fs: provide rcu-walk aware permission i_ops

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

60 files changed, 287 insertions, 146 deletions

diff --git a/Documentation/filesystems/Locking b/Documentation/filesystems/Locking
index e90ffe61eb6..977d8919cc6 100644
--- a/Documentation/filesystems/Locking
+++ b/Documentation/filesystems/Locking
@@ -47,8 +47,8 @@ ata *);
 	void * (*follow_link) (struct dentry *, struct nameidata *);
 	void (*put_link) (struct dentry *, struct nameidata *, void *);
 	void (*truncate) (struct inode *);
-	int (*permission) (struct inode *, int, struct nameidata *);
-	int (*check_acl)(struct inode *, int);
+	int (*permission) (struct inode *, int, unsigned int);
+	int (*check_acl)(struct inode *, int, unsigned int);
 	int (*setattr) (struct dentry *, struct iattr *);
 	int (*getattr) (struct vfsmount *, struct dentry *, struct kstat *);
 	int (*setxattr) (struct dentry *, const char *,const void *,size_t,int);
@@ -76,7 +76,7 @@ follow_link:	no
 put_link:	no
 truncate:	yes		(see below)
 setattr:	yes
-permission:	no
+permission:	no (may not block if called in rcu-walk mode)
 check_acl:	no
 getattr:	no
 setxattr:	yes
diff --git a/Documentation/filesystems/path-lookup.txt b/Documentation/filesystems/path-lookup.txt
index 8789d1810be..eb59c8b44be 100644
--- a/Documentation/filesystems/path-lookup.txt
+++ b/Documentation/filesystems/path-lookup.txt
@@ -316,11 +316,9 @@ The detailed design for rcu-walk is like this:
 
 The cases where rcu-walk cannot continue are:
 * NULL dentry (ie. any uncached path element)
-* parent with d_inode->i_op->permission or ACLs
 * Following links
 
-In future patches, permission checks become rcu-walk aware. It may be possible
-eventually to make following links rcu-walk aware.
+It may be possible eventually to make following links rcu-walk aware.
 
 Uncached path elements will always require dropping to ref-walk mode, at the
 very least because i_mutex needs to be grabbed, and objects allocated.
@@ -336,9 +334,49 @@ or stored into. The result is massive improvements in performance and
 scalability of path resolution.
 
 
+Interesting statistics
+======================
+
+The following table gives rcu lookup statistics for a few simple workloads
+(2s12c24t Westmere, debian non-graphical system). Ungraceful are attempts to
+drop rcu that fail due to d_seq failure and requiring the entire path lookup
+again. Other cases are successful rcu-drops that are required before the final
+element, nodentry for missing dentry, revalidate for filesystem revalidate
+routine requiring rcu drop, permission for permission check requiring drop,
+and link for symlink traversal requiring drop.
+
+     rcu-lookups     restart  nodentry          link  revalidate  permission
+bootup     47121           0      4624          1010       10283        7852
+dbench  25386793           0   6778659(26.7%)     55         549        1156
+kbuild   2696672          10     64442(2.3%)  108764(4.0%)     1        1590
+git diff   39605           0        28             2           0         106
+vfstest 24185492        4945    708725(2.9%) 1076136(4.4%)     0        2651
+
+What this shows is that failed rcu-walk lookups, ie. ones that are restarted
+entirely with ref-walk, are quite rare. Even the "vfstest" case which
+specifically has concurrent renames/mkdir/rmdir/ creat/unlink/etc to excercise
+such races is not showing a huge amount of restarts.
+
+Dropping from rcu-walk to ref-walk mean that we have encountered a dentry where
+the reference count needs to be taken for some reason. This is either because
+we have reached the target of the path walk, or because we have encountered a
+condition that can't be resolved in rcu-walk mode.  Ideally, we drop rcu-walk
+only when we have reached the target dentry, so the other statistics show where
+this does not happen.
+
+Note that a graceful drop from rcu-walk mode due to something such as the
+dentry not existing (which can be common) is not necessarily a failure of
+rcu-walk scheme, because some elements of the path may have been walked in
+rcu-walk mode. The further we get from common path elements (such as cwd or
+root), the less contended the dentry is likely to be. The closer we are to
+common path elements, the more likely they will exist in dentry cache.
+
+
 Papers and other documentation on dcache locking
 ================================================
 
