Bug 103561 - unable to handle kernel paging request with csgo in wine+nine, ttm_bo_del_from_lru
Summary: unable to handle kernel paging request with csgo in wine+nine, ttm_bo_del_fro...
Status: NEW
Alias: None
Product: Drivers
Classification: Unclassified
Component: Video(DRI - non Intel) (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: drivers_video-dri
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2015-08-26 22:15 UTC by Christoph Haag
Modified: 2016-06-12 14:22 UTC (History)
0 users

See Also:
Kernel Version: 4.2-rc8
Subsystem:
Regression: No
Bisected commit-id:

Attachments
dmesg (1017.29 KB, application/octet-stream)
2015-08-26 22:15 UTC, Christoph Haag 		Details
same with 4.1 lqx (1.03 MB, application/octet-stream)
2015-09-15 15:23 UTC, Christoph Haag 		Details
dmesg from 4.4-rc5 (208.33 KB, application/octet-stream)
2015-12-19 22:17 UTC, Christoph Haag 		Details
dmesg from 4.4-rc7 (163.17 KB, text/plain)
2016-01-01 01:22 UTC, Christoph Haag 		Details
dmesg with gpufault R600_DEBUG=nodma (109.55 KB, text/plain)
2016-06-06 22:52 UTC, Christoph Haag 		Details
Add an attachment (proposed patch, testcase, etc.)

Description Christoph Haag 2015-08-26 22:15:00 UTC 
Created attachment 185951 [details]
dmesg

Playing with PRIME on
00:02.0 VGA compatible controller: Intel Corporation 3rd Gen Core processor Graphics Controller (rev 09)
01:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Wimbledon XT [Radeon HD 7970M] (rev ff)

mesa git, llvm svn with si scheduler patch from axel davy, but I'm pretty sure I have seen the exact same thing with pure mainline too.

It takes a while of gameplay to provoke, 20-60 minutes or so.
Comment 1 Christoph Haag 2015-09-15 15:23:16 UTC 
Created attachment 187601 [details]
same with 4.1 lqx

Indeed. This one is from linux 4.1 with mainline mesa git and mainline llvm 247518.
Comment 2 Christoph Haag 2015-12-19 22:17:39 UTC 
Created attachment 197781 [details]
dmesg from 4.4-rc5

The call trace looks a bit different with linux 4.4-rc5.

There are also copious amounts of intel and btrfs issues, relevant radeon issue starts at line 2876.
Comment 3 Christoph Haag 2015-12-19 22:20:01 UTC 
On second thought, this one might have been caused by the intel hang.
Comment 4 Christoph Haag 2016-01-01 01:22:13 UTC 
Created attachment 198561 [details]
dmesg from 4.4-rc7

But this one from 4.4 rc7 is without intel hang. Yet another different backtrace.

Bonus cpu soft lockups and rcu_sched self-detected stall on CPU.

Display completely froze and normal reboot over ssh didn't work.
Comment 5 Christoph Haag 2016-01-05 21:16:07 UTC 
It's annoying because after this happens the GPU doesn't power off with runpm anymore and requires a reboot.
Comment 6 Michel Dänzer 2016-01-07 07:32:00 UTC 
This is the same as bug 96721, isn't it?

https://bugs.freedesktop.org/show_bug.cgi?id=92258 might be related.
Comment 7 Christoph Haag 2016-06-06 22:52:35 UTC 
Created attachment 219231 [details]
dmesg with gpufault R600_DEBUG=nodma

I can not confirm it 100%, but I think R600_DEBUG=nodma helps.

So far the unable to handle kernel paging request did not happen with R600_DEBUG=nodma. I only got a GPU fault (playing csgo with nine), maybe unrelated, dmesg attached.
Comment 8 Christoph Haag 2016-06-12 14:22:07 UTC 
(In reply to Christoph Haag from comment #7)
> Created attachment 219231 [details]
> dmesg with gpufault R600_DEBUG=nodma
> 
> I can not confirm it 100%, but I think R600_DEBUG=nodma helps.

False alarm, I just got

[304813.478253] BUG: unable to handle kernel paging request at ffff8001c4069120
[304813.480452] IP: [<ffffffffa05be0df>] ttm_bo_del_from_lru+0x7f/0xa0 [ttm]

with R600_DEBUG=nodma.
Note You need to log in before you can comment on or make changes to this bug.