Re: [PATCH] gpio: bcm-kona: lock IRQs when starting them

[Grygorii Strashko <grygorii.strashko@ti.com>]

On 11/21/2013 11:50 AM, Javier Martinez Canillas wrote:
> Hi,
>
> On Thu, Nov 21, 2013 at 10:04 AM, Grygorii Strashko
> <grygorii.strashko@ti.com> wrote:
>> On 11/21/2013 03:21 AM, Markus Mayer wrote:
>>>
>>> Fixing Christian's e-mail address.
>>>
>>> On 20 November 2013 17:13, Markus Mayer <markus.mayer@linaro.org> wrote:
>>>>
>>>> On 19 November 2013 05:17, Linus Walleij <linus.walleij@linaro.org>
>>>> wrote:
>>>>>
>>>>> This uses the new API for tagging GPIO lines as in use by
>>>>> IRQs. This enforces a few semantic checks on how the underlying
>>>>> GPIO line is used.
>>>>>
>>>>> Cc: Markus Mayer <markus.mayer@linaro.org>
>>>>> Cc: Christian Daudt <csd@broadcom.com>
>>>>> Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
>>>>> ---
>>>>>    drivers/gpio/gpio-bcm-kona.c | 20 ++++++++++++++++++++
>>>>>    1 file changed, 20 insertions(+)
>>>>>
>>>>> diff --git a/drivers/gpio/gpio-bcm-kona.c b/drivers/gpio/gpio-bcm-kona.c
>>>>> index 72c927dc3be1..dee6d03a7959 100644
>>>>> --- a/drivers/gpio/gpio-bcm-kona.c
>>>>> +++ b/drivers/gpio/gpio-bcm-kona.c
>>>>> @@ -449,12 +449,32 @@ static void bcm_kona_gpio_irq_handler(unsigned int
>>>>> irq, struct irq_desc *desc)
>>>>>           chained_irq_exit(chip, desc);
>>>>>    }
>>>>>
>>>>> +static unsigned int bcm_kona_gpio_irq_startup(struct irq_data *d)
>>>>> +{
>>>>> +       struct bcm_kona_gpio *kona_gpio = irq_data_get_irq_chip_data(d);
>>>>> +
>>>>> +       if (gpio_lock_as_irq(&kona_gpio->gpio_chip, d->hwirq))
>>>>> +               dev_err(kona_gpio->gpio_chip.dev,
>>>>> +                       "unable to lock HW IRQ %lu for IRQ\n",
>>>>> +                       d->hwirq);
>>>>> +       return 0;
>>>>> +}
>>>>> +
>>>>> +static void bcm_kona_gpio_irq_shutdown(struct irq_data *d)
>>>>> +{
>>>>> +       struct bcm_kona_gpio *kona_gpio = irq_data_get_irq_chip_data(d);
>>>>> +
>>>>> +       gpio_unlock_as_irq(&kona_gpio->gpio_chip, d->hwirq);
>>>>> +}
>>>>> +
>>>>>    static struct irq_chip bcm_gpio_irq_chip = {
>>>>>           .name = "bcm-kona-gpio",
>>>>>           .irq_ack = bcm_kona_gpio_irq_ack,
>>>>>           .irq_mask = bcm_kona_gpio_irq_mask,
>>>>>           .irq_unmask = bcm_kona_gpio_irq_unmask,
>>>>>           .irq_set_type = bcm_kona_gpio_irq_set_type,
>>>>> +       .irq_startup = bcm_kona_gpio_irq_startup,
>>>>> +       .irq_shutdown = bcm_kona_gpio_irq_shutdown,
>>>>>    };
>>>>>
>>>>>    static struct __initconst of_device_id bcm_kona_gpio_of_match[] = {
>>>>> --
>>>>> 1.8.3.1
>>>>>
>>>>
>>>> I don't know yet why, but as soon as I apply that patch, the cd-gpio
>>>> interrupt doesn't seem to get triggered anymore.
>>>>
>>>> $ grep bcm-kona-gpio /proc/interrupts
>>>> 270:          0  bcm-kona-gpio  14  3f1b0000.sdio cd
>>>>
>>>> The counter remains at 0. If I comment-out the irq_startup &
>>>> irq_shutdown assignments to temporarily disable the new code, the
>>>> interrupt counter increases again when I insert/remove an SD card.
>>
>>
>> Possibly, you have the same problem as I explained here
>> "[PATCH] gpio: lynxpoint: lock IRQs when starting them"
>> http://comments.gmane.org/gmane.linux.kernel.gpio/1163
>>
>
> How about locking the GPIO as IRQ on the .irq_set_type function
> handler instead of adding a .irq_startup?
>
> That's how we did on gpio-omap driver and it works pretty well.

As for me - .irq_startup/.irq_shutdown are fit the best to do gpio lock,
but they have to be implemented accordingly.
Because .irq_startup()/.irq_shutdown() are called exactly when GPIO is 
requested/freed as opposed to .irq_set_type() which may be called any 
time and many times.

Looks like, It will be enough to unmask/mask IRQ from 
bcm_kona_gpio_irq_startup/shutdown() for this driver.


Regards,
- grygorii