[PATCH] Fix tosakbd suspend

[PATCH] Fix tosakbd suspend

[Dmitry Baryshkov]

Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
---
 drivers/input/keyboard/tosakbd.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/drivers/input/keyboard/tosakbd.c b/drivers/input/keyboard/tosakbd.c
index 3884d1e..306cbe8 100644
--- a/drivers/input/keyboard/tosakbd.c
+++ b/drivers/input/keyboard/tosakbd.c
@@ -210,6 +210,9 @@ static int tosakbd_suspend(struct platform_device *dev, pm_message_t state)
 
 	del_timer_sync(&tosakbd->timer);
 
+	PGSR1 = (PGSR1 & ~TOSA_GPIO_LOW_STROBE_BIT);
+	PGSR2 = (PGSR2 & ~TOSA_GPIO_HIGH_STROBE_BIT);
+
 	return 0;
 }
 
-- 
1.5.4.4


-- 
With best wishes
Dmitry

Re: [PATCH] Fix tosakbd suspend

[Dmitry Torokhov]

Hi Dmitry,

On Tue, Apr 01, 2008 at 01:38:01AM +0400, Dmitry Baryshkov wrote:
> Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
> ---
>  drivers/input/keyboard/tosakbd.c |    3 +++
>  1 files changed, 3 insertions(+), 0 deletions(-)
> 
> diff --git a/drivers/input/keyboard/tosakbd.c b/drivers/input/keyboard/tosakbd.c
> index 3884d1e..306cbe8 100644
> --- a/drivers/input/keyboard/tosakbd.c
> +++ b/drivers/input/keyboard/tosakbd.c
> @@ -210,6 +210,9 @@ static int tosakbd_suspend(struct platform_device *dev, pm_message_t state)
>  
>  	del_timer_sync(&tosakbd->timer);
>  

What if interrupt comes here? The whole suspend path seems racy with
interrup handler/timer.

> +	PGSR1 = (PGSR1 & ~TOSA_GPIO_LOW_STROBE_BIT);
> +	PGSR2 = (PGSR2 & ~TOSA_GPIO_HIGH_STROBE_BIT);
> +
>  	return 0;
>  }
>  

-- 
Dmitry

Re: [PATCH] Fix tosakbd suspend

[Dmitry]

Hi, Dmitry,

2008/4/1, Dmitry Torokhov <dmitry.torokhov@gmail.com>:
> Hi Dmitry,
>
>
>  On Tue, Apr 01, 2008 at 01:38:01AM +0400, Dmitry Baryshkov wrote:
>  > Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
>  > ---
>  >  drivers/input/keyboard/tosakbd.c |    3 +++
>  >  1 files changed, 3 insertions(+), 0 deletions(-)
>  >
>  > diff --git a/drivers/input/keyboard/tosakbd.c b/drivers/input/keyboard/tosakbd.c
>  > index 3884d1e..306cbe8 100644
>  > --- a/drivers/input/keyboard/tosakbd.c
>  > +++ b/drivers/input/keyboard/tosakbd.c
>  > @@ -210,6 +210,9 @@ static int tosakbd_suspend(struct platform_device *dev, pm_message_t state)
>  >
>  >       del_timer_sync(&tosakbd->timer);
>  >
>
>
> What if interrupt comes here? The whole suspend path seems racy with
>  interrup handler/timer.

What is the best way to fix it? Moving from suspend to suspend_late
seems to be a hack.
Should I call a bunch of disable_irq or just introduce an atomic flag?

>
>
>  > +     PGSR1 = (PGSR1 & ~TOSA_GPIO_LOW_STROBE_BIT);
>  > +     PGSR2 = (PGSR2 & ~TOSA_GPIO_HIGH_STROBE_BIT);
>  > +
>  >       return 0;
>  >  }
>  >
>
>  --
>
> Dmitry
>


-- 
With best wishes
Dmitry

Re: [PATCH] Fix tosakbd suspend

[Dmitry Torokhov]

On Tue, Apr 01, 2008 at 04:27:38PM +0300, Dmitry wrote:
> Hi, Dmitry,
> 
> 2008/4/1, Dmitry Torokhov <dmitry.torokhov@gmail.com>:
> > Hi Dmitry,
> >
> >
> >  On Tue, Apr 01, 2008 at 01:38:01AM +0400, Dmitry Baryshkov wrote:
> >  > Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
> >  > ---
> >  >  drivers/input/keyboard/tosakbd.c |    3 +++
> >  >  1 files changed, 3 insertions(+), 0 deletions(-)
> >  >
> >  > diff --git a/drivers/input/keyboard/tosakbd.c b/drivers/input/keyboard/tosakbd.c
> >  > index 3884d1e..306cbe8 100644
> >  > --- a/drivers/input/keyboard/tosakbd.c
> >  > +++ b/drivers/input/keyboard/tosakbd.c
> >  > @@ -210,6 +210,9 @@ static int tosakbd_suspend(struct platform_device *dev, pm_message_t state)
> >  >
> >  >       del_timer_sync(&tosakbd->timer);
> >  >
> >
> >
> > What if interrupt comes here? The whole suspend path seems racy with
> >  interrup handler/timer.
> 
> What is the best way to fix it? Moving from suspend to suspend_late
> seems to be a hack.
> Should I call a bunch of disable_irq or just introduce an atomic flag?
> 

How about we take the lock in tosakbd_suspend, set flag indicating that the
picece is suspended, do the discharge and then del_timer_sync()? If you change
tosakbd_scankeyboard() to check for suspend flag and exit as soon as it sees it
it should take care of the most of the issues I think. Something like this:

Subject: Input: tosakbd - fix suspend
From: Dmitry Baryshkov <dbaryshkov@gmail.com>

Input: tosakbd - fix suspend

Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
---
 drivers/input/keyboard/tosakbd.c |   19 ++++++++++++++++---
 1 file changed, 16 insertions(+), 3 deletions(-)

Index: linux/drivers/input/keyboard/tosakbd.c
===================================================================
--- linux.orig/drivers/input/keyboard/tosakbd.c
+++ linux/drivers/input/keyboard/tosakbd.c
@@ -52,7 +52,7 @@ KEY_X, KEY_F, KEY_SPACE, KEY_APOSTROPHE,
 struct tosakbd {
 	unsigned int keycode[ARRAY_SIZE(tosakbd_keycode)];
 	struct input_dev *input;
-
+	int suspeneded;
 	spinlock_t lock; /* protect kbd scanning */
 	struct timer_list timer;
 };
@@ -133,6 +133,9 @@ static void tosakbd_scankeyboard(struct 
 
 	spin_lock_irqsave(&tosakbd->lock, flags);
 
+	if (tosakbd->suspended)
+		goto out;
+
 	for (col = 0; col < TOSA_KEY_STROBE_NUM; col++) {
 		/*
 		 * Discharge the output driver capacitatance
@@ -174,6 +177,7 @@ static void tosakbd_scankeyboard(struct 
 	if (num_pressed)
 		mod_timer(&tosakbd->timer, jiffies + SCAN_INTERVAL);
 
+ out:
 	spin_unlock_irqrestore(&tosakbd->lock, flags);
 }
 
@@ -200,6 +204,7 @@ static irqreturn_t tosakbd_interrupt(int
 static void tosakbd_timer_callback(unsigned long __dev)
 {
 	struct platform_device *dev = (struct platform_device *)__dev;
+
 	tosakbd_scankeyboard(dev);
 }
 
@@ -207,6 +212,13 @@ static void tosakbd_timer_callback(unsig
 static int tosakbd_suspend(struct platform_device *dev, pm_message_t state)
 {
 	struct tosakbd *tosakbd = platform_get_drvdata(dev);
+	unsigned long flags;
+
+	spin_lock_irqsave(&tosakbd->lock, flags);
+	PGSR1 = (PGSR1 & ~TOSA_GPIO_LOW_STROBE_BIT);
+	PGSR2 = (PGSR2 & ~TOSA_GPIO_HIGH_STROBE_BIT);
+	tosakbd->suspended = 1;
+	spin_unlock_irqrestore(&tosakbd->lock, flags);
 
 	del_timer_sync(&tosakbd->timer);
 
