Re: [PATCH 02/10] iio: gyro: hid-sensor-gyro-3d: use local struct device

[Andy Shevchenko]

On Mon, Jul 20, 2026 at 10:20:28PM +0530, Sanjay Chitroda wrote:
> On 20 July 2026 5:24:33 pm IST, Andy Shevchenko <andriy.shevchenko@xxxxxxxxx> wrote:
> >On Mon, Jul 20, 2026 at 03:10:14PM +0530, Sanjay Chitroda via B4 Relay wrote:

...

> >> -		dev_err(&pdev->dev, "failed to setup common attributes\n");
> >> +		dev_err(dev, "failed to setup common attributes\n");
> >>  		return ret;
> >
> >Nope, First one (to eliminate a lot of churn) should be to convert this and
> >similar to use
> >
> >		return dev_err_probe(...);
> 
> Thank you for your input.
> 
> I'm planning to have a incremental series: first introduce a local "dev"
> pointer, then convert the resource management to "devm_*", and finally
> replace the remaining error reporting with "dev_err_probe()".
> 
> This keeps each series focused and avoids introducing intermediate patterns such as:
> 
>    ret = dev_err_probe(dev, ret, ...);
>    goto error;
> 
> Once the cleanup paths are removed, those sites naturally become:
> 
>    return dev_err_probe(dev, ret, ...);
> 
> Does this approach sound reasonable?


No. First you should drop the dead code for sure (see above).
Second, try to convert to dev_err_probe() as many drivers as you can without an
intermediate change. If the driver requires an intermediate thing like the above
it means it requires bigger refactoring. So, you should not have intermediate
churn. With your current approach there are drivers that will have it, and for
them you need to have different strategy (see my proposal in the above lines).

-- 
With Best Regards,
Andy Shevchenko