1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2 /* 3 * VFIO API definition 4 * 5 * Copyright (C) 2012 Red Hat, Inc. All rights reserved. 6 * Author: Alex Williamson <alex.williamson@redhat.com> 7 * 8 * This program is free software; you can redistribute it and/or modify 9 * it under the terms of the GNU General Public License version 2 as 10 * published by the Free Software Foundation. 11 */ 12 #ifndef VFIO_H 13 #define VFIO_H 14 15 #include <linux/types.h> 16 #include <linux/ioctl.h> 17 18 #define VFIO_API_VERSION 0 19 20 21 /* Kernel & User level defines for VFIO IOCTLs. */ 22 23 /* Extensions */ 24 25 #define VFIO_TYPE1_IOMMU 1 26 #define VFIO_SPAPR_TCE_IOMMU 2 27 #define VFIO_TYPE1v2_IOMMU 3 28 /* 29 * IOMMU enforces DMA cache coherence (ex. PCIe NoSnoop stripping). This 30 * capability is subject to change as groups are added or removed. 31 */ 32 #define VFIO_DMA_CC_IOMMU 4 33 34 /* Check if EEH is supported */ 35 #define VFIO_EEH 5 36 37 /* Two-stage IOMMU */ 38 #define VFIO_TYPE1_NESTING_IOMMU 6 /* Implies v2 */ 39 40 #define VFIO_SPAPR_TCE_v2_IOMMU 7 41 42 /* 43 * The No-IOMMU IOMMU offers no translation or isolation for devices and 44 * supports no ioctls outside of VFIO_CHECK_EXTENSION. Use of VFIO's No-IOMMU 45 * code will taint the host kernel and should be used with extreme caution. 46 */ 47 #define VFIO_NOIOMMU_IOMMU 8 48 49 /* Supports VFIO_DMA_UNMAP_FLAG_ALL */ 50 #define VFIO_UNMAP_ALL 9 51 52 /* Supports the vaddr flag for DMA map and unmap */ 53 #define VFIO_UPDATE_VADDR 10 54 55 /* 56 * The IOCTL interface is designed for extensibility by embedding the 57 * structure length (argsz) and flags into structures passed between 58 * kernel and userspace. We therefore use the _IO() macro for these 59 * defines to avoid implicitly embedding a size into the ioctl request. 60 * As structure fields are added, argsz will increase to match and flag 61 * bits will be defined to indicate additional fields with valid data. 62 * It's *always* the caller's responsibility to indicate the size of 63 * the structure passed by setting argsz appropriately. 64 */ 65 66 #define VFIO_TYPE (';') 67 #define VFIO_BASE 100 68 69 /* 70 * For extension of INFO ioctls, VFIO makes use of a capability chain 71 * designed after PCI/e capabilities. A flag bit indicates whether 72 * this capability chain is supported and a field defined in the fixed 73 * structure defines the offset of the first capability in the chain. 74 * This field is only valid when the corresponding bit in the flags 75 * bitmap is set. This offset field is relative to the start of the 76 * INFO buffer, as is the next field within each capability header. 77 * The id within the header is a shared address space per INFO ioctl, 78 * while the version field is specific to the capability id. The 79 * contents following the header are specific to the capability id. 80 */ 81 struct vfio_info_cap_header { 82 __u16 id; /* Identifies capability */ 83 __u16 version; /* Version specific to the capability ID */ 84 __u32 next; /* Offset of next capability */ 85 }; 86 87 /* 88 * Callers of INFO ioctls passing insufficiently sized buffers will see 89 * the capability chain flag bit set, a zero value for the first capability 90 * offset (if available within the provided argsz), and argsz will be 91 * updated to report the necessary buffer size. For compatibility, the 92 * INFO ioctl will not report error in this case, but the capability chain 93 * will not be available. 94 */ 95 96 /* -------- IOCTLs for VFIO file descriptor (/dev/vfio/vfio) -------- */ 97 98 /** 99 * VFIO_GET_API_VERSION - _IO(VFIO_TYPE, VFIO_BASE + 0) 100 * 101 * Report the version of the VFIO API. This allows us to bump the entire 102 * API version should we later need to add or change features in incompatible 103 * ways. 104 * Return: VFIO_API_VERSION 105 * Availability: Always 106 */ 107 #define VFIO_GET_API_VERSION _IO(VFIO_TYPE, VFIO_BASE + 0) 108 109 /** 110 * VFIO_CHECK_EXTENSION - _IOW(VFIO_TYPE, VFIO_BASE + 1, __u32) 111 * 112 * Check whether an extension is supported. 113 * Return: 0 if not supported, 1 (or some other positive integer) if supported. 114 * Availability: Always 115 */ 116 #define VFIO_CHECK_EXTENSION _IO(VFIO_TYPE, VFIO_BASE + 1) 117 118 /** 119 * VFIO_SET_IOMMU - _IOW(VFIO_TYPE, VFIO_BASE + 2, __s32) 120 * 121 * Set the iommu to the given type. The type must be supported by an 122 * iommu driver as verified by calling CHECK_EXTENSION using the same 123 * type. A group must be set to this file descriptor before this 124 * ioctl is available. The IOMMU interfaces enabled by this call are 125 * specific to the value set. 126 * Return: 0 on success, -errno on failure 127 * Availability: When VFIO group attached 128 */ 129 #define VFIO_SET_IOMMU _IO(VFIO_TYPE, VFIO_BASE + 2) 130 131 /* -------- IOCTLs for GROUP file descriptors (/dev/vfio/$GROUP) -------- */ 132 133 /** 134 * VFIO_GROUP_GET_STATUS - _IOR(VFIO_TYPE, VFIO_BASE + 3, 135 * struct vfio_group_status) 136 * 137 * Retrieve information about the group. Fills in provided 138 * struct vfio_group_info. Caller sets argsz. 139 * Return: 0 on succes, -errno on failure. 140 * Availability: Always 141 */ 142 struct vfio_group_status { 143 __u32 argsz; 144 __u32 flags; 145 #define VFIO_GROUP_FLAGS_VIABLE (1 << 0) 146 #define VFIO_GROUP_FLAGS_CONTAINER_SET (1 << 1) 147 }; 148 #define VFIO_GROUP_GET_STATUS _IO(VFIO_TYPE, VFIO_BASE + 3) 149 150 /** 151 * VFIO_GROUP_SET_CONTAINER - _IOW(VFIO_TYPE, VFIO_BASE + 4, __s32) 152 * 153 * Set the container for the VFIO group to the open VFIO file 154 * descriptor provided. Groups may only belong to a single 155 * container. Containers may, at their discretion, support multiple 156 * groups. Only when a container is set are all of the interfaces 157 * of the VFIO file descriptor and the VFIO group file descriptor 158 * available to the user. 159 * Return: 0 on success, -errno on failure. 160 * Availability: Always 161 */ 162 #define VFIO_GROUP_SET_CONTAINER _IO(VFIO_TYPE, VFIO_BASE + 4) 163 164 /** 165 * VFIO_GROUP_UNSET_CONTAINER - _IO(VFIO_TYPE, VFIO_BASE + 5) 166 * 167 * Remove the group from the attached container. This is the 168 * opposite of the SET_CONTAINER call and returns the group to 169 * an initial state. All device file descriptors must be released 170 * prior to calling this interface. When removing the last group 171 * from a container, the IOMMU will be disabled and all state lost, 172 * effectively also returning the VFIO file descriptor to an initial 173 * state. 174 * Return: 0 on success, -errno on failure. 175 * Availability: When attached to container 176 */ 177 #define VFIO_GROUP_UNSET_CONTAINER _IO(VFIO_TYPE, VFIO_BASE + 5) 178 179 /** 180 * VFIO_GROUP_GET_DEVICE_FD - _IOW(VFIO_TYPE, VFIO_BASE + 6, char) 181 * 182 * Return a new file descriptor for the device object described by 183 * the provided string. The string should match a device listed in 184 * the devices subdirectory of the IOMMU group sysfs entry. The 185 * group containing the device must already be added to this context. 186 * Return: new file descriptor on success, -errno on failure. 