3200 Memory Crash My System

Submitted by bibitocarlos

Assigned to Default DRI bug account

Link to original bug (#111232)

Description

Created attachment 144891
dmesg

Hi,

Since 5.2 kernel, my system (raven ridge 2400G, Asrock X470 Gaming-ITX/ac, latest bios, archlinux) crashes (black screen or green screen or flickereing screen at start). I identified the issue with my RAM's speed, which is shared between the GPU and CPU.
Everything above 2133Mhz crashes.
I dont have this issue with 5.1 kernel.
Dsmeg of the crash as attachment


Bibi

Attachment 144891, "dmesg":
2900.log

added AMDgpu bugzilla labels

Alex Deucher @agd5f said:

Can you bisect?

bibitocarlos said:

(In reply to Alex Deucher from comment 1)

Can you bisect?

I'm gonna try ^^, never did it before and leave to holydays Friday.
I'm gonna start with RC, then whith commits.
You think it's a proper way ?

bibitocarlos said:

Ok, back from holidays.
Just compiled 5.12 RC1 and it doesnt work, green screen.
Any idea which commit i should test ?

bibitocarlos said:

I have finished bisecting.
This is the commit which created the bug :
" Bissection : 0 révision à tester après ceci (à peu près 0 étape)
[005440066f929ba0dca8f4e0aebfbf8daac592cc] drm/amdgpu: enable gfxoff again on raven series (v2)"

What do i do next ?

Bibi

bibitocarlos said:

Seems to be related to this bug:
https://lore.kernel.org/lkml/624BFB8F-B586-492E-EA6-4B138DAEC831@canonical.com/T/

bibitocarlos said:

Tried amd.iommu off but doesnt work, so it's something else.

bibitocarlos uploaded an attachment:

Tried this "amdgpu.ppfeaturemask=0xffff3fff" but iit doesnt work (from here: https://bugzilla.kernel.org/show_bug.cgi?id=204689).

Add new dmesg to check firmware version

Attachment 145321, "dmesg":
dmesg.log

bibitocarlos said:

And cant revert the commit with 5.3, i get git conflict...

Andrey Grodzovsky @agrodzov said:

I noticed IOMMU page fault in the log - is disabling iommu helps ? (from grub cmdline add iommu=off)

bibitocarlos said:

Tried with "iommu=offiommu=off amdgpu.ppfeaturemask=0xffff3fff", no change, green screen.

bibitocarlos said:

Any idea ?

bibitocarlos said:

Fixed with latest firmware

bibitocarlos said:

I reopened it, it wwas not fixes with latest firmware, my bad, it was my memory which was clocked at 2133MHz by the bios...

There are two other bug reports, that migth be related, since they are also about green screen crashes since 5.2 that happen with Ryzen APUs:

#983 (closed)

https://bugzilla.kernel.org/show_bug.cgi?id=205589

mentioned in issue #983 (closed)

Hi with Linux 5.4.2-1, i can now go to 3000MHZ. dont know if the other bugs are related. Bibi

Confirm that it works with 3000MHz after overclocking the RAM to 1.35V.

"echo "manual" > /sys/class/drm/card0/device/power_dpm_force_performance_level echo "1 2 3" > /sys/class/drm/card0/device/pp_dpm_mclk" doesnt change anything.

I went back to 5.1 because i want my RAM Fullspeed, and 3000MHz is not everytime stable ( it crashes sometimes when i test the RAM with memtest).

For the firmware, i went back to the one from october, there is an issue with everything after which crashes the computer when using deinterlacing VAAPI MADI (will come with another bug when i have time).

Hope we can have one day an easy option to use latest kernel and desactivate RAM Voltage clocking from the driver.

Bibi

This issue hasn't had any activity since 2019-12-25. The AMD driver stack changes rapidly and contains lots of shared code across products so it's possible that it has already been fixed. Please upgrade to a current stable kernel and userspace stack and try again. If you still experience this issue with the latest driver stack, please capture relevant logging and open a new issue referring back to this one.

closed