[PATCH 1/9] workqueue: devres: Add device-managed allocate workqueue

[Andy Shevchenko]

On Mon, Feb 23, 2026 at 12:43:42PM +0100, Krzysztof Kozlowski wrote:
> On 23/02/2026 12:34, Andy Shevchenko wrote:
> > On Mon, Feb 23, 2026 at 11:18:41AM +0100, Krzysztof Kozlowski wrote:
> >> On 23/02/2026 09:56, Andy Shevchenko wrote:
> >>> On Mon, Feb 23, 2026 at 08:27:29AM +0100, Krzysztof Kozlowski wrote:

...

> >>>> +	ptr = devres_alloc(devm_destroy_workqueue, sizeof(*ptr), GFP_KERNEL);
> >>>> +	if (!ptr)
> >>>> +		return NULL;
> >>>> +
> >>>> +	va_start(args, max_active);
> >>>> +	wq = alloc_workqueue(fmt, flags, max_active, args);
> >>>> +	va_end(args);
> >>>> +	if (wq) {
> >>>> +		*ptr = wq;
> >>>> +		devres_add(dev, ptr);
> >>>> +	} else {
> >>>> +		devres_free(ptr);
> >>>> +	}
> >>>
> >>> Why not using devm_add_action_or_reset()?
> >>
> >> Where? Here? How the code would be simpler, exactly?
> > 
> > static void devm_workqueue_release(void *wq)
> > {
> > 	destroy_workqueue(wq);
> > }
> > ...
> > {
> > 	...
> > 
> > 	va_start(args, max_active);
> > 	wq = alloc_workqueue(fmt, flags, max_active, args);
> > 	va_end(args);
> > 	if (!wq)
> > 		return NULL; // or ERR_PTR(-ENOMEM) on your choice
> > 
> > 	ret = devm_add_action_or_reset(dev, devm_workqueue_release, wq);
> > 	if (ret)
> > 		return NULL; // ERR_PTR(ret) on your choice
> > 
> > 	return wq;
> > }
> > 
> > Compare to yours :-)
> 
> Ah, so dropping the devres_alloc()? Yeah, that would be simpler.

Yep! The side effect is dropping that ugly casting in _release() along with
unused dev.

-- 
With Best Regards,
Andy Shevchenko