); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + new_tree = ocfs2_allocate_refcount_tree(osb, first_blkno); + if (!new_tree) { + ret = -ENOMEM; + mlog_errno(ret); + goto out_commit; + } + + new_bh = sb_getblk(inode->i_sb, first_blkno); + if (!new_bh) { + ret = -ENOMEM; + mlog_errno(ret); + goto out_commit; + } + ocfs2_set_new_buffer_uptodate(&new_tree->rf_ci, new_bh); + + ret = ocfs2_journal_access_rb(handle, &new_tree->rf_ci, new_bh, + OCFS2_JOURNAL_ACCESS_CREATE); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + /* Initialize ocfs2_refcount_block. */ + rb = (struct ocfs2_refcount_block *)new_bh->b_data; + memset(rb, 0, inode->i_sb->s_blocksize); + strcpy((void *)rb, OCFS2_REFCOUNT_BLOCK_SIGNATURE); + rb->rf_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot); + rb->rf_suballoc_loc = cpu_to_le64(suballoc_loc); + rb->rf_suballoc_bit = cpu_to_le16(suballoc_bit_start); + rb->rf_fs_generation = cpu_to_le32(osb->fs_generation); + rb->rf_blkno = cpu_to_le64(first_blkno); + rb->rf_count = cpu_to_le32(1); + rb->rf_records.rl_count = + cpu_to_le16(ocfs2_refcount_recs_per_rb(osb->sb)); + spin_lock(&osb->osb_lock); + rb->rf_generation = osb->s_next_generation++; + spin_unlock(&osb->osb_lock); + + ocfs2_journal_dirty(handle, new_bh); + + spin_lock(&oi->ip_lock); + oi->ip_dyn_features |= OCFS2_HAS_REFCOUNT_FL; + di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features); + di->i_refcount_loc = cpu_to_le64(first_blkno); + spin_unlock(&oi->ip_lock); + + trace_ocfs2_create_refcount_tree_blkno((unsigned long long)first_blkno); + + ocfs2_journal_dirty(handle, di_bh); + + /* + * We have to init the tree lock here since it will use + * the generation number to create it. + */ + new_tree->rf_generation = le32_to_cpu(rb->rf_generation); + ocfs2_init_refcount_tree_lock(osb, new_tree, first_blkno, + new_tree->rf_generation); + + spin_lock(&osb->osb_lock); + tree = ocfs2_find_refcount_tree(osb, first_blkno); + + /* + * We've just created a new refcount tree in this block. If + * we found a refcount tree on the ocfs2_super, it must be + * one we just deleted. We free the old tree before + * inserting the new tree. + */ + BUG_ON(tree && tree->rf_generation == new_tree->rf_generation); + if (tree) + ocfs2_erase_refcount_tree_from_list_no_lock(osb, tree); + ocfs2_insert_refcount_tree(osb, new_tree); + spin_unlock(&osb->osb_lock); + new_tree = NULL; + if (tree) + ocfs2_refcount_tree_put(tree); + +out_commit: + ocfs2_commit_trans(osb, handle); + +out: + if (new_tree) { + ocfs2_metadata_cache_exit(&new_tree->rf_ci); + kfree(new_tree); + } + + brelse(new_bh); + if (meta_ac) + ocfs2_free_alloc_context(meta_ac); + + return ret; +} + +static int ocfs2_set_refcount_tree(struct inode *inode, + struct buffer_head *di_bh, + u64 refcount_loc) +{ + int ret; + handle_t *handle = NULL; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct buffer_head *ref_root_bh = NULL; + struct ocfs2_refcount_block *rb; + struct ocfs2_refcount_tree *ref_tree; + + BUG_ON(oi->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL); + + ret = ocfs2_lock_refcount_tree(osb, refcount_loc, 1, + &ref_tree, &ref_root_bh); + if (ret) { + mlog_errno(ret); + return ret; + } + + handle = ocfs2_start_trans(osb, OCFS2_REFCOUNT_TREE_SET_CREDITS); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto out; + } + + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + ret = ocfs2_journal_access_rb(handle, &ref_tree->rf_ci, ref_root_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + rb = (struct ocfs2_refcount_block *)ref_root_bh->b_data; + le32_add_cpu(&rb->rf_count, 1); + + ocfs2_journal_dirty(handle, ref_root_bh); + + spin_lock(&oi->ip_lock); + oi->ip_dyn_features |= OCFS2_HAS_REFCOUNT_FL; + di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features); + di->i_refcount_loc = cpu_to_le64(refcount_loc); + spin_unlock(&oi->ip_lock); + ocfs2_journal_dirty(handle, di_bh); + +out_commit: + ocfs2_commit_trans(osb, handle); +out: + ocfs2_unlock_refcount_tree(osb, ref_tree, 1); + brelse(ref_root_bh); + + return ret; +} + +int ocfs2_remove_refcount_tree(struct inode *inode, struct buffer_head *di_bh) +{ + int ret, delete_tree = 0; + handle_t *handle = NULL; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_refcount_block *rb; + struct inode *alloc_inode = NULL; + struct buffer_head *alloc_bh = NULL; + struct buffer_head *blk_bh = NULL; + struct ocfs2_refcount_tree *ref_tree; + int credits = OCFS2_REFCOUNT_TREE_REMOVE_CREDITS; + u64 blk = 0, bg_blkno = 0, ref_blkno = le64_to_cpu(di->i_refcount_loc); + u16 bit = 0; + + if (!(oi->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL)) + return 0; + + BUG_ON(!ref_blkno); + ret = ocfs2_lock_refcount_tree(osb, ref_blkno, 1, &ref_tree, &blk_bh); + if (ret) { + mlog_errno(ret); + return ret; + } + + rb = (struct ocfs2_refcount_block *)blk_bh->b_data; + + /* + * If we are the last user, we need to free the block. + * So lock the allocator ahead. + */ + if (le32_to_cpu(rb->rf_count) == 1) { + blk = le64_to_cpu(rb->rf_blkno); + bit = le16_to_cpu(rb->rf_suballoc_bit); + if (rb->rf_suballoc_loc) + bg_blkno = le64_to_cpu(rb->rf_suballoc_loc); + else + bg_blkno = ocfs2_which_suballoc_group(blk, bit); + + alloc_inode = ocfs2_get_system_file_inode(osb, + EXTENT_ALLOC_SYSTEM_INODE, + le16_to_cpu(rb->rf_suballoc_slot)); + if (!alloc_inode) { + ret = -ENOMEM; + mlog_errno(ret); + goto out; + } + mutex_lock(&alloc_inode->i_mutex); + + ret = ocfs2_inode_lock(alloc_inode, &alloc_bh, 1); + if (ret) { + mlog_errno(ret); + goto out_mutex; + } + + credits += OCFS2_SUBALLOC_FREE; + } + + handle = ocfs2_start_trans(osb, credits); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto out_unlock; + } + + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + ret = ocfs2_journal_access_rb(handle, &ref_tree->rf_ci, blk_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + spin_lock(&oi->ip_lock); + oi->ip_dyn_features &= ~OCFS2_HAS_REFCOUNT_FL; + di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features); + di->i_refcount_loc = 0; + spin_unlock(&oi->ip_lock); + ocfs2_journal_dirty(handle, di_bh); + + le32_add_cpu(&rb->rf_count , -1); + ocfs2_journal_dirty(handle, blk_bh); + + if (!rb->rf_count) { + delete_tree = 1; + ocfs2_erase_refcount_tree_from_list(osb, ref_tree); + ret = ocfs2_free_suballoc_bits(handle, alloc_inode, + alloc_bh, bit, bg_blkno, 1); + if (ret) + mlog_errno(ret); + } + +out_commit: + ocfs2_commit_trans(osb, handle); +out_unlock: + if (alloc_inode) { + ocfs2_inode_unlock(alloc_inode, 1); + brelse(alloc_bh); + } +out_mutex: + if (alloc_inode) { + mutex_unlock(&alloc_inode->i_mutex); + iput(alloc_inode); + } +out: + ocfs2_unlock_refcount_tree(osb, ref_tree, 1); + if (delete_tree) + ocfs2_refcount_tree_put(ref_tree); + brelse(blk_bh); + + return ret; +} + +static void ocfs2_find_refcount_rec_in_rl(struct ocfs2_caching_info *ci, + struct buffer_head *ref_leaf_bh, + u64 cpos, unsigned int len, + struct ocfs2_refcount_rec *ret_rec, + int *index) +{ + int i = 0; + struct ocfs2_refcount_block *rb = + (struct ocfs2_refcount_block *)ref_leaf_bh->b_data; + struct ocfs2_refcount_rec *rec = NULL; + + for (; i < le16_to_cpu(rb->rf_records.rl_used); i++) { + rec = &rb->rf_records.rl_recs[i]; + + if (le64_to_cpu(rec->r_cpos) + + le32_to_cpu(rec->r_clusters) <= cpos) + continue; + else if (le64_to_cpu(rec->r_cpos) > cpos) + break; + + /* ok, cpos fail in this rec. Just return. */ + if (ret_rec) + *ret_rec = *rec; + goto out; + } + + if (ret_rec) { + /* We meet with a hole here, so fake the rec. */ + ret_rec->r_cpos = cpu_to_le64(cpos); + ret_rec->r_refcount = 0; + if (i < le16_to_cpu(rb->rf_records.rl_used) && + le64_to_cpu(rec->r_cpos) < cpos + len) + ret_rec->r_clusters = + cpu_to_le32(le64_to_cpu(rec->r_cpos) - cpos); + else + ret_rec->r_clusters = cpu_to_le32(len); + } + +out: + *index = i; +} + +/* + * Try to remove refcount tree. The mechanism is: + * 1) Check whether i_clusters == 0, if no, exit. + * 2) check whether we have i_xattr_loc in dinode. if yes, exit. + * 3) Check whether we have inline xattr stored outside, if yes, exit. + * 4) Remove the tree. + */ +int ocfs2_try_remove_refcount_tree(struct inode *inode, + struct buffer_head *di_bh) +{ + int ret; + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + + down_write(&oi->ip_xattr_sem); + down_write(&oi->ip_alloc_sem); + + if (oi->ip_clusters) + goto out; + + if ((oi->ip_dyn_features & OCFS2_HAS_XATTR_FL) && di->i_xattr_loc) + goto out; + + if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL && + ocfs2_has_inline_xattr_value_outside(inode, di)) + goto out; + + ret = ocfs2_remove_refcount_tree(inode, di_bh); + if (ret) + mlog_errno(ret); +out: + up_write(&oi->ip_alloc_sem); + up_write(&oi->ip_xattr_sem); + return 0; +} + +/* + * Find the end range for a leaf refcount block indicated by + * el->l_recs[index].e_blkno. + */ +static int ocfs2_get_refcount_cpos_end(struct ocfs2_caching_info *ci, + struct buffer_head *ref_root_bh, + struct ocfs2_extent_block *eb, + struct ocfs2_extent_list *el, + int index, u32 *cpos_end) +{ + int ret, i, subtree_root; + u32 cpos; + u64 blkno; + struct super_block *sb = ocfs2_metadata_cache_get_super(ci); + struct ocfs2_path *left_path = NULL, *right_path = NULL; + struct ocfs2_extent_tree et; + struct ocfs2_extent_list *tmp_el; + + if (index < le16_to_cpu(el->l_next_free_rec) - 1) { + /* + * We have a extent rec after index, so just use the e_cpos + * of the next extent rec. + */ + *cpos_end = le32_to_cpu(el->l_recs[index+1].e_cpos); + return 0; + } + + if (!eb || (eb && !eb->h_next_leaf_blk)) { + /* + * We are the last extent rec, so any high cpos should + * be stored in this leaf refcount block. + */ + *cpos_end = UINT_MAX; + return 0; + } + + /* + * If the extent block isn't the last one, we have to find + * the subtree root between this extent block and the next + * leaf extent block and get the corresponding e_cpos from + * the subroot. Otherwise we may corrupt the b-tree. + */ + ocfs2_init_refcount_extent_tree(&et, ci, ref_root_bh); + + left_path = ocfs2_new_path_from_et(&et); + if (!left_path) { + ret = -ENOMEM; + mlog_errno(ret); + goto out; + } + + cpos = le32_to_cpu(eb->h_list.l_recs[index].e_cpos); + ret = ocfs2_find_path(ci, left_path, cpos); + if (ret) { + mlog_errno(ret); + goto out; + } + + right_path = ocfs2_new_path_from_path(left_path); + if (!right_path) { + ret = -ENOMEM; + mlog_errno(ret); + goto out; + } + + ret = ocfs2_find_cpos_for_right_leaf(sb, left_path, &cpos); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_find_path(ci, right_path, cpos); + if (ret) { + mlog_errno(ret); + goto out; + } + + subtree_root = ocfs2_find_subtree_root(&et, left_path, + right_path); + + tmp_el = left_path->p_node[subtree_root].el; + blkno = left_path->p_node[subtree_root+1].bh->b_blocknr; + for (i = 0; i < le16_to_cpu(tmp_el->l_next_free_rec); i++) { + if (le64_to_cpu(tmp_el->l_recs[i].e_blkno) == blkno) { + *cpos_end = le32_to_cpu(tmp_el->l_recs[i+1].e_cpos); + break; + } + } + + BUG_ON(i == le16_to_cpu(tmp_el->l_next_free_rec)); + +out: + ocfs2_free_path(left_path); + ocfs2_free_path(right_path); + return ret; +} + +/* + * Given a cpos and len, try to find the refcount record which contains cpos. + * 1. If cpos can be found in one refcount record, return the record. + * 2. If cpos can't be found, return a fake record which start from cpos + * and end at a small value between cpos+len and start of the next record. + * This fake record has r_refcount = 0. + */ +static int ocfs2_get_refcount_rec(struct ocfs2_caching_info *ci, + struct buffer_head *ref_root_bh, + u64 cpos, unsigned int len, + struct ocfs2_refcount_rec *ret_rec, + int *index, + struct buffer_head **ret_bh) +{ + int ret = 0, i, found; + u32 low_cpos, uninitialized_var(cpos_end); + struct ocfs2_extent_list *el; + struct ocfs2_extent_rec *rec = NULL; + struct ocfs2_extent_block *eb = NULL; + struct buffer_head *eb_bh = NULL, *ref_leaf_bh = NULL; + struct super_block *sb = ocfs2_metadata_cache_get_super(ci); + struct ocfs2_refcount_block *rb = + (struct ocfs2_refcount_block *)ref_root_bh->b_data; + + if (!(le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL)) { + ocfs2_find_refcount_rec_in_rl(ci, ref_root_bh, cpos, len, + ret_rec, index); + *ret_bh = ref_root_bh; + get_bh(ref_root_bh); + return 0; + } + + el = &rb->rf_list; + low_cpos = cpos & OCFS2_32BIT_POS_MASK; + + if (el->l_tree_depth) { + ret = ocfs2_find_leaf(ci, el, low_cpos, &eb_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + eb = (struct ocfs2_extent_block *) eb_bh->b_data; + el = &eb->h_list; + + if (el->l_tree_depth) { + ocfs2_error(sb, + "refcount tree %llu has non zero tree " + "depth in leaf btree tree block %llu\n", + (unsigned long long)ocfs2_metadata_cache_owner(ci), + (unsigned long long)eb_bh->b_blocknr); + ret = -EROFS; + goto out; + } + } + + found = 0; + for (i = le16_to_cpu(el->l_next_free_rec) - 1; i >= 0; i--) { + rec = &el->l_recs[i]; + + if (le32_to_cpu(rec->e_cpos) <= low_cpos) { + found = 1; + break; + } + } + + if (found) { + ret = ocfs2_get_refcount_cpos_end(ci, ref_root_bh, + eb, el, i, &cpos_end); + if (ret) { + mlog_errno(ret); + goto out; + } + + if (cpos_end < low_cpos + len) + len = cpos_end - low_cpos; + } + + ret = ocfs2_read_refcount_block(ci, le64_to_cpu(rec->e_blkno), + &ref_leaf_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + ocfs2_find_refcount_rec_in_rl(ci, ref_leaf_bh, cpos, len, + ret_rec, index); + *ret_bh = ref_leaf_bh; +out: + brelse(eb_bh); + return ret; +} + +enum ocfs2_ref_rec_contig { + REF_CONTIG_NONE = 0, + REF_CONTIG_LEFT, + REF_CONTIG_RIGHT, + REF_CONTIG_LEFTRIGHT, +}; + +static enum ocfs2_ref_rec_contig + ocfs2_refcount_rec_adjacent(struct ocfs2_refcount_block *rb, + int index) +{ + if ((rb->rf_records.rl_recs[index].r_refcount == + rb->rf_records.rl_recs[index + 1].r_refcount) && + (le64_to_cpu(rb->rf_records.rl_recs[index].r_cpos) + + le32_to_cpu(rb->rf_records.rl_recs[index].r_clusters) == + le64_to_cpu(rb->rf_records.rl_recs[index + 1].r_cpos))) + return REF_CONTIG_RIGHT; + + return REF_CONTIG_NONE; +} + +static enum ocfs2_ref_rec_contig + ocfs2_refcount_rec_contig(struct ocfs2_refcount_block *rb, + int index) +{ + enum ocfs2_ref_rec_contig ret = REF_CONTIG_NONE; + + if (index < le16_to_cpu(rb->rf_records.rl_used) - 1) + ret = ocfs2_refcount_rec_adjacent(rb, index); + + if (index > 0) { + enum ocfs2_ref_rec_contig tmp; + + tmp = ocfs2_refcount_rec_adjacent(rb, index - 1); + + if (tmp == REF_CONTIG_RIGHT) { + if (ret == REF_CONTIG_RIGHT) + ret = REF_CONTIG_LEFTRIGHT; + else + ret = REF_CONTIG_LEFT; + } + } + + return ret; +} + +static void ocfs2_rotate_refcount_rec_left(struct ocfs2_refcount_block *rb, + int index) +{ + BUG_ON(rb->rf_records.rl_recs[index].r_refcount != + rb->rf_records.rl_recs[index+1].r_refcount); + + le32_add_cpu(&rb->rf_records.rl_recs[index].r_clusters, + le32_to_cpu(rb->rf_records.rl_recs[index+1].r_clusters)); + + if (index < le16_to_cpu(rb->rf_records.rl_used) - 2) + memmove(&rb->rf_records.rl_recs[index + 1], + &rb->rf_records.rl_recs[index + 2], + sizeof(struct ocfs2_refcount_rec) * + (le16_to_cpu(rb->rf_records.rl_used) - index - 2)); + + memset(&rb->rf_records.rl_recs[le16_to_cpu(rb->rf_records.rl_used) - 1], + 0, sizeof(struct ocfs2_refcount_rec)); + le16_add_cpu(&rb->rf_records.rl_used, -1); +} + +/* + * Merge the refcount rec if we are contiguous with the adjacent recs. + */ +static void ocfs2_refcount_rec_merge(struct ocfs2_refcount_block *rb, + int index) +{ + enum ocfs2_ref_rec_contig contig = + ocfs2_refcount_rec_contig(rb, index); + + if (contig == REF_CONTIG_NONE) + return; + + if (contig == REF_CONTIG_LEFT || contig == REF_CONTIG_LEFTRIGHT) { + BUG_ON(index == 0); + index--; + } + + ocfs2_rotate_refcount_rec_left(rb, index); + + if (contig == REF_CONTIG_LEFTRIGHT) + ocfs2_rotate_refcount_rec_left(rb, index); +} + +/* + * Change the refcount indexed by "index" in ref_bh. + * If refcount reaches 0, remove it. + */ +static int ocfs2_change_refcount_rec(handle_t *handle, + struct ocfs2_caching_info *ci, + struct buffer_head *ref_leaf_bh, + int index, int merge, int change) +{ + int ret; + struct ocfs2_refcount_block *rb = + (struct ocfs2_refcount_block *)ref_leaf_bh->b_data; + struct ocfs2_refcount_list *rl = &rb->rf_records; + struct ocfs2_refcount_rec *rec = &rl->rl_recs[index]; + + ret = ocfs2_journal_access_rb(handle, ci, ref_leaf_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out; + } + + trace_ocfs2_change_refcount_rec( + (unsigned long long)ocfs2_metadata_cache_owner(ci), + index, le32_to_cpu(rec->r_refcount), change); + le32_add_cpu(&rec->r_refcount, change); + + if (!rec->r_refcount) { + if (index != le16_to_cpu(rl->rl_used) - 1) { + memmove(rec, rec + 1, + (le16_to_cpu(rl->rl_used) - index - 1) * + sizeof(struct ocfs2_refcount_rec)); + memset(&rl->rl_recs[le16_to_cpu(rl->rl_used) - 1], + 0, sizeof(struct ocfs2_refcount_rec)); + } + + le16_add_cpu(&rl->rl_used, -1); + } else if (merge) + ocfs2_refcount_rec_merge(rb, index); + + ocfs2_journal_dirty(handle, ref_leaf_bh); +out: + return ret; +} + +static int ocfs2_expand_inline_ref_root(handle_t *handle, + struct ocfs2_caching_info *ci, + struct buffer_head *ref_root_bh, + struct buffer_head **ref_leaf_bh, + struct ocfs2_alloc_context *meta_ac) +{ + int ret; + u16 suballoc_bit_start; + u32 num_got; + u64 suballoc_loc, blkno; + struct super_block *sb = ocfs2_metadata_cache_get_super(ci); + struct buffer_head *new_bh = NULL; + struct ocfs2_refcount_block *new_rb; + struct ocfs2_refcount_block *root_rb = + (struct ocfs2_refcount_block *)ref_root_bh->b_data; + + ret = ocfs2_journal_access_rb(handle, ci, ref_root_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc, + &suballoc_bit_start, &num_got, + &blkno); + if (ret) { + mlog_errno(ret); + goto out; + } + + new_bh = sb_getblk(sb, blkno); + if (new_bh == NULL) { + ret = -ENOMEM; + mlog_errno(ret); + goto out; + } + ocfs2_set_new_buffer_uptodate(ci, new_bh); + + ret = ocfs2_journal_access_rb(handle, ci, new_bh, + OCFS2_JOURNAL_ACCESS_CREATE); + if (ret) { + mlog_errno(ret); + goto out; + } + + /* + * Initialize ocfs2_refcount_block. + * It should contain the same information as the old root. + * so just memcpy it and change the corresponding field. + */ + memcpy(new_bh->b_data, ref_root_bh->b_data, sb->s_blocksize); + + new_rb = (struct ocfs2_refcount_block *)new_bh->b_data; + new_rb->rf_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot); + new_rb->rf_suballoc_loc = cpu_to_le64(suballoc_loc); + new_rb->rf_suballoc_bit = cpu_to_le16(suballoc_bit_start); + new_rb->rf_blkno = cpu_to_le64(blkno); + new_rb->rf_cpos = cpu_to_le32(0); + new_rb->rf_parent = cpu_to_le64(ref_root_bh->b_blocknr); + new_rb->rf_flags = cpu_to_le32(OCFS2_REFCOUNT_LEAF_FL); + ocfs2_journal_dirty(handle, new_bh); + + /* Now change the root. */ + memset(&root_rb->rf_list, 0, sb->s_blocksize - + offsetof(struct ocfs2_refcount_block, rf_list)); + root_rb->rf_list.l_count = cpu_to_le16(ocfs2_extent_recs_per_rb(sb)); + root_rb->rf_clusters = cpu_to_le32(1); + root_rb->rf_list.l_next_free_rec = cpu_to_le16(1); + root_rb->rf_list.l_recs[0].e_blkno = cpu_to_le64(blkno); + root_rb->rf_list.l_recs[0].e_leaf_clusters = cpu_to_le16(1); + root_rb->rf_flags = cpu_to_le32(OCFS2_REFCOUNT_TREE_FL); + + ocfs2_journal_dirty(handle, ref_root_bh); + + trace_ocfs2_expand_inline_ref_root((unsigned long long)blkno, + le16_to_cpu(new_rb->rf_records.rl_used)); + + *ref_leaf_bh = new_bh; + new_bh = NULL; +out: + brelse(new_bh); + return ret; +} + +static int ocfs2_refcount_rec_no_intersect(struct ocfs2_refcount_rec *prev, + struct ocfs2_refcount_rec *next) +{ + if (ocfs2_get_ref_rec_low_cpos(prev) + le32_to_cpu(prev->r_clusters) <= + ocfs2_get_ref_rec_low_cpos(next)) + return 1; + + return 0; +} + +static int cmp_refcount_rec_by_low_cpos(const void *a, const void *b) +{ + const struct ocfs2_refcount_rec *l = a, *r = b; + u32 l_cpos = ocfs2_get_ref_rec_low_cpos(l); + u32 r_cpos = ocfs2_get_ref_rec_low_cpos(r); + + if (l_cpos > r_cpos) + return 1; + if (l_cpos < r_cpos) + return -1; + return 0; +} + +static int cmp_refcount_rec_by_cpos(const void *a, const void *b) +{ + const struct ocfs2_refcount_rec *l = a, *r = b; + u64 l_cpos = le64_to_cpu(l->r_cpos); + u64 r_cpos = le64_to_cpu(r->r_cpos); + + if (l_cpos > r_cpos) + return 1; + if (l_cpos < r_cpos) + return -1; + return 0; +} + +static void swap_refcount_rec(void *a, void *b, int size) +{ + struct ocfs2_refcount_rec *l = a, *r = b, tmp; + + tmp = *l; + *l = *r; + *r = tmp; +} + +/* + * The refcount cpos are ordered by their 64bit cpos, + * But we will use the low 32 bit to be the e_cpos in the b-tree. + * So we need to make sure that this pos isn't intersected with others. + * + * Note: The refcount block is already sorted by their low 32 bit cpos, + * So just try the middle pos first, and we will exit when we find + * the good position. + */ +static int ocfs2_find_refcount_split_pos(struct ocfs2_refcount_list *rl, + u32 *split_pos, int *split_index) +{ + int num_used = le16_to_cpu(rl->rl_used); + int delta, middle = num_used / 2; + + for (delta = 0; delta < middle; delta++) { + /* Let's check delta earlier than middle */ + if (ocfs2_refcount_rec_no_intersect( + &rl->rl_recs[middle - delta - 1], + &rl->rl_recs[middle - delta])) { + *split_index = middle - delta; + break; + } + + /* For even counts, don't walk off the end */ + if ((middle + delta + 1) == num_used) + continue; + + /* Now try delta past middle */ + if (ocfs2_refcount_rec_no_intersect( + &rl->rl_recs[middle + delta], + &rl->rl_recs[middle + delta + 1])) { + *split_index = middle + delta + 1; + break; + } + } + + if (delta >= middle) + return -ENOSPC; + + *split_pos = ocfs2_get_ref_rec_low_cpos(&rl->rl_recs[*split_index]); + return 0; +} + +static int ocfs2_divide_leaf_refcount_block(struct buffer_head *ref_leaf_bh, + struct buffer_head *new_bh, + u32 *split_cpos) +{ + int split_index = 0, num_moved, ret; + u32 cpos = 0; + struct ocfs2_refcount_block *rb = + (struct ocfs2_refcount_block *)ref_leaf_bh->b_data; + struct ocfs2_refcount_list *rl = &rb->rf_records; + struct ocfs2_refcount_block *new_rb = + (struct ocfs2_refcount_block *)new_bh->b_data; + struct ocfs2_refcount_list *new_rl = &new_rb->rf_records; + + trace_ocfs2_divide_leaf_refcount_block( + (unsigned long long)ref_leaf_bh->b_blocknr, + le16_to_cpu(rl->rl_count), le16_to_cpu(rl->rl_used)); + + /* + * XXX: Improvement later. + * If we know all the high 32 bit cpos is the same, no need to sort. + * + * In order to make the whole process safe, we do: + * 1. sort the entries by their low 32 bit cpos first so that we can + * find the split cpos easily. + * 2. call ocfs2_insert_extent to insert the new refcount block. + * 3. move the refcount rec to the new block. + * 4. sort the entries by their 64 bit cpos. + * 5. dirty the new_rb and rb. + */ + sort(&rl->rl_recs, le16_to_cpu(rl->rl_used), + sizeof(struct ocfs2_refcount_rec), + cmp_refcount_rec_by_low_cpos, swap_refcount_rec); + + ret = ocfs2_find_refcount_split_pos(rl, &cpos, &split_index); + if (ret) { + mlog_errno(ret); + return ret; + } + + new_rb->rf_cpos = cpu_to_le32(cpos); + + /* move refcount records starting from split_index to the new block. */ + num_moved = le16_to_cpu(rl->rl_used) - split_index; + memcpy(new_rl->rl_recs, &rl->rl_recs[split_index], + num_moved * sizeof(struct ocfs2_refcount_rec)); + + /*ok, remove the entries we just moved over to the other block. */ + memset(&rl->rl_recs[split_index], 0, + num_moved * sizeof(struct ocfs2_refcount_rec)); + + /* change old and new rl_used accordingly. */ + le16_add_cpu(&rl->rl_used, -num_moved); + new_rl->rl_used = cpu_to_le16(num_moved); + + sort(&rl->rl_recs, le16_to_cpu(rl->rl_used), + sizeof(struct ocfs2_refcount_rec), + cmp_refcount_rec_by_cpos, swap_refcount_rec); + + sort(&new_rl->rl_recs, le16_to_cpu(new_rl->rl_used), + sizeof(struct ocfs2_refcount_rec), + cmp_refcount_rec_by_cpos, swap_refcount_rec); + + *split_cpos = cpos; + return 0; +} + +static int ocfs2_new_leaf_refcount_block(handle_t *handle, + struct ocfs2_caching_info *ci, + struct buffer_head *ref_root_bh, + struct buffer_head *ref_leaf_bh, + struct ocfs2_alloc_context *meta_ac) +{ + int ret; + u16 suballoc_bit_start; + u32 num_got, new_cpos; + u64 suballoc_loc, blkno; + struct super_block *sb = ocfs2_metadata_cache_get_super(ci); + struct ocfs2_refcount_block *root_rb = + (struct ocfs2_refcount_block *)ref_root_bh->b_data; + struct buffer_head *new_bh = NULL; + struct ocfs2_refcount_block *new_rb; + struct ocfs2_extent_tree ref_et; + + BUG_ON(!(le32_to_cpu(root_rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL)); + + ret = ocfs2_journal_access_rb(handle, ci, ref_root_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_journal_access_rb(handle, ci, ref_leaf_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc, + &suballoc_bit_start, &num_got, + &blkno); + if (ret) { + mlog_errno(ret); + goto out; + } + + new_bh = sb_getblk(sb, blkno); + if (new_bh == NULL) { + ret = -ENOMEM; + mlog_errno(ret); + goto out; + } + ocfs2_set_new_buffer_uptodate(ci, new_bh); + + ret = ocfs2_journal_access_rb(handle, ci, new_bh, + OCFS2_JOURNAL_ACCESS_CREATE); + if (ret) { + mlog_errno(ret); + goto out; + } + + /* Initialize ocfs2_refcount_block. */ + new_rb = (struct ocfs2_refcount_block *)new_bh->b_data; + memset(new_rb, 0, sb->s_blocksize); + strcpy((void *)new_rb, OCFS2_REFCOUNT_BLOCK_SIGNATURE); + new_rb->rf_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot); + new_rb->rf_suballoc_loc = cpu_to_le64(suballoc_loc); + new_rb->rf_suballoc_bit = cpu_to_le16(suballoc_bit_start); + new_rb->rf_fs_generation = cpu_to_le32(OCFS2_SB(sb)->fs_generation); + new_rb->rf_blkno = cpu_to_le64(blkno); + new_rb->rf_parent = cpu_to_le64(ref_root_bh->b_blocknr); + new_rb->rf_flags = cpu_to_le32(OCFS2_REFCOUNT_LEAF_FL); + new_rb->rf_records.rl_count = + cpu_to_le16(ocfs2_refcount_recs_per_rb(sb)); + new_rb->rf_generation = root_rb->rf_generation; + + ret = ocfs2_divide_leaf_refcount_block(ref_leaf_bh, new_bh, &new_cpos); + if (ret) { + mlog_errno(ret); + goto out; + } + + ocfs2_journal_dirty(handle, ref_leaf_bh); + ocfs2_journal_dirty(handle, new_bh); + + ocfs2_init_refcount_extent_tree(&ref_et, ci, ref_root_bh); + + trace_ocfs2_new_leaf_refcount_block( + (unsigned long long)new_bh->b_blocknr, new_cpos); + + /* Insert the new leaf block with the specific offset cpos. */ + ret = ocfs2_insert_extent(handle, &ref_et, new_cpos, new_bh->b_blocknr, + 1, 0, meta_ac); + if (ret) + mlog_errno(ret); + +out: + brelse(new_bh); + return ret; +} + +static int ocfs2_expand_refcount_tree(handle_t *handle, + struct ocfs2_caching_info *ci, + struct buffer_head *ref_root_bh, + struct buffer_head *ref_leaf_bh, + struct ocfs2_alloc_context *meta_ac) +{ + int ret; + struct buffer_head *expand_bh = NULL; + + if (ref_root_bh == ref_leaf_bh) { + /* + * the old root bh hasn't been expanded to a b-tree, + * so expand it first. + */ + ret = ocfs2_expand_inline_ref_root(handle, ci, ref_root_bh, + &expand_bh, meta_ac); + if (ret) { + mlog_errno(ret); + goto out; + } + } else { + expand_bh = ref_leaf_bh; + get_bh(expand_bh); + } + + + /* Now add a new refcount block into the tree.*/ + ret = ocfs2_new_leaf_refcount_block(handle, ci, ref_root_bh, + expand_bh, meta_ac); + if (ret) + mlog_errno(ret); +out: + brelse(expand_bh); + return ret; +} + +/* + * Adjust the extent rec in b-tree representing ref_leaf_bh. + * + * Only called when we have inserted a new refcount rec at index 0 + * which means ocfs2_extent_rec.e_cpos may need some change. + */ +static int ocfs2_adjust_refcount_rec(handle_t *handle, + struct ocfs2_caching_info *ci, + struct buffer_head *ref_root_bh, + struct buffer_head *ref_leaf_bh, + struct ocfs2_refcount_rec *rec) +{ + int ret = 0, i; + u32 new_cpos, old_cpos; + struct ocfs2_path *path = NULL; + struct ocfs2_extent_tree et; + struct ocfs2_refcount_block *rb = + (struct ocfs2_refcount_block *)ref_root_bh->b_data; + struct ocfs2_extent_list *el; + + if (!(le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL)) + goto out; + + rb = (struct ocfs2_refcount_block *)ref_leaf_bh->b_data; + old_cpos = le32_to_cpu(rb->rf_cpos); + new_cpos = le64_to_cpu(rec->r_cpos) & OCFS2_32BIT_POS_MASK; + if (old_cpos <= new_cpos) + goto out; + + ocfs2_init_refcount_extent_tree(&et, ci, ref_root_bh); + + path = ocfs2_new_path_from_et(&et); + if (!path) { + ret = -ENOMEM; + mlog_errno(ret); + goto out; + } + + ret = ocfs2_find_path(ci, path, old_cpos); + if (ret) { + mlog_errno(ret); + goto out; + } + + /* + * 2 more credits, one for the leaf refcount block, one for + * the extent block contains the extent rec. + */ + ret = ocfs2_extend_trans(handle, 2); + if (ret < 0) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_journal_access_rb(handle, ci, ref_leaf_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret < 0) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_journal_access_eb(handle, ci, path_leaf_bh(path), + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret < 0) { + mlog_errno(ret); + goto out; + } + + /* change the leaf extent block first. */ + el = path_leaf_el(path); + + for (i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) + if (le32_to_cpu(el->l_recs[i].e_cpos) == old_cpos) + break; + + BUG_ON(i == le16_to_cpu(el->l_next_free_rec)); + + el->l_recs[i].e_cpos = cpu_to_le32(new_cpos); + + /* change the r_cpos in the leaf block. */ + rb->rf_cpos = cpu_to_le32(new_cpos); + + ocfs2_journal_dirty(handle, path_leaf_bh(path)); + ocfs2_journal_dirty(handle, ref_leaf_bh); + +out: + ocfs2_free_path(path); + return ret; +} + +static int ocfs2_insert_refcount_rec(handle_t *handle, + struct ocfs2_caching_info *ci, + struct buffer_head *ref_root_bh, + struct buffer_head *ref_leaf_bh, + struct ocfs2_refcount_rec *rec, + int index, int merge, + struct ocfs2_alloc_context *meta_ac) +{ + int ret; + struct ocfs2_refcount_block *rb = + (struct ocfs2_refcount_block *)ref_leaf_bh->b_data; + struct ocfs2_refcount_list *rf_list = &rb->rf_records; + struct buffer_head *new_bh = NULL; + + BUG_ON(le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL); + + if (rf_list->rl_used == rf_list->rl_count) { + u64 cpos = le64_to_cpu(rec->r_cpos); + u32 len = le32_to_cpu(rec->r_clusters); + + ret = ocfs2_expand_refcount_tree(handle, ci, ref_root_bh, + ref_leaf_bh, meta_ac); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_get_refcount_rec(ci, ref_root_bh, + cpos, len, NULL, &index, + &new_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + ref_leaf_bh = new_bh; + rb = (struct ocfs2_refcount_block *)ref_leaf_bh->b_data; + rf_list = &rb->rf_records; + } + + ret = ocfs2_journal_access_rb(handle, ci, ref_leaf_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out; + } + + if (index < le16_to_cpu(rf_list->rl_used)) + memmove(&rf_list->rl_recs[index + 1], + &rf_list->rl_recs[index], + (le16_to_cpu(rf_list->rl_used) - index) * + sizeof(struct ocfs2_refcount_rec)); + + trace_ocfs2_insert_refcount_rec( + (unsigned long long)ref_leaf_bh->b_blocknr, index, + (unsigned long long)le64_to_cpu(rec->r_cpos), + le32_to_cpu(rec->r_clusters), le32_to_cpu(rec->r_refcount)); + + rf_list->rl_recs[index] = *rec; + + le16_add_cpu(&rf_list->rl_used, 1); + + if (merge) + ocfs2_refcount_rec_merge(rb, index); + + ocfs2_journal_dirty(handle, ref_leaf_bh); + + if (index == 0) { + ret = ocfs2_adjust_refcount_rec(handle, ci, + ref_root_bh, + ref_leaf_bh, rec); + if (ret) + mlog_errno(ret); + } +out: + brelse(new_bh); + return ret; +} + +/* + * Split the refcount_rec indexed by "index" in ref_leaf_bh. + * This is much simple than our b-tree code. + * split_rec is the new refcount rec we want to insert. + * If split_rec->r_refcount > 0, we are changing the refcount(in case we + * increase refcount or decrease a refcount to non-zero). + * If split_rec->r_refcount == 0, we are punching a hole in current refcount + * rec( in case we decrease a refcount to zero). + */ +static int ocfs2_split_refcount_rec(handle_t *handle, + struct ocfs2_caching_info *ci, + struct buffer_head *ref_root_bh, + struct buffer_head *ref_leaf_bh, + struct ocfs2_refcount_rec *split_rec, + int index, int merge, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_cached_dealloc_ctxt *dealloc) +{ + int ret, recs_need; + u32 len; + struct ocfs2_refcount_block *rb = + (struct ocfs2_refcount_block *)ref_leaf_bh->b_data; + struct ocfs2_refcount_list *rf_list = &rb->rf_records; + struct ocfs2_refcount_rec *orig_rec = &rf_list->rl_recs[index]; + struct ocfs2_refcount_rec *tail_rec = NULL; + struct buffer_head *new_bh = NULL; + + BUG_ON(le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL); + + trace_ocfs2_split_refcount_rec(le64_to_cpu(orig_rec->r_cpos), + le32_to_cpu(orig_rec->r_clusters), + le32_to_cpu(orig_rec->r_refcount), + le64_to_cpu(split_rec->r_cpos), + le32_to_cpu(split_rec->r_clusters), + le32_to_cpu(split_rec->r_refcount)); + + /* + * If we just need to split the header or tail clusters, + * no more recs are needed, just split is OK. + * Otherwise we at least need one new recs. + */ + if (!split_rec->r_refcount && + (split_rec->r_cpos == orig_rec->r_cpos || + le64_to_cpu(split_rec->r_cpos) + + le32_to_cpu(split_rec->r_clusters) == + le64_to_cpu(orig_rec->r_cpos) + le32_to_cpu(orig_rec->r_clusters))) + recs_need = 0; + else + recs_need = 1; + + /* + * We need one more rec if we split in the middle and the new rec have + * some refcount in it. + */ + if (split_rec->r_refcount && + (split_rec->r_cpos != orig_rec->r_cpos && + le64_to_cpu(split_rec->r_cpos) + + le32_to_cpu(split_rec->r_clusters) != + le64_to_cpu(orig_rec->r_cpos) + le32_to_cpu(orig_rec->r_clusters))) + recs_need++; + + /* If the leaf block don't have enough record, expand it. */ + if (le16_to_cpu(rf_list->rl_used) + recs_need > + le16_to_cpu(rf_list->rl_count)) { + struct ocfs2_refcount_rec tmp_rec; + u64 cpos = le64_to_cpu(orig_rec->r_cpos); + len = le32_to_cpu(orig_rec->r_clusters); + ret = ocfs2_expand_refcount_tree(handle, ci, ref_root_bh, + ref_leaf_bh, meta_ac); + if (ret) { + mlog_errno(ret); + goto out; + } + + /* + * We have to re-get it since now cpos may be moved to + * another leaf block. + */ + ret = ocfs2_get_refcount_rec(ci, ref_root_bh, + cpos, len, &tmp_rec, &index, + &new_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + ref_leaf_bh = new_bh; + rb = (struct ocfs2_refcount_block *)ref_leaf_bh->b_data; + rf_list = &rb->rf_records; + orig_rec = &rf_list->rl_recs[index]; + } + + ret = ocfs2_journal_access_rb(handle, ci, ref_leaf_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out; + } + + /* + * We have calculated out how many new records we need and store + * in recs_need, so spare enough space first by moving the records + * after "index" to the end. + */ + if (index != le16_to_cpu(rf_list->rl_used) - 1) + memmove(&rf_list->rl_recs[index + 1 + recs_need], + &rf_list->rl_recs[index + 1], + (le16_to_cpu(rf_list->rl_used) - index - 1) * + sizeof(struct ocfs2_refcount_rec)); + + len = (le64_to_cpu(orig_rec->r_cpos) + + le32_to_cpu(orig_rec->r_clusters)) - + (le64_to_cpu(split_rec->r_cpos) + + le32_to_cpu(split_rec->r_clusters)); + + /* + * If we have "len", the we will split in the tail and move it + * to the end of the space we have just spared. + */ + if (len) { + tail_rec = &rf_list->rl_recs[index + recs_need]; + + memcpy(tail_rec, orig_rec, sizeof(struct ocfs2_refcount_rec)); + le64_add_cpu(&tail_rec->r_cpos, + le32_to_cpu(tail_rec->r_clusters) - len); + tail_rec->r_clusters = cpu_to_le32(len); + } + + /* + * If the split pos isn't the same as the original one, we need to + * split in the head. + * + * Note: We have the chance that split_rec.r_refcount = 0, + * recs_need = 0 and len > 0, which means we just cut the head from + * the orig_rec and in that case we have done some modification in + * orig_rec above, so the check for r_cpos is faked. + */ + if (split_rec->r_cpos != orig_rec->r_cpos && tail_rec != orig_rec) { + len = le64_to_cpu(split_rec->r_cpos) - + le64_to_cpu(orig_rec->r_cpos); + orig_rec->r_clusters = cpu_to_le32(len); + index++; + } + + le16_add_cpu(&rf_list->rl_used, recs_need); + + if (split_rec->r_refcount) { + rf_list->rl_recs[index] = *split_rec; + trace_ocfs2_split_refcount_rec_insert( + (unsigned long long)ref_leaf_bh->b_blocknr, index, + (unsigned long long)le64_to_cpu(split_rec->r_cpos), + le32_to_cpu(split_rec->r_clusters), + le32_to_cpu(split_rec->r_refcount)); + + if (merge) + ocfs2_refcount_rec_merge(rb, index); + } + + ocfs2_journal_dirty(handle, ref_leaf_bh); + +out: + brelse(new_bh); + return ret; +} + +static int __ocfs2_increase_refcount(handle_t *handle, + struct ocfs2_caching_info *ci, + struct buffer_head *ref_root_bh, + u64 cpos, u32 len, int merge, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_cached_dealloc_ctxt *dealloc) +{ + int ret = 0, index; + struct buffer_head *ref_leaf_bh = NULL; + struct ocfs2_refcount_rec rec; + unsigned int set_len = 0; + + trace_ocfs2_increase_refcount_begin( + (unsigned long long)ocfs2_metadata_cache_owner(ci), + (unsigned long long)cpos, len); + + while (len) { + ret = ocfs2_get_refcount_rec(ci, ref_root_bh, + cpos, len, &rec, &index, + &ref_leaf_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + set_len = le32_to_cpu(rec.r_clusters); + + /* + * Here we may meet with 3 situations: + * + * 1. If we find an already existing record, and the length + * is the same, cool, we just need to increase the r_refcount + * and it is OK. + * 2. If we find a hole, just insert it with r_refcount = 1. + * 3. If we are in the middle of one extent record, split + * it. + */ + if (rec.r_refcount && le64_to_cpu(rec.r_cpos) == cpos && + set_len <= len) { + trace_ocfs2_increase_refcount_change( + (unsigned long long)cpos, set_len, + le32_to_cpu(rec.r_refcount)); + ret = ocfs2_change_refcount_rec(handle, ci, + ref_leaf_bh, index, + merge, 1); + if (ret) { + mlog_errno(ret); + goto out; + } + } else if (!rec.r_refcount) { + rec.r_refcount = cpu_to_le32(1); + + trace_ocfs2_increase_refcount_insert( + (unsigned long long)le64_to_cpu(rec.r_cpos), + set_len); + ret = ocfs2_insert_refcount_rec(handle, ci, ref_root_bh, + ref_leaf_bh, + &rec, index, + merge, meta_ac); + if (ret) { + mlog_errno(ret); + goto out; + } + } else { + set_len = min((u64)(cpos + len), + le64_to_cpu(rec.r_cpos) + set_len) - cpos; + rec.r_cpos = cpu_to_le64(cpos); + rec.r_clusters = cpu_to_le32(set_len); + le32_add_cpu(&rec.r_refcount, 1); + + trace_ocfs2_increase_refcount_split( + (unsigned long long)le64_to_cpu(rec.r_cpos), + set_len, le32_to_cpu(rec.r_refcount)); + ret = ocfs2_split_refcount_rec(handle, ci, + ref_root_bh, ref_leaf_bh, + &rec, index, merge, + meta_ac, dealloc); + if (ret) { + mlog_errno(ret); + goto out; + } + } + + cpos += set_len; + len -= set_len; + brelse(ref_leaf_bh); + ref_leaf_bh = NULL; + } + +out: + brelse(ref_leaf_bh); + return ret; +} + +static int ocfs2_remove_refcount_extent(handle_t *handle, + struct ocfs2_caching_info *ci, + struct buffer_head *ref_root_bh, + struct buffer_head *ref_leaf_bh, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_cached_dealloc_ctxt *dealloc) +{ + int ret; + struct super_block *sb = ocfs2_metadata_cache_get_super(ci); + struct ocfs2_refcount_block *rb = + (struct ocfs2_refcount_block *)ref_leaf_bh->b_data; + struct ocfs2_extent_tree et; + + BUG_ON(rb->rf_records.rl_used); + + trace_ocfs2_remove_refcount_extent( + (unsigned long long)ocfs2_metadata_cache_owner(ci), + (unsigned long long)ref_leaf_bh->b_blocknr, + le32_to_cpu(rb->rf_cpos)); + + ocfs2_init_refcount_extent_tree(&et, ci, ref_root_bh); + ret = ocfs2_remove_extent(handle, &et, le32_to_cpu(rb->rf_cpos), + 1, meta_ac, dealloc); + if (ret) { + mlog_errno(ret); + goto out; + } + + ocfs2_remove_from_cache(ci, ref_leaf_bh); + + /* + * add the freed block to the dealloc so that it will be freed + * when we run dealloc. + */ + ret = ocfs2_cache_block_dealloc(dealloc, EXTENT_ALLOC_SYSTEM_INODE, + le16_to_cpu(rb->rf_suballoc_slot), + le64_to_cpu(rb->rf_suballoc_loc), + le64_to_cpu(rb->rf_blkno), + le16_to_cpu(rb->rf_suballoc_bit)); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_journal_access_rb(handle, ci, ref_root_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out; + } + + rb = (struct ocfs2_refcount_block *)ref_root_bh->b_data; + + le32_add_cpu(&rb->rf_clusters, -1); + + /* + * check whether we need to restore the root refcount block if + * there is no leaf extent block at atll. + */ + if (!rb->rf_list.l_next_free_rec) { + BUG_ON(rb->rf_clusters); + + trace_ocfs2_restore_refcount_block( + (unsigned long long)ref_root_bh->b_blocknr); + + rb->rf_flags = 0; + rb->rf_parent = 0; + rb->rf_cpos = 0; + memset(&rb->rf_records, 0, sb->s_blocksize - + offsetof(struct ocfs2_refcount_block, rf_records)); + rb->rf_records.rl_count = + cpu_to_le16(ocfs2_refcount_recs_per_rb(sb)); + } + + ocfs2_journal_dirty(handle, ref_root_bh); + +out: + return ret; +} + +int ocfs2_increase_refcount(handle_t *handle, + struct ocfs2_caching_info *ci, + struct buffer_head *ref_root_bh, + u64 cpos, u32 len, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_cached_dealloc_ctxt *dealloc) +{ + return __ocfs2_increase_refcount(handle, ci, ref_root_bh, + cpos, len, 1, + meta_ac, dealloc); +} + +static int ocfs2_decrease_refcount_rec(handle_t *handle, + struct ocfs2_caching_info *ci, + struct buffer_head *ref_root_bh, + struct buffer_head *ref_leaf_bh, + int index, u64 cpos, unsigned int len, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_cached_dealloc_ctxt *dealloc) +{ + int ret; + struct ocfs2_refcount_block *rb = + (struct ocfs2_refcount_block *)ref_leaf_bh->b_data; + struct ocfs2_refcount_rec *rec = &rb->rf_records.rl_recs[index]; + + BUG_ON(cpos < le64_to_cpu(rec->r_cpos)); + BUG_ON(cpos + len > + le64_to_cpu(rec->r_cpos) + le32_to_cpu(rec->r_clusters)); + + trace_ocfs2_decrease_refcount_rec( + (unsigned long long)ocfs2_metadata_cache_owner(ci), + (unsigned long long)cpos, len); + + if (cpos == le64_to_cpu(rec->r_cpos) && + len == le32_to_cpu(rec->r_clusters)) + ret = ocfs2_change_refcount_rec(handle, ci, + ref_leaf_bh, index, 1, -1); + else { + struct ocfs2_refcount_rec split = *rec; + split.r_cpos = cpu_to_le64(cpos); + split.r_clusters = cpu_to_le32(len); + + le32_add_cpu(&split.r_refcount, -1); + + ret = ocfs2_split_refcount_rec(handle, ci, + ref_root_bh, ref_leaf_bh, + &split, index, 1, + meta_ac, dealloc); + } + + if (ret) { + mlog_errno(ret); + goto out; + } + + /* Remove the leaf refcount block if it contains no refcount record. */ + if (!rb->rf_records.rl_used && ref_leaf_bh != ref_root_bh) { + ret = ocfs2_remove_refcount_extent(handle, ci, ref_root_bh, + ref_leaf_bh, meta_ac, + dealloc); + if (ret) + mlog_errno(ret); + } + +out: + return ret; +} + +static int __ocfs2_decrease_refcount(handle_t *handle, + struct ocfs2_caching_info *ci, + struct buffer_head *ref_root_bh, + u64 cpos, u32 len, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_cached_dealloc_ctxt *dealloc, + int delete) +{ + int ret = 0, index = 0; + struct ocfs2_refcount_rec rec; + unsigned int r_count = 0, r_len; + struct super_block *sb = ocfs2_metadata_cache_get_super(ci); + struct buffer_head *ref_leaf_bh = NULL; + + trace_ocfs2_decrease_refcount( + (unsigned long long)ocfs2_metadata_cache_owner(ci), + (unsigned long long)cpos, len, delete); + + while (len) { + ret = ocfs2_get_refcount_rec(ci, ref_root_bh, + cpos, len, &rec, &index, + &ref_leaf_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + r_count = le32_to_cpu(rec.r_refcount); + BUG_ON(r_count == 0); + if (!delete) + BUG_ON(r_count > 1); + + r_len = min((u64)(cpos + len), le64_to_cpu(rec.r_cpos) + + le32_to_cpu(rec.r_clusters)) - cpos; + + ret = ocfs2_decrease_refcount_rec(handle, ci, ref_root_bh, + ref_leaf_bh, index, + cpos, r_len, + meta_ac, dealloc); + if (ret) { + mlog_errno(ret); + goto out; + } + + if (le32_to_cpu(rec.r_refcount) == 1 && delete) { + ret = ocfs2_cache_cluster_dealloc(dealloc, + ocfs2_clusters_to_blocks(sb, cpos), + r_len); + if (ret) { + mlog_errno(ret); + goto out; + } + } + + cpos += r_len; + len -= r_len; + brelse(ref_leaf_bh); + ref_leaf_bh = NULL; + } + +out: + brelse(ref_leaf_bh); + return ret; +} + +/* Caller must hold refcount tree lock. */ +int ocfs2_decrease_refcount(struct inode *inode, + handle_t *handle, u32 cpos, u32 len, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_cached_dealloc_ctxt *dealloc, + int delete) +{ + int ret; + u64 ref_blkno; + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct buffer_head *ref_root_bh = NULL; + struct ocfs2_refcount_tree *tree; + + BUG_ON(!(oi->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL)); + + ret = ocfs2_get_refcount_block(inode, &ref_blkno); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_get_refcount_tree(OCFS2_SB(inode->i_sb), ref_blkno, &tree); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_read_refcount_block(&tree->rf_ci, tree->rf_blkno, + &ref_root_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = __ocfs2_decrease_refcount(handle, &tree->rf_ci, ref_root_bh, + cpos, len, meta_ac, dealloc, delete); + if (ret) + mlog_errno(ret); +out: + brelse(ref_root_bh); + return ret; +} + +/* + * Mark the already-existing extent at cpos as refcounted for len clusters. + * This adds the refcount extent flag. + * + * If the existing extent is larger than the request, initiate a + * split. An attempt will be made at merging with adjacent extents. + * + * The caller is responsible for passing down meta_ac if we'll need it. + */ +static int ocfs2_mark_extent_refcounted(struct inode *inode, + struct ocfs2_extent_tree *et, + handle_t *handle, u32 cpos, + u32 len, u32 phys, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_cached_dealloc_ctxt *dealloc) +{ + int ret; + + trace_ocfs2_mark_extent_refcounted(OCFS2_I(inode)->ip_blkno, + cpos, len, phys); + + if (!ocfs2_refcount_tree(OCFS2_SB(inode->i_sb))) { + ocfs2_error(inode->i_sb, "Inode %lu want to use refcount " + "tree, but the feature bit is not set in the " + "super block.", inode->i_ino); + ret = -EROFS; + goto out; + } + + ret = ocfs2_change_extent_flag(handle, et, cpos, + len, phys, meta_ac, dealloc, + OCFS2_EXT_REFCOUNTED, 0); + if (ret) + mlog_errno(ret); + +out: + return ret; +} + +/* + * Given some contiguous physical clusters, calculate what we need + * for modifying their refcount. + */ +static int ocfs2_calc_refcount_meta_credits(struct super_block *sb, + struct ocfs2_caching_info *ci, + struct buffer_head *ref_root_bh, + u64 start_cpos, + u32 clusters, + int *meta_add, + int *credits) +{ + int ret = 0, index, ref_blocks = 0, recs_add = 0; + u64 cpos = start_cpos; + struct ocfs2_refcount_block *rb; + struct ocfs2_refcount_rec rec; + struct buffer_head *ref_leaf_bh = NULL, *prev_bh = NULL; + u32 len; + + while (clusters) { + ret = ocfs2_get_refcount_rec(ci, ref_root_bh, + cpos, clusters, &rec, + &index, &ref_leaf_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + if (ref_leaf_bh != prev_bh) { + /* + * Now we encounter a new leaf block, so calculate + * whether we need to extend the old leaf. + */ + if (prev_bh) { + rb = (struct ocfs2_refcount_block *) + prev_bh->b_data; + + if (le16_to_cpu(rb->rf_records.rl_used) + + recs_add > + le16_to_cpu(rb->rf_records.rl_count)) + ref_blocks++; + } + + recs_add = 0; + *credits += 1; + brelse(prev_bh); + prev_bh = ref_leaf_bh; + get_bh(prev_bh); + } + + rb = (struct ocfs2_refcount_block *)ref_leaf_bh->b_data; + + trace_ocfs2_calc_refcount_meta_credits_iterate( + recs_add, (unsigned long long)cpos, clusters, + (unsigned long long)le64_to_cpu(rec.r_cpos), + le32_to_cpu(rec.r_clusters), + le32_to_cpu(rec.r_refcount), index); + + len = min((u64)cpos + clusters, le64_to_cpu(rec.r_cpos) + + le32_to_cpu(rec.r_clusters)) - cpos; + /* + * We record all the records which will be inserted to the + * same refcount block, so that we can tell exactly whether + * we need a new refcount block or not. + * + * If we will insert a new one, this is easy and only happens + * during adding refcounted flag to the extent, so we don't + * have a chance of spliting. We just need one record. + * + * If the refcount rec already exists, that would be a little + * complicated. we may have to: + * 1) split at the beginning if the start pos isn't aligned. + * we need 1 more record in this case. + * 2) split int the end if the end pos isn't aligned. + * we need 1 more record in this case. + * 3) split in the middle because of file system fragmentation. + * we need 2 more records in this case(we can't detect this + * beforehand, so always think of the worst case). + */ + if (rec.r_refcount) { + recs_add += 2; + /* Check whether we need a split at the beginning. */ + if (cpos == start_cpos && + cpos != le64_to_cpu(rec.r_cpos)) + recs_add++; + + /* Check whether we need a split in the end. */ + if (cpos + clusters < le64_to_cpu(rec.r_cpos) + + le32_to_cpu(rec.r_clusters)) + recs_add++; + } else + recs_add++; + + brelse(ref_leaf_bh); + ref_leaf_bh = NULL; + clusters -= len; + cpos += len; + } + + if (prev_bh) { + rb = (struct ocfs2_refcount_block *)prev_bh->b_data; + + if (le16_to_cpu(rb->rf_records.rl_used) + recs_add > + le16_to_cpu(rb->rf_records.rl_count)) + ref_blocks++; + + *credits += 1; + } + + if (!ref_blocks) + goto out; + + *meta_add += ref_blocks; + *credits += ref_blocks; + + /* + * So we may need ref_blocks to insert into the tree. + * That also means we need to change the b-tree and add that number + * of records since we never merge them. + * We need one more block for expansion since the new created leaf + * block is also full and needs split. + */ + rb = (struct ocfs2_refcount_block *)ref_root_bh->b_data; + if (le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL) { + struct ocfs2_extent_tree et; + + ocfs2_init_refcount_extent_tree(&et, ci, ref_root_bh); + *meta_add += ocfs2_extend_meta_needed(et.et_root_el); + *credits += ocfs2_calc_extend_credits(sb, + et.et_root_el); + } else { + *credits += OCFS2_EXPAND_REFCOUNT_TREE_CREDITS; + *meta_add += 1; + } + +out: + + trace_ocfs2_calc_refcount_meta_credits( + (unsigned long long)start_cpos, clusters, + *meta_add, *credits); + brelse(ref_leaf_bh); + brelse(prev_bh); + return ret; +} + +/* + * For refcount tree, we will decrease some contiguous clusters + * refcount count, so just go through it to see how many blocks + * we gonna touch and whether we need to create new blocks. + * + * Normally the refcount blocks store these refcount should be + * contiguous also, so that we can get the number easily. + * We will at most add split 2 refcount records and 2 more + * refcount blocks, so just check it in a rough way. + * + * Caller must hold refcount tree lock. + */ +int ocfs2_prepare_refcount_change_for_del(struct inode *inode, + u64 refcount_loc, + u64 phys_blkno, + u32 clusters, + int *credits, + int *ref_blocks) +{ + int ret; + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct buffer_head *ref_root_bh = NULL; + struct ocfs2_refcount_tree *tree; + u64 start_cpos = ocfs2_blocks_to_clusters(inode->i_sb, phys_blkno); + + if (!ocfs2_refcount_tree(OCFS2_SB(inode->i_sb))) { + ocfs2_error(inode->i_sb, "Inode %lu want to use refcount " + "tree, but the feature bit is not set in the " + "super block.", inode->i_ino); + ret = -EROFS; + goto out; + } + + BUG_ON(!(oi->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL)); + + ret = ocfs2_get_refcount_tree(OCFS2_SB(inode->i_sb), + refcount_loc, &tree); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_read_refcount_block(&tree->rf_ci, refcount_loc, + &ref_root_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_calc_refcount_meta_credits(inode->i_sb, + &tree->rf_ci, + ref_root_bh, + start_cpos, clusters, + ref_blocks, credits); + if (ret) { + mlog_errno(ret); + goto out; + } + + trace_ocfs2_prepare_refcount_change_for_del(*ref_blocks, *credits); + +out: + brelse(ref_root_bh); + return ret; +} + +#define MAX_CONTIG_BYTES 1048576 + +static inline unsigned int ocfs2_cow_contig_clusters(struct super_block *sb) +{ + return ocfs2_clusters_for_bytes(sb, MAX_CONTIG_BYTES); +} + +static inline unsigned int ocfs2_cow_contig_mask(struct super_block *sb) +{ + return ~(ocfs2_cow_contig_clusters(sb) - 1); +} + +/* + * Given an extent that starts at 'start' and an I/O that starts at 'cpos', + * find an offset (start + (n * contig_clusters)) that is closest to cpos + * while still being less than or equal to it. + * + * The goal is to break the extent at a multiple of contig_clusters. + */ +static inline unsigned int ocfs2_cow_align_start(struct super_block *sb, + unsigned int start, + unsigned int cpos) +{ + BUG_ON(start > cpos); + + return start + ((cpos - start) & ocfs2_cow_contig_mask(sb)); +} + +/* + * Given a cluster count of len, pad it out so that it is a multiple + * of contig_clusters. + */ +static inline unsigned int ocfs2_cow_align_length(struct super_block *sb, + unsigned int len) +{ + unsigned int padded = + (len + (ocfs2_cow_contig_clusters(sb) - 1)) & + ocfs2_cow_contig_mask(sb); + + /* Did we wrap? */ + if (padded < len) + padded = UINT_MAX; + + return padded; +} + +/* + * Calculate out the start and number of virtual clusters we need to to CoW. + * + * cpos is vitual start cluster position we want to do CoW in a + * file and write_len is the cluster length. + * max_cpos is the place where we want to stop CoW intentionally. + * + * Normal we will start CoW from the beginning of extent record cotaining cpos. + * We try to break up extents on boundaries of MAX_CONTIG_BYTES so that we + * get good I/O from the resulting extent tree. + */ +static int ocfs2_refcount_cal_cow_clusters(struct inode *inode, + struct ocfs2_extent_list *el, + u32 cpos, + u32 write_len, + u32 max_cpos, + u32 *cow_start, + u32 *cow_len) +{ + int ret = 0; + int tree_height = le16_to_cpu(el->l_tree_depth), i; + struct buffer_head *eb_bh = NULL; + struct ocfs2_extent_block *eb = NULL; + struct ocfs2_extent_rec *rec; + unsigned int want_clusters, rec_end = 0; + int contig_clusters = ocfs2_cow_contig_clusters(inode->i_sb); + int leaf_clusters; + + BUG_ON(cpos + write_len > max_cpos); + + if (tree_height > 0) { + ret = ocfs2_find_leaf(INODE_CACHE(inode), el, cpos, &eb_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + eb = (struct ocfs2_extent_block *) eb_bh->b_data; + el = &eb->h_list; + + if (el->l_tree_depth) { + ocfs2_error(inode->i_sb, + "Inode %lu has non zero tree depth in " + "leaf block %llu\n", inode->i_ino, + (unsigned long long)eb_bh->b_blocknr); + ret = -EROFS; + goto out; + } + } + + *cow_len = 0; + for (i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) { + rec = &el->l_recs[i]; + + if (ocfs2_is_empty_extent(rec)) { + mlog_bug_on_msg(i != 0, "Inode %lu has empty record in " + "index %d\n", inode->i_ino, i); + continue; + } + + if (le32_to_cpu(rec->e_cpos) + + le16_to_cpu(rec->e_leaf_clusters) <= cpos) + continue; + + if (*cow_len == 0) { + /* + * We should find a refcounted record in the + * first pass. + */ + BUG_ON(!(rec->e_flags & OCFS2_EXT_REFCOUNTED)); + *cow_start = le32_to_cpu(rec->e_cpos); + } + + /* + * If we encounter a hole, a non-refcounted record or + * pass the max_cpos, stop the search. + */ + if ((!(rec->e_flags & OCFS2_EXT_REFCOUNTED)) || + (*cow_len && rec_end != le32_to_cpu(rec->e_cpos)) || + (max_cpos <= le32_to_cpu(rec->e_cpos))) + break; + + leaf_clusters = le16_to_cpu(rec->e_leaf_clusters); + rec_end = le32_to_cpu(rec->e_cpos) + leaf_clusters; + if (rec_end > max_cpos) { + rec_end = max_cpos; + leaf_clusters = rec_end - le32_to_cpu(rec->e_cpos); + } + + /* + * How many clusters do we actually need from + * this extent? First we see how many we actually + * need to complete the write. If that's smaller + * than contig_clusters, we try for contig_clusters. + */ + if (!*cow_len) + want_clusters = write_len; + else + want_clusters = (cpos + write_len) - + (*cow_start + *cow_len); + if (want_clusters < contig_clusters) + want_clusters = contig_clusters; + + /* + * If the write does not cover the whole extent, we + * need to calculate how we're going to split the extent. + * We try to do it on contig_clusters boundaries. + * + * Any extent smaller than contig_clusters will be + * CoWed in its entirety. + */ + if (leaf_clusters <= contig_clusters) + *cow_len += leaf_clusters; + else if (*cow_len || (*cow_start == cpos)) { + /* + * This extent needs to be CoW'd from its + * beginning, so all we have to do is compute + * how many clusters to grab. We align + * want_clusters to the edge of contig_clusters + * to get better I/O. + */ + want_clusters = ocfs2_cow_align_length(inode->i_sb, + want_clusters); + + if (leaf_clusters < want_clusters) + *cow_len += leaf_clusters; + else + *cow_len += want_clusters; + } else if ((*cow_start + contig_clusters) >= + (cpos + write_len)) { + /* + * Breaking off contig_clusters at the front + * of the extent will cover our write. That's + * easy. + */ + *cow_len = contig_clusters; + } else if ((rec_end - cpos) <= contig_clusters) { + /* + * Breaking off contig_clusters at the tail of + * this extent will cover cpos. + */ + *cow_start = rec_end - contig_clusters; + *cow_len = contig_clusters; + } else if ((rec_end - cpos) <= want_clusters) { + /* + * While we can't fit the entire write in this + * extent, we know that the write goes from cpos + * to the end of the extent. Break that off. + * We try to break it at some multiple of + * contig_clusters from the front of the extent. + * Failing that (ie, cpos is within + * contig_clusters of the front), we'll CoW the + * entire extent. + */ + *cow_start = ocfs2_cow_align_start(inode->i_sb, + *cow_start, cpos); + *cow_len = rec_end - *cow_start; + } else { + /* + * Ok, the entire write lives in the middle of + * this extent. Let's try to slice the extent up + * nicely. Optimally, our CoW region starts at + * m*contig_clusters from the beginning of the + * extent and goes for n*contig_clusters, + * covering the entire write. + */ + *cow_start = ocfs2_cow_align_start(inode->i_sb, + *cow_start, cpos); + + want_clusters = (cpos + write_len) - *cow_start; + want_clusters = ocfs2_cow_align_length(inode->i_sb, + want_clusters); + if (*cow_start + want_clusters <= rec_end) + *cow_len = want_clusters; + else + *cow_len = rec_end - *cow_start; + } + + /* Have we covered our entire write yet? */ + if ((*cow_start + *cow_len) >= (cpos + write_len)) + break; + + /* + * If we reach the end of the extent block and don't get enough + * clusters, continue with the next extent block if possible. + */ + if (i + 1 == le16_to_cpu(el->l_next_free_rec) && + eb && eb->h_next_leaf_blk) { + brelse(eb_bh); + eb_bh = NULL; + + ret = ocfs2_read_extent_block(INODE_CACHE(inode), + le64_to_cpu(eb->h_next_leaf_blk), + &eb_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + eb = (struct ocfs2_extent_block *) eb_bh->b_data; + el = &eb->h_list; + i = -1; + } + } + +out: + brelse(eb_bh); + return ret; +} + +/* + * Prepare meta_ac, data_ac and calculate credits when we want to add some + * num_clusters in data_tree "et" and change the refcount for the old + * clusters(starting form p_cluster) in the refcount tree. + * + * Note: + * 1. since we may split the old tree, so we at most will need num_clusters + 2 + * more new leaf records. + * 2. In some case, we may not need to reserve new clusters(e.g, reflink), so + * just give data_ac = NULL. + */ +static int ocfs2_lock_refcount_allocators(struct super_block *sb, + u32 p_cluster, u32 num_clusters, + struct ocfs2_extent_tree *et, + struct ocfs2_caching_info *ref_ci, + struct buffer_head *ref_root_bh, + struct ocfs2_alloc_context **meta_ac, + struct ocfs2_alloc_context **data_ac, + int *credits) +{ + int ret = 0, meta_add = 0; + int num_free_extents = ocfs2_num_free_extents(OCFS2_SB(sb), et); + + if (num_free_extents < 0) { + ret = num_free_extents; + mlog_errno(ret); + goto out; + } + + if (num_free_extents < num_clusters + 2) + meta_add = + ocfs2_extend_meta_needed(et->et_root_el); + + *credits += ocfs2_calc_extend_credits(sb, et->et_root_el); + + ret = ocfs2_calc_refcount_meta_credits(sb, ref_ci, ref_root_bh, + p_cluster, num_clusters, + &meta_add, credits); + if (ret) { + mlog_errno(ret); + goto out; + } + + trace_ocfs2_lock_refcount_allocators(meta_add, *credits); + ret = ocfs2_reserve_new_metadata_blocks(OCFS2_SB(sb), meta_add, + meta_ac); + if (ret) { + mlog_errno(ret); + goto out; + } + + if (data_ac) { + ret = ocfs2_reserve_clusters(OCFS2_SB(sb), num_clusters, + data_ac); + if (ret) + mlog_errno(ret); + } + +out: + if (ret) { + if (*meta_ac) { + ocfs2_free_alloc_context(*meta_ac); + *meta_ac = NULL; + } + } + + return ret; +} + +static int ocfs2_clear_cow_buffer(handle_t *handle, struct buffer_head *bh) +{ + BUG_ON(buffer_dirty(bh)); + + clear_buffer_mapped(bh); + + return 0; +} + +int ocfs2_duplicate_clusters_by_page(handle_t *handle, + struct inode *inode, + u32 cpos, u32 old_cluster, + u32 new_cluster, u32 new_len) +{ + int ret = 0, partial; + struct super_block *sb = inode->i_sb; + u64 new_block = ocfs2_clusters_to_blocks(sb, new_cluster); + struct page *page; + pgoff_t page_index; + unsigned int from, to, readahead_pages; + loff_t offset, end, map_end; + struct address_space *mapping = inode->i_mapping; + + trace_ocfs2_duplicate_clusters_by_page(cpos, old_cluster, + new_cluster, new_len); + + readahead_pages = + (ocfs2_cow_contig_clusters(sb) << + OCFS2_SB(sb)->s_clustersize_bits) >> PAGE_CACHE_SHIFT; + offset = ((loff_t)cpos) << OCFS2_SB(sb)->s_clustersize_bits; + end = offset + (new_len << OCFS2_SB(sb)->s_clustersize_bits); + /* + * We only duplicate pages until we reach the page contains i_size - 1. + * So trim 'end' to i_size. + */ + if (end > i_size_read(inode)) + end = i_size_read(inode); + + while (offset < end) { + page_index = offset >> PAGE_CACHE_SHIFT; + map_end = ((loff_t)page_index + 1) << PAGE_CACHE_SHIFT; + if (map_end > end) + map_end = end; + + /* from, to is the offset within the page. */ + from = offset & (PAGE_CACHE_SIZE - 1); + to = PAGE_CACHE_SIZE; + if (map_end & (PAGE_CACHE_SIZE - 1)) + to = map_end & (PAGE_CACHE_SIZE - 1); + + page = find_or_create_page(mapping, page_index, GFP_NOFS); + if (!page) { + ret = -ENOMEM; + mlog_errno(ret); + break; + } + + /* + * In case PAGE_CACHE_SIZE <= CLUSTER_SIZE, This page + * can't be dirtied before we CoW it out. + */ + if (PAGE_CACHE_SIZE <= OCFS2_SB(sb)->s_clustersize) + BUG_ON(PageDirty(page)); + + if (!PageUptodate(page)) { + ret = block_read_full_page(page, ocfs2_get_block); + if (ret) { + mlog_errno(ret); + goto unlock; + } + lock_page(page); + } + + if (page_has_buffers(page)) { + ret = walk_page_buffers(handle, page_buffers(page), + from, to, &partial, + ocfs2_clear_cow_buffer); + if (ret) { + mlog_errno(ret); + goto unlock; + } + } + + ocfs2_map_and_dirty_page(inode, + handle, from, to, + page, 0, &new_block); + mark_page_accessed(page); +unlock: + unlock_page(page); + page_cache_release(page); + page = NULL; + offset = map_end; + if (ret) + break; + } + + return ret; +} + +int ocfs2_duplicate_clusters_by_jbd(handle_t *handle, + struct inode *inode, + u32 cpos, u32 old_cluster, + u32 new_cluster, u32 new_len) +{ + int ret = 0; + struct super_block *sb = inode->i_sb; + struct ocfs2_caching_info *ci = INODE_CACHE(inode); + int i, blocks = ocfs2_clusters_to_blocks(sb, new_len); + u64 old_block = ocfs2_clusters_to_blocks(sb, old_cluster); + u64 new_block = ocfs2_clusters_to_blocks(sb, new_cluster); + struct ocfs2_super *osb = OCFS2_SB(sb); + struct buffer_head *old_bh = NULL; + struct buffer_head *new_bh = NULL; + + trace_ocfs2_duplicate_clusters_by_page(cpos, old_cluster, + new_cluster, new_len); + + for (i = 0; i < blocks; i++, old_block++, new_block++) { + new_bh = sb_getblk(osb->sb, new_block); + if (new_bh == NULL) { + ret = -ENOMEM; + mlog_errno(ret); + break; + } + + ocfs2_set_new_buffer_uptodate(ci, new_bh); + + ret = ocfs2_read_block(ci, old_block, &old_bh, NULL); + if (ret) { + mlog_errno(ret); + break; + } + + ret = ocfs2_journal_access(handle, ci, new_bh, + OCFS2_JOURNAL_ACCESS_CREATE); + if (ret) { + mlog_errno(ret); + break; + } + + memcpy(new_bh->b_data, old_bh->b_data, sb->s_blocksize); + ocfs2_journal_dirty(handle, new_bh); + + brelse(new_bh); + brelse(old_bh); + new_bh = NULL; + old_bh = NULL; + } + + brelse(new_bh); + brelse(old_bh); + return ret; +} + +static int ocfs2_clear_ext_refcount(handle_t *handle, + struct ocfs2_extent_tree *et, + u32 cpos, u32 p_cluster, u32 len, + unsigned int ext_flags, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_cached_dealloc_ctxt *dealloc) +{ + int ret, index; + struct ocfs2_extent_rec replace_rec; + struct ocfs2_path *path = NULL; + struct ocfs2_extent_list *el; + struct super_block *sb = ocfs2_metadata_cache_get_super(et->et_ci); + u64 ino = ocfs2_metadata_cache_owner(et->et_ci); + + trace_ocfs2_clear_ext_refcount((unsigned long long)ino, + cpos, len, p_cluster, ext_flags); + + memset(&replace_rec, 0, sizeof(replace_rec)); + replace_rec.e_cpos = cpu_to_le32(cpos); + replace_rec.e_leaf_clusters = cpu_to_le16(len); + replace_rec.e_blkno = cpu_to_le64(ocfs2_clusters_to_blocks(sb, + p_cluster)); + replace_rec.e_flags = ext_flags; + replace_rec.e_flags &= ~OCFS2_EXT_REFCOUNTED; + + path = ocfs2_new_path_from_et(et); + if (!path) { + ret = -ENOMEM; + mlog_errno(ret); + goto out; + } + + ret = ocfs2_find_path(et->et_ci, path, cpos); + if (ret) { + mlog_errno(ret); + goto out; + } + + el = path_leaf_el(path); + + index = ocfs2_search_extent_list(el, cpos); + if (index == -1 || index >= le16_to_cpu(el->l_next_free_rec)) { + ocfs2_error(sb, + "Inode %llu has an extent at cpos %u which can no " + "longer be found.