1 /*-
2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3 *
4 * Copyright (c) 2011 NetApp, Inc.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 *
16 * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * SUCH DAMAGE.
27 *
28 * $FreeBSD$
29 */
30 /*
31 * This file and its contents are supplied under the terms of the
32 * Common Development and Distribution License ("CDDL"), version 1.0.
33 * You may only use this file in accordance with the terms of version
34 * 1.0 of the CDDL.
35 *
36 * A full copy of the text of the CDDL should have accompanied this
37 * source. A copy of the CDDL is also available via the Internet at
38 * http://www.illumos.org/license/CDDL.
39 *
40 * Copyright 2015 Pluribus Networks Inc.
41 * Copyright 2019 Joyent, Inc.
42 * Copyright 2020 Oxide Computer Company
43 */
44
45 #include <sys/cdefs.h>
46 __FBSDID("$FreeBSD$");
47
48 #include <sys/param.h>
49 #include <sys/sysctl.h>
50 #include <sys/ioctl.h>
51 #ifdef __FreeBSD__
52 #include <sys/linker.h>
53 #endif
54 #include <sys/mman.h>
55 #include <sys/module.h>
56 #include <sys/_iovec.h>
57 #include <sys/cpuset.h>
58
59 #include <x86/segments.h>
60 #include <machine/specialreg.h>
61
62 #include <errno.h>
63 #include <stdio.h>
64 #include <stdlib.h>
65 #include <assert.h>
66 #include <string.h>
67 #include <fcntl.h>
68 #include <unistd.h>
69
70 #include <libutil.h>
71
72 #include <machine/vmm.h>
73 #include <machine/vmm_dev.h>
74
75 #include "vmmapi.h"
76
77 #define MB (1024 * 1024UL)
78 #define GB (1024 * 1024 * 1024UL)
79
80 #ifndef __FreeBSD__
81 /* shim to no-op for now */
82 #define MAP_NOCORE 0
83 #define MAP_ALIGNED_SUPER 0
84
85 /* Rely on PROT_NONE for guard purposes */
86 #define MAP_GUARD (MAP_PRIVATE | MAP_ANON | MAP_NORESERVE)
87 #endif
88
89 /*
90 * Size of the guard region before and after the virtual address space
91 * mapping the guest physical memory. This must be a multiple of the
92 * superpage size for performance reasons.
93 */
94 #define VM_MMAP_GUARD_SIZE (4 * MB)
95
96 #define PROT_RW (PROT_READ | PROT_WRITE)
97 #define PROT_ALL (PROT_READ | PROT_WRITE | PROT_EXEC)
98
99 struct vmctx {
100 int fd;
101 uint32_t lowmem_limit;
102 int memflags;
103 size_t lowmem;
104 size_t highmem;
105 char *baseaddr;
106 char *name;
107 };
108
109 #ifdef __FreeBSD__
110 #define CREATE(x) sysctlbyname("hw.vmm.create", NULL, NULL, (x), strlen((x)))
111 #define DESTROY(x) sysctlbyname("hw.vmm.destroy", NULL, NULL, (x), strlen((x)))
112 #else
113 #define CREATE(x) vm_do_ctl(VMM_CREATE_VM, (x))
114 #define DESTROY(x) vm_do_ctl(VMM_DESTROY_VM, (x))
115
116 static int
117 vm_do_ctl(int cmd, const char *name)
118 {
119 int ctl_fd;
120
121 ctl_fd = open(VMM_CTL_DEV, O_EXCL | O_RDWR);
122 if (ctl_fd < 0) {
123 return (-1);
124 }
125
126 if (ioctl(ctl_fd, cmd, name) == -1) {
127 int err = errno;
128
129 /* Do not lose ioctl errno through the close(2) */
130 (void) close(ctl_fd);
131 errno = err;
132 return (-1);
133 }
134 (void) close(ctl_fd);
135
136 return (0);
137 }
138 #endif
139
140 static int
141 vm_device_open(const char *name)
142 {
143 int fd, len;
144 char *vmfile;
145
146 len = strlen("/dev/vmm/") + strlen(name) + 1;
147 vmfile = malloc(len);
148 assert(vmfile != NULL);
149 snprintf(vmfile, len, "/dev/vmm/%s", name);
150
151 /* Open the device file */
152 fd = open(vmfile, O_RDWR, 0);
153
154 free(vmfile);
155 return (fd);
156 }
157
158 int
159 vm_create(const char *name)
160 {
161 #ifdef __FreeBSD__
162 /* Try to load vmm(4) module before creating a guest. */
163 if (modfind("vmm") < 0)
164 kldload("vmm");
165 #endif
166 return (CREATE((char *)name));
167 }
168
169 struct vmctx *
170 vm_open(const char *name)
171 {
172 struct vmctx *vm;
173
174 vm = malloc(sizeof(struct vmctx) + strlen(name) + 1);
175 assert(vm != NULL);
176
177 vm->fd = -1;
178 vm->memflags = 0;
179 vm->lowmem_limit = 3 * GB;
180 vm->name = (char *)(vm + 1);
181 strcpy(vm->name, name);
182
183 if ((vm->fd = vm_device_open(vm->name)) < 0)
184 goto err;
185
186 return (vm);
187 err:
188 free(vm);
189 return (NULL);
190 }
191
192 #ifndef __FreeBSD__
193 void
194 vm_close(struct vmctx *vm)
195 {
196 assert(vm != NULL);
197 assert(vm->fd >= 0);
198
199 (void) close(vm->fd);
200
201 free(vm);
202 }
203 #endif
204
205 void
206 vm_destroy(struct vmctx *vm)
207 {
208 assert(vm != NULL);
209
210 if (vm->fd >= 0)
211 close(vm->fd);
212 DESTROY(vm->name);
213
214 free(vm);
215 }
216
217 int
218 vm_parse_memsize(const char *optarg, size_t *ret_memsize)
219 {
220 char *endptr;
221 size_t optval;
222 int error;
223
224 optval = strtoul(optarg, &endptr, 0);
225 if (*optarg != '\0' && *endptr == '\0') {
226 /*
227 * For the sake of backward compatibility if the memory size
228 * specified on the command line is less than a megabyte then
229 * it is interpreted as being in units of MB.
230 */
231 if (optval < MB)
232 optval *= MB;
233 *ret_memsize = optval;
234 error = 0;
235 } else
236 error = expand_number(optarg, ret_memsize);
237
238 return (error);
239 }
240
241 uint32_t
242 vm_get_lowmem_limit(struct vmctx *ctx)
243 {
244
245 return (ctx->lowmem_limit);
246 }
247
248 void
249 vm_set_lowmem_limit(struct vmctx *ctx, uint32_t limit)
250 {
251
252 ctx->lowmem_limit = limit;
253 }
254
255 void
256 vm_set_memflags(struct vmctx *ctx, int flags)
257 {
258
259 ctx->memflags = flags;
260 }
261
262 int
263 vm_get_memflags(struct vmctx *ctx)
264 {
265
266 return (ctx->memflags);
267 }
268
269 /*
270 * Map segment 'segid' starting at 'off' into guest address range [gpa,gpa+len).
271 */
272 int
273 vm_mmap_memseg(struct vmctx *ctx, vm_paddr_t gpa, int segid, vm_ooffset_t off,
274 size_t len, int prot)
275 {
276 struct vm_memmap memmap;
277 int error, flags;
278
279 memmap.gpa = gpa;
280 memmap.segid = segid;
281 memmap.segoff = off;
282 memmap.len = len;
283 memmap.prot = prot;
284 memmap.flags = 0;
285
286 if (ctx->memflags & VM_MEM_F_WIRED)
287 memmap.flags |= VM_MEMMAP_F_WIRED;
288
289 /*
290 * If this mapping already exists then don't create it again. This
291 * is the common case for SYSMEM mappings created by bhyveload(8).
