Pull VFIO updates from Alex Williamson: - VFIO direct character device (cdev) interface support. This extracts the vfio device fd from the container and group model, and is intended to be the native uAPI for use with IOMMUFD (Yi Liu) - Enhancements to the PCI hot reset interface in support of cdev usage (Yi Liu) - Fix a potential race between registering and unregistering vfio files in the kvm-vfio interface and extend use of a lock to avoid extra drop and acquires (Dmitry Torokhov) - A new vfio-pci variant driver for the AMD/Pensando Distributed Services Card (PDS) Ethernet device, supporting live migration (Brett Creeley) - Cleanups to remove redundant owner setup in cdx and fsl bus drivers, and simplify driver init/exit in fsl code (Li Zetao) - Fix uninitialized hole in data structure and pad capability structures for alignment (Stefan Hajnoczi) * tag 'vfio-v6.6-rc1' of https://github.com/awilliam/linux-vfio: (53 commits) vfio/pds: Send type for SUSPEND_STATUS command vfio/pds: fix return value in pds_vfio_get_lm_file() pds_core: Fix function header descriptions vfio: align capability structures vfio/type1: fix cap_migration information leak vfio/fsl-mc: Use module_fsl_mc_driver macro to simplify the code vfio/cdx: Remove redundant initialization owner in vfio_cdx_driver vfio/pds: Add Kconfig and documentation vfio/pds: Add support for firmware recovery vfio/pds: Add support for dirty page tracking vfio/pds: Add VFIO live migration support vfio/pds: register with the pds_core PF pds_core: Require callers of register/unregister to pass PF drvdata vfio/pds: Initial support for pds VFIO driver vfio: Commonize combine_ranges for use in other VFIO drivers kvm/vfio: avoid bouncing the mutex when adding and deleting groups kvm/vfio: ensure kvg instance stays around in kvm_vfio_group_add() docs: vfio: Add vfio device cdev description vfio: Compile vfio_group infrastructure optionally vfio: Move the IOMMU_CAP_CACHE_COHERENCY check in __vfio_register_dev() ...
Showing the changes to include/uapi/linux (2 of 55 files). Show all
Unified · Split
| @@ -1418,9 +1418,16 @@ struct kvm_device_attr { · 1417 unmodified lines | ||
| 1418 | 1418 | __u64 addr; /* userspace address of attr data */ |
| 1419 | 1419 | }; |
| 1420 | 1420 | |
| 1421 | −#define KVM_DEV_VFIO_GROUP 1 | |
| 1422 | −#define KVM_DEV_VFIO_GROUP_ADD 1 | |
| 1423 | −#define KVM_DEV_VFIO_GROUP_DEL 2 | |
| 1421 | +#define KVM_DEV_VFIO_FILE 1 | |
| 1422 | + | |
| 1423 | +#define KVM_DEV_VFIO_FILE_ADD 1 | |
| 1424 | +#define KVM_DEV_VFIO_FILE_DEL 2 | |
| 1425 | + | |
| 1426 | +/* KVM_DEV_VFIO_GROUP aliases are for compile time uapi compatibility */ | |
| 1427 | +#define KVM_DEV_VFIO_GROUP KVM_DEV_VFIO_FILE | |
| 1428 | + | |
| 1429 | +#define KVM_DEV_VFIO_GROUP_ADD KVM_DEV_VFIO_FILE_ADD | |
| 1430 | +#define KVM_DEV_VFIO_GROUP_DEL KVM_DEV_VFIO_FILE_DEL | |
| 1424 | 1431 | #define KVM_DEV_VFIO_GROUP_SET_SPAPR_TCE 3 |
| 1425 | 1432 | |
| 1426 | 1433 | enum kvm_device_type { |
| @@ -217,6 +217,7 @@ struct vfio_device_info { · 216 unmodified lines | ||
| 217 | 217 | __u32 num_regions; /* Max region index + 1 */ |
| 218 | 218 | __u32 num_irqs; /* Max IRQ index + 1 */ |
| 219 | 219 | __u32 cap_offset; /* Offset within info struct of first cap */ |
