[needs 7.9.1] [radeon r200] ../../radeon/radeon_cs_gem.c:181: cs_gem_write_reloc: Assertion `boi->space_accounted' failed.

[Daniel Richard G.]

This may be related to bug #602267. The failure was obtained by invoking glxgears(1) in a fully-updated Maverick RC1 live-CD environment on amd64. My graphics card is an ATI FireGL 8800, which uses the r200 driver.

glxgears starts and runs without a problem, but as soon as I maximize the window, the program aborts with an assertion failure in the driver:

ubuntu@ubuntu:~$ glxgears
Running synchronized to the vertical refresh. The framerate should be
approximately the same as the monitor refresh rate.
303 frames in 5.0 seconds = 60.488 FPS
glxgears: ../../radeon/radeon_cs_gem.c:181: cs_gem_write_reloc: Assertion `boi->space_accounted' failed.
Aborted (core dumped)

If I invoke gltron(1), I get the same assertion failure right at program startup (after an all-black window that appears briefly).

ProblemType: Crash
DistroRelease: Ubuntu 10.10
Package: mesa-utils 8.0.1-0ubuntu1
ProcVersionSignature: Ubuntu 2.6.35-22.33-generic 2.6.35.4
Uname: Linux 2.6.35-22-generic x86_64
Architecture: amd64
AssertionMessage: glxgears: ../../radeon/radeon_cs_gem.c:181: cs_gem_write_reloc: Assertion `boi->space_accounted' failed.
Date: Thu Oct 7 02:56:17 2010
ExecutablePath: /usr/bin/glxgears
LiveMediaBuild: Ubuntu 10.10 "Maverick Meerkat" - Release Candidate amd64 (20100928)
ProcCmdline: glxgears
ProcEnviron:
 SHELL=/bin/bash
 LANG=en_US.UTF-8
Signal: 6
SourcePackage: mesa-demos
StacktraceTop:
 raise () from /lib/libc.so.6
 abort () from /lib/libc.so.6
 __assert_fail () from /lib/libc.so.6
 cs_gem_write_reloc (cs=0x2248220, bo=0x22663a0, read_domain=<value optimized out>, write_domain=<value optimized out>, flags=3685209984) at ../../radeon/radeon_cs_gem.c:181
 ctx_emit_cs (ctx=<value optimized out>, atom=0x1f36bd8) at r200_state_init.c:575
Title: glxgears assert failure: glxgears: ../../radeon/radeon_cs_gem.c:181: cs_gem_write_reloc: Assertion `boi->space_accounted' failed.
UserGroups: adm admin cdrom dialout lpadmin plugdev sambashare
XsessionErrors:
 (polkit-gnome-authentication-agent-1:15954): GLib-CRITICAL **: g_once_init_leave: assertion `initialization_value != 0' failed
 (nautilus:15959): GConf-CRITICAL **: gconf_value_free: assertion `value != NULL' failed

[Apport retracing service (apport)]

StacktraceTop:
 *__GI_raise (sig=<value optimized out>)
 *__GI_abort () at abort.c:92
 *__GI___assert_fail (
 ?? () from /lib/libdrm_radeon.so.1
 ctx_emit_cs (ctx=<value optimized out>,

[Tormod Volden (tormodvolden)]

Thanks for your report. Yes, it is probably the same issue as in the other bug report. Would it be possible for you to try the i386 live CD to rule out any amd64 specific issues?

[Uwe Geuder (ubuntulp-ugeuder)]

I get the same assertio on i386 with ATI Technologies Inc Radeon RV250 [Mobility FireGL 9000] (rev 02) when trying to start stellar

[Uwe Geuder (ubuntulp-ugeuder)]