292 */
293 error = vm_mmap_getnext(ctx, &gpa, &segid, &off, &len, &prot, &flags);
294 if (error == 0 && gpa == memmap.gpa) {
295 if (segid != memmap.segid || off != memmap.segoff ||
296 prot != memmap.prot || flags != memmap.flags) {
297 errno = EEXIST;
298 return (-1);
299 } else {
300 return (0);
301 }
302 }
303
304 error = ioctl(ctx->fd, VM_MMAP_MEMSEG, &memmap);
305 return (error);
306 }
307
308 int
309 vm_munmap_memseg(struct vmctx *ctx, vm_paddr_t gpa, size_t len)
310 {
311 struct vm_munmap munmap;
312 int error;
313
314 munmap.gpa = gpa;
315 munmap.len = len;
316
317 error = ioctl(ctx->fd, VM_MUNMAP_MEMSEG, &munmap);
318 return (error);
319 }
320
321 int
322 vm_mmap_getnext(struct vmctx *ctx, vm_paddr_t *gpa, int *segid,
323 vm_ooffset_t *segoff, size_t *len, int *prot, int *flags)
324 {
325 struct vm_memmap memmap;
326 int error;
327
328 bzero(&memmap, sizeof(struct vm_memmap));
329 memmap.gpa = *gpa;
330 error = ioctl(ctx->fd, VM_MMAP_GETNEXT, &memmap);
331 if (error == 0) {
332 *gpa = memmap.gpa;
333 *segid = memmap.segid;
334 *segoff = memmap.segoff;
335 *len = memmap.len;
336 *prot = memmap.prot;
337 *flags = memmap.flags;
338 }
339 return (error);
340 }
341
342 /*
343 * Return 0 if the segments are identical and non-zero otherwise.
344 *
345 * This is slightly complicated by the fact that only device memory segments
346 * are named.
347 */
348 static int
349 cmpseg(size_t len, const char *str, size_t len2, const char *str2)
350 {
351
352 if (len == len2) {
353 if ((!str && !str2) || (str && str2 && !strcmp(str, str2)))
354 return (0);
355 }
356 return (-1);
357 }
358
359 static int
360 vm_alloc_memseg(struct vmctx *ctx, int segid, size_t len, const char *name)
361 {
362 struct vm_memseg memseg;
363 size_t n;
364 int error;
365
366 /*
367 * If the memory segment has already been created then just return.
368 * This is the usual case for the SYSMEM segment created by userspace
369 * loaders like bhyveload(8).
370 */
371 error = vm_get_memseg(ctx, segid, &memseg.len, memseg.name,
372 sizeof(memseg.name));
373 if (error)
374 return (error);
375
376 if (memseg.len != 0) {
377 if (cmpseg(len, name, memseg.len, VM_MEMSEG_NAME(&memseg))) {
378 errno = EINVAL;
379 return (-1);
380 } else {
381 return (0);
382 }
383 }
384
385 bzero(&memseg, sizeof(struct vm_memseg));
386 memseg.segid = segid;
387 memseg.len = len;
388 if (name != NULL) {
389 n = strlcpy(memseg.name, name, sizeof(memseg.name));
390 if (n >= sizeof(memseg.name)) {
391 errno = ENAMETOOLONG;
392 return (-1);
393 }
394 }
395
396 error = ioctl(ctx->fd, VM_ALLOC_MEMSEG, &memseg);
397 return (error);
398 }
399
400 int
401 vm_get_memseg(struct vmctx *ctx, int segid, size_t *lenp, char *namebuf,
402 size_t bufsize)
403 {
404 struct vm_memseg memseg;
405 size_t n;
406 int error;
407
408 memseg.segid = segid;
409 error = ioctl(ctx->fd, VM_GET_MEMSEG, &memseg);
410 if (error == 0) {
411 *lenp = memseg.len;
412 n = strlcpy(namebuf, memseg.name, bufsize);
413 if (n >= bufsize) {
414 errno = ENAMETOOLONG;
415 error = -1;
416 }
417 }
418 return (error);
419 }
420
421 static int
422 #ifdef __FreeBSD__
423 setup_memory_segment(struct vmctx *ctx, vm_paddr_t gpa, size_t len, char *base)
424 #else
425 setup_memory_segment(struct vmctx *ctx, int segid, vm_paddr_t gpa, size_t len,
426 char *base)
427 #endif
428 {
429 char *ptr;
430 int error, flags;
431
432 /* Map 'len' bytes starting at 'gpa' in the guest address space */
433 #ifdef __FreeBSD__
434 error = vm_mmap_memseg(ctx, gpa, VM_SYSMEM, gpa, len, PROT_ALL);
435 #else
436 /*
437 * As we use two segments for lowmem/highmem the offset within the
438 * segment is 0 on illumos.
439 */
440 error = vm_mmap_memseg(ctx, gpa, segid, 0, len, PROT_ALL);
441 #endif
442 if (error)
443 return (error);
444
445 flags = MAP_SHARED | MAP_FIXED;
446 if ((ctx->memflags & VM_MEM_F_INCORE) == 0)
447 flags |= MAP_NOCORE;
448
449 /* mmap into the process address space on the host */
450 ptr = mmap(base + gpa, len, PROT_RW, flags, ctx->fd, gpa);
451 if (ptr == MAP_FAILED)
452 return (-1);
453
454 return (0);
455 }
456
457 int
458 vm_setup_memory(struct vmctx *ctx, size_t memsize, enum vm_mmap_style vms)
459 {
460 size_t objsize, len;
461 vm_paddr_t gpa;
462 char *baseaddr, *ptr;
463 int error;
464
465 assert(vms == VM_MMAP_ALL);
466
467 /*
468 * If 'memsize' cannot fit entirely in the 'lowmem' segment then
469 * create another 'highmem' segment above 4GB for the remainder.
470 */
471 if (memsize > ctx->lowmem_limit) {
472 ctx->lowmem = ctx->lowmem_limit;
473 ctx->highmem = memsize - ctx->lowmem_limit;
474 objsize = 4*GB + ctx->highmem;
475 } else {
476 ctx->lowmem = memsize;
477 ctx->highmem = 0;
478 objsize = ctx->lowmem;
479 }
480
481 #ifdef __FreeBSD__
482 error = vm_alloc_memseg(ctx, VM_SYSMEM, objsize, NULL);
483 if (error)
484 return (error);
485 #endif
486
487 /*
488 * Stake out a contiguous region covering the guest physical memory
489 * and the adjoining guard regions.
490 */
491 len = VM_MMAP_GUARD_SIZE + objsize + VM_MMAP_GUARD_SIZE;
492 ptr = mmap(NULL, len, PROT_NONE, MAP_GUARD | MAP_ALIGNED_SUPER, -1, 0);
493 if (ptr == MAP_FAILED)
494 return (-1);
495
496 baseaddr = ptr + VM_MMAP_GUARD_SIZE;
497
498 #ifdef __FreeBSD__
499 if (ctx->highmem > 0) {
500 gpa = 4*GB;
501 len = ctx->highmem;
502 error = setup_memory_segment(ctx, gpa, len, baseaddr);
503 if (error)
504 return (error);
505 }
506
507 if (ctx->lowmem > 0) {
508 gpa = 0;
509 len = ctx->lowmem;
510 error = setup_memory_segment(ctx, gpa, len, baseaddr);
511 if (error)
512 return (error);
513 }
514 #else
515 if (ctx->highmem > 0) {
516 error = vm_alloc_memseg(ctx, VM_HIGHMEM, ctx->highmem, NULL);
517 if (error)
518 return (error);
519 gpa = 4*GB;
520 len = ctx->highmem;
521 error = setup_memory_segment(ctx, VM_HIGHMEM, gpa, len, baseaddr);
522 if (error)
523 return (error);
524 }
525
526 if (ctx->lowmem > 0) {
527 error = vm_alloc_memseg(ctx, VM_LOWMEM, ctx->lowmem, NULL);
528 if (error)
529 return (error);
530 gpa = 0;
531 len = ctx->lowmem;
532 error = setup_memory_segment(ctx, VM_LOWMEM, gpa, len, baseaddr);
533 if (error)
534 return (error);
535 }
536 #endif
537
538 ctx->baseaddr = baseaddr;
539
540 return (0);
541 }
542
543 /*
544 * Returns a non-NULL pointer if [gaddr, gaddr+len) is entirely contained in
545 * the lowmem or highmem regions.
546 *
547 * In particular return NULL if [gaddr, gaddr+len) falls in guest MMIO region.
548 * The instruction emulation code depends on this behavior.
