FS#39848 - [linux] [nvidia] Bumblebee + NVIDIA Binary driver failure on Linux 3.14 Kernel
Attached to Project:
Arch Linux
Opened by pavan yalamanchili (pavanky) - Monday, 14 April 2014, 13:41 GMT
Last edited by Doug Newgard (Scimmia) - Wednesday, 13 May 2015, 22:35 GMT
Opened by pavan yalamanchili (pavanky) - Monday, 14 April 2014, 13:41 GMT
Last edited by Doug Newgard (Scimmia) - Wednesday, 13 May 2015, 22:35 GMT
|
Details
Description:
Bumblebee + NVIDIA Binary driver configuration fails[1] on the Linux 3.14 Kernel. [1]: When trying to use the CUDA or OpenGL (via primus / optimus), the laptop simply shuts down. Reverting to older versions gets it back to working configuration. Additional info: Failure configuration: * linux-3.14-5-x86_64.pkg.tar.xz * nvidia-334.21-4-x86_64.pkg.tar.xz * nvidia-utils-334.21-7-x86_64.pkg.tar.xz Working configuration: * linux-3.13.7-1-x86_64.pkg.tar.xz * nvidia-334.21-2-x86_64.pkg.tar.xz * nvidia-utils-334.21-6-x86_64.pkg.tar.xz Steps to reproduce: * Get hold of a machine with bumblebee * pacman -Syu * Install bumblebee + Binary driver * optirun glxgears * *machine shuts down* |
This task depends upon
I have a GT 650m as well. Can you confirm the driver and Linux kernel versions ?
Pkg versions below, my previous sysupdate was on Apr 3:
linux-3.13.8-1 -> linux-3.14.1-1
nvidia-334.21-2 -> nvidia-334.21-4
nvidia-utils-334.21-7
graphics chip is GF119M [Quadro NVS 4200M]
/edit: sorry, i just realized i'm not even on linux-3.14-5 yet. I'd need to enable testing repo for that, right?
When the gpu fall out the bus, after few seconds the cooling fan reaches its maximum speed and i have to shutdown in order to stop it.
My working configuration is:
linux 3.13.8-1
nvidia 331.49-1
nvidia-utils 331.49-1
bbswitch 0.8-6
Failure configuration is:
linux 3.14.1-1
nvidia 337.12-1
nvidia-utils 337.12-1
bbswitch 0.8-7
graphic chip: NVIDIA Corporation GK107M [GeForce GT 650M]
-- edit: adding "rcutree.rcu_idle_gp_delay=1" to cmdline solved my issue.
FS#39092. While that bug seems to have been fixed, this one still persists. I can confirm that "rcutree.rcu_idle_gp_delay=1" works as a workaround.$ pacman -Qi linux nvidia nvidia-utils bbswitch bumblebee xorg-server | grep 'Version' -B1
Name : linux
Version : 3.15.3-1
--
Name : nvidia
Version : 337.25-4
--
Name : nvidia-utils
Version : 337.25-3
--
Name : bbswitch
Version : 0.8-12
--
Name : bumblebee
Version : 3.2.1-3
--
Name : xorg-server
Version : 1.15.2-1