[PATCH] forced recalibration for the OLPC hgpk touchpad

[PATCH] forced recalibration for the OLPC hgpk touchpad

[pgf]

The OLPC XO laptop incorporates a combination touchpad/tablet
device which unfortunately requires frequent recalibration.  The
driver will force this automatically when various suspicious
behaviors are observed, and the user can recalibrate manually
(with a special keyboard sequence).  There's currently no way,
however, for an external program to cause recalibration.

This patch creates a new node in /sys which, when written with '1',
will force a touchpad recalibration.  No other writes (or reads)
of this node are supported.

paul

--------------------


Signed-off-by: Paul Fox <pgf@foxharp.boston.ma.us>

diff --git a/drivers/input/mouse/hgpk.c b/drivers/input/mouse/hgpk.c
index a1ad2f1..e736ebd 100644
--- a/drivers/input/mouse/hgpk.c
+++ b/drivers/input/mouse/hgpk.c
@@ -369,12 +369,40 @@ static ssize_t hgpk_set_powered(struct psmouse *psmouse, void *data,
 __PSMOUSE_DEFINE_ATTR(powered, S_IWUSR | S_IRUGO, NULL,
 		      hgpk_show_powered, hgpk_set_powered, 0);
 
+static ssize_t hgpk_trigger_recal_show(struct psmouse *psmouse,
+		void *data, char *buf)
+{
+	return -EINVAL;
+}
+
+static ssize_t hgpk_trigger_recal(struct psmouse *psmouse, void *data,
+				const char *buf, size_t count)
+{
+	struct hgpk_data *priv = psmouse->private;
+	unsigned long value;
+	int err;
+
+	err = strict_strtoul(buf, 10, &value);
+	if (err || value != 1)
+		return -EINVAL;
+
+	psmouse_queue_work(psmouse, &priv->recalib_wq,
+				msecs_to_jiffies(1));
+
+	return count;
+}
+
+__PSMOUSE_DEFINE_ATTR(recalibrate, S_IWUSR | S_IRUGO, NULL,
+		      hgpk_trigger_recal_show, hgpk_trigger_recal, 0);
+
 static void hgpk_disconnect(struct psmouse *psmouse)
 {
 	struct hgpk_data *priv = psmouse->private;
 
 	device_remove_file(&psmouse->ps2dev.serio->dev,
 			   &psmouse_attr_powered.dattr);
+	device_remove_file(&psmouse->ps2dev.serio->dev,
+			   &psmouse_attr_recalibrate.dattr);
 	psmouse_reset(psmouse);
 	kfree(priv);
 }
@@ -423,8 +451,18 @@ static int hgpk_register(struct psmouse *psmouse)
 
 	err = device_create_file(&psmouse->ps2dev.serio->dev,
 				 &psmouse_attr_powered.dattr);
-	if (err)
-		hgpk_err(psmouse, "Failed to create sysfs attribute\n");
+	if (err) {
+		hgpk_err(psmouse, "Failed creating 'powered' sysfs node\n");
+	} else {
+		err = device_create_file(&psmouse->ps2dev.serio->dev,
+					 &psmouse_attr_recalibrate.dattr);
+		if (err) {
+			hgpk_err(psmouse,
+				"Failed creating 'recalibrate' sysfs node\n");
+			device_remove_file(&psmouse->ps2dev.serio->dev,
+					&psmouse_attr_powered.dattr);
+		}
+	}
 
 	return err;
 }



=---------------------
 paul fox, pgf@laptop.org

Re: [PATCH] forced recalibration for the OLPC hgpk touchpad

[Dmitry Torokhov]

Hi Paul,

On Tuesday 02 June 2009 12:47:59 pgf@laptop.org wrote:
> The OLPC XO laptop incorporates a combination touchpad/tablet
> device which unfortunately requires frequent recalibration.  The
> driver will force this automatically when various suspicious
> behaviors are observed, and the user can recalibrate manually
> (with a special keyboard sequence).  There's currently no way,
> however, for an external program to cause recalibration.
>
> This patch creates a new node in /sys which, when written with '1',
> will force a touchpad recalibration.  No other writes (or reads)
> of this node are supported.
>


Instead of creating a new sysfs attribute maybe we should make
the touchpad recalibrate upon resume (reconnect)? Userspace can
already request reconnect via sysfs.
 
