xref: /openbmc/qemu/linux-headers/linux/vfio.h (revision 55bbc861)
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 /*
50  * The IOCTL interface is designed for extensibility by embedding the
51  * structure length (argsz) and flags into structures passed between
52  * kernel and userspace.  We therefore use the _IO() macro for these
53  * defines to avoid implicitly embedding a size into the ioctl request.
54  * As structure fields are added, argsz will increase to match and flag
55  * bits will be defined to indicate additional fields with valid data.
56  * It's *always* the caller's responsibility to indicate the size of
57  * the structure passed by setting argsz appropriately.
58  */
59 
60 #define VFIO_TYPE	(';')
61 #define VFIO_BASE	100
62 
63 /*
64  * For extension of INFO ioctls, VFIO makes use of a capability chain
65  * designed after PCI/e capabilities.  A flag bit indicates whether
66  * this capability chain is supported and a field defined in the fixed
67  * structure defines the offset of the first capability in the chain.
68  * This field is only valid when the corresponding bit in the flags
69  * bitmap is set.  This offset field is relative to the start of the
70  * INFO buffer, as is the next field within each capability header.
71  * The id within the header is a shared address space per INFO ioctl,
72  * while the version field is specific to the capability id.  The
73  * contents following the header are specific to the capability id.
74  */
75 struct vfio_info_cap_header {
76 	__u16	id;		/* Identifies capability */
77 	__u16	version;	/* Version specific to the capability ID */
78 	__u32	next;		/* Offset of next capability */
79 };
80 
81 /*
82  * Callers of INFO ioctls passing insufficiently sized buffers will see
83  * the capability chain flag bit set, a zero value for the first capability
84  * offset (if available within the provided argsz), and argsz will be
85  * updated to report the necessary buffer size.  For compatibility, the
86  * INFO ioctl will not report error in this case, but the capability chain
87  * will not be available.
88  */
89 
90 /* -------- IOCTLs for VFIO file descriptor (/dev/vfio/vfio) -------- */
91 
92 /**
93  * VFIO_GET_API_VERSION - _IO(VFIO_TYPE, VFIO_BASE + 0)
94  *
95  * Report the version of the VFIO API.  This allows us to bump the entire
96  * API version should we later need to add or change features in incompatible
97  * ways.
98  * Return: VFIO_API_VERSION
99  * Availability: Always
100  */
101 #define VFIO_GET_API_VERSION		_IO(VFIO_TYPE, VFIO_BASE + 0)
102 
103 /**
104  * VFIO_CHECK_EXTENSION - _IOW(VFIO_TYPE, VFIO_BASE + 1, __u32)
105  *
106  * Check whether an extension is supported.
107  * Return: 0 if not supported, 1 (or some other positive integer) if supported.
108  * Availability: Always
109  */
110 #define VFIO_CHECK_EXTENSION		_IO(VFIO_TYPE, VFIO_BASE + 1)
111 
112 /**
113  * VFIO_SET_IOMMU - _IOW(VFIO_TYPE, VFIO_BASE + 2, __s32)
114  *
115  * Set the iommu to the given type.  The type must be supported by an
116  * iommu driver as verified by calling CHECK_EXTENSION using the same
117  * type.  A group must be set to this file descriptor before this
118  * ioctl is available.  The IOMMU interfaces enabled by this call are
119  * specific to the value set.
120  * Return: 0 on success, -errno on failure
121  * Availability: When VFIO group attached
122  */
123 #define VFIO_SET_IOMMU			_IO(VFIO_TYPE, VFIO_BASE + 2)
124 
125 /* -------- IOCTLs for GROUP file descriptors (/dev/vfio/$GROUP) -------- */
126 
127 /**
128  * VFIO_GROUP_GET_STATUS - _IOR(VFIO_TYPE, VFIO_BASE + 3,
129  *						struct vfio_group_status)
130  *
131  * Retrieve information about the group.  Fills in provided
132  * struct vfio_group_info.  Caller sets argsz.
133  * Return: 0 on succes, -errno on failure.