549 */
550 void *
551 vm_map_gpa(struct vmctx *ctx, vm_paddr_t gaddr, size_t len)
552 {
553
554 if (ctx->lowmem > 0) {
555 if (gaddr < ctx->lowmem && len <= ctx->lowmem &&
556 gaddr + len <= ctx->lowmem)
557 return (ctx->baseaddr + gaddr);
558 }
559
560 if (ctx->highmem > 0) {
561 if (gaddr >= 4*GB) {
562 if (gaddr < 4*GB + ctx->highmem &&
563 len <= ctx->highmem &&
564 gaddr + len <= 4*GB + ctx->highmem)
565 return (ctx->baseaddr + gaddr);
566 }
567 }
568
569 return (NULL);
570 }
571
572 size_t
573 vm_get_lowmem_size(struct vmctx *ctx)
574 {
575
576 return (ctx->lowmem);
577 }
578
579 size_t
580 vm_get_highmem_size(struct vmctx *ctx)
581 {
582
583 return (ctx->highmem);
584 }
585
586 #ifndef __FreeBSD__
587 int
588 vm_get_devmem_offset(struct vmctx *ctx, int segid, off_t *mapoff)
589 {
590 struct vm_devmem_offset vdo;
591 int error;
592
593 vdo.segid = segid;
594 error = ioctl(ctx->fd, VM_DEVMEM_GETOFFSET, &vdo);
595 if (error == 0)
596 *mapoff = vdo.offset;
597
598 return (error);
599 }
600 #endif
601
602 void *
603 vm_create_devmem(struct vmctx *ctx, int segid, const char *name, size_t len)
604 {
605 #ifdef __FreeBSD__
606 char pathname[MAXPATHLEN];
607 #endif
608 size_t len2;
609 char *base, *ptr;
610 int fd, error, flags;
611 off_t mapoff;
612
613 fd = -1;
614 ptr = MAP_FAILED;
615 if (name == NULL || strlen(name) == 0) {
616 errno = EINVAL;
617 goto done;
618 }
619
620 error = vm_alloc_memseg(ctx, segid, len, name);
621 if (error)
622 goto done;
623
624 #ifdef __FreeBSD__
625 strlcpy(pathname, "/dev/vmm.io/", sizeof(pathname));
626 strlcat(pathname, ctx->name, sizeof(pathname));
627 strlcat(pathname, ".", sizeof(pathname));
628 strlcat(pathname, name, sizeof(pathname));
629
630 fd = open(pathname, O_RDWR);
631 if (fd < 0)
632 goto done;
633 #else
634 if (vm_get_devmem_offset(ctx, segid, &mapoff) != 0)
635 goto done;
636 #endif
637
638 /*
639 * Stake out a contiguous region covering the device memory and the
640 * adjoining guard regions.
641 */
642 len2 = VM_MMAP_GUARD_SIZE + len + VM_MMAP_GUARD_SIZE;
643 base = mmap(NULL, len2, PROT_NONE, MAP_GUARD | MAP_ALIGNED_SUPER, -1,
644 0);
645 if (base == MAP_FAILED)
646 goto done;
647
648 flags = MAP_SHARED | MAP_FIXED;
649 if ((ctx->memflags & VM_MEM_F_INCORE) == 0)
650 flags |= MAP_NOCORE;
651
652 #ifdef __FreeBSD__
653 /* mmap the devmem region in the host address space */
654 ptr = mmap(base + VM_MMAP_GUARD_SIZE, len, PROT_RW, flags, fd, 0);
655 #else
656 /* mmap the devmem region in the host address space */
657 ptr = mmap(base + VM_MMAP_GUARD_SIZE, len, PROT_RW, flags, ctx->fd,
658 mapoff);
659 #endif
660 done:
661 if (fd >= 0)
662 close(fd);
663 return (ptr);
664 }
665
666 int
667 vm_set_desc(struct vmctx *ctx, int vcpu, int reg,
668 uint64_t base, uint32_t limit, uint32_t access)
669 {
670 int error;
671 struct vm_seg_desc vmsegdesc;
672
673 bzero(&vmsegdesc, sizeof(vmsegdesc));
674 vmsegdesc.cpuid = vcpu;
675 vmsegdesc.regnum = reg;
676 vmsegdesc.desc.base = base;
677 vmsegdesc.desc.limit = limit;
678 vmsegdesc.desc.access = access;
679
680 error = ioctl(ctx->fd, VM_SET_SEGMENT_DESCRIPTOR, &vmsegdesc);
681 return (error);
682 }
683
684 int
685 vm_get_desc(struct vmctx *ctx, int vcpu, int reg,
686 uint64_t *base, uint32_t *limit, uint32_t *access)
687 {
688 int error;
689 struct vm_seg_desc vmsegdesc;
690
691 bzero(&vmsegdesc, sizeof(vmsegdesc));
692 vmsegdesc.cpuid = vcpu;
693 vmsegdesc.regnum = reg;
694
695 error = ioctl(ctx->fd, VM_GET_SEGMENT_DESCRIPTOR, &vmsegdesc);
696 if (error == 0) {
697 *base = vmsegdesc.desc.base;
698 *limit = vmsegdesc.desc.limit;
699 *access = vmsegdesc.desc.access;
700 }
701 return (error);
702 }
703
704 int
705 vm_get_seg_desc(struct vmctx *ctx, int vcpu, int reg, struct seg_desc *seg_desc)
706 {
707 int error;
708
709 error = vm_get_desc(ctx, vcpu, reg, &seg_desc->base, &seg_desc->limit,
710 &seg_desc->access);
711 return (error);
712 }
713
714 int
715 vm_set_register(struct vmctx *ctx, int vcpu, int reg, uint64_t val)
716 {
717 int error;
718 struct vm_register vmreg;
719
720 bzero(&vmreg, sizeof(vmreg));
721 vmreg.cpuid = vcpu;
722 vmreg.regnum = reg;
723 vmreg.regval = val;
724
725 error = ioctl(ctx->fd, VM_SET_REGISTER, &vmreg);
726 return (error);
727 }
728
729 int
730 vm_get_register(struct vmctx *ctx, int vcpu, int reg, uint64_t *ret_val)
731 {
732 int error;
733 struct vm_register vmreg;
734
735 bzero(&vmreg, sizeof(vmreg));
736 vmreg.cpuid = vcpu;
737 vmreg.regnum = reg;
738
739 error = ioctl(ctx->fd, VM_GET_REGISTER, &vmreg);
740 *ret_val = vmreg.regval;
741 return (error);
742 }
743
744 int
745 vm_set_register_set(struct vmctx *ctx, int vcpu, unsigned int count,
746 const int *regnums, uint64_t *regvals)
747 {
748 int error;
749 struct vm_register_set vmregset;
750
751 bzero(&vmregset, sizeof(vmregset));
752 vmregset.cpuid = vcpu;
753 vmregset.count = count;
754 vmregset.regnums = regnums;
755 vmregset.regvals = regvals;
756
757 error = ioctl(ctx->fd, VM_SET_REGISTER_SET, &vmregset);
758 return (error);
759 }
760
761 int
762 vm_get_register_set(struct vmctx *ctx, int vcpu, unsigned int count,
763 const int *regnums, uint64_t *regvals)
764 {
765 int error;
766 struct vm_register_set vmregset;
767
768 bzero(&vmregset, sizeof(vmregset));
769 vmregset.cpuid = vcpu;
770 vmregset.count = count;
771 vmregset.regnums = regnums;
772 vmregset.regvals = regvals;
773
774 error = ioctl(ctx->fd, VM_GET_REGISTER_SET, &vmregset);
775 return (error);
776 }
777
778 int
779 vm_run(struct vmctx *ctx, int vcpu, const struct vm_entry *vm_entry,
780 struct vm_exit *vm_exit)
781 {
782 struct vm_entry entry;
783
784 bcopy(vm_entry, &entry, sizeof (entry));
785 entry.cpuid = vcpu;
786 entry.exit_data = vm_exit;
787
788 return (ioctl(ctx->fd, VM_RUN, &entry));
789 }
790
791 int
792 vm_suspend(struct vmctx *ctx, enum vm_suspend_how how)
793 {
794 struct vm_suspend vmsuspend;
795
796 bzero(&vmsuspend, sizeof(vmsuspend));
797 vmsuspend.how = how;
798 return (ioctl(ctx->fd, VM_SUSPEND, &vmsuspend));
799 }
800
801 int
802 vm_reinit(struct vmctx *ctx)
803 {
804
805 return (ioctl(ctx->fd, VM_REINIT, 0));
806 }
807
808 int
809 vm_inject_exception(struct vmctx *ctx, int vcpu, int vector, int errcode_valid,
810 uint32_t errcode, int restart_instruction)
811 {
812 struct vm_exception exc;
813
814 exc.cpuid = vcpu;
815 exc.vector = vector;
816 exc.error_code = errcode;
817 exc.error_code_valid = errcode_valid;
818 exc.restart_instruction = restart_instruction;
819
820 return (ioctl(ctx->fd, VM_INJECT_EXCEPTION, &exc));
821 }
822
823 #ifndef __FreeBSD__
824 void
825 vm_inject_fault(struct vmctx *ctx, int vcpu, int vector, int errcode_valid,
826 int errcode)
827 {
828 int error;
829 struct vm_exception exc;
830
831 exc.cpuid = vcpu;
832 exc.vector = vector;
833 exc.error_code = errcode;
834 exc.error_code_valid = errcode_valid;
835 exc.restart_instruction = 1;
836 error = ioctl(ctx->fd, VM_INJECT_EXCEPTION, &exc);
837
838 assert(error == 0);
839 }
840 #endif /* __FreeBSD__ */
841
842 int
843 vm_apicid2vcpu(struct vmctx *ctx, int apicid)
844 {
845 /*
846 * The apic id associated with the 'vcpu' has the same numerical value
847 * as the 'vcpu' itself.
