vfio: Add NvSwitches to the list of devices #150
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
There was a problem hiding this comment.
Pull request overview
This pull request adds support for binding NVIDIA NVSwitches to the vfio-pci driver for DGX/HGX systems requiring virtualization. This enables full passthrough and ServiceVM virtualization modes.
Changes:
- Added NVSwitch device discovery and binding in the
bindAll()method
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
zvonkok
force-pushed
the
enable-nvswitch
branch
from
391bbe8 to
df87aec
Compare
Collaborator
|
/ok to test df87aec |
Collaborator
|
/ok to test e16d9b6 |
zvonkok
force-pushed
the
enable-nvswitch
branch
from
e16d9b6 to
eef1c70
Compare
Contributor
Author
|
/ok to test eef1c70 |
zvonkok
force-pushed
the
enable-nvswitch
branch
from
eef1c70 to
2ed804a
Compare
Contributor
Author
|
Tested on a Viking, all four NvSwitches and all eight GPUs are bound to |
Contributor
Author
|
Add boolean to enable/disable binding On a NVLINK4 system if we want to support virt-model 1 8 GPUs bound to vfio-pci 4 NvSwitches and 8 GPUs bound to vfio-pci |
zvonkok
force-pushed
the
enable-nvswitch
branch
from
d189a3c to
498a27d
Compare
In when using DGX/HGX systems with virtualization we need to bind the NvSwitchess to vfio-pci as well. This enables us to use two virtualization modes 1. Full passthrough 2. ServiceVM We're explicitly not considering mode 3 with vGPUs. Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
Despite its name, GetGPUByPciBusID returns any NVIDIA PCI device (GPU, NVSwitch, etc.) at the specified address, not just GPUs. Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
On a NVLINK4 system if we want to support virt-model 1 we need to bind the switches to vfio-pci, supporting virt-model 3 with fabric-manager running on the host we do not need to bind the switches to vfio-pci. Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
In when using DGX/HGX systems with virtualization
we need to bind the NvSwitches to vfio-pci as well.
This enables us to use two virtualization modes
We're explicitly not considering mode 3 with vGPUs.