diff options
author | Benjamin Coddington <bcodding@redhat.com> | 2016-08-22 18:11:16 (GMT) |
---|---|---|
committer | Trond Myklebust <trond.myklebust@primarydata.com> | 2016-08-23 15:41:38 (GMT) |
commit | 41963c10c47a35185e68cb9049f7a3493c94d2d7 (patch) | |
tree | a561ace1f94987e1a66cf15d2f218b386bdd8487 /fs/nfs/blocklayout/extent_tree.c | |
parent | b88fa69eaa8649f11828158c7b65c4bcd886ebd5 (diff) | |
download | linux-41963c10c47a35185e68cb9049f7a3493c94d2d7.tar.xz |
pnfs/blocklayout: update last_write_offset atomically with extents
Block/SCSI layout write completion may add committable extents to the
extent tree before updating the layout's last-written byte under the inode
lock. If a sync happens before this value is updated, then
prepare_layoutcommit may find and encode these extents which would produce
a LAYOUTCOMMIT request whose encoded extents are larger than the request's
loca_length.
Fix this by using a last-written byte value that is updated atomically with
the extent tree so that commitable extents always match.
Signed-off-by: Benjamin Coddington <bcodding@redhat.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Diffstat (limited to 'fs/nfs/blocklayout/extent_tree.c')
-rw-r--r-- | fs/nfs/blocklayout/extent_tree.c | 10 |
1 files changed, 7 insertions, 3 deletions
diff --git a/fs/nfs/blocklayout/extent_tree.c b/fs/nfs/blocklayout/extent_tree.c index 992bcb1..c85fbfd 100644 --- a/fs/nfs/blocklayout/extent_tree.c +++ b/fs/nfs/blocklayout/extent_tree.c @@ -402,7 +402,7 @@ ext_tree_split(struct rb_root *root, struct pnfs_block_extent *be, int ext_tree_mark_written(struct pnfs_block_layout *bl, sector_t start, - sector_t len) + sector_t len, u64 lwb) { struct rb_root *root = &bl->bl_ext_rw; sector_t end = start + len; @@ -471,6 +471,8 @@ ext_tree_mark_written(struct pnfs_block_layout *bl, sector_t start, } } out: + if (bl->bl_lwb < lwb) + bl->bl_lwb = lwb; spin_unlock(&bl->bl_ext_lock); __ext_put_deviceids(&tmp); @@ -518,7 +520,7 @@ static __be32 *encode_scsi_range(struct pnfs_block_extent *be, __be32 *p) } static int ext_tree_encode_commit(struct pnfs_block_layout *bl, __be32 *p, - size_t buffer_size, size_t *count) + size_t buffer_size, size_t *count, __u64 *lastbyte) { struct pnfs_block_extent *be; int ret = 0; @@ -542,6 +544,8 @@ static int ext_tree_encode_commit(struct pnfs_block_layout *bl, __be32 *p, p = encode_block_extent(be, p); be->be_tag = EXTENT_COMMITTING; } + *lastbyte = bl->bl_lwb - 1; + bl->bl_lwb = 0; spin_unlock(&bl->bl_ext_lock); return ret; @@ -564,7 +568,7 @@ ext_tree_prepare_commit(struct nfs4_layoutcommit_args *arg) arg->layoutupdate_pages = &arg->layoutupdate_page; retry: - ret = ext_tree_encode_commit(bl, start_p + 1, buffer_size, &count); + ret = ext_tree_encode_commit(bl, start_p + 1, buffer_size, &count, &arg->lastbytewritten); if (unlikely(ret)) { ext_tree_free_commitdata(arg, buffer_size); |