xref: /freebsd/lib/libvmmapi/vmmapi.c (revision eb69d1f144a6fcc765d1b9d44a5ae8082353e70b)
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_run(struct vmctx *ctx, int vcpu, struct vm_exit *vmexit)
592 {
593 	int error;
594 	struct vm_run vmrun;
595 
596 	bzero(&vmrun, sizeof(vmrun));
597 	vmrun.cpuid = vcpu;
598 
599 	error = ioctl(ctx->fd, VM_RUN, &vmrun);
600 	bcopy(&vmrun.vm_exit, vmexit, sizeof(struct vm_exit));
601 	return (error);
602 }
603 
604 int
605 vm_suspend(struct vmctx *ctx, enum vm_suspend_how how)
606 {
607 	struct vm_suspend vmsuspend;
608 
609 	bzero(&vmsuspend, sizeof(vmsuspend));
610 	vmsuspend.how = how;
611 	return (ioctl(ctx->fd, VM_SUSPEND, &vmsuspend));
612 }
613 
614 int
615 vm_reinit(struct vmctx *ctx)
616 {
617 
618 	return (ioctl(ctx->fd, VM_REINIT, 0));
619 }
620 
621 int
622 vm_inject_exception(struct vmctx *ctx, int vcpu, int vector, int errcode_valid,
623     uint32_t errcode, int restart_instruction)
624 {
625 	struct vm_exception exc;
626 
627 	exc.cpuid = vcpu;
628 	exc.vector = vector;
629 	exc.error_code = errcode;
630 	exc.error_code_valid = errcode_valid;
631 	exc.restart_instruction = restart_instruction;
632 
633 	return (ioctl(ctx->fd, VM_INJECT_EXCEPTION, &exc));
634 }
635 
636 int
637 vm_apicid2vcpu(struct vmctx *ctx, int apicid)
638 {
639 	/*
640 	 * The apic id associated with the 'vcpu' has the same numerical value
641 	 * as the 'vcpu' itself.
642 	 */
643 	return (apicid);
644 }
645 
646 int
647 vm_lapic_irq(struct vmctx *ctx, int vcpu, int vector)
648 {
649 	struct vm_lapic_irq vmirq;
650 
651 	bzero(&vmirq, sizeof(vmirq));
652 	vmirq.cpuid = vcpu;
653 	vmirq.vector = vector;
654 
655 	return (ioctl(ctx->fd, VM_LAPIC_IRQ, &vmirq));
656 }
657 
658 int
659 vm_lapic_local_irq(struct vmctx *ctx, int vcpu, int vector)
660 {
661 	struct vm_lapic_irq vmirq;
662 
663 	bzero(&vmirq, sizeof(vmirq));
664 	vmirq.cpuid = vcpu;
665 	vmirq.vector = vector;
666 
667 	return (ioctl(ctx->fd, VM_LAPIC_LOCAL_IRQ, &vmirq));
668 }
669 
670 int
671 vm_lapic_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg)
672 {
673 	struct vm_lapic_msi vmmsi;
674 
675 	bzero(&vmmsi, sizeof(vmmsi));
676 	vmmsi.addr = addr;
677 	vmmsi.msg = msg;
678 
679 	return (ioctl(ctx->fd, VM_LAPIC_MSI, &vmmsi));
680 }
681 
682 int
683 vm_ioapic_assert_irq(struct vmctx *ctx, int irq)
684 {
685 	struct vm_ioapic_irq ioapic_irq;
686 
687 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
688 	ioapic_irq.irq = irq;
689 
690 	return (ioctl(ctx->fd, VM_IOAPIC_ASSERT_IRQ, &ioapic_irq));
691 }
692 
693 int
694 vm_ioapic_deassert_irq(struct vmctx *ctx, int irq)
695 {
696 	struct vm_ioapic_irq ioapic_irq;
697 
698 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
699 	ioapic_irq.irq = irq;
700 
701 	return (ioctl(ctx->fd, VM_IOAPIC_DEASSERT_IRQ, &ioapic_irq));
702 }
703 
704 int
705 vm_ioapic_pulse_irq(struct vmctx *ctx, int irq)
706 {
707 	struct vm_ioapic_irq ioapic_irq;
708 
709 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
710 	ioapic_irq.irq = irq;
711 
712 	return (ioctl(ctx->fd, VM_IOAPIC_PULSE_IRQ, &ioapic_irq));
713 }
714 
715 int
716 vm_ioapic_pincount(struct vmctx *ctx, int *pincount)
717 {
718 
719 	return (ioctl(ctx->fd, VM_IOAPIC_PINCOUNT, pincount));
