[How-to] BB-View on latest Debian

Perfect timing, I was just about to start looking for what pieces I need to extract in order to get the 4.3" BB-View working on a vanilla 3.13 kernel. Looks like you've identified the interesting pieces for me

When you say the touch screen is different from the standard TI 4 wire interface, how did you work that out ?  Did you find schematics for the BB-View anywhere ?

From the Angstrom source download, a diff between BB-BONE-LCD4-01-00A1.dts & BB-VIEW-LCD4-01-00A0.dts is interesting, there's a couple of differences in the pins used for LEDs & keys along with differences in numbers of leds & keys.

Other than that, there's one extra pin used for "LCD DISEN" in BB-BONE-LCD4-01-00A1.dts and for the BB-VIEW there's this

ti,touch-type = <1>;

in the LCD7 ones the BB-VIEW has that as 2

Not exactly sure what that does just yet, but it's obviously some hint to the driver to configure something a certain way. Otherwise the driver and pins used appear to be identical.  Perhaps it's a place to start looking for the problem though.

selsinork wrote:

So I'm going to try your patch exactly as-is and see what happens.

Well it works perfectly!

Now I just need to try and understand how it all works..  I may be some time

Very good!

Yeah, I think we need to know more about the low levels of the touch screen controller in the AM335x processors, in order to make the patch more universal.....see if this helps:

Taken from the AM335x Technical Reference Manual.

Hello,

I still have problem with ts_calibrate.  I read that it gived you a pile of crap....  could you explain how you managed to make it works please?

I set the environment variable like this:

export TSLIB_FBDEVICE=/dev/fb0

export TSLIB_CONFFILE=/etc/ts.conf

export TSLIB_PLUGINDIR=/usr/local/lib/ts

export TSLIB_TSDEVICE=/dev/input/event1

I tried also this :

export TSLIB_TSDEVICE=/dev/input/mouse0 (and all others input devices..... )

event1 and mouse0 are the two files that output something when I do :

cat /dev/input/event1  or cat /dev/input/mouse0 and I touch the screen.

but even if I configure the environment variable like above, I've got this message when I execute ts_calibrate:

No raw module loaded.

ts_config: No such file or directory

have you got an idea? ( in /etc/ts.conf,  the line  module_raw input (linux input layer event interface) is not commented..)

Two things of note are that the BB-View is different from the other CircuitCo LCD capes as far as the touch screen is concerned - I don't fully understand the details - and that while Louis provided a patch to the 3.8.13 kernel that Robert Nelson provides for his Debian images, I'm using a mostly vanilla 3.13-rc mainline kernel. So the details of what I did may well be different..

That said, what you need is the various patches from Louis that were posted earlier in this thread, and apply them to your kernel source, then recompile the kernel and boot it.  If your kernel has a capemgr you use it to load the appropriate device tree overlay. Otherwise, like me, you'll need to incorporate the devicetree changes into a static devicetree file and make u-boot to load it in place of the default.  There are lots of pieces to get right here and it's difficult to explain it exactly as what needs done is very dependant on what kernel, distro and u-Boot versions you have.

I can say that I didn't need any env vars to get ts_calibrate to work.

In /etc/ts.conf the only things I have un-commented are:

module_raw input
module pthres pmin=1
module variance delta=30
module dejitter delta=100
module linear

when I ran configure for tslib I used the following

./configure --prefix=/usr --sysconfdir=/etc

make install puts the various modules in /usr/lib/ts and I find things like input.so dejitter.so pthres.so in that directory.

From the error message you're seeing it would seem that the ts_calibrate can't find input.so. So you probably need to confirm that it's in /usr/local/lib/ts if that's what you're setting TSLIB_PLUGINDIR to.

My particular setup uses /dev/input/event0 and /dev/fb0.  input0 vs input1 probably depends on what order the drivers load in.

I see input devices in dmesg as follows:

(apologies for the daft formatting, the editor forces the above to be a table when pasted in, see  http://www.element14.com/community/thread/31978/l/how-do-i-un-table-pasted-text )

In my kernel .config I have

CONFIG_FRAMEBUFFER_CONSOLE=y
CONFIG_FRAMEBUFFER_CONSOLE_DETECT_PRIMARY=y

but my kernel commandline only has

console=ttyO0,115200n8 root=/dev/mmcblk0p2 ro rootfstype=ext4 rootwait

so the boot messages only go to ttyO0 not to the LCD

Hello,

Thanks for your help^^

well I used your ./configure --prefix=/usr --sysconfdir=/etc command line and now ts_calibrate works "a little"..