Architecture: i386
DRM.card0.DVI.D.1:
 status: disconnected
 enabled: disabled
 dpms: On
 modes:
 edid-base64:
DRM.card0.LVDS.1:
 status: connected
 enabled: enabled
 dpms: On
 modes: 1400x1050 1280x1024 1280x960 1280x854 1280x800 1280x720 1152x768 1024x768 800x600 848x480 720x480 640x480
 edid-base64:
DRM.card0.VGA.1:
 status: disconnected
 enabled: disabled
 dpms: On
 modes:
 edid-base64:
DistroRelease: Ubuntu 10.10
InstallationMedia: Ubuntu 10.10 "Maverick Meerkat" - Release Candidate i386 (20100928)
Lsusb:
 Bus 004 Device 001: ID 1d6b:0001 Linux Foundation 1.1 root hub
 Bus 003 Device 001: ID 1d6b:0001 Linux Foundation 1.1 root hub
 Bus 002 Device 001: ID 1d6b:0001 Linux Foundation 1.1 root hub
 Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
MachineType: IBM 23734G2
Package: mesa (not installed)
PccardctlIdent:
 Socket 0:
   no product info available
 Socket 1:
   no product info available
PccardctlStatus:
 Socket 0:
   no card
 Socket 1:
   no card
ProcCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.35-22-generic root=UUID=86ba0c39-2e59-4e72-ba14-9ee43a0d1943 ro quiet splash
ProcEnviron:
 LANG=en_US.utf8
 SHELL=/bin/bash
ProcVersionSignature: Ubuntu 2.6.35-22.33-generic 2.6.35.4
Tags: maverick maverick
Uname: Linux 2.6.35-22-generic i686
UserGroups: adm admin cdrom dialout lpadmin plugdev sambashare
dmi.bios.date: 06/18/2007
dmi.bios.vendor: IBM
dmi.bios.version: 1RETDRWW (3.23 )
dmi.board.name: 23734G2
dmi.board.vendor: IBM
dmi.board.version: Not Available
dmi.chassis.asset.tag: 995gz93
dmi.chassis.type: 10
dmi.chassis.vendor: IBM
dmi.chassis.version: Not Available
dmi.modalias: dmi:bvnIBM:bvr1RETDRWW(3.23):bd06/18/2007:svnIBM:pn23734G2:pvrThinkPadT40:rvnIBM:rn23734G2:rvrNotAvailable:cvnIBM:ct10:cvrNotAvailable:
dmi.product.name: 23734G2
dmi.product.version: ThinkPad T40
dmi.sys.vendor: IBM
glxinfo: Error: [Errno 2] No such file or directory
system:
 distro: Ubuntu
 codename: maverick
 architecture: i686
 kernel: 2.6.35-22-generic

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

apport information

[Uwe Geuder (ubuntulp-ugeuder)]

Oops sorry, I start the program with stellar<TAB>, but it's of course called stellarium.

[Tormod Volden (tormodvolden)]

Ouch. This can be... interesting. It seems from bug 602267 that this is solved upstream in the gallium driver, so we can not count on upstream help to nail this down in the classic driver. Can you also try the mesa packages from the "Radeon gallium" PPA, and check whether is fixed with gallium, or with the newest classic drivers?

[Tormod Volden (tormodvolden)]

Sorry, the "Radeon gallium" PPA has only Lucid packages at the moment, I will consider making Maverick packages later. For now maybe you can try the xorg-edgers PPA which has gallium by default as well? Or try installing those Lucid packages in Maverick, it might work.

[Uwe Geuder (ubuntulp-ugeuder)]

Sorry, I have no idea what is Gallium or the classic driver. Google lead me to http://www.x.org/wiki/RadeonFeature.

I'm not sure whether I understand the page, but it appears to me that Gallium applies only to R300 or higher. (#602267 was on R300). However, here we are talking about R200.

[Tormod Volden (tormodvolden)]

Uwe, you are absolutely right. I was too much into the other bug report. We should report this upstream, but first some of you should try the latest kernel snapshot, and also the latest mesa. I will come back later with more instructions if needed.

[Daniel Richard G. (skunk)]

Tormod, I'll be happy to do some testing, but for the time being I can only do so using a live-CD environment. I can do upgrades, pull in PPA packages and restart X, but using a different kernel might be a bit tricky.

Please let me know what you have in mind; I'm not a regular user of newer kernels or bleeding-edge drivers.

