[PATCH 0/2] Unprivileged sgl-only passthrough

[Kanchan Joshi]

On Thu, Oct 19, 2023 at 1:05 AM Keith Busch <kbusch at kernel.org> wrote:
> SGL for metadata is definitely slower, but it's the only nvme protocol
> way to directly specify how much memory is actually available for the
> command's transfer. PRP/MPTR vs SGL is like strcpy() vs strncpy().
>
> Similiar to Kanchan's earlier experience though, I haven't found real
> nvme devices that support the SGL mode for metadata. The scenarios this
> enables might be pretty limited. :(

And if not this, what should be the solution to have non-root
passthrough? This is the third one (excluding iommu) that we are
discussing.