Kyle Swenson | 8d8f654 | 2021-03-15 11:02:55 -0600 | [diff] [blame^] | 1 | /* |
| 2 | * VFIO API definition |
| 3 | * |
| 4 | * Copyright (C) 2012 Red Hat, Inc. All rights reserved. |
| 5 | * Author: Alex Williamson <alex.williamson@redhat.com> |
| 6 | * |
| 7 | * This program is free software; you can redistribute it and/or modify |
| 8 | * it under the terms of the GNU General Public License version 2 as |
| 9 | * published by the Free Software Foundation. |
| 10 | */ |
| 11 | #ifndef _UAPIVFIO_H |
| 12 | #define _UAPIVFIO_H |
| 13 | |
| 14 | #include <linux/types.h> |
| 15 | #include <linux/ioctl.h> |
| 16 | |
| 17 | #define VFIO_API_VERSION 0 |
| 18 | |
| 19 | |
| 20 | /* Kernel & User level defines for VFIO IOCTLs. */ |
| 21 | |
| 22 | /* Extensions */ |
| 23 | |
| 24 | #define VFIO_TYPE1_IOMMU 1 |
| 25 | #define VFIO_SPAPR_TCE_IOMMU 2 |
| 26 | #define VFIO_TYPE1v2_IOMMU 3 |
| 27 | /* |
| 28 | * IOMMU enforces DMA cache coherence (ex. PCIe NoSnoop stripping). This |
| 29 | * capability is subject to change as groups are added or removed. |
| 30 | */ |
| 31 | #define VFIO_DMA_CC_IOMMU 4 |
| 32 | |
| 33 | /* Check if EEH is supported */ |
| 34 | #define VFIO_EEH 5 |
| 35 | |
| 36 | /* Two-stage IOMMU */ |
| 37 | #define VFIO_TYPE1_NESTING_IOMMU 6 /* Implies v2 */ |
| 38 | |
| 39 | #define VFIO_SPAPR_TCE_v2_IOMMU 7 |
| 40 | |
| 41 | /* |
| 42 | * The IOCTL interface is designed for extensibility by embedding the |
| 43 | * structure length (argsz) and flags into structures passed between |
| 44 | * kernel and userspace. We therefore use the _IO() macro for these |
| 45 | * defines to avoid implicitly embedding a size into the ioctl request. |
| 46 | * As structure fields are added, argsz will increase to match and flag |
| 47 | * bits will be defined to indicate additional fields with valid data. |
| 48 | * It's *always* the caller's responsibility to indicate the size of |
| 49 | * the structure passed by setting argsz appropriately. |
| 50 | */ |
| 51 | |
| 52 | #define VFIO_TYPE (';') |
| 53 | #define VFIO_BASE 100 |
| 54 | |
| 55 | /* -------- IOCTLs for VFIO file descriptor (/dev/vfio/vfio) -------- */ |
| 56 | |
| 57 | /** |
| 58 | * VFIO_GET_API_VERSION - _IO(VFIO_TYPE, VFIO_BASE + 0) |
| 59 | * |
| 60 | * Report the version of the VFIO API. This allows us to bump the entire |
| 61 | * API version should we later need to add or change features in incompatible |
| 62 | * ways. |
| 63 | * Return: VFIO_API_VERSION |
| 64 | * Availability: Always |
| 65 | */ |
| 66 | #define VFIO_GET_API_VERSION _IO(VFIO_TYPE, VFIO_BASE + 0) |
| 67 | |
| 68 | /** |
| 69 | * VFIO_CHECK_EXTENSION - _IOW(VFIO_TYPE, VFIO_BASE + 1, __u32) |
| 70 | * |
| 71 | * Check whether an extension is supported. |
| 72 | * Return: 0 if not supported, 1 (or some other positive integer) if supported. |
| 73 | * Availability: Always |
| 74 | */ |
| 75 | #define VFIO_CHECK_EXTENSION _IO(VFIO_TYPE, VFIO_BASE + 1) |
| 76 | |
| 77 | /** |
| 78 | * VFIO_SET_IOMMU - _IOW(VFIO_TYPE, VFIO_BASE + 2, __s32) |
| 79 | * |
| 80 | * Set the iommu to the given type. The type must be supported by an |
| 81 | * iommu driver as verified by calling CHECK_EXTENSION using the same |
| 82 | * type. A group must be set to this file descriptor before this |
