https://bugzilla.kernel.org/show_bug.cgi?id=63101
Bug ID: 63101 Summary: Hard lockup whel launching games like TF2 on kernels 3.11.5 and 3.12 rc4 and above Product: Drivers Version: 2.5 Kernel Version: 3.11.5 Hardware: All OS: Linux Tree: Mainline Status: NEW Severity: normal Priority: P1 Component: Video(DRI - non Intel) Assignee: drivers_video-dri@kernel-bugs.osdl.org Reporter: laszlo.kertesz@gmail.com Regression: No
Created attachment 111161 --> https://bugzilla.kernel.org/attachment.cgi?id=111161&action=edit dmesg of the locked up session
I begin to experience system lockups when launching Team Fortress 2 on an A8-5500 (using its integrated 7560 Radeon video with the radeon driver) - it began starting around kernel 3.12 rc4, (i build the kernel from mainline git). I have a leftover package of 3.12 rc3+ and that works fine.
This also happens with the stable 3.11.5 kernel version.
Kernel 3.11.0 doesnt have this issue.
As i launch the game, the menus are shown fine, but as soon as i enter the actual game, either it crashes outright and throws me back to the desktop or i have the last screen frozen. In any way, i can move the mouse but cannot interact with windows, the keyboard listens only to the magic keys. From time to timethe system beeps. Dmesg is filled with warning and BUG/Oops entries.
Begins with:
Oct 15 21:21:17 laca-desktop kernel: [ 355.859798] radeon 0000:00:01.0: GPU lockup CP stall for more than 10000msec Oct 15 21:21:17 laca-desktop kernel: [ 355.859807] radeon 0000:00:01.0: GPU lockup (waiting for 0x0000000000001276 last fence id 0x0000000000001275) Oct 15 21:21:17 laca-desktop kernel: [ 355.866308] [TTM] Buffer eviction failed
Then follow warnings like:
WARNING: CPU: 3 PID: 4041 at drivers/gpu/drm/drm_mm.c:257 drm_mm_remove_node+0x104/0x110 [drm]()
After that the BUGs:
Oct 15 21:21:20 laca-desktop kernel: [ 358.208965] BUG: unable to handle kernel NULL pointer dereference at 0000000000000088
The system is unusable (i can move the mouse pointer in best case), i have to reboot with the sysrq combinations.
All of the problematic kernels share the new cpufreq code i suppose, they all work the same way - instead of powering the CPU between 0.9-1.34v as the kernels before, now i have 0.65-1.22 volts, the CPU barely reaching its nominal 3.2 GHz speed, and never above (before it was up to ~3.5GHz and boosting was working, now core boosting never works).