134  * Availability: Always
135  */
136 struct vfio_group_status {
137 	__u32	argsz;
138 	__u32	flags;
139 #define VFIO_GROUP_FLAGS_VIABLE		(1 << 0)
140 #define VFIO_GROUP_FLAGS_CONTAINER_SET	(1 << 1)
141 };
142 #define VFIO_GROUP_GET_STATUS		_IO(VFIO_TYPE, VFIO_BASE + 3)
143 
144 /**
145  * VFIO_GROUP_SET_CONTAINER - _IOW(VFIO_TYPE, VFIO_BASE + 4, __s32)
146  *
147  * Set the container for the VFIO group to the open VFIO file
148  * descriptor provided.  Groups may only belong to a single
149  * container.  Containers may, at their discretion, support multiple
150  * groups.  Only when a container is set are all of the interfaces
151  * of the VFIO file descriptor and the VFIO group file descriptor
152  * available to the user.
153  * Return: 0 on success, -errno on failure.
154  * Availability: Always
155  */
156 #define VFIO_GROUP_SET_CONTAINER	_IO(VFIO_TYPE, VFIO_BASE + 4)
157 
158 /**
159  * VFIO_GROUP_UNSET_CONTAINER - _IO(VFIO_TYPE, VFIO_BASE + 5)
160  *
161  * Remove the group from the attached container.  This is the
162  * opposite of the SET_CONTAINER call and returns the group to
163  * an initial state.  All device file descriptors must be released
164  * prior to calling this interface.  When removing the last group
165  * from a container, the IOMMU will be disabled and all state lost,
166  * effectively also returning the VFIO file descriptor to an initial
167  * state.
168  * Return: 0 on success, -errno on failure.
169  * Availability: When attached to container
170  */
171 #define VFIO_GROUP_UNSET_CONTAINER	_IO(VFIO_TYPE, VFIO_BASE + 5)
172 
173 /**
174  * VFIO_GROUP_GET_DEVICE_FD - _IOW(VFIO_TYPE, VFIO_BASE + 6, char)
175  *
176  * Return a new file descriptor for the device object described by
177  * the provided string.  The string should match a device listed in
178  * the devices subdirectory of the IOMMU group sysfs entry.  The
179  * group containing the device must already be added to this context.
180  * Return: new file descriptor on success, -errno on failure.
181  * Availability: When attached to container
182  */
183 #define VFIO_GROUP_GET_DEVICE_FD	_IO(VFIO_TYPE, VFIO_BASE + 6)
184 
185 /* --------------- IOCTLs for DEVICE file descriptors --------------- */
186 
187 /**
188  * VFIO_DEVICE_GET_INFO - _IOR(VFIO_TYPE, VFIO_BASE + 7,
189  *						struct vfio_device_info)
190  *
191  * Retrieve information about the device.  Fills in provided
192  * struct vfio_device_info.  Caller sets argsz.
193  * Return: 0 on success, -errno on failure.
194  */
195 struct vfio_device_info {
196 	__u32	argsz;
197 	__u32	flags;
198 #define VFIO_DEVICE_FLAGS_RESET	(1 << 0)	/* Device supports reset */
199 #define VFIO_DEVICE_FLAGS_PCI	(1 << 1)	/* vfio-pci device */
200 #define VFIO_DEVICE_FLAGS_PLATFORM (1 << 2)	/* vfio-platform device */
201 #define VFIO_DEVICE_FLAGS_AMBA  (1 << 3)	/* vfio-amba device */
202 #define VFIO_DEVICE_FLAGS_CCW	(1 << 4)	/* vfio-ccw device */
203 	__u32	num_regions;	/* Max region index + 1 */
204 	__u32	num_irqs;	/* Max IRQ index + 1 */
205 };
206 #define VFIO_DEVICE_GET_INFO		_IO(VFIO_TYPE, VFIO_BASE + 7)
207 
208 /*
209  * Vendor driver using Mediated device framework should provide device_api
210  * attribute in supported type attribute groups. Device API string should be one
211  * of the following corresponding to device flags in vfio_device_info structure.
212  */
213 
214 #define VFIO_DEVICE_API_PCI_STRING		"vfio-pci"
215 #define VFIO_DEVICE_API_PLATFORM_STRING		"vfio-platform"
216 #define VFIO_DEVICE_API_AMBA_STRING		"vfio-amba"
217 #define VFIO_DEVICE_API_CCW_STRING		"vfio-ccw"
218 
219 /**
220  * VFIO_DEVICE_GET_REGION_INFO - _IOWR(VFIO_TYPE, VFIO_BASE + 8,
221  *				       struct vfio_region_info)
222  *
223  * Retrieve information about a device region.  Caller provides
224  * struct vfio_region_info with index value set.  Caller sets argsz.