187 * Availability: When attached to container 188 */ 189 #define VFIO_GROUP_GET_DEVICE_FD _IO(VFIO_TYPE, VFIO_BASE + 6) 190 191 /* --------------- IOCTLs for DEVICE file descriptors --------------- */ 192 193 /** 194 * VFIO_DEVICE_GET_INFO - _IOR(VFIO_TYPE, VFIO_BASE + 7, 195 * struct vfio_device_info) 196 * 197 * Retrieve information about the device. Fills in provided 198 * struct vfio_device_info. Caller sets argsz. 199 * Return: 0 on success, -errno on failure. 200 */ 201 struct vfio_device_info { 202 __u32 argsz; 203 __u32 flags; 204 #define VFIO_DEVICE_FLAGS_RESET (1 << 0) /* Device supports reset */ 205 #define VFIO_DEVICE_FLAGS_PCI (1 << 1) /* vfio-pci device */ 206 #define VFIO_DEVICE_FLAGS_PLATFORM (1 << 2) /* vfio-platform device */ 207 #define VFIO_DEVICE_FLAGS_AMBA (1 << 3) /* vfio-amba device */ 208 #define VFIO_DEVICE_FLAGS_CCW (1 << 4) /* vfio-ccw device */ 209 #define VFIO_DEVICE_FLAGS_AP (1 << 5) /* vfio-ap device */ 210 #define VFIO_DEVICE_FLAGS_FSL_MC (1 << 6) /* vfio-fsl-mc device */ 211 #define VFIO_DEVICE_FLAGS_CAPS (1 << 7) /* Info supports caps */ 212 __u32 num_regions; /* Max region index + 1 */ 213 __u32 num_irqs; /* Max IRQ index + 1 */ 214 __u32 cap_offset; /* Offset within info struct of first cap */ 215 }; 216 #define VFIO_DEVICE_GET_INFO _IO(VFIO_TYPE, VFIO_BASE + 7) 217 218 /* 219 * Vendor driver using Mediated device framework should provide device_api 220 * attribute in supported type attribute groups. Device API string should be one 221 * of the following corresponding to device flags in vfio_device_info structure. 222 */ 223 224 #define VFIO_DEVICE_API_PCI_STRING "vfio-pci" 225 #define VFIO_DEVICE_API_PLATFORM_STRING "vfio-platform" 226 #define VFIO_DEVICE_API_AMBA_STRING "vfio-amba" 227 #define VFIO_DEVICE_API_CCW_STRING "vfio-ccw" 228 #define VFIO_DEVICE_API_AP_STRING "vfio-ap" 229 230 /* 231 * The following capabilities are unique to s390 zPCI devices. Their contents 232 * are further-defined in vfio_zdev.h 233 */ 234 #define VFIO_DEVICE_INFO_CAP_ZPCI_BASE 1 235 #define VFIO_DEVICE_INFO_CAP_ZPCI_GROUP 2 236 #define VFIO_DEVICE_INFO_CAP_ZPCI_UTIL 3 237 #define VFIO_DEVICE_INFO_CAP_ZPCI_PFIP 4 238 239 /** 240 * VFIO_DEVICE_GET_REGION_INFO - _IOWR(VFIO_TYPE, VFIO_BASE + 8, 241 * struct vfio_region_info) 242 * 243 * Retrieve information about a device region. Caller provides 244 * struct vfio_region_info with index value set. Caller sets argsz. 245 * Implementation of region mapping is bus driver specific. This is 246 * intended to describe MMIO, I/O port, as well as bus specific 247 * regions (ex. PCI config space). Zero sized regions may be used 248 * to describe unimplemented regions (ex. unimplemented PCI BARs). 249 * Return: 0 on success, -errno on failure. 250 */ 251 struct vfio_region_info { 252 __u32 argsz; 253 __u32 flags; 254 #define VFIO_REGION_INFO_FLAG_READ (1 << 0) /* Region supports read */ 255 #define VFIO_REGION_INFO_FLAG_WRITE (1 << 1) /* Region supports write */ 256 #define VFIO_REGION_INFO_FLAG_MMAP (1 << 2) /* Region supports mmap */ 257 #define VFIO_REGION_INFO_FLAG_CAPS (1 << 3) /* Info supports caps */ 258 __u32 index; /* Region index */ 259 __u32 cap_offset; /* Offset within info struct of first cap */ 260 __u64 size; /* Region size (bytes) */ 261 __u64 offset; /* Region offset from start of device fd */ 262 }; 263 #define VFIO_DEVICE_GET_REGION_INFO _IO(VFIO_TYPE, VFIO_BASE + 8) 264 265 /* 266 * The sparse mmap capability allows finer granularity of specifying areas 267 * within a region with mmap support. When specified, the user should only 268 * mmap the offset ranges specified by the areas array. mmaps outside of the 269 * areas specified may fail (such as the range covering a PCI MSI-X table) or 270 * may result in improper device behavior. 271 * 272 * The structures below define version 1 of this capability. 273 */ 274 #define VFIO_REGION_INFO_CAP_SPARSE_MMAP 1 275 276 struct vfio_region_sparse_mmap_area { 277 __u64 offset; /* Offset of mmap'able area within region */ 278 __u64 size; /* Size of mmap'able area */ 279 }; 280 281 struct vfio_region_info_cap_sparse_mmap { 282 struct vfio_info_cap_header header; 283 __u32 nr_areas; 284 __u32 reserved; 285 struct vfio_region_sparse_mmap_area areas[]; 286 }; 287 288 /* 289 * The device specific type capability allows regions unique to a specific 290 * device or class of devices to be exposed. This helps solve the problem for 291 * vfio bus drivers of defining which region indexes correspond to which region 292 * on the device, without needing to resort to static indexes, as done by 293 * vfio-pci. For instance, if we were to go back in time, we might remove 294 * VFIO_PCI_VGA_REGION_INDEX and let vfio-pci simply define that all indexes 295 * greater than or equal to VFIO_PCI_NUM_REGIONS are device specific and we'd 296 * make a "VGA" device specific type to describe the VGA access space. This 297 * means that non-VGA devices wouldn't need to waste this index, and thus the 298 * address space associated with it due to implementation of device file 299 * descriptor offsets in vfio-pci. 300 * 301 * The current implementation is now part of the user ABI, so we can't use this 302 * for VGA, but there are other upcoming use cases, such as opregions for Intel 303 * IGD devices and framebuffers for vGPU devices. We missed VGA, but we'll 304 * use this for future additions. 305 * 306 * The structure below defines version 1 of this capability. 307 */ 308 #define VFIO_REGION_INFO_CAP_TYPE 2 309 310 struct vfio_region_info_cap_type { 311 struct vfio_info_cap_header header; 312 __u32 type; /* global per bus driver */ 313 __u32 subtype; /* type specific */ 314 }; 315 316 /* 317 * List of region types, global per bus driver. 318 * If you introduce a new type, please add it here. 319 */ 320 321 /* PCI region type containing a PCI vendor part */ 322 #define VFIO_REGION_TYPE_PCI_VENDOR_TYPE (1 << 31) 323 #define VFIO_REGION_TYPE_PCI_VENDOR_MASK (0xffff) 324 #define VFIO_REGION_TYPE_GFX (1) 325 #define VFIO_REGION_TYPE_CCW (2) 326 #define VFIO_REGION_TYPE_MIGRATION (3) 327 328 /* sub-types for VFIO_REGION_TYPE_PCI_* */ 329 330 /* 8086 vendor PCI sub-types */ 331 #define VFIO_REGION_SUBTYPE_INTEL_IGD_OPREGION (1) 332 #define VFIO_REGION_SUBTYPE_INTEL_IGD_HOST_CFG (2) 333 #define VFIO_REGION_SUBTYPE_INTEL_IGD_LPC_CFG (3) 334 335 /* 10de vendor PCI sub-types */ 336 /* 337 * NVIDIA GPU NVlink2 RAM is coherent RAM mapped onto the host address space. 338 * 339 * Deprecated, region no longer provided 340 */ 341 #define VFIO_REGION_SUBTYPE_NVIDIA_NVLINK2_RAM (1) 342 343 /* 1014 vendor PCI sub-types */ 344 /* 345 * IBM NPU NVlink2 ATSD (Address Translation Shootdown) register of NPU 346 * to do TLB invalidation on a GPU. 347 * 348 * Deprecated, region no longer provided 349 */ 350 #define VFIO_REGION_SUBTYPE_IBM_NVLINK2_ATSD (1) 351 352 /* sub-types for VFIO_REGION_TYPE_GFX */ 353 #define VFIO_REGION_SUBTYPE_GFX_EDID (1) 354 355 /** 356 * struct vfio_region_gfx_edid - EDID region layout. 