xref: /freebsd/lib/libvmmapi/vmmapi.c (revision b0a5c9889882332a976fa379176c249dc55efa63)
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/mman.h>
38 #include <sys/_iovec.h>
39 #include <sys/cpuset.h>
40 
41 #include <x86/segments.h>
42 #include <machine/specialreg.h>
43 
44 #include <errno.h>
45 #include <stdio.h>
46 #include <stdlib.h>
47 #include <assert.h>
48 #include <string.h>
49 #include <fcntl.h>
50 #include <unistd.h>
51 
52 #include <libutil.h>
53 
54 #include <machine/vmm.h>
55 #include <machine/vmm_dev.h>
56 
57 #include "vmmapi.h"
58 
59 #define	MB	(1024 * 1024UL)
60 #define	GB	(1024 * 1024 * 1024UL)
61 
62 /*
63  * Size of the guard region before and after the virtual address space
64  * mapping the guest physical memory. This must be a multiple of the
65  * superpage size for performance reasons.
66  */
67 #define	VM_MMAP_GUARD_SIZE	(4 * MB)
68 
69 #define	PROT_RW		(PROT_READ | PROT_WRITE)
70 #define	PROT_ALL	(PROT_READ | PROT_WRITE | PROT_EXEC)
71 
72 struct vmctx {
73 	int	fd;
74 	uint32_t lowmem_limit;
75 	int	memflags;
76 	size_t	lowmem;
77 	size_t	highmem;
78 	char	*baseaddr;
79 	char	*name;
80 };
81 
82 #define	CREATE(x)  sysctlbyname("hw.vmm.create", NULL, NULL, (x), strlen((x)))
83 #define	DESTROY(x) sysctlbyname("hw.vmm.destroy", NULL, NULL, (x), strlen((x)))
84 
85 static int
86 vm_device_open(const char *name)
87 {
88         int fd, len;
89         char *vmfile;
90 
91 	len = strlen("/dev/vmm/") + strlen(name) + 1;
92 	vmfile = malloc(len);
93 	assert(vmfile != NULL);
94 	snprintf(vmfile, len, "/dev/vmm/%s", name);
95 
96         /* Open the device file */
97         fd = open(vmfile, O_RDWR, 0);
98 
99 	free(vmfile);
100         return (fd);
101 }
102 
103 int
104 vm_create(const char *name)
105 {
106 
107 	return (CREATE((char *)name));
108 }
109 
110 struct vmctx *
111 vm_open(const char *name)
112 {
113 	struct vmctx *vm;
114 
115 	vm = malloc(sizeof(struct vmctx) + strlen(name) + 1);
116 	assert(vm != NULL);
117 
118 	vm->fd = -1;
119 	vm->memflags = 0;
120 	vm->lowmem_limit = 3 * GB;
121 	vm->name = (char *)(vm + 1);
122 	strcpy(vm->name, name);
123 
124 	if ((vm->fd = vm_device_open(vm->name)) < 0)
125 		goto err;
126 
127 	return (vm);
128 err:
129 	vm_destroy(vm);
130 	return (NULL);
131 }
132 
133 void
134 vm_destroy(struct vmctx *vm)
135 {
136 	assert(vm != NULL);
137 
138 	if (vm->fd >= 0)
139 		close(vm->fd);
140 	DESTROY(vm->name);
141 
142 	free(vm);
143 }
144 
145 int
146 vm_parse_memsize(const char *optarg, size_t *ret_memsize)
147 {
148 	char *endptr;
149 	size_t optval;
150 	int error;
151 
152 	optval = strtoul(optarg, &endptr, 0);
153 	if (*optarg != '\0' && *endptr == '\0') {
154 		/*
155 		 * For the sake of backward compatibility if the memory size
156 		 * specified on the command line is less than a megabyte then
157 		 * it is interpreted as being in units of MB.
158 		 */
159 		if (optval < MB)
160 			optval *= MB;
161 		*ret_memsize = optval;
162 		error = 0;
163 	} else
164 		error = expand_number(optarg, ret_memsize);
165 
166 	return (error);
167 }
168 
169 uint32_t
170 vm_get_lowmem_limit(struct vmctx *ctx)
171 {
172 
173 	return (ctx->lowmem_limit);
174 }
175 
176 void
177 vm_set_lowmem_limit(struct vmctx *ctx, uint32_t limit)
178 {
179 
180 	ctx->lowmem_limit = limit;
181 }
182 
183 void
184 vm_set_memflags(struct vmctx *ctx, int flags)
185 {
186 
187 	ctx->memflags = flags;
188 }
189 
190 int
191 vm_get_memflags(struct vmctx *ctx)
192 {
193 
194 	return (ctx->memflags);
195 }
196 
197 /*
198  * Map segment 'segid' starting at 'off' into guest address range [gpa,gpa+len).
199  */
200 int
201 vm_mmap_memseg(struct vmctx *ctx, vm_paddr_t gpa, int segid, vm_ooffset_t off,
202     size_t len, int prot)
203 {
204 	struct vm_memmap memmap;
205 	int error, flags;
206 
207 	memmap.gpa = gpa;
208 	memmap.segid = segid;
209 	memmap.segoff = off;
210 	memmap.len = len;
211 	memmap.prot = prot;
212 	memmap.flags = 0;
213 
214 	if (ctx->memflags & VM_MEM_F_WIRED)
215 		memmap.flags |= VM_MEMMAP_F_WIRED;
216 
217 	/*
218 	 * If this mapping already exists then don't create it again. This
219 	 * is the common case for SYSMEM mappings created by bhyveload(8).
220 	 */
221 	error = vm_mmap_getnext(ctx, &gpa, &segid, &off, &len, &prot, &flags);
222 	if (error == 0 && gpa == memmap.gpa) {
223 		if (segid != memmap.segid || off != memmap.segoff ||
224 		    prot != memmap.prot || flags != memmap.flags) {
225 			errno = EEXIST;
226 			return (-1);
227 		} else {
228 			return (0);
229 		}
230 	}
231 
232 	error = ioctl(ctx->fd, VM_MMAP_MEMSEG, &memmap);
233 	return (error);
234 }
235 
236 int
237 vm_mmap_getnext(struct vmctx *ctx, vm_paddr_t *gpa, int *segid,
238     vm_ooffset_t *segoff, size_t *len, int *prot, int *flags)
239 {
240 	struct vm_memmap memmap;
241 	int error;
242 
243 	bzero(&memmap, sizeof(struct vm_memmap));
244 	memmap.gpa = *gpa;
245 	error = ioctl(ctx->fd, VM_MMAP_GETNEXT, &memmap);
246 	if (error == 0) {
247 		*gpa = memmap.gpa;
248 		*segid = memmap.segid;
249 		*segoff = memmap.segoff;
250 		*len = memmap.len;
251 		*prot = memmap.prot;
252 		*flags = memmap.flags;
253 	}
254 	return (error);
255 }
256 
257 /*
258  * Return 0 if the segments are identical and non-zero otherwise.
