https://bugs.freedesktop.org/show_bug.cgi?id=93360
Bug ID: 93360 Summary: [radeonsi/llvm] Segfault in CS:GO Product: Mesa Version: git Hardware: Other OS: All Status: NEW Severity: normal Priority: medium Component: Drivers/Gallium/radeonsi Assignee: dri-devel@lists.freedesktop.org Reporter: tdroste@gmx.de QA Contact: dri-devel@lists.freedesktop.org
Created attachment 120470 --> https://bugs.freedesktop.org/attachment.cgi?id=120470&action=edit Backtrace of the segfault
When loading of a level is done in CS:GO and the team selection menu should pop up CS:GO segfaults here:
#0 0xf4f2cd2d in llvm::LiveRange::find(llvm::SlotIndex) () from /usr/local/lib/dri/radeonsi_dri.so
Hardware: Hawaii PRO [Radeon R9 290]
This happens with most recent LLVM SVN and Mesa git, and also with LLVM and Mesa from October. So this seems to be some long standing error in either Mesa or LLVM that is triggered now by a change of CS:GO.
https://bugs.freedesktop.org/show_bug.cgi?id=93360
Tobias Droste tdroste@gmx.de changed:
What |Removed |Added ---------------------------------------------------------------------------- See Also| |https://bugs.freedesktop.or | |g/show_bug.cgi?id=92944
https://bugs.freedesktop.org/show_bug.cgi?id=93360
--- Comment #1 from Nicolai Hähnle nhaehnle@gmail.com --- Hi Tobias, thanks for the new report. As with bug #92944, could you please provide a log with R600_DEBUG=vs,gs,ps? This will show the segfaulting shader and should allow simple reproduction at least.
https://bugs.freedesktop.org/show_bug.cgi?id=93360
--- Comment #2 from Tobias Droste tdroste@gmx.de --- Created attachment 120476 --> https://bugs.freedesktop.org/attachment.cgi?id=120476&action=edit Output of R600_DEBUG=vs,gs,ps
https://bugs.freedesktop.org/show_bug.cgi?id=93360
sarnex commendsarnex@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |commendsarnex@gmail.com
--- Comment #3 from sarnex commendsarnex@gmail.com --- Created attachment 120488 --> https://bugs.freedesktop.org/attachment.cgi?id=120488&action=edit R600_DEBUG=vs,gs,ps
Hello. I have the exact same issue. The assert in LLVM is llvm::IndexListEntry* llvm::SlotIndex::listEntry() const: Assertion `isValid() && "Attempt to compare reserved index."' failed.
I've attached my R600_DEBUG=vs,gs,ps also. I have a HD 7950/Tahiti.
I hope this helps.
Thanks, sarnex
https://bugs.freedesktop.org/show_bug.cgi?id=93360
--- Comment #4 from Tobias Droste tdroste@gmx.de --- I found out why the last CS update caused this issue. They siwtched all graphic options to "auto" with the last update and for my graphics card it figured out that the highest preference in every setting should be the best ones. And with these settings CS:GO is crashing.
I switched backed everything to "low" and I can at least play the game again. But the driver should crash with everything set to high.
https://bugs.freedesktop.org/show_bug.cgi?id=93360
--- Comment #5 from Tobias Droste tdroste@gmx.de --- (In reply to Tobias Droste from comment #4)
I found out why the last CS update caused this issue. They siwtched all graphic options to "auto" with the last update and for my graphics card it figured out that the highest preference in every setting should be the best ones. And with these settings CS:GO is crashing.
I switched backed everything to "low" and I can at least play the game again. But the driver should crash with everything set to high.
"should *not* crash" of course :-)
https://bugs.freedesktop.org/show_bug.cgi?id=93360
--- Comment #6 from Nicolai Hähnle nhaehnle@gmail.com --- Thank you for posting the workaround!
I can confirm that the logs of both of you contain (different) shaders that crash with standalone llc, making this bug easy to reproduce at least.
https://bugs.freedesktop.org/show_bug.cgi?id=93360
--- Comment #7 from Tobias Droste tdroste@gmx.de --- Because bug 92944 was magically fixed I just tested again with everything set to high and it looks like this is not crashing anymore. This was with LLVM svn trunk and mesa git from 2 days ago.
Though the image looks wrong with everything set to high. The lighting is doesn't seem to be rendered correctly.
Sarnex is it working for you?
https://bugs.freedesktop.org/show_bug.cgi?id=93360
--- Comment #8 from sarnex commendsarnex@gmail.com --- (In reply to Tobias Droste from comment #7)
Because bug 92944 was magically fixed I just tested again with everything set to high and it looks like this is not crashing anymore. This was with LLVM svn trunk and mesa git from 2 days ago.
Though the image looks wrong with everything set to high. The lighting is doesn't seem to be rendered correctly.
Sarnex is it working for you?
Hi Tobias,
Yeah, it seems to be working for me also. I guess the underlying bug was fixed. Great works guys at AMD.
Thanks, sarnex
https://bugs.freedesktop.org/show_bug.cgi?id=93360
Michel Dänzer michel@daenzer.net changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |RESOLVED Resolution|--- |FIXED
dri-devel@lists.freedesktop.org
-
bugzilla-daemon＠freedesktop.org