| Summary: | [gm45] GPU lockup render.IPEHR: 0x78080003 playing Minecraft (mesa?) | ||
|---|---|---|---|
| Product: | Mesa | Reporter: | Bryce Harrington <bryce> |
| Component: | Drivers/DRI/i965 | Assignee: | Ian Romanick <idr> |
| Status: | RESOLVED FIXED | QA Contact: | |
| Severity: | major | ||
| Priority: | high | CC: | brot+bfdo, chris, daniel, eugeni, jdgregson, kenneth |
| Version: | 7.11 | ||
| Hardware: | x86 (IA32) | ||
| OS: | Linux (All) | ||
| Whiteboard: | |||
| i915 platform: | i915 features: | ||
| Bug Depends on: | |||
| Bug Blocks: | 42991, 44622 | ||
| Attachments: |
XorgLog.txt
i915_error_state.txt hs_err_pid5105.log CurrentDmesg.txt BootDmesg.txt i915_error_state after a gpu hang on SNB Dmesg output after most recent GUP hang |
||
|
Description
Bryce Harrington
2011-09-21 18:04:08 UTC
Created attachment 51487 [details]
XorgLog.txt
Created attachment 51488 [details]
i915_error_state.txt
Created attachment 51489 [details]
hs_err_pid5105.log
Created attachment 51490 [details]
CurrentDmesg.txt
Created attachment 51491 [details]
BootDmesg.txt
Hello. I am the original reporter. I am here if you need any more logs, reproductions, patch testing, etc. I'm not sure how to subscribe to this, so if I forget about it, you can email me directly at jdgregson@gmail.com. Dies near the start of a batch. How early into the start of minecraft does it crash? This should be amenable to bisection, if you fancy the challenge. It is kind of difficult to get it to happen at the same time. For example, last night it crashed twice, within two minutes of starting the game. Today, it crashed twice only after the game had been running for ten or fifteen minutes. And by 'crash twice' I mean, two separate incidents. The time and cause of the crash seems somewhat random. The only thing that is not random is the fact that it always crashes. Also, I'm having difficulty getting the logs, because after it crashes, Unity also goes down most of the time, so I have to reboot, which seems to clear the Xorg logs. Perhaps I could try running it in a Desktop Environment other than Unity, which is already pretty unstable at this point. What do you mean by "This should be amenable to bisection, if you fancy the challenge."? > What do you mean by "This should be amenable to bisection, if you fancy the > challenge."? It means, you could probably narrow the bug down by following the steps documented here: https://wiki.ubuntu.com/Kernel/KernelBisection You know that it worked in natty, so take as your starting point "2.6.38", and end point "3.0.0". As a shortcut, or if building things from git looks like *too* much of a challenge, there are some pre-built kernel .debs at http://kernel.ubuntu.com/~kernel-ppa/mainline/, though this won't get you down to the exact commit. Alright, I am going to try to bisect this. I was wanting to work on kernels anyway. But I'm not confident that I will get any useful results. Is it possible that one of you could bisect and build them, and than have me test them? Hi Corbin, I wish we could pre-build all the kernels to make this easier, and maybe one day we will but unfortunately we're not really set up for that currently. So, sorry... we'll try to be available for questions though (but be aware the next couple weeks are crunch time for getting oneiric finalized.) I understand. I will seek closer support elsewhere, and post back if we find anything. Thank you for your help. Created attachment 52422 [details]
i915_error_state after a gpu hang on SNB
Hello everyone,
i do also get hangs with minecraft - however i have a Intel(R) Sandybridge Mobile (GT2+) running 3.1-rc6+ (drm-intel-next git 64a742fac3a22f57303d8f1b7e347350a1c48254)
I can run minecraft, but after some time i get a gpu hang. dmesg then says:
[18291.377254] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... GPU hung
[18291.377262] [drm] capturing error event; look for more information in /debug/dri/0/i915_error_state
[18291.380448] [drm:i915_wait_request] *ERROR* i915_wait_request returns -11 (awaiting 3243224 at 3243214, next 3243229)
Said /debug/dri/0/i915_error_state is attached.
If i need to provide more informations, feel free to ask.
Thanks for your help,
Michael
Michael, Though in the same game, your hang on Sandybridge turned out to be unrelated. I fixed it today in Mesa master: commit 3cc0a7be23ab603ed40d602595f673a44e079885 Author: Kenneth Graunke <kenneth@whitecape.org> Date: Fri Oct 21 01:03:37 2011 -0700 i965: Apply post-sync non-zero workaround to homebrew workaround. The fix should appear in Mesa 7.11.1. However, that doesn't resolve Corbin's original issue on GM45. This bug report continues to track that. Kenneth, i can confirm that this fixes my problem. I just updated mesa from git, and played for some time without crashes. Too bad that this doesnt fix things for Bryce - good luck that the problem will be fixed soon :) Thanks for your work Kenneth! Michael The patch was cherry-picked to 7.11, so closing this. Wait, did the patch resolve Michael's issue, as stated above, or is there another patch that also resolves my issue? As far as I know, it's still broken on GM45. Not sure why, unfortunately. Reopening. Okay, thank you. I can provide any testing needed still, but I discovered that bisecting the kernel myself is a quite a bit outside of my skill-set. Good morning, This bug does still exist. After about an hour of testing, my tester's game-play was ended by a GUP hang (see attached terminal output). Before this was reproduced, there was a similar problem, which may or may not be related. Because I have no idea what happened, I'll talk about it here. While the tester was playing, the X server crashed, and the screen switched to a TTY-like session with kernel messages about the error, and the kernel seems to have completely crashed. I believe that the kernel crashed because the computer was taking no input. I couldn't switch TTY Sessions or anything, I just had to do a hard reboot. Because I had no X server, the only valuable output I could get was by taking pictures of the screen with my cell phone. I uploaded the images to my site here: http://jdgregson.com/temp/ubuntubug/ It may be important to note that this error occurred while the tester was using the guest session included in lightdm. Hope this helps. Created attachment 53334 [details]
Dmesg output after most recent GUP hang
Nasty...looks like something deeper than Mesa. CC'ing Chris and Daniel, maybe one of them will have some insight. Comment on attachment 52422 [details]
i915_error_state after a gpu hang on SNB
This is an error_state for another bug!!
Looked a bit at it and not noticed much. We have a few fancy reports of strange crashes, so I think another error_state would be interesting to gauge the randomness involved in this issue. Running minecraft on my g45, it's been up and rendering for about 10 minutes and I've wandered around for a bit, no GPU hang. Can you figure out some way to reliably produce the hang in just a few minutes, so that we can actually debug? Even if you can't come up with a way for us to reproduce the problem, if you can manage to do it in a short time period once with INTEL_DEBUG=aub set in the environment, the resulting intel.aub file may be useful. It requires upstream Mesa, though -- 8.0.x doesn't have that logging feature. (note, for those with theories of possible GPU hangs, minecraft isn't using FBOs) I haven't been playing this game very much anymore, but the GPU hang hasn't happened for a while. It never was predictable, but it would usually happen within an hour or two. About a month ago my friend and I played for about eight hours straight and it didn't happen. It's possible that this was fixed somehow. I definitely remember some i915 updates since this GPU hang happened last. I guess then that we can tentatively close this issue. Thanks a lot for reporting it, and please reopen if it blows up again. |
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.