259  *
260  * This is slightly complicated by the fact that only device memory segments
261  * are named.
262  */
263 static int
264 cmpseg(size_t len, const char *str, size_t len2, const char *str2)
265 {
266 
267 	if (len == len2) {
268 		if ((!str && !str2) || (str && str2 && !strcmp(str, str2)))
269 			return (0);
270 	}
271 	return (-1);
272 }
273 
274 static int
275 vm_alloc_memseg(struct vmctx *ctx, int segid, size_t len, const char *name)
276 {
277 	struct vm_memseg memseg;
278 	size_t n;
279 	int error;
280 
281 	/*
282 	 * If the memory segment has already been created then just return.
283 	 * This is the usual case for the SYSMEM segment created by userspace
284 	 * loaders like bhyveload(8).
285 	 */
286 	error = vm_get_memseg(ctx, segid, &memseg.len, memseg.name,
287 	    sizeof(memseg.name));
288 	if (error)
289 		return (error);
290 
291 	if (memseg.len != 0) {
292 		if (cmpseg(len, name, memseg.len, VM_MEMSEG_NAME(&memseg))) {
293 			errno = EINVAL;
294 			return (-1);
295 		} else {
296 			return (0);
297 		}
298 	}
299 
300 	bzero(&memseg, sizeof(struct vm_memseg));
301 	memseg.segid = segid;
302 	memseg.len = len;
303 	if (name != NULL) {
304 		n = strlcpy(memseg.name, name, sizeof(memseg.name));
305 		if (n >= sizeof(memseg.name)) {
306 			errno = ENAMETOOLONG;
307 			return (-1);
308 		}
309 	}
310 
311 	error = ioctl(ctx->fd, VM_ALLOC_MEMSEG, &memseg);
312 	return (error);
313 }
314 
315 int
316 vm_get_memseg(struct vmctx *ctx, int segid, size_t *lenp, char *namebuf,
317     size_t bufsize)
318 {
319 	struct vm_memseg memseg;
320 	size_t n;
321 	int error;
322 
323 	memseg.segid = segid;
324 	error = ioctl(ctx->fd, VM_GET_MEMSEG, &memseg);
325 	if (error == 0) {
326 		*lenp = memseg.len;
327 		n = strlcpy(namebuf, memseg.name, bufsize);
328 		if (n >= bufsize) {
329 			errno = ENAMETOOLONG;
330 			error = -1;
331 		}
332 	}
333 	return (error);
334 }
335 
336 static int
337 setup_memory_segment(struct vmctx *ctx, vm_paddr_t gpa, size_t len, char *base)
338 {
339 	char *ptr;
340 	int error, flags;
341 
342 	/* Map 'len' bytes starting at 'gpa' in the guest address space */
343 	error = vm_mmap_memseg(ctx, gpa, VM_SYSMEM, gpa, len, PROT_ALL);
344 	if (error)
345 		return (error);
346 
347 	flags = MAP_SHARED | MAP_FIXED;
348 	if ((ctx->memflags & VM_MEM_F_INCORE) == 0)
349 		flags |= MAP_NOCORE;
350 
351 	/* mmap into the process address space on the host */
352 	ptr = mmap(base + gpa, len, PROT_RW, flags, ctx->fd, gpa);
353 	if (ptr == MAP_FAILED)
354 		return (-1);
355 
356 	return (0);
357 }
358 
359 int
360 vm_setup_memory(struct vmctx *ctx, size_t memsize, enum vm_mmap_style vms)
361 {
362 	size_t objsize, len;
363 	vm_paddr_t gpa;
364 	char *baseaddr, *ptr;
365 	int error, flags;
366 
367 	assert(vms == VM_MMAP_ALL);
368 
369 	/*
370 	 * If 'memsize' cannot fit entirely in the 'lowmem' segment then
371 	 * create another 'highmem' segment above 4GB for the remainder.
372 	 */
373 	if (memsize > ctx->lowmem_limit) {
374 		ctx->lowmem = ctx->lowmem_limit;
375 		ctx->highmem = memsize - ctx->lowmem_limit;
376 		objsize = 4*GB + ctx->highmem;
377 	} else {
378 		ctx->lowmem = memsize;
379 		ctx->highmem = 0;
380 		objsize = ctx->lowmem;
381 	}
382 
383 	error = vm_alloc_memseg(ctx, VM_SYSMEM, objsize, NULL);
384 	if (error)
385 		return (error);
386 
387 	/*
388 	 * Stake out a contiguous region covering the guest physical memory
389 	 * and the adjoining guard regions.
390 	 */
391 	len = VM_MMAP_GUARD_SIZE + objsize + VM_MMAP_GUARD_SIZE;
392 	flags = MAP_PRIVATE | MAP_ANON | MAP_NOCORE | MAP_ALIGNED_SUPER;
393 	ptr = mmap(NULL, len, PROT_NONE, flags, -1, 0);
394 	if (ptr == MAP_FAILED)
395 		return (-1);
396 
397 	baseaddr = ptr + VM_MMAP_GUARD_SIZE;
398 	if (ctx->highmem > 0) {
399 		gpa = 4*GB;
400 		len = ctx->highmem;
401 		error = setup_memory_segment(ctx, gpa, len, baseaddr);
402 		if (error)
403 			return (error);
404 	}
405 
406 	if (ctx->lowmem > 0) {
407 		gpa = 0;
408 		len = ctx->lowmem;
409 		error = setup_memory_segment(ctx, gpa, len, baseaddr);
410 		if (error)
411 			return (error);
412 	}
413 
414 	ctx->baseaddr = baseaddr;
415 
416 	return (0);
417 }
418 
419 /*
420  * Returns a non-NULL pointer if [gaddr, gaddr+len) is entirely contained in
421  * the lowmem or highmem regions.
422  *
423  * In particular return NULL if [gaddr, gaddr+len) falls in guest MMIO region.