225  * Implementation of region mapping is bus driver specific.  This is
226  * intended to describe MMIO, I/O port, as well as bus specific
227  * regions (ex. PCI config space).  Zero sized regions may be used
228  * to describe unimplemented regions (ex. unimplemented PCI BARs).
229  * Return: 0 on success, -errno on failure.
230  */
231 struct vfio_region_info {
232 	__u32	argsz;
233 	__u32	flags;
234 #define VFIO_REGION_INFO_FLAG_READ	(1 << 0) /* Region supports read */
235 #define VFIO_REGION_INFO_FLAG_WRITE	(1 << 1) /* Region supports write */
236 #define VFIO_REGION_INFO_FLAG_MMAP	(1 << 2) /* Region supports mmap */
237 #define VFIO_REGION_INFO_FLAG_CAPS	(1 << 3) /* Info supports caps */
238 	__u32	index;		/* Region index */
239 	__u32	cap_offset;	/* Offset within info struct of first cap */
240 	__u64	size;		/* Region size (bytes) */
241 	__u64	offset;		/* Region offset from start of device fd */
242 };
243 #define VFIO_DEVICE_GET_REGION_INFO	_IO(VFIO_TYPE, VFIO_BASE + 8)
244 
245 /*
246  * The sparse mmap capability allows finer granularity of specifying areas
247  * within a region with mmap support.  When specified, the user should only
248  * mmap the offset ranges specified by the areas array.  mmaps outside of the
249  * areas specified may fail (such as the range covering a PCI MSI-X table) or
250  * may result in improper device behavior.
251  *
252  * The structures below define version 1 of this capability.
253  */
254 #define VFIO_REGION_INFO_CAP_SPARSE_MMAP	1
255 
256 struct vfio_region_sparse_mmap_area {
257 	__u64	offset;	/* Offset of mmap'able area within region */
258 	__u64	size;	/* Size of mmap'able area */
259 };
260 
261 struct vfio_region_info_cap_sparse_mmap {
262 	struct vfio_info_cap_header header;
263 	__u32	nr_areas;
264 	__u32	reserved;
265 	struct vfio_region_sparse_mmap_area areas[];
266 };
267 
268 /*
269  * The device specific type capability allows regions unique to a specific
270  * device or class of devices to be exposed.  This helps solve the problem for
271  * vfio bus drivers of defining which region indexes correspond to which region
272  * on the device, without needing to resort to static indexes, as done by
273  * vfio-pci.  For instance, if we were to go back in time, we might remove
274  * VFIO_PCI_VGA_REGION_INDEX and let vfio-pci simply define that all indexes
275  * greater than or equal to VFIO_PCI_NUM_REGIONS are device specific and we'd
276  * make a "VGA" device specific type to describe the VGA access space.  This
277  * means that non-VGA devices wouldn't need to waste this index, and thus the
278  * address space associated with it due to implementation of device file
279  * descriptor offsets in vfio-pci.
280  *
281  * The current implementation is now part of the user ABI, so we can't use this
282  * for VGA, but there are other upcoming use cases, such as opregions for Intel
283  * IGD devices and framebuffers for vGPU devices.  We missed VGA, but we'll
284  * use this for future additions.
285  *
286  * The structure below defines version 1 of this capability.
287  */
288 #define VFIO_REGION_INFO_CAP_TYPE	2
289 
290 struct vfio_region_info_cap_type {
291 	struct vfio_info_cap_header header;
292 	__u32 type;	/* global per bus driver */
293 	__u32 subtype;	/* type specific */
294 };
295 
296 #define VFIO_REGION_TYPE_PCI_VENDOR_TYPE	(1 << 31)
297 #define VFIO_REGION_TYPE_PCI_VENDOR_MASK	(0xffff)
298 
299 /* 8086 Vendor sub-types */
300 #define VFIO_REGION_SUBTYPE_INTEL_IGD_OPREGION	(1)
301 #define VFIO_REGION_SUBTYPE_INTEL_IGD_HOST_CFG	(2)
302 #define VFIO_REGION_SUBTYPE_INTEL_IGD_LPC_CFG	(3)
303 
304 /**
305  * VFIO_DEVICE_GET_IRQ_INFO - _IOWR(VFIO_TYPE, VFIO_BASE + 9,
306  *				    struct vfio_irq_info)
307  *
308  * Retrieve information about a device IRQ.  Caller provides
309  * struct vfio_irq_info with index value set.  Caller sets argsz.
