diff options
Diffstat (limited to 'fs/nfs/pnfs.c')
| -rw-r--r-- | fs/nfs/pnfs.c | 1308 |
1 files changed, 851 insertions, 457 deletions
diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 17149a49006..6fdcd233d6f 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -33,8 +33,10 @@ #include "internal.h" #include "pnfs.h" #include "iostat.h" +#include "nfs4trace.h" #define NFSDBG_FACILITY NFSDBG_PNFS +#define PNFS_LAYOUTGET_RETRY_TIMEOUT (120*HZ) /* Locking: * @@ -70,6 +72,10 @@ find_pnfs_driver(u32 id) spin_lock(&pnfs_spinlock); local = find_pnfs_driver_locked(id); + if (local != NULL && !try_module_get(local->owner)) { + dprintk("%s: Could not grab reference on module\n", __func__); + local = NULL; + } spin_unlock(&pnfs_spinlock); return local; } @@ -80,6 +86,9 @@ unset_pnfs_layoutdriver(struct nfs_server *nfss) if (nfss->pnfs_curr_ld) { if (nfss->pnfs_curr_ld->clear_layoutdriver) nfss->pnfs_curr_ld->clear_layoutdriver(nfss); + /* Decrement the MDS count. Purge the deviceid cache if zero */ + if (atomic_dec_and_test(&nfss->nfs_client->cl_mds_count)) + nfs4_deviceid_purge_client(nfss->nfs_client); module_put(nfss->pnfs_curr_ld->owner); } nfss->pnfs_curr_ld = NULL; @@ -101,8 +110,8 @@ set_pnfs_layoutdriver(struct nfs_server *server, const struct nfs_fh *mntfh, goto out_no_driver; if (!(server->nfs_client->cl_exchange_flags & (EXCHGID4_FLAG_USE_NON_PNFS | EXCHGID4_FLAG_USE_PNFS_MDS))) { - printk(KERN_ERR "%s: id %u cl_exchange_flags 0x%x\n", __func__, - id, server->nfs_client->cl_exchange_flags); + printk(KERN_ERR "NFS: %s: id %u cl_exchange_flags 0x%x\n", + __func__, id, server->nfs_client->cl_exchange_flags); goto out_no_driver; } ld_type = find_pnfs_driver(id); @@ -115,18 +124,16 @@ set_pnfs_layoutdriver(struct nfs_server *server, const struct nfs_fh *mntfh, goto out_no_driver; } } - if (!try_module_get(ld_type->owner)) { - dprintk("%s: Could not grab reference on module\n", __func__); - goto out_no_driver; - } server->pnfs_curr_ld = ld_type; if (ld_type->set_layoutdriver && ld_type->set_layoutdriver(server, mntfh)) { - printk(KERN_ERR "%s: Error initializing pNFS layout driver %u.\n", - __func__, id); + printk(KERN_ERR "NFS: %s: Error initializing pNFS layout " + "driver %u.\n", __func__, id); module_put(ld_type->owner); goto out_no_driver; } + /* Bump the MDS count */ + atomic_inc(&server->nfs_client->cl_mds_count); dprintk("%s: pNFS module for %u set\n", __func__, id); return; @@ -143,11 +150,11 @@ pnfs_register_layoutdriver(struct pnfs_layoutdriver_type *ld_type) struct pnfs_layoutdriver_type *tmp; if (ld_type->id == 0) { - printk(KERN_ERR "%s id 0 is reserved\n", __func__); + printk(KERN_ERR "NFS: %s id 0 is reserved\n", __func__); return status; } if (!ld_type->alloc_lseg || !ld_type->free_lseg) { - printk(KERN_ERR "%s Layout driver must provide " + printk(KERN_ERR "NFS: %s Layout driver must provide " "alloc_lseg and free_lseg.\n", __func__); return status; } @@ -160,7 +167,7 @@ pnfs_register_layoutdriver(struct pnfs_layoutdriver_type *ld_type) dprintk("%s Registering id:%u name:%s\n", __func__, ld_type->id, ld_type->name); } else { - printk(KERN_ERR "%s Module with id %d already loaded!\n", + printk(KERN_ERR "NFS: %s Module with id %d already loaded!\n", __func__, ld_type->id); } spin_unlock(&pnfs_spinlock); @@ -185,7 +192,7 @@ EXPORT_SYMBOL_GPL(pnfs_unregister_layoutdriver); /* Need to hold i_lock if caller does not already hold reference */ void -get_layout_hdr(struct pnfs_layout_hdr *lo) +pnfs_get_layout_hdr(struct pnfs_layout_hdr *lo) { atomic_inc(&lo->plh_refcount); } @@ -194,43 +201,107 @@ static struct pnfs_layout_hdr * pnfs_alloc_layout_hdr(struct inode *ino, gfp_t gfp_flags) { struct pnfs_layoutdriver_type *ld = NFS_SERVER(ino)->pnfs_curr_ld; - return ld->alloc_layout_hdr ? ld->alloc_layout_hdr(ino, gfp_flags) : - kzalloc(sizeof(struct pnfs_layout_hdr), gfp_flags); + return ld->alloc_layout_hdr(ino, gfp_flags); } static void pnfs_free_layout_hdr(struct pnfs_layout_hdr *lo) { - struct pnfs_layoutdriver_type *ld = NFS_SERVER(lo->plh_inode)->pnfs_curr_ld; + struct nfs_server *server = NFS_SERVER(lo->plh_inode); + struct pnfs_layoutdriver_type *ld = server->pnfs_curr_ld; + + if (!list_empty(&lo->plh_layouts)) { + struct nfs_client *clp = server->nfs_client; + + spin_lock(&clp->cl_lock); + list_del_init(&lo->plh_layouts); + spin_unlock(&clp->cl_lock); + } put_rpccred(lo->plh_lc_cred); - return ld->alloc_layout_hdr ? ld->free_layout_hdr(lo) : kfree(lo); + return ld->free_layout_hdr(lo); } static void -destroy_layout_hdr(struct pnfs_layout_hdr *lo) +pnfs_detach_layout_hdr(struct pnfs_layout_hdr *lo) { + struct nfs_inode *nfsi = NFS_I(lo->plh_inode); dprintk("%s: freeing layout cache %p\n", __func__, lo); - BUG_ON(!list_empty(&lo->plh_layouts)); - NFS_I(lo->plh_inode)->layout = NULL; - pnfs_free_layout_hdr(lo); + nfsi->layout = NULL; + /* Reset MDS Threshold I/O counters */ + nfsi->write_io = 0; + nfsi->read_io = 0; +} + +void +pnfs_put_layout_hdr(struct pnfs_layout_hdr *lo) +{ + struct inode *inode = lo->plh_inode; + + if (atomic_dec_and_lock(&lo->plh_refcount, &inode->i_lock)) { + pnfs_detach_layout_hdr(lo); + spin_unlock(&inode->i_lock); + pnfs_free_layout_hdr(lo); + } +} + +static int +pnfs_iomode_to_fail_bit(u32 iomode) +{ + return iomode == IOMODE_RW ? + NFS_LAYOUT_RW_FAILED : NFS_LAYOUT_RO_FAILED; } static void -put_layout_hdr_locked(struct pnfs_layout_hdr *lo) +pnfs_layout_set_fail_bit(struct pnfs_layout_hdr *lo, int fail_bit) { - if (atomic_dec_and_test(&lo->plh_refcount)) - destroy_layout_hdr(lo); + lo->plh_retry_timestamp = jiffies; + if (!test_and_set_bit(fail_bit, &lo->plh_flags)) + atomic_inc(&lo->plh_refcount); } -void -put_layout_hdr(struct pnfs_layout_hdr *lo) +static void +pnfs_layout_clear_fail_bit(struct pnfs_layout_hdr *lo, int fail_bit) +{ + if (test_and_clear_bit(fail_bit, &lo->plh_flags)) + atomic_dec(&lo->plh_refcount); +} + +static void +pnfs_layout_io_set_failed(struct pnfs_layout_hdr *lo, u32 iomode) { struct inode *inode = lo->plh_inode; + struct pnfs_layout_range range = { + .iomode = iomode, + .offset = 0, + .length = NFS4_MAX_UINT64, + }; + LIST_HEAD(head); - if (atomic_dec_and_lock(&lo->plh_refcount, &inode->i_lock)) { - destroy_layout_hdr(lo); - spin_unlock(&inode->i_lock); + spin_lock(&inode->i_lock); + pnfs_layout_set_fail_bit(lo, pnfs_iomode_to_fail_bit(iomode)); + pnfs_mark_matching_lsegs_invalid(lo, &head, &range); + spin_unlock(&inode->i_lock); + pnfs_free_lseg_list(&head); + dprintk("%s Setting layout IOMODE_%s fail bit\n", __func__, + iomode == IOMODE_RW ? "RW" : "READ"); +} + +static bool +pnfs_layout_io_test_failed(struct pnfs_layout_hdr *lo, u32 iomode) +{ + unsigned long start, end; + int fail_bit = pnfs_iomode_to_fail_bit(iomode); + + if (test_bit(fail_bit, &lo->plh_flags) == 0) + return false; + end = jiffies; + start = end - PNFS_LAYOUTGET_RETRY_TIMEOUT; + if (!time_in_range(lo->plh_retry_timestamp, start, end)) { + /* It is time to retry the failed layoutgets */ + pnfs_layout_clear_fail_bit(lo, fail_bit); + return false; } + return true; } static void @@ -244,33 +315,32 @@ init_lseg(struct pnfs_layout_hdr *lo, struct pnfs_layout_segment *lseg) lseg->pls_layout = lo; } -static void free_lseg(struct pnfs_layout_segment *lseg) +static void pnfs_free_lseg(struct pnfs_layout_segment *lseg) { struct inode *ino = lseg->pls_layout->plh_inode; NFS_SERVER(ino)->pnfs_curr_ld->free_lseg(lseg); - /* Matched by get_layout_hdr in pnfs_insert_layout */ - put_layout_hdr(NFS_I(ino)->layout); } static void -put_lseg_common(struct pnfs_layout_segment *lseg) +pnfs_layout_remove_lseg(struct pnfs_layout_hdr *lo, + struct