1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
5 * Copyright (C) 2012 Red Hat, Inc. All rights reserved.
6 * Author: Alex Williamson <alex.williamson@redhat.com>
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.
15 #include <linux/types.h>
16 #include <linux/ioctl.h>
18 #define VFIO_API_VERSION 0
21 /* Kernel & User level defines for VFIO IOCTLs. */
25 #define VFIO_TYPE1_IOMMU 1
26 #define VFIO_SPAPR_TCE_IOMMU 2
27 #define VFIO_TYPE1v2_IOMMU 3
29 * IOMMU enforces DMA cache coherence (ex. PCIe NoSnoop stripping). This
30 * capability is subject to change as groups are added or removed.
32 #define VFIO_DMA_CC_IOMMU 4
34 /* Check if EEH is supported */
38 #define VFIO_TYPE1_NESTING_IOMMU 6 /* Implies v2 */
40 #define VFIO_SPAPR_TCE_v2_IOMMU 7
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.
47 #define VFIO_NOIOMMU_IOMMU 8
49 /* Supports VFIO_DMA_UNMAP_FLAG_ALL */
50 #define VFIO_UNMAP_ALL 9
52 /* Supports the vaddr flag for DMA map and unmap */
53 #define VFIO_UPDATE_VADDR 10
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.
66 #define VFIO_TYPE (';')
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.
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 */
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.
96 /* -------- IOCTLs for VFIO file descriptor (/dev/vfio/vfio) -------- */
99 * VFIO_GET_API_VERSION - _IO(VFIO_TYPE, VFIO_BASE + 0)
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
104 * Return: VFIO_API_VERSION
105 * Availability: Always
107 #define VFIO_GET_API_VERSION _IO(VFIO_TYPE, VFIO_BASE + 0)
110 * VFIO_CHECK_EXTENSION - _IOW(VFIO_TYPE, VFIO_BASE + 1, __u32)
112 * Check whether an extension is supported.
113 * Return: 0 if not supported, 1 (or some other positive integer) if supported.
114 * Availability: Always
116 #define VFIO_CHECK_EXTENSION _IO(VFIO_TYPE, VFIO_BASE + 1)
119 * VFIO_SET_IOMMU - _IOW(VFIO_TYPE, VFIO_BASE + 2, __s32)
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
129 #define VFIO_SET_IOMMU _IO(VFIO_TYPE, VFIO_BASE + 2)
131 /* -------- IOCTLs for GROUP file descriptors (/dev/vfio/$GROUP) -------- */
134 * VFIO_GROUP_GET_STATUS - _IOR(VFIO_TYPE, VFIO_BASE + 3,
135 * struct vfio_group_status)
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
142 struct vfio_group_status {
145 #define VFIO_GROUP_FLAGS_VIABLE (1 << 0)
146 #define VFIO_GROUP_FLAGS_CONTAINER_SET (1 << 1)
148 #define VFIO_GROUP_GET_STATUS _IO(VFIO_TYPE, VFIO_BASE + 3)
151 * VFIO_GROUP_SET_CONTAINER - _IOW(VFIO_TYPE, VFIO_BASE + 4, __s32)
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
162 #define VFIO_GROUP_SET_CONTAINER _IO(VFIO_TYPE, VFIO_BASE + 4)
165 * VFIO_GROUP_UNSET_CONTAINER - _IO(VFIO_TYPE, VFIO_BASE + 5)
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
174 * Return: 0 on success, -errno on failure.
175 * Availability: When attached to container
177 #define VFIO_GROUP_UNSET_CONTAINER _IO(VFIO_TYPE, VFIO_BASE + 5)
180 * VFIO_GROUP_GET_DEVICE_FD - _IOW(VFIO_TYPE, VFIO_BASE + 6, char)
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
189 #define VFIO_GROUP_GET_DEVICE_FD _IO(VFIO_TYPE, VFIO_BASE + 6)
191 /* --------------- IOCTLs for DEVICE file descriptors --------------- */
194 * VFIO_DEVICE_GET_INFO - _IOR(VFIO_TYPE, VFIO_BASE + 7,
195 * struct vfio_device_info)
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.
201 struct vfio_device_info {
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 */
216 #define VFIO_DEVICE_GET_INFO _IO(VFIO_TYPE, VFIO_BASE + 7)
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.
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"
231 * The following capabilities are unique to s390 zPCI devices. Their contents
232 * are further-defined in vfio_zdev.h
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
240 * VFIO_DEVICE_GET_REGION_INFO - _IOWR(VFIO_TYPE, VFIO_BASE + 8,
241 * struct vfio_region_info)
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.
251 struct vfio_region_info {
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 */
263 #define VFIO_DEVICE_GET_REGION_INFO _IO(VFIO_TYPE, VFIO_BASE + 8)
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.
272 * The structures below define version 1 of this capability.
274 #define VFIO_REGION_INFO_CAP_SPARSE_MMAP 1
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 */
281 struct vfio_region_info_cap_sparse_mmap {
282 struct vfio_info_cap_header header;
285 struct vfio_region_sparse_mmap_area areas[];
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.
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.
306 * The structure below defines version 1 of this capability.
308 #define VFIO_REGION_INFO_CAP_TYPE 2
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 */
317 * List of region types, global per bus driver.
318 * If you introduce a new type, please add it here.
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_DEPRECATED (3)
328 /* sub-types for VFIO_REGION_TYPE_PCI_* */
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)
335 /* 10de vendor PCI sub-types */
337 * NVIDIA GPU NVlink2 RAM is coherent RAM mapped onto the host address space.
339 * Deprecated, region no longer provided
341 #define VFIO_REGION_SUBTYPE_NVIDIA_NVLINK2_RAM (1)
343 /* 1014 vendor PCI sub-types */
345 * IBM NPU NVlink2 ATSD (Address Translation Shootdown) register of NPU
346 * to do TLB invalidation on a GPU.
348 * Deprecated, region no longer provided
350 #define VFIO_REGION_SUBTYPE_IBM_NVLINK2_ATSD (1)
352 /* sub-types for VFIO_REGION_TYPE_GFX */
353 #define VFIO_REGION_SUBTYPE_GFX_EDID (1)
356 * struct vfio_region_gfx_edid - EDID region layout.
358 * Set display link state and EDID blob.
360 * The EDID blob has monitor information such as brand, name, serial
361 * number, physical size, supported video modes and more.
363 * This special region allows userspace (typically qemu) set a virtual
364 * EDID for the virtual monitor, which allows a flexible display
367 * For the edid blob spec look here:
368 * https://en.wikipedia.org/wiki/Extended_Display_Identification_Data
370 * On linux systems you can find the EDID blob in sysfs:
371 * /sys/class/drm/${card}/${connector}/edid
373 * You can use the edid-decode ulility (comes with xorg-x11-utils) to
374 * decode the EDID blob.
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).
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.
