Skip to contentSkip to conversation

Input: gpio-keys: Support for one-directional interrupts

PATCH In review Cory Maccarrone 3 participants 7 messages Dec 5, 2009 – Jan 9, 2010

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Input: gpio-keys: Support for one-directional interrupts

The current gpio-keys driver assumes that each given GPIO has the ability to respond to both rising and falling interrupts at the same time. For some GPIOs on certain devices, the interrupt is only one-directional -- either rising or falling, depending on what's being listened for. In particular, the HTC Herald keyboard slide switch interrupt must be switched from rising to falling and back on each event, otherwise only one transition will be detected.

This change adds a new entry to the gpio_keys_button structure to allow for this condition to be specified in board configurations. When set, the driver detects the current state of the GPIO and sets the direction of the interrupt accordingly. Then, on each event, the interrupt direction is switched.

---
 drivers/input/keyboard/gpio_keys.c |   27 +++++++++++++++++++++++----
 include/linux/gpio_keys.h          |    1 +
 2 files changed, 24 insertions(+), 4 deletions(-)
drivers/input/keyboard/gpio_keys.c +23 -4
index 77d1309..05c599a 100644
@@ -45,7 +45,18 @@ static void gpio_keys_report_event(struct work_struct *work)
 	struct gpio_keys_button *button = bdata->button;
 	struct input_dev *input = bdata->input;
 	unsigned int type = button->type ?: EV_KEY;
-	int state = (gpio_get_value(button->gpio) ? 1 : 0) ^ button->active_low;
+	int state = gpio_get_value(button->gpio) ? 1 : 0;
+
+	/* If the hardware can't do both edges, set the appropriate
+	 * interrupt control */
+	if (button->not_both_edges) {
+		if (state)
+			set_irq_type(gpio_to_irq(button->gpio), IRQ_TYPE_EDGE_FALLING);
+		else
+			set_irq_type(gpio_to_irq(button->gpio), IRQ_TYPE_EDGE_RISING);
+	}
+
+	state ^= button->active_low;
 
 	input_event(input, type, button->code, !!state);
 	input_sync(input);
@@ -79,7 +90,7 @@ static int __devinit gpio_keys_probe(struct platform_device *pdev)
 	struct gpio_keys_platform_data *pdata = pdev->dev.platform_data;
 	struct gpio_keys_drvdata *ddata;
 	struct input_dev *input;
-	int i, error;
+	int i, error, trigger;
 	int wakeup = 0;
 
 	ddata = kzalloc(sizeof(struct gpio_keys_drvdata) +
@@ -146,9 +157,17 @@ static int __devinit gpio_keys_probe(struct platform_device *pdev)
 			goto fail2;
 		}
 
+		if (button->not_both_edges) {
+			if (gpio_get_value(button->gpio))
+				trigger = IRQF_TRIGGER_FALLING;
+			else
+				trigger = IRQF_TRIGGER_RISING;
+		} else {
+			trigger = IRQF_TRIGGER_FALLING | IRQF_TRIGGER_RISING;
+		}
+
 		error = request_irq(irq, gpio_keys_isr,
-				    IRQF_SHARED |
-				    IRQF_TRIGGER_RISING | IRQF_TRIGGER_FALLING,
+				    IRQF_SHARED | trigger,
 				    button->desc ? button->desc : "gpio_keys",
 				    bdata);
 		if (error) {
include/linux/gpio_keys.h +1 -0
index 1289fa7..493426d 100644
@@ -10,6 +10,7 @@ struct gpio_keys_button {
 	int type;		/* input event type (EV_KEY, EV_SW) */
 	int wakeup;		/* configure the button as a wake-up source */
 	int debounce_interval;	/* debounce ticks interval in msecs */
+	int not_both_edges;	/* some hardware can't do interrupts on both edges */
 };
 
 struct gpio_keys_platform_data {
Signature · 1 line
1.6.3.3
#1 lore raw
Quoted 7 lines · Cory Maccarrone

The current gpio-keys driver assumes that each given GPIO has the ability to respond to both rising and falling interrupts at the same time. For some GPIOs on certain devices, the interrupt is only one-directional -- either rising or falling, depending on what's being listened for. In particular, the HTC Herald keyboard slide switch interrupt must be switched from rising to falling and back on each event, otherwise only one transition will be detected.

Interesting. Btw. I'd like to convert gpio-keys from edge-triggered to level-triggered interrupts. Would that work for your hardware?

Signature · 2 lines
Regards,
Feri.
#2 lore raw
Quoted 4 lines · Ferenc Wagner

Cory Maccarrone <[email protected]> writes:

Interesting.  Btw. I'd like to convert gpio-keys from edge-triggered to level-triggered interrupts.  Would that work for your hardware?

