Bug 36892 - page faults / general protection faults under heavy 3d load
Summary: page faults / general protection faults under heavy 3d load
Status: RESOLVED WILL_NOT_FIX
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: 2011-06-07 14:13 UTC by Niels Ole Salscheider
Modified: 2013-12-23 13:45 UTC (History)
2 users (show)

See Also:
Kernel Version: 3.6-rc1
Subsystem:
Regression: No
Bisected commit-id:

Attachments
kernel message (15.48 KB, application/octet-stream)
2011-06-07 14:13 UTC, Niels Ole Salscheider 		Details
kernel message 2 (4.37 KB, application/octet-stream)
2011-06-07 14:14 UTC, Niels Ole Salscheider 		Details
kernel message 3 (2.95 KB, application/octet-stream)
2011-06-07 14:14 UTC, Niels Ole Salscheider 		Details
kernel message 4 (3.18 KB, application/octet-stream)
2011-06-07 14:15 UTC, Niels Ole Salscheider 		Details
dmesg output for 3.6-rc1 (12.08 KB, text/plain)
2012-08-24 14:30 UTC, Niels Ole Salscheider 		Details
Show Obsolete (4) Add an attachment (proposed patch, testcase, etc.)

Description Niels Ole Salscheider 2011-06-07 14:13:13 UTC 
I get page faults or general protection faults under heavy 3d load (like "sauerbraten", but sometimes switching windows with kwin is enough) in random processes. I have a radeon hd 6870 and use r600g from mesa git.

This happens on 3.0-rc1 and 2.6.39. I am not sure if this is a problem with older kernels, too, since I get gpu lockups with them.

I am not sure what could cause these problems, maybe some bug in gem / ttm?
Comment 1 Niels Ole Salscheider 2011-06-07 14:13:59 UTC 
Created attachment 61052 [details]
kernel message
Comment 2 Niels Ole Salscheider 2011-06-07 14:14:24 UTC 
Created attachment 61062 [details]
kernel message 2
Comment 3 Niels Ole Salscheider 2011-06-07 14:14:41 UTC 
Created attachment 61072 [details]
kernel message 3
Comment 4 Niels Ole Salscheider 2011-06-07 14:15:02 UTC 
Created attachment 61082 [details]
kernel message 4
Comment 5 Niels Ole Salscheider 2011-06-07 14:18:17 UTC 
btw: I do not know why it was out of memory in the first error message. I have 8GB of ram but usually less than 4GB are used (most of it for virtuoso-t and nepomukservices)
Comment 6 Alan 2012-08-24 13:41:21 UTC 
Is this still seen on modern (3.2+) kernels ?
Comment 7 Niels Ole Salscheider 2012-08-24 14:29:42 UTC 
I have just tried to run "sauerbraten" under 3.6-rc1. It seems to work fine, but I get page allocation failure messages in dmesg, see attachment.
Comment 8 Niels Ole Salscheider 2012-08-24 14:30:17 UTC 
Created attachment 78381 [details]
dmesg output for 3.6-rc1
Comment 9 Alan 2012-08-24 16:36:07 UTC 
The page allocation warnings are fine (and just warnings that perhaps the VM needs further tuning)

The only oddity is this

[drm:radeon_cs_ioctl] *ERROR* Failed to parse relocation -12!
Comment 10 Alan 2013-12-23 13:45:13 UTC 
Which is an out of memory error so correctly just a warning
Note You need to log in before you can comment on or make changes to this bug.