xref: /freebsd/lib/libvmmapi/vmmapi.c (revision a7623790fb345e6dc986dfd31df0ace115e6f2e4)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 2011 NetApp, Inc.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  *
28  * $FreeBSD$
29  */
30 
31 #include <sys/cdefs.h>
32 __FBSDID("$FreeBSD$");
33 
34 #include <sys/param.h>
35 #include <sys/sysctl.h>
36 #include <sys/ioctl.h>
37 #include <sys/linker.h>
38 #include <sys/mman.h>
39 #include <sys/module.h>
40 #include <sys/_iovec.h>
41 #include <sys/cpuset.h>
42 
43 #include <x86/segments.h>
44 #include <machine/specialreg.h>
45 
46 #include <errno.h>
47 #include <stdbool.h>
48 #include <stdio.h>
49 #include <stdlib.h>
50 #include <assert.h>
51 #include <string.h>
52 #include <fcntl.h>
53 #include <unistd.h>
54 
55 #include <libutil.h>
56 
57 #include <vm/vm.h>
58 #include <machine/vmm.h>
59 #include <machine/vmm_dev.h>
60 #include <machine/vmm_snapshot.h>
61 
62 #include "vmmapi.h"
63 
64 #define	MB	(1024 * 1024UL)
65 #define	GB	(1024 * 1024 * 1024UL)
66 
67 /*
68  * Size of the guard region before and after the virtual address space
69  * mapping the guest physical memory. This must be a multiple of the
70  * superpage size for performance reasons.
71  */
72 #define	VM_MMAP_GUARD_SIZE	(4 * MB)
73 
74 #define	PROT_RW		(PROT_READ | PROT_WRITE)
75 #define	PROT_ALL	(PROT_READ | PROT_WRITE | PROT_EXEC)
76 
77 struct vmctx {
78 	int	fd;
79 	uint32_t lowmem_limit;
80 	int	memflags;
81 	size_t	lowmem;
82 	size_t	highmem;
83 	char	*baseaddr;
84 	char	*name;
85 };
86 
87 #define	CREATE(x)  sysctlbyname("hw.vmm.create", NULL, NULL, (x), strlen((x)))
88 #define	DESTROY(x) sysctlbyname("hw.vmm.destroy", NULL, NULL, (x), strlen((x)))
89 
90 static int
91 vm_device_open(const char *name)
92 {
93 	int fd, len;
94 	char *vmfile;
95 
96 	len = strlen("/dev/vmm/") + strlen(name) + 1;
97 	vmfile = malloc(len);
98 	assert(vmfile != NULL);
99 	snprintf(vmfile, len, "/dev/vmm/%s", name);
100 
101 	/* Open the device file */
102 	fd = open(vmfile, O_RDWR, 0);
103 
104 	free(vmfile);
105 	return (fd);
106 }
107 
108 int
109 vm_create(const char *name)
110 {
111 	/* Try to load vmm(4) module before creating a guest. */
112 	if (modfind("vmm") < 0)
113 		kldload("vmm");
114 	return (CREATE((char *)name));
115 }
116 
117 struct vmctx *
118 vm_open(const char *name)
119 {
120 	struct vmctx *vm;
121 
122 	vm = malloc(sizeof(struct vmctx) + strlen(name) + 1);
123 	assert(vm != NULL);
124 
125 	vm->fd = -1;
126 	vm->memflags = 0;
127 	vm->lowmem_limit = 3 * GB;
128 	vm->name = (char *)(vm + 1);
129 	strcpy(vm->name, name);
130 
131 	if ((vm->fd = vm_device_open(vm->name)) < 0)
132 		goto err;
133 
134 	return (vm);
135 err:
136 	vm_destroy(vm);
137 	return (NULL);
138 }
139 
140 void
141 vm_destroy(struct vmctx *vm)
142 {
143 	assert(vm != NULL);
144 
145 	if (vm->fd >= 0)
146 		close(vm->fd);
147 	DESTROY(vm->name);
148 
149 	free(vm);
150 }
151 
152 int
153 vm_parse_memsize(const char *optarg, size_t *ret_memsize)
154 {
155 	char *endptr;
156 	size_t optval;
157 	int error;
158 
159 	optval = strtoul(optarg, &endptr, 0);
160 	if (*optarg != '\0' && *endptr == '\0') {
161 		/*
162 		 * For the sake of backward compatibility if the memory size
163 		 * specified on the command line is less than a megabyte then
164 		 * it is interpreted as being in units of MB.
165 		 */
166 		if (optval < MB)
167 			optval *= MB;
168 		*ret_memsize = optval;
169 		error = 0;
170 	} else
171 		error = expand_number(optarg, ret_memsize);
172 
173 	return (error);
174 }
175 
176 uint32_t
177 vm_get_lowmem_limit(struct vmctx *ctx)
178 {
179 
180 	return (ctx->lowmem_limit);
181 }
182 
183 void
184 vm_set_lowmem_limit(struct vmctx *ctx, uint32_t limit)
185 {
186 
187 	ctx->lowmem_limit = limit;
188 }
189 
190 void
191 vm_set_memflags(struct vmctx *ctx, int flags)
192 {
193 
194 	ctx->memflags = flags;
195 }
196 
197 int
198 vm_get_memflags(struct vmctx *ctx)
199 {
200 
201 	return (ctx->memflags);
202 }
203 
204 /*
205  * Map segment 'segid' starting at 'off' into guest address range [gpa,gpa+len).
