From mboxrd@z Thu Jan 1 00:00:00 1970 From: Dave Chinner Subject: Re: [PATCH] xfstests 240: test non-aligned AIO hole-filling Date: Fri, 23 Jul 2010 20:12:22 +1000 Message-ID: <20100723101222.GH32635@dastard> References: <4C489676.2000605@redhat.com> <20100723024938.GD32635@dastard> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: ext4 development , xfs-oss To: Eric Sandeen Return-path: Received: from bld-mail19.adl2.internode.on.net ([150.101.137.104]:35601 "EHLO mail.internode.on.net" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1753718Ab0GWKMa (ORCPT ); Fri, 23 Jul 2010 06:12:30 -0400 Content-Disposition: inline In-Reply-To: <20100723024938.GD32635@dastard> Sender: linux-ext4-owner@vger.kernel.org List-ID: On Fri, Jul 23, 2010 at 12:49:38PM +1000, Dave Chinner wrote: > On Thu, Jul 22, 2010 at 02:05:26PM -0500, Eric Sandeen wrote: > > This replicates file corruption we've seen with qemu-kvm when > > we use if=virtio,cache=none,aio=native for IO to a sparse > > ext4- or xfs-hosted file, and the partitions/filesystems > > within that file image are not block-aligned. (think sector > > 63 here...) This results in AIO IOs not aligned to the > > filesystem blocks. > > > > This test modifies aiodio_sparse2.c to add an option to start > > the file IO at an offset. > > > > When we do 4k writes to a 16k file in 2 threads, starting > > at offset 512, we get 0s interspersed in the file where they > > should not be: > > > > 00000000 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................| > > * > > 00000200 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01 |................| > > * > > 00001000 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................| > > * > > 00001200 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01 |................| > > .... > > > > Signed-off-by: Eric Sandeen > > looks good - it needs to remove the $TEST_DIR/aiodio_sparse file > before the test starts, though, so that repeated tests fail > correctly. Otherwise: > > Reviewed-by: Dave Chinner > > Also, Eric, can you run this proof-of-concept test branch (on top of > the current XFS for-2.6.36 branch) and see if QEMU is happy with it? > It passes this test on my machine.... > > The following changes since commit d2ec2c790b3994bf0c4381cacc489b514705887c: > > xfs: clean up xfs_bmap_get_bp (2010-07-22 12:52:08 +1000) > > are available in the git repository at: > git://git.kernel.org/pub/scm/linux/kernel/git/dgc/xfsdev aio-unaligned I just pushed another patch onto this series that will makes test 240 pass reliably here rather than hanging. It's all a bit messy, but it does serve to illustrate the problem and a potential avenue to prevent it.... Dave Chinner (5): Merge branch 'master' into working Merge branch 'for-2.6.36' into working fs: get_blocks needs an unaligned mapping flag xfs: serialise unaligned direct IO into unwritten extents xfs: wait on IO completion inside an IO context fs/buffer.c | 17 ++++---- fs/direct-io.c | 25 ++++++++++- fs/ioctl.c | 2 +- fs/mpage.c | 6 ++- fs/xfs/linux-2.6/xfs_aops.c | 95 ++++++++++++++++++++++++++++++++++++++----- fs/xfs/xfs_inode.h | 13 +++--- include/linux/fs.h | 10 ++++- 7 files changed, 136 insertions(+), 32 deletions(-) -- Dave Chinner david@fromorbit.com