Nested translation has two stage address translations to get the final
physical addresses. Take Intel VT-d as an example, the first stage translation
structure is I/O page table. As the below diagram shows, guest I/O page
table pointer in GPA (guest physical address) is passed to host to do the
first stage translation. Along with it, guest modifications to present
mappings in the first stage page should be followed with an iotlb invalidation
to sync host iotlb.
.-------------. .---------------------------.
| vIOMMU | | Guest I/O page table |
| | '---------------------------'
.----------------/
| PASID Entry |--- PASID cache flush --+
'-------------' |
| | V
| | I/O page table pointer in GPA
'-------------'
Guest
------| Shadow |--------------------------|--------
v v v
Host
.-------------. .------------------------.
| pIOMMU | | FS for GIOVA->GPA |
| | '------------------------'
.----------------/ |
| PASID Entry | V (Nested xlate)
'----------------\.----------------------------------.
| | | SS for GPA->HPA, unmanaged domain|
| | '----------------------------------'
'-------------'
Where:
- FS = First stage page tables
- SS = Second stage page tables
<Intel VT-d Nested translation>
Different platform vendors have different first stage translation formats,
so userspace should query the underlying iommu capability before setting
first stage translation structures to host.[1]
In iommufd subsystem, I/O page tables would be tracked by hw_pagetable objects.
First stage page table is owned by userspace (guest), while second stage page
table is owned by kernel for security. So First stage page tables are tracked
by user-managed hw_pagetable, second stage page tables are tracked by kernel-
managed hw_pagetable.
This series first introduces new iommu op for allocating domains for iommufd,
and op for syncing iotlb for first stage page table modifications, and then
add the implementation of the new ops in intel-iommu driver. After this
preparation, adds kernel-managed and user-managed hw_pagetable allocation for
userspace. Last, add self-test for the new ioctls.
This series is based on "[PATCH 0/6] iommufd: Add iommu capability reporting"[1]
and Nicolin's "[PATCH v2 00/10] Add IO page table replacement support"[2]. Complete
code can be found in[3]. Draft Qemu code can be found in[4].
Basic test done with DSA device on VT-d. Where the guest has a vIOMMU built
with nested translation.
[1] https://lore.kernel.org/linux-iommu/20230209041642.9346-1-yi.l.liu@xxxxxxxxx/
[2] https://lore.kernel.org/linux-iommu/cover.1675802050.git.nicolinc@xxxxxxxxxx/
[3] https://github.com/yiliu1765/iommufd/tree/iommufd_nesting_vtd_v1
[4] https://github.com/yiliu1765/qemu/tree/wip/iommufd_rfcv3%2Bnesting
Regards,
Yi Liu
Lu Baolu (5):
iommu: Add new iommu op to create domains owned by userspace
iommu: Add nested domain support
iommu/vt-d: Extend dmar_domain to support nested domain
iommu/vt-d: Add helper to setup pasid nested translation
iommu/vt-d: Add nested domain support
Nicolin Chen (6):
iommufd: Add/del hwpt to IOAS at alloc/destroy()
iommufd/device: Move IOAS attaching and detaching operations into
helpers
iommufd/selftest: Add IOMMU_TEST_OP_MOCK_DOMAIN_REPLACE test op
iommufd/selftest: Add coverage for IOMMU_HWPT_ALLOC ioctl
iommufd/selftest: Add IOMMU_TEST_OP_MD_CHECK_IOTLB test op
iommufd/selftest: Add coverage for IOMMU_HWPT_INVALIDATE ioctl
Yi Liu (6):
iommufd/hw_pagetable: Use domain_alloc_user op for domain allocation
iommufd: Split iommufd_hw_pagetable_alloc()
iommufd: Add kernel-managed hw_pagetable allocation for userspace
iommufd: Add infrastructure for user-managed hw_pagetable allocation
iommufd: Add user-managed hw_pagetable allocation
iommufd/device: Report supported stage-1 page table types
drivers/iommu/intel/Makefile | 2 +-
drivers/iommu/intel/iommu.c | 38 ++-
drivers/iommu/intel/iommu.h | 50 +++-
drivers/iommu/intel/nested.c | 143 +++++++++
drivers/iommu/intel/pasid.c | 142 +++++++++
drivers/iommu/intel/pasid.h | 2 +
drivers/iommu/iommufd/device.c | 117 ++++----
drivers/iommu/iommufd/hw_pagetable.c | 280 +++++++++++++++++-
drivers/iommu/iommufd/iommufd_private.h | 23 +-
drivers/iommu/iommufd/iommufd_test.h | 35 +++
drivers/iommu/iommufd/main.c | 11 +
drivers/iommu/iommufd/selftest.c | 149 +++++++++-
include/linux/iommu.h | 11 +
include/uapi/linux/iommufd.h | 196 ++++++++++++
tools/testing/selftests/iommu/iommufd.c | 124 +++++++-
tools/testing/selftests/iommu/iommufd_utils.h | 106 +++++++
16 files changed, 1329 insertions(+), 100 deletions(-)
create mode 100644 drivers/iommu/intel/nested.c
--
2.34.1
