Bug 48305 - llpp gets segfault in r600g_dri.so
Summary: llpp gets segfault in r600g_dri.so
Status: RESOLVED MOVED
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/Gallium/r600 (show other bugs)
Version: git
Hardware: Other All
: medium normal
Assignee: Default DRI bug account
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2012-04-04 10:33 UTC by myxol
Modified: 2019-09-18 18:59 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
gdb output and xorg.log (133.22 KB, text/plain)
2012-04-04 13:40 UTC, myxol
Details
dmesg output (37.43 KB, text/plain)
2012-04-04 13:42 UTC, myxol
Details
mesa backtrace (27.78 KB, text/plain)
2012-04-05 05:54 UTC, myxol
Details

Description myxol 2012-04-04 10:33:58 UTC
Loading any pdf with llpp produces segfault. 
Message in dmesg:
[115638.669260] llpp[15916] general protection ip:b64f6d84 sp:bfe416dc error:0 in r600g_dri.so[b62d3000+a72000]

mupdf that is used by llpp works flawlessly on r600g. And llpp works on fglrx.
I'm using Radeon 2600 PRO AGPx4.
Comment 1 Alex Deucher 2012-04-04 11:06:35 UTC
Can you get a proper backtrace with gdb?  Can you also include your xorg log and dmesg output?
Comment 2 myxol 2012-04-04 13:40:19 UTC
Created attachment 59491 [details]
gdb output and xorg.log

gdb didn't produce any good backtrace. Could it be because llpp written in ocaml?
Comment 3 myxol 2012-04-04 13:42:36 UTC
Created attachment 59492 [details]
dmesg output
Comment 4 Alex Deucher 2012-04-04 13:45:21 UTC
Make sure you install the debugging symbols for the 3D driver/mesa.  Most distros package these separately.  Then when you hit the crash type "bt" in gdb to get the backtrace.
Comment 5 myxol 2012-04-05 05:54:49 UTC
Created attachment 59519 [details]
mesa backtrace

After playing with compilation flags to get meaningful backtrace I found that this segfault happens only with -O3… Dunno if this is considered as a bug.
Anyway, bt full attached. Mesa compiled with "-march=native -O3 -ggdb -pipe".
Comment 6 Michel Dänzer 2012-04-05 06:48:04 UTC
(In reply to comment #5)
> After playing with compilation flags to get meaningful backtrace I found that
> this segfault happens only with -O3… Dunno if this is considered as a bug.

It's a bug *somewhere*, but it looks like it could be in the compiler... Maybe you can try isolating which specific option implied by -O3 but not by -O2 makes the difference.
Comment 7 myxol 2012-04-05 17:49:05 UTC
> Maybe you can try isolating which specific option implied by -O3 but not by -O2 makes the difference.
I tried. It is -ftree-vectorize
Comment 8 Michel Dänzer 2012-04-06 04:11:52 UTC
Smells more like a compiler bug then. Which version of gcc is this? Does it also happen with an older one?
Comment 9 myxol 2012-04-07 07:28:42 UTC
I tried 3 gcc versions.
Flags: -march=native -O2 -ftree-vectorize -ggdb -pipe
4.4.7 - build fails
4.5.3 - original bugreport
4.6.2 - same problem as with 4.5.3
Comment 10 GitLab Migration User 2019-09-18 18:59:17 UTC
-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/mesa/mesa/issues/407.


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.