Re: [PATCH v2] drm/msm/dp: do not complete dp_aux_cmd_fifo_tx() if irq is not for aux transfer

[Kuogee Hsieh]

On 12/15/2022 1:15 PM, Dmitry Baryshkov wrote:
> On 15/12/2022 22:10, Stephen Boyd wrote:
> > Quoting Dmitry Baryshkov (2022-12-15 10:46:42)
> > > On 15/12/2022 20:32, Kuogee Hsieh wrote:
> > > >        if (!aux->cmd_busy)
> > > >                return;
> > > >
> > > >        if (aux->native)
> > > > -             dp_aux_native_handler(aux, isr);
> > > > +             ret = dp_aux_native_handler(aux, isr);
> > > >        else
> > > > -             dp_aux_i2c_handler(aux, isr);
> > > > +             ret = dp_aux_i2c_handler(aux, isr);
> > > >
> > > > -     complete(&aux->comp);
> > > > +     if (ret == IRQ_HANDLED)
> > > > +             complete(&aux->comp);
> > >
> > > Can you just move the complete() into the individual handling 
> > > functions?
> > > Then you won't have to return the error code from dp_aux_*_handler() at
> > > all. You can check `isr' in that function and call complete if there 
> > > was
> > > any error.
> >
> > I'd prefer we apply my patch and pass the irqreturn_t variable to the
> > caller of this function so spurious irqs are shutdown. Should I send it
> > as a proper patch?
>
> I'm for handling the spurious IRQs in a proper way. However I believe 
> that it's not related to the issue Kuogee is trying to fix.
>
> Thus I think we should have two separate patches: one fixing the EDID 
> corruption issue (for which the proper fix is !isr check, IIUC) and 
> the irqreturn_t. And for the irqreturn_t it might be beneficial to 
> move complete() call to the dp_aux_foo_handler(). Or might be not. 
> That would depend on the patch itself.
ok, I will split this patch into two.