| 220 | + __u32 pad; | |
| 220 | 221 | }; |
| 221 | 222 | #define VFIO_DEVICE_GET_INFO _IO(VFIO_TYPE, VFIO_BASE + 7) |
| 222 | 223 | |
| @@ -677,11 +678,60 @@ enum { · 454 unmodified lines | ||
| 677 | 678 | * VFIO_DEVICE_GET_PCI_HOT_RESET_INFO - _IOWR(VFIO_TYPE, VFIO_BASE + 12, |
| 678 | 679 | * struct vfio_pci_hot_reset_info) |
| 679 | 680 | * |
| 681 | + * This command is used to query the affected devices in the hot reset for | |
| 682 | + * a given device. | |
| 683 | + * | |
| 684 | + * This command always reports the segment, bus, and devfn information for | |
| 685 | + * each affected device, and selectively reports the group_id or devid per | |
| 686 | + * the way how the calling device is opened. | |
| 687 | + * | |
| 688 | + * - If the calling device is opened via the traditional group/container | |
| 689 | + * API, group_id is reported. User should check if it has owned all | |
| 690 | + * the affected devices and provides a set of group fds to prove the | |
| 691 | + * ownership in VFIO_DEVICE_PCI_HOT_RESET ioctl. | |
| 692 | + * | |
| 693 | + * - If the calling device is opened as a cdev, devid is reported. | |
| 694 | + * Flag VFIO_PCI_HOT_RESET_FLAG_DEV_ID is set to indicate this | |
| 695 | + * data type. All the affected devices should be represented in | |
| 696 | + * the dev_set, ex. bound to a vfio driver, and also be owned by | |
| 697 | + * this interface which is determined by the following conditions: | |
| 698 | + * 1) Has a valid devid within the iommufd_ctx of the calling device. | |
| 699 | + * Ownership cannot be determined across separate iommufd_ctx and | |
| 700 | + * the cdev calling conventions do not support a proof-of-ownership | |
| 701 | + * model as provided in the legacy group interface. In this case | |
| 702 | + * valid devid with value greater than zero is provided in the return | |
| 703 | + * structure. | |
| 704 | + * 2) Does not have a valid devid within the iommufd_ctx of the calling | |
| 705 | + * device, but belongs to the same IOMMU group as the calling device | |
| 706 | + * or another opened device that has a valid devid within the | |
| 707 | + * iommufd_ctx of the calling device. This provides implicit ownership | |
| 708 | + * for devices within the same DMA isolation context. In this case | |
| 709 | + * the devid value of VFIO_PCI_DEVID_OWNED is provided in the return | |
| 710 | + * structure. | |
| 711 | + * | |
| 712 | + * A devid value of VFIO_PCI_DEVID_NOT_OWNED is provided in the return | |
| 713 | + * structure for affected devices where device is NOT represented in the | |
| 714 | + * dev_set or ownership is not available. Such devices prevent the use | |
| 715 | + * of VFIO_DEVICE_PCI_HOT_RESET ioctl outside of the proof-of-ownership | |
| 716 | + * calling conventions (ie. via legacy group accessed devices). Flag | |
| 717 | + * VFIO_PCI_HOT_RESET_FLAG_DEV_ID_OWNED would be set when all the | |
| 718 | + * affected devices are represented in the dev_set and also owned by | |
| 719 | + * the user. This flag is available only when | |
| 720 | + * flag VFIO_PCI_HOT_RESET_FLAG_DEV_ID is set, otherwise reserved. | |
| 721 | + * When set, user could invoke VFIO_DEVICE_PCI_HOT_RESET with a zero | |
