On Fri, Mar 27, 2026 at 07:05:04PM +0000, Antony Kurniawan Soemardi wrote:
> On 3/27/2026 2:04 AM, Andy Shevchenko wrote:
> > On Thu, Mar 26, 2026 at 12:00:52PM +0000, Antony Kurniawan Soemardi wrote:
> > > On 3/26/2026 5:18 PM, Andy Shevchenko wrote:
...
> > > > > + if (!ch) {
> > > > > + dev_err(adc->dev, "no such channel %lu\n",
> > > > > chan->address);
> > > > > + return -EINVAL;
> > > > > + }
> > > >
> > > > Isn't it a dead code? Also poisoning dmesg with this recurrent message
> > > > is
> > > > not good idea to begin with (the user space will have a door to flood
> > > > it,
> > > > which might be considered as an assistance to hackers to clear immediate
> > > > logs after a successful attack).
> > >
> > > Good point about the successful attack hint! I was copying the existing
> > > code from pm8xxx_read_raw. Do you think those checks are unnecessary for
> > > pm8xxx_read_raw as well?
> >
> > Yes, I think they are not as the returned code should be enough to identify
> > the problem. (For no such channel I would rather see -ENOENT, but we can't
> > simply replace that in the existing code as it's part of ABI.)
>
> Just to re-clarify, do you mean for both pm8xxx_read_label &
> pm8xxx_read_raw:
> 1. if the check fails, it should only return -EINVAL without any
> logging; or
> 2. remove the checks because there's no way it's not found?
The first one. And yeah, -EINVAL in the both cases for the sake of consistency.
--
With Best Regards,
Andy Shevchenko