[Daniel Richard G. (skunk)]

Tormod: I've reproduced the bug in the i386 Maverick RC live-CD environment (with all updates applied). See bug #656709.

[Tormod Volden (tormodvolden)]

Thanks. It looks like it has been reported upstream already, in https://bugs.freedesktop.org/show_bug.cgi?id=29367

[Uwe Geuder (ubuntulp-ugeuder)]

Before reporting upstream, we probably have to try their newest version. In order to get at all started with that code I first built
the Ubuntu source.

The Maverick debian package builds mesa obviously in 9 configurations

mesa-7.9~git20100924/build/osmesa32-static/config.status
mesa-7.9~git20100924/build/swx11+glu/config.status
mesa-7.9~git20100924/build/osmesa/config.status
mesa-7.9~git20100924/build/osmesa16-static/config.status
mesa-7.9~git20100924/build/osmesa32/config.status
mesa-7.9~git20100924/build/swx11+glu-static/config.status
mesa-7.9~git20100924/build/osmesa16/config.status
mesa-7.9~git20100924/build/dri/config.status
mesa-7.9~git20100924/build/osmesa-static/config.status

I have really no clue what they are all about, so I checked what glxgears uses from the mesa source package. There are 3 libraries:

/usr/lib/mesa/libGL.so.1.2: Package: libgl1-mesa-glx Source: mesa
/usr/lib/libGLU.so.1.3.070900: Package: libglu1-mesa Source: mesa
/usr/lib/dri/r200_dri.so: Package: libgl1-mesa-dri Source: mesa

libGL.so.1.2 is built in the dri configuration.
libGLU.so.1.3.070900 is built in the wx11+glu configuration.
r200_dri.so is built in the dri configuration.

So I just guessed that the dri configuration might the relevant one here and built the newest mesa as follows:

git clone git://anongit.freedesktop.org/mesa/mesa
cd mesa
./autogen.sh '--build=i686-linux-gnu' '--with-driver=dri'
   '--with-dri-drivers=r200' '--with-egl-displays=x11 drm'
   '--enable-glx-tls' '--with-state-trackers=egl,glx,dri,vega'
   '--enable-gles-overlay' '--enable-gles1' '--enable-gles2'
   '--enable-driglx-direct' '--disable-glu' '--disable-glut' '--disable-glw'
   '--enable-debug' 'CFLAGS=-Wall -g -O0'
   '--with-dri-searchpath=/usr/lib/dri:/usr/lib32/dri'
   'build_alias=i686-linux-gnu' 'LDFLAGS=-Wl,-Bsymbolic-functions'
   'CPPFLAGS=' 'CXXFLAGS=-g -O0'
make
cd lib
LIBGL_DRIVERS_PATH=. glxgears
Mesa: Mesa 7.10-devel DEBUG build Oct 8 2010 11:17:51
Mesa warning: couldn't open libtxc_dxtn.so, software DXTn compression/decompression unavailable
Running synchronized to the vertical refresh. The framerate should be
approximately the same as the monitor refresh rate.
253 frames in 5.0 seconds = 50.547 FPS
251 frames in 5.0 seconds = 50.118 FPS
glxgears: ../../radeon/radeon_cs_gem.c:181: cs_gem_write_reloc: Assertion `boi->space_accounted' failed.
Aborted

(The assertion is sometimes hit immediately, sometimes first when resizing the window)

My configuration options were some wild guessing merge of Maverick's mesa-7.9~git20100924/build/dri/config.status and the instructions given in https://bugs.freedesktop.org/show_bug.cgi?id=29066, together with the assumption that gallium is not applicable to r200.

Not sure where the debugging info should be written. There was nothing in syslog.

This is the call stack:

#0 0x0012e416 in __kernel_vsyscall ()
#1 0x003bb941 in raise () from /lib/libc.so.6
#2 0x003bee42 in abort () from /lib/libc.so.6
#3 0x003b48e8 in __assert_fail () from /lib/libc.so.6
#4 0x00afc158 in ?? () from /lib/libdrm_radeon.so.1
#5 0x00afcd80 in radeon_cs_write_reloc () from /lib/libdrm_radeon.so.1
#6 0x0071de59 in ctx_emi...

