I’m unable to activate GPU partitions on an HGX A100 system following the Shared NVSwitch Virtualization Model (chapter 7 of https://docs.nvidia.com/datacenter/tesla/pdf/fabric-manager-user-guide.pdf).
The sample code works, we can get basic info regarding available partitions etc. When setting any partition we get error -2 if the partition is valid for our system or -1 if the partition is not valid.
We would prefer following the Full Passthrough Virtualization Model but are unable to find any information on MMIO filtering and have been trying to get support via our vendor and NVidia - to no success.