[PATCH] gpio-rcar: .irq_set_wake() support to work with suspend_device_irqs()

[PATCH] gpio-rcar: .irq_set_wake() support to work with suspend_device_irqs()

[Yoshihiro Kaneko]

From: Shinya Kuribayashi <shinya.kuribayashi.px@renesas.com>

All IRQs in the kernel get disabled by PM core prior to .suspend_noirq()
step through kernel/irq/pm.c::suspend_device_irqs().

The SYSC hardware block in the R-Car SoCs, on the other hand, uses any
interrupt as wake-up source(es) and requires them to be unmasked during
the system is suspended.

This patch implements .irq_set_wake() which makes __disable_irq() call
a NOP when a GPIO port is claimed as a wake-up source.

Signed-off-by: Shinya Kuribayashi <shinya.kuribayashi.px@renesas.com>
Signed-off-by: Koji Matsuoka <koji.matsuoka.xm@renesas.com>
Signed-off-by: Yoshihiro Kaneko <ykaneko0929@gmail.com>
---

This patch is based on for-next branch of Linus Walleij's gpio tree.

 drivers/gpio/gpio-rcar.c | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)

diff --git a/drivers/gpio/gpio-rcar.c b/drivers/gpio/gpio-rcar.c
index bf6c094..ef71ca8 100644
--- a/drivers/gpio/gpio-rcar.c
+++ b/drivers/gpio/gpio-rcar.c
@@ -38,6 +38,7 @@ struct gpio_rcar_priv {
 	struct gpio_chip gpio_chip;
 	struct irq_chip irq_chip;
 	struct irq_domain *irq_domain;
+	u32 no_suspend; /* do not disable this GPIO port during suspend */
 };
 
 #define IOINTSEL 0x00
@@ -83,6 +84,15 @@ static void gpio_rcar_irq_disable(struct irq_data *d)
 {
 	struct gpio_rcar_priv *p = irq_data_get_irq_chip_data(d);
 
+	if (p->no_suspend & BIT(irqd_to_hwirq(d))) {
+		/*
+		 * This GPIO port is claimed as a wake-up source and
+		 * currently in a no-suspend-requested state.  Don't
+		 * disable this IRQ during suspend.
+		 */
+		return;
+	}
+
 	gpio_rcar_write(p, INTMSK, ~BIT(irqd_to_hwirq(d)));
 }
 
@@ -164,6 +174,20 @@ static int gpio_rcar_irq_set_type(struct irq_data *d, unsigned int type)
 	return 0;
 }
 
+static int gpio_rcar_irq_set_wake(struct irq_data *d, unsigned int on)
+{
+	struct gpio_rcar_priv *p = irq_data_get_irq_chip_data(d);
+	unsigned int hwirq = irqd_to_hwirq(d);
+
+	dev_dbg(&p->pdev->dev, "wake irq = %u %s\n", hwirq, on ? "on" : "off");
+
+	if (on)
+		p->no_suspend |= BIT(hwirq);
+	else
+		p->no_suspend &= ~BIT(hwirq);
+	return 0;
+}
+
 static irqreturn_t gpio_rcar_irq_handler(int irq, void *dev_id)
 {
 	struct gpio_rcar_priv *p = dev_id;
@@ -416,6 +440,7 @@ static int gpio_rcar_probe(struct platform_device *pdev)
 	irq_chip->irq_mask = gpio_rcar_irq_disable;
 	irq_chip->irq_unmask = gpio_rcar_irq_enable;
 	irq_chip->irq_set_type = gpio_rcar_irq_set_type;
+	irq_chip->irq_set_wake = gpio_rcar_irq_set_wake;
 	irq_chip->flags	= IRQCHIP_SKIP_SET_WAKE | IRQCHIP_SET_TYPE_MASKED
 			 | IRQCHIP_MASK_ON_SUSPEND;
 
-- 
1.9.1

Re: [PATCH] gpio-rcar: .irq_set_wake() support to work with suspend_device_irqs()

[Linus Walleij]

On Fri, Nov 7, 2014 at 10:22 AM, Yoshihiro Kaneko <ykaneko0929@gmail.com> wrote:

> From: Shinya Kuribayashi <shinya.kuribayashi.px@renesas.com>
>
> All IRQs in the kernel get disabled by PM core prior to .suspend_noirq()
> step through kernel/irq/pm.c::suspend_device_irqs().
>
> The SYSC hardware block in the R-Car SoCs, on the other hand, uses any
> interrupt as wake-up source(es) and requires them to be unmasked during
> the system is suspended.
>
> This patch implements .irq_set_wake() which makes __disable_irq() call
> a NOP when a GPIO port is claimed as a wake-up source.
>
> Signed-off-by: Shinya Kuribayashi <shinya.kuribayashi.px@renesas.com>
> Signed-off-by: Koji Matsuoka <koji.matsuoka.xm@renesas.com>
> Signed-off-by: Yoshihiro Kaneko <ykaneko0929@gmail.com>

Patch tentatively applied unless Magnus or Laurent starts complaining...