pnfs_layout_segment *lseg) { - struct inode *inode = lseg->pls_layout->plh_inode; + struct inode *inode = lo->plh_inode; WARN_ON(test_bit(NFS_LSEG_VALID, &lseg->pls_flags)); list_del_init(&lseg->pls_list); - if (list_empty(&lseg->pls_layout->plh_segs)) { - set_bit(NFS_LAYOUT_DESTROYED, &lseg->pls_layout->plh_flags); - /* Matched by initial refcount set in alloc_init_layout_hdr */ - put_layout_hdr_locked(lseg->pls_layout); - } + /* Matched by pnfs_get_layout_hdr in pnfs_layout_insert_lseg */ + atomic_dec(&lo->plh_refcount); + if (list_empty(&lo->plh_segs)) + clear_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags); rpc_wake_up(&NFS_SERVER(inode)->roc_rpcwaitq); } void -put_lseg(struct pnfs_layout_segment *lseg) +pnfs_put_lseg(struct pnfs_layout_segment *lseg) { + struct pnfs_layout_hdr *lo; struct inode *inode; if (!lseg) @@ -279,19 +349,19 @@ put_lseg(struct pnfs_layout_segment *lseg) dprintk("%s: lseg %p ref %d valid %d\n", __func__, lseg, atomic_read(&lseg->pls_refcount), test_bit(NFS_LSEG_VALID, &lseg->pls_flags)); - inode = lseg->pls_layout->plh_inode; + lo = lseg->pls_layout; + inode = lo->plh_inode; if (atomic_dec_and_lock(&lseg->pls_refcount, &inode->i_lock)) { - LIST_HEAD(free_me); - - put_lseg_common(lseg); - list_add(&lseg->pls_list, &free_me); + pnfs_get_layout_hdr(lo); + pnfs_layout_remove_lseg(lo, lseg); spin_unlock(&inode->i_lock); - pnfs_free_lseg_list(&free_me); + pnfs_free_lseg(lseg); + pnfs_put_layout_hdr(lo); } } -EXPORT_SYMBOL_GPL(put_lseg); +EXPORT_SYMBOL_GPL(pnfs_put_lseg); -static inline u64 +static u64 end_offset(u64 start, u64 len) { u64 end; @@ -300,17 +370,6 @@ end_offset(u64 start, u64 len) return end >= start ? end : NFS4_MAX_UINT64; } -/* last octet in a range */ -static inline u64 -last_byte_offset(u64 start, u64 len) -{ - u64 end; - - BUG_ON(!len); - end = start + len; - return end > start ? end - 1 : NFS4_MAX_UINT64; -} - /* * is l2 fully contained in l1? * start1 end1 @@ -318,9 +377,9 @@ last_byte_offset(u64 start, u64 len) * start2 end2 * [----------------) */ -static inline int -lo_seg_contained(struct pnfs_layout_range *l1, - struct pnfs_layout_range *l2) +static bool +pnfs_lseg_range_contained(const struct pnfs_layout_range *l1, + const struct pnfs_layout_range *l2) { u64 start1 = l1->offset; u64 end1 = end_offset(start1, l1->length); @@ -337,9 +396,9 @@ lo_seg_contained(struct pnfs_layout_range *l1, * start2 end2 * [----------------) */ -static inline int -lo_seg_intersecting(struct pnfs_layout_range *l1, - struct pnfs_layout_range *l2) +static bool +pnfs_lseg_range_intersecting(const struct pnfs_layout_range *l1, + const struct pnfs_layout_range *l2) { u64 start1 = l1->offset; u64 end1 = end_offset(start1, l1->length); @@ -351,12 +410,22 @@ lo_seg_intersecting(struct pnfs_layout_range *l1, } static bool -should_free_lseg(struct pnfs_layout_range *lseg_range, - struct pnfs_layout_range *recall_range) +should_free_lseg(const struct pnfs_layout_range *lseg_range, + const struct pnfs_layout_range *recall_range) { return (recall_range->iomode == IOMODE_ANY || lseg_range->iomode == recall_range->iomode) && - lo_seg_intersecting(lseg_range, recall_range); + pnfs_lseg_range_intersecting(lseg_range, recall_range); +} + +static bool pnfs_lseg_dec_and_remove_zero(struct pnfs_layout_segment *lseg, + struct list_head *tmp_list) +{ + if (!atomic_dec_and_test(&lseg->pls_refcount)) + return false; + pnfs_layout_remove_lseg(lseg->pls_layout, lseg); + list_add(&lseg->pls_list, tmp_list); + return true; } /* Returns 1 if lseg is removed from list, 0 otherwise */ @@ -372,11 +441,8 @@ static int mark_lseg_invalid(struct pnfs_layout_segment *lseg, */ dprintk("%s: lseg %p ref %d\n", __func__, lseg, atomic_read(&lseg->pls_refcount)); - if (atomic_dec_and_test(&lseg->pls_refcount)) { - put_lseg_common(lseg); - list_add(&lseg->pls_list, tmp_list); + if (pnfs_lseg_dec_and_remove_zero(lseg, tmp_list)) rv = 1; - } } return rv; } @@ -385,7 +451,7 @@ static int mark_lseg_invalid(struct pnfs_layout_segment *lseg, * after call. */ int -mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo, +pnfs_mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo, struct list_head *tmp_list, struct pnfs_layout_range *recall_range) { @@ -394,11 +460,8 @@ mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo, dprintk("%s:Begin lo %p\n", __func__, lo); - if (list_empty(&lo->plh_segs)) { - if (!test_and_set_bit(NFS_LAYOUT_DESTROYED, &lo->plh_flags)) - put_layout_hdr_locked(lo); + if (list_empty(&lo->plh_segs)) return 0; - } list_for_each_entry_safe(lseg, next, &lo->plh_segs, pls_list) if (!recall_range || should_free_lseg(&lseg->pls_range, recall_range)) { @@ -418,25 +481,13 @@ void pnfs_free_lseg_list(struct list_head *free_me) { struct pnfs_layout_segment *lseg, *tmp; - struct pnfs_layout_hdr *lo; if (list_empty(free_me)) return; - lo = list_first_entry(free_me, struct pnfs_layout_segment, - pls_list)->pls_layout; - - if (test_bit(NFS_LAYOUT_DESTROYED, &lo->plh_flags)) { - struct nfs_client *clp; - - clp = NFS_SERVER(lo->plh_inode)->nfs_client; - spin_lock(&clp->cl_lock); - list_del_init(&lo->plh_layouts); - spin_unlock(&clp->cl_lock); - } list_for_each_entry_safe(lseg, tmp, free_me, pls_list) { list_del(&lseg->pls_list); - free_lseg(lseg); + pnfs_free_lseg(lseg); } } @@ -450,43 +501,179 @@ pnfs_destroy_layout(struct nfs_inode *nfsi) lo = nfsi->layout; if (lo) { lo->plh_block_lgets++; /* permanently block new LAYOUTGETs */ - mark_matching_lsegs_invalid(lo, &tmp_list, NULL); + pnfs_mark_matching_lsegs_invalid(lo, &tmp_list, NULL); + pnfs_get_layout_hdr(lo); + pnfs_layout_clear_fail_bit(lo, NFS_LAYOUT_RO_FAILED); + pnfs_layout_clear_fail_bit(lo, NFS_LAYOUT_RW_FAILED); + spin_unlock(&nfsi->vfs_inode.i_lock); + pnfs_free_lseg_list(&tmp_list); + pnfs_put_layout_hdr(lo); + } else + spin_unlock(&nfsi->vfs_inode.i_lock); +} +EXPORT_SYMBOL_GPL(pnfs_destroy_layout); + +static bool +pnfs_layout_add_bulk_destroy_list(struct inode *inode, + struct list_head *layout_list) +{ + struct pnfs_layout_hdr *lo; + bool ret = false; + + spin_lock(&inode->i_lock); + lo = NFS_I(inode)->layout; + if (lo != NULL && list_empty(&lo->plh_bulk_destroy)) { + pnfs_get_layout_hdr(lo); + list_add(&lo->plh_bulk_destroy, layout_list); + ret = true; } - spin_unlock(&nfsi->vfs_inode.i_lock); - pnfs_free_lseg_list(&tmp_list); + spin_unlock(&inode->i_lock); + return ret; } -/* - * Called by the state manger to remove all layouts established under an - * expired lease. - */ -void -pnfs_destroy_all_layouts(struct nfs_client *clp) +/* Caller must hold rcu_read_lock and clp->cl_lock */ +static int +pnfs_layout_bulk_destroy_byserver_locked(struct nfs_client *clp, + struct nfs_server *server, + struct list_head *layout_list) +{ + struct pnfs_layout_hdr *lo, *next; + struct inode *inode; + + list_for_each_entry_safe(lo, next, &server->layouts, plh_layouts) { + inode = igrab(lo->plh_inode); + if (inode == NULL) + continue; + list_del_init(&lo->plh_layouts); + if (pnfs_layout_add_bulk_destroy_list(inode, layout_list)) + continue; + rcu_read_unlock(); + spin_unlock(&clp->cl_lock); + iput(inode); + spin_lock(&clp->cl_lock); + rcu_read_lock(); + return -EAGAIN; + } + return 0; +} + +static int +pnfs_layout_free_bulk_destroy_list(struct list_head *layout_list, + bool is_bulk_recall) { - struct nfs_server *server; struct pnfs_layout_hdr *lo; - LIST_HEAD(tmp_list); + struct inode *inode; + struct pnfs_layout_range range = { + .iomode = IOMODE_ANY, + .offset = 0, + .length = NFS4_MAX_UINT64, + }; + LIST_HEAD(lseg_list); + int ret = 0; - nfs4_deviceid_mark_client_invalid(clp); - nfs4_deviceid_purge_client(clp); + while (!list_empty(layout_list)) { + lo = list_entry(layout_list->next, struct pnfs_layout_hdr, + plh_bulk_destroy); + dprintk("%s freeing layout for inode %lu\n", __func__, + lo->plh_inode->i_ino); + inode = lo->plh_inode; + spin_lock(&inode->i_lock); + list_del_init(&lo->plh_bulk_destroy); + lo->plh_block_lgets++; /* permanently block new LAYOUTGETs */ + if (is_bulk_recall) + set_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags); + if (pnfs_mark_matching_lsegs_invalid(lo, &lseg_list, &range)) + ret = -EAGAIN; + spin_unlock(&inode->i_lock); + pnfs_free_lseg_list(&lseg_list); + pnfs_put_layout_hdr(lo); + iput(inode); + } + return ret; +} + +int +pnfs_destroy_layouts_byfsid(struct nfs_client *clp, + struct nfs_fsid *fsid, + bool is_recall) +{ + struct nfs_server *server; + LIST_HEAD(layout_list); spin_lock(&clp->cl_lock); rcu_read_lock(); +restart: list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) { - if (!list_empty(&server->layouts)) - list_splice_init(&server->layouts, &tmp_list); + if (memcmp(&server->fsid, fsid, sizeof(*fsid)) != 0) + continue; + if (pnfs_layout_bulk_destroy_byserver_locked(clp, + server, + &layout_list) != 0) + goto restart; } rcu_read_unlock(); spin_unlock(&clp->cl_lock); - while (!list_empty(&tmp_list)) { - lo = list_entry(tmp_list.next, struct pnfs_layout_hdr, - plh_layouts); - dprintk("%s freeing layout for inode %lu\n", __func__, - lo->plh_inode->i_ino); - list_del_init(&lo->plh_layouts); - pnfs_destroy_layout(NFS_I(lo->plh_inode)); + if (list_empty(&layout_list)) + return 0; + return pnfs_layout_free_bulk_destroy_list(&layout_list, is_recall); +} + +int +pnfs_destroy_layouts_byclid(struct nfs_client *clp, + bool is_recall) +{ + struct nfs_server *server; + LIST_HEAD(layout_list); + + spin_lock(&clp->cl_lock); + rcu_read_lock(); +restart: + list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) { + if (pnfs_layout_bulk_destroy_byserver_locked(clp, + server, + &layout_list) != 0) + goto restart; } + rcu_read_unlock(); + spin_unlock(&clp->cl_lock); + + if (list_empty(&layout_list)) + return 0; + return pnfs_layout_free_bulk_destroy_list(&layout_list, is_recall); +} + +/* + * Called by the state manger to remove all layouts established under an + * expired lease. + */ +void +pnfs_destroy_all_layouts(struct nfs_client *clp) +{ + nfs4_deviceid_mark_client_invalid(clp); + nfs4_deviceid_purge_client(clp); + + pnfs_destroy_layouts_byclid(clp, false); +} + +/* + * Compare 2 layout stateid sequence ids, to see which is newer, + * taking into account wraparound issues. + */ +static bool pnfs_seqid_is_newer(u32 s1, u32 s2) +{ + return (s32)(s1 - s2) > 0; +} + +static void +pnfs_verify_layout_stateid(struct pnfs_layout_hdr *lo, + const nfs4_stateid *new, + struct list_head *free_me_list) +{ + if (nfs4_stateid_match_other(&lo->plh_stateid, new)) + return; + /* Layout is new! Kill existing layout segments */ + pnfs_mark_matching_lsegs_invalid(lo, free_me_list, NULL); } /* update lo->plh_stateid with new if is more recent */ @@ -494,41 +681,40 @@ void pnfs_set_layout_stateid(struct pnfs_layout_hdr *lo, const nfs4_stateid *new, bool update_barrier) { - u32 oldseq, newseq; + u32 oldseq, newseq, new_barrier; + int empty = list_empty(&lo->plh_segs); - oldseq = be32_to_cpu(lo->plh_stateid.stateid.seqid); - newseq = be32_to_cpu(new->stateid.seqid); - if ((int)(newseq - oldseq) > 0) { - memcpy(&lo->plh_stateid, &new->stateid, sizeof(new->stateid)); + oldseq = be32_to_cpu(lo->plh_stateid.seqid); + newseq = be32_to_cpu(new->seqid); + if (empty || pnfs_seqid_is_newer(newseq, oldseq)) { + nfs4_stateid_copy(&lo->plh_stateid, new); if (update_barrier) { - u32 new_barrier = be32_to_cpu(new->stateid.seqid); - - if ((int)(new_barrier - lo->plh_barrier)) - lo->plh_barrier = new_barrier; + new_barrier = be32_to_cpu(new->seqid); } else { /* Because of wraparound, we want to keep the barrier - * "close" to the current seqids. It needs to be - * within 2**31 to count as "behind", so if it - * gets too near that limit, give us a litle leeway - * and bring it to within 2**30. - * NOTE - and yes, this is all unsigned arithmetic. + * "close" to the current seqids. */ - if (unlikely((newseq - lo->plh_barrier) > (3 << 29))) - lo->plh_barrier = newseq - (1 << 30); + new_barrier = newseq - atomic_read(&lo->plh_outstanding); } + if (empty || pnfs_seqid_is_newer(new_barrier, lo->plh_barrier)) + lo->plh_barrier = new_barrier; } } +static bool +pnfs_layout_stateid_blocked(const struct pnfs_layout_hdr *lo, + const nfs4_stateid *stateid) +{ + u32 seqid = be32_to_cpu(stateid->seqid); + + return !pnfs_seqid_is_newer(seqid, lo->plh_barrier); +} + /* lget is set to 1 if called from inside send_layoutget call chain */ static bool -pnfs_layoutgets_blocked(struct pnfs_layout_hdr *lo, nfs4_stateid *stateid, - int lget) +pnfs_layoutgets_blocked(const struct pnfs_layout_hdr *lo, int lget) { - if ((stateid) && - (int)(lo->plh_barrier - be32_to_cpu(stateid->stateid.seqid)) >= 0) - return true; return lo->plh_block_lgets || - test_bit(NFS_LAYOUT_DESTROYED, &lo->plh_flags) || test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags) || (list_empty(&lo->plh_segs) && (atomic_read(&lo->plh_outstanding) > lget)); @@ -542,18 +728,19 @@ pnfs_choose_layoutget_stateid(nfs4_stateid *dst, struct pnfs_layout_hdr *lo, dprintk("--> %s\n", __func__); spin_lock(&lo->plh_inode->i_lock); - if (pnfs_layoutgets_blocked(lo, NULL, 1)) { + if (pnfs_layoutgets_blocked(lo, 1)) { status = -EAGAIN; + } else if (!nfs4_valid_open_stateid(open_state)) { + status = -EBADF; } else if (list_empty(&lo->plh_segs)) { int seq; do { seq = read_seqbegin(&open_state->seqlock); - memcpy(dst->data, open_state->stateid.data, - sizeof(open_state->stateid.data)); + nfs4_stateid_copy(dst, &open_state->stateid); } while (read_seqretry(&open_state->seqlock, seq)); } else - memcpy(dst->data, lo->plh_stateid.data, sizeof(lo->plh_stateid.data)); + nfs4_stateid_copy(dst, &lo->plh_stateid); spin_unlock(&lo->plh_inode->i_lock); dprintk("<-- %s\n", __func__); return status; @@ -574,32 +761,14 @@ send_layoutget(struct pnfs_layout_hdr *lo, struct inode *ino = lo->plh_inode; struct nfs_server *server = NFS_SERVER(ino); struct nfs4_layoutget *lgp; - struct pnfs_layout_segment *lseg = NULL; - struct page **pages = NULL; - int i; - u32 max_resp_sz, max_pages; + struct pnfs_layout_segment *lseg; dprintk("--> %s\n", __func__); - BUG_ON(ctx == NULL); lgp = kzalloc(sizeof(*lgp), gfp_flags); if (lgp == NULL) return NULL; - /* allocate pages for xdr post processing */ - max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz; - max_pages = max_resp_sz >> PAGE_SHIFT; - - pages = kzalloc(max_pages * sizeof(struct page *), gfp_flags); - if (!pages) - goto out_err_free; - - for (i = 0; i < max_pages; i++) { - pages[i] = alloc_page(gfp_flags); - if (!pages[i]) - goto out_err_free; - } - lgp->args.minlength = PAGE_CACHE_SIZE; if (lgp->args.minlength > range->length) lgp->args.minlength = range->length; @@ -608,42 +777,51 @@ send_layoutget(struct pnfs_layout_hdr *lo, lgp->args.type = server->pnfs_curr_ld->id; lgp->args.inode = ino; lgp->args.ctx = get_nfs_open_context(ctx); - lgp->args.layout.pages = pages; - lgp->args.layout.pglen = max_pages * PAGE_SIZE; - lgp->lsegpp = &lseg; lgp->gfp_flags = gfp_flags; + lgp->cred = lo->plh_lc_cred; /* Synchronously retrieve layout information from server and * store in lseg. */ - nfs4_proc_layoutget(lgp); - if (!lseg) { - /* remember that LAYOUTGET failed and suspend trying */ - set_bit(lo_fail_bit(range->iomode), &lo->plh_flags); + lseg = nfs4_proc_layoutget(lgp, gfp_flags); + if (IS_ERR(lseg)) { + switch (PTR_ERR(lseg)) { + case -ENOMEM: + case -ERESTARTSYS: + break; + default: + /* remember that LAYOUTGET failed and suspend trying */ + pnfs_layout_io_set_failed(lo, range->iomode); + } + return NULL; } - /* free xdr pages */ - for (i = 0; i < max_pages; i++) - __free_page(pages[i]); - kfree(pages); - return lseg; +} -out_err_free: - /* free any allocated xdr pages, lgp as it's not used */ - if (pages) { - for (i = 0; i < max_pages; i++) { - if (!pages[i]) - break; - __free_page(pages[i]); - } - kfree(pages); +static void pnfs_clear_layoutcommit(struct inode *inode, + struct list_head *head) +{ + struct nfs_inode *nfsi = NFS_I(inode); + struct pnfs_layout_segment *lseg, *tmp; + + if (!test_and_clear_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)) + return; + list_for_each_entry_safe(lseg, tmp, &nfsi->layout->plh_segs, pls_list) { + if (!test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT, &lseg->pls_flags)) + continue; + pnfs_lseg_dec_and_remove_zero(lseg, head); } - kfree(lgp); - return NULL; } -/* Initiates a LAYOUTRETURN(FILE) */ +/* + * Initiates a LAYOUTRETURN(FILE), and removes the pnfs_layout_hdr + * when the layout segment list is empty. + * + * Note that a pnfs_layout_hdr can exist with an empty layout segment + * list when LAYOUTGET has failed, or when LAYOUTGET succeeded, but the + * deviceid is marked invalid. + */ int _pnfs_return_layout(struct inode *ino) { @@ -652,33 +830,41 @@ _pnfs_return_layout(struct inode *ino) LIST_HEAD(tmp_list); struct nfs4_layoutreturn *lrp; nfs4_stateid stateid; - int status = 0; + int status = 0, empty; - dprintk("--> %s\n", __func__); + dprintk("NFS: %s for inode %lu\n", __func__, ino->i_ino); spin_lock(&ino->i_lock); lo = nfsi->layout; if (!lo) { spin_unlock(&ino->i_lock); - dprintk("%s: no layout to return\n", __func__); - return status; + dprintk("NFS: %s no layout to return\n", __func__); + goto out; } stateid = nfsi->layout->plh_stateid; /* Reference matched in nfs4_layoutreturn_release */ - get_layout_hdr(lo); - mark_matching_lsegs_invalid(lo, &tmp_list, NULL); + pnfs_get_layout_hdr(lo); + empty = list_empty(&lo->plh_segs); + pnfs_clear_layoutcommit(ino, &tmp_list); + pnfs_mark_matching_lsegs_invalid(lo, &tmp_list, NULL); + /* Don't send a LAYOUTRETURN if list was initially empty */ + if (empty) { + spin_unlock(&ino->i_lock); + pnfs_put_layout_hdr(lo); + dprintk("NFS: %s no layout segments to return\n", __func__); + goto out; + } lo->plh_block_lgets++; spin_unlock(&ino->i_lock); pnfs_free_lseg_list(&tmp_list); - WARN_ON(test_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)); - lrp = kzalloc(sizeof(*lrp), GFP_KERNEL); if (unlikely(lrp == NULL)) { status = -ENOMEM; - set_bit(NFS_LAYOUT_RW_FAILED, &lo->plh_flags); - set_bit(NFS_LAYOUT_RO_FAILED, &lo->plh_flags); - put_layout_hdr(lo); + spin_lock(&ino->i_lock); + lo->plh_block_lgets--; + spin_unlock(&ino->i_lock); + pnfs_put_layout_hdr(lo); goto out; } @@ -687,12 +873,41 @@ _pnfs_return_layout(struct inode *ino) lrp->args.inode = ino; lrp->args.layout = lo; lrp->clp = NFS_SERVER(ino)->nfs_client; + lrp->cred = lo->plh_lc_cred; status = nfs4_proc_layoutreturn(lrp); out: dprintk("<-- %s status: %d\n", __func__, status); return status; } +EXPORT_SYMBOL_GPL(_pnfs_return_layout); + +int +pnfs_commit_and_return_layout(struct inode *inode) +{ + struct pnfs_layout_hdr *lo; + int ret; + + spin_lock(&inode->i_lock); + lo = NFS_I(inode)->layout; + if (lo == NULL) { + spin_unlock(&inode->i_lock); + return 0; + } + pnfs_get_layout_hdr(lo); + /* Block new layoutgets and read/write to ds */ + lo->plh_block_lgets++; + spin_unlock(&inode->i_lock); + filemap_fdatawait(inode->i_mapping); + ret = pnfs_layoutcommit_inode(inode, true); + if (ret == 0) + ret = _pnfs_return_layout(inode); + spin_lock(&inode->i_lock); + lo->plh_block_lgets--; + spin_unlock(&inode->i_lock); + pnfs_put_layout_hdr(lo); + return ret; +} bool pnfs_roc(struct inode *ino) { @@ -714,7 +929,7 @@ bool pnfs_roc(struct inode *ino) if (!found) goto out_nolayout; lo->plh_block_lgets++; - get_layout_hdr(lo); /* matched in pnfs_roc_release */ + pnfs_get_layout_hdr(lo); /* matched in pnfs_roc_release */ spin_unlock(&ino->i_lock); pnfs_free_lseg_list(&tmp_list); return true; @@ -731,8 +946,12 @@ void pnfs_roc_release(struct inode *ino) spin_lock(&ino->i_lock); lo = NFS_I(ino)->layout; lo->plh_block_lgets--; - put_layout_hdr_locked(lo); - spin_unlock(&ino->i_lock); + if (atomic_dec_and_test(&lo->plh_refcount)) { + pnfs_detach_layout_hdr(lo); + spin_unlock(&ino->i_lock); + pnfs_free_layout_hdr(lo); + } else + spin_unlock(&ino->i_lock); } void pnfs_roc_set_barrier(struct inode *ino, u32 barrier) @@ -741,32 +960,34 @@ void pnfs_roc_set_barrier(struct inode *ino, u32 barrier) spin_lock(&ino->i_lock); lo = NFS_I(ino)->layout; - if ((int)(barrier - lo->plh_barrier) > 0) + if (pnfs_seqid_is_newer(barrier, lo->plh_barrier)) lo->plh_barrier = barrier; spin_unlock(&ino->i_lock); } -bool pnfs_roc_drain(struct inode *ino, u32 *barrier) +bool pnfs_roc_drain(struct inode *ino, u32 *barrier, struct rpc_task *task) { struct nfs_inode *nfsi = NFS_I(ino); + struct pnfs_layout_hdr *lo; struct pnfs_layout_segment *lseg; + u32 current_seqid; bool found = false; spin_lock(&ino->i_lock); list_for_each_entry(lseg, &nfsi->layout->plh_segs, pls_list) if (test_bit(NFS_LSEG_ROC, &lseg->pls_flags)) { + rpc_sleep_on(&NFS_SERVER(ino)->roc_rpcwaitq, task, NULL); found = true; - break; + goto out; } - if (!found) { - struct pnfs_layout_hdr *lo = nfsi->layout; - u32 current_seqid = be32_to_cpu(lo->plh_stateid.stateid.seqid); + lo = nfsi->layout; + current_seqid = be32_to_cpu(lo->plh_stateid.seqid); - /* Since close does not return a layout stateid for use as - * a barrier, we choose the worst-case barrier. - */ - *barrier = current_seqid + atomic_read(&lo->plh_outstanding); - } + /* Since close does not return a layout stateid for use as + * a barrier, we choose the worst-case barrier. + */ + *barrier = current_seqid + atomic_read(&lo->plh_outstanding); +out: spin_unlock(&ino->i_lock); return found; } @@ -777,8 +998,8 @@ bool pnfs_roc_drain(struct inode *ino, u32 *barrier) * are seen first. */ static s64 -cmp_layout(struct pnfs_layout_range *l1, - struct pnfs_layout_range *l2) +pnfs_lseg_range_cmp(const struct pnfs_layout_range *l1, + const struct pnfs_layout_range *l2) { s64 d; @@ -797,16 +1018,15 @@ cmp_layout(struct pnfs_layout_range *l1, } static void -pnfs_insert_layout(struct pnfs_layout_hdr *lo, +pnfs_layout_insert_lseg(struct pnfs_layout_hdr *lo, struct pnfs_layout_segment *lseg) { struct pnfs_layout_segment *lp; dprintk("%s:Begin\n", __func__); - assert_spin_locked(&lo->plh_inode->i_lock); list_for_each_entry(lp, &lo->plh_segs, pls_list) { - if (cmp_layout(&lseg->pls_range, &lp->pls_range) > 0) + if (pnfs_lseg_range_cmp(&lseg->pls_range, &lp->pls_range) > 0) continue; list_add_tail(&lseg->pls_list, &lp->pls_list); dprintk("%s: inserted lseg %p " @@ -824,7 +1044,7 @@ pnfs_insert_layout(struct pnfs_layout_hdr *lo, __func__, lseg, lseg->pls_range.iomode, lseg->pls_range.offset, lseg->pls_range.length); out: - get_layout_hdr(lo); + pnfs_get_layout_hdr(lo); dprintk("%s:Return\n", __func__); } @@ -842,9 +1062,9 @@ alloc_init_layout_hdr(struct inode *ino, atomic_set(&lo->plh_refcount, 1); INIT_LIST_HEAD(&lo->plh_layouts); INIT_LIST_HEAD(&lo->plh_segs); - INIT_LIST_HEAD(&lo->plh_bulk_recall); + INIT_LIST_HEAD(&lo->plh_bulk_destroy); lo->plh_inode = ino; - lo->plh_lc_cred = get_rpccred(ctx->state->owner->so_cred); + lo->plh_lc_cred = get_rpccred(ctx->cred); return lo; } @@ -858,21 +1078,19 @@ pnfs_find_alloc_layout(struct