xref: /freebsd/lib/libvmmapi/vmmapi.c (revision 184d64af1344d6f3c370499510cede49320f4863)
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(name));
115 }
116 
117 struct vmctx *
118 vm_open(const char *name)
119 {
120 	struct vmctx *vm;
121 	int saved_errno;
122 
123 	vm = malloc(sizeof(struct vmctx) + strlen(name) + 1);
124 	assert(vm != NULL);
125 
126 	vm->fd = -1;
127 	vm->memflags = 0;
128 	vm->lowmem_limit = 3 * GB;
129 	vm->name = (char *)(vm + 1);
130 	strcpy(vm->name, name);
131 
132 	if ((vm->fd = vm_device_open(vm->name)) < 0)
133 		goto err;
134 
135 	return (vm);
136 err:
137 	saved_errno = errno;
138 	free(vm);
139 	errno = saved_errno;
140 	return (NULL);
141 }
142 
143 void
144 vm_destroy(struct vmctx *vm)
145 {
146 	assert(vm != NULL);
147 
148 	if (vm->fd >= 0)
149 		close(vm->fd);
150 	DESTROY(vm->name);
151 
152 	free(vm);
153 }
154 
155 int
156 vm_parse_memsize(const char *opt, size_t *ret_memsize)
157 {
158 	char *endptr;
159 	size_t optval;
160 	int error;
161 
162 	optval = strtoul(opt, &endptr, 0);
163 	if (*opt != '\0' && *endptr == '\0') {
164 		/*
165 		 * For the sake of backward compatibility if the memory size
166 		 * specified on the command line is less than a megabyte then
167 		 * it is interpreted as being in units of MB.
168 		 */
169 		if (optval < MB)
170 			optval *= MB;
171 		*ret_memsize = optval;
172 		error = 0;
173 	} else
174 		error = expand_number(opt, ret_memsize);
175 
176 	return (error);
177 }
178 
179 uint32_t
180 vm_get_lowmem_limit(struct vmctx *ctx)
181 {
182 
183 	return (ctx->lowmem_limit);
184 }
185 
186 void
187 vm_set_lowmem_limit(struct vmctx *ctx, uint32_t limit)
188 {
189 
190 	ctx->lowmem_limit = limit;
191 }
192 
193 void
194 vm_set_memflags(struct vmctx *ctx, int flags)
195 {
196 
197 	ctx->memflags = flags;
198 }
199 
200 int
201 vm_get_memflags(struct vmctx *ctx)
202 {
203 
204 	return (ctx->memflags);
205 }
206 
207 /*
208  * Map segment 'segid' starting at 'off' into guest address range [gpa,gpa+len).
209  */
210 int
211 vm_mmap_memseg(struct vmctx *ctx, vm_paddr_t gpa, int segid, vm_ooffset_t off,
212     size_t len, int prot)
213 {
214 	struct vm_memmap memmap;
215 	int error, flags;
216 
217 	memmap.gpa = gpa;
218 	memmap.segid = segid;
219 	memmap.segoff = off;
220 	memmap.len = len;
221 	memmap.prot = prot;
222 	memmap.flags = 0;
223 
224 	if (ctx->memflags & VM_MEM_F_WIRED)
225 		memmap.flags |= VM_MEMMAP_F_WIRED;
226 
227 	/*
228 	 * If this mapping already exists then don't create it again. This
229 	 * is the common case for SYSMEM mappings created by bhyveload(8).
230 	 */
231 	error = vm_mmap_getnext(ctx, &gpa, &segid, &off, &len, &prot, &flags);
232 	if (error == 0 && gpa == memmap.gpa) {
233 		if (segid != memmap.segid || off != memmap.segoff ||
234 		    prot != memmap.prot || flags != memmap.flags) {
235 			errno = EEXIST;
236 			return (-1);
237 		} else {
238 			return (0);
239 		}
240 	}
241 
242 	error = ioctl(ctx->fd, VM_MMAP_MEMSEG, &memmap);
243 	return (error);
244 }
245 
246 int
247 vm_get_guestmem_from_ctx(struct vmctx *ctx, char **guest_baseaddr,
248     size_t *lowmem_size, size_t *highmem_size)
249 {
250 
251 	*guest_baseaddr = ctx->baseaddr;
252 	*lowmem_size = ctx->lowmem;
253 	*highmem_size = ctx->highmem;
254 	return (0);
255 }
256 
257 int
258 vm_munmap_memseg(struct vmctx *ctx, vm_paddr_t gpa, size_t len)
259 {
260 	struct vm_munmap munmap;
261 	int error;
262 
263 	munmap.gpa = gpa;
264 	munmap.len = len;
265 
266 	error = ioctl(ctx->fd, VM_MUNMAP_MEMSEG, &munmap);
267 	return (error);
268 }
269 
270 int
271 vm_mmap_getnext(struct vmctx *ctx, vm_paddr_t *gpa, int *segid,
272     vm_ooffset_t *segoff, size_t *len, int *prot, int *flags)
273 {
274 	struct vm_memmap memmap;
275 	int error;
276 
277 	bzero(&memmap, sizeof(struct vm_memmap));
278 	memmap.gpa = *gpa;
279 	error = ioctl(ctx->fd, VM_MMAP_GETNEXT, &memmap);
280 	if (error == 0) {
281 		*gpa = memmap.gpa;
282 		*segid = memmap.segid;
283 		*segoff = memmap.segoff;
284 		*len = memmap.len;
285 		*prot = memmap.prot;
286 		*flags = memmap.flags;
287 	}
288 	return (error);
289 }
290 
291 /*
292  * Return 0 if the segments are identical and non-zero otherwise.
293  *
294  * This is slightly complicated by the fact that only device memory segments
295  * are named.
296  */
297 static int
298 cmpseg(size_t len, const char *str, size_t len2, const char *str2)
299 {
300 
301 	if (len == len2) {
302 		if ((!str && !str2) || (str && str2 && !strcmp(str, str2)))
303 			return (0);
304 	}
305 	return (-1);
306 }
307 
308 static int
309 vm_alloc_memseg(struct vmctx *ctx, int segid, size_t len, const char *name)
310 {
311 	struct vm_memseg memseg;
312 	size_t n;
313 	int error;
314 
315 	/*
316 	 * If the memory segment has already been created then just return.
317 	 * This is the usual case for the SYSMEM segment created by userspace
318 	 * loaders like bhyveload(8).