| 722 | + * length fd array on the calling device as the ownership is validated | |
| 723 | + * by iommufd_ctx. | |
| 724 | + * | |
| 680 | 725 | * Return: 0 on success, -errno on failure: |
| 681 | 726 | * -enospc = insufficient buffer, -enodev = unsupported for device. |
| 682 | 727 | */ |
| 683 | 728 | struct vfio_pci_dependent_device { |
| 684 | − __u32 group_id; | |
| 729 | + union { | |
| 730 | + __u32 group_id; | |
| 731 | + __u32 devid; | |
| 732 | +#define VFIO_PCI_DEVID_OWNED 0 | |
| 733 | +#define VFIO_PCI_DEVID_NOT_OWNED -1 | |
| 734 | + }; | |
| 685 | 735 | __u16 segment; |
| 686 | 736 | __u8 bus; |
| 687 | 737 | __u8 devfn; /* Use PCI_SLOT/PCI_FUNC */ |
| @@ -690,6 +740,8 @@ struct vfio_pci_dependent_device { · 2 unmodified lines | ||
| 690 | 740 | struct vfio_pci_hot_reset_info { |
| 691 | 741 | __u32 argsz; |
| 692 | 742 | __u32 flags; |
| 743 | +#define VFIO_PCI_HOT_RESET_FLAG_DEV_ID (1 << 0) | |
| 744 | +#define VFIO_PCI_HOT_RESET_FLAG_DEV_ID_OWNED (1 << 1) | |
| 693 | 745 | __u32 count; |
| 694 | 746 | struct vfio_pci_dependent_device devices[]; |
| 695 | 747 | }; |
| @@ -700,6 +752,24 @@ struct vfio_pci_hot_reset_info { · 4 unmodified lines | ||
| 700 | 752 | * VFIO_DEVICE_PCI_HOT_RESET - _IOW(VFIO_TYPE, VFIO_BASE + 13, |
| 701 | 753 | * struct vfio_pci_hot_reset) |
| 702 | 754 | * |
| 755 | + * A PCI hot reset results in either a bus or slot reset which may affect | |
| 756 | + * other devices sharing the bus/slot. The calling user must have | |
| 757 | + * ownership of the full set of affected devices as determined by the | |
| 758 | + * VFIO_DEVICE_GET_PCI_HOT_RESET_INFO ioctl. | |
| 759 | + * | |
| 760 | + * When called on a device file descriptor acquired through the vfio | |
| 761 | + * group interface, the user is required to provide proof of ownership | |
| 762 | + * of those affected devices via the group_fds array in struct | |
| 763 | + * vfio_pci_hot_reset. | |
| 764 | + * | |
| 765 | + * When called on a direct cdev opened vfio device, the flags field of | |
| 766 | + * struct vfio_pci_hot_reset_info reports the ownership status of the | |
| 767 | + * affected devices and this ioctl must be called with an empty group_fds | |
| 768 | + * array. See above INFO ioctl definition for ownership requirements. | |
| 769 | + * | |
| 770 | + * Mixed usage of legacy groups and cdevs across the set of affected | |
| 771 | + * devices is not supported. | |
| 772 | + * | |
| 703 | 773 | * Return: 0 on success, -errno on failure. |
| 704 | 774 | */ |
| 705 | 775 | struct vfio_pci_hot_reset { |
| @@ -828,6 +898,77 @@ struct vfio_device_feature { · 122 unmodified lines | ||
| 828 | 898 | |
| 829 | 899 | #define VFIO_DEVICE_FEATURE _IO(VFIO_TYPE, VFIO_BASE + 17) |
| 830 | 900 | |
| 901 | +/* | |
| 902 | + * VFIO_DEVICE_BIND_IOMMUFD - _IOR(VFIO_TYPE, VFIO_BASE + 18, | |
| 903 | + * struct vfio_device_bind_iommufd) | |
| 904 | + * @argsz: User filled size of this data. | |
| 905 | + * @flags: Must be 0. | |
| 906 | + * @iommufd: iommufd to bind. | |
| 907 | + * @out_devid: The device id generated by this bind. devid is a handle for | |
| 908 | + * this device/iommufd bond and can be used in IOMMUFD commands. | |
| 909 | + * | |