357 * 358 * Set display link state and EDID blob. 359 * 360 * The EDID blob has monitor information such as brand, name, serial 361 * number, physical size, supported video modes and more. 362 * 363 * This special region allows userspace (typically qemu) set a virtual 364 * EDID for the virtual monitor, which allows a flexible display 365 * configuration. 366 * 367 * For the edid blob spec look here: 368 * https://en.wikipedia.org/wiki/Extended_Display_Identification_Data 369 * 370 * On linux systems you can find the EDID blob in sysfs: 371 * /sys/class/drm/${card}/${connector}/edid 372 * 373 * You can use the edid-decode ulility (comes with xorg-x11-utils) to 374 * decode the EDID blob. 375 * 376 * @edid_offset: location of the edid blob, relative to the 377 * start of the region (readonly). 378 * @edid_max_size: max size of the edid blob (readonly). 379 * @edid_size: actual edid size (read/write). 380 * @link_state: display link state (read/write). 381 * VFIO_DEVICE_GFX_LINK_STATE_UP: Monitor is turned on. 382 * VFIO_DEVICE_GFX_LINK_STATE_DOWN: Monitor is turned off. 383 * @max_xres: max display width (0 == no limitation, readonly). 384 * @max_yres: max display height (0 == no limitation, readonly). 385 * 386 * EDID update protocol: 387 * (1) set link-state to down. 388 * (2) update edid blob and size. 389 * (3) set link-state to up. 390 */ 391 struct vfio_region_gfx_edid { 392 __u32 edid_offset; 393 __u32 edid_max_size; 394 __u32 edid_size; 395 __u32 max_xres; 396 __u32 max_yres; 397 __u32 link_state; 398 #define VFIO_DEVICE_GFX_LINK_STATE_UP 1 399 #define VFIO_DEVICE_GFX_LINK_STATE_DOWN 2 400 }; 401 402 /* sub-types for VFIO_REGION_TYPE_CCW */ 403 #define VFIO_REGION_SUBTYPE_CCW_ASYNC_CMD (1) 404 #define VFIO_REGION_SUBTYPE_CCW_SCHIB (2) 405 #define VFIO_REGION_SUBTYPE_CCW_CRW (3) 406 407 /* sub-types for VFIO_REGION_TYPE_MIGRATION */ 408 #define VFIO_REGION_SUBTYPE_MIGRATION (1) 409 410 /* 411 * The structure vfio_device_migration_info is placed at the 0th offset of 412 * the VFIO_REGION_SUBTYPE_MIGRATION region to get and set VFIO device related 413 * migration information. Field accesses from this structure are only supported 414 * at their native width and alignment. Otherwise, the result is undefined and 415 * vendor drivers should return an error. 416 * 417 * device_state: (read/write) 418 * - The user application writes to this field to inform the vendor driver 419 * about the device state to be transitioned to. 420 * - The vendor driver should take the necessary actions to change the 421 * device state. After successful transition to a given state, the 422 * vendor driver should return success on write(device_state, state) 423 * system call. If the device state transition fails, the vendor driver 424 * should return an appropriate -errno for the fault condition. 425 * - On the user application side, if the device state transition fails, 426 * that is, if write(device_state, state) returns an error, read 427 * device_state again to determine the current state of the device from 428 * the vendor driver. 429 * - The vendor driver should return previous state of the device unless 430 * the vendor driver has encountered an internal error, in which case 431 * the vendor driver may report the device_state VFIO_DEVICE_STATE_ERROR. 432 * - The user application must use the device reset ioctl to recover the 433 * device from VFIO_DEVICE_STATE_ERROR state. If the device is 434 * indicated to be in a valid device state by reading device_state, the 435 * user application may attempt to transition the device to any valid 436 * state reachable from the current state or terminate itself. 437 * 438 * device_state consists of 3 bits: 439 * - If bit 0 is set, it indicates the _RUNNING state. If bit 0 is clear, 440 * it indicates the _STOP state. When the device state is changed to 441 * _STOP, driver should stop the device before write() returns. 442 * - If bit 1 is set, it indicates the _SAVING state, which means that the 443 * driver should start gathering device state information that will be 444 * provided to the VFIO user application to save the device's state. 445 * - If bit 2 is set, it indicates the _RESUMING state, which means that 446 * the driver should prepare to resume the device. Data provided through 447 * the migration region should be used to resume the device. 448 * Bits 3 - 31 are reserved for future use. To preserve them, the user 449 * application should perform a read-modify-write operation on this 450 * field when modifying the specified bits. 451 * 452 * +------- _RESUMING 453 * |+------ _SAVING 454 * ||+----- _RUNNING 455 * ||| 456 * 000b => Device Stopped, not saving or resuming 457 * 001b => Device running, which is the default state 458 * 010b => Stop the device & save the device state, stop-and-copy state 459 * 011b => Device running and save the device state, pre-copy state 460 * 100b => Device stopped and the device state is resuming 461 * 101b => Invalid state 462 * 110b => Error state 463 * 111b => Invalid state 464 * 465 * State transitions: 466 * 467 * _RESUMING _RUNNING Pre-copy Stop-and-copy _STOP 468 * (100b) (001b) (011b) (010b) (000b) 469 * 0. Running or default state 470 * | 471 * 472 * 1. Normal Shutdown (optional) 473 * |------------------------------------->| 474 * 475 * 2. Save the state or suspend 476 * |------------------------->|---------->| 477 * 478 * 3. Save the state during live migration 479 * |----------->|------------>|---------->| 480 * 481 * 4. Resuming 482 * |<---------| 483 * 484 * 5. Resumed 485 * |--------->| 486 * 487 * 0. Default state of VFIO device is _RUNNING when the user application starts. 488 * 1. During normal shutdown of the user application, the user application may 489 * optionally change the VFIO device state from _RUNNING to _STOP. This 490 * transition is optional. The vendor driver must support this transition but 491 * must not require it. 492 * 2. When the user application saves state or suspends the application, the 493 * device state transitions from _RUNNING to stop-and-copy and then to _STOP. 494 * On state transition from _RUNNING to stop-and-copy, driver must stop the 495 * device, save the device state and send it to the application through the 496 * migration region. The sequence to be followed for such transition is given 497 * below. 498 * 3. In live migration of user application, the state transitions from _RUNNING 499 * to pre-copy, to stop-and-copy, and to _STOP. 500 * On state transition from _RUNNING to pre-copy, the driver should start 501 * gathering the device state while the application is still running and send 502 * the device state data to application through the migration region. 503 * On state transition from pre-copy to stop-and-copy, the driver must stop 504 * the device, save the device state and send it to the user application 505 * through the migration region. 506 * Vendor drivers must support the pre-copy state even for implementations 507 * where no data is provided to the user before the stop-and-copy state. The 508 * user must not be required to consume all migration data before the device 509 * transitions to a new state, including the stop-and-copy state. 510 * The sequence to be followed for above two transitions is given below. 