When running the World of Padman or OpenArena games with Mesa from 7.10 branch or master (up to 2010-01-17 was last test) on SNB hardware (Core i5 2500K; ASUS P8H67-M PRO motherboard) with Linux 2.6.37 kernel, xf86-video-intel and libdrm Git, the game will crash. OpenArena makes it about one second while in Padman it will run for a few seconds before crashing in the same spot, 100% reproducible thus far. I've been running it with just a simple 'phoronix-test-suite benchmark padman' or 'phoronix-test-suite benchmark openarena' to setup either game in a standardized config and demo and execute the demo causing the crash. I haven't encountered this on non-SNB Intel hardware or with other drivers/hardware. Videos Exhibiting The Problem: http://www.phoronix.com/scan.php?page=article&item=intel_sandy_breaks&num=2 Nothing of interest is outputted, the log just shows both games 'Received signal 11, exiting...' from ioquake3.
See if QA can reproduce.
Michael, could you list the tip git commit you use for all components? Could you attach /proc/cpuinfo, lspci -vvnn output? And dmesg and intel_gpu_dump output when in failure?
Created attachment 42191 [details] cpuinfo output
Created attachment 42192 [details] dmesg ouput
Created attachment 42193 [details] lspci oupu
I don't have the specific Git revisions handy but I tested it on the 7.10 and master branches of Mesa as of yesterday (17 January) and today so it shouldn't be hard to reproduce if it's not an isolated incident to my platform. Attached is the cpuinfo and lspci. I don't have the intel_gpu_dump output handy but can try to get that when time allows if you're unable to reproduce the issue. There's the dmesg from the 2.6.37 kernel attached as well. I've had no-go with the 2.6.38 DRM as well, but I'm in the middle of some other tests so I don't have the 2.6.38 dmesg handy, but like the 2.6.37 dmesg, it doesn't appear interesting at least at the tail. I'm running latest Git for xf86-video-intel and libdrm as well.
Michael, OpenArena runs well on our side. We're downloading Padman to test. We use 2010Q4 release, and don't think 2.6.38 is necessary in this case. Can you clarify the Phoronix versions with the games?
OpenArena is 0.8.5 and Padman is 1.2.
I tried on one SNB desktop qual board with D2 CPU, which is same as Michael's stepping, but its pci id is 0x0102. I use kernel 2.6.37, and others are all git tips. The games I've run on it, openarena 0.8.5, world of padman v1.5, nexuiz 2.5.2, urban terror v4.1, alien-arena 7.40, tremulous 1.1.0. All games I run seem fine on this SNB. I played each for about several minutes without problem.
Our testing with game demos from Phoronix test suite also shows good. All these games can complete running, render correctly, and with better performance than ever: openarena 0.8.5 padman 1.2 warsow 0.5 nexuiz 2.5.2 smokin guns 1.1b4 test hardware: Sandy Bridge desktop: Core i7 2600 and Core i5 2500, pci id 0102, rev09 Sandy Bridge mobile: pci id 0126, rev08 desktop board: ASUS P8H67-M PRO
Gordon: Can you confirm what your software stack is for that working setup? Also, from the BIOS do you have it set to the normal / ASUS Optional / Power savings modes and what is the ASUS EPU enabled or disabled? In looking at the other crashes I've been experiencing and Keith and Jesse thinking it may have to do with the power management code, I found that with the EPU was set to enabled that it was slightly more stable for me (not for these ioquake3 crashes, but for the tiling corruption / lock-up) but still not perfect.
(In reply to comment #11) > Gordon: > Can you confirm what your software stack is for that working setup? We tried both 2010Q4 release and the current git (as mentioned by Zhenyu). For kernel we stick on 2.6.37. > Also, from the BIOS do you have it set to the normal / ASUS Optional / Power > savings modes and what is the ASUS EPU enabled or disabled? In looking at the > other crashes I've been experiencing and Keith and Jesse thinking it may have > to do with the power management code, I found that with the EPU was set to > enabled that it was slightly more stable for me (not for these ioquake3 > crashes, but for the tiling corruption / lock-up) but still not perfect. Zhenyu is using the ASUS board (while I'm using Intel board). He may provide that info.
Michael, in our testing we have 'System Performance' set to 'Normal' in EZ mode, and EPU Power saving mode is disabled. How about we both load bios with F5 for optimized defaults to see any difference in test result? btw, have you tried to run games by hand instead of through phonorix test suite? There might be mode setting failure issue which could cause game to quit, as it looks like from your video clip.
I've tried with power savings modes, normal, and the optimal mode, along with various other power and CPU related features from the 'advanced mode' to no avail. But enabling EPU seemed to help just a bit with the other issues. I haven't tried running the games normally but don't think it would be a mode-setting issue as I've tried running them at all available resolutions and for World of Padman there's a number of seconds in between setting the mode and when the crash occurs. But I can try running it normally anyways once I get back as I am now out of office until end of month.
Hi Michael, I found the game smokin guns(32 bit) I get from phoronix(version 1.1b4) always run with software raster, I found it was because this game smokinguns.i386 isn't a dynamic executable, and it used static library libdrm and libdrm_intel in system. So when it runs, it can't use our driver. Other games didn't have such issue. And on 64 bit system it works well. Do you have knowledge on this? Thanks. What I do is as following: 1. phoronix-test-suite install smokin-guns 2. phoronix-test-suite run smokin-guns
Close this. Reopen if there's still issue.
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.