From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1750997AbWLRBXM (ORCPT ); Sun, 17 Dec 2006 20:23:12 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1751012AbWLRBXM (ORCPT ); Sun, 17 Dec 2006 20:23:12 -0500 Received: from smtp.osdl.org ([65.172.181.25]:46634 "EHLO smtp.osdl.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750994AbWLRBXL (ORCPT ); Sun, 17 Dec 2006 20:23:11 -0500 Date: Sun, 17 Dec 2006 17:22:57 -0800 (PST) From: Linus Torvalds To: Andrew Morton cc: andrei.popa@i-neo.ro, Linux Kernel Mailing List , Peter Zijlstra , Hugh Dickins , Florian Weimer , Marc Haber , Martin Michlmayr Subject: Re: 2.6.19 file content corruption on ext3 In-Reply-To: <20061217154026.219b294f.akpm@osdl.org> Message-ID: References: <1166314399.7018.6.camel@localhost> <20061217040620.91dac272.akpm@osdl.org> <1166362772.8593.2.camel@localhost> <20061217154026.219b294f.akpm@osdl.org> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Sun, 17 Dec 2006, Andrew Morton wrote: > > From my quick reading, all callers of try_to_free_buffers() have already > unmapped the page from pagetables, and given that the reported ext3 corruption > happens on uniprocessor, non-preempt kernels, I doubt if this patch will fix > things. Hmm. One possible explanation: maybe the page actually _did_ get unmapped from the page tables, but got added back? I don't think we lock the page when faulting it in (we want it to be uptodate, but not necessarily locked). So assuming the pageout sequence always _does_ follow the rule that it only does try_to_free_buffers() on pages that aren't mapped, what actually protects them from not becoming mapped (and dirtied) during that sequence? So we should probably do a "wait_for_page()" in do_no_page()? Or maybe only do it for write accesses (since we don't really care about getting mapped readably)? If so, we need to do it in the write case of do_no_page() _and_ in the do_wp_page() case. Hmm? Linus