xref: /freebsd/lib/libvmmapi/vmmapi.c (revision b51f459a2098622c31ed54f5c1bf0e03efce403b)
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 #include <sys/cdefs.h>
32 __FBSDID("$FreeBSD$");
33 
34 #include <sys/param.h>
35 #include <sys/sysctl.h>
36 #include <sys/ioctl.h>
37 #include <sys/linker.h>
38 #include <sys/mman.h>
39 #include <sys/module.h>
40 #include <sys/_iovec.h>
41 #include <sys/cpuset.h>
42 
43 #include <x86/segments.h>
44 #include <machine/specialreg.h>
45 
46 #include <errno.h>
47 #include <stdbool.h>
48 #include <stdio.h>
49 #include <stdlib.h>
50 #include <assert.h>
51 #include <string.h>
52 #include <fcntl.h>
53 #include <unistd.h>
54 
55 #include <libutil.h>
56 
57 #include <vm/vm.h>
58 #include <machine/vmm.h>
59 #include <machine/vmm_dev.h>
60 #include <machine/vmm_snapshot.h>
61 
62 #include "vmmapi.h"
63 
64 #define	MB	(1024 * 1024UL)
65 #define	GB	(1024 * 1024 * 1024UL)
66 
67 /*
68  * Size of the guard region before and after the virtual address space
69  * mapping the guest physical memory. This must be a multiple of the
70  * superpage size for performance reasons.
71  */
72 #define	VM_MMAP_GUARD_SIZE	(4 * MB)
73 
74 #define	PROT_RW		(PROT_READ | PROT_WRITE)
75 #define	PROT_ALL	(PROT_READ | PROT_WRITE | PROT_EXEC)
76 
77 struct vmctx {
78 	int	fd;
79 	uint32_t lowmem_limit;
80 	int	memflags;
81 	size_t	lowmem;
82 	size_t	highmem;
83 	char	*baseaddr;
84 	char	*name;
85 };
86 
87 #define	CREATE(x)  sysctlbyname("hw.vmm.create", NULL, NULL, (x), strlen((x)))
88 #define	DESTROY(x) sysctlbyname("hw.vmm.destroy", NULL, NULL, (x), strlen((x)))
89 
90 static int
91 vm_device_open(const char *name)
92 {
93 	int fd, len;
94 	char *vmfile;
95 
96 	len = strlen("/dev/vmm/") + strlen(name) + 1;
97 	vmfile = malloc(len);
98 	assert(vmfile != NULL);
99 	snprintf(vmfile, len, "/dev/vmm/%s", name);
100 
101 	/* Open the device file */
102 	fd = open(vmfile, O_RDWR, 0);
103 
104 	free(vmfile);
105 	return (fd);
106 }
107 
108 int
109 vm_create(const char *name)
110 {
111 	/* Try to load vmm(4) module before creating a guest. */
112 	if (modfind("vmm") < 0)
113 		kldload("vmm");
114 	return (CREATE((char *)name));
115 }
116 
117 struct vmctx *
118 vm_open(const char *name)
119 {
120 	struct vmctx *vm;
121 
122 	vm = malloc(sizeof(struct vmctx) + strlen(name) + 1);
123 	assert(vm != NULL);
124 
125 	vm->fd = -1;
126 	vm->memflags = 0;
127 	vm->lowmem_limit = 3 * GB;
128 	vm->name = (char *)(vm + 1);
129 	strcpy(vm->name, name);
130 
131 	if ((vm->fd = vm_device_open(vm->name)) < 0)
132 		goto err;
133 
134 	return (vm);
135 err:
136 	free(vm);
137 	return (NULL);
138 }
139 
140 void
141 vm_destroy(struct vmctx *vm)
142 {
143 	assert(vm != NULL);
144 
145 	if (vm->fd >= 0)
146 		close(vm->fd);
147 	DESTROY(vm->name);
148 
149 	free(vm);
150 }
151 
152 int
153 vm_parse_memsize(const char *optarg, size_t *ret_memsize)
154 {
155 	char *endptr;
156 	size_t optval;
157 	int error;
158 
159 	optval = strtoul(optarg, &endptr, 0);
160 	if (*optarg != '\0' && *endptr == '\0') {
161 		/*
162 		 * For the sake of backward compatibility if the memory size
163 		 * specified on the command line is less than a megabyte then
164 		 * it is interpreted as being in units of MB.
165 		 */
166 		if (optval < MB)
167 			optval *= MB;
168 		*ret_memsize = optval;
169 		error = 0;
170 	} else
171 		error = expand_number(optarg, ret_memsize);
172 
173 	return (error);
174 }
175 
176 uint32_t
177 vm_get_lowmem_limit(struct vmctx *ctx)
178 {
179 
180 	return (ctx->lowmem_limit);
181 }
182 
183 void
184 vm_set_lowmem_limit(struct vmctx *ctx, uint32_t limit)
185 {
186 
187 	ctx->lowmem_limit = limit;
188 }
189 
190 void
191 vm_set_memflags(struct vmctx *ctx, int flags)
192 {
193 
194 	ctx->memflags = flags;
195 }
196 
197 int
198 vm_get_memflags(struct vmctx *ctx)
199 {
200 
201 	return (ctx->memflags);
202 }
203 
204 /*
205  * Map segment 'segid' starting at 'off' into guest address range [gpa,gpa+len).
206  */
207 int
208 vm_mmap_memseg(struct vmctx *ctx, vm_paddr_t gpa, int segid, vm_ooffset_t off,
209     size_t len, int prot)
210 {
211 	struct vm_memmap memmap;
212 	int error, flags;
213 
214 	memmap.gpa = gpa;
215 	memmap.segid = segid;
216 	memmap.segoff = off;
217 	memmap.len = len;
218 	memmap.prot = prot;
219 	memmap.flags = 0;
220 
221 	if (ctx->memflags & VM_MEM_F_WIRED)
222 		memmap.flags |= VM_MEMMAP_F_WIRED;
223 
224 	/*
225 	 * If this mapping already exists then don't create it again. This
226 	 * is the common case for SYSMEM mappings created by bhyveload(8).
227 	 */
228 	error = vm_mmap_getnext(ctx, &gpa, &segid, &off, &len, &prot, &flags);
229 	if (error == 0 && gpa == memmap.gpa) {
230 		if (segid != memmap.segid || off != memmap.segoff ||
231 		    prot != memmap.prot || flags != memmap.flags) {
232 			errno = EEXIST;
233 			return (-1);
234 		} else {
235 			return (0);
236 		}
237 	}
238 
239 	error = ioctl(ctx->fd, VM_MMAP_MEMSEG, &memmap);
240 	return (error);
241 }
242 
243 int
244 vm_get_guestmem_from_ctx(struct vmctx *ctx, char **guest_baseaddr,
245     size_t *lowmem_size, size_t *highmem_size)
246 {
247 
248 	*guest_baseaddr = ctx->baseaddr;
249 	*lowmem_size = ctx->lowmem;
250 	*highmem_size = ctx->highmem;
251 	return (0);
252 }
253 
254 int
255 vm_munmap_memseg(struct vmctx *ctx, vm_paddr_t gpa, size_t len)
256 {
257 	struct vm_munmap munmap;
258 	int error;
259 
260 	munmap.gpa = gpa;
261 	munmap.len = len;
262 
263 	error = ioctl(ctx->fd, VM_MUNMAP_MEMSEG, &munmap);
264 	return (error);
265 }
266 
267 int
268 vm_mmap_getnext(struct vmctx *ctx, vm_paddr_t *gpa, int *segid,
269     vm_ooffset_t *segoff, size_t *len, int *prot, int *flags)
270 {
271 	struct vm_memmap memmap;
272 	int error;
273 
274 	bzero(&memmap, sizeof(struct vm_memmap));
275 	memmap.gpa = *gpa;
276 	error = ioctl(ctx->fd, VM_MMAP_GETNEXT, &memmap);
277 	if (error == 0) {
278 		*gpa = memmap.gpa;
279 		*segid = memmap.segid;
280 		*segoff = memmap.segoff;
281 		*len = memmap.len;
282 		*prot = memmap.prot;
283 		*flags = memmap.flags;
284 	}
285 	return (error);
286 }
287 
288 /*
289  * Return 0 if the segments are identical and non-zero otherwise.