848 */
849 return (apicid);
850 }
851
852 int
853 vm_lapic_irq(struct vmctx *ctx, int vcpu, int vector)
854 {
855 struct vm_lapic_irq vmirq;
856
857 bzero(&vmirq, sizeof(vmirq));
858 vmirq.cpuid = vcpu;
859 vmirq.vector = vector;
860
861 return (ioctl(ctx->fd, VM_LAPIC_IRQ, &vmirq));
862 }
863
864 int
865 vm_lapic_local_irq(struct vmctx *ctx, int vcpu, int vector)
866 {
867 struct vm_lapic_irq vmirq;
868
869 bzero(&vmirq, sizeof(vmirq));
870 vmirq.cpuid = vcpu;
871 vmirq.vector = vector;
872
873 return (ioctl(ctx->fd, VM_LAPIC_LOCAL_IRQ, &vmirq));
874 }
875
876 int
877 vm_lapic_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg)
878 {
879 struct vm_lapic_msi vmmsi;
880
881 bzero(&vmmsi, sizeof(vmmsi));
882 vmmsi.addr = addr;
883 vmmsi.msg = msg;
884
885 return (ioctl(ctx->fd, VM_LAPIC_MSI, &vmmsi));
886 }
887
888 int
889 vm_ioapic_assert_irq(struct vmctx *ctx, int irq)
890 {
891 struct vm_ioapic_irq ioapic_irq;
892
893 bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
894 ioapic_irq.irq = irq;
895
896 return (ioctl(ctx->fd, VM_IOAPIC_ASSERT_IRQ, &ioapic_irq));
897 }
898
899 int
900 vm_ioapic_deassert_irq(struct vmctx *ctx, int irq)
901 {
902 struct vm_ioapic_irq ioapic_irq;
903
904 bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
905 ioapic_irq.irq = irq;
906
907 return (ioctl(ctx->fd, VM_IOAPIC_DEASSERT_IRQ, &ioapic_irq));
908 }
909
910 int
911 vm_ioapic_pulse_irq(struct vmctx *ctx, int irq)
912 {
913 struct vm_ioapic_irq ioapic_irq;
914
915 bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
916 ioapic_irq.irq = irq;
917
918 return (ioctl(ctx->fd, VM_IOAPIC_PULSE_IRQ, &ioapic_irq));
919 }
920
921 int
922 vm_ioapic_pincount(struct vmctx *ctx, int *pincount)
923 {
924
925 return (ioctl(ctx->fd, VM_IOAPIC_PINCOUNT, pincount));
926 }
927
928 int
929 vm_readwrite_kernemu_device(struct vmctx *ctx, int vcpu, vm_paddr_t gpa,
930 bool write, int size, uint64_t *value)
931 {
932 struct vm_readwrite_kernemu_device irp = {
933 .vcpuid = vcpu,
934 .access_width = fls(size) - 1,
935 .gpa = gpa,
936 .value = write ? *value : ~0ul,
937 };
938 long cmd = (write ? VM_SET_KERNEMU_DEV : VM_GET_KERNEMU_DEV);
939 int rc;
940
941 rc = ioctl(ctx->fd, cmd, &irp);
942 if (rc == 0 && !write)
943 *value = irp.value;
944 return (rc);
945 }
946
947 int
948 vm_isa_assert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
949 {
950 struct vm_isa_irq isa_irq;
951
952 bzero(&isa_irq, sizeof(struct vm_isa_irq));
953 isa_irq.atpic_irq = atpic_irq;
954 isa_irq.ioapic_irq = ioapic_irq;
955
956 return (ioctl(ctx->fd, VM_ISA_ASSERT_IRQ, &isa_irq));
957 }
958
959 int
960 vm_isa_deassert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
961 {
962 struct vm_isa_irq isa_irq;
963
964 bzero(&isa_irq, sizeof(struct vm_isa_irq));
965 isa_irq.atpic_irq = atpic_irq;
966 isa_irq.ioapic_irq = ioapic_irq;
967
968 return (ioctl(ctx->fd, VM_ISA_DEASSERT_IRQ, &isa_irq));
969 }
970
971 int
972 vm_isa_pulse_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
973 {
974 struct vm_isa_irq isa_irq;
975
976 bzero(&isa_irq, sizeof(struct vm_isa_irq));
977 isa_irq.atpic_irq = atpic_irq;
978 isa_irq.ioapic_irq = ioapic_irq;
979
980 return (ioctl(ctx->fd, VM_ISA_PULSE_IRQ, &isa_irq));
981 }
982
983 int
984 vm_isa_set_irq_trigger(struct vmctx *ctx, int atpic_irq,
985 enum vm_intr_trigger trigger)
986 {
987 struct vm_isa_irq_trigger isa_irq_trigger;
988
989 bzero(&isa_irq_trigger, sizeof(struct vm_isa_irq_trigger));
990 isa_irq_trigger.atpic_irq = atpic_irq;
991 isa_irq_trigger.trigger = trigger;
992
993 return (ioctl(ctx->fd, VM_ISA_SET_IRQ_TRIGGER, &isa_irq_trigger));
994 }
995
996 int
997 vm_inject_nmi(struct vmctx *ctx, int vcpu)
998 {
999 struct vm_nmi vmnmi;
1000
1001 bzero(&vmnmi, sizeof(vmnmi));
1002 vmnmi.cpuid = vcpu;
1003
1004 return (ioctl(ctx->fd, VM_INJECT_NMI, &vmnmi));
1005 }
1006
1007 static const char *capstrmap[] = {
1008 [VM_CAP_HALT_EXIT] = "hlt_exit",
1009 [VM_CAP_MTRAP_EXIT] = "mtrap_exit",
1010 [VM_CAP_PAUSE_EXIT] = "pause_exit",
1011 #ifdef __FreeBSD__
1012 [VM_CAP_UNRESTRICTED_GUEST] = "unrestricted_guest",
1013 #endif
1014 [VM_CAP_ENABLE_INVPCID] = "enable_invpcid",
1015 [VM_CAP_BPT_EXIT] = "bpt_exit",
1016 };
1017
1018 int
1019 vm_capability_name2type(const char *capname)
1020 {
1021 int i;
1022
1023 for (i = 0; i < nitems(capstrmap); i++) {
1024 if (strcmp(capstrmap[i], capname) == 0)
1025 return (i);
1026 }
1027
1028 return (-1);
1029 }
1030
1031 const char *
1032 vm_capability_type2name(int type)
1033 {
1034 if (type >= 0 && type < nitems(capstrmap))
1035 return (capstrmap[type]);
1036
1037 return (NULL);
1038 }
1039
1040 int
1041 vm_get_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap,
1042 int *retval)
1043 {
1044 int error;
1045 struct vm_capability vmcap;
1046
1047 bzero(&vmcap, sizeof(vmcap));
1048 vmcap.cpuid = vcpu;
1049 vmcap.captype = cap;
1050
1051 error = ioctl(ctx->fd, VM_GET_CAPABILITY, &vmcap);
1052 *retval = vmcap.capval;
1053 return (error);
1054 }
1055
1056 int
1057 vm_set_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap, int val)
1058 {
1059 struct vm_capability vmcap;
1060
1061 bzero(&vmcap, sizeof(vmcap));
1062 vmcap.cpuid = vcpu;
1063 vmcap.captype = cap;
1064 vmcap.capval = val;
1065
1066 return (ioctl(ctx->fd, VM_SET_CAPABILITY, &vmcap));
1067 }
1068
1069 #ifdef __FreeBSD__
1070 int
1071 vm_assign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
1072 {
1073 struct vm_pptdev pptdev;
1074
1075 bzero(&pptdev, sizeof(pptdev));
1076 pptdev.bus = bus;
1077 pptdev.slot = slot;
1078 pptdev.func = func;
1079
1080 return (ioctl(ctx->fd, VM_BIND_PPTDEV, &pptdev));
1081 }
1082
1083 int
1084 vm_unassign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
1085 {
1086 struct vm_pptdev pptdev;
1087
1088 bzero(&pptdev, sizeof(pptdev));
