protonscr

Vega 64 GPU hang in Waltz of the Wizard (proton)

steamvrclosed
ValveSoftware/SteamVR-for-Linux#154 · opened 2018-11-10 by ChristophHaag · updated 2018-11-27 · 3 comments · github
CChristophHaag 2018-11-10 github

Your system information

  • Steam client version (build number or date):
  • Distribution (e.g. Ubuntu): Archlinux
  • Graphics driver version (run nvidia-settings): mesa git and drm-next-4.21-wip from 2018-11-10
  • Gist for SteamVR System Information: SteamVR-2018-11-10-_02_22_29.txt
  • Opted into Steam client beta?: [Yes]
  • Opted into SteamVR beta?: [Yes]
  • Have you checked for system updates?: [Yes]

I noticed this in one Unity windows application run with proton that has been working fine on an RX 480, but caused a GPU hang on a Vega 64. The GPU hang does not happen on startup, there's first a simple scenario, but when loading another scenario, it immediately hangs.

I found another application that is similar: Waltz of the Wizard. The first tutorial room works, then you open a door, it renders one frame of the next scenario, and the GPU hangs. I have not tried that one on RX 480 but I assume it would work fine, too.

Only tested with latest mesa git and latest amdgpu development code from drm-next-4.21-wip, I will have to test with a more stable kernel later.

Nov 10 02:16:40 c-pc kernel: gmc_v9_0_process_interrupt: 132 callbacks suppressed
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: [gfxhub] VMC page fault (src_id:0 ring:171 vmid:5 pasid:32827, for process WaltzOfTheWizar pid 15942 thread WaltzOfTheWizar pid 15942)
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0:   in page starting at address 0x0000800140004000 from 27
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x00541157
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: [gfxhub] VMC page fault (src_id:0 ring:171 vmid:5 pasid:32827, for process WaltzOfTheWizar pid 15942 thread WaltzOfTheWizar pid 15942)
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0:   in page starting at address 0x0000800140004000 from 27
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x00541157
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: [gfxhub] VMC page fault (src_id:0 ring:171 vmid:5 pasid:32827, for process WaltzOfTheWizar pid 15942 thread WaltzOfTheWizar pid 15942)
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0:   in page starting at address 0x0000800140013000 from 27
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x00541157
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: [gfxhub] VMC page fault (src_id:0 ring:171 vmid:5 pasid:32827, for process WaltzOfTheWizar pid 15942 thread WaltzOfTheWizar pid 15942)
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0:   in page starting at address 0x0000800140014000 from 27
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x00000000
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: [gfxhub] VMC page fault (src_id:0 ring:171 vmid:5 pasid:32827, for process WaltzOfTheWizar pid 15942 thread WaltzOfTheWizar pid 15942)
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0:   in page starting at address 0x0000800140023000 from 27
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x00000000
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: [gfxhub] VMC page fault (src_id:0 ring:171 vmid:5 pasid:32827, for process WaltzOfTheWizar pid 15942 thread WaltzOfTheWizar pid 15942)
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0:   in page starting at address 0x0000800140024000 from 27
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x00000000
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: [gfxhub] VMC page fault (src_id:0 ring:171 vmid:5 pasid:32827, for process WaltzOfTheWizar pid 15942 thread WaltzOfTheWizar pid 15942)
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0:   in page starting at address 0x000080014000c000 from 27
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x00000000
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: [gfxhub] VMC page fault (src_id:0 ring:171 vmid:5 pasid:32827, for process WaltzOfTheWizar pid 15942 thread WaltzOfTheWizar pid 15942)
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0:   in page starting at address 0x000080014000d000 from 27
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x00000000
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: [gfxhub] VMC page fault (src_id:0 ring:171 vmid:5 pasid:32827, for process WaltzOfTheWizar pid 15942 thread WaltzOfTheWizar pid 15942)
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0:   in page starting at address 0x000080014000e000 from 27
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x00000000
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: [gfxhub] VMC page fault (src_id:0 ring:171 vmid:5 pasid:32827, for process WaltzOfTheWizar pid 15942 thread WaltzOfTheWizar pid 15942)
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0:   in page starting at address 0x000080014001b000 from 27
Nov 10 02:16:40 c-pc kernel: amdgpu 0000:0a:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x00000000
Nov 10 02:16:50 c-pc kernel: [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx timeout, signaled seq=18175598, emitted seq=18175600
Nov 10 02:16:50 c-pc kernel: amdgpu 0000:0a:00.0: GPU reset begin!
CChristophHaag 2018-11-10 github

Tested with a more reasonable setup: Linux 4.20-rc1, radv 18.3-rc2, LLVM 7.0.0.

The GPU stang still happens.

dmesg.txt

CChristophHaag 2018-11-17 github

If it helps rule out anything: This patch does not help https://patchwork.freedesktop.org/patch/256048/

CChristophHaag 2018-11-27 github

It works now on latest mesa git + llvm git.
screenshot_20181127_111323

Possible fix (not tested if it really was it): https://cgit.freedesktop.org/mesa/mesa/commit/?id=08ea6b9d9bb047603c249468dfe00d7bb9603d5e

Nothing extracted yet.