jnettlet has quit [Remote host closed the connection]
jnettlet1 has quit [Remote host closed the connection]
jnettlet has joined #linux-exynos
<amitk>
sjoerd: Got to trying out your kernel and configuring Xorg but startx fails with the following log: http://paste.ubuntu.com/9438677/ Any clues?
<sjoerd>
FPE ? that sounds special
<sjoerd>
amitk: i havne't tried X on this yet, so not sure
<amitk>
sjoerd: you tried this with wayland? Do you know if the ubuntu wayland packages work?
<amitk>
else I'll have to spend time switching to debian
<sjoerd>
I'd always recommend switching to debian :p
<sjoerd>
But the ubuntu weston package should work
<sjoerd>
if you launch it with the pixman rendering backend that is
<amitk>
sjoerd: I don't even know what that means (pixman rendering) :)
<sjoerd>
amitk: One thing to double-check is if your permissions on /dev/mali etc are correct for the user you're launching X as
<sjoerd>
amitk: Oh, that means weston uses software compositing. You can't use EGL/GL on wayland with mali just yet
<sjoerd>
amitk: but if weston properly comes up you'll at least know that the display subsystem works correctly
<amitk>
sjoerd: so how do I launch weston?
<sjoerd>
amitk: log into a VT and run weston-launch -- --use-pixman
<amitk>
hey, that seemed to work, though my touchpad isn't responding
<sjoerd>
my xu3 wip tree doesn't have the cb2 touchpad patches in
<sjoerd>
you can grab those from my wip/exynos-touchpad branch
<amitk>
sjoerd: hmm, evtest is showing the button clicks on the touchpad, not the finger movements
<amitk>
actually, everything is showing up as BTN_LEFT, will investigate the dts
<sjoerd>
amitk: is that with those patches ?
<sjoerd>
Ah, so upstream does have the dts snippet
<javier__>
sjoerd: yes, only the atmel T100 multi-touch object support in the driver is missing in mainline
<javier__>
amitk: have you cherry-picked ced0716 ("Input: atmel_mxt_ts - implement support for T100 touch object") and 5af117e ("WIP - Input: atmel_mxt_ts: potentailly register T100 as a touchpad") from sjoerd's wip/exynos-touchpad branch?
<javier__>
amitk: the peach pit and pi have different atmel chips that use different multi-touch protocols, only the one for the pit is supported in the mainline driver right now
<javier__>
amitk: iirc that is needed for the multi-touch events but support for the button event is the same which matches the behaviour you described
<javier__>
amitk: about everything reported as BTN_LEFT, daniels explained to me that it is because is a touchpad with a single button and user-space use a combination of button pressed and mt positions to figure out a right click was made for example
<daniels>
right, the tracking is too complex to figure it out in the kernel
<amitk>
javier__: thanks, with those two cherry-picks I get "unaccelerated" cursor movement
<amitk>
daniels: sjoerd: so what interesting things can I do in this weston environment? I guess running X apps is not possible...
<daniels>
amitk: xwayland is possible in theory but we don't build support for that. (should fix that!)
<daniels>
amitk: you can run gtk apps tho
<sjoerd>
amitk: you can't use the mali though, but this at least means your display subsystem is up and works
<sjoerd>
amitk: can you get a backtrace from your Xorg crash?
<daniels>
amitk: if you get the chromeos mali binaries (assuming they're released for the 5422 rather than just 5250) then you can run X11/mutter/chrome/etc
<sjoerd>
daniels: his X crashed on startup, which is why he tried weston :)
<amitk>
sjoerd: I'll try to setup all the -debug packages and capture a backtrace from Xorg
<javier__>
amitk: great, I should find some time to clean those atmel patches and post them upstream
<amitk>
javier__: feel free to cc me and I'd be happy to give my "Tested-by"
<javier__>
amitk: thanks
<sjoerd>
javier__: my outstanding thing with those is that really the init function should work for both version
<sjoerd>
currently it has different ones fro the two protocols which are functionality wise for 90% the same
<javier__>
sjoerd: I know, by clean I mean factor out the init sequence for both
<sjoerd>
;)
<javier__>
sjoerd: I should also prod to the mmc/sdio wifi support a bit more, just cherry-picking and adapting the DTS snippets is not enough so I've to dig deeper what the downstream changes in the dw_mmc host driver are
<sjoerd>
i see
<sjoerd>
javier__: does samsungs exynos-reference work ooi ?