720 }
721 
722 int
723 vm_isa_assert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
724 {
725 	struct vm_isa_irq isa_irq;
726 
727 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
728 	isa_irq.atpic_irq = atpic_irq;
729 	isa_irq.ioapic_irq = ioapic_irq;
730 
731 	return (ioctl(ctx->fd, VM_ISA_ASSERT_IRQ, &isa_irq));
732 }
733 
734 int
735 vm_isa_deassert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
736 {
737 	struct vm_isa_irq isa_irq;
738 
739 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
740 	isa_irq.atpic_irq = atpic_irq;
741 	isa_irq.ioapic_irq = ioapic_irq;
742 
743 	return (ioctl(ctx->fd, VM_ISA_DEASSERT_IRQ, &isa_irq));
744 }
745 
746 int
747 vm_isa_pulse_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
748 {
749 	struct vm_isa_irq isa_irq;
750 
751 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
752 	isa_irq.atpic_irq = atpic_irq;
753 	isa_irq.ioapic_irq = ioapic_irq;
754 
755 	return (ioctl(ctx->fd, VM_ISA_PULSE_IRQ, &isa_irq));
756 }
757 
758 int
759 vm_isa_set_irq_trigger(struct vmctx *ctx, int atpic_irq,
760     enum vm_intr_trigger trigger)
761 {
762 	struct vm_isa_irq_trigger isa_irq_trigger;
763 
764 	bzero(&isa_irq_trigger, sizeof(struct vm_isa_irq_trigger));
765 	isa_irq_trigger.atpic_irq = atpic_irq;
766 	isa_irq_trigger.trigger = trigger;
767 
768 	return (ioctl(ctx->fd, VM_ISA_SET_IRQ_TRIGGER, &isa_irq_trigger));
769 }
770 
771 int
772 vm_inject_nmi(struct vmctx *ctx, int vcpu)
773 {
774 	struct vm_nmi vmnmi;
775 
776 	bzero(&vmnmi, sizeof(vmnmi));
777 	vmnmi.cpuid = vcpu;
778 
779 	return (ioctl(ctx->fd, VM_INJECT_NMI, &vmnmi));
780 }
781 
782 static struct {
783 	const char	*name;
784 	int		type;
785 } capstrmap[] = {
786 	{ "hlt_exit",		VM_CAP_HALT_EXIT },
787 	{ "mtrap_exit",		VM_CAP_MTRAP_EXIT },
788 	{ "pause_exit",		VM_CAP_PAUSE_EXIT },
789 	{ "unrestricted_guest",	VM_CAP_UNRESTRICTED_GUEST },
790 	{ "enable_invpcid",	VM_CAP_ENABLE_INVPCID },
791 	{ 0 }
792 };
793 
794 int
795 vm_capability_name2type(const char *capname)
796 {
797 	int i;
798 
799 	for (i = 0; capstrmap[i].name != NULL && capname != NULL; i++) {
800 		if (strcmp(capstrmap[i].name, capname) == 0)
801 			return (capstrmap[i].type);
802 	}
803 
804 	return (-1);
805 }
806 
807 const char *
808 vm_capability_type2name(int type)
809 {
810 	int i;
811 
812 	for (i = 0; capstrmap[i].name != NULL; i++) {
813 		if (capstrmap[i].type == type)
814 			return (capstrmap[i].name);
815 	}
816 
817 	return (NULL);
818 }
819 
820 int
821 vm_get_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap,
822 		  int *retval)
823 {
824 	int error;
825 	struct vm_capability vmcap;
826 
827 	bzero(&vmcap, sizeof(vmcap));
828 	vmcap.cpuid = vcpu;
829 	vmcap.captype = cap;
830 
831 	error = ioctl(ctx->fd, VM_GET_CAPABILITY, &vmcap);
832 	*retval = vmcap.capval;
833 	return (error);
834 }
835 
836 int
837 vm_set_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap, int val)
838 {
839 	struct vm_capability vmcap;
840 
841 	bzero(&vmcap, sizeof(vmcap));
842 	vmcap.cpuid = vcpu;
843 	vmcap.captype = cap;
844 	vmcap.capval = val;
845 
846 	return (ioctl(ctx->fd, VM_SET_CAPABILITY, &vmcap));
847 }
848 
849 int
850 vm_assign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
