From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1758338AbZAWIBS (ORCPT ); Fri, 23 Jan 2009 03:01:18 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752713AbZAWIBE (ORCPT ); Fri, 23 Jan 2009 03:01:04 -0500 Received: from casper.infradead.org ([85.118.1.10]:54871 "EHLO casper.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752149AbZAWIBC (ORCPT ); Fri, 23 Jan 2009 03:01:02 -0500 Subject: Re: [RFC PATCH 01/19] lockdep: annotate reclaim context (__GFP_NOFS) From: Peter Zijlstra To: Nick Piggin Cc: Ingo Molnar , linux-kernel@vger.kernel.org In-Reply-To: <20090123073326.GA19986@wotan.suse.de> References: <20090122173701.674448070@chello.nl> <20090122174052.061744802@chello.nl> <1232653234.4890.135.camel@laptop> <1232656187.4890.136.camel@laptop> <20090123073326.GA19986@wotan.suse.de> Content-Type: text/plain Date: Fri, 23 Jan 2009 09:00:57 +0100 Message-Id: <1232697657.4791.8.camel@laptop> Mime-Version: 1.0 X-Mailer: Evolution 2.24.3 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, 2009-01-23 at 08:33 +0100, Nick Piggin wrote: > On Thu, Jan 22, 2009 at 09:29:47PM +0100, Peter Zijlstra wrote: > > On Thu, 2009-01-22 at 20:40 +0100, Peter Zijlstra wrote: > > > > + return; > > > > + if (DEBUG_LOCKS_WARN_ON(irqs_disabled())) > > > > + return; > > > > + > > > > + mark_held_locks(curr, RECLAIM_FS); > > > > +} > > > > > > Not quite seeing how this happens though... > > > > > > [ 6.998258] WARNING: at /usr/src/linux-2.6/kernel/lockdep.c:2266 > > > lockdep_trace_alloc+0x58/0x69() > > > > Doh, its getting there without __GFP_WAIT... > > Ah, sorry. Thanks for debugging. Should just be a matter of sticking > a check for GFP_WAIT in there somewhere? Yeah, I'll stick a check in there, that ought to suffice. > Thanks for pulling this in BTW (hopefully you don't see any more issues). > Your subsequent improvements all seem pretty nice to me too (although I > don't know the lockdep code well, but I didn't see any obviuos problems). Thanks, I'll test it some today, and hopefully stuff will just work ;-)