290  *
291  * This is slightly complicated by the fact that only device memory segments
292  * are named.
293  */
294 static int
295 cmpseg(size_t len, const char *str, size_t len2, const char *str2)
296 {
297 
298 	if (len == len2) {
299 		if ((!str && !str2) || (str && str2 && !strcmp(str, str2)))
300 			return (0);
301 	}
302 	return (-1);
303 }
304 
305 static int
306 vm_alloc_memseg(struct vmctx *ctx, int segid, size_t len, const char *name)
307 {
308 	struct vm_memseg memseg;
309 	size_t n;
310 	int error;
311 
312 	/*
313 	 * If the memory segment has already been created then just return.
314 	 * This is the usual case for the SYSMEM segment created by userspace
315 	 * loaders like bhyveload(8).
316 	 */
317 	error = vm_get_memseg(ctx, segid, &memseg.len, memseg.name,
318 	    sizeof(memseg.name));
319 	if (error)
320 		return (error);
321 
322 	if (memseg.len != 0) {
323 		if (cmpseg(len, name, memseg.len, VM_MEMSEG_NAME(&memseg))) {
324 			errno = EINVAL;
325 			return (-1);
326 		} else {
327 			return (0);
328 		}
329 	}
330 
331 	bzero(&memseg, sizeof(struct vm_memseg));
332 	memseg.segid = segid;
333 	memseg.len = len;
334 	if (name != NULL) {
335 		n = strlcpy(memseg.name, name, sizeof(memseg.name));
336 		if (n >= sizeof(memseg.name)) {
337 			errno = ENAMETOOLONG;
338 			return (-1);
339 		}
340 	}
341 
342 	error = ioctl(ctx->fd, VM_ALLOC_MEMSEG, &memseg);
343 	return (error);
344 }
345 
346 int
347 vm_get_memseg(struct vmctx *ctx, int segid, size_t *lenp, char *namebuf,
348     size_t bufsize)
349 {
350 	struct vm_memseg memseg;
351 	size_t n;
352 	int error;
353 
354 	memseg.segid = segid;
355 	error = ioctl(ctx->fd, VM_GET_MEMSEG, &memseg);
356 	if (error == 0) {
357 		*lenp = memseg.len;
358 		n = strlcpy(namebuf, memseg.name, bufsize);
359 		if (n >= bufsize) {
360 			errno = ENAMETOOLONG;
361 			error = -1;
362 		}
363 	}
364 	return (error);
365 }
366 
367 static int
368 setup_memory_segment(struct vmctx *ctx, vm_paddr_t gpa, size_t len, char *base)
369 {
370 	char *ptr;
371 	int error, flags;
372 
373 	/* Map 'len' bytes starting at 'gpa' in the guest address space */
374 	error = vm_mmap_memseg(ctx, gpa, VM_SYSMEM, gpa, len, PROT_ALL);
375 	if (error)
376 		return (error);
377 
378 	flags = MAP_SHARED | MAP_FIXED;
379 	if ((ctx->memflags & VM_MEM_F_INCORE) == 0)
380 		flags |= MAP_NOCORE;
381 
382 	/* mmap into the process address space on the host */
383 	ptr = mmap(base + gpa, len, PROT_RW, flags, ctx->fd, gpa);
384 	if (ptr == MAP_FAILED)
385 		return (-1);
386 
387 	return (0);
388 }
389 
390 int
391 vm_setup_memory(struct vmctx *ctx, size_t memsize, enum vm_mmap_style vms)
392 {
393 	size_t objsize, len;
394 	vm_paddr_t gpa;
395 	char *baseaddr, *ptr;
396 	int error;
397 
398 	assert(vms == VM_MMAP_ALL);
399 
400 	/*
401 	 * If 'memsize' cannot fit entirely in the 'lowmem' segment then
402 	 * create another 'highmem' segment above 4GB for the remainder.
403 	 */
404 	if (memsize > ctx->lowmem_limit) {
405 		ctx->lowmem = ctx->lowmem_limit;
406 		ctx->highmem = memsize - ctx->lowmem_limit;
407 		objsize = 4*GB + ctx->highmem;
408 	} else {
409 		ctx->lowmem = memsize;
410 		ctx->highmem = 0;
411 		objsize = ctx->lowmem;
412 	}
413 
414 	error = vm_alloc_memseg(ctx, VM_SYSMEM, objsize, NULL);
415 	if (error)
416 		return (error);
417 
418 	/*
419 	 * Stake out a contiguous region covering the guest physical memory
420 	 * and the adjoining guard regions.
421 	 */
422 	len = VM_MMAP_GUARD_SIZE + objsize + VM_MMAP_GUARD_SIZE;
423 	ptr = mmap(NULL, len, PROT_NONE, MAP_GUARD | MAP_ALIGNED_SUPER, -1, 0);
424 	if (ptr == MAP_FAILED)
425 		return (-1);
426 
427 	baseaddr = ptr + VM_MMAP_GUARD_SIZE;
428 	if (ctx->highmem > 0) {
429 		gpa = 4*GB;
430 		len = ctx->highmem;
431 		error = setup_memory_segment(ctx, gpa, len, baseaddr);
432 		if (error)
433 			return (error);
434 	}
435 
436 	if (ctx->lowmem > 0) {
437 		gpa = 0;
438 		len = ctx->lowmem;
439 		error = setup_memory_segment(ctx, gpa, len, baseaddr);
440 		if (error)
441 			return (error);
442 	}
443 
444 	ctx->baseaddr = baseaddr;
445 
446 	return (0);
447 }
448 
449 /*
450  * Returns a non-NULL pointer if [gaddr, gaddr+len) is entirely contained in
451  * the lowmem or highmem regions.
452  *
453  * In particular return NULL if [gaddr, gaddr+len) falls in guest MMIO region.
454  * The instruction emulation code depends on this behavior.
