From: Dave Chinner Subject: Re: hunt for 2.6.37 dm-crypt+ext4 corruption? (was: Re: dm-crypt barrier support is effective) Date: Thu, 16 Dec 2010 14:37:18 +1100 Message-ID: <20101216033718.GM9925@dastard> References: <1291747731-sup-3099@think> <1291751698-sup-9297@think> <1291754340-sup-1631@think> <1291755258-sup-8760@think> <1291810586-sup-8211@think> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: Jon Nelson , Mike Snitzer , Matt , Milan Broz , Andi Kleen , linux-btrfs , dm-devel , Linux Kernel , htd , htejun , linux-ext4 To: Chris Mason Return-path: Received: from bld-mail17.adl2.internode.on.net ([150.101.137.102]:43847 "EHLO mail.internode.on.net" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1751401Ab0LPDhg (ORCPT ); Wed, 15 Dec 2010 22:37:36 -0500 Content-Disposition: inline In-Reply-To: <1291810586-sup-8211@think> Sender: linux-ext4-owner@vger.kernel.org List-ID: On Wed, Dec 08, 2010 at 07:20:24AM -0500, Chris Mason wrote: > Excerpts from Jon Nelson's message of 2010-12-07 22:29:26 -0500: > > On Tue, Dec 7, 2010 at 3:02 PM, Chris Mason wrote: > > > Excerpts from Jon Nelson's message of 2010-12-07 15:48:58 -0500: > > >> On Tue, Dec 7, 2010 at 2:41 PM, Chris Mason wrote: > > >> > Excerpts from Jon Nelson's message of 2010-12-07 15:25:47 -050= 0: > > >> >> On Tue, Dec 7, 2010 at 2:02 PM, Chris Mason wrote: > > >> >> > Excerpts from Jon Nelson's message of 2010-12-07 14:34:40 -= 0500: > > >> >> >> On Tue, Dec 7, 2010 at 12:52 PM, Chris Mason wrote: > > >> >> >> >> postgresql errors. Typically, header corruption but fro= m the limited > > >> >> >> >> visibility I've had into this via strace, what I see is= zeroed pages > > >> >> >> >> where there shouldn't be. > > >> >> >> > > > >> >> >> > This sounds a lot like a bug higher up than dm-crypt. =A0= Zeros tend to > > >> >> >> > come from some piece of code explicitly filling a page w= ith zeros, and > > >> >> >> > that often happens in the corner cases for O_DIRECT and = a few other > > >> >> >> > places in the filesystem. > > >> >> >> > > > >> >> >> > Have you tried triggering this with a regular block devi= ce? > > >> >> >> > > >> >> >> I just tried the whole set of tests, but with /dev/sdb dir= ectly (as > > >> >> >> ext4) without any crypt-y bits. > > >> >> >> It takes more iterations but out of 6 tests I had one fail= ure: same > > >> >> >> type of thing, 'invalid page header in block ....'. > > >> >> >> > > >> >> >> I can't guarantee that it is a full-page of zeroes, just w= hat I saw > > >> >> >> from the (limited) stracing I did. > > >> >> > > > >> >> > Fantastic. Now for our usual suspects: > >=20 > > Maybe not so fantastic. I kept testing and had no more failures. At > > all. After 40+ iterations I gave up. > > I went back to trying ext4 on a LUKS volume. The 'hit' ratio went t= o > > something like 1 in 3, or better. > >=20 > > I will continue to do testing with and without LUKS. I did /not/ > > reboot between tests, but I do start with a fresh postgres database= =2E > >=20 >=20 > Once we trigger once without dm-crypt, dm-crypt is off the hook. Jus= t > to verify, when you say without luks, you mean without any crypto bit= s > in use at all on the filesystems postgres uses? >=20 > Usually the trick to reproducing filesystem corruptions is adding mem= ory > pressure. The corruption is probably a bad interaction between reads > and writes, and we need to make sure the reads actually happen. >=20 > http://oss.oracle.com/~mason/pin_ram.c >=20 > gcc -Wall -o pin_ram pin_ram.c >=20 > pin_ram -m 80%-of-your-ram-in-mb Implemented in xfstests about 10 years ago: http://git.kernel.org/?p=3Dfs/xfs/xfstests-dev.git;a=3Dblob;f=3Dsrc/use= mem.c;h=3Db8794a6b209cebf8dbf312a8ef131e2e54b18d29;hb=3DHEAD :P Cheers, Dave. --=20 Dave Chinner david@fromorbit.com -- To unsubscribe from this list: send the line "unsubscribe linux-ext4" i= n the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html