Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Mon, 31 May 2021 18:25:42 -0500
From:   Bjorn Andersson <bjorn.andersson@linaro.org>
To:     Hillf Danton <hdanton@sina.com>
Cc:     Mathieu Poirier <mathieu.poirier@linaro.org>,
        Alex Elder <elder@linaro.org>, ohad@wizery.com,
        linux-remoteproc@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH 1/1] remoteproc: use freezable workqueue for crash
 notifications
Message-ID: <YLVwdsa97jYjKKU6@yoga>
References: <20210519234418.1196387-1-elder@linaro.org>
 <20210519234418.1196387-2-elder@linaro.org>
 <YLBpmdZoGDXNz64y@builder.lan>
 <20210529024847.5164-1-hdanton@sina.com>
 <20210530030728.8340-1-hdanton@sina.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20210530030728.8340-1-hdanton@sina.com>
Precedence: bulk

On Sat 29 May 22:07 CDT 2021, Hillf Danton wrote:

> On Sat, 29 May 2021 12:28:36 -0500 Bjorn Andersson wrote:
> >
> >Can you please explain why the mutex_lock() "requires" the context
> >executing it to be "unbound"? The lock is there to protect against
> >concurrent modifications of the state coming from e.g. sysfs.
> 
> There are simple and light events pending on the bound workqueue,
> 
> static void foo_event_fn(struct work_struct *w)
> {
> 	struct bar_struct *bar = container_of(w, struct bar_struct, work);
> 
> 	spin_lock_irq(&foo_lock);
> 	list_del(&bar->list);
> 	spin_unlock_irq(&foo_lock);
> 
> 	kfree(bar);
> 	return;
> or
> 	if (bar has waiter)
> 		wake_up();
> }
> 
> and they are not tough enough to tolerate a schedule() for which the unbound
> wq is allocated.

If you have work that is so latency sensitive that it can't handle other
work items sleeping momentarily, is it really a good idea to schedule
them on the system wide queues - or even schedule them at all?

That said, the proposed patch does not move the work from an unbound to
a bound queue, it simply moves it from one bound system queue to another
and further changes to this should be done in a separate patch - backed
by some measurements/data.

Thanks,
Bjorn