310  * Implementation of IRQ mapping is bus driver specific.  Indexes
311  * using multiple IRQs are primarily intended to support MSI-like
312  * interrupt blocks.  Zero count irq blocks may be used to describe
313  * unimplemented interrupt types.
314  *
315  * The EVENTFD flag indicates the interrupt index supports eventfd based
316  * signaling.
317  *
318  * The MASKABLE flags indicates the index supports MASK and UNMASK
319  * actions described below.
320  *
321  * AUTOMASKED indicates that after signaling, the interrupt line is
322  * automatically masked by VFIO and the user needs to unmask the line
323  * to receive new interrupts.  This is primarily intended to distinguish
324  * level triggered interrupts.
325  *
326  * The NORESIZE flag indicates that the interrupt lines within the index
327  * are setup as a set and new subindexes cannot be enabled without first
328  * disabling the entire index.  This is used for interrupts like PCI MSI
329  * and MSI-X where the driver may only use a subset of the available
330  * indexes, but VFIO needs to enable a specific number of vectors
331  * upfront.  In the case of MSI-X, where the user can enable MSI-X and
332  * then add and unmask vectors, it's up to userspace to make the decision
333  * whether to allocate the maximum supported number of vectors or tear
334  * down setup and incrementally increase the vectors as each is enabled.
335  */
336 struct vfio_irq_info {
337 	__u32	argsz;
338 	__u32	flags;
339 #define VFIO_IRQ_INFO_EVENTFD		(1 << 0)
340 #define VFIO_IRQ_INFO_MASKABLE		(1 << 1)
341 #define VFIO_IRQ_INFO_AUTOMASKED	(1 << 2)
342 #define VFIO_IRQ_INFO_NORESIZE		(1 << 3)
343 	__u32	index;		/* IRQ index */
344 	__u32	count;		/* Number of IRQs within this index */
345 };
346 #define VFIO_DEVICE_GET_IRQ_INFO	_IO(VFIO_TYPE, VFIO_BASE + 9)
347 
348 /**
349  * VFIO_DEVICE_SET_IRQS - _IOW(VFIO_TYPE, VFIO_BASE + 10, struct vfio_irq_set)
350  *
351  * Set signaling, masking, and unmasking of interrupts.  Caller provides
352  * struct vfio_irq_set with all fields set.  'start' and 'count' indicate
353  * the range of subindexes being specified.
354  *
355  * The DATA flags specify the type of data provided.  If DATA_NONE, the
356  * operation performs the specified action immediately on the specified
357  * interrupt(s).  For example, to unmask AUTOMASKED interrupt [0,0]:
358  * flags = (DATA_NONE|ACTION_UNMASK), index = 0, start = 0, count = 1.
359  *
360  * DATA_BOOL allows sparse support for the same on arrays of interrupts.
361  * For example, to mask interrupts [0,1] and [0,3] (but not [0,2]):
362  * flags = (DATA_BOOL|ACTION_MASK), index = 0, start = 1, count = 3,
363  * data = {1,0,1}
364  *
365  * DATA_EVENTFD binds the specified ACTION to the provided __s32 eventfd.
366  * A value of -1 can be used to either de-assign interrupts if already
367  * assigned or skip un-assigned interrupts.  For example, to set an eventfd
368  * to be trigger for interrupts [0,0] and [0,2]:
369  * flags = (DATA_EVENTFD|ACTION_TRIGGER), index = 0, start = 0, count = 3,
370  * data = {fd1, -1, fd2}
371  * If index [0,1] is previously set, two count = 1 ioctls calls would be
372  * required to set [0,0] and [0,2] without changing [0,1].
373  *
374  * Once a signaling mechanism is set, DATA_BOOL or DATA_NONE can be used
375  * with ACTION_TRIGGER to perform kernel level interrupt loopback testing
376  * from userspace (ie. simulate hardware triggering).
