Created attachment 102452 [details]
Kernel stack dump

kernel: 3.16.0-0.rc4.git0.1.fc21.1.x86_64
mesa: 0.3.0-2.20140707.fc21.x86_64 (git master)
Xorg: 1.15.99.904-1.fc21.x86_64 (git master, no 904 build yet)

kernel command boot options:
BOOT_IMAGE=/vmlinuz-3.16.0-0.rc4.git0.1.fc21.1.x86_64
root=/dev/mapper/fedora_segfault-root ro vconsole.keymap=us rhgb slub_debug=-
cgroup_disable=memory console=tty0 console=ttyS0,115200n8 radeon.runpm=0
radeon.dpm=1 radeon.hard_reset=1 intel_iommu=on
vfio_iommu_type1.allow_unsafe_interrupts=1 LANG=en_CA.UTF-8

Kernel dump from lockup see attachment

1 [details] [review]) I am trying to use the PCI hard_reset radeon.ko option
2) pciehp decides to yank the GPU driver while it's doing a PCI reset (!)
3) When the GPU is reset, it's in a non-finished startup state.

This all happened when I tried to play Second Life with the Radeon GPU as
offload renderer on the Intel GM45 GPU which is doing the display.

To run Second Life I'm using:

$ LIBGL_DRI3_DISABLE=1
$ xrandr --setprovideroffloadsink 0x55 0xa2
$DRI_PRIME=1 ./singularity >& /dev/null &

Here is the Xrandr providers list:

Providers: number : 3
Provider 0: id: 0xa2 cap: 0xb, Source Output, Sink Output, Sink Offload crtcs:
3 outputs: 4 associated providers: 2 name:Intel
Provider 1: id: 0x55 cap: 0xf, Source Output, Sink Output, Source Offload, Sink
Offload crtcs: 2 outputs: 4 associated providers: 2 name:radeon
Provider 2: id: 0x55 cap: 0xf, Source Output, Sink Output, Source Offload, Sink
Offload crtcs: 2 outputs: 4 associated providers: 2 name:radeon