Hi,
Thanks all for giving guides and comments on previous series, this is
the remaining part of the iommufd support.
Based on Cédric's suggestion, replace old config method for IOMMUFD
with Kconfig.
This series is added some extra changes based on Cédric's changes
at https://github.com/legoater/qemu/commits/vfio-8.2
Based on Jason's suggestion, drop the implementation of manually
allocating hwpt and switch to IOAS attach/detach.
Beside current test, we also tested mdev with mtty for better cover range.
PATCH 1: Introduce iommufd object
PATCH 2-8: add IOMMUFD container and cdev support
PATCH 9-16: fd passing for cdev and linking to IOMMUFD
PATCH 17: make VFIOContainerBase parameter const
PATCH 18-20: Compile out for IOMMUFD for arm, s390x and x86
We have done wide test with different combinations, e.g:
- PCI device were tested
- FD passing and hot reset with some trick.
- device hotplug test with legacy and iommufd backends
- with or without vIOMMU for legacy and iommufd backends
- divices linked to different iommufds
- VFIO migration with a E800 net card(no dirty sync support) passthrough
- platform, ccw and ap were only compile-tested due to environment limit
- test mdev pass through with mtty and mix with real device and different BE
Given some iommufd kernel limitations, the iommufd backend is
not yet fully on par with the legacy backend w.r.t. features like:
- p2p mappings (you will see related error traces)
- dirty page sync
- and etc.
qemu code: https://github.com/yiliu1765/qemu/commits/zhenzhong/iommufd_cdev_v5
Based on vfio-next, commit id: c67cd7fe8c
--------------------------------------------------------------------------
Below are some background and graph about the design:
With the introduction of iommufd, the Linux kernel provides a generic
interface for userspace drivers to propagate their DMA mappings to kernel
for assigned devices. This series does the porting of the VFIO devices
onto the /dev/iommu uapi and let it coexist with the legacy implementation.
At QEMU level, interactions with the /dev/iommu are abstracted by a new
iommufd object (compiled in with the CONFIG_IOMMUFD option).
Any QEMU device (e.g. vfio device) wishing to use /dev/iommu must be
linked with an iommufd object. In this series, the vfio-pci device is
granted with such capability (other VFIO devices are not yet ready):
It gets a new optional parameter named iommufd which allows to pass
an iommufd object:
-object iommufd,id=iommufd0
-device vfio-pci,host=0000:02:00.0,iommufd=iommufd0
Note the /dev/iommu and vfio cdev can be externally opened by a
management layer. In such a case the fd is passed:
-object iommufd,id=iommufd0,fd=22
-device vfio-pci,iommufd=iommufd0,fd=23
If the fd parameter is not passed, the fd is opened by QEMU.
See https://www.mail-archive.com/qemu-devel@nongnu.org/msg937155.html
for detailed discuss on this requirement.
If no iommufd option is passed to the vfio-pci device, iommufd is not
used and the end-user gets the behavior based on the legacy vfio iommu
interfaces:
-device vfio-pci,host=0000:02:00.0
While the legacy kernel interface is group-centric, the new iommufd
interface is device-centric, relying on device fd and iommufd.
To support both interfaces in the QEMU VFIO device we reworked the vfio
container abstraction so that the generic VFIO code can use either
backend.
The VFIOContainer object becomes a base object derived into
a) the legacy VFIO container and
b) the new iommufd based container.
The base object implements generic code such as code related to
memory_listener and address space management whereas the derived
objects implement callbacks specific to either BE, legacy and
iommufd. Indeed each backend has its own way to setup secure context
and dma management interface. The below diagram shows how it looks
like with both BEs.
