On Mon, 1 Nov 2010, Dmitry Torokhov wrote:
> > I'm looking to dig into the usbhid driver code to get me the last
> > mile and get my HP LD4200tm touchscreen working on my Linux system.
> > [NOTE: I posted about this elsewhere, but as I think about it more,
> > this is more of a linux-kernel issue than anything else, so I'm
> > reprising that note here.]
>
> I think you'll have better luck on linux-input...
Thanks for CC, Dmitry.
> > So, in first connecting the HP touchscreen (with touchscreen hardware
> > from Nexio) to the Linux system (running Ubuntu 10.10 -- 2.6.35-22 generic)
> > things were looking up, as the unit almost worked right out of the box
> > as a plug-and-play device. Unfortunately, that last mile needs to be
> > addressed in order for the unit to work in a useful way. I think I've
> > narrowed down the problem to the usbhid driver, as I'll explain in a
> > moment, and while I expect I have the ability to debug and fix the issue,
> > this would be my first dive into Linux driver coding, so I could probably
> > use a pointer as to what steps are necessary to get the proper kernel
> > sources, compile the needed parts and restart the system (hopefully
> > just restart X-windows) to test my edits (especially since Ubuntu tends
> > to not include a not of the necessary components).
> >
> > I've collected gobs of data on what udev, hidraw, evtest and X-windows
> > report, but before I get to any of that, here's the executive summary:
> >
> > Plugging the HP LD4200tm into a Ubuntu 10.10 system will cause the
> > device to be recognized, create a usbhid device, create an event device,
> > activate the evdev driver for Xorg, and put it on the list of X's inputs.
> > So far, so good.
> >
> > Touching the device causes the mouse pointer to jump to that location,
> > even better.
> >
> > But when I'm touching the screen, what happens is that the system
> > generates a continuous stream of button presses/releases -- as if
> > I moved the mouse to a particular location and then went crazy
> > tapping on the mouse button. This is the part that makes the system
> > unusable.
> >
> > One more data point is that if I touch the screen with a second
> > finger, the button barrage stops, and it will now gracefully report
> > the movement of the first finger.
> >
> > So all I need to do is modify the driver for this device's protocol
> > to not send all the unnecessary release/press events.
> >
> >
> > Here is the data that leads me to believe the issue is at the usbhid
> > layer:
> >
> >
> > 1) Running "evtest /dev/input/event4" reveals that the button barrage
> > is already occurring at the HID event layer.
> >
> > 2) Watching the raw data on /dev/usb/hiddev2 does report a stream of
> > data as the touchscreen is touched (and no data when it's not touched),
> > looking at the output suggests that the device is only reporting touch/
> > release events when the screen is actually touched or released. And of
> > course, even if the device were reporting things poorly, I figure the
> > usbhid driver could be written to overcome this.
> >
> > Fortunately, the data seems to be a constant 14-byte stride, so I
> > wrote a simple awk script ("nexio") to reformat the data as I was
> > figuring out the raw protocol. Here's a sample of the reformatted
> > data, followed by my estimation of the protocol:
> > % sudo xxd -c 14 -g 1 /dev/hidraw2 | nexio
> > 0000000: 01 03 01 12d8 3357 02 02 15e1 1f71 01
> > 000000e: 01 03 01 12cc 332a 02 02 15e1 1f71 01
> > 000001c: 01 03 01 12de 3335 00 02 15e1 1f71 01
> > 000002a: 01 03 01 12f4 3346 00 02 15e1 1f71 01
> > 0000038: 01 03 01 1304 3352 00 02 15e1 1f71 01
> > 0000046: 01 03 01 130c 3358 00 02 15e1 1f71 01
> > [... (next I stop touching the screen, and then touch it again)]
> > 00009a0: 01 03 01 12d8 3357 00 02 15e1 1f71 01
> > 00009ae: 01 03 01 12d8 3357 00 02 15e1 1f71 01
> > 00009bc: 01 02 01 12d8 3357 00 02 15e1 1f71 01
> > 00009ca: 01 03 01 10af 3681 02 02 15e1 1f71 01
> > 00009d8: 01 03 01 10a5 3651 02 02 15e1 1f71 01
> > 00009e6: 01 03 01 10b7 3654 00 02 15e1 1f71 01
> > [... (next is a multi-touch event)]
> > 0000c78: 01 03 01 1019 3627 00 02 15e1 1f71 01
> > 0000c86: 01 03 01 1019 3627 00 02 15e1 1f71 01
> > 0000c94: 01 00 01 100b 35f8 00 02 1585 3110 00
> > 0000ca2: 01 00 01 100e 3604 00 02 158a 311b 00
> > 0000cb0: 01 00 01 1013 3615 00 02 1591 312b 00
> > 0000cbe: 01 00 01 1017 3622 00 02 1596 3136 00
> > 0000ccc: 01 00 01 1019 3628 00 02 1598 313c 00
> > 0000cda: 01 00 01 1019 3629 00 02 1599 313d 00
> > 0000ce8: 01 03 01 1019 3629 03 02 1598 313d 02
> > 0000cf6: 01 03 01 1019 3628 03 02 1598 313c 02
> > 0000d04: 01 03 01 1019 3627 03 02 1598 313b 02
> > 0000d12: 01 03 01 1018 3627 03 02 1597 313a 02
> > [... (next I release both fingers -- not quite simultaneously)]
> > 0000df2: 01 03 01 1019 3627 03 02 1598 3132 02
> > 0000e00: 01 03 01 1019 3627 03 02 1598 3132 02
> > 0000e0e: 01 03 01 1019 3627 02 02 1598 3132 01
> > 0000e1c: 01 03 01 1019 3626 02 02 1598 3132 01
> > 0000e2a: 01 03 01 1019 3626 00 02 1598 3132 01
> > 0000e38: 01 03 01 1019 3626 00 02 1598 3132 01
> > 0000e46: 01 03 01 1019 3626 00 02 1598 3132 01
> > 0000e54: 01 03 01 1019 3626 00 02 1598 3132 01
> > 0000e62: 01 03 01 1019 3625 00 02 1598 3132 01
> > 0000e70: 01 02 01 1019 3625 00 02 1598 3132 01
Thanks a lot for the detailed analysis. What would actually be interesting
to see would be the HID debugfs output. Namely
1) mount debugfs
2) cat /sys/kernel/debug/<device>/rdesc
3) obtain output of /sys/kernel/debug/<device>/events and annotate it as
you did for your previous output (i.e. "this happens when I do one-finger
move", "this happens when I do two-finger trick to make it work").
This way, it'd be much easier to see what is actually hapenning on HID
level.
Thanks,
--
Jiri Kosina
SUSE Labs, Novell Inc.
--
To unsubscribe from this list: send the line "unsubscribe linux-input" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