319 	 */
320 	error = vm_get_memseg(ctx, segid, &memseg.len, memseg.name,
321 	    sizeof(memseg.name));
322 	if (error)
323 		return (error);
324 
325 	if (memseg.len != 0) {
326 		if (cmpseg(len, name, memseg.len, VM_MEMSEG_NAME(&memseg))) {
327 			errno = EINVAL;
328 			return (-1);
329 		} else {
330 			return (0);
331 		}
332 	}
333 
334 	bzero(&memseg, sizeof(struct vm_memseg));
335 	memseg.segid = segid;
336 	memseg.len = len;
337 	if (name != NULL) {
338 		n = strlcpy(memseg.name, name, sizeof(memseg.name));
339 		if (n >= sizeof(memseg.name)) {
340 			errno = ENAMETOOLONG;
341 			return (-1);
342 		}
343 	}
344 
345 	error = ioctl(ctx->fd, VM_ALLOC_MEMSEG, &memseg);
346 	return (error);
347 }
348 
349 int
350 vm_get_memseg(struct vmctx *ctx, int segid, size_t *lenp, char *namebuf,
351     size_t bufsize)
352 {
353 	struct vm_memseg memseg;
354 	size_t n;
355 	int error;
356 
357 	memseg.segid = segid;
358 	error = ioctl(ctx->fd, VM_GET_MEMSEG, &memseg);
359 	if (error == 0) {
360 		*lenp = memseg.len;
361 		n = strlcpy(namebuf, memseg.name, bufsize);
362 		if (n >= bufsize) {
363 			errno = ENAMETOOLONG;
364 			error = -1;
365 		}
366 	}
367 	return (error);
368 }
369 
370 static int
371 setup_memory_segment(struct vmctx *ctx, vm_paddr_t gpa, size_t len, char *base)
372 {
373 	char *ptr;
374 	int error, flags;
375 
376 	/* Map 'len' bytes starting at 'gpa' in the guest address space */
377 	error = vm_mmap_memseg(ctx, gpa, VM_SYSMEM, gpa, len, PROT_ALL);
378 	if (error)
379 		return (error);
380 
381 	flags = MAP_SHARED | MAP_FIXED;
382 	if ((ctx->memflags & VM_MEM_F_INCORE) == 0)
383 		flags |= MAP_NOCORE;
384 
385 	/* mmap into the process address space on the host */
386 	ptr = mmap(base + gpa, len, PROT_RW, flags, ctx->fd, gpa);
387 	if (ptr == MAP_FAILED)
388 		return (-1);
389 
390 	return (0);
391 }
392 
393 int
394 vm_setup_memory(struct vmctx *ctx, size_t memsize, enum vm_mmap_style vms)
395 {
396 	size_t objsize, len;
397 	vm_paddr_t gpa;
398 	char *baseaddr, *ptr;
399 	int error;
400 
401 	assert(vms == VM_MMAP_ALL);
402 
403 	/*
404 	 * If 'memsize' cannot fit entirely in the 'lowmem' segment then
405 	 * create another 'highmem' segment above 4GB for the remainder.
406 	 */
407 	if (memsize > ctx->lowmem_limit) {
408 		ctx->lowmem = ctx->lowmem_limit;
409 		ctx->highmem = memsize - ctx->lowmem_limit;
410 		objsize = 4*GB + ctx->highmem;
411 	} else {
412 		ctx->lowmem = memsize;
413 		ctx->highmem = 0;
414 		objsize = ctx->lowmem;
415 	}
416 
417 	error = vm_alloc_memseg(ctx, VM_SYSMEM, objsize, NULL);
418 	if (error)
419 		return (error);
420 
421 	/*
422 	 * Stake out a contiguous region covering the guest physical memory
423 	 * and the adjoining guard regions.
424 	 */
425 	len = VM_MMAP_GUARD_SIZE + objsize + VM_MMAP_GUARD_SIZE;
426 	ptr = mmap(NULL, len, PROT_NONE, MAP_GUARD | MAP_ALIGNED_SUPER, -1, 0);
427 	if (ptr == MAP_FAILED)
428 		return (-1);
429 
430 	baseaddr = ptr + VM_MMAP_GUARD_SIZE;
431 	if (ctx->highmem > 0) {
432 		gpa = 4*GB;
433 		len = ctx->highmem;
434 		error = setup_memory_segment(ctx, gpa, len, baseaddr);
435 		if (error)
436 			return (error);
437 	}
438 
439 	if (ctx->lowmem > 0) {
440 		gpa = 0;
441 		len = ctx->lowmem;
442 		error = setup_memory_segment(ctx, gpa, len, baseaddr);
443 		if (error)
444 			return (error);
445 	}
446 
447 	ctx->baseaddr = baseaddr;
448 
449 	return (0);
450 }
451 
452 /*
453  * Returns a non-NULL pointer if [gaddr, gaddr+len) is entirely contained in
454  * the lowmem or highmem regions.
455  *
456  * In particular return NULL if [gaddr, gaddr+len) falls in guest MMIO region.
457  * The instruction emulation code depends on this behavior.
458  */
459 void *
460 vm_map_gpa(struct vmctx *ctx, vm_paddr_t gaddr, size_t len)
461 {
462 
463 	if (ctx->lowmem > 0) {
464 		if (gaddr < ctx->lowmem && len <= ctx->lowmem &&
465 		    gaddr + len <= ctx->lowmem)
466 			return (ctx->baseaddr + gaddr);
467 	}
468 
469 	if (ctx->highmem > 0) {
470                 if (gaddr >= 4*GB) {
471 			if (gaddr < 4*GB + ctx->highmem &&
472 			    len <= ctx->highmem &&
473 			    gaddr + len <= 4*GB + ctx->highmem)
474 				return (ctx->baseaddr + gaddr);
475 		}
476 	}
477 
478 	return (NULL);
479 }
480 
481 vm_paddr_t
482 vm_rev_map_gpa(struct vmctx *ctx, void *addr)
483 {
484 	vm_paddr_t offaddr;
485 
486 	offaddr = (char *)addr - ctx->baseaddr;
487 
488 	if (ctx->lowmem > 0)
489 		if (offaddr >= 0 && offaddr <= ctx->lowmem)
490 			return (offaddr);
491 
492 	if (ctx->highmem > 0)
493 		if (offaddr >= 4*GB && offaddr < 4*GB + ctx->highmem)
494 			return (offaddr);
495 
496 	return ((vm_paddr_t)-1);
497 }
498 
499 /* TODO: maximum size for vmname */
500 int
501 vm_get_name(struct vmctx *ctx, char *buf, size_t max_len)
502 {
503 
504 	if (strlcpy(buf, ctx->name, max_len) >= max_len)
505 		return (EINVAL);
506 	return (0);
507 }
508 
509 size_t
510 vm_get_lowmem_size(struct vmctx *ctx)
511 {
512 
513 	return (ctx->lowmem);
514 }
515 
516 size_t
517 vm_get_highmem_size(struct vmctx *ctx)
518 {
519 
520 	return (ctx->highmem);
521 }
522 
523 void *
524 vm_create_devmem(struct vmctx *ctx, int segid, const char *name, size_t len)
525 {
526 	char pathname[MAXPATHLEN];
527 	size_t len2;
528 	char *base, *ptr;
529 	int fd, error, flags;
530 
531 	fd = -1;
532 	ptr = MAP_FAILED;
533 	if (name == NULL || strlen(name) == 0) {
534 		errno = EINVAL;
535 		goto done;
536 	}
537 
538 	error = vm_alloc_memseg(ctx, segid, len, name);
539 	if (error)
540 		goto done;
541 
542 	strlcpy(pathname, "/dev/vmm.io/", sizeof(pathname));
543 	strlcat(pathname, ctx->name, sizeof(pathname));
544 	strlcat(pathname, ".", sizeof(pathname));
545 	strlcat(pathname, name, sizeof(pathname));
546 
547 	fd = open(pathname, O_RDWR);
548 	if (fd < 0)
549 		goto done;
550 
551 	/*
552 	 * Stake out a contiguous region covering the device memory and the
553 	 * adjoining guard regions.
