Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.debian.kernel > #93320 > unrolled thread

Bug#1141866: linux-image-7.1.3+deb14-amd64: no display on AMD Kaveri/CIK APU VGA output (amdgpu DC regression)

Back to article view | Back to linux.debian.kernel

Contents

  Bug#1141866: linux-image-7.1.3+deb14-amd64: no display on AMD Kaveri/CIK APU VGA output (amdgpu DC regression) Lionso Alejandro Pacheco Vacacela <lionso@murena.io> - 2026-07-11 21:50 +0200
    Bug#1141866: [7.1] drm/amd/display: Add dp_skip_rbr flag for NUTMEG Ben Hutchings <ben@decadent.org.uk> - 2026-07-26 16:10 +0200
      Bug#1141866: [7.1] drm/amd/display: Add dp_skip_rbr flag for NUTMEG Sasha Levin <sashal@kernel.org> - 2026-07-28 00:30 +0200

#93320 — Bug#1141866: linux-image-7.1.3+deb14-amd64: no display on AMD Kaveri/CIK APU VGA output (amdgpu DC regression)

[Multipart message — attachments visible in raw view] — view raw

Package: src:linux
Version: 7.1.3-1
Severity: important
Tags: patch upstream fixed-upstream
X-Debbugs-Cc: lionso@murena.io

Dear Maintainer,

Since the 7.1.x kernels, systems with AMD CIK/GCN1.1 APUs (Kaveri,
Kabini, Mullins) that use the analog VGA output get no display signal
at all: the monitor enters power save at KMS takeover and never wakes.
The machine is otherwise healthy (reachable over SSH). 7.0.x kernels
are fine.

Hardware here: AMD A10-7850K (Kaveri, DCE 8.1), MSI A68HM-E33, VGA
monitor on the onboard output. On these boards the VGA connector is
driven through a NUTMEG DP-to-VGA bridge chip.

Root cause (upstream): Linux 7.1 switched CIK APUs to the amdgpu DC
display stack by default (c81f35cdbab2 "drm/amdgpu: Use DC by default
on CIK APUs"). DC's support for the NUTMEG bridge requires forcing the
DP link rate to HBR, but upstream commit

  a62346043a89 ("drm/amd/display: Fix coding style issue")

inadvertently broke that quirk (the preferred link rate is no longer
applied to the link setting the algorithm iterates from), so DP link
training against the bridge fails at RBR and the output stays dark:

  [drm] enabling link 2 failed: 15   (DC_FAIL_DP_LINK_TRAINING)

Upstream fix: "drm/amd/display: Fix preferred link rate for NUTMEG" by
Timur Kristóf,
https://lore.kernel.org/amd-gfx/20260529090909.13206-1-timur.kristof@gmail.com/
Reviewed-by: Fangzhi Zuo, applied to the amdgpu maintainer tree on
2026-07-10; not yet in mainline or 7.1.y at the time of writing. Also
tracked as https://gitlab.freedesktop.org/drm/amd/-/issues/5465.

I verified the fix on the affected machine: applying just that one
patch on top of the 7.1.3 Debian kernel source (config and
Module.symvers from linux-image/headers 7.1.3+deb14-amd64, only
amdgpu.ko rebuilt) restores VGA output with DC enabled, across three
consecutive reboots.
The patch is attached; it is small (3 lines, one file:
drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c)
and applies cleanly to 7.1.3.

Please consider cherry-picking it into the next src:linux upload
rather than waiting for it to trickle through 7.1.y — the regression
makes the console unusable on all Kaveri/Kabini/Mullins VGA systems
(this affects testing/forky too, which ships the same 7.1.3-1),
and its Fixes: tag references a non-mainline hash, so the stable
autoselection may not pick it up automatically (reported upstream).

Workaround for affected users meanwhile: boot with amdgpu.dc=0.

Additional evidence available on request: full dmesg of good/bad boots, DRM
atomic state showing no active CRTC on the bad kernel.

[toc] | [next] | [standalone]

#93465 — Bug#1141866: [7.1] drm/amd/display: Add dp_skip_rbr flag for NUTMEG

[Multipart message — attachments visible in raw view] — view raw

On Sat, 11 Jul 2026 14:35:30 -0500 Lionso Alejandro Pacheco Vacacela
<lionso@murena.io> wrote:
[...]
> Since the 7.1.x kernels, systems with AMD CIK/GCN1.1 APUs (Kaveri,
> Kabini, Mullins) that use the analog VGA output get no display signal
> at all: the monitor enters power save at KMS takeover and never wakes.
> The machine is otherwise healthy (reachable over SSH). 7.0.x kernels
> are fine.
[...]
> Root cause (upstream): Linux 7.1 switched CIK APUs to the amdgpu DC
> display stack by default (c81f35cdbab2 "drm/amdgpu: Use DC by default
> on CIK APUs"). DC's support for the NUTMEG bridge requires forcing the
> DP link rate to HBR, but upstream commit
> 
>   a62346043a89 ("drm/amd/display: Fix coding style issue")
> 
> inadvertently broke that quirk (the preferred link rate is no longer
> applied to the link setting the algorithm iterates from), so DP link
> training against the bridge fails at RBR and the output stays dark:
> 
>   [drm] enabling link 2 failed: 15   (DC_FAIL_DP_LINK_TRAINING)
> 
> Upstream fix: "drm/amd/display: Fix preferred link rate for NUTMEG" by
> Timur Kristóf,
>
https://lore.kernel.org/amd-gfx/20260529090909.13206-1-timur.kristof@gmail.com/
> Reviewed-by: Fangzhi Zuo, applied to the amdgpu maintainer tree on
> 2026-07-10; not yet in mainline or 7.1.y at the time of writing. Also
> tracked as https://gitlab.freedesktop.org/drm/amd/-/issues/5465.
[...]

This ended up as commit 9fb646bc4d87f62bcbf0a7ea326430eb802c475c in
7.2-rc4.  Please queue it up for 7.1-stable.

Ben.

-- 
Ben Hutchings
Experience is what causes a person to make new mistakes
instead of old ones.

[toc] | [prev] | [next] | [standalone]

#93480 — Bug#1141866: [7.1] drm/amd/display: Add dp_skip_rbr flag for NUTMEG

On Sun, Jul 26, 2026 at 04:06:49PM +0200, Ben Hutchings wrote:
>This ended up as commit 9fb646bc4d87f62bcbf0a7ea326430eb802c475c in
>7.2-rc4.  Please queue it up for 7.1-stable.

Queued for 7.1, thanks.

I also took the dp_skip_rbr follow-up that your subject line names: it is
Cc: stable as well and it rewrites exactly the lines the fix touches, so
keeping the two together avoids a conflict in that file later on.

-- 
Thanks,
Sasha

[toc] | [prev] | [standalone]

Back to top | Article view | linux.debian.kernel

csiph-web