[PATCH 1/5] pintrl: meson: add interrupts to pinctrl data

[Heiner Kallweit]

Am 11.05.2017 um 18:08 schrieb Jerome Brunet:
> On Thu, 2017-05-11 at 16:50 +0200, Linus Walleij wrote:
>> On Sun, May 7, 2017 at 6:34 PM, Heiner Kallweit <hkallweit1 at gmail.com> wrote:
>>
>>> From: Jerome Brunet <jbrunet at baylibre.com>
>>> Add GPIO interrupt information to pinctrl data. Added to the original
>>> version from Jerome was data for Meson GXL.
>>>
>>> Signed-off-by: Jerome Brunet <jbrunet at baylibre.com>
>>> Signed-off-by: Heiner Kallweit <hkallweit1 at gmail.com>
>>
>> So what this does is:
>>
>>> diff --git a/drivers/pinctrl/meson/pinctrl-meson.h
>>> b/drivers/pinctrl/meson/pinctrl-meson.h
>>> index 1aa871d5..890f296f 100644
>>> --- a/drivers/pinctrl/meson/pinctrl-meson.h
>>> +++ b/drivers/pinctrl/meson/pinctrl-meson.h
>>> @@ -81,6 +81,7 @@ enum meson_reg_type {
>>>   * @name:      bank name
>>>   * @first:     first pin of the bank
>>>   * @last:      last pin of the bank
>>> + * @irq:       hwirq base number of the bank
>>>   * @regs:      array of register descriptors
>>>   *
>>>   * A bank represents a set of pins controlled by a contiguous set of
>>> @@ -92,6 +93,8 @@ struct meson_bank {
>>>         const char *name;
>>>         unsigned int first;
>>>         unsigned int last;
>>> +       int irq_first;
>>> +       int irq_last;
>>>         struct meson_reg_desc regs[NUM_REG];
>>>  };
>>
>> ... adds a per-bank parent IRQ.
> 
> Hi Linus,
> 
> It does not add a per-bank parent IRQ. The meson gpio irq is an IP independent
> of the gpio/pinctrl subsystem. In a nutshell, we have 8 interrupts on the GIC
> and we can route the signal of almost any pin of any bank to these parent irqs.
> The fact that there 1 gpio-irq controller and several gpio controller make the
> design a bit tricky
> 
> We already talked about this IP, here is the discussion :https://patchwork.ozlab
> s.org/patch/684208/
> 
> At the time, the problem was that I was creating the mapping within the
> gpio_to_irq callback, which is wrong.
> 
> irq_first, irq_last were introduced because there is no way to have direct
> mapping between the gpio number and the hw interrupt number. (details are in our
> previous discussion). If a more generic solution can be used for this, it would
> be nice :)
> 
> While I think having a irqdomain in the gpio driver and using the
> request_resources callback to create the mapping in the underlying domain might
> be a solution to ressouce allocation problem we had, I think the meson-gpio-irq
> should be implemented has an independent driver because  it is an independent
> device.

This kind of GPIO IRQ multiplexer doesn't really seem to match any use case
covered by the GPIO / IRQ subsystems. Especially the needed dynamic mapping
from the GPIOs to one of the eight GPIO GIC IRQs is tricky.

After Jerome's attempt from last year (using hierarchical IRQ domains) didn't
make it due to concerns of the maintainers, I decided to go another way.
I'm not stating at all that this is the right one ..

To re-use existing stuff as far as possible I go with GPIOLIB_IRQCHIP.
So far I use one irqchip per gpiochip (we have two of them).
I'll check whether I can use just one irqchip for both gpiochips.

> Eventually, this device should be the parent irq controller of the gpio
> controller.

I think this is one of the central questions here, whether this device
can be considered an irq controller.
It just manages routing of IRQs, so I'd tend to say that the gpio controller
is the irq controller with the GIC as parent.

> 
> I not sure that there is a reason for using syscon here, or if it is a good idea
> to have the data of this controller as globals for the gpio driver ... 
> 
> 
>>
>> I am just discussing with Thierry that I would like to see some code
>> in the gpiolib core to deal with this mapping so we don't have to do a
>> whole lot of custom back mapping between parent IRQs and cascaded
>> IRQ in every driver that has a multiple-bank concept.
>>
>> Please contribute to the discission, see thread subject:
>> "[PATCH v2] gpio: Add Tegra186 support"
>>
>> Yours,
>> Linus Walleij
>