On Wed, Mar 27, 2024 at 11:39:15AM -0600, Alex Williamson wrote:
> On Fri, 22 Mar 2024 12:12:10 +0530
> Vinayak Kale <vkale@xxxxxxxxxx> wrote:
>
> > In case of migration, during restore operation, qemu checks config space of the
> > pci device with the config space in the migration stream captured during save
> > operation. In case of config space data mismatch, restore operation is failed.
> >
> > config space check is done in function get_pci_config_device(). By default VSC
> > (vendor-specific-capability) in config space is checked.
> >
> > Due to qemu's config space check for VSC, live migration is broken across NVIDIA
> > vGPU devices in situation where source and destination host driver is different.
> > In this situation, Vendor Specific Information in VSC varies on the destination
> > to ensure vGPU feature capabilities exposed to the guest driver are compatible
> > with destination host.
> >
> > If a vfio-pci device is migration capable and vfio-pci vendor driver is OK with
> > volatile Vendor Specific Info in VSC then qemu should exempt config space check
> > for Vendor Specific Info. It is vendor driver's responsibility to ensure that
> > VSC is consistent across migration. Here consistency could mean that VSC format
> > should be same on source and destination, however actual Vendor Specific Info
> > may not be byte-to-byte identical.
> >
> > This patch skips the check for Vendor Specific Information in VSC for VFIO-PCI
> > device by clearing pdev->cmask[] offsets. Config space check is still enforced
> > for 3 byte VSC header. If cmask[] is not set for an offset, then qemu skips
> > config space check for that offset.
> >
> > Signed-off-by: Vinayak Kale <vkale@xxxxxxxxxx>
> > ---
> > Version History
> > v2->v3:
> > - Config space check skipped only for Vendor Specific Info in VSC, check is
> > still enforced for 3 byte VSC header.
> > - Updated commit description with live migration failure scenario.
> > v1->v2:
> > - Limited scope of change to vfio-pci devices instead of all pci devices.
> >
> > hw/vfio/pci.c | 24 ++++++++++++++++++++++++
> > 1 file changed, 24 insertions(+)
>
>
> Acked-by: Alex Williamson <alex.williamson@xxxxxxxxxx>
A very reasonable way to do it.
Reviewed-by: Michael S. Tsirkin <mst@xxxxxxxxxx>
Merge through the VFIO tree I presume?
>
> > diff --git a/hw/vfio/pci.c b/hw/vfio/pci.c
> > index d7fe06715c..1026cdba18 100644
> > --- a/hw/vfio/pci.c
> > +++ b/hw/vfio/pci.c
> > @@ -2132,6 +2132,27 @@ static void vfio_check_af_flr(VFIOPCIDevice *vdev, uint8_t pos)
> > }
> > }
> >
> > +static int vfio_add_vendor_specific_cap(VFIOPCIDevice *vdev, int pos,
> > + uint8_t size, Error **errp)
> > +{
> > + PCIDevice *pdev = &vdev->pdev;
> > +
> > + pos = pci_add_capability(pdev, PCI_CAP_ID_VNDR, pos, size, errp);
> > + if (pos < 0) {
> > + return pos;
> > + }
> > +
> > + /*
> > + * Exempt config space check for Vendor Specific Information during restore/load.
> > + * Config space check is still enforced for 3 byte VSC header.
> > + */
> > + if (size > 3) {
> > + memset(pdev->cmask + pos + 3, 0, size - 3);
> > + }
> > +
> > + return pos;
> > +}
> > +
> > static int vfio_add_std_cap(VFIOPCIDevice *vdev, uint8_t pos, Error **errp)
> > {
> > PCIDevice *pdev = &vdev->pdev;
> > @@ -2199,6 +2220,9 @@ static int vfio_add_std_cap(VFIOPCIDevice *vdev, uint8_t pos, Error **errp)
> > vfio_check_af_flr(vdev, pos);
> > ret = pci_add_capability(pdev, cap_id, pos, size, errp);
> > break;
> > + case PCI_CAP_ID_VNDR:
> > + ret = vfio_add_vendor_specific_cap(vdev, pos, size, errp);
> > + break;
> > default:
> > ret = pci_add_capability(pdev, cap_id, pos, size, errp);
> > break;
