From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756695Ab2KNLDH (ORCPT ); Wed, 14 Nov 2012 06:03:07 -0500 Received: from mail-da0-f46.google.com ([209.85.210.46]:49654 "EHLO mail-da0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752241Ab2KNLDF (ORCPT ); Wed, 14 Nov 2012 06:03:05 -0500 Date: Wed, 14 Nov 2012 03:03:02 -0800 (PST) From: David Rientjes X-X-Sender: rientjes@chino.kir.corp.google.com To: Michal Hocko cc: Andrew Morton , Greg Kroah-Hartman , KAMEZAWA Hiroyuki , KOSAKI Motohiro , linux-kernel@vger.kernel.org, linux-mm@kvack.org Subject: Re: [patch 1/4] mm, oom: ensure sysrq+f always passes valid zonelist In-Reply-To: <20121114105049.GE17111@dhcp22.suse.cz> Message-ID: References: <20121114105049.GE17111@dhcp22.suse.cz> User-Agent: Alpine 2.00 (DEB 1167 2008-08-23) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, 14 Nov 2012, Michal Hocko wrote: > > With hotpluggable and memoryless nodes, it's possible that node 0 will > > not be online, so use the first online node's zonelist rather than > > hardcoding node 0 to pass a zonelist with all zones to the oom killer. > > Makes sense although I haven't seen a machine with no 0 node yet. We routinely do testing with them, actually, just by physically removing all memory described by the SRAT that maps to node 0. You could do the same thing by making all pxms that map to node 0 to be hotpluggable in your memory affinity structure. I've been bit by it one too many times so I always keep in mind that no single node id is guaranteed to be online (although at least one node is always online); hence, first_online_node is the solution. > According to 13808910 this is indeed possible. > > > Cc: Greg Kroah-Hartman > > Signed-off-by: David Rientjes > > Reviewed-by: Michal Hocko > Thanks! From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from psmtp.com (na3sys010amx156.postini.com [74.125.245.156]) by kanga.kvack.org (Postfix) with SMTP id 851AF6B005A for ; Wed, 14 Nov 2012 06:03:05 -0500 (EST) Received: by mail-da0-f41.google.com with SMTP id i14so160983dad.14 for ; Wed, 14 Nov 2012 03:03:04 -0800 (PST) Date: Wed, 14 Nov 2012 03:03:02 -0800 (PST) From: David Rientjes Subject: Re: [patch 1/4] mm, oom: ensure sysrq+f always passes valid zonelist In-Reply-To: <20121114105049.GE17111@dhcp22.suse.cz> Message-ID: References: <20121114105049.GE17111@dhcp22.suse.cz> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-linux-mm@kvack.org List-ID: To: Michal Hocko Cc: Andrew Morton , Greg Kroah-Hartman , KAMEZAWA Hiroyuki , KOSAKI Motohiro , linux-kernel@vger.kernel.org, linux-mm@kvack.org On Wed, 14 Nov 2012, Michal Hocko wrote: > > With hotpluggable and memoryless nodes, it's possible that node 0 will > > not be online, so use the first online node's zonelist rather than > > hardcoding node 0 to pass a zonelist with all zones to the oom killer. > > Makes sense although I haven't seen a machine with no 0 node yet. We routinely do testing with them, actually, just by physically removing all memory described by the SRAT that maps to node 0. You could do the same thing by making all pxms that map to node 0 to be hotpluggable in your memory affinity structure. I've been bit by it one too many times so I always keep in mind that no single node id is guaranteed to be online (although at least one node is always online); hence, first_online_node is the solution. > According to 13808910 this is indeed possible. > > > Cc: Greg Kroah-Hartman > > Signed-off-by: David Rientjes > > Reviewed-by: Michal Hocko > Thanks! -- 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