psmouse: "bad data from KBD - bad parity" every 2ms with synaptics touchpad

psmouse: "bad data from KBD - bad parity" every 2ms with synaptics touchpad

[Jonathan Nieder]

(cc-ing some people in case they have clues)
Hi Dmitry et al,

Arvind K explains at [1]:

He owns a Synaptics touchpad and has been using it for a while.  Every
once in a while it stops responding.  Running "modprobe -r psmouse &&
modprobe psmouse" gets it working again.

On investigation, we learn that the message

	psmouse.c: bad data from KBC - bad parity

appears in dmesg approximately every 2 milliseconds when this happens.

The touchpad worked without trouble since this system was installed,
in February.  The bug only surfaced recently, when he changed the X
configuration somehow.

Userland is Debian 6.0.x (squeeze).

The problem is reproducible with:

 - Debian's 2.6.32-38 (which is based on v2.6.32.43)
 - 2.6.39-3~bpo60+1 (based on v2.6.39.2)
 - 3.1.0~rc7-1~experimental.1 (very close to v3.1-rc7)

Hardware:

	Synaptics Touchpad, model: 1, fw: 7.2, id: 0x1c0b1, caps: 0xd04731/0xa40000/0xa0000

More details are in the original report at [1].  Any ideas for
tracking this down?

Thanks for keeping the input subsystem in good shape.
Jonathan

[1] http://bugs.debian.org/647136

Re: psmouse: "bad data from KBD - bad parity" every 2ms with synaptics touchpad

[Jonathan Nieder]

Damjan Jovanovic wrote:

> Maybe that touchpad doesn't send proper parity. See what we did in commit
> 6b9d363c49d22395d0cf8729c5963f83cfbb6d69, and try to disable parity
> checking for that touchpad the same way?

Sure, though it would be nice to understand the problem better first.

So:

 - Arvind, could you say a little more about when this started
   happening[1]?  Did you change your configuration, or was there any
   upgrade around the same time (/var/log/dpkg.log* should say)?  Any
   idea why this wasn't always happening from the start?

 - Others, any idea what information would be useful to track this
   further?  Are there tools or patches for tracing what the kernel
   gets told by the touchpad, so we could see the nature of these
   parity errors?

 - Is there a dump or cache of bugzilla.kernel.org available anywhere?
   I would like to look at bug#6105 (referred to by 6b9d363c) in case
   it points to some hints or useful debugging tools.

Please forgive my laziness.
Jonathan

[1] http://bugs.debian.org/647136

Re: psmouse: "bad data from KBD - bad parity" every 2ms with synaptics touchpad

[Dmitry Torokhov]

On Tue, Nov 01, 2011 at 02:34:58PM -0500, Jonathan Nieder wrote:
> Damjan Jovanovic wrote:
> 
> > Maybe that touchpad doesn't send proper parity. See what we did in commit
> > 6b9d363c49d22395d0cf8729c5963f83cfbb6d69, and try to disable parity
> > checking for that touchpad the same way?
> 
> Sure, though it would be nice to understand the problem better first.
> 
> So:
> 
>  - Arvind, could you say a little more about when this started
>    happening[1]?  Did you change your configuration, or was there any
>    upgrade around the same time (/var/log/dpkg.log* should say)?  Any
>    idea why this wasn't always happening from the start?

If this was working and then stopped I'd try removing the keyboard and
cleaning and reseating the touchpad connector.

> 
>  - Others, any idea what information would be useful to track this
>    further?  Are there tools or patches for tracing what the kernel
>    gets told by the touchpad, so we could see the nature of these
>    parity errors?

Booting with i8042.debug=1 or doing
"echo 1 > /sys/module/i8042/parameters/debug" will dump all data from
the keyboard controller to dmesg. It is quite a bit though (80 pps @ 6
bytes/packet).

-- 
Dmitry

Re: psmouse: "bad data from KBD - bad parity" every 2ms with synaptics touchpad

[Henrik Rydberg]

On Tue, Nov 01, 2011 at 02:34:58PM -0500, Jonathan Nieder wrote:
> Damjan Jovanovic wrote:
> 
> > Maybe that touchpad doesn't send proper parity. See what we did in commit
> > 6b9d363c49d22395d0cf8729c5963f83cfbb6d69, and try to disable parity
> > checking for that touchpad the same way?
> 
> Sure, though it would be nice to understand the problem better first.
> 
> So:
> 
>  - Arvind, could you say a little more about when this started
>    happening[1]?  Did you change your configuration, or was there any
>    upgrade around the same time (/var/log/dpkg.log* should say)?  Any
>    idea why this wasn't always happening from the start?
> 
>  - Others, any idea what information would be useful to track this
>    further?  Are there tools or patches for tracing what the kernel
>    gets told by the touchpad, so we could see the nature of these
>    parity errors?
> 
>  - Is there a dump or cache of bugzilla.kernel.org available anywhere?
>    I would like to look at bug#6105 (referred to by 6b9d363c) in case
>    it points to some hints or useful debugging tools.

I would check for spurious DKMS packages as well.

Henrik