Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Sender: Tejun Heo <htejun@gmail.com>
Date:   Mon, 26 Jun 2023 11:00:29 -1000
From:   Tejun Heo <tj@kernel.org>
To:     rdunlap@infradead.org, linux-kernel@vger.kernel.org,
        jiangshanlai@gmail.com, zyhtheonly@gmail.com, zyhtheonly@yeah.net,
        zwp10758@gmail.com
Subject: Re: [PATCH v3] workqueue: add cmdline parameter
 `workqueue_unbound_cpus` to further constrain wq_unbound_cpumask at boot
 time
Message-ID: <ZJn8bQd6EdHolayS@slm.duckdns.org>
References: <20230623073033.GA6584@didi-ThinkCentre-M930t-N000>
 <20230625031502.GA22708@didi-ThinkCentre-M930t-N000>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230625031502.GA22708@didi-ThinkCentre-M930t-N000>
Precedence: bulk

Hello,

On Sun, Jun 25, 2023 at 11:15:02AM +0800, tiozhang wrote:
> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
> index a465d5242774..7f2fe8c60d5c 100644
> --- a/Documentation/admin-guide/kernel-parameters.txt
> +++ b/Documentation/admin-guide/kernel-parameters.txt
> @@ -6780,6 +6780,14 @@
>  			disables both lockup detectors. Default is 10
>  			seconds.
>  
> +	workqueue_unbound_cpus=

Please use workqueue.unbound_cpus for consistency.

> +			[KNL,SMP]
> +			Format: <cpu-list>
> +			Specify to constrain one or some CPUs to use in
> +			unbound workqueues.
> +			By default, all online CPUs are available for
> +			unbound workqueues.

and flow the paragraph.

>  	workqueue.watchdog_thresh=
>  			If CONFIG_WQ_WATCHDOG is configured, workqueue can
>  			warn stall conditions and dump internal state to
> diff --git a/kernel/workqueue.c b/kernel/workqueue.c
> index 7cd5f5e7e0a1..c247725b0873 100644
> --- a/kernel/workqueue.c
> +++ b/kernel/workqueue.c
> @@ -329,6 +329,9 @@ static bool workqueue_freezing;		/* PL: have wqs started freezing? */
>  /* PL: allowable cpus for unbound wqs and work items */
>  static cpumask_var_t wq_unbound_cpumask;
>  
> +/* for further constrain wq_unbound_cpumask by cmdline parameter*/
> +static cpumask_var_t wq_cmdline_cpumask;
> +
>  /* CPU where unbound work was last round robin scheduled from this CPU */
>  static DEFINE_PER_CPU(int, wq_rr_cpu_last);
>  
> @@ -6006,6 +6009,10 @@ void __init workqueue_init_early(void)
>  	cpumask_copy(wq_unbound_cpumask, housekeeping_cpumask(HK_TYPE_WQ));
>  	cpumask_and(wq_unbound_cpumask, wq_unbound_cpumask, housekeeping_cpumask(HK_TYPE_DOMAIN));
>  
> +	if (!cpumask_empty(wq_cmdline_cpumask))
> +		cpumask_and(wq_unbound_cpumask, wq_unbound_cpumask, wq_cmdline_cpumask);
> +	free_bootmem_cpumask_var(wq_cmdline_cpumask);

If workqueue_unbound_cpus_setup() wasn't called during boot because the
parameter wasn't set, the above would end up trying to access and fre
wq_cmdline_cpumask which hasn't been allocated. Let's just use statically
allocated __initdata struct cpumask so that we don't have to worry about
allocation and freeing during early boot.

> +static int __init workqueue_unbound_cpus_setup(char *str)
> +{
> +	cpumask_var_t cpumask;
> +
> +	alloc_bootmem_cpumask_var(&wq_cmdline_cpumask);
> +	alloc_bootmem_cpumask_var(&cpumask);
> +	if (cpulist_parse(str, cpumask) < 0)
> +		pr_warn("workqueue_unbound_cpus: incorrect CPU range\n");
> +	else
> +		cpumask_copy(wq_cmdline_cpumask, cpumask);

You can just parse direclty into wq_cmdline_cpumask and then clear it on
error.

Thanks.

-- 
tejun