851 {
852 	struct vm_pptdev pptdev;
853 
854 	bzero(&pptdev, sizeof(pptdev));
855 	pptdev.bus = bus;
856 	pptdev.slot = slot;
857 	pptdev.func = func;
858 
859 	return (ioctl(ctx->fd, VM_BIND_PPTDEV, &pptdev));
860 }
861 
862 int
863 vm_unassign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
864 {
865 	struct vm_pptdev pptdev;
866 
867 	bzero(&pptdev, sizeof(pptdev));
868 	pptdev.bus = bus;
869 	pptdev.slot = slot;
870 	pptdev.func = func;
871 
872 	return (ioctl(ctx->fd, VM_UNBIND_PPTDEV, &pptdev));
873 }
874 
875 int
876 vm_map_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func,
877 		   vm_paddr_t gpa, size_t len, vm_paddr_t hpa)
878 {
879 	struct vm_pptdev_mmio pptmmio;
880 
881 	bzero(&pptmmio, sizeof(pptmmio));
882 	pptmmio.bus = bus;
883 	pptmmio.slot = slot;
884 	pptmmio.func = func;
885 	pptmmio.gpa = gpa;
886 	pptmmio.len = len;
887 	pptmmio.hpa = hpa;
888 
889 	return (ioctl(ctx->fd, VM_MAP_PPTDEV_MMIO, &pptmmio));
890 }
891 
892 int
893 vm_setup_pptdev_msi(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
894     uint64_t addr, uint64_t msg, int numvec)
895 {
896 	struct vm_pptdev_msi pptmsi;
897 
898 	bzero(&pptmsi, sizeof(pptmsi));
899 	pptmsi.vcpu = vcpu;
900 	pptmsi.bus = bus;
901 	pptmsi.slot = slot;
902 	pptmsi.func = func;
903 	pptmsi.msg = msg;
904 	pptmsi.addr = addr;
905 	pptmsi.numvec = numvec;
906 
907 	return (ioctl(ctx->fd, VM_PPTDEV_MSI, &pptmsi));
908 }
909 
910 int
911 vm_setup_pptdev_msix(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
912     int idx, uint64_t addr, uint64_t msg, uint32_t vector_control)
913 {
914 	struct vm_pptdev_msix pptmsix;
915 
916 	bzero(&pptmsix, sizeof(pptmsix));
917 	pptmsix.vcpu = vcpu;
918 	pptmsix.bus = bus;
919 	pptmsix.slot = slot;
920 	pptmsix.func = func;
921 	pptmsix.idx = idx;
922 	pptmsix.msg = msg;
923 	pptmsix.addr = addr;
924 	pptmsix.vector_control = vector_control;
925 
926 	return ioctl(ctx->fd, VM_PPTDEV_MSIX, &pptmsix);
927 }
928 
929 uint64_t *
930 vm_get_stats(struct vmctx *ctx, int vcpu, struct timeval *ret_tv,
931 	     int *ret_entries)
932 {
933 	int error;
934 
935 	static struct vm_stats vmstats;
936 
937 	vmstats.cpuid = vcpu;
938 
939 	error = ioctl(ctx->fd, VM_STATS, &vmstats);
940 	if (error == 0) {
941 		if (ret_entries)
942 			*ret_entries = vmstats.num_entries;
943 		if (ret_tv)
944 			*ret_tv = vmstats.tv;
945 		return (vmstats.statbuf);
946 	} else
947 		return (NULL);
948 }
949 
950 const char *
951 vm_get_stat_desc(struct vmctx *ctx, int index)
952 {
953 	static struct vm_stat_desc statdesc;
954 
955 	statdesc.index = index;
956 	if (ioctl(ctx->fd, VM_STAT_DESC, &statdesc) == 0)
957 		return (statdesc.desc);
958 	else
959 		return (NULL);
960 }
961 
962 int
963 vm_get_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state *state)
964 {
965 	int error;
966 	struct vm_x2apic x2apic;
967 
968 	bzero(&x2apic, sizeof(x2apic));
969 	x2apic.cpuid = vcpu;
970 
971 	error = ioctl(ctx->fd, VM_GET_X2APIC_STATE, &x2apic);
972 	*state = x2apic.state;
973 	return (error);
974 }
975 
976 int
977 vm_set_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state state)
978 {
979 	int error;
980 	struct vm_x2apic x2apic;
981 
982 	bzero(&x2apic, sizeof(x2apic));
