ext4: fix COLLAPSE_RANGE test failure in data journalling mode
authorNamjae Jeon <namjae.jeon@samsung.com>
Fri, 11 Apr 2014 02:58:20 +0000 (22:58 -0400)
committerTheodore Ts'o <tytso@mit.edu>
Fri, 11 Apr 2014 02:58:20 +0000 (22:58 -0400)
When mounting ext4 with data=journal option, xfstest shared/002 and
shared/004 are currently failing as checksum computed for testfile
does not match with the checksum computed in other journal modes.
In case of data=journal mode, a call to filemap_write_and_wait_range
will not flush anything to disk as buffers are not marked dirty in
write_end. In collapse range this call is followed by a call to
truncate_pagecache_range. Due to this, when checksum is computed,
a portion of file is re-read from disk which replace valid data with
NULL bytes and hence the reason for the difference in checksum.

Calling ext4_force_commit before filemap_write_and_wait_range solves
the issue as it will mark the buffers dirty during commit transaction
which can be later synced by a call to filemap_write_and_wait_range.

Signed-off-by: Namjae Jeon <namjae.jeon@samsung.com>
Signed-off-by: Ashish Sangwan <a.sangwan@samsung.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
fs/ext4/extents.c

index 82df3ce9874ab7f3a65abc10e2bd2238b1ae2af3..be1e56cbbf323d7c10c8ca7979c9dd014a4a3123 100644 (file)
@@ -5383,6 +5383,13 @@ int ext4_collapse_range(struct inode *inode, loff_t offset, loff_t len)
        punch_start = offset >> EXT4_BLOCK_SIZE_BITS(sb);
        punch_stop = (offset + len) >> EXT4_BLOCK_SIZE_BITS(sb);
 
+       /* Call ext4_force_commit to flush all data in case of data=journal. */
+       if (ext4_should_journal_data(inode)) {
+               ret = ext4_force_commit(inode->i_sb);
+               if (ret)
+                       return ret;
+       }
+
        /* Write out all dirty pages */
        ret = filemap_write_and_wait_range(inode->i_mapping, offset, -1);
        if (ret)