511 * 4. To start the resuming phase, the device state should be transitioned from 512 * the _RUNNING to the _RESUMING state. 513 * In the _RESUMING state, the driver should use the device state data 514 * received through the migration region to resume the device. 515 * 5. After providing saved device data to the driver, the application should 516 * change the state from _RESUMING to _RUNNING. 517 * 518 * reserved: 519 * Reads on this field return zero and writes are ignored. 520 * 521 * pending_bytes: (read only) 522 * The number of pending bytes still to be migrated from the vendor driver. 523 * 524 * data_offset: (read only) 525 * The user application should read data_offset field from the migration 526 * region. The user application should read the device data from this 527 * offset within the migration region during the _SAVING state or write 528 * the device data during the _RESUMING state. See below for details of 529 * sequence to be followed. 530 * 531 * data_size: (read/write) 532 * The user application should read data_size to get the size in bytes of 533 * the data copied in the migration region during the _SAVING state and 534 * write the size in bytes of the data copied in the migration region 535 * during the _RESUMING state. 536 * 537 * The format of the migration region is as follows: 538 * ------------------------------------------------------------------ 539 * |vfio_device_migration_info| data section | 540 * | | /////////////////////////////// | 541 * ------------------------------------------------------------------ 542 * ^ ^ 543 * offset 0-trapped part data_offset 544 * 545 * The structure vfio_device_migration_info is always followed by the data 546 * section in the region, so data_offset will always be nonzero. The offset 547 * from where the data is copied is decided by the kernel driver. The data 548 * section can be trapped, mmapped, or partitioned, depending on how the kernel 549 * driver defines the data section. The data section partition can be defined 550 * as mapped by the sparse mmap capability. If mmapped, data_offset must be 551 * page aligned, whereas initial section which contains the 552 * vfio_device_migration_info structure, might not end at the offset, which is 553 * page aligned. The user is not required to access through mmap regardless 554 * of the capabilities of the region mmap. 555 * The vendor driver should determine whether and how to partition the data 556 * section. The vendor driver should return data_offset accordingly. 557 * 558 * The sequence to be followed while in pre-copy state and stop-and-copy state 559 * is as follows: 560 * a. Read pending_bytes, indicating the start of a new iteration to get device 561 * data. Repeated read on pending_bytes at this stage should have no side 562 * effects. 563 * If pending_bytes == 0, the user application should not iterate to get data 564 * for that device. 565 * If pending_bytes > 0, perform the following steps. 566 * b. Read data_offset, indicating that the vendor driver should make data 567 * available through the data section. The vendor driver should return this 568 * read operation only after data is available from (region + data_offset) 569 * to (region + data_offset + data_size). 570 * c. Read data_size, which is the amount of data in bytes available through 571 * the migration region. 572 * Read on data_offset and data_size should return the offset and size of 573 * the current buffer if the user application reads data_offset and 574 * data_size more than once here. 575 * d. Read data_size bytes of data from (region + data_offset) from the 576 * migration region. 577 * e. Process the data. 578 * f. Read pending_bytes, which indicates that the data from the previous 579 * iteration has been read. If pending_bytes > 0, go to step b. 580 * 581 * The user application can transition from the _SAVING|_RUNNING 582 * (pre-copy state) to the _SAVING (stop-and-copy) state regardless of the 583 * number of pending bytes. The user application should iterate in _SAVING 584 * (stop-and-copy) until pending_bytes is 0. 585 * 586 * The sequence to be followed while _RESUMING device state is as follows: 587 * While data for this device is available, repeat the following steps: 588 * a. Read data_offset from where the user application should write data. 589 * b. Write migration data starting at the migration region + data_offset for 590 * the length determined by data_size from the migration source. 591 * c. Write data_size, which indicates to the vendor driver that data is 592 * written in the migration region. Vendor driver must return this write 593 * operations on consuming data. Vendor driver should apply the 594 * user-provided migration region data to the device resume state. 595 * 596 * If an error occurs during the above sequences, the vendor driver can return 597 * an error code for next read() or write() operation, which will terminate the 598 * loop. The user application should then take the next necessary action, for 599 * example, failing migration or terminating the user application. 600 * 601 * For the user application, data is opaque. The user application should write 602 * data in the same order as the data is received and the data should be of 603 * same transaction size at the source. 604 */ 605 606 struct vfio_device_migration_info { 607 __u32 device_state; /* VFIO device state */ 608 #define VFIO_DEVICE_STATE_STOP (0) 609 #define VFIO_DEVICE_STATE_RUNNING (1 << 0) 610 #define VFIO_DEVICE_STATE_SAVING (1 << 1) 611 #define VFIO_DEVICE_STATE_RESUMING (1 << 2) 612 #define VFIO_DEVICE_STATE_MASK (VFIO_DEVICE_STATE_RUNNING | \ 613 VFIO_DEVICE_STATE_SAVING | \ 614 VFIO_DEVICE_STATE_RESUMING) 615 616 #define VFIO_DEVICE_STATE_VALID(state) \ 617 (state & VFIO_DEVICE_STATE_RESUMING ? \ 618 (state & VFIO_DEVICE_STATE_MASK) == VFIO_DEVICE_STATE_RESUMING : 1) 619 620 #define VFIO_DEVICE_STATE_IS_ERROR(state) \ 621 ((state & VFIO_DEVICE_STATE_MASK) == (VFIO_DEVICE_STATE_SAVING | \ 622 VFIO_DEVICE_STATE_RESUMING)) 623 624 #define VFIO_DEVICE_STATE_SET_ERROR(state) \ 625 ((state & ~VFIO_DEVICE_STATE_MASK) | VFIO_DEVICE_SATE_SAVING | \ 626 VFIO_DEVICE_STATE_RESUMING) 627 628 __u32 reserved; 629 __u64 pending_bytes; 630 __u64 data_offset; 631 __u64 data_size; 632 }; 633 634 /* 635 * The MSIX mappable capability informs that MSIX data of a BAR can be mmapped 636 * which allows direct access to non-MSIX registers which happened to be within 637 * the same system page. 638 * 639 * Even though the userspace gets direct access to the MSIX data, the existing 640 * VFIO_DEVICE_SET_IRQS interface must still be used for MSIX configuration. 