VFIO AddressSpace/Memory
+-------+ +----------+ +-----+ +-----+
| pci | | platform | | ap | | ccw |
+---+---+ +----+-----+ +--+--+ +--+--+ +----------------------+
| | | | | AddressSpace |
| | | | +------------+---------+
+---V-----------V-----------V--------V----+ /
| VFIOAddressSpace | <------------+
| | | MemoryListener
| VFIOContainer list |
+-------+----------------------------+----+
| |
| |
+-------V------+ +--------V----------+
| iommufd | | vfio legacy |
| container | | container |
+-------+------+ +--------+----------+
| |
| /dev/iommu | /dev/vfio/vfio
| /dev/vfio/devices/vfioX | /dev/vfio/$group_id
Userspace | |
============+============================+===========================
Kernel | device fd |
+---------------+ | group/container fd
| (BIND_IOMMUFD | | (SET_CONTAINER/SET_IOMMU)
| ATTACH_IOAS) | | device fd
| | |
| +-------V------------V-----------------+
iommufd | | vfio |
(map/unmap | +---------+--------------------+-------+
ioas_copy) | | | map/unmap
| | |
+------V------+ +-----V------+ +------V--------+
| iommfd core | | device | | vfio iommu |
+-------------+ +------------+ +---------------+
[Secure Context setup]
- iommufd BE: uses device fd and iommufd to setup secure context
(bind_iommufd, attach_ioas)
- vfio legacy BE: uses group fd and container fd to setup secure context
(set_container, set_iommu)
[Device access]
- iommufd BE: device fd is opened through /dev/vfio/devices/vfioX
- vfio legacy BE: device fd is retrieved from group fd ioctl
[DMA Mapping flow]
1. VFIOAddressSpace receives MemoryRegion add/del via MemoryListener
2. VFIO populates DMA map/unmap via the container BEs
*) iommufd BE: uses iommufd
*) vfio legacy BE: uses container fd
Changelog:
v5:
- Change to use Kconfig for CONFIG_IOMMUFD and drop stub file (Cédric)
- Add (uintptr_t) to info->allowed_iovas (Cédric)
- Switch to IOAS attach/detach and hide hwpt (Jason)
- move chardev_open.[h|c] under the IOMMUFD entry (Cédric)
- Move vfio_legacy_pci_hot_reset into container.c (Cédric)
- Add missed pgsizes initialization in vfio_get_info_iova_range
- split linking iommufd patch into three to be cleaner
- Fix comments on PCI BAR unmap
v4:
- add CONFIG_IOMMUFD check for IOMMUFDProperties (Markus)
- add doc for default case without fd (Markus)
- Fix build issue reported by Markus and Cédric
- Simply use SPDX identifier in new file (Cédric)
- make vfio_container_init/destroy helper a seperate patch (Cédric)
- make vrdl_list movement a seperate patch (Cédric)
- add const for some callback parameters (Cédric)
- add g_assert in VFIOIOMMUOps callback (Cédric)
- introduce pci_hot_reset callback (Cédric)
- remove VFIOIOMMUSpaprOps (Cédric)
- initialize g_autofree to NULL (Cédric)
- adjust func name prefix and trace event in iommufd.c (Cédric)
- add RB
v3:
- Rename base container as VFIOContainerBase and legacy container as container (Cédric)
- Drop VFIO_IOMMU_BACKEND_OPS class and use struct instead (Cédric)
- Cleanup container.c by introducing spapr backend and move spapr code out (Cédric)
- Introduce vfio_iommu_spapr_ops (Cédric)
- Add doc of iommufd in qom.json and have iommufd member sorted (Markus)
- patch19 and patch21 are splitted to two parts to facilitate review
v2:
- patch "vfio: Add base container" in v1 is split into patch1-15 per Cédric
- add fd passing to platform/ap/ccw vfio device
- add (uintptr_t) cast in iommufd_backend_map_dma() per Cédric
- rename char_dev.h to chardev_open.h for same naming scheme per Daniel
- add full copyright per Daniel and Jason
Note changelog below are from full IOMMUFD series:
v1:
- Alloc hwpt instead of using auto hwpt
- elaborate iommufd code per Nicolin
- consolidate two patches and drop as.c
- typo error fix and function rename
rfcv4:
- rebase on top of v8.0.3