| 83 | * ioctl is available. The IOMMU interfaces enabled by this call are |
| 84 | * specific to the value set. |
| 85 | * Return: 0 on success, -errno on failure |
| 86 | * Availability: When VFIO group attached |
| 87 | */ |
| 88 | #define VFIO_SET_IOMMU _IO(VFIO_TYPE, VFIO_BASE + 2) |
| 89 | |
| 90 | /* -------- IOCTLs for GROUP file descriptors (/dev/vfio/$GROUP) -------- */ |
| 91 | |
| 92 | /** |
| 93 | * VFIO_GROUP_GET_STATUS - _IOR(VFIO_TYPE, VFIO_BASE + 3, |
| 94 | * struct vfio_group_status) |
| 95 | * |
| 96 | * Retrieve information about the group. Fills in provided |
| 97 | * struct vfio_group_info. Caller sets argsz. |
| 98 | * Return: 0 on succes, -errno on failure. |
| 99 | * Availability: Always |
| 100 | */ |
| 101 | struct vfio_group_status { |
| 102 | __u32 argsz; |
| 103 | __u32 flags; |
| 104 | #define VFIO_GROUP_FLAGS_VIABLE (1 << 0) |
| 105 | #define VFIO_GROUP_FLAGS_CONTAINER_SET (1 << 1) |
| 106 | }; |
| 107 | #define VFIO_GROUP_GET_STATUS _IO(VFIO_TYPE, VFIO_BASE + 3) |
| 108 | |
| 109 | /** |
| 110 | * VFIO_GROUP_SET_CONTAINER - _IOW(VFIO_TYPE, VFIO_BASE + 4, __s32) |
| 111 | * |
| 112 | * Set the container for the VFIO group to the open VFIO file |
| 113 | * descriptor provided. Groups may only belong to a single |
| 114 | * container. Containers may, at their discretion, support multiple |
| 115 | * groups. Only when a container is set are all of the interfaces |
| 116 | * of the VFIO file descriptor and the VFIO group file descriptor |
| 117 | * available to the user. |
| 118 | * Return: 0 on success, -errno on failure. |
| 119 | * Availability: Always |
| 120 | */ |
| 121 | #define VFIO_GROUP_SET_CONTAINER _IO(VFIO_TYPE, VFIO_BASE + 4) |
| 122 | |
| 123 | /** |
| 124 | * VFIO_GROUP_UNSET_CONTAINER - _IO(VFIO_TYPE, VFIO_BASE + 5) |
| 125 | * |
| 126 | * Remove the group from the attached container. This is the |
| 127 | * opposite of the SET_CONTAINER call and returns the group to |
| 128 | * an initial state. All device file descriptors must be released |
| 129 | * prior to calling this interface. When removing the last group |
| 130 | * from a container, the IOMMU will be disabled and all state lost, |
| 131 | * effectively also returning the VFIO file descriptor to an initial |
| 132 | * state. |
| 133 | * Return: 0 on success, -errno on failure. |
| 134 | * Availability: When attached to container |
| 135 | */ |
| 136 | #define VFIO_GROUP_UNSET_CONTAINER _IO(VFIO_TYPE, VFIO_BASE + 5) |
| 137 | |
| 138 | /** |
| 139 | * VFIO_GROUP_GET_DEVICE_FD - _IOW(VFIO_TYPE, VFIO_BASE + 6, char) |
| 140 | * |
| 141 | * Return a new file descriptor for the device object described by |
| 142 | * the provided string. The string should match a device listed in |
| 143 | * the devices subdirectory of the IOMMU group sysfs entry. The |
| 144 | * group containing the device must already be added to this context. |
| 145 | * Return: new file descriptor on success, -errno on failure. |
| 146 | * Availability: When attached to container |
| 147 | */ |
| 148 | #define VFIO_GROUP_GET_DEVICE_FD _IO(VFIO_TYPE, VFIO_BASE + 6) |
| 149 | |
| 150 | /* --------------- IOCTLs for DEVICE file descriptors --------------- */ |
| 151 | |
| 152 | /** |
| 153 | * VFIO_DEVICE_GET_INFO - _IOR(VFIO_TYPE, VFIO_BASE + 7, |