554 	 */
555 	len2 = VM_MMAP_GUARD_SIZE + len + VM_MMAP_GUARD_SIZE;
556 	base = mmap(NULL, len2, PROT_NONE, MAP_GUARD | MAP_ALIGNED_SUPER, -1,
557 	    0);
558 	if (base == MAP_FAILED)
559 		goto done;
560 
561 	flags = MAP_SHARED | MAP_FIXED;
562 	if ((ctx->memflags & VM_MEM_F_INCORE) == 0)
563 		flags |= MAP_NOCORE;
564 
565 	/* mmap the devmem region in the host address space */
566 	ptr = mmap(base + VM_MMAP_GUARD_SIZE, len, PROT_RW, flags, fd, 0);
567 done:
568 	if (fd >= 0)
569 		close(fd);
570 	return (ptr);
571 }
572 
573 int
574 vm_set_desc(struct vmctx *ctx, int vcpu, int reg,
575 	    uint64_t base, uint32_t limit, uint32_t access)
576 {
577 	int error;
578 	struct vm_seg_desc vmsegdesc;
579 
580 	bzero(&vmsegdesc, sizeof(vmsegdesc));
581 	vmsegdesc.cpuid = vcpu;
582 	vmsegdesc.regnum = reg;
583 	vmsegdesc.desc.base = base;
584 	vmsegdesc.desc.limit = limit;
585 	vmsegdesc.desc.access = access;
586 
587 	error = ioctl(ctx->fd, VM_SET_SEGMENT_DESCRIPTOR, &vmsegdesc);
588 	return (error);
589 }
590 
591 int
592 vm_get_desc(struct vmctx *ctx, int vcpu, int reg,
593 	    uint64_t *base, uint32_t *limit, uint32_t *access)
594 {
595 	int error;
596 	struct vm_seg_desc vmsegdesc;
597 
598 	bzero(&vmsegdesc, sizeof(vmsegdesc));
599 	vmsegdesc.cpuid = vcpu;
600 	vmsegdesc.regnum = reg;
601 
602 	error = ioctl(ctx->fd, VM_GET_SEGMENT_DESCRIPTOR, &vmsegdesc);
603 	if (error == 0) {
604 		*base = vmsegdesc.desc.base;
605 		*limit = vmsegdesc.desc.limit;
606 		*access = vmsegdesc.desc.access;
607 	}
608 	return (error);
609 }
610 
611 int
612 vm_get_seg_desc(struct vmctx *ctx, int vcpu, int reg, struct seg_desc *seg_desc)
613 {
614 	int error;
615 
616 	error = vm_get_desc(ctx, vcpu, reg, &seg_desc->base, &seg_desc->limit,
617 	    &seg_desc->access);
618 	return (error);
619 }
620 
621 int
622 vm_set_register(struct vmctx *ctx, int vcpu, int reg, uint64_t val)
623 {
624 	int error;
625 	struct vm_register vmreg;
626 
627 	bzero(&vmreg, sizeof(vmreg));
628 	vmreg.cpuid = vcpu;
629 	vmreg.regnum = reg;
630 	vmreg.regval = val;
631 
632 	error = ioctl(ctx->fd, VM_SET_REGISTER, &vmreg);
633 	return (error);
634 }
635 
636 int
637 vm_get_register(struct vmctx *ctx, int vcpu, int reg, uint64_t *ret_val)
638 {
639 	int error;
640 	struct vm_register vmreg;
641 
642 	bzero(&vmreg, sizeof(vmreg));
643 	vmreg.cpuid = vcpu;
644 	vmreg.regnum = reg;
645 
646 	error = ioctl(ctx->fd, VM_GET_REGISTER, &vmreg);
647 	*ret_val = vmreg.regval;
648 	return (error);
649 }
650 
651 int
652 vm_set_register_set(struct vmctx *ctx, int vcpu, unsigned int count,
653     const int *regnums, uint64_t *regvals)
654 {
655 	int error;
656 	struct vm_register_set vmregset;
657 
658 	bzero(&vmregset, sizeof(vmregset));
659 	vmregset.cpuid = vcpu;
660 	vmregset.count = count;
661 	vmregset.regnums = regnums;
662 	vmregset.regvals = regvals;
663 
664 	error = ioctl(ctx->fd, VM_SET_REGISTER_SET, &vmregset);
665 	return (error);
666 }
667 
668 int
669 vm_get_register_set(struct vmctx *ctx, int vcpu, unsigned int count,
670     const int *regnums, uint64_t *regvals)
671 {
672 	int error;
673 	struct vm_register_set vmregset;
674 
675 	bzero(&vmregset, sizeof(vmregset));
676 	vmregset.cpuid = vcpu;
677 	vmregset.count = count;
678 	vmregset.regnums = regnums;
679 	vmregset.regvals = regvals;
680 
681 	error = ioctl(ctx->fd, VM_GET_REGISTER_SET, &vmregset);
682 	return (error);
683 }
684 
685 int
686 vm_run(struct vmctx *ctx, int vcpu, struct vm_exit *vmexit)
687 {
688 	int error;
689 	struct vm_run vmrun;
690 
691 	bzero(&vmrun, sizeof(vmrun));
692 	vmrun.cpuid = vcpu;
693 
694 	error = ioctl(ctx->fd, VM_RUN, &vmrun);
695 	bcopy(&vmrun.vm_exit, vmexit, sizeof(struct vm_exit));
696 	return (error);
697 }
698 
699 int
700 vm_suspend(struct vmctx *ctx, enum vm_suspend_how how)
701 {
702 	struct vm_suspend vmsuspend;
703 
704 	bzero(&vmsuspend, sizeof(vmsuspend));
705 	vmsuspend.how = how;
706 	return (ioctl(ctx->fd, VM_SUSPEND, &vmsuspend));
707 }
708 
709 int
710 vm_reinit(struct vmctx *ctx)
711 {
712 
713 	return (ioctl(ctx->fd, VM_REINIT, 0));
714 }
715 
716 int
717 vm_inject_exception(struct vmctx *ctx, int vcpu, int vector, int errcode_valid,
718     uint32_t errcode, int restart_instruction)
719 {
720 	struct vm_exception exc;
721 
722 	exc.cpuid = vcpu;
723 	exc.vector = vector;
724 	exc.error_code = errcode;
725 	exc.error_code_valid = errcode_valid;
726 	exc.restart_instruction = restart_instruction;
727 
728 	return (ioctl(ctx->fd, VM_INJECT_EXCEPTION, &exc));
729 }
730 
731 int
732 vm_apicid2vcpu(struct vmctx *ctx __unused, int apicid)
733 {
734 	/*
735 	 * The apic id associated with the 'vcpu' has the same numerical value
736 	 * as the 'vcpu' itself.