391 struct vfio_region_gfx_edid {
398 #define VFIO_DEVICE_GFX_LINK_STATE_UP 1
399 #define VFIO_DEVICE_GFX_LINK_STATE_DOWN 2
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)
407 /* sub-types for VFIO_REGION_TYPE_MIGRATION */
408 #define VFIO_REGION_SUBTYPE_MIGRATION_DEPRECATED (1)
410 struct vfio_device_migration_info {
411 __u32 device_state; /* VFIO device state */
412 #define VFIO_DEVICE_STATE_V1_STOP (0)
413 #define VFIO_DEVICE_STATE_V1_RUNNING (1 << 0)
414 #define VFIO_DEVICE_STATE_V1_SAVING (1 << 1)
415 #define VFIO_DEVICE_STATE_V1_RESUMING (1 << 2)
416 #define VFIO_DEVICE_STATE_MASK (VFIO_DEVICE_STATE_V1_RUNNING | \
417 VFIO_DEVICE_STATE_V1_SAVING | \
418 VFIO_DEVICE_STATE_V1_RESUMING)
420 #define VFIO_DEVICE_STATE_VALID(state) \
421 (state & VFIO_DEVICE_STATE_V1_RESUMING ? \
422 (state & VFIO_DEVICE_STATE_MASK) == VFIO_DEVICE_STATE_V1_RESUMING : 1)
424 #define VFIO_DEVICE_STATE_IS_ERROR(state) \
425 ((state & VFIO_DEVICE_STATE_MASK) == (VFIO_DEVICE_STATE_V1_SAVING | \
426 VFIO_DEVICE_STATE_V1_RESUMING))
428 #define VFIO_DEVICE_STATE_SET_ERROR(state) \
429 ((state & ~VFIO_DEVICE_STATE_MASK) | VFIO_DEVICE_STATE_V1_SAVING | \
430 VFIO_DEVICE_STATE_V1_RESUMING)
439 * The MSIX mappable capability informs that MSIX data of a BAR can be mmapped
440 * which allows direct access to non-MSIX registers which happened to be within
441 * the same system page.
443 * Even though the userspace gets direct access to the MSIX data, the existing
444 * VFIO_DEVICE_SET_IRQS interface must still be used for MSIX configuration.
446 #define VFIO_REGION_INFO_CAP_MSIX_MAPPABLE 3
449 * Capability with compressed real address (aka SSA - small system address)
450 * where GPU RAM is mapped on a system bus. Used by a GPU for DMA routing
451 * and by the userspace to associate a NVLink bridge with a GPU.
453 * Deprecated, capability no longer provided
455 #define VFIO_REGION_INFO_CAP_NVLINK2_SSATGT 4
457 struct vfio_region_info_cap_nvlink2_ssatgt {
458 struct vfio_info_cap_header header;
463 * Capability with an NVLink link speed. The value is read by
464 * the NVlink2 bridge driver from the bridge's "ibm,nvlink-speed"
465 * property in the device tree. The value is fixed in the hardware
466 * and failing to provide the correct value results in the link
467 * not working with no indication from the driver why.
469 * Deprecated, capability no longer provided
471 #define VFIO_REGION_INFO_CAP_NVLINK2_LNKSPD 5
473 struct vfio_region_info_cap_nvlink2_lnkspd {
474 struct vfio_info_cap_header header;
480 * VFIO_DEVICE_GET_IRQ_INFO - _IOWR(VFIO_TYPE, VFIO_BASE + 9,
481 * struct vfio_irq_info)
483 * Retrieve information about a device IRQ. Caller provides
484 * struct vfio_irq_info with index value set. Caller sets argsz.
485 * Implementation of IRQ mapping is bus driver specific. Indexes
486 * using multiple IRQs are primarily intended to support MSI-like
487 * interrupt blocks. Zero count irq blocks may be used to describe
488 * unimplemented interrupt types.
490 * The EVENTFD flag indicates the interrupt index supports eventfd based
493 * The MASKABLE flags indicates the index supports MASK and UNMASK
494 * actions described below.
496 * AUTOMASKED indicates that after signaling, the interrupt line is
497 * automatically masked by VFIO and the user needs to unmask the line
498 * to receive new interrupts. This is primarily intended to distinguish
499 * level triggered interrupts.
501 * The NORESIZE flag indicates that the interrupt lines within the index
502 * are setup as a set and new subindexes cannot be enabled without first
503 * disabling the entire index. This is used for interrupts like PCI MSI
504 * and MSI-X where the driver may only use a subset of the available
505 * indexes, but VFIO needs to enable a specific number of vectors
506 * upfront. In the case of MSI-X, where the user can enable MSI-X and
507 * then add and unmask vectors, it's up to userspace to make the decision
508 * whether to allocate the maximum supported number of vectors or tear
509 * down setup and incrementally increase the vectors as each is enabled.
511 struct vfio_irq_info {
514 #define VFIO_IRQ_INFO_EVENTFD (1 << 0)
515 #define VFIO_IRQ_INFO_MASKABLE (1 << 1)
516 #define VFIO_IRQ_INFO_AUTOMASKED (1 << 2)
517 #define VFIO_IRQ_INFO_NORESIZE (1 << 3)
518 __u32 index; /* IRQ index */
519 __u32 count; /* Number of IRQs within this index */
521 #define VFIO_DEVICE_GET_IRQ_INFO _IO(VFIO_TYPE, VFIO_BASE + 9)
524 * VFIO_DEVICE_SET_IRQS - _IOW(VFIO_TYPE, VFIO_BASE + 10, struct vfio_irq_set)
526 * Set signaling, masking, and unmasking of interrupts. Caller provides
527 * struct vfio_irq_set with all fields set. 'start' and 'count' indicate
528 * the range of subindexes being specified.
530 * The DATA flags specify the type of data provided. If DATA_NONE, the
531 * operation performs the specified action immediately on the specified
532 * interrupt(s). For example, to unmask AUTOMASKED interrupt [0,0]:
533 * flags = (DATA_NONE|ACTION_UNMASK), index = 0, start = 0, count = 1.
535 * DATA_BOOL allows sparse support for the same on arrays of interrupts.
536 * For example, to mask interrupts [0,1] and [0,3] (but not [0,2]):
537 * flags = (DATA_BOOL|ACTION_MASK), index = 0, start = 1, count = 3,
540 * DATA_EVENTFD binds the specified ACTION to the provided __s32 eventfd.
541 * A value of -1 can be used to either de-assign interrupts if already
542 * assigned or skip un-assigned interrupts. For example, to set an eventfd
543 * to be trigger for interrupts [0,0] and [0,2]:
544 * flags = (DATA_EVENTFD|ACTION_TRIGGER), index = 0, start = 0, count = 3,
545 * data = {fd1, -1, fd2}
546 * If index [0,1] is previously set, two count = 1 ioctls calls would be
547 * required to set [0,0] and [0,2] without changing [0,1].
549 * Once a signaling mechanism is set, DATA_BOOL or DATA_NONE can be used
550 * with ACTION_TRIGGER to perform kernel level interrupt loopback testing
551 * from userspace (ie. simulate hardware triggering).
