On Mon, Jan 25, 2021 at 5:22 AM Jason Gunthorpe <jgg@nvidia.com> wrote:

> SRIOV and SF's require a simple linear lookup to learn the "function"
> because the BAR space is required to be linear.

Isn't this still true even for NumVF's > 256? Wouldn't there still be
a contiguous VF BAR space? Don't the routing IDs simply carry on
incrementing by stride, with each being assigned the next slice of the
shared BAR space?

> Scaling a CAM to high sizes is physicaly infeasible, so all approaches
> to scaling PCI functions go this road of having a single large BAR
> space.

If the above is true, is there really a need to scale up CAM?

Regards,
Edwin Peer