424  * The instruction emulation code depends on this behavior.
425  */
426 void *
427 vm_map_gpa(struct vmctx *ctx, vm_paddr_t gaddr, size_t len)
428 {
429 
430 	if (ctx->lowmem > 0) {
431 		if (gaddr < ctx->lowmem && len <= ctx->lowmem &&
432 		    gaddr + len <= ctx->lowmem)
433 			return (ctx->baseaddr + gaddr);
434 	}
435 
436 	if (ctx->highmem > 0) {
437                 if (gaddr >= 4*GB) {
438 			if (gaddr < 4*GB + ctx->highmem &&
439 			    len <= ctx->highmem &&
440 			    gaddr + len <= 4*GB + ctx->highmem)
441 				return (ctx->baseaddr + gaddr);
442 		}
443 	}
444 
445 	return (NULL);
446 }
447 
448 size_t
449 vm_get_lowmem_size(struct vmctx *ctx)
450 {
451 
452 	return (ctx->lowmem);
453 }
454 
455 size_t
456 vm_get_highmem_size(struct vmctx *ctx)
457 {
458 
459 	return (ctx->highmem);
460 }
461 
462 void *
463 vm_create_devmem(struct vmctx *ctx, int segid, const char *name, size_t len)
464 {
465 	char pathname[MAXPATHLEN];
466 	size_t len2;
467 	char *base, *ptr;
468 	int fd, error, flags;
469 
470 	fd = -1;
471 	ptr = MAP_FAILED;
472 	if (name == NULL || strlen(name) == 0) {
473 		errno = EINVAL;
474 		goto done;
475 	}
476 
477 	error = vm_alloc_memseg(ctx, segid, len, name);
478 	if (error)
479 		goto done;
480 
481 	strlcpy(pathname, "/dev/vmm.io/", sizeof(pathname));
482 	strlcat(pathname, ctx->name, sizeof(pathname));
483 	strlcat(pathname, ".", sizeof(pathname));
484 	strlcat(pathname, name, sizeof(pathname));
485 
486 	fd = open(pathname, O_RDWR);
487 	if (fd < 0)
488 		goto done;
489 
490 	/*
491 	 * Stake out a contiguous region covering the device memory and the
492 	 * adjoining guard regions.
493 	 */
494 	len2 = VM_MMAP_GUARD_SIZE + len + VM_MMAP_GUARD_SIZE;
495 	flags = MAP_PRIVATE | MAP_ANON | MAP_NOCORE | MAP_ALIGNED_SUPER;
496 	base = mmap(NULL, len2, PROT_NONE, flags, -1, 0);
497 	if (base == MAP_FAILED)
498 		goto done;
499 
500 	flags = MAP_SHARED | MAP_FIXED;
501 	if ((ctx->memflags & VM_MEM_F_INCORE) == 0)
502 		flags |= MAP_NOCORE;
503 
504 	/* mmap the devmem region in the host address space */
505 	ptr = mmap(base + VM_MMAP_GUARD_SIZE, len, PROT_RW, flags, fd, 0);
506 done:
507 	if (fd >= 0)
508 		close(fd);
509 	return (ptr);
510 }
511 
512 int
513 vm_set_desc(struct vmctx *ctx, int vcpu, int reg,
514 	    uint64_t base, uint32_t limit, uint32_t access)
515 {
516 	int error;
517 	struct vm_seg_desc vmsegdesc;
518 
519 	bzero(&vmsegdesc, sizeof(vmsegdesc));
520 	vmsegdesc.cpuid = vcpu;
521 	vmsegdesc.regnum = reg;
522 	vmsegdesc.desc.base = base;
523 	vmsegdesc.desc.limit = limit;
524 	vmsegdesc.desc.access = access;
525 
526 	error = ioctl(ctx->fd, VM_SET_SEGMENT_DESCRIPTOR, &vmsegdesc);
527 	return (error);
528 }
529 
530 int
531 vm_get_desc(struct vmctx *ctx, int vcpu, int reg,
532 	    uint64_t *base, uint32_t *limit, uint32_t *access)
533 {
534 	int error;
535 	struct vm_seg_desc vmsegdesc;
536 
537 	bzero(&vmsegdesc, sizeof(vmsegdesc));
538 	vmsegdesc.cpuid = vcpu;
539 	vmsegdesc.regnum = reg;
540 
541 	error = ioctl(ctx->fd, VM_GET_SEGMENT_DESCRIPTOR, &vmsegdesc);
542 	if (error == 0) {
543 		*base = vmsegdesc.desc.base;
544 		*limit = vmsegdesc.desc.limit;
545 		*access = vmsegdesc.desc.access;
546 	}
547 	return (error);
548 }
549 
550 int
551 vm_get_seg_desc(struct vmctx *ctx, int vcpu, int reg, struct seg_desc *seg_desc)
552 {
553 	int error;
554 
555 	error = vm_get_desc(ctx, vcpu, reg, &seg_desc->base, &seg_desc->limit,
556 	    &seg_desc->access);
557 	return (error);
558 }
559 
560 int
561 vm_set_register(struct vmctx *ctx, int vcpu, int reg, uint64_t val)
562 {
563 	int error;
564 	struct vm_register vmreg;
565 
566 	bzero(&vmreg, sizeof(vmreg));
567 	vmreg.cpuid = vcpu;
568 	vmreg.regnum = reg;
569 	vmreg.regval = val;
570 
571 	error = ioctl(ctx->fd, VM_SET_REGISTER, &vmreg);
572 	return (error);
573 }
574 
575 int
576 vm_get_register(struct vmctx *ctx, int vcpu, int reg, uint64_t *ret_val)
577 {
578 	int error;
579 	struct vm_register vmreg;
580 
581 	bzero(&vmreg, sizeof(vmreg));
582 	vmreg.cpuid = vcpu;
583 	vmreg.regnum = reg;
584 
585 	error = ioctl(ctx->fd, VM_GET_REGISTER, &vmreg);
586 	*ret_val = vmreg.regval;
587 	return (error);
588 }
589 
590 int
591 vm_set_register_set(struct vmctx *ctx, int vcpu, unsigned int count,
592     const int *regnums, uint64_t *regvals)
593 {
594 	int error;
595 	struct vm_register_set vmregset;
596 
597 	bzero(&vmregset, sizeof(vmregset));
598 	vmregset.cpuid = vcpu;
599 	vmregset.count = count;
600 	vmregset.regnums = regnums;
601 	vmregset.regvals = regvals;
602 
603 	error = ioctl(ctx->fd, VM_SET_REGISTER_SET, &vmregset);
604 	return (error);
605 }
606 
607 int
608 vm_get_register_set(struct vmctx *ctx, int vcpu, unsigned int count,