206  */
207 int
208 vm_mmap_memseg(struct vmctx *ctx, vm_paddr_t gpa, int segid, vm_ooffset_t off,
209     size_t len, int prot)
210 {
211 	struct vm_memmap memmap;
212 	int error, flags;
213 
214 	memmap.gpa = gpa;
215 	memmap.segid = segid;
216 	memmap.segoff = off;
217 	memmap.len = len;
218 	memmap.prot = prot;
219 	memmap.flags = 0;
220 
221 	if (ctx->memflags & VM_MEM_F_WIRED)
222 		memmap.flags |= VM_MEMMAP_F_WIRED;
223 
224 	/*
225 	 * If this mapping already exists then don't create it again. This
226 	 * is the common case for SYSMEM mappings created by bhyveload(8).
227 	 */
228 	error = vm_mmap_getnext(ctx, &gpa, &segid, &off, &len, &prot, &flags);
229 	if (error == 0 && gpa == memmap.gpa) {
230 		if (segid != memmap.segid || off != memmap.segoff ||
231 		    prot != memmap.prot || flags != memmap.flags) {
232 			errno = EEXIST;
233 			return (-1);
234 		} else {
235 			return (0);
236 		}
237 	}
238 
239 	error = ioctl(ctx->fd, VM_MMAP_MEMSEG, &memmap);
240 	return (error);
241 }
242 
243 int
244 vm_get_guestmem_from_ctx(struct vmctx *ctx, char **guest_baseaddr,
245     size_t *lowmem_size, size_t *highmem_size)
246 {
247 
248 	*guest_baseaddr = ctx->baseaddr;
249 	*lowmem_size = ctx->lowmem;
250 	*highmem_size = ctx->highmem;
251 	return (0);
252 }
253 
254 int
255 vm_mmap_getnext(struct vmctx *ctx, vm_paddr_t *gpa, int *segid,
256     vm_ooffset_t *segoff, size_t *len, int *prot, int *flags)
257 {
258 	struct vm_memmap memmap;
259 	int error;
260 
261 	bzero(&memmap, sizeof(struct vm_memmap));
262 	memmap.gpa = *gpa;
263 	error = ioctl(ctx->fd, VM_MMAP_GETNEXT, &memmap);
264 	if (error == 0) {
265 		*gpa = memmap.gpa;
266 		*segid = memmap.segid;
267 		*segoff = memmap.segoff;
268 		*len = memmap.len;
269 		*prot = memmap.prot;
270 		*flags = memmap.flags;
271 	}
272 	return (error);
273 }
274 
275 /*
276  * Return 0 if the segments are identical and non-zero otherwise.
277  *
278  * This is slightly complicated by the fact that only device memory segments
279  * are named.
280  */
281 static int
282 cmpseg(size_t len, const char *str, size_t len2, const char *str2)
283 {
284 
285 	if (len == len2) {
286 		if ((!str && !str2) || (str && str2 && !strcmp(str, str2)))
287 			return (0);
288 	}
289 	return (-1);
290 }
291 
292 static int
293 vm_alloc_memseg(struct vmctx *ctx, int segid, size_t len, const char *name)
294 {
295 	struct vm_memseg memseg;
296 	size_t n;
297 	int error;
298 
299 	/*
300 	 * If the memory segment has already been created then just return.
301 	 * This is the usual case for the SYSMEM segment created by userspace
302 	 * loaders like bhyveload(8).
303 	 */
304 	error = vm_get_memseg(ctx, segid, &memseg.len, memseg.name,
305 	    sizeof(memseg.name));
306 	if (error)
307 		return (error);
308 
309 	if (memseg.len != 0) {
310 		if (cmpseg(len, name, memseg.len, VM_MEMSEG_NAME(&memseg))) {
311 			errno = EINVAL;
312 			return (-1);
313 		} else {
314 			return (0);
315 		}
316 	}
317 
318 	bzero(&memseg, sizeof(struct vm_memseg));
319 	memseg.segid = segid;
320 	memseg.len = len;
321 	if (name != NULL) {
322 		n = strlcpy(memseg.name, name, sizeof(memseg.name));
323 		if (n >= sizeof(memseg.name)) {
324 			errno = ENAMETOOLONG;
325 			return (-1);
326 		}
327 	}
328 
329 	error = ioctl(ctx->fd, VM_ALLOC_MEMSEG, &memseg);
330 	return (error);
331 }
332 
333 int
334 vm_get_memseg(struct vmctx *ctx, int segid, size_t *lenp, char *namebuf,
335     size_t bufsize)
336 {
337 	struct vm_memseg memseg;
338 	size_t n;
339 	int error;
340 
341 	memseg.segid = segid;
342 	error = ioctl(ctx->fd, VM_GET_MEMSEG, &memseg);
343 	if (error == 0) {
344 		*lenp = memseg.len;
345 		n = strlcpy(namebuf, memseg.name, bufsize);
346 		if (n >= bufsize) {
347 			errno = ENAMETOOLONG;
348 			error = -1;
349 		}
350 	}
351 	return (error);
352 }
353 
354 static int
355 setup_memory_segment(struct vmctx *ctx, vm_paddr_t gpa, size_t len, char *base)
356 {
357 	char *ptr;
358 	int error, flags;
359 
360 	/* Map 'len' bytes starting at 'gpa' in the guest address space */
361 	error = vm_mmap_memseg(ctx, gpa, VM_SYSMEM, gpa, len, PROT_ALL);
362 	if (error)
363 		return (error);
364 
365 	flags = MAP_SHARED | MAP_FIXED;
366 	if ((ctx->memflags & VM_MEM_F_INCORE) == 0)
367 		flags |= MAP_NOCORE;
368 
369 	/* mmap into the process address space on the host */
370 	ptr = mmap(base + gpa, len, PROT_RW, flags, ctx->fd, gpa);
371 	if (ptr == MAP_FAILED)
372 		return (-1);
373 
374 	return (0);
375 }
376 
377 int
378 vm_setup_memory(struct vmctx *ctx, size_t memsize, enum vm_mmap_style vms)
379 {
380 	size_t objsize, len;
381 	vm_paddr_t gpa;
382 	char *baseaddr, *ptr;
383 	int error;
384 
385 	assert(vms == VM_MMAP_ALL);
386 
387 	/*
388 	 * If 'memsize' cannot fit entirely in the 'lowmem' segment then
389 	 * create another 'highmem' segment above 4GB for the remainder.