553 * Setting of an event triggering mechanism to userspace for ACTION_TRIGGER
554 * enables the interrupt index for the device. Individual subindex interrupts
555 * can be disabled using the -1 value for DATA_EVENTFD or the index can be
556 * disabled as a whole with: flags = (DATA_NONE|ACTION_TRIGGER), count = 0.
558 * Note that ACTION_[UN]MASK specify user->kernel signaling (irqfds) while
559 * ACTION_TRIGGER specifies kernel->user signaling.
561 struct vfio_irq_set {
564 #define VFIO_IRQ_SET_DATA_NONE (1 << 0) /* Data not present */
565 #define VFIO_IRQ_SET_DATA_BOOL (1 << 1) /* Data is bool (u8) */
566 #define VFIO_IRQ_SET_DATA_EVENTFD (1 << 2) /* Data is eventfd (s32) */
567 #define VFIO_IRQ_SET_ACTION_MASK (1 << 3) /* Mask interrupt */
568 #define VFIO_IRQ_SET_ACTION_UNMASK (1 << 4) /* Unmask interrupt */
569 #define VFIO_IRQ_SET_ACTION_TRIGGER (1 << 5) /* Trigger interrupt */
575 #define VFIO_DEVICE_SET_IRQS _IO(VFIO_TYPE, VFIO_BASE + 10)
577 #define VFIO_IRQ_SET_DATA_TYPE_MASK (VFIO_IRQ_SET_DATA_NONE | \
578 VFIO_IRQ_SET_DATA_BOOL | \
579 VFIO_IRQ_SET_DATA_EVENTFD)
580 #define VFIO_IRQ_SET_ACTION_TYPE_MASK (VFIO_IRQ_SET_ACTION_MASK | \
581 VFIO_IRQ_SET_ACTION_UNMASK | \
582 VFIO_IRQ_SET_ACTION_TRIGGER)
584 * VFIO_DEVICE_RESET - _IO(VFIO_TYPE, VFIO_BASE + 11)
588 #define VFIO_DEVICE_RESET _IO(VFIO_TYPE, VFIO_BASE + 11)
591 * The VFIO-PCI bus driver makes use of the following fixed region and
592 * IRQ index mapping. Unimplemented regions return a size of zero.
593 * Unimplemented IRQ types return a count of zero.
597 VFIO_PCI_BAR0_REGION_INDEX,
598 VFIO_PCI_BAR1_REGION_INDEX,
599 VFIO_PCI_BAR2_REGION_INDEX,
600 VFIO_PCI_BAR3_REGION_INDEX,
601 VFIO_PCI_BAR4_REGION_INDEX,
602 VFIO_PCI_BAR5_REGION_INDEX,
603 VFIO_PCI_ROM_REGION_INDEX,
604 VFIO_PCI_CONFIG_REGION_INDEX,
606 * Expose VGA regions defined for PCI base class 03, subclass 00.
607 * This includes I/O port ranges 0x3b0 to 0x3bb and 0x3c0 to 0x3df
608 * as well as the MMIO range 0xa0000 to 0xbffff. Each implemented
609 * range is found at it's identity mapped offset from the region
610 * offset, for example 0x3b0 is region_info.offset + 0x3b0. Areas
611 * between described ranges are unimplemented.
613 VFIO_PCI_VGA_REGION_INDEX,
614 VFIO_PCI_NUM_REGIONS = 9 /* Fixed user ABI, region indexes >=9 use */
615 /* device specific cap to define content. */
619 VFIO_PCI_INTX_IRQ_INDEX,
620 VFIO_PCI_MSI_IRQ_INDEX,
621 VFIO_PCI_MSIX_IRQ_INDEX,
622 VFIO_PCI_ERR_IRQ_INDEX,
623 VFIO_PCI_REQ_IRQ_INDEX,
628 * The vfio-ccw bus driver makes use of the following fixed region and
629 * IRQ index mapping. Unimplemented regions return a size of zero.
630 * Unimplemented IRQ types return a count of zero.
634 VFIO_CCW_CONFIG_REGION_INDEX,
639 VFIO_CCW_IO_IRQ_INDEX,
640 VFIO_CCW_CRW_IRQ_INDEX,
641 VFIO_CCW_REQ_IRQ_INDEX,
646 * VFIO_DEVICE_GET_PCI_HOT_RESET_INFO - _IOWR(VFIO_TYPE, VFIO_BASE + 12,
647 * struct vfio_pci_hot_reset_info)
649 * Return: 0 on success, -errno on failure:
650 * -enospc = insufficient buffer, -enodev = unsupported for device.
652 struct vfio_pci_dependent_device {
656 __u8 devfn; /* Use PCI_SLOT/PCI_FUNC */
659 struct vfio_pci_hot_reset_info {
663 struct vfio_pci_dependent_device devices[];
666 #define VFIO_DEVICE_GET_PCI_HOT_RESET_INFO _IO(VFIO_TYPE, VFIO_BASE + 12)
669 * VFIO_DEVICE_PCI_HOT_RESET - _IOW(VFIO_TYPE, VFIO_BASE + 13,
670 * struct vfio_pci_hot_reset)
672 * Return: 0 on success, -errno on failure.
674 struct vfio_pci_hot_reset {
681 #define VFIO_DEVICE_PCI_HOT_RESET _IO(VFIO_TYPE, VFIO_BASE + 13)
684 * VFIO_DEVICE_QUERY_GFX_PLANE - _IOW(VFIO_TYPE, VFIO_BASE + 14,
685 * struct vfio_device_query_gfx_plane)
687 * Set the drm_plane_type and flags, then retrieve the gfx plane info.
690 * - VFIO_GFX_PLANE_TYPE_PROBE and VFIO_GFX_PLANE_TYPE_DMABUF are set
691 * to ask if the mdev supports dma-buf. 0 on support, -EINVAL on no
692 * support for dma-buf.
693 * - VFIO_GFX_PLANE_TYPE_PROBE and VFIO_GFX_PLANE_TYPE_REGION are set
694 * to ask if the mdev supports region. 0 on support, -EINVAL on no
695 * support for region.
696 * - VFIO_GFX_PLANE_TYPE_DMABUF or VFIO_GFX_PLANE_TYPE_REGION is set
697 * with each call to query the plane info.
698 * - Others are invalid and return -EINVAL.
701 * 1. Plane could be disabled by guest. In that case, success will be
702 * returned with zero-initialized drm_format, size, width and height
704 * 2. x_hot/y_hot is set to 0xFFFFFFFF if no hotspot information available
706 * Return: 0 on success, -errno on other failure.
