tablets using evdev

[Peter Hutterer]

On Wed, May 18, 2011 at 10:30:11PM -0400, Jimmy Hon wrote:
> The Short:
> I have a tablet and it didn't work with the evdev driver. So I
> hacked the evdev input driver to work with it. Hacks are bad, so how
> should I go about getting real fixes?
> 
> The Long:
> I bought a tablet from MonoPrice (Item 6251, http://www.monoprice.com/products/product.asp?c_id=108&cp_id=10841&cs_id=1084101&p_id=6251&seq=1&format=2).
> 
> 
> In lsusb, it identifies as:
> 5543:0064 UC-Logic Technology Corp. Aiptek HyperPen 10000U.
> 
> /proc/bus/input/devices says:
> I: Bus=0003 Vendor=5543 Product=0064 Version=0110
> N: Name="UC-LOGIC TWA60"
> P: Phys=usb-0000:00:12.2-2.1.2/input0
> S: Sysfs=/devices/pci0000:00/0000:00:12.2/usb1/1-2/1-2.1/1-2.1.2/1-2.1.2:1.0/input/input73
> U: Uniq=
> H: Handlers=mouse5 event13
> B: PROP=0
> B: EV=1f
> B: KEY=c03 3f0001 0 0 0 0
> B: REL=303
> B: ABS=100000f
> B: MSC=10
> 
> The tablet didn't work correctly when evdev drove it. Playing with
> xinput test, it turns out the tablet was reporting the x/y
> coordinates on valuator 2 and 3, and not on 0 and 1. The pen
> sensitivity is reported as evaluator 4. So valuator 2, 3, and 4 were
> reporting data when I moved then pen around.
> 
> So I wanted evdev to ignore the first two valuators. I did this by
> cheating and changed the ABS initialization code in
> EvdevAddAbsClass() to start the for loop at ABS_Z instead for ABS_X
> when it's checking the abs_bitmask). So basically, if
> /proc/bus/input/devices reported ABS=100000c instead of ABS=100000f,
> I would have been fine.

you need HID_QUIRK_MULTI_INPUT in the kernel for this device. See kernel
commit e1f092102f65e424be40c318a0fab7bb6e34194f for example.
once that's in, you get two kernel devices, the first one mute but the
second one with the right axis enumeration.

Cheers,
  Peter

> To get the wheel emulation working, in EvdevWheelEmuFilterMotion(),
> I changed the switch statement to check pEvdev->axis_map[pEv->code]
> instead of just pEv->code. It's a similar idea as the ABS
> differential measurement code just above the switch statement. You
> want to check which axis the values will be eventually reported as.
> 
> This works fine for _me_, since the tablet is the only input device
> that's in absolute mode. My normal mouse and trackball use relative
> mode, and those code paths have not been modified.
> 
> So if I want this tablet to work for real what's the proper way to do it?
> 1. Should the kernel driver not report the two unused valuators?
> (Make it report ABS=100000c instead of ABS=100000f)
> 2. Does the evdev xorg driver have an ability to remap axes?
> 3. Can the evdev xorg driver have addditional options to blacklist
> some of the valuators?
> 4. Should any of this be done in the evdev xorg driver? Or is it
> more appropriate to have a specific tablet driver?
> 
> I know there is the wizardpen xorg driver, which supports this
> tablet. But it's not as featureful as evdev. There's no wheel
> emulation, and no Properties to changes at runtime. (I've been
> changing the Calibration property, when I switch between single
> monitor vs dual monitors)
> 
> If this type of support should be a tablet specific xorg driver, is
> there a master tablet driver? (i.e. like how evdev is the master
> driver for all mice and keyboards?)
> 
> I've seen the aiptek, hyperpen, wizardpen, and wacom drivers. Why
> are there so many tablet drivers? What makes each driver so special?
> (I tried google and also perusing the xorg.freedesktop.org site, but
> I couldn't pinpoint the differences between the drivers). Having so
> many tablet drivers makes me think that there is going to be a lot
> of code duplication if they were all going to have a wheel emulation
> feature. Is there a way to put the wheel emulation logic into a
> library, and have all the drivers share that?
> 
> Jimmy
> 
> P.S. Should this be posted on xorg-devel too?
> 
> 
> My hacks look like this:
> diff -ur xf86-input-evdev-2.6.0/src/emuWheel.c
> xf86-input-evdev-2.6.0-dev/src/emuWheel.c
> --- xf86-input-evdev-2.6.0/src/emuWheel.c    2010-12-20
> 19:19:16.000000000 -0500
> +++ xf86-input-evdev-2.6.0-dev/src/emuWheel.c    2011-05-18
> 04:14:18.207068902 -0400
> @@ -125,7 +125,7 @@
>          value -= oldValue; /* make value into a differential
> measurement */
>      }
> 
> -    switch(pEv->code) {
> +    switch(pEvdev->axis_map[pEv->code]) {
> 
>      /* ABS_X has the same value as REL_X, so this case catches both */
>      case REL_X:
> diff -ur xf86-input-evdev-2.6.0/src/evdev.c
> xf86-input-evdev-2.6.0-dev/src/evdev.c
> --- xf86-input-evdev-2.6.0/src/evdev.c    2011-01-05
> 17:35:27.000000000 -0500
> +++ xf86-input-evdev-2.6.0-dev/src/evdev.c    2011-05-16
> 00:52:41.485952421 -0400
> @@ -1303,7 +1303,7 @@
>      memset(pEvdev->old_vals, -1, num_axes * sizeof(int));
>      atoms = malloc(pEvdev->num_vals * sizeof(Atom));
> 
> -    for (axis = ABS_X; i < MAX_VALUATORS && axis <= ABS_MAX; axis++) {
> +    for (axis = ABS_Z; i < MAX_VALUATORS && axis <= ABS_MAX; axis++) {
>          pEvdev->axis_map[axis] = -1;
>          if (!TestBit(axis, pEvdev->abs_bitmask))
>              continue;
> 
> 
> _______________________________________________
> xorg at lists.freedesktop.org: X.Org support
> Archives: http://lists.freedesktop.org/archives/xorg
> Info: http://lists.freedesktop.org/mailman/listinfo/xorg
> Your subscription address: peter.hutterer at who-t.net
>