609     const int *regnums, uint64_t *regvals)
610 {
611 	int error;
612 	struct vm_register_set vmregset;
613 
614 	bzero(&vmregset, sizeof(vmregset));
615 	vmregset.cpuid = vcpu;
616 	vmregset.count = count;
617 	vmregset.regnums = regnums;
618 	vmregset.regvals = regvals;
619 
620 	error = ioctl(ctx->fd, VM_GET_REGISTER_SET, &vmregset);
621 	return (error);
622 }
623 
624 int
625 vm_run(struct vmctx *ctx, int vcpu, struct vm_exit *vmexit)
626 {
627 	int error;
628 	struct vm_run vmrun;
629 
630 	bzero(&vmrun, sizeof(vmrun));
631 	vmrun.cpuid = vcpu;
632 
633 	error = ioctl(ctx->fd, VM_RUN, &vmrun);
634 	bcopy(&vmrun.vm_exit, vmexit, sizeof(struct vm_exit));
635 	return (error);
636 }
637 
638 int
639 vm_suspend(struct vmctx *ctx, enum vm_suspend_how how)
640 {
641 	struct vm_suspend vmsuspend;
642 
643 	bzero(&vmsuspend, sizeof(vmsuspend));
644 	vmsuspend.how = how;
645 	return (ioctl(ctx->fd, VM_SUSPEND, &vmsuspend));
646 }
647 
648 int
649 vm_reinit(struct vmctx *ctx)
650 {
651 
652 	return (ioctl(ctx->fd, VM_REINIT, 0));
653 }
654 
655 int
656 vm_inject_exception(struct vmctx *ctx, int vcpu, int vector, int errcode_valid,
657     uint32_t errcode, int restart_instruction)
658 {
659 	struct vm_exception exc;
660 
661 	exc.cpuid = vcpu;
662 	exc.vector = vector;
663 	exc.error_code = errcode;
664 	exc.error_code_valid = errcode_valid;
665 	exc.restart_instruction = restart_instruction;
666 
667 	return (ioctl(ctx->fd, VM_INJECT_EXCEPTION, &exc));
668 }
669 
670 int
671 vm_apicid2vcpu(struct vmctx *ctx, int apicid)
672 {
673 	/*
674 	 * The apic id associated with the 'vcpu' has the same numerical value
675 	 * as the 'vcpu' itself.
676 	 */
677 	return (apicid);
678 }
679 
680 int
681 vm_lapic_irq(struct vmctx *ctx, int vcpu, int vector)
682 {
683 	struct vm_lapic_irq vmirq;
684 
685 	bzero(&vmirq, sizeof(vmirq));
686 	vmirq.cpuid = vcpu;
687 	vmirq.vector = vector;
688 
689 	return (ioctl(ctx->fd, VM_LAPIC_IRQ, &vmirq));
690 }
691 
692 int
693 vm_lapic_local_irq(struct vmctx *ctx, int vcpu, int vector)
694 {
695 	struct vm_lapic_irq vmirq;
696 
697 	bzero(&vmirq, sizeof(vmirq));
698 	vmirq.cpuid = vcpu;
699 	vmirq.vector = vector;
700 
701 	return (ioctl(ctx->fd, VM_LAPIC_LOCAL_IRQ, &vmirq));
702 }
703 
704 int
705 vm_lapic_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg)
706 {
707 	struct vm_lapic_msi vmmsi;
708 
709 	bzero(&vmmsi, sizeof(vmmsi));
710 	vmmsi.addr = addr;
711 	vmmsi.msg = msg;
712 
713 	return (ioctl(ctx->fd, VM_LAPIC_MSI, &vmmsi));
714 }
715 
716 int
717 vm_ioapic_assert_irq(struct vmctx *ctx, int irq)
718 {
719 	struct vm_ioapic_irq ioapic_irq;
720 
721 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
722 	ioapic_irq.irq = irq;
723 
724 	return (ioctl(ctx->fd, VM_IOAPIC_ASSERT_IRQ, &ioapic_irq));
725 }
726 
727 int
728 vm_ioapic_deassert_irq(struct vmctx *ctx, int irq)
729 {
730 	struct vm_ioapic_irq ioapic_irq;
731 
732 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
733 	ioapic_irq.irq = irq;
734 
735 	return (ioctl(ctx->fd, VM_IOAPIC_DEASSERT_IRQ, &ioapic_irq));
736 }
737 
738 int
739 vm_ioapic_pulse_irq(struct vmctx *ctx, int irq)
740 {
741 	struct vm_ioapic_irq ioapic_irq;
742 
743 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
744 	ioapic_irq.irq = irq;
745 
746 	return (ioctl(ctx->fd, VM_IOAPIC_PULSE_IRQ, &ioapic_irq));
747 }
748 
749 int
750 vm_ioapic_pincount(struct vmctx *ctx, int *pincount)
751 {
752 
753 	return (ioctl(ctx->fd, VM_IOAPIC_PINCOUNT, pincount));
754 }
755 
756 int
757 vm_isa_assert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
758 {
759 	struct vm_isa_irq isa_irq;
760 
761 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
762 	isa_irq.atpic_irq = atpic_irq;
763 	isa_irq.ioapic_irq = ioapic_irq;
764 
765 	return (ioctl(ctx->fd, VM_ISA_ASSERT_IRQ, &isa_irq));
766 }
767 
768 int
769 vm_isa_deassert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
770 {
771 	struct vm_isa_irq isa_irq;
772 
773 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
774 	isa_irq.atpic_irq = atpic_irq;
775 	isa_irq.ioapic_irq = ioapic_irq;
776 
777 	return (ioctl(ctx->fd, VM_ISA_DEASSERT_IRQ, &isa_irq));
778 }
779 
780 int
781 vm_isa_pulse_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
782 {
783 	struct vm_isa_irq isa_irq;
784 
785 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
786 	isa_irq.atpic_irq = atpic_irq;
787 	isa_irq.ioapic_irq = ioapic_irq;
788 
789 	return (ioctl(ctx->fd, VM_ISA_PULSE_IRQ, &isa_irq));
790 }
791 
792 int
793 vm_isa_set_irq_trigger(struct vmctx *ctx, int atpic_irq,
794     enum vm_intr_trigger trigger)
795 {
796 	struct vm_isa_irq_trigger isa_irq_trigger;
797 
798 	bzero(&isa_irq_trigger, sizeof(struct vm_isa_irq_trigger));