-- 
Dmitry

Re: [PATCH] forced recalibration for the OLPC hgpk touchpad

[Paul Fox]

dmitry wrote:
 > Hi Paul,
 > 
 > On Tuesday 02 June 2009 12:47:59 pgf@laptop.org wrote:
 > > The OLPC XO laptop incorporates a combination touchpad/tablet
 > > device which unfortunately requires frequent recalibration.  The
 > > driver will force this automatically when various suspicious
 > > behaviors are observed, and the user can recalibrate manually
 > > (with a special keyboard sequence).  There's currently no way,
 > > however, for an external program to cause recalibration.
 > >
 > > This patch creates a new node in /sys which, when written with '1',
 > > will force a touchpad recalibration.  No other writes (or reads)
 > > of this node are supported.
 > >
 > 
 > 
 > Instead of creating a new sysfs attribute maybe we should make
 > the touchpad recalibrate upon resume (reconnect)? Userspace can
 > already request reconnect via sysfs.

dmitry -- thanks for the suggestion.

i took a look, and tried it.  the problem is that doing a full
reset of the touchpad takes quite a long time -- 1.1 or 1.2
seconds.  recalibration deprives the user of their touchpad for
long enough as it is -- i don't think we can afford that time.
(remember that this is a workaround for bad hardware -- the XO
no longer uses this touchpad at least partly due to this very
issue, but there are enough in use that we're still trying to
improve things.)

paul
=---------------------
 paul fox, pgf@laptop.org

Re: [PATCH] forced recalibration for the OLPC hgpk touchpad

[Andres Salomon]

On Wed, 03 Jun 2009 23:58:33 -0400
Paul Fox <pgf@laptop.org> wrote:

> dmitry wrote:
>  > Hi Paul,
>  > 
>  > On Tuesday 02 June 2009 12:47:59 pgf@laptop.org wrote:
>  > > The OLPC XO laptop incorporates a combination touchpad/tablet
>  > > device which unfortunately requires frequent recalibration.  The
>  > > driver will force this automatically when various suspicious
>  > > behaviors are observed, and the user can recalibrate manually
>  > > (with a special keyboard sequence).  There's currently no way,
>  > > however, for an external program to cause recalibration.
>  > >
>  > > This patch creates a new node in /sys which, when written with
>  > > '1', will force a touchpad recalibration.  No other writes (or
>  > > reads) of this node are supported.
>  > >
>  > 
>  > 
>  > Instead of creating a new sysfs attribute maybe we should make
>  > the touchpad recalibrate upon resume (reconnect)? Userspace can
>  > already request reconnect via sysfs.
> 
> dmitry -- thanks for the suggestion.
> 
> i took a look, and tried it.  the problem is that doing a full
> reset of the touchpad takes quite a long time -- 1.1 or 1.2
> seconds.  recalibration deprives the user of their touchpad for
> long enough as it is -- i don't think we can afford that time.
> (remember that this is a workaround for bad hardware -- the XO
> no longer uses this touchpad at least partly due to this very
> issue, but there are enough in use that we're still trying to
> improve things.)
> 

I do like the idea of having a sysfs trigger for recalibrations.  We've
never been successful in having the kernel driver handle recalibrations
seamlessly.  Perhaps touchpad hardware screwups would be easier to
detect in userspace, given a daemon that can keep a history of the last
N touchpad coordinates?