1089 pptdev.bus = bus;
1090 pptdev.slot = slot;
1091 pptdev.func = func;
1092
1093 return (ioctl(ctx->fd, VM_UNBIND_PPTDEV, &pptdev));
1094 }
1095
1096 int
1097 vm_map_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func,
1098 vm_paddr_t gpa, size_t len, vm_paddr_t hpa)
1099 {
1100 struct vm_pptdev_mmio pptmmio;
1101
1102 bzero(&pptmmio, sizeof(pptmmio));
1103 pptmmio.bus = bus;
1104 pptmmio.slot = slot;
1105 pptmmio.func = func;
1106 pptmmio.gpa = gpa;
1107 pptmmio.len = len;
1108 pptmmio.hpa = hpa;
1109
1110 return (ioctl(ctx->fd, VM_MAP_PPTDEV_MMIO, &pptmmio));
1111 }
1112
1113 int
1114 vm_unmap_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func,
1115 vm_paddr_t gpa, size_t len)
1116 {
1117 struct vm_pptdev_mmio pptmmio;
1118
1119 bzero(&pptmmio, sizeof(pptmmio));
1120 pptmmio.bus = bus;
1121 pptmmio.slot = slot;
1122 pptmmio.func = func;
1123 pptmmio.gpa = gpa;
1124 pptmmio.len = len;
1125
1126 return (ioctl(ctx->fd, VM_UNMAP_PPTDEV_MMIO, &pptmmio));
1127 }
1128
1129 int
1130 vm_setup_pptdev_msi(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
1131 uint64_t addr, uint64_t msg, int numvec)
1132 {
1133 struct vm_pptdev_msi pptmsi;
1134
1135 bzero(&pptmsi, sizeof(pptmsi));
1136 pptmsi.vcpu = vcpu;
1137 pptmsi.bus = bus;
1138 pptmsi.slot = slot;
1139 pptmsi.func = func;
1140 pptmsi.msg = msg;
1141 pptmsi.addr = addr;
1142 pptmsi.numvec = numvec;
1143
1144 return (ioctl(ctx->fd, VM_PPTDEV_MSI, &pptmsi));
1145 }
1146
1147 int
1148 vm_setup_pptdev_msix(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
1149 int idx, uint64_t addr, uint64_t msg, uint32_t vector_control)
1150 {
1151 struct vm_pptdev_msix pptmsix;
1152
1153 bzero(&pptmsix, sizeof(pptmsix));
1154 pptmsix.vcpu = vcpu;
1155 pptmsix.bus = bus;
1156 pptmsix.slot = slot;
1157 pptmsix.func = func;
1158 pptmsix.idx = idx;
1159 pptmsix.msg = msg;
1160 pptmsix.addr = addr;
1161 pptmsix.vector_control = vector_control;
1162
1163 return ioctl(ctx->fd, VM_PPTDEV_MSIX, &pptmsix);
1164 }
1165
1166 int
1167 vm_get_pptdev_limits(struct vmctx *ctx, int bus, int slot, int func,
1168 int *msi_limit, int *msix_limit)
1169 {
1170 struct vm_pptdev_limits pptlimits;
1171 int error;
1172
1173 bzero(&pptlimits, sizeof (pptlimits));
1174 pptlimits.bus = bus;
1175 pptlimits.slot = slot;
1176 pptlimits.func = func;
1177
1178 error = ioctl(ctx->fd, VM_GET_PPTDEV_LIMITS, &pptlimits);
1179
1180 *msi_limit = pptlimits.msi_limit;
1181 *msix_limit = pptlimits.msix_limit;
1182
1183 return (error);
1184 }
1185
1186 int
1187 vm_disable_pptdev_msix(struct vmctx *ctx, int bus, int slot, int func)
1188 {
1189 struct vm_pptdev ppt;
1190
1191 bzero(&ppt, sizeof(ppt));
1192 ppt.bus = bus;
1193 ppt.slot = slot;
1194 ppt.func = func;
1195
1196 return ioctl(ctx->fd, VM_PPTDEV_DISABLE_MSIX, &ppt);
1197 }
1198
1199 #else /* __FreeBSD__ */
1200
1201 int
1202 vm_assign_pptdev(struct vmctx *ctx, int pptfd)
1203 {
1204 struct vm_pptdev pptdev;
1205
1206 pptdev.pptfd = pptfd;
1207 return (ioctl(ctx->fd, VM_BIND_PPTDEV, &pptdev));
1208 }
1209
1210 int
1211 vm_unassign_pptdev(struct vmctx *ctx, int pptfd)
1212 {
1213 struct vm_pptdev pptdev;
1214
1215 pptdev.pptfd = pptfd;
1216 return (ioctl(ctx->fd, VM_UNBIND_PPTDEV, &pptdev));
1217 }
1218
1219 int
1220 vm_map_pptdev_mmio(struct vmctx *ctx, int pptfd, vm_paddr_t gpa, size_t len,
1221 vm_paddr_t hpa)
1222 {
1223 struct vm_pptdev_mmio pptmmio;
1224
1225 pptmmio.pptfd = pptfd;
1226 pptmmio.gpa = gpa;
1227 pptmmio.len = len;
1228 pptmmio.hpa = hpa;
1229 return (ioctl(ctx->fd, VM_MAP_PPTDEV_MMIO, &pptmmio));
1230 }
1231
1232 int
1233 vm_unmap_pptdev_mmio(struct vmctx *ctx, int pptfd, vm_paddr_t gpa, size_t len)
1234 {
1235 struct vm_pptdev_mmio pptmmio;
1236
1237 bzero(&pptmmio, sizeof(pptmmio));
1238 pptmmio.pptfd = pptfd;
1239 pptmmio.gpa = gpa;
1240 pptmmio.len = len;
1241
1242 return (ioctl(ctx->fd, VM_UNMAP_PPTDEV_MMIO, &pptmmio));
1243 }
1244
1245 int
1246 vm_setup_pptdev_msi(struct vmctx *ctx, int vcpu, int pptfd, uint64_t addr,
1247 uint64_t msg, int numvec)
1248 {
1249 struct vm_pptdev_msi pptmsi;
1250
1251 pptmsi.vcpu = vcpu;
1252 pptmsi.pptfd = pptfd;
1253 pptmsi.msg = msg;
1254 pptmsi.addr = addr;
1255 pptmsi.numvec = numvec;
1256 return (ioctl(ctx->fd, VM_PPTDEV_MSI, &pptmsi));
1257 }
1258
1259 int
1260 vm_setup_pptdev_msix(struct vmctx *ctx, int vcpu, int pptfd, int idx,
1261 uint64_t addr, uint64_t msg, uint32_t vector_control)
1262 {
1263 struct vm_pptdev_msix pptmsix;
1264
1265 pptmsix.vcpu = vcpu;
1266 pptmsix.pptfd = pptfd;
1267 pptmsix.idx = idx;
1268 pptmsix.msg = msg;
1269 pptmsix.addr = addr;
1270 pptmsix.vector_control = vector_control;
1271 return ioctl(ctx->fd, VM_PPTDEV_MSIX, &pptmsix);
1272 }
1273
1274 int
1275 vm_get_pptdev_limits(struct vmctx *ctx, int pptfd, int *msi_limit,
1276 int *msix_limit)
1277 {
1278 struct vm_pptdev_limits pptlimits;
1279 int error;
1280
1281 bzero(&pptlimits, sizeof (pptlimits));
1282 pptlimits.pptfd = pptfd;
1283 error = ioctl(ctx->fd, VM_GET_PPTDEV_LIMITS, &pptlimits);
1284
1285 *msi_limit = pptlimits.msi_limit;
1286 *msix_limit = pptlimits.msix_limit;
1287 return (error);
1288 }
1289
1290 int
1291 vm_disable_pptdev_msix(struct vmctx *ctx, int pptfd)
1292 {
1293 struct vm_pptdev pptdev;
1294
1295 pptdev.pptfd = pptfd;
1296 return (ioctl(ctx->fd, VM_PPTDEV_DISABLE_MSIX, &pptdev));
1297 }
1298 #endif /* __FreeBSD__ */
1299
1300 uint64_t *
1301 vm_get_stats(struct vmctx *ctx, int vcpu, struct timeval *ret_tv,
1302 int *ret_entries)
1303 {
1304 int error;
1305
1306 static struct vm_stats vmstats;
1307
1308 vmstats.cpuid = vcpu;
1309
1310 error = ioctl(ctx->fd, VM_STATS_IOC, &vmstats);
1311 if (error == 0) {
1312 if (ret_entries)
1313 *ret_entries = vmstats.num_entries;
1314 if (ret_tv)
1315 *ret_tv = vmstats.tv;
