Re: [PATCH v9 1/4] iommu: Move def_domain type check for untrusted device into core

[Will Deacon <will@kernel.org>]

Hi Baolu,

On Mon, Nov 23, 2020 at 08:55:17PM +0800, Lu Baolu wrote:
> On 2020/11/23 20:04, Will Deacon wrote:
> > On Sat, Nov 21, 2020 at 09:56:17PM +0800, Lu Baolu wrote:
> > > @@ -1645,13 +1655,10 @@ struct __group_domain_type {
> > >   static int probe_get_default_domain_type(struct device *dev, void *data)
> > >   {
> > > -	const struct iommu_ops *ops = dev->bus->iommu_ops;
> > >   	struct __group_domain_type *gtype = data;
> > >   	unsigned int type = 0;
> > > -	if (ops->def_domain_type)
> > > -		type = ops->def_domain_type(dev);
> > > -
> > > +	type = iommu_get_mandatory_def_domain_type(dev);
> > 
> > afaict, this code is only called from probe_alloc_default_domain(), which
> > has:
> > 
> >          /* Ask for default domain requirements of all devices in the group */
> >          __iommu_group_for_each_dev(group, &gtype,
> >                                     probe_get_default_domain_type);
> > 
> >          if (!gtype.type)
> >                  gtype.type = iommu_def_domain_type;
> > 
> > so is there actually a need to introduce the new
> > iommu_get_mandatory_def_domain_type() function, given that a type of 0
> > always ends up resolving to the default domain type?
> 
> Another consumer of this helper is in the next patch:
> 
> +	dev_def_dom = iommu_get_mandatory_def_domain_type(dev);
> +
> +	/* Check if user requested domain is supported by the device or not */
> +	if (!type) {
> +		/*
> +		 * If the user hasn't requested any specific type of domain and
> +		 * if the device supports both the domains, then default to the
> +		 * domain the device was booted with
> +		 */
> +		type = iommu_get_def_domain_type(dev);
> +	} else if (dev_def_dom && type != dev_def_dom) {
> +		dev_err_ratelimited(prev_dev, "Device cannot be in %s domain\n",
> +				    iommu_domain_type_str(type));
> +		ret = -EINVAL;
> +		goto out;
> +	}
> 
> I also added the untrusted device check in
> iommu_get_mandatory_def_domain_type(), so that we don't need to care
> about this in multiple places.

I see, but isn't this also setting the default domain type in the case that
it gets back a type of 0? I think it would be nice to avoid having both
iommu_get_mandatory_def_domain_type() and iommu_get_def_domain_type() of we
can, as it's really not clear which one to use when and what is meant by
"mandatory" imo.

Will