 1. Scaling dcache with RCU (http://linuxjournal.com/article.php?sid=7124).
 
 2. http://lse.sourceforge.net/locking/dcache/dcache.html
+
+
diff --git a/Documentation/filesystems/porting b/Documentation/filesystems/porting
index cd9756a2709..07a32b42cf9 100644
--- a/Documentation/filesystems/porting
+++ b/Documentation/filesystems/porting
@@ -380,3 +380,8 @@ the filesystem provides it), which requires dropping out of rcu-walk mode. This
 may now be called in rcu-walk mode (nd->flags & LOOKUP_RCU). -ECHILD should be
 returned if the filesystem cannot handle rcu-walk. See
 Documentation/filesystems/vfs.txt for more details.
+
+	permission and check_acl are inode permission checks that are called
+on many or all directory inodes on the way down a path walk (to check for
+exec permission). These must now be rcu-walk aware (flags & IPERM_RCU). See
+Documentation/filesystems/vfs.txt for more details.
diff --git a/Documentation/filesystems/vfs.txt b/Documentation/filesystems/vfs.txt
index c936b491238..fbb324e2bd4 100644
--- a/Documentation/filesystems/vfs.txt
+++ b/Documentation/filesystems/vfs.txt
@@ -325,7 +325,8 @@ struct inode_operations {
         void * (*follow_link) (struct dentry *, struct nameidata *);
         void (*put_link) (struct dentry *, struct nameidata *, void *);
 	void (*truncate) (struct inode *);
-	int (*permission) (struct inode *, int, struct nameidata *);
+	int (*permission) (struct inode *, int, unsigned int);
+	int (*check_acl)(struct inode *, int, unsigned int);
 	int (*setattr) (struct dentry *, struct iattr *);
 	int (*getattr) (struct vfsmount *mnt, struct dentry *, struct kstat *);
 	int (*setxattr) (struct dentry *, const char *,const void *,size_t,int);
@@ -414,6 +415,13 @@ otherwise noted.
   permission: called by the VFS to check for access rights on a POSIX-like
   	filesystem.
 
+	May be called in rcu-walk mode (flags & IPERM_RCU). If in rcu-walk
+	mode, the filesystem must check the permission without blocking or
+	storing to the inode.
+
+	If a situation is encountered that rcu-walk cannot handle, return
+	-ECHILD and it will be called again in ref-walk mode.
+
   setattr: called by the VFS to set attributes for a file. This method
   	is called by chmod(2) and related system calls.
 
diff --git a/drivers/staging/smbfs/file.c b/drivers/staging/smbfs/file.c
index 5dcd19c60eb..31372e7b12d 100644
--- a/drivers/staging/smbfs/file.c
+++ b/drivers/staging/smbfs/file.c
@@ -407,11 +407,14 @@ smb_file_release(struct inode *inode, struct file * file)
  * privileges, so we need our own check for this.
  */
 static int
-smb_file_permission(struct inode *inode, int mask)
+smb_file_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	int mode = inode->i_mode;
 	int error = 0;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	VERBOSE("mode=%x, mask=%x\n", mode, mask);
 
 	/* Look at user permissions */
diff --git a/fs/9p/acl.c b/fs/9p/acl.c
index 12d602351db..6e58c4ca1e6 100644
--- a/fs/9p/acl.c
+++ b/fs/9p/acl.c
@@ -91,11 +91,14 @@ static struct posix_acl *v9fs_get_cached_acl(struct inode *inode, int type)
 	return acl;
 }
 
-int v9fs_check_acl(struct inode *inode, int mask)
+int v9fs_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
 	struct posix_acl *acl;
 	struct v9fs_session_info *v9ses;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	v9ses = v9fs_inode2v9ses(inode);
 	if ((v9ses->flags & V9FS_ACCESS_MASK) != V9FS_ACCESS_CLIENT) {
 		/*
diff --git a/fs/9p/acl.h b/fs/9p/acl.h
index 59e18c2e8c7..7ef3ac9f6d9 100644
--- a/fs/9p/acl.h
+++ b/fs/9p/acl.h
@@ -16,7 +16,7 @@
 
 #ifdef CONFIG_9P_FS_POSIX_ACL
 extern int v9fs_get_acl(struct inode *, struct p9_fid *);
-extern int v9fs_check_acl(struct inode *inode, int mask);
+extern int v9fs_check_acl(struct inode *inode, int mask, unsigned int flags);
 extern int v9fs_acl_chmod(struct dentry *);
 extern int v9fs_set_create_acl(struct dentry *,
 			       struct posix_acl *, struct posix_acl *);
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index cca8eef736f..6d4bc1c8ff6 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -624,7 +624,7 @@ extern void afs_clear_permits(struct afs_vnode *);
 extern void afs_cache_permit(struct afs_vnode *, struct key *, long);
 extern void afs_zap_permits(struct rcu_head *);
 extern struct key *afs_request_key(struct afs_cell *);
-extern int afs_permission(struct inode *, int);
+extern int afs_permission(struct inode *, int, unsigned int);
 
