wifi: wfx: add WQ_PERCPU to alloc_workqueue users

[PATCH] wifi: wfx: add WQ_PERCPU to alloc_workqueue users

Posted by Marco Crivellari 2 months, 3 weeks ago

Currently if a user enqueues a work item using schedule_delayed_work() the
used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use
WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to
schedule_work() that is using system_wq and queue_work(), that makes use
again of WORK_CPU_UNBOUND.
This lack of consistency cannot be addressed without refactoring the API.
For more details see the Link tag below.

alloc_workqueue() treats all queues as per-CPU by default, while unbound
workqueues must opt-in via WQ_UNBOUND.

This default is suboptimal: most workloads benefit from unbound queues,
allowing the scheduler to place worker threads where they’re needed and
reducing noise when CPUs are isolated.

This continues the effort to refactor workqueue APIs, which began with
the introduction of new workqueues and a new alloc_workqueue flag in:

commit 128ea9f6ccfb ("workqueue: Add system_percpu_wq and system_dfl_wq")
commit 930c2ea566af ("workqueue: Add new WQ_PERCPU flag")

This change adds a new WQ_PERCPU flag to explicitly request
alloc_workqueue() to be per-cpu when WQ_UNBOUND has not been specified.

With the introduction of the WQ_PERCPU flag (equivalent to !WQ_UNBOUND),
any alloc_workqueue() caller that doesn’t explicitly specify WQ_UNBOUND
must now use WQ_PERCPU.

Once migration is complete, WQ_UNBOUND can be removed and unbound will
become the implicit default.

Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
Link: https://lore.kernel.org/all/20250221112003.1dSuoGyc@linutronix.de/
---
 drivers/net/wireless/silabs/wfx/main.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/wireless/silabs/wfx/main.c b/drivers/net/wireless/silabs/wfx/main.c
index a61128debbad..dda36e41eed1 100644
--- a/drivers/net/wireless/silabs/wfx/main.c
+++ b/drivers/net/wireless/silabs/wfx/main.c
@@ -364,7 +364,7 @@ int wfx_probe(struct wfx_dev *wdev)
 	wdev->pdata.gpio_wakeup = NULL;
 	wdev->poll_irq = true;
 
-	wdev->bh_wq = alloc_workqueue("wfx_bh_wq", WQ_HIGHPRI, 0);
+	wdev->bh_wq = alloc_workqueue("wfx_bh_wq", WQ_HIGHPRI | WQ_PERCPU, 0);
 	if (!wdev->bh_wq)
 		return -ENOMEM;
 
-- 
2.51.1

Re: [PATCH] wifi: wfx: add WQ_PERCPU to alloc_workqueue users

Posted by Marco Crivellari 4 weeks, 1 day ago

On Thu, Nov 13, 2025 at 5:08 PM Marco Crivellari
<marco.crivellari@suse.com> wrote:
> [...]
>  drivers/net/wireless/silabs/wfx/main.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)

Gentle ping.

Thanks!