983 	x2apic.cpuid = vcpu;
984 	x2apic.state = state;
985 
986 	error = ioctl(ctx->fd, VM_SET_X2APIC_STATE, &x2apic);
987 
988 	return (error);
989 }
990 
991 /*
992  * From Intel Vol 3a:
993  * Table 9-1. IA-32 Processor States Following Power-up, Reset or INIT
994  */
995 int
996 vcpu_reset(struct vmctx *vmctx, int vcpu)
997 {
998 	int error;
999 	uint64_t rflags, rip, cr0, cr4, zero, desc_base, rdx;
1000 	uint32_t desc_access, desc_limit;
1001 	uint16_t sel;
1002 
1003 	zero = 0;
1004 
1005 	rflags = 0x2;
1006 	error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RFLAGS, rflags);
1007 	if (error)
1008 		goto done;
1009 
1010 	rip = 0xfff0;
1011 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RIP, rip)) != 0)
1012 		goto done;
1013 
1014 	cr0 = CR0_NE;
1015 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR0, cr0)) != 0)
1016 		goto done;
1017 
1018 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR3, zero)) != 0)
1019 		goto done;
1020 
1021 	cr4 = 0;
1022 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR4, cr4)) != 0)
1023 		goto done;
1024 
1025 	/*
1026 	 * CS: present, r/w, accessed, 16-bit, byte granularity, usable
1027 	 */
1028 	desc_base = 0xffff0000;
1029 	desc_limit = 0xffff;
1030 	desc_access = 0x0093;
1031 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_CS,
1032 			    desc_base, desc_limit, desc_access);
1033 	if (error)
1034 		goto done;
1035 
1036 	sel = 0xf000;
1037 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CS, sel)) != 0)
1038 		goto done;
1039 
1040 	/*
1041 	 * SS,DS,ES,FS,GS: present, r/w, accessed, 16-bit, byte granularity
1042 	 */
1043 	desc_base = 0;
1044 	desc_limit = 0xffff;
1045 	desc_access = 0x0093;
1046 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_SS,
1047 			    desc_base, desc_limit, desc_access);
1048 	if (error)
1049 		goto done;
1050 
1051 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_DS,
1052 			    desc_base, desc_limit, desc_access);
1053 	if (error)
1054 		goto done;
1055 
1056 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_ES,
1057 			    desc_base, desc_limit, desc_access);
1058 	if (error)
1059 		goto done;
1060 
1061 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_FS,
1062 			    desc_base, desc_limit, desc_access);
1063 	if (error)
1064 		goto done;
1065 
1066 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GS,
1067 			    desc_base, desc_limit, desc_access);
1068 	if (error)
1069 		goto done;
1070 
1071 	sel = 0;
1072 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_SS, sel)) != 0)
1073 		goto done;
1074 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_DS, sel)) != 0)
1075 		goto done;
1076 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_ES, sel)) != 0)
1077 		goto done;
1078 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_FS, sel)) != 0)
1079 		goto done;
1080 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_GS, sel)) != 0)
1081 		goto done;
1082 
1083 	/* General purpose registers */
1084 	rdx = 0xf00;
1085 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RAX, zero)) != 0)
1086 		goto done;
1087 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBX, zero)) != 0)
1088 		goto done;
1089 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RCX, zero)) != 0)