1316 return (vmstats.statbuf);
1317 } else
1318 return (NULL);
1319 }
1320
1321 const char *
1322 vm_get_stat_desc(struct vmctx *ctx, int index)
1323 {
1324 static struct vm_stat_desc statdesc;
1325
1326 statdesc.index = index;
1327 if (ioctl(ctx->fd, VM_STAT_DESC, &statdesc) == 0)
1328 return (statdesc.desc);
1329 else
1330 return (NULL);
1331 }
1332
1333 int
1334 vm_get_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state *state)
1335 {
1336 int error;
1337 struct vm_x2apic x2apic;
1338
1339 bzero(&x2apic, sizeof(x2apic));
1340 x2apic.cpuid = vcpu;
1341
1342 error = ioctl(ctx->fd, VM_GET_X2APIC_STATE, &x2apic);
1343 *state = x2apic.state;
1344 return (error);
1345 }
1346
1347 int
1348 vm_set_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state state)
1349 {
1350 int error;
1351 struct vm_x2apic x2apic;
1352
1353 bzero(&x2apic, sizeof(x2apic));
1354 x2apic.cpuid = vcpu;
1355 x2apic.state = state;
1356
1357 error = ioctl(ctx->fd, VM_SET_X2APIC_STATE, &x2apic);
1358
1359 return (error);
1360 }
1361
1362 #ifndef __FreeBSD__
1363 int
1364 vcpu_reset(struct vmctx *vmctx, int vcpu)
1365 {
1366 struct vm_vcpu_reset vvr;
1367
1368 vvr.vcpuid = vcpu;
1369 vvr.kind = VRK_RESET;
1370
1371 return (ioctl(vmctx->fd, VM_RESET_CPU, &vvr));
1372 }
1373 #else /* __FreeBSD__ */
1374 /*
1375 * From Intel Vol 3a:
1376 * Table 9-1. IA-32 Processor States Following Power-up, Reset or INIT
1377 */
1378 int
1379 vcpu_reset(struct vmctx *vmctx, int vcpu)
1380 {
1381 int error;
1382 uint64_t rflags, rip, cr0, cr4, zero, desc_base, rdx;
1383 uint32_t desc_access, desc_limit;
1384 uint16_t sel;
1385
1386 zero = 0;
1387
1388 rflags = 0x2;
1389 error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RFLAGS, rflags);
1390 if (error)
1391 goto done;
1392
1393 rip = 0xfff0;
1394 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RIP, rip)) != 0)
1395 goto done;
1396
1397 cr0 = CR0_NE;
1398 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR0, cr0)) != 0)
1399 goto done;
1400
1401 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR3, zero)) != 0)
1402 goto done;
1403
1404 cr4 = 0;
1405 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR4, cr4)) != 0)
1406 goto done;
1407
1408 /*
1409 * CS: present, r/w, accessed, 16-bit, byte granularity, usable
1410 */
1411 desc_base = 0xffff0000;
1412 desc_limit = 0xffff;
1413 desc_access = 0x0093;
1414 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_CS,
1415 desc_base, desc_limit, desc_access);
1416 if (error)
1417 goto done;
1418
1419 sel = 0xf000;
1420 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CS, sel)) != 0)
1421 goto done;
1422
1423 /*
1424 * SS,DS,ES,FS,GS: present, r/w, accessed, 16-bit, byte granularity
1425 */
1426 desc_base = 0;
1427 desc_limit = 0xffff;
1428 desc_access = 0x0093;
1429 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_SS,
1430 desc_base, desc_limit, desc_access);
1431 if (error)
1432 goto done;
1433
1434 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_DS,
1435 desc_base, desc_limit, desc_access);
1436 if (error)
1437 goto done;
1438
1439 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_ES,
1440 desc_base, desc_limit, desc_access);
1441 if (error)
1442 goto done;
1443
1444 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_FS,
1445 desc_base, desc_limit, desc_access);
1446 if (error)
1447 goto done;
1448
1449 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GS,
1450 desc_base, desc_limit, desc_access);
1451 if (error)
1452 goto done;
1453
1454 sel = 0;
1455 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_SS, sel)) != 0)
1456 goto done;
1457 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_DS, sel)) != 0)
1458 goto done;
1459 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_ES, sel)) != 0)
1460 goto done;
1461 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_FS, sel)) != 0)
1462 goto done;
1463 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_GS, sel)) != 0)
1464 goto done;
1465
1466 /* General purpose registers */
1467 rdx = 0xf00;
1468 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RAX, zero)) != 0)
1469 goto done;
1470 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBX, zero)) != 0)
1471 goto done;
1472 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RCX, zero)) != 0)
1473 goto done;
1474 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDX, rdx)) != 0)
1475 goto done;
1476 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSI, zero)) != 0)
1477 goto done;
1478 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDI, zero)) != 0)
1479 goto done;
1480 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBP, zero)) != 0)
1481 goto done;
1482 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSP, zero)) != 0)
1483 goto done;
1484
1485 /* GDTR, IDTR */
1486 desc_base = 0;
1487 desc_limit = 0xffff;
1488 desc_access = 0;
1489 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GDTR,
1490 desc_base, desc_limit, desc_access);
1491 if (error != 0)
1492 goto done;
1493
1494 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_IDTR,
1495 desc_base, desc_limit, desc_access);
1496 if (error != 0)
1497 goto done;
1498
1499 /* TR */
1500 desc_base = 0;
1501 desc_limit = 0xffff;
1502 desc_access = 0x0000008b;
1503 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_TR, 0, 0, desc_access);
1504 if (error)
1505 goto done;
1506
1507 sel = 0;
1508 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_TR, sel)) != 0)
1509 goto done;
1510
1511 /* LDTR */
1512 desc_base = 0;
1513 desc_limit = 0xffff;
1514 desc_access = 0x00000082;
1515 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_LDTR, desc_base,
1516 desc_limit, desc_access);
1517 if (error)
1518 goto done;
1519
1520 sel = 0;