[Uwe Geuder (ubuntulp-ugeuder)]

Oops, my investigations obviously took some time. So I posted #39 without reading #38, which had come in the meantime.

[Tormod Volden (tormodvolden)]

Uwe, that's good sleuthing there, it is nice that you are proficient at debugging! You can get latest libdrm from the xorg-edgers PPA unless you want to build it yourself. Maybe you can, together with Andrew, look at applying Pauli's patch referenced in the upstream bug, or ask if Pauli can make a fresh one.

[Uwe Geuder (ubuntulp-ugeuder)]

I built the latest libdrm from

git clone git://anongit.freedesktop.org/git/mesa/drm

with these options:

./autogen.sh \
 '--build=i686-linux-gnu' \
 '--enable-udev' \
 '--enable-libkms' \
 '--enable-vmwgfx-experimental-api' \
 '--enable-nouveau-experimental-api' \
 '--enable-radeon' \
 '--enable-intel' \
 'CFLAGS=-Wall -g -O0' \
 'build_alias=i686-linux-gnu' \
 'LDFLAGS=-Wl,-Bsymbolic-functions' \
 'CPPFLAGS='

Not to surprisingly the same assertion is still hit. Please find the full call stack in the attachment.

I have not yet had time to look at the patch.

[Uwe Geuder (ubuntulp-ugeuder)]

As already mentioned in https://bugs.freedesktop.org/show_bug.cgi?id=29367#c5 the patch in https://bugs.freedesktop.org/show_bug.cgi?id=21374#c5 does not apply to current mesa. I searched a bit in the git history and it applies cleanly e. g. to commit 57f55a. However, this commit is 150 commits apart from today (and that is only counting files named r200 and radeon_common, there might be others less obvious ones) So I assume that that old patch is the wrong approach for us today.

So let's start from a different angle.

The assertion does not occur on Maverick. Maverick seems to based on Mesa 7.7.1 (according to the source naming, I haven't checked the contents). And 7.7.1 is much newer than the old patch. So the issue was probably corrected one way or the other.

If we look at the version history of the r200 source in git, we see that 7.7.1 was made in a side branch. The attached screen dump was produced with the command

$ gitk --all src/mesa/drivers/dri/r200/* src/mesa/drivers/dri/radeon/radeon_common.?

The side branch ends in gitk row 70 (the currently selected one), which precedes mesa_7_7_1_rc2, which in turn precedes mesa_7_7_1

So I could guess that one of the changes in the sidebranch actually fixes the assertion but it has not been applied (or applied incorrectly) to the "trunk" which lead towards Maverick. At least nothing has been visibly merged as we can easily see from gitk.

Now we are "only" some 70-80 commits away from today. Still that will require reading some source, and from what I've seen during my investigations so far I have some doubts whether anybody not familiar with the code can do it in reasonable time. I'm not sure when I have time to investigate the source further. I might be very busy the coming weeks.

Maybe you Tormod could report my findings so far upstream. (I don't have an account in freedesktop and now it's not the right time for me to get involved there.)

[Tormod Volden (tormodvolden)]

It would be very nice if someone with the hardware could follow up on the upstream bugs though! Regarding that commit from Pauli, it was also committed to trunk at 4ea694a26b99835d0b5bc814cf024850874a9a83. Sometimes the version branches are merged onto trunk, but sometimes trunk commits are cherrypicked to a version branch, so they will have another commit ID. There has been some discussion (and varying practices) on how to do this. Ideally a cherrypicked commit will contain a comment that it has been cherrypicked (from using git cherry-pick -x).

[Uwe Geuder (ubuntulp-ugeuder)]

The problem also affects this card

ATI Technologies Inc Radeon Mobility M7 LW [Radeon Mobility 7500]

which uses /usr/lib/dri/radeon_dri.so

[Alejandro R. Mosteo (mosteo)]

Just updated to Maverick and get these same exact symptoms; crash when going to fullscreen and the same assertion failures.

