xfs: make iomap_begin functions trim iomaps consistently
authorDarrick J. Wong <darrick.wong@oracle.com>
Thu, 7 Dec 2017 00:13:35 +0000 (16:13 -0800)
committerDarrick J. Wong <darrick.wong@oracle.com>
Sat, 9 Dec 2017 01:51:05 +0000 (17:51 -0800)
Historically, the XFS iomap_begin function only returned mappings for
exactly the range queried, i.e. it doesn't do XFS_BMAPI_ENTIRE lookups.
The current vfs iomap consumers are only set up to deal with trimmed
mappings.  xfs_xattr_iomap_begin does BMAPI_ENTIRE lookups, which is
inconsistent with the current iomap usage.  Remove the flag so that both
iomap_begin functions behave the same way.

FWIW this also fixes a behavioral regression in xattr FIEMAP that was
introduced in 4.8 wherein attr fork extents are no longer trimmed like
they used to be.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
fs/xfs/xfs_iomap.c

index 33eb4fb2e3fd87b0848ebe2599b0a71b05f0d082..7ab52a8bc0a9e6dff904fe10b097eb3a478db9b9 100644 (file)
@@ -1213,7 +1213,7 @@ xfs_xattr_iomap_begin(
 
        ASSERT(ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL);
        error = xfs_bmapi_read(ip, offset_fsb, end_fsb - offset_fsb, &imap,
-                              &nimaps, XFS_BMAPI_ENTIRE | XFS_BMAPI_ATTRFORK);
+                              &nimaps, XFS_BMAPI_ATTRFORK);
 out_unlock:
        xfs_iunlock(ip, lockmode);