amdgpu intermittently fails to resume correctly

[Jamie Scott]

I started seeing my system fail to resume from suspend over the last couple of weeks. It happens about 1 in 5 times maybe? I press the power button on my system to resume it and my screens stay blank. I am able to connect to the system remotely over ssh and check syslog to find out what went wrong. I then try to reboot the system with `sudo reboot`, the ssh session disconnects but it never actually reboots. I have to press the reset button after waiting a while to make it start up again.

dmesg log is attached which contains more info and the call trace. I think the interesting part is this:

kernel: amdgpu 0000:0d:00.0: [drm:amdgpu_ring_test_helper [amdgpu]] *ERROR* ring gfx test failed (-110)
kernel: [drm:amdgpu_device_ip_resume_phase2 [amdgpu]] *ERROR* resume of IP block <gfx_v8_0> failed -110
kernel: [drm:amdgpu_device_resume [amdgpu]] *ERROR* amdgpu_device_ip_resume failed (-110).
kernel: PM: dpm_run_callback(): pci_pm_resume+0x0/0xf0 returns -110
kernel: PM: Device 0000:0d:00.0 failed to resume async: error -110

It first happened on 20.04 with kernel 5.4 (can't remember exactly what version of 5.4) a couple of weeks ago. I saw 20.10 was available with 5.8 so thought I'd try upgrading as there was a variety of amdgpu fixes go in between 5.4 and 5.8 but sadly no dice.

From some research, I think what I'm seeing is similar to this bug, but that one looks like 5.4 solved it, whereas I guess some minor upgrade on 5.4 appeared to break it for me. https://bugs.freedesktop.org/show_bug.cgi?id=112221

I'm running a Ryzen 3700X on an X570 board with an AMD Radeon Pro WX 4100 GPU.
Currently using Kubuntu 20.10.

ProblemType: Bug
DistroRelease: Ubuntu 20.10
Package: linux-image-5.8.0-33-generic 5.8.0-33.36
ProcVersionSignature: Ubuntu 5.8.0-33.36-generic 5.8.17
Uname: Linux 5.8.0-33-generic x86_64
ApportVersion: 2.20.11-0ubuntu50.3
Architecture: amd64
CasperMD5CheckResult: skip
CurrentDesktop: KDE
Date: Sat Jan 2 12:40:36 2021
InstallationDate: Installed on 2020-06-18 (197 days ago)
InstallationMedia: Kubuntu 20.04 LTS "Focal Fossa" - Release amd64 (20200423)
MachineType: To Be Filled By O.E.M. To Be Filled By O.E.M.
ProcFB: 0 amdgpudrmfb
ProcKernelCmdLine: BOOT_IMAGE=/vmlinuz-5.8.0-33-generic root=/dev/mapper/vgkubuntu-root ro quiet splash usbcore.autosuspend=-1 vt.handoff=7
RelatedPackageVersions:
 linux-restricted-modules-5.8.0-33-generic N/A
 linux-backports-modules-5.8.0-33-generic N/A
 linux-firmware 1.190.2
SourcePackage: linux
UpgradeStatus: Upgraded to groovy on 2020-12-30 (3 days ago)
dmi.bios.date: 09/09/2019
dmi.bios.release: 5.14
dmi.bios.vendor: American Megatrends Inc.
dmi.bios.version: P2.10
dmi.board.name: X570 Taichi
dmi.board.vendor: ASRock
dmi.chassis.asset.tag: To Be Filled By O.E.M.
dmi.chassis.type: 3
dmi.chassis.vendor: To Be Filled By O.E.M.
dmi.chassis.version: To Be Filled By O.E.M.
dmi.modalias: dmi:bvnAmericanMegatrendsInc.:bvrP2.10:bd09/09/2019:br5.14:svnToBeFilledByO.E.M.:pnToBeFilledByO.E.M.:pvrToBeFilledByO.E.M.:rvnASRock:rnX570Taichi:rvr:cvnToBeFilledByO.E.M.:ct3:cvrToBeFilledByO.E.M.:
dmi.product.family: To Be Filled By O.E.M.
dmi.product.name: To Be Filled By O.E.M.
dmi.product.sku: To Be Filled By O.E.M.
dmi.product.version: To Be Filled By O.E.M.
dmi.sys.vendor: To Be Filled By O.E.M.

[Ubuntu Kernel Bot (ubuntu-kernel-bot)]

This change was made by a bot.

[Jamie Scott (jamiescott)]

Ah, the included dmesg output doesn't actually have the useful snippet in it. Please see attached section of log output which runs from when I attempted to resume the system to when it was rebooted.

[Kleber Sacilotto de Souza (kleber-souza)]

Hello Jamie or anyone else affected,

We have released a new Focal linux kernel (version 5.4.0-59.65) which contains several bug fixes. Could you please update to this kernel and report whether it fixes the issue?

Thank you.

[Jamie Scott (jamiescott)]

Sorry, I may not have made it obvious. The problem started on Focal but I have since upgraded to Groovy as I thought a newer kernel may help given various amdgpu fixes hit the kernel between 5.4 and 5.8.

I do see a pending update for 5.8.0-34 though which I'll install.