| 910 | + * Bind a vfio_device to the specified iommufd. | |
| 911 | + * | |
| 912 | + * User is restricted from accessing the device before the binding operation | |
| 913 | + * is completed. Only allowed on cdev fds. | |
| 914 | + * | |
| 915 | + * Unbind is automatically conducted when device fd is closed. | |
| 916 | + * | |
| 917 | + * Return: 0 on success, -errno on failure. | |
| 918 | + */ | |
| 919 | +struct vfio_device_bind_iommufd { | |
| 920 | + __u32 argsz; | |
| 921 | + __u32 flags; | |
| 922 | + __s32 iommufd; | |
| 923 | + __u32 out_devid; | |
| 924 | +}; | |
| 925 | + | |
| 926 | +#define VFIO_DEVICE_BIND_IOMMUFD _IO(VFIO_TYPE, VFIO_BASE + 18) | |
| 927 | + | |
| 928 | +/* | |
| 929 | + * VFIO_DEVICE_ATTACH_IOMMUFD_PT - _IOW(VFIO_TYPE, VFIO_BASE + 19, | |
| 930 | + * struct vfio_device_attach_iommufd_pt) | |
| 931 | + * @argsz: User filled size of this data. | |
| 932 | + * @flags: Must be 0. | |
| 933 | + * @pt_id: Input the target id which can represent an ioas or a hwpt | |
| 934 | + * allocated via iommufd subsystem. | |
| 935 | + * Output the input ioas id or the attached hwpt id which could | |
| 936 | + * be the specified hwpt itself or a hwpt automatically created | |
| 937 | + * for the specified ioas by kernel during the attachment. | |
| 938 | + * | |
| 939 | + * Associate the device with an address space within the bound iommufd. | |
| 940 | + * Undo by VFIO_DEVICE_DETACH_IOMMUFD_PT or device fd close. This is only | |
| 941 | + * allowed on cdev fds. | |
| 942 | + * | |
| 943 | + * Return: 0 on success, -errno on failure. | |
| 944 | + */ | |
| 945 | +struct vfio_device_attach_iommufd_pt { | |
| 946 | + __u32 argsz; | |
| 947 | + __u32 flags; | |
| 948 | + __u32 pt_id; | |
| 949 | +}; | |
| 950 | + | |
| 951 | +#define VFIO_DEVICE_ATTACH_IOMMUFD_PT _IO(VFIO_TYPE, VFIO_BASE + 19) | |
| 952 | + | |
| 953 | +/* | |
| 954 | + * VFIO_DEVICE_DETACH_IOMMUFD_PT - _IOW(VFIO_TYPE, VFIO_BASE + 20, | |
| 955 | + * struct vfio_device_detach_iommufd_pt) | |
| 956 | + * @argsz: User filled size of this data. | |
| 957 | + * @flags: Must be 0. | |
| 958 | + * | |
| 959 | + * Remove the association of the device and its current associated address | |
| 960 | + * space. After it, the device should be in a blocking DMA state. This is only | |
| 961 | + * allowed on cdev fds. | |
| 962 | + * | |
| 963 | + * Return: 0 on success, -errno on failure. | |
| 964 | + */ | |
| 965 | +struct vfio_device_detach_iommufd_pt { | |
| 966 | + __u32 argsz; | |
| 967 | + __u32 flags; | |
| 968 | +}; | |
| 969 | + | |
| 970 | +#define VFIO_DEVICE_DETACH_IOMMUFD_PT _IO(VFIO_TYPE, VFIO_BASE + 20) | |
| 971 | + | |
| 831 | 972 | /* |
| 832 | 973 | * Provide support for setting a PCI VF Token, which is used as a shared |
| 833 | 974 | * secret between PF and VF drivers. This feature may only be set on a |
| @@ -1304,6 +1445,7 @@ struct vfio_iommu_type1_info { · 470 unmodified lines | ||
| 1304 | 1445 | #define VFIO_IOMMU_INFO_CAPS (1 << 1) /* Info supports caps */ |
| 1305 | 1446 | __u64 iova_pgsizes; /* Bitmap of supported page sizes */ |
| 1306 | 1447 | __u32 cap_offset; /* Offset within info struct of first cap */ |
| 1448 | + __u32 pad; | |
| 1307 | 1449 | }; |
| 1308 | 1450 | |
| 1309 | 1451 | /* |