Re: [PATCH net-next v4 3/3] net: dsa: realtek: support reset controller

[Alvin Šipraga]

On Tue, Feb 20, 2024 at 01:30:33PM +0000, Alvin Šipraga wrote:
> On Tue, Feb 20, 2024 at 09:22:44AM -0300, Luiz Angelo Daros de Luca wrote:
> > Hi Alvin,
> > 
> > > On Mon, Feb 19, 2024 at 08:44:42PM -0300, Luiz Angelo Daros de Luca wrote:
> > > > +void rtl83xx_reset_assert(struct realtek_priv *priv)
> > > > +{
> > > > +     int ret;
> > > > +
> > > > +     ret = reset_control_assert(priv->reset_ctl);
> > > > +     if (!ret)
> > > > +             return;
> > >
> > > If priv->reset_ctl is NULL - i.e. if no DT property is specified - then
> > > this will always return early and the GPIO will not be asserted.
> > 
> > I made a mistake. I should be
> > 
> > if (ret) {
> >           dev_warn...
> > }
> > 
> > not returning on error (as you suggested below).
> > 
> > I was sure I was doing just that... I was surprised to see it as it
> > is.  I'll recheck my branch with all the integrated changes. It passed
> > my tests as when reset is missed, it normally does not matter. Thanks
> > for the catch.
> > 
> > >
> > > > +
> > > > +     dev_warn(priv->dev,
> > > > +              "Failed to assert the switch reset control: %pe\n",
> > > > +              ERR_PTR(ret));
> > >
> > > You only log an error if the reset controller assert fails, but not if
> > > the GPIO assert fails. Why the unequal treatment?
> > 
> > Because it does not return a value. There is no way to tell if it failed.
> 
> Ah ok, nevermind that part then.
> 
> BTW, please use gpiod_set_value_cansleep(). With that I think this is good.

OK, actually the original code wasn't doing that, so not crucial for this
change. It can be done in a follow-up.