377  *
378  * Setting of an event triggering mechanism to userspace for ACTION_TRIGGER
379  * enables the interrupt index for the device.  Individual subindex interrupts
380  * can be disabled using the -1 value for DATA_EVENTFD or the index can be
381  * disabled as a whole with: flags = (DATA_NONE|ACTION_TRIGGER), count = 0.
382  *
383  * Note that ACTION_[UN]MASK specify user->kernel signaling (irqfds) while
384  * ACTION_TRIGGER specifies kernel->user signaling.
385  */
386 struct vfio_irq_set {
387 	__u32	argsz;
388 	__u32	flags;
389 #define VFIO_IRQ_SET_DATA_NONE		(1 << 0) /* Data not present */
390 #define VFIO_IRQ_SET_DATA_BOOL		(1 << 1) /* Data is bool (u8) */
391 #define VFIO_IRQ_SET_DATA_EVENTFD	(1 << 2) /* Data is eventfd (s32) */
392 #define VFIO_IRQ_SET_ACTION_MASK	(1 << 3) /* Mask interrupt */
393 #define VFIO_IRQ_SET_ACTION_UNMASK	(1 << 4) /* Unmask interrupt */
394 #define VFIO_IRQ_SET_ACTION_TRIGGER	(1 << 5) /* Trigger interrupt */
395 	__u32	index;
396 	__u32	start;
397 	__u32	count;
398 	__u8	data[];
399 };
400 #define VFIO_DEVICE_SET_IRQS		_IO(VFIO_TYPE, VFIO_BASE + 10)
401 
402 #define VFIO_IRQ_SET_DATA_TYPE_MASK	(VFIO_IRQ_SET_DATA_NONE | \
403 					 VFIO_IRQ_SET_DATA_BOOL | \
404 					 VFIO_IRQ_SET_DATA_EVENTFD)
405 #define VFIO_IRQ_SET_ACTION_TYPE_MASK	(VFIO_IRQ_SET_ACTION_MASK | \
406 					 VFIO_IRQ_SET_ACTION_UNMASK | \
407 					 VFIO_IRQ_SET_ACTION_TRIGGER)
408 /**
409  * VFIO_DEVICE_RESET - _IO(VFIO_TYPE, VFIO_BASE + 11)
410  *
411  * Reset a device.
412  */
413 #define VFIO_DEVICE_RESET		_IO(VFIO_TYPE, VFIO_BASE + 11)
414 
415 /*
416  * The VFIO-PCI bus driver makes use of the following fixed region and
417  * IRQ index mapping.  Unimplemented regions return a size of zero.
418  * Unimplemented IRQ types return a count of zero.
419  */
420 
421 enum {
422 	VFIO_PCI_BAR0_REGION_INDEX,
423 	VFIO_PCI_BAR1_REGION_INDEX,
424 	VFIO_PCI_BAR2_REGION_INDEX,
425 	VFIO_PCI_BAR3_REGION_INDEX,
426 	VFIO_PCI_BAR4_REGION_INDEX,
427 	VFIO_PCI_BAR5_REGION_INDEX,
428 	VFIO_PCI_ROM_REGION_INDEX,
429 	VFIO_PCI_CONFIG_REGION_INDEX,
430 	/*
431 	 * Expose VGA regions defined for PCI base class 03, subclass 00.
432 	 * This includes I/O port ranges 0x3b0 to 0x3bb and 0x3c0 to 0x3df
433 	 * as well as the MMIO range 0xa0000 to 0xbffff.  Each implemented
434 	 * range is found at it's identity mapped offset from the region
435 	 * offset, for example 0x3b0 is region_info.offset + 0x3b0.  Areas
436 	 * between described ranges are unimplemented.
437 	 */
438 	VFIO_PCI_VGA_REGION_INDEX,
439 	VFIO_PCI_NUM_REGIONS = 9 /* Fixed user ABI, region indexes >=9 use */
440 				 /* device specific cap to define content. */
441 };
442 
443 enum {
444 	VFIO_PCI_INTX_IRQ_INDEX,
445 	VFIO_PCI_MSI_IRQ_INDEX,
446 	VFIO_PCI_MSIX_IRQ_INDEX,
447 	VFIO_PCI_ERR_IRQ_INDEX,
448 	VFIO_PCI_REQ_IRQ_INDEX,
449 	VFIO_PCI_NUM_IRQS
450 };
451 
452 /*
453  * The vfio-ccw bus driver makes use of the following fixed region and
454  * IRQ index mapping. Unimplemented regions return a size of zero.