| 154 | * struct vfio_device_info) |
| 155 | * |
| 156 | * Retrieve information about the device. Fills in provided |
| 157 | * struct vfio_device_info. Caller sets argsz. |
| 158 | * Return: 0 on success, -errno on failure. |
| 159 | */ |
| 160 | struct vfio_device_info { |
| 161 | __u32 argsz; |
| 162 | __u32 flags; |
| 163 | #define VFIO_DEVICE_FLAGS_RESET (1 << 0) /* Device supports reset */ |
| 164 | #define VFIO_DEVICE_FLAGS_PCI (1 << 1) /* vfio-pci device */ |
| 165 | #define VFIO_DEVICE_FLAGS_PLATFORM (1 << 2) /* vfio-platform device */ |
| 166 | #define VFIO_DEVICE_FLAGS_AMBA (1 << 3) /* vfio-amba device */ |
| 167 | __u32 num_regions; /* Max region index + 1 */ |
| 168 | __u32 num_irqs; /* Max IRQ index + 1 */ |
| 169 | }; |
| 170 | #define VFIO_DEVICE_GET_INFO _IO(VFIO_TYPE, VFIO_BASE + 7) |
| 171 | |
| 172 | /** |
| 173 | * VFIO_DEVICE_GET_REGION_INFO - _IOWR(VFIO_TYPE, VFIO_BASE + 8, |
| 174 | * struct vfio_region_info) |
| 175 | * |
| 176 | * Retrieve information about a device region. Caller provides |
| 177 | * struct vfio_region_info with index value set. Caller sets argsz. |
| 178 | * Implementation of region mapping is bus driver specific. This is |
| 179 | * intended to describe MMIO, I/O port, as well as bus specific |
| 180 | * regions (ex. PCI config space). Zero sized regions may be used |
| 181 | * to describe unimplemented regions (ex. unimplemented PCI BARs). |
| 182 | * Return: 0 on success, -errno on failure. |
| 183 | */ |
| 184 | struct vfio_region_info { |
| 185 | __u32 argsz; |
| 186 | __u32 flags; |
| 187 | #define VFIO_REGION_INFO_FLAG_READ (1 << 0) /* Region supports read */ |
| 188 | #define VFIO_REGION_INFO_FLAG_WRITE (1 << 1) /* Region supports write */ |
| 189 | #define VFIO_REGION_INFO_FLAG_MMAP (1 << 2) /* Region supports mmap */ |
| 190 | __u32 index; /* Region index */ |
| 191 | __u32 resv; /* Reserved for alignment */ |
| 192 | __u64 size; /* Region size (bytes) */ |
| 193 | __u64 offset; /* Region offset from start of device fd */ |
| 194 | }; |
| 195 | #define VFIO_DEVICE_GET_REGION_INFO _IO(VFIO_TYPE, VFIO_BASE + 8) |
| 196 | |
| 197 | /** |
| 198 | * VFIO_DEVICE_GET_IRQ_INFO - _IOWR(VFIO_TYPE, VFIO_BASE + 9, |
| 199 | * struct vfio_irq_info) |
| 200 | * |
| 201 | * Retrieve information about a device IRQ. Caller provides |
| 202 | * struct vfio_irq_info with index value set. Caller sets argsz. |
| 203 | * Implementation of IRQ mapping is bus driver specific. Indexes |
| 204 | * using multiple IRQs are primarily intended to support MSI-like |
| 205 | * interrupt blocks. Zero count irq blocks may be used to describe |
| 206 | * unimplemented interrupt types. |
| 207 | * |
| 208 | * The EVENTFD flag indicates the interrupt index supports eventfd based |
| 209 | * signaling. |
| 210 | * |
| 211 | * The MASKABLE flags indicates the index supports MASK and UNMASK |
| 212 | * actions described below. |
| 213 | * |
| 214 | * AUTOMASKED indicates that after signaling, the interrupt line is |
| 215 | * automatically masked by VFIO and the user needs to unmask the line |
| 216 | * to receive new interrupts. This is primarily intended to distinguish |
| 217 | * level triggered interrupts. |
| 218 | * |
| 219 | * The NORESIZE flag indicates that the interrupt lines within the index |
