xfs: trim eofblocks before collapse range
authorBrian Foster <bfoster@redhat.com>
Tue, 2 Sep 2014 02:12:53 +0000 (12:12 +1000)
committerDave Chinner <david@fromorbit.com>
Tue, 2 Sep 2014 02:12:53 +0000 (12:12 +1000)
xfs_collapse_file_space() currently writes back the entire file
undergoing collapse range to settle things down for the extent shift
algorithm. While this prevents changes to the extent list during the
collapse operation, the writeback itself is not enough to prevent
unnecessary collapse failures.

The current shift algorithm uses the extent index to iterate the in-core
extent list. If a post-eof delalloc extent persists after the writeback
(e.g., a prior zero range op where the end of the range aligns with eof
can separate the post-eof blocks such that they are not written back and
converted), xfs_bmap_shift_extents() becomes confused over the encoded
br_startblock value and fails the collapse.

As with the full writeback, this is a temporary fix until the algorithm
is improved to cope with a volatile extent list and avoid attempts to
shift post-eof extents.

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>
fs/xfs/xfs_bmap_util.c

index 76b6a2910e40e365dbf3cfe37d602356a12c06d6..1707980f9a4b564842cd435d9efc70571a5a9b18 100644 (file)
@@ -1471,8 +1471,10 @@ xfs_collapse_file_space(
        shift_fsb = XFS_B_TO_FSB(mp, len);
 
        /*
-        * writeback the entire file to prevent concurrent writeback of ranges
-        * outside the collapsing region from changing the extent list.
+        * Writeback the entire file and force remove any post-eof blocks. The
+        * writeback prevents changes to the extent list via concurrent
+        * writeback and the eofblocks trim prevents the extent shift algorithm
+        * from running into a post-eof delalloc extent.
         *
         * XXX: This is a temporary fix until the extent shift loop below is
         * converted to use offsets and lookups within the ILOCK rather than
@@ -1482,6 +1484,11 @@ xfs_collapse_file_space(
        error = filemap_write_and_wait(VFS_I(ip)->i_mapping);
        if (error)
                return error;
+       if (xfs_can_free_eofblocks(ip, true)) {
+               error = xfs_free_eofblocks(mp, ip, false);
+               if (error)
+                       return error;
+       }
 
        error = xfs_free_file_space(ip, offset, len);
        if (error)