455  */
456 void *
457 vm_map_gpa(struct vmctx *ctx, vm_paddr_t gaddr, size_t len)
458 {
459 
460 	if (ctx->lowmem > 0) {
461 		if (gaddr < ctx->lowmem && len <= ctx->lowmem &&
462 		    gaddr + len <= ctx->lowmem)
463 			return (ctx->baseaddr + gaddr);
464 	}
465 
466 	if (ctx->highmem > 0) {
467                 if (gaddr >= 4*GB) {
468 			if (gaddr < 4*GB + ctx->highmem &&
469 			    len <= ctx->highmem &&
470 			    gaddr + len <= 4*GB + ctx->highmem)
471 				return (ctx->baseaddr + gaddr);
472 		}
473 	}
474 
475 	return (NULL);
476 }
477 
478 vm_paddr_t
479 vm_rev_map_gpa(struct vmctx *ctx, void *addr)
480 {
481 	vm_paddr_t offaddr;
482 
483 	offaddr = (char *)addr - ctx->baseaddr;
484 
485 	if (ctx->lowmem > 0)
486 		if (offaddr >= 0 && offaddr <= ctx->lowmem)
487 			return (offaddr);
488 
489 	if (ctx->highmem > 0)
490 		if (offaddr >= 4*GB && offaddr < 4*GB + ctx->highmem)
491 			return (offaddr);
492 
493 	return ((vm_paddr_t)-1);
494 }
495 
496 /* TODO: maximum size for vmname */
497 int
498 vm_get_name(struct vmctx *ctx, char *buf, size_t max_len)
499 {
500 
501 	if (strlcpy(buf, ctx->name, max_len) >= max_len)
502 		return (EINVAL);
503 	return (0);
504 }
505 
506 size_t
507 vm_get_lowmem_size(struct vmctx *ctx)
508 {
509 
510 	return (ctx->lowmem);
511 }
512 
513 size_t
514 vm_get_highmem_size(struct vmctx *ctx)
515 {
516 
517 	return (ctx->highmem);
518 }
519 
520 void *
521 vm_create_devmem(struct vmctx *ctx, int segid, const char *name, size_t len)
522 {
523 	char pathname[MAXPATHLEN];
524 	size_t len2;
525 	char *base, *ptr;
526 	int fd, error, flags;
527 
528 	fd = -1;
529 	ptr = MAP_FAILED;
530 	if (name == NULL || strlen(name) == 0) {
531 		errno = EINVAL;
532 		goto done;
533 	}
534 
535 	error = vm_alloc_memseg(ctx, segid, len, name);
536 	if (error)
537 		goto done;
538 
539 	strlcpy(pathname, "/dev/vmm.io/", sizeof(pathname));
540 	strlcat(pathname, ctx->name, sizeof(pathname));
541 	strlcat(pathname, ".", sizeof(pathname));
542 	strlcat(pathname, name, sizeof(pathname));
543 
544 	fd = open(pathname, O_RDWR);
545 	if (fd < 0)
546 		goto done;
547 
548 	/*
549 	 * Stake out a contiguous region covering the device memory and the
550 	 * adjoining guard regions.
551 	 */
552 	len2 = VM_MMAP_GUARD_SIZE + len + VM_MMAP_GUARD_SIZE;
553 	base = mmap(NULL, len2, PROT_NONE, MAP_GUARD | MAP_ALIGNED_SUPER, -1,
554 	    0);
555 	if (base == MAP_FAILED)
556 		goto done;
557 
558 	flags = MAP_SHARED | MAP_FIXED;
559 	if ((ctx->memflags & VM_MEM_F_INCORE) == 0)
560 		flags |= MAP_NOCORE;
561 
562 	/* mmap the devmem region in the host address space */
563 	ptr = mmap(base + VM_MMAP_GUARD_SIZE, len, PROT_RW, flags, fd, 0);
564 done:
565 	if (fd >= 0)
566 		close(fd);
567 	return (ptr);
568 }
569 
570 int
571 vm_set_desc(struct vmctx *ctx, int vcpu, int reg,
572 	    uint64_t base, uint32_t limit, uint32_t access)
573 {
574 	int error;
575 	struct vm_seg_desc vmsegdesc;
576 
577 	bzero(&vmsegdesc, sizeof(vmsegdesc));
578 	vmsegdesc.cpuid = vcpu;
579 	vmsegdesc.regnum = reg;
580 	vmsegdesc.desc.base = base;
581 	vmsegdesc.desc.limit = limit;
582 	vmsegdesc.desc.access = access;
583 
584 	error = ioctl(ctx->fd, VM_SET_SEGMENT_DESCRIPTOR, &vmsegdesc);
585 	return (error);
586 }
587 
588 int
589 vm_get_desc(struct vmctx *ctx, int vcpu, int reg,
590 	    uint64_t *base, uint32_t *limit, uint32_t *access)
591 {
592 	int error;
593 	struct vm_seg_desc vmsegdesc;
594 
595 	bzero(&vmsegdesc, sizeof(vmsegdesc));
596 	vmsegdesc.cpuid = vcpu;
597 	vmsegdesc.regnum = reg;
598 
599 	error = ioctl(ctx->fd, VM_GET_SEGMENT_DESCRIPTOR, &vmsegdesc);
600 	if (error == 0) {
601 		*base = vmsegdesc.desc.base;
602 		*limit = vmsegdesc.desc.limit;
603 		*access = vmsegdesc.desc.access;
604 	}
605 	return (error);
606 }
607 
608 int
609 vm_get_seg_desc(struct vmctx *ctx, int vcpu, int reg, struct seg_desc *seg_desc)
610 {
611 	int error;
612 
613 	error = vm_get_desc(ctx, vcpu, reg, &seg_desc->base, &seg_desc->limit,
614 	    &seg_desc->access);
615 	return (error);
616 }
617 
618 int
619 vm_set_register(struct vmctx *ctx, int vcpu, int reg, uint64_t val)
620 {
621 	int error;
622 	struct vm_register vmreg;
623 
624 	bzero(&vmreg, sizeof(vmreg));
625 	vmreg.cpuid = vcpu;
626 	vmreg.regnum = reg;
627 	vmreg.regval = val;
628 
629 	error = ioctl(ctx->fd, VM_SET_REGISTER, &vmreg);
630 	return (error);
631 }
632 
633 int
634 vm_get_register(struct vmctx *ctx, int vcpu, int reg, uint64_t *ret_val)
635 {
636 	int error;
637 	struct vm_register vmreg;
638 
639 	bzero(&vmreg, sizeof(vmreg));
640 	vmreg.cpuid = vcpu;
641 	vmreg.regnum = reg;
642 
643 	error = ioctl(ctx->fd, VM_GET_REGISTER, &vmreg);
644 	*ret_val = vmreg.regval;
645 	return (error);
646 }
647 
648 int
649 vm_set_register_set(struct vmctx *ctx, int vcpu, unsigned int count,
650     const int *regnums, uint64_t *regvals)
651 {
652 	int error;
653 	struct vm_register_set vmregset;
654 
655 	bzero(&vmregset, sizeof(vmregset));
656 	vmregset.cpuid = vcpu;
657 	vmregset.count = count;
658 	vmregset.regnums = regnums;
659 	vmregset.regvals = regvals;
660 
661 	error = ioctl(ctx->fd, VM_SET_REGISTER_SET, &vmregset);
662 	return (error);
663 }
664 
665 int
666 vm_get_register_set(struct vmctx *ctx, int vcpu, unsigned int count,
667     const int *regnums, uint64_t *regvals)
668 {
669 	int error;
670 	struct vm_register_set vmregset;
671 
672 	bzero(&vmregset, sizeof(vmregset));
673 	vmregset.cpuid = vcpu;
674 	vmregset.count = count;
675 	vmregset.regnums = regnums;
676 	vmregset.regvals = regvals;
677 
678 	error = ioctl(ctx->fd, VM_GET_REGISTER_SET, &vmregset);
679 	return (error);
680 }
681 
682 int
683 vm_run(struct vmctx *ctx, int vcpu, struct vm_exit *vmexit)
684 {
685 	int error;
686 	struct vm_run vmrun;
687 
688 	bzero(&vmrun, sizeof(vmrun));
689 	vmrun.cpuid = vcpu;
690 
691 	error = ioctl(ctx->fd, VM_RUN, &vmrun);
692 	bcopy(&vmrun.vm_exit, vmexit, sizeof(struct vm_exit));
693 	return (error);
694 }
695 
696 int
697 vm_suspend(struct vmctx *ctx, enum vm_suspend_how how)
698 {
699 	struct vm_suspend vmsuspend;
700 
701 	bzero(&vmsuspend, sizeof(vmsuspend));
702 	vmsuspend.how = how;
703 	return (ioctl(ctx->fd, VM_SUSPEND, &vmsuspend));
704 }
705 
706 int
707 vm_reinit(struct vmctx *ctx)
708 {
709 
710 	return (ioctl(ctx->fd, VM_REINIT, 0));
711 }
712 
713 int
714 vm_inject_exception(struct vmctx *ctx, int vcpu, int vector, int errcode_valid,
715     uint32_t errcode, int restart_instruction)
716 {
717 	struct vm_exception exc;
718 
719 	exc.cpuid = vcpu;
720 	exc.vector = vector;
721 	exc.error_code = errcode;
722 	exc.error_code_valid = errcode_valid;
723 	exc.restart_instruction = restart_instruction;
724 
725 	return (ioctl(ctx->fd, VM_INJECT_EXCEPTION, &exc));
726 }
727 
728 int
729 vm_apicid2vcpu(struct vmctx *ctx, int apicid)
730 {
731 	/*
732 	 * The apic id associated with the 'vcpu' has the same numerical value
733 	 * as the 'vcpu' itself.
