[PATCH] pinctrl: bcm2835: Fix support for threaded level triggered IRQs

[PATCH] pinctrl: bcm2835: Fix support for threaded level triggered IRQs

[Charles Keepax]

Currently, the driver uses handle_simple_irq for all IRQ types and hard
codes the acknowledge for different IRQ types into the handler. It is
better to use the IRQ core as intended and let it handle the differences
between the various types of IRQ. For example the current system does
not work for threaded level triggered IRQs as these need to be masked
until the threaded handler has run.

Signed-off-by: Charles Keepax <ckeepax@opensource.wolfsonmicro.com>
---
 drivers/pinctrl/bcm/pinctrl-bcm2835.c |   26 +++++++++++++++++---------
 1 files changed, 17 insertions(+), 9 deletions(-)

diff --git a/drivers/pinctrl/bcm/pinctrl-bcm2835.c b/drivers/pinctrl/bcm/pinctrl-bcm2835.c
index 9aa8a3f..f0e5903 100644
--- a/drivers/pinctrl/bcm/pinctrl-bcm2835.c
+++ b/drivers/pinctrl/bcm/pinctrl-bcm2835.c
@@ -403,15 +403,7 @@ static irqreturn_t bcm2835_gpio_irq_handler(int irq, void *dev_id)
 		gpio = (32 * bank) + offset;
 		type = pc->irq_type[gpio];
 
-		/* ack edge triggered IRQs immediately */
-		if (!(type & IRQ_TYPE_LEVEL_MASK))
-			bcm2835_gpio_set_bit(pc, GPEDS0, gpio);
-
 		generic_handle_irq(irq_linear_revmap(pc->irq_domain, gpio));
-
-		/* ack level triggered IRQ after handling them */
-		if (type & IRQ_TYPE_LEVEL_MASK)
-			bcm2835_gpio_set_bit(pc, GPEDS0, gpio);
 	}
 	return events ? IRQ_HANDLED : IRQ_NONE;
 }
@@ -591,16 +583,32 @@ static int bcm2835_gpio_irq_set_type(struct irq_data *data, unsigned int type)
 	else
 		ret = __bcm2835_gpio_irq_set_type_disabled(pc, gpio, type);
 
+	if (type & IRQ_TYPE_EDGE_BOTH)
+		__irq_set_handler_locked(data->irq, handle_edge_irq);
+	else
+		__irq_set_handler_locked(data->irq, handle_level_irq);
+
 	spin_unlock_irqrestore(&pc->irq_lock[bank], flags);
 
 	return ret;
 }
 
+static void bcm2835_gpio_irq_ack(struct irq_data *data)
+{
+	struct bcm2835_pinctrl *pc = irq_data_get_irq_chip_data(data);
+	unsigned gpio = irqd_to_hwirq(data);
+
+	bcm2835_gpio_set_bit(pc, GPEDS0, gpio);
+}
+
 static struct irq_chip bcm2835_gpio_irq_chip = {
 	.name = MODULE_NAME,
 	.irq_enable = bcm2835_gpio_irq_enable,
 	.irq_disable = bcm2835_gpio_irq_disable,
 	.irq_set_type = bcm2835_gpio_irq_set_type,
+	.irq_ack = bcm2835_gpio_irq_ack,
+	.irq_mask = bcm2835_gpio_irq_disable,
+	.irq_unmask = bcm2835_gpio_irq_enable,
 };
 
 static int bcm2835_pctl_get_groups_count(struct pinctrl_dev *pctldev)
@@ -977,7 +985,7 @@ static int bcm2835_pinctrl_probe(struct platform_device *pdev)
 		int irq = irq_create_mapping(pc->irq_domain, i);
 		irq_set_lockdep_class(irq, &gpio_lock_class);
 		irq_set_chip_and_handler(irq, &bcm2835_gpio_irq_chip,
-				handle_simple_irq);
+				handle_level_irq);
 		irq_set_chip_data(irq, pc);
 		set_irq_flags(irq, IRQF_VALID);
 	}
-- 
1.7.2.5

Re: [PATCH] pinctrl: bcm2835: Fix support for threaded level triggered IRQs

[Linus Walleij]

On Tue, Apr 7, 2015 at 12:43 PM, Charles Keepax
<ckeepax@opensource.wolfsonmicro.com> wrote:

> Currently, the driver uses handle_simple_irq for all IRQ types and hard
> codes the acknowledge for different IRQ types into the handler. It is
> better to use the IRQ core as intended and let it handle the differences
> between the various types of IRQ. For example the current system does
> not work for threaded level triggered IRQs as these need to be masked
> until the threaded handler has run.
>
> Signed-off-by: Charles Keepax <ckeepax@opensource.wolfsonmicro.com>

Looks like a very fine patch so applied unless Stephen has
some problem with it.

Yours,
Linus Walleij

Re: [PATCH] pinctrl: bcm2835: Fix support for threaded level triggered IRQs

[Stephen Warren]

On 04/08/2015 08:43 AM, Linus Walleij wrote:
> On Tue, Apr 7, 2015 at 12:43 PM, Charles Keepax
> <ckeepax@opensource.wolfsonmicro.com> wrote:
> 
>> Currently, the driver uses handle_simple_irq for all IRQ types and hard
>> codes the acknowledge for different IRQ types into the handler. It is
>> better to use the IRQ core as intended and let it handle the differences
>> between the various types of IRQ. For example the current system does
>> not work for threaded level triggered IRQs as these need to be masked
>> until the threaded handler has run.
>>
>> Signed-off-by: Charles Keepax <ckeepax@opensource.wolfsonmicro.com>
> 
> Looks like a very fine patch so applied unless Stephen has
> some problem with it.

This patch sounds plausible to me:-)