641 */ 642 #define VFIO_REGION_INFO_CAP_MSIX_MAPPABLE 3 643 644 /* 645 * Capability with compressed real address (aka SSA - small system address) 646 * where GPU RAM is mapped on a system bus. Used by a GPU for DMA routing 647 * and by the userspace to associate a NVLink bridge with a GPU. 648 * 649 * Deprecated, capability no longer provided 650 */ 651 #define VFIO_REGION_INFO_CAP_NVLINK2_SSATGT 4 652 653 struct vfio_region_info_cap_nvlink2_ssatgt { 654 struct vfio_info_cap_header header; 655 __u64 tgt; 656 }; 657 658 /* 659 * Capability with an NVLink link speed. The value is read by 660 * the NVlink2 bridge driver from the bridge's "ibm,nvlink-speed" 661 * property in the device tree. The value is fixed in the hardware 662 * and failing to provide the correct value results in the link 663 * not working with no indication from the driver why. 664 * 665 * Deprecated, capability no longer provided 666 */ 667 #define VFIO_REGION_INFO_CAP_NVLINK2_LNKSPD 5 668 669 struct vfio_region_info_cap_nvlink2_lnkspd { 670 struct vfio_info_cap_header header; 671 __u32 link_speed; 672 __u32 __pad; 673 }; 674 675 /** 676 * VFIO_DEVICE_GET_IRQ_INFO - _IOWR(VFIO_TYPE, VFIO_BASE + 9, 677 * struct vfio_irq_info) 678 * 679 * Retrieve information about a device IRQ. Caller provides 680 * struct vfio_irq_info with index value set. Caller sets argsz. 681 * Implementation of IRQ mapping is bus driver specific. Indexes 682 * using multiple IRQs are primarily intended to support MSI-like 683 * interrupt blocks. Zero count irq blocks may be used to describe 684 * unimplemented interrupt types. 685 * 686 * The EVENTFD flag indicates the interrupt index supports eventfd based 687 * signaling. 688 * 689 * The MASKABLE flags indicates the index supports MASK and UNMASK 690 * actions described below. 691 * 692 * AUTOMASKED indicates that after signaling, the interrupt line is 693 * automatically masked by VFIO and the user needs to unmask the line 694 * to receive new interrupts. This is primarily intended to distinguish 695 * level triggered interrupts. 696 * 697 * The NORESIZE flag indicates that the interrupt lines within the index 698 * are setup as a set and new subindexes cannot be enabled without first 699 * disabling the entire index. This is used for interrupts like PCI MSI 700 * and MSI-X where the driver may only use a subset of the available 701 * indexes, but VFIO needs to enable a specific number of vectors 702 * upfront. In the case of MSI-X, where the user can enable MSI-X and 703 * then add and unmask vectors, it's up to userspace to make the decision 704 * whether to allocate the maximum supported number of vectors or tear 705 * down setup and incrementally increase the vectors as each is enabled. 706 */ 707 struct vfio_irq_info { 708 __u32 argsz; 709 __u32 flags; 710 #define VFIO_IRQ_INFO_EVENTFD (1 << 0) 711 #define VFIO_IRQ_INFO_MASKABLE (1 << 1) 712 #define VFIO_IRQ_INFO_AUTOMASKED (1 << 2) 713 #define VFIO_IRQ_INFO_NORESIZE (1 << 3) 714 __u32 index; /* IRQ index */ 715 __u32 count; /* Number of IRQs within this index */ 716 }; 717 #define VFIO_DEVICE_GET_IRQ_INFO _IO(VFIO_TYPE, VFIO_BASE + 9) 718 719 /** 720 * VFIO_DEVICE_SET_IRQS - _IOW(VFIO_TYPE, VFIO_BASE + 10, struct vfio_irq_set) 721 * 722 * Set signaling, masking, and unmasking of interrupts. Caller provides 723 * struct vfio_irq_set with all fields set. 'start' and 'count' indicate 724 * the range of subindexes being specified. 725 * 726 * The DATA flags specify the type of data provided. If DATA_NONE, the 727 * operation performs the specified action immediately on the specified 728 * interrupt(s). For example, to unmask AUTOMASKED interrupt [0,0]: 729 * flags = (DATA_NONE|ACTION_UNMASK), index = 0, start = 0, count = 1. 730 * 731 * DATA_BOOL allows sparse support for the same on arrays of interrupts. 732 * For example, to mask interrupts [0,1] and [0,3] (but not [0,2]): 733 * flags = (DATA_BOOL|ACTION_MASK), index = 0, start = 1, count = 3, 734 * data = {1,0,1} 735 * 736 * DATA_EVENTFD binds the specified ACTION to the provided __s32 eventfd. 737 * A value of -1 can be used to either de-assign interrupts if already 738 * assigned or skip un-assigned interrupts. For example, to set an eventfd 739 * to be trigger for interrupts [0,0] and [0,2]: 740 * flags = (DATA_EVENTFD|ACTION_TRIGGER), index = 0, start = 0, count = 3, 741 * data = {fd1, -1, fd2} 742 * If index [0,1] is previously set, two count = 1 ioctls calls would be 743 * required to set [0,0] and [0,2] without changing [0,1]. 744 * 745 * Once a signaling mechanism is set, DATA_BOOL or DATA_NONE can be used 746 * with ACTION_TRIGGER to perform kernel level interrupt loopback testing 747 * from userspace (ie. simulate hardware triggering). 748 * 749 * Setting of an event triggering mechanism to userspace for ACTION_TRIGGER 750 * enables the interrupt index for the device. Individual subindex interrupts 751 * can be disabled using the -1 value for DATA_EVENTFD or the index can be 752 * disabled as a whole with: flags = (DATA_NONE|ACTION_TRIGGER), count = 0. 753 * 754 * Note that ACTION_[UN]MASK specify user->kernel signaling (irqfds) while 755 * ACTION_TRIGGER specifies kernel->user signaling. 756 */ 757 struct vfio_irq_set { 758 __u32 argsz; 759 __u32 flags; 760 #define VFIO_IRQ_SET_DATA_NONE (1 << 0) /* Data not present */ 761 #define VFIO_IRQ_SET_DATA_BOOL (1 << 1) /* Data is bool (u8) */ 762 #define VFIO_IRQ_SET_DATA_EVENTFD (1 << 2) /* Data is eventfd (s32) */ 763 #define VFIO_IRQ_SET_ACTION_MASK (1 << 3) /* Mask interrupt */ 764 #define VFIO_IRQ_SET_ACTION_UNMASK (1 << 4) /* Unmask interrupt */ 765 #define VFIO_IRQ_SET_ACTION_TRIGGER (1 << 5) /* Trigger interrupt */ 766 __u32 index; 767 __u32 start; 768 __u32 count; 769 __u8 data[]; 770 }; 771 #define VFIO_DEVICE_SET_IRQS _IO(VFIO_TYPE, VFIO_BASE + 10) 772 773 #define VFIO_IRQ_SET_DATA_TYPE_MASK (VFIO_IRQ_SET_DATA_NONE | \ 774 VFIO_IRQ_SET_DATA_BOOL | \ 775 VFIO_IRQ_SET_DATA_EVENTFD) 776 #define VFIO_IRQ_SET_ACTION_TYPE_MASK (VFIO_IRQ_SET_ACTION_MASK | \ 777 VFIO_IRQ_SET_ACTION_UNMASK | \ 778 VFIO_IRQ_SET_ACTION_TRIGGER) 779 /** 780 * VFIO_DEVICE_RESET - _IO(VFIO_TYPE, VFIO_BASE + 11) 781 * 782 * Reset a device. 783 */ 784 #define VFIO_DEVICE_RESET _IO(VFIO_TYPE, VFIO_BASE + 11) 785 786 /* 787 * The VFIO-PCI bus driver makes use of the following fixed region and 788 * IRQ index mapping. Unimplemented regions return a size of zero. 789 * Unimplemented IRQ types return a count of zero. 790 */ 791 792 enum { 793 VFIO_PCI_BAR0_REGION_INDEX, 794 VFIO_PCI_BAR1_REGION_INDEX, 795 VFIO_PCI_BAR2_REGION_INDEX, 796 VFIO_PCI_BAR3_REGION_INDEX, 797 VFIO_PCI_BAR4_REGION_INDEX, 798 VFIO_PCI_BAR5_REGION_INDEX, 799 VFIO_PCI_ROM_REGION_INDEX, 800 VFIO_PCI_CONFIG_REGION_INDEX, 801 /* 802 * Expose VGA regions defined for PCI base class 03, subclass 00. 803 * This includes I/O port ranges 0x3b0 to 0x3bb and 0x3c0 to 0x3df 804 * as well as the MMIO range 0xa0000 to 0xbffff. Each implemented 805 * range is found at it's identity mapped offset from the region 806 * offset, for example 0x3b0 is region_info.offset + 0x3b0. Areas 807 * between described ranges are unimplemented. 