Ocfs2: Fix a regression bug from mainline commit(6b933c8e6f1a2f3118082c455eef25f9b1ac...
authorTristan Ye <tristan.ye@oracle.com>
Thu, 19 Aug 2010 07:15:00 +0000 (15:15 +0800)
committerTao Ma <tao.ma@oracle.com>
Wed, 8 Sep 2010 06:25:57 +0000 (14:25 +0800)
The patch is to fix the regression bug brought from commit 6b933c8...( 'ocfs2:
Avoid direct write if we fall back to buffered I/O'):

http://oss.oracle.com/bugzilla/show_bug.cgi?id=1285

The commit 6b933c8e6f1a2f3118082c455eef25f9b1ac7b45 changed __generic_file_aio_write
to generic_file_buffered_write, which didn't call filemap_{write,wait}_range to  flush
the pagecaches when we were falling O_DIRECT writes back to buffered ones. it did hurt
the O_DIRECT semantics somehow in extented odirect writes.

This patch tries to guarantee O_DIRECT writes of 'fall back to buffered' to be correctly
flushed.

Signed-off-by: Tristan Ye <tristan.ye@oracle.com>
Signed-off-by: Tao Ma <tao.ma@oracle.com>
fs/ocfs2/file.c

index 2caa3a7a1a39291bddd6208123858ea67ee9c43b..9a03c151b5ceabc169215d99777abb92e2d2ad36 100644 (file)
@@ -2338,7 +2338,7 @@ out_dio:
        BUG_ON(ret == -EIOCBQUEUED && !(file->f_flags & O_DIRECT));
 
        if (((file->f_flags & O_DSYNC) && !direct_io) || IS_SYNC(inode) ||
-           ((file->f_flags & O_DIRECT) && has_refcount)) {
+           ((file->f_flags & O_DIRECT) && !direct_io)) {
                ret = filemap_fdatawrite_range(file->f_mapping, pos,
                                               pos + count - 1);
                if (ret < 0)