Wayland on FreeBSD

[Jan Beich]

Frederic Chardon <chardon.frederic at gmail.com> writes:

> Le lun. 20 avr. 2020 à 20:54, Jan Beich <jbeich at freebsd.org> a écrit :
>
>>
>> Frederic Chardon <chardon.frederic at gmail.com> writes:
>>
>> > Le lun. 20 avr. 2020 à 15:20, Jan Beich <jbeich at freebsd.org> a écrit :
>> >
>> >>
>> >> Frederic Chardon <chardon.frederic at gmail.com> writes:
>> >>
>> >> > Le lun. 20 avr. 2020 à 12:20, Niclas Zeising
>> >> > <zeising+freebsd at daemonic.se> a écrit :
>> >> >
>> >> >>
>> >> >> On 2020-04-20 12:07, Greg V wrote:
>> >> >> > Apr 20, 2020 12:56:35 PM Frederic Chardon <chardon.frederic at gmail.com>:
>> >> >> >
>> >> >> >> Le dim. 19 avr. 2020 à 13:50, Jan Beich <jbeich at freebsd.org> a écrit :
>> >> >> >>
>> >> >> >>>
>> >> >> >>>
>> >> >> >>> I'm using Sway myself primarily with X11 applications. For example,
>> >> >> >>> "vblank_mode=0 glxgears" shows 3x more FPS than on real Xorg server. ;)
>> >> >> >>>
>> >> >> >>
>> >> >> >> I decided to give it a try, is hardware acceleration supposed to work?
>> >> >> >> glxgears gives 10x less fps than with X, and 3D is incredibly
>> >> >> >> sluggish.
>> >> >> >> eglinfo output that leads me to believe it is not:
>> >> >> >> GBM platform:
>> >> >> >> i965_dri.so does not support the 0xffffffff PCI ID.
>> >> >> >>
>> >> >> >> It is with sway, -current from yesterday, intel HD4000 iGPU (Ivybridge i5-3320M)
>> >> >> >> no difference between drm-devel-kmod or drm-current-kmod.
[...]
>> >> > With or without LIBGL_DRI3_ENABLE=1 in the environment the result is
>> >> > the same. I use official packages, updated yesterday.
>> >> >
>> >> > The problem I face might be due to permission issue. ktrace shows this:
>> >> > 28047 eglinfo  CALL  openat(AT_FDCWD,0x7fffffffc820,0x100002<O_RDWR|O_CLOEXEC>)
>> >> > 28047 eglinfo  NAMI  "/dev/dri/card0"
>> >> > 28047 eglinfo  RET   openat 3
>> >> > /..../
>> >> > 28047 eglinfo  CALL  ioctl(0x3,0xc0106446,0x7fffffffbc70)
>> >> > 28047 eglinfo  RET   ioctl -1 errno 13 Permission denied
[...]
>>
>> Try starting Sway via "ktrace -i" or replace Xwayland binary with a
>> wrapper (see below) that starts it via ktrace. Tracing glxgears is too
>> late as Mesa will simply give up if DRI3 is not initialized.
[...]
>
> The trace shows usage of llvmpipe after the failed ioctl on
> /dev/dri/card0, so as I understand the lack of hardware acceleration
> concerns only Xwayland, whereas wayland itself is accelerated ?

Yep. Check which ioctls fail then try to reproduce outside of Mesa or
hardcode the result. Overriding PCI ID via INTEL_DEVID_OVERRIDE is
unlikely to help e.g.,

https://lists.freebsd.org/pipermail/freebsd-x11/2019-January/022551.html

If you still have no clue try playing with sysctls under compat.linuxkpi
via /boot/loader.conf.