From: Theodore Tso Subject: Re: data corruption with ext4 (from 2.6.27.4) exposed by rtorrent Date: Wed, 5 Nov 2008 12:25:53 -0500 Message-ID: <20081105172553.GA29880@mit.edu> References: <3d3ce57e0811030442o377cf2bet212eefba79d714bb@mail.gmail.com> <20081103134008.GE29102@mit.edu> <319012f0811050534i27b7dce1t8120b3a343844adc@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Roc Valles , linux-ext4@vger.kernel.org To: Cryptooctoploid Return-path: Received: from www.church-of-our-saviour.org ([69.25.196.31]:44091 "EHLO thunker.thunk.org" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1752449AbYKERZ4 (ORCPT ); Wed, 5 Nov 2008 12:25:56 -0500 Content-Disposition: inline In-Reply-To: <319012f0811050534i27b7dce1t8120b3a343844adc@mail.gmail.com> Sender: linux-ext4-owner@vger.kernel.org List-ID: Cryptooctoploid, Based on some reflection and further discussion on #ext4, I have a theory which we need to try to prove/disprove that it may be related to memory pressure combined with the delayed allocation code. If you've already dismantled your ext4 setup, maybe you can't test this, but it would be interesting to know mounting with the -o nodelalloc makes the problem go away. Also, how much memory do you have in your system and is your system often under memory pressure? (i.e., your system often uses swap, and where the amount of free memory is often near the low water mark where it is trying to evict memory either by dropping clean pages or writing dirty pages back to the filesystem and/or swap). - Ted