glxinfo command causes Xorg to abort on Dimension E520

[Marc Tardif]

When running the glxinfo command on a Dimension E520 running Gutsy Tribe-1, Xorg aborts and then gdm presents the login prompt again.

[Idr (idr)]

Looking at that stack trace, it appears that the crash only happens with
indirect rendering. If I'm not mistaken, it should never go into libGLcore with
direct rendering.

Can you try running "LIBGL_ALWAYS_INDIRECT=y glxgears"? If my hunch is correct,
that will reproduce the crash.

[3rg34mw02 (3rg34mw02)]

I tried "env LIBGL_ALWAYS_INDIRECT=y glxgears" (I use tcsh rather than bash),
and there was no crash. (Also, when I'm running with xinerama turned on, which
disables DRI, then 3d applications don't crash X, though of course they
generally run unacceptably slowly.)

However, on a whim I tried running "env LIBGL_ALWAYS_INDIRECT=n glxgears", and I
did get a crash, though the traceback is different:

Backtrace:
0: X(xf86SigHandler+0x84) [0x80c0334]
1: [0xb7f8a420]
2: /usr/lib/dri/mga_dri.so(__driUtilUpdateDrawableInfo+0x137) [0xb3117297]
3: /usr/lib/dri/mga_dri.so(mgaUpdateRects+0xc0) [0xb311fd40]
4: /usr/lib/dri/mga_dri.so(mgaGetLock+0x7b) [0xb315f65b]
5: /usr/lib/dri/mga_dri.so [0xb311bcba]
6: /usr/lib/dri/mga_dri.so(_mesa_Clear+0x17f) [0xb327de3f]
7: /usr/lib/xorg/modules/extensions//libglx.so [0xb7c51979]
8: /usr/lib/xorg/modules/extensions//libglx.so(DoRender+0x175) [0xb7c4a575]
9: /usr/lib/xorg/modules/extensions//libglx.so [0xb7c4a5fa]
10: /usr/lib/xorg/modules/extensions//libglx.so [0xb7c4ea0c]
11: X [0x813c7de]
12: X(Dispatch+0x1ab) [0x808685b]
13: X(main+0x489) [0x806e619]
14: /lib/libc.so.6(__libc_start_main+0xd8) [0xb7d0a878]
15: X(FontFileCompleteXLFD+0xa5) [0x806d951]

Setting LIBGL_ALWAYS_INDIRECT to either y or n had no effect on the glxinfo
crash that I could tell.

[Daniel Stone (daniels)]

Sorry about the phenomenal bug spam, guys. Adding xorg-team@ to the QA contact so bugs don't get lost in future.

[Marc Tardif (cr3)]

When running the glxinfo command on a Dimension E520 running Gutsy Tribe-1, Xorg aborts and then gdm presents the login prompt again.

[Marc Tardif (cr3)]

From the xorg.conf, it seems to be that the problem is that the vesa Xorg driver is aborting when being queried by glxinfo. That is one problem, another problem is that the graphics controller is not being detected properly but I will report another bug about that.

[Timo Aaltonen (tjaalton)]

vesa does not support GLX, and your best bet is to use fglrx. I'd close this bug.

[Marc Tardif (cr3)]

Indeed, I wouldn't expect vesa to support GLX. However, I don't think Xorg should abort when the vesa driver is being queried by glxinfo.

[Ben Collins (ben-collins)]

I think Marc is correct, X shouldn't die when vesa is queried for glx support.

[Bryce Harrington (bryce)]

We're getting a similar backtrace on a Dimension E520 when running with the vesa driver: https://bugs.launchpad.net/xorg-server/+bug/119341

[Bryce Harrington (bryce)]

I'm not able to reproduce this either on Gutsy Tribe1 or on Feisty.

On an ati mobility x600 system with Gutsy Tribe1 installed, and the latest fglrx driver, I switched it to using "vesa", restarted X, and ran glxinfo. It complained about DRI missing but did not cause X to die.

Next I tried on a Dell 1505n with the factory-installed Feisty. It uses the i810 driver, but I switched it to vesa. It returns an error GLXBadContext but does not cause X to die.

However, I found this bug report, that mentions an almost identical backtrace: https://bugs.freedesktop.org/show_bug.cgi?id=9080

Can you try the test mentioned in that bug, to reproduce the crash?
"LIBGL_ALWAYS_INDIRECT=y glxgears"

[Martin Pitt (pitti)]

not super-urgent and not reproducible, moving milestone as agreed on in yesterday's meeting

[Michael Vogt (mvo)]

I can reproduce this here on a radeon r200 based card. Calling glxinfo makes my session end. This is a serious issues for the compiz-by-default spec that relies on testing for various capabilities of the gfx card used. It uses glxinfo for this currently.

[Sarah Kowalik (hobbsee-deactivatedaccount)]

Changing milestone to tribe 4.

Bryce, please debug this with mvo, when both of you are around.

[Bryce Harrington (bryce)]

mvo or marc, could either of you try running

 LIBGL_ALWAYS_INDIRECT=y glxgears

and see if that also crashes?

[Tormod Volden (tormodvolden)]

from bug 115609:
FYI, glxinfo would crash the server using the vesa driver on an ATI
X1300 in Tribe 5, but not after I installed xserver-xorg-core_1.3.0.0
.dfsg-12ubuntu4_i386.deb .

[David Tomaschik (matir)]

I can confirm that the 12ubuntu4 version of xserver-xorg-core fixes the glxinfo crash.

[Bryce Harrington (bryce)]

The bug from comment #4 was resolved due to a backported patch as of xserver 12ubuntu4

[Bryce Harrington (bryce)]

Thanks Tormod and David for confirming this as fixed.

[Jeremy Sequoia (jeremyhu)]

Is this still an issue?

[Ajax-a (ajax-a)]

glcore is no longer a thing, and neither is mga's 3d support.

[Timothy R. Chavez (timrchavez)]

The bug task for the somerville project has been removed by an automated script. This bug has been cloned on that project and is available here: https://bugs.launchpad.net/bugs/1305728