808 */ 809 VFIO_PCI_VGA_REGION_INDEX, 810 VFIO_PCI_NUM_REGIONS = 9 /* Fixed user ABI, region indexes >=9 use */ 811 /* device specific cap to define content. */ 812 }; 813 814 enum { 815 VFIO_PCI_INTX_IRQ_INDEX, 816 VFIO_PCI_MSI_IRQ_INDEX, 817 VFIO_PCI_MSIX_IRQ_INDEX, 818 VFIO_PCI_ERR_IRQ_INDEX, 819 VFIO_PCI_REQ_IRQ_INDEX, 820 VFIO_PCI_NUM_IRQS 821 }; 822 823 /* 824 * The vfio-ccw bus driver makes use of the following fixed region and 825 * IRQ index mapping. Unimplemented regions return a size of zero. 826 * Unimplemented IRQ types return a count of zero. 827 */ 828 829 enum { 830 VFIO_CCW_CONFIG_REGION_INDEX, 831 VFIO_CCW_NUM_REGIONS 832 }; 833 834 enum { 835 VFIO_CCW_IO_IRQ_INDEX, 836 VFIO_CCW_CRW_IRQ_INDEX, 837 VFIO_CCW_REQ_IRQ_INDEX, 838 VFIO_CCW_NUM_IRQS 839 }; 840 841 /** 842 * VFIO_DEVICE_GET_PCI_HOT_RESET_INFO - _IORW(VFIO_TYPE, VFIO_BASE + 12, 843 * struct vfio_pci_hot_reset_info) 844 * 845 * Return: 0 on success, -errno on failure: 846 * -enospc = insufficient buffer, -enodev = unsupported for device. 847 */ 848 struct vfio_pci_dependent_device { 849 __u32 group_id; 850 __u16 segment; 851 __u8 bus; 852 __u8 devfn; /* Use PCI_SLOT/PCI_FUNC */ 853 }; 854 855 struct vfio_pci_hot_reset_info { 856 __u32 argsz; 857 __u32 flags; 858 __u32 count; 859 struct vfio_pci_dependent_device devices[]; 860 }; 861 862 #define VFIO_DEVICE_GET_PCI_HOT_RESET_INFO _IO(VFIO_TYPE, VFIO_BASE + 12) 863 864 /** 865 * VFIO_DEVICE_PCI_HOT_RESET - _IOW(VFIO_TYPE, VFIO_BASE + 13, 866 * struct vfio_pci_hot_reset) 867 * 868 * Return: 0 on success, -errno on failure. 869 */ 870 struct vfio_pci_hot_reset { 871 __u32 argsz; 872 __u32 flags; 873 __u32 count; 874 __s32 group_fds[]; 875 }; 876 877 #define VFIO_DEVICE_PCI_HOT_RESET _IO(VFIO_TYPE, VFIO_BASE + 13) 878 879 /** 880 * VFIO_DEVICE_QUERY_GFX_PLANE - _IOW(VFIO_TYPE, VFIO_BASE + 14, 881 * struct vfio_device_query_gfx_plane) 882 * 883 * Set the drm_plane_type and flags, then retrieve the gfx plane info. 884 * 885 * flags supported: 886 * - VFIO_GFX_PLANE_TYPE_PROBE and VFIO_GFX_PLANE_TYPE_DMABUF are set 887 * to ask if the mdev supports dma-buf. 0 on support, -EINVAL on no 888 * support for dma-buf. 889 * - VFIO_GFX_PLANE_TYPE_PROBE and VFIO_GFX_PLANE_TYPE_REGION are set 890 * to ask if the mdev supports region. 0 on support, -EINVAL on no 891 * support for region. 892 * - VFIO_GFX_PLANE_TYPE_DMABUF or VFIO_GFX_PLANE_TYPE_REGION is set 893 * with each call to query the plane info. 894 * - Others are invalid and return -EINVAL. 895 * 896 * Note: 897 * 1. Plane could be disabled by guest. In that case, success will be 898 * returned with zero-initialized drm_format, size, width and height 899 * fields. 900 * 2. x_hot/y_hot is set to 0xFFFFFFFF if no hotspot information available 901 * 902 * Return: 0 on success, -errno on other failure. 903 */ 904 struct vfio_device_gfx_plane_info { 905 __u32 argsz; 906 __u32 flags; 907 #define VFIO_GFX_PLANE_TYPE_PROBE (1 << 0) 908 #define VFIO_GFX_PLANE_TYPE_DMABUF (1 << 1) 909 #define VFIO_GFX_PLANE_TYPE_REGION (1 << 2) 910 /* in */ 911 __u32 drm_plane_type; /* type of plane: DRM_PLANE_TYPE_* */ 912 /* out */ 913 __u32 drm_format; /* drm format of plane */ 914 __u64 drm_format_mod; /* tiled mode */ 915 __u32 width; /* width of plane */ 916 __u32 height; /* height of plane */ 917 __u32 stride; /* stride of plane */ 918 __u32 size; /* size of plane in bytes, align on page*/ 919 __u32 x_pos; /* horizontal position of cursor plane */ 920 __u32 y_pos; /* vertical position of cursor plane*/ 921 __u32 x_hot; /* horizontal position of cursor hotspot */ 922 __u32 y_hot; /* vertical position of cursor hotspot */ 923 union { 924 __u32 region_index; /* region index */ 925 __u32 dmabuf_id; /* dma-buf id */ 926 }; 927 }; 928 929 #define VFIO_DEVICE_QUERY_GFX_PLANE _IO(VFIO_TYPE, VFIO_BASE + 14) 930 931 /** 932 * VFIO_DEVICE_GET_GFX_DMABUF - _IOW(VFIO_TYPE, VFIO_BASE + 15, __u32) 933 * 934 * Return a new dma-buf file descriptor for an exposed guest framebuffer 935 * described by the provided dmabuf_id. The dmabuf_id is returned from VFIO_ 936 * DEVICE_QUERY_GFX_PLANE as a token of the exposed guest framebuffer. 937 */ 938 939 #define VFIO_DEVICE_GET_GFX_DMABUF _IO(VFIO_TYPE, VFIO_BASE + 15) 940 941 /** 942 * VFIO_DEVICE_IOEVENTFD - _IOW(VFIO_TYPE, VFIO_BASE + 16, 943 * struct vfio_device_ioeventfd) 944 * 945 * Perform a write to the device at the specified device fd offset, with 946 * the specified data and width when the provided eventfd is triggered. 947 * vfio bus drivers may not support this for all regions, for all widths, 948 * or at all. vfio-pci currently only enables support for BAR regions, 949 * excluding the MSI-X vector table. 950 * 951 * Return: 0 on success, -errno on failure. 952 */ 953 struct vfio_device_ioeventfd { 954 __u32 argsz; 955 __u32 flags; 956 #define VFIO_DEVICE_IOEVENTFD_8 (1 << 0) /* 1-byte write */ 957 #define VFIO_DEVICE_IOEVENTFD_16 (1 << 1) /* 2-byte write */ 958 #define VFIO_DEVICE_IOEVENTFD_32 (1 << 2) /* 4-byte write */ 959 #define VFIO_DEVICE_IOEVENTFD_64 (1 << 3) /* 8-byte write */ 960 #define VFIO_DEVICE_IOEVENTFD_SIZE_MASK (0xf) 961 __u64 offset; /* device fd offset of write */ 962 __u64 data; /* data to be written */ 963 __s32 fd; /* -1 for de-assignment */ 964 }; 965 966 #define VFIO_DEVICE_IOEVENTFD _IO(VFIO_TYPE, VFIO_BASE + 16) 967 968 /** 969 * VFIO_DEVICE_FEATURE - _IORW(VFIO_TYPE, VFIO_BASE + 17, 970 * struct vfio_device_feature) 971 * 972 * Get, set, or probe feature data of the device. The feature is selected 973 * using the FEATURE_MASK portion of the flags field. Support for a feature 974 * can be probed by setting both the FEATURE_MASK and PROBE bits. A probe 975 * may optionally include the GET and/or SET bits to determine read vs write 976 * access of the feature respectively. Probing a feature will return success 977 * if the feature is supported and all of the optionally indicated GET/SET 978 * methods are supported. The format of the data portion of the structure is 979 * specific to the given feature. The data portion is not required for 980 * probing. GET and SET are mutually exclusive, except for use with PROBE. 981 * 982 * Return 0 on success, -errno on failure. 983 */ 984 struct vfio_device_feature { 985 __u32 argsz; 986 __u32 flags; 987 #define VFIO_DEVICE_FEATURE_MASK (0xffff) /* 16-bit feature index */ 988 #define VFIO_DEVICE_FEATURE_GET (1 << 16) /* Get feature into data[] */ 989 #define VFIO_DEVICE_FEATURE_SET (1 << 17) /* Set feature from data[] */ 990 #define VFIO_DEVICE_FEATURE_PROBE (1 << 18) /* Probe feature support */ 991 __u8 data[]; 992 }; 993 994 #define VFIO_DEVICE_FEATURE _IO(VFIO_TYPE, VFIO_BASE + 17) 995 996 /* 997 * Provide support for setting a PCI VF Token, which is used as a shared 998 * secret between PF and VF drivers. This feature may only be set on a 999 * PCI SR-IOV PF when SR-IOV is enabled on the PF and there are no existing 1000 * open VFs. Data provided when setting this feature is a 16-byte array 1001 * (__u8 b[16]), representing a UUID. 1002 */ 1003 #define VFIO_DEVICE_FEATURE_PCI_VF_TOKEN (0) 1004 1005 /* -------- API for Type1 VFIO IOMMU -------- */ 1006 1007 /** 1008 * VFIO_IOMMU_GET_INFO - _IOR(VFIO_TYPE, VFIO_BASE + 12, struct vfio_iommu_info) 1009 * 1010 * Retrieve information about the IOMMU object. Fills in provided 1011 * struct vfio_iommu_info. Caller sets argsz. 