<sjoerd>
that might have more targetted dw mmc changes then the chromeos kernels
<javier__>
afaict those are the only two peripherals that are missing to have full peach pi support (modulo hw acceleration)
<sjoerd>
javier__: webcam!
<javier__>
right!
<javier__>
but that is *just* USB right?
<daniels>
heh
<sjoerd>
Yeah it just needs to be powered on i think
<javier__>
I see
* javier__
adds webcam to his TODO
<amitk>
javier__: sjoerd is right, no device support is complete these days until it can be used to take a selfie ;)
<javier__>
amitk: lol
<daniels>
sjoerd, amitk: yeah, synaptics is hosed I guess
<sjoerd>
cheese is a critical user application!
<sjoerd>
daniels: oh you suspect that crash is in synaptics?
<sjoerd>
amitk: worth trying with a working touchpad if that's the case ^
<amitk>
sjoerd: "working touchpad" from where?
<sjoerd>
amitk: i thought you mentioned that with those extra patches your touchpad works in weston now?
<amitk>
sjoerd: it does
<sjoerd>
so worth seeing if that "solved" your X problem
<amitk>
ohh
* amitk
would never have linked touchpad to X but ok, I'll try it
<sjoerd>
amitk: the syntaptic driver is the one that interacts with the touchpad and suspiciously the last log line from X before the crash is from it
<amitk>
sjoerd: daniels: thanks! Xorg started!
<sjoerd>
\o/
<amitk>
now to figure out how to get Mali acceleration
<sjoerd>
at this poitn that should be a matter of putting the blobs in the right spots
<javier__>
amitk: \o/
<amitk>
sjoerd: yes, looks like I have mesa libs installed that I should first get rid of
<amitk>
sjoerd: do you know if r4p1 blobs (mali-t62x_r4p1-00rel0_linux_1+fbdev.tar.gz, mali-t62x_r4p1-00rel0_linux_1+x11.tar.gz) will work with your kernel? I just realised I might be using something too shiny...
<sjoerd>
i think the kernel space is r4p0 but i always forget
<amitk>
sjoerd: I don't see the r4p1 string in the commit logs or code, I do see r4p0, so you might be right
* amitk
goes to find older blobs
<sjoerd>
amitk: I've always just the blobs from chromeos
<sjoerd>
those also have dmabuf import support which is nice
<Wizzup>
but does it actually give you accel that you can use, apart from opengles?
<sjoerd>
eh
<sjoerd>
GLESv2 is accell you can use
<sjoerd>
I'm not sure why you want to meaningfully accelerate on arm that can't do gles tbh
<Wizzup>
What do you mean, on arm?
<javier__>
Wizzup: big GL is not even supported by the mali GPU afaik
<sjoerd>
on an ARM based platform
<Wizzup>
There are plenty of examples of programs that only do opengl, not gles, including many visualisation toolkits, etc
<Wizzup>
javier__: I know :) but proper 2d and video accel is probably more welcome for me
<sjoerd>
On x86 there is al ot of legacy which isn't there on arm
<Wizzup>
I wouldn't call much of what I use legacy :)
<Wizzup>
but, sure
<sjoerd>
Not sure which visualisation toolkit you're referring to here
<Wizzup>
I mean, I live without accelerated opengl for a long time now and it's mostly not a problem, I just know very little that uses opengles
<sjoerd>
i would expect things like opencv to be able to do gles
<Wizzup>
maybe mpv can use it for scaling
<Wizzup>
sjoerd: vtk,paraview,blender probably too
<Wizzup>
Not sure if firefox can use it for webgl; libreoffice doesn't use it afaik
<sjoerd>
we've got fully accelerated output paths video playback with gstreamer on these devices