From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753941AbZGBR2t (ORCPT ); Thu, 2 Jul 2009 13:28:49 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752574AbZGBR23 (ORCPT ); Thu, 2 Jul 2009 13:28:29 -0400 Received: from mx2.redhat.com ([66.187.237.31]:39107 "EHLO mx2.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751535AbZGBR22 (ORCPT ); Thu, 2 Jul 2009 13:28:28 -0400 Date: Thu, 2 Jul 2009 19:25:11 +0200 From: Oleg Nesterov To: Andrew Morton Cc: Ingo Molnar , Ratan Nalumasu , Roland McGrath , Vitaly Mayatskikh , linux-kernel@vger.kernel.org Subject: [PATCH 2/3] change __wake_up_parent() to use filtered wakeup Message-ID: <20090702172511.GA24089@redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.5.18 (2008-05-17) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Ratan Nalumasu reported that in a process with many threads doing unnecessary wakeups. Every waiting thread in the process wakes up to loop through the children and see that the only ones it cares about are still not ready. Now that we have struct wait_opts we can change do_wait/__wake_up_parent to use filtered wakeups. We can make child_wait_callback() more clever later, right now it only checks eligible_child(). Signed-off-by: Oleg Nesterov --- kernel/exit.c | 25 +++++++++++++++++++++---- 1 file changed, 21 insertions(+), 4 deletions(-) --- WAIT/kernel/exit.c~2_WAKE_PARENT 2009-07-02 18:29:24.000000000 +0200 +++ WAIT/kernel/exit.c 2009-07-02 18:38:03.000000000 +0200 @@ -1089,6 +1089,7 @@ struct wait_opts { int __user *wo_stat; struct rusage __user *wo_rusage; + wait_queue_t child_wait; int notask_error; }; @@ -1558,20 +1559,35 @@ static int ptrace_do_wait(struct wait_op return 0; } +static int child_wait_callback(wait_queue_t *wait, unsigned mode, + int sync, void *key) +{ + struct wait_opts *wo = container_of(wait, struct wait_opts, + child_wait); + struct task_struct *p = key; + + if (!eligible_child(wo, p)) + return 0; + + return default_wake_function(wait, mode, sync, key); +} + void __wake_up_parent(struct task_struct *p, struct task_struct *parent) { - wake_up_interruptible_sync(&parent->signal->wait_chldexit); + __wake_up_sync_key(&parent->signal->wait_chldexit, + TASK_INTERRUPTIBLE, 1, p); } static long do_wait(struct wait_opts *wo) { - DECLARE_WAITQUEUE(wait, current); struct task_struct *tsk; int retval; trace_sched_process_wait(wo->wo_pid); - add_wait_queue(¤t->signal->wait_chldexit,&wait); + init_waitqueue_func_entry(&wo->child_wait, child_wait_callback); + wo->child_wait.private = current; + add_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); repeat: /* * If there is nothing that can match our critiera just get out. @@ -1612,7 +1628,8 @@ notask: } end: __set_current_state(TASK_RUNNING); - remove_wait_queue(¤t->signal->wait_chldexit,&wait); + remove_wait_queue(¤t->signal->wait_chldexit, &wo->child_wait); + if (wo->wo_info) { struct siginfo __user *infop = wo->wo_info;