1090 		goto done;
1091 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDX, rdx)) != 0)
1092 		goto done;
1093 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSI, zero)) != 0)
1094 		goto done;
1095 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDI, zero)) != 0)
1096 		goto done;
1097 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBP, zero)) != 0)
1098 		goto done;
1099 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSP, zero)) != 0)
1100 		goto done;
1101 
1102 	/* GDTR, IDTR */
1103 	desc_base = 0;
1104 	desc_limit = 0xffff;
1105 	desc_access = 0;
1106 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GDTR,
1107 			    desc_base, desc_limit, desc_access);
1108 	if (error != 0)
1109 		goto done;
1110 
1111 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_IDTR,
1112 			    desc_base, desc_limit, desc_access);
1113 	if (error != 0)
1114 		goto done;
1115 
1116 	/* TR */
1117 	desc_base = 0;
1118 	desc_limit = 0xffff;
1119 	desc_access = 0x0000008b;
1120 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_TR, 0, 0, desc_access);
1121 	if (error)
1122 		goto done;
1123 
1124 	sel = 0;
1125 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_TR, sel)) != 0)
1126 		goto done;
1127 
1128 	/* LDTR */
1129 	desc_base = 0;
1130 	desc_limit = 0xffff;
1131 	desc_access = 0x00000082;
1132 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_LDTR, desc_base,
1133 			    desc_limit, desc_access);
1134 	if (error)
1135 		goto done;
1136 
1137 	sel = 0;
1138 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_LDTR, 0)) != 0)
1139 		goto done;
1140 
1141 	/* XXX cr2, debug registers */
1142 
1143 	error = 0;
1144 done:
1145 	return (error);
1146 }
1147 
1148 int
1149 vm_get_gpa_pmap(struct vmctx *ctx, uint64_t gpa, uint64_t *pte, int *num)
1150 {
1151 	int error, i;
1152 	struct vm_gpa_pte gpapte;
1153 
1154 	bzero(&gpapte, sizeof(gpapte));
1155 	gpapte.gpa = gpa;
1156 
1157 	error = ioctl(ctx->fd, VM_GET_GPA_PMAP, &gpapte);
1158 
1159 	if (error == 0) {
1160 		*num = gpapte.ptenum;
1161 		for (i = 0; i < gpapte.ptenum; i++)
1162 			pte[i] = gpapte.pte[i];
1163 	}
1164 
1165 	return (error);
1166 }
1167 
1168 int
1169 vm_get_hpet_capabilities(struct vmctx *ctx, uint32_t *capabilities)
1170 {
1171 	int error;
1172 	struct vm_hpet_cap cap;
1173 
1174 	bzero(&cap, sizeof(struct vm_hpet_cap));
1175 	error = ioctl(ctx->fd, VM_GET_HPET_CAPABILITIES, &cap);
1176 	if (capabilities != NULL)
1177 		*capabilities = cap.capabilities;
1178 	return (error);
1179 }
1180 
1181 int
1182 vm_gla2gpa(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1183     uint64_t gla, int prot, uint64_t *gpa, int *fault)
1184 {
1185 	struct vm_gla2gpa gg;
1186 	int error;
1187 
1188 	bzero(&gg, sizeof(struct vm_gla2gpa));
1189 	gg.vcpuid = vcpu;
1190 	gg.prot = prot;
1191 	gg.gla = gla;
1192 	gg.paging = *paging;
1193 
1194 	error = ioctl(ctx->fd, VM_GLA2GPA, &gg);
1195 	if (error == 0) {
1196 		*fault = gg.fault;
1197 		*gpa = gg.gpa;
1198 	}
1199 	return (error);
1200 }
1201 
1202 #ifndef min
1203 #define	min(a,b)	(((a) < (b)) ? (a) : (b))
1204 #endif
1205 
1206 int
1207 vm_copy_setup(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1208     uint64_t gla, size_t len, int prot, struct iovec *iov, int iovcnt,