\n", + (unsigned long long)ino, cpos); + ret = -EROFS; + goto out; + } + + ret = ocfs2_split_extent(handle, et, path, index, + &replace_rec, meta_ac, dealloc); + if (ret) + mlog_errno(ret); + +out: + ocfs2_free_path(path); + return ret; +} + +static int ocfs2_replace_clusters(handle_t *handle, + struct ocfs2_cow_context *context, + u32 cpos, u32 old, + u32 new, u32 len, + unsigned int ext_flags) +{ + int ret; + struct ocfs2_caching_info *ci = context->data_et.et_ci; + u64 ino = ocfs2_metadata_cache_owner(ci); + + trace_ocfs2_replace_clusters((unsigned long long)ino, + cpos, old, new, len, ext_flags); + + /*If the old clusters is unwritten, no need to duplicate. */ + if (!(ext_flags & OCFS2_EXT_UNWRITTEN)) { + ret = context->cow_duplicate_clusters(handle, context->inode, + cpos, old, new, len); + if (ret) { + mlog_errno(ret); + goto out; + } + } + + ret = ocfs2_clear_ext_refcount(handle, &context->data_et, + cpos, new, len, ext_flags, + context->meta_ac, &context->dealloc); + if (ret) + mlog_errno(ret); +out: + return ret; +} + +int ocfs2_cow_sync_writeback(struct super_block *sb, + struct inode *inode, + u32 cpos, u32 num_clusters) +{ + int ret = 0; + loff_t offset, end, map_end; + pgoff_t page_index; + struct page *page; + + if (ocfs2_should_order_data(inode)) + return 0; + + offset = ((loff_t)cpos) << OCFS2_SB(sb)->s_clustersize_bits; + end = offset + (num_clusters << OCFS2_SB(sb)->s_clustersize_bits); + + ret = filemap_fdatawrite_range(inode->i_mapping, + offset, end - 1); + if (ret < 0) { + mlog_errno(ret); + return ret; + } + + while (offset < end) { + page_index = offset >> PAGE_CACHE_SHIFT; + map_end = ((loff_t)page_index + 1) << PAGE_CACHE_SHIFT; + if (map_end > end) + map_end = end; + + page = find_or_create_page(inode->i_mapping, + page_index, GFP_NOFS); + BUG_ON(!page); + + wait_on_page_writeback(page); + if (PageError(page)) { + ret = -EIO; + mlog_errno(ret); + } else + mark_page_accessed(page); + + unlock_page(page); + page_cache_release(page); + page = NULL; + offset = map_end; + if (ret) + break; + } + + return ret; +} + +static int ocfs2_di_get_clusters(struct ocfs2_cow_context *context, + u32 v_cluster, u32 *p_cluster, + u32 *num_clusters, + unsigned int *extent_flags) +{ + return ocfs2_get_clusters(context->inode, v_cluster, p_cluster, + num_clusters, extent_flags); +} + +static int ocfs2_make_clusters_writable(struct super_block *sb, + struct ocfs2_cow_context *context, + u32 cpos, u32 p_cluster, + u32 num_clusters, unsigned int e_flags) +{ + int ret, delete, index, credits = 0; + u32 new_bit, new_len, orig_num_clusters; + unsigned int set_len; + struct ocfs2_super *osb = OCFS2_SB(sb); + handle_t *handle; + struct buffer_head *ref_leaf_bh = NULL; + struct ocfs2_caching_info *ref_ci = &context->ref_tree->rf_ci; + struct ocfs2_refcount_rec rec; + + trace_ocfs2_make_clusters_writable(cpos, p_cluster, + num_clusters, e_flags); + + ret = ocfs2_lock_refcount_allocators(sb, p_cluster, num_clusters, + &context->data_et, + ref_ci, + context->ref_root_bh, + &context->meta_ac, + &context->data_ac, &credits); + if (ret) { + mlog_errno(ret); + return ret; + } + + if (context->post_refcount) + credits += context->post_refcount->credits; + + credits += context->extra_credits; + handle = ocfs2_start_trans(osb, credits); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto out; + } + + orig_num_clusters = num_clusters; + + while (num_clusters) { + ret = ocfs2_get_refcount_rec(ref_ci, context->ref_root_bh, + p_cluster, num_clusters, + &rec, &index, &ref_leaf_bh); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + BUG_ON(!rec.r_refcount); + set_len = min((u64)p_cluster + num_clusters, + le64_to_cpu(rec.r_cpos) + + le32_to_cpu(rec.r_clusters)) - p_cluster; + + /* + * There are many different situation here. + * 1. If refcount == 1, remove the flag and don't COW. + * 2. If refcount > 1, allocate clusters. + * Here we may not allocate r_len once at a time, so continue + * until we reach num_clusters. + */ + if (le32_to_cpu(rec.r_refcount) == 1) { + delete = 0; + ret = ocfs2_clear_ext_refcount(handle, + &context->data_et, + cpos, p_cluster, + set_len, e_flags, + context->meta_ac, + &context->dealloc); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + } else { + delete = 1; + + ret = __ocfs2_claim_clusters(handle, + context->data_ac, + 1, set_len, + &new_bit, &new_len); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + ret = ocfs2_replace_clusters(handle, context, + cpos, p_cluster, new_bit, + new_len, e_flags); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + set_len = new_len; + } + + ret = __ocfs2_decrease_refcount(handle, ref_ci, + context->ref_root_bh, + p_cluster, set_len, + context->meta_ac, + &context->dealloc, delete); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + cpos += set_len; + p_cluster += set_len; + num_clusters -= set_len; + brelse(ref_leaf_bh); + ref_leaf_bh = NULL; + } + + /* handle any post_cow action. */ + if (context->post_refcount && context->post_refcount->func) { + ret = context->post_refcount->func(context->inode, handle, + context->post_refcount->para); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + } + + /* + * Here we should write the new page out first if we are + * in write-back mode. + */ + if (context->get_clusters == ocfs2_di_get_clusters) { + ret = ocfs2_cow_sync_writeback(sb, context->inode, cpos, + orig_num_clusters); + if (ret) + mlog_errno(ret); + } + +out_commit: + ocfs2_commit_trans(osb, handle); + +out: + if (context->data_ac) { + ocfs2_free_alloc_context(context->data_ac); + context->data_ac = NULL; + } + if (context->meta_ac) { + ocfs2_free_alloc_context(context->meta_ac); + context->meta_ac = NULL; + } + brelse(ref_leaf_bh); + + return ret; +} + +static int ocfs2_replace_cow(struct ocfs2_cow_context *context) +{ + int ret = 0; + struct inode *inode = context->inode; + u32 cow_start = context->cow_start, cow_len = context->cow_len; + u32 p_cluster, num_clusters; + unsigned int ext_flags; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + + if (!ocfs2_refcount_tree(OCFS2_SB(inode->i_sb))) { + ocfs2_error(inode->i_sb, "Inode %lu want to use refcount " + "tree, but the feature bit is not set in the " + "super block.", inode->i_ino); + return -EROFS; + } + + ocfs2_init_dealloc_ctxt(&context->dealloc); + + while (cow_len) { + ret = context->get_clusters(context, cow_start, &p_cluster, + &num_clusters, &ext_flags); + if (ret) { + mlog_errno(ret); + break; + } + + BUG_ON(!(ext_flags & OCFS2_EXT_REFCOUNTED)); + + if (cow_len < num_clusters) + num_clusters = cow_len; + + ret = ocfs2_make_clusters_writable(inode->i_sb, context, + cow_start, p_cluster, + num_clusters, ext_flags); + if (ret) { + mlog_errno(ret); + break; + } + + cow_len -= num_clusters; + cow_start += num_clusters; + } + + if (ocfs2_dealloc_has_cluster(&context->dealloc)) { + ocfs2_schedule_truncate_log_flush(osb, 1); + ocfs2_run_deallocs(osb, &context->dealloc); + } + + return ret; +} + +/* + * Starting at cpos, try to CoW write_len clusters. Don't CoW + * past max_cpos. This will stop when it runs into a hole or an + * unrefcounted extent. + */ +static int ocfs2_refcount_cow_hunk(struct inode *inode, + struct buffer_head *di_bh, + u32 cpos, u32 write_len, u32 max_cpos) +{ + int ret; + u32 cow_start = 0, cow_len = 0; + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + struct buffer_head *ref_root_bh = NULL; + struct ocfs2_refcount_tree *ref_tree; + struct ocfs2_cow_context *context = NULL; + + BUG_ON(!(oi->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL)); + + ret = ocfs2_refcount_cal_cow_clusters(inode, &di->id2.i_list, + cpos, write_len, max_cpos, + &cow_start, &cow_len); + if (ret) { + mlog_errno(ret); + goto out; + } + + trace_ocfs2_refcount_cow_hunk(OCFS2_I(inode)->ip_blkno, + cpos, write_len, max_cpos, + cow_start, cow_len); + + BUG_ON(cow_len == 0); + + context = kzalloc(sizeof(struct ocfs2_cow_context), GFP_NOFS); + if (!context) { + ret = -ENOMEM; + mlog_errno(ret); + goto out; + } + + ret = ocfs2_lock_refcount_tree(osb, le64_to_cpu(di->i_refcount_loc), + 1, &ref_tree, &ref_root_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + context->inode = inode; + context->cow_start = cow_start; + context->cow_len = cow_len; + context->ref_tree = ref_tree; + context->ref_root_bh = ref_root_bh; + context->cow_duplicate_clusters = ocfs2_duplicate_clusters_by_page; + context->get_clusters = ocfs2_di_get_clusters; + + ocfs2_init_dinode_extent_tree(&context->data_et, + INODE_CACHE(inode), di_bh); + + ret = ocfs2_replace_cow(context); + if (ret) + mlog_errno(ret); + + /* + * truncate the extent map here since no matter whether we meet with + * any error during the action, we shouldn't trust cached extent map + * any more. + */ + ocfs2_extent_map_trunc(inode, cow_start); + + ocfs2_unlock_refcount_tree(osb, ref_tree, 1); + brelse(ref_root_bh); +out: + kfree(context); + return ret; +} + +/* + * CoW any and all clusters between cpos and cpos+write_len. + * Don't CoW past max_cpos. If this returns successfully, all + * clusters between cpos and cpos+write_len are safe to modify. + */ +int ocfs2_refcount_cow(struct inode *inode, + struct buffer_head *di_bh, + u32 cpos, u32 write_len, u32 max_cpos) +{ + int ret = 0; + u32 p_cluster, num_clusters; + unsigned int ext_flags; + + while (write_len) { + ret = ocfs2_get_clusters(inode, cpos, &p_cluster, + &num_clusters, &ext_flags); + if (ret) { + mlog_errno(ret); + break; + } + + if (write_len < num_clusters) + num_clusters = write_len; + + if (ext_flags & OCFS2_EXT_REFCOUNTED) { + ret = ocfs2_refcount_cow_hunk(inode, di_bh, cpos, + num_clusters, max_cpos); + if (ret) { + mlog_errno(ret); + break; + } + } + + write_len -= num_clusters; + cpos += num_clusters; + } + + return ret; +} + +static int ocfs2_xattr_value_get_clusters(struct ocfs2_cow_context *context, + u32 v_cluster, u32 *p_cluster, + u32 *num_clusters, + unsigned int *extent_flags) +{ + struct inode *inode = context->inode; + struct ocfs2_xattr_value_root *xv = context->cow_object; + + return ocfs2_xattr_get_clusters(inode, v_cluster, p_cluster, + num_clusters, &xv->xr_list, + extent_flags); +} + +/* + * Given a xattr value root, calculate the most meta/credits we need for + * refcount tree change if we truncate it to 0. + */ +int ocfs2_refcounted_xattr_delete_need(struct inode *inode, + struct ocfs2_caching_info *ref_ci, + struct buffer_head *ref_root_bh, + struct ocfs2_xattr_value_root *xv, + int *meta_add, int *credits) +{ + int ret = 0, index, ref_blocks = 0; + u32 p_cluster, num_clusters; + u32 cpos = 0, clusters = le32_to_cpu(xv->xr_clusters); + struct ocfs2_refcount_block *rb; + struct ocfs2_refcount_rec rec; + struct buffer_head *ref_leaf_bh = NULL; + + while (cpos < clusters) { + ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster, + &num_clusters, &xv->xr_list, + NULL); + if (ret) { + mlog_errno(ret); + goto out; + } + + cpos += num_clusters; + + while (num_clusters) { + ret = ocfs2_get_refcount_rec(ref_ci, ref_root_bh, + p_cluster, num_clusters, + &rec, &index, + &ref_leaf_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + BUG_ON(!rec.r_refcount); + + rb = (struct ocfs2_refcount_block *)ref_leaf_bh->b_data; + + /* + * We really don't know whether the other clusters is in + * this refcount block or not, so just take the worst + * case that all the clusters are in this block and each + * one will split a refcount rec, so totally we need + * clusters * 2 new refcount rec. + */ + if (le16_to_cpu(rb->rf_records.rl_used) + clusters * 2 > + le16_to_cpu(rb->rf_records.rl_count)) + ref_blocks++; + + *credits += 1; + brelse(ref_leaf_bh); + ref_leaf_bh = NULL; + + if (num_clusters <= le32_to_cpu(rec.r_clusters)) + break; + else + num_clusters -= le32_to_cpu(rec.r_clusters); + p_cluster += num_clusters; + } + } + + *meta_add += ref_blocks; + if (!ref_blocks) + goto out; + + rb = (struct ocfs2_refcount_block *)ref_root_bh->b_data; + if (le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL) + *credits += OCFS2_EXPAND_REFCOUNT_TREE_CREDITS; + else { + struct ocfs2_extent_tree et; + + ocfs2_init_refcount_extent_tree(&et, ref_ci, ref_root_bh); + *credits += ocfs2_calc_extend_credits(inode->i_sb, + et.et_root_el); + } + +out: + brelse(ref_leaf_bh); + return ret; +} + +/* + * Do CoW for xattr. + */ +int ocfs2_refcount_cow_xattr(struct inode *inode, + struct ocfs2_dinode *di, + struct ocfs2_xattr_value_buf *vb, + struct ocfs2_refcount_tree *ref_tree, + struct buffer_head *ref_root_bh, + u32 cpos, u32 write_len, + struct ocfs2_post_refcount *post) +{ + int ret; + struct ocfs2_xattr_value_root *xv = vb->vb_xv; + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct ocfs2_cow_context *context = NULL; + u32 cow_start, cow_len; + + BUG_ON(!(oi->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL)); + + ret = ocfs2_refcount_cal_cow_clusters(inode, &xv->xr_list, + cpos, write_len, UINT_MAX, + &cow_start, &cow_len); + if (ret) { + mlog_errno(ret); + goto out; + } + + BUG_ON(cow_len == 0); + + context = kzalloc(sizeof(struct ocfs2_cow_context), GFP_NOFS); + if (!context) { + ret = -ENOMEM; + mlog_errno(ret); + goto out; + } + + context->inode = inode; + context->cow_start = cow_start; + context->cow_len = cow_len; + context->ref_tree = ref_tree; + context->ref_root_bh = ref_root_bh; + context->cow_object = xv; + + context->cow_duplicate_clusters = ocfs2_duplicate_clusters_by_jbd; + /* We need the extra credits for duplicate_clusters by jbd. */ + context->extra_credits = + ocfs2_clusters_to_blocks(inode->i_sb, 1) * cow_len; + context->get_clusters = ocfs2_xattr_value_get_clusters; + context->post_refcount = post; + + ocfs2_init_xattr_value_extent_tree(&context->data_et, + INODE_CACHE(inode), vb); + + ret = ocfs2_replace_cow(context); + if (ret) + mlog_errno(ret); + +out: + kfree(context); + return ret; +} + +/* + * Insert a new extent into refcount tree and mark a extent rec + * as refcounted in the dinode tree. + */ +int ocfs2_add_refcount_flag(struct inode *inode, + struct ocfs2_extent_tree *data_et, + struct ocfs2_caching_info *ref_ci, + struct buffer_head *ref_root_bh, + u32 cpos, u32 p_cluster, u32 num_clusters, + struct ocfs2_cached_dealloc_ctxt *dealloc, + struct ocfs2_post_refcount *post) +{ + int ret; + handle_t *handle; + int credits = 1, ref_blocks = 0; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_alloc_context *meta_ac = NULL; + + ret = ocfs2_calc_refcount_meta_credits(inode->i_sb, + ref_ci, ref_root_bh, + p_cluster, num_clusters, + &ref_blocks, &credits); + if (ret) { + mlog_errno(ret); + goto out; + } + + trace_ocfs2_add_refcount_flag(ref_blocks, credits); + + if (ref_blocks) { + ret = ocfs2_reserve_new_metadata_blocks(OCFS2_SB(inode->i_sb), + ref_blocks, &meta_ac); + if (ret) { + mlog_errno(ret); + goto out; + } + } + + if (post) + credits += post->credits; + + handle = ocfs2_start_trans(osb, credits); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto out; + } + + ret = ocfs2_mark_extent_refcounted(inode, data_et, handle, + cpos, num_clusters, p_cluster, + meta_ac, dealloc); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + ret = __ocfs2_increase_refcount(handle, ref_ci, ref_root_bh, + p_cluster, num_clusters, 0, + meta_ac, dealloc); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + if (post && post->func) { + ret = post->func(inode, handle, post->para); + if (ret) + mlog_errno(ret); + } + +out_commit: + ocfs2_commit_trans(osb, handle); +out: + if (meta_ac) + ocfs2_free_alloc_context(meta_ac); + return ret; +} + +static int ocfs2_change_ctime(struct inode *inode, + struct buffer_head *di_bh) +{ + int ret; + handle_t *handle; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + + handle = ocfs2_start_trans(OCFS2_SB(inode->i_sb), + OCFS2_INODE_UPDATE_CREDITS); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto out; + } + + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + inode->i_ctime = CURRENT_TIME; + di->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec); + di->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec); + + ocfs2_journal_dirty(handle, di_bh); + +out_commit: + ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle); +out: + return ret; +} + +static int ocfs2_attach_refcount_tree(struct inode *inode, + struct buffer_head *di_bh) +{ + int ret, data_changed = 0; + struct buffer_head *ref_root_bh = NULL; + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_refcount_tree *ref_tree; + unsigned int ext_flags; + loff_t size; + u32 cpos, num_clusters, clusters, p_cluster; + struct ocfs2_cached_dealloc_ctxt dealloc; + struct ocfs2_extent_tree di_et; + + ocfs2_init_dealloc_ctxt(&dealloc); + + if (!(oi->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL)) { + ret = ocfs2_create_refcount_tree(inode, di_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + } + + BUG_ON(!di->i_refcount_loc); + ret = ocfs2_lock_refcount_tree(osb, + le64_to_cpu(di->i_refcount_loc), 1, + &ref_tree, &ref_root_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) + goto attach_xattr; + + ocfs2_init_dinode_extent_tree(&di_et, INODE_CACHE(inode), di_bh); + + size = i_size_read(inode); + clusters = ocfs2_clusters_for_bytes(inode->i_sb, size); + + cpos = 0; + while (cpos < clusters) { + ret = ocfs2_get_clusters(inode, cpos, &p_cluster, + &num_clusters, &ext_flags); + if (ret) { + mlog_errno(ret); + goto unlock; + } + if (p_cluster && !(ext_flags & OCFS2_EXT_REFCOUNTED)) { + ret = ocfs2_add_refcount_flag(inode, &di_et, + &ref_tree->rf_ci, + ref_root_bh, cpos, + p_cluster, num_clusters, + &dealloc, NULL); + if (ret) { + mlog_errno(ret); + goto unlock; + } + + data_changed = 1; + } + cpos += num_clusters; + } + +attach_xattr: + if (oi->ip_dyn_features & OCFS2_HAS_XATTR_FL) { + ret = ocfs2_xattr_attach_refcount_tree(inode, di_bh, + &ref_tree->rf_ci, + ref_root_bh, + &dealloc); + if (ret) { + mlog_errno(ret); + goto unlock; + } + } + + if (data_changed) { + ret = ocfs2_change_ctime(inode, di_bh); + if (ret) + mlog_errno(ret); + } + +unlock: + ocfs2_unlock_refcount_tree(osb, ref_tree, 1); + brelse(ref_root_bh); + + if (!ret && ocfs2_dealloc_has_cluster(&dealloc)) { + ocfs2_schedule_truncate_log_flush(osb, 1); + ocfs2_run_deallocs(osb, &dealloc); + } +out: + /* + * Empty the extent map so that we may get the right extent + * record from the disk. + */ + ocfs2_extent_map_trunc(inode, 0); + + return ret; +} + +static int ocfs2_add_refcounted_extent(struct inode *inode, + struct ocfs2_extent_tree *et, + struct ocfs2_caching_info *ref_ci, + struct buffer_head *ref_root_bh, + u32 cpos, u32 p_cluster, u32 num_clusters, + unsigned int ext_flags, + struct ocfs2_cached_dealloc_ctxt *dealloc) +{ + int ret; + handle_t *handle; + int credits = 0; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_alloc_context *meta_ac = NULL; + + ret = ocfs2_lock_refcount_allocators(inode->i_sb, + p_cluster, num_clusters, + et, ref_ci, + ref_root_bh, &meta_ac, + NULL, &credits); + if (ret) { + mlog_errno(ret); + goto out; + } + + handle = ocfs2_start_trans(osb, credits); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto out; + } + + ret = ocfs2_insert_extent(handle, et, cpos, + ocfs2_clusters_to_blocks(inode->i_sb, p_cluster), + num_clusters, ext_flags, meta_ac); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + ret = ocfs2_increase_refcount(handle, ref_ci, ref_root_bh, + p_cluster, num_clusters, + meta_ac, dealloc); + if (ret) + mlog_errno(ret); + +out_commit: + ocfs2_commit_trans(osb, handle); +out: + if (meta_ac) + ocfs2_free_alloc_context(meta_ac); + return ret; +} + +static int ocfs2_duplicate_inline_data(struct inode *s_inode, + struct buffer_head *s_bh, + struct inode *t_inode, + struct buffer_head *t_bh) +{ + int ret; + handle_t *handle; + struct ocfs2_super *osb = OCFS2_SB(s_inode->i_sb); + struct ocfs2_dinode *s_di = (struct ocfs2_dinode *)s_bh->b_data; + struct ocfs2_dinode *t_di = (struct ocfs2_dinode *)t_bh->b_data; + + BUG_ON(!(OCFS2_I(s_inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL)); + + handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto out; + } + + ret = ocfs2_journal_access_di(handle, INODE_CACHE(t_inode), t_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + t_di->id2.i_data.id_count = s_di->id2.i_data.id_count; + memcpy(t_di->id2.i_data.id_data, s_di->id2.i_data.id_data, + le16_to_cpu(s_di->id2.i_data.id_count)); + spin_lock(&OCFS2_I(t_inode)->ip_lock); + OCFS2_I(t_inode)->ip_dyn_features |= OCFS2_INLINE_DATA_FL; + t_di->i_dyn_features = cpu_to_le16(OCFS2_I(t_inode)->ip_dyn_features); + spin_unlock(&OCFS2_I(t_inode)->ip_lock); + + ocfs2_journal_dirty(handle, t_bh); + +out_commit: + ocfs2_commit_trans(osb, handle); +out: + return ret; +} + +static int ocfs2_duplicate_extent_list(struct inode *s_inode, + struct inode *t_inode, + struct buffer_head *t_bh, + struct ocfs2_caching_info *ref_ci, + struct buffer_head *ref_root_bh, + struct ocfs2_cached_dealloc_ctxt *dealloc) +{ + int ret = 0; + u32 p_cluster, num_clusters, clusters, cpos; + loff_t size; + unsigned int ext_flags; + struct ocfs2_extent_tree et; + + ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(t_inode), t_bh); + + size = i_size_read(s_inode); + clusters = ocfs2_clusters_for_bytes(s_inode->i_sb, size); + + cpos = 0; + while (cpos < clusters) { + ret = ocfs2_get_clusters(s_inode, cpos, &p_cluster, + &num_clusters, &ext_flags); + if (ret) { + mlog_errno(ret); + goto out; + } + if (p_cluster) { + ret = ocfs2_add_refcounted_extent(t_inode, &et, + ref_ci, ref_root_bh, + cpos, p_cluster, + num_clusters, + ext_flags, + dealloc); + if (ret) { + mlog_errno(ret); + goto out; + } + } + + cpos += num_clusters; + } + +out: + return ret; +} + +/* + * change the new file's attributes to the src. + * + * reflink creates a snapshot of a file, that means the attributes + * must be identical except for three exceptions - nlink, ino, and ctime. + */ +static int ocfs2_complete_reflink(struct inode *s_inode, + struct buffer_head *s_bh, + struct inode *t_inode, + struct buffer_head *t_bh, + bool preserve) +{ + int ret; + handle_t *handle; + struct ocfs2_dinode *s_di = (struct ocfs2_dinode *)s_bh->b_data; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)t_bh->b_data; + loff_t size = i_size_read(s_inode); + + handle = ocfs2_start_trans(OCFS2_SB(t_inode->i_sb), + OCFS2_INODE_UPDATE_CREDITS); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + return ret; + } + + ret = ocfs2_journal_access_di(handle, INODE_CACHE(t_inode), t_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + spin_lock(&OCFS2_I(t_inode)->ip_lock); + OCFS2_I(t_inode)->ip_clusters = OCFS2_I(s_inode)->ip_clusters; + OCFS2_I(t_inode)->ip_attr = OCFS2_I(s_inode)->ip_attr; + OCFS2_I(t_inode)->ip_dyn_features = OCFS2_I(s_inode)->ip_dyn_features; + spin_unlock(&OCFS2_I(t_inode)->ip_lock); + i_size_write(t_inode, size); + t_inode->i_blocks = s_inode->i_blocks; + + di->i_xattr_inline_size = s_di->i_xattr_inline_size; + di->i_clusters = s_di->i_clusters; + di->i_size = s_di->i_size; + di->i_dyn_features = s_di->i_dyn_features; + di->i_attr = s_di->i_attr; + + if (preserve) { + t_inode->i_uid = s_inode->i_uid; + t_inode->i_gid = s_inode->i_gid; + t_inode->i_mode = s_inode->i_mode; + di->i_uid = s_di->i_uid; + di->i_gid = s_di->i_gid; + di->i_mode = s_di->i_mode; + + /* + * update time. + * we want mtime to appear identical to the source and + * update ctime. + */ + t_inode->i_ctime = CURRENT_TIME; + + di->i_ctime = cpu_to_le64(t_inode->i_ctime.tv_sec); + di->i_ctime_nsec = cpu_to_le32(t_inode->i_ctime.tv_nsec); + + t_inode->i_mtime = s_inode->i_mtime; + di->i_mtime = s_di->i_mtime; + di->i_mtime_nsec = s_di->i_mtime_nsec; + } + + ocfs2_journal_dirty(handle, t_bh); + +out_commit: + ocfs2_commit_trans(OCFS2_SB(t_inode->i_sb), handle); + return ret; +} + +static int ocfs2_create_reflink_node(struct inode *s_inode, + struct buffer_head *s_bh, + struct inode *t_inode, + struct buffer_head *t_bh, + bool preserve) +{ + int ret; + struct buffer_head *ref_root_bh = NULL; + struct ocfs2_cached_dealloc_ctxt dealloc; + struct ocfs2_super *osb = OCFS2_SB(s_inode->i_sb); + struct ocfs2_refcount_block *rb; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)s_bh->b_data; + struct ocfs2_refcount_tree *ref_tree; + + ocfs2_init_dealloc_ctxt(&dealloc); + + ret = ocfs2_set_refcount_tree(t_inode, t_bh, + le64_to_cpu(di->i_refcount_loc)); + if (ret) { + mlog_errno(ret); + goto out; + } + + if (OCFS2_I(s_inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) { + ret = ocfs2_duplicate_inline_data(s_inode, s_bh, + t_inode, t_bh); + if (ret) + mlog_errno(ret); + goto out; + } + + ret = ocfs2_lock_refcount_tree(osb, le64_to_cpu(di->i_refcount_loc), + 1, &ref_tree, &ref_root_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + rb = (struct ocfs2_refcount_block *)ref_root_bh->b_data; + + ret = ocfs2_duplicate_extent_list(s_inode, t_inode, t_bh, + &ref_tree->rf_ci, ref_root_bh, + &dealloc); + if (ret) { + mlog_errno(ret); + goto out_unlock_refcount; + } + +out_unlock_refcount: + ocfs2_unlock_refcount_tree(osb, ref_tree, 1); + brelse(ref_root_bh); +out: + if (ocfs2_dealloc_has_cluster(&dealloc)) { + ocfs2_schedule_truncate_log_flush(osb, 1); + ocfs2_run_deallocs(osb, &dealloc); + } + + return ret; +} + +static int __ocfs2_reflink(struct dentry *old_dentry, + struct buffer_head *old_bh, + struct inode *new_inode, + bool preserve) +{ + int ret; + struct inode *inode = old_dentry->d_inode; + struct buffer_head *new_bh = NULL; + + if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SYSTEM_FILE) { + ret = -EINVAL; + mlog_errno(ret); + goto out; + } + + ret = filemap_fdatawrite(inode->i_mapping); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_attach_refcount_tree(inode, old_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + mutex_lock_nested(&new_inode->i_mutex, I_MUTEX_CHILD); + ret = ocfs2_inode_lock_nested(new_inode, &new_bh, 1, + OI_LS_REFLINK_TARGET); + if (ret) { + mlog_errno(ret); + goto out_unlock; + } + + ret = ocfs2_create_reflink_node(inode, old_bh, + new_inode, new_bh, preserve); + if (ret) { + mlog_errno(ret); + goto inode_unlock; + } + + if (OCFS2_I(inode)->ip_dyn_features & OCFS2_HAS_XATTR_FL) { + ret = ocfs2_reflink_xattrs(inode, old_bh, + new_inode, new_bh, + preserve); + if (ret) { + mlog_errno(ret); + goto inode_unlock; + } + } + + ret = ocfs2_complete_reflink(inode, old_bh, + new_inode, new_bh, preserve); + if (ret) + mlog_errno(ret); + +inode_unlock: + ocfs2_inode_unlock(new_inode, 1); + brelse(new_bh); +out_unlock: + mutex_unlock(&new_inode->i_mutex); +out: + if (!ret) { + ret = filemap_fdatawait(inode->i_mapping); + if (ret) + mlog_errno(ret); + } + return ret; +} + +static int ocfs2_reflink(struct dentry *old_dentry, struct inode *dir, + struct dentry *new_dentry, bool preserve) +{ + int error; + struct inode *inode = old_dentry->d_inode; + struct buffer_head *old_bh = NULL; + struct inode *new_orphan_inode = NULL; + struct posix_acl *default_acl, *acl; + umode_t mode; + + if (!ocfs2_refcount_tree(OCFS2_SB(inode->i_sb))) + return -EOPNOTSUPP; + + mode = inode->i_mode; + error = posix_acl_create(dir, &mode, &default_acl, &acl); + if (error) { + mlog_errno(error); + goto out; + } + + error = ocfs2_create_inode_in_orphan(dir, mode, + &new_orphan_inode); + if (error) { + mlog_errno(error); + goto out; + } + + error = ocfs2_rw_lock(inode, 1); + if (error) { + mlog_errno(error); + goto out; + } + + error = ocfs2_inode_lock(inode, &old_bh, 1); + if (error) { + mlog_errno(error); + ocfs2_rw_unlock(inode, 1); + goto out; + } + + down_write(&OCFS2_I(inode)->ip_xattr_sem); + down_write(&OCFS2_I(inode)->ip_alloc_sem); + error = __ocfs2_reflink(old_dentry, old_bh, + new_orphan_inode, preserve); + up_write(&OCFS2_I(inode)->ip_alloc_sem); + up_write(&OCFS2_I(inode)->ip_xattr_sem); + + ocfs2_inode_unlock(inode, 1); + ocfs2_rw_unlock(inode, 1); + brelse(old_bh); + + if (error) { + mlog_errno(error); + goto out; + } + + /* If the security isn't preserved, we need to re-initialize them. */ + if (!preserve) { + error = ocfs2_init_security_and_acl(dir, new_orphan_inode, + &new_dentry->d_name, + default_acl, acl); + if (error) + mlog_errno(error); + } +out: + if (default_acl) + posix_acl_release(default_acl); + if (acl) + posix_acl_release(acl); + if (!error) { + error = ocfs2_mv_orphaned_inode_to_new(dir, new_orphan_inode, + new_dentry); + if (error) + mlog_errno(error); + } + + if (new_orphan_inode) { + /* + * We need to open_unlock the inode no matter whether we + * succeed or not, so that other nodes can delete it later. + */ + ocfs2_open_unlock(new_orphan_inode); + if (error) + iput(new_orphan_inode); + } + + return error; +} + +/* + * Below here are the bits used by OCFS2_IOC_REFLINK() to fake + * sys_reflink(). This will go away when vfs_reflink() exists in + * fs/namei.c. + */ + +/* copied from may_create in VFS. */ +static inline int ocfs2_may_create(struct inode *dir, struct dentry *child) +{ + if (child->d_inode) + return -EEXIST; + if (IS_DEADDIR(dir)) + return -ENOENT; + return inode_permission(dir, MAY_WRITE | MAY_EXEC); +} + +/** + * ocfs2_vfs_reflink - Create a reference-counted link + * + * @old_dentry: source dentry + inode + * @dir: directory to create the target + * @new_dentry: target dentry + * @preserve: if true, preserve all file attributes + */ +static int ocfs2_vfs_reflink(struct dentry *old_dentry, struct inode *dir, + struct dentry *new_dentry, bool preserve) +{ + struct inode *inode = old_dentry->d_inode; + int error; + + if (!inode) + return -ENOENT; + + error = ocfs2_may_create(dir, new_dentry); + if (error) + return error; + + if (dir->i_sb != inode->i_sb) + return -EXDEV; + + /* + * A reflink to an append-only or immutable file cannot be created. + */ + if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) + return -EPERM; + + /* Only regular files can be reflinked. */ + if (!S_ISREG(inode->i_mode)) + return -EPERM; + + /* + * If the caller wants to preserve ownership, they require the + * rights to do so. + */ + if (preserve) { + if (!uid_eq(current_fsuid(), inode->i_uid) && !capable(CAP_CHOWN)) + return -EPERM; + if (!in_group_p(inode->i_gid) && !capable(CAP_CHOWN)) + return -EPERM; + } + + /* + * If the caller is modifying any aspect of the attributes, they + * are not creating a snapshot. They need read permission on the + * file. + */ + if (!preserve) { + error = inode_permission(inode, MAY_READ); + if (error) + return error; + } + + mutex_lock(&inode->i_mutex); + dquot_initialize(dir); + error = ocfs2_reflink(old_dentry, dir, new_dentry, preserve); + mutex_unlock(&inode->i_mutex); + if (!error) + fsnotify_create(dir, new_dentry); + return error; +} +/* + * Most codes are copied from sys_linkat. + */ +int ocfs2_reflink_ioctl(struct inode *inode, + const char __user *oldname, + const char __user *newname, + bool preserve) +{ + struct dentry *new_dentry; + struct path old_path, new_path; + int error; + + if (!ocfs2_refcount_tree(OCFS2_SB(inode->i_sb))) + return -EOPNOTSUPP; + + error = user_path_at(AT_FDCWD, oldname, 0, &old_path); + if (error) { + mlog_errno(error); + return error; + } + + new_dentry = user_path_create(AT_FDCWD, newname, &new_path, 0); + error = PTR_ERR(new_dentry); + if (IS_ERR(new_dentry)) { + mlog_errno(error); + goto out; + } + + error = -EXDEV; + if (old_path.mnt != new_path.mnt) { + mlog_errno(error); + goto out_dput; + } + + error = ocfs2_vfs_reflink(old_path.dentry, + new_path.dentry->d_inode, + new_dentry, preserve); +out_dput: + done_path_create(&new_path, new_dentry); +out: + path_put(&old_path); + + return error; +} diff --git a/fs/ocfs2/refcounttree.h b/fs/ocfs2/refcounttree.h new file mode 100644 index 00000000000..6422bbcdb52 --- /dev/null +++ b/fs/ocfs2/refcounttree.h @@ -0,0 +1,118 @@ +/* -*- mode: c; c-basic-offset: 8; -*- + * vim: noexpandtab sw=8 ts=8 sts=0: + * + * refcounttree.h + * + * Copyright (C) 2009 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License version 2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + */ +#ifndef OCFS2_REFCOUNTTREE_H +#define OCFS2_REFCOUNTTREE_H + +struct ocfs2_refcount_tree { + struct rb_node rf_node; + u64 rf_blkno; + u32 rf_generation; + struct kref rf_getcnt; + struct rw_semaphore rf_sem; + struct ocfs2_lock_res rf_lockres; + int rf_removed; + + /* the following 4 fields are used by caching_info. */ + spinlock_t rf_lock; + struct ocfs2_caching_info rf_ci; + struct mutex rf_io_mutex; + struct super_block *rf_sb; +}; + +void ocfs2_purge_refcount_trees(struct ocfs2_super *osb); +int ocfs2_lock_refcount_tree(struct ocfs2_super *osb, u64 ref_blkno, int rw, + struct ocfs2_refcount_tree **tree, + struct buffer_head **ref_bh); +void ocfs2_unlock_refcount_tree(struct ocfs2_super *osb, + struct ocfs2_refcount_tree *tree, + int rw); + +int ocfs2_decrease_refcount(struct inode *inode, + handle_t *handle, u32 cpos, u32 len, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_cached_dealloc_ctxt *dealloc, + int delete); +int ocfs2_prepare_refcount_change_for_del(struct inode *inode, + u64 refcount_loc, + u64 phys_blkno, + u32 clusters, + int *credits, + int *ref_blocks); +int ocfs2_refcount_cow(struct inode *inode, + struct buffer_head *di_bh, + u32 cpos, u32 write_len, u32 max_cpos); + +typedef int (ocfs2_post_refcount_func)(struct inode *inode, + handle_t *handle, + void *para); +/* + * Some refcount caller need to do more work after we modify the data b-tree + * during refcount operation(including CoW and add refcount flag), and make the + * transaction complete. So it must give us this structure so that we can do it + * within our transaction. + * + */ +struct ocfs2_post_refcount { + int credits; /* credits it need for journal. */ + ocfs2_post_refcount_func *func; /* real function. */ + void *para; +}; + +int ocfs2_refcounted_xattr_delete_need(struct inode *inode, + struct ocfs2_caching_info *ref_ci, + struct buffer_head *ref_root_bh, + struct ocfs2_xattr_value_root *xv, + int *meta_add, int *credits); +int ocfs2_refcount_cow_xattr(struct inode *inode, + struct ocfs2_dinode *di, + struct ocfs2_xattr_value_buf *vb, + struct ocfs2_refcount_tree *ref_tree, + struct buffer_head *ref_root_bh, + u32 cpos, u32 write_len, + struct ocfs2_post_refcount *post); +int ocfs2_duplicate_clusters_by_page(handle_t *handle, + struct inode *inode, + u32 cpos, u32 old_cluster, + u32 new_cluster, u32 new_len); +int ocfs2_duplicate_clusters_by_jbd(handle_t *handle, + struct inode *inode, + u32 cpos, u32 old_cluster, + u32 new_cluster, u32 new_len); +int ocfs2_cow_sync_writeback(struct super_block *sb, + struct inode *inode, + u32 cpos, u32 num_clusters); +int ocfs2_add_refcount_flag(struct inode *inode, + struct ocfs2_extent_tree *data_et, + struct ocfs2_caching_info *ref_ci, + struct buffer_head *ref_root_bh, + u32 cpos, u32 p_cluster, u32 num_clusters, + struct ocfs2_cached_dealloc_ctxt *dealloc, + struct ocfs2_post_refcount *post); +int ocfs2_remove_refcount_tree(struct inode *inode, struct buffer_head *di_bh); +int ocfs2_try_remove_refcount_tree(struct inode *inode, + struct buffer_head *di_bh); +int ocfs2_increase_refcount(handle_t *handle, + struct ocfs2_caching_info *ci, + struct buffer_head *ref_root_bh, + u64 cpos, u32 len, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_cached_dealloc_ctxt *dealloc); +int ocfs2_reflink_ioctl(struct inode *inode, + const char __user *oldname, + const char __user *newname, + bool preserve); +#endif /* OCFS2_REFCOUNTTREE_H */ diff --git a/fs/ocfs2/reservations.c b/fs/ocfs2/reservations.c new file mode 100644 index 00000000000..41ffd36c689 --- /dev/null +++ b/fs/ocfs2/reservations.c @@ -0,0 +1,839 @@ +/* -*- mode: c; c-basic-offset: 8; -*- + * vim: noexpandtab sw=8 ts=8 sts=0: + * + * reservations.c + * + * Allocation reservations implementation + * + * Some code borrowed from fs/ext3/balloc.c and is: + * + * Copyright (C) 1992, 1993, 1994, 1995 + * Remy Card (card@masi.ibp.fr) + * Laboratoire MASI - Institut Blaise Pascal + * Universite Pierre et Marie Curie (Paris VI) + * + * The rest is copyright (C) 2010 Novell. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License version 2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + */ + +#include <linux/fs.h> +#include <linux/types.h> +#include <linux/highmem.h> +#include <linux/bitops.h> +#include <linux/list.h> + +#include <cluster/masklog.h> + +#include "ocfs2.h" +#include "ocfs2_trace.h" + +#ifdef CONFIG_OCFS2_DEBUG_FS +#define OCFS2_CHECK_RESERVATIONS +#endif + +DEFINE_SPINLOCK(resv_lock); + +#define OCFS2_MIN_RESV_WINDOW_BITS 8 +#define OCFS2_MAX_RESV_WINDOW_BITS 1024 + +int ocfs2_dir_resv_allowed(struct ocfs2_super *osb) +{ + return (osb->osb_resv_level && osb->osb_dir_resv_level); +} + +static unsigned int ocfs2_resv_window_bits(struct ocfs2_reservation_map *resmap, + struct ocfs2_alloc_reservation *resv) +{ + struct ocfs2_super *osb = resmap->m_osb; + unsigned int bits; + + if (!(resv->r_flags & OCFS2_RESV_FLAG_DIR)) { + /* 8, 16, 32, 64, 128, 256, 512, 1024 */ + bits = 4 << osb->osb_resv_level; + } else { + bits = 4 << osb->osb_dir_resv_level; + } + return bits; +} + +static inline unsigned int ocfs2_resv_end(struct ocfs2_alloc_reservation *resv) +{ + if (resv->r_len) + return resv->r_start + resv->r_len - 1; + return resv->r_start; +} + +static inline int ocfs2_resv_empty(struct ocfs2_alloc_reservation *resv) +{ + return !!(resv->r_len == 0); +} + +static inline int ocfs2_resmap_disabled(struct ocfs2_reservation_map *resmap) +{ + if (resmap->m_osb->osb_resv_level == 0) + return 1; + return 0; +} + +static void ocfs2_dump_resv(struct ocfs2_reservation_map *resmap) +{ + struct ocfs2_super *osb = resmap->m_osb; + struct rb_node *node; + struct ocfs2_alloc_reservation *resv; + int i = 0; + + mlog(ML_NOTICE, "Dumping resmap for device %s. Bitmap length: %u\n", + osb->dev_str, resmap->m_bitmap_len); + + node = rb_first(&resmap->m_reservations); + while (node) { + resv = rb_entry(node, struct ocfs2_alloc_reservation, r_node); + + mlog(ML_NOTICE, "start: %u\tend: %u\tlen: %u\tlast_start: %u" + "\tlast_len: %u\n", resv->r_start, + ocfs2_resv_end(resv), resv->r_len, resv->r_last_start, + resv->r_last_len); + + node = rb_next(node); + i++; + } + + mlog(ML_NOTICE, "%d reservations found. LRU follows\n", i); + + i = 0; + list_for_each_entry(resv, &resmap->m_lru, r_lru) { + mlog(ML_NOTICE, "LRU(%d) start: %u\tend: %u\tlen: %u\t" + "last_start: %u\tlast_len: %u\n", i, resv->r_start, + ocfs2_resv_end(resv), resv->r_len, resv->r_last_start, + resv->r_last_len); + + i++; + } +} + +#ifdef OCFS2_CHECK_RESERVATIONS +static int ocfs2_validate_resmap_bits(struct ocfs2_reservation_map *resmap, + int i, + struct ocfs2_alloc_reservation *resv) +{ + char *disk_bitmap = resmap->m_disk_bitmap; + unsigned int start = resv->r_start; + unsigned int end = ocfs2_resv_end(resv); + + while (start <= end) { + if (ocfs2_test_bit(start, disk_bitmap)) { + mlog(ML_ERROR, + "reservation %d covers an allocated area " + "starting at bit %u!\n", i, start); + return 1; + } + + start++; + } + return 0; +} + +static void ocfs2_check_resmap(struct ocfs2_reservation_map *resmap) +{ + unsigned int off = 0; + int i = 0; + struct rb_node *node; + struct ocfs2_alloc_reservation *resv; + + node = rb_first(&resmap->m_reservations); + while (node) { + resv = rb_entry(node, struct ocfs2_alloc_reservation, r_node); + + if (i > 0 && resv->r_start <= off) { + mlog(ML_ERROR, "reservation %d has bad start off!\n", + i); + goto bad; + } + + if (resv->r_len == 0) { + mlog(ML_ERROR, "reservation %d has no length!\n", + i); + goto bad; + } + + if (resv->r_start > ocfs2_resv_end(resv)) { + mlog(ML_ERROR, "reservation %d has invalid range!\n", + i); + goto bad; + } + + if (ocfs2_resv_end(resv) >= resmap->m_bitmap_len) { + mlog(ML_ERROR, "reservation %d extends past bitmap!\n", + i); + goto bad; + } + + if (ocfs2_validate_resmap_bits(resmap, i, resv)) + goto bad; + + off = ocfs2_resv_end(resv); + node = rb_next(node); + + i++; + } + return; + +bad: + ocfs2_dump_resv(resmap); + BUG(); +} +#else +static inline void ocfs2_check_resmap(struct ocfs2_reservation_map *resmap) +{ + +} +#endif + +void ocfs2_resv_init_once(struct ocfs2_alloc_reservation *resv) +{ + memset(resv, 0, sizeof(*resv)); + INIT_LIST_HEAD(&resv->r_lru); +} + +void ocfs2_resv_set_type(struct ocfs2_alloc_reservation *resv, + unsigned int flags) +{ + BUG_ON(flags & ~OCFS2_RESV_TYPES); + + resv->r_flags |= flags; +} + +int ocfs2_resmap_init(struct ocfs2_super *osb, + struct ocfs2_reservation_map *resmap) +{ + memset(resmap, 0, sizeof(*resmap)); + + resmap->m_osb = osb; + resmap->m_reservations = RB_ROOT; + /* m_bitmap_len is initialized to zero by the above memset. */ + INIT_LIST_HEAD(&resmap->m_lru); + + return 0; +} + +static void ocfs2_resv_mark_lru(struct ocfs2_reservation_map *resmap, + struct ocfs2_alloc_reservation *resv) +{ + assert_spin_locked(&resv_lock); + + if (!list_empty(&resv->r_lru)) + list_del_init(&resv->r_lru); + + list_add_tail(&resv->r_lru, &resmap->m_lru); +} + +static void __ocfs2_resv_trunc(struct ocfs2_alloc_reservation *resv) +{ + resv->r_len = 0; + resv->r_start = 0; +} + +static void ocfs2_resv_remove(struct ocfs2_reservation_map *resmap, + struct ocfs2_alloc_reservation *resv) +{ + if (resv->r_flags & OCFS2_RESV_FLAG_INUSE) { + list_del_init(&resv->r_lru); + rb_erase(&resv->r_node, &resmap->m_reservations); + resv->r_flags &= ~OCFS2_RESV_FLAG_INUSE; + } +} + +static void __ocfs2_resv_discard(struct ocfs2_reservation_map *resmap, + struct ocfs2_alloc_reservation *resv) +{ + assert_spin_locked(&resv_lock); + + __ocfs2_resv_trunc(resv); + /* + * last_len and last_start no longer make sense if + * we're changing the range of our allocations. + */ + resv->r_last_len = resv->r_last_start = 0; + + ocfs2_resv_remove(resmap, resv); +} + +/* does nothing if 'resv' is null */ +void ocfs2_resv_discard(struct ocfs2_reservation_map *resmap, + struct ocfs2_alloc_reservation *resv) +{ + if (resv) { + spin_lock(&resv_lock); + __ocfs2_resv_discard(resmap, resv); + spin_unlock(&resv_lock); + } +} + +static void ocfs2_resmap_clear_all_resv(struct ocfs2_reservation_map *resmap) +{ + struct rb_node *node; + struct ocfs2_alloc_reservation *resv; + + assert_spin_locked(&resv_lock); + + while ((node = rb_last(&resmap->m_reservations)) != NULL) { + resv = rb_entry(node, struct ocfs2_alloc_reservation, r_node); + + __ocfs2_resv_discard(resmap, resv); + } +} + +void ocfs2_resmap_restart(struct ocfs2_reservation_map *resmap, + unsigned int clen, char *disk_bitmap) +{ + if (ocfs2_resmap_disabled(resmap)) + return; + + spin_lock(&resv_lock); + + ocfs2_resmap_clear_all_resv(resmap); + resmap->m_bitmap_len = clen; + resmap->m_disk_bitmap = disk_bitmap; + + spin_unlock(&resv_lock); +} + +void ocfs2_resmap_uninit(struct ocfs2_reservation_map *resmap) +{ + /* Does nothing for now. Keep this around for API symmetry */ +} + +static void ocfs2_resv_insert(struct ocfs2_reservation_map *resmap, + struct ocfs2_alloc_reservation *new) +{ + struct rb_root *root = &resmap->m_reservations; + struct rb_node *parent = NULL; + struct rb_node **p = &root->rb_node; + struct ocfs2_alloc_reservation *tmp; + + assert_spin_locked(&resv_lock); + + trace_ocfs2_resv_insert(new->r_start, new->r_len); + + while (*p) { + parent = *p; + + tmp = rb_entry(parent, struct ocfs2_alloc_reservation, r_node); + + if (new->r_start < tmp->r_start) { + p = &(*p)->rb_left; + + /* + * This is a good place to check for + * overlapping reservations. + */ + BUG_ON(ocfs2_resv_end(new) >= tmp->r_start); + } else if (new->r_start > ocfs2_resv_end(tmp)) { + p = &(*p)->rb_right; + } else { + /* This should never happen! */ + mlog(ML_ERROR, "Duplicate reservation window!\n"); + BUG(); + } + } + + rb_link_node(&new->r_node, parent, p); + rb_insert_color(&new->r_node, root); + new->r_flags |= OCFS2_RESV_FLAG_INUSE; + + ocfs2_resv_mark_lru(resmap, new); + + ocfs2_check_resmap(resmap); +} + +/** + * ocfs2_find_resv_lhs() - find the window which contains goal + * @resmap: reservation map to search + * @goal: which bit to search for + * + * If a window containing that goal is not found, we return the window + * which comes before goal. Returns NULL on empty rbtree or no window + * before goal. + */ +static struct ocfs2_alloc_reservation * +ocfs2_find_resv_lhs(struct ocfs2_reservation_map *resmap, unsigned int goal) +{ + struct ocfs2_alloc_reservation *resv = NULL; + struct ocfs2_alloc_reservation *prev_resv = NULL; + struct rb_node *node = resmap->m_reservations.rb_node; + + assert_spin_locked(&resv_lock); + + if (!node) + return NULL; + + node = rb_first(&resmap->m_reservations); + while (node) { + resv = rb_entry(node, struct ocfs2_alloc_reservation, r_node); + + if (resv->r_start <= goal && ocfs2_resv_end(resv) >= goal) + break; + + /* Check if we overshot the reservation just before goal? */ + if (resv->r_start > goal) { + resv = prev_resv; + break; + } + + prev_resv = resv; + node = rb_next(node); + } + + return resv; +} + +/* + * We are given a range within the bitmap, which corresponds to a gap + * inside the reservations tree (search_start, search_len). The range + * can be anything from the whole bitmap, to a gap between + * reservations. + * + * The start value of *rstart is insignificant. + * + * This function searches the bitmap range starting at search_start + * with length search_len for a set of contiguous free bits. We try + * to find up to 'wanted' bits, but can sometimes return less. + * + * Returns the length of allocation, 0 if no free bits are found. + * + * *cstart and *clen will also be populated with the result. + */ +static int ocfs2_resmap_find_free_bits(struct ocfs2_reservation_map *resmap, + unsigned int wanted, + unsigned int search_start, + unsigned int search_len, + unsigned int *rstart, + unsigned int *rlen) +{ + void *bitmap = resmap->m_disk_bitmap; + unsigned int best_start, best_len = 0; + int offset, start, found; + + trace_ocfs2_resmap_find_free_bits_begin(search_start, search_len, + wanted, resmap->m_bitmap_len); + + found = best_start = best_len = 0; + + start = search_start; + while ((offset = ocfs2_find_next_zero_bit(bitmap, resmap->m_bitmap_len, + start)) != -1) { + /* Search reached end of the region */ + if (offset >= (search_start + search_len)) + break; + + if (offset == start) { + /* we found a zero */ + found++; + /* move start to the next bit to test */ + start++; + } else { + /* got a zero after some ones */ + found = 1; + start = offset + 1; + } + if (found > best_len) { + best_len = found; + best_start = start - found; + } + + if (found >= wanted) + break; + } + + if (best_len == 0) + return 0; + + if (best_len >= wanted) + best_len = wanted; + + *rlen = best_len; + *rstart = best_start; + + trace_ocfs2_resmap_find_free_bits_end(best_start, best_len); + + return *rlen; +} + +static void __ocfs2_resv_find_window(struct ocfs2_reservation_map *resmap, + struct ocfs2_alloc_reservation *resv, + unsigned int goal, unsigned int wanted) +{ + struct rb_root *root = &resmap->m_reservations; + unsigned int gap_start, gap_end, gap_len; + struct ocfs2_alloc_reservation *prev_resv, *next_resv; + struct rb_node *prev, *next; + unsigned int cstart, clen; + unsigned int best_start = 0, best_len = 0; + + /* + * Nasty cases to consider: + * + * - rbtree is empty + * - our window should be first in all reservations + * - our window should be last in all reservations + * - need to make sure we don't go past end of bitmap + */ + trace_ocfs2_resv_find_window_begin(resv->r_start, ocfs2_resv_end(resv), + goal, wanted, RB_EMPTY_ROOT(root)); + + assert_spin_locked(&resv_lock); + + if (RB_EMPTY_ROOT(root)) { + /* + * Easiest case - empty tree. We can just take + * whatever window of free bits we want. + */ + clen = ocfs2_resmap_find_free_bits(resmap, wanted, goal, + resmap->m_bitmap_len - goal, + &cstart, &clen); + + /* + * This should never happen - the local alloc window + * will always have free bits when we're called. + */ + BUG_ON(goal == 0 && clen == 0); + + if (clen == 0) + return; + + resv->r_start = cstart; + resv->r_len = clen; + + ocfs2_resv_insert(resmap, resv); + return; + } + + prev_resv = ocfs2_find_resv_lhs(resmap, goal); + + if (prev_resv == NULL) { + /* + * A NULL here means that the search code couldn't + * find a window that starts before goal. + * + * However, we can take the first window after goal, + * which is also by definition, the leftmost window in + * the entire tree. If we can find free bits in the + * gap between goal and the LHS window, then the + * reservation can safely be placed there. + * + * Otherwise we fall back to a linear search, checking + * the gaps in between windows for a place to + * allocate. + */ + + next = rb_first(root); + next_resv = rb_entry(next, struct ocfs2_alloc_reservation, + r_node); + + /* + * The search should never return such a window. (see + * comment above + */ + if (next_resv->r_start <= goal) { + mlog(ML_ERROR, "goal: %u next_resv: start %u len %u\n", + goal, next_resv->r_start, next_resv->r_len); + ocfs2_dump_resv(resmap); + BUG(); + } + + clen = ocfs2_resmap_find_free_bits(resmap, wanted, goal, + next_resv->r_start - goal, + &cstart, &clen); + if (clen) { + best_len = clen; + best_start = cstart; + if (best_len == wanted) + goto out_insert; + } + + prev_resv = next_resv; + next_resv = NULL; + } + + trace_ocfs2_resv_find_window_prev(prev_resv->r_start, + ocfs2_resv_end(prev_resv)); + + prev = &prev_resv->r_node; + + /* Now we do a linear search for a window, starting at 'prev_rsv' */ + while (1) { + next = rb_next(prev); + if (next) { + next_resv = rb_entry(next, + struct ocfs2_alloc_reservation, + r_node); + + gap_start = ocfs2_resv_end(prev_resv) + 1; + gap_end = next_resv->r_start - 1; + gap_len = gap_end - gap_start + 1; + } else { + /* + * We're at the rightmost edge of the + * tree. See if a reservation between this + * window and the end of the bitmap will work. + */ + gap_start = ocfs2_resv_end(prev_resv) + 1; + gap_len = resmap->m_bitmap_len - gap_start; + gap_end = resmap->m_bitmap_len - 1; + } + + trace_ocfs2_resv_find_window_next(next ? next_resv->r_start: -1, + next ? ocfs2_resv_end(next_resv) : -1); + /* + * No need to check this gap if we have already found + * a larger region of free bits. + */ + if (gap_len <= best_len) + goto next_resv; + + clen = ocfs2_resmap_find_free_bits(resmap, wanted, gap_start, + gap_len, &cstart, &clen); + if (clen == wanted) { + best_len = clen; + best_start = cstart; + goto out_insert; + } else if (clen > best_len) { + best_len = clen; + best_start = cstart; + } + +next_resv: + if (!next) + break; + + prev = next; + prev_resv = rb_entry(prev, struct ocfs2_alloc_reservation, + r_node); + } + +out_insert: + if (best_len) { + resv->r_start = best_start; + resv->r_len = best_len; + ocfs2_resv_insert(resmap, resv); + } +} + +static void ocfs2_cannibalize_resv(struct ocfs2_reservation_map *resmap, + struct ocfs2_alloc_reservation *resv, + unsigned int wanted) +{ + struct ocfs2_alloc_reservation *lru_resv; + int tmpwindow = !!(resv->r_flags & OCFS2_RESV_FLAG_TMP); + unsigned int min_bits; + + if (!tmpwindow) + min_bits = ocfs2_resv_window_bits(resmap, resv) >> 1; + else + min_bits = wanted; /* We at know the temp window will use all + * of these bits */ + + /* + * Take the first reservation off the LRU as our 'target'. We + * don't try to be smart about it. There might be a case for + * searching based on size but I don't have enough data to be + * sure. --Mark (3/16/2010) + */ + lru_resv = list_first_entry(&resmap->m_lru, + struct ocfs2_alloc_reservation, r_lru); + + trace_ocfs2_cannibalize_resv_begin(lru_resv->r_start, + lru_resv->r_len, + ocfs2_resv_end(lru_resv)); + + /* + * Cannibalize (some or all) of the target reservation and + * feed it to the current window. + */ + if (lru_resv->r_len <= min_bits) { + /* + * Discard completely if size is less than or equal to a + * reasonable threshold - 50% of window bits for non temporary + * windows. + */ + resv->r_start = lru_resv->r_start; + resv->r_len = lru_resv->r_len; + + __ocfs2_resv_discard(resmap, lru_resv); + } else { + unsigned int shrink; + if (tmpwindow) + shrink = min_bits; + else + shrink = lru_resv->r_len / 2; + + lru_resv->r_len -= shrink; + + resv->r_start = ocfs2_resv_end(lru_resv) + 1; + resv->r_len = shrink; + } + + trace_ocfs2_cannibalize_resv_end(resv->r_start, ocfs2_resv_end(resv), + resv->r_len, resv->r_last_start, + resv->r_last_len); + + ocfs2_resv_insert(resmap, resv); +} + +static void ocfs2_resv_find_window(struct ocfs2_reservation_map *resmap, + struct ocfs2_alloc_reservation *resv, + unsigned int wanted) +{ + unsigned int goal = 0; + + BUG_ON(!ocfs2_resv_empty(resv)); + + /* + * Begin by trying to get a window as close to the previous + * one as possible. Using the most recent allocation as a + * start goal makes sense. + */ + if (resv->r_last_len) { + goal = resv->r_last_start + resv->r_last_len; + if (goal >= resmap->m_bitmap_len) + goal = 0; + } + + __ocfs2_resv_find_window(resmap, resv, goal, wanted); + + /* Search from last alloc didn't work, try once more from beginning. */ + if (ocfs2_resv_empty(resv) && goal != 0) + __ocfs2_resv_find_window(resmap, resv, 0, wanted); + + if (ocfs2_resv_empty(resv)) { + /* + * Still empty? Pull oldest one off the LRU, remove it from + * tree, put this one in it's place. + */ + ocfs2_cannibalize_resv(resmap, resv, wanted); + } + + BUG_ON(ocfs2_resv_empty(resv)); +} + +int ocfs2_resmap_resv_bits(struct ocfs2_reservation_map *resmap, + struct ocfs2_alloc_reservation *resv, + int *cstart, int *clen) +{ + if (resv == NULL || ocfs2_resmap_disabled(resmap)) + return -ENOSPC; + + spin_lock(&resv_lock); + + if (ocfs2_resv_empty(resv)) { + /* + * We don't want to over-allocate for temporary + * windows. Otherwise, we run the risk of fragmenting the + * allocation space. + */ + unsigned int wanted = ocfs2_resv_window_bits(resmap, resv); + + if ((resv->r_flags & OCFS2_RESV_FLAG_TMP) || wanted < *clen) + wanted = *clen; + + /* + * Try to get a window here. If it works, we must fall + * through and test the bitmap . This avoids some + * ping-ponging of windows due to non-reserved space + * being allocation before we initialize a window for + * that inode. + */ + ocfs2_resv_find_window(resmap, resv, wanted); + trace_ocfs2_resmap_resv_bits(resv->r_start, resv->r_len); + } + + BUG_ON(ocfs2_resv_empty(resv)); + + *cstart = resv->r_start; + *clen = resv->r_len; + + spin_unlock(&resv_lock); + return 0; +} + +static void + ocfs2_adjust_resv_from_alloc(struct ocfs2_reservation_map *resmap, + struct ocfs2_alloc_reservation *resv, + unsigned int start, unsigned int end) +{ + unsigned int rhs = 0; + unsigned int old_end = ocfs2_resv_end(resv); + + BUG_ON(start != resv->r_start || old_end < end); + + /* + * Completely used? We can remove it then. + */ + if (old_end == end) { + __ocfs2_resv_discard(resmap, resv); + return; + } + + rhs = old_end - end; + + /* + * This should have been trapped above. + */ + BUG_ON(rhs == 0); + + resv->r_start = end + 1; + resv->r_len = old_end - resv->r_start + 1; +} + +void ocfs2_resmap_claimed_bits(struct ocfs2_reservation_map *resmap, + struct ocfs2_alloc_reservation *resv, + u32 cstart, u32 clen) +{ + unsigned int cend = cstart + clen - 1; + + if (resmap == NULL || ocfs2_resmap_disabled(resmap)) + return; + + if (resv == NULL) + return; + + BUG_ON(cstart != resv->r_start); + + spin_lock(&resv_lock); + + trace_ocfs2_resmap_claimed_bits_begin(cstart, cend, clen, resv->r_start, + ocfs2_resv_end(resv), resv->r_len, + resv->r_last_start, + resv->r_last_len); + + BUG_ON(cstart < resv->r_start); + BUG_ON(cstart > ocfs2_resv_end(resv)); + BUG_ON(cend > ocfs2_resv_end(resv)); + + ocfs2_adjust_resv_from_alloc(resmap, resv, cstart, cend); + resv->r_last_start = cstart; + resv->r_last_len = clen; + + /* + * May have been discarded above from + * ocfs2_adjust_resv_from_alloc(). + */ + if (!ocfs2_resv_empty(resv)) + ocfs2_resv_mark_lru(resmap, resv); + + trace_ocfs2_resmap_claimed_bits_end(resv->r_start, ocfs2_resv_end(resv), + resv->r_len, resv->r_last_start, + resv->r_last_len); + + ocfs2_check_resmap(resmap); + + spin_unlock(&resv_lock); +} diff --git a/fs/ocfs2/reservations.h b/fs/ocfs2/reservations.h new file mode 100644 index 00000000000..42c2b804f3f --- /dev/null +++ b/fs/ocfs2/reservations.h @@ -0,0 +1,159 @@ +/* -*- mode: c; c-basic-offset: 8; -*- + * vim: noexpandtab sw=8 ts=8 sts=0: + * + * reservations.h + * + * Allocation reservations function prototypes and structures. + * + * Copyright (C) 2010 Novell. