GFS2: Fix AIL flush issue during fsync
authorSteven Whitehouse <swhiteho@redhat.com>
Wed, 7 Sep 2011 09:33:25 +0000 (10:33 +0100)
committerSteven Whitehouse <swhiteho@redhat.com>
Fri, 21 Oct 2011 11:39:41 +0000 (12:39 +0100)
commitb5b24d7aeb9608935786369ac2d3e9f362877d55
tree2043123df7a6e6ca367b24cb3442ee50314cdee0
parent70b0c3656f12964a6dac104214c904c66e626058
GFS2: Fix AIL flush issue during fsync

Unfortunately, it is not enough to just ignore locked buffers during
the AIL flush from fsync. We need to be able to ignore all buffers
which are locked, dirty or pinned at this stage as they might have
been added subsequent to the log flush earlier in the fsync function.

In addition, this means that we no longer need to rely on i_mutex to
keep out writes during fsync, so we can, as a side-effect, remove
that protection too.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Tested-By: Abhijith Das <adas@redhat.com>
fs/gfs2/file.c
fs/gfs2/glops.c
fs/gfs2/glops.h
fs/gfs2/super.c