Re: Crash caused by hid-rmi and no longer working touchscreen

[Andrew Duggan]

On 04/10/2017 11:38 AM, Dmitry Torokhov wrote:
> On Mon, Apr 10, 2017 at 03:18:03PM +0200, Jiri Kosina wrote:
> > [ a couple more people from "the other regression thread" added to CC ]
> >
> > On Mon, 10 Apr 2017, Gabriele Mazzotta wrote:
> >
> > > Hi,
> > >
> > > I'm having troubles with hid-rmi using 4.11-rc6. I'm getting some
> > > "rmi_hid_read_block: timeout elapsed" errors and ultimately the
> > > crash you can see at the end of this mail. I will reply to this
> > > mail with a full log that shows all the errors.
> > >
> > > The problem seems to be caused by commit 279967a65b320d174a507498ae
> > > ("HID: rmi: Handle all Synaptics touchpads using hid-rmi").
> > >
> > > My laptop has a Synaptics touchpad and a Synaptics touchscreen.
> > > Because of that commit, hid-rmi is used for both the devices, while
> > > it has always been used only for the touchpad. Needless to say that
> > > the touchscreen no longer works.
> > >
> > > If I revert that commit, everything goes back to normal.
> > Alright, given this is a second regression cause by this switch, I suggest
> > to revert this one for 4.11 still, and work on fixing this properly for
> > 4.12.
> >
> > If noone raises any objections (a few more people added to Cc), I'll queue
> > a 4.11-rc revert tomorrow.
> Yes, we also have this:
>
> 	https://bugzilla.kernel.org/show_bug.cgi?id=195287

I suspect that the issue might be that both of these touchscreens have 
the HID_GD_POINTER usage defined in their HID report descriptor. This 
might be because the firmware is configured to support pens. It looks 
like the Asus T302CA from the bug does have a pen and the XPS13 9333 may 
simply have firmware which has pen support enabled. The HID_GD_POINTER 
usage is used in hid-core to differentiate between touchpads and 
touchscreens. So if a touchscreen has HID_GD_POINTER defined then 
hid-core will bind hid-rmi to it. None, of my test systems have pen 
support so I did not anticipate this.

Also, the reason why we can't just support touchscreens using hid-rmi is 
because the firmware for HID touchscreens lacks the functionality of 
reporting the attention data (finger positions) in a single report. It 
ends up being less efficient to run HID touchscreens in RMI mode.

> so reverting seems prudent at this time.

Yes, I agree we should revert the commit. There simply isn't enough time 
to come up with and test a new method for differentiating touchpads and 
touchscreens before 4.11 is released.

Andrew

> Thanks.