Nvidia Tesla vGPU mdevctl

[Jacob Nollette]

Hi there,

I installed the currently nvidia drivers from the enterprise nvidia enterprise portal. Black listed radeon, etc, not nvidia. Installed .8 (something) mdevctl. IOMMU is enabled, and I have been using PCI passthrough successfully. Trying to split apart this tesla p4 into multiple gpus.

MDEVCTL uuid and instance types seem to be showing data correctly. Once I start mdevs, my available instances go down.

In proxmox, I try to add the pci passthrough device; chose pci adapter 84. The mdev shows up, and lists instances for my device, but nothing is available.

Any ideas?

I understand tesla p4 should be able to passthrough; could this be a proxmox issue?

 

[rrprime]

please explain "but nothing is available."
do you see the mdevs types ?:


what does "mdevctl list" show ?

 

[Jacob Nollette]

Hi there,
sorry for the delay...

First and foremost, in mdevctl I can see types, and available (available count). I can define and activate types. But available count doesn't show up in proxmox. In proxmox I see the same available types, but the count isn't there.

 

[rrprime]

check you DMs

 

[kromberg]

I am trying to do something similar here. When I run 'mdevctl types' all the 'Available instances: ' counts are zero. What does that mean?