Re: [PATCH] forced recalibration for the OLPC hgpk touchpad

[Paul Fox]

hi dmitry --

have you had a chance to think any more about this?  i'm open to
suggestions on implementation detail, but i think that a full
reset of the device won't really work, because of the time it
adds to the recalibration.

paul

andres wrote:
 > On Wed, 03 Jun 2009 23:58:33 -0400
 > Paul Fox <pgf@laptop.org> wrote:
 > 
 > > dmitry wrote:
 > >  > Hi Paul,
 > >  > 
 > >  > On Tuesday 02 June 2009 12:47:59 pgf@laptop.org wrote:
 > >  > > The OLPC XO laptop incorporates a combination touchpad/tablet
 > >  > > device which unfortunately requires frequent recalibration.  The
 > >  > > driver will force this automatically when various suspicious
 > >  > > behaviors are observed, and the user can recalibrate manually
 > >  > > (with a special keyboard sequence).  There's currently no way,
 > >  > > however, for an external program to cause recalibration.
 > >  > >
 > >  > > This patch creates a new node in /sys which, when written with
 > >  > > '1', will force a touchpad recalibration.  No other writes (or
 > >  > > reads) of this node are supported.
 > >  > >
 > >  > 
 > >  > 
 > >  > Instead of creating a new sysfs attribute maybe we should make
 > >  > the touchpad recalibrate upon resume (reconnect)? Userspace can
 > >  > already request reconnect via sysfs.
 > > 
 > > dmitry -- thanks for the suggestion.
 > > 
 > > i took a look, and tried it.  the problem is that doing a full
 > > reset of the touchpad takes quite a long time -- 1.1 or 1.2
 > > seconds.  recalibration deprives the user of their touchpad for
 > > long enough as it is -- i don't think we can afford that time.
 > > (remember that this is a workaround for bad hardware -- the XO
 > > no longer uses this touchpad at least partly due to this very
 > > issue, but there are enough in use that we're still trying to
 > > improve things.)
 > > 
 > 
 > I do like the idea of having a sysfs trigger for recalibrations.  We've
 > never been successful in having the kernel driver handle recalibrations
 > seamlessly.  Perhaps touchpad hardware screwups would be easier to
 > detect in userspace, given a daemon that can keep a history of the last
 > N touchpad coordinates?

=---------------------
 paul fox, pgf@laptop.org

Re: [PATCH] forced recalibration for the OLPC hgpk touchpad

[Paul Fox]

this is a resubmission, in the hopes of jump-starting the merge
discussion.  it was last discussed in early june of this year.

the OLPC XO laptop incorporates a combination touchpad/tablet
device which unfortunately requires frequent recalibration.  the
driver will force this automatically when various suspicious
behaviors are observed, and the user can recalibrate manually
(with a special keyboard sequence).  there's currently no way,
however, for an external program to cause recalibration.

this patch creates a new node in /sys which, when written with '1',
will force a touchpad recalibration.  no other writes (or reads)
of this node are supported.

during the previous discussion, dmitry suggested that we instead
use the reconnect capability which is already available in /sys. 
i experimented with this, and unfortunately a full reset of
the touchpad takes quite a long time -- 1.1 or 1.2 seconds. 
recalibration deprives the user of their touchpad for long enough
as it is -- i don't think we can afford that time.  (note
that this is a workaround for bad hardware -- the XO no longer
uses this touchpad at least partly due to this very issue, but
there are enough in use that we're still trying to improve
things.)

paul

Signed-off-by: Paul Fox <pgf@laptop.org>
Acked-by: Andres Salomon <dilinger@collabora.co.uk>

diff --git a/drivers/input/mouse/hgpk.c b/drivers/input/mouse/hgpk.c
index a1ad2f1..e736ebd 100644
--- a/drivers/input/mouse/hgpk.c
+++ b/drivers/input/mouse/hgpk.c
@@ -369,12 +369,40 @@ static ssize_t hgpk_set_powered(struct psmouse *psmouse, void *data,
 __PSMOUSE_DEFINE_ATTR(powered, S_IWUSR | S_IRUGO, NULL,
 		      hgpk_show_powered, hgpk_set_powered, 0);
 