734 	 */
735 	return (apicid);
736 }
737 
738 int
739 vm_lapic_irq(struct vmctx *ctx, int vcpu, int vector)
740 {
741 	struct vm_lapic_irq vmirq;
742 
743 	bzero(&vmirq, sizeof(vmirq));
744 	vmirq.cpuid = vcpu;
745 	vmirq.vector = vector;
746 
747 	return (ioctl(ctx->fd, VM_LAPIC_IRQ, &vmirq));
748 }
749 
750 int
751 vm_lapic_local_irq(struct vmctx *ctx, int vcpu, int vector)
752 {
753 	struct vm_lapic_irq vmirq;
754 
755 	bzero(&vmirq, sizeof(vmirq));
756 	vmirq.cpuid = vcpu;
757 	vmirq.vector = vector;
758 
759 	return (ioctl(ctx->fd, VM_LAPIC_LOCAL_IRQ, &vmirq));
760 }
761 
762 int
763 vm_lapic_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg)
764 {
765 	struct vm_lapic_msi vmmsi;
766 
767 	bzero(&vmmsi, sizeof(vmmsi));
768 	vmmsi.addr = addr;
769 	vmmsi.msg = msg;
770 
771 	return (ioctl(ctx->fd, VM_LAPIC_MSI, &vmmsi));
772 }
773 
774 int
775 vm_ioapic_assert_irq(struct vmctx *ctx, int irq)
776 {
777 	struct vm_ioapic_irq ioapic_irq;
778 
779 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
780 	ioapic_irq.irq = irq;
781 
782 	return (ioctl(ctx->fd, VM_IOAPIC_ASSERT_IRQ, &ioapic_irq));
783 }
784 
785 int
786 vm_ioapic_deassert_irq(struct vmctx *ctx, int irq)
787 {
788 	struct vm_ioapic_irq ioapic_irq;
789 
790 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
791 	ioapic_irq.irq = irq;
792 
793 	return (ioctl(ctx->fd, VM_IOAPIC_DEASSERT_IRQ, &ioapic_irq));
794 }
795 
796 int
797 vm_ioapic_pulse_irq(struct vmctx *ctx, int irq)
798 {
799 	struct vm_ioapic_irq ioapic_irq;
800 
801 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
802 	ioapic_irq.irq = irq;
803 
804 	return (ioctl(ctx->fd, VM_IOAPIC_PULSE_IRQ, &ioapic_irq));
805 }
806 
807 int
808 vm_ioapic_pincount(struct vmctx *ctx, int *pincount)
809 {
810 
811 	return (ioctl(ctx->fd, VM_IOAPIC_PINCOUNT, pincount));
812 }
813 
814 int
815 vm_readwrite_kernemu_device(struct vmctx *ctx, int vcpu, vm_paddr_t gpa,
816     bool write, int size, uint64_t *value)
817 {
818 	struct vm_readwrite_kernemu_device irp = {
819 		.vcpuid = vcpu,
820 		.access_width = fls(size) - 1,
821 		.gpa = gpa,
822 		.value = write ? *value : ~0ul,
823 	};
824 	long cmd = (write ? VM_SET_KERNEMU_DEV : VM_GET_KERNEMU_DEV);
825 	int rc;
826 
827 	rc = ioctl(ctx->fd, cmd, &irp);
828 	if (rc == 0 && !write)
829 		*value = irp.value;
830 	return (rc);
831 }
832 
833 int
834 vm_isa_assert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
835 {
836 	struct vm_isa_irq isa_irq;
837 
838 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
839 	isa_irq.atpic_irq = atpic_irq;
840 	isa_irq.ioapic_irq = ioapic_irq;
841 
842 	return (ioctl(ctx->fd, VM_ISA_ASSERT_IRQ, &isa_irq));
843 }
844 
845 int
846 vm_isa_deassert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
847 {
848 	struct vm_isa_irq isa_irq;
849 
850 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
851 	isa_irq.atpic_irq = atpic_irq;
852 	isa_irq.ioapic_irq = ioapic_irq;
853 
854 	return (ioctl(ctx->fd, VM_ISA_DEASSERT_IRQ, &isa_irq));
855 }
856 
857 int
858 vm_isa_pulse_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
859 {
860 	struct vm_isa_irq isa_irq;
861 
862 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
863 	isa_irq.atpic_irq = atpic_irq;
864 	isa_irq.ioapic_irq = ioapic_irq;
865 
866 	return (ioctl(ctx->fd, VM_ISA_PULSE_IRQ, &isa_irq));
867 }
868 
869 int
870 vm_isa_set_irq_trigger(struct vmctx *ctx, int atpic_irq,
871     enum vm_intr_trigger trigger)
872 {
873 	struct vm_isa_irq_trigger isa_irq_trigger;
874 
875 	bzero(&isa_irq_trigger, sizeof(struct vm_isa_irq_trigger));
876 	isa_irq_trigger.atpic_irq = atpic_irq;
877 	isa_irq_trigger.trigger = trigger;
878 
879 	return (ioctl(ctx->fd, VM_ISA_SET_IRQ_TRIGGER, &isa_irq_trigger));
880 }
881 
882 int
883 vm_inject_nmi(struct vmctx *ctx, int vcpu)
884 {
885 	struct vm_nmi vmnmi;
886 
887 	bzero(&vmnmi, sizeof(vmnmi));
888 	vmnmi.cpuid = vcpu;
889 
890 	return (ioctl(ctx->fd, VM_INJECT_NMI, &vmnmi));
891 }
892 
893 static const char *capstrmap[] = {
894 	[VM_CAP_HALT_EXIT]  = "hlt_exit",
895 	[VM_CAP_MTRAP_EXIT] = "mtrap_exit",
896 	[VM_CAP_PAUSE_EXIT] = "pause_exit",
897 	[VM_CAP_UNRESTRICTED_GUEST] = "unrestricted_guest",
898 	[VM_CAP_ENABLE_INVPCID] = "enable_invpcid",
899 	[VM_CAP_BPT_EXIT] = "bpt_exit",
900 };
901 
902 int
903 vm_capability_name2type(const char *capname)
904 {
905 	int i;
906 
907 	for (i = 0; i < nitems(capstrmap); i++) {
908 		if (strcmp(capstrmap[i], capname) == 0)
909 			return (i);
910 	}
911 
912 	return (-1);
913 }
914 
915 const char *
916 vm_capability_type2name(int type)
917 {
918 	if (type >= 0 && type < nitems(capstrmap))
919 		return (capstrmap[type]);
920 
921 	return (NULL);
922 }
923 
924 int
925 vm_get_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap,
926 		  int *retval)
927 {
928 	int error;
929 	struct vm_capability vmcap;
930 
931 	bzero(&vmcap, sizeof(vmcap));
932 	vmcap.cpuid = vcpu;
933 	vmcap.captype = cap;
934 
935 	error = ioctl(ctx->fd, VM_GET_CAPABILITY, &vmcap);
936 	*retval = vmcap.capval;
937 	return (error);
938 }
939 
940 int