- Add one patch from Yi which is about vfio device add in kvm
- Remove IOAS_COPY optimization and focus on functions in this patchset
- Fix wrong name issue reported and fix suggested by Matthew
- Fix compilation issue reported and fix sugggsted by Nicolin
- Use query_dirty_bitmap callback to replace get_dirty_bitmap for better
granularity
- Add dev_iter_next() callback to avoid adding so many callback
at container scope, add VFIODevice.hwpt to support that
- Restore all functions back to common from container whenever possible,
mainly migration and reset related functions
- Add --enable/disable-iommufd config option, enabled by default in linux
- Remove VFIODevice.hwpt_next as it's redundant with VFIODevice.next
- Adapt new VFIO_DEVICE_PCI_HOT_RESET uAPI for IOMMUFD backed device
- vfio_kvm_device_add/del_group call vfio_kvm_device_add/del_fd to remove
redundant code
- Add FD passing support for vfio device backed by IOMMUFD
- Fix hot unplug resource leak issue in vfio_legacy_detach_device()
- Fix FD leak in vfio_get_devicefd()
rfcv3:
- rebase on top of v7.2.0
- Fix the compilation with CONFIG_IOMMUFD unset by using true classes for
VFIO backends
- Fix use after free in error path, reported by Alister
- Split common.c in several steps to ease the review
rfcv2:
- remove the first three patches of rfcv1
- add open cdev helper suggested by Jason
- remove the QOMification of the VFIOContainer and simply use standard ops
(David)
- add "-object iommufd" suggested by Alex
Thanks
Zhenzhong
Cédric Le Goater (3):
hw/arm: Activate IOMMUFD for virt machines
kconfig: Activate IOMMUFD for s390x machines
hw/i386: Activate IOMMUFD for q35 machines
Eric Auger (2):
backends/iommufd: Introduce the iommufd object
vfio/pci: Allow the selection of a given iommu backend
Yi Liu (2):
util/char_dev: Add open_cdev()
vfio/iommufd: Implement the iommufd backend
Zhenzhong Duan (13):
vfio/iommufd: Relax assert check for iommufd backend
vfio/iommufd: Add support for iova_ranges and pgsizes
vfio/pci: Extract out a helper vfio_pci_get_pci_hot_reset_info
vfio/pci: Introduce a vfio pci hot reset interface
vfio/iommufd: Enable pci hot reset through iommufd cdev interface
vfio/pci: Make vfio cdev pre-openable by passing a file handle
vfio/platform: Allow the selection of a given iommu backend
vfio/platform: Make vfio cdev pre-openable by passing a file handle
vfio/ap: Allow the selection of a given iommu backend
vfio/ap: Make vfio cdev pre-openable by passing a file handle
vfio/ccw: Allow the selection of a given iommu backend
vfio/ccw: Make vfio cdev pre-openable by passing a file handle
vfio: Make VFIOContainerBase poiner parameter const in VFIOIOMMUOps
callbacks
MAINTAINERS | 10 +
qapi/qom.json | 19 +
hw/vfio/pci.h | 5 +
include/hw/vfio/vfio-common.h | 24 +-
include/hw/vfio/vfio-container-base.h | 15 +-
include/hw/vfio/vfio-platform.h | 1 +
include/qemu/chardev_open.h | 16 +
include/sysemu/iommufd.h | 46 ++
backends/iommufd.c | 264 +++++++++++
hw/vfio/ap.c | 38 +-
hw/vfio/ccw.c | 40 +-
hw/vfio/common.c | 29 +-
hw/vfio/container-base.c | 6 +-
hw/vfio/container.c | 215 ++++++++-
hw/vfio/helpers.c | 34 ++
hw/vfio/iommufd.c | 623 ++++++++++++++++++++++++++
hw/vfio/pci.c | 232 +++-------
hw/vfio/platform.c | 46 +-
util/chardev_open.c | 81 ++++
backends/Kconfig | 4 +
backends/meson.build | 1 +
backends/trace-events | 13 +
hw/arm/Kconfig | 1 +
hw/i386/Kconfig | 1 +
hw/s390x/Kconfig | 1 +
hw/vfio/meson.build | 3 +
hw/vfio/trace-events | 11 +
qemu-options.hx | 12 +
util/meson.build | 1 +
29 files changed, 1563 insertions(+), 229 deletions(-)
create mode 100644 include/qemu/chardev_open.h
create mode 100644 include/sysemu/iommufd.h
create mode 100644 backends/iommufd.c
create mode 100644 hw/vfio/iommufd.c
create mode 100644 util/chardev_open.c
--
2.34.1