1521 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_LDTR, 0)) != 0)
1522 goto done;
1523
1524 /* XXX cr2, debug registers */
1525
1526 error = 0;
1527 done:
1528 return (error);
1529 }
1530 #endif /* __FreeBSD__ */
1531
1532 int
1533 vm_get_gpa_pmap(struct vmctx *ctx, uint64_t gpa, uint64_t *pte, int *num)
1534 {
1535 int error, i;
1536 struct vm_gpa_pte gpapte;
1537
1538 bzero(&gpapte, sizeof(gpapte));
1539 gpapte.gpa = gpa;
1540
1541 error = ioctl(ctx->fd, VM_GET_GPA_PMAP, &gpapte);
1542
1543 if (error == 0) {
1544 *num = gpapte.ptenum;
1545 for (i = 0; i < gpapte.ptenum; i++)
1546 pte[i] = gpapte.pte[i];
1547 }
1548
1549 return (error);
1550 }
1551
1552 int
1553 vm_get_hpet_capabilities(struct vmctx *ctx, uint32_t *capabilities)
1554 {
1555 int error;
1556 struct vm_hpet_cap cap;
1557
1558 bzero(&cap, sizeof(struct vm_hpet_cap));
1559 error = ioctl(ctx->fd, VM_GET_HPET_CAPABILITIES, &cap);
1560 if (capabilities != NULL)
1561 *capabilities = cap.capabilities;
1562 return (error);
1563 }
1564
1565 int
1566 vm_gla2gpa(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1567 uint64_t gla, int prot, uint64_t *gpa, int *fault)
1568 {
1569 struct vm_gla2gpa gg;
1570 int error;
1571
1572 bzero(&gg, sizeof(struct vm_gla2gpa));
1573 gg.vcpuid = vcpu;
1574 gg.prot = prot;
1575 gg.gla = gla;
1576 gg.paging = *paging;
1577
1578 error = ioctl(ctx->fd, VM_GLA2GPA, &gg);
1579 if (error == 0) {
1580 *fault = gg.fault;
1581 *gpa = gg.gpa;
1582 }
1583 return (error);
1584 }
1585
1586 int
1587 vm_gla2gpa_nofault(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1588 uint64_t gla, int prot, uint64_t *gpa, int *fault)
1589 {
1590 struct vm_gla2gpa gg;
1591 int error;
1592
1593 bzero(&gg, sizeof(struct vm_gla2gpa));
1594 gg.vcpuid = vcpu;
1595 gg.prot = prot;
1596 gg.gla = gla;
1597 gg.paging = *paging;
1598
1599 error = ioctl(ctx->fd, VM_GLA2GPA_NOFAULT, &gg);
1600 if (error == 0) {
1601 *fault = gg.fault;
1602 *gpa = gg.gpa;
1603 }
1604 return (error);
1605 }
1606
1607 #ifndef min
1608 #define min(a,b) (((a) < (b)) ? (a) : (b))
1609 #endif
1610
1611 int
1612 vm_copy_setup(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1613 uint64_t gla, size_t len, int prot, struct iovec *iov, int iovcnt,
1614 int *fault)
1615 {
1616 void *va;
1617 uint64_t gpa;
1618 int error, i, n, off;
1619
1620 for (i = 0; i < iovcnt; i++) {
1621 iov[i].iov_base = 0;
1622 iov[i].iov_len = 0;
1623 }
1624
1625 while (len) {
1626 assert(iovcnt > 0);
1627 error = vm_gla2gpa(ctx, vcpu, paging, gla, prot, &gpa, fault);
1628 if (error || *fault)
1629 return (error);
1630
1631 off = gpa & PAGE_MASK;
1632 n = min(len, PAGE_SIZE - off);
1633
1634 va = vm_map_gpa(ctx, gpa, n);
1635 if (va == NULL)
1636 return (EFAULT);
1637
1638 iov->iov_base = va;
1639 iov->iov_len = n;
1640 iov++;
1641 iovcnt--;
1642
1643 gla += n;
1644 len -= n;
1645 }
1646 return (0);
1647 }
1648
1649 void
1650 vm_copy_teardown(struct vmctx *ctx, int vcpu, struct iovec *iov, int iovcnt)
1651 {
1652
1653 return;
1654 }
1655
1656 void
1657 vm_copyin(struct vmctx *ctx, int vcpu, struct iovec *iov, void *vp, size_t len)
1658 {
1659 const char *src;
1660 char *dst;
1661 size_t n;
1662
1663 dst = vp;
1664 while (len) {
1665 assert(iov->iov_len);
1666 n = min(len, iov->iov_len);
1667 src = iov->iov_base;
1668 bcopy(src, dst, n);
1669
1670 iov++;
1671 dst += n;
1672 len -= n;
1673 }
1674 }
1675
1676 void
1677 vm_copyout(struct vmctx *ctx, int vcpu, const void *vp, struct iovec *iov,
1678 size_t len)
1679 {
1680 const char *src;
1681 char *dst;
1682 size_t n;
1683
1684 src = vp;
1685 while (len) {
1686 assert(iov->iov_len);
1687 n = min(len, iov->iov_len);
1688 dst = iov->iov_base;
1689 bcopy(src, dst, n);
1690
1691 iov++;
1692 src += n;
1693 len -= n;
1694 }
1695 }
1696
1697 static int
1698 vm_get_cpus(struct vmctx *ctx, int which, cpuset_t *cpus)
1699 {
1700 struct vm_cpuset vm_cpuset;
1701 int error;
1702
1703 bzero(&vm_cpuset, sizeof(struct vm_cpuset));
1704 vm_cpuset.which = which;
1705 vm_cpuset.cpusetsize = sizeof(cpuset_t);
1706 vm_cpuset.cpus = cpus;
1707
1708 error = ioctl(ctx->fd, VM_GET_CPUS, &vm_cpuset);
1709 return (error);
1710 }
1711
1712 int
1713 vm_active_cpus(struct vmctx *ctx, cpuset_t *cpus)
1714 {
1715
1716 return (vm_get_cpus(ctx, VM_ACTIVE_CPUS, cpus));
1717 }
1718
1719 int
1720 vm_suspended_cpus(struct vmctx *ctx, cpuset_t *cpus)
1721 {
1722
1723 return (vm_get_cpus(ctx, VM_SUSPENDED_CPUS, cpus));
1724 }
1725
1726 int
1727 vm_debug_cpus(struct vmctx *ctx, cpuset_t *cpus)
1728 {
1729
1730 return (vm_get_cpus(ctx, VM_DEBUG_CPUS, cpus));
1731 }
1732
1733 int
1734 vm_activate_cpu(struct vmctx *ctx, int vcpu)
1735 {
1736 struct vm_activate_cpu ac;
1737 int error;
1738
1739 bzero(&ac, sizeof(struct vm_activate_cpu));
1740 ac.vcpuid = vcpu;
1741 error = ioctl(ctx->fd, VM_ACTIVATE_CPU, &ac);
1742 return (error);
1743 }
1744
1745 int
1746 vm_suspend_cpu(struct vmctx *ctx, int vcpu)
1747 {
1748 struct vm_activate_cpu ac;
1749 int error;
1750
1751 bzero(&ac, sizeof(struct vm_activate_cpu));
1752 ac.vcpuid = vcpu;
1753 error = ioctl(ctx->fd, VM_SUSPEND_CPU, &ac);
1754 return (error);
1755 }
1756
1757 int
1758 vm_resume_cpu(struct vmctx *ctx, int vcpu)
1759 {
1760 struct vm_activate_cpu ac;
1761 int error;
1762
1763 bzero(&ac, sizeof(struct vm_activate_cpu));
1764 ac.vcpuid = vcpu;
1765 error = ioctl(ctx->fd, VM_RESUME_CPU, &ac);
1766 return (error);
1767 }
1768
1769 int
1770 vm_get_intinfo(struct vmctx *ctx, int vcpu, uint64_t *info1, uint64_t *info2)
1771 {
1772 struct vm_intinfo vmii;
1773 int error;
1774
1775 bzero(&vmii, sizeof(struct vm_intinfo));
1776 vmii.vcpuid = vcpu;
1777 error = ioctl(ctx->fd, VM_GET_INTINFO, &vmii);
1778 if (error == 0) {
1779 *info1 = vmii.info1;
1780 *info2 = vmii.info2;
1781 }
1782 return (error);
