Bug 93436

Summary:

[radeonsi][regression] GPUVM faults playing ARK: Survival Evolved, started 4.4-rc3, -rc5 is worse

Product:

DRI

Reporter:

Shawn Starr <shawn.starr>

Component:

DRM/Radeon

Assignee:

Default DRI bug account <dri-devel>

Status:

RESOLVED DUPLICATE

QA Contact:

Severity:

normal

Priority:

medium

Version:

XOrg git

Hardware:

Other

OS:

All

Whiteboard:

i915 platform:

i915 features:

Attachments:

Description	Flags
R600_DEBUG=check_vm capture of VM fault	none

Description Shawn Starr 2015-12-18 07:58:05 UTC

Kernel is spamming non-fatal GPU VM faults:

[ 8792.444427] VM fault (0x02, vmid 5) at page 600320, read from 'TC2' (0x54433200) (72)
[ 8792.475656] radeon 0000:01:00.0: GPU fault detected: 147 0x000a4802
[ 8792.475976] radeon 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x00092900
[ 8792.476343] radeon 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0A048002
[ 8792.476718] VM fault (0x02, vmid 5) at page 600320, read from 'TC2' (0x54433200) (72)
[ 8792.505901] radeon 0000:01:00.0: GPU fault detected: 147 0x000a4802
[ 8792.506232] radeon 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x00092900
[ 8792.506596] radeon 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0A048002
[ 8792.506970] VM fault (0x02, vmid 5) at page 600320, read from 'TC2' (0x54433200) (72)
[ 8792.537730] radeon 0000:01:00.0: GPU fault detected: 147 0x000a4802
[ 8792.538049] radeon 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x00092900
[ 8792.538416] radeon 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0A048002
[ 8792.538783] VM fault (0x02, vmid 5) at page 600320, read from 'TC2' (0x54433200) (72)
[ 8792.567524] radeon 0000:01:00.0: GPU fault detected: 147 0x000a4802
[ 8792.567849] radeon 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x00092900
[ 8792.568216] radeon 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0A048002
[ 8792.568586] VM fault (0x02, vmid 5) at page 600320, read from 'TC2' (0x54433200) (72)
[ 8795.015367] radeon 0000:01:00.0: GPU fault detected: 147 0x0c0a4802
[ 8795.015698] radeon 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x000928E0
[ 8795.016065] radeon 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0A048002
[ 8795.016438] VM fault (0x02, vmid 5) at page 600288, read from 'TC2' (0x54433200) (72)
...

Mesa from git master: 2015-12-11
kernel: 4.4.0-0.rc5.git1.2.fc24.x86_64

Can I use the new debugging in mesa to capture what is causing fault?

Comment 1 Alex Deucher 2015-12-18 15:09:58 UTC

Is this a regression?  If so, can you bisect?

Comment 2 Shawn Starr 2015-12-18 23:58:59 UTC

Created attachment 120588 [details]
R600_DEBUG=check_vm capture of VM fault

Comment 3 Shawn Starr 2015-12-19 01:50:23 UTC

Google drive link to apitrace: https://drive.google.com/file/d/0Bze7CJKD12nOWXZ0WnhOSS00aW8/view?usp=sharing

Comment 4 Nicolai Hähnle 2015-12-19 07:22:49 UTC

For what it's worth, I cannot reproduce on Tonga with latest Mesa and LLVM master. Can you confirm that you still get faults after LLVM r256072? Your ddebug_dump contains very large shaders, it would not be surprising if this turns out to be a duplicate of bug 93264.

Comment 5 Shawn Starr 2015-12-19 10:38:39 UTC

Yes, appears to be duplicate, will continue playing game later today but it looked good in glretrace and actual game play no GPU VM faults noted.

*** This bug has been marked as a duplicate of bug 93264 ***

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.