941 vm_set_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap, int val)
942 {
943 	struct vm_capability vmcap;
944 
945 	bzero(&vmcap, sizeof(vmcap));
946 	vmcap.cpuid = vcpu;
947 	vmcap.captype = cap;
948 	vmcap.capval = val;
949 
950 	return (ioctl(ctx->fd, VM_SET_CAPABILITY, &vmcap));
951 }
952 
953 int
954 vm_assign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
955 {
956 	struct vm_pptdev pptdev;
957 
958 	bzero(&pptdev, sizeof(pptdev));
959 	pptdev.bus = bus;
960 	pptdev.slot = slot;
961 	pptdev.func = func;
962 
963 	return (ioctl(ctx->fd, VM_BIND_PPTDEV, &pptdev));
964 }
965 
966 int
967 vm_unassign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
968 {
969 	struct vm_pptdev pptdev;
970 
971 	bzero(&pptdev, sizeof(pptdev));
972 	pptdev.bus = bus;
973 	pptdev.slot = slot;
974 	pptdev.func = func;
975 
976 	return (ioctl(ctx->fd, VM_UNBIND_PPTDEV, &pptdev));
977 }
978 
979 int
980 vm_map_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func,
981 		   vm_paddr_t gpa, size_t len, vm_paddr_t hpa)
982 {
983 	struct vm_pptdev_mmio pptmmio;
984 
985 	bzero(&pptmmio, sizeof(pptmmio));
986 	pptmmio.bus = bus;
987 	pptmmio.slot = slot;
988 	pptmmio.func = func;
989 	pptmmio.gpa = gpa;
990 	pptmmio.len = len;
991 	pptmmio.hpa = hpa;
992 
993 	return (ioctl(ctx->fd, VM_MAP_PPTDEV_MMIO, &pptmmio));
994 }
995 
996 int
997 vm_unmap_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func,
998 		     vm_paddr_t gpa, size_t len)
999 {
1000 	struct vm_pptdev_mmio pptmmio;
1001 
1002 	bzero(&pptmmio, sizeof(pptmmio));
1003 	pptmmio.bus = bus;
1004 	pptmmio.slot = slot;
1005 	pptmmio.func = func;
1006 	pptmmio.gpa = gpa;
1007 	pptmmio.len = len;
1008 
1009 	return (ioctl(ctx->fd, VM_UNMAP_PPTDEV_MMIO, &pptmmio));
1010 }
1011 
1012 int
1013 vm_setup_pptdev_msi(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
1014     uint64_t addr, uint64_t msg, int numvec)
1015 {
1016 	struct vm_pptdev_msi pptmsi;
1017 
1018 	bzero(&pptmsi, sizeof(pptmsi));
1019 	pptmsi.vcpu = vcpu;
1020 	pptmsi.bus = bus;
1021 	pptmsi.slot = slot;
1022 	pptmsi.func = func;
1023 	pptmsi.msg = msg;
1024 	pptmsi.addr = addr;
1025 	pptmsi.numvec = numvec;
1026 
1027 	return (ioctl(ctx->fd, VM_PPTDEV_MSI, &pptmsi));
1028 }
1029 
1030 int
1031 vm_setup_pptdev_msix(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
1032     int idx, uint64_t addr, uint64_t msg, uint32_t vector_control)
1033 {
1034 	struct vm_pptdev_msix pptmsix;
1035 
1036 	bzero(&pptmsix, sizeof(pptmsix));
1037 	pptmsix.vcpu = vcpu;
1038 	pptmsix.bus = bus;
1039 	pptmsix.slot = slot;
1040 	pptmsix.func = func;
1041 	pptmsix.idx = idx;
1042 	pptmsix.msg = msg;
1043 	pptmsix.addr = addr;
1044 	pptmsix.vector_control = vector_control;
1045 
1046 	return ioctl(ctx->fd, VM_PPTDEV_MSIX, &pptmsix);
1047 }
1048 
1049 int
1050 vm_disable_pptdev_msix(struct vmctx *ctx, int bus, int slot, int func)
1051 {
1052 	struct vm_pptdev ppt;
1053 
1054 	bzero(&ppt, sizeof(ppt));
1055 	ppt.bus = bus;
1056 	ppt.slot = slot;
1057 	ppt.func = func;
1058 
1059 	return ioctl(ctx->fd, VM_PPTDEV_DISABLE_MSIX, &ppt);
1060 }
1061 
1062 uint64_t *
1063 vm_get_stats(struct vmctx *ctx, int vcpu, struct timeval *ret_tv,
1064 	     int *ret_entries)
1065 {
1066 	int error;
1067 
1068 	static struct vm_stats vmstats;
1069 
1070 	vmstats.cpuid = vcpu;
1071 
1072 	error = ioctl(ctx->fd, VM_STATS, &vmstats);
1073 	if (error == 0) {
1074 		if (ret_entries)
1075 			*ret_entries = vmstats.num_entries;
1076 		if (ret_tv)
1077 			*ret_tv = vmstats.tv;
1078 		return (vmstats.statbuf);
1079 	} else
1080 		return (NULL);
1081 }
1082 
1083 const char *
1084 vm_get_stat_desc(struct vmctx *ctx, int index)
1085 {
1086 	static struct vm_stat_desc statdesc;
1087 
1088 	statdesc.index = index;
1089 	if (ioctl(ctx->fd, VM_STAT_DESC, &statdesc) == 0)
1090 		return (statdesc.desc);
1091 	else
1092 		return (NULL);
1093 }
1094 
1095 int
1096 vm_get_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state *state)
1097 {
1098 	int error;
1099 	struct vm_x2apic x2apic;
1100 
1101 	bzero(&x2apic, sizeof(x2apic));
1102 	x2apic.cpuid = vcpu;
1103 
1104 	error = ioctl(ctx->fd, VM_GET_X2APIC_STATE, &x2apic);
1105 	*state = x2apic.state;
1106 	return (error);
1107 }
1108 
1109 int
1110 vm_set_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state state)
1111 {
1112 	int error;
1113 	struct vm_x2apic x2apic;
1114 
1115 	bzero(&x2apic, sizeof(x2apic));
1116 	x2apic.cpuid = vcpu;
1117 	x2apic.state = state;
1118 
1119 	error = ioctl(ctx->fd, VM_SET_X2APIC_STATE, &x2apic);
1120 
1121 	return (error);
1122 }
1123 
1124 /*
1125  * From Intel Vol 3a:
1126  * Table 9-1. IA-32 Processor States Following Power-up, Reset or INIT
1127  */
1128 int
1129 vcpu_reset(struct vmctx *vmctx, int vcpu)
1130 {
1131 	int error;
1132 	uint64_t rflags, rip, cr0, cr4, zero, desc_base, rdx;
1133 	uint32_t desc_access, desc_limit;
1134 	uint16_t sel;
1135 
1136 	zero = 0;
1137 
1138 	rflags = 0x2;
1139 	error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RFLAGS, rflags);
1140 	if (error)
1141 		goto done;
1142 
1143 	rip = 0xfff0;