799 	isa_irq_trigger.atpic_irq = atpic_irq;
800 	isa_irq_trigger.trigger = trigger;
801 
802 	return (ioctl(ctx->fd, VM_ISA_SET_IRQ_TRIGGER, &isa_irq_trigger));
803 }
804 
805 int
806 vm_inject_nmi(struct vmctx *ctx, int vcpu)
807 {
808 	struct vm_nmi vmnmi;
809 
810 	bzero(&vmnmi, sizeof(vmnmi));
811 	vmnmi.cpuid = vcpu;
812 
813 	return (ioctl(ctx->fd, VM_INJECT_NMI, &vmnmi));
814 }
815 
816 static struct {
817 	const char	*name;
818 	int		type;
819 } capstrmap[] = {
820 	{ "hlt_exit",		VM_CAP_HALT_EXIT },
821 	{ "mtrap_exit",		VM_CAP_MTRAP_EXIT },
822 	{ "pause_exit",		VM_CAP_PAUSE_EXIT },
823 	{ "unrestricted_guest",	VM_CAP_UNRESTRICTED_GUEST },
824 	{ "enable_invpcid",	VM_CAP_ENABLE_INVPCID },
825 	{ 0 }
826 };
827 
828 int
829 vm_capability_name2type(const char *capname)
830 {
831 	int i;
832 
833 	for (i = 0; capstrmap[i].name != NULL && capname != NULL; i++) {
834 		if (strcmp(capstrmap[i].name, capname) == 0)
835 			return (capstrmap[i].type);
836 	}
837 
838 	return (-1);
839 }
840 
841 const char *
842 vm_capability_type2name(int type)
843 {
844 	int i;
845 
846 	for (i = 0; capstrmap[i].name != NULL; i++) {
847 		if (capstrmap[i].type == type)
848 			return (capstrmap[i].name);
849 	}
850 
851 	return (NULL);
852 }
853 
854 int
855 vm_get_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap,
856 		  int *retval)
857 {
858 	int error;
859 	struct vm_capability vmcap;
860 
861 	bzero(&vmcap, sizeof(vmcap));
862 	vmcap.cpuid = vcpu;
863 	vmcap.captype = cap;
864 
865 	error = ioctl(ctx->fd, VM_GET_CAPABILITY, &vmcap);
866 	*retval = vmcap.capval;
867 	return (error);
868 }
869 
870 int
871 vm_set_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap, int val)
872 {
873 	struct vm_capability vmcap;
874 
875 	bzero(&vmcap, sizeof(vmcap));
876 	vmcap.cpuid = vcpu;
877 	vmcap.captype = cap;
878 	vmcap.capval = val;
879 
880 	return (ioctl(ctx->fd, VM_SET_CAPABILITY, &vmcap));
881 }
882 
883 int
884 vm_assign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
885 {
886 	struct vm_pptdev pptdev;
887 
888 	bzero(&pptdev, sizeof(pptdev));
889 	pptdev.bus = bus;
890 	pptdev.slot = slot;
891 	pptdev.func = func;
892 
893 	return (ioctl(ctx->fd, VM_BIND_PPTDEV, &pptdev));
894 }
895 
896 int
897 vm_unassign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
898 {
899 	struct vm_pptdev pptdev;
900 
901 	bzero(&pptdev, sizeof(pptdev));
902 	pptdev.bus = bus;
903 	pptdev.slot = slot;
904 	pptdev.func = func;
905 
906 	return (ioctl(ctx->fd, VM_UNBIND_PPTDEV, &pptdev));
907 }
908 
909 int
910 vm_map_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func,
911 		   vm_paddr_t gpa, size_t len, vm_paddr_t hpa)
912 {
913 	struct vm_pptdev_mmio pptmmio;
914 
915 	bzero(&pptmmio, sizeof(pptmmio));
916 	pptmmio.bus = bus;
917 	pptmmio.slot = slot;
918 	pptmmio.func = func;
919 	pptmmio.gpa = gpa;
920 	pptmmio.len = len;
921 	pptmmio.hpa = hpa;
922 
923 	return (ioctl(ctx->fd, VM_MAP_PPTDEV_MMIO, &pptmmio));
924 }
925 
926 int
927 vm_setup_pptdev_msi(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
928     uint64_t addr, uint64_t msg, int numvec)
929 {
930 	struct vm_pptdev_msi pptmsi;
931 
932 	bzero(&pptmsi, sizeof(pptmsi));
933 	pptmsi.vcpu = vcpu;
934 	pptmsi.bus = bus;
935 	pptmsi.slot = slot;
936 	pptmsi.func = func;
937 	pptmsi.msg = msg;
938 	pptmsi.addr = addr;
939 	pptmsi.numvec = numvec;
940 
941 	return (ioctl(ctx->fd, VM_PPTDEV_MSI, &pptmsi));
942 }
943 
944 int
945 vm_setup_pptdev_msix(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
946     int idx, uint64_t addr, uint64_t msg, uint32_t vector_control)
947 {
948 	struct vm_pptdev_msix pptmsix;
949 
950 	bzero(&pptmsix, sizeof(pptmsix));
951 	pptmsix.vcpu = vcpu;
952 	pptmsix.bus = bus;
953 	pptmsix.slot = slot;
954 	pptmsix.func = func;
955 	pptmsix.idx = idx;
956 	pptmsix.msg = msg;
957 	pptmsix.addr = addr;
958 	pptmsix.vector_control = vector_control;
959 
960 	return ioctl(ctx->fd, VM_PPTDEV_MSIX, &pptmsix);
961 }
962 
963 uint64_t *
964 vm_get_stats(struct vmctx *ctx, int vcpu, struct timeval *ret_tv,
965 	     int *ret_entries)
966 {
967 	int error;
968 
969 	static struct vm_stats vmstats;
970 
971 	vmstats.cpuid = vcpu;
972 
973 	error = ioctl(ctx->fd, VM_STATS, &vmstats);
974 	if (error == 0) {
975 		if (ret_entries)
976 			*ret_entries = vmstats.num_entries;
977 		if (ret_tv)
978 			*ret_tv = vmstats.tv;
979 		return (vmstats.statbuf);
980 	} else
981 		return (NULL);
982 }
983 
984 const char *
985 vm_get_stat_desc(struct vmctx *ctx, int index)
986 {
987 	static struct vm_stat_desc statdesc;
988 
989 	statdesc.index = index;
990 	if (ioctl(ctx->fd, VM_STAT_DESC, &statdesc) == 0)
991 		return (statdesc.desc);
992 	else
