From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ted Ts'o Subject: Re: Oops while going into hibernate Date: Thu, 13 Jan 2011 00:56:12 -0500 Message-ID: <20110113055612.GF13496@thunk.org> References: <20110112162655.GA13496@thunk.org> <20110112172646.GB13496@thunk.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: "linux-ext4@vger.kernel.org development" , LKML Kernel , pm list To: Sebastian Ott Return-path: Received: from thunk.org ([69.25.196.29]:50318 "EHLO thunker.thunk.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750912Ab1AMF4S (ORCPT ); Thu, 13 Jan 2011 00:56:18 -0500 Content-Disposition: inline In-Reply-To: Sender: linux-ext4-owner@vger.kernel.org List-ID: On Wed, Jan 12, 2011 at 07:44:17PM -0500, Theodore Tso wrote: > > You said originally that the oops was happening "while going into > hibernation right after resuming with...". So that means you did a > successful suspend/resume, and then the second suspend caused the > oops? It looks like somehow the pages were left marked as dirty, so > the writeback daemons attempted writing back a page to an inode > which was never opened read/write (and in fact as a text page for > /usr/bin/killall, was mapped read/only). Given that ext4 > initializes jinode only when the file is opened read/write, the fact > that it is null, and the fact that it makes no sense that a program > would be modifying /usr/bin/killall as part of a suspend/resume, it > looks very much like we just unmasked a software suspend bug.... ... and I think I've found the problem. In kernel/power/block_io.c, in the function submit(), we see this: if (bio_chain == NULL) { submit_bio(bio_rw, bio); wait_on_page_locked(page); if (rw == READ) bio_set_pages_dirty(bio); <==== bio_put(bio); So when we read in pages from the software suspend device, we end up marking the pages as dirty(!). I'm guessing this was caused by a copy and paste from the only other caller of bio_set_pages_dirty(), which is the direct I/O code, which needs this when we are writing from a file into a user-provided buffer. But for restoring from a software suspend case, this is as far as I can tell wholely inappropriate. This causes needless writes, which is bad even before ext4 unmasked the problem. I will send a patch under separate cover; could you give it a try and see if it fixes your crash? I will look into bulletproofing ext4 by adding checks for this case and printing warning messages, but neverthe less, I think the root cause is actually in the hibernation's bio code. - Ted