Arch Linux

Please read this before reporting a bug:
https://wiki.archlinux.org/title/Bug_reporting_guidelines

Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.

REPEAT: Do NOT report bugs for outdated packages!
Tasklist

FS#17695 - [xorg] radeon driver very slow (and unstable) acceleration for RV250 chipset

Attached to Project: Arch Linux
Opened by Andre Herbst (moormaster) - Sunday, 03 January 2010, 11:16 GMT
Last edited by Andreas Radke (AndyRTR) - Monday, 03 May 2010, 16:47 GMT
Task Type Bug Report
Category Packages: Extra
Status Closed
Assigned To Jan de Groot (JGC)
Andreas Radke (AndyRTR)
Architecture i686
Severity High
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 2
Private No

Details

Description:
* glxgears uses 100% CPU in fullscreen even though glxinfo is showing that the MESA DRI R200 renderer is used.

* running nexuiz leads to unplayable slow framerate ( < 0.2 fps )

* hedgewars does not even start... it just opens up a window and closes it again (and this is not a 3d game...)

* googleearth has many graphic errors and sometimes causes Xorg to lock up... or even more weird: the mounted filesystems suddenly get unusable... reading files causes I/O Errors and writing fails due to not enough space left (even though there is...)


I have also tried to use EXA AccelMethod... this lead glxgears to work fine in fullscreen with ~5% CPU usage
hedgewars is able to start games... but with similiar graphic errors like googleearth (but still very slow)
googleearth is as unstable as it was using XAA before :D

Though it has graphic errors googleearth is the only program which seems to run with a usable framerate (besides glxgears).

Additional info:
* package version(s)
xf86-video-ati 6.12.4-3
mesa 7.7-1
ati-dri 7.7-1

* config and/or log files etc.


Steps to reproduce:
This task depends upon

Closed by  Andreas Radke (AndyRTR)
Monday, 03 May 2010, 16:47 GMT
Reason for closing:  Works for me
Additional comments about closing:  no response for months and many updates later it seems solved.
Comment by Andreas Radke (AndyRTR) - Sunday, 03 January 2010, 11:53 GMT
Please give full Xorg.0.log, dmesg and glxinfo output.
Comment by Andre Herbst (moormaster) - Sunday, 03 January 2010, 12:56 GMT
I took my time to do some experiments with the bug...

Since Google Earth causes the biggest problems I have tried running it to get the log files when the errors are happening. If I disable DRI google earth works fine ( but slow of course) without any errors shown in dmesg... Having DRI enabled seems to almost always crush critical memory ranges in RAM ( corrupted r/w cache of mounted file systems etc... ). If I should get googleearth to work before the filesystems stop working I get other memory corruption errors in dmesg if I continue to use the application.

I hope the log files are helping :D
Comment by Andre Herbst (moormaster) - Sunday, 03 January 2010, 12:57 GMT
PS: I am using Google Earth for GNU/Linux 5.1.3533.1731 :D
Comment by fabio (fabioamd) - Wednesday, 13 January 2010, 11:48 GMT
have you tried: glxinfo | grep rendering ?
Comment by Andre Herbst (moormaster) - Wednesday, 13 January 2010, 18:23 GMT
yes direct rendering is active and I need to disable it to have a stable system... having dri activated causes ugly ram corruptions additionally to the bad performance as already stated above :D
Comment by Andreas Radke (AndyRTR) - Friday, 15 January 2010, 05:47 GMT
please file your bug upstream. it's probably no packaging issue.
Comment by Andre Herbst (moormaster) - Friday, 15 January 2010, 15:31 GMT
I think I have found a similiar bug at freedesktop.org:
https://bugs.freedesktop.org/show_bug.cgi?id=24414

The comments tell that the problems are caused by mesa 7.6.0... a downgrade seems to fix the problems... I need to try that to see if that also works for me...
Comment by Andre Herbst (moormaster) - Saturday, 16 January 2010, 14:49 GMT
yep i can confirm this... compiling and installing mesa 7.5.2 solves my stability problems... the performance of hedgewars and nexuiz is still not very good but at least without errors and crashes
Comment by daimonion (demon) - Wednesday, 03 February 2010, 11:01 GMT
Guys, the problem is in libdrm package! It is compiled without "--enable-radeon-experimental-api"!!! Recompile this package with this falg and everything should be OK.
Comment by Andre Herbst (moormaster) - Wednesday, 03 February 2010, 22:22 GMT
this didn't change a single thing for me... the system still crashes if i use a mesa version >= 7.6.0 while trying to run hedgewars or google-earth...
Comment by Andreas Radke (AndyRTR) - Wednesday, 03 March 2010, 20:25 GMT
Please test the testing packages with either KMS or UMS.
Comment by Andreas Radke (AndyRTR) - Thursday, 18 March 2010, 18:12 GMT
there's also now mesa 7.7.1rc1 in testing. please give feedback.
Comment by Andreas Radke (AndyRTR) - Wednesday, 31 March 2010, 07:05 GMT
state? you should check that DRI2 driven packages in testing work for you. DRI1 will soon be gone.
Comment by Andreas Radke (AndyRTR) - Saturday, 10 April 2010, 10:21 GMT
Can we close it? We now ship latest stable upstream stuff?

Loading...