Built-in display doesn't work with AMDGPU on Linux 5.6 Macbook (MacBookPro13,3)

[dmarinuswoodwing]

When I update to the latest kernel (5.6.6) using Fedora 31 my built-in display shows a (backlighted) black screen.

Could someone please share their experience with a AMDGPU Macbook and linux-5.6?

[roadrunner2]

Yes, the amdgpu driver apparently got broken in the 5.6 kernel - I noticed because the /sys/kernel/debug/vgaswitcheroo/switch entry is missing (which means only the iGPU is being registered - the entry only shows up if there are at least 2 GPU's registered) and therefore I can't power down the dGPU anymore. (Note that I boot with the iGPU enabled and active, so the system would still be usable, just sucking way more power than I'd like).

I haven't had time to file a bug report, but it would be a good idea to do so.

[fannullone]

I am facing same problem on fc31 after last dnf upgrade which installed kernel-5.6
I can open the bug, but please can you give me some directions as I have never done it before.
I can provide log and such to facilitate the fix

[andyholst]

Linux kernel 5.7.0-050700rc3-generic

My eDP (builtin display) has no modes for 5.6+

xrandr --query | grep '\bconnected\b'

xrandr: Output eDP is not disconnected but has no modes
eDP connected (normal left inverted right x axis y axis)
DisplayPort-0 connected primary 1920x1080+0+0 (normal left inverted right x axis y axis) 521mm x 293mm

[fannullone]

opened bug: https://bugzilla.kernel.org/show_bug.cgi?id=207589

[meeuw]

I've also opened a bug for Fedora: https://bugzilla.redhat.com/show_bug.cgi?id=1829174

[meeuw]

I've posted a question at amd-gfx@lists.freedesktop.org and got the following reply:
https://lists.freedesktop.org/archives/amd-gfx/2020-May/049363.html

So I'll try to do a bisect in the next days...

[meeuw]

I've bisected the problem to:

b9f1246df179522bc28fda50b720553c845863db is the first bad commit
commit b9f1246df179522bc28fda50b720553c845863db
Author: Noah Abradjian noah.abradjian@amd.com
Date: Fri Nov 22 16:07:24 2019 -0500

drm/amd/display: Collapse resource arrays when pipe is disabled

[Why]
Currently, pipe resources are assigned to an index that matches the pipe position.
However, if pipe 1 or 2 is disabled, there will be a gap in the arrays which causes a crash when iterating based on pipe_count.

[How]
Fix resource construct to assign resources to minimum available array index.

Signed-off-by: Noah Abradjian <noah.abradjian@amd.com>
Reviewed-by: Yongqiang Sun <yongqiang.sun@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

.../gpu/drm/amd/display/dc/dcn21/dcn21_resource.c | 20 ++++++++++----------
1 file changed, 10 insertions(+), 10 deletions(-)

https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=b9f1246df179522bc28fda50b720553c845863db

[aunali1]

You guys might have luck with this patch. Its a quick and dirty hack to force all DPCD link rate values to whats expected by Retina panels. Currently, this will prevent externally attached displays from working.

[meeuw]

You guys might have luck with this patch. Its a quick and dirty hack to force all DPCD link rate values to whats expected by Retina panels. Currently, this will prevent externally attached displays from working.

Thanks I'll try that! Could you please report your findings at:
https://bugzilla.kernel.org/show_bug.cgi?id=207589

[andyholst]

I have reported my findings with the bisected kernel commits between v5.5.0 and v5.6.0 at https://bugzilla.kernel.org/show_bug.cgi?id=207589 with the related Xorg and dmesg logs.

[dmarinuswoodwing]

Hi @aunali1 thanks for posting your patch, that seems to fix it on my system!

Updated URL:
https://github.com/aunali1/linux-mbp-arch/blob/master/2001-drm-amd-display-Force-link_rate-as-LINK_RATE_RBR2-fo.patch

[Dunedan]

Is somebody currently working to get a fix for that upstreamed? What's the status there?

[aunali1]

@dmarinuswoodwing Thanks for testing! Great to see that it works for other models aswell.

@Dunedan Unfortunately, this patch is not upstream ready since it causes issues with 16inch MBPs. Additional testing is also needed for other models.