Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756181AbZKKCtz (ORCPT ); Tue, 10 Nov 2009 21:49:55 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1753253AbZKKCty (ORCPT ); Tue, 10 Nov 2009 21:49:54 -0500 Received: from smtp-out.google.com ([216.239.45.13]:17700 "EHLO smtp-out.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753232AbZKKCty (ORCPT ); Tue, 10 Nov 2009 21:49:54 -0500 DomainKey-Signature: a=rsa-sha1; s=beta; d=google.com; c=nofws; q=dns; h=date:from:x-x-sender:to:cc:subject:in-reply-to:message-id: references:user-agent:mime-version:content-type:x-system-of-record; b=dy+6uJdVHpMIJybrAojYT5WS2PnpB8PEWhzYTVeAG5dRn8VRgz4F7uDqPsvP6KDEv i3Me79FpqegXkTm1OkNEA== Date: Tue, 10 Nov 2009 18:49:51 -0800 (PST) From: David Rientjes X-X-Sender: rientjes@chino.kir.corp.google.com To: KAMEZAWA Hiroyuki cc: Daisuke Nishimura , KOSAKI Motohiro , linux-mm@kvack.org, linux-kernel@vger.kernel.org, Andrew Morton , Christoph Lameter Subject: Re: [BUGFIX][PATCH] oom-kill: fix NUMA consraint check with nodemask v3 In-Reply-To: <20091111112404.0026e601.kamezawa.hiroyu@jp.fujitsu.com> Message-ID: References: <20091110162121.361B.A69D9226@jp.fujitsu.com> <20091110162445.c6db7521.kamezawa.hiroyu@jp.fujitsu.com> <20091110163419.361E.A69D9226@jp.fujitsu.com> <20091110164055.a1b44a4b.kamezawa.hiroyu@jp.fujitsu.com> <20091110170338.9f3bb417.nishimura@mxp.nes.nec.co.jp> <20091110171704.3800f081.kamezawa.hiroyu@jp.fujitsu.com> <20091111112404.0026e601.kamezawa.hiroyu@jp.fujitsu.com> User-Agent: Alpine 2.00 (DEB 1167 2008-08-23) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-System-Of-Record: true Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 5632 Lines: 151 On Wed, 11 Nov 2009, KAMEZAWA Hiroyuki wrote: > Index: mm-test-kernel/drivers/char/sysrq.c > =================================================================== > --- mm-test-kernel.orig/drivers/char/sysrq.c > +++ mm-test-kernel/drivers/char/sysrq.c > @@ -339,7 +339,7 @@ static struct sysrq_key_op sysrq_term_op > > static void moom_callback(struct work_struct *ignored) > { > - out_of_memory(node_zonelist(0, GFP_KERNEL), GFP_KERNEL, 0); > + out_of_memory(node_zonelist(0, GFP_KERNEL), GFP_KERNEL, 0, NULL); > } > > static DECLARE_WORK(moom_work, moom_callback); > Index: mm-test-kernel/mm/oom_kill.c > =================================================================== > --- mm-test-kernel.orig/mm/oom_kill.c > +++ mm-test-kernel/mm/oom_kill.c > @@ -196,27 +196,45 @@ unsigned long badness(struct task_struct > /* > * Determine the type of allocation constraint. > */ > +#ifdef CONFIG_NUMA > static inline enum oom_constraint constrained_alloc(struct zonelist *zonelist, > - gfp_t gfp_mask) > + gfp_t gfp_mask, nodemask_t *nodemask) We should probably remove the inline specifier, there's only one caller currently and if additional ones were added in the future this function should probably not be replicated. > { > -#ifdef CONFIG_NUMA > struct zone *zone; > struct zoneref *z; > enum zone_type high_zoneidx = gfp_zone(gfp_mask); > - nodemask_t nodes = node_states[N_HIGH_MEMORY]; > + int ret = CONSTRAINT_NONE; > > - for_each_zone_zonelist(zone, z, zonelist, high_zoneidx) > - if (cpuset_zone_allowed_softwall(zone, gfp_mask)) > - node_clear(zone_to_nid(zone), nodes); > - else > + /* > + * The nodemask here is a nodemask passed to alloc_pages(). Now, > + * cpuset doesn't use this nodemask for its hardwall/softwall/hierarchy > + * feature. mempolicy is an only user of nodemask