Created attachment 117230 [details] Crash dump Hello! A few minutes ago, I experienced a momentary stuttering, my video player crashed, and the game I was playing (Eduke32) stopped for a second. I checked dmesg and saw: [ 4678.947120] [drm] GPU HANG: ecode 6:0:0xf288fff8, in eduke32 [1828], reason: Ring hung, action: reset [ 4678.947123] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace. [ 4678.947123] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel [ 4678.947124] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue. [ 4678.947125] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it. [ 4678.947126] [drm] GPU crash dump saved to /sys/class/drm/card0/error [ 4678.953044] drm/i915: Resetting chip after gpu hang ... so here I am. The operating system is an up-to-date version of Arch Linux. The specific package versions of everything I believe to be relevant: linux 4.0.7-2 linux-firmware 20150527.3161bfa-1 xf86-video-intel 1:2.99.917+364+gb24e758-1 xorg-server 1.17.2-2 mesa-libgl 10.6.2-1 The crash dump is attached along with the dmesg of the machine.
Created attachment 117231 [details] dmesg
Please let me know if you need more information.
That looks like a bad shader. Is it reproducible? If so could you capture an apitrace of the gpu hang?
The problem hasn't happened again. I took a look into eduke32, and it seems that it actually relies entirely on the fixed-function pipeline. I've published a trace of about ten seconds of gameplay, I don't see any calls to glUseProgram in that time: http://waste.io7m.com/2015/07/18/eduke32.trace.xz
To clarify, given that the game essentially recreates a software renderer from the 90s, I'd expect that the OpenGL calls made in the first frame are essentially the same calls that would be made in any frame.
Please reopen if you can still reproduce with Mesa 17.0.
Will do! Apologies for the scarcity of information, but it was an issue triggered by software that I didn't write with no obvious way to reproduce it. I've not seen it happen since, so let's assume that it's fixed for now.
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.