-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 2/23/12 7:23 AM, Philipp Hahn wrote:

Hello Ted, hello Eric,

On Monday February 7th 2011 16:59:36 Ted Ts'o wrote:

quoted

commit 7520bb0f2980ef79d17dcbec2783760b37490ffc

upstream is actually e9e3bcecf44c04b9e6b505fd8e2eb9cea58fb94d but you probably knew that.

quoted

Author: Eric Sandeen [off-list ref]
Date:   Mon Feb 7 10:57:28 2011 -0500

    ext4: serialize unaligned asynchronous DIO

    ext4 has a data corruption case when doing non-block-aligned
    asynchronous direct IO into a sparse file, as demonstrated
    by xfstest 240.

I hope you remember that bug, because I encountered this data corruption bug 
on Debians 2.6.32(.51) kernel as well.

I remember it well ;)

I also backported it to RHEL6, but that "2.6.32" kernel also had a few of ext4 updates.

Still, grabbing a centos6 kernel src.rpm and looking might help you.

FWIW I also backported f46c483357c2d87606bbefb511321e3efd4baae0 and
f2d28a2ebcb525a6ec7e2152106ddb385ef52b73 as helpers.

On the other hand RedHat seems to have back-ported that fix to RHEL5 (2.6.18)  
and probably RHEL6 (2.6.32) as well, but I don't have a subscription, so I 
can't verify that:
<http://rpmfind.net/linux/RPM/centos/updates/5.7/x86_64/RPMS/kernel-devel-2.6.18-274.12.1.el5.x86_64.html>
<https://bugzilla.redhat.com/show_bug.cgi?id=689830>

615309 is the RHEL6 bug.  Sadly it's marked private.

The Xen-people also encountered it and asked for someone to backport it:
<http://osdir.com/ml/xen-development/2011-07/msg00474.html>

I tried to backport it from 2.6.38~rc5 to 2.6.32.51 and thus far it seems to 
fix the bug. But several other things were re-named and re-organized between 
those versions, so it was not slreight forward.

Since I'm no ext4 expert, I'd like to ask you to have a look at this backport. 
Is it sound or are there some tests I can throw at it to get it tested more 
thoroughly?

xfstests test #240 tests it specifically:

# FS QA Test No. 240
#
# Test that non-block-aligned aio+dio into holes does not leave
# zero'd out portions of the file
#
# QEMU IO to a file-backed device with misaligned partitions
# can send this sort of IO
#
# This test need only be run in the case where the logical block size
# of the device can be smaller than the file system block size.

and running all the xfstests over your result would probably be good.

FWIW, there was a related xfs fix as well:

https://bugzilla.redhat.com/show_bug.cgi?id=669272

that one's not marked private <eyeroll>

Does is classify for <mailto:stable@vger.kernel.org>?

probably so.

To be honest I am really pressed for time right now, and this was somewhat tricky code.  I won't be able to review it right now, anyway, but can try to get to it at some point if my schedule lightens up... :(

- -Eric

Thanks in advance
Philipp Hahn

-----BEGIN PGP SIGNATURE-----
Version: GnuPG/MacGPG2 v2.0.17 (Darwin)
Comment: GPGTools - http://gpgtools.org
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iQIcBAEBAgAGBQJPRlgFAAoJECCuFpLhPd7gU58QAJwAzv2e9WVT2zs741Jhv0Fl
NCAJ7t7+EYECA5Qtbzm/LCojmvf2mYKVA2MmHiLIG0jxSAX6DQ+bHsjx0N3DbFCA
nabZFBwjiNfbII3ut4lHTWXfi6hZ0yqs6/qZTCnm4janwQN9ffR7+kuwTfuJGTaI
a0pelgiXQTIVQcx/togd9qezrEwVGd/7Z/sw67o1/hpc76fsELXYnZVkQ4jzXKwc
gvAgjFKSdkY0sMCq/owwiA6lgZydMeGzkXYbDlvYx7lfPp6n8ZPpupa2UKAeSF/P
4T8cweTK/XNvlr7KcXx9zHoD3ZLRTYaVvvIlaOa5n0S+v71wGV/AOV2wLNeK3s2x
jn91Zsf1sKTpvUQsh5P1UZKgOEXVgQ+gu4+15Ggk5LPOrSd8j4wjrGbrHHu4MfRv
udLZP1lE+RINflzdkL6nx6UeeI+X4LOU5McjSgs4gaKInTK9U090vpsCDijEK+mJ
ku9nfX+pBiXQgcHFA8fTe0KwnBxvA+AuY0n4w4zzj96bh9tqlYK2MH3qfVJTrMrs
pl6gZQTUbt8/UwaP8ooa0gBBG9aGVaBsRtBng9hqIb5vU+TImR8PBlL3QQZPvcZa
g+wYq3UmIl5z+A17Ify37FgbU6lsuFmleU0fRglV/ofAywMGatKNUJZn4yz+7QTv
ADMfCQ8Gnpy8Ue+LxQkO
=cYPJ
-----END PGP SIGNATURE-----