WM97xx touchscreen updates

WM97xx touchscreen updates

[Mark Brown]

This patch series updates the WM97xx touchscreen drivers with a couple
of small cleanups and the addition of support for use of the WM97xx
touchpanels to wake up from suspend.  Since the last submission I have:

 - Updated the commit message for scheduling of the interrupt handler. 
 - Hooked the touchscreen into the driver model wakeup infrastructure.

It is also available in the git repository at:

  git://opensource.wolfsonmicro.com/linux-2.6-touch upstream

Mark Brown (3):
      wm97xx-core: Only schedule interrupt handler if not already scheduled
      wm97xx-core: Use IRQF_SAMPLE_RANDOM
      wm97xx-core: Support use as a wakeup source

 drivers/input/touchscreen/wm97xx-core.c |   56 ++++++++++++++++++++++++++----
 include/linux/wm97xx.h                  |    3 ++
 2 files changed, 51 insertions(+), 8 deletions(-)

[PATCH 1/3] wm97xx-core: Only schedule interrupt handler if not already scheduled

[Mark Brown]

As well as clarifying the fact that the driver can cope if a second
interrupt occurs before the IRQ work is scheduled this also ensures
that calls to the machine irq_enable() are balanced, making that easier
to implement.  Normally this is redundant due to the interrupt disabling
but some unusal board configurations can trigger it.

Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
---
 drivers/input/touchscreen/wm97xx-core.c |   14 +++++++-------
 1 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/drivers/input/touchscreen/wm97xx-core.c b/drivers/input/touchscreen/wm97xx-core.c
index 2910999..e27b1e0 100644
--- a/drivers/input/touchscreen/wm97xx-core.c
+++ b/drivers/input/touchscreen/wm97xx-core.c
@@ -328,18 +328,18 @@ static void wm97xx_pen_irq_worker(struct work_struct *work)
  *
  * We have to disable the codec interrupt in the handler because it
  * can take upto 1ms to clear the interrupt source. We schedule a task
- * in a work queue to do the actual interaction with the chip (it
- * doesn't matter if we end up reenqueing it before it is executed
- * since we don't touch the chip until it has run).  The interrupt is
- * then enabled again in the slow handler when the source has been
- * cleared.
+ * in a work queue to do the actual interaction with the chip.  The
+ * interrupt is then enabled again in the slow handler when the source
+ * has been cleared.
  */
 static irqreturn_t wm97xx_pen_interrupt(int irq, void *dev_id)
 {
 	struct wm97xx *wm = dev_id;
 
-	wm->mach_ops->irq_enable(wm, 0);
-	queue_work(wm->ts_workq, &wm->pen_event_work);
+	if (!work_pending(&wm->pen_event_work)) {
+		wm->mach_ops->irq_enable(wm, 0);
+		queue_work(wm->ts_workq, &wm->pen_event_work);
+	}
 
 	return IRQ_HANDLED;
 }
-- 
1.5.5

[PATCH 2/3] wm97xx-core: Use IRQF_SAMPLE_RANDOM

[Mark Brown]

The touchscreen interrupt is driven by human input which can reasonably
be used to provide entropy.

Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
---
 drivers/input/touchscreen/wm97xx-core.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/drivers/input/touchscreen/wm97xx-core.c b/drivers/input/touchscreen/wm97xx-core.c
index e27b1e0..fec07c2 100644
--- a/drivers/input/touchscreen/wm97xx-core.c
+++ b/drivers/input/touchscreen/wm97xx-core.c
@@ -355,7 +355,8 @@ static int wm97xx_init_pen_irq(struct wm97xx *wm)
 	 * provided. */
 	BUG_ON(!wm->mach_ops->irq_enable);
 
-	if (request_irq(wm->pen_irq, wm97xx_pen_interrupt, IRQF_SHARED,
+	if (request_irq(wm->pen_irq, wm97xx_pen_interrupt,
+			IRQF_SHARED | IRQF_SAMPLE_RANDOM,
 			"wm97xx-pen", wm)) {
 		dev_err(wm->dev,
 			"Failed to register pen down interrupt, polling");
-- 
1.5.5

[PATCH 3/3] wm97xx-core: Support use as a wakeup source

[Mark Brown]

The WM97xx touch screen controllers can be used to generate a wakeup
event when the system is suspended. Provide a new core API call
wm97xx_set_suspend_mode() allowing machine drivers to enable this. If no
suspend_mode is provided then the touch panel will be powered down when
the system is suspended.

Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
---
 drivers/input/touchscreen/wm97xx-core.c |   39 +++++++++++++++++++++++++++++++
 include/linux/wm97xx.h                  |    3 ++
 2 files changed, 42 insertions(+), 0 deletions(-)

diff --git a/drivers/input/touchscreen/wm97xx-core.c b/drivers/input/touchscreen/wm97xx-core.c
index fec07c2..e9c7ea4 100644
--- a/drivers/input/touchscreen/wm97xx-core.c
+++ b/drivers/input/touchscreen/wm97xx-core.c
@@ -262,6 +262,23 @@ void wm97xx_config_gpio(struct wm97xx *wm, u32 gpio, enum wm97xx_gpio_dir dir,
 EXPORT_SYMBOL_GPL(wm97xx_config_gpio);
 
 /*
+ * Configure the WM97XX_PRP value to use while system is suspended.
+ * If a value other than 0 is set then WM97xx pen detection will be
+ * left enabled in the configured mode while the system is in suspend,
+ * the device has users and suspend has not been disabled via the
+ * wakeup sysfs entries.
+ *
+ * @wm:   WM97xx device to configure
+ * @mode: WM97XX_PRP value to configure while suspended
+ */
+void wm97xx_set_suspend_mode(struct wm97xx *wm, u16 mode)
+{
+	wm->suspend_mode = mode;
+	device_init_wakeup(&wm->input_dev->dev, mode != 0);
+}
+EXPORT_SYMBOL_GPL(wm97xx_set_suspend_mode);
+
+/*
  * Handle a pen down interrupt.
  */
 static void wm97xx_pen_irq_worker(struct work_struct *work)
@@ -689,10 +706,32 @@ static int wm97xx_remove(struct device *dev)
 static int wm97xx_suspend(struct device *dev, pm_message_t state)
 {
 	struct wm97xx *wm = dev_get_drvdata(dev);
+	u16 reg;
+	int suspend_mode;
+
+	if (device_may_wakeup(&wm->input_dev->dev))
+		suspend_mode = wm->suspend_mode;
+	else
+		suspend_mode = 0;
 
 	if (wm->input_dev->users)
 		cancel_delayed_work_sync(&wm->ts_reader);
 
+	/* Power down the digitiser (bypassing the cache for resume) */
+	reg = wm97xx_reg_read(wm, AC97_WM97XX_DIGITISER2);
+	reg &= ~WM97XX_PRP_DET_DIG;
+	if (wm->input_dev->users)
+		reg |= suspend_mode;
+	wm->ac97->bus->ops->write(wm->ac97, AC97_WM97XX_DIGITISER2, reg);
+
+	/* WM9713 has an additional power bit - turn it off if there
+	 * are no users or if suspend mode is zero. */
+	if (wm->id == WM9713_ID2 &&
+	    (!wm->input_dev->users || !suspend_mode)) {
+		reg = wm97xx_reg_read(wm, AC97_EXTENDED_MID) | 0x8000;
+		wm97xx_reg_write(wm, AC97_EXTENDED_MID, reg);
+	}
+
 	return 0;
 }
 
diff --git a/include/linux/wm97xx.h b/include/linux/wm97xx.h
index ed01c7d..4d13732 100644
--- a/include/linux/wm97xx.h
+++ b/include/linux/wm97xx.h
@@ -282,6 +282,7 @@ struct wm97xx {
 	unsigned pen_is_down:1;		/* Pen is down */
 	unsigned aux_waiting:1;		/* aux measurement waiting */
 	unsigned pen_probably_down:1;	/* used in polling mode */
+	u16 suspend_mode;               /* PRP in suspend mode */
 };
 