+static ssize_t hgpk_trigger_recal_show(struct psmouse *psmouse,
+		void *data, char *buf)
+{
+	return -EINVAL;
+}
+
+static ssize_t hgpk_trigger_recal(struct psmouse *psmouse, void *data,
+				const char *buf, size_t count)
+{
+	struct hgpk_data *priv = psmouse->private;
+	unsigned long value;
+	int err;
+
+	err = strict_strtoul(buf, 10, &value);
+	if (err || value != 1)
+		return -EINVAL;
+
+	psmouse_queue_work(psmouse, &priv->recalib_wq,
+				msecs_to_jiffies(1));
+
+	return count;
+}
+
+__PSMOUSE_DEFINE_ATTR(recalibrate, S_IWUSR | S_IRUGO, NULL,
+		      hgpk_trigger_recal_show, hgpk_trigger_recal, 0);
+
 static void hgpk_disconnect(struct psmouse *psmouse)
 {
 	struct hgpk_data *priv = psmouse->private;
 
 	device_remove_file(&psmouse->ps2dev.serio->dev,
 			   &psmouse_attr_powered.dattr);
+	device_remove_file(&psmouse->ps2dev.serio->dev,
+			   &psmouse_attr_recalibrate.dattr);
 	psmouse_reset(psmouse);
 	kfree(priv);
 }
@@ -423,8 +451,18 @@ static int hgpk_register(struct psmouse *psmouse)
 
 	err = device_create_file(&psmouse->ps2dev.serio->dev,
 				 &psmouse_attr_powered.dattr);
-	if (err)
-		hgpk_err(psmouse, "Failed to create sysfs attribute\n");
+	if (err) {
+		hgpk_err(psmouse, "Failed creating 'powered' sysfs node\n");
+	} else {
+		err = device_create_file(&psmouse->ps2dev.serio->dev,
+					 &psmouse_attr_recalibrate.dattr);
+		if (err) {
+			hgpk_err(psmouse,
+				"Failed creating 'recalibrate' sysfs node\n");
+			device_remove_file(&psmouse->ps2dev.serio->dev,
+					&psmouse_attr_powered.dattr);
+		}
+	}
 
 	return err;
 }
=---------------------
 paul fox, pgf@laptop.org

Re: [PATCH] forced recalibration for the OLPC hgpk touchpad

[Dmitry Torokhov]

Hi Paul,

On Wed, Jul 29, 2009 at 06:08:50PM -0400, Paul Fox wrote:
> this is a resubmission, in the hopes of jump-starting the merge
> discussion.  it was last discussed in early june of this year.
> 
> the OLPC XO laptop incorporates a combination touchpad/tablet
> device which unfortunately requires frequent recalibration.  the
> driver will force this automatically when various suspicious
> behaviors are observed, and the user can recalibrate manually
> (with a special keyboard sequence).  there's currently no way,
> however, for an external program to cause recalibration.
> 
> this patch creates a new node in /sys which, when written with '1',
> will force a touchpad recalibration.  no other writes (or reads)
> of this node are supported.
> 
> during the previous discussion, dmitry suggested that we instead
> use the reconnect capability which is already available in /sys. 
> i experimented with this, and unfortunately a full reset of
> the touchpad takes quite a long time -- 1.1 or 1.2 seconds. 
> recalibration deprives the user of their touchpad for long enough
> as it is -- i don't think we can afford that time.  (note
> that this is a workaround for bad hardware -- the XO no longer
> uses this touchpad at least partly due to this very issue, but
> there are enough in use that we're still trying to improve
> things.)
> 

Sorry, I completely forgot about this patch... I am going to apply
it, however I think that we should not be creating the attribute
if the touchpad does not support recalibration. Also, I don't see
the reason why we can't schedule recalibration immediately instead
of waiting for 1 msec.

Could you please try updated version of the patch below?

Thanks!

-- 
Dmitry


Input: hgpk - forced recalibration for the OLPC touchpad

From: Paul Fox <pgf@laptop.org>

The OLPC XO laptop incorporates a combination touchpad/tablet device
which unfortunately requires frequent recalibration.  The driver will
force this automatically when various suspicious behaviors are
observed, and the user can recalibrate manually (with a special
keyboard sequence). There's currently no way, however, for an external
program to cause recalibration. We can not use the reconnect
capability which is already available in /sys because full reset of
the touchpad takes 1.1 - 1.2 secons which is too long.

This patch creates a new node in /sys which, when written with '1',
will force a touchpad recalibration; no other writes (or reads)
of this node are supported.