1209     int *fault)
1210 {
1211 	void *va;
1212 	uint64_t gpa;
1213 	int error, i, n, off;
1214 
1215 	for (i = 0; i < iovcnt; i++) {
1216 		iov[i].iov_base = 0;
1217 		iov[i].iov_len = 0;
1218 	}
1219 
1220 	while (len) {
1221 		assert(iovcnt > 0);
1222 		error = vm_gla2gpa(ctx, vcpu, paging, gla, prot, &gpa, fault);
1223 		if (error || *fault)
1224 			return (error);
1225 
1226 		off = gpa & PAGE_MASK;
1227 		n = min(len, PAGE_SIZE - off);
1228 
1229 		va = vm_map_gpa(ctx, gpa, n);
1230 		if (va == NULL)
1231 			return (EFAULT);
1232 
1233 		iov->iov_base = va;
1234 		iov->iov_len = n;
1235 		iov++;
1236 		iovcnt--;
1237 
1238 		gla += n;
1239 		len -= n;
1240 	}
1241 	return (0);
1242 }
1243 
1244 void
1245 vm_copy_teardown(struct vmctx *ctx, int vcpu, struct iovec *iov, int iovcnt)
1246 {
1247 
1248 	return;
1249 }
1250 
1251 void
1252 vm_copyin(struct vmctx *ctx, int vcpu, struct iovec *iov, void *vp, size_t len)
1253 {
1254 	const char *src;
1255 	char *dst;
1256 	size_t n;
1257 
1258 	dst = vp;
1259 	while (len) {
1260 		assert(iov->iov_len);
1261 		n = min(len, iov->iov_len);
1262 		src = iov->iov_base;
1263 		bcopy(src, dst, n);
1264 
1265 		iov++;
1266 		dst += n;
1267 		len -= n;
1268 	}
1269 }
1270 
1271 void
1272 vm_copyout(struct vmctx *ctx, int vcpu, const void *vp, struct iovec *iov,
1273     size_t len)
1274 {
1275 	const char *src;
1276 	char *dst;
1277 	size_t n;
1278 
1279 	src = vp;
1280 	while (len) {
1281 		assert(iov->iov_len);
1282 		n = min(len, iov->iov_len);
1283 		dst = iov->iov_base;
1284 		bcopy(src, dst, n);
1285 
1286 		iov++;
1287 		src += n;
1288 		len -= n;
1289 	}
1290 }
1291 
1292 static int
1293 vm_get_cpus(struct vmctx *ctx, int which, cpuset_t *cpus)
1294 {
1295 	struct vm_cpuset vm_cpuset;
1296 	int error;
1297 
1298 	bzero(&vm_cpuset, sizeof(struct vm_cpuset));
1299 	vm_cpuset.which = which;
1300 	vm_cpuset.cpusetsize = sizeof(cpuset_t);
1301 	vm_cpuset.cpus = cpus;
1302 
1303 	error = ioctl(ctx->fd, VM_GET_CPUS, &vm_cpuset);
1304 	return (error);
1305 }
1306 
1307 int
1308 vm_active_cpus(struct vmctx *ctx, cpuset_t *cpus)
1309 {
1310 
1311 	return (vm_get_cpus(ctx, VM_ACTIVE_CPUS, cpus));
1312 }
1313 
1314 int
1315 vm_suspended_cpus(struct vmctx *ctx, cpuset_t *cpus)
1316 {
1317 
1318 	return (vm_get_cpus(ctx, VM_SUSPENDED_CPUS, cpus));
1319 }
1320 
1321 int
1322 vm_activate_cpu(struct vmctx *ctx, int vcpu)
1323 {
1324 	struct vm_activate_cpu ac;
1325 	int error;
1326 
1327 	bzero(&ac, sizeof(struct vm_activate_cpu));
1328 	ac.vcpuid = vcpu;
1329 	error = ioctl(ctx->fd, VM_ACTIVATE_CPU, &ac);
1330 	return (error);
1331 }
1332 
1333 int
1334 vm_get_intinfo(struct vmctx *ctx, int vcpu, uint64_t *info1, uint64_t *info2)
1335 {
1336 	struct vm_intinfo vmii;
1337 	int error;
1338 
1339 	bzero(&vmii, sizeof(struct vm_intinfo));
1340 	vmii.vcpuid = vcpu;
1341 	error = ioctl(ctx->fd, VM_GET_INTINFO, &vmii);
1342 	if (error == 0) {
1343 		*info1 = vmii.info1;
1344 		*info2 = vmii.info2;
1345 	}
1346 	return (error);
1347 }
1348 
1349 int
1350 vm_set_intinfo(struct vmctx *ctx, int vcpu, uint64_t info1)
1351 {
1352 	struct vm_intinfo vmii;
1353 	int error;
1354 
