[PATCH v2] iio: temperature: tsys02d: Fix state desynchronization on resolution write error

[PATCH v2] iio: temperature: tsys02d: Fix state desynchronization on resolution write error

[Salah Triki]

In tsys02d_write_raw(), dev_data->res_index is updated before checking
the return value of ms_sensors_write_resolution(). If the hardware I2C
write operation fails, dev_data->res_index remains updated with the
new index despite the physical sensor remaining in its previous state.

This leads to a state desynchronization where subsequent reads via
tsys02d_read_raw() return an incorrect sampling frequency.

Fix this by handling the error path first, updating dev_data->res_index
only on success, and using guard(mutex) to clean up locking.

This was found through manual code review.

Fixes: 53bf4d067d51 ("Add tsys02d meas-spec driver support")
Signed-off-by: Salah Triki <salah.triki@gmail.com>
---
v1 -> v2:
 - Use error-first pattern (if (ret) return ret) instead of if (!ret).
 - Use guard(mutex) for clean locking and simplified return path.
 - Add note stating the issue was found through manual code review.
 - Add proper blank lines.

 drivers/iio/temperature/tsys02d.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/drivers/iio/temperature/tsys02d.c b/drivers/iio/temperature/tsys02d.c
index 3ef72347456e..20f67134affd 100644
--- a/drivers/iio/temperature/tsys02d.c
+++ b/drivers/iio/temperature/tsys02d.c
@@ -72,12 +72,14 @@ static int tsys02d_write_raw(struct iio_dev *indio_dev,
 				break;
 		if (i < 0)
 			return -EINVAL;
-		mutex_lock(&dev_data->lock);
-		dev_data->res_index = i;
+
+		guard(mutex)(&dev_data->lock);
+
 		ret = ms_sensors_write_resolution(dev_data, i);
-		mutex_unlock(&dev_data->lock);
+		if (ret)
+			return ret;
 
-		return ret;
+		dev_data->res_index = i;
 	default:
 		return -EINVAL;
 	}
-- 
2.43.0

Re: [PATCH v2] iio: temperature: tsys02d: Fix state desynchronization on resolution write error

[Joshua Crofts]

On Thu, 17 Sep 2026 10:16:08 +0100
Salah Triki <salah.triki@gmail.com> wrote:

> In tsys02d_write_raw(), dev_data->res_index is updated before checking
> the return value of ms_sensors_write_resolution(). If the hardware I2C
> write operation fails, dev_data->res_index remains updated with the
> new index despite the physical sensor remaining in its previous state.
> 
> This leads to a state desynchronization where subsequent reads via
> tsys02d_read_raw() return an incorrect sampling frequency.
> 
> Fix this by handling the error path first, updating dev_data->res_index
> only on success, and using guard(mutex) to clean up locking.
> 
> This was found through manual code review.
> 
> Fixes: 53bf4d067d51 ("Add tsys02d meas-spec driver support")
> Signed-off-by: Salah Triki <salah.triki@gmail.com>
> ---
> v1 -> v2:
>  - Use error-first pattern (if (ret) return ret) instead of if (!ret).
>  - Use guard(mutex) for clean locking and simplified return path.
>  - Add note stating the issue was found through manual code review.
>  - Add proper blank lines.
> 
>  drivers/iio/temperature/tsys02d.c | 10 ++++++----
>  1 file changed, 6 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/iio/temperature/tsys02d.c b/drivers/iio/temperature/tsys02d.c
> index 3ef72347456e..20f67134affd 100644
> --- a/drivers/iio/temperature/tsys02d.c
> +++ b/drivers/iio/temperature/tsys02d.c
> @@ -72,12 +72,14 @@ static int tsys02d_write_raw(struct iio_dev *indio_dev,
>  				break;
>  		if (i < 0)
>  			return -EINVAL;
> -		mutex_lock(&dev_data->lock);
> -		dev_data->res_index = i;
> +
> +		guard(mutex)(&dev_data->lock);

Please wrap the case in curly brackets to define the scope for guard().

Additionally, add a <linux/cleanup.h> include. I recently removed kernel.h
from this driver and cleaned the headers, so I'd recommend basing this on
the iio/togreg branch to get the changes if you haven't yet.

https://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio.git/commit/?id=6b5856d16c2e9213fd7eed8a4121aa3ad41081b4

>  		ret = ms_sensors_write_resolution(dev_data, i);
> -		mutex_unlock(&dev_data->lock);
> +		if (ret)
> +			return ret;
>  
> -		return ret;
> +		dev_data->res_index = i;

You're missing a return, this will just fall through to the default
case below.

>  	default:
>  		return -EINVAL;
>  	}



-- 
Kind regards,
Joshua Crofts

Re: [PATCH v2] iio: temperature: tsys02d: Fix state desynchronization on resolution write error

[Andy Shevchenko]

On Thu, Sep 17, 2026 at 01:16:38PM +0200, Joshua Crofts wrote:
> On Thu, 17 Sep 2026 10:16:08 +0100
> Salah Triki <salah.triki@gmail.com> wrote:

...

> > -		return ret;
> > +		dev_data->res_index = i;
> 
> You're missing a return, this will just fall through to the default
> case below.
> 
> >  	default:
> >  		return -EINVAL;

...and having a such without explicit annotation will lead to a compiler
warning.

-- 
With Best Regards,
Andy Shevchenko