 /*
  * server.c
diff --git a/fs/afs/security.c b/fs/afs/security.c
index bb4ed144d0e..f44b9d35537 100644
--- a/fs/afs/security.c
+++ b/fs/afs/security.c
@@ -285,13 +285,16 @@ static int afs_check_permit(struct afs_vnode *vnode, struct key *key,
  * - AFS ACLs are attached to directories only, and a file is controlled by its
  *   parent directory's ACL
  */
-int afs_permission(struct inode *inode, int mask)
+int afs_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	struct afs_vnode *vnode = AFS_FS_I(inode);
 	afs_access_t uninitialized_var(access);
 	struct key *key;
 	int ret;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	_enter("{{%x:%u},%lx},%x,",
 	       vnode->fid.vid, vnode->fid.vnode, vnode->flags, mask);
 
@@ -347,7 +350,7 @@ int afs_permission(struct inode *inode, int mask)
 	}
 
 	key_put(key);
-	ret = generic_permission(inode, mask, NULL);
+	ret = generic_permission(inode, mask, flags, NULL);
 	_leave(" = %d", ret);
 	return ret;
 
diff --git a/fs/bad_inode.c b/fs/bad_inode.c
index f024d8aadde..9ad2369d9e3 100644
--- a/fs/bad_inode.c
+++ b/fs/bad_inode.c
@@ -229,8 +229,11 @@ static int bad_inode_readlink(struct dentry *dentry, char __user *buffer,
 	return -EIO;
 }
 
-static int bad_inode_permission(struct inode *inode, int mask)
+static int bad_inode_permission(struct inode *inode, int mask, unsigned int flags)
 {
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	return -EIO;
 }
 
diff --git a/fs/btrfs/acl.c b/fs/btrfs/acl.c
index 2222d161c7b..cb518a4b917 100644
--- a/fs/btrfs/acl.c
+++ b/fs/btrfs/acl.c
@@ -185,13 +185,15 @@ static int btrfs_xattr_acl_set(struct dentry *dentry, const char *name,
 	return ret;
 }
 
-int btrfs_check_acl(struct inode *inode, int mask)
+int btrfs_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
 	struct posix_acl *acl;
 	int error = -EAGAIN;
 
-	acl = btrfs_get_acl(inode, ACL_TYPE_ACCESS);
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
+	acl = btrfs_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
 		return PTR_ERR(acl);
 	if (acl) {
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index af52f6d7a4d..a142d204b52 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -2544,7 +2544,7 @@ int btrfs_sync_fs(struct super_block *sb, int wait);
 
 /* acl.c */
 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
-int btrfs_check_acl(struct inode *inode, int mask);
+int btrfs_check_acl(struct inode *inode, int mask, unsigned int flags);
 #else
 #define btrfs_check_acl NULL
 #endif
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 63e4546b478..5cf0db0081f 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -7211,11 +7211,14 @@ static int btrfs_set_page_dirty(struct page *page)
 	return __set_page_dirty_nobuffers(page);
 }
 
-static int btrfs_permission(struct inode *inode, int mask)
+static int btrfs_permission(struct inode *inode, int mask, unsigned int flags)
 {
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	if ((BTRFS_I(inode)->flags & BTRFS_INODE_READONLY) && (mask & MAY_WRITE))
 		return -EACCES;
-	return generic_permission(inode, mask, btrfs_check_acl);
+	return generic_permission(inode, mask, flags, btrfs_check_acl);
 }
 
 static const struct inode_operations btrfs_dir_inode_operations = {
diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c
index 47f8c8baf3b..e61de4f7b99 100644
--- a/fs/ceph/inode.c
+++ b/fs/ceph/inode.c
@@ -1781,12 +1781,17 @@ int ceph_do_getattr(struct inode *inode, int mask)
  * Check inode permissions.  We verify we have a valid value for
  * the AUTH cap, then call the generic handler.
  */
-int ceph_permission(struct inode *inode, int mask)
+int ceph_permission(struct inode *inode, int mask, unsigned int flags)
 {
-	int err = ceph_do_getattr(inode, CEPH_CAP_AUTH_SHARED);
+	int err;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
+	err = ceph_do_getattr(inode, CEPH_CAP_AUTH_SHARED);
 
