Bug 21896 - DRM and EXA fails to initialize
Summary: DRM and EXA fails to initialize
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/radeonhd (show other bugs)
Version: 7.4 (2008.09)
Hardware: Other All
: medium normal
Assignee: Luc Verhaegen
QA Contact: Xorg Project Team
Depends on:
Reported: 2009-05-23 11:19 UTC by Elmar Stellnberger
Modified: 2011-11-07 15:38 UTC (History)
0 users

See Also:
i915 platform:
i915 features:

xorg.conf (10.62 KB, text/plain)
2009-05-23 11:21 UTC, Elmar Stellnberger
no flags Details
Xorg.0.log.old (50.62 KB, patch)
2009-05-23 12:10 UTC, Elmar Stellnberger
no flags Details | Splinter Review
updated xorg.conf (10.64 KB, application/x-kdeuser2)
2009-05-24 07:36 UTC, Elmar Stellnberger
no flags Details
xorg.log of corresponding session (55.60 KB, text/x-log)
2009-05-24 07:38 UTC, Elmar Stellnberger
no flags Details
Xorg.0.log kernel 2.6.30, DRI option set (52.07 KB, text/x-log)
2009-05-25 02:23 UTC, Elmar Stellnberger
no flags Details
xorg-noaccelaration.tar (70.00 KB, application/x-tar)
2009-10-09 03:01 UTC, Elmar Stellnberger
no flags Details
/var/log/Xorg.0.log + without and with second radeon-X (56.88 KB, text/x-log)
2009-12-21 12:56 UTC, Elmar Stellnberger
no flags Details
dmesg (244.73 KB, text/plain)
2009-12-21 13:23 UTC, Elmar Stellnberger
no flags Details

Description Elmar Stellnberger 2009-05-23 11:19:33 UTC
Moving and Scrolling windows was extremely slow with radeonhd-1.2.5_20090506-4be5f71-5.5/-3.1 as if there was no hw accellaration at all. A downgrade has resolved this.

used xorg-x11-server-7.4-17.4.1, unichrome-20081229-5.1; distro method: opensuse build service
Comment 1 Elmar Stellnberger 2009-05-23 11:21:12 UTC
Created attachment 26144 [details]
Comment 2 Yang Zhao 2009-05-23 11:41:46 UTC
Please post the Xorg.0.log from a problematic X session.
Comment 3 Elmar Stellnberger 2009-05-23 12:01:24 UTC
Besides the current log there are Xorg.1/2/99.log & Xorg.0/1/2/99.log.old files. How to determine which logs apply to the specified radeonhd version?
Comment 4 Yang Zhao 2009-05-23 12:05:50 UTC
(In reply to comment #3)
> Besides the current log there are Xorg.1/2/99.log & Xorg.0/1/2/99.log.old
> files. How to determine which logs apply to the specified radeonhd version?

Look for the string RADEONHD in the logs.
Comment 5 Elmar Stellnberger 2009-05-23 12:10:24 UTC
Created attachment 26146 [details] [review]

I hope this is the right log file (It should be because I think that I am booting for the first time with the newly downgraded driver.). Nonetheless some of the other Xorg logs may provide supplementary informations on this issue.
Comment 6 Yang Zhao 2009-05-23 12:46:02 UTC
Please add

  Option "DRI" "true"

to the appropriate Device section of your xorg.conf.

EXA on r6xx requires DRI to be enabled. I think the driver is silently falling back to no acceleration.
Comment 7 Elmar Stellnberger 2009-05-24 07:35:46 UTC
  Unfortunately adding the DRI option hasn`t changed anything. Moving in general and scrolling at OpenOffice (works for Konqueror&Konsole) are as slow as without HW accelaration.
Comment 8 Elmar Stellnberger 2009-05-24 07:36:18 UTC
Created attachment 26168 [details]
updated xorg.conf
Comment 9 Elmar Stellnberger 2009-05-24 07:38:03 UTC
Created attachment 26169 [details]
xorg.log of corresponding session
Comment 10 Yang Zhao 2009-05-24 10:29:48 UTC
You do not have the necessary DRM modules installed. DRM for r6xx is available in linux-2.6.30.

