From: Theodore Ts'o Subject: Re: [PATCH v4 2/3] ext4: fix races between changing inode journal mode and ext4_writepages Date: Wed, 27 Apr 2016 10:30:32 -0400 Message-ID: <20160427143032.GG30021@thunk.org> References: <1319414639.152111461744250568.JavaMail.weblogic@ep2mlwas01c> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: "jack@suse.cz" , "linux-ext4@vger.kernel.org" To: Daeho Jeong Return-path: Received: from imap.thunk.org ([74.207.234.97]:48118 "EHLO imap.thunk.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751550AbcD0Oai (ORCPT ); Wed, 27 Apr 2016 10:30:38 -0400 Content-Disposition: inline In-Reply-To: <1319414639.152111461744250568.JavaMail.weblogic@ep2mlwas01c> Sender: linux-ext4-owner@vger.kernel.org List-ID: On Wed, Apr 27, 2016 at 08:04:10AM +0000, Daeho Jeong wrote: > Oops, sorry about causing the regression. > > But in the version I worked on, we passed the test case and > we still pass that. I executed the test case as belows. > > kvm-xfstests -c 1k -C 10 generic/231 > > I am going to retry the test using GCE with the latest kernel version. Hi, sorry, it looks like I jumped to conclusions too hastily. I let the tests finish last night and generic/231 is failing around 10-20% of the time prior to your patch series. I just didn't notice it at first because the failure happened at the very end: xfstests results 201604270033 - 4.6.0-rc4-ext4 Passed all 1 tests Passed all 1 tests Passed all 1 tests Passed all 1 tests Passed all 1 tests Passed all 1 tests Passed all 1 tests Passed all 1 tests Failures: generic/231 Failures: generic/231 With your patches, it remains at about the same level, but when the tests failed caused me to assume they were the caused by your patches. xfstests results 201604270044 - 4.6.0-rc4-ext4-00007-g76bee81 (Just before your patches) Passed all 1 tests Passed all 1 tests Passed all 1 tests Passed all 1 tests Failures: generic/231 Passed all 1 tests Passed all 1 tests Passed all 1 tests Failures: generic/231 Passed all 1 tests xfstests results 201604270054 - 4.6.0-rc4-ext4-00008-g31cedaa (Your 1/3 patch) Passed all 1 tests Passed all 1 tests Passed all 1 tests Failures: generic/231 Passed all 1 tests Passed all 1 tests Passed all 1 tests Passed all 1 tests Passed all 1 tests Passed all 1 tests xfstests results 201604270052 - 4.6.0-rc4-ext4-00009-g46ceec9 (Your 2/3 patch) Failures: generic/231 <=== this is why I thought this commit was the regression Passed all 1 tests Passed all 1 tests Passed all 1 tests Passed all 1 tests Passed all 1 tests Passed all 1 tests Passed all 1 tests Passed all 1 tests Failures: generic/231 xfstests results 201604270046 - 4.6.0-rc4-ext4-00010-geed525e (Your 3/3 patch) Failures: generic/231 Failures: generic/231 Failures: generic/231 Failures: generic/231 Failures: generic/231 Passed all 1 tests Passed all 1 tests Passed all 1 tests Passed all 1 tests Failures: generic/231 So enabling per-file data journalling and delalloc is definitely makes the problem much more visible, but it looks like it was always there. Apologies for the initial misdiagnosis.... - Ted