390 	 */
391 	if (memsize > ctx->lowmem_limit) {
392 		ctx->lowmem = ctx->lowmem_limit;
393 		ctx->highmem = memsize - ctx->lowmem_limit;
394 		objsize = 4*GB + ctx->highmem;
395 	} else {
396 		ctx->lowmem = memsize;
397 		ctx->highmem = 0;
398 		objsize = ctx->lowmem;
399 	}
400 
401 	error = vm_alloc_memseg(ctx, VM_SYSMEM, objsize, NULL);
402 	if (error)
403 		return (error);
404 
405 	/*
406 	 * Stake out a contiguous region covering the guest physical memory
407 	 * and the adjoining guard regions.
408 	 */
409 	len = VM_MMAP_GUARD_SIZE + objsize + VM_MMAP_GUARD_SIZE;
410 	ptr = mmap(NULL, len, PROT_NONE, MAP_GUARD | MAP_ALIGNED_SUPER, -1, 0);
411 	if (ptr == MAP_FAILED)
412 		return (-1);
413 
414 	baseaddr = ptr + VM_MMAP_GUARD_SIZE;
415 	if (ctx->highmem > 0) {
416 		gpa = 4*GB;
417 		len = ctx->highmem;
418 		error = setup_memory_segment(ctx, gpa, len, baseaddr);
419 		if (error)
420 			return (error);
421 	}
422 
423 	if (ctx->lowmem > 0) {
424 		gpa = 0;
425 		len = ctx->lowmem;
426 		error = setup_memory_segment(ctx, gpa, len, baseaddr);
427 		if (error)
428 			return (error);
429 	}
430 
431 	ctx->baseaddr = baseaddr;
432 
433 	return (0);
434 }
435 
436 /*
437  * Returns a non-NULL pointer if [gaddr, gaddr+len) is entirely contained in
438  * the lowmem or highmem regions.
439  *
440  * In particular return NULL if [gaddr, gaddr+len) falls in guest MMIO region.
441  * The instruction emulation code depends on this behavior.
442  */
443 void *
444 vm_map_gpa(struct vmctx *ctx, vm_paddr_t gaddr, size_t len)
445 {
446 
447 	if (ctx->lowmem > 0) {
448 		if (gaddr < ctx->lowmem && len <= ctx->lowmem &&
449 		    gaddr + len <= ctx->lowmem)
450 			return (ctx->baseaddr + gaddr);
451 	}
452 
453 	if (ctx->highmem > 0) {
454                 if (gaddr >= 4*GB) {
455 			if (gaddr < 4*GB + ctx->highmem &&
456 			    len <= ctx->highmem &&
457 			    gaddr + len <= 4*GB + ctx->highmem)
458 				return (ctx->baseaddr + gaddr);
459 		}
460 	}
461 
462 	return (NULL);
463 }
464 
465 vm_paddr_t
466 vm_rev_map_gpa(struct vmctx *ctx, void *addr)
467 {
468 	vm_paddr_t offaddr;
469 
470 	offaddr = (char *)addr - ctx->baseaddr;
471 
472 	if (ctx->lowmem > 0)
473 		if (offaddr >= 0 && offaddr <= ctx->lowmem)
474 			return (offaddr);
475 
476 	if (ctx->highmem > 0)
477 		if (offaddr >= 4*GB && offaddr < 4*GB + ctx->highmem)
478 			return (offaddr);
479 
480 	return ((vm_paddr_t)-1);
481 }
482 
483 /* TODO: maximum size for vmname */
484 int
485 vm_get_name(struct vmctx *ctx, char *buf, size_t max_len)
486 {
487 
488 	if (strlcpy(buf, ctx->name, max_len) >= max_len)
489 		return (EINVAL);
490 	return (0);
491 }
492 
493 size_t
494 vm_get_lowmem_size(struct vmctx *ctx)
495 {
496 
497 	return (ctx->lowmem);
498 }
499 
500 size_t
501 vm_get_highmem_size(struct vmctx *ctx)
502 {
503 
504 	return (ctx->highmem);
505 }
506 
507 void *
508 vm_create_devmem(struct vmctx *ctx, int segid, const char *name, size_t len)
509 {
510 	char pathname[MAXPATHLEN];
511 	size_t len2;
512 	char *base, *ptr;
513 	int fd, error, flags;
514 
515 	fd = -1;
516 	ptr = MAP_FAILED;
517 	if (name == NULL || strlen(name) == 0) {
518 		errno = EINVAL;
519 		goto done;
520 	}
521 
522 	error = vm_alloc_memseg(ctx, segid, len, name);
523 	if (error)
524 		goto done;
525 
526 	strlcpy(pathname, "/dev/vmm.io/", sizeof(pathname));
527 	strlcat(pathname, ctx->name, sizeof(pathname));
528 	strlcat(pathname, ".", sizeof(pathname));
529 	strlcat(pathname, name, sizeof(pathname));
530 
531 	fd = open(pathname, O_RDWR);
532 	if (fd < 0)
533 		goto done;
534 
535 	/*
536 	 * Stake out a contiguous region covering the device memory and the
537 	 * adjoining guard regions.
