Bug 41124

Can you install debugging symbols and get a proper backtrace with gdb?

Created attachment 51539 [details]
stack trace

X.Org X Server 1.11.0
Release Date: 2011-08-26
X Protocol Version 11, Revision 0
Build Operating System: Linux 3.0.0-1-amd64 x86_64 Debian
Build Date: 28 August 2011  10:58:30AM
xorg-server 2:1.11.0-1 (Cyril Brulebois <kibi@debian.org>) 
Current version of pixman: 0.22.2

Created attachment 51770 [details] [review]
st/dri: Guard against NULL drawable

Does this Mesa patch fix the problem?

P.S. Are you running piglit with indirect rendering on purpose?

I am running with direct rendering enabled:

libGL: OpenDriver: trying /usr/lib/x86_64-linux-gnu/dri/tls/r600_dri.so
libGL: OpenDriver: trying /usr/lib/x86_64-linux-gnu/dri/r600_dri.so
libGL: Can't open configuration file /etc/drirc: No such file or directory.
libGL: Can't open configuration file /root/.drirc: No such file or directory.
libGL: Can't open configuration file /etc/drirc: No such file or directory.
libGL: Can't open configuration file /root/.drirc: No such file or directory.
name of display: :1.0
display: :1  screen: 0
direct rendering: Yes
server glx vendor string: SGI
server glx version string: 1.4

Backtrace:
0: X (xorg_backtrace+0x26) [0x5689d6]
1: X (0x400000+0x16c5c9) [0x56c5c9]
2: /lib/x86_64-linux-gnu/libpthread.so.0 (0x7fe049340000+0xf020) [0x7fe04934f020]
3: /usr/lib/x86_64-linux-gnu/dri/r600_dri.so (0x7fe0452bd000+0x45755) [0x7fe045302755]
4: /usr/lib/xorg/modules/extensions/libdri2.so (0x7fe0462b0000+0x1aca) [0x7fe0462b1aca]
5: /usr/lib/xorg/modules/extensions/libdri2.so (0x7fe0462b0000+0x1b97) [0x7fe0462b1b97]
6: X (0x400000+0xbc783) [0x4bc783]
7: X (ConfigureWindow+0x222) [0x45e2d2]
8: X (0x400000+0x3257b) [0x43257b]
9: X (0x400000+0x37ae9) [0x437ae9]
10: X (0x400000+0x26eaa) [0x426eaa]
11: /lib/x86_64-linux-gnu/libc.so.6 (__libc_start_main+0xfd) [0x7fe04807dead]
12: X (0x400000+0x2719d) [0x42719d]
Segmentation fault at address (nil)

The patch does not seem to make any difference unless the X server needs to load the patched mesa somehow.

I run only the application with LD_LIBRARY_PATH.

unpatched:


Backtrace:
0: X (xorg_backtrace+0x26) [0x5689d6]
1: X (0x400000+0x16c5c9) [0x56c5c9]
2: /lib/x86_64-linux-gnu/libpthread.so.0 (0x7f1fa6277000+0xf020) [0x7f1fa6286020]
3: /usr/lib/xorg/modules/extensions/libglx.so (0x7f1fa3a0a000+0x3e124) [0x7f1fa3a48124]
4: /usr/lib/xorg/modules/extensions/libdri2.so (0x7f1fa31e7000+0x1aca) [0x7f1fa31e8aca]
5: /usr/lib/xorg/modules/extensions/libdri2.so (0x7f1fa31e7000+0x1b97) [0x7f1fa31e8b97]
6: X (0x400000+0xbc783) [0x4bc783]
7: X (ConfigureWindow+0x222) [0x45e2d2]
8: X (0x400000+0x3257b) [0x43257b]
9: X (0x400000+0x37ae9) [0x437ae9]
10: X (0x400000+0x26eaa) [0x426eaa]
11: /lib/x86_64-linux-gnu/libc.so.6 (__libc_start_main+0xfd) [0x7f1fa4fb4ead]
12: X (0x400000+0x2719d) [0x42719d]
Segmentation fault at address 0xa8


patched:

Backtrace:
0: X (xorg_backtrace+0x26) [0x5689d6]
1: X (0x400000+0x16c5c9) [0x56c5c9]
2: /lib/x86_64-linux-gnu/libpthread.so.0 (0x7f264a877000+0xf020) [0x7f264a886020]
3: /usr/lib/xorg/modules/extensions/libglx.so (0x7f264800a000+0x3e124) [0x7f2648048124]
4: /usr/lib/xorg/modules/extensions/libdri2.so (0x7f26477e7000+0x1aca) [0x7f26477e8aca]
5: /usr/lib/xorg/modules/extensions/libdri2.so (0x7f26477e7000+0x1b97) [0x7f26477e8b97]
6: X (0x400000+0xbc783) [0x4bc783]
7: X (ConfigureWindow+0x222) [0x45e2d2]
8: X (0x400000+0x3257b) [0x43257b]
9: X (0x400000+0x37ae9) [0x437ae9]
10: X (0x400000+0x26eaa) [0x426eaa]
11: /lib/x86_64-linux-gnu/libc.so.6 (__libc_start_main+0xfd) [0x7f26495b4ead]
12: X (0x400000+0x2719d) [0x42719d]
Segmentation fault at address (nil)

(In reply to comment #5)
> The patch does not seem to make any difference unless the X server needs to
> load the patched mesa somehow.

It does.

(In reply to comment #4)
> I am running with direct rendering enabled:

The crash can only happen with indirect rendering, so at least some OpenGL app(s) end up using indirect rendering or you.

(In reply to comment #6)
> (In reply to comment #5)
> > The patch does not seem to make any difference unless the X server needs to
> > load the patched mesa somehow.
> 
> It does.

Running the X server with LD_LIBRARY_PATH set to the patched mesa does not make a difference either.

> 
> (In reply to comment #4)
> > I am running with direct rendering enabled:
> 
> The crash can only happen with indirect rendering, so at least some OpenGL
> app(s) end up using indirect rendering or you.

Then it is a bug in itself that some applications use indirect rendering at random.

(In reply to comment #7)
> Running the X server with LD_LIBRARY_PATH set to the patched mesa does not make
> a difference either.

Did you check the log file that it actually loaded the patched r600_dri.so?


> Then it is a bug in itself that some applications use indirect rendering at
> random.

More likely a configuration/setup issue.

(In reply to comment #8)
> More likely a configuration/setup issue.

Or maybe some piglit tests force indirect rendering.

I copied the patched libraries over the system libraries and the X server still crashes.

It might be crashing in different place, though. The original stack trace does not seem to include the glx module.

Created attachment 51796 [details]
stack trace with glx module

Created attachment 51797 [details] [review]
st/dri: Guard against NULL drawable (v2)

How about this patch instead?

Created attachment 51800 [details]
backtrace with the new patch

still crashes in glx

Looks like an xserver GLX/DRI2 issue then.

Looking at the test logs it seems one of the tests running at the time X crashed was a glx (=possibly indirect) test.

So it might be that X server fails at running indirect applications together with direct applications (possibly after some non-trivial number of attempts).

I expect that this is still an issue in 1.11.1, but can you please confirm.  
Thanks.

yes, the last stack traces were with 1.11.1

cannot reproduce with X server 1.12.2/mesa 8.1-dev