From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1759441AbYHULdx (ORCPT ); Thu, 21 Aug 2008 07:33:53 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1751744AbYHULdp (ORCPT ); Thu, 21 Aug 2008 07:33:45 -0400 Received: from gir.skynet.ie ([193.1.99.77]:33029 "EHLO gir.skynet.ie" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750828AbYHULdo (ORCPT ); Thu, 21 Aug 2008 07:33:44 -0400 Date: Thu, 21 Aug 2008 12:33:39 +0100 From: Mel Gorman To: Adam Litke Cc: Dave Hansen , linux-mm , linux-kernel , Andrew Morton , nacc , apw , agl Subject: Re: [BUG] [PATCH v2] Make setup_zone_migrate_reserve() aware of overlapping nodes Message-ID: <20080821113338.GA29950@csn.ul.ie> References: <1218837685.12953.11.camel@localhost.localdomain> <1219252134.13885.25.camel@localhost.localdomain> <1219255911.8960.41.camel@nimitz> <1219262152.13885.27.camel@localhost.localdomain> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-15 Content-Disposition: inline In-Reply-To: <1219262152.13885.27.camel@localhost.localdomain> User-Agent: Mutt/1.5.13 (2006-08-11) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On (20/08/08 14:55), Adam Litke didst pronounce: > Changes since V1 > - Fix build for !NUMA > - Add VM_BUG_ON() to catch this problem at the source > > I have gotten to the root cause of the hugetlb badness I reported back on > August 15th. My system has the following memory topology (note the > overlapping node): > > Node 0 Memory: 0x8000000-0x44000000 > Node 1 Memory: 0x0-0x8000000 0x44000000-0x80000000 > > setup_zone_migrate_reserve() scans the address range 0x0-0x8000000 looking > for a pageblock to move onto the MIGRATE_RESERVE list. Finding no > candidates, it happily continues the scan into 0x8000000-0x44000000. When > a pageblock is found, the pages are moved to the MIGRATE_RESERVE list on > the wrong zone. Oops. > > (Andrew: once the proper fix is agreed upon, this should also be a > candidate for -stable.) > > setup_zone_migrate_reserve() should skip pageblocks in overlapping nodes. > > Signed-off-by: Adam Litke > zone_to_nid(zone) is called every time in the loop even though it will never change. This is less than optimal but setup_zone_migrate_reserve() is only called during init and when min_free_kbytes is adjusted so it's not worth worrying about. Otherwise it looks good. Acked-by: Mel Gorman > diff --git a/mm/page_alloc.c b/mm/page_alloc.c > index af982f7..feb7916 100644 > --- a/mm/page_alloc.c > +++ b/mm/page_alloc.c > @@ -694,6 +694,9 @@ static int move_freepages(struct zone *zone, > #endif > > for (page = start_page; page <= end_page;) { > + /* Make sure we are not inadvertently changing nodes */ > + VM_BUG_ON(page_to_nid(page) != zone_to_nid(zone)); > + > if (!pfn_valid_within(page_to_pfn(page))) { > page++; > continue; > @@ -2516,6 +2519,10 @@ static void setup_zone_migrate_reserve(struct zone *zone) > continue; > page = pfn_to_page(pfn); > > + /* Watch out for overlapping nodes */ > + if (page_to_nid(page) != zone_to_nid(zone)) > + continue; > + > /* Blocks with reserved pages will never free, skip them. */ > if (PageReserved(page)) > continue; > > -- > Adam Litke - (agl at us.ibm.com) > IBM Linux Technology Center > -- Mel Gorman Part-time Phd Student Linux Technology Center University of Limerick IBM Dublin Software Lab