-
Notifications
You must be signed in to change notification settings - Fork 176
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add vDPA support #305
Comments
Hey Adrián Moreno, I have recently (today) signed-up on this repository. I am interested in this project as I want to make virtio devices available in K8s pod. Could you please let me know how I can understand and contribute? |
Hi @arunsharma75. Welcome! |
Thanks a lot Adrián |
@arunsharma75, the vDPA kernel framework was introduced in 5.7. If you need to work with an old kernel, you can try to make it work with the DPDK framework. It's still WIP but we have made it work on PoCs. Shoot me an email or ping me on slack if you need assistance setting it up |
vDPA technology allows Pods to consume accelerated device interfaces while using standard virtio / vhost drivers. This is possible because vdpa-capable devices implement the standard VirtIO ring layout that is used for the datapath. As for the control path, a framework was added to the linux kernel (first introduced in 5.7 and enhanced in 5.8 and 5.9) to translate virtio/vhost commands to vendor specific operations.
The vdpa kernel framework is exposed to userspace as a system bus (
/sys/bus/vdpa
). Vendor drivers, once bound to their VFs, create vdpa devices (that can be listed under/sys/bus/vdpa/devices
). Currently two vdpa drivers are available (under/sys/bus/vdpa/drivers
):vhost_vdpa
andvirtio_vdpa
What would you like to be added?
vdpa support should be added to the SR-IOV Device Plugin so Pods can consume vdpa devices that are created on top of SR-IOV VFs
What is the use case for this feature / enhancement?
vDPA has many usecases, mainly the capability to consume accelerated network devices (which theoretically could be as performant as direct SR-IOV devices) using standard drivers (e.g: virtio-net or virtio-user PMD driver in DPDK)
The text was updated successfully, but these errors were encountered: