Re: [RFC 00/18] vfio: Adopt iommufd

[Eric Auger <eric.auger@redhat.com>]

Hi Kevin,

On 4/18/22 10:49 AM, Tian, Kevin wrote:
>> From: Liu, Yi L <yi.l.liu@intel.com>
>> Sent: Thursday, April 14, 2022 6:47 PM
>>
>> With the introduction of iommufd[1], 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.
>> Other devices like vpda, vfio mdev and etc. are not considered yet.
> vfio mdev has no special support in Qemu. Just that it's not supported
> by iommufd yet thus can only be operated in legacy container interface at
> this point. Later once it's supported by the kernel suppose no additional
> enabling work is required for mdev in Qemu.
>
>> For vfio devices, the new interface is tied with device fd and iommufd
>> as the iommufd solution is device-centric. This is different from legacy
>> vfio which is group-centric. To support both interfaces in QEMU, this
>> series introduces the iommu backend concept in the form of different
>> container classes. The existing vfio container is named legacy container
>> (equivalent with legacy iommu backend in this series), while the new
>> iommufd based container is named as iommufd container (may also be
>> mentioned
>> as iommufd backend in this series). The two backend types have their own
>> way to setup secure context and dma management interface. 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   |
>>      +-------------+    +------------+      +---------------+
> last row: s/iommfd/iommufd/
>
> overall this sounds a reasonable abstraction. Later when vdpa starts
> supporting iommufd probably the iommufd BE will become even
> smaller with more logic shareable between vfio and vdpa.
>
>> [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]
>> - VFIOAddressSpace receives MemoryRegion add/del via MemoryListener
>> - VFIO populates DMA map/unmap via the container BEs
>>   *) iommufd BE: uses iommufd
>>   *) vfio legacy BE: uses container fd
>>
>> This series qomifies the VFIOContainer object which acts as a base class
> what does 'qomify' mean? I didn't find this word from dictionary...
sorry this is pure QEMU terminology. This stands for "QEMU Object Model"
additional info at:
https://qemu.readthedocs.io/en/latest/devel/qom.html

Eric
>
>> for a container. This base class is derived into the legacy VFIO container
>> and the new iommufd based container. The base class implements generic
>> code
>> such as code related to memory_listener and address space management
>> whereas
>> the derived class implements callbacks that depend on the kernel user space
> 'the kernel user space'?
>
>> being used.
>>
>> The selection of the backend is made on a device basis using the new
>> iommufd option (on/off/auto). By default the iommufd backend is selected
>> if supported by the host and by QEMU (iommufd KConfig). This option is
>> currently available only for the vfio-pci device. For other types of
>> devices, it does not yet exist and the legacy BE is chosen by default.
>>
>> Test done:
>> - PCI and Platform device were tested
> In this case PCI uses iommufd while platform device uses legacy?
>
>> - ccw and ap were only compile-tested
>> - limited device hotplug test
>> - vIOMMU test run for both legacy and iommufd backends (limited tests)
>>
>> This series was co-developed by Eric Auger and me based on the exploration
>> iommufd kernel[2], complete code of this series is available in[3]. As
>> iommufd kernel is in the early step (only iommufd generic interface is in
>> mailing list), so this series hasn't made the iommufd backend fully on par
>> with legacy backend w.r.t. features like p2p mappings, coherency tracking,
> what does 'coherency tracking' mean here? if related to iommu enforce
> snoop it is fully handled by the kernel so far. I didn't find any use of
> VFIO_DMA_CC_IOMMU in current Qemu.
>
>> live migration, etc. This series hasn't supported PCI devices without FLR
>> neither as the kernel doesn't support VFIO_DEVICE_PCI_HOT_RESET when
>> userspace
>> is using iommufd. The kernel needs to be updated to accept device fd list for
>> reset when userspace is using iommufd. Related work is in progress by
>> Jason[4].
>>
>> TODOs:
>> - Add DMA alias check for iommufd BE (group level)
>> - Make pci.c to be BE agnostic. Needs kernel change as well to fix the
>>   VFIO_DEVICE_PCI_HOT_RESET gap
>> - Cleanup the VFIODevice fields as it's used in both BEs
>> - Add locks
>> - Replace list with g_tree
>> - More tests
>>
>> Patch Overview:
>>
>> - Preparation:
>>   0001-scripts-update-linux-headers-Add-iommufd.h.patch
>>   0002-linux-headers-Import-latest-vfio.h-and-iommufd.h.patch
>>   0003-hw-vfio-pci-fix-vfio_pci_hot_reset_result-trace-poin.patch
>>   0004-vfio-pci-Use-vbasedev-local-variable-in-vfio_realize.patch
>>   0005-vfio-common-Rename-VFIOGuestIOMMU-iommu-into-
>> iommu_m.patch
> 3-5 are pure cleanups which could be sent out separately 
>
>>   0006-vfio-common-Split-common.c-into-common.c-container.c.patch
>>
>> - Introduce container object and covert existing vfio to use it:
>>   0007-vfio-Add-base-object-for-VFIOContainer.patch
>>   0008-vfio-container-Introduce-vfio_attach-detach_device.patch
>>   0009-vfio-platform-Use-vfio_-attach-detach-_device.patch
>>   0010-vfio-ap-Use-vfio_-attach-detach-_device.patch
>>   0011-vfio-ccw-Use-vfio_-attach-detach-_device.patch
>>   0012-vfio-container-obj-Introduce-attach-detach-_device-c.patch
>>   0013-vfio-container-obj-Introduce-VFIOContainer-reset-cal.patch
>>
>> - Introduce iommufd based container:
>>   0014-hw-iommufd-Creation.patch
>>   0015-vfio-iommufd-Implement-iommufd-backend.patch
>>   0016-vfio-iommufd-Add-IOAS_COPY_DMA-support.patch
>>
>> - Add backend selection for vfio-pci:
>>   0017-vfio-as-Allow-the-selection-of-a-given-iommu-backend.patch
>>   0018-vfio-pci-Add-an-iommufd-option.patch
>>
>> [1] https://lore.kernel.org/kvm/0-v1-e79cd8d168e8+6-
>> iommufd_jgg@nvidia.com/
>> [2] https://github.com/luxis1999/iommufd/tree/iommufd-v5.17-rc6
>> [3] https://github.com/luxis1999/qemu/tree/qemu-for-5.17-rc6-vm-rfcv1
>> [4] https://lore.kernel.org/kvm/0-v1-a8faf768d202+125dd-
>> vfio_mdev_no_group_jgg@nvidia.com/
> Following is probably more relevant to [4]:
>
> https://lore.kernel.org/all/10-v1-33906a626da1+16b0-vfio_kvm_no_group_jgg@nvidia.com/
>
> Thanks
> Kevin
>