993 		return (NULL);
994 }
995 
996 int
997 vm_get_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state *state)
998 {
999 	int error;
1000 	struct vm_x2apic x2apic;
1001 
1002 	bzero(&x2apic, sizeof(x2apic));
1003 	x2apic.cpuid = vcpu;
1004 
1005 	error = ioctl(ctx->fd, VM_GET_X2APIC_STATE, &x2apic);
1006 	*state = x2apic.state;
1007 	return (error);
1008 }
1009 
1010 int
1011 vm_set_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state state)
1012 {
1013 	int error;
1014 	struct vm_x2apic x2apic;
1015 
1016 	bzero(&x2apic, sizeof(x2apic));
1017 	x2apic.cpuid = vcpu;
1018 	x2apic.state = state;
1019 
1020 	error = ioctl(ctx->fd, VM_SET_X2APIC_STATE, &x2apic);
1021 
1022 	return (error);
1023 }
1024 
1025 /*
1026  * From Intel Vol 3a:
1027  * Table 9-1. IA-32 Processor States Following Power-up, Reset or INIT
1028  */
1029 int
1030 vcpu_reset(struct vmctx *vmctx, int vcpu)
1031 {
1032 	int error;
1033 	uint64_t rflags, rip, cr0, cr4, zero, desc_base, rdx;
1034 	uint32_t desc_access, desc_limit;
1035 	uint16_t sel;
1036 
1037 	zero = 0;
1038 
1039 	rflags = 0x2;
1040 	error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RFLAGS, rflags);
1041 	if (error)
1042 		goto done;
1043 
1044 	rip = 0xfff0;
1045 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RIP, rip)) != 0)
1046 		goto done;
1047 
1048 	cr0 = CR0_NE;
1049 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR0, cr0)) != 0)
1050 		goto done;
1051 
1052 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR3, zero)) != 0)
1053 		goto done;
1054 
1055 	cr4 = 0;
1056 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR4, cr4)) != 0)
1057 		goto done;
1058 
1059 	/*
1060 	 * CS: present, r/w, accessed, 16-bit, byte granularity, usable
1061 	 */
1062 	desc_base = 0xffff0000;
1063 	desc_limit = 0xffff;
1064 	desc_access = 0x0093;
1065 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_CS,
1066 			    desc_base, desc_limit, desc_access);
1067 	if (error)
1068 		goto done;
1069 
1070 	sel = 0xf000;
1071 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CS, sel)) != 0)
1072 		goto done;
1073 
1074 	/*
1075 	 * SS,DS,ES,FS,GS: present, r/w, accessed, 16-bit, byte granularity
1076 	 */
1077 	desc_base = 0;
1078 	desc_limit = 0xffff;
1079 	desc_access = 0x0093;
1080 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_SS,
1081 			    desc_base, desc_limit, desc_access);
1082 	if (error)
1083 		goto done;
1084 
1085 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_DS,
1086 			    desc_base, desc_limit, desc_access);
1087 	if (error)
1088 		goto done;
1089 
1090 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_ES,
1091 			    desc_base, desc_limit, desc_access);
1092 	if (error)
1093 		goto done;
1094 
1095 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_FS,
1096 			    desc_base, desc_limit, desc_access);
1097 	if (error)
1098 		goto done;
1099 
1100 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GS,
1101 			    desc_base, desc_limit, desc_access);
1102 	if (error)
1103 		goto done;
1104 
1105 	sel = 0;
1106 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_SS, sel)) != 0)
1107 		goto done;
1108 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_DS, sel)) != 0)
1109 		goto done;
1110 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_ES, sel)) != 0)
1111 		goto done;
1112 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_FS, sel)) != 0)
1113 		goto done;
1114 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_GS, sel)) != 0)
1115 		goto done;
1116 
1117 	/* General purpose registers */
1118 	rdx = 0xf00;
1119 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RAX, zero)) != 0)
1120 		goto done;
1121 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBX, zero)) != 0)
1122 		goto done;
1123 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RCX, zero)) != 0)
1124 		goto done;
1125 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDX, rdx)) != 0)
1126 		goto done;
1127 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSI, zero)) != 0)
1128 		goto done;
1129 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDI, zero)) != 0)
1130 		goto done;
1131 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBP, zero)) != 0)
1132 		goto done;
1133 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSP, zero)) != 0)
1134 		goto done;
1135 
1136 	/* GDTR, IDTR */
1137 	desc_base = 0;
1138 	desc_limit = 0xffff;
1139 	desc_access = 0;
1140 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GDTR,
1141 			    desc_base, desc_limit, desc_access);
1142 	if (error != 0)
1143 		goto done;
1144 
1145 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_IDTR,
1146 			    desc_base, desc_limit, desc_access);
