From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754925Ab3FGNLd (ORCPT ); Fri, 7 Jun 2013 09:11:33 -0400 Received: from mx1.redhat.com ([209.132.183.28]:65194 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752382Ab3FGNLc (ORCPT ); Fri, 7 Jun 2013 09:11:32 -0400 Date: Fri, 7 Jun 2013 15:07:23 +0200 From: Oleg Nesterov To: Tejun Heo Cc: Andrew Morton , Daniel Vetter , Dave Jones , David Howells , Imre Deak , Jens Axboe , Linus Torvalds , Lukas Czerner , "Paul E. McKenney" , linux-kernel@vger.kernel.org Subject: Re: [PATCH 2/2] wait: introduce prepare_to_wait_event() Message-ID: <20130607130723.GA2777@redhat.com> References: <20130606200257.GA23628@redhat.com> <20130606200334.GC23628@redhat.com> <20130606213636.GJ5045@htj.dyndns.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20130606213636.GJ5045@htj.dyndns.org> 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 Hi Tejun, On 06/06, Tejun Heo wrote: > > Which tree are the patches based on? I'm getting conflicts on > linus#master, mmotd and next. Hmm. linus#master. I just verified that the patches apply cleanly after git-pull. > Reviewed-by: Tejun Heo Thanks, > But a couple nits. > > > +int > > I don't think we need to keep the unnecessary line break here. All > other functions in the file don't do it except for the two > prepare_to_wait functions. No need to give the weirdos more power. :) Yes, I tried to follow the style around the new helper, but the extra line is not needed. Please see v2 below, no other changes. > > +prepare_to_wait_event(wait_queue_head_t *q, wait_queue_t *wait, int state) > > +{ > > + if (signal_pending_state(state, current)) > > + return -ERESTARTSYS; > > + > > + wait->private = current; > > + wait->func = autoremove_wake_function; > > + prepare_to_wait(q, wait, state); > > I wonder whether it'd be worthwhile to make prepare_to_wait() inline > so that it can be inlined into the above. I think gcc is smart enough > to emit inline for in-file stuff and then build a proper function for > external references. No biggie. Just wondering. Not sure... As you can see from this patch, I like -Os more than -O2 ;) ------------------------------------------------------------------------------ Subject: [PATCH v2 2/2] wait: introduce prepare_to_wait_event() Add the new helper, prepare_to_wait_event() which should only be used by wait_event_common/etc. prepare_to_wait_event() returns -ERESTARTSYS if signal_pending_state() is true, otherwise it calls prepare_to_wait(). This allows to uninline the signal-pending checks in wait_event_*. Also, it can initialize wait->private/func. We do not care they were already initialized, the values are the same. This also shaves a couple of insns from the inlined code. Unlike the previous change, this patch "reliably" shrinks the size of generated code for every wait_event*() call. Signed-off-by: Oleg Nesterov Reviewed-by: Tejun Heo --- include/linux/wait.h | 22 +++++++++++----------- kernel/wait.c | 13 +++++++++++++ 2 files changed, 24 insertions(+), 11 deletions(-) diff --git a/include/linux/wait.h b/include/linux/wait.h index 392c54d..a2314dd 100644 --- a/include/linux/wait.h +++ b/include/linux/wait.h @@ -176,28 +176,27 @@ wait_queue_head_t *bit_waitqueue(void *, int); #define __wait_no_timeout(tout) \ (__builtin_constant_p(tout) && (tout) == MAX_SCHEDULE_TIMEOUT) -/* uglified signal_pending_state() optimized for constant state */ -#define __wait_signal_pending(state) \ - ((state == TASK_INTERRUPTIBLE) ? signal_pending(current) : \ - (state == TASK_KILLABLE) ? fatal_signal_pending(current) : \ - 0) +#define __wait_interruptible(state) \ + (!__builtin_constant_p(state) || \ + state == TASK_INTERRUPTIBLE || state == TASK_KILLABLE) #define __wait_event_common(wq, condition, state, tout) \ ({ \ - DEFINE_WAIT(__wait); \ - long __ret = 0, __tout = tout; \ + long __ret, __tout = tout; \ + wait_queue_t __wait; \ + \ + INIT_LIST_HEAD(&__wait.task_list); \ + __wait.flags = 0; \ \ for (;;) { \ - prepare_to_wait(&wq, &__wait, state); \ + __ret = prepare_to_wait_event(&wq, &__wait, state); \ if (condition) { \ __ret = __wait_no_timeout(tout) ?: __tout ?: 1; \ break; \ } \ \ - if (__wait_signal_pending(state)) { \ - __ret = -ERESTARTSYS; \ + if (__wait_interruptible(state) && __ret) \ break; \ - } \ \ if (__wait_no_timeout(tout)) \ schedule(); \ @@ -781,6 +780,7 @@ extern long interruptible_sleep_on_timeout(wait_queue_head_t *q, * Waitqueues which are removed from the waitqueue_head at wakeup time */ void prepare_to_wait(wait_queue_head_t *q, wait_queue_t *wait, int state); +int prepare_to_wait_event(wait_queue_head_t *q, wait_queue_t *wait, int state); void prepare_to_wait_exclusive(wait_queue_head_t *q, wait_queue_t *wait, int state); void finish_wait(wait_queue_head_t *q, wait_queue_t *wait); void abort_exclusive_wait(wait_queue_head_t *q, wait_queue_t *wait, diff --git a/kernel/wait.c b/kernel/wait.c index 6698e0c..3b8619a 100644 --- a/kernel/wait.c +++ b/kernel/wait.c @@ -78,6 +78,19 @@ prepare_to_wait(wait_queue_head_t *q, wait_queue_t *wait, int state) } EXPORT_SYMBOL(prepare_to_wait); +int prepare_to_wait_event(wait_queue_head_t *q, wait_queue_t *wait, int state) +{ + if (signal_pending_state(state, current)) + return -ERESTARTSYS; + + wait->private = current; + wait->func = autoremove_wake_function; + prepare_to_wait(q, wait, state); + + return 0; +} +EXPORT_SYMBOL(prepare_to_wait_event); + void prepare_to_wait_exclusive(wait_queue_head_t *q, wait_queue_t *wait, int state) { -- 1.5.5.1