538 	 */
539 	len2 = VM_MMAP_GUARD_SIZE + len + VM_MMAP_GUARD_SIZE;
540 	base = mmap(NULL, len2, PROT_NONE, MAP_GUARD | MAP_ALIGNED_SUPER, -1,
541 	    0);
542 	if (base == MAP_FAILED)
543 		goto done;
544 
545 	flags = MAP_SHARED | MAP_FIXED;
546 	if ((ctx->memflags & VM_MEM_F_INCORE) == 0)
547 		flags |= MAP_NOCORE;
548 
549 	/* mmap the devmem region in the host address space */
550 	ptr = mmap(base + VM_MMAP_GUARD_SIZE, len, PROT_RW, flags, fd, 0);
551 done:
552 	if (fd >= 0)
553 		close(fd);
554 	return (ptr);
555 }
556 
557 int
558 vm_set_desc(struct vmctx *ctx, int vcpu, int reg,
559 	    uint64_t base, uint32_t limit, uint32_t access)
560 {
561 	int error;
562 	struct vm_seg_desc vmsegdesc;
563 
564 	bzero(&vmsegdesc, sizeof(vmsegdesc));
565 	vmsegdesc.cpuid = vcpu;
566 	vmsegdesc.regnum = reg;
567 	vmsegdesc.desc.base = base;
568 	vmsegdesc.desc.limit = limit;
569 	vmsegdesc.desc.access = access;
570 
571 	error = ioctl(ctx->fd, VM_SET_SEGMENT_DESCRIPTOR, &vmsegdesc);
572 	return (error);
573 }
574 
575 int
576 vm_get_desc(struct vmctx *ctx, int vcpu, int reg,
577 	    uint64_t *base, uint32_t *limit, uint32_t *access)
578 {
579 	int error;
580 	struct vm_seg_desc vmsegdesc;
581 
582 	bzero(&vmsegdesc, sizeof(vmsegdesc));
583 	vmsegdesc.cpuid = vcpu;
584 	vmsegdesc.regnum = reg;
585 
586 	error = ioctl(ctx->fd, VM_GET_SEGMENT_DESCRIPTOR, &vmsegdesc);
587 	if (error == 0) {
588 		*base = vmsegdesc.desc.base;
589 		*limit = vmsegdesc.desc.limit;
590 		*access = vmsegdesc.desc.access;
591 	}
592 	return (error);
593 }
594 
595 int
596 vm_get_seg_desc(struct vmctx *ctx, int vcpu, int reg, struct seg_desc *seg_desc)
597 {
598 	int error;
599 
600 	error = vm_get_desc(ctx, vcpu, reg, &seg_desc->base, &seg_desc->limit,
601 	    &seg_desc->access);
602 	return (error);
603 }
604 
605 int
606 vm_set_register(struct vmctx *ctx, int vcpu, int reg, uint64_t val)
607 {
608 	int error;
609 	struct vm_register vmreg;
610 
611 	bzero(&vmreg, sizeof(vmreg));
612 	vmreg.cpuid = vcpu;
613 	vmreg.regnum = reg;
614 	vmreg.regval = val;
615 
616 	error = ioctl(ctx->fd, VM_SET_REGISTER, &vmreg);
617 	return (error);
618 }
619 
620 int
621 vm_get_register(struct vmctx *ctx, int vcpu, int reg, uint64_t *ret_val)
622 {
623 	int error;
624 	struct vm_register vmreg;
625 
626 	bzero(&vmreg, sizeof(vmreg));
627 	vmreg.cpuid = vcpu;
628 	vmreg.regnum = reg;
629 
630 	error = ioctl(ctx->fd, VM_GET_REGISTER, &vmreg);
631 	*ret_val = vmreg.regval;
632 	return (error);
633 }
634 
635 int
636 vm_set_register_set(struct vmctx *ctx, int vcpu, unsigned int count,
637     const int *regnums, uint64_t *regvals)
638 {
639 	int error;
640 	struct vm_register_set vmregset;
641 
642 	bzero(&vmregset, sizeof(vmregset));
643 	vmregset.cpuid = vcpu;
644 	vmregset.count = count;
645 	vmregset.regnums = regnums;
646 	vmregset.regvals = regvals;
647 
648 	error = ioctl(ctx->fd, VM_SET_REGISTER_SET, &vmregset);
649 	return (error);
650 }
651 
652 int
653 vm_get_register_set(struct vmctx *ctx, int vcpu, unsigned int count,
654     const int *regnums, uint64_t *regvals)
655 {
656 	int error;
657 	struct vm_register_set vmregset;
658 
659 	bzero(&vmregset, sizeof(vmregset));
660 	vmregset.cpuid = vcpu;
661 	vmregset.count = count;
662 	vmregset.regnums = regnums;
663 	vmregset.regvals = regvals;
664 
665 	error = ioctl(ctx->fd, VM_GET_REGISTER_SET, &vmregset);
666 	return (error);
667 }
668 
669 int
670 vm_run(struct vmctx *ctx, int vcpu, struct vm_exit *vmexit)
671 {
672 	int error;
673 	struct vm_run vmrun;
674 
675 	bzero(&vmrun, sizeof(vmrun));
676 	vmrun.cpuid = vcpu;
677 
678 	error = ioctl(ctx->fd, VM_RUN, &vmrun);
679 	bcopy(&vmrun.vm_exit, vmexit, sizeof(struct vm_exit));
680 	return (error);
681 }
682 
683 int
684 vm_suspend(struct vmctx *ctx, enum vm_suspend_how how)
685 {
686 	struct vm_suspend vmsuspend;
687 
688 	bzero(&vmsuspend, sizeof(vmsuspend));
689 	vmsuspend.how = how;
690 	return (ioctl(ctx->fd, VM_SUSPEND, &vmsuspend));
691 }
692 
693 int
694 vm_reinit(struct vmctx *ctx)
695 {
696 
697 	return (ioctl(ctx->fd, VM_REINIT, 0));
698 }
699 
