How to have accelerated graphics with Intel Celeron J1800 (Bay Trail) ?

[Natasha Kerensikova]

Hello,

on Monday 18 July 2016 at 09:45, Juan Ramón Molina Menor wrote:
> Hi Natasha.

Thanks a lot for taking the time to answer.

> Matthew Macy has done a lot of work recently to support relatively 
> recent Intel GPUs in CURRENT. The development branch is in GitHub, I 
> suggest you to start here:
> https://github.com/FreeBSDDesktop/freebsd-base-graphics/wiki

I did encounter that page, though I admit digging seriously into it after
posting to the ML.

However, it looks like rather experimental stuff, which is fine by me
but slightly besides my original questions.

So just to be clear and be sure I correctly understood everything I
came across:

 - Is vanilla FreeBSD 11.0 supposed/planned to support Haswell, or have
   I misunderstood something at some point?
 - Is there any definite information one way or the other about whether
   vanilla FreeBSD 11.0 is supposed to support Bay Trail ?
 - If not, was it a stretch to assume Bay Trail support when the
   both immediate successor (Haswell) and immediate predecessor (Ivy
   Bridge) are supported?

I honestly thought that my 11.0-BETA1 was supposed to handle my Bay
Trail, albeit with a beta-level of quality. When that outright didn't
work, I tried with the vanilla 11.0-BETA1 kernel and userland the newer
ports from http://www.bsddesktop.com/images/xserver-next-pkgs/, with
similar results.

Then I did try to compile the drm-next-4.6 branch, using wiki
instructions, and found a build error beyond my understand a few hours
later.

> As for your Bay Trail chip, there is an open bug:
> https://github.com/FreeBSDDesktop/freebsd-base-graphics/issues/6

Only an hour ago did I finish downloading the CFT image labelled
2016062423, and I think I can reproduce the bug. The funny thing is that
the corruption I see "heals" over time, so I first I thought it was a
fancy effect from the desktop environment (I'm not well acquainted to
these fancy UIs).

I guess it does dampen the motivation to try and make it work on a more
vanilla release, if that's the best that can be achieved with current
code...

In case it is unexpected for anyone, here the difference in Xorg.0.log
between CFT and 11.0-BETA1. The following line are the last common
lines:
[   140.072] (II) intel: Driver for Intel(R) Integrated Graphics Chipsets: [long list snipped]
[   140.074] (II) intel: Driver for Intel(R) HD Graphics: 2000-6000
[   140.074] (II) intel: Driver for Intel(R) Iris(TM) Graphics: 5100, 6100
[   140.074] (II) intel: Driver for Intel(R) Iris(TM) Pro Graphics: 5200, 6200, P6300
[   140.074] (--) Using syscons driver with X support (version 16777218.0)
[   140.074] (--) using VT number 9

following that, 11.0-BETA1 immediately complains that there is no
device and no screen and aborts, while the CFT gives:
[   140.078] (II) intel(0): Using Kernel Mode Setting driver: i915, version 1.6.0 20160229

Also, it might or might not be related to the corruption healing, I have

(EE) [mi] EQ overflowing.  Additional events will be discarded until existing events are processed.
(EE) [mi] These backtraces from mieqEnqueue may point to a culprit higher up the stack.
(EE) [mi] mieq is *NOT* the cause.  It is a victim.
[   164.765] [mi] Increasing EQ size to 1024 to prevent dropped events.
[   164.767] [mi] EQ processing has resumed after 857 dropped events.
[   164.767] [mi] This may be caused by a misbehaving driver monopolizing the server's resources.
[   170.463] (EE) intel(0): Failed to submit rendering commands (Input/output error), disabling acceleration.

> Concerning the scfb driver, I’ve used it with success (but only in a 
> Haswell chip) by just adding the following lines to xorg.conf:
> 
> Section "Device"
>       Identifier "Generic FB"
>       Driver "scfb"
> #    Option "NoAccel" "True" # Both work, test it
> EndSection

Thanks exactly what I did, but X couldn't start, again complaining about
lack of screen. scfb driver itself reported an error "FBIOGTYPE" about
an inappropriate ioctl.

Unfortunately I lost the exact line when installing the CFT image, I
thought it was in the backuped data but I was wrong. I can rebuild the
11.0-BETA1 given some time, if there is interest.

Which leads me to the naive question: is there any reason to use scfb
when vesa fits my need?
The only thing I miss from 10.3-RELEASE vesa is the 2D acceleration (and
maybe the video decoding, not sure how the celeron CPU fares), but I
thought I read somewhere that scfb isn't really accelerated either.
Am I missing something?


Thanks a lot for your help,
Natasha