Re: [PATCH v3 13/17] iommu: Revise iommu_group_alloc_default_domain()

[Jason Gunthorpe <jgg@nvidia.com>]

On Thu, Apr 06, 2023 at 11:40:12AM +0800, Baolu Lu wrote:
> On 4/6/23 7:38 AM, Jason Gunthorpe wrote:
> > diff --git a/drivers/iommu/iommu.c b/drivers/iommu/iommu.c
> > index 8623d3edc61014..5f6aecfd92c657 100644
> > --- a/drivers/iommu/iommu.c
> > +++ b/drivers/iommu/iommu.c
> > @@ -91,8 +91,9 @@ static const char * const iommu_group_resv_type_string[] = {
> >   static int iommu_bus_notifier(struct notifier_block *nb,
> >   			      unsigned long action, void *data);
> > -static int iommu_alloc_default_domain(struct iommu_group *group,
> > -				      struct device *dev);
> > +static struct iommu_domain *
> > +iommu_group_alloc_default_domain(struct iommu_group *group, int req_type);
> > +static int iommu_get_def_domain_type(struct device *dev);
> >   static struct iommu_domain *__iommu_domain_alloc(struct bus_type *bus,
> >   						 unsigned type);
> >   static int __iommu_attach_device(struct iommu_domain *domain,
> > @@ -444,13 +445,16 @@ int iommu_probe_device(struct device *dev)
> >   		 * support default domains, so the return value is not yet
> >   		 * checked.
> >   		 */
> > -		iommu_alloc_default_domain(group, dev);
> > -		if (group->default_domain) {
> > -			iommu_create_device_direct_mappings(group, dev);
> > -			ret = __iommu_group_set_domain_internal(
> > -				group, group->default_domain,
> > -				IOMMU_SET_DOMAIN_WITH_DEFERRED);
> > +		group->default_domain = iommu_group_alloc_default_domain(
> > +			group, iommu_get_def_domain_type(dev));
> > +		if (!group->default_domain) {
> > +			ret = -EINVAL;
> > +			goto err_unlock;
> >   		}
> 
> There are still some drivers that don't support default domains. It was
> allowed previously, but above seems to change that behavior. The
> iommu_probe_device() fails now on those platforms. Or perhaps I
> overlooked anything?

Blah, yes, this is fixed up in the next patches it is an artifact of
trying to split patches up. Let me think about it

Jason