xfs: skip bmbt block ino validation during owner change
authorBrian Foster <bfoster@redhat.com>
Tue, 29 Aug 2017 17:08:39 +0000 (10:08 -0700)
committerDarrick J. Wong <darrick.wong@oracle.com>
Fri, 1 Sep 2017 17:55:30 +0000 (10:55 -0700)
Extent swap uses xfs_btree_visit_blocks() to fix up bmbt block
owners on v5 (!rmapbt) filesystems. The bmbt scan uses
xfs_btree_lookup_get_block() to read bmbt blocks which verifies the
current owner of the block against the parent inode of the bmbt.
This works during extent swap because the bmbt owners are updated to
the opposite inode number before the inode extent forks are swapped.

The modified bmbt blocks are marked as ordered buffers which allows
everything to commit in a single transaction. If the transaction
commits to the log and the system crashes such that recovery of the
extent swap is required, log recovery restarts the bmbt scan to fix
up any bmbt blocks that may have not been written back before the
crash. The log recovery bmbt scan occurs after the inode forks have
been swapped, however. This causes the bmbt block owner verification
to fail, leads to log recovery failure and requires xfs_repair to
zap the log to recover.

Define a new invalid inode owner flag to inform the btree block
lookup mechanism that the current inode may be invalid with respect
to the current owner of the bmbt block. Set this flag on the cursor
used for change owner scans to allow this operation to work at
runtime and during log recovery.

Signed-off-by: Brian Foster <bfoster@redhat.com>
Fixes: bb3be7e7c ("xfs: check for bogus values in btree block headers")
Cc: stable@vger.kernel.org
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
fs/xfs/libxfs/xfs_bmap_btree.c
fs/xfs/libxfs/xfs_btree.c
fs/xfs/libxfs/xfs_btree.h

index 85de2251301457244181bec3a6f2993010a6dccd..a6331ffa51e3e69f04a789de7f335da99ab856dd 100644 (file)
@@ -858,6 +858,7 @@ xfs_bmbt_change_owner(
        cur = xfs_bmbt_init_cursor(ip->i_mount, tp, ip, whichfork);
        if (!cur)
                return -ENOMEM;
+       cur->bc_private.b.flags |= XFS_BTCUR_BPRV_INVALID_OWNER;
 
        error = xfs_btree_change_owner(cur, new_owner, buffer_list);
        xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
index 0b7905a4e27a86593735b869f957f8bb47e68e42..1d15d04e29a5568d051e41896858c5fb17a40a1e 100644 (file)
@@ -1791,6 +1791,7 @@ xfs_btree_lookup_get_block(
 
        /* Check the inode owner since the verifiers don't. */
        if (xfs_sb_version_hascrc(&cur->bc_mp->m_sb) &&
+           !(cur->bc_private.b.flags & XFS_BTCUR_BPRV_INVALID_OWNER) &&
            (cur->bc_flags & XFS_BTREE_LONG_PTRS) &&
            be64_to_cpu((*blkp)->bb_u.l.bb_owner) !=
                        cur->bc_private.b.ip->i_ino)
index 9c95e965cfe53e3c61edbab6a0e6f0a076f82a4f..f2a88c3b11595222d5290781ce91005cd150cdd8 100644 (file)
@@ -233,7 +233,8 @@ typedef struct xfs_btree_cur
                        short           forksize;       /* fork's inode space */
                        char            whichfork;      /* data or attr fork */
                        char            flags;          /* flags */
-#define        XFS_BTCUR_BPRV_WASDEL   1                       /* was delayed */
+#define        XFS_BTCUR_BPRV_WASDEL           (1<<0)          /* was delayed */
+#define        XFS_BTCUR_BPRV_INVALID_OWNER    (1<<1)          /* for ext swap */
                } b;
        }               bc_private;     /* per-btree type data */
 } xfs_btree_cur_t;