From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751084AbbI2Wjz (ORCPT ); Tue, 29 Sep 2015 18:39:55 -0400 Received: from mail-pa0-f51.google.com ([209.85.220.51]:33370 "EHLO mail-pa0-f51.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750813AbbI2Wjs (ORCPT ); Tue, 29 Sep 2015 18:39:48 -0400 Date: Tue, 29 Sep 2015 15:39:45 -0700 (PDT) From: David Rientjes X-X-Sender: rientjes@chino.kir.corp.google.com To: Oleg Nesterov cc: Andrew Morton , Kyle Walker , Michal Hocko , Stanislav Kozina , Tetsuo Handa , linux-kernel@vger.kernel.org Subject: Re: [PATCH -mm 2/3] mm/oom_kill: cleanup the "kill sharing same memory" In-Reply-To: <20150929141818.GA10955@redhat.com> Message-ID: References: <20150929141818.GA10955@redhat.com> User-Agent: Alpine 2.10 (DEB 1266 2009-07-14) 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 Tue, 29 Sep 2015, Oleg Nesterov wrote: > Purely cosmetic, but the complex "if" condition looks annoying to me. > Especially because it is not consistent with OOM_SCORE_ADJ_MIN check > which adds another if/continue. > > Signed-off-by: Oleg Nesterov > --- > mm/oom_kill.c | 22 +++++++++++++--------- > 1 file changed, 13 insertions(+), 9 deletions(-) > > diff --git a/mm/oom_kill.c b/mm/oom_kill.c > index 0d581c6..8e7bed2 100644 > --- a/mm/oom_kill.c > +++ b/mm/oom_kill.c > @@ -583,16 +583,20 @@ void oom_kill_process(struct oom_control *oc, struct task_struct *p, > * pending fatal signal. > */ > rcu_read_lock(); > - for_each_process(p) > - if (p->mm == mm && !same_thread_group(p, victim) && > - !(p->flags & PF_KTHREAD)) { > - if (p->signal->oom_score_adj == OOM_SCORE_ADJ_MIN) > - continue; > + for_each_process(p) { > + if (unlikely(p->flags & PF_KTHREAD)) > + continue; > + if (same_thread_group(p, victim)) > + continue; > + if (p->mm != mm) > + continue; This ordering is a little weird to me, I think we would eliminate the majority of processes by checking for p->mm != mm first. There are certainly pathological cases where that can be defeated, but in practice it seems to happen more often than not. Unless you object, I think the ordering should be p->mm != mm, same_thread_group(), unlikely(PF_KTHREAD) as it originally was (thanks for adding the unlikely). I agree your cleanup looks much better than the nested conditional.