inode *ino, dprintk("%s Begin ino=%p layout=%p\n", __func__, ino, nfsi->layout); - assert_spin_locked(&ino->i_lock); - if (nfsi->layout) { - if (test_bit(NFS_LAYOUT_DESTROYED, &nfsi->layout->plh_flags)) - return NULL; - else - return nfsi->layout; - } + if (nfsi->layout != NULL) + goto out_existing; spin_unlock(&ino->i_lock); new = alloc_init_layout_hdr(ino, ctx, gfp_flags); spin_lock(&ino->i_lock); - if (likely(nfsi->layout == NULL)) /* Won the race? */ + if (likely(nfsi->layout == NULL)) { /* Won the race? */ nfsi->layout = new; - else + return new; + } else if (new != NULL) pnfs_free_layout_hdr(new); +out_existing: + pnfs_get_layout_hdr(nfsi->layout); return nfsi->layout; } @@ -887,21 +1105,21 @@ pnfs_find_alloc_layout(struct inode *ino, * READ READ true * READ RW true */ -static int -is_matching_lseg(struct pnfs_layout_range *ls_range, - struct pnfs_layout_range *range) +static bool +pnfs_lseg_range_match(const struct pnfs_layout_range *ls_range, + const struct pnfs_layout_range *range) { struct pnfs_layout_range range1; if ((range->iomode == IOMODE_RW && ls_range->iomode != IOMODE_RW) || - !lo_seg_intersecting(ls_range, range)) + !pnfs_lseg_range_intersecting(ls_range, range)) return 0; /* range1 covers only the first byte in the range */ range1 = *range; range1.length = 1; - return lo_seg_contained(ls_range, &range1); + return pnfs_lseg_range_contained(ls_range, &range1); } /* @@ -915,11 +1133,10 @@ pnfs_find_lseg(struct pnfs_layout_hdr *lo, dprintk("%s:Begin\n", __func__); - assert_spin_locked(&lo->plh_inode->i_lock); list_for_each_entry(lseg, &lo->plh_segs, pls_list) { if (test_bit(NFS_LSEG_VALID, &lseg->pls_flags) && - is_matching_lseg(&lseg->pls_range, range)) { - ret = get_lseg(lseg); + pnfs_lseg_range_match(&lseg->pls_range, range)) { + ret = pnfs_get_lseg(lseg); break; } if (lseg->pls_range.offset > range->offset) @@ -932,6 +1149,81 @@ pnfs_find_lseg(struct pnfs_layout_hdr *lo, } /* + * Use mdsthreshold hints set at each OPEN to determine if I/O should go + * to the MDS or over pNFS + * + * The nfs_inode read_io and write_io fields are cumulative counters reset + * when there are no layout segments. Note that in pnfs_update_layout iomode + * is set to IOMODE_READ for a READ request, and set to IOMODE_RW for a + * WRITE request. + * + * A return of true means use MDS I/O. + * + * From rfc 5661: + * If a file's size is smaller than the file size threshold, data accesses + * SHOULD be sent to the metadata server. If an I/O request has a length that + * is below the I/O size threshold, the I/O SHOULD be sent to the metadata + * server. If both file size and I/O size are provided, the client SHOULD + * reach or exceed both thresholds before sending its read or write + * requests to the data server. + */ +static bool pnfs_within_mdsthreshold(struct nfs_open_context *ctx, + struct inode *ino, int iomode) +{ + struct nfs4_threshold *t = ctx->mdsthreshold; + struct nfs_inode *nfsi = NFS_I(ino); + loff_t fsize = i_size_read(ino); + bool size = false, size_set = false, io = false, io_set = false, ret = false; + + if (t == NULL) + return ret; + + dprintk("%s bm=0x%x rd_sz=%llu wr_sz=%llu rd_io=%llu wr_io=%llu\n", + __func__, t->bm, t->rd_sz, t->wr_sz, t->rd_io_sz, t->wr_io_sz); + + switch (iomode) { + case IOMODE_READ: + if (t->bm & THRESHOLD_RD) { + dprintk("%s fsize %llu\n", __func__, fsize); + size_set = true; + if (fsize < t->rd_sz) + size = true; + } + if (t->bm & THRESHOLD_RD_IO) { + dprintk("%s nfsi->read_io %llu\n", __func__, + nfsi->read_io); + io_set = true; + if (nfsi->read_io < t->rd_io_sz) + io = true; + } + break; + case IOMODE_RW: + if (t->bm & THRESHOLD_WR) { + dprintk("%s fsize %llu\n", __func__, fsize); + size_set = true; + if (fsize < t->wr_sz) + size = true; + } + if (t->bm & THRESHOLD_WR_IO) { + dprintk("%s nfsi->write_io %llu\n", __func__, + nfsi->write_io); + io_set = true; + if (nfsi->write_io < t->wr_io_sz) + io = true; + } + break; + } + if (size_set && io_set) { + if (size && io) + ret = true; + } else if (size || io) + ret = true; + + dprintk("<-- %s size %d io %d ret %d\n", __func__, size, io, ret); + return ret; +} + +/* * Layout segment is retreived from the server if not cached. * The appropriate layout segment is referenced and returned to the caller. */ @@ -949,31 +1241,33 @@ pnfs_update_layout(struct inode *ino, .length = count, }; unsigned pg_offset; - struct nfs_inode *nfsi = NFS_I(ino); struct nfs_server *server = NFS_SERVER(ino); struct nfs_client *clp = server->nfs_client; struct pnfs_layout_hdr *lo; struct pnfs_layout_segment *lseg = NULL; - bool first = false; + bool first; if (!pnfs_enabled_sb(NFS_SERVER(ino))) - return NULL; + goto out; + + if (pnfs_within_mdsthreshold(ctx, ino, iomode)) + goto out; + spin_lock(&ino->i_lock); lo = pnfs_find_alloc_layout(ino, ctx, gfp_flags); if (lo == NULL) { - dprintk("%s ERROR: can't get pnfs_layout_hdr\n", __func__); - goto out_unlock; + spin_unlock(&ino->i_lock); + goto out; } /* Do we even need to bother with this? */ - if (test_bit(NFS4CLNT_LAYOUTRECALL, &clp->cl_state) || - test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags)) { + if (test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags)) { dprintk("%s matches recall, use MDS\n", __func__); goto out_unlock; } /* if LAYOUTGET already failed once we don't try again */ - if (test_bit(lo_fail_bit(iomode), &nfsi->layout->plh_flags)) + if (pnfs_layout_io_test_failed(lo, iomode)) goto out_unlock; /* Check to see if the layout for the given range already exists */ @@ -981,20 +1275,18 @@ pnfs_update_layout(struct inode *ino, if (lseg) goto out_unlock; - if (pnfs_layoutgets_blocked(lo, NULL, 0)) + if (pnfs_layoutgets_blocked(lo, 0)) goto out_unlock; atomic_inc(&lo->plh_outstanding); - get_layout_hdr(lo); - if (list_empty(&lo->plh_segs)) - first = true; + first = list_empty(&lo->plh_layouts) ? true : false; spin_unlock(&ino->i_lock); + if (first) { /* The lo must be on the clp list if there is any * chance of a CB_LAYOUTRECALL(FILE) coming in. */ spin_lock(&clp->cl_lock); - BUG_ON(!list_empty(&lo->plh_layouts)); list_add_tail(&lo->plh_layouts, &server->layouts); spin_unlock(&clp->cl_lock); } @@ -1008,31 +1300,33 @@ pnfs_update_layout(struct inode *ino, arg.length = PAGE_CACHE_ALIGN(arg.length); lseg = send_layoutget(lo, ctx, &arg, gfp_flags); - if (!lseg && first) { - spin_lock(&clp->cl_lock); - list_del_init(&lo->plh_layouts); - spin_unlock(&clp->cl_lock); - } atomic_dec(&lo->plh_outstanding); - put_layout_hdr(lo); +out_put_layout_hdr: + pnfs_put_layout_hdr(lo); out: - dprintk("%s end, state 0x%lx lseg %p\n", __func__, - nfsi->layout ? nfsi->layout->plh_flags : -1, lseg); + dprintk("%s: inode %s/%llu pNFS layout segment %s for " + "(%s, offset: %llu, length: %llu)\n", + __func__, ino->i_sb->s_id, + (unsigned long long)NFS_FILEID(ino), + lseg == NULL ? "not found" : "found", + iomode==IOMODE_RW ? "read/write" : "read-only", + (unsigned long long)pos, + (unsigned long long)count); return lseg; out_unlock: spin_unlock(&ino->i_lock); - goto out; + goto out_put_layout_hdr; } EXPORT_SYMBOL_GPL(pnfs_update_layout); -int +struct pnfs_layout_segment * pnfs_layout_process(struct nfs4_layoutget *lgp) { struct pnfs_layout_hdr *lo = NFS_I(lgp->args.inode)->layout; struct nfs4_layoutget_res *res = &lgp->res; struct pnfs_layout_segment *lseg; struct inode *ino = lo->plh_inode; - struct nfs_client *clp = NFS_SERVER(ino)->nfs_client; + LIST_HEAD(free_me); int status = 0; /* Inject layout blob into I/O device driver */ @@ -1048,31 +1342,37 @@ pnfs_layout_process(struct nfs4_layoutget *lgp) } spin_lock(&ino->i_lock); - if (test_bit(NFS4CLNT_LAYOUTRECALL, &clp->cl_state) || - test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags)) { + if (test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags)) { dprintk("%s forget reply due to recall\n", __func__); goto out_forget_reply; } - if (pnfs_layoutgets_blocked(lo, &res->stateid, 1)) { + if (pnfs_layoutgets_blocked(lo, 1) || + pnfs_layout_stateid_blocked(lo, &res->stateid)) { dprintk("%s forget reply due to state\n", __func__); goto out_forget_reply; } + + /* Check that the new stateid matches the old stateid */ + pnfs_verify_layout_stateid(lo, &res->stateid, &free_me); + /* Done processing layoutget. Set the layout stateid */ + pnfs_set_layout_stateid(lo, &res->stateid, false); + init_lseg(lo, lseg); lseg->pls_range = res->range; - *lgp->lsegpp = get_lseg(lseg); - pnfs_insert_layout(lo, lseg); + pnfs_get_lseg(lseg); + pnfs_layout_insert_lseg(lo, lseg); if (res->return_on_close) { set_bit(NFS_LSEG_ROC, &lseg->pls_flags); set_bit(NFS_LAYOUT_ROC, &lo->plh_flags); } - /* Done processing layoutget. Set the layout stateid */ - pnfs_set_layout_stateid(lo, &res->stateid, false); spin_unlock(&ino->i_lock); + pnfs_free_lseg_list(&free_me); + return lseg; out: - return status; + return ERR_PTR(status); out_forget_reply: spin_unlock(&ino->i_lock); @@ -1084,12 +1384,19 @@ out_forget_reply: void pnfs_generic_pg_init_read(struct nfs_pageio_descriptor *pgio, struct nfs_page *req) { - BUG_ON(pgio->pg_lseg != NULL); + u64 rd_size = req->wb_bytes; + + WARN_ON_ONCE(pgio->pg_lseg != NULL); + + if (pgio->pg_dreq == NULL) + rd_size = i_size_read(pgio->pg_inode) - req_offset(req); + else + rd_size = nfs_dreq_bytes_left(pgio->pg_dreq); pgio->pg_lseg = pnfs_update_layout(pgio->pg_inode, req->wb_context, req_offset(req), - req->wb_bytes, + rd_size, IOMODE_READ, GFP_KERNEL); /* If no lseg, fall back to read through mds */ @@ -1100,14 +1407,15 @@ pnfs_generic_pg_init_read(struct nfs_pageio_descriptor *pgio, struct nfs_page *r EXPORT_SYMBOL_GPL(pnfs_generic_pg_init_read); void -pnfs_generic_pg_init_write(struct nfs_pageio_descriptor *pgio, struct nfs_page *req) +pnfs_generic_pg_init_write(struct nfs_pageio_descriptor *pgio, + struct nfs_page *req, u64 wb_size) { - BUG_ON(pgio->pg_lseg != NULL); + WARN_ON_ONCE(pgio->pg_lseg != NULL); pgio->pg_lseg = pnfs_update_layout(pgio->pg_inode, req->wb_context, req_offset(req), - req->wb_bytes, + wb_size, IOMODE_RW, GFP_NOFS); /* If no lseg, fall back to write through mds */ @@ -1116,63 +1424,63 @@ pnfs_generic_pg_init_write(struct nfs_pageio_descriptor *pgio, struct nfs_page * } EXPORT_SYMBOL_GPL(pnfs_generic_pg_init_write); -bool -pnfs_pageio_init_read(struct nfs_pageio_descriptor *pgio, struct inode *inode) -{ - struct nfs_server *server = NFS_SERVER(inode); - struct pnfs_layoutdriver_type *ld = server->pnfs_curr_ld; - - if (ld == NULL) - return false; - nfs_pageio_init(pgio, inode, ld->pg_read_ops, server->rsize, 0); - return true; -} - -bool -pnfs_pageio_init_write(struct nfs_pageio_descriptor *pgio, struct inode *inode, int ioflags) -{ - struct nfs_server *server = NFS_SERVER(inode); - struct pnfs_layoutdriver_type *ld = server->pnfs_curr_ld; - - if (ld == NULL) - return false; - nfs_pageio_init(pgio, inode, ld->pg_write_ops, server->wsize, ioflags); - return true; -} - -bool +/* + * Return 0 if @req cannot be coalesced into @pgio, otherwise return the number + * of bytes (maximum @req->wb_bytes) that can be coalesced. + */ +size_t pnfs_generic_pg_test(struct nfs_pageio_descriptor *pgio, struct nfs_page *prev, struct nfs_page *req) { - if (pgio->pg_lseg == NULL) - return nfs_generic_pg_test(pgio, prev, req); + unsigned int size; + u64 seg_end, req_start, seg_left; + + size = nfs_generic_pg_test(pgio, prev, req); + if (!size) + return 0; /* - * Test if a nfs_page is fully contained in the pnfs_layout_range. - * Note that this test makes several assumptions: - * - that the previous nfs_page in the struct nfs_pageio_descriptor - * is known to lie within the range. - * - that the nfs_page being tested is known to be contiguous with the - * previous nfs_page. - * - Layout ranges are page aligned, so we only have to test the - * start offset of the request. + * 'size' contains the number of bytes left in the current page (up + * to the original size asked for in @req->wb_bytes). + * + * Calculate how many bytes are left in the layout segment + * and if there are less bytes than 'size', return that instead. * * Please also note that 'end_offset' is actually the offset of the * first byte that lies outside the pnfs_layout_range. FIXME? * */ - return req_offset(req) < end_offset(pgio->pg_lseg->pls_range.offset, - pgio->pg_lseg->pls_range.length); + if (pgio->pg_lseg) { + seg_end = end_offset(pgio->pg_lseg->pls_range.offset, + pgio->pg_lseg->pls_range.length); + req_start = req_offset(req); + WARN_ON_ONCE(req_start > seg_end); + /* start of request is past the last byte of this segment */ + if (req_start >= seg_end) + return 0; + + /* adjust 'size' iff there are fewer bytes left in the + * segment than what nfs_generic_pg_test returned */ + seg_left = seg_end - req_start; + if (seg_left < size) + size = (unsigned int)seg_left; + } + + return size; } EXPORT_SYMBOL_GPL(pnfs_generic_pg_test); -static int pnfs_write_done_resend_to_mds(struct inode *inode, struct list_head *head) +int pnfs_write_done_resend_to_mds(struct inode *inode, + struct list_head *head, + const struct nfs_pgio_completion_ops *compl_ops, + struct nfs_direct_req *dreq) { struct nfs_pageio_descriptor pgio; LIST_HEAD(failed); /* Resend all requests through the MDS */ - nfs_pageio_init_write_mds(&pgio, inode, FLUSH_STABLE); + nfs_pageio_init_write(&pgio, inode, FLUSH_STABLE, true, compl_ops); + pgio.pg_dreq = dreq; while (!list_empty(head)) { struct nfs_page *req = nfs_list_entry(head->next); @@ -1192,224 +1500,292 @@ static int pnfs_write_done_resend_to_mds(struct inode *inode, struct list_head * } return 0; } +EXPORT_SYMBOL_GPL(pnfs_write_done_resend_to_mds); + +static void pnfs_ld_handle_write_error(struct nfs_pgio_data *data) +{ + struct nfs_pgio_header *hdr = data->header; + + dprintk("pnfs write error = %d\n", hdr->pnfs_error); + if (NFS_SERVER(hdr->inode)->pnfs_curr_ld->flags & + PNFS_LAYOUTRET_ON_ERROR) { + pnfs_return_layout(hdr->inode); + } + if (!test_and_set_bit(NFS_IOHDR_REDO, &hdr->flags)) + data->task.tk_status = pnfs_write_done_resend_to_mds(hdr->inode, + &hdr->pages, + hdr->completion_ops, + hdr->dreq); +} /* * Called by non rpc-based layout drivers */ -void pnfs_ld_write_done(struct nfs_write_data *data) +void pnfs_ld_write_done(struct nfs_pgio_data *data) { - if (likely(!data->pnfs_error)) { + struct nfs_pgio_header *hdr = data->header; + + trace_nfs4_pnfs_write(data, hdr->pnfs_error); + if (!hdr->pnfs_error) { pnfs_set_layoutcommit(data); - data->mds_ops->rpc_call_done(&data->task, data); - } else { - dprintk("pnfs write error = %d\n", data->pnfs_error); - if (NFS_SERVER(data->inode)->pnfs_curr_ld->flags & - PNFS_LAYOUTRET_ON_ERROR) { - /* Don't lo_commit on error, Server will needs to - * preform a file recovery. - */ - clear_bit(NFS_INO_LAYOUTCOMMIT, - &NFS_I(data->inode)->flags); - pnfs_return_layout(data->inode); - } - data->task.tk_status = pnfs_write_done_resend_to_mds(data->inode, &data->pages); - } - data->mds_ops->rpc_release(data); + hdr->mds_ops->rpc_call_done(&data->task, data); + } else + pnfs_ld_handle_write_error(data); + hdr->mds_ops->rpc_release(data); } EXPORT_SYMBOL_GPL(pnfs_ld_write_done); static void pnfs_write_through_mds(struct nfs_pageio_descriptor *desc, - struct nfs_write_data *data) + struct nfs_pgio_data *data) { - list_splice_tail_init(&data->pages, &desc->pg_list); - if (data->req && list_empty(&data->req->wb_list)) - nfs_list_add_request(data->req, &desc->pg_list); - nfs_pageio_reset_write_mds(desc); - desc->pg_recoalesce = 1; - nfs_writedata_release(data); + struct nfs_pgio_header *hdr = data->header; + + if (!test_and_set_bit(NFS_IOHDR_REDO, &hdr->flags)) { + list_splice_tail_init(&hdr->pages, &desc->pg_list); + nfs_pageio_reset_write_mds(desc); + desc->pg_recoalesce = 1; + } + nfs_pgio_data_release(data); } static enum pnfs_try_status -pnfs_try_to_write_data(struct nfs_write_data *wdata, +pnfs_try_to_write_data(struct nfs_pgio_data *wdata, const struct rpc_call_ops *call_ops, struct pnfs_layout_segment *lseg, int how) { - struct inode *inode = wdata->inode; + struct nfs_pgio_header *hdr = wdata->header; + struct inode *inode = hdr->inode; enum pnfs_try_status trypnfs; struct nfs_server *nfss = NFS_SERVER(inode); - wdata->mds_ops = call_ops; - wdata->lseg = get_lseg(lseg); + hdr->mds_ops = call_ops; dprintk("%s: Writing ino:%lu %u@%llu (how %d)\n", __func__, inode->i_ino, wdata->args.count, wdata->args.offset, how); - trypnfs = nfss->pnfs_curr_ld->write_pagelist(wdata, how); - if (trypnfs == PNFS_NOT_ATTEMPTED) { - put_lseg(wdata->lseg); - wdata->lseg = NULL; - } else + if (trypnfs != PNFS_NOT_ATTEMPTED) nfs_inc_stats(inode, NFSIOS_PNFS_WRITE); - dprintk("%s End (trypnfs:%d)\n", __func__, trypnfs); return trypnfs; } static void -pnfs_do_multiple_writes(struct nfs_pageio_descriptor *desc, struct list_head *head, int how) +pnfs_do_write(struct nfs_pageio_descriptor *desc, + struct nfs_pgio_header *hdr, int how) { - struct nfs_write_data *data; + struct nfs_pgio_data *data = hdr->data; const struct rpc_call_ops *call_ops = desc->pg_rpc_callops; struct pnfs_layout_segment *lseg = desc->pg_lseg; + enum pnfs_try_status trypnfs; desc->pg_lseg = NULL; - while (!list_empty(head)) { - enum pnfs_try_status trypnfs; - - data = list_entry(head->next, struct nfs_write_data, list); - list_del_init(&data->list); + trypnfs = pnfs_try_to_write_data(data, call_ops, lseg, how); + if (trypnfs == PNFS_NOT_ATTEMPTED) + pnfs_write_through_mds(desc, data); + pnfs_put_lseg(lseg); +} - trypnfs = pnfs_try_to_write_data(data, call_ops, lseg, how); - if (trypnfs == PNFS_NOT_ATTEMPTED) - pnfs_write_through_mds(desc, data); - } - put_lseg(lseg); +static void pnfs_writehdr_free(struct nfs_pgio_header *hdr) +{ + pnfs_put_lseg(hdr->lseg); + nfs_rw_header_free(hdr); } +EXPORT_SYMBOL_GPL(pnfs_writehdr_free); int pnfs_generic_pg_writepages(struct nfs_pageio_descriptor *desc) { - LIST_HEAD(head); + struct nfs_rw_header *whdr; + struct nfs_pgio_header *hdr; int ret; - ret = nfs_generic_flush(desc, &head); - if (ret != 0) { - put_lseg(desc->pg_lseg); + whdr = nfs_rw_header_alloc(desc->pg_rw_ops); + if (!whdr) { + desc->pg_completion_ops->error_cleanup(&desc->pg_list); + pnfs_put_lseg(desc->pg_lseg); desc->pg_lseg = NULL; - return ret; + return -ENOMEM; } - pnfs_do_multiple_writes(desc, &head, desc->pg_ioflags); - return 0; + hdr = &whdr->header; + nfs_pgheader_init(desc, hdr, pnfs_writehdr_free); + hdr->lseg = pnfs_get_lseg(desc->pg_lseg); + atomic_inc(&hdr->refcnt); + ret = nfs_generic_pgio(desc, hdr); + if (ret != 0) { + pnfs_put_lseg(desc->pg_lseg); + desc->pg_lseg = NULL; + } else + pnfs_do_write(desc, hdr, desc->pg_ioflags); + if (atomic_dec_and_test(&hdr->refcnt)) + hdr->completion_ops->completion(hdr); + return ret; } EXPORT_SYMBOL_GPL(pnfs_generic_pg_writepages); -static void pnfs_ld_handle_read_error(struct nfs_read_data *data) +int pnfs_read_done_resend_to_mds(struct inode *inode, + struct list_head *head, + const struct nfs_pgio_completion_ops *compl_ops, + struct nfs_direct_req *dreq) { struct nfs_pageio_descriptor pgio; + LIST_HEAD(failed); - put_lseg(data->lseg); - data->lseg = NULL; - dprintk("pnfs write error = %d\n", data->pnfs_error); - if (NFS_SERVER(data->inode)->pnfs_curr_ld->flags & - PNFS_LAYOUTRET_ON_ERROR) - pnfs_return_layout(data->inode); - - nfs_pageio_init_read_mds(&pgio, data->inode); - - while (!list_empty(&data->pages)) { - struct nfs_page *req = nfs_list_entry(data->pages.next); + /* Resend all requests through the MDS */ + nfs_pageio_init_read(&pgio, inode, true, compl_ops); + pgio.pg_dreq = dreq; + while (!list_empty(head)) { + struct nfs_page *req = nfs_list_entry(head->next); nfs_list_remove_request(req); - nfs_pageio_add_request(&pgio, req); + if (!nfs_pageio_add_request(&pgio, req)) + nfs_list_add_request(req, &failed); } nfs_pageio_complete(&pgio); + + if (!list_empty(&failed)) { + list_move(&failed, head); + return -EIO; + } + return 0; +} +EXPORT_SYMBOL_GPL(pnfs_read_done_resend_to_mds); + +static void pnfs_ld_handle_read_error(struct nfs_pgio_data *data) +{ + struct nfs_pgio_header *hdr = data->header; + + dprintk("pnfs read error = %d\n", hdr->pnfs_error); + if (NFS_SERVER(hdr->inode)->pnfs_curr_ld->flags & + PNFS_LAYOUTRET_ON_ERROR) { + pnfs_return_layout(hdr->inode); + } + if (!test_and_set_bit(NFS_IOHDR_REDO, &hdr->flags)) + data->task.tk_status = pnfs_read_done_resend_to_mds(hdr->inode, + &hdr->pages, + hdr->completion_ops, + hdr->dreq); } /* * Called by non rpc-based layout drivers */ -void pnfs_ld_read_done(struct nfs_read_data *data) +void pnfs_ld_read_done(struct nfs_pgio_data *data) { - if (likely(!data->pnfs_error)) { + struct nfs_pgio_header *hdr = data->header; + + trace_nfs4_pnfs_read(data, hdr->pnfs_error); + if (likely(!hdr->pnfs_error)) { __nfs4_read_done_cb(data); - data->mds_ops->rpc_call_done(&data->task, data); + hdr->mds_ops->rpc_call_done(&data->task, data); } else pnfs_ld_handle_read_error(data); - data->mds_ops->rpc_release(data); + hdr->mds_ops->rpc_release(data); } EXPORT_SYMBOL_GPL(pnfs_ld_read_done); static void pnfs_read_through_mds(struct nfs_pageio_descriptor *desc, - struct nfs_read_data *data) + struct nfs_pgio_data *data) { - list_splice_tail_init(&data->pages, &desc->pg_list); - if (data->req && list_empty(&data->req->wb_list)) - nfs_list_add_request(data->req, &desc->pg_list); - nfs_pageio_reset_read_mds(desc); - desc->pg_recoalesce = 1; - nfs_readdata_release(data); + struct nfs_pgio_header *hdr = data->header; + + if (!test_and_set_bit(NFS_IOHDR_REDO, &hdr->flags)) { + list_splice_tail_init(&hdr->pages, &desc->pg_list); + nfs_pageio_reset_read_mds(desc); + desc->pg_recoalesce = 1; + } + nfs_pgio_data_release(data); } /* * Call the appropriate parallel I/O subsystem read function. */ static enum pnfs_try_status -pnfs_try_to_read_data(struct nfs_read_data *rdata, +pnfs_try_to_read_data(struct nfs_pgio_data *rdata, const struct rpc_call_ops *call_ops, struct pnfs_layout_segment *lseg) { - struct inode *inode = rdata->inode; + struct nfs_pgio_header *hdr = rdata->header; + struct inode *inode = hdr->inode; struct nfs_server *nfss = NFS_SERVER(inode); enum pnfs_try_status trypnfs; - rdata->mds_ops = call_ops; - rdata->lseg = get_lseg(lseg); + hdr->mds_ops = call_ops; dprintk("%s: Reading ino:%lu %u@%llu\n", __func__, inode->i_ino, rdata->args.count, rdata->args.offset); trypnfs = nfss->pnfs_curr_ld->read_pagelist(rdata); - if (trypnfs == PNFS_NOT_ATTEMPTED) { - put_lseg(rdata->lseg); - rdata->lseg = NULL; - } else { + if (trypnfs != PNFS_NOT_ATTEMPTED) nfs_inc_stats(inode, NFSIOS_PNFS_READ); - } dprintk("%s End (trypnfs:%d)\n", __func__, trypnfs); return trypnfs; } static void -pnfs_do_multiple_reads(struct nfs_pageio_descriptor *desc, struct list_head *head) +pnfs_do_read(struct nfs_pageio_descriptor *desc, struct nfs_pgio_header *hdr) { - struct nfs_read_data *data; + struct nfs_pgio_data *data = hdr->data; const struct