1144 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RIP, rip)) != 0)
1145 		goto done;
1146 
1147 	cr0 = CR0_NE;
1148 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR0, cr0)) != 0)
1149 		goto done;
1150 
1151 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR3, zero)) != 0)
1152 		goto done;
1153 
1154 	cr4 = 0;
1155 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR4, cr4)) != 0)
1156 		goto done;
1157 
1158 	/*
1159 	 * CS: present, r/w, accessed, 16-bit, byte granularity, usable
1160 	 */
1161 	desc_base = 0xffff0000;
1162 	desc_limit = 0xffff;
1163 	desc_access = 0x0093;
1164 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_CS,
1165 			    desc_base, desc_limit, desc_access);
1166 	if (error)
1167 		goto done;
1168 
1169 	sel = 0xf000;
1170 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CS, sel)) != 0)
1171 		goto done;
1172 
1173 	/*
1174 	 * SS,DS,ES,FS,GS: present, r/w, accessed, 16-bit, byte granularity
1175 	 */
1176 	desc_base = 0;
1177 	desc_limit = 0xffff;
1178 	desc_access = 0x0093;
1179 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_SS,
1180 			    desc_base, desc_limit, desc_access);
1181 	if (error)
1182 		goto done;
1183 
1184 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_DS,
1185 			    desc_base, desc_limit, desc_access);
1186 	if (error)
1187 		goto done;
1188 
1189 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_ES,
1190 			    desc_base, desc_limit, desc_access);
1191 	if (error)
1192 		goto done;
1193 
1194 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_FS,
1195 			    desc_base, desc_limit, desc_access);
1196 	if (error)
1197 		goto done;
1198 
1199 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GS,
1200 			    desc_base, desc_limit, desc_access);
1201 	if (error)
1202 		goto done;
1203 
1204 	sel = 0;
1205 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_SS, sel)) != 0)
1206 		goto done;
1207 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_DS, sel)) != 0)
1208 		goto done;
1209 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_ES, sel)) != 0)
1210 		goto done;
1211 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_FS, sel)) != 0)
1212 		goto done;
1213 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_GS, sel)) != 0)
1214 		goto done;
1215 
1216 	/* General purpose registers */
1217 	rdx = 0xf00;
1218 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RAX, zero)) != 0)
1219 		goto done;
1220 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBX, zero)) != 0)
1221 		goto done;
1222 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RCX, zero)) != 0)
1223 		goto done;
1224 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDX, rdx)) != 0)
1225 		goto done;
1226 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSI, zero)) != 0)
1227 		goto done;
1228 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDI, zero)) != 0)
1229 		goto done;
1230 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBP, zero)) != 0)
1231 		goto done;
1232 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSP, zero)) != 0)
1233 		goto done;
1234 
1235 	/* GDTR, IDTR */
1236 	desc_base = 0;
1237 	desc_limit = 0xffff;
1238 	desc_access = 0;
1239 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GDTR,
1240 			    desc_base, desc_limit, desc_access);
1241 	if (error != 0)
1242 		goto done;
1243 
1244 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_IDTR,
1245 			    desc_base, desc_limit, desc_access);
1246 	if (error != 0)
1247 		goto done;
1248 
1249 	/* TR */
1250 	desc_base = 0;
1251 	desc_limit = 0xffff;
1252 	desc_access = 0x0000008b;
1253 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_TR, 0, 0, desc_access);
1254 	if (error)
1255 		goto done;
1256 
1257 	sel = 0;
1258 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_TR, sel)) != 0)
1259 		goto done;
1260 
1261 	/* LDTR */
1262 	desc_base = 0;
1263 	desc_limit = 0xffff;
1264 	desc_access = 0x00000082;
1265 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_LDTR, desc_base,
1266 			    desc_limit, desc_access);
1267 	if (error)
1268 		goto done;
1269 
1270 	sel = 0;
1271 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_LDTR, 0)) != 0)
1272 		goto done;
1273 
1274 	/* XXX cr2, debug registers */
1275 
1276 	error = 0;
1277 done:
1278 	return (error);
1279 }
1280 
1281 int
1282 vm_get_gpa_pmap(struct vmctx *ctx, uint64_t gpa, uint64_t *pte, int *num)
1283 {
1284 	int error, i;
1285 	struct vm_gpa_pte gpapte;
1286 
1287 	bzero(&gpapte, sizeof(gpapte));
1288 	gpapte.gpa = gpa;
1289 
1290 	error = ioctl(ctx->fd, VM_GET_GPA_PMAP, &gpapte);
1291 
1292 	if (error == 0) {
1293 		*num = gpapte.ptenum;
1294 		for (i = 0; i < gpapte.ptenum; i++)
1295 			pte[i] = gpapte.pte[i];
1296 	}
1297 
1298 	return (error);
1299 }
1300 
1301 int
1302 vm_get_hpet_capabilities(struct vmctx *ctx, uint32_t *capabilities)
1303 {
1304 	int error;
1305 	struct vm_hpet_cap cap;
1306 
1307 	bzero(&cap, sizeof(struct vm_hpet_cap));
1308 	error = ioctl(ctx->fd, VM_GET_HPET_CAPABILITIES, &cap);
1309 	if (capabilities != NULL)
1310 		*capabilities = cap.capabilities;
