On Wed, 5 Feb 2020 07:57:36 +0000 "Liu, Yi L" <yi.l.liu@xxxxxxxxx> wrote: > > From: Alex Williamson <alex.williamson@xxxxxxxxxx> > > Sent: Wednesday, February 5, 2020 7:18 AM > > To: kvm@xxxxxxxxxxxxxxx > > Subject: Re: [RFC PATCH 0/7] vfio/pci: SR-IOV support > > > > > > Promised example QEMU test case... > > > > commit 3557c63bcb286c71f3f7242cad632edd9e297d26 > > Author: Alex Williamson <alex.williamson@xxxxxxxxxx> > > Date: Tue Feb 4 13:47:41 2020 -0700 > > > > vfio-pci: QEMU support for vfio-pci VF tokens > > > > Example support for using a vf_token to gain access to a device as > > well as using the VFIO_DEVICE_FEATURE interface to set the VF token. > > Note that the kernel will disregard the additional option where it's > > not required, such as opening the PF with no VF users, so we can > > always provide it. > > > > NB. It's unclear whether there's value to this QEMU support without > > further exposure of SR-IOV within a VM. This is meant mostly as a > > test case where the real initial users will likely be DPDK drivers. > > > > Signed-off-by: Alex Williamson <alex.williamson@xxxxxxxxxx> > > Just curious how UUID is used across the test. Should the QEMU > which opens VFs add the vfio_token=UUID or the QEMU which > opens PF add the vfio_token=UUID? or both should add vfio_token=UUID. In this example we do both as this covers the case where there are existing VF users, which requires the PF to also provide the vf_token. If there are no VF users, the PF is not required to provide a vf_token and vfio-pci will not fail the device match if a vf_token is provided but not needed. In fact, when a PF is probed by vfio-pci a random vf_token is set, so it's required to use a PF driver to set a known vf_token before any VF users can access their VFs. Thanks, Alex
