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>

diff --git a/fs/nfs/blocklayout/blocklayout.c b/fs/nfs/blocklayout/blocklayout.c
index f55a4e7560470d7a1ef8ab790beb6db59232b35f..217847679f0eac675492ac049cbc0a42a6f75066 100644 (file)
--- a/fs/nfs/blocklayout/blocklayout.c
+++ b/fs/nfs/blocklayout/blocklayout.c
@@ -346,7 +346,7 @@ static void bl_write_cleanup(struct work_struct *work)
                        PAGE_SIZE - 1) & (loff_t)PAGE_MASK;
 
                ext_tree_mark_written(bl, start >> SECTOR_SHIFT,
-                                       (end - start) >> SECTOR_SHIFT);
+                                       (end - start) >> SECTOR_SHIFT, end);
        }
 
        pnfs_ld_write_done(hdr);

diff --git a/fs/nfs/blocklayout/blocklayout.h b/fs/nfs/blocklayout/blocklayout.h
index 18e6fd0b9506e931a62acd3f91303574433a1113..efc007f007427c84bf0e56961b51f07073b78865 100644 (file)
--- a/fs/nfs/blocklayout/blocklayout.h
+++ b/fs/nfs/blocklayout/blocklayout.h
@@ -141,6 +141,7 @@ struct pnfs_block_layout {
        struct rb_root          bl_ext_ro;
        spinlock_t              bl_ext_lock;   /* Protects list manipulation */
        bool                    bl_scsi_layout;
+       u64                     bl_lwb;
 };
 
 static inline struct pnfs_block_layout *
@@ -182,7 +183,7 @@ int ext_tree_insert(struct pnfs_block_layout *bl,
 int ext_tree_remove(struct pnfs_block_layout *bl, bool rw, sector_t start,
                sector_t end);
 int ext_tree_mark_written(struct pnfs_block_layout *bl, sector_t start,
-               sector_t len);
+               sector_t len, u64 lwb);
 bool ext_tree_lookup(struct pnfs_block_layout *bl, sector_t isect,
                struct pnfs_block_extent *ret, bool rw);
 int ext_tree_prepare_commit(struct nfs4_layoutcommit_args *arg);

diff --git a/fs/nfs/blocklayout/extent_tree.c b/fs/nfs/blocklayout/extent_tree.c
index 992bcb19c11e744bed4c390e133ea3ebe7cf0916..c85fbfd2d0d99e7cd5e3cc902ebd4b75b8e1b02c 100644 (file)
--- 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);