Rather than always using handle_simple_irq() as the gpio_irq_chip
handler, set a more appropriate handler based on the IRQ trigger type
requested. This is important for level triggered interrupts which need
to be masked during handling. Also, always acknowledge the interrupt
regardless of whether it is edge or level triggered.
Signed-off-by: Mark Tomlinson <[email protected]>
---
drivers/pinctrl/bcm/pinctrl-nsp-gpio.c | 13 ++++++++-----
1 file changed, 8 insertions(+), 5 deletions(-)
diff --git a/drivers/pinctrl/bcm/pinctrl-nsp-gpio.c b/drivers/pinctrl/bcm/pinctrl-nsp-gpio.c
index bed0124388c0..349fb384113e 100644
--- a/drivers/pinctrl/bcm/pinctrl-nsp-gpio.c
+++ b/drivers/pinctrl/bcm/pinctrl-nsp-gpio.c
@@ -174,11 +174,8 @@ static void nsp_gpio_irq_ack(struct irq_data *d)
struct nsp_gpio *chip = gpiochip_get_data(gc);
unsigned gpio = d->hwirq;
u32 val = BIT(gpio);
- u32 trigger_type;
- trigger_type = irq_get_trigger_type(d->irq);
- if (trigger_type & (IRQ_TYPE_EDGE_FALLING | IRQ_TYPE_EDGE_RISING))
- nsp_set_bit(chip, REG, NSP_GPIO_EVENT, gpio, val);
+ nsp_set_bit(chip, REG, NSP_GPIO_EVENT, gpio, val);
}
/*
@@ -262,6 +259,12 @@ static int nsp_gpio_irq_set_type(struct irq_data *d, unsigned int type)
nsp_set_bit(chip, REG, NSP_GPIO_EVENT_INT_POLARITY, gpio, falling);
nsp_set_bit(chip, REG, NSP_GPIO_INT_POLARITY, gpio, level_low);
+
+ if (type & IRQ_TYPE_EDGE_BOTH)
+ irq_set_handler_locked(d, handle_edge_irq);
+ else
+ irq_set_handler_locked(d, handle_level_irq);
+
raw_spin_unlock_irqrestore(&chip->lock, flags);
dev_dbg(chip->dev, "gpio:%u level_low:%s falling:%s\n", gpio,
@@ -691,7 +694,7 @@ static int nsp_gpio_probe(struct platform_device *pdev)
girq->num_parents = 0;
girq->parents = NULL;
girq->default_type = IRQ_TYPE_NONE;
- girq->handler = handle_simple_irq;
+ girq->handler = handle_bad_irq;
}
ret = devm_gpiochip_add_data(dev, gc, chip);
--
2.27.0
Hi Mark,
On 6/30/2020 1:47 PM, Mark Tomlinson wrote:
> Rather than always using handle_simple_irq() as the gpio_irq_chip
> handler, set a more appropriate handler based on the IRQ trigger type
> requested. This is important for level triggered interrupts which need
> to be masked during handling. Also, always acknowledge the interrupt
> regardless of whether it is edge or level triggered.
>
> Signed-off-by: Mark Tomlinson <[email protected]>
> ---
> drivers/pinctrl/bcm/pinctrl-nsp-gpio.c | 13 ++++++++-----
> 1 file changed, 8 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/pinctrl/bcm/pinctrl-nsp-gpio.c b/drivers/pinctrl/bcm/pinctrl-nsp-gpio.c
> index bed0124388c0..349fb384113e 100644
> --- a/drivers/pinctrl/bcm/pinctrl-nsp-gpio.c
> +++ b/drivers/pinctrl/bcm/pinctrl-nsp-gpio.c
> @@ -174,11 +174,8 @@ static void nsp_gpio_irq_ack(struct irq_data *d)
> struct nsp_gpio *chip = gpiochip_get_data(gc);
> unsigned gpio = d->hwirq;
> u32 val = BIT(gpio);
> - u32 trigger_type;
>
> - trigger_type = irq_get_trigger_type(d->irq);
> - if (trigger_type & (IRQ_TYPE_EDGE_FALLING | IRQ_TYPE_EDGE_RISING))
> - nsp_set_bit(chip, REG, NSP_GPIO_EVENT, gpio, val);
> + nsp_set_bit(chip, REG, NSP_GPIO_EVENT, gpio, val);
I have a question here. I assume writing a bit to this register will
result in clearing that bit, is that true?
Based on the driver, the 'nsp_gpio_irq_handler' seems to rely on
'NSP_GPIO_EVENT' register to figure out which GPIO the interrupt is for.
And if so, and if this is cleared here that is invoked before the actual
IRQ handler, how does this work?
I could be missing something here, so please help to explain it in more
details here.
Thanks,
Ray
On Tue, 2020-06-30 at 15:26 -0700, Ray Jui wrote:
> - u32 trigger_type;
> >
> > - trigger_type = irq_get_trigger_type(d->irq);
> > - if (trigger_type & (IRQ_TYPE_EDGE_FALLING | IRQ_TYPE_EDGE_RISING))
> > - nsp_set_bit(chip, REG, NSP_GPIO_EVENT, gpio, val);
> > + nsp_set_bit(chip, REG, NSP_GPIO_EVENT, gpio, val);
>
>
> I have a question here. I assume writing a bit to this register will
> result in clearing that bit, is that true?
>
> Based on the driver, the 'nsp_gpio_irq_handler' seems to rely on
> 'NSP_GPIO_EVENT' register to figure out which GPIO the interrupt is for.
> And if so, and if this is cleared here that is invoked before the actual
> IRQ handler, how does this work?
It seems that this change masked another issue I was having. However,
the original code is still wrong as using nsp_set_bit() will do a
read/modify/write and clear all events regardless of what gpio is.
I have found another issue, where I'm getting many more edge-triggered
interrupts than I think I should be, so I'll sort that and send a v2 of
this patch.