Re: [PATCH v2] iio: gyro: adxrs290: Use guard(mutex) in lieu of manual lock+unlock

Re: [PATCH v2] iio: gyro: adxrs290: Use guard(mutex) in lieu of manual lock+unlock

[Guilherme Dias]

Em sex., 24 de abr. de 2026 às 07:39, Jonathan Cameron
<jic23@kernel.org> escreveu:
>
> On Fri, 24 Apr 2026 12:24:07 +0300
> Andy Shevchenko <andriy.shevchenko@intel.com> wrote:
>
> > On Thu, Apr 23, 2026 at 07:37:51PM -0300, Guilherme Dias wrote:
> > > Use guard(mutex) to automatically release the lock on scope exit,
> > > simplifying the error handling path and removing the need for
> > > explicit unlock and goto-based cleanup.
...
> > >     /* exercise a bulk data capture starting from reg DATAX0... */
> > >     ret = spi_write_then_read(st->spi, &tx, sizeof(tx), st->buffer.channels,
> > >                               sizeof(st->buffer.channels));
> > > -   if (ret < 0)
> > > -           goto out_unlock_notify;
> > > -
> > > -   iio_push_to_buffers_with_timestamp(indio_dev, &st->buffer,
> > > +   if (ret >= 0)
> No to this as well. Keep the error out of line.  Which makes the use
> of guard() here tricky.  You could do a do { guard(); } while(0) loop so that
> you can use break with out it being odd looking.
>
> (I don't like breaks out of scoped_guard() because it's non obvious scoped
> guard is actually a loop.

Thanks for the suggestion — just to confirm I understood correctly, do
you mean something like this?

...
 do {

 guard(mutex)(&st->lock);

 /* exercise a bulk data capture starting from reg DATAX0... */
 ret = spi_write_then_read(st->spi, &tx, sizeof(tx), st->buffer.channels,
                          sizeof(st->buffer.channels));
 if (ret < 0)
            break;

 iio_push_to_buffers_with_timestamp(indio_dev, &st->buffer,
                         pf->timestamp);
 } while (0);

 iio_trigger_notify_done(indio_dev->trig);
 return IRQ_HANDLED;
}

Keeping the error out of line again.

--
Kind regards,
Guilherme Dias

Re: [PATCH v2] iio: gyro: adxrs290: Use guard(mutex) in lieu of manual lock+unlock

[Andy Shevchenko]

On Mon, Apr 27, 2026 at 07:04:23PM -0300, Guilherme Dias wrote:
> Em sex., 24 de abr. de 2026 às 07:39, Jonathan Cameron
> <jic23@kernel.org> escreveu:
> > On Fri, 24 Apr 2026 12:24:07 +0300
> > Andy Shevchenko <andriy.shevchenko@intel.com> wrote:
> > > On Thu, Apr 23, 2026 at 07:37:51PM -0300, Guilherme Dias wrote:

...

> > > >     /* exercise a bulk data capture starting from reg DATAX0... */
> > > >     ret = spi_write_then_read(st->spi, &tx, sizeof(tx), st->buffer.channels,
> > > >                               sizeof(st->buffer.channels));
> > > > -   if (ret < 0)
> > > > -           goto out_unlock_notify;

> > > > +   if (ret >= 0)
> > No to this as well. Keep the error out of line.  Which makes the use
> > of guard() here tricky.  You could do a do { guard(); } while(0) loop so that
> > you can use break with out it being odd looking.
> >
> > (I don't like breaks out of scoped_guard() because it's non obvious scoped
> > guard is actually a loop.
> 
> Thanks for the suggestion — just to confirm I understood correctly, do
> you mean something like this?

Yes, with proper indentation made.

-- 
With Best Regards,
Andy Shevchenko