Re: [PATCH] iio: mlx90614: fix missing GPIO direction return value checks

[Crt Mori]

On Tue, 5 May 2026 at 09:31, Salah Triki <salah.triki@xxxxxxxxx> wrote:
>
> On Tue, Apr 28, 2026 at 06:33:05PM +0200, Crt Mori wrote:
> > On Tue, 28 Apr 2026 at 18:04, Jonathan Cameron <jic23@xxxxxxxxxx> wrote:
> > >
> > > On Tue, 28 Apr 2026 09:38:47 +0100
> > > Salah Triki <salah.triki@xxxxxxxxx> wrote:
> > >
> > > > On Tue, Apr 28, 2026 at 11:00:04AM +0300, Andy Shevchenko wrote:
> > > > > On Mon, Apr 27, 2026 at 10:58:00PM +0100, Salah Triki wrote:
> > > > > > The functions gpiod_direction_output() and gpiod_direction_input() can
> > > > > > fail, but their return values were previously ignored.
> > > > > >
> > > > > > If an error occurs during the GPIO configuration, the function should
> > > > > > abort the wake-up sequence and return the error code. More importantly,
> > > > > > failing to check these values could lead to the I2C bus remaining
> > > > > > locked if an error occurs after i2c_lock_bus() is called.
> > > > > >
> > > > > > Add return value checks and ensure the I2C bus is properly unlocked
> > > > > > via a goto label in case of failure.
> > > > >
> > > > > ...
> > > > >
> > > > > > - gpiod_direction_output(data->wakeup_gpio, 0);
> > > > > > +
> > > > > > + ret = gpiod_direction_output(data->wakeup_gpio, 0);
> > > > > > + if (ret)
> > > > > > +         goto out_unlock;
> > > > > > +
> > > > > >   msleep(chip_info->wakeup_delay_ms);
> > > > > > - gpiod_direction_input(data->wakeup_gpio);
> > > > > > +
> > > > > > + ret = gpiod_direction_input(data->wakeup_gpio);
> > > > > > + if (ret)
> > > > > > +         goto out_unlock;
> > > > >
> > > > > While technically it sounds correct, the potential problem here is that you may
> > > > > fail this in case CONFIG_GPIOLIB=n. Is this GPIO optional?
> > > > > What may happen if GPIO is not optional, but for some reason setting it fails?
> > > > >
> > > > > TL;DR:
> > > > > I am not sure about this patch. At least I'm not comfortable to take it without
> > > > > testing on real HW.
> > > > >
> > > > > --
> > > > > With Best Regards,
> > > > > Andy Shevchenko
> > > > >
> > > > >
> > > >
> > > > Thank you for your feedback. Since I don't have the physical hardware to
> > > > perform the necessary tests and ensure there are no regressions, I agree
> > > > that it is safer to drop this patch. I don't want to risk breaking the
> > > > driver for a minor cleanup.
> > > >
> > > Maybe Crt has the hardware to test. For now I'll assume this is not
> > > going anywhere wrt to tracking in patchwork.
> > >
> > I will dig the hardware to test as it was off my desk for some time
> > now. It seems like backwards compatible, but I am wondering why would
> > we want to be informed of this failing? The problem is that sensor
> > uses wakeup line (same as for i2c) to receive wakeup pulse, so in
> > theory this is already a workaround. @Salah did you see any issues in
> > the wild?
> >
>
> Thanks for digging out the hardware to test this!
>
> To answer your question: I haven't observed this failing in the wild on
> this specific sensor. The patch was initially motivated by a static
> analysis and general robustness cleanup, ensuring that we don't ignore
> errors from gpiod_* functions (especially in the context of keeping the
> I2C bus locked).
>
> However, as Andy pointed out, turning this into a fatal error might not
> be desirable if the GPIO is optional or if the failure is transient on some
> platforms. The goal was simply to improve visibility if the GPIO
> configuration fails.

This GPIO in this case is basically used for generating wakeup pulse
(so when sensor is sleeping) and I doubt there is usefulness of the
GPIO error failure for this case from user perspective - there is a
use case when debugging.

At the moment I am fairly busy. I have found the sensor, but I need to
dig out the some board to port the latest kernel to.  I hope I get a
reminder from Jonathan as the window gets closed, but I also have
trouble

>
> Best Regards,
> --
> Salah Triki