700 int
701 vm_inject_exception(struct vmctx *ctx, int vcpu, int vector, int errcode_valid,
702     uint32_t errcode, int restart_instruction)
703 {
704 	struct vm_exception exc;
705 
706 	exc.cpuid = vcpu;
707 	exc.vector = vector;
708 	exc.error_code = errcode;
709 	exc.error_code_valid = errcode_valid;
710 	exc.restart_instruction = restart_instruction;
711 
712 	return (ioctl(ctx->fd, VM_INJECT_EXCEPTION, &exc));
713 }
714 
715 int
716 vm_apicid2vcpu(struct vmctx *ctx, int apicid)
717 {
718 	/*
719 	 * The apic id associated with the 'vcpu' has the same numerical value
720 	 * as the 'vcpu' itself.
721 	 */
722 	return (apicid);
723 }
724 
725 int
726 vm_lapic_irq(struct vmctx *ctx, int vcpu, int vector)
727 {
728 	struct vm_lapic_irq vmirq;
729 
730 	bzero(&vmirq, sizeof(vmirq));
731 	vmirq.cpuid = vcpu;
732 	vmirq.vector = vector;
733 
734 	return (ioctl(ctx->fd, VM_LAPIC_IRQ, &vmirq));
735 }
736 
737 int
738 vm_lapic_local_irq(struct vmctx *ctx, int vcpu, int vector)
739 {
740 	struct vm_lapic_irq vmirq;
741 
742 	bzero(&vmirq, sizeof(vmirq));
743 	vmirq.cpuid = vcpu;
744 	vmirq.vector = vector;
745 
746 	return (ioctl(ctx->fd, VM_LAPIC_LOCAL_IRQ, &vmirq));
747 }
748 
749 int
750 vm_lapic_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg)
751 {
752 	struct vm_lapic_msi vmmsi;
753 
754 	bzero(&vmmsi, sizeof(vmmsi));
755 	vmmsi.addr = addr;
756 	vmmsi.msg = msg;
757 
758 	return (ioctl(ctx->fd, VM_LAPIC_MSI, &vmmsi));
759 }
760 
761 int
762 vm_ioapic_assert_irq(struct vmctx *ctx, int irq)
763 {
764 	struct vm_ioapic_irq ioapic_irq;
765 
766 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
767 	ioapic_irq.irq = irq;
768 
769 	return (ioctl(ctx->fd, VM_IOAPIC_ASSERT_IRQ, &ioapic_irq));
770 }
771 
772 int
773 vm_ioapic_deassert_irq(struct vmctx *ctx, int irq)
774 {
775 	struct vm_ioapic_irq ioapic_irq;
776 
777 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
778 	ioapic_irq.irq = irq;
779 
780 	return (ioctl(ctx->fd, VM_IOAPIC_DEASSERT_IRQ, &ioapic_irq));
781 }
782 
783 int
784 vm_ioapic_pulse_irq(struct vmctx *ctx, int irq)
785 {
786 	struct vm_ioapic_irq ioapic_irq;
787 
788 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
789 	ioapic_irq.irq = irq;
790 
791 	return (ioctl(ctx->fd, VM_IOAPIC_PULSE_IRQ, &ioapic_irq));
792 }
793 
794 int
795 vm_ioapic_pincount(struct vmctx *ctx, int *pincount)
796 {
797 
798 	return (ioctl(ctx->fd, VM_IOAPIC_PINCOUNT, pincount));
799 }
800 
801 int
802 vm_readwrite_kernemu_device(struct vmctx *ctx, int vcpu, vm_paddr_t gpa,
803     bool write, int size, uint64_t *value)
804 {
805 	struct vm_readwrite_kernemu_device irp = {
806 		.vcpuid = vcpu,
807 		.access_width = fls(size) - 1,
808 		.gpa = gpa,
809 		.value = write ? *value : ~0ul,
810 	};
811 	long cmd = (write ? VM_SET_KERNEMU_DEV : VM_GET_KERNEMU_DEV);
812 	int rc;
813 
814 	rc = ioctl(ctx->fd, cmd, &irp);
815 	if (rc == 0 && !write)
816 		*value = irp.value;
817 	return (rc);
818 }
819 
820 int
821 vm_isa_assert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
822 {
823 	struct vm_isa_irq isa_irq;
824 
825 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
826 	isa_irq.atpic_irq = atpic_irq;
827 	isa_irq.ioapic_irq = ioapic_irq;
828 
829 	return (ioctl(ctx->fd, VM_ISA_ASSERT_IRQ, &isa_irq));
830 }
831 
832 int
833 vm_isa_deassert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
834 {
835 	struct vm_isa_irq isa_irq;
836 
837 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
838 	isa_irq.atpic_irq = atpic_irq;
839 	isa_irq.ioapic_irq = ioapic_irq;
840 
841 	return (ioctl(ctx->fd, VM_ISA_DEASSERT_IRQ, &isa_irq));
842 }
843 
844 int
845 vm_isa_pulse_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
846 {
847 	struct vm_isa_irq isa_irq;
848 
849 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
850 	isa_irq.atpic_irq = atpic_irq;
851 	isa_irq.ioapic_irq = ioapic_irq;
852 
853 	return (ioctl(ctx->fd, VM_ISA_PULSE_IRQ, &isa_irq));
854 }
855 
856 int
857 vm_isa_set_irq_trigger(struct vmctx *ctx, int atpic_irq,
858     enum vm_intr_trigger trigger)
859 {
860 	struct vm_isa_irq_trigger isa_irq_trigger;