737 	 */
738 	return (apicid);
739 }
740 
741 int
742 vm_lapic_irq(struct vmctx *ctx, int vcpu, int vector)
743 {
744 	struct vm_lapic_irq vmirq;
745 
746 	bzero(&vmirq, sizeof(vmirq));
747 	vmirq.cpuid = vcpu;
748 	vmirq.vector = vector;
749 
750 	return (ioctl(ctx->fd, VM_LAPIC_IRQ, &vmirq));
751 }
752 
753 int
754 vm_lapic_local_irq(struct vmctx *ctx, int vcpu, int vector)
755 {
756 	struct vm_lapic_irq vmirq;
757 
758 	bzero(&vmirq, sizeof(vmirq));
759 	vmirq.cpuid = vcpu;
760 	vmirq.vector = vector;
761 
762 	return (ioctl(ctx->fd, VM_LAPIC_LOCAL_IRQ, &vmirq));
763 }
764 
765 int
766 vm_lapic_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg)
767 {
768 	struct vm_lapic_msi vmmsi;
769 
770 	bzero(&vmmsi, sizeof(vmmsi));
771 	vmmsi.addr = addr;
772 	vmmsi.msg = msg;
773 
774 	return (ioctl(ctx->fd, VM_LAPIC_MSI, &vmmsi));
775 }
776 
777 int
778 vm_ioapic_assert_irq(struct vmctx *ctx, int irq)
779 {
780 	struct vm_ioapic_irq ioapic_irq;
781 
782 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
783 	ioapic_irq.irq = irq;
784 
785 	return (ioctl(ctx->fd, VM_IOAPIC_ASSERT_IRQ, &ioapic_irq));
786 }
787 
788 int
789 vm_ioapic_deassert_irq(struct vmctx *ctx, int irq)
790 {
791 	struct vm_ioapic_irq ioapic_irq;
792 
793 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
794 	ioapic_irq.irq = irq;
795 
796 	return (ioctl(ctx->fd, VM_IOAPIC_DEASSERT_IRQ, &ioapic_irq));
797 }
798 
799 int
800 vm_ioapic_pulse_irq(struct vmctx *ctx, int irq)
801 {
802 	struct vm_ioapic_irq ioapic_irq;
803 
804 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
805 	ioapic_irq.irq = irq;
806 
807 	return (ioctl(ctx->fd, VM_IOAPIC_PULSE_IRQ, &ioapic_irq));
808 }
809 
810 int
811 vm_ioapic_pincount(struct vmctx *ctx, int *pincount)
812 {
813 
814 	return (ioctl(ctx->fd, VM_IOAPIC_PINCOUNT, pincount));
815 }
816 
817 int
818 vm_readwrite_kernemu_device(struct vmctx *ctx, int vcpu, vm_paddr_t gpa,
819     bool write, int size, uint64_t *value)
820 {
821 	struct vm_readwrite_kernemu_device irp = {
822 		.vcpuid = vcpu,
823 		.access_width = fls(size) - 1,
824 		.gpa = gpa,
825 		.value = write ? *value : ~0ul,
826 	};
827 	long cmd = (write ? VM_SET_KERNEMU_DEV : VM_GET_KERNEMU_DEV);
828 	int rc;
829 
830 	rc = ioctl(ctx->fd, cmd, &irp);
831 	if (rc == 0 && !write)
832 		*value = irp.value;
833 	return (rc);
834 }
835 
836 int
837 vm_isa_assert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
838 {
839 	struct vm_isa_irq isa_irq;
840 
841 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
842 	isa_irq.atpic_irq = atpic_irq;
843 	isa_irq.ioapic_irq = ioapic_irq;
844 
845 	return (ioctl(ctx->fd, VM_ISA_ASSERT_IRQ, &isa_irq));
846 }
847 
848 int
849 vm_isa_deassert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
850 {
851 	struct vm_isa_irq isa_irq;
852 
853 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
854 	isa_irq.atpic_irq = atpic_irq;
855 	isa_irq.ioapic_irq = ioapic_irq;
856 
857 	return (ioctl(ctx->fd, VM_ISA_DEASSERT_IRQ, &isa_irq));
858 }
859 
860 int
861 vm_isa_pulse_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
862 {
863 	struct vm_isa_irq isa_irq;
864 
865 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
866 	isa_irq.atpic_irq = atpic_irq;
867 	isa_irq.ioapic_irq = ioapic_irq;
868 
869 	return (ioctl(ctx->fd, VM_ISA_PULSE_IRQ, &isa_irq));
870 }
871 
872 int
873 vm_isa_set_irq_trigger(struct vmctx *ctx, int atpic_irq,
874     enum vm_intr_trigger trigger)
875 {
876 	struct vm_isa_irq_trigger isa_irq_trigger;
877 
878 	bzero(&isa_irq_trigger, sizeof(struct vm_isa_irq_trigger));
879 	isa_irq_trigger.atpic_irq = atpic_irq;
880 	isa_irq_trigger.trigger = trigger;
881 
882 	return (ioctl(ctx->fd, VM_ISA_SET_IRQ_TRIGGER, &isa_irq_trigger));
883 }
884 
885 int
886 vm_inject_nmi(struct vmctx *ctx, int vcpu)
887 {
888 	struct vm_nmi vmnmi;
889 
890 	bzero(&vmnmi, sizeof(vmnmi));
891 	vmnmi.cpuid = vcpu;
892 
893 	return (ioctl(ctx->fd, VM_INJECT_NMI, &vmnmi));
894 }
895 
896 static const char *capstrmap[] = {
897 	[VM_CAP_HALT_EXIT]  = "hlt_exit",
898 	[VM_CAP_MTRAP_EXIT] = "mtrap_exit",
899 	[VM_CAP_PAUSE_EXIT] = "pause_exit",
900 	[VM_CAP_UNRESTRICTED_GUEST] = "unrestricted_guest",
901 	[VM_CAP_ENABLE_INVPCID] = "enable_invpcid",
902 	[VM_CAP_BPT_EXIT] = "bpt_exit",
903 };
904 
905 int
906 vm_capability_name2type(const char *capname)
907 {
908 	int i;
909 
910 	for (i = 0; i < (int)nitems(capstrmap); i++) {
911 		if (strcmp(capstrmap[i], capname) == 0)
912 			return (i);
913 	}
914 
915 	return (-1);
916 }
917 
918 const char *
919 vm_capability_type2name(int type)
920 {
921 	if (type >= 0 && type < (int)nitems(capstrmap))
922 		return (capstrmap[type]);
923 
924 	return (NULL);
925 }
926 
927 int
928 vm_get_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap,
929 		  int *retval)
930 {
931 	int error;
932 	struct vm_capability vmcap;
933 
934 	bzero(&vmcap, sizeof(vmcap));
935 	vmcap.cpuid = vcpu;
936 	vmcap.captype = cap;
937 
938 	error = ioctl(ctx->fd, VM_GET_CAPABILITY, &vmcap);
939 	*retval = vmcap.capval;
940 	return (error);
941 }
942 
943 int