1147 	if (error != 0)
1148 		goto done;
1149 
1150 	/* TR */
1151 	desc_base = 0;
1152 	desc_limit = 0xffff;
1153 	desc_access = 0x0000008b;
1154 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_TR, 0, 0, desc_access);
1155 	if (error)
1156 		goto done;
1157 
1158 	sel = 0;
1159 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_TR, sel)) != 0)
1160 		goto done;
1161 
1162 	/* LDTR */
1163 	desc_base = 0;
1164 	desc_limit = 0xffff;
1165 	desc_access = 0x00000082;
1166 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_LDTR, desc_base,
1167 			    desc_limit, desc_access);
1168 	if (error)
1169 		goto done;
1170 
1171 	sel = 0;
1172 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_LDTR, 0)) != 0)
1173 		goto done;
1174 
1175 	/* XXX cr2, debug registers */
1176 
1177 	error = 0;
1178 done:
1179 	return (error);
1180 }
1181 
1182 int
1183 vm_get_gpa_pmap(struct vmctx *ctx, uint64_t gpa, uint64_t *pte, int *num)
1184 {
1185 	int error, i;
1186 	struct vm_gpa_pte gpapte;
1187 
1188 	bzero(&gpapte, sizeof(gpapte));
1189 	gpapte.gpa = gpa;
1190 
1191 	error = ioctl(ctx->fd, VM_GET_GPA_PMAP, &gpapte);
1192 
1193 	if (error == 0) {
1194 		*num = gpapte.ptenum;
1195 		for (i = 0; i < gpapte.ptenum; i++)
1196 			pte[i] = gpapte.pte[i];
1197 	}
1198 
1199 	return (error);
1200 }
1201 
1202 int
1203 vm_get_hpet_capabilities(struct vmctx *ctx, uint32_t *capabilities)
1204 {
1205 	int error;
1206 	struct vm_hpet_cap cap;
1207 
1208 	bzero(&cap, sizeof(struct vm_hpet_cap));
1209 	error = ioctl(ctx->fd, VM_GET_HPET_CAPABILITIES, &cap);
1210 	if (capabilities != NULL)
1211 		*capabilities = cap.capabilities;
1212 	return (error);
1213 }
1214 
1215 int
1216 vm_gla2gpa(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1217     uint64_t gla, int prot, uint64_t *gpa, int *fault)
1218 {
1219 	struct vm_gla2gpa gg;
1220 	int error;
1221 
1222 	bzero(&gg, sizeof(struct vm_gla2gpa));
1223 	gg.vcpuid = vcpu;
1224 	gg.prot = prot;
1225 	gg.gla = gla;
1226 	gg.paging = *paging;
1227 
1228 	error = ioctl(ctx->fd, VM_GLA2GPA, &gg);
1229 	if (error == 0) {
1230 		*fault = gg.fault;
1231 		*gpa = gg.gpa;
1232 	}
1233 	return (error);
1234 }
1235 
1236 int
1237 vm_gla2gpa_nofault(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1238     uint64_t gla, int prot, uint64_t *gpa, int *fault)
1239 {
1240 	struct vm_gla2gpa gg;
1241 	int error;
1242 
1243 	bzero(&gg, sizeof(struct vm_gla2gpa));
1244 	gg.vcpuid = vcpu;
1245 	gg.prot = prot;
1246 	gg.gla = gla;
1247 	gg.paging = *paging;
1248 
1249 	error = ioctl(ctx->fd, VM_GLA2GPA_NOFAULT, &gg);
1250 	if (error == 0) {
1251 		*fault = gg.fault;
1252 		*gpa = gg.gpa;
1253 	}
1254 	return (error);
1255 }
1256 
1257 #ifndef min
1258 #define	min(a,b)	(((a) < (b)) ? (a) : (b))
1259 #endif
1260 
1261 int
1262 vm_copy_setup(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1263     uint64_t gla, size_t len, int prot, struct iovec *iov, int iovcnt,
1264     int *fault)
1265 {
1266 	void *va;
1267 	uint64_t gpa;
1268 	int error, i, n, off;
1269 
1270 	for (i = 0; i < iovcnt; i++) {
1271 		iov[i].iov_base = 0;
1272 		iov[i].iov_len = 0;
1273 	}
1274 
1275 	while (len) {
1276 		assert(iovcnt > 0);
1277 		error = vm_gla2gpa(ctx, vcpu, paging, gla, prot, &gpa, fault);
1278 		if (error || *fault)
1279 			return (error);
1280 
1281 		off = gpa & PAGE_MASK;
1282 		n = min(len, PAGE_SIZE - off);
1283 
1284 		va = vm_map_gpa(ctx, gpa, n);
1285 		if (va == NULL)
1286 			return (EFAULT);
1287 
1288 		iov->iov_base = va;
1289 		iov->iov_len = n;
1290 		iov++;
1291 		iovcnt--;
1292 
1293 		gla += n;
1294 		len -= n;
1295 	}
1296 	return (0);
1297 }
1298 
1299 void
1300 vm_copy_teardown(struct vmctx *ctx, int vcpu, struct iovec *iov, int iovcnt)
1301 {
1302 
1303 	return;
1304 }
1305 
1306 void
1307 vm_copyin(struct vmctx *ctx, int vcpu, struct iovec *iov, void *vp, size_t len)
1308 {
1309 	const char *src;
1310 	char *dst;
1311 	size_t n;
1312 
1313 	dst = vp;
1314 	while (len) {
1315 		assert(iov->iov_len);
1316 		n = min(len, iov->iov_len);
1317 		src = iov->iov_base;
1318 		bcopy(src, dst, n);
1319 
1320 		iov++;
1321 		dst += n;
1322 		len -= n;
1323 	}
1324 }
1325 
1326 void
1327 vm_copyout(struct vmctx *ctx, int vcpu, const void *vp, struct iovec *iov,
1328     size_t len)
1329 {
1330 	const char *src;
1331 	char *dst;
1332 	size_t n;
1333 
1334 	src = vp;
1335 	while (len) {
1336 		assert(iov->iov_len);
1337 		n = min(len, iov->iov_len);
1338 		dst = iov->iov_base;
1339 		bcopy(src, dst, n);
1340 
1341 		iov++;
1342 		src += n;
1343 		len -= n;
1344 	}
1345 }
1346 
1347 static int
1348 vm_get_cpus(struct vmctx *ctx, int which, cpuset_t *cpus)
1349 {
1350 	struct vm_cpuset vm_cpuset;
1351 	int error;