708 struct vfio_device_gfx_plane_info {
711 #define VFIO_GFX_PLANE_TYPE_PROBE (1 << 0)
712 #define VFIO_GFX_PLANE_TYPE_DMABUF (1 << 1)
713 #define VFIO_GFX_PLANE_TYPE_REGION (1 << 2)
715 __u32 drm_plane_type; /* type of plane: DRM_PLANE_TYPE_* */
717 __u32 drm_format; /* drm format of plane */
718 __u64 drm_format_mod; /* tiled mode */
719 __u32 width; /* width of plane */
720 __u32 height; /* height of plane */
721 __u32 stride; /* stride of plane */
722 __u32 size; /* size of plane in bytes, align on page*/
723 __u32 x_pos; /* horizontal position of cursor plane */
724 __u32 y_pos; /* vertical position of cursor plane*/
725 __u32 x_hot; /* horizontal position of cursor hotspot */
726 __u32 y_hot; /* vertical position of cursor hotspot */
728 __u32 region_index; /* region index */
729 __u32 dmabuf_id; /* dma-buf id */
733 #define VFIO_DEVICE_QUERY_GFX_PLANE _IO(VFIO_TYPE, VFIO_BASE + 14)
736 * VFIO_DEVICE_GET_GFX_DMABUF - _IOW(VFIO_TYPE, VFIO_BASE + 15, __u32)
738 * Return a new dma-buf file descriptor for an exposed guest framebuffer
739 * described by the provided dmabuf_id. The dmabuf_id is returned from VFIO_
740 * DEVICE_QUERY_GFX_PLANE as a token of the exposed guest framebuffer.
743 #define VFIO_DEVICE_GET_GFX_DMABUF _IO(VFIO_TYPE, VFIO_BASE + 15)
746 * VFIO_DEVICE_IOEVENTFD - _IOW(VFIO_TYPE, VFIO_BASE + 16,
747 * struct vfio_device_ioeventfd)
749 * Perform a write to the device at the specified device fd offset, with
750 * the specified data and width when the provided eventfd is triggered.
751 * vfio bus drivers may not support this for all regions, for all widths,
752 * or at all. vfio-pci currently only enables support for BAR regions,
753 * excluding the MSI-X vector table.
755 * Return: 0 on success, -errno on failure.
757 struct vfio_device_ioeventfd {
760 #define VFIO_DEVICE_IOEVENTFD_8 (1 << 0) /* 1-byte write */
761 #define VFIO_DEVICE_IOEVENTFD_16 (1 << 1) /* 2-byte write */
762 #define VFIO_DEVICE_IOEVENTFD_32 (1 << 2) /* 4-byte write */
763 #define VFIO_DEVICE_IOEVENTFD_64 (1 << 3) /* 8-byte write */
764 #define VFIO_DEVICE_IOEVENTFD_SIZE_MASK (0xf)
765 __u64 offset; /* device fd offset of write */
766 __u64 data; /* data to be written */
767 __s32 fd; /* -1 for de-assignment */
770 #define VFIO_DEVICE_IOEVENTFD _IO(VFIO_TYPE, VFIO_BASE + 16)
773 * VFIO_DEVICE_FEATURE - _IOWR(VFIO_TYPE, VFIO_BASE + 17,
774 * struct vfio_device_feature)
776 * Get, set, or probe feature data of the device. The feature is selected
777 * using the FEATURE_MASK portion of the flags field. Support for a feature
778 * can be probed by setting both the FEATURE_MASK and PROBE bits. A probe
779 * may optionally include the GET and/or SET bits to determine read vs write
780 * access of the feature respectively. Probing a feature will return success
781 * if the feature is supported and all of the optionally indicated GET/SET
782 * methods are supported. The format of the data portion of the structure is
783 * specific to the given feature. The data portion is not required for
784 * probing. GET and SET are mutually exclusive, except for use with PROBE.
786 * Return 0 on success, -errno on failure.
788 struct vfio_device_feature {
791 #define VFIO_DEVICE_FEATURE_MASK (0xffff) /* 16-bit feature index */
792 #define VFIO_DEVICE_FEATURE_GET (1 << 16) /* Get feature into data[] */
793 #define VFIO_DEVICE_FEATURE_SET (1 << 17) /* Set feature from data[] */
794 #define VFIO_DEVICE_FEATURE_PROBE (1 << 18) /* Probe feature support */
798 #define VFIO_DEVICE_FEATURE _IO(VFIO_TYPE, VFIO_BASE + 17)
801 * Provide support for setting a PCI VF Token, which is used as a shared
802 * secret between PF and VF drivers. This feature may only be set on a
803 * PCI SR-IOV PF when SR-IOV is enabled on the PF and there are no existing
804 * open VFs. Data provided when setting this feature is a 16-byte array
805 * (__u8 b[16]), representing a UUID.
807 #define VFIO_DEVICE_FEATURE_PCI_VF_TOKEN (0)
810 * Indicates the device can support the migration API through
811 * VFIO_DEVICE_FEATURE_MIG_DEVICE_STATE. If this GET succeeds, the RUNNING and
812 * ERROR states are always supported. Support for additional states is
813 * indicated via the flags field; at least VFIO_MIGRATION_STOP_COPY must be
816 * VFIO_MIGRATION_STOP_COPY means that STOP, STOP_COPY and
817 * RESUMING are supported.
819 * VFIO_MIGRATION_STOP_COPY | VFIO_MIGRATION_P2P means that RUNNING_P2P
820 * is supported in addition to the STOP_COPY states.
822 * Other combinations of flags have behavior to be defined in the future.
824 struct vfio_device_feature_migration {
826 #define VFIO_MIGRATION_STOP_COPY (1 << 0)
827 #define VFIO_MIGRATION_P2P (1 << 1)
829 #define VFIO_DEVICE_FEATURE_MIGRATION 1
832 * Upon VFIO_DEVICE_FEATURE_SET, execute a migration state change on the VFIO
833 * device. The new state is supplied in device_state, see enum
834 * vfio_device_mig_state for details
836 * The kernel migration driver must fully transition the device to the new state
837 * value before the operation returns to the user.
839 * The kernel migration driver must not generate asynchronous device state
840 * transitions outside of manipulation by the user or the VFIO_DEVICE_RESET
841 * ioctl as described above.
843 * If this function fails then current device_state may be the original
844 * operating state or some other state along the combination transition path.
845 * The user can then decide if it should execute a VFIO_DEVICE_RESET, attempt
846 * to return to the original state, or attempt to return to some other state
847 * such as RUNNING or STOP.
849 * If the new_state starts a new data transfer session then the FD associated
850 * with that session is returned in data_fd. The user is responsible to close
851 * this FD when it is finished. The user must consider the migration data stream
852 * carried over the FD to be opaque and must preserve the byte order of the
853 * stream. The user is not required to preserve buffer segmentation when writing
854 * the data stream during the RESUMING operation.
856 * Upon VFIO_DEVICE_FEATURE_GET, get the current migration state of the VFIO
857 * device, data_fd will be -1.
859 struct vfio_device_feature_mig_state {
860 __u32 device_state; /* From enum vfio_device_mig_state */
863 #define VFIO_DEVICE_FEATURE_MIG_DEVICE_STATE 2
866 * The device migration Finite State Machine is described by the enum
867 * vfio_device_mig_state. Some of the FSM arcs will create a migration data
868 * transfer session by returning a FD, in this case the migration data will
869 * flow over the FD using read() and write() as discussed below.