1355 	bzero(&vmii, sizeof(struct vm_intinfo));
1356 	vmii.vcpuid = vcpu;
1357 	vmii.info1 = info1;
1358 	error = ioctl(ctx->fd, VM_SET_INTINFO, &vmii);
1359 	return (error);
1360 }
1361 
1362 int
1363 vm_rtc_write(struct vmctx *ctx, int offset, uint8_t value)
1364 {
1365 	struct vm_rtc_data rtcdata;
1366 	int error;
1367 
1368 	bzero(&rtcdata, sizeof(struct vm_rtc_data));
1369 	rtcdata.offset = offset;
1370 	rtcdata.value = value;
1371 	error = ioctl(ctx->fd, VM_RTC_WRITE, &rtcdata);
1372 	return (error);
1373 }
1374 
1375 int
1376 vm_rtc_read(struct vmctx *ctx, int offset, uint8_t *retval)
1377 {
1378 	struct vm_rtc_data rtcdata;
1379 	int error;
1380 
1381 	bzero(&rtcdata, sizeof(struct vm_rtc_data));
1382 	rtcdata.offset = offset;
1383 	error = ioctl(ctx->fd, VM_RTC_READ, &rtcdata);
1384 	if (error == 0)
1385 		*retval = rtcdata.value;
1386 	return (error);
1387 }
1388 
1389 int
1390 vm_rtc_settime(struct vmctx *ctx, time_t secs)
1391 {
1392 	struct vm_rtc_time rtctime;
1393 	int error;
1394 
1395 	bzero(&rtctime, sizeof(struct vm_rtc_time));
1396 	rtctime.secs = secs;
1397 	error = ioctl(ctx->fd, VM_RTC_SETTIME, &rtctime);
1398 	return (error);
1399 }
1400 
1401 int
1402 vm_rtc_gettime(struct vmctx *ctx, time_t *secs)
1403 {
1404 	struct vm_rtc_time rtctime;
1405 	int error;
1406 
1407 	bzero(&rtctime, sizeof(struct vm_rtc_time));
1408 	error = ioctl(ctx->fd, VM_RTC_GETTIME, &rtctime);
1409 	if (error == 0)
1410 		*secs = rtctime.secs;
1411 	return (error);
1412 }
1413 
1414 int
1415 vm_restart_instruction(void *arg, int vcpu)
1416 {
1417 	struct vmctx *ctx = arg;
1418 
1419 	return (ioctl(ctx->fd, VM_RESTART_INSTRUCTION, &vcpu));
1420 }
1421 
1422 int
1423 vm_get_device_fd(struct vmctx *ctx)
1424 {
1425 
1426 	return (ctx->fd);
1427 }
1428 
1429 const cap_ioctl_t *
1430 vm_get_ioctls(size_t *len)
1431 {
1432 	cap_ioctl_t *cmds;
1433 	/* keep in sync with machine/vmm_dev.h */
1434 	static const cap_ioctl_t vm_ioctl_cmds[] = { VM_RUN, VM_SUSPEND, VM_REINIT,
1435 	    VM_ALLOC_MEMSEG, VM_GET_MEMSEG, VM_MMAP_MEMSEG, VM_MMAP_MEMSEG,
1436 	    VM_MMAP_GETNEXT, VM_SET_REGISTER, VM_GET_REGISTER,
1437 	    VM_SET_SEGMENT_DESCRIPTOR, VM_GET_SEGMENT_DESCRIPTOR,
1438 	    VM_INJECT_EXCEPTION, VM_LAPIC_IRQ, VM_LAPIC_LOCAL_IRQ,
1439 	    VM_LAPIC_MSI, VM_IOAPIC_ASSERT_IRQ, VM_IOAPIC_DEASSERT_IRQ,
1440 	    VM_IOAPIC_PULSE_IRQ, VM_IOAPIC_PINCOUNT, VM_ISA_ASSERT_IRQ,
1441 	    VM_ISA_DEASSERT_IRQ, VM_ISA_PULSE_IRQ, VM_ISA_SET_IRQ_TRIGGER,
1442 	    VM_SET_CAPABILITY, VM_GET_CAPABILITY, VM_BIND_PPTDEV,
1443 	    VM_UNBIND_PPTDEV, VM_MAP_PPTDEV_MMIO, VM_PPTDEV_MSI,
1444 	    VM_PPTDEV_MSIX, VM_INJECT_NMI, VM_STATS, VM_STAT_DESC,
1445 	    VM_SET_X2APIC_STATE, VM_GET_X2APIC_STATE,
1446 	    VM_GET_HPET_CAPABILITIES, VM_GET_GPA_PMAP, VM_GLA2GPA,
1447 	    VM_ACTIVATE_CPU, VM_GET_CPUS, VM_SET_INTINFO, VM_GET_INTINFO,
1448 	    VM_RTC_WRITE, VM_RTC_READ, VM_RTC_SETTIME, VM_RTC_GETTIME,
1449 	    VM_RESTART_INSTRUCTION };
1450 
1451 	if (len == NULL) {
1452 		cmds = malloc(sizeof(vm_ioctl_cmds));
1453 		if (cmds == NULL)
1454 			return (NULL);
1455 		bcopy(vm_ioctl_cmds, cmds, sizeof(vm_ioctl_cmds));
1456 		return (cmds);
1457 	}
1458 
1459 	*len = nitems(vm_ioctl_cmds);
1460 	return (NULL);
1461 }
1462 
1463