Re: [PATCH] thermal: core: Add WQ_PERCPU to alloc_workqueue users

[Rafael J. Wysocki]

On Thu, May 14, 2026 at 4:03 PM Marco Crivellari
<marco.crivellari@xxxxxxxx> wrote:
>
> On Thu, May 14, 2026 at 3:53 PM Rafael J. Wysocki <rafael@xxxxxxxxxx> wrote:
> > [...]
> > > diff --git a/drivers/thermal/thermal_core.c b/drivers/thermal/thermal_core.c
> > > index 2f4e2dc46b8f..691c5e749c3e 100644
> > > --- a/drivers/thermal/thermal_core.c
> > > +++ b/drivers/thermal/thermal_core.c
> > > @@ -1889,7 +1889,8 @@ static int __init thermal_init(void)
> > >         if (result)
> > >                 goto error;
> > >
> > > -       thermal_wq = alloc_workqueue("thermal_events", WQ_POWER_EFFICIENT, 0);
> > > +       thermal_wq = alloc_workqueue("thermal_events",
> > > +                                    WQ_POWER_EFFICIENT | WQ_PERCPU, 0);
> >
> > No, we actually don't want that here.  See
> >
> > 26fd03effa94 thermal: core: Adjust thermal_wq allocation flags
> >
> > >         if (!thermal_wq) {
> > >                 result = -ENOMEM;
> > >                 goto unregister_netlink;
> > > --
>
> Thanks for sharing the commit, I didn't see it.
> Quoting from the commit log:
>
> > The thermal workqueue doesn't need to be freezable or per-CPU
> > [...]
>
> Currently, alloc_workqueue() without WQ_PERCPU is already per-cpu,
> so drop this flag it doesn't make the workqueue unbound.
>
> If you want it unbound WQ_UNBOUND must be used.
> If it's okay with you, I will send the v2 with WQ_UNBOUND.

Yes, please!