On Fri, 8 Dec 2017 12:40:17 -0800
Jacob Pan <jacob.jun....@linux.intel.com> wrote:
> On Fri, 8 Dec 2017 13:52:00 +0000
> Jean-Philippe Brucker <jean-philippe.bruc...@arm.com> wrote:
>
> > On 07/12/17 21:51, Alex Williamson wrote:
> > >> Agree, IOMMU driver cannot enforce it. I think VFIO layer can make
> > >> sure page response come from the assigned device and its
> > >> guest/container.
> > >
> > > Can we enforce it via the IOMMU/VFIO interface? If the response is
> > > for a struct device, and not an rid/did embedded in a structure,
> > > then vfio can pass it through w/o worrying about it, ie. response
> > > comes in via ioctl with association to vfio device fd -> struct
> > > vfio_device -> struct device, iommu driver fills in rid/did.
> > > Thanks,
> >
> > Yes that's probably the best way, reporting faults and receiving
> > responses on the device fd.
> >
> Just to put these ideas in to code. The IOMMU API used by VFIO has
> struct device* (derived from fd), no did/rid (to be derived from
> struct device by IOMMU driver.)
>
> int intel_iommu_page_response(struct iommu_domain *domain, struct device *dev,
> struct page_response_msg *msg)
>
> IOMMU driver can further sanitize by checking whether this is a pending
> page request for the device, and refcount outstanding PRQs.
>
> Does it sound right?
Yep. Thanks,
Alex
> /**
> * Generic page response information based on PCI ATS and PASID spec.
> * @addr: servicing page address
> * @pasid: contains process address space ID, used in shared virtual
> memory(SVM)
> * @resp_code: response code
> * @page_req_group_id: page request group index
> * @type: group or stream/single page response
> * @private_data: uniquely identify device-specific private data for an
> * individual page response
>
> */
> struct page_response_msg {
> u64 addr;
> u32 pasid;
> u32 resp_code:4;
> #define IOMMU_PAGE_RESP_SUCCESS 0
> #define IOMMU_PAGE_RESP_INVALID 1
> #define IOMMU_PAGE_RESP_FAILURE 0xF
>
> u32 pasid_present:1;
> u32 page_req_group_id : 9;
> enum page_response_type type;
> u32 private_data;
> };