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License version 2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + */ + +#ifndef OCFS2_RESERVATIONS_H +#define OCFS2_RESERVATIONS_H + +#include <linux/rbtree.h> + +#define OCFS2_DEFAULT_RESV_LEVEL 2 +#define OCFS2_MAX_RESV_LEVEL 9 +#define OCFS2_MIN_RESV_LEVEL 0 + +struct ocfs2_alloc_reservation { + struct rb_node r_node; + + unsigned int r_start; /* Beginning of current window */ + unsigned int r_len; /* Length of the window */ + + unsigned int r_last_len; /* Length of most recent alloc */ + unsigned int r_last_start; /* Start of most recent alloc */ + struct list_head r_lru; /* LRU list head */ + + unsigned int r_flags; +}; + +#define OCFS2_RESV_FLAG_INUSE 0x01 /* Set when r_node is part of a btree */ +#define OCFS2_RESV_FLAG_TMP 0x02 /* Temporary reservation, will be + * destroyed immedately after use */ +#define OCFS2_RESV_FLAG_DIR 0x04 /* Reservation is for an unindexed + * directory btree */ + +struct ocfs2_reservation_map { + struct rb_root m_reservations; + char *m_disk_bitmap; + + struct ocfs2_super *m_osb; + + /* The following are not initialized to meaningful values until a disk + * bitmap is provided. */ + u32 m_bitmap_len; /* Number of valid + * bits available */ + + struct list_head m_lru; /* LRU of reservations + * structures. */ + +}; + +void ocfs2_resv_init_once(struct ocfs2_alloc_reservation *resv); + +#define OCFS2_RESV_TYPES (OCFS2_RESV_FLAG_TMP|OCFS2_RESV_FLAG_DIR) +void ocfs2_resv_set_type(struct ocfs2_alloc_reservation *resv, + unsigned int flags); + +int ocfs2_dir_resv_allowed(struct ocfs2_super *osb); + +/** + * ocfs2_resv_discard() - truncate a reservation + * @resmap: + * @resv: the reservation to truncate. + * + * After this function is called, the reservation will be empty, and + * unlinked from the rbtree. + */ +void ocfs2_resv_discard(struct ocfs2_reservation_map *resmap, + struct ocfs2_alloc_reservation *resv); + + +/** + * ocfs2_resmap_init() - Initialize fields of a reservations bitmap + * @resmap: struct ocfs2_reservation_map to initialize + * @obj: unused for now + * @ops: unused for now + * @max_bitmap_bytes: Maximum size of the bitmap (typically blocksize) + * + * Only possible return value other than '0' is -ENOMEM for failure to + * allocation mirror bitmap. + */ +int ocfs2_resmap_init(struct ocfs2_super *osb, + struct ocfs2_reservation_map *resmap); + +/** + * ocfs2_resmap_restart() - "restart" a reservation bitmap + * @resmap: reservations bitmap + * @clen: Number of valid bits in the bitmap + * @disk_bitmap: the disk bitmap this resmap should refer to. + * + * Re-initialize the parameters of a reservation bitmap. This is + * useful for local alloc window slides. + * + * This function will call ocfs2_trunc_resv against all existing + * reservations. A future version will recalculate existing + * reservations based on the new bitmap. + */ +void ocfs2_resmap_restart(struct ocfs2_reservation_map *resmap, + unsigned int clen, char *disk_bitmap); + +/** + * ocfs2_resmap_uninit() - uninitialize a reservation bitmap structure + * @resmap: the struct ocfs2_reservation_map to uninitialize + */ +void ocfs2_resmap_uninit(struct ocfs2_reservation_map *resmap); + +/** + * ocfs2_resmap_resv_bits() - Return still-valid reservation bits + * @resmap: reservations bitmap + * @resv: reservation to base search from + * @cstart: start of proposed allocation + * @clen: length (in clusters) of proposed allocation + * + * Using the reservation data from resv, this function will compare + * resmap and resmap->m_disk_bitmap to determine what part (if any) of + * the reservation window is still clear to use. If resv is empty, + * this function will try to allocate a window for it. + * + * On success, zero is returned and the valid allocation area is set in cstart + * and clen. + * + * Returns -ENOSPC if reservations are disabled. + */ +int ocfs2_resmap_resv_bits(struct ocfs2_reservation_map *resmap, + struct ocfs2_alloc_reservation *resv, + int *cstart, int *clen); + +/** + * ocfs2_resmap_claimed_bits() - Tell the reservation code that bits were used. + * @resmap: reservations bitmap + * @resv: optional reservation to recalulate based on new bitmap + * @cstart: start of allocation in clusters + * @clen: end of allocation in clusters. + * + * Tell the reservation code that bits were used to fulfill allocation in + * resmap. The bits don't have to have been part of any existing + * reservation. But we must always call this function when bits are claimed. + * Internally, the reservations code will use this information to mark the + * reservations bitmap. If resv is passed, it's next allocation window will be + * calculated. It also expects that 'cstart' is the same as we passed back + * from ocfs2_resmap_resv_bits(). + */ +void ocfs2_resmap_claimed_bits(struct ocfs2_reservation_map *resmap, + struct ocfs2_alloc_reservation *resv, + u32 cstart, u32 clen); + +#endif /* OCFS2_RESERVATIONS_H */ diff --git a/fs/ocfs2/resize.c b/fs/ocfs2/resize.c index 424adaa5f90..d5da6f62414 100644 --- a/fs/ocfs2/resize.c +++ b/fs/ocfs2/resize.c @@ -27,7 +27,6 @@ #include <linux/fs.h> #include <linux/types.h> -#define MLOG_MASK_PREFIX ML_DISK_ALLOC #include <cluster/masklog.h> #include "ocfs2.h" @@ -39,6 +38,7 @@ #include "super.h" #include "sysfile.h" #include "uptodate.h" +#include "ocfs2_trace.h" #include "buffer_head_io.h" #include "suballoc.h" @@ -53,8 +53,6 @@ */ static u16 ocfs2_calc_new_backup_super(struct inode *inode, struct ocfs2_group_desc *gd, - int new_clusters, - u32 first_new_cluster, u16 cl_cpg, int set) { @@ -82,7 +80,6 @@ static u16 ocfs2_calc_new_backup_super(struct inode *inode, backups++; } - mlog_exit_void(); return backups; } @@ -103,11 +100,11 @@ static int ocfs2_update_last_group_and_inode(handle_t *handle, u16 cl_bpc = le16_to_cpu(cl->cl_bpc); u16 cl_cpg = le16_to_cpu(cl->cl_cpg); - mlog_entry("(new_clusters=%d, first_new_cluster = %u)\n", - new_clusters, first_new_cluster); + trace_ocfs2_update_last_group_and_inode(new_clusters, + first_new_cluster); - ret = ocfs2_journal_access_gd(handle, bm_inode, group_bh, - OCFS2_JOURNAL_ACCESS_WRITE); + ret = ocfs2_journal_access_gd(handle, INODE_CACHE(bm_inode), + group_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret < 0) { mlog_errno(ret); goto out; @@ -128,20 +125,14 @@ static int ocfs2_update_last_group_and_inode(handle_t *handle, OCFS2_FEATURE_COMPAT_BACKUP_SB)) { backups = ocfs2_calc_new_backup_super(bm_inode, group, - new_clusters, - first_new_cluster, cl_cpg, 1); le16_add_cpu(&group->bg_free_bits_count, -1 * backups); } - ret = ocfs2_journal_dirty(handle, group_bh); - if (ret < 0) { - mlog_errno(ret); - goto out_rollback; - } + ocfs2_journal_dirty(handle, group_bh); /* update the inode accordingly. */ - ret = ocfs2_journal_access_di(handle, bm_inode, bm_bh, + ret = ocfs2_journal_access_di(handle, INODE_CACHE(bm_inode), bm_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret < 0) { mlog_errno(ret); @@ -162,7 +153,7 @@ static int ocfs2_update_last_group_and_inode(handle_t *handle, spin_lock(&OCFS2_I(bm_inode)->ip_lock); OCFS2_I(bm_inode)->ip_clusters = le32_to_cpu(fe->i_clusters); - le64_add_cpu(&fe->i_size, new_clusters << osb->s_clustersize_bits); + le64_add_cpu(&fe->i_size, (u64)new_clusters << osb->s_clustersize_bits); spin_unlock(&OCFS2_I(bm_inode)->ip_lock); i_size_write(bm_inode, le64_to_cpu(fe->i_size)); @@ -172,15 +163,14 @@ out_rollback: if (ret < 0) { ocfs2_calc_new_backup_super(bm_inode, group, - new_clusters, - first_new_cluster, cl_cpg, 0); le16_add_cpu(&group->bg_free_bits_count, backups); le16_add_cpu(&group->bg_bits, -1 * num_bits); le16_add_cpu(&group->bg_free_bits_count, -1 * num_bits); } out: - mlog_exit(ret); + if (ret) + mlog_errno(ret); return ret; } @@ -285,8 +275,6 @@ int ocfs2_group_extend(struct inode * inode, int new_clusters) u32 first_new_cluster; u64 lgd_blkno; - mlog_entry_void(); - if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb)) return -EROFS; @@ -319,7 +307,8 @@ int ocfs2_group_extend(struct inode * inode, int new_clusters) BUG_ON(!OCFS2_IS_VALID_DINODE(fe)); if (le16_to_cpu(fe->id2.i_chain.cl_cpg) != - ocfs2_group_bitmap_size(osb->sb) * 8) { + ocfs2_group_bitmap_size(osb->sb, 0, + osb->s_feature_incompat) * 8) { mlog(ML_ERROR, "The disk is too old and small. " "Force to do offline resize."); ret = -EINVAL; @@ -345,7 +334,8 @@ int ocfs2_group_extend(struct inode * inode, int new_clusters) goto out_unlock; } - mlog(0, "extend the last group at %llu, new clusters = %d\n", + + trace_ocfs2_group_extend( (unsigned long long)le64_to_cpu(group->bg_blkno), new_clusters); handle = ocfs2_start_trans(osb, OCFS2_GROUP_EXTEND_CREDITS); @@ -380,7 +370,6 @@ out_mutex: iput(main_bm_inode); out: - mlog_exit_void(); return ret; } @@ -474,8 +463,7 @@ int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input) struct ocfs2_chain_list *cl; struct ocfs2_chain_rec *cr; u16 cl_bpc; - - mlog_entry_void(); + u64 bg_ptr; if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb)) return -EROFS; @@ -500,7 +488,8 @@ int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input) fe = (struct ocfs2_dinode *)main_bm_bh->b_data; if (le16_to_cpu(fe->id2.i_chain.cl_cpg) != - ocfs2_group_bitmap_size(osb->sb) * 8) { + ocfs2_group_bitmap_size(osb->sb, 0, + osb->s_feature_incompat) * 8) { mlog(ML_ERROR, "The disk is too old and small." " Force to do offline resize."); ret = -EINVAL; @@ -514,47 +503,44 @@ int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input) goto out_unlock; } - ocfs2_set_new_buffer_uptodate(inode, group_bh); + ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode), group_bh); ret = ocfs2_verify_group_and_input(main_bm_inode, fe, input, group_bh); if (ret) { mlog_errno(ret); - goto out_unlock; + goto out_free_group_bh; } - mlog(0, "Add a new group %llu in chain = %u, length = %u\n", - (unsigned long long)input->group, input->chain, input->clusters); + trace_ocfs2_group_add((unsigned long long)input->group, + input->chain, input->clusters, input->frees); handle = ocfs2_start_trans(osb, OCFS2_GROUP_ADD_CREDITS); if (IS_ERR(handle)) { mlog_errno(PTR_ERR(handle)); ret = -EINVAL; - goto out_unlock; + goto out_free_group_bh; } cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc); cl = &fe->id2.i_chain; cr = &cl->cl_recs[input->chain]; - ret = ocfs2_journal_access_gd(handle, main_bm_inode, group_bh, - OCFS2_JOURNAL_ACCESS_WRITE); + ret = ocfs2_journal_access_gd(handle, INODE_CACHE(main_bm_inode), + group_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret < 0) { mlog_errno(ret); goto out_commit; } group = (struct ocfs2_group_desc *)group_bh->b_data; + bg_ptr = le64_to_cpu(group->bg_next_group); group->bg_next_group = cr->c_blkno; + ocfs2_journal_dirty(handle, group_bh); - ret = ocfs2_journal_dirty(handle, group_bh); - if (ret < 0) { - mlog_errno(ret); - goto out_commit; - } - - ret = ocfs2_journal_access_di(handle, main_bm_inode, main_bm_bh, - OCFS2_JOURNAL_ACCESS_WRITE); + ret = ocfs2_journal_access_di(handle, INODE_CACHE(main_bm_inode), + main_bm_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret < 0) { + group->bg_next_group = cpu_to_le64(bg_ptr); mlog_errno(ret); goto out_commit; } @@ -577,7 +563,7 @@ int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input) spin_lock(&OCFS2_I(main_bm_inode)->ip_lock); OCFS2_I(main_bm_inode)->ip_clusters = le32_to_cpu(fe->i_clusters); - le64_add_cpu(&fe->i_size, input->clusters << osb->s_clustersize_bits); + le64_add_cpu(&fe->i_size, (u64)input->clusters << osb->s_clustersize_bits); spin_unlock(&OCFS2_I(main_bm_inode)->ip_lock); i_size_write(main_bm_inode, le64_to_cpu(fe->i_size)); @@ -585,8 +571,11 @@ int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input) out_commit: ocfs2_commit_trans(osb, handle); -out_unlock: + +out_free_group_bh: brelse(group_bh); + +out_unlock: brelse(main_bm_bh); ocfs2_inode_unlock(main_bm_inode, 1); @@ -596,6 +585,5 @@ out_mutex: iput(main_bm_inode); out: - mlog_exit_void(); return ret; } diff --git a/fs/ocfs2/slot_map.c b/fs/ocfs2/slot_map.c index 40661e7824e..1424c151ccc 100644 --- a/fs/ocfs2/slot_map.c +++ b/fs/ocfs2/slot_map.c @@ -27,7 +27,6 @@ #include <linux/slab.h> #include <linux/highmem.h> -#define MLOG_MASK_PREFIX ML_SUPER #include <cluster/masklog.h> #include "ocfs2.h" @@ -39,6 +38,7 @@ #include "slot_map.h" #include "super.h" #include "sysfile.h" +#include "ocfs2_trace.h" #include "buffer_head_io.h" @@ -142,16 +142,15 @@ int ocfs2_refresh_slot_info(struct ocfs2_super *osb) BUG_ON(si->si_blocks == 0); BUG_ON(si->si_bh == NULL); - mlog(0, "Refreshing slot map, reading %u block(s)\n", - si->si_blocks); + trace_ocfs2_refresh_slot_info(si->si_blocks); /* * We pass -1 as blocknr because we expect all of si->si_bh to * be !NULL. Thus, ocfs2_read_blocks() will ignore blocknr. If * this is not true, the read of -1 (UINT64_MAX) will fail. */ - ret = ocfs2_read_blocks(si->si_inode, -1, si->si_blocks, si->si_bh, - OCFS2_BH_IGNORE_CACHE, NULL); + ret = ocfs2_read_blocks(INODE_CACHE(si->si_inode), -1, si->si_blocks, + si->si_bh, OCFS2_BH_IGNORE_CACHE, NULL); if (ret == 0) { spin_lock(&osb->osb_lock); ocfs2_update_slot_info(si); @@ -213,7 +212,7 @@ static int ocfs2_update_disk_slot(struct ocfs2_super *osb, ocfs2_update_disk_slot_old(si, slot_num, &bh); spin_unlock(&osb->osb_lock); - status = ocfs2_write_block(osb, bh, si->si_inode); + status = ocfs2_write_block(osb, bh, INODE_CACHE(si->si_inode)); if (status < 0) mlog_errno(status); @@ -357,7 +356,7 @@ static int ocfs2_map_slot_buffers(struct ocfs2_super *osb, { int status = 0; u64 blkno; - unsigned long long blocks, bytes; + unsigned long long blocks, bytes = 0; unsigned int i; struct buffer_head *bh; @@ -381,8 +380,7 @@ static int ocfs2_map_slot_buffers(struct ocfs2_super *osb, /* The size checks above should ensure this */ BUG_ON((osb->max_slots / si->si_slots_per_block) > blocks); - mlog(0, "Slot map needs %u buffers for %llu bytes\n", - si->si_blocks, bytes); + trace_ocfs2_map_slot_buffers(bytes, si->si_blocks); si->si_bh = kzalloc(sizeof(struct buffer_head *) * si->si_blocks, GFP_KERNEL); @@ -400,12 +398,11 @@ static int ocfs2_map_slot_buffers(struct ocfs2_super *osb, goto bail; } - mlog(0, "Reading slot map block %u at %llu\n", i, - (unsigned long long)blkno); + trace_ocfs2_map_slot_buffers_block((unsigned long long)blkno, i); bh = NULL; /* Acquire a fresh bh */ - status = ocfs2_read_blocks(si->si_inode, blkno, 1, &bh, - OCFS2_BH_IGNORE_CACHE, NULL); + status = ocfs2_read_blocks(INODE_CACHE(si->si_inode), blkno, + 1, &bh, OCFS2_BH_IGNORE_CACHE, NULL); if (status < 0) { mlog_errno(status); goto bail; @@ -475,8 +472,6 @@ int ocfs2_find_slot(struct ocfs2_super *osb) int slot; struct ocfs2_slot_info *si; - mlog_entry_void(); - si = osb->slot_info; spin_lock(&osb->osb_lock); @@ -498,21 +493,20 @@ int ocfs2_find_slot(struct ocfs2_super *osb) goto bail; } } else - mlog(ML_NOTICE, "slot %d is already allocated to this node!\n", - slot); + printk(KERN_INFO "ocfs2: Slot %d on device (%s) was already " + "allocated to this node!\n", slot, osb->dev_str); ocfs2_set_slot(si, slot, osb->node_num); osb->slot_num = slot; spin_unlock(&osb->osb_lock); - mlog(0, "taking node slot %d\n", osb->slot_num); + trace_ocfs2_find_slot(osb->slot_num); status = ocfs2_update_disk_slot(osb, si, osb->slot_num); if (status < 0) mlog_errno(status); bail: - mlog_exit(status); return status; } diff --git a/fs/ocfs2/stack_o2cb.c b/fs/ocfs2/stack_o2cb.c index fcd120f1493..1724d43d3da 100644 --- a/fs/ocfs2/stack_o2cb.c +++ b/fs/ocfs2/stack_o2cb.c @@ -17,7 +17,9 @@ * General Public License for more details. */ +#include <linux/kernel.h> #include <linux/crc32.h> +#include <linux/slab.h> #include <linux/module.h> /* Needed for AOP_TRUNCATED_PAGE in mlog_errno() */ @@ -26,6 +28,7 @@ #include "cluster/masklog.h" #include "cluster/nodemanager.h" #include "cluster/heartbeat.h" +#include "cluster/tcp.h" #include "stackglue.h" @@ -153,31 +156,30 @@ static int status_map[] = { static int dlm_status_to_errno(enum dlm_status status) { - BUG_ON(status > (sizeof(status_map) / sizeof(status_map[0]))); + BUG_ON(status < 0 || status >= ARRAY_SIZE(status_map)); return status_map[status]; } static void o2dlm_lock_ast_wrapper(void *astarg) { - BUG_ON(o2cb_stack.sp_proto == NULL); + struct ocfs2_dlm_lksb *lksb = astarg; - o2cb_stack.sp_proto->lp_lock_ast(astarg); + lksb->lksb_conn->cc_proto->lp_lock_ast(lksb); } static void o2dlm_blocking_ast_wrapper(void *astarg, int level) { - BUG_ON(o2cb_stack.sp_proto == NULL); + struct ocfs2_dlm_lksb *lksb = astarg; - o2cb_stack.sp_proto->lp_blocking_ast(astarg, level); + lksb->lksb_conn->cc_proto->lp_blocking_ast(lksb, level); } static void o2dlm_unlock_ast_wrapper(void *astarg, enum dlm_status status) { + struct ocfs2_dlm_lksb *lksb = astarg; int error = dlm_status_to_errno(status); - BUG_ON(o2cb_stack.sp_proto == NULL); - /* * In o2dlm, you can get both the lock_ast() for the lock being * granted and the unlock_ast() for the CANCEL failing. A @@ -192,16 +194,15 @@ static void o2dlm_unlock_ast_wrapper(void *astarg, enum dlm_status status) if (status == DLM_CANCELGRANT) return; - o2cb_stack.sp_proto->lp_unlock_ast(astarg, error); + lksb->lksb_conn->cc_proto->lp_unlock_ast(lksb, error); } static int o2cb_dlm_lock(struct ocfs2_cluster_connection *conn, int mode, - union ocfs2_dlm_lksb *lksb, + struct ocfs2_dlm_lksb *lksb, u32 flags, void *name, - unsigned int namelen, - void *astarg) + unsigned int namelen) { enum dlm_status status; int o2dlm_mode = mode_to_o2dlm(mode); @@ -210,43 +211,107 @@ static int o2cb_dlm_lock(struct ocfs2_cluster_connection *conn, status = dlmlock(conn->cc_lockspace, o2dlm_mode, &lksb->lksb_o2dlm, o2dlm_flags, name, namelen, - o2dlm_lock_ast_wrapper, astarg, + o2dlm_lock_ast_wrapper, lksb, o2dlm_blocking_ast_wrapper); ret = dlm_status_to_errno(status); return ret; } static int o2cb_dlm_unlock(struct ocfs2_cluster_connection *conn, - union ocfs2_dlm_lksb *lksb, - u32 flags, - void *astarg) + struct ocfs2_dlm_lksb *lksb, + u32 flags) { enum dlm_status status; int o2dlm_flags = flags_to_o2dlm(flags); int ret; status = dlmunlock(conn->cc_lockspace, &lksb->lksb_o2dlm, - o2dlm_flags, o2dlm_unlock_ast_wrapper, astarg); + o2dlm_flags, o2dlm_unlock_ast_wrapper, lksb); ret = dlm_status_to_errno(status); return ret; } -static int o2cb_dlm_lock_status(union ocfs2_dlm_lksb *lksb) +static int o2cb_dlm_lock_status(struct ocfs2_dlm_lksb *lksb) { return dlm_status_to_errno(lksb->lksb_o2dlm.status); } -static void *o2cb_dlm_lvb(union ocfs2_dlm_lksb *lksb) +/* + * o2dlm aways has a "valid" LVB. If the dlm loses track of the LVB + * contents, it will zero out the LVB. Thus the caller can always trust + * the contents. + */ +static int o2cb_dlm_lvb_valid(struct ocfs2_dlm_lksb *lksb) +{ + return 1; +} + +static void *o2cb_dlm_lvb(struct ocfs2_dlm_lksb *lksb) { return (void *)(lksb->lksb_o2dlm.lvb); } -static void o2cb_dump_lksb(union ocfs2_dlm_lksb *lksb) +static void o2cb_dump_lksb(struct ocfs2_dlm_lksb *lksb) { dlm_print_one_lock(lksb->lksb_o2dlm.lockid); } /* + * Check if this node is heartbeating and is connected to all other + * heartbeating nodes. + */ +static int o2cb_cluster_check(void) +{ + u8 node_num; + int i; + unsigned long hbmap[BITS_TO_LONGS(O2NM_MAX_NODES)]; + unsigned long netmap[BITS_TO_LONGS(O2NM_MAX_NODES)]; + + node_num = o2nm_this_node(); + if (node_num == O2NM_MAX_NODES) { + printk(KERN_ERR "o2cb: This node has not been configured.\n"); + return -EINVAL; + } + + /* + * o2dlm expects o2net sockets to be created. If not, then + * dlm_join_domain() fails with a stack of errors which are both cryptic + * and incomplete. The idea here is to detect upfront whether we have + * managed to connect to all nodes or not. If not, then list the nodes + * to allow the user to check the configuration (incorrect IP, firewall, + * etc.) Yes, this is racy. But its not the end of the world. + */ +#define O2CB_MAP_STABILIZE_COUNT 60 + for (i = 0; i < O2CB_MAP_STABILIZE_COUNT; ++i) { + o2hb_fill_node_map(hbmap, sizeof(hbmap)); + if (!test_bit(node_num, hbmap)) { + printk(KERN_ERR "o2cb: %s heartbeat has not been " + "started.\n", (o2hb_global_heartbeat_active() ? + "Global" : "Local")); + return -EINVAL; + } + o2net_fill_node_map(netmap, sizeof(netmap)); + /* Force set the current node to allow easy compare */ + set_bit(node_num, netmap); + if (!memcmp(hbmap, netmap, sizeof(hbmap))) + return 0; + if (i < O2CB_MAP_STABILIZE_COUNT) + msleep(1000); + } + + printk(KERN_ERR "o2cb: This node could not connect to nodes:"); + i = -1; + while ((i = find_next_bit(hbmap, O2NM_MAX_NODES, + i + 1)) < O2NM_MAX_NODES) { + if (!test_bit(i, netmap)) + printk(" %u", i); + } + printk(".\n"); + + return -ENOTCONN; +} + +/* * Called from the dlm when it's about to evict a node. This is how the * classic stack signals node death. */ @@ -254,8 +319,8 @@ static void o2dlm_eviction_cb(int node_num, void *data) { struct ocfs2_cluster_connection *conn = data; - mlog(ML_NOTICE, "o2dlm has evicted node %d from group %.*s\n", - node_num, conn->cc_namelen, conn->cc_name); + printk(KERN_NOTICE "o2cb: o2dlm has evicted node %d from domain %.*s\n", + node_num, conn->cc_namelen, conn->cc_name); conn->cc_recovery_handler(node_num, conn->cc_recovery_data); } @@ -266,15 +331,16 @@ static int o2cb_cluster_connect(struct ocfs2_cluster_connection *conn) u32 dlm_key; struct dlm_ctxt *dlm; struct o2dlm_private *priv; - struct dlm_protocol_version dlm_version; + struct dlm_protocol_version fs_version; BUG_ON(conn == NULL); - BUG_ON(o2cb_stack.sp_proto == NULL); + BUG_ON(conn->cc_proto == NULL); - /* for now we only have one cluster/node, make sure we see it - * in the heartbeat universe */ - if (!o2hb_check_local_node_heartbeating()) { - rc = -EINVAL; + /* Ensure cluster stack is up and all nodes are connected */ + rc = o2cb_cluster_check(); + if (rc) { + printk(KERN_ERR "o2cb: Cluster check failed. Fix errors " + "before retrying.\n"); goto out; } @@ -293,24 +359,24 @@ static int o2cb_cluster_connect(struct ocfs2_cluster_connection *conn) /* used by the dlm code to make message headers unique, each * node in this domain must agree on this. */ dlm_key = crc32_le(0, conn->cc_name, conn->cc_namelen); - dlm_version.pv_major = conn->cc_version.pv_major; - dlm_version.pv_minor = conn->cc_version.pv_minor; + fs_version.pv_major = conn->cc_version.pv_major; + fs_version.pv_minor = conn->cc_version.pv_minor; - dlm = dlm_register_domain(conn->cc_name, dlm_key, &dlm_version); + dlm = dlm_register_domain(conn->cc_name, dlm_key, &fs_version); if (IS_ERR(dlm)) { rc = PTR_ERR(dlm); mlog_errno(rc); goto out_free; } - conn->cc_version.pv_major = dlm_version.pv_major; - conn->cc_version.pv_minor = dlm_version.pv_minor; + conn->cc_version.pv_major = fs_version.pv_major; + conn->cc_version.pv_minor = fs_version.pv_minor; conn->cc_lockspace = dlm; dlm_register_eviction_cb(dlm, &priv->op_eviction_cb); out_free: - if (rc && conn->cc_private) + if (rc) kfree(conn->cc_private); out: @@ -332,7 +398,8 @@ static int o2cb_cluster_disconnect(struct ocfs2_cluster_connection *conn) return 0; } -static int o2cb_cluster_this_node(unsigned int *node) +static int o2cb_cluster_this_node(struct ocfs2_cluster_connection *conn, + unsigned int *node) { int node_num; @@ -354,6 +421,7 @@ static struct ocfs2_stack_operations o2cb_stack_ops = { .dlm_lock = o2cb_dlm_lock, .dlm_unlock = o2cb_dlm_unlock, .lock_status = o2cb_dlm_lock_status, + .lvb_valid = o2cb_dlm_lvb_valid, .lock_lvb = o2cb_dlm_lvb, .dump_lksb = o2cb_dump_lksb, }; diff --git a/fs/ocfs2/stack_user.c b/fs/ocfs2/stack_user.c index 9b76d41a8ac..13a8537d8e8 100644 --- a/fs/ocfs2/stack_user.c +++ b/fs/ocfs2/stack_user.c @@ -21,11 +21,11 @@ #include <linux/fs.h> #include <linux/miscdevice.h> #include <linux/mutex.h> -#include <linux/smp_lock.h> +#include <linux/slab.h> #include <linux/reboot.h> +#include <linux/sched.h> #include <asm/uaccess.h> -#include "ocfs2.h" /* For struct ocfs2_lock_res */ #include "stackglue.h" #include <linux/dlm_plock.h> @@ -63,8 +63,8 @@ * negotiated by the client. The client negotiates based on the maximum * version advertised in /sys/fs/ocfs2/max_locking_protocol. The major * number from the "SETV" message must match - * ocfs2_user_plugin.sp_proto->lp_max_version.pv_major, and the minor number - * must be less than or equal to ...->lp_max_version.pv_minor. + * ocfs2_user_plugin.sp_max_proto.pv_major, and the minor number + * must be less than or equal to ...sp_max_version.pv_minor. * * Once this information has been set, mounts will be allowed. From this * point on, the "DOWN" message can be sent for node down notification. @@ -103,6 +103,12 @@ #define OCFS2_TEXT_UUID_LEN 32 #define OCFS2_CONTROL_MESSAGE_VERNUM_LEN 2 #define OCFS2_CONTROL_MESSAGE_NODENUM_LEN 8 +#define VERSION_LOCK "version_lock" + +enum ocfs2_connection_type { + WITH_CONTROLD, + NO_CONTROLD +}; /* * ocfs2_live_connection is refcounted because the filesystem and @@ -111,6 +117,13 @@ struct ocfs2_live_connection { struct list_head oc_list; struct ocfs2_cluster_connection *oc_conn; + enum ocfs2_connection_type oc_type; + atomic_t oc_this_node; + int oc_our_slot; + struct dlm_lksb oc_version_lksb; + char oc_lvb[DLM_LVB_LEN]; + struct completion oc_sync_wait; + wait_queue_head_t oc_wait; }; struct ocfs2_control_private { @@ -191,7 +204,7 @@ static struct ocfs2_live_connection *ocfs2_connection_find(const char *name) return c; } - return c; + return NULL; } /* @@ -199,20 +212,15 @@ static struct ocfs2_live_connection *ocfs2_connection_find(const char *name) * mount path. Since the VFS prevents multiple calls to * fill_super(), we can't get dupes here. */ -static int ocfs2_live_connection_new(struct ocfs2_cluster_connection *conn, - struct ocfs2_live_connection **c_ret) +static int ocfs2_live_connection_attach(struct ocfs2_cluster_connection *conn, + struct ocfs2_live_connection *c) { int rc = 0; - struct ocfs2_live_connection *c; - - c = kzalloc(sizeof(struct ocfs2_live_connection), GFP_KERNEL); - if (!c) - return -ENOMEM; mutex_lock(&ocfs2_control_lock); c->oc_conn = conn; - if (atomic_read(&ocfs2_control_opened)) + if ((c->oc_type == NO_CONTROLD) || atomic_read(&ocfs2_control_opened)) list_add(&c->oc_list, &ocfs2_live_connection_list); else { printk(KERN_ERR @@ -221,12 +229,6 @@ static int ocfs2_live_connection_new(struct ocfs2_cluster_connection *conn, } mutex_unlock(&ocfs2_control_lock); - - if (!rc) - *c_ret = c; - else - kfree(c); - return rc; } @@ -401,7 +403,7 @@ static int ocfs2_control_do_setversion_msg(struct file *file, char *ptr = NULL; struct ocfs2_control_private *p = file->private_data; struct ocfs2_protocol_version *max = - &ocfs2_user_plugin.sp_proto->lp_max_version; + &ocfs2_user_plugin.sp_max_proto; if (ocfs2_control_get_handshake_state(file) != OCFS2_CONTROL_HANDSHAKE_PROTOCOL) @@ -612,12 +614,10 @@ static int ocfs2_control_open(struct inode *inode, struct file *file) return -ENOMEM; p->op_this_node = -1; - lock_kernel(); mutex_lock(&ocfs2_control_lock); file->private_data = p; list_add(&p->op_list, &ocfs2_control_private_list); mutex_unlock(&ocfs2_control_lock); - unlock_kernel(); return 0; } @@ -628,6 +628,7 @@ static const struct file_operations ocfs2_control_fops = { .read = ocfs2_control_read, .write = ocfs2_control_write, .owner = THIS_MODULE, + .llseek = default_llseek, }; static struct miscdevice ocfs2_control_device = { @@ -664,18 +665,10 @@ static void ocfs2_control_exit(void) -rc); } -static struct dlm_lksb *fsdlm_astarg_to_lksb(void *astarg) -{ - struct ocfs2_lock_res *res = astarg; - return &res->l_lksb.lksb_fsdlm; -} - static void fsdlm_lock_ast_wrapper(void *astarg) { - struct dlm_lksb *lksb = fsdlm_astarg_to_lksb(astarg); - int status = lksb->sb_status; - - BUG_ON(ocfs2_user_plugin.sp_proto == NULL); + struct ocfs2_dlm_lksb *lksb = astarg; + int status = lksb->lksb_fsdlm.sb_status; /* * For now we're punting on the issue of other non-standard errors @@ -688,25 +681,24 @@ static void fsdlm_lock_ast_wrapper(void *astarg) */ if (status == -DLM_EUNLOCK || status == -DLM_ECANCEL) - ocfs2_user_plugin.sp_proto->lp_unlock_ast(astarg, 0); + lksb->lksb_conn->cc_proto->lp_unlock_ast(lksb, 0); else - ocfs2_user_plugin.sp_proto->lp_lock_ast(astarg); + lksb->lksb_conn->cc_proto->lp_lock_ast(lksb); } static void fsdlm_blocking_ast_wrapper(void *astarg, int level) { - BUG_ON(ocfs2_user_plugin.sp_proto == NULL); + struct ocfs2_dlm_lksb *lksb = astarg; - ocfs2_user_plugin.sp_proto->lp_blocking_ast(astarg, level); + lksb->lksb_conn->cc_proto->lp_blocking_ast(lksb, level); } static int user_dlm_lock(struct ocfs2_cluster_connection *conn, int mode, - union ocfs2_dlm_lksb *lksb, + struct ocfs2_dlm_lksb *lksb, u32 flags, void *name, - unsigned int namelen, - void *astarg) + unsigned int namelen) { int ret; @@ -716,29 +708,35 @@ static int user_dlm_lock(struct ocfs2_cluster_connection *conn, ret = dlm_lock(conn->cc_lockspace, mode, &lksb->lksb_fsdlm, flags|DLM_LKF_NODLCKWT, name, namelen, 0, - fsdlm_lock_ast_wrapper, astarg, + fsdlm_lock_ast_wrapper, lksb, fsdlm_blocking_ast_wrapper); return ret; } static int user_dlm_unlock(struct ocfs2_cluster_connection *conn, - union ocfs2_dlm_lksb *lksb, - u32 flags, - void *astarg) + struct ocfs2_dlm_lksb *lksb, + u32 flags) { int ret; ret = dlm_unlock(conn->cc_lockspace, lksb->lksb_fsdlm.sb_lkid, - flags, &lksb->lksb_fsdlm, astarg); + flags, &lksb->lksb_fsdlm, lksb); return ret; } -static int user_dlm_lock_status(union ocfs2_dlm_lksb *lksb) +static int user_dlm_lock_status(struct ocfs2_dlm_lksb *lksb) { return lksb->lksb_fsdlm.sb_status; } -static void *user_dlm_lvb(union ocfs2_dlm_lksb *lksb) +static int user_dlm_lvb_valid(struct ocfs2_dlm_lksb *lksb) +{ + int invalid = lksb->lksb_fsdlm.sb_flags & DLM_SBF_VALNOTVALID; + + return !invalid; +} + +static void *user_dlm_lvb(struct ocfs2_dlm_lksb *lksb) { if (!lksb->lksb_fsdlm.sb_lvbptr) lksb->lksb_fsdlm.sb_lvbptr = (char *)lksb + @@ -746,7 +744,7 @@ static void *user_dlm_lvb(union ocfs2_dlm_lksb *lksb) return (void *)(lksb->lksb_fsdlm.sb_lvbptr); } -static void user_dlm_dump_lksb(union ocfs2_dlm_lksb *lksb) +static void user_dlm_dump_lksb(struct ocfs2_dlm_lksb *lksb) { } @@ -804,18 +802,251 @@ static int fs_protocol_compare(struct ocfs2_protocol_version *existing, return 0; } +static void lvb_to_version(char *lvb, struct ocfs2_protocol_version *ver) +{ + struct ocfs2_protocol_version *pv = + (struct ocfs2_protocol_version *)lvb; + /* + * ocfs2_protocol_version has two u8 variables, so we don't + * need any endian conversion. + */ + ver->pv_major = pv->pv_major; + ver->pv_minor = pv->pv_minor; +} + +static void version_to_lvb(struct ocfs2_protocol_version *ver, char *lvb) +{ + struct ocfs2_protocol_version *pv = + (struct ocfs2_protocol_version *)lvb; + /* + * ocfs2_protocol_version has two u8 variables, so we don't + * need any endian conversion. + */ + pv->pv_major = ver->pv_major; + pv->pv_minor = ver->pv_minor; +} + +static void sync_wait_cb(void *arg) +{ + struct ocfs2_cluster_connection *conn = arg; + struct ocfs2_live_connection *lc = conn->cc_private; + complete(&lc->oc_sync_wait); +} + +static int sync_unlock(struct ocfs2_cluster_connection *conn, + struct dlm_lksb *lksb, char *name) +{ + int error; + struct ocfs2_live_connection *lc = conn->cc_private; + + error = dlm_unlock(conn->cc_lockspace, lksb->sb_lkid, 0, lksb, conn); + if (error) { + printk(KERN_ERR "%s lkid %x error %d\n", + name, lksb->sb_lkid, error); + return error; + } + + wait_for_completion(&lc->oc_sync_wait); + + if (lksb->sb_status != -DLM_EUNLOCK) { + printk(KERN_ERR "%s lkid %x status %d\n", + name, lksb->sb_lkid, lksb->sb_status); + return -1; + } + return 0; +} + +static int sync_lock(struct ocfs2_cluster_connection *conn, + int mode, uint32_t flags, + struct dlm_lksb *lksb, char *name) +{ + int error, status; + struct ocfs2_live_connection *lc = conn->cc_private; + + error = dlm_lock(conn->cc_lockspace, mode, lksb, flags, + name, strlen(name), + 0, sync_wait_cb, conn, NULL); + if (error) { + printk(KERN_ERR "%s lkid %x flags %x mode %d error %d\n", + name, lksb->sb_lkid, flags, mode, error); + return error; + } + + wait_for_completion(&lc->oc_sync_wait); + + status = lksb->sb_status; + + if (status && status != -EAGAIN) { + printk(KERN_ERR "%s lkid %x flags %x mode %d status %d\n", + name, lksb->sb_lkid, flags, mode, status); + } + + return status; +} + + +static int version_lock(struct ocfs2_cluster_connection *conn, int mode, + int flags) +{ + struct ocfs2_live_connection *lc = conn->cc_private; + return sync_lock(conn, mode, flags, + &lc->oc_version_lksb, VERSION_LOCK); +} + +static int version_unlock(struct ocfs2_cluster_connection *conn) +{ + struct ocfs2_live_connection *lc = conn->cc_private; + return sync_unlock(conn, &lc->oc_version_lksb, VERSION_LOCK); +} + +/* get_protocol_version() + * + * To exchange ocfs2 versioning, we use the LVB of the version dlm lock. + * The algorithm is: + * 1. Attempt to take the lock in EX mode (non-blocking). + * 2. If successful (which means it is the first mount), write the + * version number and downconvert to PR lock. + * 3. If unsuccessful (returns -EAGAIN), read the version from the LVB after + * taking the PR lock. + */ + +static int get_protocol_version(struct ocfs2_cluster_connection *conn) +{ + int ret; + struct ocfs2_live_connection *lc = conn->cc_private; + struct ocfs2_protocol_version pv; + + running_proto.pv_major = + ocfs2_user_plugin.sp_max_proto.pv_major; + running_proto.pv_minor = + ocfs2_user_plugin.sp_max_proto.pv_minor; + + lc->oc_version_lksb.sb_lvbptr = lc->oc_lvb; + ret = version_lock(conn, DLM_LOCK_EX, + DLM_LKF_VALBLK|DLM_LKF_NOQUEUE); + if (!ret) { + conn->cc_version.pv_major = running_proto.pv_major; + conn->cc_version.pv_minor = running_proto.pv_minor; + version_to_lvb(&running_proto, lc->oc_lvb); + version_lock(conn, DLM_LOCK_PR, DLM_LKF_CONVERT|DLM_LKF_VALBLK); + } else if (ret == -EAGAIN) { + ret = version_lock(conn, DLM_LOCK_PR, DLM_LKF_VALBLK); + if (ret) + goto out; + lvb_to_version(lc->oc_lvb, &pv); + + if ((pv.pv_major != running_proto.pv_major) || + (pv.pv_minor > running_proto.pv_minor)) { + ret = -EINVAL; + goto out; + } + + conn->cc_version.pv_major = pv.pv_major; + conn->cc_version.pv_minor = pv.pv_minor; + } +out: + return ret; +} + +static void user_recover_prep(void *arg) +{ +} + +static void user_recover_slot(void *arg, struct dlm_slot *slot) +{ + struct ocfs2_cluster_connection *conn = arg; + printk(KERN_INFO "ocfs2: Node %d/%d down. Initiating recovery.\n", + slot->nodeid, slot->slot); + conn->cc_recovery_handler(slot->nodeid, conn->cc_recovery_data); + +} + +static void user_recover_done(void *arg, struct dlm_slot *slots, + int num_slots, int our_slot, + uint32_t generation) +{ + struct ocfs2_cluster_connection *conn = arg; + struct ocfs2_live_connection *lc = conn->cc_private; + int i; + + for (i = 0; i < num_slots; i++) + if (slots[i].slot == our_slot) { + atomic_set(&lc->oc_this_node, slots[i].nodeid); + break; + } + + lc->oc_our_slot = our_slot; + wake_up(&lc->oc_wait); +} + +static const struct dlm_lockspace_ops ocfs2_ls_ops = { + .recover_prep = user_recover_prep, + .recover_slot = user_recover_slot, + .recover_done = user_recover_done, +}; + +static int user_cluster_disconnect(struct ocfs2_cluster_connection *conn) +{ + version_unlock(conn); + dlm_release_lockspace(conn->cc_lockspace, 2); + conn->cc_lockspace = NULL; + ocfs2_live_connection_drop(conn->cc_private); + conn->cc_private = NULL; + return 0; +} + static int user_cluster_connect(struct ocfs2_cluster_connection *conn) { dlm_lockspace_t *fsdlm; - struct ocfs2_live_connection *control; - int rc = 0; + struct ocfs2_live_connection *lc; + int rc, ops_rv; BUG_ON(conn == NULL); - rc = ocfs2_live_connection_new(conn, &control); + lc = kzalloc(sizeof(struct ocfs2_live_connection), GFP_KERNEL); + if (!lc) { + rc = -ENOMEM; + goto out; + } + + init_waitqueue_head(&lc->oc_wait); + init_completion(&lc->oc_sync_wait); + atomic_set(&lc->oc_this_node, 0); + conn->cc_private = lc; + lc->oc_type = NO_CONTROLD; + + rc = dlm_new_lockspace(conn->cc_name, conn->cc_cluster_name, + DLM_LSFL_FS, DLM_LVB_LEN, + &ocfs2_ls_ops, conn, &ops_rv, &fsdlm); if (rc) goto out; + if (ops_rv == -EOPNOTSUPP) { + lc->oc_type = WITH_CONTROLD; + printk(KERN_NOTICE "ocfs2: You seem to be using an older " + "version of dlm_controld and/or ocfs2-tools." + " Please consider upgrading.\n"); + } else if (ops_rv) { + rc = ops_rv; + goto out; + } + conn->cc_lockspace = fsdlm; + + rc = ocfs2_live_connection_attach(conn, lc); + if (rc) + goto out; + + if (lc->oc_type == NO_CONTROLD) { + rc = get_protocol_version(conn); + if (rc) { + printk(KERN_ERR "ocfs2: Could not determine" + " locking version\n"); + user_cluster_disconnect(conn); + goto out; + } + wait_event(lc->oc_wait, (atomic_read(&lc->oc_this_node) > 0)); + } + /* * running_proto must have been set before we allowed any mounts * to proceed. @@ -823,42 +1054,34 @@ static int user_cluster_connect(struct ocfs2_cluster_connection *conn) if (fs_protocol_compare(&running_proto, &conn->cc_version)) { printk(KERN_ERR "Unable to mount with fs locking protocol version " - "%u.%u because the userspace control daemon has " - "negotiated %u.%u\n", + "%u.%u because negotiated protocol is %u.%u\n", conn->cc_version.pv_major, conn->cc_version.pv_minor, running_proto.pv_major, running_proto.pv_minor); rc = -EPROTO; - ocfs2_live_connection_drop(control); - goto out; + ocfs2_live_connection_drop(lc); + lc = NULL; } - rc = dlm_new_lockspace(conn->cc_name, strlen(conn->cc_name), - &fsdlm, DLM_LSFL_FS, DLM_LVB_LEN); - if (rc) { - ocfs2_live_connection_drop(control); - goto out; - } - - conn->cc_private = control; - conn->cc_lockspace = fsdlm; out: + if (rc && lc) + kfree(lc); return rc; } -static int user_cluster_disconnect(struct ocfs2_cluster_connection *conn) -{ - dlm_release_lockspace(conn->cc_lockspace, 2); - conn->cc_lockspace = NULL; - ocfs2_live_connection_drop(conn->cc_private); - conn->cc_private = NULL; - return 0; -} -static int user_cluster_this_node(unsigned int *this_node) +static int user_cluster_this_node(struct ocfs2_cluster_connection *conn, + unsigned int *this_node) { int rc; + struct ocfs2_live_connection *lc = conn->cc_private; + + if (lc->oc_type == WITH_CONTROLD) + rc = ocfs2_control_get_this_node(); + else if (lc->oc_type == NO_CONTROLD) + rc = atomic_read(&lc->oc_this_node); + else + rc = -EINVAL; - rc = ocfs2_control_get_this_node(); if (rc < 0) return rc; @@ -873,6 +1096,7 @@ static struct ocfs2_stack_operations ocfs2_user_plugin_ops = { .dlm_lock = user_dlm_lock, .dlm_unlock = user_dlm_unlock, .lock_status = user_dlm_lock_status, + .lvb_valid = user_dlm_lvb_valid, .lock_lvb = user_dlm_lvb, .plock = user_plock, .dump_lksb = user_dlm_dump_lksb, diff --git a/fs/ocfs2/stackglue.c b/fs/ocfs2/stackglue.c index 68b668b0e60..5d965e83bd4 100644 --- a/fs/ocfs2/stackglue.c +++ b/fs/ocfs2/stackglue.c @@ -6,7 +6,7 @@ * Code which implements an OCFS2 specific interface to underlying * cluster stacks. * - * Copyright (C) 2007 Oracle. All rights reserved. + * Copyright (C) 2007, 2009 Oracle. All rights reserved. * * This program is free software; you can redistribute it and/or * modify it under the terms of the GNU General Public @@ -36,7 +36,7 @@ #define OCFS2_STACK_PLUGIN_USER "user" #define OCFS2_MAX_HB_CTL_PATH 256 -static struct ocfs2_locking_protocol *lproto; +static struct ocfs2_protocol_version locking_max_version; static DEFINE_SPINLOCK(ocfs2_stack_lock); static LIST_HEAD(ocfs2_stack_list); static char cluster_stack_name[OCFS2_STACK_LABEL_LEN + 1]; @@ -176,7 +176,7 @@ int ocfs2_stack_glue_register(struct ocfs2_stack_plugin *plugin) spin_lock(&ocfs2_stack_lock); if (!ocfs2_stack_lookup(plugin->sp_name)) { plugin->sp_count = 0; - plugin->sp_proto = lproto; + plugin->sp_max_proto = locking_max_version; list_add(&plugin->sp_list, &ocfs2_stack_list); printk(KERN_INFO "ocfs2: Registered cluster interface %s\n", plugin->sp_name); @@ -213,76 +213,76 @@ void ocfs2_stack_glue_unregister(struct ocfs2_stack_plugin *plugin) } EXPORT_SYMBOL_GPL(ocfs2_stack_glue_unregister); -void ocfs2_stack_glue_set_locking_protocol(struct ocfs2_locking_protocol *proto) +void ocfs2_stack_glue_set_max_proto_version(struct ocfs2_protocol_version *max_proto) { struct ocfs2_stack_plugin *p; - BUG_ON(proto == NULL); - spin_lock(&ocfs2_stack_lock); - BUG_ON(active_stack != NULL); + if (memcmp(max_proto, &locking_max_version, + sizeof(struct ocfs2_protocol_version))) { + BUG_ON(locking_max_version.pv_major != 0); - lproto = proto; - list_for_each_entry(p, &ocfs2_stack_list, sp_list) { - p->sp_proto = lproto; + locking_max_version = *max_proto; + list_for_each_entry(p, &ocfs2_stack_list, sp_list) { + p->sp_max_proto = locking_max_version; + } } - spin_unlock(&ocfs2_stack_lock); } -EXPORT_SYMBOL_GPL(ocfs2_stack_glue_set_locking_protocol); +EXPORT_SYMBOL_GPL(ocfs2_stack_glue_set_max_proto_version); /* - * The ocfs2_dlm_lock() and ocfs2_dlm_unlock() functions take - * "struct ocfs2_lock_res *astarg" instead of "void *astarg" because the - * underlying stack plugins need to pilfer the lksb off of the lock_res. - * If some other structure needs to be passed as an astarg, the plugins - * will need to be given a different avenue to the lksb. + * The ocfs2_dlm_lock() and ocfs2_dlm_unlock() functions take no argument + * for the ast and bast functions. They will pass the lksb to the ast + * and bast. The caller can wrap the lksb with their own structure to + * get more information. */ int ocfs2_dlm_lock(struct ocfs2_cluster_connection *conn, int mode, - union ocfs2_dlm_lksb *lksb, + struct ocfs2_dlm_lksb *lksb, u32 flags, void *name, - unsigned int namelen, - struct ocfs2_lock_res *astarg) + unsigned int namelen) { - BUG_ON(lproto == NULL); - + if (!lksb->lksb_conn) + lksb->lksb_conn = conn; + else + BUG_ON(lksb->lksb_conn != conn); return active_stack->sp_ops->dlm_lock(conn, mode, lksb, flags, - name, namelen, astarg); + name, namelen); } EXPORT_SYMBOL_GPL(ocfs2_dlm_lock); int ocfs2_dlm_unlock(struct ocfs2_cluster_connection *conn, - union ocfs2_dlm_lksb *lksb, - u32 flags, - struct ocfs2_lock_res *astarg) + struct ocfs2_dlm_lksb *lksb, + u32 flags) { - BUG_ON(lproto == NULL); + BUG_ON(lksb->lksb_conn == NULL); - return active_stack->sp_ops->dlm_unlock(conn, lksb, flags, astarg); + return active_stack->sp_ops->dlm_unlock(conn, lksb, flags); } EXPORT_SYMBOL_GPL(ocfs2_dlm_unlock); -int ocfs2_dlm_lock_status(union ocfs2_dlm_lksb *lksb) +int ocfs2_dlm_lock_status(struct ocfs2_dlm_lksb *lksb) { return active_stack->sp_ops->lock_status(lksb); } EXPORT_SYMBOL_GPL(ocfs2_dlm_lock_status); -/* - * Why don't we cast to ocfs2_meta_lvb? The "clean" answer is that we - * don't cast at the glue level. The real answer is that the header - * ordering is nigh impossible. - */ -void *ocfs2_dlm_lvb(union ocfs2_dlm_lksb *lksb) +int ocfs2_dlm_lvb_valid(struct ocfs2_dlm_lksb *lksb) +{ + return active_stack->sp_ops->lvb_valid(lksb); +} +EXPORT_SYMBOL_GPL(ocfs2_dlm_lvb_valid); + +void *ocfs2_dlm_lvb(struct ocfs2_dlm_lksb *lksb) { return active_stack->sp_ops->lock_lvb(lksb); } EXPORT_SYMBOL_GPL(ocfs2_dlm_lvb); -void ocfs2_dlm_dump_lksb(union ocfs2_dlm_lksb *lksb) +void ocfs2_dlm_dump_lksb(struct ocfs2_dlm_lksb *lksb) { active_stack->sp_ops->dump_lksb(lksb); } @@ -309,8 +309,11 @@ int ocfs2_plock(struct ocfs2_cluster_connection *conn, u64 ino, EXPORT_SYMBOL_GPL(ocfs2_plock); int ocfs2_cluster_connect(const char *stack_name, + const char *cluster_name, + int cluster_name_len, const char *group, int grouplen, + struct ocfs2_locking_protocol *lproto, void (*recovery_handler)(int node_num, void *recovery_data), void *recovery_data, @@ -328,6 +331,12 @@ int ocfs2_cluster_connect(const char *stack_name, goto out; } + if (memcmp(&lproto->lp_max_version, &locking_max_version, + sizeof(struct ocfs2_protocol_version))) { + rc = -EINVAL; + goto out; + } + new_conn = kzalloc(sizeof(struct ocfs2_cluster_connection), GFP_KERNEL); if (!new_conn) { @@ -335,11 +344,16 @@ int ocfs2_cluster_connect(const char *stack_name, goto out; } - memcpy(new_conn->cc_name, group, grouplen); + strlcpy(new_conn->cc_name, group, GROUP_NAME_MAX + 1); new_conn->cc_namelen = grouplen; + if (cluster_name_len) + strlcpy(new_conn->cc_cluster_name, cluster_name, + CLUSTER_NAME_MAX + 1); + new_conn->cc_cluster_name_len = cluster_name_len; new_conn->cc_recovery_handler = recovery_handler; new_conn->cc_recovery_data = recovery_data; + new_conn->cc_proto = lproto; /* Start the new connection at our maximum compatibility level */ new_conn->cc_version = lproto->lp_max_version; @@ -365,6 +379,25 @@ out: } EXPORT_SYMBOL_GPL(ocfs2_cluster_connect); +/* The caller will ensure all nodes have the same cluster stack */ +int ocfs2_cluster_connect_agnostic(const char *group, + int grouplen, + struct ocfs2_locking_protocol *lproto, + void (*recovery_handler)(int node_num, + void *recovery_data), + void *recovery_data, + struct ocfs2_cluster_connection **conn) +{ + char *stack_name = NULL; + + if (cluster_stack_name[0]) + stack_name = cluster_stack_name; + return ocfs2_cluster_connect(stack_name, NULL, 0, group, grouplen, + lproto, recovery_handler, recovery_data, + conn); +} +EXPORT_SYMBOL_GPL(ocfs2_cluster_connect_agnostic); + /* If hangup_pending is 0, the stack driver will be dropped */ int ocfs2_cluster_disconnect(struct ocfs2_cluster_connection *conn, int hangup_pending) @@ -434,9 +467,10 @@ void ocfs2_cluster_hangup(const char *group, int grouplen) } EXPORT_SYMBOL_GPL(ocfs2_cluster_hangup); -int ocfs2_cluster_this_node(unsigned int *node) +int ocfs2_cluster_this_node(struct ocfs2_cluster_connection *conn, + unsigned int *node) { - return active_stack->sp_ops->this_node(node); + return active_stack->sp_ops->this_node(conn, node); } EXPORT_SYMBOL_GPL(ocfs2_cluster_this_node); @@ -452,17 +486,17 @@ static ssize_t ocfs2_max_locking_protocol_show(struct kobject *kobj, ssize_t ret = 0; spin_lock(&ocfs2_stack_lock); - if (lproto) + if (locking_max_version.pv_major) ret = snprintf(buf, PAGE_SIZE, "%u.%u\n", - lproto->lp_max_version.pv_major, - lproto->lp_max_version.pv_minor); + locking_max_version.pv_major, + locking_max_version.pv_minor); spin_unlock(&ocfs2_stack_lock); return ret; } static struct kobj_attribute ocfs2_attr_max_locking_protocol = - __ATTR(max_locking_protocol, S_IFREG | S_IRUGO, + __ATTR(max_locking_protocol, S_IRUGO, ocfs2_max_locking_protocol_show, NULL); static ssize_t ocfs2_loaded_cluster_plugins_show(struct kobject *kobj, @@ -494,7 +528,7 @@ static ssize_t ocfs2_loaded_cluster_plugins_show(struct kobject *kobj, } static struct kobj_attribute ocfs2_attr_loaded_cluster_plugins = - __ATTR(loaded_cluster_plugins, S_IFREG | S_IRUGO, + __ATTR(loaded_cluster_plugins, S_IRUGO, ocfs2_loaded_cluster_plugins_show, NULL); static ssize_t ocfs2_active_cluster_plugin_show(struct kobject *kobj, @@ -516,7 +550,7 @@ static ssize_t ocfs2_active_cluster_plugin_show(struct kobject *kobj, } static struct kobj_attribute ocfs2_attr_active_cluster_plugin = - __ATTR(active_cluster_plugin, S_IFREG | S_IRUGO, + __ATTR(active_cluster_plugin, S_IRUGO, ocfs2_active_cluster_plugin_show, NULL); static ssize_t ocfs2_cluster_stack_show(struct kobject *kobj, @@ -565,15 +599,29 @@ static ssize_t ocfs2_cluster_stack_store(struct kobject *kobj, static struct kobj_attribute ocfs2_attr_cluster_stack = - __ATTR(cluster_stack, S_IFREG | S_IRUGO | S_IWUSR, + __ATTR(cluster_stack, S_IRUGO | S_IWUSR, ocfs2_cluster_stack_show, ocfs2_cluster_stack_store); + + +static ssize_t ocfs2_dlm_recover_show(struct kobject *kobj, + struct kobj_attribute *attr, + char *buf) +{ + return snprintf(buf, PAGE_SIZE, "1\n"); +} + +static struct kobj_attribute ocfs2_attr_dlm_recover_support = + __ATTR(dlm_recover_callback_support, S_IRUGO, + ocfs2_dlm_recover_show, NULL); + static struct attribute *ocfs2_attrs[] = { &ocfs2_attr_max_locking_protocol.attr, &ocfs2_attr_loaded_cluster_plugins.attr, &ocfs2_attr_active_cluster_plugin.attr, &ocfs2_attr_cluster_stack.attr, + &ocfs2_attr_dlm_recover_support.attr, NULL, }; @@ -617,56 +665,51 @@ error: #define FS_OCFS2_NM 1 -static ctl_table ocfs2_nm_table[] = { +static struct ctl_table ocfs2_nm_table[] = { { - .ctl_name = 1, .procname = "hb_ctl_path", .data = ocfs2_hb_ctl_path, .maxlen = OCFS2_MAX_HB_CTL_PATH, .mode = 0644, - .proc_handler = &proc_dostring, - .strategy = &sysctl_string, + .proc_handler = proc_dostring, }, - { .ctl_name = 0 } + { } }; -static ctl_table ocfs2_mod_table[] = { +static struct ctl_table ocfs2_mod_table[] = { { - .ctl_name = FS_OCFS2_NM, .procname = "nm", .data = NULL, .maxlen = 0, .mode = 0555, .child = ocfs2_nm_table }, - { .ctl_name = 0} + { } }; -static ctl_table ocfs2_kern_table[] = { +static struct ctl_table ocfs2_kern_table[] = { { - .ctl_name = FS_OCFS2, .procname = "ocfs2", .data = NULL, .maxlen = 0, .mode = 0555, .child = ocfs2_mod_table }, - { .ctl_name = 0} + { } }; -static ctl_table ocfs2_root_table[] = { +static struct ctl_table ocfs2_root_table[] = { { - .ctl_name = CTL_FS, .procname = "fs", .data = NULL, .maxlen = 0, .mode = 0555, .child = ocfs2_kern_table }, - { .ctl_name = 0 } + { } }; -static struct ctl_table_header *ocfs2_table_header = NULL; +static struct ctl_table_header *ocfs2_table_header; /* @@ -689,7 +732,10 @@ static int __init ocfs2_stack_glue_init(void) static void __exit ocfs2_stack_glue_exit(void) { - lproto = NULL; + memset(&locking_max_version, 0, + sizeof(struct ocfs2_protocol_version)); + locking_max_version.pv_major = 0; + locking_max_version.pv_minor = 0; ocfs2_sysfs_exit(); if (ocfs2_table_header) unregister_sysctl_table(ocfs2_table_header); diff --git a/fs/ocfs2/stackglue.h b/fs/ocfs2/stackglue.h index c571af375ef..66334a30cea 100644 --- a/fs/ocfs2/stackglue.h +++ b/fs/ocfs2/stackglue.h @@ -45,6 +45,9 @@ struct file_lock; */ #define GROUP_NAME_MAX 64 +/* This shadows OCFS2_CLUSTER_NAME_LEN */ +#define CLUSTER_NAME_MAX 16 + /* * ocfs2_protocol_version changes when ocfs2 does something different in @@ -56,17 +59,6 @@ struct ocfs2_protocol_version { }; /* - * The ocfs2_locking_protocol defines the handlers called on ocfs2's behalf. - */ -struct ocfs2_locking_protocol { - struct ocfs2_protocol_version lp_max_version; - void (*lp_lock_ast)(void *astarg); - void (*lp_blocking_ast)(void *astarg, int level); - void (*lp_unlock_ast)(void *astarg, int error); -}; - - -/* * The dlm_lockstatus struct includes lvb space, but the dlm_lksb struct only * has a pointer to separately allocated lvb space. This struct exists only to * include in the lksb union to make space for a combined dlm_lksb and lvb. @@ -81,21 +73,39 @@ struct fsdlm_lksb_plus_lvb { * size of the union is known. Lock status structures are embedded in * ocfs2 inodes. */ -union ocfs2_dlm_lksb { - struct dlm_lockstatus lksb_o2dlm; - struct dlm_lksb lksb_fsdlm; - struct fsdlm_lksb_plus_lvb padding; +struct ocfs2_cluster_connection; +struct ocfs2_dlm_lksb { + union { + struct dlm_lockstatus lksb_o2dlm; + struct dlm_lksb lksb_fsdlm; + struct fsdlm_lksb_plus_lvb padding; + }; + struct ocfs2_cluster_connection *lksb_conn; +}; + +/* + * The ocfs2_locking_protocol defines the handlers called on ocfs2's behalf. + */ +struct ocfs2_locking_protocol { + struct ocfs2_protocol_version lp_max_version; + void (*lp_lock_ast)(struct ocfs2_dlm_lksb *lksb); + void (*lp_blocking_ast)(struct ocfs2_dlm_lksb *lksb, int level); + void (*lp_unlock_ast)(struct ocfs2_dlm_lksb *lksb, int error); }; + /* * A cluster connection. Mostly opaque to ocfs2, the connection holds * state for the underlying stack. ocfs2 does use cc_version to determine * locking compatibility. */ struct ocfs2_cluster_connection { - char cc_name[GROUP_NAME_MAX]; + char cc_name[GROUP_NAME_MAX + 1]; int cc_namelen; + char cc_cluster_name[CLUSTER_NAME_MAX + 1]; + int cc_cluster_name_len; struct ocfs2_protocol_version cc_version; + struct ocfs2_locking_protocol *cc_proto; void (*cc_recovery_handler)(int node_num, void *recovery_data); void *cc_recovery_data; void *cc_lockspace; @@ -121,7 +131,7 @@ struct ocfs2_stack_operations { * * ->connect() must not return until it is guaranteed that * - * - Node down notifications for the filesystem will be recieved + * - Node down notifications for the filesystem will be received * and passed to conn->cc_recovery_handler(). * - Locking requests for the filesystem will be processed. */ @@ -147,7 +157,8 @@ struct ocfs2_stack_operations { * ->this_node() returns the cluster's unique identifier for the * local node. */ - int (*this_node)(unsigned int *node); + int (*this_node)(struct ocfs2_cluster_connection *conn, + unsigned int *node); /* * Call the underlying dlm lock function. The ->dlm_lock() @@ -155,27 +166,29 @@ struct ocfs2_stack_operations { * * ast and bast functions are not part of the call because the * stack will likely want to wrap ast and bast calls before passing - * them to stack->sp_proto. + * them to stack->sp_proto. There is no astarg. The lksb will + * be passed back to the ast and bast functions. The caller can + * use this to find their object. */ int (*dlm_lock)(struct ocfs2_cluster_connection *conn, int mode, - union ocfs2_dlm_lksb *lksb, + struct ocfs2_dlm_lksb *lksb, u32 flags, void *name, - unsigned int namelen, - void *astarg); + unsigned int namelen); /* * Call the underlying dlm unlock function. The ->dlm_unlock() * function should convert the flags as appropriate. * * The unlock ast is not passed, as the stack will want to wrap - * it before calling stack->sp_proto->lp_unlock_ast(). + * it before calling stack->sp_proto->lp_unlock_ast(). There is + * no astarg. The lksb will be passed back to the unlock ast + * function. The caller can use this to find their object. */ int (*dlm_unlock)(struct ocfs2_cluster_connection *conn, - union ocfs2_dlm_lksb *lksb, - u32 flags, - void *astarg); + struct ocfs2_dlm_lksb *lksb, + u32 flags); /* * Return the status of the current lock status block. The fs @@ -183,12 +196,17 @@ struct ocfs2_stack_operations { * callback pulls out the stack-specific lksb, converts the status * to a proper errno, and returns it. */ - int (*lock_status)(union ocfs2_dlm_lksb *lksb); + int (*lock_status)(struct ocfs2_dlm_lksb *lksb); + + /* + * Return non-zero if the LVB is valid. + */ + int (*lvb_valid)(struct ocfs2_dlm_lksb *lksb); /* * Pull the lvb pointer off of the stack-specific lksb. */ - void *(*lock_lvb)(union ocfs2_dlm_lksb *lksb); + void *(*lock_lvb)(struct ocfs2_dlm_lksb *lksb); /* * Cluster-aware posix locks @@ -205,7 +223,7 @@ struct ocfs2_stack_operations { * This is an optoinal debugging hook. If provided, the * stack can dump debugging information about this lock. */ - void (*dump_lksb)(union ocfs2_dlm_lksb *lksb); + void (*dump_lksb)(struct ocfs2_dlm_lksb *lksb); }; /* @@ -221,45 +239,59 @@ struct ocfs2_stack_plugin { /* These are managed by the stackglue code. */ struct list_head sp_list; unsigned int sp_count; - struct ocfs2_locking_protocol *sp_proto; + struct ocfs2_protocol_version sp_max_proto; }; /* Used by the filesystem */ int ocfs2_cluster_connect(const char *stack_name, + const char *cluster_name, + int cluster_name_len, const char *group, int grouplen, + struct ocfs2_locking_protocol *lproto, void (*recovery_handler)(int node_num, void *recovery_data), void *recovery_data, struct ocfs2_cluster_connection **conn); +/* + * Used by callers that don't store their stack name. They must ensure + * all nodes have the same stack. + */ +int ocfs2_cluster_connect_agnostic(const char *group, + int grouplen, + struct ocfs2_locking_protocol *lproto, + void (*recovery_handler)(int node_num, + void *recovery_data), + void *recovery_data, + struct ocfs2_cluster_connection **conn); int ocfs2_cluster_disconnect(struct ocfs2_cluster_connection *conn, int hangup_pending); void ocfs2_cluster_hangup(const char *group, int grouplen); -int ocfs2_cluster_this_node(unsigned int *node); +int ocfs2_cluster_this_node(struct ocfs2_cluster_connection *conn, + unsigned int *node); struct ocfs2_lock_res; int ocfs2_dlm_lock(struct ocfs2_cluster_connection *conn, int mode, - union ocfs2_dlm_lksb *lksb, + struct ocfs2_dlm_lksb *lksb, u32 flags, void *name, - unsigned int namelen, - struct ocfs2_lock_res *astarg); + unsigned int namelen); int ocfs2_dlm_unlock(struct ocfs2_cluster_connection *conn, - union ocfs2_dlm_lksb *lksb, - u32 flags, - struct ocfs2_lock_res *astarg); + struct ocfs2_dlm_lksb *lksb, + u32 flags); -int ocfs2_dlm_lock_status(union ocfs2_dlm_lksb *lksb); -void *ocfs2_dlm_lvb(union ocfs2_dlm_lksb *lksb); -void ocfs2_dlm_dump_lksb(union ocfs2_dlm_lksb *lksb); +int ocfs2_dlm_lock_status(struct ocfs2_dlm_lksb *lksb); +int ocfs2_dlm_lvb_valid(struct ocfs2_dlm_lksb *lksb); +void *ocfs2_dlm_lvb(struct ocfs2_dlm_lksb *lksb); +void ocfs2_dlm_dump_lksb(struct ocfs2_dlm_lksb *lksb); int ocfs2_stack_supports_plocks(void); int ocfs2_plock(struct ocfs2_cluster_connection *conn, u64 ino, struct file *file, int cmd, struct file_lock *fl); -void ocfs2_stack_glue_set_locking_protocol(struct ocfs2_locking_protocol *proto); +void ocfs2_stack_glue_set_max_proto_version(struct ocfs2_protocol_version *max_proto); /* Used by stack plugins */ diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c index 8439f6b324b..0cb889a17ae 100644 --- a/fs/ocfs2/suballoc.c +++ b/fs/ocfs2/suballoc.c @@ -29,7 +29,6 @@ #include <linux/slab.h> #include <linux/highmem.h> -#define MLOG_MASK_PREFIX ML_DISK_ALLOC #include <cluster/masklog.h> #include "ocfs2.h" @@ -44,6 +43,7 @@ #include "super.h" #include "sysfile.h" #include "uptodate.h" +#include "ocfs2_trace.h" #include "buffer_head_io.h" @@ -51,7 +51,33 @@ #define ALLOC_NEW_GROUP 0x1 #define ALLOC_GROUPS_FROM_GLOBAL 0x2 -#define OCFS2_MAX_INODES_TO_STEAL 1024 +#define OCFS2_MAX_TO_STEAL 1024 + +struct ocfs2_suballoc_result { + u64 sr_bg_blkno; /* The bg we allocated from. Set + to 0 when a block group is + contiguous. */ + u64 sr_bg_stable_blkno; /* + * Doesn't change, always + * set to target block + * group descriptor + * block. + */ + u64 sr_blkno; /* The first allocated block */ + unsigned int sr_bit_offset; /* The bit in the bg */ + unsigned int sr_bits; /* How many bits we claimed */ +}; + +static u64 ocfs2_group_from_res(struct ocfs2_suballoc_result *res) +{ + if (res->sr_blkno == 0) + return 0; + + if (res->sr_bg_blkno) + return res->sr_bg_blkno; + + return ocfs2_which_suballoc_group(res->sr_blkno, res->sr_bit_offset); +} static inline void ocfs2_debug_bg(struct ocfs2_group_desc *bg); static inline void ocfs2_debug_suballoc_inode(struct ocfs2_dinode *fe); @@ -60,6 +86,7 @@ static int ocfs2_block_group_fill(handle_t *handle, struct inode *alloc_inode, struct buffer_head *bg_bh, u64 group_blkno, + unsigned int group_clusters, u16 my_chain, struct ocfs2_chain_list *cl); static int ocfs2_block_group_alloc(struct ocfs2_super *osb, @@ -73,35 +100,19 @@ static int ocfs2_cluster_group_search(struct inode *inode, struct buffer_head *group_bh, u32 bits_wanted, u32 min_bits, u64 max_block, - u16 *bit_off, u16 *bits_found); + struct ocfs2_suballoc_result *res); static int ocfs2_block_group_search(struct inode *inode, struct buffer_head *group_bh, u32 bits_wanted, u32 min_bits, u64 max_block, - u16 *bit_off, u16 *bits_found); -static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb, - struct ocfs2_alloc_context *ac, + struct ocfs2_suballoc_result *res); +static int ocfs2_claim_suballoc_bits(struct ocfs2_alloc_context *ac, handle_t *handle, u32 bits_wanted, u32 min_bits, - u16 *bit_off, - unsigned int *num_bits, - u64 *bg_blkno); + struct ocfs2_suballoc_result *res); static int ocfs2_test_bg_bit_allocatable(struct buffer_head *bg_bh, int nr); -static inline int ocfs2_block_group_set_bits(handle_t *handle, - struct inode *alloc_inode, - struct ocfs2_group_desc *bg, - struct buffer_head *group_bh, - unsigned int bit_off, - unsigned int num_bits); -static inline int ocfs2_block_group_clear_bits(handle_t *handle, - struct inode *alloc_inode, - struct ocfs2_group_desc *bg, - struct buffer_head *group_bh, - unsigned int bit_off, - unsigned int num_bits); - static int ocfs2_relink_block_group(handle_t *handle, struct inode *alloc_inode, struct buffer_head *fe_bh, @@ -137,6 +148,11 @@ void ocfs2_free_ac_resource(struct ocfs2_alloc_context *ac) } brelse(ac->ac_bh); ac->ac_bh = NULL; + ac->ac_resv = NULL; + if (ac->ac_find_loc_priv) { + kfree(ac->ac_find_loc_priv); + ac->ac_find_loc_priv = NULL; + } } void ocfs2_free_alloc_context(struct ocfs2_alloc_context *ac) @@ -152,7 +168,7 @@ static u32 ocfs2_bits_per_group(struct ocfs2_chain_list *cl) #define do_error(fmt, ...) \ do{ \ - if (clean_error) \ + if (resize) \ mlog(ML_ERROR, fmt "\n", ##__VA_ARGS__); \ else \ ocfs2_error(sb, fmt, ##__VA_ARGS__); \ @@ -160,7 +176,7 @@ static u32 ocfs2_bits_per_group(struct ocfs2_chain_list *cl) static int ocfs2_validate_gd_self(struct super_block *sb, struct buffer_head *bh, - int clean_error) + int resize) { struct ocfs2_group_desc *gd = (struct ocfs2_group_desc *)bh->b_data; @@ -211,7 +227,7 @@ static int ocfs2_validate_gd_self(struct super_block *sb, static int ocfs2_validate_gd_parent(struct super_block *sb, struct ocfs2_dinode *di, struct buffer_head *bh, - int clean_error) + int resize) { unsigned int max_bits; struct ocfs2_group_desc *gd = (struct ocfs2_group_desc *)bh->b_data; @@ -233,8 +249,11 @@ static int ocfs2_validate_gd_parent(struct super_block *sb, return -EINVAL; } - if (le16_to_cpu(gd->bg_chain) >= - le16_to_cpu(di->id2.i_chain.cl_next_free_rec)) { + /* In resize, we may meet the case bg_chain == cl_next_free_rec. */ + if ((le16_to_cpu(gd->bg_chain) > + le16_to_cpu(di->id2.i_chain.cl_next_free_rec)) || + ((le16_to_cpu(gd->bg_chain) == + le16_to_cpu(di->id2.i_chain.cl_next_free_rec)) && !resize)) { do_error("Group descriptor #%llu has bad chain %u", (unsigned long long)bh->b_blocknr, le16_to_cpu(gd->bg_chain)); @@ -283,8 +302,8 @@ static int ocfs2_validate_group_descriptor(struct super_block *sb, int rc; struct ocfs2_group_desc *gd = (struct ocfs2_group_desc *)bh->b_data; - mlog(0, "Validating group descriptor %llu\n", - (unsigned long long)bh->b_blocknr); + trace_ocfs2_validate_group_descriptor( + (unsigned long long)bh->b_blocknr); BUG_ON(!buffer_uptodate(bh)); @@ -310,7 +329,7 @@ int ocfs2_read_group_descriptor(struct inode *inode, struct ocfs2_dinode *di, int rc; struct buffer_head *tmp = *bh; - rc = ocfs2_read_block(inode, gd_blkno, &tmp, + rc = ocfs2_read_block(INODE_CACHE(inode), gd_blkno, &tmp, ocfs2_validate_group_descriptor); if (rc) goto out; @@ -329,19 +348,41 @@ out: return rc; } +static void ocfs2_bg_discontig_add_extent(struct ocfs2_super *osb, + struct ocfs2_group_desc *bg, + struct ocfs2_chain_list *cl, + u64 p_blkno, unsigned int clusters) +{ + struct ocfs2_extent_list *el = &bg->bg_list; + struct ocfs2_extent_rec *rec; + + BUG_ON(!ocfs2_supports_discontig_bg(osb)); + if (!el->l_next_free_rec) + el->l_count = cpu_to_le16(ocfs2_extent_recs_per_gd(osb->sb)); + rec = &el->l_recs[le16_to_cpu(el->l_next_free_rec)]; + rec->e_blkno = cpu_to_le64(p_blkno); + rec->e_cpos = cpu_to_le32(le16_to_cpu(bg->bg_bits) / + le16_to_cpu(cl->cl_bpc)); + rec->e_leaf_clusters = cpu_to_le16(clusters); + le16_add_cpu(&bg->bg_bits, clusters * le16_to_cpu(cl->cl_bpc)); + le16_add_cpu(&bg->bg_free_bits_count, + clusters * le16_to_cpu(cl->cl_bpc)); + le16_add_cpu(&el->l_next_free_rec, 1); +} + static int ocfs2_block_group_fill(handle_t *handle, struct inode *alloc_inode, struct buffer_head *bg_bh, u64 group_blkno, + unsigned int group_clusters, u16 my_chain, struct ocfs2_chain_list *cl) { int status = 0; + struct ocfs2_super *osb = OCFS2_SB(alloc_inode->i_sb); struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) bg_bh->b_data; struct super_block * sb = alloc_inode->i_sb; - mlog_entry_void(); - if (((unsigned long long) bg_bh->b_blocknr) != group_blkno) { ocfs2_error(alloc_inode->i_sb, "group block (%llu) != " "b_blocknr (%llu)", @@ -352,7 +393,7 @@ static int ocfs2_block_group_fill(handle_t *handle, } status = ocfs2_journal_access_gd(handle, - alloc_inode, + INODE_CACHE(alloc_inode), bg_bh, OCFS2_JOURNAL_ACCESS_CREATE); if (status < 0) { @@ -363,19 +404,23 @@ static int ocfs2_block_group_fill(handle_t *handle, memset(bg, 0, sb->s_blocksize); strcpy(bg->bg_signature, OCFS2_GROUP_DESC_SIGNATURE); bg->bg_generation = cpu_to_le32(OCFS2_SB(sb)->fs_generation); - bg->bg_size = cpu_to_le16(ocfs2_group_bitmap_size(sb)); - bg->bg_bits = cpu_to_le16(ocfs2_bits_per_group(cl)); + bg->bg_size = cpu_to_le16(ocfs2_group_bitmap_size(sb, 1, + osb->s_feature_incompat)); bg->bg_chain = cpu_to_le16(my_chain); bg->bg_next_group = cl->cl_recs[my_chain].c_blkno; bg->bg_parent_dinode = cpu_to_le64(OCFS2_I(alloc_inode)->ip_blkno); bg->bg_blkno = cpu_to_le64(group_blkno); + if (group_clusters == le16_to_cpu(cl->cl_cpg)) + bg->bg_bits = cpu_to_le16(ocfs2_bits_per_group(cl)); + else + ocfs2_bg_discontig_add_extent(osb, bg, cl, group_blkno, + group_clusters); + /* set the 1st bit in the bitmap to account for the descriptor block */ ocfs2_set_bit(0, (unsigned long *)bg->bg_bitmap); bg->bg_free_bits_count = cpu_to_le16(le16_to_cpu(bg->bg_bits) - 1); - status = ocfs2_journal_dirty(handle, bg_bh); - if (status < 0) - mlog_errno(status); + ocfs2_journal_dirty(handle, bg_bh); /* There is no need to zero out or otherwise initialize the * other blocks in a group - All valid FS metadata in a block @@ -383,7 +428,8 @@ static int ocfs2_block_group_fill(handle_t *handle, * allocation time. */ bail: - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -401,6 +447,238 @@ static inline u16 ocfs2_find_smallest_chain(struct ocfs2_chain_list *cl) return best; } +static struct buffer_head * +ocfs2_block_group_alloc_contig(struct ocfs2_super *osb, handle_t *handle, + struct inode *alloc_inode, + struct ocfs2_alloc_context *ac, + struct ocfs2_chain_list *cl) +{ + int status; + u32 bit_off, num_bits; + u64 bg_blkno; + struct buffer_head *bg_bh; + unsigned int alloc_rec = ocfs2_find_smallest_chain(cl); + + status = ocfs2_claim_clusters(handle, ac, + le16_to_cpu(cl->cl_cpg), &bit_off, + &num_bits); + if (status < 0) { + if (status != -ENOSPC) + mlog_errno(status); + goto bail; + } + + /* setup the group */ + bg_blkno = ocfs2_clusters_to_blocks(osb->sb, bit_off); + trace_ocfs2_block_group_alloc_contig( + (unsigned long long)bg_blkno, alloc_rec); + + bg_bh = sb_getblk(osb->sb, bg_blkno); + if (!bg_bh) { + status = -ENOMEM; + mlog_errno(status); + goto bail; + } + ocfs2_set_new_buffer_uptodate(INODE_CACHE(alloc_inode), bg_bh); + + status = ocfs2_block_group_fill(handle, alloc_inode, bg_bh, + bg_blkno, num_bits, alloc_rec, cl); + if (status < 0) { + brelse(bg_bh); + mlog_errno(status); + } + +bail: + return status ? ERR_PTR(status) : bg_bh; +} + +static int ocfs2_block_group_claim_bits(struct ocfs2_super *osb, + handle_t *handle, + struct ocfs2_alloc_context *ac, + unsigned int min_bits, + u32 *bit_off, u32 *num_bits) +{ + int status = 0; + + while (min_bits) { + status = ocfs2_claim_clusters(handle, ac, min_bits, + bit_off, num_bits); + if (status != -ENOSPC) + break; + + min_bits >>= 1; + } + + return status; +} + +static int ocfs2_block_group_grow_discontig(handle_t *handle, + struct inode *alloc_inode, + struct buffer_head *bg_bh, + struct ocfs2_alloc_context *ac, + struct ocfs2_chain_list *cl, + unsigned int min_bits) +{ + int status; + struct ocfs2_super *osb = OCFS2_SB(alloc_inode->i_sb); + struct ocfs2_group_desc *bg = + (struct ocfs2_group_desc *)bg_bh->b_data; + unsigned int needed = le16_to_cpu(cl->cl_cpg) - + le16_to_cpu(bg->bg_bits) / le16_to_cpu(cl->cl_bpc); + u32 p_cpos, clusters; + u64 p_blkno; + struct ocfs2_extent_list *el = &bg->bg_list; + + status = ocfs2_journal_access_gd(handle, + INODE_CACHE(alloc_inode), + bg_bh, + OCFS2_JOURNAL_ACCESS_CREATE); + if (status < 0) { + mlog_errno(status); + goto bail; + } + + while ((needed > 0) && (le16_to_cpu(el->l_next_free_rec) < + le16_to_cpu(el->l_count))) { + if (min_bits > needed) + min_bits = needed; + status = ocfs2_block_group_claim_bits(osb, handle, ac, + min_bits, &p_cpos, + &clusters); + if (status < 0) { + if (status != -ENOSPC) + mlog_errno(status); + goto bail; + } + p_blkno = ocfs2_clusters_to_blocks(osb->sb, p_cpos); + ocfs2_bg_discontig_add_extent(osb, bg, cl, p_blkno, + clusters); + + min_bits = clusters; + needed = le16_to_cpu(cl->cl_cpg) - + le16_to_cpu(bg->bg_bits) / le16_to_cpu(cl->cl_bpc); + } + + if (needed > 0) { + /* + * We have used up all the extent rec but can't fill up + * the cpg. So bail out. + */ + status = -ENOSPC; + goto bail; + } + + ocfs2_journal_dirty(handle, bg_bh); + +bail: + return status; +} + +static void ocfs2_bg_alloc_cleanup(handle_t *handle, + struct ocfs2_alloc_context *cluster_ac, + struct inode *alloc_inode, + struct buffer_head *bg_bh) +{ + int i, ret; + struct ocfs2_group_desc *bg; + struct ocfs2_extent_list *el; + struct ocfs2_extent_rec *rec; + + if (!bg_bh) + return; + + bg = (struct ocfs2_group_desc *)bg_bh->b_data; + el = &bg->bg_list; + for (i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) { + rec = &el->l_recs[i]; + ret = ocfs2_free_clusters(handle, cluster_ac->ac_inode, + cluster_ac->ac_bh, + le64_to_cpu(rec->e_blkno), + le16_to_cpu(rec->e_leaf_clusters)); + if (ret) + mlog_errno(ret); + /* Try all the clusters to free */ + } + + ocfs2_remove_from_cache(INODE_CACHE(alloc_inode), bg_bh); + brelse(bg_bh); +} + +static struct buffer_head * +ocfs2_block_group_alloc_discontig(handle_t *handle, + struct inode *alloc_inode, + struct ocfs2_alloc_context *ac, + struct ocfs2_chain_list *cl) +{ + int status; + u32 bit_off, num_bits; + u64 bg_blkno; + unsigned int min_bits = le16_to_cpu(cl->cl_cpg) >> 1; + struct buffer_head *bg_bh = NULL; + unsigned int alloc_rec = ocfs2_find_smallest_chain(cl); + struct ocfs2_super *osb = OCFS2_SB(alloc_inode->i_sb); + + if (!ocfs2_supports_discontig_bg(osb)) { + status = -ENOSPC; + goto bail; + } + + status = ocfs2_extend_trans(handle, + ocfs2_calc_bg_discontig_credits(osb->sb)); + if (status) { + mlog_errno(status); + goto bail; + } + + /* + * We're going to be grabbing from multiple cluster groups. + * We don't have enough credits to relink them all, and the + * cluster groups will be staying in cache for the duration of + * this operation. + */ + ac->ac_disable_chain_relink = 1; + + /* Claim the first region */ + status = ocfs2_block_group_claim_bits(osb, handle, ac, min_bits, + &bit_off, &num_bits); + if (status < 0) { + if (status != -ENOSPC) + mlog_errno(status); + goto bail; + } + min_bits = num_bits; + + /* setup the group */ + bg_blkno = ocfs2_clusters_to_blocks(osb->sb, bit_off); + trace_ocfs2_block_group_alloc_discontig( + (unsigned long long)bg_blkno, alloc_rec); + + bg_bh = sb_getblk(osb->sb, bg_blkno); + if (!bg_bh) { + status = -ENOMEM; + mlog_errno(status); + goto bail; + } + ocfs2_set_new_buffer_uptodate(INODE_CACHE(alloc_inode), bg_bh); + + status = ocfs2_block_group_fill(handle, alloc_inode, bg_bh, + bg_blkno, num_bits, alloc_rec, cl); + if (status < 0) { + mlog_errno(status); + goto bail; + } + + status = ocfs2_block_group_grow_discontig(handle, alloc_inode, + bg_bh, ac, cl, min_bits); + if (status) + mlog_errno(status); + +bail: + if (status) + ocfs2_bg_alloc_cleanup(handle, ac, alloc_inode, bg_bh); + return status ? ERR_PTR(status) : bg_bh; +} + /* * We expect the block group allocator to already be locked. */ @@ -416,16 +694,12 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb, struct ocfs2_chain_list *cl; struct ocfs2_alloc_context *ac = NULL; handle_t *handle = NULL; - u32 bit_off, num_bits; u16 alloc_rec; - u64 bg_blkno; struct buffer_head *bg_bh = NULL; struct ocfs2_group_desc *bg; BUG_ON(ocfs2_is_cluster_bitmap(alloc_inode)); - mlog_entry_void(); - cl = &fe->id2.i_chain; status = ocfs2_reserve_clusters_with_limit(osb, le16_to_cpu(cl->cl_cpg), @@ -447,61 +721,39 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb, } if (last_alloc_group && *last_alloc_group != 0) { - mlog(0, "use old allocation group %llu for block group alloc\n", - (unsigned long long)*last_alloc_group); + trace_ocfs2_block_group_alloc( + (unsigned long long)*last_alloc_group); ac->ac_last_group = *last_alloc_group; } - status = ocfs2_claim_clusters(osb, - handle, - ac, - le16_to_cpu(cl->cl_cpg), - &bit_off, - &num_bits); - if (status < 0) { + + bg_bh = ocfs2_block_group_alloc_contig(osb, handle, alloc_inode, + ac, cl); + if (IS_ERR(bg_bh) && (PTR_ERR(bg_bh) == -ENOSPC)) + bg_bh = ocfs2_block_group_alloc_discontig(handle, + alloc_inode, + ac, cl); + if (IS_ERR(bg_bh)) { + status = PTR_ERR(bg_bh); + bg_bh = NULL; if (status != -ENOSPC) mlog_errno(status); goto bail; } - - alloc_rec = ocfs2_find_smallest_chain(cl); - - /* setup the group */ - bg_blkno = ocfs2_clusters_to_blocks(osb->sb, bit_off); - mlog(0, "new descriptor, record %u, at block %llu\n", - alloc_rec, (unsigned long long)bg_blkno); - - bg_bh = sb_getblk(osb->sb, bg_blkno); - if (!bg_bh) { - status = -EIO; - mlog_errno(status); - goto bail; - } - ocfs2_set_new_buffer_uptodate(alloc_inode, bg_bh); - - status = ocfs2_block_group_fill(handle, - alloc_inode, - bg_bh, - bg_blkno, - alloc_rec, - cl); - if (status < 0) { - mlog_errno(status); - goto bail; - } - bg = (struct ocfs2_group_desc *) bg_bh->b_data; - status = ocfs2_journal_access_di(handle, alloc_inode, + status = ocfs2_journal_access_di(handle, INODE_CACHE(alloc_inode), bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); goto bail; } + alloc_rec = le16_to_cpu(bg->bg_chain); le32_add_cpu(&cl->cl_recs[alloc_rec].c_free, le16_to_cpu(bg->bg_free_bits_count)); - le32_add_cpu(&cl->cl_recs[alloc_rec].c_total, le16_to_cpu(bg->bg_bits)); - cl->cl_recs[alloc_rec].c_blkno = cpu_to_le64(bg_blkno); + le32_add_cpu(&cl->cl_recs[alloc_rec].c_total, + le16_to_cpu(bg->bg_bits)); + cl->cl_recs[alloc_rec].c_blkno = bg->bg_blkno; if (le16_to_cpu(cl->cl_next_free_rec) < le16_to_cpu(cl->cl_count)) le16_add_cpu(&cl->cl_next_free_rec, 1); @@ -510,11 +762,7 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb, le32_add_cpu(&fe->id1.bitmap1.i_total, le16_to_cpu(bg->bg_bits)); le32_add_cpu(&fe->i_clusters, le16_to_cpu(cl->cl_cpg)); - status = ocfs2_journal_dirty(handle, bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } + ocfs2_journal_dirty(handle, bh); spin_lock(&OCFS2_I(alloc_inode)->ip_lock); OCFS2_I(alloc_inode)->ip_clusters = le32_to_cpu(fe->i_clusters); @@ -523,6 +771,7 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb, spin_unlock(&OCFS2_I(alloc_inode)->ip_lock); i_size_write(alloc_inode, le64_to_cpu(fe->i_size)); alloc_inode->i_blocks = ocfs2_inode_sector_count(alloc_inode); + ocfs2_update_inode_fsync_trans(handle, alloc_inode, 0); status = 0; @@ -539,7 +788,8 @@ bail: brelse(bg_bh); - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -557,8 +807,6 @@ static int ocfs2_reserve_suballoc_bits(struct ocfs2_super *osb, struct ocfs2_dinode *fe; u32 free_bits; - mlog_entry_void(); - alloc_inode = ocfs2_get_system_file_inode(osb, type, slot); if (!alloc_inode) { mlog_errno(-EINVAL); @@ -598,16 +846,15 @@ static int ocfs2_reserve_suballoc_bits(struct ocfs2_super *osb, if (bits_wanted > free_bits) { /* cluster bitmap never grows */ if (ocfs2_is_cluster_bitmap(alloc_inode)) { - mlog(0, "Disk Full: wanted=%u, free_bits=%u\n", - bits_wanted, free_bits); + trace_ocfs2_reserve_suballoc_bits_nospc(bits_wanted, + free_bits); status = -ENOSPC; goto bail; } if (!