1352 
1353 	bzero(&vm_cpuset, sizeof(struct vm_cpuset));
1354 	vm_cpuset.which = which;
1355 	vm_cpuset.cpusetsize = sizeof(cpuset_t);
1356 	vm_cpuset.cpus = cpus;
1357 
1358 	error = ioctl(ctx->fd, VM_GET_CPUS, &vm_cpuset);
1359 	return (error);
1360 }
1361 
1362 int
1363 vm_active_cpus(struct vmctx *ctx, cpuset_t *cpus)
1364 {
1365 
1366 	return (vm_get_cpus(ctx, VM_ACTIVE_CPUS, cpus));
1367 }
1368 
1369 int
1370 vm_suspended_cpus(struct vmctx *ctx, cpuset_t *cpus)
1371 {
1372 
1373 	return (vm_get_cpus(ctx, VM_SUSPENDED_CPUS, cpus));
1374 }
1375 
1376 int
1377 vm_activate_cpu(struct vmctx *ctx, int vcpu)
1378 {
1379 	struct vm_activate_cpu ac;
1380 	int error;
1381 
1382 	bzero(&ac, sizeof(struct vm_activate_cpu));
1383 	ac.vcpuid = vcpu;
1384 	error = ioctl(ctx->fd, VM_ACTIVATE_CPU, &ac);
1385 	return (error);
1386 }
1387 
1388 int
1389 vm_get_intinfo(struct vmctx *ctx, int vcpu, uint64_t *info1, uint64_t *info2)
1390 {
1391 	struct vm_intinfo vmii;
1392 	int error;
1393 
1394 	bzero(&vmii, sizeof(struct vm_intinfo));
1395 	vmii.vcpuid = vcpu;
1396 	error = ioctl(ctx->fd, VM_GET_INTINFO, &vmii);
1397 	if (error == 0) {
1398 		*info1 = vmii.info1;
1399 		*info2 = vmii.info2;
1400 	}
1401 	return (error);
1402 }
1403 
1404 int
1405 vm_set_intinfo(struct vmctx *ctx, int vcpu, uint64_t info1)
1406 {
1407 	struct vm_intinfo vmii;
1408 	int error;
1409 
1410 	bzero(&vmii, sizeof(struct vm_intinfo));
1411 	vmii.vcpuid = vcpu;
1412 	vmii.info1 = info1;
1413 	error = ioctl(ctx->fd, VM_SET_INTINFO, &vmii);
1414 	return (error);
1415 }
1416 
1417 int
1418 vm_rtc_write(struct vmctx *ctx, int offset, uint8_t value)
1419 {
1420 	struct vm_rtc_data rtcdata;
1421 	int error;
1422 
1423 	bzero(&rtcdata, sizeof(struct vm_rtc_data));
1424 	rtcdata.offset = offset;
1425 	rtcdata.value = value;
1426 	error = ioctl(ctx->fd, VM_RTC_WRITE, &rtcdata);
1427 	return (error);
1428 }
1429 
1430 int
1431 vm_rtc_read(struct vmctx *ctx, int offset, uint8_t *retval)
1432 {
1433 	struct vm_rtc_data rtcdata;
1434 	int error;
1435 
1436 	bzero(&rtcdata, sizeof(struct vm_rtc_data));
1437 	rtcdata.offset = offset;
1438 	error = ioctl(ctx->fd, VM_RTC_READ, &rtcdata);
1439 	if (error == 0)
1440 		*retval = rtcdata.value;
1441 	return (error);
1442 }
1443 
1444 int
1445 vm_rtc_settime(struct vmctx *ctx, time_t secs)
1446 {
1447 	struct vm_rtc_time rtctime;
1448 	int error;
1449 
1450 	bzero(&rtctime, sizeof(struct vm_rtc_time));
1451 	rtctime.secs = secs;
1452 	error = ioctl(ctx->fd, VM_RTC_SETTIME, &rtctime);
1453 	return (error);
1454 }
1455 
1456 int
1457 vm_rtc_gettime(struct vmctx *ctx, time_t *secs)
1458 {
1459 	struct vm_rtc_time rtctime;
1460 	int error;
1461 
1462 	bzero(&rtctime, sizeof(struct vm_rtc_time));
1463 	error = ioctl(ctx->fd, VM_RTC_GETTIME, &rtctime);
1464 	if (error == 0)
1465 		*secs = rtctime.secs;
1466 	return (error);
1467 }
1468 
1469 int
1470 vm_restart_instruction(void *arg, int vcpu)
1471 {
1472 	struct vmctx *ctx = arg;
1473 
1474 	return (ioctl(ctx->fd, VM_RESTART_INSTRUCTION, &vcpu));
1475 }
1476 
1477 int
1478 vm_get_device_fd(struct vmctx *ctx)
1479 {
1480 
1481 	return (ctx->fd);
1482 }
1483 
1484 const cap_ioctl_t *
1485 vm_get_ioctls(size_t *len)
1486 {
1487 	cap_ioctl_t *cmds;
1488 	/* keep in sync with machine/vmm_dev.h */
1489 	static const cap_ioctl_t vm_ioctl_cmds[] = { VM_RUN, VM_SUSPEND, VM_REINIT,
1490 	    VM_ALLOC_MEMSEG, VM_GET_MEMSEG, VM_MMAP_MEMSEG, VM_MMAP_MEMSEG,
1491 	    VM_MMAP_GETNEXT, VM_SET_REGISTER, VM_GET_REGISTER,
1492 	    VM_SET_SEGMENT_DESCRIPTOR, VM_GET_SEGMENT_DESCRIPTOR,
1493 	    VM_SET_REGISTER_SET, VM_GET_REGISTER_SET,
1494 	    VM_INJECT_EXCEPTION, VM_LAPIC_IRQ, VM_LAPIC_LOCAL_IRQ,
1495 	    VM_LAPIC_MSI, VM_IOAPIC_ASSERT_IRQ, VM_IOAPIC_DEASSERT_IRQ,
1496 	    VM_IOAPIC_PULSE_IRQ, VM_IOAPIC_PINCOUNT, VM_ISA_ASSERT_IRQ,
1497 	    VM_ISA_DEASSERT_IRQ, VM_ISA_PULSE_IRQ, VM_ISA_SET_IRQ_TRIGGER,
1498 	    VM_SET_CAPABILITY, VM_GET_CAPABILITY, VM_BIND_PPTDEV,
1499 	    VM_UNBIND_PPTDEV, VM_MAP_PPTDEV_MMIO, VM_PPTDEV_MSI,
1500 	    VM_PPTDEV_MSIX, VM_INJECT_NMI, VM_STATS, VM_STAT_DESC,
1501 	    VM_SET_X2APIC_STATE, VM_GET_X2APIC_STATE,
1502 	    VM_GET_HPET_CAPABILITIES, VM_GET_GPA_PMAP, VM_GLA2GPA,
1503 	    VM_GLA2GPA_NOFAULT,
1504 	    VM_ACTIVATE_CPU, VM_GET_CPUS, VM_SET_INTINFO, VM_GET_INTINFO,
1505 	    VM_RTC_WRITE, VM_RTC_READ, VM_RTC_SETTIME, VM_RTC_GETTIME,
1506 	    VM_RESTART_INSTRUCTION };
1507 
1508 	if (len == NULL) {
1509 		cmds = malloc(sizeof(vm_ioctl_cmds));
1510 		if (cmds == NULL)
1511 			return (NULL);
1512 		bcopy(vm_ioctl_cmds, cmds, sizeof(vm_ioctl_cmds));
1513 		return (cmds);
1514 	}
1515 
1516 	*len = nitems(vm_ioctl_cmds);
1517 	return (NULL);
1518 }
1519 
1520