944 vm_set_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap, int val)
945 {
946 	struct vm_capability vmcap;
947 
948 	bzero(&vmcap, sizeof(vmcap));
949 	vmcap.cpuid = vcpu;
950 	vmcap.captype = cap;
951 	vmcap.capval = val;
952 
953 	return (ioctl(ctx->fd, VM_SET_CAPABILITY, &vmcap));
954 }
955 
956 int
957 vm_assign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
958 {
959 	struct vm_pptdev pptdev;
960 
961 	bzero(&pptdev, sizeof(pptdev));
962 	pptdev.bus = bus;
963 	pptdev.slot = slot;
964 	pptdev.func = func;
965 
966 	return (ioctl(ctx->fd, VM_BIND_PPTDEV, &pptdev));
967 }
968 
969 int
970 vm_unassign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
971 {
972 	struct vm_pptdev pptdev;
973 
974 	bzero(&pptdev, sizeof(pptdev));
975 	pptdev.bus = bus;
976 	pptdev.slot = slot;
977 	pptdev.func = func;
978 
979 	return (ioctl(ctx->fd, VM_UNBIND_PPTDEV, &pptdev));
980 }
981 
982 int
983 vm_map_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func,
984 		   vm_paddr_t gpa, size_t len, vm_paddr_t hpa)
985 {
986 	struct vm_pptdev_mmio pptmmio;
987 
988 	bzero(&pptmmio, sizeof(pptmmio));
989 	pptmmio.bus = bus;
990 	pptmmio.slot = slot;
991 	pptmmio.func = func;
992 	pptmmio.gpa = gpa;
993 	pptmmio.len = len;
994 	pptmmio.hpa = hpa;
995 
996 	return (ioctl(ctx->fd, VM_MAP_PPTDEV_MMIO, &pptmmio));
997 }
998 
999 int
1000 vm_unmap_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func,
1001 		     vm_paddr_t gpa, size_t len)
1002 {
1003 	struct vm_pptdev_mmio pptmmio;
1004 
1005 	bzero(&pptmmio, sizeof(pptmmio));
1006 	pptmmio.bus = bus;
1007 	pptmmio.slot = slot;
1008 	pptmmio.func = func;
1009 	pptmmio.gpa = gpa;
1010 	pptmmio.len = len;
1011 
1012 	return (ioctl(ctx->fd, VM_UNMAP_PPTDEV_MMIO, &pptmmio));
1013 }
1014 
1015 int
1016 vm_setup_pptdev_msi(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
1017     uint64_t addr, uint64_t msg, int numvec)
1018 {
1019 	struct vm_pptdev_msi pptmsi;
1020 
1021 	bzero(&pptmsi, sizeof(pptmsi));
1022 	pptmsi.vcpu = vcpu;
1023 	pptmsi.bus = bus;
1024 	pptmsi.slot = slot;
1025 	pptmsi.func = func;
1026 	pptmsi.msg = msg;
1027 	pptmsi.addr = addr;
1028 	pptmsi.numvec = numvec;
1029 
1030 	return (ioctl(ctx->fd, VM_PPTDEV_MSI, &pptmsi));
1031 }
1032 
1033 int
1034 vm_setup_pptdev_msix(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
1035     int idx, uint64_t addr, uint64_t msg, uint32_t vector_control)
1036 {
1037 	struct vm_pptdev_msix pptmsix;
1038 
1039 	bzero(&pptmsix, sizeof(pptmsix));
1040 	pptmsix.vcpu = vcpu;
1041 	pptmsix.bus = bus;
1042 	pptmsix.slot = slot;
1043 	pptmsix.func = func;
1044 	pptmsix.idx = idx;
1045 	pptmsix.msg = msg;
1046 	pptmsix.addr = addr;
1047 	pptmsix.vector_control = vector_control;
1048 
1049 	return ioctl(ctx->fd, VM_PPTDEV_MSIX, &pptmsix);
1050 }
1051 
1052 int
1053 vm_disable_pptdev_msix(struct vmctx *ctx, int bus, int slot, int func)
1054 {
1055 	struct vm_pptdev ppt;
1056 
1057 	bzero(&ppt, sizeof(ppt));
1058 	ppt.bus = bus;
1059 	ppt.slot = slot;
1060 	ppt.func = func;
1061 
1062 	return ioctl(ctx->fd, VM_PPTDEV_DISABLE_MSIX, &ppt);
1063 }
1064 
1065 uint64_t *
1066 vm_get_stats(struct vmctx *ctx, int vcpu, struct timeval *ret_tv,
1067 	     int *ret_entries)
1068 {
1069 	int error;
1070 
1071 	static struct vm_stats vmstats;
1072 
1073 	vmstats.cpuid = vcpu;
1074 
1075 	error = ioctl(ctx->fd, VM_STATS, &vmstats);
1076 	if (error == 0) {
1077 		if (ret_entries)
1078 			*ret_entries = vmstats.num_entries;
1079 		if (ret_tv)
1080 			*ret_tv = vmstats.tv;
1081 		return (vmstats.statbuf);
1082 	} else
1083 		return (NULL);
1084 }
1085 
1086 const char *
1087 vm_get_stat_desc(struct vmctx *ctx, int index)
1088 {
1089 	static struct vm_stat_desc statdesc;
1090 
1091 	statdesc.index = index;
1092 	if (ioctl(ctx->fd, VM_STAT_DESC, &statdesc) == 0)
1093 		return (statdesc.desc);
1094 	else
1095 		return (NULL);
1096 }
1097 
1098 int
1099 vm_get_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state *state)
1100 {
1101 	int error;
1102 	struct vm_x2apic x2apic;
1103 
1104 	bzero(&x2apic, sizeof(x2apic));
1105 	x2apic.cpuid = vcpu;
1106 
1107 	error = ioctl(ctx->fd, VM_GET_X2APIC_STATE, &x2apic);
1108 	*state = x2apic.state;
1109 	return (error);
1110 }
1111 
1112 int
1113 vm_set_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state state)
1114 {
1115 	int error;
1116 	struct vm_x2apic x2apic;
1117 
1118 	bzero(&x2apic, sizeof(x2apic));
1119 	x2apic.cpuid = vcpu;
1120 	x2apic.state = state;
1121 
1122 	error = ioctl(ctx->fd, VM_SET_X2APIC_STATE, &x2apic);
1123 
1124 	return (error);
1125 }
1126 
1127 /*
1128  * From Intel Vol 3a:
1129  * Table 9-1. IA-32 Processor States Following Power-up, Reset or INIT
1130  */
1131 int
1132 vcpu_reset(struct vmctx *vmctx, int vcpu)
1133 {
1134 	int error;
1135 	uint64_t rflags, rip, cr0, cr4, zero, desc_base, rdx;
1136 	uint32_t desc_access, desc_limit;
1137 	uint16_t sel;
1138 
1139 	zero = 0;
1140 
1141 	rflags = 0x2;
1142 	error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RFLAGS, rflags);
1143 	if (error)
1144 		goto done;
1145 
1146 	rip = 0xfff0;