Signed-off-by: Paul Fox <pgf@laptop.org>
Acked-by: Andres Salomon <dilinger@collabora.co.uk>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
---

 drivers/input/mouse/hgpk.c |   55 ++++++++++++++++++++++++++++++++++++++++++--
 1 files changed, 52 insertions(+), 3 deletions(-)


diff --git a/drivers/input/mouse/hgpk.c b/drivers/input/mouse/hgpk.c
index a1ad2f1..f5aa035 100644
--- a/drivers/input/mouse/hgpk.c
+++ b/drivers/input/mouse/hgpk.c
@@ -369,12 +369,46 @@ static ssize_t hgpk_set_powered(struct psmouse *psmouse, void *data,
 __PSMOUSE_DEFINE_ATTR(powered, S_IWUSR | S_IRUGO, NULL,
 		      hgpk_show_powered, hgpk_set_powered, 0);
 
+static ssize_t hgpk_trigger_recal_show(struct psmouse *psmouse,
+		void *data, char *buf)
+{
+	return -EINVAL;
+}
+
+static ssize_t hgpk_trigger_recal(struct psmouse *psmouse, void *data,
+				const char *buf, size_t count)
+{
+	struct hgpk_data *priv = psmouse->private;
+	unsigned long value;
+	int err;
+
+	err = strict_strtoul(buf, 10, &value);
+	if (err || value != 1)
+		return -EINVAL;
+
+	/*
+	 * We queue work instead of doing recalibration right here
+	 * to avoid adding locking to to hgpk_force_recalibrate()
+	 * since workqueue provides serialization.
+	 */
+	psmouse_queue_work(psmouse, &priv->recalib_wq, 0);
+	return count;
+}
+
+__PSMOUSE_DEFINE_ATTR(recalibrate, S_IWUSR | S_IRUGO, NULL,
+		      hgpk_trigger_recal_show, hgpk_trigger_recal, 0);
+
 static void hgpk_disconnect(struct psmouse *psmouse)
 {
 	struct hgpk_data *priv = psmouse->private;
 
 	device_remove_file(&psmouse->ps2dev.serio->dev,
 			   &psmouse_attr_powered.dattr);
+
+	if (psmouse->model >= HGPK_MODEL_C)
+		device_remove_file(&psmouse->ps2dev.serio->dev,
+				   &psmouse_attr_recalibrate.dattr);
+
 	psmouse_reset(psmouse);
 	kfree(priv);
 }
@@ -423,10 +457,25 @@ static int hgpk_register(struct psmouse *psmouse)
 
 	err = device_create_file(&psmouse->ps2dev.serio->dev,
 				 &psmouse_attr_powered.dattr);
-	if (err)
-		hgpk_err(psmouse, "Failed to create sysfs attribute\n");
+	if (err) {
+		hgpk_err(psmouse, "Failed creating 'powered' sysfs node\n");
+		return err;
+	}
 
-	return err;
+	/* C-series touchpads added the recalibrate command */
+	if (psmouse->model >= HGPK_MODEL_C) {
+		err = device_create_file(&psmouse->ps2dev.serio->dev,
+					 &psmouse_attr_recalibrate.dattr);
+		if (err) {
+			hgpk_err(psmouse,
+				"Failed creating 'recalibrate' sysfs node\n");
+			device_remove_file(&psmouse->ps2dev.serio->dev,
+					&psmouse_attr_powered.dattr);
+			return err;
+		}
+	}
+
+	return 0;
 }
 
 int hgpk_init(struct psmouse *psmouse)

Re: [PATCH] forced recalibration for the OLPC hgpk touchpad

[Paul Fox]

hi dmitry --

dmitry wrote:
 > Sorry, I completely forgot about this patch... I am going to apply
 > it, however I think that we should not be creating the attribute
 > if the touchpad does not support recalibration. Also, I don't see
 > the reason why we can't schedule recalibration immediately instead
 > of waiting for 1 msec.
 > 
 > Could you please try updated version of the patch below?

looks and works fine.  thanks!

paul
=---------------------
 paul fox, pgf@laptop.org