871 * There are 5 states to support VFIO_MIGRATION_STOP_COPY:
872 * RUNNING - The device is running normally
873 * STOP - The device does not change the internal or external state
874 * STOP_COPY - The device internal state can be read out
875 * RESUMING - The device is stopped and is loading a new internal state
876 * ERROR - The device has failed and must be reset
878 * And 1 optional state to support VFIO_MIGRATION_P2P:
879 * RUNNING_P2P - RUNNING, except the device cannot do peer to peer DMA
881 * The FSM takes actions on the arcs between FSM states. The driver implements
882 * the following behavior for the FSM arcs:
884 * RUNNING_P2P -> STOP
886 * While in STOP the device must stop the operation of the device. The device
887 * must not generate interrupts, DMA, or any other change to external state.
888 * It must not change its internal state. When stopped the device and kernel
889 * migration driver must accept and respond to interaction to support external
890 * subsystems in the STOP state, for example PCI MSI-X and PCI config space.
891 * Failure by the user to restrict device access while in STOP must not result
892 * in error conditions outside the user context (ex. host system faults).
894 * The STOP_COPY arc will terminate a data transfer session.
897 * Leaving RESUMING terminates a data transfer session and indicates the
898 * device should complete processing of the data delivered by write(). The
899 * kernel migration driver should complete the incorporation of data written
900 * to the data transfer FD into the device internal state and perform
901 * final validity and consistency checking of the new device state. If the
902 * user provided data is found to be incomplete, inconsistent, or otherwise
903 * invalid, the migration driver must fail the SET_STATE ioctl and
904 * optionally go to the ERROR state as described below.
906 * While in STOP the device has the same behavior as other STOP states
909 * To abort a RESUMING session the device must be reset.
911 * RUNNING_P2P -> RUNNING
912 * While in RUNNING the device is fully operational, the device may generate
913 * interrupts, DMA, respond to MMIO, all vfio device regions are functional,
914 * and the device may advance its internal state.
916 * RUNNING -> RUNNING_P2P
917 * STOP -> RUNNING_P2P
918 * While in RUNNING_P2P the device is partially running in the P2P quiescent
919 * state defined below.
922 * This arc begin the process of saving the device state and will return a
925 * While in the STOP_COPY state the device has the same behavior as STOP
926 * with the addition that the data transfers session continues to stream the
927 * migration state. End of stream on the FD indicates the entire device
928 * state has been transferred.
930 * The user should take steps to restrict access to vfio device regions while
931 * the device is in STOP_COPY or risk corruption of the device migration data
935 * Entering the RESUMING state starts a process of restoring the device state
936 * and will return a new data_fd. The data stream fed into the data_fd should
937 * be taken from the data transfer output of a single FD during saving from
938 * a compatible device. The migration driver may alter/reset the internal
939 * device state for this arc if required to prepare the device to receive the
943 * ERROR cannot be specified as a device state, however any transition request
944 * can be failed with an errno return and may then move the device_state into
945 * ERROR. In this case the device was unable to execute the requested arc and
946 * was also unable to restore the device to any valid device_state.
947 * To recover from ERROR VFIO_DEVICE_RESET must be used to return the
948 * device_state back to RUNNING.
950 * The optional peer to peer (P2P) quiescent state is intended to be a quiescent
951 * state for the device for the purposes of managing multiple devices within a
952 * user context where peer-to-peer DMA between devices may be active. The
953 * RUNNING_P2P states must prevent the device from initiating
954 * any new P2P DMA transactions. If the device can identify P2P transactions
955 * then it can stop only P2P DMA, otherwise it must stop all DMA. The migration
956 * driver must complete any such outstanding operations prior to completing the
957 * FSM arc into a P2P state. For the purpose of specification the states
958 * behave as though the device was fully running if not supported. Like while in
959 * STOP or STOP_COPY the user must not touch the device, otherwise the state
962 * The remaining possible transitions are interpreted as combinations of the
963 * above FSM arcs. As there are multiple paths through the FSM arcs the path
964 * should be selected based on the following rules:
965 * - Select the shortest path.
966 * Refer to vfio_mig_get_next_state() for the result of the algorithm.
968 * The automatic transit through the FSM arcs that make up the combination
969 * transition is invisible to the user. When working with combination arcs the
970 * user may see any step along the path in the device_state if SET_STATE
971 * fails. When handling these types of errors users should anticipate future
972 * revisions of this protocol using new states and those states becoming
973 * visible in this case.
975 * The optional states cannot be used with SET_STATE if the device does not
976 * support them. The user can discover if these states are supported by using
977 * VFIO_DEVICE_FEATURE_MIGRATION. By using combination transitions the user can
978 * avoid knowing about these optional states if the kernel driver supports them.
980 enum vfio_device_mig_state {
981 VFIO_DEVICE_STATE_ERROR = 0,
982 VFIO_DEVICE_STATE_STOP = 1,
983 VFIO_DEVICE_STATE_RUNNING = 2,
984 VFIO_DEVICE_STATE_STOP_COPY = 3,
985 VFIO_DEVICE_STATE_RESUMING = 4,
986 VFIO_DEVICE_STATE_RUNNING_P2P = 5,
990 * Upon VFIO_DEVICE_FEATURE_SET, allow the device to be moved into a low power
991 * state with the platform-based power management. Device use of lower power
992 * states depends on factors managed by the runtime power management core,
993 * including system level support and coordinating support among dependent
994 * devices. Enabling device low power entry does not guarantee lower power
995 * usage by the device, nor is a mechanism provided through this feature to
996 * know the current power state of the device. If any device access happens
997 * (either from the host or through the vfio uAPI) when the device is in the
998 * low power state, then the host will move the device out of the low power
999 * state as necessary prior to the access. Once the access is completed, the
1000 * device may re-enter the low power state. For single shot low power support
1001 * with wake-up notification, see
1002 * VFIO_DEVICE_FEATURE_LOW_POWER_ENTRY_WITH_WAKEUP below. Access to mmap'd
1003 * device regions is disabled on LOW_POWER_ENTRY and may only be resumed after
1004 * calling LOW_POWER_EXIT.
1006 #define VFIO_DEVICE_FEATURE_LOW_POWER_ENTRY 3
1009 * This device feature has the same behavior as
1010 * VFIO_DEVICE_FEATURE_LOW_POWER_ENTRY with the exception that the user
1011 * provides an eventfd for wake-up notification. When the device moves out of
1012 * the low power state for the wake-up, the host will not allow the device to
1013 * re-enter a low power state without a subsequent user call to one of the low
1014 * power entry device feature IOCTLs. Access to mmap'd device regions is
1015 * disabled on LOW_POWER_ENTRY_WITH_WAKEUP and may only be resumed after the
1016 * low power exit. The low power exit can happen either through LOW_POWER_EXIT
1017 * or through any other access (where the wake-up notification has been
1018 * generated). The access to mmap'd device regions will not trigger low power
1021 * The notification through the provided eventfd will be generated only when
1022 * the device has entered and is resumed from a low power state after
1023 * calling this device feature IOCTL. A device that has not entered low power
1024 * state, as managed through the runtime power management core, will not
1025 * generate a notification through the provided eventfd on access. Calling the
1026 * LOW_POWER_EXIT feature is optional in the case where notification has been
1027 * signaled on the provided eventfd that a resume from low power has occurred.