1012 * 1013 * XXX Should we do these by CHECK_EXTENSION too? 1014 */ 1015 struct vfio_iommu_type1_info { 1016 __u32 argsz; 1017 __u32 flags; 1018 #define VFIO_IOMMU_INFO_PGSIZES (1 << 0) /* supported page sizes info */ 1019 #define VFIO_IOMMU_INFO_CAPS (1 << 1) /* Info supports caps */ 1020 __u64 iova_pgsizes; /* Bitmap of supported page sizes */ 1021 __u32 cap_offset; /* Offset within info struct of first cap */ 1022 }; 1023 1024 /* 1025 * The IOVA capability allows to report the valid IOVA range(s) 1026 * excluding any non-relaxable reserved regions exposed by 1027 * devices attached to the container. Any DMA map attempt 1028 * outside the valid iova range will return error. 1029 * 1030 * The structures below define version 1 of this capability. 1031 */ 1032 #define VFIO_IOMMU_TYPE1_INFO_CAP_IOVA_RANGE 1 1033 1034 struct vfio_iova_range { 1035 __u64 start; 1036 __u64 end; 1037 }; 1038 1039 struct vfio_iommu_type1_info_cap_iova_range { 1040 struct vfio_info_cap_header header; 1041 __u32 nr_iovas; 1042 __u32 reserved; 1043 struct vfio_iova_range iova_ranges[]; 1044 }; 1045 1046 /* 1047 * The migration capability allows to report supported features for migration. 1048 * 1049 * The structures below define version 1 of this capability. 1050 * 1051 * The existence of this capability indicates that IOMMU kernel driver supports 1052 * dirty page logging. 1053 * 1054 * pgsize_bitmap: Kernel driver returns bitmap of supported page sizes for dirty 1055 * page logging. 1056 * max_dirty_bitmap_size: Kernel driver returns maximum supported dirty bitmap 1057 * size in bytes that can be used by user applications when getting the dirty 1058 * bitmap. 1059 */ 1060 #define VFIO_IOMMU_TYPE1_INFO_CAP_MIGRATION 2 1061 1062 struct vfio_iommu_type1_info_cap_migration { 1063 struct vfio_info_cap_header header; 1064 __u32 flags; 1065 __u64 pgsize_bitmap; 1066 __u64 max_dirty_bitmap_size; /* in bytes */ 1067 }; 1068 1069 /* 1070 * The DMA available capability allows to report the current number of 1071 * simultaneously outstanding DMA mappings that are allowed. 1072 * 1073 * The structure below defines version 1 of this capability. 1074 * 1075 * avail: specifies the current number of outstanding DMA mappings allowed. 1076 */ 1077 #define VFIO_IOMMU_TYPE1_INFO_DMA_AVAIL 3 1078 1079 struct vfio_iommu_type1_info_dma_avail { 1080 struct vfio_info_cap_header header; 1081 __u32 avail; 1082 }; 1083 1084 #define VFIO_IOMMU_GET_INFO _IO(VFIO_TYPE, VFIO_BASE + 12) 1085 1086 /** 1087 * VFIO_IOMMU_MAP_DMA - _IOW(VFIO_TYPE, VFIO_BASE + 13, struct vfio_dma_map) 1088 * 1089 * Map process virtual addresses to IO virtual addresses using the 1090 * provided struct vfio_dma_map. Caller sets argsz. READ &/ WRITE required. 1091 * 1092 * If flags & VFIO_DMA_MAP_FLAG_VADDR, update the base vaddr for iova, and 1093 * unblock translation of host virtual addresses in the iova range. The vaddr 1094 * must have previously been invalidated with VFIO_DMA_UNMAP_FLAG_VADDR. To 1095 * maintain memory consistency within the user application, the updated vaddr 1096 * must address the same memory object as originally mapped. Failure to do so 1097 * will result in user memory corruption and/or device misbehavior. iova and 1098 * size must match those in the original MAP_DMA call. Protection is not 1099 * changed, and the READ & WRITE flags must be 0. 1100 */ 1101 struct vfio_iommu_type1_dma_map { 1102 __u32 argsz; 1103 __u32 flags; 1104 #define VFIO_DMA_MAP_FLAG_READ (1 << 0) /* readable from device */ 1105 #define VFIO_DMA_MAP_FLAG_WRITE (1 << 1) /* writable from device */ 1106 #define VFIO_DMA_MAP_FLAG_VADDR (1 << 2) 1107 __u64 vaddr; /* Process virtual address */ 1108 __u64 iova; /* IO virtual address */ 1109 __u64 size; /* Size of mapping (bytes) */ 1110 }; 1111 1112 #define VFIO_IOMMU_MAP_DMA _IO(VFIO_TYPE, VFIO_BASE + 13) 1113 1114 struct vfio_bitmap { 1115 __u64 pgsize; /* page size for bitmap in bytes */ 1116 __u64 size; /* in bytes */ 1117 __u64 *data; /* one bit per page */ 1118 }; 1119 1120 /** 1121 * VFIO_IOMMU_UNMAP_DMA - _IOWR(VFIO_TYPE, VFIO_BASE + 14, 1122 * struct vfio_dma_unmap) 1123 * 1124 * Unmap IO virtual addresses using the provided struct vfio_dma_unmap. 1125 * Caller sets argsz. The actual unmapped size is returned in the size 1126 * field. No guarantee is made to the user that arbitrary unmaps of iova 1127 * or size different from those used in the original mapping call will 1128 * succeed. 1129 * 1130 * VFIO_DMA_UNMAP_FLAG_GET_DIRTY_BITMAP should be set to get the dirty bitmap 1131 * before unmapping IO virtual addresses. When this flag is set, the user must 1132 * provide a struct vfio_bitmap in data[]. User must provide zero-allocated 1133 * memory via vfio_bitmap.data and its size in the vfio_bitmap.size field. 1134 * A bit in the bitmap represents one page, of user provided page size in 1135 * vfio_bitmap.pgsize field, consecutively starting from iova offset. Bit set 1136 * indicates that the page at that offset from iova is dirty. A Bitmap of the 1137 * pages in the range of unmapped size is returned in the user-provided 1138 * vfio_bitmap.data. 1139 * 1140 * If flags & VFIO_DMA_UNMAP_FLAG_ALL, unmap all addresses. iova and size 1141 * must be 0. This cannot be combined with the get-dirty-bitmap flag. 1142 * 1143 * If flags & VFIO_DMA_UNMAP_FLAG_VADDR, do not unmap, but invalidate host 1144 * virtual addresses in the iova range. Tasks that attempt to translate an 1145 * iova's vaddr will block. DMA to already-mapped pages continues. This 1146 * cannot be combined with the get-dirty-bitmap flag. 1147 */ 1148 struct vfio_iommu_type1_dma_unmap { 1149 __u32 argsz; 1150 __u32 flags; 1151 #define VFIO_DMA_UNMAP_FLAG_GET_DIRTY_BITMAP (1 << 0) 1152 #define VFIO_DMA_UNMAP_FLAG_ALL (1 << 1) 1153 #define VFIO_DMA_UNMAP_FLAG_VADDR (1 << 2) 1154 __u64 iova; /* IO virtual address */ 1155 __u64 size; /* Size of mapping (bytes) */ 1156 __u8 data[]; 1157 }; 1158 1159 #define VFIO_IOMMU_UNMAP_DMA _IO(VFIO_TYPE, VFIO_BASE + 14) 1160 1161 /* 1162 * IOCTLs to enable/disable IOMMU container usage. 1163 * No parameters are supported. 1164 */ 1165 #define VFIO_IOMMU_ENABLE _IO(VFIO_TYPE, VFIO_BASE + 15) 1166 #define VFIO_IOMMU_DISABLE _IO(VFIO_TYPE, VFIO_BASE + 16) 1167 1168 /** 1169 * VFIO_IOMMU_DIRTY_PAGES - _IOWR(VFIO_TYPE, VFIO_BASE + 17, 1170 * struct vfio_iommu_type1_dirty_bitmap) 1171 * IOCTL is used for dirty pages logging. 1172 * Caller should set flag depending on which operation to perform, details as 1173 * below: 1174 * 1175 * Calling the IOCTL with VFIO_IOMMU_DIRTY_PAGES_FLAG_START flag set, instructs 1176 * the IOMMU driver to log pages that are dirtied or potentially dirtied by 1177 * the device; designed to be used when a migration is in progress. Dirty pages 1178 * are logged until logging is disabled by user application by calling the IOCTL 1179 * with VFIO_IOMMU_DIRTY_PAGES_FLAG_STOP flag. 1180 * 1181 * Calling the IOCTL with VFIO_IOMMU_DIRTY_PAGES_FLAG_STOP flag set, instructs 1182 * the IOMMU driver to stop logging dirtied pages. 1183 * 1184 * Calling the IOCTL with VFIO_IOMMU_DIRTY_PAGES_FLAG_GET_BITMAP flag set 1185 * returns the dirty pages bitmap for IOMMU container for a given IOVA range. 