(flags & ALLOC_NEW_GROUP)) { - mlog(0, "Alloc File %u Full: wanted=%u, free_bits=%u, " - "and we don't alloc a new group for it.\n", - slot, bits_wanted, free_bits); + trace_ocfs2_reserve_suballoc_bits_no_new_group( + slot, bits_wanted, free_bits); status = -ENOSPC; goto bail; } @@ -633,16 +880,118 @@ static int ocfs2_reserve_suballoc_bits(struct ocfs2_super *osb, bail: brelse(bh); - mlog_exit(status); + if (status) + mlog_errno(status); + return status; +} + +static void ocfs2_init_inode_steal_slot(struct ocfs2_super *osb) +{ + spin_lock(&osb->osb_lock); + osb->s_inode_steal_slot = OCFS2_INVALID_SLOT; + spin_unlock(&osb->osb_lock); + atomic_set(&osb->s_num_inodes_stolen, 0); +} + +static void ocfs2_init_meta_steal_slot(struct ocfs2_super *osb) +{ + spin_lock(&osb->osb_lock); + osb->s_meta_steal_slot = OCFS2_INVALID_SLOT; + spin_unlock(&osb->osb_lock); + atomic_set(&osb->s_num_meta_stolen, 0); +} + +void ocfs2_init_steal_slots(struct ocfs2_super *osb) +{ + ocfs2_init_inode_steal_slot(osb); + ocfs2_init_meta_steal_slot(osb); +} + +static void __ocfs2_set_steal_slot(struct ocfs2_super *osb, int slot, int type) +{ + spin_lock(&osb->osb_lock); + if (type == INODE_ALLOC_SYSTEM_INODE) + osb->s_inode_steal_slot = slot; + else if (type == EXTENT_ALLOC_SYSTEM_INODE) + osb->s_meta_steal_slot = slot; + spin_unlock(&osb->osb_lock); +} + +static int __ocfs2_get_steal_slot(struct ocfs2_super *osb, int type) +{ + int slot = OCFS2_INVALID_SLOT; + + spin_lock(&osb->osb_lock); + if (type == INODE_ALLOC_SYSTEM_INODE) + slot = osb->s_inode_steal_slot; + else if (type == EXTENT_ALLOC_SYSTEM_INODE) + slot = osb->s_meta_steal_slot; + spin_unlock(&osb->osb_lock); + + return slot; +} + +static int ocfs2_get_inode_steal_slot(struct ocfs2_super *osb) +{ + return __ocfs2_get_steal_slot(osb, INODE_ALLOC_SYSTEM_INODE); +} + +static int ocfs2_get_meta_steal_slot(struct ocfs2_super *osb) +{ + return __ocfs2_get_steal_slot(osb, EXTENT_ALLOC_SYSTEM_INODE); +} + +static int ocfs2_steal_resource(struct ocfs2_super *osb, + struct ocfs2_alloc_context *ac, + int type) +{ + int i, status = -ENOSPC; + int slot = __ocfs2_get_steal_slot(osb, type); + + /* Start to steal resource from the first slot after ours. */ + if (slot == OCFS2_INVALID_SLOT) + slot = osb->slot_num + 1; + + for (i = 0; i < osb->max_slots; i++, slot++) { + if (slot == osb->max_slots) + slot = 0; + + if (slot == osb->slot_num) + continue; + + status = ocfs2_reserve_suballoc_bits(osb, ac, + type, + (u32)slot, NULL, + NOT_ALLOC_NEW_GROUP); + if (status >= 0) { + __ocfs2_set_steal_slot(osb, slot, type); + break; + } + + ocfs2_free_ac_resource(ac); + } + return status; } +static int ocfs2_steal_inode(struct ocfs2_super *osb, + struct ocfs2_alloc_context *ac) +{ + return ocfs2_steal_resource(osb, ac, INODE_ALLOC_SYSTEM_INODE); +} + +static int ocfs2_steal_meta(struct ocfs2_super *osb, + struct ocfs2_alloc_context *ac) +{ + return ocfs2_steal_resource(osb, ac, EXTENT_ALLOC_SYSTEM_INODE); +} + int ocfs2_reserve_new_metadata_blocks(struct ocfs2_super *osb, int blocks, struct ocfs2_alloc_context **ac) { int status; - u32 slot; + int slot = ocfs2_get_meta_steal_slot(osb); *ac = kzalloc(sizeof(struct ocfs2_alloc_context), GFP_KERNEL); if (!(*ac)) { @@ -653,12 +1002,34 @@ int ocfs2_reserve_new_metadata_blocks(struct ocfs2_super *osb, (*ac)->ac_bits_wanted = blocks; (*ac)->ac_which = OCFS2_AC_USE_META; - slot = osb->slot_num; (*ac)->ac_group_search = ocfs2_block_group_search; + if (slot != OCFS2_INVALID_SLOT && + atomic_read(&osb->s_num_meta_stolen) < OCFS2_MAX_TO_STEAL) + goto extent_steal; + + atomic_set(&osb->s_num_meta_stolen, 0); status = ocfs2_reserve_suballoc_bits(osb, (*ac), EXTENT_ALLOC_SYSTEM_INODE, - slot, NULL, ALLOC_NEW_GROUP); + (u32)osb->slot_num, NULL, + ALLOC_GROUPS_FROM_GLOBAL|ALLOC_NEW_GROUP); + + + if (status >= 0) { + status = 0; + if (slot != OCFS2_INVALID_SLOT) + ocfs2_init_meta_steal_slot(osb); + goto bail; + } else if (status < 0 && status != -ENOSPC) { + mlog_errno(status); + goto bail; + } + + ocfs2_free_ac_resource(*ac); + +extent_steal: + status = ocfs2_steal_meta(osb, *ac); + atomic_inc(&osb->s_num_meta_stolen); if (status < 0) { if (status != -ENOSPC) mlog_errno(status); @@ -672,7 +1043,8 @@ bail: *ac = NULL; } - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -685,43 +1057,11 @@ int ocfs2_reserve_new_metadata(struct ocfs2_super *osb, ac); } -static int ocfs2_steal_inode_from_other_nodes(struct ocfs2_super *osb, - struct ocfs2_alloc_context *ac) -{ - int i, status = -ENOSPC; - s16 slot = ocfs2_get_inode_steal_slot(osb); - - /* Start to steal inodes from the first slot after ours. */ - if (slot == OCFS2_INVALID_SLOT) - slot = osb->slot_num + 1; - - for (i = 0; i < osb->max_slots; i++, slot++) { - if (slot == osb->max_slots) - slot = 0; - - if (slot == osb->slot_num) - continue; - - status = ocfs2_reserve_suballoc_bits(osb, ac, - INODE_ALLOC_SYSTEM_INODE, - slot, NULL, - NOT_ALLOC_NEW_GROUP); - if (status >= 0) { - ocfs2_set_inode_steal_slot(osb, slot); - break; - } - - ocfs2_free_ac_resource(ac); - } - - return status; -} - int ocfs2_reserve_new_inode(struct ocfs2_super *osb, struct ocfs2_alloc_context **ac) { int status; - s16 slot = ocfs2_get_inode_steal_slot(osb); + int slot = ocfs2_get_inode_steal_slot(osb); u64 alloc_group; *ac = kzalloc(sizeof(struct ocfs2_alloc_context), GFP_KERNEL); @@ -754,14 +1094,14 @@ int ocfs2_reserve_new_inode(struct ocfs2_super *osb, * need to check our slots to see whether there is some space for us. */ if (slot != OCFS2_INVALID_SLOT && - atomic_read(&osb->s_num_inodes_stolen) < OCFS2_MAX_INODES_TO_STEAL) + atomic_read(&osb->s_num_inodes_stolen) < OCFS2_MAX_TO_STEAL) goto inode_steal; atomic_set(&osb->s_num_inodes_stolen, 0); alloc_group = osb->osb_inode_alloc_group; status = ocfs2_reserve_suballoc_bits(osb, *ac, INODE_ALLOC_SYSTEM_INODE, - osb->slot_num, + (u32)osb->slot_num, &alloc_group, ALLOC_NEW_GROUP | ALLOC_GROUPS_FROM_GLOBAL); @@ -771,8 +1111,8 @@ int ocfs2_reserve_new_inode(struct ocfs2_super *osb, spin_lock(&osb->osb_lock); osb->osb_inode_alloc_group = alloc_group; spin_unlock(&osb->osb_lock); - mlog(0, "after reservation, new allocation group is " - "%llu\n", (unsigned long long)alloc_group); + trace_ocfs2_reserve_new_inode_new_group( + (unsigned long long)alloc_group); /* * Some inodes must be freed by us, so try to allocate @@ -789,7 +1129,7 @@ int ocfs2_reserve_new_inode(struct ocfs2_super *osb, ocfs2_free_ac_resource(*ac); inode_steal: - status = ocfs2_steal_inode_from_other_nodes(osb, *ac); + status = ocfs2_steal_inode(osb, *ac); atomic_inc(&osb->s_num_inodes_stolen); if (status < 0) { if (status != -ENOSPC) @@ -804,7 +1144,8 @@ bail: *ac = NULL; } - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -841,8 +1182,6 @@ static int ocfs2_reserve_clusters_with_limit(struct ocfs2_super *osb, { int status; - mlog_entry_void(); - *ac = kzalloc(sizeof(struct ocfs2_alloc_context), GFP_KERNEL); if (!(*ac)) { status = -ENOMEM; @@ -859,11 +1198,7 @@ static int ocfs2_reserve_clusters_with_limit(struct ocfs2_super *osb, status = ocfs2_reserve_local_alloc_bits(osb, bits_wanted, *ac); - if (status == -EFBIG) { - /* The local alloc window is outside ac_max_block. - * use the main bitmap. */ - status = -ENOSPC; - } else if ((status < 0) && (status != -ENOSPC)) { + if ((status < 0) && (status != -ENOSPC)) { mlog_errno(status); goto bail; } @@ -885,7 +1220,8 @@ bail: *ac = NULL; } - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -923,22 +1259,30 @@ static int ocfs2_test_bg_bit_allocatable(struct buffer_head *bg_bh, int nr) { struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) bg_bh->b_data; + int ret; if (ocfs2_test_bit(nr, (unsigned long *)bg->bg_bitmap)) return 0; - if (!buffer_jbd(bg_bh) || !bh2jh(bg_bh)->b_committed_data) + + if (!buffer_jbd(bg_bh)) return 1; + jbd_lock_bh_state(bg_bh); bg = (struct ocfs2_group_desc *) bh2jh(bg_bh)->b_committed_data; - return !ocfs2_test_bit(nr, (unsigned long *)bg->bg_bitmap); + if (bg) + ret = !ocfs2_test_bit(nr, (unsigned long *)bg->bg_bitmap); + else + ret = 1; + jbd_unlock_bh_state(bg_bh); + + return ret; } static int ocfs2_block_group_find_clear_bits(struct ocfs2_super *osb, struct buffer_head *bg_bh, unsigned int bits_wanted, unsigned int total_bits, - u16 *bit_off, - u16 *bits_found) + struct ocfs2_suballoc_result *res) { void *bitmap; u16 best_offset, best_size; @@ -982,14 +1326,9 @@ static int ocfs2_block_group_find_clear_bits(struct ocfs2_super *osb, } } - /* XXX: I think the first clause is equivalent to the second - * - jlbec */ - if (found == bits_wanted) { - *bit_off = start - found; - *bits_found = found; - } else if (best_size) { - *bit_off = best_offset; - *bits_found = best_size; + if (best_size) { + res->sr_bit_offset = best_offset; + res->sr_bits = best_size; } else { status = -ENOSPC; /* No error log here -- see the comment above @@ -999,7 +1338,7 @@ static int ocfs2_block_group_find_clear_bits(struct ocfs2_super *osb, return status; } -static inline int ocfs2_block_group_set_bits(handle_t *handle, +int ocfs2_block_group_set_bits(handle_t *handle, struct inode *alloc_inode, struct ocfs2_group_desc *bg, struct buffer_head *group_bh, @@ -1010,21 +1349,18 @@ static inline int ocfs2_block_group_set_bits(handle_t *handle, void *bitmap = bg->bg_bitmap; int journal_type = OCFS2_JOURNAL_ACCESS_WRITE; - mlog_entry_void(); - /* All callers get the descriptor via * ocfs2_read_group_descriptor(). Any corruption is a code bug. */ BUG_ON(!OCFS2_IS_VALID_GROUP_DESC(bg)); BUG_ON(le16_to_cpu(bg->bg_free_bits_count) < num_bits); - mlog(0, "block_group_set_bits: off = %u, num = %u\n", bit_off, - num_bits); + trace_ocfs2_block_group_set_bits(bit_off, num_bits); if (ocfs2_is_cluster_bitmap(alloc_inode)) journal_type = OCFS2_JOURNAL_ACCESS_UNDO; status = ocfs2_journal_access_gd(handle, - alloc_inode, + INODE_CACHE(alloc_inode), group_bh, journal_type); if (status < 0) { @@ -1033,19 +1369,20 @@ static inline int ocfs2_block_group_set_bits(handle_t *handle, } le16_add_cpu(&bg->bg_free_bits_count, -num_bits); - + if (le16_to_cpu(bg->bg_free_bits_count) > le16_to_cpu(bg->bg_bits)) { + ocfs2_error(alloc_inode->i_sb, "Group descriptor # %llu has bit" + " count %u but claims %u are freed. num_bits %d", + (unsigned long long)le64_to_cpu(bg->bg_blkno), + le16_to_cpu(bg->bg_bits), + le16_to_cpu(bg->bg_free_bits_count), num_bits); + return -EROFS; + } while(num_bits--) ocfs2_set_bit(bit_off++, bitmap); - status = ocfs2_journal_dirty(handle, - group_bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } + ocfs2_journal_dirty(handle, group_bh); bail: - mlog_exit(status); return status; } @@ -1078,7 +1415,7 @@ static int ocfs2_relink_block_group(handle_t *handle, int status; /* there is a really tiny chance the journal calls could fail, * but we wouldn't want inconsistent blocks in *any* case. */ - u64 fe_ptr, bg_ptr, prev_bg_ptr; + u64 bg_ptr, prev_bg_ptr; struct ocfs2_dinode *fe = (struct ocfs2_dinode *) fe_bh->b_data; struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) bg_bh->b_data; struct ocfs2_group_desc *prev_bg = (struct ocfs2_group_desc *) prev_bg_bh->b_data; @@ -1088,70 +1425,49 @@ static int ocfs2_relink_block_group(handle_t *handle, BUG_ON(!OCFS2_IS_VALID_GROUP_DESC(bg)); BUG_ON(!OCFS2_IS_VALID_GROUP_DESC(prev_bg)); - mlog(0, "Suballoc %llu, chain %u, move group %llu to top, prev = %llu\n", - (unsigned long long)le64_to_cpu(fe->i_blkno), chain, - (unsigned long long)le64_to_cpu(bg->bg_blkno), - (unsigned long long)le64_to_cpu(prev_bg->bg_blkno)); + trace_ocfs2_relink_block_group( + (unsigned long long)le64_to_cpu(fe->i_blkno), chain, + (unsigned long long)le64_to_cpu(bg->bg_blkno), + (unsigned long long)le64_to_cpu(prev_bg->bg_blkno)); - fe_ptr = le64_to_cpu(fe->id2.i_chain.cl_recs[chain].c_blkno); bg_ptr = le64_to_cpu(bg->bg_next_group); prev_bg_ptr = le64_to_cpu(prev_bg->bg_next_group); - status = ocfs2_journal_access_gd(handle, alloc_inode, prev_bg_bh, + status = ocfs2_journal_access_gd(handle, INODE_CACHE(alloc_inode), + prev_bg_bh, OCFS2_JOURNAL_ACCESS_WRITE); - if (status < 0) { - mlog_errno(status); - goto out_rollback; - } + if (status < 0) + goto out; prev_bg->bg_next_group = bg->bg_next_group; + ocfs2_journal_dirty(handle, prev_bg_bh); - status = ocfs2_journal_dirty(handle, prev_bg_bh); - if (status < 0) { - mlog_errno(status); - goto out_rollback; - } - - status = ocfs2_journal_access_gd(handle, alloc_inode, bg_bh, - OCFS2_JOURNAL_ACCESS_WRITE); - if (status < 0) { - mlog_errno(status); - goto out_rollback; - } + status = ocfs2_journal_access_gd(handle, INODE_CACHE(alloc_inode), + bg_bh, OCFS2_JOURNAL_ACCESS_WRITE); + if (status < 0) + goto out_rollback_prev_bg; bg->bg_next_group = fe->id2.i_chain.cl_recs[chain].c_blkno; + ocfs2_journal_dirty(handle, bg_bh); - status = ocfs2_journal_dirty(handle, bg_bh); - if (status < 0) { - mlog_errno(status); - goto out_rollback; - } - - status = ocfs2_journal_access_di(handle, alloc_inode, fe_bh, - OCFS2_JOURNAL_ACCESS_WRITE); - if (status < 0) { - mlog_errno(status); - goto out_rollback; - } + status = ocfs2_journal_access_di(handle, INODE_CACHE(alloc_inode), + fe_bh, OCFS2_JOURNAL_ACCESS_WRITE); + if (status < 0) + goto out_rollback_bg; fe->id2.i_chain.cl_recs[chain].c_blkno = bg->bg_blkno; + ocfs2_journal_dirty(handle, fe_bh); - status = ocfs2_journal_dirty(handle, fe_bh); - if (status < 0) { +out: + if (status < 0) mlog_errno(status); - goto out_rollback; - } - - status = 0; -out_rollback: - if (status < 0) { - fe->id2.i_chain.cl_recs[chain].c_blkno = cpu_to_le64(fe_ptr); - bg->bg_next_group = cpu_to_le64(bg_ptr); - prev_bg->bg_next_group = cpu_to_le64(prev_bg_ptr); - } - - mlog_exit(status); return status; + +out_rollback_bg: + bg->bg_next_group = cpu_to_le64(bg_ptr); +out_rollback_prev_bg: + prev_bg->bg_next_group = cpu_to_le64(prev_bg_ptr); + goto out; } static inline int ocfs2_block_group_reasonably_empty(struct ocfs2_group_desc *bg, @@ -1166,14 +1482,13 @@ static int ocfs2_cluster_group_search(struct inode *inode, struct buffer_head *group_bh, u32 bits_wanted, u32 min_bits, u64 max_block, - u16 *bit_off, u16 *bits_found) + struct ocfs2_suballoc_result *res) { int search = -ENOSPC; int ret; u64 blkoff; struct ocfs2_group_desc *gd = (struct ocfs2_group_desc *) group_bh->b_data; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - u16 tmp_off, tmp_found; unsigned int max_bits, gd_cluster_off; BUG_ON(!ocfs2_is_cluster_bitmap(inode)); @@ -1182,7 +1497,7 @@ static int ocfs2_cluster_group_search(struct inode *inode, max_bits = le16_to_cpu(gd->bg_bits); /* Tail groups in cluster bitmaps which aren't cpg - * aligned are prone to partial extention by a failed + * aligned are prone to partial extension by a failed * fs resize. If the file system resize never got to * update the dinode cluster count, then we don't want * to trust any clusters past it, regardless of what @@ -1192,26 +1507,26 @@ static int ocfs2_cluster_group_search(struct inode *inode, if ((gd_cluster_off + max_bits) > OCFS2_I(inode)->ip_clusters) { max_bits = OCFS2_I(inode)->ip_clusters - gd_cluster_off; - mlog(0, "Desc %llu, bg_bits %u, clusters %u, use %u\n", - (unsigned long long)le64_to_cpu(gd->bg_blkno), - le16_to_cpu(gd->bg_bits), - OCFS2_I(inode)->ip_clusters, max_bits); + trace_ocfs2_cluster_group_search_wrong_max_bits( + (unsigned long long)le64_to_cpu(gd->bg_blkno), + le16_to_cpu(gd->bg_bits), + OCFS2_I(inode)->ip_clusters, max_bits); } ret = ocfs2_block_group_find_clear_bits(OCFS2_SB(inode->i_sb), group_bh, bits_wanted, - max_bits, - &tmp_off, &tmp_found); + max_bits, res); if (ret) return ret; if (max_block) { blkoff = ocfs2_clusters_to_blocks(inode->i_sb, gd_cluster_off + - tmp_off + tmp_found); - mlog(0, "Checking %llu against %llu\n", - (unsigned long long)blkoff, - (unsigned long long)max_block); + res->sr_bit_offset + + res->sr_bits); + trace_ocfs2_cluster_group_search_max_block( + (unsigned long long)blkoff, + (unsigned long long)max_block); if (blkoff > max_block) return -ENOSPC; } @@ -1220,16 +1535,14 @@ static int ocfs2_cluster_group_search(struct inode *inode, * return success, but we still want to return * -ENOSPC unless it found the minimum number * of bits. */ - if (min_bits <= tmp_found) { - *bit_off = tmp_off; - *bits_found = tmp_found; + if (min_bits <= res->sr_bits) search = 0; /* success */ - } else if (tmp_found) { + else if (res->sr_bits) { /* * Don't show bits which we'll be returning * for allocation to the local alloc bitmap. */ - ocfs2_local_alloc_seen_free_bits(osb, tmp_found); + ocfs2_local_alloc_seen_free_bits(osb, res->sr_bits); } } @@ -1240,7 +1553,7 @@ static int ocfs2_block_group_search(struct inode *inode, struct buffer_head *group_bh, u32 bits_wanted, u32 min_bits, u64 max_block, - u16 *bit_off, u16 *bits_found) + struct ocfs2_suballoc_result *res) { int ret = -ENOSPC; u64 blkoff; @@ -1253,13 +1566,13 @@ static int ocfs2_block_group_search(struct inode *inode, ret = ocfs2_block_group_find_clear_bits(OCFS2_SB(inode->i_sb), group_bh, bits_wanted, le16_to_cpu(bg->bg_bits), - bit_off, bits_found); + res); if (!ret && max_block) { - blkoff = le64_to_cpu(bg->bg_blkno) + *bit_off + - *bits_found; - mlog(0, "Checking %llu against %llu\n", - (unsigned long long)blkoff, - (unsigned long long)max_block); + blkoff = le64_to_cpu(bg->bg_blkno) + + res->sr_bit_offset + res->sr_bits; + trace_ocfs2_block_group_search_max_block( + (unsigned long long)blkoff, + (unsigned long long)max_block); if (blkoff > max_block) ret = -ENOSPC; } @@ -1268,7 +1581,7 @@ static int ocfs2_block_group_search(struct inode *inode, return ret; } -static int ocfs2_alloc_dinode_update_counts(struct inode *inode, +int ocfs2_alloc_dinode_update_counts(struct inode *inode, handle_t *handle, struct buffer_head *di_bh, u32 num_bits, @@ -1279,7 +1592,7 @@ static int ocfs2_alloc_dinode_update_counts(struct inode *inode, struct ocfs2_dinode *di = (struct ocfs2_dinode *) di_bh->b_data; struct ocfs2_chain_list *cl = (struct ocfs2_chain_list *) &di->id2.i_chain; - ret = ocfs2_journal_access_di(handle, inode, di_bh, + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret < 0) { mlog_errno(ret); @@ -1289,33 +1602,91 @@ static int ocfs2_alloc_dinode_update_counts(struct inode *inode, tmp_used = le32_to_cpu(di->id1.bitmap1.i_used); di->id1.bitmap1.i_used = cpu_to_le32(num_bits + tmp_used); le32_add_cpu(&cl->cl_recs[chain].c_free, -num_bits); - - ret = ocfs2_journal_dirty(handle, di_bh); - if (ret < 0) - mlog_errno(ret); + ocfs2_journal_dirty(handle, di_bh); out: return ret; } +void ocfs2_rollback_alloc_dinode_counts(struct inode *inode, + struct buffer_head *di_bh, + u32 num_bits, + u16 chain) +{ + u32 tmp_used; + struct ocfs2_dinode *di = (struct ocfs2_dinode *) di_bh->b_data; + struct ocfs2_chain_list *cl; + + cl = (struct ocfs2_chain_list *)&di->id2.i_chain; + tmp_used = le32_to_cpu(di->id1.bitmap1.i_used); + di->id1.bitmap1.i_used = cpu_to_le32(tmp_used - num_bits); + le32_add_cpu(&cl->cl_recs[chain].c_free, num_bits); +} + +static int ocfs2_bg_discontig_fix_by_rec(struct ocfs2_suballoc_result *res, + struct ocfs2_extent_rec *rec, + struct ocfs2_chain_list *cl) +{ + unsigned int bpc = le16_to_cpu(cl->cl_bpc); + unsigned int bitoff = le32_to_cpu(rec->e_cpos) * bpc; + unsigned int bitcount = le16_to_cpu(rec->e_leaf_clusters) * bpc; + + if (res->sr_bit_offset < bitoff) + return 0; + if (res->sr_bit_offset >= (bitoff + bitcount)) + return 0; + res->sr_blkno = le64_to_cpu(rec->e_blkno) + + (res->sr_bit_offset - bitoff); + if ((res->sr_bit_offset + res->sr_bits) > (bitoff + bitcount)) + res->sr_bits = (bitoff + bitcount) - res->sr_bit_offset; + return 1; +} + +static void ocfs2_bg_discontig_fix_result(struct ocfs2_alloc_context *ac, + struct ocfs2_group_desc *bg, + struct ocfs2_suballoc_result *res) +{ + int i; + u64 bg_blkno = res->sr_bg_blkno; /* Save off */ + struct ocfs2_extent_rec *rec; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)ac->ac_bh->b_data; + struct ocfs2_chain_list *cl = &di->id2.i_chain; + + if (ocfs2_is_cluster_bitmap(ac->ac_inode)) { + res->sr_blkno = 0; + return; + } + + res->sr_blkno = res->sr_bg_blkno + res->sr_bit_offset; + res->sr_bg_blkno = 0; /* Clear it for contig block groups */ + if (!ocfs2_supports_discontig_bg(OCFS2_SB(ac->ac_inode->i_sb)) || + !bg->bg_list.l_next_free_rec) + return; + + for (i = 0; i < le16_to_cpu(bg->bg_list.l_next_free_rec); i++) { + rec = &bg->bg_list.l_recs[i]; + if (ocfs2_bg_discontig_fix_by_rec(res, rec, cl)) { + res->sr_bg_blkno = bg_blkno; /* Restore */ + break; + } + } +} + static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac, handle_t *handle, u32 bits_wanted, u32 min_bits, - u16 *bit_off, - unsigned int *num_bits, - u64 gd_blkno, + struct ocfs2_suballoc_result *res, u16 *bits_left) { int ret; - u16 found; struct buffer_head *group_bh = NULL; struct ocfs2_group_desc *gd; struct ocfs2_dinode *di = (struct ocfs2_dinode *)ac->ac_bh->b_data; struct inode *alloc_inode = ac->ac_inode; - ret = ocfs2_read_group_descriptor(alloc_inode, di, gd_blkno, - &group_bh); + ret = ocfs2_read_group_descriptor(alloc_inode, di, + res->sr_bg_blkno, &group_bh); if (ret < 0) { mlog_errno(ret); return ret; @@ -1323,17 +1694,27 @@ static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac, gd = (struct ocfs2_group_desc *) group_bh->b_data; ret = ac->ac_group_search(alloc_inode, group_bh, bits_wanted, min_bits, - ac->ac_max_block, bit_off, &found); + ac->ac_max_block, res); if (ret < 0) { if (ret != -ENOSPC) mlog_errno(ret); goto out; } - *num_bits = found; + if (!ret) + ocfs2_bg_discontig_fix_result(ac, gd, res); + + /* + * sr_bg_blkno might have been changed by + * ocfs2_bg_discontig_fix_result + */ + res->sr_bg_stable_blkno = group_bh->b_blocknr; + + if (ac->ac_find_loc_only) + goto out_loc_only; ret = ocfs2_alloc_dinode_update_counts(alloc_inode, handle, ac->ac_bh, - *num_bits, + res->sr_bits, le16_to_cpu(gd->bg_chain)); if (ret < 0) { mlog_errno(ret); @@ -1341,10 +1722,15 @@ static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac, } ret = ocfs2_block_group_set_bits(handle, alloc_inode, gd, group_bh, - *bit_off, *num_bits); - if (ret < 0) + res->sr_bit_offset, res->sr_bits); + if (ret < 0) { + ocfs2_rollback_alloc_dinode_counts(alloc_inode, ac->ac_bh, + res->sr_bits, + le16_to_cpu(gd->bg_chain)); mlog_errno(ret); + } +out_loc_only: *bits_left = le16_to_cpu(gd->bg_free_bits_count); out: @@ -1357,14 +1743,11 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac, handle_t *handle, u32 bits_wanted, u32 min_bits, - u16 *bit_off, - unsigned int *num_bits, - u64 *bg_blkno, + struct ocfs2_suballoc_result *res, u16 *bits_left) { int status; - u16 chain, tmp_bits; - u32 tmp_used; + u16 chain; u64 next_group; struct inode *alloc_inode = ac->ac_inode; struct buffer_head *group_bh = NULL; @@ -1374,9 +1757,9 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac, struct ocfs2_group_desc *bg; chain = ac->ac_chain; - mlog(0, "trying to alloc %u bits from chain %u, inode %llu\n", - bits_wanted, chain, - (unsigned long long)OCFS2_I(alloc_inode)->ip_blkno); + trace_ocfs2_search_chain_begin( + (unsigned long long)OCFS2_I(alloc_inode)->ip_blkno, + bits_wanted, chain); status = ocfs2_read_group_descriptor(alloc_inode, fe, le64_to_cpu(cl->cl_recs[chain].c_blkno), @@ -1392,8 +1775,8 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac, * the 1st group with any empty bits. */ while ((status = ac->ac_group_search(alloc_inode, group_bh, bits_wanted, min_bits, - ac->ac_max_block, bit_off, - &tmp_bits)) == -ENOSPC) { + ac->ac_max_block, + res)) == -ENOSPC) { if (!bg->bg_next_group) break; @@ -1417,12 +1800,20 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac, goto bail; } - mlog(0, "alloc succeeds: we give %u bits from block group %llu\n", - tmp_bits, (unsigned long long)le64_to_cpu(bg->bg_blkno)); + trace_ocfs2_search_chain_succ( + (unsigned long long)le64_to_cpu(bg->bg_blkno), res->sr_bits); - *num_bits = tmp_bits; + res->sr_bg_blkno = le64_to_cpu(bg->bg_blkno); - BUG_ON(*num_bits == 0); + BUG_ON(res->sr_bits == 0); + if (!status) + ocfs2_bg_discontig_fix_result(ac, bg, res); + + /* + * sr_bg_blkno might have been changed by + * ocfs2_bg_discontig_fix_result + */ + res->sr_bg_stable_blkno = group_bh->b_blocknr; /* * Keep track of previous block descriptor read. When @@ -1437,9 +1828,9 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac, * Do this *after* figuring out how many bits we're taking out * of our target group. */ - if (ac->ac_allow_chain_relink && + if (!ac->ac_disable_chain_relink && (prev_group_bh) && - (ocfs2_block_group_reasonably_empty(bg, *num_bits))) { + (ocfs2_block_group_reasonably_empty(bg, res->sr_bits))) { status = ocfs2_relink_block_group(handle, alloc_inode, ac->ac_bh, group_bh, prev_group_bh, chain); @@ -1449,24 +1840,13 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac, } } - /* Ok, claim our bits now: set the info on dinode, chainlist - * and then the group */ - status = ocfs2_journal_access_di(handle, - alloc_inode, - ac->ac_bh, - OCFS2_JOURNAL_ACCESS_WRITE); - if (status < 0) { - mlog_errno(status); - goto bail; - } - - tmp_used = le32_to_cpu(fe->id1.bitmap1.i_used); - fe->id1.bitmap1.i_used = cpu_to_le32(*num_bits + tmp_used); - le32_add_cpu(&cl->cl_recs[chain].c_free, -(*num_bits)); + if (ac->ac_find_loc_only) + goto out_loc_only; - status = ocfs2_journal_dirty(handle, - ac->ac_bh); - if (status < 0) { + status = ocfs2_alloc_dinode_update_counts(alloc_inode, handle, + ac->ac_bh, res->sr_bits, + chain); + if (status) { mlog_errno(status); goto bail; } @@ -1475,45 +1855,44 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac, alloc_inode, bg, group_bh, - *bit_off, - *num_bits); + res->sr_bit_offset, + res->sr_bits); if (status < 0) { + ocfs2_rollback_alloc_dinode_counts(alloc_inode, + ac->ac_bh, res->sr_bits, chain); mlog_errno(status); goto bail; } - mlog(0, "Allocated %u bits from suballocator %llu\n", *num_bits, - (unsigned long long)le64_to_cpu(fe->i_blkno)); + trace_ocfs2_search_chain_end( + (unsigned long long)le64_to_cpu(fe->i_blkno), + res->sr_bits); - *bg_blkno = le64_to_cpu(bg->bg_blkno); +out_loc_only: *bits_left = le16_to_cpu(bg->bg_free_bits_count); bail: brelse(group_bh); brelse(prev_group_bh); - mlog_exit(status); + if (status) + mlog_errno(status); return status; } /* will give out up to bits_wanted contiguous bits. */ -static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb, - struct ocfs2_alloc_context *ac, +static int ocfs2_claim_suballoc_bits(struct ocfs2_alloc_context *ac, handle_t *handle, u32 bits_wanted, u32 min_bits, - u16 *bit_off, - unsigned int *num_bits, - u64 *bg_blkno) + struct ocfs2_suballoc_result *res) { int status; u16 victim, i; u16 bits_left = 0; - u64 hint_blkno = ac->ac_last_group; + u64 hint = ac->ac_last_group; struct ocfs2_chain_list *cl; struct ocfs2_dinode *fe; - mlog_entry_void(); - BUG_ON(ac->ac_bits_given >= ac->ac_bits_wanted); BUG_ON(bits_wanted > (ac->ac_bits_wanted - ac->ac_bits_given)); BUG_ON(!ac->ac_bh); @@ -1526,7 +1905,8 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb, if (le32_to_cpu(fe->id1.bitmap1.i_used) >= le32_to_cpu(fe->id1.bitmap1.i_total)) { - ocfs2_error(osb->sb, "Chain allocator dinode %llu has %u used " + ocfs2_error(ac->ac_inode->i_sb, + "Chain allocator dinode %llu has %u used " "bits but only %u total.", (unsigned long long)le64_to_cpu(fe->i_blkno), le32_to_cpu(fe->id1.bitmap1.i_used), @@ -1535,22 +1915,16 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb, goto bail; } - if (hint_blkno) { + res->sr_bg_blkno = hint; + if (res->sr_bg_blkno) { /* Attempt to short-circuit the usual search mechanism * by jumping straight to the most recently used - * allocation group. This helps us mantain some + * allocation group. This helps us maintain some * contiguousness across allocations. */ status = ocfs2_search_one_group(ac, handle, bits_wanted, - min_bits, bit_off, num_bits, - hint_blkno, &bits_left); - if (!status) { - /* Be careful to update *bg_blkno here as the - * caller is expecting it to be filled in, and - * ocfs2_search_one_group() won't do that for - * us. */ - *bg_blkno = hint_blkno; + min_bits, res, &bits_left); + if (!status) goto set_hint; - } if (status < 0 && status != -ENOSPC) { mlog_errno(status); goto bail; @@ -1561,25 +1935,25 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb, victim = ocfs2_find_victim_chain(cl); ac->ac_chain = victim; - ac->ac_allow_chain_relink = 1; - status = ocfs2_search_chain(ac, handle, bits_wanted, min_bits, bit_off, - num_bits, bg_blkno, &bits_left); - if (!status) + status = ocfs2_search_chain(ac, handle, bits_wanted, min_bits, + res, &bits_left); + if (!status) { + hint = ocfs2_group_from_res(res); goto set_hint; + } if (status < 0 && status != -ENOSPC) { mlog_errno(status); goto bail; } - mlog(0, "Search of victim chain %u came up with nothing, " - "trying all chains now.\n", victim); + trace_ocfs2_claim_suballoc_bits(victim); /* If we didn't pick a good victim, then just default to * searching each chain in order. Don't allow chain relinking * because we only calculate enough journal credits for one * relink per alloc. */ - ac->ac_allow_chain_relink = 0; + ac->ac_disable_chain_relink = 1; for (i = 0; i < le16_to_cpu(cl->cl_next_free_rec); i ++) { if (i == victim) continue; @@ -1588,10 +1962,11 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb, ac->ac_chain = i; status = ocfs2_search_chain(ac, handle, bits_wanted, min_bits, - bit_off, num_bits, bg_blkno, - &bits_left); - if (!status) + res, &bits_left); + if (!status) { + hint = ocfs2_group_from_res(res); break; + } if (status < 0 && status != -ENOSPC) { mlog_errno(status); goto bail; @@ -1606,56 +1981,58 @@ set_hint: if (bits_left < min_bits) ac->ac_last_group = 0; else - ac->ac_last_group = *bg_blkno; + ac->ac_last_group = hint; } bail: - mlog_exit(status); + if (status) + mlog_errno(status); return status; } -int ocfs2_claim_metadata(struct ocfs2_super *osb, - handle_t *handle, +int ocfs2_claim_metadata(handle_t *handle, struct ocfs2_alloc_context *ac, u32 bits_wanted, + u64 *suballoc_loc, u16 *suballoc_bit_start, unsigned int *num_bits, u64 *blkno_start) { int status; - u64 bg_blkno; + struct ocfs2_suballoc_result res = { .sr_blkno = 0, }; BUG_ON(!ac); BUG_ON(ac->ac_bits_wanted < (ac->ac_bits_given + bits_wanted)); BUG_ON(ac->ac_which != OCFS2_AC_USE_META); - status = ocfs2_claim_suballoc_bits(osb, - ac, + status = ocfs2_claim_suballoc_bits(ac, handle, bits_wanted, 1, - suballoc_bit_start, - num_bits, - &bg_blkno); + &res); if (status < 0) { mlog_errno(status); goto bail; } - atomic_inc(&osb->alloc_stats.bg_allocs); + atomic_inc(&OCFS2_SB(ac->ac_inode->i_sb)->alloc_stats.bg_allocs); - *blkno_start = bg_blkno + (u64) *suballoc_bit_start; - ac->ac_bits_given += (*num_bits); + *suballoc_loc = res.sr_bg_blkno; + *suballoc_bit_start = res.sr_bit_offset; + *blkno_start = res.sr_blkno; + ac->ac_bits_given += res.sr_bits; + *num_bits = res.sr_bits; status = 0; bail: - mlog_exit(status); + if (status) + mlog_errno(status); return status; } static void ocfs2_init_inode_ac_group(struct inode *dir, - struct buffer_head *parent_fe_bh, + struct buffer_head *parent_di_bh, struct ocfs2_alloc_context *ac) { - struct ocfs2_dinode *fe = (struct ocfs2_dinode *)parent_fe_bh->b_data; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)parent_di_bh->b_data; /* * Try to allocate inodes from some specific group. * @@ -1669,10 +2046,14 @@ static void ocfs2_init_inode_ac_group(struct inode *dir, if (OCFS2_I(dir)->ip_last_used_group && OCFS2_I(dir)->ip_last_used_slot == ac->ac_alloc_slot) ac->ac_last_group = OCFS2_I(dir)->ip_last_used_group; - else if (le16_to_cpu(fe->i_suballoc_slot) == ac->ac_alloc_slot) - ac->ac_last_group = ocfs2_which_suballoc_group( - le64_to_cpu(fe->i_blkno), - le16_to_cpu(fe->i_suballoc_bit)); + else if (le16_to_cpu(di->i_suballoc_slot) == ac->ac_alloc_slot) { + if (di->i_suballoc_loc) + ac->ac_last_group = le64_to_cpu(di->i_suballoc_loc); + else + ac->ac_last_group = ocfs2_which_suballoc_group( + le64_to_cpu(di->i_blkno), + le16_to_cpu(di->i_suballoc_bit)); + } } static inline void ocfs2_save_inode_ac_group(struct inode *dir, @@ -1682,19 +2063,148 @@ static inline void ocfs2_save_inode_ac_group(struct inode *dir, OCFS2_I(dir)->ip_last_used_slot = ac->ac_alloc_slot; } -int ocfs2_claim_new_inode(struct ocfs2_super *osb, - handle_t *handle, +int ocfs2_find_new_inode_loc(struct inode *dir, + struct buffer_head *parent_fe_bh, + struct ocfs2_alloc_context *ac, + u64 *fe_blkno) +{ + int ret; + handle_t *handle = NULL; + struct ocfs2_suballoc_result *res; + + BUG_ON(!ac); + BUG_ON(ac->ac_bits_given != 0); + BUG_ON(ac->ac_bits_wanted != 1); + BUG_ON(ac->ac_which != OCFS2_AC_USE_INODE); + + res = kzalloc(sizeof(*res), GFP_NOFS); + if (res == NULL) { + ret = -ENOMEM; + mlog_errno(ret); + goto out; + } + + ocfs2_init_inode_ac_group(dir, parent_fe_bh, ac); + + /* + * The handle started here is for chain relink. Alternatively, + * we could just disable relink for these calls. + */ + handle = ocfs2_start_trans(OCFS2_SB(dir->i_sb), OCFS2_SUBALLOC_ALLOC); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + handle = NULL; + mlog_errno(ret); + goto out; + } + + /* + * This will instruct ocfs2_claim_suballoc_bits and + * ocfs2_search_one_group to search but save actual allocation + * for later. + */ + ac->ac_find_loc_only = 1; + + ret = ocfs2_claim_suballoc_bits(ac, handle, 1, 1, res); + if (ret < 0) { + mlog_errno(ret); + goto out; + } + + ac->ac_find_loc_priv = res; + *fe_blkno = res->sr_blkno; + ocfs2_update_inode_fsync_trans(handle, dir, 0); +out: + if (handle) + ocfs2_commit_trans(OCFS2_SB(dir->i_sb), handle); + + if (ret) + kfree(res); + + return ret; +} + +int ocfs2_claim_new_inode_at_loc(handle_t *handle, + struct inode *dir, + struct ocfs2_alloc_context *ac, + u64 *suballoc_loc, + u16 *suballoc_bit, + u64 di_blkno) +{ + int ret; + u16 chain; + struct ocfs2_suballoc_result *res = ac->ac_find_loc_priv; + struct buffer_head *bg_bh = NULL; + struct ocfs2_group_desc *bg; + struct ocfs2_dinode *di = (struct ocfs2_dinode *) ac->ac_bh->b_data; + + /* + * Since di_blkno is being passed back in, we check for any + * inconsistencies which may have happened between + * calls. These are code bugs as di_blkno is not expected to + * change once returned from ocfs2_find_new_inode_loc() + */ + BUG_ON(res->sr_blkno != di_blkno); + + ret = ocfs2_read_group_descriptor(ac->ac_inode, di, + res->sr_bg_stable_blkno, &bg_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + bg = (struct ocfs2_group_desc *) bg_bh->b_data; + chain = le16_to_cpu(bg->bg_chain); + + ret = ocfs2_alloc_dinode_update_counts(ac->ac_inode, handle, + ac->ac_bh, res->sr_bits, + chain); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_block_group_set_bits(handle, + ac->ac_inode, + bg, + bg_bh, + res->sr_bit_offset, + res->sr_bits); + if (ret < 0) { + ocfs2_rollback_alloc_dinode_counts(ac->ac_inode, + ac->ac_bh, res->sr_bits, chain); + mlog_errno(ret); + goto out; + } + + trace_ocfs2_claim_new_inode_at_loc((unsigned long long)di_blkno, + res->sr_bits); + + atomic_inc(&OCFS2_SB(ac->ac_inode->i_sb)->alloc_stats.bg_allocs); + + BUG_ON(res->sr_bits != 1); + + *suballoc_loc = res->sr_bg_blkno; + *suballoc_bit = res->sr_bit_offset; + ac->ac_bits_given++; + ocfs2_save_inode_ac_group(dir, ac); + +out: + brelse(bg_bh); + + return ret; +} + +int ocfs2_claim_new_inode(handle_t *handle, struct inode *dir, struct buffer_head *parent_fe_bh, struct ocfs2_alloc_context *ac, + u64 *suballoc_loc, u16 *suballoc_bit, u64 *fe_blkno) { int status; - unsigned int num_bits; - u64 bg_blkno; - - mlog_entry_void(); + struct ocfs2_suballoc_result res; BUG_ON(!ac); BUG_ON(ac->ac_bits_given != 0); @@ -1703,28 +2213,28 @@ int ocfs2_claim_new_inode(struct ocfs2_super *osb, ocfs2_init_inode_ac_group(dir, parent_fe_bh, ac); - status = ocfs2_claim_suballoc_bits(osb, - ac, + status = ocfs2_claim_suballoc_bits(ac, handle, 1, 1, - suballoc_bit, - &num_bits, - &bg_blkno); + &res); if (status < 0) { mlog_errno(status); goto bail; } - atomic_inc(&osb->alloc_stats.bg_allocs); + atomic_inc(&OCFS2_SB(ac->ac_inode->i_sb)->alloc_stats.bg_allocs); - BUG_ON(num_bits != 1); + BUG_ON(res.sr_bits != 1); - *fe_blkno = bg_blkno + (u64) (*suballoc_bit); + *suballoc_loc = res.sr_bg_blkno; + *suballoc_bit = res.sr_bit_offset; + *fe_blkno = res.sr_blkno; ac->ac_bits_given++; ocfs2_save_inode_ac_group(dir, ac); status = 0; bail: - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -1789,8 +2299,7 @@ static inline void ocfs2_block_to_cluster_group(struct inode *inode, * contig. allocation, set to '1' to indicate we can deal with extents * of any size. */ -int __ocfs2_claim_clusters(struct ocfs2_super *osb, - handle_t *handle, +int __ocfs2_claim_clusters(handle_t *handle, struct ocfs2_alloc_context *ac, u32 min_clusters, u32 max_clusters, @@ -1799,10 +2308,8 @@ int __ocfs2_claim_clusters(struct ocfs2_super *osb, { int status; unsigned int bits_wanted = max_clusters; - u64 bg_blkno = 0; - u16 bg_bit_off; - - mlog_entry_void(); + struct ocfs2_suballoc_result res = { .sr_blkno = 0, }; + struct ocfs2_super *osb = OCFS2_SB(ac->ac_inode->i_sb); BUG_ON(ac->ac_bits_given >= ac->ac_bits_wanted); @@ -1810,6 +2317,8 @@ int __ocfs2_claim_clusters(struct ocfs2_super *osb, && ac->ac_which != OCFS2_AC_USE_MAIN); if (ac->ac_which == OCFS2_AC_USE_LOCAL) { + WARN_ON(min_clusters > 1); + status = ocfs2_claim_local_alloc_bits(osb, handle, ac, @@ -1832,20 +2341,19 @@ int __ocfs2_claim_clusters(struct ocfs2_super *osb, if (bits_wanted > (osb->bitmap_cpg - 1)) bits_wanted = osb->bitmap_cpg - 1; - status = ocfs2_claim_suballoc_bits(osb, - ac, + status = ocfs2_claim_suballoc_bits(ac, handle, bits_wanted, min_clusters, - &bg_bit_off, - num_clusters, - &bg_blkno); + &res); if (!status) { + BUG_ON(res.sr_blkno); /* cluster alloc can't set */ *cluster_start = ocfs2_desc_bitmap_to_cluster_off(ac->ac_inode, - bg_blkno, - bg_bit_off); + res.sr_bg_blkno, + res.sr_bit_offset); atomic_inc(&osb->alloc_stats.bitmap_data); + *num_clusters = res.sr_bits; } } if (status < 0) { @@ -1857,12 +2365,12 @@ int __ocfs2_claim_clusters(struct ocfs2_super *osb, ac->ac_bits_given += *num_clusters; bail: - mlog_exit(status); + if (status) + mlog_errno(status); return status; } -int ocfs2_claim_clusters(struct ocfs2_super *osb, - handle_t *handle, +int ocfs2_claim_clusters(handle_t *handle, struct ocfs2_alloc_context *ac, u32 min_clusters, u32 *cluster_start, @@ -1870,56 +2378,69 @@ int ocfs2_claim_clusters(struct ocfs2_super *osb, { unsigned int bits_wanted = ac->ac_bits_wanted - ac->ac_bits_given; - return __ocfs2_claim_clusters(osb, handle, ac, min_clusters, + return __ocfs2_claim_clusters(handle, ac, min_clusters, bits_wanted, cluster_start, num_clusters); } -static inline int ocfs2_block_group_clear_bits(handle_t *handle, - struct inode *alloc_inode, - struct ocfs2_group_desc *bg, - struct buffer_head *group_bh, - unsigned int bit_off, - unsigned int num_bits) +static int ocfs2_block_group_clear_bits(handle_t *handle, + struct inode *alloc_inode, + struct ocfs2_group_desc *bg, + struct buffer_head *group_bh, + unsigned int bit_off, + unsigned int num_bits, + void (*undo_fn)(unsigned int bit, + unsigned long *bmap)) { int status; unsigned int tmp; - int journal_type = OCFS2_JOURNAL_ACCESS_WRITE; struct ocfs2_group_desc *undo_bg = NULL; - mlog_entry_void(); - /* The caller got this descriptor from * ocfs2_read_group_descriptor(). Any corruption is a code bug. */ BUG_ON(!OCFS2_IS_VALID_GROUP_DESC(bg)); - mlog(0, "off = %u, num = %u\n", bit_off, num_bits); - - if (ocfs2_is_cluster_bitmap(alloc_inode)) - journal_type = OCFS2_JOURNAL_ACCESS_UNDO; + trace_ocfs2_block_group_clear_bits(bit_off, num_bits); - status = ocfs2_journal_access_gd(handle, alloc_inode, group_bh, - journal_type); + BUG_ON(undo_fn && !ocfs2_is_cluster_bitmap(alloc_inode)); + status = ocfs2_journal_access_gd(handle, INODE_CACHE(alloc_inode), + group_bh, + undo_fn ? + OCFS2_JOURNAL_ACCESS_UNDO : + OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); goto bail; } - if (ocfs2_is_cluster_bitmap(alloc_inode)) - undo_bg = (struct ocfs2_group_desc *) bh2jh(group_bh)->b_committed_data; + if (undo_fn) { + jbd_lock_bh_state(group_bh); + undo_bg = (struct ocfs2_group_desc *) + bh2jh(group_bh)->b_committed_data; + BUG_ON(!undo_bg); + } tmp = num_bits; while(tmp--) { ocfs2_clear_bit((bit_off + tmp), (unsigned long *) bg->bg_bitmap); - if (ocfs2_is_cluster_bitmap(alloc_inode)) - ocfs2_set_bit(bit_off + tmp, - (unsigned long *) undo_bg->bg_bitmap); + if (undo_fn) + undo_fn(bit_off + tmp, + (unsigned long *) undo_bg->bg_bitmap); } le16_add_cpu(&bg->bg_free_bits_count, num_bits); + if (le16_to_cpu(bg->bg_free_bits_count) > le16_to_cpu(bg->bg_bits)) { + ocfs2_error(alloc_inode->i_sb, "Group descriptor # %llu has bit" + " count %u but claims %u are freed. num_bits %d", + (unsigned long long)le64_to_cpu(bg->bg_blkno), + le16_to_cpu(bg->bg_bits), + le16_to_cpu(bg->bg_free_bits_count), num_bits); + return -EROFS; + } - status = ocfs2_journal_dirty(handle, group_bh); - if (status < 0) - mlog_errno(status); + if (undo_fn) + jbd_unlock_bh_state(group_bh); + + ocfs2_journal_dirty(handle, group_bh); bail: return status; } @@ -1927,12 +2448,14 @@ bail: /* * expects the suballoc inode to already be locked. */ -int ocfs2_free_suballoc_bits(handle_t *handle, - struct inode *alloc_inode, - struct buffer_head *alloc_bh, - unsigned int start_bit, - u64 bg_blkno, - unsigned int count) +static int _ocfs2_free_suballoc_bits(handle_t *handle, + struct inode *alloc_inode, + struct buffer_head *alloc_bh, + unsigned int start_bit, + u64 bg_blkno, + unsigned int count, + void (*undo_fn)(unsigned int bit, + unsigned long *bitmap)) { int status = 0; u32 tmp_used; @@ -1941,19 +2464,18 @@ int ocfs2_free_suballoc_bits(handle_t *handle, struct buffer_head *group_bh = NULL; struct ocfs2_group_desc *group; - mlog_entry_void(); - /* The alloc_bh comes from ocfs2_free_dinode() or * ocfs2_free_clusters(). The callers have all locked the * allocator and gotten alloc_bh from the lock call. This - * validates the dinode buffer. Any corruption that has happended + * validates the dinode buffer. Any corruption that has happened * is a code bug. */ BUG_ON(!OCFS2_IS_VALID_DINODE(fe)); BUG_ON((count + start_bit) > ocfs2_bits_per_group(cl)); - mlog(0, "%llu: freeing %u bits from group %llu, starting at %u\n", - (unsigned long long)OCFS2_I(alloc_inode)->ip_blkno, count, - (unsigned long long)bg_blkno, start_bit); + trace_ocfs2_free_suballoc_bits( + (unsigned long long)OCFS2_I(alloc_inode)->ip_blkno, + (unsigned long long)bg_blkno, + start_bit, count); status = ocfs2_read_group_descriptor(alloc_inode, fe, bg_blkno, &group_bh); @@ -1967,14 +2489,14 @@ int ocfs2_free_suballoc_bits(handle_t *handle, status = ocfs2_block_group_clear_bits(handle, alloc_inode, group, group_bh, - start_bit, count); + start_bit, count, undo_fn); if (status < 0) { mlog_errno(status); goto bail; } - status = ocfs2_journal_access_di(handle, alloc_inode, alloc_bh, - OCFS2_JOURNAL_ACCESS_WRITE); + status = ocfs2_journal_access_di(handle, INODE_CACHE(alloc_inode), + alloc_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); goto bail; @@ -1984,20 +2506,27 @@ int ocfs2_free_suballoc_bits(handle_t *handle, count); tmp_used = le32_to_cpu(fe->id1.bitmap1.i_used); fe->id1.bitmap1.i_used = cpu_to_le32(tmp_used - count); - - status = ocfs2_journal_dirty(handle, alloc_bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } + ocfs2_journal_dirty(handle, alloc_bh); bail: brelse(group_bh); - mlog_exit(status); + if (status) + mlog_errno(status); return status; } +int ocfs2_free_suballoc_bits(handle_t *handle, + struct inode *alloc_inode, + struct buffer_head *alloc_bh, + unsigned int start_bit, + u64 bg_blkno, + unsigned int count) +{ + return _ocfs2_free_suballoc_bits(handle, alloc_inode, alloc_bh, + start_bit, bg_blkno, count, NULL); +} + int ocfs2_free_dinode(handle_t *handle, struct inode *inode_alloc_inode, struct buffer_head *inode_alloc_bh, @@ -2007,15 +2536,19 @@ int ocfs2_free_dinode(handle_t *handle, u16 bit = le16_to_cpu(di->i_suballoc_bit); u64 bg_blkno = ocfs2_which_suballoc_group(blk, bit); + if (di->i_suballoc_loc) + bg_blkno = le64_to_cpu(di->i_suballoc_loc); return ocfs2_free_suballoc_bits(handle, inode_alloc_inode, inode_alloc_bh, bit, bg_blkno, 1); } -int ocfs2_free_clusters(handle_t *handle, - struct inode *bitmap_inode, - struct buffer_head *bitmap_bh, - u64 start_blk, - unsigned int num_clusters) +static int _ocfs2_free_clusters(handle_t *handle, + struct inode *bitmap_inode, + struct buffer_head *bitmap_bh, + u64 start_blk, + unsigned int num_clusters, + void (*undo_fn)(unsigned int bit, + unsigned long *bitmap)) { int status; u16 bg_start_bit; @@ -2024,11 +2557,8 @@ int ocfs2_free_clusters(handle_t *handle, /* You can't ever have a contiguous set of clusters * bigger than a block group bitmap so we never have to worry - * about looping on them. */ - - mlog_entry_void(); - - /* This is expensive. We can safely remove once this stuff has + * about looping on them. + * This is expensive. We can safely remove once this stuff has * gotten tested really well. */ BUG_ON(start_blk != ocfs2_clusters_to_blocks(bitmap_inode->i_sb, ocfs2_blocks_to_clusters(bitmap_inode->i_sb, start_blk))); @@ -2037,14 +2567,13 @@ int ocfs2_free_clusters(handle_t *handle, ocfs2_block_to_cluster_group(bitmap_inode, start_blk, &bg_blkno, &bg_start_bit); - mlog(0, "want to free %u clusters starting at block %llu\n", - num_clusters, (unsigned long long)start_blk); - mlog(0, "bg_blkno = %llu, bg_start_bit = %u\n", - (unsigned long long)bg_blkno, bg_start_bit); + trace_ocfs2_free_clusters((unsigned long long)bg_blkno, + (unsigned long long)start_blk, + bg_start_bit, num_clusters); - status = ocfs2_free_suballoc_bits(handle, bitmap_inode, bitmap_bh, - bg_start_bit, bg_blkno, - num_clusters); + status = _ocfs2_free_suballoc_bits(handle, bitmap_inode, bitmap_bh, + bg_start_bit, bg_blkno, + num_clusters, undo_fn); if (status < 0) { mlog_errno(status); goto out; @@ -2054,10 +2583,37 @@ int ocfs2_free_clusters(handle_t *handle, num_clusters); out: - mlog_exit(status); + if (status) + mlog_errno(status); return status; } +int ocfs2_free_clusters(handle_t *handle, + struct inode *bitmap_inode, + struct buffer_head *bitmap_bh, + u64 start_blk, + unsigned int num_clusters) +{ + return _ocfs2_free_clusters(handle, bitmap_inode, bitmap_bh, + start_blk, num_clusters, + _ocfs2_set_bit); +} + +/* + * Give never-used clusters back to the global bitmap. We don't need + * to protect these bits in the undo buffer. + */ +int ocfs2_release_clusters(handle_t *handle, + struct inode *bitmap_inode, + struct buffer_head *bitmap_bh, + u64 start_blk, + unsigned int num_clusters) +{ + return _ocfs2_free_clusters(handle, bitmap_inode, bitmap_bh, + start_blk, num_clusters, + _ocfs2_clear_bit); +} + static inline void ocfs2_debug_bg(struct ocfs2_group_desc *bg) { printk("Block Group:\n"); @@ -2131,7 +2687,7 @@ int ocfs2_lock_allocators(struct inode *inode, BUG_ON(clusters_to_add != 0 && data_ac == NULL); - num_free_extents = ocfs2_num_free_extents(osb, inode, et); + num_free_extents = ocfs2_num_free_extents(osb, et); if (num_free_extents < 0) { ret = num_free_extents; mlog_errno(ret); @@ -2191,13 +2747,14 @@ out: * suballoc_bit. */ static int ocfs2_get_suballoc_slot_bit(struct ocfs2_super *osb, u64 blkno, - u16 *suballoc_slot, u16 *suballoc_bit) + u16 *suballoc_slot, u64 *group_blkno, + u16 *suballoc_bit) { int status; struct buffer_head *inode_bh = NULL; struct ocfs2_dinode *inode_fe; - mlog_entry("blkno: %llu\n", (unsigned long long)blkno); + trace_ocfs2_get_suballoc_slot_bit((unsigned long long)blkno); /* dirty read disk */ status = ocfs2_read_blocks_sync(osb, blkno, 1, &inode_bh); @@ -2228,11 +2785,14 @@ static int ocfs2_get_suballoc_slot_bit(struct ocfs2_super *osb, u64 blkno, *suballoc_slot = le16_to_cpu(inode_fe->i_suballoc_slot); if (suballoc_bit) *suballoc_bit = le16_to_cpu(inode_fe->i_suballoc_bit); + if (group_blkno) + *group_blkno = le64_to_cpu(inode_fe->i_suballoc_loc); bail: brelse(inode_bh); - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -2245,29 +2805,31 @@ bail: */ static int ocfs2_test_suballoc_bit(struct ocfs2_super *osb, struct inode *suballoc, - struct buffer_head *alloc_bh, u64 blkno, + struct buffer_head *alloc_bh, + u64 group_blkno, u64 blkno, u16 bit, int *res) { - struct ocfs2_dinode *alloc_fe; + struct ocfs2_dinode *alloc_di; struct ocfs2_group_desc *group; struct buffer_head *group_bh = NULL; u64 bg_blkno; int status; - mlog_entry("blkno: %llu bit: %u\n", (unsigned long long)blkno, - (unsigned int)bit); + trace_ocfs2_test_suballoc_bit((unsigned long long)blkno, + (unsigned int)bit); - alloc_fe = (struct ocfs2_dinode *)alloc_bh->b_data; - if ((bit + 1) > ocfs2_bits_per_group(&alloc_fe->id2.i_chain)) { + alloc_di = (struct ocfs2_dinode *)alloc_bh->b_data; + if ((bit + 1) > ocfs2_bits_per_group(&alloc_di->id2.i_chain)) { mlog(ML_ERROR, "suballoc bit %u out of range of %u\n", (unsigned int)bit, - ocfs2_bits_per_group(&alloc_fe->id2.i_chain)); + ocfs2_bits_per_group(&alloc_di->id2.i_chain)); status = -EINVAL; goto bail; } - bg_blkno = ocfs2_which_suballoc_group(blkno, bit); - status = ocfs2_read_group_descriptor(suballoc, alloc_fe, bg_blkno, + bg_blkno = group_blkno ? group_blkno : + ocfs2_which_suballoc_group(blkno, bit); + status = ocfs2_read_group_descriptor(suballoc, alloc_di, bg_blkno, &group_bh); if (status < 0) { mlog(ML_ERROR, "read group %llu failed %d\n", @@ -2281,7 +2843,8 @@ static int ocfs2_test_suballoc_bit(struct ocfs2_super *osb, bail: brelse(group_bh); - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -2301,14 +2864,15 @@ bail: int ocfs2_test_inode_bit(struct ocfs2_super *osb, u64 blkno, int *res) { int status; + u64 group_blkno = 0; u16 suballoc_bit = 0, suballoc_slot = 0; struct inode *inode_alloc_inode; struct buffer_head *alloc_bh = NULL; - mlog_entry("blkno: %llu", (unsigned long long)blkno); + trace_ocfs2_test_inode_bit((unsigned long long)blkno); status = ocfs2_get_suballoc_slot_bit(osb, blkno, &suballoc_slot, - &suballoc_bit); + &group_blkno, &suballoc_bit); if (status < 0) { mlog(ML_ERROR, "get alloc slot and bit failed %d\n", status); goto bail; @@ -2330,13 +2894,14 @@ int ocfs2_test_inode_bit(struct ocfs2_super *osb, u64 blkno, int *res) status = ocfs2_inode_lock(inode_alloc_inode, &alloc_bh, 0); if (status < 0) { mutex_unlock(&inode_alloc_inode->i_mutex); + iput(inode_alloc_inode); mlog(ML_ERROR, "lock on alloc inode on slot %u failed %d\n", (u32)suballoc_slot, status); goto bail; } status = ocfs2_test_suballoc_bit(osb, inode_alloc_inode, alloc_bh, - blkno, suballoc_bit, res); + group_blkno, blkno, suballoc_bit, res); if (status < 0) mlog(ML_ERROR, "test suballoc bit failed %d\n", status); @@ -2346,6 +2911,7 @@ int ocfs2_test_inode_bit(struct ocfs2_super *osb, u64 blkno, int *res) iput(inode_alloc_inode); brelse(alloc_bh); bail: - mlog_exit(status); + if (status) + mlog_errno(status); return status; } diff --git a/fs/ocfs2/suballoc.h b/fs/ocfs2/suballoc.h index 8c9a78a4316..2d2501767c0 100644 --- a/fs/ocfs2/suballoc.h +++ b/fs/ocfs2/suballoc.h @@ -26,13 +26,14 @@ #ifndef _CHAINALLOC_H_ #define _CHAINALLOC_H_ +struct ocfs2_suballoc_result; typedef int (group_search_t)(struct inode *, struct buffer_head *, u32, /* bits_wanted */ u32, /* min_bits */ u64, /* max_block */ - u16 *, /* *bit_off */ - u16 *); /* *bits_found */ + struct ocfs2_suballoc_result *); + /* found bits */ struct ocfs2_alloc_context { struct inode *ac_inode; /* which bitmap are we allocating from? */ @@ -48,14 +49,20 @@ struct ocfs2_alloc_context { /* these are used by the chain search */ u16 ac_chain; - int ac_allow_chain_relink; + int ac_disable_chain_relink; group_search_t *ac_group_search; u64 ac_last_group; u64 ac_max_block; /* Highest block number to allocate. 0 is is the same as ~0 - unlimited */ + + int ac_find_loc_only; /* hack for reflink operation ordering */ + struct ocfs2_suballoc_result *ac_find_loc_priv; /* */ + + struct ocfs2_alloc_reservation *ac_resv; }; +void ocfs2_init_steal_slots(struct ocfs2_super *osb); void ocfs2_free_alloc_context(struct ocfs2_alloc_context *ac); static inline int ocfs2_alloc_context_bits_left(struct ocfs2_alloc_context *ac) { @@ -79,22 +86,37 @@ int ocfs2_reserve_clusters(struct ocfs2_super *osb, u32 bits_wanted, struct ocfs2_alloc_context **ac); -int ocfs2_claim_metadata(struct ocfs2_super *osb, +int ocfs2_alloc_dinode_update_counts(struct inode *inode, handle_t *handle, + struct buffer_head *di_bh, + u32 num_bits, + u16 chain); +void ocfs2_rollback_alloc_dinode_counts(struct inode *inode, + struct buffer_head *di_bh, + u32 num_bits, + u16 chain); +int ocfs2_block_group_set_bits(handle_t *handle, + struct inode *alloc_inode, + struct ocfs2_group_desc *bg, + struct buffer_head *group_bh, + unsigned int bit_off, + unsigned int num_bits); + +int ocfs2_claim_metadata(handle_t *handle, struct ocfs2_alloc_context *ac, u32 bits_wanted, + u64 *suballoc_loc, u16 *suballoc_bit_start, u32 *num_bits, u64 *blkno_start); -int ocfs2_claim_new_inode(struct ocfs2_super *osb, - handle_t *handle, +int ocfs2_claim_new_inode(handle_t *handle, struct inode *dir, struct buffer_head *parent_fe_bh, struct ocfs2_alloc_context *ac, + u64 *suballoc_loc, u16 *suballoc_bit, u64 *fe_blkno); -int ocfs2_claim_clusters(struct ocfs2_super *osb, - handle_t *handle, +int ocfs2_claim_clusters(handle_t *handle, struct ocfs2_alloc_context *ac, u32 min_clusters, u32 *cluster_start, @@ -103,8 +125,7 @@ int ocfs2_claim_clusters(struct ocfs2_super *osb, * Use this variant of ocfs2_claim_clusters to specify a maxiumum * number of clusters smaller than the allocation reserved. */ -int __ocfs2_claim_clusters(struct ocfs2_super *osb, - handle_t *handle, +int __ocfs2_claim_clusters(handle_t *handle, struct ocfs2_alloc_context *ac, u32 min_clusters, u32 max_clusters, @@ -126,6 +147,11 @@ int ocfs2_free_clusters(handle_t *handle, struct buffer_head *bitmap_bh, u64 start_blk, unsigned int num_clusters); +int ocfs2_release_clusters(handle_t *handle, + struct inode *bitmap_inode, + struct buffer_head *bitmap_bh, + u64 start_blk, + unsigned int num_clusters); static inline u64 ocfs2_which_suballoc_group(u64 block, unsigned int bit) { @@ -190,4 +216,22 @@ int ocfs2_lock_allocators(struct inode *inode, struct ocfs2_extent_tree *et, struct ocfs2_alloc_context **meta_ac); int ocfs2_test_inode_bit(struct ocfs2_super *osb, u64 blkno, int *res); + + + +/* + * The following two interfaces are for ocfs2_create_inode_in_orphan(). + */ +int ocfs2_find_new_inode_loc(struct inode *dir, + struct buffer_head *parent_fe_bh, + struct ocfs2_alloc_context *ac, + u64 *fe_blkno); + +int ocfs2_claim_new_inode_at_loc(handle_t *handle, + struct inode *dir, + struct ocfs2_alloc_context *ac, + u64 *suballoc_loc, + u16 *suballoc_bit, + u64 di_blkno); + #endif /* _CHAINALLOC_H_ */ diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c index 5c6163f5503..ddb662b3244 100644 --- a/fs/ocfs2/super.c +++ b/fs/ocfs2/super.c @@ -28,7 +28,6 @@ #include <linux/types.h> #include <linux/slab.h> #include <linux/highmem.h> -#include <linux/utsname.h> #include <linux/init.h> #include <linux/random.h> #include <linux/statfs.h> @@ -42,8 +41,11 @@ #include <linux/mount.h> #include <linux/seq_file.h> #include <linux/quotaops.h> +#include <linux/cleancache.h> + +#define CREATE_TRACE_POINTS +#include "ocfs2_trace.h" -#define MLOG_MASK_PREFIX ML_SUPER #include <cluster/masklog.h> #include "ocfs2.h" @@ -52,6 +54,7 @@ #include "ocfs1_fs_compat.h" #include "alloc.h" +#include "aops.h" #include "blockcheck.h" #include "dlmglue.h" #include "export.h" @@ -65,26 +68,28 @@ #include "super.h" #include "sysfile.h" #include "uptodate.h" -#include "ver.h" #include "xattr.h" #include "quota.h" +#include "refcounttree.h" +#include "suballoc.h" #include "buffer_head_io.h" -static struct kmem_cache *ocfs2_inode_cachep = NULL; +static struct kmem_cache *ocfs2_inode_cachep; struct kmem_cache *ocfs2_dquot_cachep; struct kmem_cache *ocfs2_qf_chunk_cachep; -/* OCFS2 needs to schedule several differnt types of work which +/* OCFS2 needs to schedule several different types of work which * require cluster locking, disk I/O, recovery waits, etc. Since these * types of work tend to be heavy we avoid using the kernel events * workqueue and schedule on our own. */ struct workqueue_struct *ocfs2_wq = NULL; -static struct dentry *ocfs2_debugfs_root = NULL; +static struct dentry *ocfs2_debugfs_root; MODULE_AUTHOR("Oracle"); MODULE_LICENSE("GPL"); +MODULE_DESCRIPTION("OCFS2 cluster file system"); struct mount_options { @@ -92,14 +97,18 @@ struct mount_options unsigned long mount_opt; unsigned int atime_quantum; signed short slot; - unsigned int localalloc_opt; + int localalloc_opt; + unsigned int resv_level; + int dir_resv_level; char cluster_stack[OCFS2_STACK_LABEL_LEN + 1]; }; static int ocfs2_parse_options(struct super_block *sb, char *options, struct mount_options *mopt, int is_remount); -static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt); +static int ocfs2_check_set_options(struct super_block *sb, + struct mount_options *options); +static int ocfs2_show_options(struct seq_file *s, struct dentry *root); static void ocfs2_put_super(struct super_block *sb); static int ocfs2_mount_volume(struct super_block *sb); static int ocfs2_remount(struct super_block *sb, int *flags, char *data); @@ -118,15 +127,16 @@ static void ocfs2_release_system_inodes(struct ocfs2_super *osb); static int ocfs2_check_volume(struct ocfs2_super *osb); static int ocfs2_verify_volume(struct ocfs2_dinode *di, struct buffer_head *bh, - u32 sectsize); + u32 sectsize, + struct ocfs2_blockcheck_stats *stats); static int ocfs2_initialize_super(struct super_block *sb, struct buffer_head *bh, - int sector_size); + int sector_size, + struct ocfs2_blockcheck_stats *stats); static int ocfs2_get_sector(struct super_block *sb, struct buffer_head **bh, int block, int sect_size); -static void ocfs2_write_super(struct super_block *sb); static struct inode *ocfs2_alloc_inode(struct super_block *sb); static void ocfs2_destroy_inode(struct inode *inode); static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend); @@ -138,10 +148,8 @@ static const struct super_operations ocfs2_sops = { .alloc_inode = ocfs2_alloc_inode, .destroy_inode = ocfs2_destroy_inode, .drop_inode = ocfs2_drop_inode, - .clear_inode = ocfs2_clear_inode, - .delete_inode = ocfs2_delete_inode, + .evict_inode = ocfs2_evict_inode, .sync_fs = ocfs2_sync_fs, - .write_super = ocfs2_write_super, .put_super = ocfs2_put_super, .remount_fs = ocfs2_remount, .show_options = ocfs2_show_options, @@ -157,6 +165,7 @@ enum { Opt_nointr, Opt_hb_none, Opt_hb_local, + Opt_hb_global, Opt_data_ordered, Opt_data_writeback, Opt_atime_quantum, @@ -172,6 +181,10 @@ enum { Opt_noacl, Opt_usrquota, Opt_grpquota, + Opt_coherency_buffered, + Opt_coherency_full, + Opt_resv_level, + Opt_dir_resv_level, Opt_err, }; @@ -183,6 +196,7 @@ static const match_table_t tokens = { {Opt_nointr, "nointr"}, {Opt_hb_none, OCFS2_HB_NONE}, {Opt_hb_local, OCFS2_HB_LOCAL}, + {Opt_hb_global, OCFS2_HB_GLOBAL}, {Opt_data_ordered, "data=ordered"}, {Opt_data_writeback, "data=writeback"}, {Opt_atime_quantum, "atime_quantum=%u"}, @@ -198,16 +212,20 @@ static const match_table_t tokens = { {Opt_noacl, "noacl"}, {Opt_usrquota, "usrquota"}, {Opt_grpquota, "grpquota"}, + {Opt_coherency_buffered, "coherency=buffered"}, + {Opt_coherency_full, "coherency=full"}, + {Opt_resv_level, "resv_level=%u"}, + {Opt_dir_resv_level, "dir_resv_level=%u"}, {Opt_err, NULL} }; #ifdef CONFIG_DEBUG_FS static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len) { - int out = 0; - int i; struct ocfs2_cluster_connection *cconn = osb->cconn; struct ocfs2_recovery_map *rm = osb->recovery_map; + struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan; + int i, out = 0; out += snprintf(buf + out, len - out, "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n", @@ -232,20 +250,24 @@ static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len) "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount", osb->s_mount_opt, osb->s_atime_quantum); - out += snprintf(buf + out, len - out, - "%10s => Stack: %s Name: %*s Version: %d.