1029 struct vfio_device_low_power_entry_with_wakeup {
1030 __s32 wakeup_eventfd;
1034 #define VFIO_DEVICE_FEATURE_LOW_POWER_ENTRY_WITH_WAKEUP 4
1037 * Upon VFIO_DEVICE_FEATURE_SET, disallow use of device low power states as
1038 * previously enabled via VFIO_DEVICE_FEATURE_LOW_POWER_ENTRY or
1039 * VFIO_DEVICE_FEATURE_LOW_POWER_ENTRY_WITH_WAKEUP device features.
1040 * This device feature IOCTL may itself generate a wakeup eventfd notification
1041 * in the latter case if the device had previously entered a low power state.
1043 #define VFIO_DEVICE_FEATURE_LOW_POWER_EXIT 5
1046 * Upon VFIO_DEVICE_FEATURE_SET start/stop device DMA logging.
1047 * VFIO_DEVICE_FEATURE_PROBE can be used to detect if the device supports
1050 * DMA logging allows a device to internally record what DMAs the device is
1051 * initiating and report them back to userspace. It is part of the VFIO
1052 * migration infrastructure that allows implementing dirty page tracking
1053 * during the pre copy phase of live migration. Only DMA WRITEs are logged,
1054 * and this API is not connected to VFIO_DEVICE_FEATURE_MIG_DEVICE_STATE.
1056 * When DMA logging is started a range of IOVAs to monitor is provided and the
1057 * device can optimize its logging to cover only the IOVA range given. Each
1058 * DMA that the device initiates inside the range will be logged by the device
1059 * for later retrieval.
1061 * page_size is an input that hints what tracking granularity the device
1062 * should try to achieve. If the device cannot do the hinted page size then
1063 * it's the driver choice which page size to pick based on its support.
1064 * On output the device will return the page size it selected.
1066 * ranges is a pointer to an array of
1067 * struct vfio_device_feature_dma_logging_range.
1069 * The core kernel code guarantees to support by minimum num_ranges that fit
1070 * into a single kernel page. User space can try higher values but should give
1071 * up if the above can't be achieved as of some driver limitations.
1073 * A single call to start device DMA logging can be issued and a matching stop
1074 * should follow at the end. Another start is not allowed in the meantime.
1076 struct vfio_device_feature_dma_logging_control {
1077 __aligned_u64 page_size;
1080 __aligned_u64 ranges;
1083 struct vfio_device_feature_dma_logging_range {
1085 __aligned_u64 length;
1088 #define VFIO_DEVICE_FEATURE_DMA_LOGGING_START 6
1091 * Upon VFIO_DEVICE_FEATURE_SET stop device DMA logging that was started
1092 * by VFIO_DEVICE_FEATURE_DMA_LOGGING_START
1094 #define VFIO_DEVICE_FEATURE_DMA_LOGGING_STOP 7
1097 * Upon VFIO_DEVICE_FEATURE_GET read back and clear the device DMA log
1099 * Query the device's DMA log for written pages within the given IOVA range.
1100 * During querying the log is cleared for the IOVA range.
1102 * bitmap is a pointer to an array of u64s that will hold the output bitmap
1103 * with 1 bit reporting a page_size unit of IOVA. The mapping of IOVA to bits
1105 * bitmap[(addr - iova)/page_size] & (1ULL << (addr % 64))
1107 * The input page_size can be any power of two value and does not have to
1108 * match the value given to VFIO_DEVICE_FEATURE_DMA_LOGGING_START. The driver
1109 * will format its internal logging to match the reporting page size, possibly
1110 * by replicating bits if the internal page size is lower than requested.
1112 * The LOGGING_REPORT will only set bits in the bitmap and never clear or
1113 * perform any initialization of the user provided bitmap.
1115 * If any error is returned userspace should assume that the dirty log is
1116 * corrupted. Error recovery is to consider all memory dirty and try to
1117 * restart the dirty tracking, or to abort/restart the whole migration.
1119 * If DMA logging is not enabled, an error will be returned.
1122 struct vfio_device_feature_dma_logging_report {
1124 __aligned_u64 length;
1125 __aligned_u64 page_size;
1126 __aligned_u64 bitmap;
1129 #define VFIO_DEVICE_FEATURE_DMA_LOGGING_REPORT 8
1131 /* -------- API for Type1 VFIO IOMMU -------- */
1134 * VFIO_IOMMU_GET_INFO - _IOR(VFIO_TYPE, VFIO_BASE + 12, struct vfio_iommu_info)
1136 * Retrieve information about the IOMMU object. Fills in provided
1137 * struct vfio_iommu_info. Caller sets argsz.
1139 * XXX Should we do these by CHECK_EXTENSION too?
1141 struct vfio_iommu_type1_info {
1144 #define VFIO_IOMMU_INFO_PGSIZES (1 << 0) /* supported page sizes info */
1145 #define VFIO_IOMMU_INFO_CAPS (1 << 1) /* Info supports caps */
1146 __u64 iova_pgsizes; /* Bitmap of supported page sizes */
1147 __u32 cap_offset; /* Offset within info struct of first cap */
1151 * The IOVA capability allows to report the valid IOVA range(s)
1152 * excluding any non-relaxable reserved regions exposed by
1153 * devices attached to the container. Any DMA map attempt
1154 * outside the valid iova range will return error.
1156 * The structures below define version 1 of this capability.
1158 #define VFIO_IOMMU_TYPE1_INFO_CAP_IOVA_RANGE 1
1160 struct vfio_iova_range {
1165 struct vfio_iommu_type1_info_cap_iova_range {
1166 struct vfio_info_cap_header header;
1169 struct vfio_iova_range iova_ranges[];
1173 * The migration capability allows to report supported features for migration.
1175 * The structures below define version 1 of this capability.
1177 * The existence of this capability indicates that IOMMU kernel driver supports
1178 * dirty page logging.
1180 * pgsize_bitmap: Kernel driver returns bitmap of supported page sizes for dirty
1182 * max_dirty_bitmap_size: Kernel driver returns maximum supported dirty bitmap
1183 * size in bytes that can be used by user applications when getting the dirty
1186 #define VFIO_IOMMU_TYPE1_INFO_CAP_MIGRATION 2
1188 struct vfio_iommu_type1_info_cap_migration {
1189 struct vfio_info_cap_header header;
1191 __u64 pgsize_bitmap;
1192 __u64 max_dirty_bitmap_size; /* in bytes */
1196 * The DMA available capability allows to report the current number of
1197 * simultaneously outstanding DMA mappings that are allowed.