rpc_call_ops *call_ops = desc->pg_rpc_callops; struct pnfs_layout_segment *lseg = desc->pg_lseg; + enum pnfs_try_status trypnfs; desc->pg_lseg = NULL; - while (!list_empty(head)) { - enum pnfs_try_status trypnfs; - - data = list_entry(head->next, struct nfs_read_data, list); - list_del_init(&data->list); + trypnfs = pnfs_try_to_read_data(data, call_ops, lseg); + if (trypnfs == PNFS_NOT_ATTEMPTED) + pnfs_read_through_mds(desc, data); + pnfs_put_lseg(lseg); +} - trypnfs = pnfs_try_to_read_data(data, call_ops, lseg); - if (trypnfs == PNFS_NOT_ATTEMPTED) - pnfs_read_through_mds(desc, data); - } - put_lseg(lseg); +static void pnfs_readhdr_free(struct nfs_pgio_header *hdr) +{ + pnfs_put_lseg(hdr->lseg); + nfs_rw_header_free(hdr); } +EXPORT_SYMBOL_GPL(pnfs_readhdr_free); int pnfs_generic_pg_readpages(struct nfs_pageio_descriptor *desc) { - LIST_HEAD(head); + struct nfs_rw_header *rhdr; + struct nfs_pgio_header *hdr; int ret; - ret = nfs_generic_pagein(desc, &head); - if (ret != 0) { - put_lseg(desc->pg_lseg); + rhdr = nfs_rw_header_alloc(desc->pg_rw_ops); + if (!rhdr) { + desc->pg_completion_ops->error_cleanup(&desc->pg_list); + ret = -ENOMEM; + pnfs_put_lseg(desc->pg_lseg); desc->pg_lseg = NULL; return ret; } - pnfs_do_multiple_reads(desc, &head); - return 0; + hdr = &rhdr->header; + nfs_pgheader_init(desc, hdr, pnfs_readhdr_free); + hdr->lseg = pnfs_get_lseg(desc->pg_lseg); + atomic_inc(&hdr->refcnt); + ret = nfs_generic_pgio(desc, hdr); + if (ret != 0) { + pnfs_put_lseg(desc->pg_lseg); + desc->pg_lseg = NULL; + } else + pnfs_do_read(desc, hdr); + if (atomic_dec_and_test(&hdr->refcnt)) + hdr->completion_ops->completion(hdr); + return ret; } EXPORT_SYMBOL_GPL(pnfs_generic_pg_readpages); +static void pnfs_clear_layoutcommitting(struct inode *inode) +{ + unsigned long *bitlock = &NFS_I(inode)->flags; + + clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock); + smp_mb__after_atomic(); + wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING); +} + /* * There can be multiple RW segments. */ @@ -1419,50 +1795,59 @@ static void pnfs_list_write_lseg(struct inode *inode, struct list_head *listp) list_for_each_entry(lseg, &NFS_I(inode)->layout->plh_segs, pls_list) { if (lseg->pls_range.iomode == IOMODE_RW && - test_bit(NFS_LSEG_LAYOUTCOMMIT, &lseg->pls_flags)) + test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT, &lseg->pls_flags)) list_add(&lseg->pls_lc_list, listp); } } -void pnfs_set_lo_fail(struct pnfs_layout_segment *lseg) +static void pnfs_list_write_lseg_done(struct inode *inode, struct list_head *listp) { - if (lseg->pls_range.iomode == IOMODE_RW) { - dprintk("%s Setting layout IOMODE_RW fail bit\n", __func__); - set_bit(lo_fail_bit(IOMODE_RW), &lseg->pls_layout->plh_flags); - } else { - dprintk("%s Setting layout IOMODE_READ fail bit\n", __func__); - set_bit(lo_fail_bit(IOMODE_READ), &lseg->pls_layout->plh_flags); + struct pnfs_layout_segment *lseg, *tmp; + + /* Matched by references in pnfs_set_layoutcommit */ + list_for_each_entry_safe(lseg, tmp, listp, pls_lc_list) { + list_del_init(&lseg->pls_lc_list); + pnfs_put_lseg(lseg); } + + pnfs_clear_layoutcommitting(inode); +} + +void pnfs_set_lo_fail(struct pnfs_layout_segment *lseg) +{ + pnfs_layout_io_set_failed(lseg->pls_layout, lseg->pls_range.iomode); } EXPORT_SYMBOL_GPL(pnfs_set_lo_fail); void -pnfs_set_layoutcommit(struct nfs_write_data *wdata) +pnfs_set_layoutcommit(struct nfs_pgio_data *wdata) { - struct nfs_inode *nfsi = NFS_I(wdata->inode); + struct nfs_pgio_header *hdr = wdata->header; + struct inode *inode = hdr->inode; + struct nfs_inode *nfsi = NFS_I(inode); loff_t end_pos = wdata->mds_offset + wdata->res.count; bool mark_as_dirty = false; - spin_lock(&nfsi->vfs_inode.i_lock); + spin_lock(&inode->i_lock); if (!test_and_set_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)) { mark_as_dirty = true; dprintk("%s: Set layoutcommit for inode %lu ", - __func__, wdata->inode->i_ino); + __func__, inode->i_ino); } - if (!test_and_set_bit(NFS_LSEG_LAYOUTCOMMIT, &wdata->lseg->pls_flags)) { + if (!test_and_set_bit(NFS_LSEG_LAYOUTCOMMIT, &hdr->lseg->pls_flags)) { /* references matched in nfs4_layoutcommit_release */ - get_lseg(wdata->lseg); + pnfs_get_lseg(hdr->lseg); } if (end_pos > nfsi->layout->plh_lwb) nfsi->layout->plh_lwb = end_pos; - spin_unlock(&nfsi->vfs_inode.i_lock); + spin_unlock(&inode->i_lock); dprintk("%s: lseg %p end_pos %llu\n", - __func__, wdata->lseg, nfsi->layout->plh_lwb); + __func__, hdr->lseg, nfsi->layout->plh_lwb); /* if pnfs_layoutcommit_inode() runs between inode locks, the next one * will be a noop because NFS_INO_LAYOUTCOMMIT will not be set */ if (mark_as_dirty) - mark_inode_dirty_sync(wdata->inode); + mark_inode_dirty_sync(inode); } EXPORT_SYMBOL_GPL(pnfs_set_layoutcommit); @@ -1472,6 +1857,7 @@ void pnfs_cleanup_layoutcommit(struct nfs4_layoutcommit_data *data) if (nfss->pnfs_curr_ld->cleanup_layoutcommit) nfss->pnfs_curr_ld->cleanup_layoutcommit(data); + pnfs_list_write_lseg_done(data->args.inode, &data->lseg_list); } /* @@ -1488,50 +1874,43 @@ pnfs_layoutcommit_inode(struct inode *inode, bool sync) struct nfs4_layoutcommit_data *data; struct nfs_inode *nfsi = NFS_I(inode); loff_t end_pos; - int status = 0; - - dprintk("--> %s inode %lu\n", __func__, inode->i_ino); + int status; - if (!test_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)) + if (!pnfs_layoutcommit_outstanding(inode)) return 0; - /* Note kzalloc ensures data->res.seq_res.sr_slot == NULL */ - data = kzalloc(sizeof(*data), GFP_NOFS); - if (!data) { - status = -ENOMEM; - goto out; - } - - if (!test_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)) - goto out_free; + dprintk("--> %s inode %lu\n", __func__, inode->i_ino); + status = -EAGAIN; if (test_and_set_bit(NFS_INO_LAYOUTCOMMITTING, &nfsi->flags)) { - if (!sync) { - status = -EAGAIN; - goto out_free; - } - status = wait_on_bit_lock(&nfsi->flags, NFS_INO_LAYOUTCOMMITTING, - nfs_wait_bit_killable, TASK_KILLABLE); + if (!sync) + goto out; + status = wait_on_bit_lock(&nfsi->flags, + NFS_INO_LAYOUTCOMMITTING, + nfs_wait_bit_killable, + TASK_KILLABLE); if (status) - goto out_free; + goto out; } - INIT_LIST_HEAD(&data->lseg_list); + status = -ENOMEM; + /* Note kzalloc ensures data->res.seq_res.sr_slot == NULL */ + data = kzalloc(sizeof(*data), GFP_NOFS); + if (!data) + goto clear_layoutcommitting; + + status = 0; spin_lock(&inode->i_lock); - if (!test_and_clear_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)) { - clear_bit(NFS_INO_LAYOUTCOMMITTING, &nfsi->flags); - spin_unlock(&inode->i_lock); - wake_up_bit(&nfsi->flags, NFS_INO_LAYOUTCOMMITTING); - goto out_free; - } + if (!test_and_clear_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)) + goto out_unlock; + INIT_LIST_HEAD(&data->lseg_list); pnfs_list_write_lseg(inode, &data->lseg_list); end_pos = nfsi->layout->plh_lwb; nfsi->layout->plh_lwb = 0; - memcpy(&data->args.stateid.data, nfsi->layout->plh_stateid.data, - sizeof(nfsi->layout->plh_stateid.data)); + nfs4_stateid_copy(&data->args.stateid, &nfsi->layout->plh_stateid); spin_unlock(&inode->i_lock); data->args.inode = inode; @@ -1548,7 +1927,22 @@ out: mark_inode_dirty_sync(inode); dprintk("<-- %s status %d\n", __func__, status); return status; -out_free: +out_unlock: + spin_unlock(&inode->i_lock); kfree(data); +clear_layoutcommitting: + pnfs_clear_layoutcommitting(inode); goto out; } + +struct nfs4_threshold *pnfs_mdsthreshold_alloc(void) +{ + struct nfs4_threshold *thp; + + thp = kzalloc(sizeof(*thp), GFP_NOFS); + if (!thp) { + dprintk("%s mdsthreshold allocation failed\n", __func__); + return NULL; + } + return thp; +} |