861 
862 	bzero(&isa_irq_trigger, sizeof(struct vm_isa_irq_trigger));
863 	isa_irq_trigger.atpic_irq = atpic_irq;
864 	isa_irq_trigger.trigger = trigger;
865 
866 	return (ioctl(ctx->fd, VM_ISA_SET_IRQ_TRIGGER, &isa_irq_trigger));
867 }
868 
869 int
870 vm_inject_nmi(struct vmctx *ctx, int vcpu)
871 {
872 	struct vm_nmi vmnmi;
873 
874 	bzero(&vmnmi, sizeof(vmnmi));
875 	vmnmi.cpuid = vcpu;
876 
877 	return (ioctl(ctx->fd, VM_INJECT_NMI, &vmnmi));
878 }
879 
880 static const char *capstrmap[] = {
881 	[VM_CAP_HALT_EXIT]  = "hlt_exit",
882 	[VM_CAP_MTRAP_EXIT] = "mtrap_exit",
883 	[VM_CAP_PAUSE_EXIT] = "pause_exit",
884 	[VM_CAP_UNRESTRICTED_GUEST] = "unrestricted_guest",
885 	[VM_CAP_ENABLE_INVPCID] = "enable_invpcid",
886 	[VM_CAP_BPT_EXIT] = "bpt_exit",
887 };
888 
889 int
890 vm_capability_name2type(const char *capname)
891 {
892 	int i;
893 
894 	for (i = 0; i < nitems(capstrmap); i++) {
895 		if (strcmp(capstrmap[i], capname) == 0)
896 			return (i);
897 	}
898 
899 	return (-1);
900 }
901 
902 const char *
903 vm_capability_type2name(int type)
904 {
905 	if (type >= 0 && type < nitems(capstrmap))
906 		return (capstrmap[type]);
907 
908 	return (NULL);
909 }
910 
911 int
912 vm_get_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap,
913 		  int *retval)
914 {
915 	int error;
916 	struct vm_capability vmcap;
917 
918 	bzero(&vmcap, sizeof(vmcap));
919 	vmcap.cpuid = vcpu;
920 	vmcap.captype = cap;
921 
922 	error = ioctl(ctx->fd, VM_GET_CAPABILITY, &vmcap);
923 	*retval = vmcap.capval;
924 	return (error);
925 }
926 
927 int
928 vm_set_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap, int val)
929 {
930 	struct vm_capability vmcap;
931 
932 	bzero(&vmcap, sizeof(vmcap));
933 	vmcap.cpuid = vcpu;
934 	vmcap.captype = cap;
935 	vmcap.capval = val;
936 
937 	return (ioctl(ctx->fd, VM_SET_CAPABILITY, &vmcap));
938 }
939 
940 int
941 vm_assign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
942 {
943 	struct vm_pptdev pptdev;
944 
945 	bzero(&pptdev, sizeof(pptdev));
946 	pptdev.bus = bus;
947 	pptdev.slot = slot;
948 	pptdev.func = func;
949 
950 	return (ioctl(ctx->fd, VM_BIND_PPTDEV, &pptdev));
951 }
952 
953 int
954 vm_unassign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
955 {
956 	struct vm_pptdev pptdev;
957 
958 	bzero(&pptdev, sizeof(pptdev));
959 	pptdev.bus = bus;
960 	pptdev.slot = slot;
961 	pptdev.func = func;
962 
963 	return (ioctl(ctx->fd, VM_UNBIND_PPTDEV, &pptdev));
964 }
965 
966 int
967 vm_map_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func,
968 		   vm_paddr_t gpa, size_t len, vm_paddr_t hpa)
969 {
970 	struct vm_pptdev_mmio pptmmio;
971 
972 	bzero(&pptmmio, sizeof(pptmmio));
973 	pptmmio.bus = bus;
974 	pptmmio.slot = slot;
975 	pptmmio.func = func;
976 	pptmmio.gpa = gpa;
977 	pptmmio.len = len;
978 	pptmmio.hpa = hpa;
979 
980 	return (ioctl(ctx->fd, VM_MAP_PPTDEV_MMIO, &pptmmio));
981 }
982 
983 int
984 vm_setup_pptdev_msi(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
985     uint64_t addr, uint64_t msg, int numvec)
986 {
987 	struct vm_pptdev_msi pptmsi;
988 
989 	bzero(&pptmsi, sizeof(pptmsi));
990 	pptmsi.vcpu = vcpu;
991 	pptmsi.bus = bus;
992 	pptmsi.slot = slot;
993 	pptmsi.func = func;
994 	pptmsi.msg = msg;
995 	pptmsi.addr = addr;
996 	pptmsi.numvec = numvec;
997 
998 	return (ioctl(ctx->fd, VM_PPTDEV_MSI, &pptmsi));
999 }
1000 
1001 int
1002 vm_setup_pptdev_msix(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
1003     int idx, uint64_t addr, uint64_t msg, uint32_t vector_control)
1004 {
1005 	struct vm_pptdev_msix pptmsix;
1006 
1007 	bzero(&pptmsix, sizeof(pptmsix));
1008 	pptmsix.vcpu = vcpu;
1009 	pptmsix.bus = bus;
1010 	pptmsix.slot = slot;
1011 	pptmsix.func = func;
1012 	pptmsix.idx = idx;
1013 	pptmsix.msg = msg;
1014 	pptmsix.addr = addr;
1015 	pptmsix.vector_control = vector_control;
1016 
1017 	return ioctl(ctx->fd, VM_PPTDEV_MSIX, &pptmsix);
1018 }
1019 
1020 int
1021 vm_disable_pptdev_msix(struct vmctx *ctx, int bus, int slot, int func)
1022 {
1023 	struct vm_pptdev ppt;
1024 
1025 	bzero(&ppt, sizeof(ppt));