| 220 | * are setup as a set and new subindexes cannot be enabled without first |
| 221 | * disabling the entire index. This is used for interrupts like PCI MSI |
| 222 | * and MSI-X where the driver may only use a subset of the available |
| 223 | * indexes, but VFIO needs to enable a specific number of vectors |
| 224 | * upfront. In the case of MSI-X, where the user can enable MSI-X and |
| 225 | * then add and unmask vectors, it's up to userspace to make the decision |
| 226 | * whether to allocate the maximum supported number of vectors or tear |
| 227 | * down setup and incrementally increase the vectors as each is enabled. |
| 228 | */ |
| 229 | struct vfio_irq_info { |
| 230 | __u32 argsz; |
| 231 | __u32 flags; |
| 232 | #define VFIO_IRQ_INFO_EVENTFD (1 << 0) |
| 233 | #define VFIO_IRQ_INFO_MASKABLE (1 << 1) |
| 234 | #define VFIO_IRQ_INFO_AUTOMASKED (1 << 2) |
| 235 | #define VFIO_IRQ_INFO_NORESIZE (1 << 3) |
| 236 | __u32 index; /* IRQ index */ |
| 237 | __u32 count; /* Number of IRQs within this index */ |
| 238 | }; |
| 239 | #define VFIO_DEVICE_GET_IRQ_INFO _IO(VFIO_TYPE, VFIO_BASE + 9) |
| 240 | |
| 241 | /** |
| 242 | * VFIO_DEVICE_SET_IRQS - _IOW(VFIO_TYPE, VFIO_BASE + 10, struct vfio_irq_set) |
| 243 | * |
| 244 | * Set signaling, masking, and unmasking of interrupts. Caller provides |
| 245 | * struct vfio_irq_set with all fields set. 'start' and 'count' indicate |
| 246 | * the range of subindexes being specified. |
| 247 | * |
| 248 | * The DATA flags specify the type of data provided. If DATA_NONE, the |
| 249 | * operation performs the specified action immediately on the specified |
| 250 | * interrupt(s). For example, to unmask AUTOMASKED interrupt [0,0]: |
| 251 | * flags = (DATA_NONE|ACTION_UNMASK), index = 0, start = 0, count = 1. |
| 252 | * |
| 253 | * DATA_BOOL allows sparse support for the same on arrays of interrupts. |
| 254 | * For example, to mask interrupts [0,1] and [0,3] (but not [0,2]): |
| 255 | * flags = (DATA_BOOL|ACTION_MASK), index = 0, start = 1, count = 3, |
| 256 | * data = {1,0,1} |
| 257 | * |
| 258 | * DATA_EVENTFD binds the specified ACTION to the provided __s32 eventfd. |
| 259 | * A value of -1 can be used to either de-assign interrupts if already |
| 260 | * assigned or skip un-assigned interrupts. For example, to set an eventfd |
| 261 | * to be trigger for interrupts [0,0] and [0,2]: |
| 262 | * flags = (DATA_EVENTFD|ACTION_TRIGGER), index = 0, start = 0, count = 3, |
| 263 | * data = {fd1, -1, fd2} |
| 264 | * If index [0,1] is previously set, two count = 1 ioctls calls would be |
| 265 | * required to set [0,0] and [0,2] without changing [0,1]. |
| 266 | * |
| 267 | * Once a signaling mechanism is set, DATA_BOOL or DATA_NONE can be used |
| 268 | * with ACTION_TRIGGER to perform kernel level interrupt loopback testing |
| 269 | * from userspace (ie. simulate hardware triggering). |
| 270 | * |
| 271 | * Setting of an event triggering mechanism to userspace for ACTION_TRIGGER |
| 272 | * enables the interrupt index for the device. Individual subindex interrupts |
| 273 | * can be disabled using the -1 value for DATA_EVENTFD or the index can be |
| 274 | * disabled as a whole with: flags = (DATA_NONE|ACTION_TRIGGER), count = 0. |
| 275 | * |
