[PATCH] iio: accel: st_accel: Fix invalid mount_matrix on devices without ACPI _ONT method

[PATCH] iio: accel: st_accel: Fix invalid mount_matrix on devices without ACPI _ONT method

[Hans de Goede]

When apply_acpi_orientation() fails, st_accel_common_probe() will fall back
to iio_read_mount_matrix(), which checks for a mount-matrix device property
and if that is not set falls back to the identity matrix.

But when a sensor has no ACPI companion fwnode, or when the ACPI fwnode
does not have a "_ONT" method apply_acpi_orientation() was returning 0,
causing iio_read_mount_matrix() to never get called resulting in an
invalid mount_matrix:

[root@fedora ~]# cat /sys/bus/iio/devices/iio\:device0/mount_matrix
(null), (null), (null); (null), (null), (null); (null), (null), (null)

Fix this by making apply_acpi_orientation() always return an error when
it did not set the mount_matrix.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
---
 drivers/iio/accel/st_accel_core.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/iio/accel/st_accel_core.c b/drivers/iio/accel/st_accel_core.c
index 1628b177d0ed..f7b823ebc96b 100644
--- a/drivers/iio/accel/st_accel_core.c
+++ b/drivers/iio/accel/st_accel_core.c
@@ -1291,12 +1291,12 @@ static int apply_acpi_orientation(struct iio_dev *indio_dev)
 
 	adev = ACPI_COMPANION(indio_dev->dev.parent);
 	if (!adev)
-		return 0;
+		return -ENXIO;
 
 	/* Read _ONT data, which should be a package of 6 integers. */
 	status = acpi_evaluate_object(adev->handle, "_ONT", NULL, &buffer);
 	if (status == AE_NOT_FOUND) {
-		return 0;
+		return -ENXIO;
 	} else if (ACPI_FAILURE(status)) {
 		dev_warn(&indio_dev->dev, "failed to execute _ONT: %d\n",
 			 status);
-- 
2.39.2

Re: [PATCH] iio: accel: st_accel: Fix invalid mount_matrix on devices without ACPI _ONT method

[Marius Hoch]

On 16/04/2023 23:24, Hans de Goede wrote:
> When apply_acpi_orientation() fails, st_accel_common_probe() will fall back
> to iio_read_mount_matrix(), which checks for a mount-matrix device property
> and if that is not set falls back to the identity matrix.
>
> But when a sensor has no ACPI companion fwnode, or when the ACPI fwnode
> does not have a "_ONT" method apply_acpi_orientation() was returning 0,
> causing iio_read_mount_matrix() to never get called resulting in an
> invalid mount_matrix:
>
> [root@fedora ~]# cat /sys/bus/iio/devices/iio\:device0/mount_matrix
> (null), (null), (null); (null), (null), (null); (null), (null), (null)
>
> Fix this by making apply_acpi_orientation() always return an error when
> it did not set the mount_matrix.
>
> Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Tested, with the LSM303D patches applied, on a Lenovo Yoga Tablet 2 1051-F.

Tested-by: Marius Hoch <mail@mariushoch.de>


> ---
>   drivers/iio/accel/st_accel_core.c | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/iio/accel/st_accel_core.c b/drivers/iio/accel/st_accel_core.c
> index 1628b177d0ed..f7b823ebc96b 100644
> --- a/drivers/iio/accel/st_accel_core.c
> +++ b/drivers/iio/accel/st_accel_core.c
> @@ -1291,12 +1291,12 @@ static int apply_acpi_orientation(struct iio_dev *indio_dev)
>   
>   	adev = ACPI_COMPANION(indio_dev->dev.parent);
>   	if (!adev)
> -		return 0;
> +		return -ENXIO;
>   
>   	/* Read _ONT data, which should be a package of 6 integers. */
>   	status = acpi_evaluate_object(adev->handle, "_ONT", NULL, &buffer);
>   	if (status == AE_NOT_FOUND) {
> -		return 0;
> +		return -ENXIO;
>   	} else if (ACPI_FAILURE(status)) {
>   		dev_warn(&indio_dev->dev, "failed to execute _ONT: %d\n",
>   			 status);