The driver can't fall back to software 2D acceleration when DRI is specified but not available.
Comment 11 Elmar Stellnberger 2009-05-25 01:53:09 UTC
oops, this time I have used an elder kernel because of a bug in the opensuse specific 2.6.30 kernel. Is there anything else I should take care of / additionally test for since always reinstalling/purging the radeonhd driver is a bit long-winded.
Comment 12 Elmar Stellnberger 2009-05-25 02:21:46 UTC
  With kernel 2.6.30 and the DRI option set it really works better. However moving a window in front of the background bitmap displayed by kdesktop is still significantly slower than before.
Comment 13 Elmar Stellnberger 2009-05-25 02:23:51 UTC
Created attachment 26192 [details]
Xorg.0.log kernel 2.6.30, DRI option set
Comment 14 Elmar Stellnberger 2009-06-18 04:21:54 UTC
  Have now switched to AccelMethod ShadowFB. Interestingly moving and scrolling windows work really fine in ShadowFB-mode with current X-servers as soon and as long as I start a second X-server with DRI in exa mode (former X-servers worked always well with ShadowFB). I can avoid bug 22114 by using ShadowFB.
  Unfortunately switching between multiple radeon-X-servers can still lead to an unretractable hang while Google Earth initializes or may desire an explicit chvt if Ctrl-Alt-Fx does not work; sometimes it requires a Ctrl-Alt-Fx followed by a Ctrl-Alt. Running games on a separate X-Server makes sense because they require different resolutions and often require clone instead of xinerama mode (see also bug 20628, bug 20888).
Comment 15 Elmar Stellnberger 2009-09-06 03:28:38 UTC
resolved with xorg-x11-driver-video-radeonhd-1.2.5_20090506_4be5f71-7.16.
Comment 16 Elmar Stellnberger 2009-10-09 03:00:55 UTC
  Unfortunately with xorg-x11-driver-video-radeonhd-1.2.5_20091001_be7216f-2.2 I have again no hardware accelaration (xorg even reports this). Might perhaps be a packaging issue. Please have a look at my attachement.
Comment 17 Elmar Stellnberger 2009-10-09 03:01:28 UTC
Created attachment 30212 [details]
Comment 18 Yang Zhao 2009-10-09 08:53:31 UTC
X is reporting that everything is fine.

Just in case you're talking about:

(EE) AIGLX error: dlopen of /usr/lib64/dri/r600_dri.so failed (/usr/lib64/dri/r600_dri.so: cannot open shared object file: No such file or directory)
(EE) AIGLX: reverting to software rendering

This is expected as support for that doesn't exist yet (hence the lack of said .so file), and it's the domain of Mesa in any case.
Comment 19 Elmar Stellnberger 2009-10-13 09:09:23 UTC
now tested in three settings:

1. radeon instead of radeonhd driver (xorg-x11-driver-video-7.4-83.3):
   scrolling is perfectly smooth
2. xorg-x11-driver-video-radeonhd-1.2.5_20090506_4be5f71-2.13.x86_64:
   quite good results though window stays flurry during move
3. xorg-x11-driver-video-radeonhd-1.2.5_20091001_be7216f-13.1
   almost no hardware accelaration: jerking in wide distances
Comment 20 Elmar Stellnberger 2009-10-15 06:59:54 UTC
with xorg-x11-driver-video-radeonhd-1.3.0_20091009_8cbff7b-2.1 again as good as with 2.; radeon is still a bit better.
Comment 21 Elmar Stellnberger 2009-11-23 09:23:22 UTC
  Please have a look at the radeon driver. It implements hardware accellerated scrolling and moving very well. Interestingly it is sufficient to run a second X-server with radeon to leverage a good radeonhd-scroll-speed even at a high cpu load. This does not apply to moving windows with radeonhd.
Comment 22 Elmar Stellnberger 2009-11-23 09:28:29 UTC
Comment 23 Elmar Stellnberger 2009-12-21 11:51:19 UTC
  Please do also think about those who do not yet have hardware accellaration enabled for their graphics card as the owners of the Radeon HD 2000 - HD 4000 series. Window scroll and move are basic features that should work everywhere.
  This should not be hard to implement since the only thing I need to do in order to make scroll and move work for screen :0 is to start another X-server with the radeon driver on display :1 (radeon does not have this problem.).
Comment 24 Yang Zhao 2009-12-21 12:06:09 UTC
I have no idea what you are going on about anymore.

