Re: [PATCH 01/20] ASoC: Intel: catpt: ipc: Use guard() for mutex & spin locks

[Cezary Rojewski]

On 6/12/2026 6:05 AM, Bui Duc Phuc wrote:
> Hi,
>
> > The catpt_dsp_send_msg_timeout is already part of Mark's tree [1].  At
> > the same time I'm against using scoped_guard() for
> > catpt_dsp_do_send_msg().  The existing code is small and transparent,
> > there is no need to tabify it.
>
> > > @@ -97,10 +96,10 @@ static int catpt_dsp_do_send_msg(struct catpt_dev *cdev,
> > >            (reply && reply->size > ipc->config.outbox_size))
> > >                return -EINVAL;
> > >
> > > -     spin_lock_irqsave(&ipc->lock, flags);
> > > -     catpt_ipc_msg_init(ipc, reply);
> > > -     catpt_dsp_send_tx(cdev, &request);
> > > -     spin_unlock_irqrestore(&ipc->lock, flags);
> > > +     scoped_guard(spinlock_irqsave, &ipc->lock) {
> > > +             catpt_ipc_msg_init(ipc, reply);
> > > +             catpt_dsp_send_tx(cdev, &request);
> > > +     }
> > >
> > >        ret = catpt_wait_msg_completion(cdev, timeout);
>
> So I understand that, for this conversion effort, you would prefer a
> case-by-case approach rather than converting everything to
> guard()/scoped_guard() purely for consistency. Is that correct?
Yeah.  As stated previously, I see no reason to update the catpt-driver, 
simply drop 1/20 and 2/20.  Let's focus on the avs- and the atom- drivers.