@@ -215,6 +227,7 @@ static int tosakbd_suspend(struct platfo
 
 static int tosakbd_resume(struct platform_device *dev)
 {
+	tosakbd->suspended = 0;
 	tosakbd_scankeyboard(dev);
 
 	return 0;
@@ -365,8 +378,8 @@ fail:
 	return error;
 }
 
-static int __devexit tosakbd_remove(struct platform_device *dev) {
-
+static int __devexit tosakbd_remove(struct platform_device *dev)
+{
 	int i;
 	struct tosakbd *tosakbd = platform_get_drvdata(dev);
 
-- 
Dmitry

Re: [PATCH] Fix tosakbd suspend

[Dmitry]

Hi,

2008/4/1, Dmitry Torokhov <dmitry.torokhov@gmail.com>:
> On Tue, Apr 01, 2008 at 04:27:38PM +0300, Dmitry wrote:
>  > Hi, Dmitry,
>  >
>  > 2008/4/1, Dmitry Torokhov <dmitry.torokhov@gmail.com>:
>  > > Hi Dmitry,
>  > >
>  > >
>  > >  On Tue, Apr 01, 2008 at 01:38:01AM +0400, Dmitry Baryshkov wrote:
>  > >  > Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
>  > >  > ---
>  > >  >  drivers/input/keyboard/tosakbd.c |    3 +++
>  > >  >  1 files changed, 3 insertions(+), 0 deletions(-)
>  > >  >
>  > >  > diff --git a/drivers/input/keyboard/tosakbd.c b/drivers/input/keyboard/tosakbd.c
>  > >  > index 3884d1e..306cbe8 100644
>  > >  > --- a/drivers/input/keyboard/tosakbd.c
>  > >  > +++ b/drivers/input/keyboard/tosakbd.c
>  > >  > @@ -210,6 +210,9 @@ static int tosakbd_suspend(struct platform_device *dev, pm_message_t state)
>  > >  >
>  > >  >       del_timer_sync(&tosakbd->timer);
>  > >  >
>  > >
>  > >
>  > > What if interrupt comes here? The whole suspend path seems racy with
>  > >  interrup handler/timer.
>  >
>  > What is the best way to fix it? Moving from suspend to suspend_late
>  > seems to be a hack.
>  > Should I call a bunch of disable_irq or just introduce an atomic flag?
>  >
>
>
> How about we take the lock in tosakbd_suspend, set flag indicating that the
>  picece is suspended, do the discharge and then del_timer_sync()? If you change
>  tosakbd_scankeyboard() to check for suspend flag and exit as soon as it sees it
>  it should take care of the most of the issues I think. Something like this:

This should do the trick. I'll test the patch tomorrow.

-- 
With best wishes
Dmitry

[PATCH] Fix tosakbd suspend

[Dmitry Torokhov]

Hi,

2008/4/1, Dmitry Torokhov <dmitry.torokhov@gmail.com>:
> On Tue, Apr 01, 2008 at 04:27:38PM +0300, Dmitry wrote:
> > Hi, Dmitry,
> >
> > 2008/4/1, Dmitry Torokhov <dmitry.torokhov@gmail.com>:
> > > Hi Dmitry,
> > >
> > >
> > >  On Tue, Apr 01, 2008 at 01:38:01AM +0400, Dmitry Baryshkov wrote:
> > >  > Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
> > >  > ---
> > >  >  drivers/input/keyboard/tosakbd.c |    3 +++
> > >  >  1 files changed, 3 insertions(+), 0 deletions(-)
> > >  >
> > >  > diff --git a/drivers/input/keyboard/tosakbd.c b/drivers/input/keyboard/tosakbd.c
> > >  > index 3884d1e..306cbe8 100644
> > >  > --- a/drivers/input/keyboard/tosakbd.c
> > >  > +++ b/drivers/input/keyboard/tosakbd.c
> > >  > @@ -210,6 +210,9 @@ static int tosakbd_suspend(struct platform_device *dev, pm_message_t state)
> > >  >
> > >  >       del_timer_sync(&tosakbd->timer);
> > >  >
> > >
> > >
> > > What if interrupt comes here? The whole suspend path seems racy with
> > >  interrup handler/timer.
> >
> > What is the best way to fix it? Moving from suspend to suspend_late
> > seems to be a hack.
> > Should I call a bunch of disable_irq or just introduce an atomic flag?
> >
> 
> How about we take the lock in tosakbd_suspend, set flag indicating that the
> picece is suspended, do the discharge and then del_timer_sync()? If you change
> tosakbd_scankeyboard() to check for suspend flag and exit as soon as it sees it
> it should take care of the most of the issues I think. Something like this:

Fixed two minor glitches. Works for me. Please add it to your tree.

Subject: Input: tosakbd - fix suspend
Input: tosakbd - fix suspend

Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
---
 drivers/input/keyboard/tosakbd.c |   21 ++++++++++++++++++---
 1 files changed, 18 insertions(+), 3 deletions(-)

diff --git a/drivers/input/keyboard/tosakbd.c b/drivers/input/keyboard/tosakbd.c
index 3884d1e..a247006 100644
--- a/drivers/input/keyboard/tosakbd.c
+++ b/drivers/input/keyboard/tosakbd.c
@@ -52,7 +52,7 @@ KEY_X, KEY_F, KEY_SPACE, KEY_APOSTROPHE, TOSA_KEY_MAIL, KEY_LEFT, KEY_DOWN, KEY_
 struct tosakbd {
 	unsigned int keycode[ARRAY_SIZE(tosakbd_keycode)];
 	struct input_dev *input;
-
+	int suspended;
 	spinlock_t lock; /* protect kbd scanning */
 	struct timer_list timer;
 };
@@ -133,6 +133,9 @@ static void tosakbd_scankeyboard(struct platform_device *dev)
 
 	spin_lock_irqsave(&tosakbd->lock, flags);
 
+	if (tosakbd->suspended)
+		goto out;
+
 	for (col = 0; col < TOSA_KEY_STROBE_NUM; col++) {
 		/*
 		 * Discharge the output driver capacitatance
@@ -174,6 +177,7 @@ static void tosakbd_scankeyboard(struct platform_device *dev)
 	if (num_pressed)
 		mod_timer(&tosakbd->timer, jiffies + SCAN_INTERVAL);
 
+ out:
 	spin_unlock_irqrestore(&tosakbd->lock, flags);
 }
 
@@ -200,6 +204,7 @@ static irqreturn_t tosakbd_interrupt(int irq, void *__dev)
 static void tosakbd_timer_callback(unsigned long __dev)
 {
 	struct platform_device *dev = (struct platform_device *)__dev;
+
 	tosakbd_scankeyboard(dev);
 }
 
@@ -207,6 +212,13 @@ static void tosakbd_timer_callback(unsigned long __dev)
 static int tosakbd_suspend(struct platform_device *dev, pm_message_t state)
 {
 	struct tosakbd *tosakbd = platform_get_drvdata(dev);
+	unsigned long flags;
+
+	spin_lock_irqsave(&tosakbd->lock, flags);
+	PGSR1 = (PGSR1 & ~TOSA_GPIO_LOW_STROBE_BIT);
+	PGSR2 = (PGSR2 & ~TOSA_GPIO_HIGH_STROBE_BIT);
+	tosakbd->suspended = 1;
+	spin_unlock_irqrestore(&tosakbd->lock, flags);
 
 	del_timer_sync(&tosakbd->timer);
 
@@ -215,6 +227,9 @@ static int tosakbd_suspend(struct platform_device *dev, pm_message_t state)
 
 static int tosakbd_resume(struct platform_device *dev)
 {
+	struct tosakbd *tosakbd = platform_get_drvdata(dev);
+
+	tosakbd->suspended = 0;
 	tosakbd_scankeyboard(dev);
 
 	return 0;
@@ -365,8 +380,8 @@ fail:
 	return error;
 }
 
-static int __devexit tosakbd_remove(struct platform_device *dev) {
-
+static int __devexit tosakbd_remove(struct platform_device *dev)
+{
 	int i;
 	struct tosakbd *tosakbd = platform_get_drvdata(dev);
 
-- 
1.5.4.4


-- 
With best wishes
Dmitry

Re: [PATCH] Fix tosakbd suspend

[Dmitry]

Hi,

Sorry, that was me... Sometimes I hate git-am & git-format-patch

2008/4/2, Dmitry Torokhov <dbaryshkov@gmail.com>:
> Hi,
>
>
>  2008/4/1, Dmitry Torokhov <dmitry.torokhov@gmail.com>:
>  > On Tue, Apr 01, 2008 at 04:27:38PM +0300, Dmitry wrote:
>  > > Hi, Dmitry,
>  > >
>  > > 2008/4/1, Dmitry Torokhov <dmitry.torokhov@gmail.com>:
>  > > > Hi Dmitry,
>  > > >
>  > > >
>  > > >  On Tue, Apr 01, 2008 at 01:38:01AM +0400, Dmitry Baryshkov wrote:
>  > > >  > Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
>  > > >  > ---
>  > > >  >  drivers/input/keyboard/tosakbd.c |    3 +++
>  > > >  >  1 files changed, 3 insertions(+), 0 deletions(-)
>  > > >  >
>  > > >  > diff --git a/drivers/input/keyboard/tosakbd.c b/drivers/input/keyboard/tosakbd.c
>  > > >  > index 3884d1e..306cbe8 100644
>  > > >  > --- a/drivers/input/keyboard/tosakbd.c
>  > > >  > +++ b/drivers/input/keyboard/tosakbd.c
>  > > >  > @@ -210,6 +210,9 @@ static int tosakbd_suspend(struct platform_device *dev, pm_message_t state)
>  > > >  >
>  > > >  >       del_timer_sync(&tosakbd->timer);
>  > > >  >
>  > > >
>  > > >
>  > > > What if interrupt comes here? The whole suspend path seems racy with
>  > > >  interrup handler/timer.
>  > >
>  > > What is the best way to fix it? Moving from suspend to suspend_late
>  > > seems to be a hack.
>  > > Should I call a bunch of disable_irq or just introduce an atomic flag?
>  > >
>  >
>  > How about we take the lock in tosakbd_suspend, set flag indicating that the
>  > picece is suspended, do the discharge and then del_timer_sync()? If you change
>  > tosakbd_scankeyboard() to check for suspend flag and exit as soon as it sees it
>  > it should take care of the most of the issues I think. Something like this:
>
>
> Fixed two minor glitches. Works for me. Please add it to your tree.
>
>
>  Subject: Input: tosakbd - fix suspend
>
> Input: tosakbd - fix suspend
>
>  Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
>  Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
>  ---
>
>  drivers/input/keyboard/tosakbd.c |   21 ++++++++++++++++++---
>   1 files changed, 18 insertions(+), 3 deletions(-)
>
>
>  diff --git a/drivers/input/keyboard/tosakbd.c b/drivers/input/keyboard/tosakbd.c
>
> index 3884d1e..a247006 100644
>
> --- a/drivers/input/keyboard/tosakbd.c
>  +++ b/drivers/input/keyboard/tosakbd.c
>
> @@ -52,7 +52,7 @@ KEY_X, KEY_F, KEY_SPACE, KEY_APOSTROPHE, TOSA_KEY_MAIL, KEY_LEFT, KEY_DOWN, KEY_
>
>  struct tosakbd {
>         unsigned int keycode[ARRAY_SIZE(tosakbd_keycode)];
>         struct input_dev *input;
>  -
>
> +       int suspended;
>
>         spinlock_t lock; /* protect kbd scanning */
>         struct timer_list timer;
>   };
>
> @@ -133,6 +133,9 @@ static void tosakbd_scankeyboard(struct platform_device *dev)
>
>
>         spin_lock_irqsave(&tosakbd->lock, flags);
>
>  +       if (tosakbd->suspended)
>  +               goto out;
>  +
>         for (col = 0; col < TOSA_KEY_STROBE_NUM; col++) {
>                 /*
>                  * Discharge the output driver capacitatance
>
> @@ -174,6 +177,7 @@ static void tosakbd_scankeyboard(struct platform_device *dev)
>
>         if (num_pressed)
>                 mod_timer(&tosakbd->timer, jiffies + SCAN_INTERVAL);
>
>  + out:
>         spin_unlock_irqrestore(&tosakbd->lock, flags);
>   }
>
>
> @@ -200,6 +204,7 @@ static irqreturn_t tosakbd_interrupt(int irq, void *__dev)
>
>  static void tosakbd_timer_callback(unsigned long __dev)
>   {
>         struct platform_device *dev = (struct platform_device *)__dev;
>  +
>         tosakbd_scankeyboard(dev);
>   }
>
>
> @@ -207,6 +212,13 @@ static void tosakbd_timer_callback(unsigned long __dev)
>
>  static int tosakbd_suspend(struct platform_device *dev, pm_message_t state)
>   {
>         struct tosakbd *tosakbd = platform_get_drvdata(dev);
>  +       unsigned long flags;
>  +
>  +       spin_lock_irqsave(&tosakbd->lock, flags);
>  +       PGSR1 = (PGSR1 & ~TOSA_GPIO_LOW_STROBE_BIT);
>  +       PGSR2 = (PGSR2 & ~TOSA_GPIO_HIGH_STROBE_BIT);
>  +       tosakbd->suspended = 1;
>  +       spin_unlock_irqrestore(&tosakbd->lock, flags);
>
>         del_timer_sync(&tosakbd->timer);
>
>
> @@ -215,6 +227,9 @@ static int tosakbd_suspend(struct platform_device *dev, pm_message_t state)
>
>
>   static int tosakbd_resume(struct platform_device *dev)
>   {
>
> +       struct tosakbd *tosakbd = platform_get_drvdata(dev);
>
> +
>  +       tosakbd->suspended = 0;
>         tosakbd_scankeyboard(dev);
>
>         return 0;
>
> @@ -365,8 +380,8 @@ fail:
>
>         return error;
>   }
>
>  -static int __devexit tosakbd_remove(struct platform_device *dev) {
>  -
>  +static int __devexit tosakbd_remove(struct platform_device *dev)
>  +{
>         int i;
>         struct tosakbd *tosakbd = platform_get_drvdata(dev);
>
>  --
>
> 1.5.4.4
>
>
>  --
>  With best wishes
>  Dmitry
>
>