1311 	return (error);
1312 }
1313 
1314 int
1315 vm_gla2gpa(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1316     uint64_t gla, int prot, uint64_t *gpa, int *fault)
1317 {
1318 	struct vm_gla2gpa gg;
1319 	int error;
1320 
1321 	bzero(&gg, sizeof(struct vm_gla2gpa));
1322 	gg.vcpuid = vcpu;
1323 	gg.prot = prot;
1324 	gg.gla = gla;
1325 	gg.paging = *paging;
1326 
1327 	error = ioctl(ctx->fd, VM_GLA2GPA, &gg);
1328 	if (error == 0) {
1329 		*fault = gg.fault;
1330 		*gpa = gg.gpa;
1331 	}
1332 	return (error);
1333 }
1334 
1335 int
1336 vm_gla2gpa_nofault(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1337     uint64_t gla, int prot, uint64_t *gpa, int *fault)
1338 {
1339 	struct vm_gla2gpa gg;
1340 	int error;
1341 
1342 	bzero(&gg, sizeof(struct vm_gla2gpa));
1343 	gg.vcpuid = vcpu;
1344 	gg.prot = prot;
1345 	gg.gla = gla;
1346 	gg.paging = *paging;
1347 
1348 	error = ioctl(ctx->fd, VM_GLA2GPA_NOFAULT, &gg);
1349 	if (error == 0) {
1350 		*fault = gg.fault;
1351 		*gpa = gg.gpa;
1352 	}
1353 	return (error);
1354 }
1355 
1356 #ifndef min
1357 #define	min(a,b)	(((a) < (b)) ? (a) : (b))
1358 #endif
1359 
1360 int
1361 vm_copy_setup(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1362     uint64_t gla, size_t len, int prot, struct iovec *iov, int iovcnt,
1363     int *fault)
1364 {
1365 	void *va;
1366 	uint64_t gpa;
1367 	int error, i, n, off;
1368 
1369 	for (i = 0; i < iovcnt; i++) {
1370 		iov[i].iov_base = 0;
1371 		iov[i].iov_len = 0;
1372 	}
1373 
1374 	while (len) {
1375 		assert(iovcnt > 0);
1376 		error = vm_gla2gpa(ctx, vcpu, paging, gla, prot, &gpa, fault);
1377 		if (error || *fault)
1378 			return (error);
1379 
1380 		off = gpa & PAGE_MASK;
1381 		n = min(len, PAGE_SIZE - off);
1382 
1383 		va = vm_map_gpa(ctx, gpa, n);
1384 		if (va == NULL)
1385 			return (EFAULT);
1386 
1387 		iov->iov_base = va;
1388 		iov->iov_len = n;
1389 		iov++;
1390 		iovcnt--;
1391 
1392 		gla += n;
1393 		len -= n;
1394 	}
1395 	return (0);
1396 }
1397 
1398 void
1399 vm_copy_teardown(struct vmctx *ctx, int vcpu, struct iovec *iov, int iovcnt)
1400 {
1401 
1402 	return;
1403 }
1404 
1405 void
1406 vm_copyin(struct vmctx *ctx, int vcpu, struct iovec *iov, void *vp, size_t len)
1407 {
1408 	const char *src;
1409 	char *dst;
1410 	size_t n;
1411 
1412 	dst = vp;
1413 	while (len) {
1414 		assert(iov->iov_len);
1415 		n = min(len, iov->iov_len);
1416 		src = iov->iov_base;
1417 		bcopy(src, dst, n);
1418 
1419 		iov++;
1420 		dst += n;
1421 		len -= n;
1422 	}
1423 }
1424 
1425 void
1426 vm_copyout(struct vmctx *ctx, int vcpu, const void *vp, struct iovec *iov,
1427     size_t len)
1428 {
1429 	const char *src;
1430 	char *dst;
1431 	size_t n;
1432 
1433 	src = vp;
1434 	while (len) {
1435 		assert(iov->iov_len);
1436 		n = min(len, iov->iov_len);
1437 		dst = iov->iov_base;
1438 		bcopy(src, dst, n);
1439 
1440 		iov++;
1441 		src += n;
1442 		len -= n;
1443 	}
1444 }
1445 
1446 static int
1447 vm_get_cpus(struct vmctx *ctx, int which, cpuset_t *cpus)
1448 {
1449 	struct vm_cpuset vm_cpuset;
1450 	int error;
1451 
1452 	bzero(&vm_cpuset, sizeof(struct vm_cpuset));
1453 	vm_cpuset.which = which;
1454 	vm_cpuset.cpusetsize = sizeof(cpuset_t);
1455 	vm_cpuset.cpus = cpus;
1456 
1457 	error = ioctl(ctx->fd, VM_GET_CPUS, &vm_cpuset);
1458 	return (error);
1459 }
1460 
1461 int
1462 vm_active_cpus(struct vmctx *ctx, cpuset_t *cpus)
1463 {
1464 
1465 	return (vm_get_cpus(ctx, VM_ACTIVE_CPUS, cpus));
1466 }
1467 
1468 int
1469 vm_suspended_cpus(struct vmctx *ctx, cpuset_t *cpus)
1470 {
1471 
1472 	return (vm_get_cpus(ctx, VM_SUSPENDED_CPUS, cpus));
1473 }
1474 
1475 int
1476 vm_debug_cpus(struct vmctx *ctx, cpuset_t *cpus)
1477 {
1478 
1479 	return (vm_get_cpus(ctx, VM_DEBUG_CPUS, cpus));
1480 }
1481 
1482 int
1483 vm_activate_cpu(struct vmctx *ctx, int vcpu)
1484 {
1485 	struct vm_activate_cpu ac;
1486 	int error;
1487 
1488 	bzero(&ac, sizeof(struct vm_activate_cpu));
1489 	ac.vcpuid = vcpu;
1490 	error = ioctl(ctx->fd, VM_ACTIVATE_CPU, &ac);
1491 	return (error);
1492 }
1493 
1494 int
1495 vm_suspend_cpu(struct vmctx *ctx, int vcpu)
1496 {
1497 	struct vm_activate_cpu ac;
1498 	int error;
1499 
1500 	bzero(&ac, sizeof(struct vm_activate_cpu));
1501 	ac.vcpuid = vcpu;
1502 	error = ioctl(ctx->fd, VM_SUSPEND_CPU, &ac);
1503 	return (error);
1504 }
1505 
1506 int
1507 vm_resume_cpu(struct vmctx *ctx, int vcpu)
1508 {
1509 	struct vm_activate_cpu ac;
1510 	int error;
1511 
1512 	bzero(&ac, sizeof(struct vm_activate_cpu));
1513 	ac.vcpuid = vcpu;
1514 	error = ioctl(ctx->fd, VM_RESUME_CPU, &ac);
1515 	return (error);
1516 }
1517 
1518 int
1519 vm_get_intinfo(struct vmctx *ctx, int vcpu, uint64_t *info1, uint64_t *info2)
1520 {
1521 	struct vm_intinfo vmii;
1522 	int error;
1523 
1524 	bzero(&vmii, sizeof(struct vm_intinfo));
1525 	vmii.vcpuid = vcpu;
1526 	error = ioctl(ctx->fd, VM_GET_INTINFO, &vmii);
1527 	if (error == 0) {
1528 		*info1 = vmii.info1;
1529 		*info2 = vmii.info2;
1530 	}
1531 	return (error);
1532 }
1533 
1534 int
1535 vm_set_intinfo(struct vmctx *ctx, int vcpu, uint64_t info1)
