Manjaro on Btrfs not bootable after a system freeze (caused by Blender / AMD driver?)

Hi!

I was working on a project in Blender using the realtime rendering-engine Eevee (entirely running on the GPU) when suddenly my laptop has frozen.

I’ve rebooted it and realized my autosave was in /tmp/ so it’s gone.

I’ve redone my progress and in the exact same moment, when I connected two things in Blender - my system froze again. But this time - it failed to boot up again.

I’ve investigated what’s going on and found that the root Btrfs filesystem is damaged and is unmountable.

I suspect AMD GPU driver has a bug that was triggered by Blender’s GPU renderer, but how could that corrupt my root filesystem?

Note that I also had another Btrs filesystem mounted (for data) and that one is fine (thankfully).

I’m currently imaging the system drive (dual booting with Windows) and can’t mount it so I can’t provide any logs.

I am running Manjaro Linux KDE edition on Kernel 5.9 (AFAIK).

EDIT:

I’ve since had to completely trash the whole Manjaro Install and redo it - my backups were not perfect, and I was unable to restore the filesystem to fully working order. I’ve since made a proper disk image after the fresh installation in case I’ll need to do this again.

Still: it’s kinda scary that a few clicks in Blender can blow up my entire OS. I have to complete this project on Windows 10, to avoid unnecessary downtime.

Have anybody had similar issues?

Details (I can’t include links in my posts).

https: // developer . blender . org / T81547

Recently I answered to this post:

And i believe it is related to this bug: 206017 – Kernel 5.4.x unusable with GUI due to crashes (some hard)

Caused by the error msg:

[amdgpu]] *ERROR* Waiting for fences timed out!

I saw this also in your logs on ⚓ T81547 Blender has froze my Manajro Linux, which has corrupted my root Btrfs filesystem, rendering the machine unable to boot

Maybe this could be a direction. There is a kernel parameter, that could avoid this problem:

amdgpu.noretry=0

For more information:

modinfo amdgpu

PS: Since I don’t use a amd gpu, i cant test it, but maybe it is a direction how to workaround this issue.

1 Like