Card is (from lspci):

02:00.0 VGA compatible controller: ATI Technologies Inc RV280 [Radeon 9200] (rev 01)

[Tormod Volden (tormodvolden)]

The latest mesa git has some fixes for R200, although I am not sure it is the same issue. Can you please try the mesa packages for maverick from the Radeon Gallium PPA, https://launchpad.net/~xorg-edgers/+archive/radeon ? Since all drivers (gallium and classic) are included this is the easiest way to test latest git.

[Oibaf (oibaf)]

It should be fixed in 7.9.1 with:
http://cgit.freedesktop.org/mesa/mesa/commit/?h=7.9&id=6936fbc03999354075da4d5b91b1ee7ce190fa80

[Uwe Geuder (ubuntulp-ugeuder)]

Excellent!!! After

git checkout 6936fb

and building and starting applications as in comment #42 the problem is gone.

Tested with running stellarium and resizing glxgears.

[Uwe Geuder (ubuntulp-ugeuder)]

To clarify: I have only built and tested r200_dri.so on

01:00.0 VGA compatible controller: ATI Technologies Inc Radeon RV250 [Mobility FireGL 9000] (rev 02)

The comment says that the fix is for r200.

No idea whether it helps also for the different card & driver mentioned in #45

[Uwe Geuder (ubuntulp-ugeuder)]

Now I've built both radeon_dri.so and r200_dri.so for the benfit of those who are not so familiar with git and building. (or those who just want to save the time to do it)

1. Extract the attached file.
tar xf mesafix-6936fb.tbz

2, go to the newly created directory

cd mesafix-6936fb

3. run OpenGL applications like this

LIBGL_DRIVERS_PATH=. glxgears

LIBGL_DRIVERS_PATH=. stellarium

If it does not work make sure that the applications really use the fixed libraries and not the previously installed ones

1. Find the process id of the application

ps -e

2. display the libraries used

grep xp /proc/1234/maps

(replace 1234 by the process id you got in step 1)

3. Make sure the dri library comes from your newly created directory instead of /usr/lib/dri

Elias: could you please test on your laptop to see whether the r200 fix also helps for the radeon driver used for your graphics adapter.

[Oibaf (oibaf)]

r100 and r300 are still not fixed, they should need a similar patch. See also the thread at:
http://lists.freedesktop.org/archives/mesa-dev/2010-October/003491.html

[Barry Grumbine (barry-grumbine)]

I was getting the same error when trying to run Blender 2.5 Beta.

I added Tormod Volden's ppa and updated to his newer drivers.

glxgears seems to work now, I can re-size the window without it crashing, and Blender starts up but has other issues. The text in Blender is all jumbled up. It looks almost like Chinese or Japanese characters until you look real close and can barely make out the English words. I attached a screen shot.

I see another version of the drivers are scheduled to be built soon on Tom's ppa. Will test them tomorrow.

Some information:
Xubuntu 10.10 i386

$ uname -a
Linux vaio 2.6.35-22-generic #35-Ubuntu SMP Sat Oct 16 20:36:48 UTC 2010 i686 GNU/Linux

$ grep DRI /var/log/Xorg.0.log
[ 30.584] (II) Loading extension XFree86-DRI
[ 30.585] (II) Loading extension DRI2
[ 31.011] (II) RADEON(0): [DRI2] Setup complete
[ 31.011] (II) RADEON(0): [DRI2] DRI driver: r200
[ 31.665] (II) GLX: Initialized DRI2 GL provider for screen 0

[ideasman42 (ideasman42)]

@Barry Grumbine.
I found that removing glClear(GL_COLOR_BUFFER_BIT); from wm_window.c fixes it and blender starts fine.

Once blender is started this can be called again for new windows.
looking into it further, will try fix for the next beta.

[ideasman42 (ideasman42)]

@Barry Grumbine.
Committed workaround for this bug into blender r32642.

[Barry Grumbine (barry-grumbine)]

@ideasman42
My first comment (#53) was using blender r32532. I built r32644 and r32646, both revisions still have jumbled text.

