Bug 202667 - amdgpu fails to boot: atombios stuck executing D6A0
Summary: amdgpu fails to boot: atombios stuck executing D6A0
Status: RESOLVED INVALID
Alias: None
Product: Drivers
Classification: Unclassified
Component: Video(DRI - non Intel) (show other bugs)
Hardware: x86-64 Linux
: P1 normal
Assignee: drivers_video-dri
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2019-02-24 10:27 UTC by Dennis Wagelaar
Modified: 2019-02-25 12:09 UTC (History)
0 users

See Also:
Kernel Version: 4.20.5
Subsystem:
Regression: No
Bisected commit-id:


Attachments
journalctl --no-hostname -k for kernel-4.20.4-100.fc28.x86_64 (92.81 KB, text/plain)
2019-02-25 11:20 UTC, Dennis Wagelaar
Details
journalctl --no-hostname -k for kernel-4.20.5-100.fc28.x86_64 (87.57 KB, text/plain)
2019-02-25 11:20 UTC, Dennis Wagelaar
Details
Active kernel module parameters for 4.20.4 (11.25 KB, text/plain)
2019-02-25 11:44 UTC, Dennis Wagelaar
Details
journalctl --no-hostname -k for kernel-4.20.8-100.fc28.x86_64 with amdgpu.dc=1 (87.46 KB, text/plain)
2019-02-25 12:08 UTC, Dennis Wagelaar
Details

Description Dennis Wagelaar 2019-02-24 10:27:22 UTC
Starting from kernel 4.20.5, the plymouth bootup screen, as well as the textual framebuffer is garbled (see https://bugzilla.redhat.com/show_bug.cgi?id=1674065 for attachments), and the wayland graphical login never starts up. Instead, the monitor goes into power save (no video signal). There is no recovery from this state, and I have to power cycle the computer.

All is fine on kernel-4.20.4, albeit that I have to disable amdgpu dc (options amdgpu dc=0), because it causes screen flicker. It appears dc was enabled by default since kernel 4.20, whereas it was disabled by default before that.

Suspicious dmesg output:

feb 08 22:50:34 kernel: [drm:atom_op_jump [amdgpu]] *ERROR* atombios stuck in loop for more than 5secs aborting
feb 08 22:50:34 kernel: [drm:amdgpu_atom_execute_table_locked [amdgpu]] *ERROR* atombios stuck executing D6A0 (len 978, WS 0, PS 0) @ 0xD7EF
feb 08 22:50:34 kernel: [drm:amdgpu_atom_execute_table_locked [amdgpu]] *ERROR* atombios stuck executing D55A (len 326, WS 0, PS 0) @ 0xD5F6

Hardware:
- AMD Ryzen 1600X CPU
- Asus B350M-A PRIME motherboard, BIOS version 4207
- MSI Radeon RX 560 AERO ITX 4G OC graphics card
- Philips 276E8VJSB (4K) monitor

Software:
- Fedora 28 x86_64
Comment 1 Michel Dänzer 2019-02-25 09:18:34 UTC
Please attach the corresponding full output of dmesg.

Can you bisect between 4.20.4 and 4.20.5?
Comment 2 Dennis Wagelaar 2019-02-25 11:20:00 UTC
Created attachment 281333 [details]
journalctl --no-hostname -k for kernel-4.20.4-100.fc28.x86_64
Comment 3 Dennis Wagelaar 2019-02-25 11:20:39 UTC
Created attachment 281335 [details]
journalctl --no-hostname -k for kernel-4.20.5-100.fc28.x86_64
Comment 4 Michel Dänzer 2019-02-25 11:31:50 UTC
The attached output from 4.20.4 shows DC being enabled. Does the problem really not happen with 4.20.4 and DC disabled?
Comment 5 Dennis Wagelaar 2019-02-25 11:44:06 UTC
Created attachment 281337 [details]
Active kernel module parameters for 4.20.4

Hmm, I have:

options amdgpu dc=0

in /etc/modprobe.d/amdgpu.conf, but my active module parameters say DC is enabled anyway...
Comment 6 Dennis Wagelaar 2019-02-25 12:08:13 UTC
Created attachment 281339 [details]
journalctl --no-hostname -k for kernel-4.20.8-100.fc28.x86_64 with amdgpu.dc=1

Just found out changing stuff in /etc/modprobe.d/ has no effect until the next run of dracut :$

It turns out I can no longer use amdgpu without DC=1. With that enabled, the newer kernel run as normal.

What probably happened is that I saw the flicker once, which is mentioned here: https://wiki.archlinux.org/index.php/AMDGPU : "If you experience flickering add amdgpu.dc=0 to your kernel parameters." That's when trouble began.

Note You need to log in before you can comment on or make changes to this bug.