[alsa-devel] [PATCH] ASoC: Use delayed work for debounce of GPIO based jacks
Rather than block the workqueue by sleeping to do the debounce use delayed work to implement the debounce time. This should also mean that we extend the debounce time on each new bounce, potentially allowing shorter debounce times for clean insertions.
Signed-off-by: Mark Brown broonie@opensource.wolfsonmicro.com ---
This is completely untested as I do not have any systems to hand which can make use of this functionality but I believe it should work - I've CCed a couple of the OMAP people since they seem to be the main users.
include/sound/soc.h | 2 +- sound/soc/soc-jack.c | 11 +++++------ 2 files changed, 6 insertions(+), 7 deletions(-)
diff --git a/include/sound/soc.h b/include/sound/soc.h index 493b3a4..4fb079e 100644 --- a/include/sound/soc.h +++ b/include/sound/soc.h @@ -385,7 +385,7 @@ struct snd_soc_jack_gpio { int invert; int debounce_time; struct snd_soc_jack *jack; - struct work_struct work; + struct delayed_work work;
int (*jack_status_check)(void); }; diff --git a/sound/soc/soc-jack.c b/sound/soc/soc-jack.c index 8862770..8a0a920 100644 --- a/sound/soc/soc-jack.c +++ b/sound/soc/soc-jack.c @@ -188,9 +188,6 @@ static void snd_soc_jack_gpio_detect(struct snd_soc_jack_gpio *gpio) int enable; int report;
- if (gpio->debounce_time > 0) - mdelay(gpio->debounce_time); - enable = gpio_get_value(gpio->gpio); if (gpio->invert) enable = !enable; @@ -211,7 +208,8 @@ static irqreturn_t gpio_handler(int irq, void *data) { struct snd_soc_jack_gpio *gpio = data;
- schedule_work(&gpio->work); + schedule_delayed_work(&gpio->work, + msecs_to_jiffies(gpio->debounce_time));
return IRQ_HANDLED; } @@ -221,7 +219,7 @@ static void gpio_work(struct work_struct *work) { struct snd_soc_jack_gpio *gpio;
- gpio = container_of(work, struct snd_soc_jack_gpio, work); + gpio = container_of(work, struct snd_soc_jack_gpio, work.work); snd_soc_jack_gpio_detect(gpio); }
@@ -262,7 +260,7 @@ int snd_soc_jack_add_gpios(struct snd_soc_jack *jack, int count, if (ret) goto err;
- INIT_WORK(&gpios[i].work, gpio_work); + INIT_DELAYED_WORK(&gpios[i].work, gpio_work); gpios[i].jack = jack;
ret = request_irq(gpio_to_irq(gpios[i].gpio), @@ -312,6 +310,7 @@ void snd_soc_jack_free_gpios(struct snd_soc_jack *jack, int count, gpio_unexport(gpios[i].gpio); #endif free_irq(gpio_to_irq(gpios[i].gpio), &gpios[i]); + cancel_delayed_work_sync(&gpios[i].work); gpio_free(gpios[i].gpio); gpios[i].jack = NULL; }
On Wed, 6 Oct 2010 16:05:01 -0700 Mark Brown broonie@opensource.wolfsonmicro.com wrote:
Rather than block the workqueue by sleeping to do the debounce use delayed work to implement the debounce time. This should also mean that we extend the debounce time on each new bounce, potentially allowing shorter debounce times for clean insertions.
Signed-off-by: Mark Brown broonie@opensource.wolfsonmicro.com
This is completely untested as I do not have any systems to hand which can make use of this functionality but I believe it should work - I've CCed a couple of the OMAP people since they seem to be the main users.
include/sound/soc.h | 2 +- sound/soc/soc-jack.c | 11 +++++------ 2 files changed, 6 insertions(+), 7 deletions(-)
Works fine on N900 when testing with hexdump /dev/input/by-path/platform-soc-audio-event
I have a vague memory that user space and actual state can get soon out of sync if using delay based debouncing only and doing some insane plug in/out cycle. So this is a step to right direction for implementing more advanced debouncing.
Acked-by: Jarkko Nikula jhnikula@gmail.com
On Thu, 2010-10-07 at 08:54 +0300, Jarkko Nikula wrote:
On Wed, 6 Oct 2010 16:05:01 -0700 Mark Brown broonie@opensource.wolfsonmicro.com wrote:
Rather than block the workqueue by sleeping to do the debounce use delayed work to implement the debounce time. This should also mean that we extend the debounce time on each new bounce, potentially allowing shorter debounce times for clean insertions.
Signed-off-by: Mark Brown broonie@opensource.wolfsonmicro.com
This is completely untested as I do not have any systems to hand which can make use of this functionality but I believe it should work - I've CCed a couple of the OMAP people since they seem to be the main users.
include/sound/soc.h | 2 +- sound/soc/soc-jack.c | 11 +++++------ 2 files changed, 6 insertions(+), 7 deletions(-)
Works fine on N900 when testing with hexdump /dev/input/by-path/platform-soc-audio-event
I have a vague memory that user space and actual state can get soon out of sync if using delay based debouncing only and doing some insane plug in/out cycle. So this is a step to right direction for implementing more advanced debouncing.
Acked-by: Jarkko Nikula jhnikula@gmail.com
Acked-by: Liam Girdwood lrg@slimlogic.co.uk
On Wed, 2010-10-06 at 16:05 -0700, Mark Brown wrote:
Rather than block the workqueue by sleeping to do the debounce use delayed work to implement the debounce time. This should also mean that we extend the debounce time on each new bounce, potentially allowing shorter debounce times for clean insertions.
Signed-off-by: Mark Brown broonie@opensource.wolfsonmicro.com
Acked-by: Liam Girdwood lrg@slimlogic.co.uk
participants (3)
-
Jarkko Nikula -
Liam Girdwood -
Mark Brown