1147 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RIP, rip)) != 0)
1148 		goto done;
1149 
1150 	cr0 = CR0_NE;
1151 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR0, cr0)) != 0)
1152 		goto done;
1153 
1154 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR3, zero)) != 0)
1155 		goto done;
1156 
1157 	cr4 = 0;
1158 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR4, cr4)) != 0)
1159 		goto done;
1160 
1161 	/*
1162 	 * CS: present, r/w, accessed, 16-bit, byte granularity, usable
1163 	 */
1164 	desc_base = 0xffff0000;
1165 	desc_limit = 0xffff;
1166 	desc_access = 0x0093;
1167 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_CS,
1168 			    desc_base, desc_limit, desc_access);
1169 	if (error)
1170 		goto done;
1171 
1172 	sel = 0xf000;
1173 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CS, sel)) != 0)
1174 		goto done;
1175 
1176 	/*
1177 	 * SS,DS,ES,FS,GS: present, r/w, accessed, 16-bit, byte granularity
1178 	 */
1179 	desc_base = 0;
1180 	desc_limit = 0xffff;
1181 	desc_access = 0x0093;
1182 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_SS,
1183 			    desc_base, desc_limit, desc_access);
1184 	if (error)
1185 		goto done;
1186 
1187 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_DS,
1188 			    desc_base, desc_limit, desc_access);
1189 	if (error)
1190 		goto done;
1191 
1192 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_ES,
1193 			    desc_base, desc_limit, desc_access);
1194 	if (error)
1195 		goto done;
1196 
1197 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_FS,
1198 			    desc_base, desc_limit, desc_access);
1199 	if (error)
1200 		goto done;
1201 
1202 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GS,
1203 			    desc_base, desc_limit, desc_access);
1204 	if (error)
1205 		goto done;
1206 
1207 	sel = 0;
1208 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_SS, sel)) != 0)
1209 		goto done;
1210 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_DS, sel)) != 0)
1211 		goto done;
1212 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_ES, sel)) != 0)
1213 		goto done;
1214 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_FS, sel)) != 0)
1215 		goto done;
1216 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_GS, sel)) != 0)
1217 		goto done;
1218 
1219 	/* General purpose registers */
1220 	rdx = 0xf00;
1221 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RAX, zero)) != 0)
1222 		goto done;
1223 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBX, zero)) != 0)
1224 		goto done;
1225 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RCX, zero)) != 0)
1226 		goto done;
1227 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDX, rdx)) != 0)
1228 		goto done;
1229 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSI, zero)) != 0)
1230 		goto done;
1231 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDI, zero)) != 0)
1232 		goto done;
1233 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBP, zero)) != 0)
1234 		goto done;
1235 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSP, zero)) != 0)
1236 		goto done;
1237 
1238 	/* GDTR, IDTR */
1239 	desc_base = 0;
1240 	desc_limit = 0xffff;
1241 	desc_access = 0;
1242 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GDTR,
1243 			    desc_base, desc_limit, desc_access);
1244 	if (error != 0)
1245 		goto done;
1246 
1247 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_IDTR,
1248 			    desc_base, desc_limit, desc_access);
1249 	if (error != 0)
1250 		goto done;
1251 
1252 	/* TR */
1253 	desc_base = 0;
1254 	desc_limit = 0xffff;
1255 	desc_access = 0x0000008b;
1256 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_TR, 0, 0, desc_access);
1257 	if (error)
1258 		goto done;
1259 
1260 	sel = 0;
1261 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_TR, sel)) != 0)
1262 		goto done;
1263 
1264 	/* LDTR */
1265 	desc_base = 0;
1266 	desc_limit = 0xffff;
1267 	desc_access = 0x00000082;
1268 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_LDTR, desc_base,
1269 			    desc_limit, desc_access);
1270 	if (error)
1271 		goto done;
1272 
1273 	sel = 0;
1274 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_LDTR, 0)) != 0)
1275 		goto done;
1276 
1277 	/* XXX cr2, debug registers */
1278 
1279 	error = 0;
1280 done:
1281 	return (error);
1282 }
1283 
1284 int
1285 vm_get_gpa_pmap(struct vmctx *ctx, uint64_t gpa, uint64_t *pte, int *num)
1286 {
1287 	int error, i;
1288 	struct vm_gpa_pte gpapte;
1289 
1290 	bzero(&gpapte, sizeof(gpapte));
1291 	gpapte.gpa = gpa;
1292 
1293 	error = ioctl(ctx->fd, VM_GET_GPA_PMAP, &gpapte);
1294 
1295 	if (error == 0) {
1296 		*num = gpapte.ptenum;
1297 		for (i = 0; i < gpapte.ptenum; i++)
1298 			pte[i] = gpapte.pte[i];
1299 	}
1300 
1301 	return (error);
1302 }
1303 
1304 int
1305 vm_get_hpet_capabilities(struct vmctx *ctx, uint32_t *capabilities)
1306 {
1307 	int error;
1308 	struct vm_hpet_cap cap;
1309 
1310 	bzero(&cap, sizeof(struct vm_hpet_cap));
1311 	error = ioctl(ctx->fd, VM_GET_HPET_CAPABILITIES, &cap);
1312 	if (capabilities != NULL)
1313 		*capabilities = cap.capabilities;
1314 	return (error);
1315 }