%d\n", - "Cluster", - (*osb->osb_cluster_stack == '\0' ? - "o2cb" : osb->osb_cluster_stack), - cconn->cc_namelen, cconn->cc_name, - cconn->cc_version.pv_major, cconn->cc_version.pv_minor); + if (cconn) { + out += snprintf(buf + out, len - out, + "%10s => Stack: %s Name: %*s " + "Version: %d.%d\n", "Cluster", + (*osb->osb_cluster_stack == '\0' ? + "o2cb" : osb->osb_cluster_stack), + cconn->cc_namelen, cconn->cc_name, + cconn->cc_version.pv_major, + cconn->cc_version.pv_minor); + } spin_lock(&osb->dc_task_lock); out += snprintf(buf + out, len - out, "%10s => Pid: %d Count: %lu WakeSeq: %lu " "WorkSeq: %lu\n", "DownCnvt", - task_pid_nr(osb->dc_task), osb->blocked_lock_count, - osb->dc_wake_sequence, osb->dc_work_sequence); + (osb->dc_task ? task_pid_nr(osb->dc_task) : -1), + osb->blocked_lock_count, osb->dc_wake_sequence, + osb->dc_work_sequence); spin_unlock(&osb->dc_task_lock); spin_lock(&osb->osb_lock); @@ -264,15 +286,15 @@ static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len) spin_unlock(&osb->osb_lock); out += snprintf(buf + out, len - out, - "%10s => Pid: %d Interval: %lu Needs: %d\n", "Commit", - task_pid_nr(osb->commit_task), osb->osb_commit_interval, - atomic_read(&osb->needs_checkpoint)); + "%10s => Pid: %d Interval: %lu\n", "Commit", + (osb->commit_task ? task_pid_nr(osb->commit_task) : -1), + osb->osb_commit_interval); out += snprintf(buf + out, len - out, - "%10s => State: %d NumTxns: %d TxnId: %lu\n", + "%10s => State: %d TxnId: %lu NumTxns: %d\n", "Journal", osb->journal->j_state, - atomic_read(&osb->journal->j_num_trans), - osb->journal->j_trans_id); + osb->journal->j_trans_id, + atomic_read(&osb->journal->j_num_trans)); out += snprintf(buf + out, len - out, "%10s => GlobalAllocs: %d LocalAllocs: %d " @@ -293,14 +315,26 @@ static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len) spin_lock(&osb->osb_lock); out += snprintf(buf + out, len - out, - "%10s => Slot: %d NumStolen: %d\n", "Steal", + "%10s => InodeSlot: %d StolenInodes: %d, " + "MetaSlot: %d StolenMeta: %d\n", "Steal", osb->s_inode_steal_slot, - atomic_read(&osb->s_num_inodes_stolen)); + atomic_read(&osb->s_num_inodes_stolen), + osb->s_meta_steal_slot, + atomic_read(&osb->s_num_meta_stolen)); spin_unlock(&osb->osb_lock); + out += snprintf(buf + out, len - out, "OrphanScan => "); + out += snprintf(buf + out, len - out, "Local: %u Global: %u ", + os->os_count, os->os_seqno); + out += snprintf(buf + out, len - out, " Last Scan: "); + if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE) + out += snprintf(buf + out, len - out, "Disabled\n"); + else + out += snprintf(buf + out, len - out, "%lu seconds ago\n", + (get_seconds() - os->os_scantime.tv_sec)); + out += snprintf(buf + out, len - out, "%10s => %3s %10s\n", "Slots", "Num", "RecoGen"); - for (i = 0; i < osb->max_slots; ++i) { out += snprintf(buf + out, len - out, "%10s %c %3d %10d\n", @@ -358,31 +392,19 @@ static ssize_t ocfs2_debug_read(struct file *file, char __user *buf, } #endif /* CONFIG_DEBUG_FS */ -static struct file_operations ocfs2_osb_debug_fops = { +static const struct file_operations ocfs2_osb_debug_fops = { .open = ocfs2_osb_debug_open, .release = ocfs2_debug_release, .read = ocfs2_debug_read, .llseek = generic_file_llseek, }; -/* - * write_super and sync_fs ripped right out of ext3. - */ -static void ocfs2_write_super(struct super_block *sb) -{ - if (mutex_trylock(&sb->s_lock) != 0) - BUG(); - sb->s_dirt = 0; -} - static int ocfs2_sync_fs(struct super_block *sb, int wait) { int status; tid_t target; struct ocfs2_super *osb = OCFS2_SB(sb); - sb->s_dirt = 0; - if (ocfs2_is_hard_readonly(osb)) return -EROFS; @@ -422,8 +444,6 @@ static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb) int status = 0; int i; - mlog_entry_void(); - new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0); if (IS_ERR(new)) { status = PTR_ERR(new); @@ -459,7 +479,8 @@ static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb) } bail: - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -469,8 +490,6 @@ static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb) int status = 0; int i; - mlog_entry_void(); - for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1; i < NUM_SYSTEM_INODES; i++) { @@ -489,7 +508,8 @@ static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb) } bail: - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -498,13 +518,11 @@ static void ocfs2_release_system_inodes(struct ocfs2_super *osb) int i; struct inode *inode; - mlog_entry_void(); - - for (i = 0; i < NUM_SYSTEM_INODES; i++) { - inode = osb->system_inodes[i]; + for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) { + inode = osb->global_system_inodes[i]; if (inode) { iput(inode); - osb->system_inodes[i] = NULL; + osb->global_system_inodes[i] = NULL; } } @@ -520,7 +538,18 @@ static void ocfs2_release_system_inodes(struct ocfs2_super *osb) osb->root_inode = NULL; } - mlog_exit(0); + if (!osb->local_system_inodes) + return; + + for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) { + if (osb->local_system_inodes[i]) { + iput(osb->local_system_inodes[i]); + osb->local_system_inodes[i] = NULL; + } + } + + kfree(osb->local_system_inodes); + osb->local_system_inodes = NULL; } /* We're allocating fs objects, use GFP_NOFS */ @@ -532,15 +561,24 @@ static struct inode *ocfs2_alloc_inode(struct super_block *sb) if (!oi) return NULL; + oi->i_sync_tid = 0; + oi->i_datasync_tid = 0; + jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode); return &oi->vfs_inode; } -static void ocfs2_destroy_inode(struct inode *inode) +static void ocfs2_i_callback(struct rcu_head *head) { + struct inode *inode = container_of(head, struct inode, i_rcu); kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode)); } +static void ocfs2_destroy_inode(struct inode *inode) +{ + call_rcu(&inode->i_rcu, ocfs2_i_callback); +} + static unsigned long long ocfs2_max_file_offset(unsigned int bbits, unsigned int cbits) { @@ -555,7 +593,7 @@ static unsigned long long ocfs2_max_file_offset(unsigned int bbits, */ #if BITS_PER_LONG == 32 -# if defined(CONFIG_LBD) +# if defined(CONFIG_LBDAF) BUILD_BUG_ON(sizeof(sector_t) != 8); /* * We might be limited by page cache size. @@ -594,14 +632,19 @@ static int ocfs2_remount(struct super_block *sb, int *flags, char *data) int ret = 0; struct mount_options parsed_options; struct ocfs2_super *osb = OCFS2_SB(sb); + u32 tmp; + + sync_filesystem(sb); - if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) { + if (!ocfs2_parse_options(sb, data, &parsed_options, 1) || + !ocfs2_check_set_options(sb, &parsed_options)) { ret = -EINVAL; goto out; } - if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) != - (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) { + tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL | + OCFS2_MOUNT_HB_NONE; + if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) { ret = -EINVAL; mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n"); goto out; @@ -641,12 +684,9 @@ static int ocfs2_remount(struct super_block *sb, int *flags, char *data) } if (*flags & MS_RDONLY) { - mlog(0, "Going to ro mode.\n"); sb->s_flags |= MS_RDONLY; osb->osb_flags |= OCFS2_OSB_SOFT_RO; } else { - mlog(0, "Making ro filesystem writeable.\n"); - if (osb->osb_flags & OCFS2_OSB_ERROR_FS) { mlog(ML_ERROR, "Cannot remount RDWR " "filesystem due to previous errors.\n"); @@ -664,6 +704,7 @@ static int ocfs2_remount(struct super_block *sb, int *flags, char *data) sb->s_flags &= ~MS_RDONLY; osb->osb_flags &= ~OCFS2_OSB_SOFT_RO; } + trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags); unlock_osb: spin_unlock(&osb->osb_lock); /* Enable quota accounting after remounting RW */ @@ -686,8 +727,6 @@ unlock_osb: if (!ret) { /* Only save off the new mount options in case of a successful * remount. */ - if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR)) - parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL; osb->s_mount_opt = parsed_options.mount_opt; osb->s_atime_quantum = parsed_options.atime_quantum; osb->preferred_slot = parsed_options.slot; @@ -696,6 +735,10 @@ unlock_osb: if (!ocfs2_is_hard_readonly(osb)) ocfs2_set_journal_params(osb); + + sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | + ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? + MS_POSIXACL : 0); } out: return ret; @@ -703,7 +746,8 @@ out: static int ocfs2_sb_probe(struct super_block *sb, struct buffer_head **bh, - int *sector_size) + int *sector_size, + struct ocfs2_blockcheck_stats *stats) { int status, tmpstat; struct ocfs1_vol_disk_hdr *hdr; @@ -766,16 +810,20 @@ static int ocfs2_sb_probe(struct super_block *sb, if (tmpstat < 0) { status = tmpstat; mlog_errno(status); - goto bail; + break; } di = (struct ocfs2_dinode *) (*bh)->b_data; - status = ocfs2_verify_volume(di, *bh, blksize); - if (status >= 0) - goto bail; - brelse(*bh); - *bh = NULL; - if (status != -EAGAIN) + memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats)); + spin_lock_init(&stats->b_lock); + tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats); + if (tmpstat < 0) { + brelse(*bh); + *bh = NULL; + } + if (tmpstat != -EAGAIN) { + status = tmpstat; break; + } } bail: @@ -784,23 +832,29 @@ bail: static int ocfs2_verify_heartbeat(struct ocfs2_super *osb) { - if (ocfs2_mount_local(osb)) { - if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { + u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL; + + if (osb->s_mount_opt & hb_enabled) { + if (ocfs2_mount_local(osb)) { mlog(ML_ERROR, "Cannot heartbeat on a locally " "mounted device.\n"); return -EINVAL; } - } - - if (ocfs2_userspace_stack(osb)) { - if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { + if (ocfs2_userspace_stack(osb)) { mlog(ML_ERROR, "Userspace stack expected, but " "o2cb heartbeat arguments passed to mount\n"); return -EINVAL; } + if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) && + !ocfs2_cluster_o2cb_global_heartbeat(osb)) || + ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) && + ocfs2_cluster_o2cb_global_heartbeat(osb))) { + mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n"); + return -EINVAL; + } } - if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) { + if (!(osb->s_mount_opt & hb_enabled)) { if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) && !ocfs2_userspace_stack(osb)) { mlog(ML_ERROR, "Heartbeat has to be started to mount " @@ -853,13 +907,15 @@ static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend) if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type])) continue; if (unsuspend) - status = vfs_quota_enable( - sb_dqopt(sb)->files[type], - type, QFMT_OCFS2, - DQUOT_SUSPENDED); - else - status = vfs_quota_disable(sb, type, - DQUOT_SUSPENDED); + status = dquot_resume(sb, type); + else { + struct ocfs2_mem_dqinfo *oinfo; + + /* Cancel periodic syncing before suspending */ + oinfo = sb_dqinfo(sb, type)->dqi_priv; + cancel_delayed_work_sync(&oinfo->dqi_sync_work); + status = dquot_suspend(sb, type); + } if (status < 0) break; } @@ -890,8 +946,8 @@ static int ocfs2_enable_quotas(struct ocfs2_super *osb) status = -ENOENT; goto out_quota_off; } - status = vfs_quota_enable(inode[type], type, QFMT_OCFS2, - DQUOT_USAGE_ENABLED); + status = dquot_enable(inode[type], type, QFMT_OCFS2, + DQUOT_USAGE_ENABLED); if (status < 0) goto out_quota_off; } @@ -912,18 +968,22 @@ static void ocfs2_disable_quotas(struct ocfs2_super *osb) int type; struct inode *inode; struct super_block *sb = osb->sb; + struct ocfs2_mem_dqinfo *oinfo; /* We mostly ignore errors in this function because there's not much * we can do when we see them */ for (type = 0; type < MAXQUOTAS; type++) { if (!sb_has_quota_loaded(sb, type)) continue; + /* Cancel periodic syncing before we grab dqonoff_mutex */ + oinfo = sb_dqinfo(sb, type)->dqi_priv; + cancel_delayed_work_sync(&oinfo->dqi_sync_work); inode = igrab(sb->s_dquot.files[type]); /* Turn off quotas. This will remove all dquot structures from * memory and so they will be automatically synced to global * quota files */ - vfs_quota_disable(sb, type, DQUOT_USAGE_ENABLED | - DQUOT_LIMITS_ENABLED); + dquot_disable(sb, type, DQUOT_USAGE_ENABLED | + DQUOT_LIMITS_ENABLED); if (!inode) continue; iput(inode); @@ -931,8 +991,7 @@ static void ocfs2_disable_quotas(struct ocfs2_super *osb) } /* Handle quota on quotactl */ -static int ocfs2_quota_on(struct super_block *sb, int type, int format_id, - char *path, int remount) +static int ocfs2_quota_on(struct super_block *sb, int type, int format_id) { unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA}; @@ -940,30 +999,24 @@ static int ocfs2_quota_on(struct super_block *sb, int type, int format_id, if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type])) return -EINVAL; - if (remount) - return 0; /* Just ignore it has been handled in - * ocfs2_remount() */ - return vfs_quota_enable(sb_dqopt(sb)->files[type], type, - format_id, DQUOT_LIMITS_ENABLED); + return dquot_enable(sb_dqopt(sb)->files[type], type, + format_id, DQUOT_LIMITS_ENABLED); } /* Handle quota off quotactl */ -static int ocfs2_quota_off(struct super_block *sb, int type, int remount) +static int ocfs2_quota_off(struct super_block *sb, int type) { - if (remount) - return 0; /* Ignore now and handle later in - * ocfs2_remount() */ - return vfs_quota_disable(sb, type, DQUOT_LIMITS_ENABLED); + return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED); } -static struct quotactl_ops ocfs2_quotactl_ops = { - .quota_on = ocfs2_quota_on, +static const struct quotactl_ops ocfs2_quotactl_ops = { + .quota_on_meta = ocfs2_quota_on, .quota_off = ocfs2_quota_off, - .quota_sync = vfs_quota_sync, - .get_info = vfs_get_dqinfo, - .set_info = vfs_set_dqinfo, - .get_dqblk = vfs_get_dqblk, - .set_dqblk = vfs_set_dqblk, + .quota_sync = dquot_quota_sync, + .get_info = dquot_get_dqinfo, + .set_info = dquot_set_dqinfo, + .get_dqblk = dquot_get_dqblk, + .set_dqblk = dquot_set_dqblk, }; static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) @@ -974,9 +1027,10 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) struct inode *inode = NULL; struct ocfs2_super *osb = NULL; struct buffer_head *bh = NULL; - char nodestr[8]; + char nodestr[12]; + struct ocfs2_blockcheck_stats stats; - mlog_entry("%p, %p, %i", sb, data, silent); + trace_ocfs2_fill_super(sb, data, silent); if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) { status = -EINVAL; @@ -984,13 +1038,13 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) } /* probe for superblock */ - status = ocfs2_sb_probe(sb, &bh, §or_size); + status = ocfs2_sb_probe(sb, &bh, §or_size, &stats); if (status < 0) { mlog(ML_ERROR, "superblock probe failed!\n"); goto read_super_error; } - status = ocfs2_initialize_super(sb, bh, sector_size); + status = ocfs2_initialize_super(sb, bh, sector_size, &stats); osb = OCFS2_SB(sb); if (status < 0) { mlog_errno(status); @@ -999,31 +1053,22 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) brelse(bh); bh = NULL; - if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR)) - parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL; - + if (!ocfs2_check_set_options(sb, &parsed_options)) { + status = -EINVAL; + goto read_super_error; + } osb->s_mount_opt = parsed_options.mount_opt; osb->s_atime_quantum = parsed_options.atime_quantum; osb->preferred_slot = parsed_options.slot; osb->osb_commit_interval = parsed_options.commit_interval; - osb->local_alloc_default_bits = ocfs2_megabytes_to_clusters(sb, parsed_options.localalloc_opt); - osb->local_alloc_bits = osb->local_alloc_default_bits; - if (osb->s_mount_opt & OCFS2_MOUNT_USRQUOTA && - !OCFS2_HAS_RO_COMPAT_FEATURE(sb, - OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) { - status = -EINVAL; - mlog(ML_ERROR, "User quotas were requested, but this " - "filesystem does not have the feature enabled.\n"); - goto read_super_error; - } - if (osb->s_mount_opt & OCFS2_MOUNT_GRPQUOTA && - !OCFS2_HAS_RO_COMPAT_FEATURE(sb, - OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) { - status = -EINVAL; - mlog(ML_ERROR, "Group quotas were requested, but this " - "filesystem does not have the feature enabled.\n"); - goto read_super_error; - } + + ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt); + osb->osb_resv_level = parsed_options.resv_level; + osb->osb_dir_resv_level = parsed_options.resv_level; + if (parsed_options.dir_resv_level == -1) + osb->osb_dir_resv_level = parsed_options.resv_level; + else + osb->osb_dir_resv_level = parsed_options.dir_resv_level; status = ocfs2_verify_userspace_stack(osb, &parsed_options); if (status) @@ -1031,7 +1076,7 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) sb->s_magic = OCFS2_SUPER_MAGIC; - sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | + sb->s_flags = (sb->s_flags & ~(MS_POSIXACL | MS_NOSEC)) | ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0); /* Hard readonly mode only if: bdev_read_only, MS_RDONLY, @@ -1060,15 +1105,15 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) "file system, but write access is " "unavailable.\n"); else - mlog_errno(status); + mlog_errno(status); goto read_super_error; } ocfs2_set_ro_flag(osb, 1); - printk(KERN_NOTICE "Readonly device detected. No cluster " - "services will be utilized for this mount. Recovery " - "will be skipped.\n"); + printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. " + "Cluster services will not be used for this mount. " + "Recovery will be skipped.\n", osb->dev_str); } if (!ocfs2_is_hard_readonly(osb)) { @@ -1100,20 +1145,32 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) goto read_super_error; } - status = ocfs2_mount_volume(sb); - if (osb->root_inode) - inode = igrab(osb->root_inode); + if (ocfs2_meta_ecc(osb)) { + status = ocfs2_blockcheck_stats_debugfs_install( + &osb->osb_ecc_stats, + osb->osb_debug_root); + if (status) { + mlog(ML_ERROR, + "Unable to create blockcheck statistics " + "files\n"); + goto read_super_error; + } + } + status = ocfs2_mount_volume(sb); if (status < 0) goto read_super_error; + if (osb->root_inode) + inode = igrab(osb->root_inode); + if (!inode) { status = -EIO; mlog_errno(status); goto read_super_error; } - root = d_alloc_root(inode); + root = d_make_root(inode); if (!root) { status = -ENOMEM; mlog_errno(status); @@ -1149,7 +1206,6 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) mlog_errno(status); atomic_set(&osb->vol_state, VOLUME_DISABLED); wake_up(&osb->osb_mount_event); - mlog_exit(status); return status; } } @@ -1160,64 +1216,96 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS); wake_up(&osb->osb_mount_event); - mlog_exit(status); + /* Start this when the mount is almost sure of being successful */ + ocfs2_orphan_scan_start(osb); + return status; read_super_error: brelse(bh); - if (inode) - iput(inode); - if (osb) { atomic_set(&osb->vol_state, VOLUME_DISABLED); wake_up(&osb->osb_mount_event); ocfs2_dismount_volume(sb, 1); } - mlog_exit(status); + if (status) + mlog_errno(status); return status; } -static int ocfs2_get_sb(struct file_system_type *fs_type, +static struct dentry *ocfs2_mount(struct file_system_type *fs_type, int flags, const char *dev_name, - void *data, - struct vfsmount *mnt) + void *data) { - return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super, - mnt); + return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super); } static struct file_system_type ocfs2_fs_type = { .owner = THIS_MODULE, .name = "ocfs2", - .get_sb = ocfs2_get_sb, /* is this called when we mount - * the fs? */ - .kill_sb = kill_block_super, /* set to the generic one - * right now, but do we - * need to change that? */ + .mount = ocfs2_mount, + .kill_sb = kill_block_super, .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE, .next = NULL }; +MODULE_ALIAS_FS("ocfs2"); + +static int ocfs2_check_set_options(struct super_block *sb, + struct mount_options *options) +{ + if (options->mount_opt & OCFS2_MOUNT_USRQUOTA && + !OCFS2_HAS_RO_COMPAT_FEATURE(sb, + OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) { + mlog(ML_ERROR, "User quotas were requested, but this " + "filesystem does not have the feature enabled.\n"); + return 0; + } + if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA && + !OCFS2_HAS_RO_COMPAT_FEATURE(sb, + OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) { + mlog(ML_ERROR, "Group quotas were requested, but this " + "filesystem does not have the feature enabled.\n"); + return 0; + } + if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL && + !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) { + mlog(ML_ERROR, "ACL support requested but extended attributes " + "feature is not enabled\n"); + return 0; + } + /* No ACL setting specified? Use XATTR feature... */ + if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL | + OCFS2_MOUNT_NO_POSIX_ACL))) { + if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) + options->mount_opt |= OCFS2_MOUNT_POSIX_ACL; + else + options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL; + } + return 1; +} static int ocfs2_parse_options(struct super_block *sb, char *options, struct mount_options *mopt, int is_remount) { - int status; + int status, user_stack = 0; char *p; + u32 tmp; - mlog_entry("remount: %d, options: \"%s\"\n", is_remount, - options ? options : "(none)"); + trace_ocfs2_parse_options(is_remount, options ? options : "(none)"); mopt->commit_interval = 0; - mopt->mount_opt = 0; + mopt->mount_opt = OCFS2_MOUNT_NOINTR; mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; mopt->slot = OCFS2_INVALID_SLOT; - mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE; + mopt->localalloc_opt = -1; mopt->cluster_stack[0] = '\0'; + mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL; + mopt->dir_resv_level = -1; if (!options) { status = 1; @@ -1237,7 +1325,10 @@ static int ocfs2_parse_options(struct super_block *sb, mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL; break; case Opt_hb_none: - mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL; + mopt->mount_opt |= OCFS2_MOUNT_HB_NONE; + break; + case Opt_hb_global: + mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL; break; case Opt_barrier: if (match_int(&args[0], &option)) { @@ -1308,7 +1399,7 @@ static int ocfs2_parse_options(struct super_block *sb, status = 0; goto bail; } - if (option >= 0 && (option <= ocfs2_local_alloc_size(sb) * 8)) + if (option >= 0) mopt->localalloc_opt = option; break; case Opt_localflocks: @@ -1343,45 +1434,61 @@ static int ocfs2_parse_options(struct super_block *sb, memcpy(mopt->cluster_stack, args[0].from, OCFS2_STACK_LABEL_LEN); mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0'; + /* + * Open code the memcmp here as we don't have + * an osb to pass to + * ocfs2_userspace_stack(). + */ + if (memcmp(mopt->cluster_stack, + OCFS2_CLASSIC_CLUSTER_STACK, + OCFS2_STACK_LABEL_LEN)) + user_stack = 1; break; case Opt_inode64: mopt->mount_opt |= OCFS2_MOUNT_INODE64; break; case Opt_usrquota: - /* We check only on remount, otherwise features - * aren't yet initialized. */ - if (is_remount && !OCFS2_HAS_RO_COMPAT_FEATURE(sb, - OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) { - mlog(ML_ERROR, "User quota requested but " - "filesystem feature is not set\n"); - status = 0; - goto bail; - } mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA; break; case Opt_grpquota: - if (is_remount && !OCFS2_HAS_RO_COMPAT_FEATURE(sb, - OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) { - mlog(ML_ERROR, "Group quota requested but " - "filesystem feature is not set\n"); - status = 0; - goto bail; - } mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA; break; -#ifdef CONFIG_OCFS2_FS_POSIX_ACL + case Opt_coherency_buffered: + mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED; + break; + case Opt_coherency_full: + mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED; + break; case Opt_acl: mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL; + mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL; break; case Opt_noacl: + mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL; mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL; break; -#else - case Opt_acl: - case Opt_noacl: - printk(KERN_INFO "ocfs2 (no)acl options not supported\n"); + case Opt_resv_level: + if (is_remount) + break; + if (match_int(&args[0], &option)) { + status = 0; + goto bail; + } + if (option >= OCFS2_MIN_RESV_LEVEL && + option < OCFS2_MAX_RESV_LEVEL) + mopt->resv_level = option; + break; + case Opt_dir_resv_level: + if (is_remount) + break; + if (match_int(&args[0], &option)) { + status = 0; + goto bail; + } + if (option >= OCFS2_MIN_RESV_LEVEL && + option < OCFS2_MAX_RESV_LEVEL) + mopt->dir_resv_level = option; break; -#endif default: mlog(ML_ERROR, "Unrecognized mount option \"%s\" " @@ -1391,23 +1498,38 @@ static int ocfs2_parse_options(struct super_block *sb, } } + if (user_stack == 0) { + /* Ensure only one heartbeat mode */ + tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL | + OCFS2_MOUNT_HB_GLOBAL | + OCFS2_MOUNT_HB_NONE); + if (hweight32(tmp) != 1) { + mlog(ML_ERROR, "Invalid heartbeat mount options\n"); + status = 0; + goto bail; + } + } + status = 1; bail: - mlog_exit(status); return status; } -static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt) +static int ocfs2_show_options(struct seq_file *s, struct dentry *root) { - struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb); + struct ocfs2_super *osb = OCFS2_SB(root->d_sb); unsigned long opts = osb->s_mount_opt; unsigned int local_alloc_megs; - if (opts & OCFS2_MOUNT_HB_LOCAL) - seq_printf(s, ",_netdev,heartbeat=local"); - else - seq_printf(s, ",heartbeat=none"); + if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) { + seq_printf(s, ",_netdev"); + if (opts & OCFS2_MOUNT_HB_LOCAL) + seq_printf(s, ",%s", OCFS2_HB_LOCAL); + else + seq_printf(s, ",%s", OCFS2_HB_GLOBAL); + } else + seq_printf(s, ",%s", OCFS2_HB_NONE); if (opts & OCFS2_MOUNT_NOINTR) seq_printf(s, ",nointr"); @@ -1428,15 +1550,14 @@ static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt) if (osb->preferred_slot != OCFS2_INVALID_SLOT) seq_printf(s, ",preferred_slot=%d", osb->preferred_slot); - if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM) - seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum); + seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum); if (osb->osb_commit_interval) seq_printf(s, ",commit=%u", (unsigned) (osb->osb_commit_interval / HZ)); local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits); - if (local_alloc_megs != OCFS2_DEFAULT_LOCAL_ALLOC_SIZE) + if (local_alloc_megs != ocfs2_la_default_mb(osb)) seq_printf(s, ",localalloc=%d", local_alloc_megs); if (opts & OCFS2_MOUNT_LOCALFLOCKS) @@ -1450,6 +1571,11 @@ static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt) if (opts & OCFS2_MOUNT_GRPQUOTA) seq_printf(s, ",grpquota"); + if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED) + seq_printf(s, ",coherency=buffered"); + else + seq_printf(s, ",coherency=full"); + if (opts & OCFS2_MOUNT_NOUSERXATTR) seq_printf(s, ",nouser_xattr"); else @@ -1458,12 +1584,16 @@ static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt) if (opts & OCFS2_MOUNT_INODE64) seq_printf(s, ",inode64"); -#ifdef CONFIG_OCFS2_FS_POSIX_ACL if (opts & OCFS2_MOUNT_POSIX_ACL) seq_printf(s, ",acl"); else seq_printf(s, ",noacl"); -#endif + + if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL) + seq_printf(s, ",resv_level=%d", osb->osb_resv_level); + + if (osb->osb_dir_resv_level != osb->osb_resv_level) + seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level); return 0; } @@ -1472,26 +1602,18 @@ static int __init ocfs2_init(void) { int status; - mlog_entry_void(); - - ocfs2_print_version(); - status = init_ocfs2_uptodate_cache(); - if (status < 0) { - mlog_errno(status); - goto leave; - } + if (status < 0) + goto out1; status = ocfs2_initialize_mem_caches(); - if (status < 0) { - mlog_errno(status); - goto leave; - } + if (status < 0) + goto out2; ocfs2_wq = create_singlethread_workqueue("ocfs2_wq"); if (!ocfs2_wq) { status = -ENOMEM; - goto leave; + goto out3; } ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL); @@ -1500,34 +1622,30 @@ static int __init ocfs2_init(void) mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n"); } - status = ocfs2_quota_setup(); - if (status) - goto leave; - ocfs2_set_locking_protocol(); status = register_quota_format(&ocfs2_quota_format); -leave: - if (status < 0) { - ocfs2_quota_shutdown(); - ocfs2_free_mem_caches(); - exit_ocfs2_uptodate_cache(); - } - - mlog_exit(status); + if (status < 0) + goto out4; + status = register_filesystem(&ocfs2_fs_type); + if (!status) + return 0; - if (status >= 0) { - return register_filesystem(&ocfs2_fs_type); - } else - return -1; + unregister_quota_format(&ocfs2_quota_format); +out4: + destroy_workqueue(ocfs2_wq); + debugfs_remove(ocfs2_debugfs_root); +out3: + ocfs2_free_mem_caches(); +out2: + exit_ocfs2_uptodate_cache(); +out1: + mlog_errno(status); + return status; } static void __exit ocfs2_exit(void) { - mlog_entry_void(); - - ocfs2_quota_shutdown(); - if (ocfs2_wq) { flush_workqueue(ocfs2_wq); destroy_workqueue(ocfs2_wq); @@ -1542,18 +1660,14 @@ static void __exit ocfs2_exit(void) unregister_filesystem(&ocfs2_fs_type); exit_ocfs2_uptodate_cache(); - - mlog_exit_void(); } static void ocfs2_put_super(struct super_block *sb) { - mlog_entry("(0x%p)\n", sb); + trace_ocfs2_put_super(sb); ocfs2_sync_blockdev(sb); ocfs2_dismount_volume(sb, 0); - - mlog_exit_void(); } static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf) @@ -1565,7 +1679,7 @@ static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf) struct buffer_head *bh = NULL; struct inode *inode = NULL; - mlog_entry("(%p, %p)\n", dentry->d_sb, buf); + trace_ocfs2_statfs(dentry->d_sb, buf); osb = OCFS2_SB(dentry->d_sb); @@ -1599,6 +1713,10 @@ static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf) buf->f_bavail = buf->f_bfree; buf->f_files = numbits; buf->f_ffree = freebits; + buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN) + & 0xFFFFFFFFUL; + buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN, + OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL; brelse(bh); @@ -1608,7 +1726,8 @@ bail: if (inode) iput(inode); - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -1622,10 +1741,8 @@ static void ocfs2_inode_init_once(void *data) spin_lock_init(&oi->ip_lock); ocfs2_extent_map_init(&oi->vfs_inode); INIT_LIST_HEAD(&oi->ip_io_markers); - oi->ip_created_trans = 0; - oi->ip_last_trans = 0; oi->ip_dir_start_lookup = 0; - + mutex_init(&oi->ip_unaligned_aio); init_rwsem(&oi->ip_alloc_sem); init_rwsem(&oi->ip_xattr_sem); mutex_init(&oi->ip_io_mutex); @@ -1633,11 +1750,14 @@ static void ocfs2_inode_init_once(void *data) oi->ip_blkno = 0ULL; oi->ip_clusters = 0; + ocfs2_resv_init_once(&oi->ip_la_data_resv); + ocfs2_lock_res_init_once(&oi->ip_rw_lockres); ocfs2_lock_res_init_once(&oi->ip_inode_lockres); ocfs2_lock_res_init_once(&oi->ip_open_lockres); - ocfs2_metadata_cache_init(&oi->vfs_inode); + ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode), + &ocfs2_inode_caching_ops); inode_init_once(&oi->vfs_inode); } @@ -1677,6 +1797,11 @@ static int ocfs2_initialize_mem_caches(void) static void ocfs2_free_mem_caches(void) { + /* + * Make sure all delayed rcu free inodes are flushed before we + * destroy cache. + */ + rcu_barrier(); if (ocfs2_inode_cachep) kmem_cache_destroy(ocfs2_inode_cachep); ocfs2_inode_cachep = NULL; @@ -1702,8 +1827,8 @@ static int ocfs2_get_sector(struct super_block *sb, *bh = sb_getblk(sb, block); if (!*bh) { - mlog_errno(-EIO); - return -EIO; + mlog_errno(-ENOMEM); + return -ENOMEM; } lock_buffer(*bh); if (!buffer_dirty(*bh)) @@ -1727,8 +1852,6 @@ static int ocfs2_mount_volume(struct super_block *sb) int unlock_super = 0; struct ocfs2_super *osb = OCFS2_SB(sb); - mlog_entry_void(); - if (ocfs2_is_hard_readonly(osb)) goto leave; @@ -1766,19 +1889,13 @@ static int ocfs2_mount_volume(struct super_block *sb) } status = ocfs2_truncate_log_init(osb); - if (status < 0) { + if (status < 0) mlog_errno(status); - goto leave; - } - - if (ocfs2_mount_local(osb)) - goto leave; leave: if (unlock_super) ocfs2_super_unlock(osb, 1); - mlog_exit(status); return status; } @@ -1786,9 +1903,9 @@ static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err) { int tmp, hangup_needed = 0; struct ocfs2_super *osb = NULL; - char nodestr[8]; + char nodestr[12]; - mlog_entry("(0x%p)\n", sb); + trace_ocfs2_dismount_volume(sb); BUG_ON(!sb); osb = OCFS2_SB(sb); @@ -1796,8 +1913,16 @@ static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err) debugfs_remove(osb->osb_ctxt); + /* Orphan scan should be stopped as early as possible */ + ocfs2_orphan_scan_stop(osb); + ocfs2_disable_quotas(osb); + /* All dquots should be freed by now */ + WARN_ON(!llist_empty(&osb->dquot_drop_list)); + /* Wait for worker to be done with the work structure in osb */ + cancel_work_sync(&osb->dquot_drop_work); + ocfs2_shutdown_local_alloc(osb); ocfs2_truncate_log_shutdown(osb); @@ -1809,6 +1934,8 @@ static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err) ocfs2_sync_blockdev(sb); + ocfs2_purge_refcount_trees(osb); + /* No cluster connection means we've failed during mount, so skip * all the steps which depended on that to complete. */ if (osb->cconn) { @@ -1833,12 +1960,14 @@ static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err) * If we failed before we got a uuid_str yet, we can't stop * heartbeat. Otherwise, do it. */ - if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str) + if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str && + !ocfs2_is_hard_readonly(osb)) hangup_needed = 1; if (osb->cconn) ocfs2_dlm_shutdown(osb, hangup_needed); + ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats); debugfs_remove(osb->osb_debug_root); if (hangup_needed) @@ -1884,19 +2013,48 @@ static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uu return 0; } +/* Make sure entire volume is addressable by our journal. Requires + osb_clusters_at_boot to be valid and for the journal to have been + initialized by ocfs2_journal_init(). */ +static int ocfs2_journal_addressable(struct ocfs2_super *osb) +{ + int status = 0; + u64 max_block = + ocfs2_clusters_to_blocks(osb->sb, + osb->osb_clusters_at_boot) - 1; + + /* 32-bit block number is always OK. */ + if (max_block <= (u32)~0ULL) + goto out; + + /* Volume is "huge", so see if our journal is new enough to + support it. */ + if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb, + OCFS2_FEATURE_COMPAT_JBD2_SB) && + jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0, + JBD2_FEATURE_INCOMPAT_64BIT))) { + mlog(ML_ERROR, "The journal cannot address the entire volume. " + "Enable the 'block64' journal option with tunefs.ocfs2"); + status = -EFBIG; + goto out; + } + + out: + return status; +} + static int ocfs2_initialize_super(struct super_block *sb, struct buffer_head *bh, - int sector_size) + int sector_size, + struct ocfs2_blockcheck_stats *stats) { int status; int i, cbits, bbits; struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; struct inode *inode = NULL; struct ocfs2_journal *journal; - __le32 uuid_net_key; struct ocfs2_super *osb; - - mlog_entry_void(); + u64 total_blocks; osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL); if (!osb) { @@ -1907,6 +2065,7 @@ static int ocfs2_initialize_super(struct super_block *sb, sb->s_fs_info = osb; sb->s_op = &ocfs2_sops; + sb->s_d_op = &ocfs2_dentry_ops; sb->s_export_op = &ocfs2_export_ops; sb->s_qcop = &ocfs2_quotactl_ops; sb->dq_op = &ocfs2_quota_operations; @@ -1937,7 +2096,9 @@ static int ocfs2_initialize_super(struct super_block *sb, osb->blocked_lock_count = 0; spin_lock_init(&osb->osb_lock); spin_lock_init(&osb->osb_xattr_lock); - ocfs2_init_inode_steal_slot(osb); + ocfs2_init_steal_slots(osb); + + mutex_init(&osb->system_file_mutex); atomic_set(&osb->alloc_stats.moves, 0); atomic_set(&osb->alloc_stats.local_data, 0); @@ -1945,11 +2106,24 @@ static int ocfs2_initialize_super(struct super_block *sb, atomic_set(&osb->alloc_stats.bg_allocs, 0); atomic_set(&osb->alloc_stats.bg_extends, 0); + /* Copy the blockcheck stats from the superblock probe */ + osb->osb_ecc_stats = *stats; + ocfs2_init_node_maps(osb); snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u", MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev)); + osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots); + if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) { + mlog(ML_ERROR, "Invalid number of node slots (%u)\n", + osb->max_slots); + status = -EINVAL; + goto bail; + } + + ocfs2_orphan_scan_init(osb); + status = ocfs2_recovery_init(osb); if (status) { mlog(ML_ERROR, "Unable to initialize recovery state\n"); @@ -1958,7 +2132,6 @@ static int ocfs2_initialize_super(struct super_block *sb, } init_waitqueue_head(&osb->checkpoint_event); - atomic_set(&osb->needs_checkpoint, 0); osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; @@ -1973,6 +2146,12 @@ static int ocfs2_initialize_super(struct super_block *sb, init_waitqueue_head(&osb->osb_mount_event); + status = ocfs2_resmap_init(osb, &osb->osb_la_resmap); + if (status) { + mlog_errno(status); + goto bail; + } + osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL); if (!osb->vol_label) { mlog(ML_ERROR, "unable to alloc vol label\n"); @@ -1980,15 +2159,6 @@ static int ocfs2_initialize_super(struct super_block *sb, goto bail; } - osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots); - if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) { - mlog(ML_ERROR, "Invalid number of node slots (%u)\n", - osb->max_slots); - status = -EINVAL; - goto bail; - } - mlog(0, "max_slots for this device: %u\n", osb->max_slots); - osb->slot_recovery_generations = kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations), GFP_KERNEL); @@ -2008,6 +2178,8 @@ static int ocfs2_initialize_super(struct super_block *sb, goto bail; } + osb->osb_rf_lock_tree = RB_ROOT; + osb->s_feature_compat = le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat); osb->s_feature_ro_compat = @@ -2029,11 +2201,12 @@ static int ocfs2_initialize_super(struct super_block *sb, goto bail; } - if (ocfs2_userspace_stack(osb)) { - memcpy(osb->osb_cluster_stack, + if (ocfs2_clusterinfo_valid(osb)) { + osb->osb_stackflags = + OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags; + strlcpy(osb->osb_cluster_stack, OCFS2_RAW_SB(di)->s_cluster_info.ci_stack, - OCFS2_STACK_LABEL_LEN); - osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0'; + OCFS2_STACK_LABEL_LEN + 1); if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) { mlog(ML_ERROR, "couldn't mount because of an invalid " @@ -2042,6 +2215,9 @@ static int ocfs2_initialize_super(struct super_block *sb, status = -EINVAL; goto bail; } + strlcpy(osb->osb_cluster_name, + OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster, + OCFS2_CLUSTER_NAME_LEN + 1); } else { /* The empty string is identical with classic tools that * don't know about s_cluster_info. */ @@ -2077,14 +2253,13 @@ static int ocfs2_initialize_super(struct super_block *sb, INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery); journal->j_state = OCFS2_JOURNAL_FREE; - INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes); - osb->dentry_lock_list = NULL; + INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs); + init_llist_head(&osb->dquot_drop_list); /* get some pseudo constants for clustersize bits */ osb->s_clustersize_bits = le32_to_cpu(di->id2.i_super.s_clustersize_bits); osb->s_clustersize = 1 << osb->s_clustersize_bits; - mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits); if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE || osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) { @@ -2094,11 +2269,15 @@ static int ocfs2_initialize_super(struct super_block *sb, goto bail; } - if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1) - > (u32)~0UL) { - mlog(ML_ERROR, "Volume might try to write to blocks beyond " - "what jbd can address in 32 bits.\n"); - status = -EINVAL; + total_blocks = ocfs2_clusters_to_blocks(osb->sb, + le32_to_cpu(di->i_clusters)); + + status = generic_check_addressable(osb->sb->s_blocksize_bits, + total_blocks); + if (status) { + mlog(ML_ERROR, "Volume too large " + "to mount safely on this system"); + status = -EFBIG; goto bail; } @@ -2109,21 +2288,18 @@ static int ocfs2_initialize_super(struct super_block *sb, goto bail; } - memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key)); - - strncpy(osb->vol_label, di->id2.i_super.s_label, 63); - osb->vol_label[63] = '\0'; + strlcpy(osb->vol_label, di->id2.i_super.s_label, + OCFS2_MAX_VOL_LABEL_LEN); osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno); osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno); osb->first_cluster_group_blkno = le64_to_cpu(di->id2.i_super.s_first_cluster_group); osb->fs_generation = le32_to_cpu(di->i_fs_generation); osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash); - mlog(0, "vol_label: %s\n", osb->vol_label); - mlog(0, "uuid: %s\n", osb->uuid_str); - mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n", - (unsigned long long)osb->root_blkno, - (unsigned long long)osb->system_dir_blkno); + trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str, + (unsigned long long)osb->root_blkno, + (unsigned long long)osb->system_dir_blkno, + osb->s_clustersize_bits); osb->osb_dlm_debug = ocfs2_new_dlm_debug(); if (!osb->osb_dlm_debug) { @@ -2153,18 +2329,20 @@ static int ocfs2_initialize_super(struct super_block *sb, } osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno; + osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters; iput(inode); - osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8; + osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0, + osb->s_feature_incompat) * 8; status = ocfs2_init_slot_info(osb); if (status < 0) { mlog_errno(status); goto bail; } + cleancache_init_shared_fs((char *)&di->id2.i_super.s_uuid, sb); bail: - mlog_exit(status); return status; } @@ -2175,12 +2353,11 @@ bail: */ static int ocfs2_verify_volume(struct ocfs2_dinode *di, struct buffer_head *bh, - u32 blksz) + u32 blksz, + struct ocfs2_blockcheck_stats *stats) { int status = -EAGAIN; - mlog_entry_void(); - if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE, strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) { /* We have to do a raw check of the feature here */ @@ -2188,7 +2365,8 @@ static int ocfs2_verify_volume(struct ocfs2_dinode *di, OCFS2_FEATURE_INCOMPAT_META_ECC) { status = ocfs2_block_check_validate(bh->b_data, bh->b_size, - &di->i_check); + &di->i_check, + stats); if (status) goto out; } @@ -2234,7 +2412,8 @@ static int ocfs2_verify_volume(struct ocfs2_dinode *di, } out: - mlog_exit(status); + if (status && status != -EAGAIN) + mlog_errno(status); return status; } @@ -2247,8 +2426,6 @@ static int ocfs2_check_volume(struct ocfs2_super *osb) * recover * ourselves. */ - mlog_entry_void(); - /* Init our journal object. */ status = ocfs2_journal_init(osb->journal, &dirty); if (status < 0) { @@ -2256,6 +2433,12 @@ static int ocfs2_check_volume(struct ocfs2_super *osb) goto finally; } + /* Now that journal has been initialized, check to make sure + entire volume is addressable. */ + status = ocfs2_journal_addressable(osb); + if (status) + goto finally; + /* If the journal was unmounted cleanly then we don't want to * recover anything. Otherwise, journal_load will do that * dirty work for us :) */ @@ -2266,8 +2449,8 @@ static int ocfs2_check_volume(struct ocfs2_super *osb) goto finally; } } else { - mlog(ML_NOTICE, "File system was not unmounted cleanly, " - "recovering volume.\n"); + printk(KERN_NOTICE "ocfs2: File system on device (%s) was not " + "unmounted cleanly, recovering it.\n", osb->dev_str); } local = ocfs2_mount_local(osb); @@ -2292,8 +2475,6 @@ static int ocfs2_check_volume(struct ocfs2_super *osb) * ourselves as mounted. */ } - mlog(0, "Journal loaded.\n"); - status = ocfs2_load_local_alloc(osb); if (status < 0) { mlog_errno(status); @@ -2322,10 +2503,10 @@ static int ocfs2_check_volume(struct ocfs2_super *osb) mlog_errno(status); finally: - if (local_alloc) - kfree(local_alloc); + kfree(local_alloc); - mlog_exit(status); + if (status) + mlog_errno(status); return status; } @@ -2337,8 +2518,6 @@ finally: */ static void ocfs2_delete_osb(struct ocfs2_super *osb) { - mlog_entry_void(); - /* This function assumes that the caller has the main osb resource */ ocfs2_free_slot_info(osb); @@ -2347,17 +2526,14 @@ static void ocfs2_delete_osb(struct ocfs2_super *osb) kfree(osb->slot_recovery_generations); /* FIXME * This belongs in journal shutdown, but because we have to - * allocate osb->journal at the start of ocfs2_initalize_osb(), + * allocate osb->journal at the start of ocfs2_initialize_osb(), * we free it here. */ kfree(osb->journal); - if (osb->local_alloc_copy) - kfree(osb->local_alloc_copy); + kfree(osb->local_alloc_copy); kfree(osb->uuid_str); ocfs2_put_dlm_debug(osb->osb_dlm_debug); memset(osb, 0, sizeof(struct ocfs2_super)); - - mlog_exit_void(); } /* Put OCFS2 into a readonly state, or (if the user specifies it), @@ -2431,9 +2607,30 @@ void __ocfs2_abort(struct super_block* sb, /* Force a panic(). This stinks, but it's better than letting * things continue without having a proper hard readonly * here. */ - OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; + if (!ocfs2_mount_local(OCFS2_SB(sb))) + OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; ocfs2_handle_error(sb); } +/* + * Void signal blockers, because in-kernel sigprocmask() only fails + * when SIG_* is wrong. + */ +void ocfs2_block_signals(sigset_t *oldset) +{ + int rc; + sigset_t blocked; + + sigfillset(&blocked); + rc = sigprocmask(SIG_BLOCK, &blocked, oldset); + BUG_ON(rc); +} + +void ocfs2_unblock_signals(sigset_t *oldset) +{ + int rc = sigprocmask(SIG_SETMASK, oldset, NULL); + BUG_ON(rc); +} + module_init(ocfs2_init); module_exit(ocfs2_exit); diff --git a/fs/ocfs2/super.h b/fs/ocfs2/super.h index 783f5270f2a..74ff74cf78f 100644 --- a/fs/ocfs2/super.h +++ b/fs/ocfs2/super.h @@ -31,18 +31,23 @@ extern struct workqueue_struct *ocfs2_wq; int ocfs2_publish_get_mount_state(struct ocfs2_super *osb, int node_num); -void __ocfs2_error(struct super_block *sb, - const char *function, - const char *fmt, ...) - __attribute__ ((format (printf, 3, 4))); +__printf(3, 4) +void __ocfs2_error(struct super_block *sb, const char *function, + const char *fmt, ...); #define ocfs2_error(sb, fmt, args...) __ocfs2_error(sb, __PRETTY_FUNCTION__, fmt, ##args) -void __ocfs2_abort(struct super_block *sb, - const char *function, - const char *fmt, ...) - __attribute__ ((format (printf, 3, 4))); +__printf(3, 4) +void __ocfs2_abort(struct super_block *sb, const char *function, + const char *fmt, ...); #define ocfs2_abort(sb, fmt, args...) __ocfs2_abort(sb, __PRETTY_FUNCTION__, fmt, ##args) +/* + * Void signal blockers, because in-kernel sigprocmask() only fails + * when SIG_* is wrong. + */ +void ocfs2_block_signals(sigset_t *oldset); +void ocfs2_unblock_signals(sigset_t *oldset); + #endif /* OCFS2_SUPER_H */ diff --git a/fs/ocfs2/symlink.c b/fs/ocfs2/symlink.c index 579dd1b1110..66edce7ecfd 100644 --- a/fs/ocfs2/symlink.c +++ b/fs/ocfs2/symlink.c @@ -38,10 +38,8 @@ #include <linux/types.h> #include <linux/slab.h> #include <linux/pagemap.h> -#include <linux/utsname.h> #include <linux/namei.h> -#define MLOG_MASK_PREFIX ML_NAMEI #include <cluster/masklog.h> #include "ocfs2.h" @@ -56,106 +54,40 @@ #include "buffer_head_io.h" -static char *ocfs2_fast_symlink_getlink(struct inode *inode, - struct buffer_head **bh) +static int ocfs2_fast_symlink_readpage(struct file *unused, struct page *page) { - int status; - char *link = NULL; + struct inode *inode = page->mapping->host; + struct buffer_head *bh = NULL; + int status = ocfs2_read_inode_block(inode, &bh); struct ocfs2_dinode *fe; + const char *link; + void *kaddr; + size_t len; - mlog_entry_void(); - - status = ocfs2_read_inode_block(inode, bh); if (status < 0) { mlog_errno(status); - link = ERR_PTR(status); - goto bail; + return status; } - fe = (struct ocfs2_dinode *) (*bh)->b_data; + fe = (struct ocfs2_dinode *) bh->b_data; link = (char *) fe->id2.i_symlink; -bail: - mlog_exit(status); - - return link; -} - -static int ocfs2_readlink(struct dentry *dentry, - char __user *buffer, - int buflen) -{ - int ret; - char *link; - struct buffer_head *bh = NULL; - struct inode *inode = dentry->d_inode; - - mlog_entry_void(); - - link = ocfs2_fast_symlink_getlink(inode, &bh); - if (IS_ERR(link)) { - ret = PTR_ERR(link); - goto out; - } - - /* - * Without vfsmount we can't update atime now, - * but we will update atime here ultimately. - */ - ret = vfs_readlink(dentry, buffer, buflen, link); - - brelse(bh); -out: - mlog_exit(ret); - return ret; -} - -static void *ocfs2_fast_follow_link(struct dentry *dentry, - struct nameidata *nd) -{ - int status = 0; - int len; - char *target, *link = ERR_PTR(-ENOMEM); - struct inode *inode = dentry->d_inode; - struct buffer_head *bh = NULL; - - mlog_entry_void(); - - BUG_ON(!ocfs2_inode_is_fast_symlink(inode)); - target = ocfs2_fast_symlink_getlink(inode, &bh); - if (IS_ERR(target)) { - status = PTR_ERR(target); - mlog_errno(status); - goto bail; - } - - /* Fast symlinks can't be large */ - len = strlen(target); - link = kzalloc(len + 1, GFP_NOFS); - if (!link) { - status = -ENOMEM; - mlog_errno(status); - goto bail; - } - - memcpy(link, target, len); - nd_set_link(nd, link); - -bail: + /* will be less than a page size */ + len = strnlen(link, ocfs2_fast_symlink_chars(inode->i_sb)); + kaddr = kmap_atomic(page); + memcpy(kaddr, link, len + 1); + kunmap_atomic(kaddr); + SetPageUptodate(page); + unlock_page(page); brelse(bh); - - mlog_exit(status); - return status ? ERR_PTR(status) : link; + return 0; } -static void ocfs2_fast_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) -{ - char *link = cookie; - - kfree(link); -} +const struct address_space_operations ocfs2_fast_symlink_aops = { + .readpage = ocfs2_fast_symlink_readpage, +}; const struct inode_operations ocfs2_symlink_inode_operations = { - .readlink = page_readlink, + .readlink = generic_readlink, .follow_link = page_follow_link_light, .put_link = page_put_link, .getattr = ocfs2_getattr, @@ -164,15 +96,5 @@ const struct inode_operations ocfs2_symlink_inode_operations = { .getxattr = generic_getxattr, .listxattr = ocfs2_listxattr, .removexattr = generic_removexattr, -}; -const struct inode_operations ocfs2_fast_symlink_inode_operations = { - .readlink = ocfs2_readlink, - .follow_link = ocfs2_fast_follow_link, - .put_link = ocfs2_fast_put_link, - .getattr = ocfs2_getattr, - .setattr = ocfs2_setattr, - .setxattr = generic_setxattr, - .getxattr = generic_getxattr, - .listxattr = ocfs2_listxattr, - .removexattr = generic_removexattr, + .fiemap = ocfs2_fiemap, }; diff --git a/fs/ocfs2/symlink.h b/fs/ocfs2/symlink.h index 65a6c9c6ad5..71ee4245e91 100644 --- a/fs/ocfs2/symlink.h +++ b/fs/ocfs2/symlink.h @@ -27,7 +27,7 @@ #define OCFS2_SYMLINK_H extern const struct inode_operations ocfs2_symlink_inode_operations; -extern const struct inode_operations ocfs2_fast_symlink_inode_operations; +extern const struct address_space_operations ocfs2_fast_symlink_aops; /* * Test whether an inode is a fast symlink. diff --git a/fs/ocfs2/sysfile.c b/fs/ocfs2/sysfile.c index ab713ebdd54..af155c18312 100644 --- a/fs/ocfs2/sysfile.c +++ b/fs/ocfs2/sysfile.c @@ -25,10 +25,8 @@ #include <linux/fs.h> #include <linux/types.h> -#include <linux/slab.h> #include <linux/highmem.h> -#define MLOG_MASK_PREFIX ML_INODE #include <cluster/masklog.h> #include "ocfs2.h" @@ -45,10 +43,9 @@ static struct inode * _ocfs2_get_system_file_inode(struct ocfs2_super *osb, int type, u32 slot); -static inline int is_global_system_inode(int type); -static inline int is_in_system_inode_array(struct ocfs2_super *osb, - int type, - u32 slot); +#ifdef CONFIG_DEBUG_LOCK_ALLOC +static struct lock_class_key ocfs2_sysfile_cluster_lock_key[NUM_SYSTEM_INODES]; +#endif static inline int is_global_system_inode(int type) { @@ -56,11 +53,51 @@ static inline int is_global_system_inode(int type) type <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; } -static inline int is_in_system_inode_array(struct ocfs2_super *osb, - int type, - u32 slot) +static struct inode **get_local_system_inode(struct ocfs2_super *osb, + int type, + u32 slot) { - return slot == osb->slot_num || is_global_system_inode(type); + int index; + struct inode **local_system_inodes, **free = NULL; + + BUG_ON(slot == OCFS2_INVALID_SLOT); + BUG_ON(type < OCFS2_FIRST_LOCAL_SYSTEM_INODE || + type > OCFS2_LAST_LOCAL_SYSTEM_INODE); + + spin_lock(&osb->osb_lock); + local_system_inodes = osb->local_system_inodes; + spin_unlock(&osb->osb_lock); + + if (unlikely(!local_system_inodes)) { + local_system_inodes = kzalloc(sizeof(struct inode *) * + NUM_LOCAL_SYSTEM_INODES * + osb->max_slots, + GFP_NOFS); + if (!local_system_inodes) { + mlog_errno(-ENOMEM); + /* + * return NULL here so that ocfs2_get_sytem_file_inodes + * will try to create an inode and use it. We will try + * to initialize local_system_inodes next time. + */ + return NULL; + } + + spin_lock(&osb->osb_lock); + if (osb->local_system_inodes) { + /* Someone has initialized it for us. */ + free = local_system_inodes; + local_system_inodes = osb->local_system_inodes; + } else + osb->local_system_inodes = local_system_inodes; + spin_unlock(&osb->osb_lock); + kfree(free); + } + + index = (slot * NUM_LOCAL_SYSTEM_INODES) + + (type - OCFS2_FIRST_LOCAL_SYSTEM_INODE); + + return &local_system_inodes[index]; } struct inode *ocfs2_get_system_file_inode(struct ocfs2_super *osb, @@ -71,12 +108,16 @@ struct inode *ocfs2_get_system_file_inode(struct ocfs2_super *osb, struct inode **arr = NULL; /* avoid the lookup if cached in local system file array */ - if (is_in_system_inode_array(osb, type, slot)) - arr = &(osb->system_inodes[type]); + if (is_global_system_inode(type)) { + arr = &(osb->global_system_inodes[type]); + } else + arr = get_local_system_inode(osb, type, slot); + mutex_lock(&osb->system_file_mutex); if (arr && ((inode = *arr) != NULL)) { /* get a ref in addition to the array ref */ inode = igrab(inode); + mutex_unlock(&osb->system_file_mutex); BUG_ON(!inode); return inode; @@ -90,6 +131,7 @@ struct inode *ocfs2_get_system_file_inode(struct ocfs2_super *osb, *arr = igrab(inode); BUG_ON(!*arr); } + mutex_unlock(&osb->system_file_mutex); return inode; } @@ -118,6 +160,21 @@ static struct inode * _ocfs2_get_system_file_inode(struct ocfs2_super *osb, inode = NULL; goto bail; } +#ifdef CONFIG_DEBUG_LOCK_ALLOC + if (type == LOCAL_USER_QUOTA_SYSTEM_INODE || + type == LOCAL_GROUP_QUOTA_SYSTEM_INODE || + type == JOURNAL_SYSTEM_INODE) { + /* Ignore inode lock on these inodes as the lock does not + * really belong to any process and lockdep cannot handle + * that */ + OCFS2_I(inode)->ip_inode_lockres.l_lockdep_map.key = NULL; + } else { + lockdep_init_map(&OCFS2_I(inode)->ip_inode_lockres. + l_lockdep_map, + ocfs2_system_inodes[type].si_name, + &ocfs2_sysfile_cluster_lock_key[type], 0); + } +#endif bail: return inode; diff --git a/fs/ocfs2/uptodate.c b/fs/ocfs2/uptodate.c index 187b99ff036..82e17b076ce 100644 --- a/fs/ocfs2/uptodate.c +++ b/fs/ocfs2/uptodate.c @@ -53,13 +53,6 @@ #include <linux/highmem.h> #include <linux/buffer_head.h> #include <linux/rbtree.h> -#ifndef CONFIG_OCFS2_COMPAT_JBD -# include <linux/jbd2.h> -#else -# include <linux/jbd.h> -#endif - -#define MLOG_MASK_PREFIX ML_UPTODATE #include <cluster/masklog.h> @@ -67,23 +60,86 @@ #include "inode.h" #include "uptodate.h" +#include "ocfs2_trace.h" struct ocfs2_meta_cache_item { struct rb_node c_node; sector_t c_block; }; -static struct kmem_cache *ocfs2_uptodate_cachep = NULL; +static struct kmem_cache *ocfs2_uptodate_cachep; + +u64 ocfs2_metadata_cache_owner(struct ocfs2_caching_info *ci) +{ + BUG_ON(!ci || !ci->ci_ops); + + return ci->ci_ops->co_owner(ci); +} + +struct super_block *ocfs2_metadata_cache_get_super(struct ocfs2_caching_info *ci) +{ + BUG_ON(!ci || !ci->ci_ops); + + return ci->ci_ops->co_get_super(ci); +} + +static void ocfs2_metadata_cache_lock(struct ocfs2_caching_info *ci) +{ + BUG_ON(!ci || !ci->ci_ops); + + ci->ci_ops->co_cache_lock(ci); +} + +static void ocfs2_metadata_cache_unlock(struct ocfs2_caching_info *ci) +{ + BUG_ON(!ci || !ci->ci_ops); + + ci->ci_ops->co_cache_unlock(ci); +} + +void ocfs2_metadata_cache_io_lock(struct ocfs2_caching_info *ci) +{ + BUG_ON(!ci || !ci->ci_ops); + + ci->ci_ops->co_io_lock(ci); +} -void ocfs2_metadata_cache_init(struct inode *inode) +void ocfs2_metadata_cache_io_unlock(struct ocfs2_caching_info *ci) { - struct ocfs2_inode_info *oi = OCFS2_I(inode); - struct ocfs2_caching_info *ci = &oi->ip_metadata_cache; + BUG_ON(!ci || !ci->ci_ops); - oi->ip_flags |= OCFS2_INODE_CACHE_INLINE; + ci->ci_ops->co_io_unlock(ci); +} + + +static void ocfs2_metadata_cache_reset(struct ocfs2_caching_info *ci, + int clear) +{ + ci->ci_flags |= OCFS2_CACHE_FL_INLINE; ci->ci_num_cached = 0; + + if (clear) { + ci->ci_created_trans = 0; + ci->ci_last_trans = 0; + } } +void ocfs2_metadata_cache_init(struct ocfs2_caching_info *ci, + const struct ocfs2_caching_operations *ops) +{ + BUG_ON(!ops); + + ci->ci_ops = ops; + ocfs2_metadata_cache_reset(ci, 1); +} + +void ocfs2_metadata_cache_exit(struct ocfs2_caching_info *ci) +{ + ocfs2_metadata_cache_purge(ci); + ocfs2_metadata_cache_reset(ci, 1); +} + + /* No lock taken here as 'root' is not expected to be visible to other * processes. */ static unsigned int ocfs2_purge_copied_metadata_tree(struct rb_root *root) @@ -95,8 +151,8 @@ static unsigned int ocfs2_purge_copied_metadata_tree(struct rb_root *root) while ((node = rb_last(root)) != NULL) { item = rb_entry(node, struct ocfs2_meta_cache_item, c_node); - mlog(0, "Purge item %llu\n", - (unsigned long long) item->c_block); + trace_ocfs2_purge_copied_metadata_tree( + (unsigned long long) item->c_block); rb_erase(&item->c_node, root); kmem_cache_free(ocfs2_uptodate_cachep, item); @@ -112,19 +168,20 @@ static unsigned int ocfs2_purge_copied_metadata_tree(struct rb_root *root) * This function is a few more lines longer than necessary due to some * accounting done here, but I think it's worth tracking down those * bugs sooner -- Mark */ -void ocfs2_metadata_cache_purge(struct inode *inode) +void ocfs2_metadata_cache_purge(struct ocfs2_caching_info *ci) { - struct ocfs2_inode_info *oi = OCFS2_I(inode); unsigned int tree, to_purge, purged; - struct ocfs2_caching_info *ci = &oi->ip_metadata_cache; struct rb_root root = RB_ROOT; - spin_lock(&oi->ip_lock); - tree = !(oi->ip_flags & OCFS2_INODE_CACHE_INLINE); + BUG_ON(!ci || !ci->ci_ops); + + ocfs2_metadata_cache_lock(ci); + tree = !(ci->ci_flags & OCFS2_CACHE_FL_INLINE); to_purge = ci->ci_num_cached; - mlog(0, "Purge %u %s items from Inode %llu\n", to_purge, - tree ? "array" : "tree", (unsigned long long)oi->ip_blkno); + trace_ocfs2_metadata_cache_purge( + (unsigned long long)ocfs2_metadata_cache_owner(ci), + to_purge, tree); /* If we're a tree, save off the root so that we can safely * initialize the cache. We do the work to free tree members @@ -132,16 +189,17 @@ void ocfs2_metadata_cache_purge(struct inode *inode) if (tree) root = ci->ci_cache.ci_tree; - ocfs2_metadata_cache_init(inode); - spin_unlock(&oi->ip_lock); + ocfs2_metadata_cache_reset(ci, 0); + ocfs2_metadata_cache_unlock(ci); purged = ocfs2_purge_copied_metadata_tree(&root); /* If possible, track the number wiped so that we can more * easily detect counting errors. Unfortunately, this is only * meaningful for trees. */ if (tree && purged != to_purge) - mlog(ML_ERROR, "Inode %llu, count = %u, purged = %u\n", - (unsigned long long)oi->ip_blkno, to_purge, purged); + mlog(ML_ERROR, "Owner %llu, count = %u, purged = %u\n", + (unsigned long long)ocfs2_metadata_cache_owner(ci), + to_purge, purged); } /* Returns the index in the cache array, -1 if not found. @@ -182,39 +240,37 @@ ocfs2_search_cache_tree(struct ocfs2_caching_info *ci, return NULL; } -static int ocfs2_buffer_cached(struct ocfs2_inode_info *oi, +static int ocfs2_buffer_cached(struct ocfs2_caching_info *ci, struct buffer_head *bh) { int index = -1; struct ocfs2_meta_cache_item *item = NULL; - spin_lock(&oi->ip_lock); + ocfs2_metadata_cache_lock(ci); - mlog(0, "Inode %llu, query block %llu (inline = %u)\n", - (unsigned long long)oi->ip_blkno, - (unsigned long long) bh->b_blocknr, - !!(oi->ip_flags & OCFS2_INODE_CACHE_INLINE)); + trace_ocfs2_buffer_cached_begin( + (unsigned long long)ocfs2_metadata_cache_owner(ci), + (unsigned long long) bh->b_blocknr, + !!(ci->ci_flags & OCFS2_CACHE_FL_INLINE)); - if (oi->ip_flags & OCFS2_INODE_CACHE_INLINE) - index = ocfs2_search_cache_array(&oi->ip_metadata_cache, - bh->b_blocknr); + if (ci->ci_flags & OCFS2_CACHE_FL_INLINE) + index = ocfs2_search_cache_array(ci, bh->b_blocknr); else - item = ocfs2_search_cache_tree(&oi->ip_metadata_cache, - bh->b_blocknr); + item = ocfs2_search_cache_tree(ci, bh->b_blocknr); - spin_unlock(&oi->ip_lock); + ocfs2_metadata_cache_unlock(ci); - mlog(0, "index = %d, item = %p\n", index, item); + trace_ocfs2_buffer_cached_end(index, item); return (index != -1) || (item != NULL); } /* Warning: even if it returns true, this does *not* guarantee that - * the block is stored in our inode metadata cache. - * + * the block is stored in our inode metadata cache. + * * This can be called under lock_buffer() */ -int ocfs2_buffer_uptodate(struct inode *inode, +int ocfs2_buffer_uptodate(struct ocfs2_caching_info *ci, struct buffer_head *bh) { /* Doesn't matter if the bh is in our cache or not -- if it's @@ -230,27 +286,28 @@ int ocfs2_buffer_uptodate(struct inode *inode, /* Ok, locally the buffer is marked as up to date, now search * our cache to see if we can trust that. */ - return ocfs2_buffer_cached(OCFS2_I(inode), bh); + return ocfs2_buffer_cached(ci, bh); } -/* +/* * Determine whether a buffer is currently out on a read-ahead request. - * ip_io_sem should be held to serialize submitters with the logic here. + * ci_io_sem should be held to serialize submitters with the logic here. */ -int ocfs2_buffer_read_ahead(struct inode *inode, +int ocfs2_buffer_read_ahead(struct ocfs2_caching_info *ci, struct buffer_head *bh) { - return buffer_locked(bh) && ocfs2_buffer_cached(OCFS2_I(inode), bh); + return buffer_locked(bh) && ocfs2_buffer_cached(ci, bh); } /* Requires ip_lock */ static void ocfs2_append_cache_array(struct ocfs2_caching_info *ci, sector_t block) { - BUG_ON(ci->ci_num_cached >= OCFS2_INODE_MAX_CACHE_ARRAY); + BUG_ON(ci->ci_num_cached >= OCFS2_CACHE_INFO_MAX_ARRAY); - mlog(0, "block %llu takes position %u\n", (unsigned long long) block, - ci->ci_num_cached); + trace_ocfs2_append_cache_array( + (unsigned long long)ocfs2_metadata_cache_owner(ci), + (unsigned long long)block, ci->ci_num_cached); ci->ci_cache.ci_array[ci->ci_num_cached] = block; ci->ci_num_cached++; @@ -267,8 +324,9 @@ static void __ocfs2_insert_cache_tree(struct ocfs2_caching_info *ci, struct rb_node **p = &ci->ci_cache.ci_tree.rb_node; struct ocfs2_meta_cache_item *tmp; - mlog(0, "Insert block %llu num = %u\n", (unsigned long long) block, - ci->ci_num_cached); + trace_ocfs2_insert_cache_tree( + (unsigned long long)ocfs2_metadata_cache_owner(ci), + (unsigned long long)block, ci->ci_num_cached); while(*p) { parent = *p; @@ -292,67 +350,65 @@ static void __ocfs2_insert_cache_tree(struct ocfs2_caching_info *ci, ci->ci_num_cached++; } -static inline int ocfs2_insert_can_use_array(struct ocfs2_inode_info *oi, - struct ocfs2_caching_info *ci) +/* co_cache_lock() must be held */ +static inline int ocfs2_insert_can_use_array(struct ocfs2_caching_info *ci) { - assert_spin_locked(&oi->ip_lock); - - return (oi->ip_flags & OCFS2_INODE_CACHE_INLINE) && - (ci->ci_num_cached < OCFS2_INODE_MAX_CACHE_ARRAY); + return (ci->ci_flags & OCFS2_CACHE_FL_INLINE) && + (ci->ci_num_cached < OCFS2_CACHE_INFO_MAX_ARRAY); } -/* tree should be exactly OCFS2_INODE_MAX_CACHE_ARRAY wide. NULL the +/* tree should be exactly OCFS2_CACHE_INFO_MAX_ARRAY wide. NULL the * pointers in tree after we use them - this allows caller to detect - * when to free in case of error. */ -static void ocfs2_expand_cache(struct ocfs2_inode_info *oi, + * when to free in case of error. + * + * The co_cache_lock() must be held. */ +static void ocfs2_expand_cache(struct ocfs2_caching_info *ci, struct ocfs2_meta_cache_item **tree) { int i; - struct ocfs2_caching_info *ci = &oi->ip_metadata_cache; - mlog_bug_on_msg(ci->ci_num_cached != OCFS2_INODE_MAX_CACHE_ARRAY, - "Inode %llu, num cached = %u, should be %u\n", - (unsigned long long)oi->ip_blkno, ci->ci_num_cached, - OCFS2_INODE_MAX_CACHE_ARRAY); - mlog_bug_on_msg(!