STOP THE MADNESS

[Rob Herring]

On Tue, Nov 26, 2013 at 4:09 PM, Linus Walleij <linus.walleij at linaro.org> wrote:
> On Tue, Nov 26, 2013 at 6:29 PM, Rob Herring <robherring2 at gmail.com> wrote:
>
>> IMO, the whole commit causing this problem should be reverted for
>> additional reasons. Irq generic chp and irqdomains are orthogonal
>> features/infrastructure and should not be mutually exclusive. We need
>> to fix irq generic chip.
>
> This cannot be easily reverted due to a lot of code and fixes building
> on top of it, and even if it could I guess it will also break device tree
> use on these: git grep pl061 arch/arm/boot/dts
>
> As DT mandates that the irqdomain be used, or it cannot translate
> the IRQs as before this patch it uses an hard-coded base from
> platform data. This is one of those odd drivers that don't need
> one line of DT-specific code in it to work with device tree, yet
> it works.

Correct, interrupts did not work with DT before this. It is needed
functionality, but it is how it was implemented that I have issue
with.

> I guess what I have to do is take out the very board Russell has
> and that I happen to have also and just get to the bottom of this
> bug and fix it and test it.
>
> A pointer to an "ideal" GPIO chip + irq_chip driver would be highly
> appreciated ...

I'm not sure if one exists. Shouldn't be too hard to find if anything
uses generic chip and irq domains together. I had done some work to
address generic chip issues with irq domains and pl061 support[1], but
Grant was never really happy with it and I never got back to looking
at it. I think the conclusion was we don't want to add irqdomains into
generic chip as my series did, but we need to at least make them
coexist.

Rob

[1]  git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux.git
irqdomain-for-grant