-- 

Marco Crivellari

L3 Support Engineer

Re: [PATCH] wifi: wfx: add WQ_PERCPU to alloc_workqueue users

Posted by Jérôme Pouiller 2 months, 3 weeks ago

On Thursday 13 November 2025 17:08:25 Central European Standard Time Marco Crivellari wrote:
> CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you recognize the sender and know the content is safe.
> 
> 
> Currently if a user enqueues a work item using schedule_delayed_work() the
> used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use
> WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to
> schedule_work() that is using system_wq and queue_work(), that makes use
> again of WORK_CPU_UNBOUND.
> This lack of consistency cannot be addressed without refactoring the API.
> For more details see the Link tag below.
> 
> alloc_workqueue() treats all queues as per-CPU by default, while unbound
> workqueues must opt-in via WQ_UNBOUND.
> 
> This default is suboptimal: most workloads benefit from unbound queues,
> allowing the scheduler to place worker threads where they’re needed and
> reducing noise when CPUs are isolated.
> 
> This continues the effort to refactor workqueue APIs, which began with
> the introduction of new workqueues and a new alloc_workqueue flag in:
> 
> commit 128ea9f6ccfb ("workqueue: Add system_percpu_wq and system_dfl_wq")
> commit 930c2ea566af ("workqueue: Add new WQ_PERCPU flag")
> 
> This change adds a new WQ_PERCPU flag to explicitly request
> alloc_workqueue() to be per-cpu when WQ_UNBOUND has not been specified.
> 
> With the introduction of the WQ_PERCPU flag (equivalent to !WQ_UNBOUND),
> any alloc_workqueue() caller that doesn’t explicitly specify WQ_UNBOUND
> must now use WQ_PERCPU.
> 
> Once migration is complete, WQ_UNBOUND can be removed and unbound will
> become the implicit default.
> 
> Suggested-by: Tejun Heo <tj@kernel.org>
> Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
> Link: https://urldefense.com/v3/__https://lore.kernel.org/all/20250221112003.1dSuoGyc@linutronix.de/__;!!N30Cs7Jr!XKe1YoeqdRNkCRNrXBDpd3dyHXefzYb-3Cpd2WmhtEb9vvp1pYw6DF7umzTJlLkLvSdeRGYOrAZAgyS9L-15V9i67-I_-g$
> ---
>  drivers/net/wireless/silabs/wfx/main.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/net/wireless/silabs/wfx/main.c b/drivers/net/wireless/silabs/wfx/main.c
> index a61128debbad..dda36e41eed1 100644
> --- a/drivers/net/wireless/silabs/wfx/main.c
> +++ b/drivers/net/wireless/silabs/wfx/main.c
> @@ -364,7 +364,7 @@ int wfx_probe(struct wfx_dev *wdev)
>         wdev->pdata.gpio_wakeup = NULL;
>         wdev->poll_irq = true;
> 
> -       wdev->bh_wq = alloc_workqueue("wfx_bh_wq", WQ_HIGHPRI, 0);
> +       wdev->bh_wq = alloc_workqueue("wfx_bh_wq", WQ_HIGHPRI | WQ_PERCPU, 0);
>         if (!wdev->bh_wq)
>                 return -ENOMEM;
> 
> --
> 2.51.1
> 
> 

Reviewed-by: Jérôme Pouiller <jerome.pouiller@silabs.com>

BTW, this workqueue has changed multiple times. I though it was already
unbound (and I believe it should).

I also think the HIGHPRI is not required (the device perform better
with HIGHPRI, but this is only because we steal the CPU of the other
tasks).

(Anyway, these changes are out of the scope of your PR).


-- 
Jérôme Pouiller

Re: [PATCH] wifi: wfx: add WQ_PERCPU to alloc_workqueue users

Posted by Sebastian Andrzej Siewior 2 months, 3 weeks ago

On 2025-11-13 17:57:40 [+0100], Jérôme Pouiller wrote:
> > -       wdev->bh_wq = alloc_workqueue("wfx_bh_wq", WQ_HIGHPRI, 0);
> > +       wdev->bh_wq = alloc_workqueue("wfx_bh_wq", WQ_HIGHPRI | WQ_PERCPU, 0);
…
> BTW, this workqueue has changed multiple times. I though it was already
> unbound (and I believe it should).

It is not unbound. It can be enqueued on any CPU but the work item/
worker not migrate to another CPU should there be resources available)
compared to the CPU performing the enqueue.

> I also think the HIGHPRI is not required (the device perform better
> with HIGHPRI, but this is only because we steal the CPU of the other
> tasks).

HIGHPRI sets the nice level as per nice(2).
I did look a bit and one of the users is a threaded interrupt as of
wfx_spi_irq_handler(). This is already a thread with SCHED_FIFO 50.
You can sleep here or do anything else that you could in a workqueue.
This thread will be preferred to other threads in the system which run
usually at SCHED_OTHER including the worker.

Not sure if the requirement for the additional offload is to be able to
serve another interrupt before the worker completes.

Sebastian