On 8/12/2023 3:49 AM, Christoph Hellwig wrote:
Caution: This message originated from an External Source. Use proper caution when opening attachments, clicking links, or responding.
On Thu, Aug 10, 2023 at 02:19:40PM -0300, Jason Gunthorpe wrote:
It's somewhat a strange requirement since we have no expectation of
compatibility between vendors for any other device type, but how far
are we going to take it? Is it enough that the device table here only
includes the Ethernet VF ID or do we want to actively prevent what
might be a trivial enabling of migration for another device type
because we envision it happening through an industry standard that
currently doesn't exist? Sorry if I'm not familiar with the dynamics
of the NVMe working group or previous agreements. Thanks,
I don't really have a solid answer. Christoph and others in the NVMe
space are very firm that NVMe related things must go through
standards, I think that is their right.
Yes, anything that uses a class code needs a standardized way of
being managed. That is very different from say mlx5 which is obviously
controlled by Mellanox.
So I don't think any vfio driver except for the plain passthrough ones
should bind anything but very specific PCI IDs.
And AMD really needs to join the NVMe working group where the passthrough
work is happening right now. If you need help finding the right persons
at AMD to work with NVMe send me a mail offline, I can point you to them.
Hi Christoph,
We have folks at AMD participating in NVMe working groups and are aware
of TPARs related to NVMe live migration. We’re checking to be sure they
are up to speed on the discussions and will reach out to you if they
need help getting further involved.
As I mentioned in another response, I've been out for a few days so
apologies for the delayed response.
Thanks for your help,
Brett
