Bug 24634 - X server crash with rs780
Summary: X server crash with rs780
Status: RESOLVED FIXED
Alias: None
Product: xorg
Classification: Unclassified
Component: Server/General (show other bugs)
Version: git
Hardware: x86 (IA32) Linux (All)
: medium normal
Assignee: Xorg Project Team
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2009-10-19 22:35 UTC by Andrew Randrianasulu
Modified: 2009-12-01 00:06 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
gdb.txt - bt full (2.33 KB, text/plain)
2009-10-19 22:35 UTC, Andrew Randrianasulu
no flags Details
X log (132.82 KB, text/plain)
2009-10-19 22:36 UTC, Andrew Randrianasulu
no flags Details
xorg.conf (10.43 KB, application/octet-stream)
2009-10-19 22:36 UTC, Andrew Randrianasulu
no flags Details
gdb.txt - hopefully better log (2.64 KB, text/plain)
2009-10-20 08:54 UTC, Andrew Randrianasulu
no flags Details
Handle NULL return value from fbpict.c:copy_drawable(). (547 bytes, patch)
2009-11-05 06:23 UTC, Michel Dänzer
no flags Details | Splinter Review

Description Andrew Randrianasulu 2009-10-19 22:35:27 UTC
Created attachment 30578 [details]
gdb.txt - bt full

I have this hardware:

01:05.0 VGA compatible controller: ATI Technologies Inc Radeon 3100 Graphics (prog-if 00 [VGA controller])
        Subsystem: ASUSTeK Computer Inc. Device 82ee
        Flags: bus master, fast devsel, latency 0, IRQ 18
        Memory at d0000000 (32-bit, prefetchable) [size=256M]
        I/O ports at d000 [size=256]
        Memory at fbef0000 (32-bit, non-prefetchable) [size=64K]
        Memory at fbd00000 (32-bit, non-prefetchable) [size=1M]
        Expansion ROM at <unassigned> [disabled]
        Capabilities: [50] Power Management version 3
        Capabilities: [a0] MSI: Mask- 64bit+ Count=1/1 Enable-
        Kernel driver in use: radeon
        Kernel modules: radeon

software used:


kernel - Linus's tree up to commit 2fdc246aaf9a7fa088451ad2a72e9119b5f7f029
pixman - commit 358f96d20219b4460bfd8ecf88e69ff10044b577
xserver - commit 63f4bf39170eb2262617ef2dc95fd6d337b9dad5
mesa - commit ec5c23551cdb4c369d8f8f392208f4d4bf29911b
libdrm - commit 3a387a983ec40cd443e22c1f8d9a6b5b5a8fa0d1
xf86-video-ati - commit bd89b7501f294ac645390ef144df569953c81dc4

E16 1.0.0 - window manager with built-in compositor. (crash observed only with compositing on)
wine 1.2.25 (yes, its old, but even old Wine shouldn't crash whole X server?)
3DMark2000 (win32 3D benchmark, used for tracking basic wine performance)

If I run default benchmark (only change  mode to 1024x768x32 and buffering to double), let it run, note result (some 1500-2900, depending on compositor on/off), and try to leave app with menu -> exit - it crashes.
Comment 1 Andrew Randrianasulu 2009-10-19 22:36:14 UTC
Created attachment 30579 [details]
X log
Comment 2 Andrew Randrianasulu 2009-10-19 22:36:40 UTC
Created attachment 30580 [details]
xorg.conf
Comment 3 Andrew Randrianasulu 2009-10-19 22:41:43 UTC
For me, it looks like crash in some fallback codepath .... because Wine tries to create very big window (i can see it with compositor off), and  E16 pager tries to read it back at the same time (it doesn't crash if i set E16 pager to another mode, not 'Live Update'. Just some  black rectangle in place of window, slowly fading in).
Comment 4 Michel Dänzer 2009-10-20 06:24:35 UTC
The log file appears to be from an old Git snapshot of xserver - in current Git, line 307 of fbpict.c is no longer even in the function image_from_pict(). Does the problem still occur with a current Git snapshot? If so, please attach an updated backtrace.

Also reassigning away from EXA as the evidence so far points to fb.
Comment 5 Michel Dänzer 2009-10-20 06:34:20 UTC
Actually, the backtrace may just be confusing due to function inlining... try getting another backtrace with at least the fb module rebuilt without optimization.
Comment 6 Andrew Randrianasulu 2009-10-20 08:54:35 UTC
Created attachment 30588 [details]
gdb.txt - hopefully better log

X server was compiled with CFLAGS="-O0 -g"
Comment 7 Michel Dänzer 2009-11-05 06:23:22 UTC
Created attachment 30982 [details] [review]
Handle NULL return value from fbpict.c:copy_drawable().

Looks like this goes back to commit c3d6799cee7ff8411b3a05a7ab7e2a9e80c95059 ('Bug #594: CAN-2005-2495: Fix exploitable integer overflow in pixmap creation, [...]'), which prevents the allocation of pixmaps with width or height > 32767.

Does this patch fix the crash? It probably results in lost rendering though.
Comment 8 Andrew Randrianasulu 2009-11-05 22:37:05 UTC
(In reply to comment #7)
> Created an attachment (id=30982) [details]
> Handle NULL return value from fbpict.c:copy_drawable().
> 
> Looks like this goes back to commit c3d6799cee7ff8411b3a05a7ab7e2a9e80c95059
> ('Bug #594: CAN-2005-2495: Fix exploitable integer overflow in pixmap creation,
> [...]'), which prevents the allocation of pixmaps with width or height > 32767.
> 
> Does this patch fix the crash? It probably results in lost rendering though.
> 

Yes,  this patch helps, and super-long window was nearly always hardly visible with compositor, anyway.
Comment 9 Michel Dänzer 2009-12-01 00:06:13 UTC
The fix has landed in xserver master (and the crashing code has been removed on server-1.7-branch).


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.