support for the Picasso APU in amdgpu

[Bakul Shah]

On Thu, 23 Jan 2020 11:23:00 +0000 Greg V <greg at unrelenting.technology> wrote:
Greg V writes:
> 
>
>
> Jan 23, 2020 5:58:29 AM Bakul Shah :
>
> > On the wiki.freebsd.org/Graphics I see that updating drm-devel
> > to 4.20 is 20% complete but I suspect this page is outdated!
> > Looking at drm-deel-kmod/Makefile it seems it is based on
> > linux 5.0? Ideally I don't want to upgrade to -current.
> >
> > Any advice?
>
> Use -current. Always use -current, especially on desktops.

Thnks all.

With -current and drm-devel-kmod I see /dev/dri but Xorg still
fails. Details below. I haven't touched the FreeBSD kernel in
a while but I may be able to help with the missing pieces.
[Though I suspect there are other problems as to why Xorg fails]

Xorg.0.log:
	...
      [  1837.611] (EE) open /dev/dri/card0: Operation not supported by device
      [  1837.611] (WW) Falling back to old probe method for modesetting
      [  1837.612] (EE) open /dev/dri/card0: Operation not supported by device
      [  1837.612] (WW) Falling back to old probe method for scfb
      [  1837.612] scfb trace: probe start
      [  1837.612] (II) scfb(1): using default device
      [  1837.612] scfb trace: probe done
      [  1837.612] (WW) VGA arbiter: cannot open kernel arbiter, no multi-card support
      [  1837.612] (EE) Screen 0 deleted because of no matching config section.
      [  1837.612] (II) UnloadModule: "modesetting"
      [  1837.612] (EE)
      Fatal server error:
      [  1837.612] (EE) Cannot run in framebuffer mode. Please specify busIDs        for all framebuffer devices

/dev/dri/card0 does exist.

dmesg reveals messages like
    ...
    i_size_write not implemented -- see your local kernel hacker
    ...
    get_nr_swap_pages not implemented -- see your local kernel hacker
    ...
    sched_setscheduler not implemented -- see your local kernel hacker
    ...
    pm_runtime_mark_last_busy not implemented -- see your local kernel hacker
    drmn0: [drm:*ERROR* ] amdgpu_ring_test_helper[drm ERROR :gfx_v9_0_kiq_kcq_enable] KCQ enable failed
    [drm ERROR :amdgpu_device_ip_hw_init_phase2] hw_init of IP block <gfx_v9_0> failed -60
    drmn0: amdgpu_device_ip_init failed
    drmn0: Fatal error during GPU init
    unregister_acpi_notifier not implemented -- see your local kernel hacker
    [drm] amdgpu: finishing device.
    Memory manager not clean during takedown.

I am not sure why inode size is being written (may be
something to do with sysfs in linux?).