#1042818 firmware-amd-graphics: Random display freezes on certain AMD GPUs due to "Error waiting for DMUB idle: status=3" #1042818
- Package:
- firmware-amd-graphics
- Source:
- firmware-amd-graphics
- Submitter:
- Gregor Riepl
- Date:
- 2026-05-13 18:19:02 UTC
- Severity:
- normal
- Tags:
Dear Maintainer, The current AMDGPU firmware in Debian has compatibility issues with 6.3+ kernels. These errors manifest themselves with kernel messages like these: [ +0.226777] [drm:dc_dmub_srv_wait_idle [amdgpu]] *ERROR* Error waiting for DMUB idle: status=3 [ +4.020959] [drm:dc_dmub_setup_subvp_dmub_command [amdgpu]] *ERROR* Error waiting for DMUB idle: status=3 Furthermore, they cause sudden display freezes and even GPU lock-ups that require power-cycling the system. It's not clear why these problems occur, but they might have to do with certain optimizations that AMD had to do to reduce power consumption at idle on RX 7000 series GPUs. See https://gitlab.freedesktop.org/drm/amd/-/issues/2315 for more information about this issue. As a temporary workaround, it's possible to avoid the power management optimizations by reducing the overall pixel clock rate or creating modelines with a longer blanking delay, as long as the display supports this. For example, reducing the refresh rate from 120Hz to 60Hz has helped in one case for me. Another workaround is to disable optimizations in the affected firmware with the kernel option drm.vblankoffdelay=0 . As stated in https://gitlab.freedesktop.org/drm/amd/-/issues/1887#note_1993615 , the problematic firmware changes were reverted in https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux- firmware.git/commit/?id=d3f66064cf43bd7338a79174bd0ff60c4ecbdf6d , and there have been several amdgpu firmware commits since. Please update linux-firmware as soon a release containing the revert or a permanent fix is available. Thanks.
linux-firmware 20230804 has been released and contains the mentioned reverts for amdgpu firmware. This is not a permanent fix of the underlying problem, but it will at least allow systems to function normally again. Please update as soon as possible. This should also fix #1040185, but for that issue, a backport to bookworm may be required. Thanks.
Hi, Thanks for looking into this, but I have some questions. Partially revert "amdgpu: DMCUB updates for DCN 3.1.4 and 3.1.5" That does indeed seem relevant to the error msgs you posted in this bug. That commit links to https://gitlab.freedesktop.org/drm/amd/-/issues/2666 which is rather different from the 'Forwarded' URL/issue. Was the initial Forwarded incorrect or am I missing something? "This partially reverts commit ade163aaaeae0c1ad20cb3dd8ce878bf61c91b3a" is also part of that commit msg, but that commit was NOT part of version 20230515, but 20230625 which has not (yet?) been uploaded to Debian. Which leads me to conclude that you shouldn't have been affected as Debian's firmware package (version) does not include the commit which caused the issue. Can you shed some light on this discrepancy? Similar as above as it shouldn't have the commit _causing_ the issue.
Dear Maintainer, I think the most relevant new information is that after installing 20230625-1 from 20230515-3 on kernel version 6.0.0-2, I had the random freezes mentioned above. Also, only my DP monitor would output anything, and my wayland compositor would try to make sense of them in a loop, leading to a very unresponsive system (freezes, missed keystrokes, repeated keystrokes) and other things like keyboard settings not being applied because the config reload process got interrupted. I tried upgrading to linux 6.5.0-5, which didn't do anything. Still on that version, I manually installed 20230515-4 (instead of the original -3, but I don't think this makes a difference) and everything is working fine. The workaround for drm kernel parameters did not work for me unfortunately. Thanks, Raphaël Gomès