(oi->ip_flags & OCFS2_INODE_CACHE_INLINE), - "Inode %llu not marked as inline anymore!\n", - (unsigned long long)oi->ip_blkno); - assert_spin_locked(&oi->ip_lock); + mlog_bug_on_msg(ci->ci_num_cached != OCFS2_CACHE_INFO_MAX_ARRAY, + "Owner %llu, num cached = %u, should be %u\n", + (unsigned long long)ocfs2_metadata_cache_owner(ci), + ci->ci_num_cached, OCFS2_CACHE_INFO_MAX_ARRAY); + mlog_bug_on_msg(!(ci->ci_flags & OCFS2_CACHE_FL_INLINE), + "Owner %llu not marked as inline anymore!\n", + (unsigned long long)ocfs2_metadata_cache_owner(ci)); /* Be careful to initialize the tree members *first* because * once the ci_tree is used, the array is junk... */ - for(i = 0; i < OCFS2_INODE_MAX_CACHE_ARRAY; i++) + for (i = 0; i < OCFS2_CACHE_INFO_MAX_ARRAY; i++) tree[i]->c_block = ci->ci_cache.ci_array[i]; - oi->ip_flags &= ~OCFS2_INODE_CACHE_INLINE; + ci->ci_flags &= ~OCFS2_CACHE_FL_INLINE; ci->ci_cache.ci_tree = RB_ROOT; /* this will be set again by __ocfs2_insert_cache_tree */ ci->ci_num_cached = 0; - for(i = 0; i < OCFS2_INODE_MAX_CACHE_ARRAY; i++) { + for (i = 0; i < OCFS2_CACHE_INFO_MAX_ARRAY; i++) { __ocfs2_insert_cache_tree(ci, tree[i]); tree[i] = NULL; } - mlog(0, "Expanded %llu to a tree cache: flags 0x%x, num = %u\n", - (unsigned long long)oi->ip_blkno, oi->ip_flags, ci->ci_num_cached); + trace_ocfs2_expand_cache( + (unsigned long long)ocfs2_metadata_cache_owner(ci), + ci->ci_flags, ci->ci_num_cached); } /* Slow path function - memory allocation is necessary. See the * comment above ocfs2_set_buffer_uptodate for more information. */ -static void __ocfs2_set_buffer_uptodate(struct ocfs2_inode_info *oi, +static void __ocfs2_set_buffer_uptodate(struct ocfs2_caching_info *ci, sector_t block, int expand_tree) { int i; - struct ocfs2_caching_info *ci = &oi->ip_metadata_cache; struct ocfs2_meta_cache_item *new = NULL; - struct ocfs2_meta_cache_item *tree[OCFS2_INODE_MAX_CACHE_ARRAY] = + struct ocfs2_meta_cache_item *tree[OCFS2_CACHE_INFO_MAX_ARRAY] = { NULL, }; - mlog(0, "Inode %llu, block %llu, expand = %d\n", - (unsigned long long)oi->ip_blkno, - (unsigned long long)block, expand_tree); + trace_ocfs2_set_buffer_uptodate( + (unsigned long long)ocfs2_metadata_cache_owner(ci), + (unsigned long long)block, expand_tree); new = kmem_cache_alloc(ocfs2_uptodate_cachep, GFP_NOFS); if (!new) { @@ -364,7 +420,7 @@ static void __ocfs2_set_buffer_uptodate(struct ocfs2_inode_info *oi, if (expand_tree) { /* Do *not* allocate an array here - the removal code * has no way of tracking that. */ - for(i = 0; i < OCFS2_INODE_MAX_CACHE_ARRAY; i++) { + for (i = 0; i < OCFS2_CACHE_INFO_MAX_ARRAY; i++) { tree[i] = kmem_cache_alloc(ocfs2_uptodate_cachep, GFP_NOFS); if (!tree[i]) { @@ -376,21 +432,20 @@ static void __ocfs2_set_buffer_uptodate(struct ocfs2_inode_info *oi, } } - spin_lock(&oi->ip_lock); - if (ocfs2_insert_can_use_array(oi, ci)) { - mlog(0, "Someone cleared the tree underneath us\n"); + ocfs2_metadata_cache_lock(ci); + if (ocfs2_insert_can_use_array(ci)) { /* Ok, items were removed from the cache in between * locks. Detect this and revert back to the fast path */ ocfs2_append_cache_array(ci, block); - spin_unlock(&oi->ip_lock); + ocfs2_metadata_cache_unlock(ci); goto out_free; } if (expand_tree) - ocfs2_expand_cache(oi, tree); + ocfs2_expand_cache(ci, tree); __ocfs2_insert_cache_tree(ci, new); - spin_unlock(&oi->ip_lock); + ocfs2_metadata_cache_unlock(ci); new = NULL; out_free: @@ -400,14 +455,14 @@ out_free: /* If these were used, then ocfs2_expand_cache re-set them to * NULL for us. */ if (tree[0]) { - for(i = 0; i < OCFS2_INODE_MAX_CACHE_ARRAY; i++) + for (i = 0; i < OCFS2_CACHE_INFO_MAX_ARRAY; i++) if (tree[i]) kmem_cache_free(ocfs2_uptodate_cachep, tree[i]); } } -/* Item insertion is guarded by ip_io_mutex, so the insertion path takes +/* Item insertion is guarded by co_io_lock(), so the insertion path takes * advantage of this by not rechecking for a duplicate insert during * the slow case. Additionally, if the cache needs to be bumped up to * a tree, the code will not recheck after acquiring the lock -- @@ -425,59 +480,55 @@ out_free: * Readahead buffers can be passed in here before the I/O request is * completed. */ -void ocfs2_set_buffer_uptodate(struct inode *inode, +void ocfs2_set_buffer_uptodate(struct ocfs2_caching_info *ci, struct buffer_head *bh) { int expand; - struct ocfs2_inode_info *oi = OCFS2_I(inode); - struct ocfs2_caching_info *ci = &oi->ip_metadata_cache; /* The block may very well exist in our cache already, so avoid * doing any more work in that case. */ - if (ocfs2_buffer_cached(oi, bh)) + if (ocfs2_buffer_cached(ci, bh)) return; - mlog(0, "Inode %llu, inserting block %llu\n", - (unsigned long long)oi->ip_blkno, - (unsigned long long)bh->b_blocknr); + trace_ocfs2_set_buffer_uptodate_begin( + (unsigned long long)ocfs2_metadata_cache_owner(ci), + (unsigned long long)bh->b_blocknr); /* No need to recheck under spinlock - insertion is guarded by - * ip_io_mutex */ - spin_lock(&oi->ip_lock); - if (ocfs2_insert_can_use_array(oi, ci)) { + * co_io_lock() */ + ocfs2_metadata_cache_lock(ci); + if (ocfs2_insert_can_use_array(ci)) { /* Fast case - it's an array and there's a free * spot. */ ocfs2_append_cache_array(ci, bh->b_blocknr); - spin_unlock(&oi->ip_lock); + ocfs2_metadata_cache_unlock(ci); return; } expand = 0; - if (oi->ip_flags & OCFS2_INODE_CACHE_INLINE) { + if (ci->ci_flags & OCFS2_CACHE_FL_INLINE) { /* We need to bump things up to a tree. */ expand = 1; } - spin_unlock(&oi->ip_lock); + ocfs2_metadata_cache_unlock(ci); - __ocfs2_set_buffer_uptodate(oi, bh->b_blocknr, expand); + __ocfs2_set_buffer_uptodate(ci, bh->b_blocknr, expand); } /* Called against a newly allocated buffer. Most likely nobody should * be able to read this sort of metadata while it's still being - * allocated, but this is careful to take ip_io_mutex anyway. */ -void ocfs2_set_new_buffer_uptodate(struct inode *inode, + * allocated, but this is careful to take co_io_lock() anyway. */ +void ocfs2_set_new_buffer_uptodate(struct ocfs2_caching_info *ci, struct buffer_head *bh) { - struct ocfs2_inode_info *oi = OCFS2_I(inode); - /* This should definitely *not* exist in our cache */ - BUG_ON(ocfs2_buffer_cached(oi, bh)); + BUG_ON(ocfs2_buffer_cached(ci, bh)); set_buffer_uptodate(bh); - mutex_lock(&oi->ip_io_mutex); - ocfs2_set_buffer_uptodate(inode, bh); - mutex_unlock(&oi->ip_io_mutex); + ocfs2_metadata_cache_io_lock(ci); + ocfs2_set_buffer_uptodate(ci, bh); + ocfs2_metadata_cache_io_unlock(ci); } /* Requires ip_lock. */ @@ -487,12 +538,13 @@ static void ocfs2_remove_metadata_array(struct ocfs2_caching_info *ci, sector_t *array = ci->ci_cache.ci_array; int bytes; - BUG_ON(index < 0 || index >= OCFS2_INODE_MAX_CACHE_ARRAY); + BUG_ON(index < 0 || index >= OCFS2_CACHE_INFO_MAX_ARRAY); BUG_ON(index >= ci->ci_num_cached); BUG_ON(!ci->ci_num_cached); - mlog(0, "remove index %d (num_cached = %u\n", index, - ci->ci_num_cached); + trace_ocfs2_remove_metadata_array( + (unsigned long long)ocfs2_metadata_cache_owner(ci), + index, ci->ci_num_cached); ci->ci_num_cached--; @@ -508,28 +560,27 @@ static void ocfs2_remove_metadata_array(struct ocfs2_caching_info *ci, static void ocfs2_remove_metadata_tree(struct ocfs2_caching_info *ci, struct ocfs2_meta_cache_item *item) { - mlog(0, "remove block %llu from tree\n", - (unsigned long long) item->c_block); + trace_ocfs2_remove_metadata_tree( + (unsigned long long)ocfs2_metadata_cache_owner(ci), + (unsigned long long)item->c_block); rb_erase(&item->c_node, &ci->ci_cache.ci_tree); ci->ci_num_cached--; } -static void ocfs2_remove_block_from_cache(struct inode *inode, +static void ocfs2_remove_block_from_cache(struct ocfs2_caching_info *ci, sector_t block) { int index; struct ocfs2_meta_cache_item *item = NULL; - struct ocfs2_inode_info *oi = OCFS2_I(inode); - struct ocfs2_caching_info *ci = &oi->ip_metadata_cache; - spin_lock(&oi->ip_lock); - mlog(0, "Inode %llu, remove %llu, items = %u, array = %u\n", - (unsigned long long)oi->ip_blkno, - (unsigned long long) block, ci->ci_num_cached, - oi->ip_flags & OCFS2_INODE_CACHE_INLINE); + ocfs2_metadata_cache_lock(ci); + trace_ocfs2_remove_block_from_cache( + (unsigned long long)ocfs2_metadata_cache_owner(ci), + (unsigned long long) block, ci->ci_num_cached, + ci->ci_flags); - if (oi->ip_flags & OCFS2_INODE_CACHE_INLINE) { + if (ci->ci_flags & OCFS2_CACHE_FL_INLINE) { index = ocfs2_search_cache_array(ci, block); if (index != -1) ocfs2_remove_metadata_array(ci, index); @@ -538,7 +589,7 @@ static void ocfs2_remove_block_from_cache(struct inode *inode, if (item) ocfs2_remove_metadata_tree(ci, item); } - spin_unlock(&oi->ip_lock); + ocfs2_metadata_cache_unlock(ci); if (item) kmem_cache_free(ocfs2_uptodate_cachep, item); @@ -549,23 +600,24 @@ static void ocfs2_remove_block_from_cache(struct inode *inode, * bother reverting things to an inlined array in the case of a remove * which moves us back under the limit. */ -void ocfs2_remove_from_cache(struct inode *inode, +void ocfs2_remove_from_cache(struct ocfs2_caching_info *ci, struct buffer_head *bh) { sector_t block = bh->b_blocknr; - ocfs2_remove_block_from_cache(inode, block); + ocfs2_remove_block_from_cache(ci, block); } /* Called when we remove xattr clusters from an inode. */ -void ocfs2_remove_xattr_clusters_from_cache(struct inode *inode, +void ocfs2_remove_xattr_clusters_from_cache(struct ocfs2_caching_info *ci, sector_t block, u32 c_len) { - unsigned int i, b_len = ocfs2_clusters_to_blocks(inode->i_sb, 1) * c_len; + struct super_block *sb = ocfs2_metadata_cache_get_super(ci); + unsigned int i, b_len = ocfs2_clusters_to_blocks(sb, 1) * c_len; for (i = 0; i < b_len; i++, block++) - ocfs2_remove_block_from_cache(inode, block); + ocfs2_remove_block_from_cache(ci, block); } int __init init_ocfs2_uptodate_cache(void) @@ -576,9 +628,6 @@ int __init init_ocfs2_uptodate_cache(void) if (!ocfs2_uptodate_cachep) return -ENOMEM; - mlog(0, "%u inlined cache items per inode.\n", - OCFS2_INODE_MAX_CACHE_ARRAY); - return 0; } diff --git a/fs/ocfs2/uptodate.h b/fs/ocfs2/uptodate.h index 531b4b3a0c4..0d826fe2da0 100644 --- a/fs/ocfs2/uptodate.h +++ b/fs/ocfs2/uptodate.h @@ -26,24 +26,59 @@ #ifndef OCFS2_UPTODATE_H #define OCFS2_UPTODATE_H +/* + * The caching code relies on locking provided by the user of + * struct ocfs2_caching_info. These operations connect that up. + */ +struct ocfs2_caching_operations { + /* + * A u64 representing the owning structure. Usually this + * is the block number (i_blkno or whatnot). This is used so + * that caching log messages can identify the owning structure. + */ + u64 (*co_owner)(struct ocfs2_caching_info *ci); + + /* The superblock is needed during I/O. */ + struct super_block *(*co_get_super)(struct ocfs2_caching_info *ci); + /* + * Lock and unlock the caching data. These will not sleep, and + * should probably be spinlocks. + */ + void (*co_cache_lock)(struct ocfs2_caching_info *ci); + void (*co_cache_unlock)(struct ocfs2_caching_info *ci); + + /* + * Lock and unlock for disk I/O. These will sleep, and should + * be mutexes. + */ + void (*co_io_lock)(struct ocfs2_caching_info *ci); + void (*co_io_unlock)(struct ocfs2_caching_info *ci); +}; + int __init init_ocfs2_uptodate_cache(void); void exit_ocfs2_uptodate_cache(void); -void ocfs2_metadata_cache_init(struct inode *inode); -void ocfs2_metadata_cache_purge(struct inode *inode); +void ocfs2_metadata_cache_init(struct ocfs2_caching_info *ci, + const struct ocfs2_caching_operations *ops); +void ocfs2_metadata_cache_purge(struct ocfs2_caching_info *ci); +void ocfs2_metadata_cache_exit(struct ocfs2_caching_info *ci); + +u64 ocfs2_metadata_cache_owner(struct ocfs2_caching_info *ci); +void ocfs2_metadata_cache_io_lock(struct ocfs2_caching_info *ci); +void ocfs2_metadata_cache_io_unlock(struct ocfs2_caching_info *ci); -int ocfs2_buffer_uptodate(struct inode *inode, +int ocfs2_buffer_uptodate(struct ocfs2_caching_info *ci, struct buffer_head *bh); -void ocfs2_set_buffer_uptodate(struct inode *inode, +void ocfs2_set_buffer_uptodate(struct ocfs2_caching_info *ci, struct buffer_head *bh); -void ocfs2_set_new_buffer_uptodate(struct inode *inode, +void ocfs2_set_new_buffer_uptodate(struct ocfs2_caching_info *ci, struct buffer_head *bh); -void ocfs2_remove_from_cache(struct inode *inode, +void ocfs2_remove_from_cache(struct ocfs2_caching_info *ci, struct buffer_head *bh); -void ocfs2_remove_xattr_clusters_from_cache(struct inode *inode, +void ocfs2_remove_xattr_clusters_from_cache(struct ocfs2_caching_info *ci, sector_t block, u32 c_len); -int ocfs2_buffer_read_ahead(struct inode *inode, +int ocfs2_buffer_read_ahead(struct ocfs2_caching_info *ci, struct buffer_head *bh); #endif /* OCFS2_UPTODATE_H */ diff --git a/fs/ocfs2/ver.c b/fs/ocfs2/ver.c deleted file mode 100644 index e2488f4128a..00000000000 --- a/fs/ocfs2/ver.c +++ /dev/null @@ -1,43 +0,0 @@ -/* -*- mode: c; c-basic-offset: 8; -*- - * vim: noexpandtab sw=8 ts=8 sts=0: - * - * ver.c - * - * version string - * - * Copyright (C) 2002, 2005 Oracle. All rights reserved. - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public - * License as published by the Free Software Foundation; either - * version 2 of the License, or (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - * General Public License for more details. - * - * You should have received a copy of the GNU General Public - * License along with this program; if not, write to the - * Free Software Foundation, Inc., 59 Temple Place - Suite 330, - * Boston, MA 021110-1307, USA. - */ - -#include <linux/module.h> -#include <linux/string.h> -#include <linux/kernel.h> - -#include "ver.h" - -#define OCFS2_BUILD_VERSION "1.5.0" - -#define VERSION_STR "OCFS2 " OCFS2_BUILD_VERSION - -void ocfs2_print_version(void) -{ - printk(KERN_INFO "%s\n", VERSION_STR); -} - -MODULE_DESCRIPTION(VERSION_STR); - -MODULE_VERSION(OCFS2_BUILD_VERSION); diff --git a/fs/ocfs2/ver.h b/fs/ocfs2/ver.h deleted file mode 100644 index d7395cb91d2..00000000000 --- a/fs/ocfs2/ver.h +++ /dev/null @@ -1,31 +0,0 @@ -/* -*- mode: c; c-basic-offset: 8; -*- - * vim: noexpandtab sw=8 ts=8 sts=0: - * - * ver.h - * - * Function prototypes - * - * Copyright (C) 2002, 2004 Oracle. All rights reserved. - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public - * License as published by the Free Software Foundation; either - * version 2 of the License, or (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - * General Public License for more details. - * - * You should have received a copy of the GNU General Public - * License along with this program; if not, write to the - * Free Software Foundation, Inc., 59 Temple Place - Suite 330, - * Boston, MA 021110-1307, USA. - */ - -#ifndef OCFS2_VER_H -#define OCFS2_VER_H - -void ocfs2_print_version(void); - -#endif /* OCFS2_VER_H */ diff --git a/fs/ocfs2/xattr.c b/fs/ocfs2/xattr.c index 15631019dc6..016f01df382 100644 --- a/fs/ocfs2/xattr.c +++ b/fs/ocfs2/xattr.c @@ -37,7 +37,6 @@ #include <linux/string.h> #include <linux/security.h> -#define MLOG_MASK_PREFIX ML_XATTR #include <cluster/masklog.h> #include "ocfs2.h" @@ -55,7 +54,9 @@ #include "buffer_head_io.h" #include "super.h" #include "xattr.h" - +#include "refcounttree.h" +#include "acl.h" +#include "ocfs2_trace.h" struct ocfs2_xattr_def_value_root { struct ocfs2_xattr_value_root xv; @@ -78,6 +79,7 @@ struct ocfs2_xattr_set_ctxt { struct ocfs2_alloc_context *meta_ac; struct ocfs2_alloc_context *data_ac; struct ocfs2_cached_dealloc_ctxt dealloc; + int set_abort; }; #define OCFS2_XATTR_ROOT_SIZE (sizeof(struct ocfs2_xattr_def_value_root)) @@ -95,34 +97,31 @@ static struct ocfs2_xattr_def_value_root def_xv = { .xv.xr_list.l_count = cpu_to_le16(1), }; -struct xattr_handler *ocfs2_xattr_handlers[] = { +const struct xattr_handler *ocfs2_xattr_handlers[] = { &ocfs2_xattr_user_handler, -#ifdef CONFIG_OCFS2_FS_POSIX_ACL - &ocfs2_xattr_acl_access_handler, - &ocfs2_xattr_acl_default_handler, -#endif + &posix_acl_access_xattr_handler, + &posix_acl_default_xattr_handler, &ocfs2_xattr_trusted_handler, &ocfs2_xattr_security_handler, NULL }; -static struct xattr_handler *ocfs2_xattr_handler_map[OCFS2_XATTR_MAX] = { +static const struct xattr_handler *ocfs2_xattr_handler_map[OCFS2_XATTR_MAX] = { [OCFS2_XATTR_INDEX_USER] = &ocfs2_xattr_user_handler, -#ifdef CONFIG_OCFS2_FS_POSIX_ACL [OCFS2_XATTR_INDEX_POSIX_ACL_ACCESS] - = &ocfs2_xattr_acl_access_handler, + = &posix_acl_access_xattr_handler, [OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT] - = &ocfs2_xattr_acl_default_handler, -#endif + = &posix_acl_default_xattr_handler, [OCFS2_XATTR_INDEX_TRUSTED] = &ocfs2_xattr_trusted_handler, [OCFS2_XATTR_INDEX_SECURITY] = &ocfs2_xattr_security_handler, }; struct ocfs2_xattr_info { - int name_index; - const char *name; - const void *value; - size_t value_len; + int xi_name_index; + const char *xi_name; + int xi_name_len; + const void *xi_value; + size_t xi_value_len; }; struct ocfs2_xattr_search { @@ -140,7 +139,116 @@ struct ocfs2_xattr_search { int not_found; }; -static int ocfs2_xattr_bucket_get_name_value(struct inode *inode, +/* Operations on struct ocfs2_xa_entry */ +struct ocfs2_xa_loc; +struct ocfs2_xa_loc_operations { + /* + * Journal functions + */ + int (*xlo_journal_access)(handle_t *handle, struct ocfs2_xa_loc *loc, + int type); + void (*xlo_journal_dirty)(handle_t *handle, struct ocfs2_xa_loc *loc); + + /* + * Return a pointer to the appropriate buffer in loc->xl_storage + * at the given offset from loc->xl_header. + */ + void *(*xlo_offset_pointer)(struct ocfs2_xa_loc *loc, int offset); + + /* Can we reuse the existing entry for the new value? */ + int (*xlo_can_reuse)(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_info *xi); + + /* How much space is needed for the new value? */ + int (*xlo_check_space)(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_info *xi); + + /* + * Return the offset of the first name+value pair. This is + * the start of our downward-filling free space. + */ + int (*xlo_get_free_start)(struct ocfs2_xa_loc *loc); + + /* + * Remove the name+value at this location. Do whatever is + * appropriate with the remaining name+value pairs. + */ + void (*xlo_wipe_namevalue)(struct ocfs2_xa_loc *loc); + + /* Fill xl_entry with a new entry */ + void (*xlo_add_entry)(struct ocfs2_xa_loc *loc, u32 name_hash); + + /* Add name+value storage to an entry */ + void (*xlo_add_namevalue)(struct ocfs2_xa_loc *loc, int size); + + /* + * Initialize the value buf's access and bh fields for this entry. + * ocfs2_xa_fill_value_buf() will handle the xv pointer. + */ + void (*xlo_fill_value_buf)(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_value_buf *vb); +}; + +/* + * Describes an xattr entry location. This is a memory structure + * tracking the on-disk structure. + */ +struct ocfs2_xa_loc { + /* This xattr belongs to this inode */ + struct inode *xl_inode; + + /* The ocfs2_xattr_header inside the on-disk storage. Not NULL. */ + struct ocfs2_xattr_header *xl_header; + + /* Bytes from xl_header to the end of the storage */ + int xl_size; + + /* + * The ocfs2_xattr_entry this location describes. If this is + * NULL, this location describes the on-disk structure where it + * would have been. + */ + struct ocfs2_xattr_entry *xl_entry; + + /* + * Internal housekeeping + */ + + /* Buffer(s) containing this entry */ + void *xl_storage; + + /* Operations on the storage backing this location */ + const struct ocfs2_xa_loc_operations *xl_ops; +}; + +/* + * Convenience functions to calculate how much space is needed for a + * given name+value pair + */ +static int namevalue_size(int name_len, uint64_t value_len) +{ + if (value_len > OCFS2_XATTR_INLINE_SIZE) + return OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE; + else + return OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_SIZE(value_len); +} + +static int namevalue_size_xi(struct ocfs2_xattr_info *xi) +{ + return namevalue_size(xi->xi_name_len, xi->xi_value_len); +} + +static int namevalue_size_xe(struct ocfs2_xattr_entry *xe) +{ + u64 value_len = le64_to_cpu(xe->xe_value_size); + + BUG_ON((value_len > OCFS2_XATTR_INLINE_SIZE) && + ocfs2_xattr_is_local(xe)); + return namevalue_size(xe->xe_name_len, value_len); +} + + +static int ocfs2_xattr_bucket_get_name_value(struct super_block *sb, struct ocfs2_xattr_header *xh, int index, int *block_off, @@ -157,7 +265,7 @@ static int ocfs2_xattr_index_block_find(struct inode *inode, struct ocfs2_xattr_search *xs); static int ocfs2_xattr_tree_list_index_block(struct inode *inode, - struct ocfs2_xattr_tree_root *xt, + struct buffer_head *blk_bh, char *buffer, size_t buffer_size); @@ -170,12 +278,40 @@ static int ocfs2_xattr_set_entry_index_block(struct inode *inode, struct ocfs2_xattr_search *xs, struct ocfs2_xattr_set_ctxt *ctxt); -static int ocfs2_delete_xattr_index_block(struct inode *inode, - struct buffer_head *xb_bh); +typedef int (xattr_tree_rec_func)(struct inode *inode, + struct buffer_head *root_bh, + u64 blkno, u32 cpos, u32 len, void *para); +static int ocfs2_iterate_xattr_index_block(struct inode *inode, + struct buffer_head *root_bh, + xattr_tree_rec_func *rec_func, + void *para); +static int ocfs2_delete_xattr_in_bucket(struct inode *inode, + struct ocfs2_xattr_bucket *bucket, + void *para); +static int ocfs2_rm_xattr_cluster(struct inode *inode, + struct buffer_head *root_bh, + u64 blkno, + u32 cpos, + u32 len, + void *para); + static int ocfs2_mv_xattr_buckets(struct inode *inode, handle_t *handle, u64 src_blk, u64 last_blk, u64 to_blk, unsigned int start_bucket, u32 *first_hash); +static int ocfs2_prepare_refcount_xattr(struct inode *inode, + struct ocfs2_dinode *di, + struct ocfs2_xattr_info *xi, + struct ocfs2_xattr_search *xis, + struct ocfs2_xattr_search *xbs, + struct ocfs2_refcount_tree **ref_tree, + int *meta_need, + int *credits); +static int ocfs2_get_xattr_tree_value_root(struct super_block *sb, + struct ocfs2_xattr_bucket *bucket, + int offset, + struct ocfs2_xattr_value_root **xv, + struct buffer_head **bh); static inline u16 ocfs2_xattr_buckets_per_cluster(struct ocfs2_super *osb) { @@ -187,14 +323,6 @@ static inline u16 ocfs2_blocks_per_xattr_bucket(struct super_block *sb) return OCFS2_XATTR_BUCKET_SIZE / (1 << sb->s_blocksize_bits); } -static inline u16 ocfs2_xattr_max_xe_in_bucket(struct super_block *sb) -{ - u16 len = sb->s_blocksize - - offsetof(struct ocfs2_xattr_header, xh_entries); - - return len / sizeof(struct ocfs2_xattr_entry); -} - #define bucket_blkno(_b) ((_b)->bu_bhs[0]->b_blocknr) #define bucket_block(_b, _n) ((_b)->bu_bhs[(_n)]->b_data) #define bucket_xh(_b) ((struct ocfs2_xattr_header *)bucket_block((_b), 0)) @@ -241,7 +369,7 @@ static void ocfs2_xattr_bucket_free(struct ocfs2_xattr_bucket *bucket) * them fully. */ static int ocfs2_init_xattr_bucket(struct ocfs2_xattr_bucket *bucket, - u64 xb_blkno) + u64 xb_blkno, int new) { int i, rc = 0; @@ -249,15 +377,22 @@ static int ocfs2_init_xattr_bucket(struct ocfs2_xattr_bucket *bucket, bucket->bu_bhs[i] = sb_getblk(bucket->bu_inode->i_sb, xb_blkno + i); if (!bucket->bu_bhs[i]) { - rc = -EIO; + rc = -ENOMEM; mlog_errno(rc); break; } - if (!ocfs2_buffer_uptodate(bucket->bu_inode, - bucket->bu_bhs[i])) - ocfs2_set_new_buffer_uptodate(bucket->bu_inode, - bucket->bu_bhs[i]); + if (!ocfs2_buffer_uptodate(INODE_CACHE(bucket->bu_inode), + bucket->bu_bhs[i])) { + if (new) + ocfs2_set_new_buffer_uptodate(INODE_CACHE(bucket->bu_inode), + bucket->bu_bhs[i]); + else { + set_buffer_uptodate(bucket->bu_bhs[i]); + ocfs2_set_buffer_uptodate(INODE_CACHE(bucket->bu_inode), + bucket->bu_bhs[i]); + } + } } if (rc) @@ -271,7 +406,7 @@ static int ocfs2_read_xattr_bucket(struct ocfs2_xattr_bucket *bucket, { int rc; - rc = ocfs2_read_blocks(bucket->bu_inode, xb_blkno, + rc = ocfs2_read_blocks(INODE_CACHE(bucket->bu_inode), xb_blkno, bucket->bu_blocks, bucket->bu_bhs, 0, NULL); if (!rc) { @@ -297,7 +432,8 @@ static int ocfs2_xattr_bucket_journal_access(handle_t *handle, int i, rc = 0; for (i = 0; i < bucket->bu_blocks; i++) { - rc = ocfs2_journal_access(handle, bucket->bu_inode, + rc = ocfs2_journal_access(handle, + INODE_CACHE(bucket->bu_inode), bucket->bu_bhs[i], type); if (rc) { mlog_errno(rc); @@ -345,8 +481,7 @@ static int ocfs2_validate_xattr_block(struct super_block *sb, struct ocfs2_xattr_block *xb = (struct ocfs2_xattr_block *)bh->b_data; - mlog(0, "Validating xattr block %llu\n", - (unsigned long long)bh->b_blocknr); + trace_ocfs2_validate_xattr_block((unsigned long long)bh->b_blocknr); BUG_ON(!buffer_uptodate(bh)); @@ -399,7 +534,7 @@ static int ocfs2_read_xattr_block(struct inode *inode, u64 xb_blkno, int rc; struct buffer_head *tmp = *bh; - rc = ocfs2_read_block(inode, xb_blkno, &tmp, + rc = ocfs2_read_block(INODE_CACHE(inode), xb_blkno, &tmp, ocfs2_validate_xattr_block); /* If ocfs2_read_block() got us a new bh, pass it up. */ @@ -411,7 +546,7 @@ static int ocfs2_read_xattr_block(struct inode *inode, u64 xb_blkno, static inline const char *ocfs2_xattr_prefix(int name_index) { - struct xattr_handler *handler = NULL; + const struct xattr_handler *handler = NULL; if (name_index > 0 && name_index < OCFS2_XATTR_MAX) handler = ocfs2_xattr_handler_map[name_index]; @@ -437,35 +572,22 @@ static u32 ocfs2_xattr_name_hash(struct inode *inode, return hash; } -/* - * ocfs2_xattr_hash_entry() - * - * Compute the hash of an extended attribute. - */ -static void ocfs2_xattr_hash_entry(struct inode *inode, - struct ocfs2_xattr_header *header, - struct ocfs2_xattr_entry *entry) +static int ocfs2_xattr_entry_real_size(int name_len, size_t value_len) { - u32 hash = 0; - char *name = (char *)header + le16_to_cpu(entry->xe_name_offset); - - hash = ocfs2_xattr_name_hash(inode, name, entry->xe_name_len); - entry->xe_name_hash = cpu_to_le32(hash); - - return; + return namevalue_size(name_len, value_len) + + sizeof(struct ocfs2_xattr_entry); } -static int ocfs2_xattr_entry_real_size(int name_len, size_t value_len) +static int ocfs2_xi_entry_usage(struct ocfs2_xattr_info *xi) { - int size = 0; - - if (value_len <= OCFS2_XATTR_INLINE_SIZE) - size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_SIZE(value_len); - else - size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE; - size += sizeof(struct ocfs2_xattr_entry); + return namevalue_size_xi(xi) + + sizeof(struct ocfs2_xattr_entry); +} - return size; +static int ocfs2_xe_entry_usage(struct ocfs2_xattr_entry *xe) +{ + return namevalue_size_xe(xe) + + sizeof(struct ocfs2_xattr_entry); } int ocfs2_calc_security_init(struct inode *dir, @@ -508,7 +630,7 @@ int ocfs2_calc_security_init(struct inode *dir, int ocfs2_calc_xattr_init(struct inode *dir, struct buffer_head *dir_bh, - int mode, + umode_t mode, struct ocfs2_security_xattr_info *si, int *want_clusters, int *xattr_credits, @@ -593,55 +715,61 @@ static int ocfs2_xattr_extend_allocation(struct inode *inode, struct ocfs2_xattr_value_buf *vb, struct ocfs2_xattr_set_ctxt *ctxt) { - int status = 0; + int status = 0, credits; handle_t *handle = ctxt->handle; enum ocfs2_alloc_restarted why; - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); u32 prev_clusters, logical_start = le32_to_cpu(vb->vb_xv->xr_clusters); struct ocfs2_extent_tree et; - mlog(0, "(clusters_to_add for xattr= %u)\n", clusters_to_add); - - ocfs2_init_xattr_value_extent_tree(&et, inode, vb); + ocfs2_init_xattr_value_extent_tree(&et, INODE_CACHE(inode), vb); - status = vb->vb_access(handle, inode, vb->vb_bh, - OCFS2_JOURNAL_ACCESS_WRITE); - if (status < 0) { - mlog_errno(status); - goto leave; - } + while (clusters_to_add) { + trace_ocfs2_xattr_extend_allocation(clusters_to_add); - prev_clusters = le32_to_cpu(vb->vb_xv->xr_clusters); - status = ocfs2_add_clusters_in_btree(osb, - inode, - &logical_start, - clusters_to_add, - 0, - &et, - handle, - ctxt->data_ac, - ctxt->meta_ac, - &why); - if (status < 0) { - mlog_errno(status); - goto leave; - } + status = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (status < 0) { + mlog_errno(status); + break; + } - status = ocfs2_journal_dirty(handle, vb->vb_bh); - if (status < 0) { - mlog_errno(status); - goto leave; - } + prev_clusters = le32_to_cpu(vb->vb_xv->xr_clusters); + status = ocfs2_add_clusters_in_btree(handle, + &et, + &logical_start, + clusters_to_add, + 0, + ctxt->data_ac, + ctxt->meta_ac, + &why); + if ((status < 0) && (status != -EAGAIN)) { + if (status != -ENOSPC) + mlog_errno(status); + break; + } - clusters_to_add -= le32_to_cpu(vb->vb_xv->xr_clusters) - prev_clusters; + ocfs2_journal_dirty(handle, vb->vb_bh); - /* - * We should have already allocated enough space before the transaction, - * so no need to restart. - */ - BUG_ON(why != RESTART_NONE || clusters_to_add); + clusters_to_add -= le32_to_cpu(vb->vb_xv->xr_clusters) - + prev_clusters; -leave: + if (why != RESTART_NONE && clusters_to_add) { + /* + * We can only fail in case the alloc file doesn't give + * up enough clusters. + */ + BUG_ON(why == RESTART_META); + + credits = ocfs2_calc_extend_credits(inode->i_sb, + &vb->vb_xv->xr_list); + status = ocfs2_extend_trans(handle, credits); + if (status < 0) { + status = -ENOMEM; + mlog_errno(status); + break; + } + } + } return status; } @@ -649,6 +777,7 @@ leave: static int __ocfs2_remove_xattr_range(struct inode *inode, struct ocfs2_xattr_value_buf *vb, u32 cpos, u32 phys_cpos, u32 len, + unsigned int ext_flags, struct ocfs2_xattr_set_ctxt *ctxt) { int ret; @@ -656,16 +785,16 @@ static int __ocfs2_remove_xattr_range(struct inode *inode, handle_t *handle = ctxt->handle; struct ocfs2_extent_tree et; - ocfs2_init_xattr_value_extent_tree(&et, inode, vb); + ocfs2_init_xattr_value_extent_tree(&et, INODE_CACHE(inode), vb); - ret = vb->vb_access(handle, inode, vb->vb_bh, + ret = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); goto out; } - ret = ocfs2_remove_extent(inode, &et, cpos, len, handle, ctxt->meta_ac, + ret = ocfs2_remove_extent(handle, &et, cpos, len, ctxt->meta_ac, &ctxt->dealloc); if (ret) { mlog_errno(ret); @@ -673,14 +802,16 @@ static int __ocfs2_remove_xattr_range(struct inode *inode, } le32_add_cpu(&vb->vb_xv->xr_clusters, -len); + ocfs2_journal_dirty(handle, vb->vb_bh); - ret = ocfs2_journal_dirty(handle, vb->vb_bh); - if (ret) { - mlog_errno(ret); - goto out; - } - - ret = ocfs2_cache_cluster_dealloc(&ctxt->dealloc, phys_blkno, len); + if (ext_flags & OCFS2_EXT_REFCOUNTED) + ret = ocfs2_decrease_refcount(inode, handle, + ocfs2_blocks_to_clusters(inode->i_sb, + phys_blkno), + len, ctxt->meta_ac, &ctxt->dealloc, 1); + else + ret = ocfs2_cache_cluster_dealloc(&ctxt->dealloc, + phys_blkno, len); if (ret) mlog_errno(ret); @@ -695,6 +826,7 @@ static int ocfs2_xattr_shrink_size(struct inode *inode, struct ocfs2_xattr_set_ctxt *ctxt) { int ret = 0; + unsigned int ext_flags; u32 trunc_len, cpos, phys_cpos, alloc_size; u64 block; @@ -706,7 +838,7 @@ static int ocfs2_xattr_shrink_size(struct inode *inode, while (trunc_len) { ret = ocfs2_xattr_get_clusters(inode, cpos, &phys_cpos, &alloc_size, - &vb->vb_xv->xr_list); + &vb->vb_xv->xr_list, &ext_flags); if (ret) { mlog_errno(ret); goto out; @@ -717,15 +849,15 @@ static int ocfs2_xattr_shrink_size(struct inode *inode, ret = __ocfs2_remove_xattr_range(inode, vb, cpos, phys_cpos, alloc_size, - ctxt); + ext_flags, ctxt); if (ret) { mlog_errno(ret); goto out; } block = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos); - ocfs2_remove_xattr_clusters_from_cache(inode, block, - alloc_size); + ocfs2_remove_xattr_clusters_from_cache(INODE_CACHE(inode), + block, alloc_size); cpos += alloc_size; trunc_len -= alloc_size; } @@ -810,6 +942,23 @@ static int ocfs2_xattr_list_entries(struct inode *inode, return result; } +int ocfs2_has_inline_xattr_value_outside(struct inode *inode, + struct ocfs2_dinode *di) +{ + struct ocfs2_xattr_header *xh; + int i; + + xh = (struct ocfs2_xattr_header *) + ((void *)di + inode->i_sb->s_blocksize - + le16_to_cpu(di->i_xattr_inline_size)); + + for (i = 0; i < le16_to_cpu(xh->xh_count); i++) + if (!ocfs2_xattr_is_local(&xh->xh_entries[i])) + return 1; + + return 0; +} + static int ocfs2_xattr_ibody_list(struct inode *inode, struct ocfs2_dinode *di, char *buffer, @@ -855,11 +1004,9 @@ static int ocfs2_xattr_block_list(struct inode *inode, struct ocfs2_xattr_header *header = &xb->xb_attrs.xb_header; ret = ocfs2_xattr_list_entries(inode, header, buffer, buffer_size); - } else { - struct ocfs2_xattr_tree_root *xt = &xb->xb_attrs.xb_root; - ret = ocfs2_xattr_tree_list_index_block(inode, xt, + } else + ret = ocfs2_xattr_tree_list_index_block(inode, blk_bh, buffer, buffer_size); - } brelse(blk_bh); @@ -961,7 +1108,7 @@ static int ocfs2_xattr_get_value_outside(struct inode *inode, cpos = 0; while (cpos < clusters) { ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster, - &num_clusters, el); + &num_clusters, el, NULL); if (ret) { mlog_errno(ret); goto out; @@ -970,7 +1117,8 @@ static int ocfs2_xattr_get_value_outside(struct inode *inode, blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster); /* Copy ocfs2_xattr_value */ for (i = 0; i < num_clusters * bpc; i++, blkno++) { - ret = ocfs2_read_block(inode, blkno, &bh, NULL); + ret = ocfs2_read_block(INODE_CACHE(inode), blkno, + &bh, NULL); if (ret) { mlog_errno(ret); goto out; @@ -1052,7 +1200,8 @@ static int ocfs2_xattr_block_get(struct inode *inode, struct ocfs2_xattr_block *xb; struct ocfs2_xattr_value_root *xv; size_t size; - int ret = -ENODATA, name_offset, name_len, block_off, i; + int ret = -ENODATA, name_offset, name_len, i; + int uninitialized_var(block_off); xs->bucket = ocfs2_xattr_bucket_new(inode); if (!xs->bucket) { @@ -1084,7 +1233,7 @@ static int ocfs2_xattr_block_get(struct inode *inode, i = xs->here - xs->header->xh_entries; if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { - ret = ocfs2_xattr_bucket_get_name_value(inode, + ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, bucket_xh(xs->bucket), i, &block_off, @@ -1140,13 +1289,11 @@ int ocfs2_xattr_get_nolock(struct inode *inode, xis.inode_bh = xbs.inode_bh = di_bh; di = (struct ocfs2_dinode *)di_bh->b_data; - down_read(&oi->ip_xattr_sem); ret = ocfs2_xattr_ibody_get(inode, name_index, name, buffer, buffer_size, &xis); if (ret == -ENODATA && di->i_xattr_loc) ret = ocfs2_xattr_block_get(inode, name_index, name, buffer, buffer_size, &xbs); - up_read(&oi->ip_xattr_sem); return ret; } @@ -1170,8 +1317,10 @@ static int ocfs2_xattr_get(struct inode *inode, mlog_errno(ret); return ret; } + down_read(&OCFS2_I(inode)->ip_xattr_sem); ret = ocfs2_xattr_get_nolock(inode, di_bh, name_index, name, buffer, buffer_size); + up_read(&OCFS2_I(inode)->ip_xattr_sem); ocfs2_inode_unlock(inode, 0); @@ -1182,7 +1331,7 @@ static int ocfs2_xattr_get(struct inode *inode, static int __ocfs2_xattr_set_value_outside(struct inode *inode, handle_t *handle, - struct ocfs2_xattr_value_root *xv, + struct ocfs2_xattr_value_buf *vb, const void *value, int value_len) { @@ -1193,28 +1342,34 @@ static int __ocfs2_xattr_set_value_outside(struct inode *inode, u32 clusters = ocfs2_clusters_for_bytes(inode->i_sb, value_len); u64 blkno; struct buffer_head *bh = NULL; + unsigned int ext_flags; + struct ocfs2_xattr_value_root *xv = vb->vb_xv; BUG_ON(clusters > le32_to_cpu(xv->xr_clusters)); while (cpos < clusters) { ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster, - &num_clusters, &xv->xr_list); + &num_clusters, &xv->xr_list, + &ext_flags); if (ret) { mlog_errno(ret); goto out; } + BUG_ON(ext_flags & OCFS2_EXT_REFCOUNTED); + blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster); for (i = 0; i < num_clusters * bpc; i++, blkno++) { - ret = ocfs2_read_block(inode, blkno, &bh, NULL); + ret = ocfs2_read_block(INODE_CACHE(inode), blkno, + &bh, NULL); if (ret) { mlog_errno(ret); goto out; } ret = ocfs2_journal_access(handle, - inode, + INODE_CACHE(inode), bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret < 0) { @@ -1230,11 +1385,7 @@ static int __ocfs2_xattr_set_value_outside(struct inode *inode, memset(bh->b_data + cp_len, 0, blocksize - cp_len); - ret = ocfs2_journal_dirty(handle, bh); - if (ret < 0) { - mlog_errno(ret); - goto out; - } + ocfs2_journal_dirty(handle, bh); brelse(bh); bh = NULL; @@ -1253,497 +1404,1009 @@ out: return ret; } -static int ocfs2_xattr_cleanup(struct inode *inode, - handle_t *handle, - struct ocfs2_xattr_info *xi, - struct ocfs2_xattr_search *xs, - struct ocfs2_xattr_value_buf *vb, - size_t offs) +static int ocfs2_xa_check_space_helper(int needed_space, int free_start, + int num_entries) { - int ret = 0; - size_t name_len = strlen(xi->name); - void *val = xs->base + offs; - size_t size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE; + int free_space; - ret = vb->vb_access(handle, inode, vb->vb_bh, - OCFS2_JOURNAL_ACCESS_WRITE); - if (ret) { - mlog_errno(ret); - goto out; - } - /* Decrease xattr count */ - le16_add_cpu(&xs->header->xh_count, -1); - /* Remove the xattr entry and tree root which has already be set*/ - memset((void *)xs->here, 0, sizeof(struct ocfs2_xattr_entry)); - memset(val, 0, size); + if (!needed_space) + return 0; - ret = ocfs2_journal_dirty(handle, vb->vb_bh); - if (ret < 0) - mlog_errno(ret); -out: - return ret; + free_space = free_start - + sizeof(struct ocfs2_xattr_header) - + (num_entries * sizeof(struct ocfs2_xattr_entry)) - + OCFS2_XATTR_HEADER_GAP; + if (free_space < 0) + return -EIO; + if (free_space < needed_space) + return -ENOSPC; + + return 0; } -static int ocfs2_xattr_update_entry(struct inode *inode, - handle_t *handle, - struct ocfs2_xattr_info *xi, - struct ocfs2_xattr_search *xs, - struct ocfs2_xattr_value_buf *vb, - size_t offs) +static int ocfs2_xa_journal_access(handle_t *handle, struct ocfs2_xa_loc *loc, + int type) { - int ret; + return loc->xl_ops->xlo_journal_access(handle, loc, type); +} - ret = vb->vb_access(handle, inode, vb->vb_bh, - OCFS2_JOURNAL_ACCESS_WRITE); - if (ret) { - mlog_errno(ret); - goto out; - } +static void ocfs2_xa_journal_dirty(handle_t *handle, struct ocfs2_xa_loc *loc) +{ + loc->xl_ops->xlo_journal_dirty(handle, loc); +} - xs->here->xe_name_offset = cpu_to_le16(offs); - xs->here->xe_value_size = cpu_to_le64(xi->value_len); - if (xi->value_len <= OCFS2_XATTR_INLINE_SIZE) - ocfs2_xattr_set_local(xs->here, 1); - else - ocfs2_xattr_set_local(xs->here, 0); - ocfs2_xattr_hash_entry(inode, xs->header, xs->here); +/* Give a pointer into the storage for the given offset */ +static void *ocfs2_xa_offset_pointer(struct ocfs2_xa_loc *loc, int offset) +{ + BUG_ON(offset >= loc->xl_size); + return loc->xl_ops->xlo_offset_pointer(loc, offset); +} - ret = ocfs2_journal_dirty(handle, vb->vb_bh); - if (ret < 0) - mlog_errno(ret); -out: - return ret; +/* + * Wipe the name+value pair and allow the storage to reclaim it. This + * must be followed by either removal of the entry or a call to + * ocfs2_xa_add_namevalue(). + */ +static void ocfs2_xa_wipe_namevalue(struct ocfs2_xa_loc *loc) +{ + loc->xl_ops->xlo_wipe_namevalue(loc); } /* - * ocfs2_xattr_set_value_outside() - * - * Set large size value in B tree. + * Find lowest offset to a name+value pair. This is the start of our + * downward-growing free space. */ -static int ocfs2_xattr_set_value_outside(struct inode *inode, - struct ocfs2_xattr_info *xi, - struct ocfs2_xattr_search *xs, - struct ocfs2_xattr_set_ctxt *ctxt, - struct ocfs2_xattr_value_buf *vb, - size_t offs) +static int ocfs2_xa_get_free_start(struct ocfs2_xa_loc *loc) { - size_t name_len = strlen(xi->name); - void *val = xs->base + offs; - struct ocfs2_xattr_value_root *xv = NULL; - size_t size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE; - int ret = 0; + return loc->xl_ops->xlo_get_free_start(loc); +} - memset(val, 0, size); - memcpy(val, xi->name, name_len); - xv = (struct ocfs2_xattr_value_root *) - (val + OCFS2_XATTR_SIZE(name_len)); - xv->xr_clusters = 0; - xv->xr_last_eb_blk = 0; - xv->xr_list.l_tree_depth = 0; - xv->xr_list.l_count = cpu_to_le16(1); - xv->xr_list.l_next_free_rec = 0; - vb->vb_xv = xv; - - ret = ocfs2_xattr_value_truncate(inode, vb, xi->value_len, ctxt); - if (ret < 0) { - mlog_errno(ret); - return ret; +/* Can we reuse loc->xl_entry for xi? */ +static int ocfs2_xa_can_reuse_entry(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_info *xi) +{ + return loc->xl_ops->xlo_can_reuse(loc, xi); +} + +/* How much free space is needed to set the new value */ +static int ocfs2_xa_check_space(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_info *xi) +{ + return loc->xl_ops->xlo_check_space(loc, xi); +} + +static void ocfs2_xa_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash) +{ + loc->xl_ops->xlo_add_entry(loc, name_hash); + loc->xl_entry->xe_name_hash = cpu_to_le32(name_hash); + /* + * We can't leave the new entry's xe_name_offset at zero or + * add_namevalue() will go nuts. We set it to the size of our + * storage so that it can never be less than any other entry. + */ + loc->xl_entry->xe_name_offset = cpu_to_le16(loc->xl_size); +} + +static void ocfs2_xa_add_namevalue(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_info *xi) +{ + int size = namevalue_size_xi(xi); + int nameval_offset; + char *nameval_buf; + + loc->xl_ops->xlo_add_namevalue(loc, size); + loc->xl_entry->xe_value_size = cpu_to_le64(xi->xi_value_len); + loc->xl_entry->xe_name_len = xi->xi_name_len; + ocfs2_xattr_set_type(loc->xl_entry, xi->xi_name_index); + ocfs2_xattr_set_local(loc->xl_entry, + xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE); + + nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset); + nameval_buf = ocfs2_xa_offset_pointer(loc, nameval_offset); + memset(nameval_buf, 0, size); + memcpy(nameval_buf, xi->xi_name, xi->xi_name_len); +} + +static void ocfs2_xa_fill_value_buf(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_value_buf *vb) +{ + int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset); + int name_size = OCFS2_XATTR_SIZE(loc->xl_entry->xe_name_len); + + /* Value bufs are for value trees */ + BUG_ON(ocfs2_xattr_is_local(loc->xl_entry)); + BUG_ON(namevalue_size_xe(loc->xl_entry) != + (name_size + OCFS2_XATTR_ROOT_SIZE)); + + loc->xl_ops->xlo_fill_value_buf(loc, vb); + vb->vb_xv = + (struct ocfs2_xattr_value_root *)ocfs2_xa_offset_pointer(loc, + nameval_offset + + name_size); +} + +static int ocfs2_xa_block_journal_access(handle_t *handle, + struct ocfs2_xa_loc *loc, int type) +{ + struct buffer_head *bh = loc->xl_storage; + ocfs2_journal_access_func access; + + if (loc->xl_size == (bh->b_size - + offsetof(struct ocfs2_xattr_block, + xb_attrs.xb_header))) + access = ocfs2_journal_access_xb; + else + access = ocfs2_journal_access_di; + return access(handle, INODE_CACHE(loc->xl_inode), bh, type); +} + +static void ocfs2_xa_block_journal_dirty(handle_t *handle, + struct ocfs2_xa_loc *loc) +{ + struct buffer_head *bh = loc->xl_storage; + + ocfs2_journal_dirty(handle, bh); +} + +static void *ocfs2_xa_block_offset_pointer(struct ocfs2_xa_loc *loc, + int offset) +{ + return (char *)loc->xl_header + offset; +} + +static int ocfs2_xa_block_can_reuse(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_info *xi) +{ + /* + * Block storage is strict. If the sizes aren't exact, we will + * remove the old one and reinsert the new. + */ + return namevalue_size_xe(loc->xl_entry) == + namevalue_size_xi(xi); +} + +static int ocfs2_xa_block_get_free_start(struct ocfs2_xa_loc *loc) +{ + struct ocfs2_xattr_header *xh = loc->xl_header; + int i, count = le16_to_cpu(xh->xh_count); + int offset, free_start = loc->xl_size; + + for (i = 0; i < count; i++) { + offset = le16_to_cpu(xh->xh_entries[i].xe_name_offset); + if (offset < free_start) + free_start = offset; } - ret = ocfs2_xattr_update_entry(inode, ctxt->handle, xi, xs, vb, offs); - if (ret < 0) { - mlog_errno(ret); - return ret; + + return free_start; +} + +static int ocfs2_xa_block_check_space(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_info *xi) +{ + int count = le16_to_cpu(loc->xl_header->xh_count); + int free_start = ocfs2_xa_get_free_start(loc); + int needed_space = ocfs2_xi_entry_usage(xi); + + /* + * Block storage will reclaim the original entry before inserting + * the new value, so we only need the difference. If the new + * entry is smaller than the old one, we don't need anything. + */ + if (loc->xl_entry) { + /* Don't need space if we're reusing! */ + if (ocfs2_xa_can_reuse_entry(loc, xi)) + needed_space = 0; + else + needed_space -= ocfs2_xe_entry_usage(loc->xl_entry); } - ret = __ocfs2_xattr_set_value_outside(inode, ctxt->handle, vb->vb_xv, - xi->value, xi->value_len); - if (ret < 0) - mlog_errno(ret); + if (needed_space < 0) + needed_space = 0; + return ocfs2_xa_check_space_helper(needed_space, free_start, count); +} - return ret; +/* + * Block storage for xattrs keeps the name+value pairs compacted. When + * we remove one, we have to shift any that preceded it towards the end. + */ +static void ocfs2_xa_block_wipe_namevalue(struct ocfs2_xa_loc *loc) +{ + int i, offset; + int namevalue_offset, first_namevalue_offset, namevalue_size; + struct ocfs2_xattr_entry *entry = loc->xl_entry; + struct ocfs2_xattr_header *xh = loc->xl_header; + int count = le16_to_cpu(xh->xh_count); + + namevalue_offset = le16_to_cpu(entry->xe_name_offset); + namevalue_size = namevalue_size_xe(entry); + first_namevalue_offset = ocfs2_xa_get_free_start(loc); + + /* Shift the name+value pairs */ + memmove((char *)xh + first_namevalue_offset + namevalue_size, + (char *)xh + first_namevalue_offset, + namevalue_offset - first_namevalue_offset); + memset((char *)xh + first_namevalue_offset, 0, namevalue_size); + + /* Now tell xh->xh_entries about it */ + for (i = 0; i < count; i++) { + offset = le16_to_cpu(xh->xh_entries[i].xe_name_offset); + if (offset <= namevalue_offset) + le16_add_cpu(&xh->xh_entries[i].xe_name_offset, + namevalue_size); + } + + /* + * Note that we don't update xh_free_start or xh_name_value_len + * because they're not used in block-stored xattrs. + */ +} + +static void ocfs2_xa_block_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash) +{ + int count = le16_to_cpu(loc->xl_header->xh_count); + loc->xl_entry = &(loc->xl_header->xh_entries[count]); + le16_add_cpu(&loc->xl_header->xh_count, 1); + memset(loc->xl_entry, 0, sizeof(struct ocfs2_xattr_entry)); +} + +static void ocfs2_xa_block_add_namevalue(struct ocfs2_xa_loc *loc, int size) +{ + int free_start = ocfs2_xa_get_free_start(loc); + + loc->xl_entry->xe_name_offset = cpu_to_le16(free_start - size); +} + +static void ocfs2_xa_block_fill_value_buf(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_value_buf *vb) +{ + struct buffer_head *bh = loc->xl_storage; + + if (loc->xl_size == (bh->b_size - + offsetof(struct ocfs2_xattr_block, + xb_attrs.xb_header))) + vb->vb_access = ocfs2_journal_access_xb; + else + vb->vb_access = ocfs2_journal_access_di; + vb->vb_bh = bh; } /* - * ocfs2_xattr_set_entry_local() - * - * Set, replace or remove extended attribute in local. + * Operations for xattrs stored in blocks. This includes inline inode + * storage and unindexed ocfs2_xattr_blocks. */ -static void ocfs2_xattr_set_entry_local(struct inode *inode, - struct ocfs2_xattr_info *xi, - struct ocfs2_xattr_search *xs, - struct ocfs2_xattr_entry *last, - size_t min_offs) +static const struct ocfs2_xa_loc_operations ocfs2_xa_block_loc_ops = { + .xlo_journal_access = ocfs2_xa_block_journal_access, + .xlo_journal_dirty = ocfs2_xa_block_journal_dirty, + .xlo_offset_pointer = ocfs2_xa_block_offset_pointer, + .xlo_check_space = ocfs2_xa_block_check_space, + .xlo_can_reuse = ocfs2_xa_block_can_reuse, + .xlo_get_free_start = ocfs2_xa_block_get_free_start, + .xlo_wipe_namevalue = ocfs2_xa_block_wipe_namevalue, + .xlo_add_entry = ocfs2_xa_block_add_entry, + .xlo_add_namevalue = ocfs2_xa_block_add_namevalue, + .xlo_fill_value_buf = ocfs2_xa_block_fill_value_buf, +}; + +static int ocfs2_xa_bucket_journal_access(handle_t *handle, + struct ocfs2_xa_loc *loc, int type) { - size_t name_len = strlen(xi->name); - int i; + struct ocfs2_xattr_bucket *bucket = loc->xl_storage; - if (xi->value && xs->not_found) { - /* Insert the new xattr entry. */ - le16_add_cpu(&xs->header->xh_count, 1); - ocfs2_xattr_set_type(last, xi->name_index); - ocfs2_xattr_set_local(last, 1); - last->xe_name_len = name_len; - } else { - void *first_val; - void *val; - size_t offs, size; - - first_val = xs->base + min_offs; - offs = le16_to_cpu(xs->here->xe_name_offset); - val = xs->base + offs; - - if (le64_to_cpu(xs->here->xe_value_size) > - OCFS2_XATTR_INLINE_SIZE) - size = OCFS2_XATTR_SIZE(name_len) + - OCFS2_XATTR_ROOT_SIZE; + return ocfs2_xattr_bucket_journal_access(handle, bucket, type); +} + +static void ocfs2_xa_bucket_journal_dirty(handle_t *handle, + struct ocfs2_xa_loc *loc) +{ + struct ocfs2_xattr_bucket *bucket = loc->xl_storage; + + ocfs2_xattr_bucket_journal_dirty(handle, bucket); +} + +static void *ocfs2_xa_bucket_offset_pointer(struct ocfs2_xa_loc *loc, + int offset) +{ + struct ocfs2_xattr_bucket *bucket = loc->xl_storage; + int block, block_offset; + + /* The header is at the front of the bucket */ + block = offset >> loc->xl_inode->i_sb->s_blocksize_bits; + block_offset = offset % loc->xl_inode->i_sb->s_blocksize; + + return bucket_block(bucket, block) + block_offset; +} + +static int ocfs2_xa_bucket_can_reuse(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_info *xi) +{ + return namevalue_size_xe(loc->xl_entry) >= + namevalue_size_xi(xi); +} + +static int ocfs2_xa_bucket_get_free_start(struct ocfs2_xa_loc *loc) +{ + struct ocfs2_xattr_bucket *bucket = loc->xl_storage; + return le16_to_cpu(bucket_xh(bucket)->xh_free_start); +} + +static int ocfs2_bucket_align_free_start(struct super_block *sb, + int free_start, int size) +{ + /* + * We need to make sure that the name+value pair fits within + * one block. + */ + if (((free_start - size) >> sb->s_blocksize_bits) != + ((free_start - 1) >> sb->s_blocksize_bits)) + free_start -= free_start % sb->s_blocksize; + + return free_start; +} + +static int ocfs2_xa_bucket_check_space(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_info *xi) +{ + int rc; + int count = le16_to_cpu(loc->xl_header->xh_count); + int free_start = ocfs2_xa_get_free_start(loc); + int needed_space = ocfs2_xi_entry_usage(xi); + int size = namevalue_size_xi(xi); + struct super_block *sb = loc->xl_inode->i_sb; + + /* + * Bucket storage does not reclaim name+value pairs it cannot + * reuse. They live as holes until the bucket fills, and then + * the bucket is defragmented. However, the bucket can reclaim + * the ocfs2_xattr_entry. + */ + if (loc->xl_entry) { + /* Don't need space if we're reusing! */ + if (ocfs2_xa_can_reuse_entry(loc, xi)) + needed_space = 0; else - size = OCFS2_XATTR_SIZE(name_len) + - OCFS2_XATTR_SIZE(le64_to_cpu(xs->here->xe_value_size)); - - if (xi->value && size == OCFS2_XATTR_SIZE(name_len) + - OCFS2_XATTR_SIZE(xi->value_len)) { - /* The old and the new value have the - same size. Just replace the value. */ - ocfs2_xattr_set_local(xs->here, 1); - xs->here->xe_value_size = cpu_to_le64(xi->value_len); - /* Clear value bytes. */ - memset(val + OCFS2_XATTR_SIZE(name_len), - 0, - OCFS2_XATTR_SIZE(xi->value_len)); - memcpy(val + OCFS2_XATTR_SIZE(name_len), - xi->value, - xi->value_len); - return; - } - /* Remove the old name+value. */ - memmove(first_val + size, first_val, val - first_val); - memset(first_val, 0, size); - xs->here->xe_name_hash = 0; - xs->here->xe_name_offset = 0; - ocfs2_xattr_set_local(xs->here, 1); - xs->here->xe_value_size = 0; - - min_offs += size; - - /* Adjust all value offsets. */ - last = xs->header->xh_entries; - for (i = 0 ; i < le16_to_cpu(xs->header->xh_count); i++) { - size_t o = le16_to_cpu(last->xe_name_offset); - - if (o < offs) - last->xe_name_offset = cpu_to_le16(o + size); - last += 1; - } + needed_space -= sizeof(struct ocfs2_xattr_entry); + } + BUG_ON(needed_space < 0); - if (!xi->value) { - /* Remove the old entry. */ - last -= 1; - memmove(xs->here, xs->here + 1, - (void *)last - (void *)xs->here); - memset(last, 0, sizeof(struct ocfs2_xattr_entry)); - le16_add_cpu(&xs->header->xh_count, -1); - } + if (free_start < size) { + if (needed_space) + return -ENOSPC; + } else { + /* + * First we check if it would fit in the first place. + * Below, we align the free start to a block. This may + * slide us below the minimum gap. By checking unaligned + * first, we avoid that error. + */ + rc = ocfs2_xa_check_space_helper(needed_space, free_start, + count); + if (rc) + return rc; + free_start = ocfs2_bucket_align_free_start(sb, free_start, + size); } - if (xi->value) { - /* Insert the new name+value. */ - size_t size = OCFS2_XATTR_SIZE(name_len) + - OCFS2_XATTR_SIZE(xi->value_len); - void *val = xs->base + min_offs - size; + return ocfs2_xa_check_space_helper(needed_space, free_start, count); +} - xs->here->xe_name_offset = cpu_to_le16(min_offs - size); - memset(val, 0, size); - memcpy(val, xi->name, name_len); - memcpy(val + OCFS2_XATTR_SIZE(name_len), - xi->value, - xi->value_len); - xs->here->xe_value_size = cpu_to_le64(xi->value_len); - ocfs2_xattr_set_local(xs->here, 1); - ocfs2_xattr_hash_entry(inode, xs->header, xs->here); +static void ocfs2_xa_bucket_wipe_namevalue(struct ocfs2_xa_loc *loc) +{ + le16_add_cpu(&loc->xl_header->xh_name_value_len, + -namevalue_size_xe(loc->xl_entry)); +} + +static void ocfs2_xa_bucket_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash) +{ + struct ocfs2_xattr_header *xh = loc->xl_header; + int count = le16_to_cpu(xh->xh_count); + int low = 0, high = count - 1, tmp; + struct ocfs2_xattr_entry *tmp_xe; + + /* + * We keep buckets sorted by name_hash, so we need to find + * our insert place. + */ + while (low <= high && count) { + tmp = (low + high) / 2; + tmp_xe = &xh->xh_entries[tmp]; + + if (name_hash > le32_to_cpu(tmp_xe->xe_name_hash)) + low = tmp + 1; + else if (name_hash < le32_to_cpu(tmp_xe->xe_name_hash)) + high = tmp - 1; + else { + low = tmp; + break; + } } - return; + if (low != count) + memmove(&xh->xh_entries[low + 1], + &xh->xh_entries[low], + ((count - low) * sizeof(struct ocfs2_xattr_entry))); + + le16_add_cpu(&xh->xh_count, 1); + loc->xl_entry = &xh->xh_entries[low]; + memset(loc->xl_entry, 0, sizeof(struct ocfs2_xattr_entry)); +} + +static void ocfs2_xa_bucket_add_namevalue(struct ocfs2_xa_loc *loc, int size) +{ + int free_start = ocfs2_xa_get_free_start(loc); + struct ocfs2_xattr_header *xh = loc->xl_header; + struct super_block *sb = loc->xl_inode->i_sb; + int nameval_offset; + + free_start = ocfs2_bucket_align_free_start(sb, free_start, size); + nameval_offset = free_start - size; + loc->xl_entry->xe_name_offset = cpu_to_le16(nameval_offset); + xh->xh_free_start = cpu_to_le16(nameval_offset); + le16_add_cpu(&xh->xh_name_value_len, size); + +} + +static void ocfs2_xa_bucket_fill_value_buf(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_value_buf *vb) +{ + struct ocfs2_xattr_bucket *bucket = loc->xl_storage; + struct super_block *sb = loc->xl_inode->i_sb; + int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset); + int size = namevalue_size_xe(loc->xl_entry); + int block_offset = nameval_offset >> sb->s_blocksize_bits; + + /* Values are not allowed to straddle block boundaries */ + BUG_ON(block_offset != + ((nameval_offset + size - 1) >> sb->s_blocksize_bits)); + /* We expect the bucket to be filled in */ + BUG_ON(!bucket->bu_bhs[block_offset]); + + vb->vb_access = ocfs2_journal_access; + vb->vb_bh = bucket->bu_bhs[block_offset]; +} + +/* Operations for xattrs stored in buckets. */ +static const struct ocfs2_xa_loc_operations ocfs2_xa_bucket_loc_ops = { + .xlo_journal_access = ocfs2_xa_bucket_journal_access, + .xlo_journal_dirty = ocfs2_xa_bucket_journal_dirty, + .xlo_offset_pointer = ocfs2_xa_bucket_offset_pointer, + .xlo_check_space = ocfs2_xa_bucket_check_space, + .xlo_can_reuse = ocfs2_xa_bucket_can_reuse, + .xlo_get_free_start = ocfs2_xa_bucket_get_free_start, + .xlo_wipe_namevalue = ocfs2_xa_bucket_wipe_namevalue, + .xlo_add_entry = ocfs2_xa_bucket_add_entry, + .xlo_add_namevalue = ocfs2_xa_bucket_add_namevalue, + .xlo_fill_value_buf = ocfs2_xa_bucket_fill_value_buf, +}; + +static unsigned int ocfs2_xa_value_clusters(struct ocfs2_xa_loc *loc) +{ + struct ocfs2_xattr_value_buf vb; + + if (ocfs2_xattr_is_local(loc->xl_entry)) + return 0; + + ocfs2_xa_fill_value_buf(loc, &vb); + return le32_to_cpu(vb.vb_xv->xr_clusters); +} + +static int ocfs2_xa_value_truncate(struct ocfs2_xa_loc *loc, u64 bytes, + struct ocfs2_xattr_set_ctxt *ctxt) +{ + int trunc_rc, access_rc; + struct ocfs2_xattr_value_buf vb; + + ocfs2_xa_fill_value_buf(loc, &vb); + trunc_rc = ocfs2_xattr_value_truncate(loc->xl_inode, &vb, bytes, + ctxt); + + /* + * The caller of ocfs2_xa_value_truncate() has already called + * ocfs2_xa_journal_access on the loc. However, The truncate code + * calls ocfs2_extend_trans(). This may commit the previous + * transaction and open a new one. If this is a bucket, truncate + * could leave only vb->vb_bh set up for journaling. Meanwhile, + * the caller is expecting to dirty the entire bucket. So we must + * reset the journal work. We do this even if truncate has failed, + * as it could have failed after committing the extend. + */ + access_rc = ocfs2_xa_journal_access(ctxt->handle, loc, + OCFS2_JOURNAL_ACCESS_WRITE); + + /* Errors in truncate take precedence */ + return trunc_rc ? trunc_rc : access_rc; +} + +static void ocfs2_xa_remove_entry(struct ocfs2_xa_loc *loc) +{ + int index, count; + struct ocfs2_xattr_header *xh = loc->xl_header; + struct ocfs2_xattr_entry *entry = loc->xl_entry; + + ocfs2_xa_wipe_namevalue(loc); + loc->xl_entry = NULL; + + le16_add_cpu(&xh->xh_count, -1); + count = le16_to_cpu(xh->xh_count); + + /* + * Only zero out the entry if there are more remaining. This is + * important for an empty bucket, as it keeps track of the + * bucket's hash value. It doesn't hurt empty block storage. + */ + if (count) { + index = ((char *)entry - (char *)&xh->xh_entries) / + sizeof(struct ocfs2_xattr_entry); + memmove(&xh->xh_entries[index], &xh->xh_entries[index + 1], + (count - index) * sizeof(struct ocfs2_xattr_entry)); + memset(&xh->xh_entries[count], 0, + sizeof(struct ocfs2_xattr_entry)); + } } /* - * ocfs2_xattr_set_entry() + * If we have a problem adjusting the size of an external value during + * ocfs2_xa_prepare_entry() or ocfs2_xa_remove(), we may have an xattr + * in an intermediate state. For example, the value may be partially + * truncated. * - * Set extended attribute entry into inode or block. + * If the value tree hasn't changed, the extend/truncate went nowhere. + * We have nothing to do. The caller can treat it as a straight error. * - * If extended attribute value size > OCFS2_XATTR_INLINE_SIZE, - * We first insert tree root(ocfs2_xattr_value_root) with set_entry_local(), - * then set value in B tree with set_value_outside(). + * If the value tree got partially truncated, we now have a corrupted + * extended attribute. We're going to wipe its entry and leak the + * clusters. Better to leak some storage than leave a corrupt entry. + * + * If the value tree grew, it obviously didn't grow enough for the + * new entry. We're not going to try and reclaim those clusters either. + * If there was already an external value there (orig_clusters != 0), + * the new clusters are attached safely and we can just leave the old + * value in place. If there was no external value there, we remove + * the entry. + * + * This way, the xattr block we store in the journal will be consistent. + * If the size change broke because of the journal, no changes will hit + * disk anyway. */ -static int ocfs2_xattr_set_entry(struct inode *inode, - struct ocfs2_xattr_info *xi, - struct ocfs2_xattr_search *xs, - struct ocfs2_xattr_set_ctxt *ctxt, - int flag) +static void ocfs2_xa_cleanup_value_truncate(struct ocfs2_xa_loc *loc, + const char *what, + unsigned int orig_clusters) { - struct ocfs2_xattr_entry *last; - struct ocfs2_inode_info *oi = OCFS2_I(inode); - struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; - size_t min_offs = xs->end - xs->base, name_len = strlen(xi->name); - size_t size_l = 0; - handle_t *handle = ctxt->handle; - int free, i, ret; - struct ocfs2_xattr_info xi_l = { - .name_index = xi->name_index, - .name = xi->name, - .value = xi->value, - .value_len = xi->value_len, - }; - struct ocfs2_xattr_value_buf vb = { - .vb_bh = xs->xattr_bh, - .vb_access = ocfs2_journal_access_di, - }; - - if (!(flag & OCFS2_INLINE_XATTR_FL)) { - BUG_ON(xs->xattr_bh == xs->inode_bh); - vb.vb_access = ocfs2_journal_access_xb; - } else - BUG_ON(xs->xattr_bh != xs->inode_bh); + unsigned int new_clusters = ocfs2_xa_value_clusters(loc); + char *nameval_buf = ocfs2_xa_offset_pointer(loc, + le16_to_cpu(loc->xl_entry->xe_name_offset)); + + if (new_clusters < orig_clusters) { + mlog(ML_ERROR, + "Partial truncate while %s xattr %.*s. Leaking " + "%u clusters and removing the entry\n", + what, loc->xl_entry->xe_name_len, nameval_buf, + orig_clusters - new_clusters); + ocfs2_xa_remove_entry(loc); + } else if (!orig_clusters) { + mlog(ML_ERROR, + "Unable to allocate an external value for xattr " + "%.*s safely. Leaking %u clusters and removing the " + "entry\n", + loc->xl_entry->xe_name_len, nameval_buf, + new_clusters - orig_clusters); + ocfs2_xa_remove_entry(loc); + } else if (new_clusters > orig_clusters) + mlog(ML_ERROR, + "Unable to grow xattr %.