 	if (!err)
-		err = generic_permission(inode, mask, NULL);
+		err = generic_permission(inode, mask, flags, NULL);
 	return err;
 }
 
diff --git a/fs/ceph/super.h b/fs/ceph/super.h
index 7f01728a465..4553d8829ed 100644
--- a/fs/ceph/super.h
+++ b/fs/ceph/super.h
@@ -665,7 +665,7 @@ extern void ceph_queue_invalidate(struct inode *inode);
 extern void ceph_queue_writeback(struct inode *inode);
 
 extern int ceph_do_getattr(struct inode *inode, int mask);
-extern int ceph_permission(struct inode *inode, int mask);
+extern int ceph_permission(struct inode *inode, int mask, unsigned int flags);
 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
 			struct kstat *stat);
diff --git a/fs/cifs/cifsfs.c b/fs/cifs/cifsfs.c
index 223717dcc40..8e21e0fe65d 100644
--- a/fs/cifs/cifsfs.c
+++ b/fs/cifs/cifsfs.c
@@ -283,10 +283,13 @@ cifs_statfs(struct dentry *dentry, struct kstatfs *buf)
 	return 0;
 }
 
-static int cifs_permission(struct inode *inode, int mask)
+static int cifs_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	struct cifs_sb_info *cifs_sb;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	cifs_sb = CIFS_SB(inode->i_sb);
 
 	if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NO_PERM) {
@@ -298,7 +301,7 @@ static int cifs_permission(struct inode *inode, int mask)
 		on the client (above and beyond ACL on servers) for
 		servers which do not support setting and viewing mode bits,
 		so allowing client to check permissions is useful */
-		return generic_permission(inode, mask, NULL);
+		return generic_permission(inode, mask, flags, NULL);
 }
 
 static struct kmem_cache *cifs_inode_cachep;
diff --git a/fs/coda/dir.c b/fs/coda/dir.c
index 619a8303766..29badd91360 100644
--- a/fs/coda/dir.c
+++ b/fs/coda/dir.c
@@ -135,10 +135,13 @@ exit:
 }
 
 
-int coda_permission(struct inode *inode, int mask)
+int coda_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	int error;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+
 	mask &= MAY_READ | MAY_WRITE | MAY_EXEC;
  
 	if (!mask)
diff --git a/fs/coda/pioctl.c b/fs/coda/pioctl.c
index 2fd89b5c5c7..741f0bd0391 100644
--- a/fs/coda/pioctl.c
+++ b/fs/coda/pioctl.c
@@ -24,7 +24,7 @@
 #include <linux/coda_psdev.h>
 
 /* pioctl ops */
-static int coda_ioctl_permission(struct inode *inode, int mask);
+static int coda_ioctl_permission(struct inode *inode, int mask, unsigned int flags);
 static long coda_pioctl(struct file *filp, unsigned int cmd,
 			unsigned long user_data);
 
@@ -41,8 +41,10 @@ const struct file_operations coda_ioctl_operations = {
 };
 
 /* the coda pioctl inode ops */
-static int coda_ioctl_permission(struct inode *inode, int mask)
+static int coda_ioctl_permission(struct inode *inode, int mask, unsigned int flags)
 {
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 	return (mask & MAY_EXEC) ? -EACCES : 0;
 }
 
diff --git a/fs/ecryptfs/inode.c b/fs/ecryptfs/inode.c
index f91b35db4c6..337352a9475 100644
--- a/fs/ecryptfs/inode.c
+++ b/fs/ecryptfs/inode.c
@@ -980,8 +980,10 @@ int ecryptfs_truncate(struct dentry *dentry, loff_t new_length)
 }
 
 static int
-ecryptfs_permission(struct inode *inode, int mask)
+ecryptfs_permission(struct inode *inode, int mask, unsigned int flags)
 {
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 	return inode_permission(ecryptfs_inode_to_lower(inode), mask);
 }
 
diff --git a/fs/ext2/acl.c b/fs/ext2/acl.c
index 2bcc0431bad..dd9bb3f0c8d 100644
--- a/fs/ext2/acl.c
+++ b/fs/ext2/acl.c
@@ -232,10 +232,14 @@ ext2_set_acl(struct inode *inode, int type, struct posix_acl *acl)
 }
 
