Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754295AbbHCTHr (ORCPT ); Mon, 3 Aug 2015 15:07:47 -0400 Received: from casper.infradead.org ([85.118.1.10]:50680 "EHLO casper.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753901AbbHCTHo (ORCPT ); Mon, 3 Aug 2015 15:07:44 -0400 Date: Mon, 3 Aug 2015 21:07:40 +0200 From: Peter Zijlstra To: Frederic Weisbecker Cc: Chris Metcalf , LKML , Thomas Gleixner , Preeti U Murthy , Christoph Lameter , Ingo Molnar , Viresh Kumar , Rik van Riel Subject: Re: [PATCH 08/10] posix-cpu-timers: Migrate to use new tick dependency mask model Message-ID: <20150803190740.GB25159@twins.programming.kicks-ass.net> References: <55B26E74.5040803@ezchip.com> <20150729132343.GC11554@lerouge> <55B90C40.5090000@ezchip.com> <20150730004444.GA14744@lerouge> <55BA7C6A.1050602@ezchip.com> <20150730194519.GA24607@lerouge> <55BA8096.7030601@ezchip.com> <20150731144954.GB27875@lerouge> <20150803171243.GW25159@twins.programming.kicks-ass.net> <20150803173936.GC26022@lerouge> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20150803173936.GC26022@lerouge> User-Agent: Mutt/1.5.21 (2012-12-30) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 2815 Lines: 81 On Mon, Aug 03, 2015 at 07:39:37PM +0200, Frederic Weisbecker wrote: > On Mon, Aug 03, 2015 at 07:12:43PM +0200, Peter Zijlstra wrote: > > On Fri, Jul 31, 2015 at 04:49:55PM +0200, Frederic Weisbecker wrote: > > > Instead of doing a per signal dependency, I'm going to use a per task > > > one. > > > > Urgh, does this mean you'll keep the horrid tick_nohz_task_switch() > > thing? > > I thought I would drop it, but now that I think about it more, I think I > need to keep it because if we enqueue a posix timer to a sleeping task > and that task later wakes up, we must restart the tick, and that is only > possible with a check on context switch :-( > > This current patchset removed the need for that with a global dependency > for posix timers: as long as there is one enqueued we keep the tick. But > Chris and Luiz fear that Tilera users have posix timers on housekeepers. > They also suggested we offline the posix timers. I fear it's going to be > a high overhead as it means polling on the target task context of execution. > Unless we move the task itself to housekeepers... At least do something like the below, that irq save/restore is expensive and can be trivially avoided. Also, tick_nohz_full_kick() checks that tick_nohz_full_cpu() thing again. --- kernel/sched/core.c | 2 +- kernel/time/tick-sched.c | 9 +-------- 2 files changed, 2 insertions(+), 9 deletions(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 4d34035bb3ee..57d1af7c0660 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -2471,6 +2471,7 @@ static struct rq *finish_task_switch(struct task_struct *prev) vtime_task_switch(prev); finish_arch_switch(prev); perf_event_task_sched_in(prev, current); + tick_nohz_task_switch(); finish_lock_switch(rq, prev); finish_arch_post_lock_switch(); @@ -2489,7 +2490,6 @@ static struct rq *finish_task_switch(struct task_struct *prev) put_task_struct(prev); } - tick_nohz_task_switch(); return rq; } diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c index 3319e16f31e5..9f2225ef230c 100644 --- a/kernel/time/tick-sched.c +++ b/kernel/time/tick-sched.c @@ -260,18 +260,11 @@ void tick_nohz_full_kick_all(void) */ void __tick_nohz_task_switch(void) { - unsigned long flags; - - local_irq_save(flags); - if (!tick_nohz_full_cpu(smp_processor_id())) - goto out; + return; if (tick_nohz_tick_stopped() && !can_stop_full_tick()) tick_nohz_full_kick(); - -out: - local_irq_restore(flags); } /* Parse the boot-time nohz CPU list from the kernel parameters. */ -- 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/