[PATCH 2/3] xfs: don't leak linked inodes during log recovery

["Darrick J. Wong" <darrick.wong@oracle.com>]

When we introduced the bmap redo log items, we set MS_ACTIVE on the
mountpoint and XFS_IRECOVERY on the inode to prevent unlinked inodes
from being truncated prematurely during log recovery.  However, we
neglected to drop linked inodes that are recovered, and if we don't use
the inode between recovery and unmount, the inode will never be marked
reclaimable and thus we fail to free it at umount time.  If we're in
log recovery but IRECOVERY is /not/ set, the inode is linked and can be
reclaimed.

Fixes: 17c12bcd30 ("xfs: when replaying bmap operations, don't let unlinked inodes get reaped")
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/xfs_super.c |    7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index 38aaacd..9b06ca2 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -1040,6 +1040,13 @@ xfs_fs_drop_inode(
 	if (ip->i_flags & XFS_IRECOVERY) {
 		ASSERT(ip->i_mount->m_log->l_flags & XLOG_RECOVERY_NEEDED);
 		return 0;
+	} else if (ip->i_mount->m_log->l_flags & XLOG_RECOVERY_NEEDED) {
+		/*
+		 * This inode was loaded during recovery but is not
+		 * being unlinked, so we can free it without fear of
+		 * premature truncation.
+		 */
+		return 1;
 	}
 
 	return generic_drop_inode(inode) || (ip->i_flags & XFS_IDONTCACHE);