| 276 | * Note that ACTION_[UN]MASK specify user->kernel signaling (irqfds) while |
| 277 | * ACTION_TRIGGER specifies kernel->user signaling. |
| 278 | */ |
| 279 | struct vfio_irq_set { |
| 280 | __u32 argsz; |
| 281 | __u32 flags; |
| 282 | #define VFIO_IRQ_SET_DATA_NONE (1 << 0) /* Data not present */ |
| 283 | #define VFIO_IRQ_SET_DATA_BOOL (1 << 1) /* Data is bool (u8) */ |
| 284 | #define VFIO_IRQ_SET_DATA_EVENTFD (1 << 2) /* Data is eventfd (s32) */ |
| 285 | #define VFIO_IRQ_SET_ACTION_MASK (1 << 3) /* Mask interrupt */ |
| 286 | #define VFIO_IRQ_SET_ACTION_UNMASK (1 << 4) /* Unmask interrupt */ |
| 287 | #define VFIO_IRQ_SET_ACTION_TRIGGER (1 << 5) /* Trigger interrupt */ |
| 288 | __u32 index; |
| 289 | __u32 start; |
| 290 | __u32 count; |
| 291 | __u8 data[]; |
| 292 | }; |
| 293 | #define VFIO_DEVICE_SET_IRQS _IO(VFIO_TYPE, VFIO_BASE + 10) |
| 294 | |
| 295 | #define VFIO_IRQ_SET_DATA_TYPE_MASK (VFIO_IRQ_SET_DATA_NONE | \ |
| 296 | VFIO_IRQ_SET_DATA_BOOL | \ |
| 297 | VFIO_IRQ_SET_DATA_EVENTFD) |
| 298 | #define VFIO_IRQ_SET_ACTION_TYPE_MASK (VFIO_IRQ_SET_ACTION_MASK | \ |
| 299 | VFIO_IRQ_SET_ACTION_UNMASK | \ |
| 300 | VFIO_IRQ_SET_ACTION_TRIGGER) |
| 301 | /** |
| 302 | * VFIO_DEVICE_RESET - _IO(VFIO_TYPE, VFIO_BASE + 11) |
| 303 | * |
| 304 | * Reset a device. |
| 305 | */ |
| 306 | #define VFIO_DEVICE_RESET _IO(VFIO_TYPE, VFIO_BASE + 11) |
| 307 | |
| 308 | /* |
| 309 | * The VFIO-PCI bus driver makes use of the following fixed region and |
| 310 | * IRQ index mapping. Unimplemented regions return a size of zero. |
| 311 | * Unimplemented IRQ types return a count of zero. |
| 312 | */ |
| 313 | |
| 314 | enum { |
| 315 | VFIO_PCI_BAR0_REGION_INDEX, |
| 316 | VFIO_PCI_BAR1_REGION_INDEX, |
| 317 | VFIO_PCI_BAR2_REGION_INDEX, |
| 318 | VFIO_PCI_BAR3_REGION_INDEX, |
| 319 | VFIO_PCI_BAR4_REGION_INDEX, |
| 320 | VFIO_PCI_BAR5_REGION_INDEX, |
| 321 | VFIO_PCI_ROM_REGION_INDEX, |
| 322 | VFIO_PCI_CONFIG_REGION_INDEX, |
| 323 | /* |
| 324 | * Expose VGA regions defined for PCI base class 03, subclass 00. |
| 325 | * This includes I/O port ranges 0x3b0 to 0x3bb and 0x3c0 to 0x3df |
| 326 | * as well as the MMIO range 0xa0000 to 0xbffff. Each implemented |
| 327 | * range is found at it's identity mapped offset from the region |
| 328 | * offset, for example 0x3b0 is region_info.offset + 0x3b0. Areas |
| 329 | * between described ranges are unimplemented. |
| 330 | */ |
| 331 | VFIO_PCI_VGA_REGION_INDEX, |
| 332 | VFIO_PCI_NUM_REGIONS |
| 333 | }; |
| 334 | |
| 335 | enum { |
| 336 | VFIO_PCI_INTX_IRQ_INDEX, |
| 337 | VFIO_PCI_MSI_IRQ_INDEX, |
| 338 | VFIO_PCI_MSIX_IRQ_INDEX, |
| 339 | VFIO_PCI_ERR_IRQ_INDEX, |
| 340 | VFIO_PCI_REQ_IRQ_INDEX, |
| 341 | VFIO_PCI_NUM_IRQS |
| 342 | }; |
| 343 | |
| 344 | /** |
| 345 | * VFIO_DEVICE_GET_PCI_HOT_RESET_INFO - _IORW(VFIO_TYPE, VFIO_BASE + 12, |
| 346 | * struct vfio_pci_hot_reset_info) |
| 347 | * |
| 348 | * Return: 0 on success, -errno on failure: |
| 349 | * -enospc = insufficient buffer, -enodev = unsupported for device. |
| 350 | */ |
| 351 | struct vfio_pci_dependent_device { |
| 352 | __u32 group_id; |
| 353 | __u16 segment; |
| 354 | __u8 bus; |
| 355 | __u8 devfn; /* Use PCI_SLOT/PCI_FUNC */ |
| 356 | }; |
| 357 | |
| 358 | struct vfio_pci_hot_reset_info { |