-- 
With best wishes
Dmitry

Re: [PATCH] Fix tosakbd suspend

[Dmitry Torokhov]

On Wed, Apr 02, 2008 at 06:38:26PM +0400, Dmitry wrote:
> Hi,
> 
> Sorry, that was me... Sometimes I hate git-am & git-format-patch
> 

Heh, you had me scratching my head for a moment ;)

> 2008/4/2, <dbaryshkov@gmail.com>:
> >
> > Fixed two minor glitches. Works for me. Please add it to your tree.
> >

What were they? I dont see any difference from the patch I sent to you,
so I want to make sure I have the correct version.

-- 
Dmitry

Re: [PATCH] Fix tosakbd suspend

[Dmitry]

Hi,

2008/4/2, Dmitry Torokhov <dmitry.torokhov@gmail.com>:
> On Wed, Apr 02, 2008 at 06:38:26PM +0400, Dmitry wrote:
>  > Hi,
>  >
>  > Sorry, that was me... Sometimes I hate git-am & git-format-patch
>  >
>
>
> Heh, you had me scratching my head for a moment ;)

:)

>
>  > 2008/4/2, <dbaryshkov@gmail.com>:
>
> > >
>  > > Fixed two minor glitches. Works for me. Please add it to your tree.
>  > >
>
>
> What were they? I dont see any difference from the patch I sent to you,
>  so I want to make sure I have the correct version.

suspended instead of suspeneded in struct tosakbd definition,
declaration of tosakbd in tosakbd_resume.

-- 
With best wishes
Dmitry