Re: [RT PATCH 2/2] block: blk-mq: move blk_queue_usage_counter_release() into process context

From: Peter Zijlstra <peterz@infradead.org>
To: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: linux-rt-users <linux-rt-users@vger.kernel.org>,
	Corey Minyard <cminyard@mvista.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	Steven Rostedt <rostedt@goodmis.org>,
	linux-kernel <linux-kernel@vger.kernel.org>,
	Tejun Heo <tj@kernel.org>
Subject: Re: [RT PATCH 2/2] block: blk-mq: move blk_queue_usage_counter_release() into process context
Date: Tue, 13 Mar 2018 21:10:39 +0100	[thread overview]
Message-ID: <20180313201039.GB4082@hirez.programming.kicks-ass.net> (raw)
In-Reply-To: <20180313184240.ndq2pw2fmvhkwidc@linutronix.de>

On Tue, Mar 13, 2018 at 07:42:41PM +0100, Sebastian Andrzej Siewior wrote:
> +static void blk_queue_usage_counter_release_swork(struct swork_event *sev)
> +{
> +	struct request_queue *q =
> +		container_of(sev, struct request_queue, mq_pcpu_wake);
> +
> +	wake_up_all(&q->mq_freeze_wq);
> +}
> +
>  static void blk_queue_usage_counter_release(struct percpu_ref *ref)
>  {
>  	struct request_queue *q =
>  		container_of(ref, struct request_queue, q_usage_counter);
>  
> -	wake_up_all(&q->mq_freeze_wq);
> +	swork_queue(&q->mq_pcpu_wake);
>  }

Depending on if we expect there to actually be wakeups, you could do
something like:

	if (wq_has_sleepers(&q->mq_freeze_wq))
		swork_queue(&q->mq_pcpu_wake));

avoiding the whole workqueue thing in the case there wasn't anybody
waiting for it. But since I don't know this code, I can't say if it
makes sense or not. Tejun?