Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932093AbWH0Mav (ORCPT ); Sun, 27 Aug 2006 08:30:51 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S932099AbWH0Mav (ORCPT ); Sun, 27 Aug 2006 08:30:51 -0400 Received: from taganka54-host.corbina.net ([213.234.233.54]:13710 "EHLO screens.ru") by vger.kernel.org with ESMTP id S932093AbWH0Mav (ORCPT ); Sun, 27 Aug 2006 08:30:51 -0400 Date: Sun, 27 Aug 2006 20:54:44 +0400 From: Oleg Nesterov To: Andrew Morton Cc: Nick Piggin , linux-kernel@vger.kernel.org Subject: [PATCH -mm] select_bad_process: kill a bogus PF_DEAD/TASK_DEAD check Message-ID: <20060827165444.GA1162@oleg> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.5.11 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1763 Lines: 62 (Compile tested, on top of introduce-task_dead-state.patch) The only one usage of TASK_DEAD outside of last schedule path, select_bad_process: for_each_task(p) { if (!p->mm) continue; ... if (p->state == TASK_DEAD) continue; ... TASK_DEAD state is set at the end of do_exit(), this means that p->mm was already set == NULL by exit_mm(), so this task was already rejected by 'if (!p->mm)' above. Note also that the caller holds tasklist_lock, this means that p can't pass exit_notify() and then set TASK_DEAD when p->mm != NULL. Also, remove open-coded is_init(). Signed-off-by: Oleg Nesterov --- 2.6.18-rc4/mm/oom_kill.c~ 2006-08-27 19:49:27.000000000 +0400 +++ 2.6.18-rc4/mm/oom_kill.c 2006-08-27 20:16:28.000000000 +0400 @@ -207,11 +207,14 @@ static struct task_struct *select_bad_pr unsigned long points; int releasing; - /* skip kernel threads */ + /* + * skip kernel threads and tasks which have already released + * their mm. + */ if (!p->mm) continue; - /* skip the init task with pid == 1 */ - if (p->pid == 1) + /* skip the init task */ + if (is_init(p)) continue; /* @@ -227,9 +230,6 @@ static struct task_struct *select_bad_pr releasing = test_tsk_thread_flag(p, TIF_MEMDIE) || p->flags & PF_EXITING; if (releasing) { - /* TASK_DEAD tasks have already released their mm */ - if (p->state == TASK_DEAD) - continue; if (p->flags & PF_EXITING && p == current) { chosen = p; *ppoints = ULONG_MAX; - 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/