Re: [PATCH] sched: Do not release current rq lock on non contended double_lock_balance()

From: Peter Zijlstra <peterz@infradead.org>
To: Steven Rostedt <rostedt@goodmis.org>
Cc: LKML <linux-kernel@vger.kernel.org>,
	Ingo Molnar <mingo@kernel.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	Clark Williams <williams@redhat.com>,
	Andrew Morton <akpm@linux-foundation.org>,
	Nick Piggin <nickpiggin@yahoo.com.au>
Subject: Re: [PATCH] sched: Do not release current rq lock on non contended double_lock_balance()
Date: Wed, 15 Jun 2016 13:14:53 +0200	[thread overview]
Message-ID: <20160615111453.GG30909@twins.programming.kicks-ass.net> (raw)
In-Reply-To: <20160614140228.0ecf15af@grimm.local.home>

On Tue, Jun 14, 2016 at 02:02:28PM -0400, Steven Rostedt wrote:
> > For instance; pull_rt_task() does:
> > 
> > 	for_each_cpu() {
> > 		double_lock_balance(this, that);
> > 		...
> > 		double_unlock_balance(this, that);
> > 	}
> > 
> > Which, with the trylock, ends up with a max possible hold time of
> > O(nr_cpus).
> 
> Sure, but I think we should try to limit that loop too, because that
> loop itself is what is triggering the large latency for me, because
> it constantly releases a spinlock and has to wait. This loop is done
> with preemption disabled.

OK, so should not the whole HAVE_RT_PUSH_IPI thing have avoided that
loop entirely? And therefore made the point moot?

In any case, can't we add another cpupri for pushable tasks and use that
to find the highest priority task to pull and avoid the loop thus?