| 359 | __u32 argsz; |
| 360 | __u32 flags; |
| 361 | __u32 count; |
| 362 | struct vfio_pci_dependent_device devices[]; |
| 363 | }; |
| 364 | |
| 365 | #define VFIO_DEVICE_GET_PCI_HOT_RESET_INFO _IO(VFIO_TYPE, VFIO_BASE + 12) |
| 366 | |
| 367 | /** |
| 368 | * VFIO_DEVICE_PCI_HOT_RESET - _IOW(VFIO_TYPE, VFIO_BASE + 13, |
| 369 | * struct vfio_pci_hot_reset) |
| 370 | * |
| 371 | * Return: 0 on success, -errno on failure. |
| 372 | */ |
| 373 | struct vfio_pci_hot_reset { |
| 374 | __u32 argsz; |
| 375 | __u32 flags; |
| 376 | __u32 count; |
| 377 | __s32 group_fds[]; |
| 378 | }; |
| 379 | |
| 380 | #define VFIO_DEVICE_PCI_HOT_RESET _IO(VFIO_TYPE, VFIO_BASE + 13) |
| 381 | |
| 382 | /* -------- API for Type1 VFIO IOMMU -------- */ |
| 383 | |
| 384 | /** |
| 385 | * VFIO_IOMMU_GET_INFO - _IOR(VFIO_TYPE, VFIO_BASE + 12, struct vfio_iommu_info) |
| 386 | * |
| 387 | * Retrieve information about the IOMMU object. Fills in provided |
| 388 | * struct vfio_iommu_info. Caller sets argsz. |
| 389 | * |
| 390 | * XXX Should we do these by CHECK_EXTENSION too? |
| 391 | */ |
| 392 | struct vfio_iommu_type1_info { |
| 393 | __u32 argsz; |
| 394 | __u32 flags; |
| 395 | #define VFIO_IOMMU_INFO_PGSIZES (1 << 0) /* supported page sizes info */ |
| 396 | __u64 iova_pgsizes; /* Bitmap of supported page sizes */ |
| 397 | }; |
| 398 | |
| 399 | #define VFIO_IOMMU_GET_INFO _IO(VFIO_TYPE, VFIO_BASE + 12) |
| 400 | |
| 401 | /** |
| 402 | * VFIO_IOMMU_MAP_DMA - _IOW(VFIO_TYPE, VFIO_BASE + 13, struct vfio_dma_map) |
| 403 | * |
| 404 | * Map process virtual addresses to IO virtual addresses using the |
| 405 | * provided struct vfio_dma_map. Caller sets argsz. READ &/ WRITE required. |
| 406 | */ |
| 407 | struct vfio_iommu_type1_dma_map { |
| 408 | __u32 argsz; |
| 409 | __u32 flags; |
| 410 | #define VFIO_DMA_MAP_FLAG_READ (1 << 0) /* readable from device */ |
| 411 | #define VFIO_DMA_MAP_FLAG_WRITE (1 << 1) /* writable from device */ |
| 412 | __u64 vaddr; /* Process virtual address */ |
| 413 | __u64 iova; /* IO virtual address */ |
| 414 | __u64 size; /* Size of mapping (bytes) */ |
| 415 | }; |
| 416 | |
| 417 | #define VFIO_IOMMU_MAP_DMA _IO(VFIO_TYPE, VFIO_BASE + 13) |
| 418 | |
| 419 | /** |
| 420 | * VFIO_IOMMU_UNMAP_DMA - _IOWR(VFIO_TYPE, VFIO_BASE + 14, |
| 421 | * struct vfio_dma_unmap) |
| 422 | * |
| 423 | * Unmap IO virtual addresses using the provided struct vfio_dma_unmap. |
| 424 | * Caller sets argsz. The actual unmapped size is returned in the size |
| 425 | * field. No guarantee is made to the user that arbitrary unmaps of iova |
| 426 | * or size different from those used in the original mapping call will |
| 427 | * succeed. |
| 428 | */ |
| 429 | struct vfio_iommu_type1_dma_unmap { |
| 430 | __u32 argsz; |
| 431 | __u32 flags; |
| 432 | __u64 iova; /* IO virtual address */ |
| 433 | __u64 size; /* Size of mapping (bytes) */ |
| 434 | }; |
| 435 | |
| 436 | #define VFIO_IOMMU_UNMAP_DMA _IO(VFIO_TYPE, VFIO_BASE + 14) |
| 437 | |
| 438 | /* |
| 439 | * IOCTLs to enable/disable IOMMU container usage. |
| 440 | * No parameters are supported. |
| 441 | */ |
| 442 | #define VFIO_IOMMU_ENABLE _IO(VFIO_TYPE, VFIO_BASE + 15) |
| 443 | #define VFIO_IOMMU_DISABLE _IO(VFIO_TYPE, VFIO_BASE + 16) |