Re: [PATCH] iio: accel: st_accel: Fix invalid mount_matrix on devices without ACPI _ONT method

[Jonathan Cameron]

On Tue, 18 Apr 2023 14:32:42 +0200
Marius Hoch <mail@mariushoch.de> wrote:

> On 16/04/2023 23:24, Hans de Goede wrote:
> > When apply_acpi_orientation() fails, st_accel_common_probe() will fall back
> > to iio_read_mount_matrix(), which checks for a mount-matrix device property
> > and if that is not set falls back to the identity matrix.
> >
> > But when a sensor has no ACPI companion fwnode, or when the ACPI fwnode
> > does not have a "_ONT" method apply_acpi_orientation() was returning 0,
> > causing iio_read_mount_matrix() to never get called resulting in an
> > invalid mount_matrix:
> >
> > [root@fedora ~]# cat /sys/bus/iio/devices/iio\:device0/mount_matrix
> > (null), (null), (null); (null), (null), (null); (null), (null), (null)
> >
> > Fix this by making apply_acpi_orientation() always return an error when
> > it did not set the mount_matrix.
> >
> > Signed-off-by: Hans de Goede <hdegoede@redhat.com>  
> Tested, with the LSM303D patches applied, on a Lenovo Yoga Tablet 2 1051-F.
> 
> Tested-by: Marius Hoch <mail@mariushoch.de>

Needs a fixes tag:. I think...

Fixes: 3d8ad94bb175 ("iio: accel: st_sensors: Support generic mounting matrix")

actually introduced the issue rather than the patch that added the _ONT support

Linus, that's one of yours.  Could you also take a glance at this.

Seems 'obviously' correct but I've been wrong before :)
> 
> 
> > ---
> >   drivers/iio/accel/st_accel_core.c | 4 ++--
> >   1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/iio/accel/st_accel_core.c b/drivers/iio/accel/st_accel_core.c
> > index 1628b177d0ed..f7b823ebc96b 100644
> > --- a/drivers/iio/accel/st_accel_core.c
> > +++ b/drivers/iio/accel/st_accel_core.c
> > @@ -1291,12 +1291,12 @@ static int apply_acpi_orientation(struct iio_dev *indio_dev)
> >   
> >   	adev = ACPI_COMPANION(indio_dev->dev.parent);
> >   	if (!adev)
> > -		return 0;
> > +		return -ENXIO;
> >   
> >   	/* Read _ONT data, which should be a package of 6 integers. */
> >   	status = acpi_evaluate_object(adev->handle, "_ONT", NULL, &buffer);
> >   	if (status == AE_NOT_FOUND) {
> > -		return 0;
> > +		return -ENXIO;
> >   	} else if (ACPI_FAILURE(status)) {
> >   		dev_warn(&indio_dev->dev, "failed to execute _ONT: %d\n",
> >   			 status);  
> 

Re: [PATCH] iio: accel: st_accel: Fix invalid mount_matrix on devices without ACPI _ONT method

[Linus Walleij]

On Sun, Apr 23, 2023 at 12:33 PM Jonathan Cameron <jic23@kernel.org> wrote:

> Linus, that's one of yours.  Could you also take a glance at this.