I've attempted to find other OpenGL projects that are affected by this, but none of them appear to render text the same as blender 2.50 Beta.

Oddly enough before updating to Maverick, I went to blender 2.50 Beta because blender 2.49b was not working properly (i.e. click "File" and the menu does not appear until the cursor is moved, very annoying) but now all is well with 2.49b, even the text.

Is this still related to freedesktop/mesa/opengl or should a new bug be filed with blender?

Is there anything I can do to help test this other than building and testing blender 2.50 Beta?

[Daniel Richard G. (skunk)]

Tormod, Fabio,

I'm not having much luck here. I am running an up-to-date Maverick on amd64, with the following:

* Current packages from the xorg-edgers PPA at https://launchpad.net/~xorg-edgers/+archive/radeon

* The 2.6.36-020636-generic mainline kernel from http://kernel.ubuntu.com/~kernel-ppa/mainline/v2.6.36-maverick/

I'm seeing the following problems:

* First, gltron(1) and other apps fail with "drmRadeonCmdBuffer: -22. Kernel failed to parse or rejected command stream. See dmesg for more info" as reported in bug #557266, even though that one was believed to have been fixed.

* If I build mesa from git, as Uwe did above, and use the resulting r200_dri.so, I eventually get a GPU lockup. I'm attaching the relevant /var/log/messages excerpt from that.

* If I have a GL app running (e.g. glxgears), and I click on the desktop pager to switch to another desktop, the X server segfaults. But I can't get a core dump from this for some reason---see bug #666601. (Note, I'm running the standard Ubuntu desktop, i.e. GNOME with Compiz et al.)

* And as if that weren't enough, there's random wonky text, in what could well be a milder form of what Barry is seeing. It comes up regularly when, say, I scroll a page in Firefox. Scroll up/down with a keypress, the text is "rough"; scroll again, it's perfect; scroll again, it's rough. I'll attach a partial screenshot showing this.

[Daniel Richard G. (skunk)]

Okay, an update:

I'm still running Maverick, with current packages from xorg-edgers/radeon (mesa @ git20101130), and the 2.6.37-020637rc3-generic mainline kernel. Many improvements:

* The "drmRadeonCmdBuffer: -22" error is history, thanks to Alex Deucher: https://bugs.freedesktop.org/show_bug.cgi?id=25544

* I'm not getting GPU lockups anymore!

* I still get the X server crash if I try to switch desktops while a GL app is running in (Ubuntu) GNOME. I was ultimately somewhat able to get a core dump, and send in bug #671293, but the use of PPA packages didn't play well with the Apport tracing service.

* Still seeing the wonky text, but this only shows up in the (Ubuntu) GNOME environment. My normal desktop is Xfce, and it's now running on Maverick with all the latest stuff, and I don't see anything like it there. Maybe it's a GNOME/Compiz thing?

[Timo Aaltonen (tjaalton)]

this bug has been fixed upstream, and natty.

[Santiago Roland (santiago-roland)]

what about maverick users?

santiago@pandora:/$ stellarium
Using default graphics system specified at build time: raster
QProcess: Destroyed while process is still running.
 -------------------------------------------------------
[ This is Stellarium 0.10.5 - http://www.stellarium.org ]
[ Copyright (C) 2000-2010 Fabien Chereau et al ]
 -------------------------------------------------------
Writing log file to: "/home/santiago/.stellarium/log.txt"
File search paths:
  0 . "/home/santiago/.stellarium"
  1 . "/usr/share/stellarium"
Config file is: "/home/santiago/.stellarium/config.ini"
Qt GL paint engine is: "OpenGL"
stellarium: ../../radeon/radeon_cs_gem.c:181: cs_gem_write_reloc: Assertion `boi->space_accounted' failed.
Aborted
santiago@pandora:/$

[dllud (dllud)]

Distro: maverick
Graphics card: ATI Technologies Inc RV280 [Radeon 9200 SE]
X Driver: radeon

Adding Tormod Volden's PPA (sudo apt-add-repository ppa:xorg-edgers/radeon) and updating to his latest drivers solves the issue with stellarium.