 int
-ext2_check_acl(struct inode *inode, int mask)
+ext2_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct posix_acl *acl = ext2_get_acl(inode, ACL_TYPE_ACCESS);
+	struct posix_acl *acl;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
+	acl = ext2_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
 		return PTR_ERR(acl);
 	if (acl) {
diff --git a/fs/ext2/acl.h b/fs/ext2/acl.h
index 3ff6cbb9ac4..c939b7b1209 100644
--- a/fs/ext2/acl.h
+++ b/fs/ext2/acl.h
@@ -54,7 +54,7 @@ static inline int ext2_acl_count(size_t size)
 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 
 /* acl.c */
-extern int ext2_check_acl (struct inode *, int);
+extern int ext2_check_acl (struct inode *, int, unsigned int);
 extern int ext2_acl_chmod (struct inode *);
 extern int ext2_init_acl (struct inode *, struct inode *);
 
diff --git a/fs/ext3/acl.c b/fs/ext3/acl.c
index 8a11fe21218..9e49da8302d 100644
--- a/fs/ext3/acl.c
+++ b/fs/ext3/acl.c
@@ -240,10 +240,14 @@ ext3_set_acl(handle_t *handle, struct inode *inode, int type,
 }
 
 int
-ext3_check_acl(struct inode *inode, int mask)
+ext3_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct posix_acl *acl = ext3_get_acl(inode, ACL_TYPE_ACCESS);
+	struct posix_acl *acl;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
+	acl = ext3_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
 		return PTR_ERR(acl);
 	if (acl) {
diff --git a/fs/ext3/acl.h b/fs/ext3/acl.h
index 597334626de..5faf8048e90 100644
--- a/fs/ext3/acl.h
+++ b/fs/ext3/acl.h
@@ -54,7 +54,7 @@ static inline int ext3_acl_count(size_t size)
 #ifdef CONFIG_EXT3_FS_POSIX_ACL
 
 /* acl.c */
-extern int ext3_check_acl (struct inode *, int);
+extern int ext3_check_acl (struct inode *, int, unsigned int);
 extern int ext3_acl_chmod (struct inode *);
 extern int ext3_init_acl (handle_t *, struct inode *, struct inode *);
 
diff --git a/fs/ext4/acl.c b/fs/ext4/acl.c
index 5e2ed4504ea..373dcaeedba 100644
--- a/fs/ext4/acl.c
+++ b/fs/ext4/acl.c
@@ -238,10 +238,14 @@ ext4_set_acl(handle_t *handle, struct inode *inode, int type,
 }
 
 int
-ext4_check_acl(struct inode *inode, int mask)
+ext4_check_acl(struct inode *inode, int mask, unsigned int flags)
 {
-	struct posix_acl *acl = ext4_get_acl(inode, ACL_TYPE_ACCESS);
+	struct posix_acl *acl;
+
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
 
+	acl = ext4_get_acl(inode, ACL_TYPE_ACCESS);
 	if (IS_ERR(acl))
 		return PTR_ERR(acl);
 	if (acl) {
diff --git a/fs/ext4/acl.h b/fs/ext4/acl.h
index 9d843d5deac..dec821168fd 100644
--- a/fs/ext4/acl.h
+++ b/fs/ext4/acl.h
@@ -54,7 +54,7 @@ static inline int ext4_acl_count(size_t size)
 #ifdef CONFIG_EXT4_FS_POSIX_ACL
 
 /* acl.c */
-extern int ext4_check_acl(struct inode *, int);
+extern int ext4_check_acl(struct inode *, int, unsigned int);
 extern int ext4_acl_chmod(struct inode *);
 extern int ext4_init_acl(handle_t *, struct inode *, struct inode *);
 
diff --git a/fs/fuse/dir.c b/fs/fuse/dir.c
index 07f4b5e675f..f738599fd8c 100644
--- a/fs/fuse/dir.c
+++ b/fs/fuse/dir.c
@@ -985,12 +985,15 @@ static int fuse_access(struct inode *inode, int mask)
  * access request is sent.  Execute permission is still checked
  * locally based on file mode.
  */
-static int fuse_permission(struct inode *inode, int mask)
+static int fuse_permission(struct inode *inode, int mask, unsigned int flags)
 {
 	struct fuse_conn *fc = get_fuse_conn(inode);
 	bool refreshed = false;
 	int err = 0;
 
+	if (flags & IPERM_FLAG_RCU)
+		return -ECHILD;
+