I'm fairly certain it wouldn't. Each of the interrupts on my hardware has a corresponding bit in a control register that determines if it's rising or falling-edge triggered -- thus the need for my patch. To capture both directions, it's necessary to modify the control register when a falling edge is detected, so that the corresponding rising edge can be collected afterward. May I ask why you're thinking of converting to level-triggering?

Perhaps it would be better to provide an option in the platform_device structure to set edge- or level-triggering, similar to the change I'm proposing for interrupts that can only signal one way at a time.

- Cory
--
To unsubscribe from this list: send the line "unsubscribe linux-input" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

maintainer
#3 lore raw
Quoted 6 lines · Cory Maccarrone
Quoted 4 lines · Ferenc Wagner

Cory Maccarrone <[email protected]> writes:

Interesting.  Btw. I'd like to convert gpio-keys from edge-triggered to level-triggered interrupts.  Would that work for your hardware?

Quoted 6 lines

I'm fairly certain it wouldn't. Each of the interrupts on my hardware has a corresponding bit in a control register that determines if it's rising or falling-edge triggered -- thus the need for my patch. To capture both directions, it's necessary to modify the control register when a falling edge is detected, so that the corresponding rising edge can be collected afterward.

This kind of ugliness should be hidden in irqchip driver. See mfd/asic3.c for an example.

Quoted 7 lines

May I ask why you're thinking of
converting to level-triggering?

Perhaps it would be better to provide an option in the platform_device structure to set edge- or level-triggering, similar to the change I'm proposing for interrupts that can only signal one way at a time.

Yes, we need a way fro platform code to specify desired interrupt flags but I don't believe we should be reconfiguring them on the fly.

Signature · 5 lines
Dmitry
--
To unsubscribe from this list: send the line "unsubscribe linux-input" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#4 lore raw
Quoted 10 lines · Dmitry Torokhov
Quoted 8 lines · Cory Maccarrone

On Wed, Dec 9, 2009 at 3:32 AM, Ferenc Wagner <[email protected]> wrote:

I'm fairly certain it wouldn't.  Each of the interrupts on my hardware has a corresponding bit in a control register that determines if it's rising or falling-edge triggered -- thus the need for my patch.  To capture both directions, it's necessary to modify the control register when a falling edge is detected, so that the corresponding rising edge can be collected afterward.

Quoted 3 lines

This kind of ugliness should be hidden in irqchip driver. See mfd/asic3.c for an example.

I would agree with that -- it should be possible to hide that away as part of the functionality of the interrupt driver. Perhaps a fix to the OMAP1 IRQ handling is warranted. I know there are some interrupts that can do both directions out of the box, but that shouldn't be the responsibility of each driver to know.

Quoted 11 lines
Quoted 7 lines

May I ask why you're thinking of
converting to level-triggering?

Perhaps it would be better to provide an option in the platform_device structure to set edge- or level-triggering, similar to the change I'm proposing for interrupts that can only signal one way at a time.

Yes, we need a way fro platform code to specify desired interrupt flags but I don't believe we should be reconfiguring them on the fly.

If the ability to handle this type of interrupt transparently was added to our board IRQ chip driver, this whole thing would be a non-issue, as gpio-keys could request edge triggered falling and rising at the same time, and the driver would Just Work.

I'll take a look and see how possible that would be to do. Looking through the code for OMAP1's GPIO IRQ handlers, there's a note that OMAP1 only supports edge triggering, so if gpio-keys were to move exclusively to that, itwould stop working with those boards. But, it may be possible to do the trigger flip in the chip driver, which would make this patch unnecessary.

- Cory
--
To unsubscribe from this list: send the line "unsubscribe linux-input" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#5 lore raw
Quoted 6 lines · Cory Maccarrone

I'll take a look and see how possible that would be to do.  Looking through the code for OMAP1's GPIO IRQ handlers, there's a note that OMAP1 only supports edge triggering, so if gpio-keys were to move exclusively to that, itwould stop working with those boards.  But, it may be possible to do the trigger flip in the chip driver, which would make this patch unnecessary.

I've just submitted a patch to linux-omap for this against the gpio/irq handling code. It implements the ICR flip in there automatically whenever both RISING and FALLING edge are requested of one of these gpios. With it, the patch I've submitted here is moot. We'll see if it gets accepted.

- Cory
--
To unsubscribe from this list: send the line "unsubscribe linux-input" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#6 lore raw
Quoted 6 lines · Cory Maccarrone

I've just submitted a patch to linux-omap for this against the gpio/irq handling code. It implements the ICR flip in there automatically whenever both RISING and FALLING edge are requested of one of these gpios.  With it, the patch I've submitted here is moot. We'll see if it gets accepted.

OK, looks like the patch will get approved, so I'm dropping this patch. Thanks!

- Cory
--
To unsubscribe from this list: send the line "unsubscribe linux-input" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html