I still have an issue, when I execute ts_calibrate I've got this message:

tslib: selected device is not a touchscreen (must support ABS and Key event types)

but it takes samples and write calibrations constant (with bad values) in pointercal.

my ti-tsc is input1 and bb_view_gpio_keys is input4

Can you tell me what have you got in pointercal please? (I'll try to force it "by hand"..)

I saw this on the web, http://stackoverflow.com/questions/11937287/touchscreen-and-driver-installed-but-tslib-cannot-calibrate

someone advise to modify the source code like this:

input_dev = input_allocate_device();
[..]
set_bit(EV_ABS, input_dev->evbit);
set_bit(EV_KEY, input_dev->evbit);

But it seems that it doesn't work..

If yours in input1 then perhaps you need to export TSLIB_TSDEVICE with something appropriate ?

My /etc/pointercal with a BB-View-4.3 is

-8179 -5 32323404 11 -5081 18944446 65536 480 272

Thanks for your answer, maybe I have a clue :

For the moment, I compiled tslib directly on the beagleboard.. Maybe should I compile it on my PC desktop with the same cross compile chain that I used to compile the kernell..?

Edit: well I compiled it on my PC desktop with the cross tool chain :

./configure --prefix=/usr --sysconfdir=/etc

make ARCH=arm CROSS_COMPILE=/home/ana/Téléchargements/BBB/gcc-linaro-arm-linux-gnueabihf-4.8-2013.10_linux/bin/arm-linux-gnueabihf-

and copy all files in the "good" directories on the BBB . I mean all needed *.la and *.so in /usr/lib/ts and the all needed binaries like _s_calibrate in /usr/bin .. and I copy also tslib directory with source files (with all *.o thats also seems to be needed) .,

and now I have this error message:

ts_calibrate.o: file not recognized: File format not recognized

collect2: ld return 1 exit statutsn

It's really a pain to make this works ... Well I think I will stop with my Ubuntu file system and try the [How-to] of this post... I'm really confused because I don't understand what I'm doing wrong...

IMHO cross compiling is always a bad idea, it's notoriously hard to get it right.  I natively compile all my code.

Anyway, your problem is likely to be that you didn't tell ./configure that you're cross compiling so it created makefiles for compiling on the host instead.  ARCH=arm CROSS_COMPILE= are things that the kernel build system knows how to handle. Nothing that I can see in the tslib sources uses either of those environment vars.

./configure --help has this:

so if you really must cross compile, you'll have to work out exactly what to pass to configure using that option as a start. You may also need to set CC to your cross compiler.  Even then, it may not work as configure often tries to compile and run small pieces of code while testing for features. Getting those Arm-compiled tests to run on your x86 system isn't impossible, but it's an extra challenge. 

Hello selsinork,

Thanks for your answer ^^

well I tried a cross-compile of tslib  just because my natively compiled on the BBB wasn't working...

I understand what you mean by passing options to configure.. I'll try to do this..  (just to be clearer, I'm not trying to run on an x86, but just to compile on it and run it on the BBB)

Hello again,

Well I managed to cross compile tslib, but I've got exactly the same problem than with the native compile.

So I think that maybe the problems comes from my *dtb  file because I took the Angstrom ones (I didn't compiled them like I did with the kernell)...  (I've just discovered dtb files, so I'm a little bit disapointed with it.. )

Hello again,

Well I managed to cross compile tslib, but I've got exactly the same problem than with the native compile.

So I think that maybe the problems comes from my *dtb  file because I took the Angstrom ones (I didn't compiled them like I did with the kernell)...  (I've just discovered dtb files, so I'm a little bit disapointed with it.. )

Devicetree has been a painful experience for a lot of us, however since they're used to describe what hardware is present it's important to get them right.  On the BBB, depending on which kernel you're using, you'll also have the added challenge of device tree overlays (*.dtbo files).

While the idea behind overlays could be an interesting one, I've personally come to the conclusion that overlays are unnecessary and mostly just an added complexity. For the large majority of people who will struggle with normal device tree stuff, overlays just are not helpful at all, they're just another barrier.