Bug 108825

Created attachment 142561 [details]
dmesg

bpc seems to be detected as 6, but visually there does not appear to be any kind of dithering going on, resulting in bad banding during movies etc in addition to the test I posted in the last message.

Created attachment 142562 [details]
xorg log

Created attachment 142563 [details]
dmesg with drm.debug=0x1

Looks like this is a regression. Dithering works in 4.19.2 but does not work in amd-staging-drm-next.

I have bisected it to commit: drm/amd/display: Support amdgpu "max bpc" connector property
307638884f721b02b6a54ee8b351c5b4434bd4a6 Author: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>

(In reply to Samantha McVey from comment #5)
> I have bisected it to commit: drm/amd/display: Support amdgpu "max bpc"
> connector property
> 307638884f721b02b6a54ee8b351c5b4434bd4a6 Author: Nicholas Kazlauskas
> <nicholas.kazlauskas@amd.com>

This isn't a regression from 4.19 but rather a fix for 4.18 where bpc was uncapped from 8 unconditionally. It locked out high refresh modes for high resolution panels with max bpc greater than 8.

The reported 6bpc in your dmesg log comes from DRM itself and not amdgpu. It's what's being reported as being supported by your panel from its EDID. I'm not sure if this is the panel itself reporting itself wrong or if there's a quirk affecting it from the edid qurik list in drm_edid.c.

You can try reverting my patch on amd-staging-drm-next to see if you still get the banding.

Or you could try setting the "max bpc" property to something higher (like 16). You should be able to add this as an "Option" to a conf file in /etc/X11/xorg.conf.d for your Monitor.

Nicholas,

Reverting that commit (307638884f721b02b6a54ee8b351c5b4434bd4a6) on the top of amd-staging-drm-next resolves the banding issues. So it is that commit that is causing the problem.

Created attachment 142652 [details] [review]
0001-drm-amd-display-Fix-unintialized-max_bpc-state-value.patch

Does this help fix the issue?

It doesn't compile with that patch:

drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm.c: In function ‘dm_crtc_duplicate_state’:
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm.c:2991:7: error: ‘struct dm_crtc_state’ has no member named ‘max_bpc’
  state->max_bpc = cur->max_bpc;
       ^~
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm.c:2991:22: error: ‘struct dm_crtc_state’ has no member named ‘max_bpc’
  state->max_bpc = cur->max_bpc;
                      ^~

Created attachment 142653 [details] [review]
v2-0001-drm-amd-display-Fix-unintialized-max_bpc-state-va.patch

I meant to copy the property for the dm connector state, sorry.

Try this patch instead.

new_state->max_bpc = cur->max_bpc;

cur doesn't exist here (fails to compile with v2 patch). Did you mean state->max_bpc?

Nicholas,

Your last patch but with `new_state->max_bpc = state->max_bpc;` instead of `new_state->max_bpc = cur->max_bpc;` resolves the issue.

Going to close this one as fixed.