1316 
1317 int
1318 vm_gla2gpa(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1319     uint64_t gla, int prot, uint64_t *gpa, int *fault)
1320 {
1321 	struct vm_gla2gpa gg;
1322 	int error;
1323 
1324 	bzero(&gg, sizeof(struct vm_gla2gpa));
1325 	gg.vcpuid = vcpu;
1326 	gg.prot = prot;
1327 	gg.gla = gla;
1328 	gg.paging = *paging;
1329 
1330 	error = ioctl(ctx->fd, VM_GLA2GPA, &gg);
1331 	if (error == 0) {
1332 		*fault = gg.fault;
1333 		*gpa = gg.gpa;
1334 	}
1335 	return (error);
1336 }
1337 
1338 int
1339 vm_gla2gpa_nofault(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1340     uint64_t gla, int prot, uint64_t *gpa, int *fault)
1341 {
1342 	struct vm_gla2gpa gg;
1343 	int error;
1344 
1345 	bzero(&gg, sizeof(struct vm_gla2gpa));
1346 	gg.vcpuid = vcpu;
1347 	gg.prot = prot;
1348 	gg.gla = gla;
1349 	gg.paging = *paging;
1350 
1351 	error = ioctl(ctx->fd, VM_GLA2GPA_NOFAULT, &gg);
1352 	if (error == 0) {
1353 		*fault = gg.fault;
1354 		*gpa = gg.gpa;
1355 	}
1356 	return (error);
1357 }
1358 
1359 #ifndef min
1360 #define	min(a,b)	(((a) < (b)) ? (a) : (b))
1361 #endif
1362 
1363 int
1364 vm_copy_setup(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1365     uint64_t gla, size_t len, int prot, struct iovec *iov, int iovcnt,
1366     int *fault)
1367 {
1368 	void *va;
1369 	uint64_t gpa, off;
1370 	int error, i, n;
1371 
1372 	for (i = 0; i < iovcnt; i++) {
1373 		iov[i].iov_base = 0;
1374 		iov[i].iov_len = 0;
1375 	}
1376 
1377 	while (len) {
1378 		assert(iovcnt > 0);
1379 		error = vm_gla2gpa(ctx, vcpu, paging, gla, prot, &gpa, fault);
1380 		if (error || *fault)
1381 			return (error);
1382 
1383 		off = gpa & PAGE_MASK;
1384 		n = MIN(len, PAGE_SIZE - off);
1385 
1386 		va = vm_map_gpa(ctx, gpa, n);
1387 		if (va == NULL)
1388 			return (EFAULT);
1389 
1390 		iov->iov_base = va;
1391 		iov->iov_len = n;
1392 		iov++;
1393 		iovcnt--;
1394 
1395 		gla += n;
1396 		len -= n;
1397 	}
1398 	return (0);
1399 }
1400 
1401 void
1402 vm_copy_teardown(struct vmctx *ctx __unused, int vcpu __unused,
1403     struct iovec *iov __unused, int iovcnt __unused)
1404 {
1405 }
1406 
1407 void
1408 vm_copyin(struct vmctx *ctx __unused, int vcpu __unused, struct iovec *iov,
1409     void *vp, size_t len)
1410 {
1411 	const char *src;
1412 	char *dst;
1413 	size_t n;
1414 
1415 	dst = vp;
1416 	while (len) {
1417 		assert(iov->iov_len);
1418 		n = min(len, iov->iov_len);
1419 		src = iov->iov_base;
1420 		bcopy(src, dst, n);
1421 
1422 		iov++;
1423 		dst += n;
1424 		len -= n;
1425 	}
1426 }
1427 
1428 void
1429 vm_copyout(struct vmctx *ctx __unused, int vcpu __unused, const void *vp,
1430     struct iovec *iov, size_t len)
1431 {
1432 	const char *src;
1433 	char *dst;
1434 	size_t n;
1435 
1436 	src = vp;
1437 	while (len) {
1438 		assert(iov->iov_len);
1439 		n = min(len, iov->iov_len);
1440 		dst = iov->iov_base;
1441 		bcopy(src, dst, n);
1442 
1443 		iov++;
1444 		src += n;
1445 		len -= n;
1446 	}
1447 }
1448 
1449 static int
1450 vm_get_cpus(struct vmctx *ctx, int which, cpuset_t *cpus)
1451 {
1452 	struct vm_cpuset vm_cpuset;
1453 	int error;
1454 
1455 	bzero(&vm_cpuset, sizeof(struct vm_cpuset));
1456 	vm_cpuset.which = which;
1457 	vm_cpuset.cpusetsize = sizeof(cpuset_t);
1458 	vm_cpuset.cpus = cpus;
1459 
1460 	error = ioctl(ctx->fd, VM_GET_CPUS, &vm_cpuset);
1461 	return (error);
1462 }
1463 
1464 int
1465 vm_active_cpus(struct vmctx *ctx, cpuset_t *cpus)
1466 {
1467 
1468 	return (vm_get_cpus(ctx, VM_ACTIVE_CPUS, cpus));
1469 }
1470 
1471 int
1472 vm_suspended_cpus(struct vmctx *ctx, cpuset_t *cpus)
1473 {
1474 
1475 	return (vm_get_cpus(ctx, VM_SUSPENDED_CPUS, cpus));
1476 }
1477 
1478 int
1479 vm_debug_cpus(struct vmctx *ctx, cpuset_t *cpus)
1480 {
1481 
1482 	return (vm_get_cpus(ctx, VM_DEBUG_CPUS, cpus));
1483 }
1484 
1485 int
1486 vm_activate_cpu(struct vmctx *ctx, int vcpu)
1487 {
1488 	struct vm_activate_cpu ac;
1489 	int error;
1490 
1491 	bzero(&ac, sizeof(struct vm_activate_cpu));
1492 	ac.vcpuid = vcpu;
1493 	error = ioctl(ctx->fd, VM_ACTIVATE_CPU, &ac);
1494 	return (error);
1495 }
1496 
1497 int
1498 vm_suspend_cpu(struct vmctx *ctx, int vcpu)
1499 {
1500 	struct vm_activate_cpu ac;
1501 	int error;
1502 
1503 	bzero(&ac, sizeof(struct vm_activate_cpu));
1504 	ac.vcpuid = vcpu;
1505 	error = ioctl(ctx->fd, VM_SUSPEND_CPU, &ac);
1506 	return (error);
1507 }
1508 
1509 int
1510 vm_resume_cpu(struct vmctx *ctx, int vcpu)
1511 {
1512 	struct vm_activate_cpu ac;
1513 	int error;
1514 
1515 	bzero(&ac, sizeof(struct vm_activate_cpu));
1516 	ac.vcpuid = vcpu;
1517 	error = ioctl(ctx->fd, VM_RESUME_CPU, &ac);
1518 	return (error);
1519 }
1520 
1521 int
1522 vm_get_intinfo(struct vmctx *ctx, int vcpu, uint64_t *info1, uint64_t *info2)
1523 {
1524 	struct vm_intinfo vmii;
1525 	int error;
1526 
1527 	bzero(&vmii, sizeof(struct vm_intinfo));
1528 	vmii.vcpuid = vcpu;
1529 	error = ioctl(ctx->fd, VM_GET_INTINFO, &vmii);
1530 	if (error == 0) {
1531 		*info1 = vmii.info1;
1532 		*info2 = vmii.info2;
1533 	}
1534 	return (error);
1535 }
1536 
1537 int