455  * Unimplemented IRQ types return a count of zero.
456  */
457 
458 enum {
459 	VFIO_CCW_CONFIG_REGION_INDEX,
460 	VFIO_CCW_NUM_REGIONS
461 };
462 
463 enum {
464 	VFIO_CCW_IO_IRQ_INDEX,
465 	VFIO_CCW_NUM_IRQS
466 };
467 
468 /**
469  * VFIO_DEVICE_GET_PCI_HOT_RESET_INFO - _IORW(VFIO_TYPE, VFIO_BASE + 12,
470  *					      struct vfio_pci_hot_reset_info)
471  *
472  * Return: 0 on success, -errno on failure:
473  *	-enospc = insufficient buffer, -enodev = unsupported for device.
474  */
475 struct vfio_pci_dependent_device {
476 	__u32	group_id;
477 	__u16	segment;
478 	__u8	bus;
479 	__u8	devfn; /* Use PCI_SLOT/PCI_FUNC */
480 };
481 
482 struct vfio_pci_hot_reset_info {
483 	__u32	argsz;
484 	__u32	flags;
485 	__u32	count;
486 	struct vfio_pci_dependent_device	devices[];
487 };
488 
489 #define VFIO_DEVICE_GET_PCI_HOT_RESET_INFO	_IO(VFIO_TYPE, VFIO_BASE + 12)
490 
491 /**
492  * VFIO_DEVICE_PCI_HOT_RESET - _IOW(VFIO_TYPE, VFIO_BASE + 13,
493  *				    struct vfio_pci_hot_reset)
494  *
495  * Return: 0 on success, -errno on failure.
496  */
497 struct vfio_pci_hot_reset {
498 	__u32	argsz;
499 	__u32	flags;
500 	__u32	count;
501 	__s32	group_fds[];
502 };
503 
504 #define VFIO_DEVICE_PCI_HOT_RESET	_IO(VFIO_TYPE, VFIO_BASE + 13)
505 
506 /* -------- API for Type1 VFIO IOMMU -------- */
507 
508 /**
509  * VFIO_IOMMU_GET_INFO - _IOR(VFIO_TYPE, VFIO_BASE + 12, struct vfio_iommu_info)
510  *
511  * Retrieve information about the IOMMU object. Fills in provided
512  * struct vfio_iommu_info. Caller sets argsz.
513  *
514  * XXX Should we do these by CHECK_EXTENSION too?
515  */
516 struct vfio_iommu_type1_info {
517 	__u32	argsz;
518 	__u32	flags;
519 #define VFIO_IOMMU_INFO_PGSIZES (1 << 0)	/* supported page sizes info */
520 	__u64	iova_pgsizes;		/* Bitmap of supported page sizes */
521 };
522 
523 #define VFIO_IOMMU_GET_INFO _IO(VFIO_TYPE, VFIO_BASE + 12)
524 
525 /**
526  * VFIO_IOMMU_MAP_DMA - _IOW(VFIO_TYPE, VFIO_BASE + 13, struct vfio_dma_map)
527  *
528  * Map process virtual addresses to IO virtual addresses using the
529  * provided struct vfio_dma_map. Caller sets argsz. READ &/ WRITE required.
530  */
531 struct vfio_iommu_type1_dma_map {
532 	__u32	argsz;
533 	__u32	flags;
534 #define VFIO_DMA_MAP_FLAG_READ (1 << 0)		/* readable from device */
535 #define VFIO_DMA_MAP_FLAG_WRITE (1 << 1)	/* writable from device */
536 	__u64	vaddr;				/* Process virtual address */
537 	__u64	iova;				/* IO virtual address */
538 	__u64	size;				/* Size of mapping (bytes) */
539 };
540 
541 #define VFIO_IOMMU_MAP_DMA _IO(VFIO_TYPE, VFIO_BASE + 13)
542 
543 /**
544  * VFIO_IOMMU_UNMAP_DMA - _IOWR(VFIO_TYPE, VFIO_BASE + 14,
545  *							struct vfio_dma_unmap)
546  *
547  * Unmap IO virtual addresses using the provided struct vfio_dma_unmap.