*s safely. %u new clusters " + "have been added, but the value will not be " + "modified\n", + loc->xl_entry->xe_name_len, nameval_buf, + new_clusters - orig_clusters); +} - /* Compute min_offs, last and free space. */ - last = xs->header->xh_entries; +static int ocfs2_xa_remove(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_set_ctxt *ctxt) +{ + int rc = 0; + unsigned int orig_clusters; - for (i = 0 ; i < le16_to_cpu(xs->header->xh_count); i++) { - size_t offs = le16_to_cpu(last->xe_name_offset); - if (offs < min_offs) - min_offs = offs; - last += 1; + if (!ocfs2_xattr_is_local(loc->xl_entry)) { + orig_clusters = ocfs2_xa_value_clusters(loc); + rc = ocfs2_xa_value_truncate(loc, 0, ctxt); + if (rc) { + mlog_errno(rc); + /* + * Since this is remove, we can return 0 if + * ocfs2_xa_cleanup_value_truncate() is going to + * wipe the entry anyway. So we check the + * cluster count as well. + */ + if (orig_clusters != ocfs2_xa_value_clusters(loc)) + rc = 0; + ocfs2_xa_cleanup_value_truncate(loc, "removing", + orig_clusters); + if (rc) + goto out; + } } - free = min_offs - ((void *)last - xs->base) - OCFS2_XATTR_HEADER_GAP; - if (free < 0) - return -EIO; + ocfs2_xa_remove_entry(loc); - if (!xs->not_found) { - size_t size = 0; - if (ocfs2_xattr_is_local(xs->here)) - size = OCFS2_XATTR_SIZE(name_len) + - OCFS2_XATTR_SIZE(le64_to_cpu(xs->here->xe_value_size)); - else - size = OCFS2_XATTR_SIZE(name_len) + - OCFS2_XATTR_ROOT_SIZE; - free += (size + sizeof(struct ocfs2_xattr_entry)); - } - /* Check free space in inode or block */ - if (xi->value && xi->value_len > OCFS2_XATTR_INLINE_SIZE) { - if (free < sizeof(struct ocfs2_xattr_entry) + - OCFS2_XATTR_SIZE(name_len) + - OCFS2_XATTR_ROOT_SIZE) { - ret = -ENOSPC; - goto out; +out: + return rc; +} + +static void ocfs2_xa_install_value_root(struct ocfs2_xa_loc *loc) +{ + int name_size = OCFS2_XATTR_SIZE(loc->xl_entry->xe_name_len); + char *nameval_buf; + + nameval_buf = ocfs2_xa_offset_pointer(loc, + le16_to_cpu(loc->xl_entry->xe_name_offset)); + memcpy(nameval_buf + name_size, &def_xv, OCFS2_XATTR_ROOT_SIZE); +} + +/* + * Take an existing entry and make it ready for the new value. This + * won't allocate space, but it may free space. It should be ready for + * ocfs2_xa_prepare_entry() to finish the work. + */ +static int ocfs2_xa_reuse_entry(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_info *xi, + struct ocfs2_xattr_set_ctxt *ctxt) +{ + int rc = 0; + int name_size = OCFS2_XATTR_SIZE(xi->xi_name_len); + unsigned int orig_clusters; + char *nameval_buf; + int xe_local = ocfs2_xattr_is_local(loc->xl_entry); + int xi_local = xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE; + + BUG_ON(OCFS2_XATTR_SIZE(loc->xl_entry->xe_name_len) != + name_size); + + nameval_buf = ocfs2_xa_offset_pointer(loc, + le16_to_cpu(loc->xl_entry->xe_name_offset)); + if (xe_local) { + memset(nameval_buf + name_size, 0, + namevalue_size_xe(loc->xl_entry) - name_size); + if (!xi_local) + ocfs2_xa_install_value_root(loc); + } else { + orig_clusters = ocfs2_xa_value_clusters(loc); + if (xi_local) { + rc = ocfs2_xa_value_truncate(loc, 0, ctxt); + if (rc < 0) + mlog_errno(rc); + else + memset(nameval_buf + name_size, 0, + namevalue_size_xe(loc->xl_entry) - + name_size); + } else if (le64_to_cpu(loc->xl_entry->xe_value_size) > + xi->xi_value_len) { + rc = ocfs2_xa_value_truncate(loc, xi->xi_value_len, + ctxt); + if (rc < 0) + mlog_errno(rc); } - size_l = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE; - xi_l.value = (void *)&def_xv; - xi_l.value_len = OCFS2_XATTR_ROOT_SIZE; - } else if (xi->value) { - if (free < sizeof(struct ocfs2_xattr_entry) + - OCFS2_XATTR_SIZE(name_len) + - OCFS2_XATTR_SIZE(xi->value_len)) { - ret = -ENOSPC; + + if (rc) { + ocfs2_xa_cleanup_value_truncate(loc, "reusing", + orig_clusters); goto out; } } - if (!xs->not_found) { - /* For existing extended attribute */ - size_t size = OCFS2_XATTR_SIZE(name_len) + - OCFS2_XATTR_SIZE(le64_to_cpu(xs->here->xe_value_size)); - size_t offs = le16_to_cpu(xs->here->xe_name_offset); - void *val = xs->base + offs; + loc->xl_entry->xe_value_size = cpu_to_le64(xi->xi_value_len); + ocfs2_xattr_set_local(loc->xl_entry, xi_local); - if (ocfs2_xattr_is_local(xs->here) && size == size_l) { - /* Replace existing local xattr with tree root */ - ret = ocfs2_xattr_set_value_outside(inode, xi, xs, - ctxt, &vb, offs); - if (ret < 0) - mlog_errno(ret); - goto out; - } else if (!ocfs2_xattr_is_local(xs->here)) { - /* For existing xattr which has value outside */ - vb.vb_xv = (struct ocfs2_xattr_value_root *) - (val + OCFS2_XATTR_SIZE(name_len)); +out: + return rc; +} - if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) { - /* - * If new value need set outside also, - * first truncate old value to new value, - * then set new value with set_value_outside(). - */ - ret = ocfs2_xattr_value_truncate(inode, - &vb, - xi->value_len, - ctxt); - if (ret < 0) { - mlog_errno(ret); - goto out; - } +/* + * Prepares loc->xl_entry to receive the new xattr. This includes + * properly setting up the name+value pair region. If loc->xl_entry + * already exists, it will take care of modifying it appropriately. + * + * Note that this modifies the data. You did journal_access already, + * right? + */ +static int ocfs2_xa_prepare_entry(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_info *xi, + u32 name_hash, + struct ocfs2_xattr_set_ctxt *ctxt) +{ + int rc = 0; + unsigned int orig_clusters; + __le64 orig_value_size = 0; - ret = ocfs2_xattr_update_entry(inode, - handle, - xi, - xs, - &vb, - offs); - if (ret < 0) { - mlog_errno(ret); - goto out; - } + rc = ocfs2_xa_check_space(loc, xi); + if (rc) + goto out; - ret = __ocfs2_xattr_set_value_outside(inode, - handle, - vb.vb_xv, - xi->value, - xi->value_len); - if (ret < 0) - mlog_errno(ret); + if (loc->xl_entry) { + if (ocfs2_xa_can_reuse_entry(loc, xi)) { + orig_value_size = loc->xl_entry->xe_value_size; + rc = ocfs2_xa_reuse_entry(loc, xi, ctxt); + if (rc) goto out; - } else { - /* - * If new value need set in local, - * just trucate old value to zero. - */ - ret = ocfs2_xattr_value_truncate(inode, - &vb, - 0, - ctxt); - if (ret < 0) - mlog_errno(ret); + goto alloc_value; + } + + if (!ocfs2_xattr_is_local(loc->xl_entry)) { + orig_clusters = ocfs2_xa_value_clusters(loc); + rc = ocfs2_xa_value_truncate(loc, 0, ctxt); + if (rc) { + mlog_errno(rc); + ocfs2_xa_cleanup_value_truncate(loc, + "overwriting", + orig_clusters); + goto out; + } + } + ocfs2_xa_wipe_namevalue(loc); + } else + ocfs2_xa_add_entry(loc, name_hash); + + /* + * If we get here, we have a blank entry. Fill it. We grow our + * name+value pair back from the end. + */ + ocfs2_xa_add_namevalue(loc, xi); + if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) + ocfs2_xa_install_value_root(loc); + +alloc_value: + if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { + orig_clusters = ocfs2_xa_value_clusters(loc); + rc = ocfs2_xa_value_truncate(loc, xi->xi_value_len, ctxt); + if (rc < 0) { + ctxt->set_abort = 1; + ocfs2_xa_cleanup_value_truncate(loc, "growing", + orig_clusters); + /* + * If we were growing an existing value, + * ocfs2_xa_cleanup_value_truncate() won't remove + * the entry. We need to restore the original value + * size. + */ + if (loc->xl_entry) { + BUG_ON(!orig_value_size); + loc->xl_entry->xe_value_size = orig_value_size; } + mlog_errno(rc); } } - ret = ocfs2_journal_access_di(handle, inode, xs->inode_bh, +out: + return rc; +} + +/* + * Store the value portion of the name+value pair. This will skip + * values that are stored externally. Their tree roots were set up + * by ocfs2_xa_prepare_entry(). + */ +static int ocfs2_xa_store_value(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_info *xi, + struct ocfs2_xattr_set_ctxt *ctxt) +{ + int rc = 0; + int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset); + int name_size = OCFS2_XATTR_SIZE(xi->xi_name_len); + char *nameval_buf; + struct ocfs2_xattr_value_buf vb; + + nameval_buf = ocfs2_xa_offset_pointer(loc, nameval_offset); + if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { + ocfs2_xa_fill_value_buf(loc, &vb); + rc = __ocfs2_xattr_set_value_outside(loc->xl_inode, + ctxt->handle, &vb, + xi->xi_value, + xi->xi_value_len); + } else + memcpy(nameval_buf + name_size, xi->xi_value, xi->xi_value_len); + + return rc; +} + +static int ocfs2_xa_set(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_info *xi, + struct ocfs2_xattr_set_ctxt *ctxt) +{ + int ret; + u32 name_hash = ocfs2_xattr_name_hash(loc->xl_inode, xi->xi_name, + xi->xi_name_len); + + ret = ocfs2_xa_journal_access(ctxt->handle, loc, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); goto out; } - if (!(flag & OCFS2_INLINE_XATTR_FL)) { - ret = vb.vb_access(handle, inode, vb.vb_bh, - OCFS2_JOURNAL_ACCESS_WRITE); - if (ret) { - mlog_errno(ret); - goto out; - } - } - /* - * Set value in local, include set tree root in local. - * This is the first step for value size >INLINE_SIZE. + * From here on out, everything is going to modify the buffer a + * little. Errors are going to leave the xattr header in a + * sane state. Thus, even with errors we dirty the sucker. */ - ocfs2_xattr_set_entry_local(inode, &xi_l, xs, last, min_offs); - if (!(flag & OCFS2_INLINE_XATTR_FL)) { - ret = ocfs2_journal_dirty(handle, xs->xattr_bh); - if (ret < 0) { + /* Don't worry, we are never called with !xi_value and !xl_entry */ + if (!xi->xi_value) { + ret = ocfs2_xa_remove(loc, ctxt); + goto out_dirty; + } + + ret = ocfs2_xa_prepare_entry(loc, xi, name_hash, ctxt); + if (ret) { + if (ret != -ENOSPC) mlog_errno(ret); - goto out; - } + goto out_dirty; } - if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) && - (flag & OCFS2_INLINE_XATTR_FL)) { - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - unsigned int xattrsize = osb->s_xattr_inline_size; + ret = ocfs2_xa_store_value(loc, xi, ctxt); + if (ret) + mlog_errno(ret); - /* - * Adjust extent record count or inline data size - * to reserve space for extended attribute. - */ - if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) { - struct ocfs2_inline_data *idata = &di->id2.i_data; - le16_add_cpu(&idata->id_count, -xattrsize); - } else if (!(ocfs2_inode_is_fast_symlink(inode))) { - struct ocfs2_extent_list *el = &di->id2.i_list; - le16_add_cpu(&el->l_count, -(xattrsize / - sizeof(struct ocfs2_extent_rec))); - } - di->i_xattr_inline_size = cpu_to_le16(xattrsize); - } - /* Update xattr flag */ - spin_lock(&oi->ip_lock); - oi->ip_dyn_features |= flag; - di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features); - spin_unlock(&oi->ip_lock); +out_dirty: + ocfs2_xa_journal_dirty(ctxt->handle, loc); - ret = ocfs2_journal_dirty(handle, xs->inode_bh); - if (ret < 0) +out: + return ret; +} + +static void ocfs2_init_dinode_xa_loc(struct ocfs2_xa_loc *loc, + struct inode *inode, + struct buffer_head *bh, + struct ocfs2_xattr_entry *entry) +{ + struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; + + BUG_ON(!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_XATTR_FL)); + + loc->xl_inode = inode; + loc->xl_ops = &ocfs2_xa_block_loc_ops; + loc->xl_storage = bh; + loc->xl_entry = entry; + loc->xl_size = le16_to_cpu(di->i_xattr_inline_size); + loc->xl_header = + (struct ocfs2_xattr_header *)(bh->b_data + bh->b_size - + loc->xl_size); +} + +static void ocfs2_init_xattr_block_xa_loc(struct ocfs2_xa_loc *loc, + struct inode *inode, + struct buffer_head *bh, + struct ocfs2_xattr_entry *entry) +{ + struct ocfs2_xattr_block *xb = + (struct ocfs2_xattr_block *)bh->b_data; + + BUG_ON(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED); + + loc->xl_inode = inode; + loc->xl_ops = &ocfs2_xa_block_loc_ops; + loc->xl_storage = bh; + loc->xl_header = &(xb->xb_attrs.xb_header); + loc->xl_entry = entry; + loc->xl_size = bh->b_size - offsetof(struct ocfs2_xattr_block, + xb_attrs.xb_header); +} + +static void ocfs2_init_xattr_bucket_xa_loc(struct ocfs2_xa_loc *loc, + struct ocfs2_xattr_bucket *bucket, + struct ocfs2_xattr_entry *entry) +{ + loc->xl_inode = bucket->bu_inode; + loc->xl_ops = &ocfs2_xa_bucket_loc_ops; + loc->xl_storage = bucket; + loc->xl_header = bucket_xh(bucket); + loc->xl_entry = entry; + loc->xl_size = OCFS2_XATTR_BUCKET_SIZE; +} + +/* + * In xattr remove, if it is stored outside and refcounted, we may have + * the chance to split the refcount tree. So need the allocators. + */ +static int ocfs2_lock_xattr_remove_allocators(struct inode *inode, + struct ocfs2_xattr_value_root *xv, + struct ocfs2_caching_info *ref_ci, + struct buffer_head *ref_root_bh, + struct ocfs2_alloc_context **meta_ac, + int *ref_credits) +{ + int ret, meta_add = 0; + u32 p_cluster, num_clusters; + unsigned int ext_flags; + + *ref_credits = 0; + ret = ocfs2_xattr_get_clusters(inode, 0, &p_cluster, + &num_clusters, + &xv->xr_list, + &ext_flags); + if (ret) { mlog_errno(ret); + goto out; + } - if (!ret && xi->value_len > OCFS2_XATTR_INLINE_SIZE) { - /* - * Set value outside in B tree. - * This is the second step for value size > INLINE_SIZE. - */ - size_t offs = le16_to_cpu(xs->here->xe_name_offset); - ret = ocfs2_xattr_set_value_outside(inode, xi, xs, ctxt, - &vb, offs); - if (ret < 0) { - int ret2; + if (!(ext_flags & OCFS2_EXT_REFCOUNTED)) + goto out; - mlog_errno(ret); - /* - * If set value outside failed, we have to clean - * the junk tree root we have already set in local. - */ - ret2 = ocfs2_xattr_cleanup(inode, ctxt->handle, - xi, xs, &vb, offs); - if (ret2 < 0) - mlog_errno(ret2); - } + ret = ocfs2_refcounted_xattr_delete_need(inode, ref_ci, + ref_root_bh, xv, + &meta_add, ref_credits); + if (ret) { + mlog_errno(ret); + goto out; } + + ret = ocfs2_reserve_new_metadata_blocks(OCFS2_SB(inode->i_sb), + meta_add, meta_ac); + if (ret) + mlog_errno(ret); + out: return ret; } static int ocfs2_remove_value_outside(struct inode*inode, struct ocfs2_xattr_value_buf *vb, - struct ocfs2_xattr_header *header) + struct ocfs2_xattr_header *header, + struct ocfs2_caching_info *ref_ci, + struct buffer_head *ref_root_bh) { - int ret = 0, i; + int ret = 0, i, ref_credits; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); struct ocfs2_xattr_set_ctxt ctxt = { NULL, NULL, }; + void *val; ocfs2_init_dealloc_ctxt(&ctxt.dealloc); - ctxt.handle = ocfs2_start_trans(osb, - ocfs2_remove_extent_credits(osb->sb)); - if (IS_ERR(ctxt.handle)) { - ret = PTR_ERR(ctxt.handle); - mlog_errno(ret); - goto out; - } - for (i = 0; i < le16_to_cpu(header->xh_count); i++) { struct ocfs2_xattr_entry *entry = &header->xh_entries[i]; - if (!ocfs2_xattr_is_local(entry)) { - void *val; + if (ocfs2_xattr_is_local(entry)) + continue; - val = (void *)header + - le16_to_cpu(entry->xe_name_offset); - vb->vb_xv = (struct ocfs2_xattr_value_root *) - (val + OCFS2_XATTR_SIZE(entry->xe_name_len)); - ret = ocfs2_xattr_value_truncate(inode, vb, 0, &ctxt); - if (ret < 0) { - mlog_errno(ret); - break; - } + val = (void *)header + + le16_to_cpu(entry->xe_name_offset); + vb->vb_xv = (struct ocfs2_xattr_value_root *) + (val + OCFS2_XATTR_SIZE(entry->xe_name_len)); + + ret = ocfs2_lock_xattr_remove_allocators(inode, vb->vb_xv, + ref_ci, ref_root_bh, + &ctxt.meta_ac, + &ref_credits); + + ctxt.handle = ocfs2_start_trans(osb, ref_credits + + ocfs2_remove_extent_credits(osb->sb)); + if (IS_ERR(ctxt.handle)) { + ret = PTR_ERR(ctxt.handle); + mlog_errno(ret); + break; } + + ret = ocfs2_xattr_value_truncate(inode, vb, 0, &ctxt); + + ocfs2_commit_trans(osb, ctxt.handle); + if (ctxt.meta_ac) { + ocfs2_free_alloc_context(ctxt.meta_ac); + ctxt.meta_ac = NULL; + } + + if (ret < 0) { + mlog_errno(ret); + break; + } + } - ocfs2_commit_trans(osb, ctxt.handle); + if (ctxt.meta_ac) + ocfs2_free_alloc_context(ctxt.meta_ac); ocfs2_schedule_truncate_log_flush(osb, 1); ocfs2_run_deallocs(osb, &ctxt.dealloc); -out: return ret; } static int ocfs2_xattr_ibody_remove(struct inode *inode, - struct buffer_head *di_bh) + struct buffer_head *di_bh, + struct ocfs2_caching_info *ref_ci, + struct buffer_head *ref_root_bh) { struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; @@ -1758,13 +2421,21 @@ static int ocfs2_xattr_ibody_remove(struct inode *inode, ((void *)di + inode->i_sb->s_blocksize - le16_to_cpu(di->i_xattr_inline_size)); - ret = ocfs2_remove_value_outside(inode, &vb, header); + ret = ocfs2_remove_value_outside(inode, &vb, header, + ref_ci, ref_root_bh); return ret; } +struct ocfs2_rm_xattr_bucket_para { + struct ocfs2_caching_info *ref_ci; + struct buffer_head *ref_root_bh; +}; + static int ocfs2_xattr_block_remove(struct inode *inode, - struct buffer_head *blk_bh) + struct buffer_head *blk_bh, + struct ocfs2_caching_info *ref_ci, + struct buffer_head *ref_root_bh) { struct ocfs2_xattr_block *xb; int ret = 0; @@ -1772,19 +2443,29 @@ static int ocfs2_xattr_block_remove(struct inode *inode, .vb_bh = blk_bh, .vb_access = ocfs2_journal_access_xb, }; + struct ocfs2_rm_xattr_bucket_para args = { + .ref_ci = ref_ci, + .ref_root_bh = ref_root_bh, + }; xb = (struct ocfs2_xattr_block *)blk_bh->b_data; if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) { struct ocfs2_xattr_header *header = &(xb->xb_attrs.xb_header); - ret = ocfs2_remove_value_outside(inode, &vb, header); + ret = ocfs2_remove_value_outside(inode, &vb, header, + ref_ci, ref_root_bh); } else - ret = ocfs2_delete_xattr_index_block(inode, blk_bh); + ret = ocfs2_iterate_xattr_index_block(inode, + blk_bh, + ocfs2_rm_xattr_cluster, + &args); return ret; } static int ocfs2_xattr_free_block(struct inode *inode, - u64 block) + u64 block, + struct ocfs2_caching_info *ref_ci, + struct buffer_head *ref_root_bh) { struct inode *xb_alloc_inode; struct buffer_head *xb_alloc_bh = NULL; @@ -1802,7 +2483,7 @@ static int ocfs2_xattr_free_block(struct inode *inode, goto out; } - ret = ocfs2_xattr_block_remove(inode, blk_bh); + ret = ocfs2_xattr_block_remove(inode, blk_bh, ref_ci, ref_root_bh); if (ret < 0) { mlog_errno(ret); goto out; @@ -1811,7 +2492,10 @@ static int ocfs2_xattr_free_block(struct inode *inode, xb = (struct ocfs2_xattr_block *)blk_bh->b_data; blk = le64_to_cpu(xb->xb_blkno); bit = le16_to_cpu(xb->xb_suballoc_bit); - bg_blkno = ocfs2_which_suballoc_group(blk, bit); + if (xb->xb_suballoc_loc) + bg_blkno = le64_to_cpu(xb->xb_suballoc_loc); + else + bg_blkno = ocfs2_which_suballoc_group(blk, bit); xb_alloc_inode = ocfs2_get_system_file_inode(osb, EXTENT_ALLOC_SYSTEM_INODE, @@ -1862,6 +2546,9 @@ int ocfs2_xattr_remove(struct inode *inode, struct buffer_head *di_bh) { struct ocfs2_inode_info *oi = OCFS2_I(inode); struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + struct ocfs2_refcount_tree *ref_tree = NULL; + struct buffer_head *ref_root_bh = NULL; + struct ocfs2_caching_info *ref_ci = NULL; handle_t *handle; int ret; @@ -1871,8 +2558,21 @@ int ocfs2_xattr_remove(struct inode *inode, struct buffer_head *di_bh) if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) return 0; + if (OCFS2_I(inode)->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL) { + ret = ocfs2_lock_refcount_tree(OCFS2_SB(inode->i_sb), + le64_to_cpu(di->i_refcount_loc), + 1, &ref_tree, &ref_root_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + ref_ci = &ref_tree->rf_ci; + + } + if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { - ret = ocfs2_xattr_ibody_remove(inode, di_bh); + ret = ocfs2_xattr_ibody_remove(inode, di_bh, + ref_ci, ref_root_bh); if (ret < 0) { mlog_errno(ret); goto out; @@ -1881,7 +2581,8 @@ int ocfs2_xattr_remove(struct inode *inode, struct buffer_head *di_bh) if (di->i_xattr_loc) { ret = ocfs2_xattr_free_block(inode, - le64_to_cpu(di->i_xattr_loc)); + le64_to_cpu(di->i_xattr_loc), + ref_ci, ref_root_bh); if (ret < 0) { mlog_errno(ret); goto out; @@ -1895,7 +2596,7 @@ int ocfs2_xattr_remove(struct inode *inode, struct buffer_head *di_bh) mlog_errno(ret); goto out; } - ret = ocfs2_journal_access_di(handle, inode, di_bh, + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); @@ -1908,13 +2609,15 @@ int ocfs2_xattr_remove(struct inode *inode, struct buffer_head *di_bh) oi->ip_dyn_features &= ~(OCFS2_INLINE_XATTR_FL | OCFS2_HAS_XATTR_FL); di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features); spin_unlock(&oi->ip_lock); + ocfs2_update_inode_fsync_trans(handle, inode, 0); - ret = ocfs2_journal_dirty(handle, di_bh); - if (ret < 0) - mlog_errno(ret); + ocfs2_journal_dirty(handle, di_bh); out_commit: ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle); out: + if (ref_tree) + ocfs2_unlock_refcount_tree(OCFS2_SB(inode->i_sb), ref_tree, 1); + brelse(ref_root_bh); return ret; } @@ -1995,6 +2698,53 @@ static int ocfs2_xattr_ibody_find(struct inode *inode, return 0; } +static int ocfs2_xattr_ibody_init(struct inode *inode, + struct buffer_head *di_bh, + struct ocfs2_xattr_set_ctxt *ctxt) +{ + int ret; + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + unsigned int xattrsize = osb->s_xattr_inline_size; + + if (!ocfs2_xattr_has_space_inline(inode, di)) { + ret = -ENOSPC; + goto out; + } + + ret = ocfs2_journal_access_di(ctxt->handle, INODE_CACHE(inode), di_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out; + } + + /* + * Adjust extent record count or inline data size + * to reserve space for extended attribute. + */ + if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) { + struct ocfs2_inline_data *idata = &di->id2.i_data; + le16_add_cpu(&idata->id_count, -xattrsize); + } else if (!(ocfs2_inode_is_fast_symlink(inode))) { + struct ocfs2_extent_list *el = &di->id2.i_list; + le16_add_cpu(&el->l_count, -(xattrsize / + sizeof(struct ocfs2_extent_rec))); + } + di->i_xattr_inline_size = cpu_to_le16(xattrsize); + + spin_lock(&oi->ip_lock); + oi->ip_dyn_features |= OCFS2_INLINE_XATTR_FL|OCFS2_HAS_XATTR_FL; + di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features); + spin_unlock(&oi->ip_lock); + + ocfs2_journal_dirty(ctxt->handle, di_bh); + +out: + return ret; +} + /* * ocfs2_xattr_ibody_set() * @@ -2006,23 +2756,33 @@ static int ocfs2_xattr_ibody_set(struct inode *inode, struct ocfs2_xattr_search *xs, struct ocfs2_xattr_set_ctxt *ctxt) { - struct ocfs2_inode_info *oi = OCFS2_I(inode); - struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; int ret; + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct ocfs2_xa_loc loc; if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE) return -ENOSPC; down_write(&oi->ip_alloc_sem); if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) { - if (!ocfs2_xattr_has_space_inline(inode, di)) { - ret = -ENOSPC; + ret = ocfs2_xattr_ibody_init(inode, xs->inode_bh, ctxt); + if (ret) { + if (ret != -ENOSPC) + mlog_errno(ret); goto out; } } - ret = ocfs2_xattr_set_entry(inode, xi, xs, ctxt, - (OCFS2_INLINE_XATTR_FL | OCFS2_HAS_XATTR_FL)); + ocfs2_init_dinode_xa_loc(&loc, inode, xs->inode_bh, + xs->not_found ? NULL : xs->here); + ret = ocfs2_xa_set(&loc, xi, ctxt); + if (ret) { + if (ret != -ENOSPC) + mlog_errno(ret); + goto out; + } + xs->here = loc.xl_entry; + out: up_write(&oi->ip_alloc_sem); @@ -2082,6 +2842,92 @@ cleanup: return ret; } +static int ocfs2_create_xattr_block(struct inode *inode, + struct buffer_head *inode_bh, + struct ocfs2_xattr_set_ctxt *ctxt, + int indexed, + struct buffer_head **ret_bh) +{ + int ret; + u16 suballoc_bit_start; + u32 num_got; + u64 suballoc_loc, first_blkno; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)inode_bh->b_data; + struct buffer_head *new_bh = NULL; + struct ocfs2_xattr_block *xblk; + + ret = ocfs2_journal_access_di(ctxt->handle, INODE_CACHE(inode), + inode_bh, OCFS2_JOURNAL_ACCESS_CREATE); + if (ret < 0) { + mlog_errno(ret); + goto end; + } + + ret = ocfs2_claim_metadata(ctxt->handle, ctxt->meta_ac, 1, + &suballoc_loc, &suballoc_bit_start, + &num_got, &first_blkno); + if (ret < 0) { + mlog_errno(ret); + goto end; + } + + new_bh = sb_getblk(inode->i_sb, first_blkno); + if (!new_bh) { + ret = -ENOMEM; + mlog_errno(ret); + goto end; + } + + ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode), new_bh); + + ret = ocfs2_journal_access_xb(ctxt->handle, INODE_CACHE(inode), + new_bh, + OCFS2_JOURNAL_ACCESS_CREATE); + if (ret < 0) { + mlog_errno(ret); + goto end; + } + + /* Initialize ocfs2_xattr_block */ + xblk = (struct ocfs2_xattr_block *)new_bh->b_data; + memset(xblk, 0, inode->i_sb->s_blocksize); + strcpy((void *)xblk, OCFS2_XATTR_BLOCK_SIGNATURE); + xblk->xb_suballoc_slot = cpu_to_le16(ctxt->meta_ac->ac_alloc_slot); + xblk->xb_suballoc_loc = cpu_to_le64(suballoc_loc); + xblk->xb_suballoc_bit = cpu_to_le16(suballoc_bit_start); + xblk->xb_fs_generation = + cpu_to_le32(OCFS2_SB(inode->i_sb)->fs_generation); + xblk->xb_blkno = cpu_to_le64(first_blkno); + if (indexed) { + struct ocfs2_xattr_tree_root *xr = &xblk->xb_attrs.xb_root; + xr->xt_clusters = cpu_to_le32(1); + xr->xt_last_eb_blk = 0; + xr->xt_list.l_tree_depth = 0; + xr->xt_list.l_count = cpu_to_le16( + ocfs2_xattr_recs_per_xb(inode->i_sb)); + xr->xt_list.l_next_free_rec = cpu_to_le16(1); + xblk->xb_flags = cpu_to_le16(OCFS2_XATTR_INDEXED); + } + ocfs2_journal_dirty(ctxt->handle, new_bh); + + /* Add it to the inode */ + di->i_xattr_loc = cpu_to_le64(first_blkno); + + spin_lock(&OCFS2_I(inode)->ip_lock); + OCFS2_I(inode)->ip_dyn_features |= OCFS2_HAS_XATTR_FL; + di->i_dyn_features = cpu_to_le16(OCFS2_I(inode)->ip_dyn_features); + spin_unlock(&OCFS2_I(inode)->ip_lock); + + ocfs2_journal_dirty(ctxt->handle, inode_bh); + + *ret_bh = new_bh; + new_bh = NULL; + +end: + brelse(new_bh); + return ret; +} + /* * ocfs2_xattr_block_set() * @@ -2094,82 +2940,47 @@ static int ocfs2_xattr_block_set(struct inode *inode, struct ocfs2_xattr_set_ctxt *ctxt) { struct buffer_head *new_bh = NULL; - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; - handle_t *handle = ctxt->handle; struct ocfs2_xattr_block *xblk = NULL; - u16 suballoc_bit_start; - u32 num_got; - u64 first_blkno; int ret; + struct ocfs2_xa_loc loc; if (!xs->xattr_bh) { - ret = ocfs2_journal_access_di(handle, inode, xs->inode_bh, - OCFS2_JOURNAL_ACCESS_CREATE); - if (ret < 0) { - mlog_errno(ret); - goto end; - } - - ret = ocfs2_claim_metadata(osb, handle, ctxt->meta_ac, 1, - &suballoc_bit_start, &num_got, - &first_blkno); - if (ret < 0) { - mlog_errno(ret); - goto end; - } - - new_bh = sb_getblk(inode->i_sb, first_blkno); - ocfs2_set_new_buffer_uptodate(inode, new_bh); - - ret = ocfs2_journal_access_xb(handle, inode, new_bh, - OCFS2_JOURNAL_ACCESS_CREATE); - if (ret < 0) { + ret = ocfs2_create_xattr_block(inode, xs->inode_bh, ctxt, + 0, &new_bh); + if (ret) { mlog_errno(ret); goto end; } - /* Initialize ocfs2_xattr_block */ xs->xattr_bh = new_bh; - xblk = (struct ocfs2_xattr_block *)new_bh->b_data; - memset(xblk, 0, inode->i_sb->s_blocksize); - strcpy((void *)xblk, OCFS2_XATTR_BLOCK_SIGNATURE); - xblk->xb_suballoc_slot = cpu_to_le16(osb->slot_num); - xblk->xb_suballoc_bit = cpu_to_le16(suballoc_bit_start); - xblk->xb_fs_generation = cpu_to_le32(osb->fs_generation); - xblk->xb_blkno = cpu_to_le64(first_blkno); - + xblk = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data; xs->header = &xblk->xb_attrs.xb_header; xs->base = (void *)xs->header; xs->end = (void *)xblk + inode->i_sb->s_blocksize; xs->here = xs->header->xh_entries; - - ret = ocfs2_journal_dirty(handle, new_bh); - if (ret < 0) { - mlog_errno(ret); - goto end; - } - di->i_xattr_loc = cpu_to_le64(first_blkno); - ocfs2_journal_dirty(handle, xs->inode_bh); } else xblk = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data; if (!(le16_to_cpu(xblk->xb_flags) & OCFS2_XATTR_INDEXED)) { - /* Set extended attribute into external block */ - ret = ocfs2_xattr_set_entry(inode, xi, xs, ctxt, - OCFS2_HAS_XATTR_FL); - if (!ret || ret != -ENOSPC) - goto end; + ocfs2_init_xattr_block_xa_loc(&loc, inode, xs->xattr_bh, + xs->not_found ? NULL : xs->here); - ret = ocfs2_xattr_create_index_block(inode, xs, ctxt); - if (ret) + ret = ocfs2_xa_set(&loc, xi, ctxt); + if (!ret) + xs->here = loc.xl_entry; + else if ((ret != -ENOSPC) || ctxt->set_abort) goto end; + else { + ret = ocfs2_xattr_create_index_block(inode, xs, ctxt); + if (ret) + goto end; + } } - ret = ocfs2_xattr_set_entry_index_block(inode, xi, xs, ctxt); + if (le16_to_cpu(xblk->xb_flags) & OCFS2_XATTR_INDEXED) + ret = ocfs2_xattr_set_entry_index_block(inode, xi, xs, ctxt); end: - return ret; } @@ -2178,7 +2989,6 @@ static int ocfs2_xattr_can_be_in_inode(struct inode *inode, struct ocfs2_xattr_info *xi, struct ocfs2_xattr_search *xs) { - u64 value_size; struct ocfs2_xattr_entry *last; int free, i; size_t min_offs = xs->end - xs->base; @@ -2201,13 +3011,7 @@ static int ocfs2_xattr_can_be_in_inode(struct inode *inode, BUG_ON(!xs->not_found); - if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) - value_size = OCFS2_XATTR_ROOT_SIZE; - else - value_size = OCFS2_XATTR_SIZE(xi->value_len); - - if (free >= sizeof(struct ocfs2_xattr_entry) + - OCFS2_XATTR_SIZE(strlen(xi->name)) + value_size) + if (free >= (sizeof(struct ocfs2_xattr_entry) + namevalue_size_xi(xi))) return 1; return 0; @@ -2231,7 +3035,7 @@ static int ocfs2_calc_xattr_set_need(struct inode *inode, char *base = NULL; int name_offset, name_len = 0; u32 new_clusters = ocfs2_clusters_for_bytes(inode->i_sb, - xi->value_len); + xi->xi_value_len); u64 value_size; /* @@ -2239,18 +3043,17 @@ static int ocfs2_calc_xattr_set_need(struct inode *inode, * No matter whether we replace an old one or add a new one, * we need this for writing. */ - if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) + if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) credits += new_clusters * ocfs2_clusters_to_blocks(inode->i_sb, 1); if (xis->not_found && xbs->not_found) { credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb); - if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) { + if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { clusters_add += new_clusters; credits += ocfs2_calc_extend_credits(inode->i_sb, - &def_xv.xv.xr_list, - new_clusters); + &def_xv.xv.xr_list); } goto meta_guess; @@ -2272,7 +3075,7 @@ static int ocfs2_calc_xattr_set_need(struct inode *inode, old_in_xb = 1; if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { - ret = ocfs2_xattr_bucket_get_name_value(inode, + ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, bucket_xh(xbs->bucket), i, &block_off, &name_offset); @@ -2291,7 +3094,7 @@ static int ocfs2_calc_xattr_set_need(struct inode *inode, * The credits for removing the value tree will be extended * by ocfs2_remove_extent itself. */ - if (!xi->value) { + if (!xi->xi_value) { if (!ocfs2_xattr_is_local(xe)) credits += ocfs2_remove_extent_credits(inode->i_sb); @@ -2315,13 +3118,12 @@ static int ocfs2_calc_xattr_set_need(struct inode *inode, if (!ocfs2_xattr_is_local(xe)) credits += ocfs2_calc_extend_credits( inode->i_sb, - &def_xv.xv.xr_list, - new_clusters); + &def_xv.xv.xr_list); goto out; } } - if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) { + if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { /* the new values will be stored outside. */ u32 old_clusters = 0; @@ -2341,9 +3143,7 @@ static int ocfs2_calc_xattr_set_need(struct inode *inode, meta_add += ocfs2_extend_meta_needed(&xv->xr_list); clusters_add += new_clusters - old_clusters; credits += ocfs2_calc_extend_credits(inode->i_sb, - &xv->xr_list, - new_clusters - - old_clusters); + &xv->xr_list); if (value_size >= OCFS2_XATTR_ROOT_SIZE) goto out; } @@ -2354,9 +3154,10 @@ static int ocfs2_calc_xattr_set_need(struct inode *inode, * value, we don't need any allocation, otherwise we have * to guess metadata allocation. */ - if ((ocfs2_xattr_is_local(xe) && value_size >= xi->value_len) || + if ((ocfs2_xattr_is_local(xe) && + (value_size >= xi->xi_value_len)) || (!ocfs2_xattr_is_local(xe) && - OCFS2_XATTR_ROOT_SIZE >= xi->value_len)) + OCFS2_XATTR_ROOT_SIZE >= xi->xi_value_len)) goto out; } @@ -2388,7 +3189,7 @@ meta_guess: &xb->xb_attrs.xb_root.xt_list; meta_add += ocfs2_extend_meta_needed(el); credits += ocfs2_calc_extend_credits(inode->i_sb, - el, 1); + el); } else credits += OCFS2_SUBALLOC_ALLOC + 1; @@ -2407,8 +3208,15 @@ meta_guess: clusters_add += 1; } } else { - meta_add += 1; credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS; + if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { + struct ocfs2_extent_list *el = &def_xv.xv.xr_list; + meta_add += ocfs2_extend_meta_needed(el); + credits += ocfs2_calc_extend_credits(inode->i_sb, + el); + } else { + meta_add += 1; + } } out: if (clusters_need) @@ -2427,6 +3235,7 @@ static int ocfs2_init_xattr_set_ctxt(struct inode *inode, struct ocfs2_xattr_search *xis, struct ocfs2_xattr_search *xbs, struct ocfs2_xattr_set_ctxt *ctxt, + int extra_meta, int *credits) { int clusters_add, meta_add, ret; @@ -2443,8 +3252,9 @@ static int ocfs2_init_xattr_set_ctxt(struct inode *inode, return ret; } - mlog(0, "Set xattr %s, reserve meta blocks = %d, clusters = %d, " - "credits = %d\n", xi->name, meta_add, clusters_add, *credits); + meta_add += extra_meta; + trace_ocfs2_init_xattr_set_ctxt(xi->xi_name, meta_add, + clusters_add, *credits); if (meta_add) { ret = ocfs2_reserve_new_metadata_blocks(osb, meta_add, @@ -2484,7 +3294,7 @@ static int __ocfs2_xattr_set_handle(struct inode *inode, { int ret = 0, credits, old_found; - if (!xi->value) { + if (!xi->xi_value) { /* Remove existing extended attribute */ if (!xis->not_found) ret = ocfs2_xattr_ibody_set(inode, xi, xis, ctxt); @@ -2498,8 +3308,8 @@ static int __ocfs2_xattr_set_handle(struct inode *inode, * If succeed and that extended attribute existing in * external block, then we will remove it. */ - xi->value = NULL; - xi->value_len = 0; + xi->xi_value = NULL; + xi->xi_value_len = 0; old_found = xis->not_found; xis->not_found = -ENODATA; @@ -2517,18 +3327,17 @@ static int __ocfs2_xattr_set_handle(struct inode *inode, goto out; } - ret = ocfs2_extend_trans(ctxt->handle, credits + - ctxt->handle->h_buffer_credits); + ret = ocfs2_extend_trans(ctxt->handle, credits); if (ret) { mlog_errno(ret); goto out; } ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt); - } else if (ret == -ENOSPC) { + } else if ((ret == -ENOSPC) && !ctxt->set_abort) { if (di->i_xattr_loc && !xbs->xattr_bh) { ret = ocfs2_xattr_block_find(inode, - xi->name_index, - xi->name, xbs); + xi->xi_name_index, + xi->xi_name, xbs); if (ret) goto out; @@ -2548,8 +3357,7 @@ static int __ocfs2_xattr_set_handle(struct inode *inode, goto out; } - ret = ocfs2_extend_trans(ctxt->handle, credits + - ctxt->handle->h_buffer_credits); + ret = ocfs2_extend_trans(ctxt->handle, credits); if (ret) { mlog_errno(ret); goto out; @@ -2567,8 +3375,8 @@ static int __ocfs2_xattr_set_handle(struct inode *inode, * If succeed and that extended attribute * existing in inode, we will remove it. */ - xi->value = NULL; - xi->value_len = 0; + xi->xi_value = NULL; + xi->xi_value_len = 0; xbs->not_found = -ENODATA; ret = ocfs2_calc_xattr_set_need(inode, di, @@ -2583,8 +3391,7 @@ static int __ocfs2_xattr_set_handle(struct inode *inode, goto out; } - ret = ocfs2_extend_trans(ctxt->handle, credits + - ctxt->handle->h_buffer_credits); + ret = ocfs2_extend_trans(ctxt->handle, credits); if (ret) { mlog_errno(ret); goto out; @@ -2597,7 +3404,7 @@ static int __ocfs2_xattr_set_handle(struct inode *inode, if (!ret) { /* Update inode ctime. */ - ret = ocfs2_journal_access_di(ctxt->handle, inode, + ret = ocfs2_journal_access_di(ctxt->handle, INODE_CACHE(inode), xis->inode_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { @@ -2634,10 +3441,11 @@ int ocfs2_xattr_set_handle(handle_t *handle, int ret; struct ocfs2_xattr_info xi = { - .name_index = name_index, - .name = name, - .value = value, - .value_len = value_len, + .xi_name_index = name_index, + .xi_name = name, + .xi_name_len = strlen(name), + .xi_value = value, + .xi_value_len = value_len, }; struct ocfs2_xattr_search xis = { @@ -2710,16 +3518,18 @@ int ocfs2_xattr_set(struct inode *inode, { struct buffer_head *di_bh = NULL; struct ocfs2_dinode *di; - int ret, credits; + int ret, credits, ref_meta = 0, ref_credits = 0; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); struct inode *tl_inode = osb->osb_tl_inode; - struct ocfs2_xattr_set_ctxt ctxt = { NULL, NULL, }; + struct ocfs2_xattr_set_ctxt ctxt = { NULL, NULL, NULL, }; + struct ocfs2_refcount_tree *ref_tree = NULL; struct ocfs2_xattr_info xi = { - .name_index = name_index, - .name = name, - .value = value, - .value_len = value_len, + .xi_name_index = name_index, + .xi_name = name, + .xi_name_len = strlen(name), + .xi_value = value, + .xi_value_len = value_len, }; struct ocfs2_xattr_search xis = { @@ -2754,7 +3564,7 @@ int ocfs2_xattr_set(struct inode *inode, down_write(&OCFS2_I(inode)->ip_xattr_sem); /* * Scan inode and external block to find the same name - * extended attribute and collect search infomation. + * extended attribute and collect search information. */ ret = ocfs2_xattr_ibody_find(inode, name_index, name, &xis); if (ret) @@ -2778,6 +3588,17 @@ int ocfs2_xattr_set(struct inode *inode, goto cleanup; } + /* Check whether the value is refcounted and do some preparation. */ + if (OCFS2_I(inode)->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL && + (!xis.not_found || !xbs.not_found)) { + ret = ocfs2_prepare_refcount_xattr(inode, di, &xi, + &xis, &xbs, &ref_tree, + &ref_meta, &ref_credits); + if (ret) { + mlog_errno(ret); + goto cleanup; + } + } mutex_lock(&tl_inode->i_mutex); @@ -2792,7 +3613,7 @@ int ocfs2_xattr_set(struct inode *inode, mutex_unlock(&tl_inode->i_mutex); ret = ocfs2_init_xattr_set_ctxt(inode, di, &xi, &xis, - &xbs, &ctxt, &credits); + &xbs, &ctxt, ref_meta, &credits); if (ret) { mlog_errno(ret); goto cleanup; @@ -2800,17 +3621,19 @@ int ocfs2_xattr_set(struct inode *inode, /* we need to update inode's ctime field, so add credit for it. */ credits += OCFS2_INODE_UPDATE_CREDITS; - ctxt.handle = ocfs2_start_trans(osb, credits); + ctxt.handle = ocfs2_start_trans(osb, credits + ref_credits); if (IS_ERR(ctxt.handle)) { ret = PTR_ERR(ctxt.handle); mlog_errno(ret); - goto cleanup; + goto out_free_ac; } ret = __ocfs2_xattr_set_handle(inode, di, &xi, &xis, &xbs, &ctxt); + ocfs2_update_inode_fsync_trans(ctxt.handle, inode, 0); ocfs2_commit_trans(osb, ctxt.handle); +out_free_ac: if (ctxt.data_ac) ocfs2_free_alloc_context(ctxt.data_ac); if (ctxt.meta_ac) @@ -2818,8 +3641,16 @@ int ocfs2_xattr_set(struct inode *inode, if (ocfs2_dealloc_has_cluster(&ctxt.dealloc)) ocfs2_schedule_truncate_log_flush(osb, 1); ocfs2_run_deallocs(osb, &ctxt.dealloc); + cleanup: + if (ref_tree) + ocfs2_unlock_refcount_tree(osb, ref_tree, 1); up_write(&OCFS2_I(inode)->ip_xattr_sem); + if (!value && !ret) { + ret = ocfs2_try_remove_refcount_tree(inode, di_bh); + if (ret) + mlog_errno(ret); + } ocfs2_inode_unlock(inode, 1); cleanup_nolock: brelse(di_bh); @@ -2848,7 +3679,8 @@ static int ocfs2_xattr_get_rec(struct inode *inode, u64 e_blkno = 0; if (el->l_tree_depth) { - ret = ocfs2_find_leaf(inode, el, name_hash, &eb_bh); + ret = ocfs2_find_leaf(INODE_CACHE(inode), el, name_hash, + &eb_bh); if (ret) { mlog_errno(ret); goto out; @@ -2930,7 +3762,7 @@ static int ocfs2_find_xe_in_bucket(struct inode *inode, if (cmp) continue; - ret = ocfs2_xattr_bucket_get_name_value(inode, + ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, xh, i, &block_off, @@ -3064,8 +3896,10 @@ static int ocfs2_xattr_bucket_find(struct inode *inode, if (found) { xs->here = &xs->header->xh_entries[index]; - mlog(0, "find xattr %s in bucket %llu, entry = %u\n", name, - (unsigned long long)bucket_blkno(xs->bucket), index); + trace_ocfs2_xattr_bucket_find(OCFS2_I(inode)->ip_blkno, + name, name_index, name_hash, + (unsigned long long)bucket_blkno(xs->bucket), + index); } else ret = -ENODATA; @@ -3092,8 +3926,10 @@ static int ocfs2_xattr_index_block_find(struct inode *inode, if (le16_to_cpu(el->l_next_free_rec) == 0) return -ENODATA; - mlog(0, "find xattr %s, hash = %u, index = %d in xattr tree\n", - name, name_hash, name_index); + trace_ocfs2_xattr_index_block_find(OCFS2_I(inode)->ip_blkno, + name, name_index, name_hash, + (unsigned long long)root_bh->b_blocknr, + -1); ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, &first_hash, &num_clusters, el); @@ -3104,9 +3940,10 @@ static int ocfs2_xattr_index_block_find(struct inode *inode, BUG_ON(p_blkno == 0 || num_clusters == 0 || first_hash > name_hash); - mlog(0, "find xattr extent rec %u clusters from %llu, the first hash " - "in the rec is %u\n", num_clusters, (unsigned long long)p_blkno, - first_hash); + trace_ocfs2_xattr_index_block_find_rec(OCFS2_I(inode)->ip_blkno, + name, name_index, first_hash, + (unsigned long long)p_blkno, + num_clusters); ret = ocfs2_xattr_bucket_find(inode, name_index, name, name_hash, p_blkno, first_hash, num_clusters, xs); @@ -3132,8 +3969,9 @@ static int ocfs2_iterate_xattr_buckets(struct inode *inode, return -ENOMEM; } - mlog(0, "iterating xattr buckets in %u clusters starting from %llu\n", - clusters, (unsigned long long)blkno); + trace_ocfs2_iterate_xattr_buckets( + (unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)blkno, clusters); for (i = 0; i < num_buckets; i++, blkno += bucket->bu_blocks) { ret = ocfs2_read_xattr_bucket(bucket, blkno); @@ -3149,12 +3987,11 @@ static int ocfs2_iterate_xattr_buckets(struct inode *inode, if (i == 0) num_buckets = le16_to_cpu(bucket_xh(bucket)->xh_num_buckets); - mlog(0, "iterating xattr bucket %llu, first hash %u\n", - (unsigned long long)blkno, + trace_ocfs2_iterate_xattr_bucket((unsigned long long)blkno, le32_to_cpu(bucket_xh(bucket)->xh_entries[0].xe_name_hash)); if (func) { ret = func(inode, bucket, para); - if (ret) + if (ret && ret != -ERANGE) mlog_errno(ret); /* Fall through to bucket_relse() */ } @@ -3174,7 +4011,7 @@ struct ocfs2_xattr_tree_list { size_t result; }; -static int ocfs2_xattr_bucket_get_name_value(struct inode *inode, +static int ocfs2_xattr_bucket_get_name_value(struct super_block *sb, struct ocfs2_xattr_header *xh, int index, int *block_off, @@ -3187,8 +4024,8 @@ static int ocfs2_xattr_bucket_get_name_value(struct inode *inode, name_offset = le16_to_cpu(xh->xh_entries[index].xe_name_offset); - *block_off = name_offset >> inode->i_sb->s_blocksize_bits; - *new_offset = name_offset % inode->i_sb->s_blocksize; + *block_off = name_offset >> sb->s_blocksize_bits; + *new_offset = name_offset % sb->s_blocksize; return 0; } @@ -3208,7 +4045,7 @@ static int ocfs2_list_xattr_bucket(struct inode *inode, prefix = ocfs2_xattr_prefix(type); if (prefix) { - ret = ocfs2_xattr_bucket_get_name_value(inode, + ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, bucket_xh(bucket), i, &block_off, @@ -3231,22 +4068,19 @@ static int ocfs2_list_xattr_bucket(struct inode *inode, return ret; } -static int ocfs2_xattr_tree_list_index_block(struct inode *inode, - struct ocfs2_xattr_tree_root *xt, - char *buffer, - size_t buffer_size) +static int ocfs2_iterate_xattr_index_block(struct inode *inode, + struct buffer_head *blk_bh, + xattr_tree_rec_func *rec_func, + void *para) { - struct ocfs2_extent_list *el = &xt->xt_list; + struct ocfs2_xattr_block *xb = + (struct ocfs2_xattr_block *)blk_bh->b_data; + struct ocfs2_extent_list *el = &xb->xb_attrs.xb_root.xt_list; int ret = 0; u32 name_hash = UINT_MAX, e_cpos = 0, num_clusters = 0; u64 p_blkno = 0; - struct ocfs2_xattr_tree_list xl = { - .buffer = buffer, - .buffer_size = buffer_size, - .result = 0, - }; - if (le16_to_cpu(el->l_next_free_rec) == 0) + if (!el->l_next_free_rec || !rec_func) return 0; while (name_hash > 0) { @@ -3254,15 +4088,15 @@ static int ocfs2_xattr_tree_list_index_block(struct inode *inode, &e_cpos, &num_clusters, el); if (ret) { mlog_errno(ret); - goto out; + break; } - ret = ocfs2_iterate_xattr_buckets(inode, p_blkno, num_clusters, - ocfs2_list_xattr_bucket, - &xl); + ret = rec_func(inode, blk_bh, p_blkno, e_cpos, + num_clusters, para); if (ret) { - mlog_errno(ret); - goto out; + if (ret != -ERANGE) + mlog_errno(ret); + break; } if (e_cpos == 0) @@ -3271,6 +4105,37 @@ static int ocfs2_xattr_tree_list_index_block(struct inode *inode, name_hash = e_cpos - 1; } + return ret; + +} + +static int ocfs2_list_xattr_tree_rec(struct inode *inode, + struct buffer_head *root_bh, + u64 blkno, u32 cpos, u32 len, void *para) +{ + return ocfs2_iterate_xattr_buckets(inode, blkno, len, + ocfs2_list_xattr_bucket, para); +} + +static int ocfs2_xattr_tree_list_index_block(struct inode *inode, + struct buffer_head *blk_bh, + char *buffer, + size_t buffer_size) +{ + int ret; + struct ocfs2_xattr_tree_list xl = { + .buffer = buffer, + .buffer_size = buffer_size, + .result = 0, + }; + + ret = ocfs2_iterate_xattr_index_block(inode, blk_bh, + ocfs2_list_xattr_tree_rec, &xl); + if (ret) { + mlog_errno(ret); + goto out; + } + ret = xl.result; out: return ret; @@ -3322,9 +4187,9 @@ static void ocfs2_cp_xattr_block_to_bucket(struct inode *inode, char *src = xb_bh->b_data; char *target = bucket_block(bucket, blks - 1); - mlog(0, "cp xattr from block %llu to bucket %llu\n", - (unsigned long long)xb_bh->b_blocknr, - (unsigned long long)bucket_blkno(bucket)); + trace_ocfs2_cp_xattr_block_to_bucket_begin( + (unsigned long long)xb_bh->b_blocknr, + (unsigned long long)bucket_blkno(bucket)); for (i = 0; i < blks; i++) memset(bucket_block(bucket, i), 0, blocksize); @@ -3360,8 +4225,7 @@ static void ocfs2_cp_xattr_block_to_bucket(struct inode *inode, for (i = 0; i < count; i++) le16_add_cpu(&xh->xh_entries[i].xe_name_offset, off_change); - mlog(0, "copy entry: start = %u, size = %u, offset_change = %u\n", - offset, size, off_change); + trace_ocfs2_cp_xattr_block_to_bucket_end(offset, size, off_change); sort(target + offset, count, sizeof(struct ocfs2_xattr_entry), cmp_xe, swap_xe); @@ -3403,7 +4267,6 @@ static int ocfs2_xattr_create_index_block(struct inode *inode, u32 bit_off, len; u64 blkno; handle_t *handle = ctxt->handle; - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); struct ocfs2_inode_info *oi = OCFS2_I(inode); struct buffer_head *xb_bh = xs->xattr_bh; struct ocfs2_xattr_block *xb = @@ -3411,8 +4274,8 @@ static int ocfs2_xattr_create_index_block(struct inode *inode, struct ocfs2_xattr_tree_root *xr; u16 xb_flags = le16_to_cpu(xb->xb_flags); - mlog(0, "create xattr index block for %llu\n", - (unsigned long long)xb_bh->b_blocknr); + trace_ocfs2_xattr_create_index_block_begin( + (unsigned long long)xb_bh->b_blocknr); BUG_ON(xb_flags & OCFS2_XATTR_INDEXED); BUG_ON(!xs->bucket); @@ -3424,14 +4287,14 @@ static int ocfs2_xattr_create_index_block(struct inode *inode, */ down_write(&oi->ip_alloc_sem); - ret = ocfs2_journal_access_xb(handle, inode, xb_bh, + ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode), xb_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); goto out; } - ret = __ocfs2_claim_clusters(osb, handle, ctxt->data_ac, + ret = __ocfs2_claim_clusters(handle, ctxt->data_ac, 1, 1, &bit_off, &len); if (ret) { mlog_errno(ret); @@ -3445,10 +4308,9 @@ static int ocfs2_xattr_create_index_block(struct inode *inode, */ blkno = ocfs2_clusters_to_blocks(inode->i_sb, bit_off); - mlog(0, "allocate 1 cluster from %llu to xattr block\n", - (unsigned long long)blkno); + trace_ocfs2_xattr_create_index_block((unsigned long long)blkno); - ret = ocfs2_init_xattr_bucket(xs->bucket, blkno); + ret = ocfs2_init_xattr_bucket(xs->bucket, blkno, 1); if (ret) { mlog_errno(ret); goto out; @@ -3515,7 +4377,7 @@ static int ocfs2_defrag_xattr_bucket(struct inode *inode, struct ocfs2_xattr_bucket *bucket) { int ret, i; - size_t end, offset, len, value_len; + size_t end, offset, len; struct ocfs2_xattr_header *xh; char *entries, *buf, *bucket_buf = NULL; u64 blkno = bucket_blkno(bucket); @@ -3550,8 +4412,7 @@ static int ocfs2_defrag_xattr_bucket(struct inode *inode, entries = (char *)xh->xh_entries; xh_free_start = le16_to_cpu(xh->xh_free_start); - mlog(0, "adjust xattr bucket in %llu, count = %u, " - "xh_free_start = %u, xh_name_value_len = %u.\n", + trace_ocfs2_defrag_xattr_bucket( (unsigned long long)blkno, le16_to_cpu(xh->xh_count), xh_free_start, le16_to_cpu(xh->xh_name_value_len)); @@ -3569,12 +4430,7 @@ static int ocfs2_defrag_xattr_bucket(struct inode *inode, end = OCFS2_XATTR_BUCKET_SIZE; for (i = 0; i < le16_to_cpu(xh->xh_count); i++, xe++) { offset = le16_to_cpu(xe->xe_name_offset); - if (ocfs2_xattr_is_local(xe)) - value_len = OCFS2_XATTR_SIZE( - le64_to_cpu(xe->xe_value_size)); - else - value_len = OCFS2_XATTR_ROOT_SIZE; - len = OCFS2_XATTR_SIZE(xe->xe_name_len) + value_len; + len = namevalue_size_xe(xe); /* * We must make sure that the name/value pair @@ -3658,8 +4514,9 @@ static int ocfs2_mv_xattr_bucket_cross_cluster(struct inode *inode, BUG_ON(le16_to_cpu(bucket_xh(first)->xh_num_buckets) < num_buckets); BUG_ON(OCFS2_XATTR_BUCKET_SIZE == OCFS2_SB(sb)->s_clustersize); - mlog(0, "move half of xattrs in cluster %llu to %llu\n", - (unsigned long long)last_cluster_blkno, (unsigned long long)new_blkno); + trace_ocfs2_mv_xattr_bucket_cross_cluster( + (unsigned long long)last_cluster_blkno, + (unsigned long long)new_blkno); ret = ocfs2_mv_xattr_buckets(inode, handle, bucket_blkno(first), last_cluster_blkno, new_blkno, @@ -3763,14 +4620,14 @@ static int ocfs2_divide_xattr_bucket(struct inode *inode, int new_bucket_head) { int ret, i; - int count, start, len, name_value_len = 0, xe_len, name_offset = 0; + int count, start, len, name_value_len = 0, name_offset = 0; struct ocfs2_xattr_bucket *s_bucket = NULL, *t_bucket = NULL; struct ocfs2_xattr_header *xh; struct ocfs2_xattr_entry *xe; int blocksize = inode->i_sb->s_blocksize; - mlog(0, "move some of xattrs from bucket %llu to %llu\n", - (unsigned long long)blk, (unsigned long long)new_blk); + trace_ocfs2_divide_xattr_bucket_begin((unsigned long long)blk, + (unsigned long long)new_blk); s_bucket = ocfs2_xattr_bucket_new(inode); t_bucket = ocfs2_xattr_bucket_new(inode); @@ -3797,7 +4654,7 @@ static int ocfs2_divide_xattr_bucket(struct inode *inode, * Even if !new_bucket_head, we're overwriting t_bucket. Thus, * there's no need to read it. */ - ret = ocfs2_init_xattr_bucket(t_bucket, new_blk); + ret = ocfs2_init_xattr_bucket(t_bucket, new_blk, new_bucket_head); if (ret) { mlog_errno(ret); goto out; @@ -3854,13 +4711,7 @@ static int ocfs2_divide_xattr_bucket(struct inode *inode, name_value_len = 0; for (i = 0; i < start; i++) { xe = &xh->xh_entries[i]; - xe_len = OCFS2_XATTR_SIZE(xe->xe_name_len); - if (ocfs2_xattr_is_local(xe)) - xe_len += - OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size)); - else - xe_len += OCFS2_XATTR_ROOT_SIZE; - name_value_len += xe_len; + name_value_len += namevalue_size_xe(xe); if (le16_to_cpu(xe->xe_name_offset) < name_offset) name_offset = le16_to_cpu(xe->xe_name_offset); } @@ -3875,9 +4726,9 @@ static int ocfs2_divide_xattr_bucket(struct inode *inode, */ xe = &xh->xh_entries[start]; len = sizeof(struct ocfs2_xattr_entry) * (count - start); - mlog(0, "mv xattr entry len %d from %d to %d\n", len, - (int)((char *)xe - (char *)xh), - (int)((char *)xh->xh_entries - (char *)xh)); + trace_ocfs2_divide_xattr_bucket_move(len, + (int)((char *)xe - (char *)xh), + (int)((char *)xh->xh_entries - (char *)xh)); memmove((char *)xh->xh_entries, (char *)xe, len); xe = &xh->xh_entries[count - start]; len = sizeof(struct ocfs2_xattr_entry) * start; @@ -3890,12 +4741,6 @@ static int ocfs2_divide_xattr_bucket(struct inode *inode, xh->xh_free_start = cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE); for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { xe = &xh->xh_entries[i]; - xe_len = OCFS2_XATTR_SIZE(xe->xe_name_len); - if (ocfs2_xattr_is_local(xe)) - xe_len += - OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size)); - else - xe_len += OCFS2_XATTR_ROOT_SIZE; if (le16_to_cpu(xe->xe_name_offset) < le16_to_cpu(xh->xh_free_start)) xh->xh_free_start = xe->xe_name_offset; @@ -3955,9 +4800,9 @@ static int ocfs2_cp_xattr_bucket(struct inode *inode, BUG_ON(s_blkno == t_blkno); - mlog(0, "cp bucket %llu to %llu, target is %d\n", - (unsigned long long)s_blkno, (unsigned long long)t_blkno, - t_is_new); + trace_ocfs2_cp_xattr_bucket((unsigned long long)s_blkno, + (unsigned long long)t_blkno, + t_is_new); s_bucket = ocfs2_xattr_bucket_new(inode); t_bucket = ocfs2_xattr_bucket_new(inode); @@ -3975,7 +4820,7 @@ static int ocfs2_cp_xattr_bucket(struct inode *inode, * Even if !t_is_new, we're overwriting t_bucket. Thus, * there's no need to read it. */ - ret = ocfs2_init_xattr_bucket(t_bucket, t_blkno); + ret = ocfs2_init_xattr_bucket(t_bucket, t_blkno, t_is_new); if (ret) goto out; @@ -4029,8 +4874,8 @@ static int ocfs2_mv_xattr_buckets(struct inode *inode, handle_t *handle, int num_buckets = ocfs2_xattr_buckets_per_cluster(osb); struct ocfs2_xattr_bucket *old_first, *new_first; - mlog(0, "mv xattrs from cluster %llu to %llu\n", - (unsigned long long)last_blk, (unsigned long long)to_blk); + trace_ocfs2_mv_xattr_buckets((unsigned long long)last_blk, + (unsigned long long)to_blk); BUG_ON(start_bucket >= num_buckets); if (start_bucket) { @@ -4058,8 +4903,7 @@ static int ocfs2_mv_xattr_buckets(struct inode *inode, handle_t *handle, * We need to update the first bucket of the old extent and all * the buckets going to the new extent. */ - credits = ((num_buckets + 1) * blks_per_bucket) + - handle->h_buffer_credits; + credits = ((num_buckets + 1) * blks_per_bucket); ret = ocfs2_extend_trans(handle, credits); if (ret) { mlog_errno(ret); @@ -4129,7 +4973,7 @@ static int ocfs2_divide_xattr_cluster(struct inode *inode, u32 *first_hash) { u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); - int ret, credits = 2 * blk_per_bucket + handle->h_buffer_credits; + int ret, credits = 2 * blk_per_bucket; BUG_ON(OCFS2_XATTR_BUCKET_SIZE < OCFS2_SB(inode->i_sb)->s_clustersize); @@ -4181,9 +5025,9 @@ static int ocfs2_adjust_xattr_cross_cluster(struct inode *inode, { int ret; - mlog(0, "adjust xattrs from cluster %llu len %u to %llu\n", - (unsigned long long)bucket_blkno(first), prev_clusters, - (unsigned long long)new_blk); + trace_ocfs2_adjust_xattr_cross_cluster( + (unsigned long long)bucket_blkno(first), + (unsigned long long)new_blk, prev_clusters); if (ocfs2_xattr_buckets_per_cluster(OCFS2_SB(inode->i_sb)) > 1) { ret = ocfs2_mv_xattr_bucket_cross_cluster(inode, @@ -4256,21 +5100,21 @@ static int ocfs2_add_new_xattr_cluster(struct inode *inode, struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); struct ocfs2_extent_tree et; - mlog(0, "Add new xattr cluster for %llu, previous xattr hash = %u, " - "previous xattr blkno = %llu\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, - prev_cpos, (unsigned long long)bucket_blkno(first)); + trace_ocfs2_add_new_xattr_cluster_begin( + (unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)bucket_blkno(first), + prev_cpos, prev_clusters); - ocfs2_init_xattr_tree_extent_tree(&et, inode, root_bh); + ocfs2_init_xattr_tree_extent_tree(&et, INODE_CACHE(inode), root_bh); - ret = ocfs2_journal_access_xb(handle, inode, root_bh, + ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode), root_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret < 0) { mlog_errno(ret); goto leave; } - ret = __ocfs2_claim_clusters(osb, handle, ctxt->data_ac, 1, + ret = __ocfs2_claim_clusters(handle, ctxt->data_ac, 1, clusters_to_add, &bit_off, &num_bits); if (ret < 0) { if (ret != -ENOSPC) @@ -4281,8 +5125,7 @@ static int ocfs2_add_new_xattr_cluster(struct inode *inode, BUG_ON(num_bits > clusters_to_add); block = ocfs2_clusters_to_blocks(osb->sb, bit_off); - mlog(0, "Allocating %u clusters at block %u for xattr in inode %llu\n", - num_bits, bit_off, (unsigned long long)OCFS2_I(inode)->ip_blkno); + trace_ocfs2_add_new_xattr_cluster((unsigned long long)block, num_bits); if (bucket_blkno(first) + (prev_clusters * bpc) == block && (prev_clusters + num_bits) << osb->s_clustersize_bits <= @@ -4298,8 +5141,6 @@ static int ocfs2_add_new_xattr_cluster(struct inode *inode, */ v_start = prev_cpos + prev_clusters; *num_clusters = prev_clusters + num_bits; - mlog(0, "Add contiguous %u clusters to previous extent rec.\n", - num_bits); } else { ret = ocfs2_adjust_xattr_cross_cluster(inode, handle, @@ -4315,18 +5156,16 @@ static int ocfs2_add_new_xattr_cluster(struct inode *inode, } } - mlog(0, "Insert %u clusters at block %llu for xattr at %u\n", - num_bits, (unsigned long long)block, v_start); - ret = ocfs2_insert_extent(osb, handle, inode, &et, v_start, block, + trace_ocfs2_add_new_xattr_cluster_insert((unsigned long long)block, + v_start, num_bits); + ret = ocfs2_insert_extent(handle, &et, v_start, block, num_bits, 0, ctxt->meta_ac); if (ret < 0) { mlog_errno(ret); goto leave; } - ret = ocfs2_journal_dirty(handle, root_bh); - if (ret < 0) - mlog_errno(ret); + ocfs2_journal_dirty(handle, root_bh); leave: return ret; @@ -4353,9 +5192,9 @@ static int ocfs2_extend_xattr_bucket(struct inode *inode, u64 end_blk; u16 new_bucket = le16_to_cpu(bucket_xh(first)->xh_num_buckets); - mlog(0, "extend xattr bucket in %llu, xattr extend rec starting " - "from %llu, len = %u\n", (unsigned long long)target_blk, - (unsigned long long)bucket_blkno(first), num_clusters); + trace_ocfs2_extend_xattr_bucket((unsigned long long)target_blk, + (unsigned long long)bucket_blkno(first), + num_clusters, new_bucket); /* The extent must have room for an additional bucket */ BUG_ON(new_bucket >= @@ -4371,8 +5210,7 @@ static int ocfs2_extend_xattr_bucket(struct inode *inode, * existing bucket. Then we add the last existing bucket, the * new bucket, and the first bucket (3 * blk_per_bucket). */ - credits = (end_blk - target_blk) + (3 * blk_per_bucket) + - handle->h_buffer_credits; + credits = (end_blk - target_blk) + (3 * blk_per_bucket); ret = ocfs2_extend_trans(handle, credits); if (ret) { mlog_errno(ret); @@ -4436,8 +5274,8 @@ static int ocfs2_add_new_xattr_bucket(struct inode *inode, /* The bucket at the front of the extent */ struct ocfs2_xattr_bucket *first; - mlog(0, "Add new xattr bucket starting from %llu\n", - (unsigned long long)bucket_blkno(target)); + trace_ocfs2_add_new_xattr_bucket( + (unsigned long long)bucket_blkno(target)); /* The first bucket of the original extent */ first = ocfs2_xattr_bucket_new(inode); @@ -4507,195 +5345,6 @@ static inline char *ocfs2_xattr_bucket_get_val(struct inode *inode, } /* - * Handle the normal xattr set, including replace, delete and new. - * - * Note: "local" indicates the real data's locality. So we can't - * just its bucket locality by its length. - */ -static void ocfs2_xattr_set_entry_normal(struct inode *inode, - struct ocfs2_xattr_info *xi, - struct ocfs2_xattr_search *xs, - u32 name_hash, - int local) -{ - struct ocfs2_xattr_entry *last, *xe; - int name_len = strlen(xi->name); - struct ocfs2_xattr_header *xh = xs->header; - u16 count = le16_to_cpu(xh->xh_count), start; - size_t blocksize = inode->i_sb->s_blocksize; - char *val; - size_t offs, size, new_size; - - last = &xh->xh_entries[count]; - if (!xs->not_found) { - xe = xs->here; - offs = le16_to_cpu(xe->xe_name_offset); - if (ocfs2_xattr_is_local(xe)) - size = OCFS2_XATTR_SIZE(name_len) + - OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size)); - else - size = OCFS2_XATTR_SIZE(name_len) + - OCFS2_XATTR_SIZE(OCFS2_XATTR_ROOT_SIZE); - - /* - * If the new value will be stored outside, xi->value has been - * initalized as an empty ocfs2_xattr_value_root, and the same - * goes with xi->value_len, so we can set new_size safely here. - * See ocfs2_xattr_set_in_bucket. - */ - new_size = OCFS2_XATTR_SIZE(name_len) + - OCFS2_XATTR_SIZE(xi->value_len); - - le16_add_cpu(&xh->xh_name_value_len, -size); - if (xi->value) { - if (new_size > size) - goto set_new_name_value; - - /* Now replace the old value with new one. */ - if (local) - xe->xe_value_size = cpu_to_le64(xi->value_len); - else - xe->xe_value_size = 0; - - val = ocfs2_xattr_bucket_get_val(inode, - xs->bucket, offs); - memset(val + OCFS2_XATTR_SIZE(name_len), 0, - size - OCFS2_XATTR_SIZE(name_len)); - if (OCFS2_XATTR_SIZE(xi->value_len) > 0) - memcpy(val + OCFS2_XATTR_SIZE(name_len), - xi->value, xi->value_len); - - le16_add_cpu(&xh->xh_name_value_len, new_size); - ocfs2_xattr_set_local(xe, local); - return; - } else { - /* - * Remove the old entry if there is more than one. - * We don't remove the last entry so that we can - * use it to indicate the hash value of the empty - * bucket. - */ - last -= 1; - le16_add_cpu(&xh->xh_count, -1); - if (xh->xh_count) { - memmove(xe, xe + 1, - (void *)last - (void *)xe); - memset(last, 0, - sizeof(struct ocfs2_xattr_entry)); - } else - xh->xh_free_start = - cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE); - - return; - } - } else { - /* find a new entry for insert. */ - int low = 0, high = count - 1, tmp; - struct ocfs2_xattr_entry *tmp_xe; - - while (low <= high && count) { - tmp = (low + high) / 2; - tmp_xe = &xh->xh_entries[tmp]; - - if (name_hash > le32_to_cpu(tmp_xe->xe_name_hash)) - low = tmp + 1; - else if (name_hash < - le32_to_cpu(tmp_xe->xe_name_hash)) - high = tmp - 1; - else { - low = tmp; - break; - } - } - - xe = &xh->xh_entries[low]; - if (low != count) - memmove(xe + 1, xe, (void *)last - (void *)xe); - - le16_add_cpu(&xh->xh_count, 1); - memset(xe, 0, sizeof(struct ocfs2_xattr_entry)); - xe->xe_name_hash = cpu_to_le32(name_hash); - xe->xe_name_len = name_len; - ocfs2_xattr_set_type(xe, xi->name_index); - } - -set_new_name_value: - /* Insert the new name+value. */ - size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_SIZE(xi->value_len); - - /* - * We must make sure that the name/value pair - * exists in the same block. - */ - offs = le16_to_cpu(xh->xh_free_start); - start = offs - size; - - if (start >> inode->i_sb->s_blocksize_bits != - (offs - 1) >> inode->i_sb->s_blocksize_bits) { - offs = offs - offs % blocksize; - xh->xh_free_start = cpu_to_le16(offs); - } - - val = ocfs2_xattr_bucket_get_val(inode, xs->bucket, offs - size); - xe->xe_name_offset = cpu_to_le16(offs - size); - - memset(val, 0, size); - memcpy(val, xi->name, name_len); - memcpy(val + OCFS2_XATTR_SIZE(name_len), xi->value, xi->value_len); - - xe->xe_value_size = cpu_to_le64(xi->value_len); - ocfs2_xattr_set_local(xe, local); - xs->here = xe; - le16_add_cpu(&xh->xh_free_start, -size); - le16_add_cpu(&xh->xh_name_value_len, size); - - return; -} - -/* - * Set the xattr entry in the specified bucket. - * The bucket is indicated by xs->bucket and it should have the enough - * space for the xattr insertion. - */ -static int ocfs2_xattr_set_entry_in_bucket(struct inode *inode, - handle_t *handle, - struct ocfs2_xattr_info *xi, - struct ocfs2_xattr_search *xs, - u32 name_hash, - int local) -{ - int ret; - u64 blkno; - - mlog(0, "Set xattr entry len = %lu index = %d in bucket %llu\n", - (unsigned long)xi->value_len, xi->name_index, - (unsigned long long)bucket_blkno(xs->bucket)); - - if (!xs->bucket->bu_bhs[1]) { - blkno = bucket_blkno(xs->bucket); - ocfs2_xattr_bucket_relse(xs->bucket); - ret = ocfs2_read_xattr_bucket(xs->bucket, blkno); - if (ret) { - mlog_errno(ret); - goto out; - } - } - - ret = ocfs2_xattr_bucket_journal_access(handle, xs->bucket, - OCFS2_JOURNAL_ACCESS_WRITE); - if (ret < 0) { - mlog_errno(ret); - goto out; - } - - ocfs2_xattr_set_entry_normal(inode, xi, xs, name_hash, local); - ocfs2_xattr_bucket_journal_dirty(handle, xs->bucket); - -out: - return ret; -} - -/* * Truncate the specified xe_off entry in xattr bucket. * bucket is indicated by header_bh and len is the new length. * Both the ocfs2_xattr_value_root and the entry will be updated here. @@ -4742,8 +5391,8 @@ static int ocfs2_xattr_bucket_value_truncate(struct inode *inode, * modified something. We have to assume they did, and dirty * the whole bucket. This leaves us in a consistent state. */ - mlog(0, "truncate %u in xattr bucket %llu to %d bytes.