Re: [RFC PATCH] vfio/type1: Do not support IOMMUs that allow bypass

[Alex Williamson <alex.williamson@redhat.com>]

On Fri, 2015-10-16 at 16:03 +0200, Eric Auger wrote:
> Hi Alex,
> On 10/15/2015 10:52 PM, Alex Williamson wrote:
> > We can only provide isolation if DMA is forced through the IOMMU
> > aperture.  Don't allow type1 to be used if this is not the case.
> > 
> > Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
> > ---
> > 
> > Eric, I see a number of IOMMU drivers enable this, do the ones you
> > care about for ARM set geometry.force_aperture?  Thanks,
> I am currently using arm-smmu.c. I don't see force_aperture being set.

Hi Will,

Would it be possible to add iommu_domain_geometry support to arm-smmu.c?
In addition to this test to verify that DMA cannot bypass the IOMMU, I'd
eventually like to pass the aperture information out through the VFIO
API.  Thanks,

Alex
 
> >  drivers/vfio/vfio_iommu_type1.c |   12 ++++++++++++
> >  1 file changed, 12 insertions(+)
> > 
> > diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
> > index 57d8c37..6afa9d4 100644
> > --- a/drivers/vfio/vfio_iommu_type1.c
> > +++ b/drivers/vfio/vfio_iommu_type1.c
> > @@ -728,6 +728,7 @@ static int vfio_iommu_type1_attach_group(void *iommu_data,
> >  	struct vfio_group *group, *g;
> >  	struct vfio_domain *domain, *d;
> >  	struct bus_type *bus = NULL;
> > +	struct iommu_domain_geometry geometry;
> >  	int ret;
> >  
> >  	mutex_lock(&iommu->lock);
> > @@ -762,6 +763,17 @@ static int vfio_iommu_type1_attach_group(void *iommu_data,
> >  		goto out_free;
> >  	}
> >  
> > +	/*
> > +	 * If a domain does not force DMA within the aperture, devices are not
> > +	 * isolated and type1 is not an appropriate IOMMU model.
> > +	 */
> > +	ret = iommu_domain_get_attr(domain->domain,
> > +				    DOMAIN_ATTR_GEOMETRY, &geometry);
> > +	if (ret || !geometry.force_aperture) {
> > +		ret = -EPERM;
> > +		goto out_domain;
> > +	}
> > +
> >  	if (iommu->nesting) {
> >  		int attr = 1;
> >  
> > 
>