548  * Caller sets argsz.  The actual unmapped size is returned in the size
549  * field.  No guarantee is made to the user that arbitrary unmaps of iova
550  * or size different from those used in the original mapping call will
551  * succeed.
552  */
553 struct vfio_iommu_type1_dma_unmap {
554 	__u32	argsz;
555 	__u32	flags;
556 	__u64	iova;				/* IO virtual address */
557 	__u64	size;				/* Size of mapping (bytes) */
558 };
559 
560 #define VFIO_IOMMU_UNMAP_DMA _IO(VFIO_TYPE, VFIO_BASE + 14)
561 
562 /*
563  * IOCTLs to enable/disable IOMMU container usage.
564  * No parameters are supported.
565  */
566 #define VFIO_IOMMU_ENABLE	_IO(VFIO_TYPE, VFIO_BASE + 15)
567 #define VFIO_IOMMU_DISABLE	_IO(VFIO_TYPE, VFIO_BASE + 16)
568 
569 /* -------- Additional API for SPAPR TCE (Server POWERPC) IOMMU -------- */
570 
571 /*
572  * The SPAPR TCE DDW info struct provides the information about
573  * the details of Dynamic DMA window capability.
574  *
575  * @pgsizes contains a page size bitmask, 4K/64K/16M are supported.
576  * @max_dynamic_windows_supported tells the maximum number of windows
577  * which the platform can create.
578  * @levels tells the maximum number of levels in multi-level IOMMU tables;
579  * this allows splitting a table into smaller chunks which reduces
580  * the amount of physically contiguous memory required for the table.
581  */
582 struct vfio_iommu_spapr_tce_ddw_info {
583 	__u64 pgsizes;			/* Bitmap of supported page sizes */
584 	__u32 max_dynamic_windows_supported;
585 	__u32 levels;
586 };
587 
588 /*
589  * The SPAPR TCE info struct provides the information about the PCI bus
590  * address ranges available for DMA, these values are programmed into
591  * the hardware so the guest has to know that information.
592  *
593  * The DMA 32 bit window start is an absolute PCI bus address.
594  * The IOVA address passed via map/unmap ioctls are absolute PCI bus
595  * addresses too so the window works as a filter rather than an offset
596  * for IOVA addresses.
597  *
598  * Flags supported:
599  * - VFIO_IOMMU_SPAPR_INFO_DDW: informs the userspace that dynamic DMA windows
600  *   (DDW) support is present. @ddw is only supported when DDW is present.
601  */
602 struct vfio_iommu_spapr_tce_info {
603 	__u32 argsz;
604 	__u32 flags;
605 #define VFIO_IOMMU_SPAPR_INFO_DDW	(1 << 0)	/* DDW supported */
606 	__u32 dma32_window_start;	/* 32 bit window start (bytes) */
607 	__u32 dma32_window_size;	/* 32 bit window size (bytes) */
608 	struct vfio_iommu_spapr_tce_ddw_info ddw;
609 };
610 
611 #define VFIO_IOMMU_SPAPR_TCE_GET_INFO	_IO(VFIO_TYPE, VFIO_BASE + 12)
612 
613 /*
614  * EEH PE operation struct provides ways to:
615  * - enable/disable EEH functionality;
616  * - unfreeze IO/DMA for frozen PE;
617  * - read PE state;
618  * - reset PE;
619  * - configure PE;
620  * - inject EEH error.
