Message-ID: <1437709139.3520.73.camel@gmail.com>
Subject: Re: [patch] workqueue: schedule WORK_CPU_UNBOUND work on
 wq_unbound_cpumask CPUs
From: Mike Galbraith <umgwanakikbuti@gmail.com>
To: Tejun Heo <tj@kernel.org>
Cc: Lai Jiangshan <jiangshanlai@gmail.com>,
        Frederic Weisbecker <fweisbec@gmail.com>,
        Daniel Bristot de Oliveira <bristot@redhat.com>,
        LKML <linux-kernel@vger.kernel.org>, Rik van Riel <riel@redhat.com>,
        "Luis Claudio R. Goncalves" <lclaudio@uudg.org>
Date: Fri, 24 Jul 2015 05:38:59 +0200
In-Reply-To: <20150722152500.GN15934@mtj.duckdns.org>
References: <1437107190.3438.23.camel@gmail.com>
	 <20150717152720.GD15934@mtj.duckdns.org>
	 <1437153348.5860.32.camel@gmail.com> <20150718133602.GA3041@lerouge>
	 <1437292973.3505.83.camel@gmail.com> <1437468925.12755.57.camel@gmail.com>
	 <1437542686.3106.55.camel@gmail.com>
	 <20150722141148.GL15934@mtj.duckdns.org>
	 <1437576912.3484.8.camel@gmail.com> <1437578340.3484.15.camel@gmail.com>
	 <20150722152500.GN15934@mtj.duckdns.org>
Content-Type: text/plain; charset="UTF-8"
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2262
Lines: 54

On Wed, 2015-07-22 at 11:25 -0400, Tejun Heo wrote:
> Hello,
> 
> On Wed, Jul 22, 2015 at 05:19:00PM +0200, Mike Galbraith wrote:
> > Ew, looking at the numbers, they may prefer to either a) pretend to not
> > notice, or b) scurry off to HPC'R'US store if a) won't fly ;-)
> 
> Yeah, there are a lot of them.  The sad part is that only very few of
> them would actually need local binding for correctness. :(

I'm going to end this thread with an under my breath mutter.  Before I
do that, note that I went and beat up three different boxen over days
before turning trivial little patchlet loose, ie I was quite paranoid
despite the below, "No way Jose" was not entirely unexpected.

Mutter: WORK_CPU_UNBOUND is about as far from a local execution
guarantee as it gets, and not only is it prominently displayed for all
to see...

/**
 * queue_work - queue work on a workqueue
 * @wq: workqueue to use
 * @work: work to queue
 *
 * Returns %false if @work was already on a queue, %true otherwise.
 *
 * We queue the work to the CPU on which it was submitted, but if the CPU dies
 * it can be processed by another CPU.
 */
static inline bool queue_work(struct workqueue_struct *wq,
                              struct work_struct *work)
{
        return queue_work_on(WORK_CPU_UNBOUND, wq, work);
}

...there's nothing ambiguous about that.  While it states that we queue
locally, it also clearly states that local execution is NOT guaranteed,
making assumption thereof a BUG, which in turn makes it seem a bit odd
to tie fixing up nohz_full a little to auditing and fixing every booboo
ever made (unlikely to ever happen).  It would appear that the only way
anyone could ever have sanely assume local execution, WORK_CPU_UNBOUND
aside, is to have called get_online_cpus() before queueing any work.

Mutter expressed, it's not a big hairy deal, anyone using an SMP kernel
with an expectation of zero perturbation is doomed to be disappointed
anyway, patchlet just reduced noise a bit.

	-Mike (mutter mutter mutter;)

--
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/