Re: [PATCH v8 02/10] iommu/vt-d: Add nested translation helper function

[Lu Baolu]

Hi Jacob,

On 1/10/20 2:39 AM, Jacob Pan wrote:
> On Wed, 18 Dec 2019 10:41:53 +0800
> Lu Baolu <baolu.lu@xxxxxxxxxxxxxxx> wrote:
>
> > Hi again,
> >
> > On 12/17/19 3:24 AM, Jacob Pan wrote:
> > > +/**
> > > + * intel_pasid_setup_nested() - Set up PASID entry for nested
> > > translation
> > > + * which is used for vSVA. The first level page tables are used for
> > > + * GVA-GPA or GIOVA-GPA translation in the guest, second level
> > > page tables
> > > + *  are used for GPA-HPA translation.
> > > + *
> > > + * @iommu:      Iommu which the device belong to
> > > + * @dev:        Device to be set up for translation
> > > + * @gpgd:       FLPTPTR: First Level Page translation pointer in
> > > GPA
> > > + * @pasid:      PASID to be programmed in the device PASID table
> > > + * @pasid_data: Additional PASID info from the guest bind request
> > > + * @domain:     Domain info for setting up second level page tables
> > > + * @addr_width: Address width of the first level (guest)
> > > + */
> > > +int intel_pasid_setup_nested(struct intel_iommu *iommu,
> > > +			struct device *dev, pgd_t *gpgd,
> > > +			int pasid, struct
> > > iommu_gpasid_bind_data_vtd *pasid_data,
> > > +			struct dmar_domain *domain,
> > > +			int addr_width)
> > > +{
> > > +	struct pasid_entry *pte;
> > > +	struct dma_pte *pgd;
> > > +	u64 pgd_val;
> > > +	int agaw;
> > > +	u16 did;
> > > +
> > > +	if (!ecap_nest(iommu->ecap)) {
> > > +		pr_err("IOMMU: %s: No nested translation
> > > support\n",
> > > +		       iommu->name);
> > > +		return -EINVAL;
> > > +	}
> > > +
> > > +	pte = intel_pasid_get_entry(dev, pasid);
> > > +	if (WARN_ON(!pte))
> > > +		return -EINVAL;
> > > +
> > > +	pasid_clear_entry(pte);
> >
> > In some cases, e.g. nested mode for GIOVA-HPA, the PASID entry might
> > have already been setup for second level translation. (This could be
> > checked with the Present bit.) Hence, it's safe to flush caches here.
> >
> > Or, maybe intel_pasid_tear_down_entry() is more suitable?
> We don't allow binding the same device-PASID twice, so if the PASID
> entry was used for GIOVA/RID2PASID, it should unbind first, and
> teardown flush included, right?

Fair enough. Can you please add this as a comment to this function? So
that the caller of this interface can know this. Or add a check in this
function which returns error if the pasid entry has already been bond.

Best regards,
baolu