Re: [PATCH v4 1/3] iio: adc: at91_adc: fix NULL pointer dereference on trigger remove
From: Jonathan Cameron
Date: Sat Nov 21 2026 - 19:21:31 EST
On Tue, 17 Nov 2026 18:45:29 +0200
Andy Shevchenko <andriy.shevchenko@xxxxxxxxx> wrote:
> On Tue, Nov 17, 2026 at 06:44:37PM +0200, Andy Shevchenko wrote:
> > On Tue, Nov 17, 2026 at 04:05:08PM +0200, Abdelnasser Hussein wrote:
>
> ...
>
> > > for (i = 0; i < st->caps->trigger_number; i++) {
> > > + if (!st->trig[i])
> > > + continue;
> > > iio_trigger_unregister(st->trig[i]);
> > > iio_trigger_free(st->trig[i]);
> > > }
> >
> > Yes, but no. Please, rather fix the above mentioned IIO trigger APIs to be
> > NULL-aware.
>
> Actually the iio_trigger_free() is already NULL-aware, so make it only for
> iio_trigger_unregister().
I'd rather we didn't. Any code that is unregistering a trigger
that it didn't register is probably really very very buggy.
I'd like them to crash or at least guarantee that sashiko
flags it. We could add a warn if it is called on NULL I suppose
but that is ugly.
I'd rather deal with this by fixing at source. That doesn't mean
papering over it except as a potential easy to backport change
which is the intent of this one. It think the description should
call that out though. Something like
"Make a minimal fix to simplify backporting. Moving over to
device managed cleanup will shortly remove this code."
Jonathan
>
>