1783 }
1784
1785 int
1786 vm_set_intinfo(struct vmctx *ctx, int vcpu, uint64_t info1)
1787 {
1788 struct vm_intinfo vmii;
1789 int error;
1790
1791 bzero(&vmii, sizeof(struct vm_intinfo));
1792 vmii.vcpuid = vcpu;
1793 vmii.info1 = info1;
1794 error = ioctl(ctx->fd, VM_SET_INTINFO, &vmii);
1795 return (error);
1796 }
1797
1798 int
1799 vm_rtc_write(struct vmctx *ctx, int offset, uint8_t value)
1800 {
1801 struct vm_rtc_data rtcdata;
1802 int error;
1803
1804 bzero(&rtcdata, sizeof(struct vm_rtc_data));
1805 rtcdata.offset = offset;
1806 rtcdata.value = value;
1807 error = ioctl(ctx->fd, VM_RTC_WRITE, &rtcdata);
1808 return (error);
1809 }
1810
1811 int
1812 vm_rtc_read(struct vmctx *ctx, int offset, uint8_t *retval)
1813 {
1814 struct vm_rtc_data rtcdata;
1815 int error;
1816
1817 bzero(&rtcdata, sizeof(struct vm_rtc_data));
1818 rtcdata.offset = offset;
1819 error = ioctl(ctx->fd, VM_RTC_READ, &rtcdata);
1820 if (error == 0)
1821 *retval = rtcdata.value;
1822 return (error);
1823 }
1824
1825 int
1826 vm_rtc_settime(struct vmctx *ctx, time_t secs)
1827 {
1828 struct vm_rtc_time rtctime;
1829 int error;
1830
1831 bzero(&rtctime, sizeof(struct vm_rtc_time));
1832 rtctime.secs = secs;
1833 error = ioctl(ctx->fd, VM_RTC_SETTIME, &rtctime);
1834 return (error);
1835 }
1836
1837 int
1838 vm_rtc_gettime(struct vmctx *ctx, time_t *secs)
1839 {
1840 struct vm_rtc_time rtctime;
1841 int error;
1842
1843 bzero(&rtctime, sizeof(struct vm_rtc_time));
1844 error = ioctl(ctx->fd, VM_RTC_GETTIME, &rtctime);
1845 if (error == 0)
1846 *secs = rtctime.secs;
1847 return (error);
1848 }
1849
1850 int
1851 vm_restart_instruction(void *arg, int vcpu)
1852 {
1853 struct vmctx *ctx = arg;
1854
1855 return (ioctl(ctx->fd, VM_RESTART_INSTRUCTION, &vcpu));
1856 }
1857
1858 int
1859 vm_set_topology(struct vmctx *ctx,
1860 uint16_t sockets, uint16_t cores, uint16_t threads, uint16_t maxcpus)
1861 {
1862 struct vm_cpu_topology topology;
1863
1864 bzero(&topology, sizeof (struct vm_cpu_topology));
1865 topology.sockets = sockets;
1866 topology.cores = cores;
1867 topology.threads = threads;
1868 topology.maxcpus = maxcpus;
1869 return (ioctl(ctx->fd, VM_SET_TOPOLOGY, &topology));
1870 }
1871
1872 int
1873 vm_get_topology(struct vmctx *ctx,
1874 uint16_t *sockets, uint16_t *cores, uint16_t *threads, uint16_t *maxcpus)
1875 {
1876 struct vm_cpu_topology topology;
1877 int error;
1878
1879 bzero(&topology, sizeof (struct vm_cpu_topology));
1880 error = ioctl(ctx->fd, VM_GET_TOPOLOGY, &topology);
1881 if (error == 0) {
1882 *sockets = topology.sockets;
1883 *cores = topology.cores;
1884 *threads = topology.threads;
1885 *maxcpus = topology.maxcpus;
1886 }
1887 return (error);
1888 }
1889
1890 int
1891 vm_get_device_fd(struct vmctx *ctx)
1892 {
1893
1894 return (ctx->fd);
1895 }
1896
1897 #ifndef __FreeBSD__
1898 int
1899 vm_pmtmr_set_location(struct vmctx *ctx, uint16_t ioport)
1900 {
1901 return (ioctl(ctx->fd, VM_PMTMR_LOCATE, ioport));
1902 }
1903
1904 int
1905 vm_wrlock_cycle(struct vmctx *ctx)
1906 {
1907 if (ioctl(ctx->fd, VM_WRLOCK_CYCLE, 0) != 0) {
1908 return (errno);
1909 }
1910 return (0);
1911 }
1912
1913 int
1914 vm_get_run_state(struct vmctx *ctx, int vcpu, enum vcpu_run_state *state,
1915 uint8_t *sipi_vector)
1916 {
1917 struct vm_run_state data;
1918
1919 data.vcpuid = vcpu;
1920 if (ioctl(ctx->fd, VM_GET_RUN_STATE, &data) != 0) {
1921 return (errno);
1922 }
1923
1924 *state = data.state;
1925 *sipi_vector = data.sipi_vector;
1926 return (0);
1927 }
1928
1929 int
1930 vm_set_run_state(struct vmctx *ctx, int vcpu, enum vcpu_run_state state,
1931 uint8_t sipi_vector)
1932 {
1933 struct vm_run_state data;
1934
1935 data.vcpuid = vcpu;
1936 data.state = state;
1937 data.sipi_vector = sipi_vector;
1938 if (ioctl(ctx->fd, VM_SET_RUN_STATE, &data) != 0) {
1939 return (errno);
1940 }
1941
1942 return (0);
1943 }
1944
1945 int
1946 vm_arc_resv(struct vmctx *ctx, size_t len)
1947 {
1948 if (ioctl(ctx->fd, VM_ARC_RESV, (uint64_t)len) != 0) {
1949 return (errno);
1950 }
1951 return (0);
1952 }
1953 #endif /* __FreeBSD__ */
1954
1955 #ifdef __FreeBSD__
1956 const cap_ioctl_t *
1957 vm_get_ioctls(size_t *len)
1958 {
1959 cap_ioctl_t *cmds;
1960 /* keep in sync with machine/vmm_dev.h */
1961 static const cap_ioctl_t vm_ioctl_cmds[] = { VM_RUN, VM_SUSPEND, VM_REINIT,
1962 VM_ALLOC_MEMSEG, VM_GET_MEMSEG, VM_MMAP_MEMSEG, VM_MMAP_MEMSEG,
1963 VM_MMAP_GETNEXT, VM_MUNMAP_MEMSEG, VM_SET_REGISTER, VM_GET_REGISTER,
1964 VM_SET_SEGMENT_DESCRIPTOR, VM_GET_SEGMENT_DESCRIPTOR,
1965 VM_SET_REGISTER_SET, VM_GET_REGISTER_SET,
1966 VM_SET_KERNEMU_DEV, VM_GET_KERNEMU_DEV,
1967 VM_INJECT_EXCEPTION, VM_LAPIC_IRQ, VM_LAPIC_LOCAL_IRQ,
1968 VM_LAPIC_MSI, VM_IOAPIC_ASSERT_IRQ, VM_IOAPIC_DEASSERT_IRQ,
1969 VM_IOAPIC_PULSE_IRQ, VM_IOAPIC_PINCOUNT, VM_ISA_ASSERT_IRQ,
1970 VM_ISA_DEASSERT_IRQ, VM_ISA_PULSE_IRQ, VM_ISA_SET_IRQ_TRIGGER,
1971 VM_SET_CAPABILITY, VM_GET_CAPABILITY, VM_BIND_PPTDEV,
1972 VM_UNBIND_PPTDEV, VM_MAP_PPTDEV_MMIO, VM_PPTDEV_MSI,
1973 VM_PPTDEV_MSIX, VM_UNMAP_PPTDEV_MMIO, VM_PPTDEV_DISABLE_MSIX,
1974 VM_INJECT_NMI, VM_STATS, VM_STAT_DESC,
1975 VM_SET_X2APIC_STATE, VM_GET_X2APIC_STATE,
1976 VM_GET_HPET_CAPABILITIES, VM_GET_GPA_PMAP, VM_GLA2GPA,
1977 VM_GLA2GPA_NOFAULT,
1978 VM_ACTIVATE_CPU, VM_GET_CPUS, VM_SUSPEND_CPU, VM_RESUME_CPU,
1979 VM_SET_INTINFO, VM_GET_INTINFO,
1980 VM_RTC_WRITE, VM_RTC_READ, VM_RTC_SETTIME, VM_RTC_GETTIME,
1981 VM_RESTART_INSTRUCTION, VM_SET_TOPOLOGY, VM_GET_TOPOLOGY };
1982
1983 if (len == NULL) {
1984 cmds = malloc(sizeof(vm_ioctl_cmds));
1985 if (cmds == NULL)
1986 return (NULL);
1987 bcopy(vm_ioctl_cmds, cmds, sizeof(vm_ioctl_cmds));
1988 return (cmds);
1989 }
1990
1991 *len = nitems(vm_ioctl_cmds);
1992 return (NULL);
1993 }
1994 #endif /* __FreeBSD__ */