| 444 | |
| 445 | /* -------- Additional API for SPAPR TCE (Server POWERPC) IOMMU -------- */ |
| 446 | |
| 447 | /* |
| 448 | * The SPAPR TCE DDW info struct provides the information about |
| 449 | * the details of Dynamic DMA window capability. |
| 450 | * |
| 451 | * @pgsizes contains a page size bitmask, 4K/64K/16M are supported. |
| 452 | * @max_dynamic_windows_supported tells the maximum number of windows |
| 453 | * which the platform can create. |
| 454 | * @levels tells the maximum number of levels in multi-level IOMMU tables; |
| 455 | * this allows splitting a table into smaller chunks which reduces |
| 456 | * the amount of physically contiguous memory required for the table. |
| 457 | */ |
| 458 | struct vfio_iommu_spapr_tce_ddw_info { |
| 459 | __u64 pgsizes; /* Bitmap of supported page sizes */ |
| 460 | __u32 max_dynamic_windows_supported; |
| 461 | __u32 levels; |
| 462 | }; |
| 463 | |
| 464 | /* |
| 465 | * The SPAPR TCE info struct provides the information about the PCI bus |
| 466 | * address ranges available for DMA, these values are programmed into |
| 467 | * the hardware so the guest has to know that information. |
| 468 | * |
| 469 | * The DMA 32 bit window start is an absolute PCI bus address. |
| 470 | * The IOVA address passed via map/unmap ioctls are absolute PCI bus |
| 471 | * addresses too so the window works as a filter rather than an offset |
| 472 | * for IOVA addresses. |
| 473 | * |
| 474 | * Flags supported: |
| 475 | * - VFIO_IOMMU_SPAPR_INFO_DDW: informs the userspace that dynamic DMA windows |
| 476 | * (DDW) support is present. @ddw is only supported when DDW is present. |
| 477 | */ |
| 478 | struct vfio_iommu_spapr_tce_info { |
| 479 | __u32 argsz; |
| 480 | __u32 flags; |
| 481 | #define VFIO_IOMMU_SPAPR_INFO_DDW (1 << 0) /* DDW supported */ |
| 482 | __u32 dma32_window_start; /* 32 bit window start (bytes) */ |
| 483 | __u32 dma32_window_size; /* 32 bit window size (bytes) */ |
| 484 | struct vfio_iommu_spapr_tce_ddw_info ddw; |
| 485 | }; |
| 486 | |
| 487 | #define VFIO_IOMMU_SPAPR_TCE_GET_INFO _IO(VFIO_TYPE, VFIO_BASE + 12) |
| 488 | |
| 489 | /* |
| 490 | * EEH PE operation struct provides ways to: |
| 491 | * - enable/disable EEH functionality; |
| 492 | * - unfreeze IO/DMA for frozen PE; |
| 493 | * - read PE state; |
| 494 | * - reset PE; |
| 495 | * - configure PE; |
| 496 | * - inject EEH error. |
| 497 | */ |
| 498 | struct vfio_eeh_pe_err { |
| 499 | __u32 type; |
| 500 | __u32 func; |
| 501 | __u64 addr; |
| 502 | __u64 mask; |
| 503 | }; |
| 504 | |
| 505 | struct vfio_eeh_pe_op { |
| 506 | __u32 argsz; |
| 507 | __u32 flags; |
| 508 | __u32 op; |
| 509 | union { |
| 510 | struct vfio_eeh_pe_err err; |
| 511 | }; |
| 512 | }; |
| 513 | |
| 514 | #define VFIO_EEH_PE_DISABLE 0 /* Disable EEH functionality */ |
| 515 | #define VFIO_EEH_PE_ENABLE 1 /* Enable EEH functionality */ |
| 516 | #define VFIO_EEH_PE_UNFREEZE_IO 2 /* Enable IO for frozen PE */ |
| 517 | #define VFIO_EEH_PE_UNFREEZE_DMA 3 /* Enable DMA for frozen PE */ |
| 518 | #define VFIO_EEH_PE_GET_STATE 4 /* PE state retrieval */ |
| 519 | #define VFIO_EEH_PE_STATE_NORMAL 0 /* PE in functional state */ |
| 520 | #define VFIO_EEH_PE_STATE_RESET 1 /* PE reset in progress */ |