1199 * The structure below defines version 1 of this capability.
1201 * avail: specifies the current number of outstanding DMA mappings allowed.
1203 #define VFIO_IOMMU_TYPE1_INFO_DMA_AVAIL 3
1205 struct vfio_iommu_type1_info_dma_avail {
1206 struct vfio_info_cap_header header;
1210 #define VFIO_IOMMU_GET_INFO _IO(VFIO_TYPE, VFIO_BASE + 12)
1213 * VFIO_IOMMU_MAP_DMA - _IOW(VFIO_TYPE, VFIO_BASE + 13, struct vfio_dma_map)
1215 * Map process virtual addresses to IO virtual addresses using the
1216 * provided struct vfio_dma_map. Caller sets argsz. READ &/ WRITE required.
1218 * If flags & VFIO_DMA_MAP_FLAG_VADDR, update the base vaddr for iova, and
1219 * unblock translation of host virtual addresses in the iova range. The vaddr
1220 * must have previously been invalidated with VFIO_DMA_UNMAP_FLAG_VADDR. To
1221 * maintain memory consistency within the user application, the updated vaddr
1222 * must address the same memory object as originally mapped. Failure to do so
1223 * will result in user memory corruption and/or device misbehavior. iova and
1224 * size must match those in the original MAP_DMA call. Protection is not
1225 * changed, and the READ & WRITE flags must be 0.
1227 struct vfio_iommu_type1_dma_map {
1230 #define VFIO_DMA_MAP_FLAG_READ (1 << 0) /* readable from device */
1231 #define VFIO_DMA_MAP_FLAG_WRITE (1 << 1) /* writable from device */
1232 #define VFIO_DMA_MAP_FLAG_VADDR (1 << 2)
1233 __u64 vaddr; /* Process virtual address */
1234 __u64 iova; /* IO virtual address */
1235 __u64 size; /* Size of mapping (bytes) */
1238 #define VFIO_IOMMU_MAP_DMA _IO(VFIO_TYPE, VFIO_BASE + 13)
1240 struct vfio_bitmap {
1241 __u64 pgsize; /* page size for bitmap in bytes */
1242 __u64 size; /* in bytes */
1243 __u64 __user *data; /* one bit per page */
1247 * VFIO_IOMMU_UNMAP_DMA - _IOWR(VFIO_TYPE, VFIO_BASE + 14,
1248 * struct vfio_dma_unmap)
1250 * Unmap IO virtual addresses using the provided struct vfio_dma_unmap.
1251 * Caller sets argsz. The actual unmapped size is returned in the size
1252 * field. No guarantee is made to the user that arbitrary unmaps of iova
1253 * or size different from those used in the original mapping call will
1256 * VFIO_DMA_UNMAP_FLAG_GET_DIRTY_BITMAP should be set to get the dirty bitmap
1257 * before unmapping IO virtual addresses. When this flag is set, the user must
1258 * provide a struct vfio_bitmap in data[]. User must provide zero-allocated
1259 * memory via vfio_bitmap.data and its size in the vfio_bitmap.size field.
1260 * A bit in the bitmap represents one page, of user provided page size in
1261 * vfio_bitmap.pgsize field, consecutively starting from iova offset. Bit set
1262 * indicates that the page at that offset from iova is dirty. A Bitmap of the
1263 * pages in the range of unmapped size is returned in the user-provided
1266 * If flags & VFIO_DMA_UNMAP_FLAG_ALL, unmap all addresses. iova and size
1267 * must be 0. This cannot be combined with the get-dirty-bitmap flag.
1269 * If flags & VFIO_DMA_UNMAP_FLAG_VADDR, do not unmap, but invalidate host
1270 * virtual addresses in the iova range. Tasks that attempt to translate an
1271 * iova's vaddr will block. DMA to already-mapped pages continues. This
1272 * cannot be combined with the get-dirty-bitmap flag.
1274 struct vfio_iommu_type1_dma_unmap {
1277 #define VFIO_DMA_UNMAP_FLAG_GET_DIRTY_BITMAP (1 << 0)
1278 #define VFIO_DMA_UNMAP_FLAG_ALL (1 << 1)
1279 #define VFIO_DMA_UNMAP_FLAG_VADDR (1 << 2)
1280 __u64 iova; /* IO virtual address */
1281 __u64 size; /* Size of mapping (bytes) */
1285 #define VFIO_IOMMU_UNMAP_DMA _IO(VFIO_TYPE, VFIO_BASE + 14)
1288 * IOCTLs to enable/disable IOMMU container usage.
1289 * No parameters are supported.
1291 #define VFIO_IOMMU_ENABLE _IO(VFIO_TYPE, VFIO_BASE + 15)
1292 #define VFIO_IOMMU_DISABLE _IO(VFIO_TYPE, VFIO_BASE + 16)
1295 * VFIO_IOMMU_DIRTY_PAGES - _IOWR(VFIO_TYPE, VFIO_BASE + 17,
1296 * struct vfio_iommu_type1_dirty_bitmap)
1297 * IOCTL is used for dirty pages logging.
1298 * Caller should set flag depending on which operation to perform, details as
1301 * Calling the IOCTL with VFIO_IOMMU_DIRTY_PAGES_FLAG_START flag set, instructs
1302 * the IOMMU driver to log pages that are dirtied or potentially dirtied by
1303 * the device; designed to be used when a migration is in progress. Dirty pages
1304 * are logged until logging is disabled by user application by calling the IOCTL
1305 * with VFIO_IOMMU_DIRTY_PAGES_FLAG_STOP flag.
1307 * Calling the IOCTL with VFIO_IOMMU_DIRTY_PAGES_FLAG_STOP flag set, instructs
1308 * the IOMMU driver to stop logging dirtied pages.
1310 * Calling the IOCTL with VFIO_IOMMU_DIRTY_PAGES_FLAG_GET_BITMAP flag set
1311 * returns the dirty pages bitmap for IOMMU container for a given IOVA range.
1312 * The user must specify the IOVA range and the pgsize through the structure
1313 * vfio_iommu_type1_dirty_bitmap_get in the data[] portion. This interface
1314 * supports getting a bitmap of the smallest supported pgsize only and can be
1315 * modified in future to get a bitmap of any specified supported pgsize. The
1316 * user must provide a zeroed memory area for the bitmap memory and specify its
1317 * size in bitmap.size. One bit is used to represent one page consecutively
1318 * starting from iova offset. The user should provide page size in bitmap.pgsize
1319 * field. A bit set in the bitmap indicates that the page at that offset from
1320 * iova is dirty. The caller must set argsz to a value including the size of
1321 * structure vfio_iommu_type1_dirty_bitmap_get, but excluding the size of the
1322 * actual bitmap. If dirty pages logging is not enabled, an error will be
1325 * Only one of the flags _START, _STOP and _GET may be specified at a time.