Heh it should really be looked after by Denis Ciocca or Mario Tesi I think.
I don't know if ST MEMS has an official Linux contact point, but I would
like to nudge them to get one :)

However it looks good to me.
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>

Yours,
Linus Walleij

Re: [PATCH] iio: accel: st_accel: Fix invalid mount_matrix on devices without ACPI _ONT method

[Jonathan Cameron]

On Mon, 24 Apr 2023 09:59:47 +0200
Linus Walleij <linus.walleij@linaro.org> wrote:

> On Sun, Apr 23, 2023 at 12:33 PM Jonathan Cameron <jic23@kernel.org> wrote:
> 
> > Linus, that's one of yours.  Could you also take a glance at this.  
> 
> Heh it should really be looked after by Denis Ciocca or Mario Tesi I think.
> I don't know if ST MEMS has an official Linux contact point, but I would
> like to nudge them to get one :)

Always good to get input from Denis or Mario.
> 
> However it looks good to me.
> Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Meh, I consider this one obvious enough that the current set of eyes
would probably have noticed if it was horribly broken.

Applied to the fixes-togreg branch of iio.git and marked for stable.

There's about a week left before rc1 so time for people to scream if
I'm wrong :)

Jonathan

> 
> Yours,
> Linus Walleij

Re: [PATCH] iio: accel: st_accel: Fix invalid mount_matrix on devices without ACPI _ONT method

[Hans de Goede]

Hi,

On 4/23/23 12:48, Jonathan Cameron wrote:
> On Tue, 18 Apr 2023 14:32:42 +0200
> Marius Hoch <mail@mariushoch.de> wrote:
> 
>> On 16/04/2023 23:24, Hans de Goede wrote:
>>> When apply_acpi_orientation() fails, st_accel_common_probe() will fall back
>>> to iio_read_mount_matrix(), which checks for a mount-matrix device property
>>> and if that is not set falls back to the identity matrix.
>>>
>>> But when a sensor has no ACPI companion fwnode, or when the ACPI fwnode
>>> does not have a "_ONT" method apply_acpi_orientation() was returning 0,
>>> causing iio_read_mount_matrix() to never get called resulting in an
>>> invalid mount_matrix:
>>>
>>> [root@fedora ~]# cat /sys/bus/iio/devices/iio\:device0/mount_matrix
>>> (null), (null), (null); (null), (null), (null); (null), (null), (null)
>>>
>>> Fix this by making apply_acpi_orientation() always return an error when
>>> it did not set the mount_matrix.
>>>
>>> Signed-off-by: Hans de Goede <hdegoede@redhat.com>  
>> Tested, with the LSM303D patches applied, on a Lenovo Yoga Tablet 2 1051-F.
>>
>> Tested-by: Marius Hoch <mail@mariushoch.de>
> 
> Needs a fixes tag:. I think...
> 
> Fixes: 3d8ad94bb175 ("iio: accel: st_sensors: Support generic mounting matrix")

Right, ack for adding that fixes tag.

Regards,

Hans





> 
> actually introduced the issue rather than the patch that added the _ONT support
> 
> Linus, that's one of yours.  Could you also take a glance at this.
> 
> Seems 'obviously' correct but I've been wrong before :)
>>
>>
>>> ---
>>>   drivers/iio/accel/st_accel_core.c | 4 ++--
>>>   1 file changed, 2 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/iio/accel/st_accel_core.c b/drivers/iio/accel/st_accel_core.c
>>> index 1628b177d0ed..f7b823ebc96b 100644
>>> --- a/drivers/iio/accel/st_accel_core.c
>>> +++ b/drivers/iio/accel/st_accel_core.c
>>> @@ -1291,12 +1291,12 @@ static int apply_acpi_orientation(struct iio_dev *indio_dev)
>>>   
>>>   	adev = ACPI_COMPANION(indio_dev->dev.parent);
>>>   	if (!adev)
>>> -		return 0;
>>> +		return -ENXIO;
>>>   
>>>   	/* Read _ONT data, which should be a package of 6 integers. */
>>>   	status = acpi_evaluate_object(adev->handle, "_ONT", NULL, &buffer);
>>>   	if (status == AE_NOT_FOUND) {
>>> -		return 0;
>>> +		return -ENXIO;
>>>   	} else if (ACPI_FAILURE(status)) {
>>>   		dev_warn(&indio_dev->dev, "failed to execute _ONT: %d\n",
>>>   			 status);  
>>
>