1026 	ppt.bus = bus;
1027 	ppt.slot = slot;
1028 	ppt.func = func;
1029 
1030 	return ioctl(ctx->fd, VM_PPTDEV_DISABLE_MSIX, &ppt);
1031 }
1032 
1033 uint64_t *
1034 vm_get_stats(struct vmctx *ctx, int vcpu, struct timeval *ret_tv,
1035 	     int *ret_entries)
1036 {
1037 	int error;
1038 
1039 	static struct vm_stats vmstats;
1040 
1041 	vmstats.cpuid = vcpu;
1042 
1043 	error = ioctl(ctx->fd, VM_STATS, &vmstats);
1044 	if (error == 0) {
1045 		if (ret_entries)
1046 			*ret_entries = vmstats.num_entries;
1047 		if (ret_tv)
1048 			*ret_tv = vmstats.tv;
1049 		return (vmstats.statbuf);
1050 	} else
1051 		return (NULL);
1052 }
1053 
1054 const char *
1055 vm_get_stat_desc(struct vmctx *ctx, int index)
1056 {
1057 	static struct vm_stat_desc statdesc;
1058 
1059 	statdesc.index = index;
1060 	if (ioctl(ctx->fd, VM_STAT_DESC, &statdesc) == 0)
1061 		return (statdesc.desc);
1062 	else
1063 		return (NULL);
1064 }
1065 
1066 int
1067 vm_get_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state *state)
1068 {
1069 	int error;
1070 	struct vm_x2apic x2apic;
1071 
1072 	bzero(&x2apic, sizeof(x2apic));
1073 	x2apic.cpuid = vcpu;
1074 
1075 	error = ioctl(ctx->fd, VM_GET_X2APIC_STATE, &x2apic);
1076 	*state = x2apic.state;
1077 	return (error);
1078 }
1079 
1080 int
1081 vm_set_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state state)
1082 {
1083 	int error;
1084 	struct vm_x2apic x2apic;
1085 
1086 	bzero(&x2apic, sizeof(x2apic));
1087 	x2apic.cpuid = vcpu;
1088 	x2apic.state = state;
1089 
1090 	error = ioctl(ctx->fd, VM_SET_X2APIC_STATE, &x2apic);
1091 
1092 	return (error);
1093 }
1094 
1095 /*
1096  * From Intel Vol 3a:
1097  * Table 9-1. IA-32 Processor States Following Power-up, Reset or INIT
1098  */
1099 int
1100 vcpu_reset(struct vmctx *vmctx, int vcpu)
1101 {
1102 	int error;
1103 	uint64_t rflags, rip, cr0, cr4, zero, desc_base, rdx;
1104 	uint32_t desc_access, desc_limit;
1105 	uint16_t sel;
1106 
1107 	zero = 0;
1108 
1109 	rflags = 0x2;
1110 	error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RFLAGS, rflags);
1111 	if (error)
1112 		goto done;
1113 
1114 	rip = 0xfff0;
1115 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RIP, rip)) != 0)
1116 		goto done;
1117 
1118 	cr0 = CR0_NE;
1119 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR0, cr0)) != 0)
1120 		goto done;
1121 
1122 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR3, zero)) != 0)
1123 		goto done;
1124 
1125 	cr4 = 0;
1126 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR4, cr4)) != 0)
1127 		goto done;
1128 
1129 	/*
1130 	 * CS: present, r/w, accessed, 16-bit, byte granularity, usable
1131 	 */
1132 	desc_base = 0xffff0000;
1133 	desc_limit = 0xffff;
1134 	desc_access = 0x0093;
1135 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_CS,
1136 			    desc_base, desc_limit, desc_access);
1137 	if (error)
1138 		goto done;
1139 
1140 	sel = 0xf000;
1141 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CS, sel)) != 0)
1142 		goto done;
1143 
1144 	/*
1145 	 * SS,DS,ES,FS,GS: present, r/w, accessed, 16-bit, byte granularity
1146 	 */
1147 	desc_base = 0;
1148 	desc_limit = 0xffff;
1149 	desc_access = 0x0093;
1150 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_SS,
1151 			    desc_base, desc_limit, desc_access);
1152 	if (error)
1153 		goto done;
1154 
1155 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_DS,
1156 			    desc_base, desc_limit, desc_access);
1157 	if (error)
1158 		goto done;
1159 
1160 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_ES,
1161 			    desc_base, desc_limit, desc_access);
1162 	if (error)
1163 		goto done;
1164 
1165 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_FS,
1166 			    desc_base, desc_limit, desc_access);
1167 	if (error)
1168 		goto done;
1169 
1170 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GS,
1171 			    desc_base, desc_limit, desc_access);
1172 	if (error)
1173 		goto done;
1174 
1175 	sel = 0;
1176 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_SS, sel)) != 0)
1177 		goto done;