 /*
@@ -297,6 +298,8 @@ void wm97xx_config_gpio(struct wm97xx *wm, u32 gpio,
 				     enum wm97xx_gpio_sticky sticky,
 				     enum wm97xx_gpio_wake wake);
 
+void wm97xx_set_suspend_mode(struct wm97xx *wm, u16 mode);
+
 /* codec AC97 IO access */
 int wm97xx_reg_read(struct wm97xx *wm, u16 reg);
 void wm97xx_reg_write(struct wm97xx *wm, u16 reg, u16 val);
-- 
1.5.5

Re: WM97xx touchscreen updates

[Dmitry Torokhov]

On Tue, Apr 15, 2008 at 03:45:35PM +0100, Mark Brown wrote:
> This patch series updates the WM97xx touchscreen drivers with a couple
> of small cleanups and the addition of support for use of the WM97xx
> touchpanels to wake up from suspend.  Since the last submission I have:
> 
>  - Updated the commit message for scheduling of the interrupt handler. 
>  - Hooked the touchscreen into the driver model wakeup infrastructure.
> 
> It is also available in the git repository at:
> 
>   git://opensource.wolfsonmicro.com/linux-2.6-touch upstream
> 
> Mark Brown (3):
>       wm97xx-core: Only schedule interrupt handler if not already scheduled
>       wm97xx-core: Use IRQF_SAMPLE_RANDOM
>       wm97xx-core: Support use as a wakeup source
> 
>  drivers/input/touchscreen/wm97xx-core.c |   56 ++++++++++++++++++++++++++----
>  include/linux/wm97xx.h                  |    3 ++
>  2 files changed, 51 insertions(+), 8 deletions(-)
> 

Applied the lot, thank you Mark.

-- 
Dmitry

WM97xx touchscreen updates

[Mark Brown]

This patch series updates the WM97xx touchscreen drivers with a couple
of small cleanups and the addition of support for use of the WM97xx
touchpanels to wake up from suspend.

It is also available in the git repository at:

  git://opensource.wolfsonmicro.com/linux-2.6-touch upstream

Mark Brown (3):
      wm97xx-core: Only schedule interrupt handler if not already scheduled
      wm97xx-core: Use IRQF_SAMPLE_RANDOM
      wm97xx-core: Support use as a wakeup source

 drivers/input/touchscreen/wm97xx-core.c |   39 ++++++++++++++++++++++++------
 include/linux/wm97xx.h                  |    3 ++
 2 files changed, 34 insertions(+), 8 deletions(-)

[PATCH 1/3] wm97xx-core: Only schedule interrupt handler if not already scheduled

[Mark Brown]

Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
---
 drivers/input/touchscreen/wm97xx-core.c |   14 +++++++-------
 1 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/drivers/input/touchscreen/wm97xx-core.c b/drivers/input/touchscreen/wm97xx-core.c
index 2910999..e27b1e0 100644
--- a/drivers/input/touchscreen/wm97xx-core.c
+++ b/drivers/input/touchscreen/wm97xx-core.c
@@ -328,18 +328,18 @@ static void wm97xx_pen_irq_worker(struct work_struct *work)
  *
  * We have to disable the codec interrupt in the handler because it
  * can take upto 1ms to clear the interrupt source. We schedule a task
- * in a work queue to do the actual interaction with the chip (it
- * doesn't matter if we end up reenqueing it before it is executed
- * since we don't touch the chip until it has run).  The interrupt is
- * then enabled again in the slow handler when the source has been
- * cleared.
+ * in a work queue to do the actual interaction with the chip.  The
+ * interrupt is then enabled again in the slow handler when the source
+ * has been cleared.
  */
 static irqreturn_t wm97xx_pen_interrupt(int irq, void *dev_id)
 {
 	struct wm97xx *wm = dev_id;
 
