Re: [PATCH v2] iio: adc: mt6359: fix unchecked return value in mt6358_read_imp

[Andy Shevchenko <andriy.shevchenko@intel.com>]

On Mon, Apr 27, 2026 at 06:18:56PM +0100, Jonathan Cameron wrote:
> On Mon, 27 Apr 2026 17:57:59 +0100
> Salah Triki <salah.triki@gmail.com> wrote:
> 
> > In mt6358_read_imp(), the return value of regmap_read() is currently
> > ignored. This is problematic because if the bus read fails the variable
> > val_v remains uninitialized.
> > 
> > The function subsequently assigns this uninitialized stack value to
> > *vbat, leading to incorrect measurement results being reported to
> > the IIO subsystem without any error indication.
> > 
> > Update the function to check the return value of regmap_read(). Ensure
> > that mt6358_stop_imp_conv() is still called to clean up the hardware
> > state before returning the error code.

> I'd be more concerned about what it might otherwise report.. Will expose
> something random that was on the stack previously.
> 
> Ok on this one having a fixes tag given the uninitialized, though as I mentioned
> I'd have been tempted to take the approach of just initialising it.
> 
> Even with the code as you have it here a static analyzer may not be able
> to see far enough to tell it is initialized and so throw false positive warnings.

True.

> Still, let's see what Andy thinks before you change anything.

...

> >  	/* Read the params before stopping */
> > -	regmap_read(regmap, reg_adc0 + (cinfo->imp_adc_num << 1), &val_v);
> > +	ret = regmap_read(regmap, reg_adc0 + (cinfo->imp_adc_num << 1), &val_v);

> >  	mt6358_stop_imp_conv(adc_dev);

This still uses unchecked IO.

> > +	if (ret)
> > +		return ret;

So, assume we got it right and error code is 0, but failed to stop conversion.
What would be the consequences?

As I said, the whole driver needs to be re-think of. If you going to check this
you need to add checks into _stop_imp_conv() which may lead to adding checks to
the complementary functions and it will go as a rabbit hole.

I agree with Jonathan that quick fix of initialising it explicitly with
a comment will be better approach taking into account the state of affairs
of this driver as a whole.

-- 
With Best Regards,
Andy Shevchenko