1536 {
1537 	struct vm_intinfo vmii;
1538 	int error;
1539 
1540 	bzero(&vmii, sizeof(struct vm_intinfo));
1541 	vmii.vcpuid = vcpu;
1542 	vmii.info1 = info1;
1543 	error = ioctl(ctx->fd, VM_SET_INTINFO, &vmii);
1544 	return (error);
1545 }
1546 
1547 int
1548 vm_rtc_write(struct vmctx *ctx, int offset, uint8_t value)
1549 {
1550 	struct vm_rtc_data rtcdata;
1551 	int error;
1552 
1553 	bzero(&rtcdata, sizeof(struct vm_rtc_data));
1554 	rtcdata.offset = offset;
1555 	rtcdata.value = value;
1556 	error = ioctl(ctx->fd, VM_RTC_WRITE, &rtcdata);
1557 	return (error);
1558 }
1559 
1560 int
1561 vm_rtc_read(struct vmctx *ctx, int offset, uint8_t *retval)
1562 {
1563 	struct vm_rtc_data rtcdata;
1564 	int error;
1565 
1566 	bzero(&rtcdata, sizeof(struct vm_rtc_data));
1567 	rtcdata.offset = offset;
1568 	error = ioctl(ctx->fd, VM_RTC_READ, &rtcdata);
1569 	if (error == 0)
1570 		*retval = rtcdata.value;
1571 	return (error);
1572 }
1573 
1574 int
1575 vm_rtc_settime(struct vmctx *ctx, time_t secs)
1576 {
1577 	struct vm_rtc_time rtctime;
1578 	int error;
1579 
1580 	bzero(&rtctime, sizeof(struct vm_rtc_time));
1581 	rtctime.secs = secs;
1582 	error = ioctl(ctx->fd, VM_RTC_SETTIME, &rtctime);
1583 	return (error);
1584 }
1585 
1586 int
1587 vm_rtc_gettime(struct vmctx *ctx, time_t *secs)
1588 {
1589 	struct vm_rtc_time rtctime;
1590 	int error;
1591 
1592 	bzero(&rtctime, sizeof(struct vm_rtc_time));
1593 	error = ioctl(ctx->fd, VM_RTC_GETTIME, &rtctime);
1594 	if (error == 0)
1595 		*secs = rtctime.secs;
1596 	return (error);
1597 }
1598 
1599 int
1600 vm_restart_instruction(void *arg, int vcpu)
1601 {
1602 	struct vmctx *ctx = arg;
1603 
1604 	return (ioctl(ctx->fd, VM_RESTART_INSTRUCTION, &vcpu));
1605 }
1606 
1607 int
1608 vm_snapshot_req(struct vm_snapshot_meta *meta)
1609 {
1610 
1611 	if (ioctl(meta->ctx->fd, VM_SNAPSHOT_REQ, meta) == -1) {
1612 #ifdef SNAPSHOT_DEBUG
1613 		fprintf(stderr, "%s: snapshot failed for %s: %d\r\n",
1614 		    __func__, meta->dev_name, errno);
1615 #endif
1616 		return (-1);
1617 	}
1618 	return (0);
1619 }
1620 
1621 int
1622 vm_restore_time(struct vmctx *ctx)
1623 {
1624 	int dummy;
1625 
1626 	dummy = 0;
1627 	return (ioctl(ctx->fd, VM_RESTORE_TIME, &dummy));
1628 }
1629 
1630 int
1631 vm_set_topology(struct vmctx *ctx,
1632     uint16_t sockets, uint16_t cores, uint16_t threads, uint16_t maxcpus)
1633 {
1634 	struct vm_cpu_topology topology;
1635 
1636 	bzero(&topology, sizeof (struct vm_cpu_topology));
1637 	topology.sockets = sockets;
1638 	topology.cores = cores;
1639 	topology.threads = threads;
1640 	topology.maxcpus = maxcpus;
1641 	return (ioctl(ctx->fd, VM_SET_TOPOLOGY, &topology));
1642 }
1643 
1644 int
1645 vm_get_topology(struct vmctx *ctx,
1646     uint16_t *sockets, uint16_t *cores, uint16_t *threads, uint16_t *maxcpus)
1647 {
1648 	struct vm_cpu_topology topology;
1649 	int error;
1650 
1651 	bzero(&topology, sizeof (struct vm_cpu_topology));
1652 	error = ioctl(ctx->fd, VM_GET_TOPOLOGY, &topology);
1653 	if (error == 0) {
1654 		*sockets = topology.sockets;
1655 		*cores = topology.cores;
1656 		*threads = topology.threads;
1657 		*maxcpus = topology.maxcpus;
1658 	}
1659 	return (error);
1660 }
1661 
1662 int
1663 vm_get_device_fd(struct vmctx *ctx)
1664 {
1665 
1666 	return (ctx->fd);
1667 }
1668 
1669 const cap_ioctl_t *
1670 vm_get_ioctls(size_t *len)
1671 {
1672 	cap_ioctl_t *cmds;
1673 	/* keep in sync with machine/vmm_dev.h */
1674 	static const cap_ioctl_t vm_ioctl_cmds[] = { VM_RUN, VM_SUSPEND, VM_REINIT,
1675 	    VM_ALLOC_MEMSEG, VM_GET_MEMSEG, VM_MMAP_MEMSEG, VM_MMAP_MEMSEG,
1676 	    VM_MMAP_GETNEXT, VM_MUNMAP_MEMSEG, VM_SET_REGISTER, VM_GET_REGISTER,
1677 	    VM_SET_SEGMENT_DESCRIPTOR, VM_GET_SEGMENT_DESCRIPTOR,
1678 	    VM_SET_REGISTER_SET, VM_GET_REGISTER_SET,
1679 	    VM_SET_KERNEMU_DEV, VM_GET_KERNEMU_DEV,
1680 	    VM_INJECT_EXCEPTION, VM_LAPIC_IRQ, VM_LAPIC_LOCAL_IRQ,
1681 	    VM_LAPIC_MSI, VM_IOAPIC_ASSERT_IRQ, VM_IOAPIC_DEASSERT_IRQ,
1682 	    VM_IOAPIC_PULSE_IRQ, VM_IOAPIC_PINCOUNT, VM_ISA_ASSERT_IRQ,
1683 	    VM_ISA_DEASSERT_IRQ, VM_ISA_PULSE_IRQ, VM_ISA_SET_IRQ_TRIGGER,
1684 	    VM_SET_CAPABILITY, VM_GET_CAPABILITY, VM_BIND_PPTDEV,
1685 	    VM_UNBIND_PPTDEV, VM_MAP_PPTDEV_MMIO, VM_PPTDEV_MSI,
1686 	    VM_PPTDEV_MSIX, VM_UNMAP_PPTDEV_MMIO, VM_PPTDEV_DISABLE_MSIX,
1687 	    VM_INJECT_NMI, VM_STATS, VM_STAT_DESC,
1688 	    VM_SET_X2APIC_STATE, VM_GET_X2APIC_STATE,
1689 	    VM_GET_HPET_CAPABILITIES, VM_GET_GPA_PMAP, VM_GLA2GPA,
1690 	    VM_GLA2GPA_NOFAULT,
1691 	    VM_ACTIVATE_CPU, VM_GET_CPUS, VM_SUSPEND_CPU, VM_RESUME_CPU,
1692 	    VM_SET_INTINFO, VM_GET_INTINFO,
1693 	    VM_RTC_WRITE, VM_RTC_READ, VM_RTC_SETTIME, VM_RTC_GETTIME,
1694 	    VM_RESTART_INSTRUCTION, VM_SET_TOPOLOGY, VM_GET_TOPOLOGY };
1695 
1696 	if (len == NULL) {
1697 		cmds = malloc(sizeof(vm_ioctl_cmds));
1698 		if (cmds == NULL)
1699 			return (NULL);
1700 		bcopy(vm_ioctl_cmds, cmds, sizeof(vm_ioctl_cmds));
1701 		return (cmds);
1702 	}
1703 
1704 	*len = nitems(vm_ioctl_cmds);
1705 	return (NULL);
1706 }
1707