[vfio-users] CPU stall on dynamic rebinding with nvidia proprietary drivers

[Torbjorn Jansson]

On 2018-07-26 19:21, Alex John wrote:
> On Thu, Jul 26, 2018 at 06:56:48PM +0800, Alex John wrote:
>> Hello!
>>
>> I was experimenting with dynamically rebinding my GPU (nvidia <-> vfio-pci) and
>> it works exactly two times and crashes on the third time. More details and
>> kernel logs as follows:
>>
>> I boot the system using the 1080 GTX as the boot GPU, X starts fine, everything
> 
> I've tested that setting the i915 as the boot GPU in bios (IGD for those with
> MSI boards) works fine and allows me to keep switching GPUs without any
> problems.
> 
> So it has something to do with how the framebuffer/modesetting behaves when the
> nvidia GPU is set to boot GPU. I can take this to the linux-fbdev list to see if
> they have any answers as well.
> 

it is probably related to nvidia proprietary driver.

i have had issues with it in the past when console was using framebuffer and i 
still have.
sometimes it doesn't restore the text console when exiting X, everything still 
works, it is just that i cant see whats going on for example during a reboot.
if i disabled graphical framebuffer console for text mode i bet it would work 
better (plain old 80x25 character mode)

in my experience you are better off having the host not touch the nvidia gfx 
card you intend to use with a vm.
since i use a xeon cpu (for good iommu grouping) i don't have any built in gfx 
so i bought a cheap nvidia 1030 or something for host then bound vfio-pci to 
the other card.