Created attachment 100377 [details] dmesg of nouveau before during and after the crash Pro-longed play of BZFlag or joining a very popular room for < 30mins where there are a lot of explosions will cause nouveau to crash. There is brief jitter followed by a prompt coredump.Attached is the dmesg log and the game for use as a test case. I am using the latest 3.14.4 kernel with mesa 10.1 on Arch.
Did this happen with earlier versions of mesa? If you can reproduce at will, an apitrace would be nice. (30 mins would be ridiculous though.) I'm not aware of any opcode emission bugs with nvc0 cards. Another thing to try would be to make a debug mesa build, and run the game with NV50_PROG_OPTIMIZE=0 -- that turns off all shader code optimizations, which would make it even less likely to hit a "weird" instruction emission situation. This might also be some vram corruption/buffer copy issues/something else unrelated to opcode emission...
The bug is reproducible at will within the time given. I have captured an APITrace but it was over 30GB. Would you like a trimmed version? The trim seems to break the apitrace replay function but does retain various technical information. It's only about 70MB for the last 2k frames. I am not sure how to make a debug build, but if you could point me in the right direction I can try. Please note that some site I found was claiming over 2GB to compile. My ISP doesn't even allow that data much per month on the current plan. Thanks.
-- 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/1065.
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.