From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail144.messagelabs.com (mail144.messagelabs.com [216.82.254.51]) by kanga.kvack.org (Postfix) with ESMTP id DE8896B012A for ; Mon, 21 Sep 2009 05:09:19 -0400 (EDT) Date: Mon, 21 Sep 2009 10:09:23 +0100 From: Mel Gorman Subject: Re: [PATCH 1/3] slqb: Do not use DEFINE_PER_CPU for per-node data Message-ID: <20090921090923.GH12726@csn.ul.ie> References: <1253302451-27740-1-git-send-email-mel@csn.ul.ie> <1253302451-27740-2-git-send-email-mel@csn.ul.ie> <84144f020909200145w74037ab9vb66dae65d3b8a048@mail.gmail.com> <4AB5FD4D.3070005@kernel.org> <4AB5FFF8.7000602@cs.helsinki.fi> <4AB6508C.4070602@kernel.org> <4AB739A6.5060807@in.ibm.com> <20090921084248.GC12726@csn.ul.ie> <4AB74129.90402@in.ibm.com> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-15 Content-Disposition: inline In-Reply-To: <4AB74129.90402@in.ibm.com> Sender: owner-linux-mm@kvack.org To: Sachin Sant Cc: Tejun Heo , Pekka Enberg , Nick Piggin , Christoph Lameter , heiko.carstens@de.ibm.com, linux-kernel@vger.kernel.org, linux-mm@kvack.org, Andrew Morton , Benjamin Herrenschmidt List-ID: On Mon, Sep 21, 2009 at 02:32:33PM +0530, Sachin Sant wrote: > Mel Gorman wrote: >>> I applied the three patches from Mel and one from Tejun. >>> >> >> Thanks Sachin >> >> Was there any useful result from Tejun's patch applied on its own? >> > Haven't tried with just the patch from Tejun. I will give this a try. > I might not get a chance to test this until late in the evening my time. > (Today being a holiday for me ) > Ok, I've blatently nabbed your machine again then (yoink haha) and a relevant kernel is building at the moment. I'll put the dmesg when it becomes available. Go enjoy your holiday >>> Tejun, the above hang looks exactly the same as the one >>> i have reported here : >>> >>> http://lists.ozlabs.org/pipermail/linuxppc-dev/2009-September/075791.html >>> >>> This particular hang was bisected to the following patch >>> >>> powerpc64: convert to dynamic percpu allocator >>> >>> This hang can be recreated without SLQB. So i think this is a different >>> problem. >>> >>> >> >> Was that bug ever resolved? >> > The bug was still present with git9(78f28b..). With latest git > git10(ebc79c4 ..)i haven't tested it yet because of perf counter > build errors. > Ok, so right now there are three separate bugs with this machine 1. SLQB + Memoryless shots itself in the face due to suspected per-cpu area corruption 2. SLQB with memoryless nodes frees pages remotely and allocates locally with a similar effect as if it was a memory leak. 3. pcpu_alloc can hang but is not related to SLQB Thanks -- Mel Gorman Part-time Phd Student Linux Technology Center University of Limerick IBM Dublin Software Lab -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: email@kvack.org