On Tue, Feb 02, 2021 at 06:06:34PM -0500, Michael S. Tsirkin wrote:
> On Tue, Feb 02, 2021 at 03:13:14PM +1100, David Gibson wrote:
> > The default behaviour for virtio devices is not to use the platforms normal
> > DMA paths, but instead to use the fact that it's running in a hypervisor
> > to directly access guest memory. That doesn't work if the guest's memory
> > is protected from hypervisor access, such as with AMD's SEV or POWER's PEF.
> >
> > So, if a confidential guest mechanism is enabled, then apply the
> > iommu_platform=on option so it will go through normal DMA mechanisms.
> > Those will presumably have some way of marking memory as shared with
> > the hypervisor or hardware so that DMA will work.
> >
> > Signed-off-by: David Gibson <david@xxxxxxxxxxxxxxxxxxxxx>
> > Reviewed-by: Dr. David Alan Gilbert <dgilbert@xxxxxxxxxx>
> > Reviewed-by: Cornelia Huck <cohuck@xxxxxxxxxx>
> > Reviewed-by: Greg Kurz <groug@xxxxxxxx>
>
>
> > ---
> > hw/core/machine.c | 13 +++++++++++++
> > 1 file changed, 13 insertions(+)
> >
> > diff --git a/hw/core/machine.c b/hw/core/machine.c
> > index 94194ab82d..497949899b 100644
> > --- a/hw/core/machine.c
> > +++ b/hw/core/machine.c
> > @@ -33,6 +33,8 @@
> > #include "migration/global_state.h"
> > #include "migration/vmstate.h"
> > #include "exec/confidential-guest-support.h"
> > +#include "hw/virtio/virtio.h"
> > +#include "hw/virtio/virtio-pci.h"
> >
> > GlobalProperty hw_compat_5_2[] = {};
> > const size_t hw_compat_5_2_len = G_N_ELEMENTS(hw_compat_5_2);
> > @@ -1196,6 +1198,17 @@ void machine_run_board_init(MachineState *machine)
> > * areas.
> > */
> > machine_set_mem_merge(OBJECT(machine), false, &error_abort);
> > +
> > + /*
> > + * Virtio devices can't count on directly accessing guest
> > + * memory, so they need iommu_platform=on to use normal DMA
> > + * mechanisms. That requires also disabling legacy virtio
> > + * support for those virtio pci devices which allow it.
> > + */
> > + object_register_sugar_prop(TYPE_VIRTIO_PCI, "disable-legacy",
> > + "on", true);
> > + object_register_sugar_prop(TYPE_VIRTIO_DEVICE, "iommu_platform",
> > + "on", false);
>
> So overriding a boolean property always poses a problem:
> if user does set iommu_platform=off we are ignoring this
> silently.
No, we don't. That's why this is register_sugar_prop() rather than an
outright set_prop(). An explicitly given option will take precedence.
> Can we change iommu_platform to on/off/auto? Then we can
> change how does auto behave.
I've never had a satisfactory explanation of what the semantics of
"auto" need to be.
>
> Bonus points for adding "access_platform" and making it
> a synonym of platform_iommu.
>
> > }
> >
> > machine_class->init(machine);
>
--
David Gibson | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_
| _way_ _around_!
http://www.ozlabs.org/~dgibson
Attachment:
signature.asc
Description: PGP signature
