[PATCH] input: psmouse: fix input_dev leak in lifebook driver

The lifebook driver may register a second input device, but it never
unregisters it.  This fixes that.

Signed-off-by: Andres Salomon <redacted>
---
 drivers/input/mouse/lifebook.c |    7 ++++++-
 1 files changed, 6 insertions(+), 1 deletions(-)

diff --git a/drivers/input/mouse/lifebook.c b/drivers/input/mouse/lifebook.c
index 9ec57d8..df81b0a 100644
--- a/drivers/input/mouse/lifebook.c
+++ b/drivers/input/mouse/lifebook.c

@@ -225,8 +225,13 @@ static void lifebook_set_resolution(struct psmouse *psmouse, unsigned int resolu
 
 static void lifebook_disconnect(struct psmouse *psmouse)
 {
+	struct lifebook_data *priv = psmouse->private;
+
 	psmouse_reset(psmouse);
-	kfree(psmouse->private);
+	if (priv) {
+		input_unregister_device(priv->dev2);
+		kfree(priv);
+	}
 	psmouse->private = NULL;
 }
 

-- 
1.5.3.5

On Sat, Jan 12, 2008 at 03:12:52PM -0500, Andres Salomon wrote:

The lifebook driver may register a second input device, but it never
unregisters it.  This fixes that.

Signed-off-by: Andres Salomon <redacted>

Applied, thank you Andres.

-- 
Dmitry

On Tue, 15 Jan 2008 17:04:01 -0500
Dmitry Torokhov [off-list ref] wrote:

On Sat, Jan 12, 2008 at 03:12:52PM -0500, Andres Salomon wrote:

quoted

The lifebook driver may register a second input device, but it never
unregisters it.  This fixes that.

Signed-off-by: Andres Salomon <redacted>

Applied, thank you Andres.

Hi Dmitry,
 
There's one additional patch, I don't know if you saw it; the subject
was "check return value of input_register_device() in hil_ptr.c's init".


Also, I've found myself needing to provide my own device_attribute for
the OLPC psmouse driver; I couldn't use PSMOUSE_DEFINE_ATTR(), as its
set_helper callback calls psmouse_disable.

static DEVICE_ATTR(powered, S_IWUSR | S_IRUGO, hgpk_show_powered,
                hgpk_set_powered);

Unfortunately, in order to not be racy, hgpk_set_powered needs to
deal with psmouse_mutex.  Which method of dealing with this would
you prefer?  I could either make psmouse_mutex no longer static,
provide psmouse_mutex locking functions (declared in psmouse.h), or
put psmouse_mutex into 'struct psmouse'.  Or, if you have any other
ideas..

On Wed, Jan 16, 2008 at 12:58:01PM -0500, Andres Salomon wrote:

On Tue, 15 Jan 2008 17:04:01 -0500
Dmitry Torokhov [off-list ref] wrote:

quoted

On Sat, Jan 12, 2008 at 03:12:52PM -0500, Andres Salomon wrote:

quoted

The lifebook driver may register a second input device, but it never
unregisters it.  This fixes that.

Signed-off-by: Andres Salomon <redacted>

Applied, thank you Andres.

Hi Dmitry,
 
There's one additional patch, I don't know if you saw it; the subject
was "check return value of input_register_device() in hil_ptr.c's init".

I have seen it but postponed for later. Right now I am trying to get though
2.6.24 material.

Also, I've found myself needing to provide my own device_attribute for
the OLPC psmouse driver; I couldn't use PSMOUSE_DEFINE_ATTR(), as its
set_helper callback calls psmouse_disable.

static DEVICE_ATTR(powered, S_IWUSR | S_IRUGO, hgpk_show_powered,
                hgpk_set_powered);

Unfortunately, in order to not be racy, hgpk_set_powered needs to
deal with psmouse_mutex.  Which method of dealing with this would
you prefer?  I could either make psmouse_mutex no longer static,
provide psmouse_mutex locking functions (declared in psmouse.h), or
put psmouse_mutex into 'struct psmouse'.  Or, if you have any other
ideas..

Could you just send me the code? I tend to think better when I see it...

Thanks!

-- 
Dmitry