This driver should be simple to convert to GPIOLIB_IRQCHIP and get
rid of some code overhead.

Yoshihiro, can you look into this?

Yours,
Linus Walleij

Re: [PATCH] gpio-rcar: .irq_set_wake() support to work with suspend_device_irqs()

[Magnus Damm]

Hi Linus, Kaneko-san,

On Fri, Nov 14, 2014 at 6:53 PM, Linus Walleij <linus.walleij@linaro.org> wrote:
> On Fri, Nov 7, 2014 at 10:22 AM, Yoshihiro Kaneko <ykaneko0929@gmail.com> wrote:
>
>> From: Shinya Kuribayashi <shinya.kuribayashi.px@renesas.com>
>>
>> All IRQs in the kernel get disabled by PM core prior to .suspend_noirq()
>> step through kernel/irq/pm.c::suspend_device_irqs().
>>
>> The SYSC hardware block in the R-Car SoCs, on the other hand, uses any
>> interrupt as wake-up source(es) and requires them to be unmasked during
>> the system is suspended.
>>
>> This patch implements .irq_set_wake() which makes __disable_irq() call
>> a NOP when a GPIO port is claimed as a wake-up source.
>>
>> Signed-off-by: Shinya Kuribayashi <shinya.kuribayashi.px@renesas.com>
>> Signed-off-by: Koji Matsuoka <koji.matsuoka.xm@renesas.com>
>> Signed-off-by: Yoshihiro Kaneko <ykaneko0929@gmail.com>
>
> Patch tentatively applied unless Magnus or Laurent starts complaining...

I'll bite. =)

Thanks for the efforts, but I doubt that this patch is needed. Unless
I'm mistaken this feature is implemented already in the following
commits:

4039611 gpio: rcar: Enable mask on suspend
fba968a gpio: rcar: Use lazy disable

> This driver should be simple to convert to GPIOLIB_IRQCHIP and get
> rid of some code overhead.

Perhaps Geert [CC:ed] can have a look at this in the not so distant future?

Thanks,

/ magnus

Re: [PATCH] gpio-rcar: .irq_set_wake() support to work with suspend_device_irqs()

[Geert Uytterhoeven]

Hi Magnus, Kaneko-san,

On Mon, Nov 17, 2014 at 2:06 AM, Magnus Damm <magnus.damm@gmail.com> wrote:
> On Fri, Nov 14, 2014 at 6:53 PM, Linus Walleij <linus.walleij@linaro.org> wrote:
>> On Fri, Nov 7, 2014 at 10:22 AM, Yoshihiro Kaneko <ykaneko0929@gmail.com> wrote:
>>> From: Shinya Kuribayashi <shinya.kuribayashi.px@renesas.com>
>>>
>>> All IRQs in the kernel get disabled by PM core prior to .suspend_noirq()
>>> step through kernel/irq/pm.c::suspend_device_irqs().
>>>
>>> The SYSC hardware block in the R-Car SoCs, on the other hand, uses any
>>> interrupt as wake-up source(es) and requires them to be unmasked during
>>> the system is suspended.
>>>
>>> This patch implements .irq_set_wake() which makes __disable_irq() call
>>> a NOP when a GPIO port is claimed as a wake-up source.
>>>
>>> Signed-off-by: Shinya Kuribayashi <shinya.kuribayashi.px@renesas.com>
>>> Signed-off-by: Koji Matsuoka <koji.matsuoka.xm@renesas.com>
>>> Signed-off-by: Yoshihiro Kaneko <ykaneko0929@gmail.com>
>>
>> Patch tentatively applied unless Magnus or Laurent starts complaining...
>
> I'll bite. =)
>
> Thanks for the efforts, but I doubt that this patch is needed. Unless
> I'm mistaken this feature is implemented already in the following
> commits:
>
> 4039611 gpio: rcar: Enable mask on suspend
> fba968a gpio: rcar: Use lazy disable

Good to see you (as the author of the driver and those two fixes) have some
reservations. I also had my doubts about this patch, but I didn't get to
testing it earlier...

I can confirm resume from s2ram works without and with this patch.
So unless this patch is meant to fix something else (is it?), I think it can be
dropped.

>> This driver should be simple to convert to GPIOLIB_IRQCHIP and get
>> rid of some code overhead.
>
> Perhaps Geert [CC:ed] can have a look at this in the not so distant future?

Sure. Will do.

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH] gpio-rcar: .irq_set_wake() support to work with suspend_device_irqs()

[Linus Walleij]

On Mon, Nov 17, 2014 at 1:05 PM, Geert Uytterhoeven
<geert@linux-m68k.org> wrote:

> I can confirm resume from s2ram works without and with this patch.
> So unless this patch is meant to fix something else (is it?), I think it can be
> dropped.

OK patch is dropped from my queue.

>>> This driver should be simple to convert to GPIOLIB_IRQCHIP and get
>>> rid of some code overhead.
>>
>> Perhaps Geert [CC:ed] can have a look at this in the not so distant future?
>
> Sure. Will do.

THANKS!

Yours,
Linus Walleij