On Sunday 18 January 2015 08:22:45 Dmitry Torokhov wrote:
> On Sat, Jan 17, 2015 at 11:01:56AM +0100, Pali Rohár wrote:
> > On Thursday 15 January 2015 22:02:16 Dmitry Torokhov wrote:
> > > On Thu, Jan 15, 2015 at 09:28:41PM +0100, Pali Rohár wrote:
> > > > On Thursday 15 January 2015 20:38:18 Dmitry Torokhov 
wrote:
> > > > > On Thu, Jan 15, 2015 at 08:19:59PM +0100, Pali Rohár 
wrote:
> > > > > > On Thursday 15 January 2015 19:18:20 Dmitry Torokhov
> > 
> > wrote:
> > > > > > > On Thu, Jan 15, 2015 at 11:49:32AM +0100, Pali
> > > > > > > Rohár
> > 
> > wrote:
> > > > > > > > On Wednesday 14 January 2015 23:55:48 Dmitry
> > > > > > > > Torokhov
> > > > 
> > > > wrote:
> > > > > > > > > Hi Pali,
> > > > > > > > > 
> > > > > > > > > This series try to address the issue you
> > > > > > > > > brought regarding trackstick initialization
> > > > > > > > > on Dell Latitudes in a different way than the
> > > > > > > > > patches you proposed. Basically in this
> > > > > > > > > series we move resetting and all detection in
> > > > > > > > > alps_detect() and make sure we keep the state
> > > > > > > > > so alps_init() can reuse it and not perform
> > > > > > > > > the detection all over again. Doing this
> > > > > > > > > allows us to set up device characteristics
> > > > > > > > > (name, version, etc) properly from the get go
> > > > > > > > > while still performing reset only once.
> > > > > > > > > 
> > > > > > > > > This is untested as I do not have any ALPS
> > > > > > > > > devices anymore so I'd appreciate you giving
> > > > > > > > > it a spin.
> > > > > > > > > 
> > > > > > > > > Thanks!
> > > > > > > > 
> > > > > > > > Hi Dmitry,
> > > > > > > > 
> > > > > > > > on top of which branch/repository should I apply
> > > > > > > > your patches?
> > > > > > > 
> > > > > > > Should be applicable to my 'next' branch (which I
> > > > > > > just upreved to 3.19-rc4).
> > > > > > > 
> > > > > > > Thanks.
> > > > > > 
> > > > > > Not working at top of next (0c3e994).
> > > > > > 
> > > > > > Applying: Input: ALPS - renumber protocol numbers
> > > > > > Applying: Input: ALPS - make Rushmore a separate
> > > > > > protocol error: patch failed:
> > > > > > drivers/input/mouse/alps.c:1275 error:
> > > > > > drivers/input/mouse/alps.c: patch does not apply
> > > > > > Patch failed at 0002 Input: ALPS - make Rushmore a
> > > > > > separate protocol
> > > > > 
> > > > > Hmm.. I created a new alps branch (based on 3.19-rc4),
> > > > > can you try it?
> > > > > 
> > > > > Thanks.
> > > > 
> > > > Compiled from your new alps branch (with "if (!priv)"
> > > > fix) and modprobing psmouse.ko caused laptop freeze :-(
> > > > Even sysrq not responded. So something is not
> > > > working...
> > > 
> > > Hmm, is it on text console or in X? Any chance you could
> > > go through pathes - there are only 8 of them including 2
> > > of yours that should be unmodified.
> > > 
> > > Thanks.
> > 
> > Hi, now I tested patch by patch and kernel crash is caused
> > only by last patch 6/6 and only after I touch touchpad or
> > trackstick.
> > 
> > In text console it prints lot of panic messages and because
> > it prints lot of messages I cannot read (or record) more
> > then last.
> > 
> > In last call trace I see that alps_register_bare_ps2_mouse()
> > was called and it generated page_fault.
> 
> That happens because while you added priv->psmouse pointer it
> looks like you forgot to initialize it and I missed that
> too...
> 

Right your patch 6/6 does not initialize priv->psmouse. I looked 
into my original patch and it initialize it, so there was some
copy-paste error.

Look at other emails... can you fix problems and send new version 
of your patches for testing?

> > I do not understand why that
> > function was ever called (I do not have connected any PS/2
> > mouse which can generate bare 3 bytes PS/2 packet) and also
> > why that function caused page fault.
> 
> Hmm... my memory might be hazy but I seem to recall that
> trackstick data can come as either dedicated packets or bare
> PS/2 format, I am not sure if we can actually split the data
> streams into 2 devices.
> 
> I do not have boxes with ALPS devices though so I am unable to
> experiment.
> 
> Thanks.

From documentation and also source code it looks like trackstick 
data are reported in ALPS 6bytes packets and not in bare 3bytes.

3 months ago Vadim Klishko (CCed) wrote me:

"I don't think there is an Alps device that would mix 3-byte and 
6-byte packets."

-- 
Pali Rohár
pali.rohar@gmail.com