-	wm->mach_ops->irq_enable(wm, 0);
-	queue_work(wm->ts_workq, &wm->pen_event_work);
+	if (!work_pending(&wm->pen_event_work)) {
+		wm->mach_ops->irq_enable(wm, 0);
+		queue_work(wm->ts_workq, &wm->pen_event_work);
+	}
 
 	return IRQ_HANDLED;
 }
-- 
1.5.5

Re: [PATCH 1/3] wm97xx-core: Only schedule interrupt handler if not already scheduled

[Dmitry Torokhov]

Hi Mark,

On Mon, Apr 14, 2008 at 06:39:33PM +0100, Mark Brown wrote:
>  static irqreturn_t wm97xx_pen_interrupt(int irq, void *dev_id)
>  {
>  	struct wm97xx *wm = dev_id;
>  
> -	wm->mach_ops->irq_enable(wm, 0);
> -	queue_work(wm->ts_workq, &wm->pen_event_work);
> +	if (!work_pending(&wm->pen_event_work)) {
> +		wm->mach_ops->irq_enable(wm, 0);
> +		queue_work(wm->ts_workq, &wm->pen_event_work);
> +	}

Given the fact that work will not be queued if it is pending anyway
why is this change needed?

-- 
Dmitry

Re: [PATCH 1/3] wm97xx-core: Only schedule interrupt handler if not already scheduled

[Mark Brown]

On Mon, Apr 14, 2008 at 02:10:02PM -0400, Dmitry Torokhov wrote:
> On Mon, Apr 14, 2008 at 06:39:33PM +0100, Mark Brown wrote:

> > +	if (!work_pending(&wm->pen_event_work)) {
> > +		wm->mach_ops->irq_enable(wm, 0);
> > +		queue_work(wm->ts_workq, &wm->pen_event_work);
> > +	}

> Given the fact that work will not be queued if it is pending anyway
> why is this change needed?

As well as not queuing the work it ensures that the calls to
irq_enable() are balanced which helps with implementing that operation.

Re: [PATCH 1/3] wm97xx-core: Only schedule interrupt handler if not already scheduled

[Dmitry Torokhov]

On Tue, Apr 15, 2008 at 09:48:19AM +0100, Mark Brown wrote:
> On Mon, Apr 14, 2008 at 02:10:02PM -0400, Dmitry Torokhov wrote:
> > On Mon, Apr 14, 2008 at 06:39:33PM +0100, Mark Brown wrote:
> 
> > > +	if (!work_pending(&wm->pen_event_work)) {
> > > +		wm->mach_ops->irq_enable(wm, 0);
> > > +		queue_work(wm->ts_workq, &wm->pen_event_work);
> > > +	}
> 
> > Given the fact that work will not be queued if it is pending anyway
> > why is this change needed?
> 
> As well as not queuing the work it ensures that the calls to
> irq_enable() are balanced which helps with implementing that operation.

Hmm... another question then - what would fire up wm97xx_pen_interrupt()
again, before wm97xx_pen_irq_worker had a chance to run, if the very
first thing we do after getting interrupt is irq_enable(wm, 0)?

-- 
Dmitry

Re: [PATCH 1/3] wm97xx-core: Only schedule interrupt handler if not already scheduled

[Mark Brown]

On Tue, Apr 15, 2008 at 09:02:35AM -0400, Dmitry Torokhov wrote:
> On Tue, Apr 15, 2008 at 09:48:19AM +0100, Mark Brown wrote:

> > As well as not queuing the work it ensures that the calls to
> > irq_enable() are balanced which helps with implementing that operation.

> Hmm... another question then - what would fire up wm97xx_pen_interrupt()
> again, before wm97xx_pen_irq_worker had a chance to run, if the very
> first thing we do after getting interrupt is irq_enable(wm, 0)?

Normally nothing - in most systems the interrupt will not be shared
and/or the board irq_enable() operation will completely disable the IRQ
line.  The main use case would be if the interrupt is shared and the
board irq_enable() operation doesn't affect other sources, though boards
may also decide that they can do without disabling the interrupt.