| 521 | #define VFIO_EEH_PE_STATE_STOPPED 2 /* Stopped DMA and IO */ |
| 522 | #define VFIO_EEH_PE_STATE_STOPPED_DMA 4 /* Stopped DMA only */ |
| 523 | #define VFIO_EEH_PE_STATE_UNAVAIL 5 /* State unavailable */ |
| 524 | #define VFIO_EEH_PE_RESET_DEACTIVATE 5 /* Deassert PE reset */ |
| 525 | #define VFIO_EEH_PE_RESET_HOT 6 /* Assert hot reset */ |
| 526 | #define VFIO_EEH_PE_RESET_FUNDAMENTAL 7 /* Assert fundamental reset */ |
| 527 | #define VFIO_EEH_PE_CONFIGURE 8 /* PE configuration */ |
| 528 | #define VFIO_EEH_PE_INJECT_ERR 9 /* Inject EEH error */ |
| 529 | |
| 530 | #define VFIO_EEH_PE_OP _IO(VFIO_TYPE, VFIO_BASE + 21) |
| 531 | |
| 532 | /** |
| 533 | * VFIO_IOMMU_SPAPR_REGISTER_MEMORY - _IOW(VFIO_TYPE, VFIO_BASE + 17, struct vfio_iommu_spapr_register_memory) |
| 534 | * |
| 535 | * Registers user space memory where DMA is allowed. It pins |
| 536 | * user pages and does the locked memory accounting so |
| 537 | * subsequent VFIO_IOMMU_MAP_DMA/VFIO_IOMMU_UNMAP_DMA calls |
| 538 | * get faster. |
| 539 | */ |
| 540 | struct vfio_iommu_spapr_register_memory { |
| 541 | __u32 argsz; |
| 542 | __u32 flags; |
| 543 | __u64 vaddr; /* Process virtual address */ |
| 544 | __u64 size; /* Size of mapping (bytes) */ |
| 545 | }; |
| 546 | #define VFIO_IOMMU_SPAPR_REGISTER_MEMORY _IO(VFIO_TYPE, VFIO_BASE + 17) |
| 547 | |
| 548 | /** |
| 549 | * VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - _IOW(VFIO_TYPE, VFIO_BASE + 18, struct vfio_iommu_spapr_register_memory) |
| 550 | * |
| 551 | * Unregisters user space memory registered with |
| 552 | * VFIO_IOMMU_SPAPR_REGISTER_MEMORY. |
| 553 | * Uses vfio_iommu_spapr_register_memory for parameters. |
| 554 | */ |
| 555 | #define VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY _IO(VFIO_TYPE, VFIO_BASE + 18) |
| 556 | |
| 557 | /** |
| 558 | * VFIO_IOMMU_SPAPR_TCE_CREATE - _IOWR(VFIO_TYPE, VFIO_BASE + 19, struct vfio_iommu_spapr_tce_create) |
| 559 | * |
| 560 | * Creates an additional TCE table and programs it (sets a new DMA window) |
| 561 | * to every IOMMU group in the container. It receives page shift, window |
| 562 | * size and number of levels in the TCE table being created. |
| 563 | * |
| 564 | * It allocates and returns an offset on a PCI bus of the new DMA window. |
| 565 | */ |
| 566 | struct vfio_iommu_spapr_tce_create { |
| 567 | __u32 argsz; |
| 568 | __u32 flags; |
| 569 | /* in */ |
| 570 | __u32 page_shift; |
| 571 | __u64 window_size; |
| 572 | __u32 levels; |
| 573 | /* out */ |
| 574 | __u64 start_addr; |
| 575 | }; |
| 576 | #define VFIO_IOMMU_SPAPR_TCE_CREATE _IO(VFIO_TYPE, VFIO_BASE + 19) |
| 577 | |
| 578 | /** |
| 579 | * VFIO_IOMMU_SPAPR_TCE_REMOVE - _IOW(VFIO_TYPE, VFIO_BASE + 20, struct vfio_iommu_spapr_tce_remove) |
| 580 | * |
| 581 | * Unprograms a TCE table from all groups in the container and destroys it. |
| 582 | * It receives a PCI bus offset as a window id. |
| 583 | */ |
| 584 | struct vfio_iommu_spapr_tce_remove { |
| 585 | __u32 argsz; |
| 586 | __u32 flags; |
| 587 | /* in */ |
| 588 | __u64 start_addr; |
| 589 | }; |
| 590 | #define VFIO_IOMMU_SPAPR_TCE_REMOVE _IO(VFIO_TYPE, VFIO_BASE + 20) |
| 591 | |
| 592 | /* ***************************************************************** */ |
| 593 | |
| 594 | #endif /* _UAPIVFIO_H */ |