Bug 101961

It might be related to the CLEAR_STATE changes, can you checkout to 5c1241268ba9b240cb79ab9a30c5255b176c83c9 and try to reproduce the issue? Thanks!

(In reply to Samuel Pitoiset from comment #1)
> It might be related to the CLEAR_STATE changes, can you checkout to
> 5c1241268ba9b240cb79ab9a30c5255b176c83c9 and try to reproduce the issue?
> Thanks!

It still hangs on 5c1241268ba9b240cb79ab9a30c5255b176c83c9

(In reply to network723 from comment #2)
 
> It still hangs on 5c1241268ba9b240cb79ab9a30c5255b176c83c9

I've tried bisecting, and looks like the bad commit somewhere between 	7257c171e9eadc05903140cffa26a253f0d0178a and 5c1241268ba9b240cb79ab9a30c5255b176c83c9

Unfortunately, I have no time to bisect it further.

It looks weird, but can you try reverting 5c1241268b ?

(In reply to Samuel Pitoiset from comment #4)
> It looks weird, but can you try reverting 5c1241268b ?

reverting 5c1241268b helps, no hangs so far.

I reverted 5c1241268b. Please reopen if you still get the hang.

Actually, I may have been wrong about revering 5c1241268b fixes the crash. After a really long session in the game system froze again. Here are last lines from the journal:

kernel: gmc_v8_0_process_interrupt: 7 callbacks suppressed
kernel: amdgpu 0000:28:00.0: GPU fault detected: 147 0x00004801
kernel: amdgpu 0000:28:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x05500000
kernel: amdgpu 0000:28:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x02048001
kernel: amdgpu 0000:28:00.0: VM fault (0x01, vmid 1) at page 89128960, read from 'TC4' (0x54433400) (72)
kernel: amdgpu 0000:28:00.0: GPU fault detected: 147 0x00004801
kernel: amdgpu 0000:28:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x05500000
kernel: amdgpu 0000:28:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x02048001
kernel: amdgpu 0000:28:00.0: VM fault (0x01, vmid 1) at page 89128960, read from 'TC4' (0x54433400) (72)
kernel: amdgpu 0000:28:00.0: GPU fault detected: 147 0x00004801
kernel: amdgpu 0000:28:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x05500000
kernel: amdgpu 0000:28:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x02048001
kernel: amdgpu 0000:28:00.0: VM fault (0x01, vmid 1) at page 89128960, read from 'TC4' (0x54433400) (72)

Okay, I should be able to reproduce tomorrow, I will let you know.

By the way, do you have some hints about how to reproduce, like which settings, which steps, etc? Also, what version is exactly, I mean the Steam appid?

(In reply to Samuel Pitoiset from comment #9)
> By the way, do you have some hints about how to reproduce, like which
> settings, which steps, etc? Also, what version is exactly, I mean the Steam
> appid?

It's Serious Sam Fusion 2017 steam current version.
Steam AppID: 564310
DLC: 564324 564323 612560 564321 564322 564320

Graphics options
Graphics API: OpenGL
Resolution: 1920x1200
Window Mode: Fullscreen
Max 3D Rendering MPIX: 3.7(QHD 1440)
MSAA: Medium 4x
VSync: disabled

Performance options
CPU Speed: Medium
GPU Speed: Medium
GPU Memory: Medium
Level Caching: High
MAX FPS: Unlimited

The last time it crashed yesterday I was playing multiplayer, but it should crash on 'Death Canyon' map in single player too (Single Player -> Survival -> Multiplayer Levels BFE -> Death Canyon).

Unfortunately, I can't reproduce the issue with 3f38e64270c03c9a9eb5368c06dcfd1896fbf6d0 and the same settings, etc.

Can you try again with latest mesa and boot your kernel with amdgpu.vm_debug=1? It should be easier to catch all VM faults.

Though, ideally it would be better to record an apitrace which reproduces the hang.

(In reply to Samuel Pitoiset from comment #11)
> Unfortunately, I can't reproduce the issue with
> 3f38e64270c03c9a9eb5368c06dcfd1896fbf6d0 and the same settings, etc.
> 
> Can you try again with latest mesa and boot your kernel with
> amdgpu.vm_debug=1? It should be easier to catch all VM faults.
> 
> Though, ideally it would be better to record an apitrace which reproduces
> the hang.

Sorry it took so long, I've been updating entire distro, and had trouble building Mesa. Currently I'm on git master 4468764ef0cd0e71db03e14aaed7c650ffa1f77d & llvm revision 310200, and I'm not able to reproduce hangs myself.

Did you boot with amdgpu.vm_debug=1? Anyway, I think the previous hangs were related to the CLEAR_state changes which are now fixed in master.

As I already said, this was probably related to the CLEAR_state changes. I can't reproduce myself, closing. Please re-open if you still have the issue. Thanks!