1328 struct vfio_iommu_type1_dirty_bitmap {
1331 #define VFIO_IOMMU_DIRTY_PAGES_FLAG_START (1 << 0)
1332 #define VFIO_IOMMU_DIRTY_PAGES_FLAG_STOP (1 << 1)
1333 #define VFIO_IOMMU_DIRTY_PAGES_FLAG_GET_BITMAP (1 << 2)
1337 struct vfio_iommu_type1_dirty_bitmap_get {
1338 __u64 iova; /* IO virtual address */
1339 __u64 size; /* Size of iova range */
1340 struct vfio_bitmap bitmap;
1343 #define VFIO_IOMMU_DIRTY_PAGES _IO(VFIO_TYPE, VFIO_BASE + 17)
1345 /* -------- Additional API for SPAPR TCE (Server POWERPC) IOMMU -------- */
1348 * The SPAPR TCE DDW info struct provides the information about
1349 * the details of Dynamic DMA window capability.
1351 * @pgsizes contains a page size bitmask, 4K/64K/16M are supported.
1352 * @max_dynamic_windows_supported tells the maximum number of windows
1353 * which the platform can create.
1354 * @levels tells the maximum number of levels in multi-level IOMMU tables;
1355 * this allows splitting a table into smaller chunks which reduces
1356 * the amount of physically contiguous memory required for the table.
1358 struct vfio_iommu_spapr_tce_ddw_info {
1359 __u64 pgsizes; /* Bitmap of supported page sizes */
1360 __u32 max_dynamic_windows_supported;
1365 * The SPAPR TCE info struct provides the information about the PCI bus
1366 * address ranges available for DMA, these values are programmed into
1367 * the hardware so the guest has to know that information.
1369 * The DMA 32 bit window start is an absolute PCI bus address.
1370 * The IOVA address passed via map/unmap ioctls are absolute PCI bus
1371 * addresses too so the window works as a filter rather than an offset
1372 * for IOVA addresses.
1375 * - VFIO_IOMMU_SPAPR_INFO_DDW: informs the userspace that dynamic DMA windows
1376 * (DDW) support is present. @ddw is only supported when DDW is present.
1378 struct vfio_iommu_spapr_tce_info {
1381 #define VFIO_IOMMU_SPAPR_INFO_DDW (1 << 0) /* DDW supported */
1382 __u32 dma32_window_start; /* 32 bit window start (bytes) */
1383 __u32 dma32_window_size; /* 32 bit window size (bytes) */
1384 struct vfio_iommu_spapr_tce_ddw_info ddw;
1387 #define VFIO_IOMMU_SPAPR_TCE_GET_INFO _IO(VFIO_TYPE, VFIO_BASE + 12)
1390 * EEH PE operation struct provides ways to:
1391 * - enable/disable EEH functionality;
1392 * - unfreeze IO/DMA for frozen PE;
1396 * - inject EEH error.
1398 struct vfio_eeh_pe_err {
1405 struct vfio_eeh_pe_op {
1410 struct vfio_eeh_pe_err err;
1414 #define VFIO_EEH_PE_DISABLE 0 /* Disable EEH functionality */
1415 #define VFIO_EEH_PE_ENABLE 1 /* Enable EEH functionality */
1416 #define VFIO_EEH_PE_UNFREEZE_IO 2 /* Enable IO for frozen PE */
1417 #define VFIO_EEH_PE_UNFREEZE_DMA 3 /* Enable DMA for frozen PE */
1418 #define VFIO_EEH_PE_GET_STATE 4 /* PE state retrieval */
1419 #define VFIO_EEH_PE_STATE_NORMAL 0 /* PE in functional state */
1420 #define VFIO_EEH_PE_STATE_RESET 1 /* PE reset in progress */
1421 #define VFIO_EEH_PE_STATE_STOPPED 2 /* Stopped DMA and IO */
1422 #define VFIO_EEH_PE_STATE_STOPPED_DMA 4 /* Stopped DMA only */
1423 #define VFIO_EEH_PE_STATE_UNAVAIL 5 /* State unavailable */
1424 #define VFIO_EEH_PE_RESET_DEACTIVATE 5 /* Deassert PE reset */
1425 #define VFIO_EEH_PE_RESET_HOT 6 /* Assert hot reset */
1426 #define VFIO_EEH_PE_RESET_FUNDAMENTAL 7 /* Assert fundamental reset */
1427 #define VFIO_EEH_PE_CONFIGURE 8 /* PE configuration */
1428 #define VFIO_EEH_PE_INJECT_ERR 9 /* Inject EEH error */
1430 #define VFIO_EEH_PE_OP _IO(VFIO_TYPE, VFIO_BASE + 21)
1433 * VFIO_IOMMU_SPAPR_REGISTER_MEMORY - _IOW(VFIO_TYPE, VFIO_BASE + 17, struct vfio_iommu_spapr_register_memory)
1435 * Registers user space memory where DMA is allowed. It pins
1436 * user pages and does the locked memory accounting so
1437 * subsequent VFIO_IOMMU_MAP_DMA/VFIO_IOMMU_UNMAP_DMA calls
1440 struct vfio_iommu_spapr_register_memory {
1443 __u64 vaddr; /* Process virtual address */
1444 __u64 size; /* Size of mapping (bytes) */
1446 #define VFIO_IOMMU_SPAPR_REGISTER_MEMORY _IO(VFIO_TYPE, VFIO_BASE + 17)
1449 * VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - _IOW(VFIO_TYPE, VFIO_BASE + 18, struct vfio_iommu_spapr_register_memory)
1451 * Unregisters user space memory registered with
1452 * VFIO_IOMMU_SPAPR_REGISTER_MEMORY.
1453 * Uses vfio_iommu_spapr_register_memory for parameters.
1455 #define VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY _IO(VFIO_TYPE, VFIO_BASE + 18)
1458 * VFIO_IOMMU_SPAPR_TCE_CREATE - _IOWR(VFIO_TYPE, VFIO_BASE + 19, struct vfio_iommu_spapr_tce_create)
1460 * Creates an additional TCE table and programs it (sets a new DMA window)
1461 * to every IOMMU group in the container. It receives page shift, window
1462 * size and number of levels in the TCE table being created.
1464 * It allocates and returns an offset on a PCI bus of the new DMA window.
1466 struct vfio_iommu_spapr_tce_create {
1478 #define VFIO_IOMMU_SPAPR_TCE_CREATE _IO(VFIO_TYPE, VFIO_BASE + 19)
1481 * VFIO_IOMMU_SPAPR_TCE_REMOVE - _IOW(VFIO_TYPE, VFIO_BASE + 20, struct vfio_iommu_spapr_tce_remove)
1483 * Unprograms a TCE table from all groups in the container and destroys it.
1484 * It receives a PCI bus offset as a window id.
1486 struct vfio_iommu_spapr_tce_remove {
1492 #define VFIO_IOMMU_SPAPR_TCE_REMOVE _IO(VFIO_TYPE, VFIO_BASE + 20)
1494 /* ***************************************************************** */
1496 #endif /* _UAPIVFIO_H */