From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1030607AbWBHUOo (ORCPT ); Wed, 8 Feb 2006 15:14:44 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1030609AbWBHUOo (ORCPT ); Wed, 8 Feb 2006 15:14:44 -0500 Received: from science.horizon.com ([192.35.100.1]:11849 "HELO science.horizon.com") by vger.kernel.org with SMTP id S1030607AbWBHUOn (ORCPT ); Wed, 8 Feb 2006 15:14:43 -0500 Date: 8 Feb 2006 15:14:36 -0500 Message-ID: <20060208201436.14693.qmail@science.horizon.com> From: linux@horizon.com To: clameter@engr.sgi.com Subject: Re: Terminate process that fails on a constrained allocation Cc: linux-kernel@vger.kernel.org Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Would perhaps a less drastic solution, that at least supports the common partitioned-system configuration, be to limit the oom killer to processes whose nodes are a *subset* of ours? That way, a limited process won't kill any unlimited processes, but it can fight with other processes with the same limits. However, an unlimited process discovering oom can kill anything on the system if necessary. (This requires a modified version of cpuset_excl_nodes_overlap that calls nodes_subset() instead of nodes_intersects(), but it's pretty straightforward.)