621  */
622 struct vfio_eeh_pe_err {
623 	__u32 type;
624 	__u32 func;
625 	__u64 addr;
626 	__u64 mask;
627 };
628 
629 struct vfio_eeh_pe_op {
630 	__u32 argsz;
631 	__u32 flags;
632 	__u32 op;
633 	union {
634 		struct vfio_eeh_pe_err err;
635 	};
636 };
637 
638 #define VFIO_EEH_PE_DISABLE		0	/* Disable EEH functionality */
639 #define VFIO_EEH_PE_ENABLE		1	/* Enable EEH functionality  */
640 #define VFIO_EEH_PE_UNFREEZE_IO		2	/* Enable IO for frozen PE   */
641 #define VFIO_EEH_PE_UNFREEZE_DMA	3	/* Enable DMA for frozen PE  */
642 #define VFIO_EEH_PE_GET_STATE		4	/* PE state retrieval        */
643 #define  VFIO_EEH_PE_STATE_NORMAL	0	/* PE in functional state    */
644 #define  VFIO_EEH_PE_STATE_RESET	1	/* PE reset in progress      */
645 #define  VFIO_EEH_PE_STATE_STOPPED	2	/* Stopped DMA and IO        */
646 #define  VFIO_EEH_PE_STATE_STOPPED_DMA	4	/* Stopped DMA only          */
647 #define  VFIO_EEH_PE_STATE_UNAVAIL	5	/* State unavailable         */
648 #define VFIO_EEH_PE_RESET_DEACTIVATE	5	/* Deassert PE reset         */
649 #define VFIO_EEH_PE_RESET_HOT		6	/* Assert hot reset          */
650 #define VFIO_EEH_PE_RESET_FUNDAMENTAL	7	/* Assert fundamental reset  */
651 #define VFIO_EEH_PE_CONFIGURE		8	/* PE configuration          */
652 #define VFIO_EEH_PE_INJECT_ERR		9	/* Inject EEH error          */
653 
654 #define VFIO_EEH_PE_OP			_IO(VFIO_TYPE, VFIO_BASE + 21)
655 
656 /**
657  * VFIO_IOMMU_SPAPR_REGISTER_MEMORY - _IOW(VFIO_TYPE, VFIO_BASE + 17, struct vfio_iommu_spapr_register_memory)
658  *
659  * Registers user space memory where DMA is allowed. It pins
660  * user pages and does the locked memory accounting so
661  * subsequent VFIO_IOMMU_MAP_DMA/VFIO_IOMMU_UNMAP_DMA calls
662  * get faster.
663  */
664 struct vfio_iommu_spapr_register_memory {
665 	__u32	argsz;
666 	__u32	flags;
667 	__u64	vaddr;				/* Process virtual address */
668 	__u64	size;				/* Size of mapping (bytes) */
669 };
670 #define VFIO_IOMMU_SPAPR_REGISTER_MEMORY	_IO(VFIO_TYPE, VFIO_BASE + 17)
671 
672 /**
673  * VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY - _IOW(VFIO_TYPE, VFIO_BASE + 18, struct vfio_iommu_spapr_register_memory)
674  *
675  * Unregisters user space memory registered with
676  * VFIO_IOMMU_SPAPR_REGISTER_MEMORY.
677  * Uses vfio_iommu_spapr_register_memory for parameters.
678  */
679 #define VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY	_IO(VFIO_TYPE, VFIO_BASE + 18)
680 
681 /**
682  * VFIO_IOMMU_SPAPR_TCE_CREATE - _IOWR(VFIO_TYPE, VFIO_BASE + 19, struct vfio_iommu_spapr_tce_create)
683  *
684  * Creates an additional TCE table and programs it (sets a new DMA window)
685  * to every IOMMU group in the container. It receives page shift, window
686  * size and number of levels in the TCE table being created.
687  *
688  * It allocates and returns an offset on a PCI bus of the new DMA window.
689  */
690 struct vfio_iommu_spapr_tce_create {
691 	__u32 argsz;
692 	__u32 flags;
693 	/* in */
694 	__u32 page_shift;
695 	__u32 __resv1;
696 	__u64 window_size;
697 	__u32 levels;
698 	__u32 __resv2;
699 	/* out */
700 	__u64 start_addr;
701 };
702 #define VFIO_IOMMU_SPAPR_TCE_CREATE	_IO(VFIO_TYPE, VFIO_BASE + 19)
703 
704 /**
705  * VFIO_IOMMU_SPAPR_TCE_REMOVE - _IOW(VFIO_TYPE, VFIO_BASE + 20, struct vfio_iommu_spapr_tce_remove)
706  *
707  * Unprograms a TCE table from all groups in the container and destroys it.
708  * It receives a PCI bus offset as a window id.
709  */
710 struct vfio_iommu_spapr_tce_remove {
711 	__u32 argsz;
712 	__u32 flags;
713 	/* in */
714 	__u64 start_addr;
715 };
716 #define VFIO_IOMMU_SPAPR_TCE_REMOVE	_IO(VFIO_TYPE, VFIO_BASE + 20)
717 
718 /* ***************************************************************** */
719 
720 #endif /* VFIO_H */
721