Created attachment 97947 [details]
When I tried to start a WebGL application (triggerrally.com), the GPU crashed while loading the game.
This happened with Chromium Nightly and Firefox Stable (28) and is reproducable.
I've attached the full log from journal - sorry, I don't know how much you guys need to debug. If you need any more information please let me know.
The error file (gpu crash dump at /sys/class/drm/card0/error) is empty / 0 bytes..?
HP Compaq 6510b w. Intel GMA X3100
Arch Linux x64,
The error state is essential; cat /sys/class/drm/card0/error | xz -9 > error.xz
$ sudo cat /sys/class/drm/card0/error
no error state collected
How can I make sure the error state is being collected (as I said, the bug is reproducable)?
It is collected. It only lasts until the next reboot however.
Created attachment 97976 [details]
the ominous error file
I wonder what happens when the system completely crashes and you cannot ssh into the machine and copy the file? :D
A bit unfriendly for users but I managed it... there you go!
Is there any update on this?
Do you need anything else?
I can reproduce this. The first time I went to the site, I got an assertion failure in the blitter code. The second time, it seemed to work, but I then got a GPU hang on 3DSTATE_CONSTANT_COLOR. I figured it was the first non-pipelined command, but it wasn't - 3DTATE_GLOBAL_DEPTH_OFFSET_CLAMP happened earlier in the batch, and there was no intervening drawing. Odd...
Please reopen if you can still reproduce with Mesa 17.0.