here. > + */ > + if (nodemask) { > + nodemask_t mask; > + /* check mempolicy's nodemask contains all N_HIGH_MEMORY */ > + nodes_and(mask, *nodemask, node_states[N_HIGH_MEMORY]); > + if (!nodes_equal(mask, node_states[N_HIGH_MEMORY])) > + return CONSTRAINT_MEMORY_POLICY; > + } Although a nodemask_t was previously allocated on the stack, we should probably change this to use NODEMASK_ALLOC() for kernels with higher CONFIG_NODES_SHIFT since allocations can happen very deep into the stack. There should be a way around that, however. Shouldn't if (nodes_subset(node_states[N_HIGH_MEMORY], *nodemask)) return CONSTRAINT_MEMORY_POLICY; be sufficient? > + > + /* Check this allocation failure is caused by cpuset's wall function */ > + for_each_zone_zonelist_nodemask(zone, z, zonelist, > + high_zoneidx, nodemask) > + if (!cpuset_zone_allowed_softwall(zone, gfp_mask)) > return CONSTRAINT_CPUSET; > > - if (!nodes_empty(nodes)) > - return CONSTRAINT_MEMORY_POLICY; > -#endif > + /* __GFP_THISNODE never calls OOM.*/ > > return CONSTRAINT_NONE; > } > +#else > +static inline enum oom_constraint constrained_alloc(struct zonelist *zonelist, > + gfp_t gfp_mask, nodemask_t *nodemask) > +{ > + return CONSTRAINT_NONE; > +} > +#endif > > /* > * Simple selection loop. We chose the process with the highest > @@ -603,7 +621,8 @@ rest_and_return: > * OR try to be smart about which process to kill. Note that we > * don't have to be perfect here, we just have to be good. > */ > -void out_of_memory(struct zonelist *zonelist, gfp_t gfp_mask, int order) > +void out_of_memory(struct zonelist *zonelist, gfp_t gfp_mask, > + int order, nodemask_t *nodemask) > { > unsigned long freed = 0; > enum oom_constraint constraint; > @@ -622,11 +641,12 @@ void out_of_memory(struct zonelist *zone > * Check if there were limitations on the allocation (only relevant for > * NUMA) that may require different handling. > */ > - constraint = constrained_alloc(zonelist, gfp_mask); > + constraint = constrained_alloc(zonelist, gfp_mask, nodemask); > read_lock(&tasklist_lock); > > switch (constraint) { > case CONSTRAINT_MEMORY_POLICY: > + /* kill by process's its own memory alloc limitation */ I don't understand this comment. > oom_kill_process(current, gfp_mask, order, 0, NULL, > "No available memory (MPOL_BIND)"); > break; > Index: mm-test-kernel/mm/page_alloc.c > =================================================================== > --- mm-test-kernel.orig/mm/page_alloc.c > +++ mm-test-kernel/mm/page_alloc.c > @@ -1667,9 +1667,15 @@ __alloc_pages_may_oom(gfp_t gfp_mask, un > /* The OOM killer will not help higher order allocs */ > if (order > PAGE_ALLOC_COSTLY_ORDER && !(gfp_mask & __GFP_NOFAIL)) > goto out; > - > + /* > + * In usual, GFP_THISNODE contains __GFP_NORETRY and we never hit this. > + * Sanity check for bare calls of __GFP_THISNODE, not real OOM. > + * Note: Hugepage uses it but will hit PAGE_ALLOC_COSTLY_ORDER. > + */ > + if (gfp_mask & __GFP_THISNODE) > + goto out; > /* Exhausted what can be done so it's blamo time */ > - out_of_memory(zonelist, gfp_mask, order); > + out_of_memory(zonelist, gfp_mask, order, nodemask); > > out: > clear_zonelist_oom(zonelist, gfp_mask); This doesn't seem like the right place for this check; should we even try direct reclaim for bare users of __GFP_THISNODE? If we're adding it for sanity even though no callers would currently hit it, it also is a potential escape route for __GFP_NOFAIL. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/