In which a touchscreen is rehabilitated, or: How I learned to stop being scared and just hack at /usr/src/sys/dev

[Large Hadron Collider]

(Apologies if this doesn't line break at 79 chars - full formatting in 
HTML but this may be lost - shouldn't lose any info though)

Good day subscribers to this list.

I'm here with what could be described as a success story and a patch in 
the same e-mail.

Please do stop me if WACF00E has already been slated for the next major 
release - but I would like to share how I got my HP Elitebook 2760p's 
touchscreen working.

So I, a former and now again user of FreeBSD (I got hacked the first 
time... silly Ellie shouldn't give shells to strangers, should she now?) 
have a laptop whose screen is touch-capable, and whose touchscreen 
subsystem is based on a serial Wacom tablet.

It worked under Linux and too I presume Windows (with which the laptop 
shipped), but not FreeBSD. I thought, what was going on? What was I 
doing wrong? So after some poking around I discovered that the screen is 
a WACF00E - not supported in 11.0-RELEASE-p1 by the driver that handles 
the UART.

It showed

     unknown pnpinfo _HID=WACF00E _UID=0 at handle=\_SB_.PCI0.LPCB.SIO_.DIGI

as the devinfo line.

Intriguingly, there was this line in uart_bus_acpi.c:static struct 
isa_pnp_id acpi_ns8250_ids[]:

     {0x04f0235c, "Wacom Tablet PC Screen"},         /* WACF004 */

So I thought what the hell, I'd copy that line under itself and change 
04f0 (which is byte-swapped, counterintuitively) to 0ef0, representing 
WACF00E.

Adding this:

     {0x0ef0235c, "Wacom Tablet PC Screen 00e"}, /* WACF00e */

to uart_bus_acpi.c and this:

     {0x0ef0235c, NULL},     /* WACF004 - Wacom Tablet PC Screen*/

(Yes it should read WACF00E in the comment) under the WACF004 entry in 
uart_bus_isa.c, then recompiling and installing in whatever way your 
configuration might demand seems to make the kernel detect the tablet as 
a UART.

So it detected it, and the dev file was /dev/cuau4, for uart4, the 
WACF00E (it was ttyS4 under Linux).

Great. X didn't detect it on its own, but that let me debug it using 
Minicom, which I promptly installed.

After telling Minicom to use /dev/cuau4 as the modem, and telling it to 
use 38400 8N1, touches to the screen resulted in what can only be 
described as euphoric garbage, indicating that this ugly hack on top of 
hack alert worked.

So I set up /usr/local/etc/X11/xorg.conf.d/wacom.conf to include 
(slightly amended from my actual setup, which only has ISDV4 in the 
stylus but still works for touch, haven't tested for stylus):

Section "InputDevice"
     Identifier    "wacom stylus"
     Driver        "wacom"
     Option        "Type"            "stylus"
     Option        "Device"        "/dev/cuau4"
     Option        "ForceDevice"        "ISDV4"
     Option          "AutoServerLayout"    "true"
EndSection

Section "InputDevice"
     Identifier    "wacom eraser"
     Driver        "wacom"
     Option        "Type"            "eraser"
     Option        "Device"        "/dev/cuau4"
     Option        "ForceDevice"        "ISDV4"
     Option          "AutoServerLayout"    "true"
EndSection

Section "InputDevice"
     Identifier    "wacom touch"
     Driver        "wacom"
     Option        "Type"            "touch"
     Option        "Touch"            "on"
     Option        "Device"        "/dev/cuau4"
     Option        "ForceDevice"        "ISDV4"
     Option          "AutoServerLayout"    "true"
EndSection

Restarted X, and after

     % xsetwacom set "wacom touch" Touch on

(I didn't initially have Touch on in the options list for "wacom touch") 
it was almost like striking platinum in a gold mine or something when 
the mouse just followed my finger the way I was used to it doing so 
under Linux.

To those of you who say that FreeBSD will never be ready for the 
desktop, you're only right when you're talking to newbs. And this is 
living proof that if you know some C and you're intrepid enough, 
miracles really can happen.