The original cause of the slowdown was already explained to you and was apparently resolved.  The relevant acceleration codes have not been touched in a long time, and you claim the latest package by your distro works.

Acceleration fallback when lacking appropriate DRM is already in place. If you think this is broken, investigate and file another bug.

Resolving as NOTABUG.
Comment 25 Elmar Stellnberger 2009-12-21 12:19:28 UTC
  No, it definitely does NOT work. I have definitely erred in reporting that things are better. They never were. The only thing that I did not notice first was that running a second radeon-X-server on :1 can also improve things for :0.
Comment 26 Yang Zhao 2009-12-21 12:25:15 UTC
Post your Xorg.0.log and output from dmesg.
Comment 27 Elmar Stellnberger 2009-12-21 12:36:40 UTC
Do you need an Xorg.0.log where scroll & move work fast because of a second X-server or one where no such server is running and hence scroll & move are very slow.
Comment 28 Elmar Stellnberger 2009-12-21 12:36:53 UTC
Comment 29 Alex Deucher 2009-12-21 12:39:07 UTC
The DRI is required for accel on r6xx+ and it only works with one Xserver, so starting a second xserver will not get acceleration.
Comment 30 Elmar Stellnberger 2009-12-21 12:43:58 UTC
 Strange, but true. At me the accellaration only works with a second X server.
So what kind of log will you need?
Comment 31 Elmar Stellnberger 2009-12-21 12:56:46 UTC
Created attachment 32227 [details]
/var/log/Xorg.0.log + without and with second radeon-X

I hope this is what you expect. At the beginning scroll&move were slow. After starting a second radeon-X they became fast.
Comment 32 Yang Zhao 2009-12-21 13:12:34 UTC
(In reply to comment #31)
> Created an attachment (id=32227) [details]
> /var/log/Xorg.0.log + without and with second radeon-X

(WW) RADEONHD(0): Falling back to ShadowFB acceleration

DRIPreInit is failing.  DRM initialization messages should precede that line, but they don't, so there's something wrong with your setup.

Post the output of dmesg.
Comment 33 Elmar Stellnberger 2009-12-21 13:23:05 UTC
Created attachment 32228 [details]
Comment 34 Yang Zhao 2009-12-21 13:29:53 UTC
OK, the verbosity of your dmesg has pushed pushed what's useful in this case out of the buffer.

We need to see the sections where the radeon kernel module is being initialized, and also where X is started.
Comment 35 Jeremy Huddleston Sequoia 2011-10-16 15:57:47 UTC
Does this issue occur with the preferred ati driver (xf86-vide-ati)?  If so, please move this to the Driver/Radeon component.  

Development of radeonhd has pretty much halted and development focus is on the ati driver.  Please see http://www.x.org/wiki/radeonhd

If the issue does not exist in the ati driver (or if there is no response to this message), this bug will be closed as WONTFIX unless someone contributes a patch.
Comment 36 Jeremy Huddleston Sequoia 2011-11-07 15:38:26 UTC
Closing due to lack of response.  Please reopen and move to the Driver/Radeon 
component if this issue persists with xf86-video-ati

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.