Re: [PATCH 2/3] intel-iommu: fail DEVIOTLB_UNMAP without dt mode

[Peter Xu <peterx@redhat.com>]

On Tue, Feb 07, 2023 at 05:17:42PM +0100, Laurent Vivier wrote:
> On 2/3/23 10:08, Laurent Vivier wrote:
> > On 11/29/22 09:10, Jason Wang wrote:
> > > Without dt mode, device IOTLB notifier won't work since guest won't
> > > send device IOTLB invalidation descriptor in this case. Let's fail
> > > early instead of misbehaving silently.
> > > 
> > > Signed-off-by: Jason Wang <jasowang@redhat.com>
> > > ---
> > >   hw/i386/intel_iommu.c | 8 ++++++++
> > >   1 file changed, 8 insertions(+)
> > > 
> > > diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> > > index 9143376677..d025ef2873 100644
> > > --- a/hw/i386/intel_iommu.c
> > > +++ b/hw/i386/intel_iommu.c
> > > @@ -3179,6 +3179,7 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
> > >   {
> > >       VTDAddressSpace *vtd_as = container_of(iommu, VTDAddressSpace, iommu);
> > >       IntelIOMMUState *s = vtd_as->iommu_state;
> > > +    X86IOMMUState *x86_iommu = X86_IOMMU_DEVICE(s);
> > >       /* TODO: add support for VFIO and vhost users */
> > >       if (s->snoop_control) {
> > > @@ -3193,6 +3194,13 @@ static int vtd_iommu_notify_flag_changed(IOMMUMemoryRegion *iommu,
> > >                            PCI_FUNC(vtd_as->devfn));
> > >           return -ENOTSUP;
> > >       }
> > > +    if (!x86_iommu->dt_supported && (new & IOMMU_NOTIFIER_DEVIOTLB_UNMAP)) {
> > > +        error_setg_errno(errp, ENOTSUP,
> > > +                         "device %02x.%02x.%x requires device IOTLB mode",
> > > +                         pci_bus_num(vtd_as->bus), PCI_SLOT(vtd_as->devfn),
> > > +                         PCI_FUNC(vtd_as->devfn));
> > > +        return -ENOTSUP;
> > > +    }
> > >       /* Update per-address-space notifier flags */
> > >       vtd_as->notifier_flags = new;
> > 
> > Reviewed-by: Laurent Vivier <lvivier@redhat.com>
> > Tested-by: Laurent Vivier <lvivier@redhat.com>
> > Buglink: https://bugzilla.redhat.com/2156876
> 
> Is this possible to have this patch merged?
> It fixes a real problem and it is really trivial.

AFAIU Jason will post a new version soon for this whole set.  But I also
agree if Michael has an earlier pull we can add this in even earlier.

Thanks,

-- 
Peter Xu