1186 * The user must specify the IOVA range and the pgsize through the structure 1187 * vfio_iommu_type1_dirty_bitmap_get in the data[] portion. This interface 1188 * supports getting a bitmap of the smallest supported pgsize only and can be 1189 * modified in future to get a bitmap of any specified supported pgsize. The 1190 * user must provide a zeroed memory area for the bitmap memory and specify its 1191 * size in bitmap.size. One bit is used to represent one page consecutively 1192 * starting from iova offset. The user should provide page size in bitmap.pgsize 1193 * field. A bit set in the bitmap indicates that the page at that offset from 1194 * iova is dirty. The caller must set argsz to a value including the size of 1195 * structure vfio_iommu_type1_dirty_bitmap_get, but excluding the size of the 1196 * actual bitmap. If dirty pages logging is not enabled, an error will be 1197 * returned. 1198 * 1199 * Only one of the flags _START, _STOP and _GET may be specified at a time. 1200 * 1201 */ 1202 struct vfio_iommu_type1_dirty_bitmap { 1203 __u32 argsz; 1204 __u32 flags; 1205 #define VFIO_IOMMU_DIRTY_PAGES_FLAG_START (1 << 0) 1206 #define VFIO_IOMMU_DIRTY_PAGES_FLAG_STOP (1 << 1) 1207 #define VFIO_IOMMU_DIRTY_PAGES_FLAG_GET_BITMAP (1 << 2) 1208 __u8 data[]; 1209 }; 1210 1211 struct vfio_iommu_type1_dirty_bitmap_get { 1212 __u64 iova; /* IO virtual address */ 1213 __u64 size; /* Size of iova range */ 1214 struct vfio_bitmap bitmap; 1215 }; 1216 1217 #define VFIO_IOMMU_DIRTY_PAGES _IO(VFIO_TYPE, VFIO_BASE + 17) 1218 1219 /* -------- Additional API for SPAPR TCE (Server POWERPC) IOMMU -------- */ 1220 1221 /* 1222 * The SPAPR TCE DDW info struct provides the information about 1223 * the details of Dynamic DMA window capability. 1224 * 1225 * @pgsizes contains a page size bitmask, 4K/64K/16M are supported. 1226 * @max_dynamic_windows_supported tells the maximum number of windows 1227 * which the platform can create. 1228 * @levels tells the maximum number of levels in multi-level IOMMU tables; 1229 * this allows splitting a table into smaller chunks which reduces 1230 * the amount of physically contiguous memory required for the table. 1231 */ 1232 struct vfio_iommu_spapr_tce_ddw_info { 1233 __u64 pgsizes; /* Bitmap of supported page sizes */ 1234 __u32 max_dynamic_windows_supported; 1235 __u32 levels; 1236 }; 1237 1238 /* 1239 * The SPAPR TCE info struct provides the information about the PCI bus 1240 * address ranges available for DMA, these values are programmed into 1241 * the hardware so the guest has to know that information. 1242 * 1243 * The DMA 32 bit window start is an absolute PCI bus address. 1244 * The IOVA address passed via map/unmap ioctls are absolute PCI bus 1245 * addresses too so the window works as a filter rather than an offset 1246 * for IOVA addresses. 1247 * 1248 * Flags supported: 1249 * - VFIO_IOMMU_SPAPR_INFO_DDW: informs the userspace that dynamic DMA windows 1250 * (DDW) support is present. @ddw is only supported when DDW is present. 1251 */ 1252 struct vfio_iommu_spapr_tce_info { 1253 __u32 argsz; 1254 __u32 flags; 1255 #define VFIO_IOMMU_SPAPR_INFO_DDW (1 << 0) /* DDW supported */ 1256 __u32 dma32_window_start; /* 32 bit window start (bytes) */ 1257 __u32 dma32_window_size; /* 32 bit window size (bytes) */ 1258 struct vfio_iommu_spapr_tce_ddw_info ddw; 1259 }; 1260 1261 #define VFIO_IOMMU_SPAPR_TCE_GET_INFO _IO(VFIO_TYPE, VFIO_BASE + 12) 1262 1263 /* 1264 * EEH PE operation struct provides ways to: 1265 * - enable/disable EEH functionality; 1266 * - unfreeze IO/DMA for frozen PE; 1267 * - read PE state; 1268 * - reset PE; 1269 * - configure PE; 1270 * - inject EEH error. 1271 */ 1272 struct vfio_eeh_pe_err { 1273 __u32 type; 1274 __u32 func; 1275 __u64 addr; 1276 __u64 mask; 1277 }; 1278 1279 struct vfio_eeh_pe_op { 1280 __u32 argsz; 1281 __u32 flags; 1282 __u32 op; 1283 union { 1284 struct vfio_eeh_pe_err err; 1285 }; 1286 }; 1287 1288 #define VFIO_EEH_PE_DISABLE 0 /* Disable EEH functionality */ 1289 #define VFIO_EEH_PE_ENABLE 1 /* Enable EEH functionality */ 1290 #define VFIO_EEH_PE_UNFREEZE_IO 2 /* Enable IO for frozen PE */ 1291 #define VFIO_EEH_PE_UNFREEZE_DMA 3 /* Enable DMA for frozen PE */ 1292 #define VFIO_EEH_PE_GET_STATE 4 /* PE state retrieval */ 1293 #define VFIO_EEH_PE_STATE_NORMAL 0 /* PE in functional state */ 1294 #define VFIO_EEH_PE_STATE_RESET 1 /* PE reset in progress */ 1295 #define VFIO_EEH_PE_STATE_STOPPED 2 /* Stopped DMA and IO */ 1296 #define VFIO_EEH_PE_STATE_STOPPED_DMA 4 /* Stopped DMA only */ 1297 #define VFIO_EEH_PE_STATE_UNAVAIL 5 /* State unavailable */ 1298 #define VFIO_EEH_PE_RESET_DEACTIVATE 5 /* Deassert PE reset */ 1299 #define VFIO_EEH_PE_RESET_HOT 6 /* Assert hot reset */ 1300 #define VFIO_EEH_PE_RESET_FUNDAMENTAL 7 /* Assert fundamental reset */ 1301 #define VFIO_EEH_PE_CONFIGURE 8 /* PE configuration */ 1302 #define VFIO_EEH_PE_INJECT_ERR 9 /* Inject EEH error */ 1303 1304 #define VFIO_EEH_PE_OP _IO(VFIO_TYPE, VFIO_BASE + 21) 1305 1306 /** 1307 * VFIO_IOMMU_SPAPR_REGISTER_MEMORY - _IOW(VFIO_TYPE, VFIO_BASE + 17, struct vfio_iommu_spapr_register_memory) 1308 * 1309 * Registers user space memory where DMA is allowed. It pins 1310 * user pages and does the locked memory accounting so 1311 * subsequent VFIO_IOMMU_MAP_DMA/VFIO_IOMMU_UNMAP_DMA calls 1312 * get faster. 1313 */ 1314 struct vfio_iommu_spapr_register_memory { 1315 __u32 argsz; 1316 __u32 flags; 1317 __u64 vaddr; /* Process virtual address */ 1318 __u64 size; /* Size of mapping (bytes) */ 1319 }; 1320 #define VFIO_IOMMU_SPAPR_REGISTER_MEMORY _IO(VFIO_TYPE, VFIO_BASE + 17) 1321 1322 /** 1323 * VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - _IOW(VFIO_TYPE, VFIO_BASE + 18, struct vfio_iommu_spapr_register_memory) 1324 * 1325 * Unregisters user space memory registered with 1326 * VFIO_IOMMU_SPAPR_REGISTER_MEMORY. 1327 * Uses vfio_iommu_spapr_register_memory for parameters. 1328 */ 1329 #define VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY _IO(VFIO_TYPE, VFIO_BASE + 18) 1330 1331 /** 1332 * VFIO_IOMMU_SPAPR_TCE_CREATE - _IOWR(VFIO_TYPE, VFIO_BASE + 19, struct vfio_iommu_spapr_tce_create) 1333 * 1334 * Creates an additional TCE table and programs it (sets a new DMA window) 1335 * to every IOMMU group in the container. It receives page shift, window 1336 * size and number of levels in the TCE table being created. 1337 * 1338 * It allocates and returns an offset on a PCI bus of the new DMA window. 1339 */ 1340 struct vfio_iommu_spapr_tce_create { 1341 __u32 argsz; 1342 __u32 flags; 1343 /* in */ 1344 __u32 page_shift; 1345 __u32 __resv1; 1346 __u64 window_size; 1347 __u32 levels; 1348 __u32 __resv2; 1349 /* out */ 1350 __u64 start_addr; 1351 }; 1352 #define VFIO_IOMMU_SPAPR_TCE_CREATE _IO(VFIO_TYPE, VFIO_BASE + 19) 1353 1354 /** 1355 * VFIO_IOMMU_SPAPR_TCE_REMOVE - _IOW(VFIO_TYPE, VFIO_BASE + 20, struct vfio_iommu_spapr_tce_remove) 1356 * 1357 * Unprograms a TCE table from all groups in the container and destroys it. 1358 * It receives a PCI bus offset as a window id. 1359 */ 1360 struct vfio_iommu_spapr_tce_remove { 1361 __u32 argsz; 1362 __u32 flags; 1363 /* in */ 1364 __u64 start_addr; 1365 }; 1366 #define VFIO_IOMMU_SPAPR_TCE_REMOVE _IO(VFIO_TYPE, VFIO_BASE + 20) 1367 1368 /* ***************************************************************** */ 1369 1370 #endif /* VFIO_H */ 1371