AMD GPU freezes randomly

Question

I have a ThinkPad A285 with Ubuntu 18.04.2 installed. It's equipped with Ryzen 7 2700U CPU. It seems that GPU randomly freezes. It doesn't matter if there's any load on it or not.

I tried to upgrade the kernel from 4.18.0 to 5.0.2, but the problem persists.

I also tried to install AMD drivers from here, but it doesn't seem to install and from further research it looks like it only works with 4.16 kernel. I tried downgrading, but then I'd get stuck in a login loop.

/var/log/syslog just before the crash:

Mar 14 19:35:14 kernel: [ 1505.210444] amdgpu 0000:05:00.0: [gfxhub] VMC page fault (src_id:0 ring:24 vmid:4 pasid:32779)
Mar 14 19:35:14 kernel: [ 1505.210447] amdgpu 0000:05:00.0:   at page 0x000000010a208000 from 27
Mar 14 19:35:14 kernel: [ 1505.210451] amdgpu 0000:05:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x00000000
Mar 14 19:35:14 kernel: [ 1505.210459] amdgpu 0000:05:00.0: [gfxhub] VMC page fault (src_id:0 ring:24 vmid:4 pasid:32779)
Mar 14 19:35:14 kernel: [ 1505.210462] amdgpu 0000:05:00.0:   at page 0x000000010a208000 from 27
Mar 14 19:35:14 kernel: [ 1505.210465] amdgpu 0000:05:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x00000000
Mar 14 19:35:24 kernel: [ 1515.228395] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx timeout, last signaled seq=92425, last emitted seq=92427
Mar 14 19:35:24 kernel: [ 1515.228402] [drm] GPU recovery disabled.

Answer 1

I had the same issue for a long time. It was driving me crazy. Especially annoying as it would freeze randomly during occasional gaming session (using Steam). Freezing was reduced a lot after adding a kernel parameter, amdgpu.gpu_recovery=1. in the boot string. However, this is only the case if you are using the amdgpu driver.

1. Edit /etc/default/grub.

2. Modify the value of the variable GRUB_CMDLINE_LINUX_DEFAULT to include amdgpu.gpu_recovery=1.

3. Run:

   sudo grub-update
   

4. Reboot.

My GRUB config line looks like this, as I'm forcing the use of amdgpu driver, even though my R9 280 is supposedly not supported (works great):

GRUB_CMDLINE_LINUX_DEFAULT="quiet splash amdgpu.si_support=1 radeon.si_support=0 amdgpu.cik_support=1 radeon.cik_support=0 amdgpu.gpu_recovery=1 elevator=noop"

If you are wondering what the elevator=noop is, it's supposed to help with SSD wear.

Hope this helps.

Answer 2

I'm using amdgpu 19.10 and it works fine for me.

https://www.amd.com/en/support/kb/release-notes/rn-rad-lin-19-10-unified

I recommend you to use amdgpu(All-open driver) not amgdpu-pro driver. Using pro-driver causes some applications crash frequently (firefox etc...)

• ThinkCenter M715q Tiny/Ryzen 5 Pro 2500U (BIOS M1XKT41A)
• Kernel linux-image-4.18.0-21-generic
• Driver amdgpu-core 19.10-785425
• Ubuntu on Wayland

Until then I had used ppa:oibaf/graphics-drivers. It seemed to work very fine. But I cannot launch GUI session from this month. I'm not sure it caused by updated kernel, amdgpu-driver or etc.

19.20 a new version has been released. I'll try it later. https://www.amd.com/en/support/kb/release-notes/rn-rad-lin-19-20-unified