Bug 33369 - [RADEON:KMS:M9:RV250] instability with EXA acceleration enabled
Summary: [RADEON:KMS:M9:RV250] instability with EXA acceleration enabled
Status: RESOLVED INVALID
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/Radeon (show other bugs)
Version: 7.5 (2009.10)
Hardware: x86 (IA32) Linux (All)
: medium major
Assignee: xf86-video-ati maintainers
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2011-01-22 11:52 UTC by Lorenzo Marcantonio
Modified: 2018-06-12 19:07 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
GPU Lockup log (2.63 KB, text/plain)
2011-01-22 11:52 UTC, Lorenzo Marcantonio
no flags Details
Working Xorg.conf (only without kms) (1.88 KB, text/plain)
2011-01-24 03:21 UTC, Lorenzo Marcantonio
no flags Details
Log for working configuration (XAA in ums) (44.46 KB, text/plain)
2011-01-24 03:22 UTC, Lorenzo Marcantonio
no flags Details
Configuration using EXA (unstable with both kms and ums) (1.88 KB, text/plain)
2011-01-24 03:24 UTC, Lorenzo Marcantonio
no flags Details
Log for EXA using ums (unstable!) (44.00 KB, text/plain)
2011-01-24 03:25 UTC, Lorenzo Marcantonio
no flags Details
Log for EXA using kms (unstable!) (34.50 KB, text/plain)
2011-01-24 03:25 UTC, Lorenzo Marcantonio
no flags Details

Description Lorenzo Marcantonio 2011-01-22 11:52:07 UTC
Created attachment 42316 [details]
GPU Lockup log

System: various kernel from 2.6.33 to 2.6.38-pre1
Driver: xf86-video-ati-6.13.2
X-Server: xorg-server-1.9.3
GPU: Radeon 9000 mobility on a GoBook 3

With EXA acceleration enabled (radeon kernel module loaded, irrespectively of the modeset options) there are multiple issues, like:
- xvideo is slow, often skips a frame and shows the previous one
- GPU lockups on heavy activity (like flash based sited)
- -vo gl_nosw in mplayer works fine but often locks up on exit

I already have agp_mode=-1 in the kernel modeline.

Using "AccelMethod" "XAA" with modeset=0 works fine with everything except xvideo with 16bpp, it only shows a blue screen (24bpp is OK)

Attached is a typical lockup log (it repeats and I need to reboot with SysRq). Of course I can give any other log or information needed to solve the issue
Comment 1 Michel Dänzer 2011-01-24 02:45:49 UTC
(In reply to comment #1)
> Driver: xf86-video-ati-6.13.2

Does a newer Git snapshot work better by any chance?


> I already have agp_mode=-1 in the kernel modeline.

That only has an effect with KMS (and is known to seriously degrade performance then). Are you using the corresponding option in xorg.conf for UMS? Maybe attach the full xorg.conf file and Xorg.0.log files for KMS and UMS to clarify this and other potential questions.
Comment 2 Lorenzo Marcantonio 2011-01-24 03:21:30 UTC
Created attachment 42359 [details]
Working Xorg.conf (only without kms)

This works putting X in XAA mode, and it works fine, no signs of instability.
The AGPMode=1 is needed because this machine is known to need it! In fact in 4x AGP mode there is severe display corruption
Comment 3 Lorenzo Marcantonio 2011-01-24 03:22:49 UTC
Created attachment 42360 [details]
Log for working configuration (XAA in ums)

Log for working configuration (see conf above)
Comment 4 Lorenzo Marcantonio 2011-01-24 03:24:19 UTC
Created attachment 42361 [details]
Configuration using EXA (unstable with both kms and ums)

In fact the only difference is the acceleration mode option missing
Comment 5 Lorenzo Marcantonio 2011-01-24 03:25:09 UTC
Created attachment 42362 [details]
Log for EXA using ums (unstable!)
Comment 6 Lorenzo Marcantonio 2011-01-24 03:25:43 UTC
Created attachment 42363 [details]
Log for EXA using kms (unstable!)
Comment 7 Lorenzo Marcantonio 2011-01-24 03:30:10 UTC
I posted the relevant configurations and logs as requested. There is apparently no difference in stability between ums and kms. The gobook iii is known to require limiting the AGP rate to 1x (the module option was set trying to stabilize the thing), btw, and the input devices need to be manually configurated (because there is a semtech touch in there and it confuses things...)

I tried with the driver git head just before submitting the bug, nothing seemed to change.
Comment 8 Adam Jackson 2018-06-12 19:07:53 UTC
Mass closure: This bug has been untouched for more than six years, and is not
obviously still valid. Please reopen this bug or file a new report if you continue to experience issues with current releases.


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.