1538 vm_set_intinfo(struct vmctx *ctx, int vcpu, uint64_t info1)
1539 {
1540 	struct vm_intinfo vmii;
1541 	int error;
1542 
1543 	bzero(&vmii, sizeof(struct vm_intinfo));
1544 	vmii.vcpuid = vcpu;
1545 	vmii.info1 = info1;
1546 	error = ioctl(ctx->fd, VM_SET_INTINFO, &vmii);
1547 	return (error);
1548 }
1549 
1550 int
1551 vm_rtc_write(struct vmctx *ctx, int offset, uint8_t value)
1552 {
1553 	struct vm_rtc_data rtcdata;
1554 	int error;
1555 
1556 	bzero(&rtcdata, sizeof(struct vm_rtc_data));
1557 	rtcdata.offset = offset;
1558 	rtcdata.value = value;
1559 	error = ioctl(ctx->fd, VM_RTC_WRITE, &rtcdata);
1560 	return (error);
1561 }
1562 
1563 int
1564 vm_rtc_read(struct vmctx *ctx, int offset, uint8_t *retval)
1565 {
1566 	struct vm_rtc_data rtcdata;
1567 	int error;
1568 
1569 	bzero(&rtcdata, sizeof(struct vm_rtc_data));
1570 	rtcdata.offset = offset;
1571 	error = ioctl(ctx->fd, VM_RTC_READ, &rtcdata);
1572 	if (error == 0)
1573 		*retval = rtcdata.value;
1574 	return (error);
1575 }
1576 
1577 int
1578 vm_rtc_settime(struct vmctx *ctx, time_t secs)
1579 {
1580 	struct vm_rtc_time rtctime;
1581 	int error;
1582 
1583 	bzero(&rtctime, sizeof(struct vm_rtc_time));
1584 	rtctime.secs = secs;
1585 	error = ioctl(ctx->fd, VM_RTC_SETTIME, &rtctime);
1586 	return (error);
1587 }
1588 
1589 int
1590 vm_rtc_gettime(struct vmctx *ctx, time_t *secs)
1591 {
1592 	struct vm_rtc_time rtctime;
1593 	int error;
1594 
1595 	bzero(&rtctime, sizeof(struct vm_rtc_time));
1596 	error = ioctl(ctx->fd, VM_RTC_GETTIME, &rtctime);
1597 	if (error == 0)
1598 		*secs = rtctime.secs;
1599 	return (error);
1600 }
1601 
1602 int
1603 vm_restart_instruction(void *arg, int vcpu)
1604 {
1605 	struct vmctx *ctx = arg;
1606 
1607 	return (ioctl(ctx->fd, VM_RESTART_INSTRUCTION, &vcpu));
1608 }
1609 
1610 int
1611 vm_snapshot_req(struct vm_snapshot_meta *meta)
1612 {
1613 
1614 	if (ioctl(meta->ctx->fd, VM_SNAPSHOT_REQ, meta) == -1) {
1615 #ifdef SNAPSHOT_DEBUG
1616 		fprintf(stderr, "%s: snapshot failed for %s: %d\r\n",
1617 		    __func__, meta->dev_name, errno);
1618 #endif
1619 		return (-1);
1620 	}
1621 	return (0);
1622 }
1623 
1624 int
1625 vm_restore_time(struct vmctx *ctx)
1626 {
1627 	int dummy;
1628 
1629 	dummy = 0;
1630 	return (ioctl(ctx->fd, VM_RESTORE_TIME, &dummy));
1631 }
1632 
1633 int
1634 vm_set_topology(struct vmctx *ctx,
1635     uint16_t sockets, uint16_t cores, uint16_t threads, uint16_t maxcpus)
1636 {
1637 	struct vm_cpu_topology topology;
1638 
1639 	bzero(&topology, sizeof (struct vm_cpu_topology));
1640 	topology.sockets = sockets;
1641 	topology.cores = cores;
1642 	topology.threads = threads;
1643 	topology.maxcpus = maxcpus;
1644 	return (ioctl(ctx->fd, VM_SET_TOPOLOGY, &topology));
1645 }
1646 
1647 int
1648 vm_get_topology(struct vmctx *ctx,
1649     uint16_t *sockets, uint16_t *cores, uint16_t *threads, uint16_t *maxcpus)
1650 {
1651 	struct vm_cpu_topology topology;
1652 	int error;
1653 
1654 	bzero(&topology, sizeof (struct vm_cpu_topology));
1655 	error = ioctl(ctx->fd, VM_GET_TOPOLOGY, &topology);
1656 	if (error == 0) {
1657 		*sockets = topology.sockets;
1658 		*cores = topology.cores;
1659 		*threads = topology.threads;
1660 		*maxcpus = topology.maxcpus;
1661 	}
1662 	return (error);
1663 }
1664 
1665 int
1666 vm_get_device_fd(struct vmctx *ctx)
1667 {
1668 
1669 	return (ctx->fd);
1670 }
1671 
1672 const cap_ioctl_t *
1673 vm_get_ioctls(size_t *len)
1674 {
1675 	cap_ioctl_t *cmds;
1676 	/* keep in sync with machine/vmm_dev.h */
1677 	static const cap_ioctl_t vm_ioctl_cmds[] = { VM_RUN, VM_SUSPEND, VM_REINIT,
1678 	    VM_ALLOC_MEMSEG, VM_GET_MEMSEG, VM_MMAP_MEMSEG, VM_MMAP_MEMSEG,
1679 	    VM_MMAP_GETNEXT, VM_MUNMAP_MEMSEG, VM_SET_REGISTER, VM_GET_REGISTER,
1680 	    VM_SET_SEGMENT_DESCRIPTOR, VM_GET_SEGMENT_DESCRIPTOR,
1681 	    VM_SET_REGISTER_SET, VM_GET_REGISTER_SET,
1682 	    VM_SET_KERNEMU_DEV, VM_GET_KERNEMU_DEV,
1683 	    VM_INJECT_EXCEPTION, VM_LAPIC_IRQ, VM_LAPIC_LOCAL_IRQ,
1684 	    VM_LAPIC_MSI, VM_IOAPIC_ASSERT_IRQ, VM_IOAPIC_DEASSERT_IRQ,
1685 	    VM_IOAPIC_PULSE_IRQ, VM_IOAPIC_PINCOUNT, VM_ISA_ASSERT_IRQ,
1686 	    VM_ISA_DEASSERT_IRQ, VM_ISA_PULSE_IRQ, VM_ISA_SET_IRQ_TRIGGER,
1687 	    VM_SET_CAPABILITY, VM_GET_CAPABILITY, VM_BIND_PPTDEV,
1688 	    VM_UNBIND_PPTDEV, VM_MAP_PPTDEV_MMIO, VM_PPTDEV_MSI,
1689 	    VM_PPTDEV_MSIX, VM_UNMAP_PPTDEV_MMIO, VM_PPTDEV_DISABLE_MSIX,
1690 	    VM_INJECT_NMI, VM_STATS, VM_STAT_DESC,
1691 	    VM_SET_X2APIC_STATE, VM_GET_X2APIC_STATE,
1692 	    VM_GET_HPET_CAPABILITIES, VM_GET_GPA_PMAP, VM_GLA2GPA,
1693 	    VM_GLA2GPA_NOFAULT,
1694 	    VM_ACTIVATE_CPU, VM_GET_CPUS, VM_SUSPEND_CPU, VM_RESUME_CPU,
1695 	    VM_SET_INTINFO, VM_GET_INTINFO,
1696 	    VM_RTC_WRITE, VM_RTC_READ, VM_RTC_SETTIME, VM_RTC_GETTIME,
1697 	    VM_RESTART_INSTRUCTION, VM_SET_TOPOLOGY, VM_GET_TOPOLOGY };
1698 
1699 	if (len == NULL) {
1700 		cmds = malloc(sizeof(vm_ioctl_cmds));
1701 		if (cmds == NULL)
1702 			return (NULL);
1703 		bcopy(vm_ioctl_cmds, cmds, sizeof(vm_ioctl_cmds));
1704 		return (cmds);
1705 	}
1706 
1707 	*len = nitems(vm_ioctl_cmds);
1708 	return (NULL);
1709 }
1710