\n", - xe_off, (unsigned long long)bucket_blkno(bucket), len); + trace_ocfs2_xattr_bucket_value_truncate( + (unsigned long long)bucket_blkno(bucket), xe_off, len); ret = ocfs2_xattr_value_truncate(inode, &vb, len, ctxt); if (ret) { mlog_errno(ret); @@ -4765,66 +5414,12 @@ out: return ret; } -static int ocfs2_xattr_bucket_value_truncate_xs(struct inode *inode, - struct ocfs2_xattr_search *xs, - int len, - struct ocfs2_xattr_set_ctxt *ctxt) -{ - int ret, offset; - struct ocfs2_xattr_entry *xe = xs->here; - struct ocfs2_xattr_header *xh = (struct ocfs2_xattr_header *)xs->base; - - BUG_ON(!xs->bucket->bu_bhs[0] || !xe || ocfs2_xattr_is_local(xe)); - - offset = xe - xh->xh_entries; - ret = ocfs2_xattr_bucket_value_truncate(inode, xs->bucket, - offset, len, ctxt); - if (ret) - mlog_errno(ret); - - return ret; -} - -static int ocfs2_xattr_bucket_set_value_outside(struct inode *inode, - handle_t *handle, - struct ocfs2_xattr_search *xs, - char *val, - int value_len) -{ - int ret, offset, block_off; - struct ocfs2_xattr_value_root *xv; - struct ocfs2_xattr_entry *xe = xs->here; - struct ocfs2_xattr_header *xh = bucket_xh(xs->bucket); - void *base; - - BUG_ON(!xs->base || !xe || ocfs2_xattr_is_local(xe)); - - ret = ocfs2_xattr_bucket_get_name_value(inode, xh, - xe - xh->xh_entries, - &block_off, - &offset); - if (ret) { - mlog_errno(ret); - goto out; - } - - base = bucket_block(xs->bucket, block_off); - xv = (struct ocfs2_xattr_value_root *)(base + offset + - OCFS2_XATTR_SIZE(xe->xe_name_len)); - - ret = __ocfs2_xattr_set_value_outside(inode, handle, - xv, val, value_len); - if (ret) - mlog_errno(ret); -out: - return ret; -} - static int ocfs2_rm_xattr_cluster(struct inode *inode, struct buffer_head *root_bh, u64 blkno, u32 cpos, - u32 len) + u32 len, + void *para) { int ret; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); @@ -4836,14 +5431,23 @@ static int ocfs2_rm_xattr_cluster(struct inode *inode, struct ocfs2_cached_dealloc_ctxt dealloc; struct ocfs2_extent_tree et; - ocfs2_init_xattr_tree_extent_tree(&et, inode, root_bh); + ret = ocfs2_iterate_xattr_buckets(inode, blkno, len, + ocfs2_delete_xattr_in_bucket, para); + if (ret) { + mlog_errno(ret); + return ret; + } + + ocfs2_init_xattr_tree_extent_tree(&et, INODE_CACHE(inode), root_bh); ocfs2_init_dealloc_ctxt(&dealloc); - mlog(0, "rm xattr extent rec at %u len = %u, start from %llu\n", - cpos, len, (unsigned long long)blkno); + trace_ocfs2_rm_xattr_cluster( + (unsigned long long)OCFS2_I(inode)->ip_blkno, + (unsigned long long)blkno, cpos, len); - ocfs2_remove_xattr_clusters_from_cache(inode, blkno, len); + ocfs2_remove_xattr_clusters_from_cache(INODE_CACHE(inode), blkno, + len); ret = ocfs2_lock_allocators(inode, &et, 0, 1, NULL, &meta_ac); if (ret) { @@ -4868,14 +5472,14 @@ static int ocfs2_rm_xattr_cluster(struct inode *inode, goto out; } - ret = ocfs2_journal_access_xb(handle, inode, root_bh, + ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode), root_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); goto out_commit; } - ret = ocfs2_remove_extent(inode, &et, cpos, len, handle, meta_ac, + ret = ocfs2_remove_extent(handle, &et, cpos, len, meta_ac, &dealloc); if (ret) { mlog_errno(ret); @@ -4883,16 +5487,12 @@ static int ocfs2_rm_xattr_cluster(struct inode *inode, } le32_add_cpu(&xb->xb_attrs.xb_root.xt_clusters, -len); - - ret = ocfs2_journal_dirty(handle, root_bh); - if (ret) { - mlog_errno(ret); - goto out_commit; - } + ocfs2_journal_dirty(handle, root_bh); ret = ocfs2_truncate_log_append(osb, handle, blkno, len); if (ret) mlog_errno(ret); + ocfs2_update_inode_fsync_trans(handle, inode, 0); out_commit: ocfs2_commit_trans(osb, handle); @@ -4909,394 +5509,1689 @@ out: return ret; } -static void ocfs2_xattr_bucket_remove_xs(struct inode *inode, - handle_t *handle, - struct ocfs2_xattr_search *xs) +/* + * check whether the xattr bucket is filled up with the same hash value. + * If we want to insert the xattr with the same hash, return -ENOSPC. + * If we want to insert a xattr with different hash value, go ahead + * and ocfs2_divide_xattr_bucket will handle this. + */ +static int ocfs2_check_xattr_bucket_collision(struct inode *inode, + struct ocfs2_xattr_bucket *bucket, + const char *name) { - struct ocfs2_xattr_header *xh = bucket_xh(xs->bucket); - struct ocfs2_xattr_entry *last = &xh->xh_entries[ - le16_to_cpu(xh->xh_count) - 1]; - int ret = 0; + struct ocfs2_xattr_header *xh = bucket_xh(bucket); + u32 name_hash = ocfs2_xattr_name_hash(inode, name, strlen(name)); - ret = ocfs2_xattr_bucket_journal_access(handle, xs->bucket, - OCFS2_JOURNAL_ACCESS_WRITE); + if (name_hash != le32_to_cpu(xh->xh_entries[0].xe_name_hash)) + return 0; + + if (xh->xh_entries[le16_to_cpu(xh->xh_count) - 1].xe_name_hash == + xh->xh_entries[0].xe_name_hash) { + mlog(ML_ERROR, "Too much hash collision in xattr bucket %llu, " + "hash = %u\n", + (unsigned long long)bucket_blkno(bucket), + le32_to_cpu(xh->xh_entries[0].xe_name_hash)); + return -ENOSPC; + } + + return 0; +} + +/* + * Try to set the entry in the current bucket. If we fail, the caller + * will handle getting us another bucket. + */ +static int ocfs2_xattr_set_entry_bucket(struct inode *inode, + struct ocfs2_xattr_info *xi, + struct ocfs2_xattr_search *xs, + struct ocfs2_xattr_set_ctxt *ctxt) +{ + int ret; + struct ocfs2_xa_loc loc; + + trace_ocfs2_xattr_set_entry_bucket(xi->xi_name); + + ocfs2_init_xattr_bucket_xa_loc(&loc, xs->bucket, + xs->not_found ? NULL : xs->here); + ret = ocfs2_xa_set(&loc, xi, ctxt); + if (!ret) { + xs->here = loc.xl_entry; + goto out; + } + if (ret != -ENOSPC) { + mlog_errno(ret); + goto out; + } + + /* Ok, we need space. Let's try defragmenting the bucket. */ + ret = ocfs2_defrag_xattr_bucket(inode, ctxt->handle, + xs->bucket); if (ret) { mlog_errno(ret); - return; + goto out; } - /* Remove the old entry. */ - memmove(xs->here, xs->here + 1, - (void *)last - (void *)xs->here); - memset(last, 0, sizeof(struct ocfs2_xattr_entry)); - le16_add_cpu(&xh->xh_count, -1); + ret = ocfs2_xa_set(&loc, xi, ctxt); + if (!ret) { + xs->here = loc.xl_entry; + goto out; + } + if (ret != -ENOSPC) + mlog_errno(ret); - ocfs2_xattr_bucket_journal_dirty(handle, xs->bucket); + +out: + return ret; +} + +static int ocfs2_xattr_set_entry_index_block(struct inode *inode, + struct ocfs2_xattr_info *xi, + struct ocfs2_xattr_search *xs, + struct ocfs2_xattr_set_ctxt *ctxt) +{ + int ret; + + trace_ocfs2_xattr_set_entry_index_block(xi->xi_name); + + ret = ocfs2_xattr_set_entry_bucket(inode, xi, xs, ctxt); + if (!ret) + goto out; + if (ret != -ENOSPC) { + mlog_errno(ret); + goto out; + } + + /* Ack, need more space. Let's try to get another bucket! */ + + /* + * We do not allow for overlapping ranges between buckets. And + * the maximum number of collisions we will allow for then is + * one bucket's worth, so check it here whether we need to + * add a new bucket for the insert. + */ + ret = ocfs2_check_xattr_bucket_collision(inode, + xs->bucket, + xi->xi_name); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_add_new_xattr_bucket(inode, + xs->xattr_bh, + xs->bucket, + ctxt); + if (ret) { + mlog_errno(ret); + goto out; + } + + /* + * ocfs2_add_new_xattr_bucket() will have updated + * xs->bucket if it moved, but it will not have updated + * any of the other search fields. Thus, we drop it and + * re-search. Everything should be cached, so it'll be + * quick. + */ + ocfs2_xattr_bucket_relse(xs->bucket); + ret = ocfs2_xattr_index_block_find(inode, xs->xattr_bh, + xi->xi_name_index, + xi->xi_name, xs); + if (ret && ret != -ENODATA) + goto out; + xs->not_found = ret; + + /* Ok, we have a new bucket, let's try again */ + ret = ocfs2_xattr_set_entry_bucket(inode, xi, xs, ctxt); + if (ret && (ret != -ENOSPC)) + mlog_errno(ret); + +out: + return ret; +} + +static int ocfs2_delete_xattr_in_bucket(struct inode *inode, + struct ocfs2_xattr_bucket *bucket, + void *para) +{ + int ret = 0, ref_credits; + struct ocfs2_xattr_header *xh = bucket_xh(bucket); + u16 i; + struct ocfs2_xattr_entry *xe; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_xattr_set_ctxt ctxt = {NULL, NULL,}; + int credits = ocfs2_remove_extent_credits(osb->sb) + + ocfs2_blocks_per_xattr_bucket(inode->i_sb); + struct ocfs2_xattr_value_root *xv; + struct ocfs2_rm_xattr_bucket_para *args = + (struct ocfs2_rm_xattr_bucket_para *)para; + + ocfs2_init_dealloc_ctxt(&ctxt.dealloc); + + for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { + xe = &xh->xh_entries[i]; + if (ocfs2_xattr_is_local(xe)) + continue; + + ret = ocfs2_get_xattr_tree_value_root(inode->i_sb, bucket, + i, &xv, NULL); + + ret = ocfs2_lock_xattr_remove_allocators(inode, xv, + args->ref_ci, + args->ref_root_bh, + &ctxt.meta_ac, + &ref_credits); + + ctxt.handle = ocfs2_start_trans(osb, credits + ref_credits); + if (IS_ERR(ctxt.handle)) { + ret = PTR_ERR(ctxt.handle); + mlog_errno(ret); + break; + } + + ret = ocfs2_xattr_bucket_value_truncate(inode, bucket, + i, 0, &ctxt); + + ocfs2_commit_trans(osb, ctxt.handle); + if (ctxt.meta_ac) { + ocfs2_free_alloc_context(ctxt.meta_ac); + ctxt.meta_ac = NULL; + } + if (ret) { + mlog_errno(ret); + break; + } + } + + if (ctxt.meta_ac) + ocfs2_free_alloc_context(ctxt.meta_ac); + ocfs2_schedule_truncate_log_flush(osb, 1); + ocfs2_run_deallocs(osb, &ctxt.dealloc); + return ret; } /* - * Set the xattr name/value in the bucket specified in xs. + * Whenever we modify a xattr value root in the bucket(e.g, CoW + * or change the extent record flag), we need to recalculate + * the metaecc for the whole bucket. So it is done here. * - * As the new value in xi may be stored in the bucket or in an outside cluster, - * we divide the whole process into 3 steps: - * 1. insert name/value in the bucket(ocfs2_xattr_set_entry_in_bucket) - * 2. truncate of the outside cluster(ocfs2_xattr_bucket_value_truncate_xs) - * 3. Set the value to the outside cluster(ocfs2_xattr_bucket_set_value_outside) - * 4. If the clusters for the new outside value can't be allocated, we need - * to free the xattr we allocated in set. + * Note: + * We have to give the extra credits for the caller. */ -static int ocfs2_xattr_set_in_bucket(struct inode *inode, - struct ocfs2_xattr_info *xi, - struct ocfs2_xattr_search *xs, - struct ocfs2_xattr_set_ctxt *ctxt) +static int ocfs2_xattr_bucket_post_refcount(struct inode *inode, + handle_t *handle, + void *para) { - int ret, local = 1; - size_t value_len; - char *val = (char *)xi->value; - struct ocfs2_xattr_entry *xe = xs->here; - u32 name_hash = ocfs2_xattr_name_hash(inode, xi->name, - strlen(xi->name)); + int ret; + struct ocfs2_xattr_bucket *bucket = + (struct ocfs2_xattr_bucket *)para; - if (!xs->not_found && !ocfs2_xattr_is_local(xe)) { - /* - * We need to truncate the xattr storage first. - * - * If both the old and new value are stored to - * outside block, we only need to truncate - * the storage and then set the value outside. - * - * If the new value should be stored within block, - * we should free all the outside block first and - * the modification to the xattr block will be done - * by following steps. - */ - if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) - value_len = xi->value_len; - else - value_len = 0; + ret = ocfs2_xattr_bucket_journal_access(handle, bucket, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + return ret; + } - ret = ocfs2_xattr_bucket_value_truncate_xs(inode, xs, - value_len, - ctxt); - if (ret) - goto out; + ocfs2_xattr_bucket_journal_dirty(handle, bucket); - if (value_len) - goto set_value_outside; - } + return 0; +} - value_len = xi->value_len; - /* So we have to handle the inside block change now. */ - if (value_len > OCFS2_XATTR_INLINE_SIZE) { - /* - * If the new value will be stored outside of block, - * initalize a new empty value root and insert it first. - */ - local = 0; - xi->value = &def_xv; - xi->value_len = OCFS2_XATTR_ROOT_SIZE; +/* + * Special action we need if the xattr value is refcounted. + * + * 1. If the xattr is refcounted, lock the tree. + * 2. CoW the xattr if we are setting the new value and the value + * will be stored outside. + * 3. In other case, decrease_refcount will work for us, so just + * lock the refcount tree, calculate the meta and credits is OK. + * + * We have to do CoW before ocfs2_init_xattr_set_ctxt since + * currently CoW is a completed transaction, while this function + * will also lock the allocators and let us deadlock. So we will + * CoW the whole xattr value. + */ +static int ocfs2_prepare_refcount_xattr(struct inode *inode, + struct ocfs2_dinode *di, + struct ocfs2_xattr_info *xi, + struct ocfs2_xattr_search *xis, + struct ocfs2_xattr_search *xbs, + struct ocfs2_refcount_tree **ref_tree, + int *meta_add, + int *credits) +{ + int ret = 0; + struct ocfs2_xattr_block *xb; + struct ocfs2_xattr_entry *xe; + char *base; + u32 p_cluster, num_clusters; + unsigned int ext_flags; + int name_offset, name_len; + struct ocfs2_xattr_value_buf vb; + struct ocfs2_xattr_bucket *bucket = NULL; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_post_refcount refcount; + struct ocfs2_post_refcount *p = NULL; + struct buffer_head *ref_root_bh = NULL; + + if (!xis->not_found) { + xe = xis->here; + name_offset = le16_to_cpu(xe->xe_name_offset); + name_len = OCFS2_XATTR_SIZE(xe->xe_name_len); + base = xis->base; + vb.vb_bh = xis->inode_bh; + vb.vb_access = ocfs2_journal_access_di; + } else { + int i, block_off = 0; + xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data; + xe = xbs->here; + name_offset = le16_to_cpu(xe->xe_name_offset); + name_len = OCFS2_XATTR_SIZE(xe->xe_name_len); + i = xbs->here - xbs->header->xh_entries; + + if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { + ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, + bucket_xh(xbs->bucket), + i, &block_off, + &name_offset); + if (ret) { + mlog_errno(ret); + goto out; + } + base = bucket_block(xbs->bucket, block_off); + vb.vb_bh = xbs->bucket->bu_bhs[block_off]; + vb.vb_access = ocfs2_journal_access; + + if (ocfs2_meta_ecc(osb)) { + /*create parameters for ocfs2_post_refcount. */ + bucket = xbs->bucket; + refcount.credits = bucket->bu_blocks; + refcount.para = bucket; + refcount.func = + ocfs2_xattr_bucket_post_refcount; + p = &refcount; + } + } else { + base = xbs->base; + vb.vb_bh = xbs->xattr_bh; + vb.vb_access = ocfs2_journal_access_xb; + } } - ret = ocfs2_xattr_set_entry_in_bucket(inode, ctxt->handle, xi, xs, - name_hash, local); + if (ocfs2_xattr_is_local(xe)) + goto out; + + vb.vb_xv = (struct ocfs2_xattr_value_root *) + (base + name_offset + name_len); + + ret = ocfs2_xattr_get_clusters(inode, 0, &p_cluster, + &num_clusters, &vb.vb_xv->xr_list, + &ext_flags); if (ret) { mlog_errno(ret); goto out; } - if (value_len <= OCFS2_XATTR_INLINE_SIZE) + /* + * We just need to check the 1st extent record, since we always + * CoW the whole xattr. So there shouldn't be a xattr with + * some REFCOUNT extent recs after the 1st one. + */ + if (!(ext_flags & OCFS2_EXT_REFCOUNTED)) goto out; - /* allocate the space now for the outside block storage. */ - ret = ocfs2_xattr_bucket_value_truncate_xs(inode, xs, - value_len, ctxt); + ret = ocfs2_lock_refcount_tree(osb, le64_to_cpu(di->i_refcount_loc), + 1, ref_tree, &ref_root_bh); if (ret) { mlog_errno(ret); + goto out; + } - if (xs->not_found) { - /* - * We can't allocate enough clusters for outside - * storage and we have allocated xattr already, - * so need to remove it. - */ - ocfs2_xattr_bucket_remove_xs(inode, ctxt->handle, xs); - } + /* + * If we are deleting the xattr or the new size will be stored inside, + * cool, leave it there, the xattr truncate process will remove them + * for us(it still needs the refcount tree lock and the meta, credits). + * And the worse case is that every cluster truncate will split the + * refcount tree, and make the original extent become 3. So we will need + * 2 * cluster more extent recs at most. + */ + if (!xi->xi_value || xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE) { + + ret = ocfs2_refcounted_xattr_delete_need(inode, + &(*ref_tree)->rf_ci, + ref_root_bh, vb.vb_xv, + meta_add, credits); + if (ret) + mlog_errno(ret); goto out; } -set_value_outside: - ret = ocfs2_xattr_bucket_set_value_outside(inode, ctxt->handle, - xs, val, value_len); + ret = ocfs2_refcount_cow_xattr(inode, di, &vb, + *ref_tree, ref_root_bh, 0, + le32_to_cpu(vb.vb_xv->xr_clusters), p); + if (ret) + mlog_errno(ret); + out: + brelse(ref_root_bh); return ret; } /* - * check whether the xattr bucket is filled up with the same hash value. - * If we want to insert the xattr with the same hash, return -ENOSPC. - * If we want to insert a xattr with different hash value, go ahead - * and ocfs2_divide_xattr_bucket will handle this. + * Add the REFCOUNTED flags for all the extent rec in ocfs2_xattr_value_root. + * The physical clusters will be added to refcount tree. */ -static int ocfs2_check_xattr_bucket_collision(struct inode *inode, - struct ocfs2_xattr_bucket *bucket, - const char *name) +static int ocfs2_xattr_value_attach_refcount(struct inode *inode, + struct ocfs2_xattr_value_root *xv, + struct ocfs2_extent_tree *value_et, + struct ocfs2_caching_info *ref_ci, + struct buffer_head *ref_root_bh, + struct ocfs2_cached_dealloc_ctxt *dealloc, + struct ocfs2_post_refcount *refcount) { - struct ocfs2_xattr_header *xh = bucket_xh(bucket); - u32 name_hash = ocfs2_xattr_name_hash(inode, name, strlen(name)); + int ret = 0; + u32 clusters = le32_to_cpu(xv->xr_clusters); + u32 cpos, p_cluster, num_clusters; + struct ocfs2_extent_list *el = &xv->xr_list; + unsigned int ext_flags; - if (name_hash != le32_to_cpu(xh->xh_entries[0].xe_name_hash)) - return 0; + cpos = 0; + while (cpos < clusters) { + ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster, + &num_clusters, el, &ext_flags); + if (ret) { + mlog_errno(ret); + break; + } - if (xh->xh_entries[le16_to_cpu(xh->xh_count) - 1].xe_name_hash == - xh->xh_entries[0].xe_name_hash) { - mlog(ML_ERROR, "Too much hash collision in xattr bucket %llu, " - "hash = %u\n", - (unsigned long long)bucket_blkno(bucket), - le32_to_cpu(xh->xh_entries[0].xe_name_hash)); - return -ENOSPC; + cpos += num_clusters; + if ((ext_flags & OCFS2_EXT_REFCOUNTED)) + continue; + + BUG_ON(!p_cluster); + + ret = ocfs2_add_refcount_flag(inode, value_et, + ref_ci, ref_root_bh, + cpos - num_clusters, + p_cluster, num_clusters, + dealloc, refcount); + if (ret) { + mlog_errno(ret); + break; + } } - return 0; + return ret; } -static int ocfs2_xattr_set_entry_index_block(struct inode *inode, - struct ocfs2_xattr_info *xi, - struct ocfs2_xattr_search *xs, - struct ocfs2_xattr_set_ctxt *ctxt) +/* + * Given a normal ocfs2_xattr_header, refcount all the entries which + * have value stored outside. + * Used for xattrs stored in inode and ocfs2_xattr_block. + */ +static int ocfs2_xattr_attach_refcount_normal(struct inode *inode, + struct ocfs2_xattr_value_buf *vb, + struct ocfs2_xattr_header *header, + struct ocfs2_caching_info *ref_ci, + struct buffer_head *ref_root_bh, + struct ocfs2_cached_dealloc_ctxt *dealloc) { - struct ocfs2_xattr_header *xh; + struct ocfs2_xattr_entry *xe; - u16 count, header_size, xh_free_start; - int free, max_free, need, old; - size_t value_size = 0, name_len = strlen(xi->name); - size_t blocksize = inode->i_sb->s_blocksize; - int ret, allocation = 0; + struct ocfs2_xattr_value_root *xv; + struct ocfs2_extent_tree et; + int i, ret = 0; - mlog_entry("Set xattr %s in xattr index block\n", xi->name); + for (i = 0; i < le16_to_cpu(header->xh_count); i++) { + xe = &header->xh_entries[i]; -try_again: - xh = xs->header; - count = le16_to_cpu(xh->xh_count); - xh_free_start = le16_to_cpu(xh->xh_free_start); - header_size = sizeof(struct ocfs2_xattr_header) + - count * sizeof(struct ocfs2_xattr_entry); - max_free = OCFS2_XATTR_BUCKET_SIZE - header_size - - le16_to_cpu(xh->xh_name_value_len) - OCFS2_XATTR_HEADER_GAP; + if (ocfs2_xattr_is_local(xe)) + continue; - mlog_bug_on_msg(header_size > blocksize, "bucket %llu has header size " - "of %u which exceed block size\n", - (unsigned long long)bucket_blkno(xs->bucket), - header_size); + xv = (struct ocfs2_xattr_value_root *)((void *)header + + le16_to_cpu(xe->xe_name_offset) + + OCFS2_XATTR_SIZE(xe->xe_name_len)); - if (xi->value && xi->value_len > OCFS2_XATTR_INLINE_SIZE) - value_size = OCFS2_XATTR_ROOT_SIZE; - else if (xi->value) - value_size = OCFS2_XATTR_SIZE(xi->value_len); + vb->vb_xv = xv; + ocfs2_init_xattr_value_extent_tree(&et, INODE_CACHE(inode), vb); - if (xs->not_found) - need = sizeof(struct ocfs2_xattr_entry) + - OCFS2_XATTR_SIZE(name_len) + value_size; - else { - need = value_size + OCFS2_XATTR_SIZE(name_len); + ret = ocfs2_xattr_value_attach_refcount(inode, xv, &et, + ref_ci, ref_root_bh, + dealloc, NULL); + if (ret) { + mlog_errno(ret); + break; + } + } + + return ret; +} + +static int ocfs2_xattr_inline_attach_refcount(struct inode *inode, + struct buffer_head *fe_bh, + struct ocfs2_caching_info *ref_ci, + struct buffer_head *ref_root_bh, + struct ocfs2_cached_dealloc_ctxt *dealloc) +{ + struct ocfs2_dinode *di = (struct ocfs2_dinode *)fe_bh->b_data; + struct ocfs2_xattr_header *header = (struct ocfs2_xattr_header *) + (fe_bh->b_data + inode->i_sb->s_blocksize - + le16_to_cpu(di->i_xattr_inline_size)); + struct ocfs2_xattr_value_buf vb = { + .vb_bh = fe_bh, + .vb_access = ocfs2_journal_access_di, + }; + + return ocfs2_xattr_attach_refcount_normal(inode, &vb, header, + ref_ci, ref_root_bh, dealloc); +} + +struct ocfs2_xattr_tree_value_refcount_para { + struct ocfs2_caching_info *ref_ci; + struct buffer_head *ref_root_bh; + struct ocfs2_cached_dealloc_ctxt *dealloc; +}; + +static int ocfs2_get_xattr_tree_value_root(struct super_block *sb, + struct ocfs2_xattr_bucket *bucket, + int offset, + struct ocfs2_xattr_value_root **xv, + struct buffer_head **bh) +{ + int ret, block_off, name_offset; + struct ocfs2_xattr_header *xh = bucket_xh(bucket); + struct ocfs2_xattr_entry *xe = &xh->xh_entries[offset]; + void *base; + + ret = ocfs2_xattr_bucket_get_name_value(sb, + bucket_xh(bucket), + offset, + &block_off, + &name_offset); + if (ret) { + mlog_errno(ret); + goto out; + } + + base = bucket_block(bucket, block_off); + + *xv = (struct ocfs2_xattr_value_root *)(base + name_offset + + OCFS2_XATTR_SIZE(xe->xe_name_len)); + + if (bh) + *bh = bucket->bu_bhs[block_off]; +out: + return ret; +} + +/* + * For a given xattr bucket, refcount all the entries which + * have value stored outside. + */ +static int ocfs2_xattr_bucket_value_refcount(struct inode *inode, + struct ocfs2_xattr_bucket *bucket, + void *para) +{ + int i, ret = 0; + struct ocfs2_extent_tree et; + struct ocfs2_xattr_tree_value_refcount_para *ref = + (struct ocfs2_xattr_tree_value_refcount_para *)para; + struct ocfs2_xattr_header *xh = + (struct ocfs2_xattr_header *)bucket->bu_bhs[0]->b_data; + struct ocfs2_xattr_entry *xe; + struct ocfs2_xattr_value_buf vb = { + .vb_access = ocfs2_journal_access, + }; + struct ocfs2_post_refcount refcount = { + .credits = bucket->bu_blocks, + .para = bucket, + .func = ocfs2_xattr_bucket_post_refcount, + }; + struct ocfs2_post_refcount *p = NULL; + + /* We only need post_refcount if we support metaecc. */ + if (ocfs2_meta_ecc(OCFS2_SB(inode->i_sb))) + p = &refcount; + + trace_ocfs2_xattr_bucket_value_refcount( + (unsigned long long)bucket_blkno(bucket), + le16_to_cpu(xh->xh_count)); + for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { + xe = &xh->xh_entries[i]; + + if (ocfs2_xattr_is_local(xe)) + continue; + + ret = ocfs2_get_xattr_tree_value_root(inode->i_sb, bucket, i, + &vb.vb_xv, &vb.vb_bh); + if (ret) { + mlog_errno(ret); + break; + } + + ocfs2_init_xattr_value_extent_tree(&et, + INODE_CACHE(inode), &vb); + + ret = ocfs2_xattr_value_attach_refcount(inode, vb.vb_xv, + &et, ref->ref_ci, + ref->ref_root_bh, + ref->dealloc, p); + if (ret) { + mlog_errno(ret); + break; + } + } + + return ret; + +} + +static int ocfs2_refcount_xattr_tree_rec(struct inode *inode, + struct buffer_head *root_bh, + u64 blkno, u32 cpos, u32 len, void *para) +{ + return ocfs2_iterate_xattr_buckets(inode, blkno, len, + ocfs2_xattr_bucket_value_refcount, + para); +} + +static int ocfs2_xattr_block_attach_refcount(struct inode *inode, + struct buffer_head *blk_bh, + struct ocfs2_caching_info *ref_ci, + struct buffer_head *ref_root_bh, + struct ocfs2_cached_dealloc_ctxt *dealloc) +{ + int ret = 0; + struct ocfs2_xattr_block *xb = + (struct ocfs2_xattr_block *)blk_bh->b_data; + + if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) { + struct ocfs2_xattr_header *header = &xb->xb_attrs.xb_header; + struct ocfs2_xattr_value_buf vb = { + .vb_bh = blk_bh, + .vb_access = ocfs2_journal_access_xb, + }; + + ret = ocfs2_xattr_attach_refcount_normal(inode, &vb, header, + ref_ci, ref_root_bh, + dealloc); + } else { + struct ocfs2_xattr_tree_value_refcount_para para = { + .ref_ci = ref_ci, + .ref_root_bh = ref_root_bh, + .dealloc = dealloc, + }; + + ret = ocfs2_iterate_xattr_index_block(inode, blk_bh, + ocfs2_refcount_xattr_tree_rec, + ¶); + } + + return ret; +} + +int ocfs2_xattr_attach_refcount_tree(struct inode *inode, + struct buffer_head *fe_bh, + struct ocfs2_caching_info *ref_ci, + struct buffer_head *ref_root_bh, + struct ocfs2_cached_dealloc_ctxt *dealloc) +{ + int ret = 0; + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct ocfs2_dinode *di = (struct ocfs2_dinode *)fe_bh->b_data; + struct buffer_head *blk_bh = NULL; + + if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { + ret = ocfs2_xattr_inline_attach_refcount(inode, fe_bh, + ref_ci, ref_root_bh, + dealloc); + if (ret) { + mlog_errno(ret); + goto out; + } + } + + if (!di->i_xattr_loc) + goto out; + + ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc), + &blk_bh); + if (ret < 0) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_xattr_block_attach_refcount(inode, blk_bh, ref_ci, + ref_root_bh, dealloc); + if (ret) + mlog_errno(ret); + + brelse(blk_bh); +out: + + return ret; +} + +typedef int (should_xattr_reflinked)(struct ocfs2_xattr_entry *xe); +/* + * Store the information we need in xattr reflink. + * old_bh and new_bh are inode bh for the old and new inode. + */ +struct ocfs2_xattr_reflink { + struct inode *old_inode; + struct inode *new_inode; + struct buffer_head *old_bh; + struct buffer_head *new_bh; + struct ocfs2_caching_info *ref_ci; + struct buffer_head *ref_root_bh; + struct ocfs2_cached_dealloc_ctxt *dealloc; + should_xattr_reflinked *xattr_reflinked; +}; + +/* + * Given a xattr header and xe offset, + * return the proper xv and the corresponding bh. + * xattr in inode, block and xattr tree have different implementaions. + */ +typedef int (get_xattr_value_root)(struct super_block *sb, + struct buffer_head *bh, + struct ocfs2_xattr_header *xh, + int offset, + struct ocfs2_xattr_value_root **xv, + struct buffer_head **ret_bh, + void *para); + +/* + * Calculate all the xattr value root metadata stored in this xattr header and + * credits we need if we create them from the scratch. + * We use get_xattr_value_root so that all types of xattr container can use it. + */ +static int ocfs2_value_metas_in_xattr_header(struct super_block *sb, + struct buffer_head *bh, + struct ocfs2_xattr_header *xh, + int *metas, int *credits, + int *num_recs, + get_xattr_value_root *func, + void *para) +{ + int i, ret = 0; + struct ocfs2_xattr_value_root *xv; + struct ocfs2_xattr_entry *xe; + + for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { + xe = &xh->xh_entries[i]; + if (ocfs2_xattr_is_local(xe)) + continue; + + ret = func(sb, bh, xh, i, &xv, NULL, para); + if (ret) { + mlog_errno(ret); + break; + } + + *metas += le16_to_cpu(xv->xr_list.l_tree_depth) * + le16_to_cpu(xv->xr_list.l_next_free_rec); + + *credits += ocfs2_calc_extend_credits(sb, + &def_xv.xv.xr_list); /* - * We only replace the old value if the new length is smaller - * than the old one. Otherwise we will allocate new space in the - * bucket to store it. + * If the value is a tree with depth > 1, We don't go deep + * to the extent block, so just calculate a maximum record num. */ - xe = xs->here; - if (ocfs2_xattr_is_local(xe)) - old = OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size)); + if (!xv->xr_list.l_tree_depth) + *num_recs += le16_to_cpu(xv->xr_list.l_next_free_rec); else - old = OCFS2_XATTR_SIZE(OCFS2_XATTR_ROOT_SIZE); + *num_recs += ocfs2_clusters_for_bytes(sb, + XATTR_SIZE_MAX); + } + + return ret; +} + +/* Used by xattr inode and block to return the right xv and buffer_head. */ +static int ocfs2_get_xattr_value_root(struct super_block *sb, + struct buffer_head *bh, + struct ocfs2_xattr_header *xh, + int offset, + struct ocfs2_xattr_value_root **xv, + struct buffer_head **ret_bh, + void *para) +{ + struct ocfs2_xattr_entry *xe = &xh->xh_entries[offset]; + + *xv = (struct ocfs2_xattr_value_root *)((void *)xh + + le16_to_cpu(xe->xe_name_offset) + + OCFS2_XATTR_SIZE(xe->xe_name_len)); + + if (ret_bh) + *ret_bh = bh; - if (old >= value_size) - need = 0; + return 0; +} + +/* + * Lock the meta_ac and caculate how much credits we need for reflink xattrs. + * It is only used for inline xattr and xattr block. + */ +static int ocfs2_reflink_lock_xattr_allocators(struct ocfs2_super *osb, + struct ocfs2_xattr_header *xh, + struct buffer_head *ref_root_bh, + int *credits, + struct ocfs2_alloc_context **meta_ac) +{ + int ret, meta_add = 0, num_recs = 0; + struct ocfs2_refcount_block *rb = + (struct ocfs2_refcount_block *)ref_root_bh->b_data; + + *credits = 0; + + ret = ocfs2_value_metas_in_xattr_header(osb->sb, NULL, xh, + &meta_add, credits, &num_recs, + ocfs2_get_xattr_value_root, + NULL); + if (ret) { + mlog_errno(ret); + goto out; } - free = xh_free_start - header_size - OCFS2_XATTR_HEADER_GAP; /* - * We need to make sure the new name/value pair - * can exist in the same block. + * We need to add/modify num_recs in refcount tree, so just calculate + * an approximate number we need for refcount tree change. + * Sometimes we need to split the tree, and after split, half recs + * will be moved to the new block, and a new block can only provide + * half number of recs. So we multiple new blocks by 2. */ - if (xh_free_start % blocksize < need) - free -= xh_free_start % blocksize; - - mlog(0, "xs->not_found = %d, in xattr bucket %llu: free = %d, " - "need = %d, max_free = %d, xh_free_start = %u, xh_name_value_len =" - " %u\n", xs->not_found, - (unsigned long long)bucket_blkno(xs->bucket), - free, need, max_free, le16_to_cpu(xh->xh_free_start), - le16_to_cpu(xh->xh_name_value_len)); - - if (free < need || - (xs->not_found && - count == ocfs2_xattr_max_xe_in_bucket(inode->i_sb))) { - if (need <= max_free && - count < ocfs2_xattr_max_xe_in_bucket(inode->i_sb)) { + num_recs = num_recs / ocfs2_refcount_recs_per_rb(osb->sb) * 2; + meta_add += num_recs; + *credits += num_recs + num_recs * OCFS2_EXPAND_REFCOUNT_TREE_CREDITS; + if (le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL) + *credits += le16_to_cpu(rb->rf_list.l_tree_depth) * + le16_to_cpu(rb->rf_list.l_next_free_rec) + 1; + else + *credits += 1; + + ret = ocfs2_reserve_new_metadata_blocks(osb, meta_add, meta_ac); + if (ret) + mlog_errno(ret); + +out: + return ret; +} + +/* + * Given a xattr header, reflink all the xattrs in this container. + * It can be used for inode, block and bucket. + * + * NOTE: + * Before we call this function, the caller has memcpy the xattr in + * old_xh to the new_xh. + * + * If args.xattr_reflinked is set, call it to decide whether the xe should + * be reflinked or not. If not, remove it from the new xattr header. + */ +static int ocfs2_reflink_xattr_header(handle_t *handle, + struct ocfs2_xattr_reflink *args, + struct buffer_head *old_bh, + struct ocfs2_xattr_header *xh, + struct buffer_head *new_bh, + struct ocfs2_xattr_header *new_xh, + struct ocfs2_xattr_value_buf *vb, + struct ocfs2_alloc_context *meta_ac, + get_xattr_value_root *func, + void *para) +{ + int ret = 0, i, j; + struct super_block *sb = args->old_inode->i_sb; + struct buffer_head *value_bh; + struct ocfs2_xattr_entry *xe, *last; + struct ocfs2_xattr_value_root *xv, *new_xv; + struct ocfs2_extent_tree data_et; + u32 clusters, cpos, p_cluster, num_clusters; + unsigned int ext_flags = 0; + + trace_ocfs2_reflink_xattr_header((unsigned long long)old_bh->b_blocknr, + le16_to_cpu(xh->xh_count)); + + last = &new_xh->xh_entries[le16_to_cpu(new_xh->xh_count)]; + for (i = 0, j = 0; i < le16_to_cpu(xh->xh_count); i++, j++) { + xe = &xh->xh_entries[i]; + + if (args->xattr_reflinked && !args->xattr_reflinked(xe)) { + xe = &new_xh->xh_entries[j]; + + le16_add_cpu(&new_xh->xh_count, -1); + if (new_xh->xh_count) { + memmove(xe, xe + 1, + (void *)last - (void *)xe); + memset(last, 0, + sizeof(struct ocfs2_xattr_entry)); + } + /* - * We can create the space by defragment. Since only the - * name/value will be moved, the xe shouldn't be changed - * in xs. + * We don't want j to increase in the next round since + * it is already moved ahead. */ - ret = ocfs2_defrag_xattr_bucket(inode, ctxt->handle, - xs->bucket); + j--; + continue; + } + + if (ocfs2_xattr_is_local(xe)) + continue; + + ret = func(sb, old_bh, xh, i, &xv, NULL, para); + if (ret) { + mlog_errno(ret); + break; + } + + ret = func(sb, new_bh, new_xh, j, &new_xv, &value_bh, para); + if (ret) { + mlog_errno(ret); + break; + } + + /* + * For the xattr which has l_tree_depth = 0, all the extent + * recs have already be copied to the new xh with the + * propriate OCFS2_EXT_REFCOUNTED flag we just need to + * increase the refount count int the refcount tree. + * + * For the xattr which has l_tree_depth > 0, we need + * to initialize it to the empty default value root, + * and then insert the extents one by one. + */ + if (xv->xr_list.l_tree_depth) { + memcpy(new_xv, &def_xv, sizeof(def_xv)); + vb->vb_xv = new_xv; + vb->vb_bh = value_bh; + ocfs2_init_xattr_value_extent_tree(&data_et, + INODE_CACHE(args->new_inode), vb); + } + + clusters = le32_to_cpu(xv->xr_clusters); + cpos = 0; + while (cpos < clusters) { + ret = ocfs2_xattr_get_clusters(args->old_inode, + cpos, + &p_cluster, + &num_clusters, + &xv->xr_list, + &ext_flags); if (ret) { mlog_errno(ret); goto out; } - xh_free_start = le16_to_cpu(xh->xh_free_start); - free = xh_free_start - header_size - - OCFS2_XATTR_HEADER_GAP; - if (xh_free_start % blocksize < need) - free -= xh_free_start % blocksize; + BUG_ON(!p_cluster); - if (free >= need) - goto xattr_set; + if (xv->xr_list.l_tree_depth) { + ret = ocfs2_insert_extent(handle, + &data_et, cpos, + ocfs2_clusters_to_blocks( + args->old_inode->i_sb, + p_cluster), + num_clusters, ext_flags, + meta_ac); + if (ret) { + mlog_errno(ret); + goto out; + } + } + + ret = ocfs2_increase_refcount(handle, args->ref_ci, + args->ref_root_bh, + p_cluster, num_clusters, + meta_ac, args->dealloc); + if (ret) { + mlog_errno(ret); + goto out; + } - mlog(0, "Can't get enough space for xattr insert by " - "defragment. Need %u bytes, but we have %d, so " - "allocate new bucket for it.\n", need, free); + cpos += num_clusters; } + } - /* - * We have to add new buckets or clusters and one - * allocation should leave us enough space for insert. - */ - BUG_ON(allocation); +out: + return ret; +} - /* - * We do not allow for overlapping ranges between buckets. And - * the maximum number of collisions we will allow for then is - * one bucket's worth, so check it here whether we need to - * add a new bucket for the insert. - */ - ret = ocfs2_check_xattr_bucket_collision(inode, - xs->bucket, - xi->name); +static int ocfs2_reflink_xattr_inline(struct ocfs2_xattr_reflink *args) +{ + int ret = 0, credits = 0; + handle_t *handle; + struct ocfs2_super *osb = OCFS2_SB(args->old_inode->i_sb); + struct ocfs2_dinode *di = (struct ocfs2_dinode *)args->old_bh->b_data; + int inline_size = le16_to_cpu(di->i_xattr_inline_size); + int header_off = osb->sb->s_blocksize - inline_size; + struct ocfs2_xattr_header *xh = (struct ocfs2_xattr_header *) + (args->old_bh->b_data + header_off); + struct ocfs2_xattr_header *new_xh = (struct ocfs2_xattr_header *) + (args->new_bh->b_data + header_off); + struct ocfs2_alloc_context *meta_ac = NULL; + struct ocfs2_inode_info *new_oi; + struct ocfs2_dinode *new_di; + struct ocfs2_xattr_value_buf vb = { + .vb_bh = args->new_bh, + .vb_access = ocfs2_journal_access_di, + }; + + ret = ocfs2_reflink_lock_xattr_allocators(osb, xh, args->ref_root_bh, + &credits, &meta_ac); + if (ret) { + mlog_errno(ret); + goto out; + } + + handle = ocfs2_start_trans(osb, credits); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto out; + } + + ret = ocfs2_journal_access_di(handle, INODE_CACHE(args->new_inode), + args->new_bh, OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + memcpy(args->new_bh->b_data + header_off, + args->old_bh->b_data + header_off, inline_size); + + new_di = (struct ocfs2_dinode *)args->new_bh->b_data; + new_di->i_xattr_inline_size = cpu_to_le16(inline_size); + + ret = ocfs2_reflink_xattr_header(handle, args, args->old_bh, xh, + args->new_bh, new_xh, &vb, meta_ac, + ocfs2_get_xattr_value_root, NULL); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + new_oi = OCFS2_I(args->new_inode); + /* + * Adjust extent record count to reserve space for extended attribute. + * Inline data count had been adjusted in ocfs2_duplicate_inline_data(). + */ + if (!(new_oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) && + !(ocfs2_inode_is_fast_symlink(args->new_inode))) { + struct ocfs2_extent_list *el = &new_di->id2.i_list; + le16_add_cpu(&el->l_count, -(inline_size / + sizeof(struct ocfs2_extent_rec))); + } + spin_lock(&new_oi->ip_lock); + new_oi->ip_dyn_features |= OCFS2_HAS_XATTR_FL | OCFS2_INLINE_XATTR_FL; + new_di->i_dyn_features = cpu_to_le16(new_oi->ip_dyn_features); + spin_unlock(&new_oi->ip_lock); + + ocfs2_journal_dirty(handle, args->new_bh); + +out_commit: + ocfs2_commit_trans(osb, handle); + +out: + if (meta_ac) + ocfs2_free_alloc_context(meta_ac); + return ret; +} + +static int ocfs2_create_empty_xattr_block(struct inode *inode, + struct buffer_head *fe_bh, + struct buffer_head **ret_bh, + int indexed) +{ + int ret; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_xattr_set_ctxt ctxt; + + memset(&ctxt, 0, sizeof(ctxt)); + ret = ocfs2_reserve_new_metadata_blocks(osb, 1, &ctxt.meta_ac); + if (ret < 0) { + mlog_errno(ret); + return ret; + } + + ctxt.handle = ocfs2_start_trans(osb, OCFS2_XATTR_BLOCK_CREATE_CREDITS); + if (IS_ERR(ctxt.handle)) { + ret = PTR_ERR(ctxt.handle); + mlog_errno(ret); + goto out; + } + + trace_ocfs2_create_empty_xattr_block( + (unsigned long long)fe_bh->b_blocknr, indexed); + ret = ocfs2_create_xattr_block(inode, fe_bh, &ctxt, indexed, + ret_bh); + if (ret) + mlog_errno(ret); + + ocfs2_commit_trans(osb, ctxt.handle); +out: + ocfs2_free_alloc_context(ctxt.meta_ac); + return ret; +} + +static int ocfs2_reflink_xattr_block(struct ocfs2_xattr_reflink *args, + struct buffer_head *blk_bh, + struct buffer_head *new_blk_bh) +{ + int ret = 0, credits = 0; + handle_t *handle; + struct ocfs2_inode_info *new_oi = OCFS2_I(args->new_inode); + struct ocfs2_dinode *new_di; + struct ocfs2_super *osb = OCFS2_SB(args->new_inode->i_sb); + int header_off = offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header); + struct ocfs2_xattr_block *xb = + (struct ocfs2_xattr_block *)blk_bh->b_data; + struct ocfs2_xattr_header *xh = &xb->xb_attrs.xb_header; + struct ocfs2_xattr_block *new_xb = + (struct ocfs2_xattr_block *)new_blk_bh->b_data; + struct ocfs2_xattr_header *new_xh = &new_xb->xb_attrs.xb_header; + struct ocfs2_alloc_context *meta_ac; + struct ocfs2_xattr_value_buf vb = { + .vb_bh = new_blk_bh, + .vb_access = ocfs2_journal_access_xb, + }; + + ret = ocfs2_reflink_lock_xattr_allocators(osb, xh, args->ref_root_bh, + &credits, &meta_ac); + if (ret) { + mlog_errno(ret); + return ret; + } + + /* One more credits in case we need to add xattr flags in new inode. */ + handle = ocfs2_start_trans(osb, credits + 1); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto out; + } + + if (!(new_oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) { + ret = ocfs2_journal_access_di(handle, + INODE_CACHE(args->new_inode), + args->new_bh, + OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); - goto out; + goto out_commit; } + } - ret = ocfs2_add_new_xattr_bucket(inode, - xs->xattr_bh, - xs->bucket, - ctxt); + ret = ocfs2_journal_access_xb(handle, INODE_CACHE(args->new_inode), + new_blk_bh, OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + memcpy(new_blk_bh->b_data + header_off, blk_bh->b_data + header_off, + osb->sb->s_blocksize - header_off); + + ret = ocfs2_reflink_xattr_header(handle, args, blk_bh, xh, + new_blk_bh, new_xh, &vb, meta_ac, + ocfs2_get_xattr_value_root, NULL); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + ocfs2_journal_dirty(handle, new_blk_bh); + + if (!(new_oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) { + new_di = (struct ocfs2_dinode *)args->new_bh->b_data; + spin_lock(&new_oi->ip_lock); + new_oi->ip_dyn_features |= OCFS2_HAS_XATTR_FL; + new_di->i_dyn_features = cpu_to_le16(new_oi->ip_dyn_features); + spin_unlock(&new_oi->ip_lock); + + ocfs2_journal_dirty(handle, args->new_bh); + } + +out_commit: + ocfs2_commit_trans(osb, handle); + +out: + ocfs2_free_alloc_context(meta_ac); + return ret; +} + +struct ocfs2_reflink_xattr_tree_args { + struct ocfs2_xattr_reflink *reflink; + struct buffer_head *old_blk_bh; + struct buffer_head *new_blk_bh; + struct ocfs2_xattr_bucket *old_bucket; + struct ocfs2_xattr_bucket *new_bucket; +}; + +/* + * NOTE: + * We have to handle the case that both old bucket and new bucket + * will call this function to get the right ret_bh. + * So The caller must give us the right bh. + */ +static int ocfs2_get_reflink_xattr_value_root(struct super_block *sb, + struct buffer_head *bh, + struct ocfs2_xattr_header *xh, + int offset, + struct ocfs2_xattr_value_root **xv, + struct buffer_head **ret_bh, + void *para) +{ + struct ocfs2_reflink_xattr_tree_args *args = + (struct ocfs2_reflink_xattr_tree_args *)para; + struct ocfs2_xattr_bucket *bucket; + + if (bh == args->old_bucket->bu_bhs[0]) + bucket = args->old_bucket; + else + bucket = args->new_bucket; + + return ocfs2_get_xattr_tree_value_root(sb, bucket, offset, + xv, ret_bh); +} + +struct ocfs2_value_tree_metas { + int num_metas; + int credits; + int num_recs; +}; + +static int ocfs2_value_tree_metas_in_bucket(struct super_block *sb, + struct buffer_head *bh, + struct ocfs2_xattr_header *xh, + int offset, + struct ocfs2_xattr_value_root **xv, + struct buffer_head **ret_bh, + void *para) +{ + struct ocfs2_xattr_bucket *bucket = + (struct ocfs2_xattr_bucket *)para; + + return ocfs2_get_xattr_tree_value_root(sb, bucket, offset, + xv, ret_bh); +} + +static int ocfs2_calc_value_tree_metas(struct inode *inode, + struct ocfs2_xattr_bucket *bucket, + void *para) +{ + struct ocfs2_value_tree_metas *metas = + (struct ocfs2_value_tree_metas *)para; + struct ocfs2_xattr_header *xh = + (struct ocfs2_xattr_header *)bucket->bu_bhs[0]->b_data; + + /* Add the credits for this bucket first. */ + metas->credits += bucket->bu_blocks; + return ocfs2_value_metas_in_xattr_header(inode->i_sb, bucket->bu_bhs[0], + xh, &metas->num_metas, + &metas->credits, &metas->num_recs, + ocfs2_value_tree_metas_in_bucket, + bucket); +} + +/* + * Given a xattr extent rec starting from blkno and having len clusters, + * iterate all the buckets calculate how much metadata we need for reflinking + * all the ocfs2_xattr_value_root and lock the allocators accordingly. + */ +static int ocfs2_lock_reflink_xattr_rec_allocators( + struct ocfs2_reflink_xattr_tree_args *args, + struct ocfs2_extent_tree *xt_et, + u64 blkno, u32 len, int *credits, + struct ocfs2_alloc_context **meta_ac, + struct ocfs2_alloc_context **data_ac) +{ + int ret, num_free_extents; + struct ocfs2_value_tree_metas metas; + struct ocfs2_super *osb = OCFS2_SB(args->reflink->old_inode->i_sb); + struct ocfs2_refcount_block *rb; + + memset(&metas, 0, sizeof(metas)); + + ret = ocfs2_iterate_xattr_buckets(args->reflink->old_inode, blkno, len, + ocfs2_calc_value_tree_metas, &metas); + if (ret) { + mlog_errno(ret); + goto out; + } + + *credits = metas.credits; + + /* + * Calculate we need for refcount tree change. + * + * We need to add/modify num_recs in refcount tree, so just calculate + * an approximate number we need for refcount tree change. + * Sometimes we need to split the tree, and after split, half recs + * will be moved to the new block, and a new block can only provide + * half number of recs. So we multiple new blocks by 2. + * In the end, we have to add credits for modifying the already + * existed refcount block. + */ + rb = (struct ocfs2_refcount_block *)args->reflink->ref_root_bh->b_data; + metas.num_recs = + (metas.num_recs + ocfs2_refcount_recs_per_rb(osb->sb) - 1) / + ocfs2_refcount_recs_per_rb(osb->sb) * 2; + metas.num_metas += metas.num_recs; + *credits += metas.num_recs + + metas.num_recs * OCFS2_EXPAND_REFCOUNT_TREE_CREDITS; + if (le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL) + *credits += le16_to_cpu(rb->rf_list.l_tree_depth) * + le16_to_cpu(rb->rf_list.l_next_free_rec) + 1; + else + *credits += 1; + + /* count in the xattr tree change. */ + num_free_extents = ocfs2_num_free_extents(osb, xt_et); + if (num_free_extents < 0) { + ret = num_free_extents; + mlog_errno(ret); + goto out; + } + + if (num_free_extents < len) + metas.num_metas += ocfs2_extend_meta_needed(xt_et->et_root_el); + + *credits += ocfs2_calc_extend_credits(osb->sb, + xt_et->et_root_el); + + if (metas.num_metas) { + ret = ocfs2_reserve_new_metadata_blocks(osb, metas.num_metas, + meta_ac); if (ret) { mlog_errno(ret); goto out; } - - /* - * ocfs2_add_new_xattr_bucket() will have updated - * xs->bucket if it moved, but it will not have updated - * any of the other search fields. Thus, we drop it and - * re-search. Everything should be cached, so it'll be - * quick. - */ - ocfs2_xattr_bucket_relse(xs->bucket); - ret = ocfs2_xattr_index_block_find(inode, xs->xattr_bh, - xi->name_index, - xi->name, xs); - if (ret && ret != -ENODATA) - goto out; - xs->not_found = ret; - allocation = 1; - goto try_again; } -xattr_set: - ret = ocfs2_xattr_set_in_bucket(inode, xi, xs, ctxt); + if (len) { + ret = ocfs2_reserve_clusters(osb, len, data_ac); + if (ret) + mlog_errno(ret); + } out: - mlog_exit(ret); + if (ret) { + if (*meta_ac) { + ocfs2_free_alloc_context(*meta_ac); + *meta_ac = NULL; + } + } + return ret; } -static int ocfs2_delete_xattr_in_bucket(struct inode *inode, - struct ocfs2_xattr_bucket *bucket, - void *para) +static int ocfs2_reflink_xattr_bucket(handle_t *handle, + u64 blkno, u64 new_blkno, u32 clusters, + u32 *cpos, int num_buckets, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_alloc_context *data_ac, + struct ocfs2_reflink_xattr_tree_args *args) { - int ret = 0; - struct ocfs2_xattr_header *xh = bucket_xh(bucket); - u16 i; - struct ocfs2_xattr_entry *xe; - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - struct ocfs2_xattr_set_ctxt ctxt = {NULL, NULL,}; - int credits = ocfs2_remove_extent_credits(osb->sb) + - ocfs2_blocks_per_xattr_bucket(inode->i_sb); - + int i, j, ret = 0; + struct super_block *sb = args->reflink->old_inode->i_sb; + int bpb = args->old_bucket->bu_blocks; + struct ocfs2_xattr_value_buf vb = { + .vb_access = ocfs2_journal_access, + }; - ocfs2_init_dealloc_ctxt(&ctxt.dealloc); + for (i = 0; i < num_buckets; i++, blkno += bpb, new_blkno += bpb) { + ret = ocfs2_read_xattr_bucket(args->old_bucket, blkno); + if (ret) { + mlog_errno(ret); + break; + } - for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { - xe = &xh->xh_entries[i]; - if (ocfs2_xattr_is_local(xe)) - continue; + ret = ocfs2_init_xattr_bucket(args->new_bucket, new_blkno, 1); + if (ret) { + mlog_errno(ret); + break; + } - ctxt.handle = ocfs2_start_trans(osb, credits); - if (IS_ERR(ctxt.handle)) { - ret = PTR_ERR(ctxt.handle); + ret = ocfs2_xattr_bucket_journal_access(handle, + args->new_bucket, + OCFS2_JOURNAL_ACCESS_CREATE); + if (ret) { mlog_errno(ret); break; } - ret = ocfs2_xattr_bucket_value_truncate(inode, bucket, - i, 0, &ctxt); + for (j = 0; j < bpb; j++) + memcpy(bucket_block(args->new_bucket, j), + bucket_block(args->old_bucket, j), + sb->s_blocksize); - ocfs2_commit_trans(osb, ctxt.handle); + /* + * Record the start cpos so that we can use it to initialize + * our xattr tree we also set the xh_num_bucket for the new + * bucket. + */ + if (i == 0) { + *cpos = le32_to_cpu(bucket_xh(args->new_bucket)-> + xh_entries[0].xe_name_hash); + bucket_xh(args->new_bucket)->xh_num_buckets = + cpu_to_le16(num_buckets); + } + + ocfs2_xattr_bucket_journal_dirty(handle, args->new_bucket); + + ret = ocfs2_reflink_xattr_header(handle, args->reflink, + args->old_bucket->bu_bhs[0], + bucket_xh(args->old_bucket), + args->new_bucket->bu_bhs[0], + bucket_xh(args->new_bucket), + &vb, meta_ac, + ocfs2_get_reflink_xattr_value_root, + args); + if (ret) { + mlog_errno(ret); + break; + } + + /* + * Re-access and dirty the bucket to calculate metaecc. + * Because we may extend the transaction in reflink_xattr_header + * which will let the already accessed block gone. + */ + ret = ocfs2_xattr_bucket_journal_access(handle, + args->new_bucket, + OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); break; } + + ocfs2_xattr_bucket_journal_dirty(handle, args->new_bucket); + + ocfs2_xattr_bucket_relse(args->old_bucket); + ocfs2_xattr_bucket_relse(args->new_bucket); } - ocfs2_schedule_truncate_log_flush(osb, 1); - ocfs2_run_deallocs(osb, &ctxt.dealloc); + ocfs2_xattr_bucket_relse(args->old_bucket); + ocfs2_xattr_bucket_relse(args->new_bucket); return ret; } -static int ocfs2_delete_xattr_index_block(struct inode *inode, - struct buffer_head *xb_bh) +static int ocfs2_reflink_xattr_buckets(handle_t *handle, + struct inode *inode, + struct ocfs2_reflink_xattr_tree_args *args, + struct ocfs2_extent_tree *et, + struct ocfs2_alloc_context *meta_ac, + struct ocfs2_alloc_context *data_ac, + u64 blkno, u32 cpos, u32 len) { - struct ocfs2_xattr_block *xb = - (struct ocfs2_xattr_block *)xb_bh->b_data; - struct ocfs2_extent_list *el = &xb->xb_attrs.xb_root.xt_list; - int ret = 0; - u32 name_hash = UINT_MAX, e_cpos, num_clusters; - u64 p_blkno; - - if (le16_to_cpu(el->l_next_free_rec) == 0) - return 0; + int ret, first_inserted = 0; + u32 p_cluster, num_clusters, reflink_cpos = 0; + u64 new_blkno; + unsigned int num_buckets, reflink_buckets; + unsigned int bpc = + ocfs2_xattr_buckets_per_cluster(OCFS2_SB(inode->i_sb)); + + ret = ocfs2_read_xattr_bucket(args->old_bucket, blkno); + if (ret) { + mlog_errno(ret); + goto out; + } + num_buckets = le16_to_cpu(bucket_xh(args->old_bucket)->xh_num_buckets); + ocfs2_xattr_bucket_relse(args->old_bucket); - while (name_hash > 0) { - ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, - &e_cpos, &num_clusters, el); + while (len && num_buckets) { + ret = ocfs2_claim_clusters(handle, data_ac, + 1, &p_cluster, &num_clusters); if (ret) { mlog_errno(ret); goto out; } - ret = ocfs2_iterate_xattr_buckets(inode, p_blkno, num_clusters, - ocfs2_delete_xattr_in_bucket, - NULL); + new_blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster); + reflink_buckets = min(num_buckets, bpc * num_clusters); + + ret = ocfs2_reflink_xattr_bucket(handle, blkno, + new_blkno, num_clusters, + &reflink_cpos, reflink_buckets, + meta_ac, data_ac, args); if (ret) { mlog_errno(ret); goto out; } - ret = ocfs2_rm_xattr_cluster(inode, xb_bh, - p_blkno, e_cpos, num_clusters); + /* + * For the 1st allocated cluster, we make it use the same cpos + * so that the xattr tree looks the same as the original one + * in the most case. + */ + if (!first_inserted) { + reflink_cpos = cpos; + first_inserted = 1; + } + ret = ocfs2_insert_extent(handle, et, reflink_cpos, new_blkno, + num_clusters, 0, meta_ac); + if (ret) + mlog_errno(ret); + + trace_ocfs2_reflink_xattr_buckets((unsigned long long)new_blkno, + num_clusters, reflink_cpos); + + len -= num_clusters; + blkno += ocfs2_clusters_to_blocks(inode->i_sb, num_clusters); + num_buckets -= reflink_buckets; + } +out: + return ret; +} + +/* + * Create the same xattr extent record in the new inode's xattr tree. + */ +static int ocfs2_reflink_xattr_rec(struct inode *inode, + struct buffer_head *root_bh, + u64 blkno, + u32 cpos, + u32 len, + void *para) +{ + int ret, credits = 0; + handle_t *handle; + struct ocfs2_reflink_xattr_tree_args *args = + (struct ocfs2_reflink_xattr_tree_args *)para; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_alloc_context *meta_ac = NULL; + struct ocfs2_alloc_context *data_ac = NULL; + struct ocfs2_extent_tree et; + + trace_ocfs2_reflink_xattr_rec((unsigned long long)blkno, len); + + ocfs2_init_xattr_tree_extent_tree(&et, + INODE_CACHE(args->reflink->new_inode), + args->new_blk_bh); + + ret = ocfs2_lock_reflink_xattr_rec_allocators(args, &et, blkno, + len, &credits, + &meta_ac, &data_ac); + if (ret) { + mlog_errno(ret); + goto out; + } + + handle = ocfs2_start_trans(osb, credits); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto out; + } + + ret = ocfs2_reflink_xattr_buckets(handle, inode, args, &et, + meta_ac, data_ac, + blkno, cpos, len); + if (ret) + mlog_errno(ret); + + ocfs2_commit_trans(osb, handle); + +out: + if (meta_ac) + ocfs2_free_alloc_context(meta_ac); + if (data_ac) + ocfs2_free_alloc_context(data_ac); + return ret; +} + +/* + * Create reflinked xattr buckets. + * We will add bucket one by one, and refcount all the xattrs in the bucket + * if they are stored outside. + */ +static int ocfs2_reflink_xattr_tree(struct ocfs2_xattr_reflink *args, + struct buffer_head *blk_bh, + struct buffer_head *new_blk_bh) +{ + int ret; + struct ocfs2_reflink_xattr_tree_args para; + + memset(¶, 0, sizeof(para)); + para.reflink = args; + para.old_blk_bh = blk_bh; + para.new_blk_bh = new_blk_bh; + + para.old_bucket = ocfs2_xattr_bucket_new(args->old_inode); + if (!para.old_bucket) { + mlog_errno(-ENOMEM); + return -ENOMEM; + } + + para.new_bucket = ocfs2_xattr_bucket_new(args->new_inode); + if (!para.new_bucket) { + ret = -ENOMEM; + mlog_errno(ret); + goto out; + } + + ret = ocfs2_iterate_xattr_index_block(args->old_inode, blk_bh, + ocfs2_reflink_xattr_rec, + ¶); + if (ret) + mlog_errno(ret); + +out: + ocfs2_xattr_bucket_free(para.old_bucket); + ocfs2_xattr_bucket_free(para.new_bucket); + return ret; +} + +static int ocfs2_reflink_xattr_in_block(struct ocfs2_xattr_reflink *args, + struct buffer_head *blk_bh) +{ + int ret, indexed = 0; + struct buffer_head *new_blk_bh = NULL; + struct ocfs2_xattr_block *xb = + (struct ocfs2_xattr_block *)blk_bh->b_data; + + + if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) + indexed = 1; + + ret = ocfs2_create_empty_xattr_block(args->new_inode, args->new_bh, + &new_blk_bh, indexed); + if (ret) { + mlog_errno(ret); + goto out; + } + + if (!indexed) + ret = ocfs2_reflink_xattr_block(args, blk_bh, new_blk_bh); + else + ret = ocfs2_reflink_xattr_tree(args, blk_bh, new_blk_bh); + if (ret) + mlog_errno(ret); + +out: + brelse(new_blk_bh); + return ret; +} + +static int ocfs2_reflink_xattr_no_security(struct ocfs2_xattr_entry *xe) +{ + int type = ocfs2_xattr_get_type(xe); + + return type != OCFS2_XATTR_INDEX_SECURITY && + type != OCFS2_XATTR_INDEX_POSIX_ACL_ACCESS && + type != OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT; +} + +int ocfs2_reflink_xattrs(struct inode *old_inode, + struct buffer_head *old_bh, + struct inode *new_inode, + struct buffer_head *new_bh, + bool preserve_security) +{ + int ret; + struct ocfs2_xattr_reflink args; + struct ocfs2_inode_info *oi = OCFS2_I(old_inode); + struct ocfs2_dinode *di = (struct ocfs2_dinode *)old_bh->b_data; + struct buffer_head *blk_bh = NULL; + struct ocfs2_cached_dealloc_ctxt dealloc; + struct ocfs2_refcount_tree *ref_tree; + struct buffer_head *ref_root_bh = NULL; + + ret = ocfs2_lock_refcount_tree(OCFS2_SB(old_inode->i_sb), + le64_to_cpu(di->i_refcount_loc), + 1, &ref_tree, &ref_root_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + ocfs2_init_dealloc_ctxt(&dealloc); + + args.old_inode = old_inode; + args.new_inode = new_inode; + args.old_bh = old_bh; + args.new_bh = new_bh; + args.ref_ci = &ref_tree->rf_ci; + args.ref_root_bh = ref_root_bh; + args.dealloc = &dealloc; + if (preserve_security) + args.xattr_reflinked = NULL; + else + args.xattr_reflinked = ocfs2_reflink_xattr_no_security; + + if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { + ret = ocfs2_reflink_xattr_inline(&args); if (ret) { mlog_errno(ret); - break; + goto out_unlock; } + } - if (e_cpos == 0) - break; + if (!di->i_xattr_loc) + goto out_unlock; - name_hash = e_cpos - 1; + ret = ocfs2_read_xattr_block(old_inode, le64_to_cpu(di->i_xattr_loc), + &blk_bh); + if (ret < 0) { + mlog_errno(ret); + goto out_unlock; + } + + ret = ocfs2_reflink_xattr_in_block(&args, blk_bh); + if (ret) + mlog_errno(ret); + + brelse(blk_bh); + +out_unlock: + ocfs2_unlock_refcount_tree(OCFS2_SB(old_inode->i_sb), + ref_tree, 1); + brelse(ref_root_bh); + + if (ocfs2_dealloc_has_cluster(&dealloc)) { + ocfs2_schedule_truncate_log_flush(OCFS2_SB(old_inode->i_sb), 1); + ocfs2_run_deallocs(OCFS2_SB(old_inode->i_sb), &dealloc); } out: @@ -5304,11 +7199,49 @@ out: } /* + * Initialize security and acl for a already created inode. + * Used for reflink a non-preserve-security file. + * + * It uses common api like ocfs2_xattr_set, so the caller + * must not hold any lock expect i_mutex. + */ +int ocfs2_init_security_and_acl(struct inode *dir, + struct inode *inode, + const struct qstr *qstr, + struct posix_acl *default_acl, + struct posix_acl *acl) +{ + struct buffer_head *dir_bh = NULL; + int ret = 0; + + ret = ocfs2_init_security_get(inode, dir, qstr, NULL); + if (ret) { + mlog_errno(ret); + goto leave; + } + + ret = ocfs2_inode_lock(dir, &dir_bh, 0); + if (ret) { + mlog_errno(ret); + goto leave; + } + + if (!ret && default_acl) + ret = ocfs2_iop_set_acl(inode, default_acl, ACL_TYPE_DEFAULT); + if (!ret && acl) + ret = ocfs2_iop_set_acl(inode, acl, ACL_TYPE_ACCESS); + + ocfs2_inode_unlock(dir, 0); + brelse(dir_bh); +leave: + return ret; +} +/* * 'security' attributes support */ -static size_t ocfs2_xattr_security_list(struct inode *inode, char *list, +static size_t ocfs2_xattr_security_list(struct dentry *dentry, char *list, size_t list_size, const char *name, - size_t name_len) + size_t name_len, int type) { const size_t prefix_len = XATTR_SECURITY_PREFIX_LEN; const size_t total_len = prefix_len + name_len + 1; @@ -5321,34 +7254,56 @@ static size_t ocfs2_xattr_security_list(struct inode *inode, char *list, return total_len; } -static int ocfs2_xattr_security_get(struct inode *inode, const char *name, - void *buffer, size_t size) +static int ocfs2_xattr_security_get(struct dentry *dentry, const char *name, + void *buffer, size_t size, int type) { if (strcmp(name, "") == 0) return -EINVAL; - return ocfs2_xattr_get(inode, OCFS2_XATTR_INDEX_SECURITY, name, - buffer, size); + return ocfs2_xattr_get(dentry->d_inode, OCFS2_XATTR_INDEX_SECURITY, + name, buffer, size); } -static int ocfs2_xattr_security_set(struct inode *inode, const char *name, - const void *value, size_t size, int flags) +static int ocfs2_xattr_security_set(struct dentry *dentry, const char *name, + const void *value, size_t size, int flags, int type) { if (strcmp(name, "") == 0) return -EINVAL; - return ocfs2_xattr_set(inode, OCFS2_XATTR_INDEX_SECURITY, name, value, - size, flags); + return ocfs2_xattr_set(dentry->d_inode, OCFS2_XATTR_INDEX_SECURITY, + name, value, size, flags); +} + +int ocfs2_initxattrs(struct inode *inode, const struct xattr *xattr_array, + void *fs_info) +{ + const struct xattr *xattr; + int err = 0; + + for (xattr = xattr_array; xattr->name != NULL; xattr++) { + err = ocfs2_xattr_set(inode, OCFS2_XATTR_INDEX_SECURITY, + xattr->name, xattr->value, + xattr->value_len, XATTR_CREATE); + if (err) + break; + } + return err; } int ocfs2_init_security_get(struct inode *inode, struct inode *dir, + const struct qstr *qstr, struct ocfs2_security_xattr_info *si) { /* check whether ocfs2 support feature xattr */ if (!ocfs2_supports_xattr(OCFS2_SB(dir->i_sb))) return -EOPNOTSUPP; - return security_inode_init_security(inode, dir, &si->name, &si->value, - &si->value_len); + if (si) + return security_old_inode_init_security(inode, dir, qstr, + &si->name, &si->value, + &si->value_len); + + return security_inode_init_security(inode, dir, qstr, + &ocfs2_initxattrs, NULL); } int ocfs2_init_security_set(handle_t *handle, @@ -5364,7 +7319,7 @@ int ocfs2_init_security_set(handle_t *handle, xattr_ac, data_ac); } -struct xattr_handler ocfs2_xattr_security_handler = { +const struct xattr_handler ocfs2_xattr_security_handler = { .prefix = XATTR_SECURITY_PREFIX, .list = ocfs2_xattr_security_list, .get = ocfs2_xattr_security_get, @@ -5374,9 +7329,9 @@ struct xattr_handler ocfs2_xattr_security_handler = { /* * 'trusted' attributes support */ -static size_t ocfs2_xattr_trusted_list(struct inode *inode, char *list, +static size_t ocfs2_xattr_trusted_list(struct dentry *dentry, char *list, size_t list_size, const char *name, - size_t name_len) + size_t name_len, int type) { const size_t prefix_len = XATTR_TRUSTED_PREFIX_LEN; const size_t total_len = prefix_len + name_len + 1; @@ -5389,26 +7344,26 @@ static size_t ocfs2_xattr_trusted_list(struct inode *inode, char *list, return total_len; } -static int ocfs2_xattr_trusted_get(struct inode *inode, const char *name, - void *buffer, size_t size) +static int ocfs2_xattr_trusted_get(struct dentry *dentry, const char *name, + void *buffer, size_t size, int type) { if (strcmp(name, "") == 0) return -EINVAL; - return ocfs2_xattr_get(inode, OCFS2_XATTR_INDEX_TRUSTED, name, - buffer, size); + return ocfs2_xattr_get(dentry->d_inode, OCFS2_XATTR_INDEX_TRUSTED, + name, buffer, size); } -static int ocfs2_xattr_trusted_set(struct inode *inode, const char *name, - const void *value, size_t size, int flags) +static int ocfs2_xattr_trusted_set(struct dentry *dentry, const char *name, + const void *value, size_t size, int flags, int type) { if (strcmp(name, "") == 0) return -EINVAL; - return ocfs2_xattr_set(inode, OCFS2_XATTR_INDEX_TRUSTED, name, value, - size, flags); + return ocfs2_xattr_set(dentry->d_inode, OCFS2_XATTR_INDEX_TRUSTED, + name, value, size, flags); } -struct xattr_handler ocfs2_xattr_trusted_handler = { +const struct xattr_handler ocfs2_xattr_trusted_handler = { .prefix = XATTR_TRUSTED_PREFIX, .list = ocfs2_xattr_trusted_list, .get = ocfs2_xattr_trusted_get, @@ -5418,13 +7373,13 @@ struct xattr_handler ocfs2_xattr_trusted_handler = { /* * 'user' attributes support */ -static size_t ocfs2_xattr_user_list(struct inode *inode, char *list, +static size_t ocfs2_xattr_user_list(struct dentry *dentry, char *list, size_t list_size, const char *name, - size_t name_len) + size_t name_len, int type) { const size_t prefix_len = XATTR_USER_PREFIX_LEN; const size_t total_len = prefix_len + name_len + 1; - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR) return 0; @@ -5437,34 +7392,34 @@ static size_t ocfs2_xattr_user_list(struct inode *inode, char *list, return total_len; } -static int ocfs2_xattr_user_get(struct inode *inode, const char *name, - void *buffer, size_t size) +static int ocfs2_xattr_user_get(struct dentry *dentry, const char *name, + void *buffer, size_t size, int type) { - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); if (strcmp(name, "") == 0) return -EINVAL; if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR) return -EOPNOTSUPP; - return ocfs2_xattr_get(inode, OCFS2_XATTR_INDEX_USER, name, + return ocfs2_xattr_get(dentry->d_inode, OCFS2_XATTR_INDEX_USER, name, buffer, size); } -static int ocfs2_xattr_user_set(struct inode *inode, const char *name, - const void *value, size_t size, int flags) +static int ocfs2_xattr_user_set(struct dentry *dentry, const char *name, + const void *value, size_t size, int flags, int type) { - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); if (strcmp(name, "") == 0) return -EINVAL; if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR) return -EOPNOTSUPP; - return ocfs2_xattr_set(inode, OCFS2_XATTR_INDEX_USER, name, value, - size, flags); + return ocfs2_xattr_set(dentry->d_inode, OCFS2_XATTR_INDEX_USER, + name, value, size, flags); } -struct xattr_handler ocfs2_xattr_user_handler = { +const struct xattr_handler ocfs2_xattr_user_handler = { .prefix = XATTR_USER_PREFIX, .list = ocfs2_xattr_user_list, .get = ocfs2_xattr_user_get, diff --git a/fs/ocfs2/xattr.h b/fs/ocfs2/xattr.h index 1ca7e9a1b7b..f10d5b93c36 100644 --- a/fs/ocfs2/xattr.h +++ b/fs/ocfs2/xattr.h @@ -32,19 +32,15 @@ enum ocfs2_xattr_type { struct ocfs2_security_xattr_info { int enable; - char *name; + const char *name; void *value; size_t value_len; }; -extern struct xattr_handler ocfs2_xattr_user_handler; -extern struct xattr_handler ocfs2_xattr_trusted_handler; -extern struct xattr_handler ocfs2_xattr_security_handler; -#ifdef CONFIG_OCFS2_FS_POSIX_ACL -extern struct xattr_handler ocfs2_xattr_acl_access_handler; -extern struct xattr_handler ocfs2_xattr_acl_default_handler; -#endif -extern struct xattr_handler *ocfs2_xattr_handlers[]; +extern const struct xattr_handler ocfs2_xattr_user_handler; +extern const struct xattr_handler ocfs2_xattr_trusted_handler; +extern const struct xattr_handler ocfs2_xattr_security_handler; +extern const struct xattr_handler *ocfs2_xattr_handlers[]; ssize_t ocfs2_listxattr(struct dentry *, char *, size_t); int ocfs2_xattr_get_nolock(struct inode *, struct buffer_head *, int, @@ -55,8 +51,11 @@ int ocfs2_xattr_set_handle(handle_t *, struct inode *, struct buffer_head *, int, const char *, const void *, size_t, int, struct ocfs2_alloc_context *, struct ocfs2_alloc_context *); +int ocfs2_has_inline_xattr_value_outside(struct inode *inode, + struct ocfs2_dinode *di); int ocfs2_xattr_remove(struct inode *, struct buffer_head *); int ocfs2_init_security_get(struct inode *, struct inode *, + const struct qstr *, struct ocfs2_security_xattr_info *); int ocfs2_init_security_set(handle_t *, struct inode *, struct buffer_head *, @@ -67,7 +66,7 @@ int ocfs2_calc_security_init(struct inode *, struct ocfs2_security_xattr_info *, int *, int *, struct ocfs2_alloc_context **); int ocfs2_calc_xattr_init(struct inode *, struct buffer_head *, - int, struct ocfs2_security_xattr_info *, + umode_t, struct ocfs2_security_xattr_info *, int *, int *, int *); /* @@ -83,5 +82,19 @@ struct ocfs2_xattr_value_buf { struct ocfs2_xattr_value_root *vb_xv; }; - +int ocfs2_xattr_attach_refcount_tree(struct inode *inode, + struct buffer_head *fe_bh, + struct ocfs2_caching_info *ref_ci, + struct buffer_head *ref_root_bh, + struct ocfs2_cached_dealloc_ctxt *dealloc); +int ocfs2_reflink_xattrs(struct inode *old_inode, + struct buffer_head *old_bh, + struct inode *new_inode, + struct buffer_head *new_bh, + bool preserve_security); +int ocfs2_init_security_and_acl(struct inode *dir, + struct inode *inode, + const struct qstr *qstr, + struct posix_acl *default_acl, + struct posix_acl *acl); #endif /* OCFS2_XATTR_H */ |