Re: [PATCH] gpio: lynxpoint: Pass irqchip when adding gpiochip

[Linus Walleij <linus.walleij@xxxxxxxxxx>]

On Wed, Aug 14, 2019 at 11:40 AM Andy Shevchenko
<andriy.shevchenko@xxxxxxxxxxxxxxx> wrote:
> On Wed, Aug 14, 2019 at 10:46:49AM +0200, Linus Walleij wrote:
> > On Mon, Aug 12, 2019 at 12:58 PM Andy Shevchenko
> > <andriy.shevchenko@xxxxxxxxxxxxxxx> wrote:
> > > On Mon, Aug 12, 2019 at 10:13:51AM +0200, Linus Walleij wrote:
> >
> > > > +             girq->num_parents = 1;
> > > > +             girq->parents = devm_kcalloc(&pdev->dev, 1,
> > > > +                                          sizeof(*girq->parents),
> > > > +                                          GFP_KERNEL);
> > > > +             if (!girq->parents)
> > > > +                     return -ENOMEM;
> > >
> > > I understand the point to use kcalloc() for one entry, though I would make
> > > intention more explicitly, i.e. use girq->num_parents in it instead of hard
> > > coded value.
> >
> > That is better, but I have a loose plan to get rid of this
> > and just set parents to a fixed width because all the allocation
> > is annoying.
>
> I see your intentions, though for current state I think the less hard coded
> constants the better. In any case I pushed updated versions to my trees.

Thanks a lot. Yeah we live with the API we have and work from there.

> > If you are sure that every consumer will call .set_type() you can
> > use handle_bad_irq, and that is preferred.
>
> They should do this. Let me prepare the patch for next cycle (v5.5) and I put
> it to my tree after merge window. If we see any complains from linux-next
> testers, we will act accordingly.

Sounds like a plan!

Thanks,
Linus Walleij