On Wed, 11 Aug 2021 23:43:43 +0000 "Rustad, Mark D" <mark.d.rustad@xxxxxxxxx> wrote: > at 3:00 PM, Bjorn Helgaas <helgaas@xxxxxxxxxx> wrote: > > > So I wonder if we should just disallow VPD access for these NICs > > except on function 0. There was a little bit of discussion in that > > direction at [2]. > > If this is done, you'll have to be sure that any non-0 functions assigned > to guests (which will appear as function 0 to the guest!) does not appear > to have VPD and block access to those resources. That seems kind of messy, > but should work. One hopes that no guest drivers are hard-wired to assume > VPD presence based on the device type. A bit messy, yes. But if we're saying VPD is not essential for operation in the VM, I'd rather we mask it on all functions rather than create scenarios where different identical functions produce different userspace results. Thanks, Alex
