ocfs2: remove filesize checks for sync I/O journal commit
authorGoldwyn Rodrigues <rgoldwyn@suse.de>
Wed, 10 Dec 2014 23:41:53 +0000 (15:41 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 11 Dec 2014 01:41:03 +0000 (17:41 -0800)
Filesize is not a good indication that the file needs to be synced.
An example where this breaks is:
 1. Open the file in O_SYNC|O_RDWR
 2. Read a small portion of the file (say 64 bytes)
 3. Lseek to starting of the file
 4. Write 64 bytes

If the node crashes, it is not written out to disk because this was not
committed in the journal and the other node which reads the file after
recovery reads stale data (even if the write on the other node was
successful)

Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.de>
Reviewed-by: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/ocfs2/file.c

index 324dc93ac896c073886846e0d883d1d8f5bd30ca..69fb9f75b0825f14527a8e2ec3f8e99c72f5372d 100644 (file)
@@ -2381,9 +2381,7 @@ out_dio:
                if (ret < 0)
                        written = ret;
 
-               if (!ret && ((old_size != i_size_read(inode)) ||
-                            (old_clusters != OCFS2_I(inode)->ip_clusters) ||
-                            has_refcount)) {
+               if (!ret) {
                        ret = jbd2_journal_force_commit(osb->journal->j_journal);
                        if (ret < 0)
                                written = ret;