Re: [PATCH] pinctrl: ingenic: Convert to immutable irq chip

From: Marc Zyngier
Date: Thu Jun 09 2022 - 08:09:06 EST


On 2022-06-09 11:00, Paul Cercueil wrote:
Hi Aidan,

Le mar., juin 7 2022 at 17:47:19 +0100, Aidan MacDonald
<aidanmacdonald.0x0@xxxxxxxxx> a écrit :

Paul Cercueil <paul@xxxxxxxxxxxxxxx> writes:

Hi Aidan,

Le mar., juin 7 2022 at 12:05:25 +0100, Aidan MacDonald
<aidanmacdonald.0x0@xxxxxxxxx> a écrit :
Update the driver to use an immutable IRQ chip to fix this warning:
"not an immutable chip, please consider fixing it!"
Signed-off-by: Aidan MacDonald <aidanmacdonald.0x0@xxxxxxxxx>
---
drivers/pinctrl/pinctrl-ingenic.c | 33 ++++++++++++++++++-------------
1 file changed, 19 insertions(+), 14 deletions(-)
diff --git a/drivers/pinctrl/pinctrl-ingenic.c
b/drivers/pinctrl/pinctrl-ingenic.c
index 1ca11616db74..37258fb05be3 100644
--- a/drivers/pinctrl/pinctrl-ingenic.c
+++ b/drivers/pinctrl/pinctrl-ingenic.c
@@ -135,7 +135,6 @@ struct ingenic_pinctrl {
struct ingenic_gpio_chip {
struct ingenic_pinctrl *jzpc;
struct gpio_chip gc;
- struct irq_chip irq_chip;
unsigned int irq, reg_base;
};
@@ -3419,6 +3418,8 @@ static void ingenic_gpio_irq_enable(struct irq_data
*irqd)
struct ingenic_gpio_chip *jzgc = gpiochip_get_data(gc);
int irq = irqd->hwirq;
+ gpiochip_enable_irq(gc, irq);
+
if (is_soc_or_above(jzgc->jzpc, ID_JZ4770))
ingenic_gpio_set_bit(jzgc, JZ4770_GPIO_INT, irq, true);
else if (is_soc_or_above(jzgc->jzpc, ID_JZ4740))
@@ -3443,6 +3444,8 @@ static void ingenic_gpio_irq_disable(struct irq_data
*irqd)
ingenic_gpio_set_bit(jzgc, JZ4740_GPIO_SELECT, irq, false);
else
ingenic_gpio_set_bit(jzgc, JZ4730_GPIO_GPIER, irq, false);
+
+ gpiochip_disable_irq(gc, irq);
}
static void ingenic_gpio_irq_ack(struct irq_data *irqd)
@@ -3684,6 +3687,20 @@ static void ingenic_gpio_irq_release(struct irq_data
*data)
return gpiochip_relres_irq(gpio_chip, data->hwirq);
}
+static const struct irq_chip ingenic_gpio_irqchip = {
+ .name = "gpio",
+ .irq_enable = ingenic_gpio_irq_enable,
+ .irq_disable = ingenic_gpio_irq_disable,
+ .irq_unmask = ingenic_gpio_irq_unmask,
+ .irq_mask = ingenic_gpio_irq_mask,
+ .irq_ack = ingenic_gpio_irq_ack,
+ .irq_set_type = ingenic_gpio_irq_set_type,
+ .irq_set_wake = ingenic_gpio_irq_set_wake,
+ .irq_request_resources = ingenic_gpio_irq_request,
+ .irq_release_resources = ingenic_gpio_irq_release,
+ .flags = IRQCHIP_MASK_ON_SUSPEND | IRQCHIP_IMMUTABLE,
+};
+
static int ingenic_pinmux_set_pin_fn(struct ingenic_pinctrl *jzpc,
int pin, int func)
{
@@ -4172,20 +4189,8 @@ static int __init ingenic_gpio_probe(struct
ingenic_pinctrl *jzpc,
if (!jzgc->irq)
return -EINVAL;
- jzgc->irq_chip.name = jzgc->gc.label;
- jzgc->irq_chip.irq_enable = ingenic_gpio_irq_enable;
- jzgc->irq_chip.irq_disable = ingenic_gpio_irq_disable;
- jzgc->irq_chip.irq_unmask = ingenic_gpio_irq_unmask;
- jzgc->irq_chip.irq_mask = ingenic_gpio_irq_mask;
- jzgc->irq_chip.irq_ack = ingenic_gpio_irq_ack;
- jzgc->irq_chip.irq_set_type = ingenic_gpio_irq_set_type;
- jzgc->irq_chip.irq_set_wake = ingenic_gpio_irq_set_wake;
- jzgc->irq_chip.irq_request_resources = ingenic_gpio_irq_request;
- jzgc->irq_chip.irq_release_resources = ingenic_gpio_irq_release;
- jzgc->irq_chip.flags = IRQCHIP_MASK_ON_SUSPEND;
-
girq = &jzgc->gc.irq;
- girq->chip = &jzgc->irq_chip;
+ gpio_irq_chip_set_chip(girq, &ingenic_gpio_irqchip);

This will change each irq_chip's name to "gpio", do we want that?

You didn't remove jzgc->irq_chip, so maybe what you could do is
jzgc->irq_chip = ingenic_gpio_irqchip;
jzgc->irq_chip.name = jzgc->gc.label;
gpio_irq_chip_set_chip(girq, &jzgc->irq_chip);

Thoughts?

Cheers,
-Paul


I wondered that myself, but it doesn't seem to affect anything except
what is displayed in /proc/interrupts. Is the name used anywhere else
where it might cause confusion?

I don't really know. If it only really affects the display in
/proc/interrupts then I'm fine with it. In doubt, I'd prefer to keep
the existing names.

The only similar case I could find was pinctrl-microchip-sgpio.c where
microchip_sgpio_register_bank() is called in a loop and registers the
same irq chip repeatedly, so it's probably(?) okay to do this here. It
seems to defeat the point of immutable irqchips if they just have to be
copied anyway...

The point of immutable irqchips is that they aren't modified by the
core, if I understand it correctly. Immutable doesn't mean it has to
be static const.

I want these to be made const. I agree that the fancy string should
be kept (sadly), as it is a userspace visible change, and we don't
do that.

You can solve it using the irq_print_chip() callback as part of
your irq_chip structures. See 3344265a2692 for an example.

Thanks,

M.
--
Jazz is not dead. It just smells funny...