1178 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_DS, sel)) != 0)
1179 		goto done;
1180 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_ES, sel)) != 0)
1181 		goto done;
1182 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_FS, sel)) != 0)
1183 		goto done;
1184 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_GS, sel)) != 0)
1185 		goto done;
1186 
1187 	/* General purpose registers */
1188 	rdx = 0xf00;
1189 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RAX, zero)) != 0)
1190 		goto done;
1191 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBX, zero)) != 0)
1192 		goto done;
1193 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RCX, zero)) != 0)
1194 		goto done;
1195 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDX, rdx)) != 0)
1196 		goto done;
1197 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSI, zero)) != 0)
1198 		goto done;
1199 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDI, zero)) != 0)
1200 		goto done;
1201 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBP, zero)) != 0)
1202 		goto done;
1203 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSP, zero)) != 0)
1204 		goto done;
1205 
1206 	/* GDTR, IDTR */
1207 	desc_base = 0;
1208 	desc_limit = 0xffff;
1209 	desc_access = 0;
1210 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GDTR,
1211 			    desc_base, desc_limit, desc_access);
1212 	if (error != 0)
1213 		goto done;
1214 
1215 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_IDTR,
1216 			    desc_base, desc_limit, desc_access);
1217 	if (error != 0)
1218 		goto done;
1219 
1220 	/* TR */
1221 	desc_base = 0;
1222 	desc_limit = 0xffff;
1223 	desc_access = 0x0000008b;
1224 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_TR, 0, 0, desc_access);
1225 	if (error)
1226 		goto done;
1227 
1228 	sel = 0;
1229 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_TR, sel)) != 0)
1230 		goto done;
1231 
1232 	/* LDTR */
1233 	desc_base = 0;
1234 	desc_limit = 0xffff;
1235 	desc_access = 0x00000082;
1236 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_LDTR, desc_base,
1237 			    desc_limit, desc_access);
1238 	if (error)
1239 		goto done;
1240 
1241 	sel = 0;
1242 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_LDTR, 0)) != 0)
1243 		goto done;
1244 
1245 	/* XXX cr2, debug registers */
1246 
1247 	error = 0;
1248 done:
1249 	return (error);
1250 }
1251 
1252 int
1253 vm_get_gpa_pmap(struct vmctx *ctx, uint64_t gpa, uint64_t *pte, int *num)
1254 {
1255 	int error, i;
1256 	struct vm_gpa_pte gpapte;
1257 
1258 	bzero(&gpapte, sizeof(gpapte));
1259 	gpapte.gpa = gpa;
1260 
1261 	error = ioctl(ctx->fd, VM_GET_GPA_PMAP, &gpapte);
1262 
1263 	if (error == 0) {
1264 		*num = gpapte.ptenum;
1265 		for (i = 0; i < gpapte.ptenum; i++)
1266 			pte[i] = gpapte.pte[i];
1267 	}
1268 
1269 	return (error);
1270 }
1271 
1272 int
1273 vm_get_hpet_capabilities(struct vmctx *ctx, uint32_t *capabilities)
1274 {
1275 	int error;
1276 	struct vm_hpet_cap cap;
1277 
1278 	bzero(&cap, sizeof(struct vm_hpet_cap));
1279 	error = ioctl(ctx->fd, VM_GET_HPET_CAPABILITIES, &cap);
1280 	if (capabilities != NULL)
1281 		*capabilities = cap.capabilities;
1282 	return (error);
1283 }
1284 
1285 int
1286 vm_gla2gpa(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1287     uint64_t gla, int prot, uint64_t *gpa, int *fault)
1288 {
1289 	struct vm_gla2gpa gg;
1290 	int error;
1291 
1292 	bzero(&gg, sizeof(struct vm_gla2gpa));
1293 	gg.vcpuid = vcpu;
1294 	gg.prot = prot;
1295 	gg.gla = gla;
1296 	gg.paging = *paging;
1297 
1298 	error = ioctl(ctx->fd, VM_GLA2GPA, &gg);
1299 	if (error == 0) {
1300 		*fault = gg.fault;
1301 		*gpa = gg.gpa;
1302 	}
1303 	return (error);
1304 }
1305 
1306 int
1307 vm_gla2gpa_nofault(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1308     uint64_t gla, int prot, uint64_t *gpa, int *fault)
1309 {
1310 	struct vm_gla2gpa gg;
1311 	int error;
1312 
1313 	bzero(&gg, sizeof(struct vm_gla2gpa));
1314 	gg.vcpuid = vcpu;
1315 	gg.prot = prot;
1316 	gg.gla = gla;
1317 	gg.paging = *paging;
1318 
1319 	error = ioctl(ctx->fd, VM_GLA2GPA_NOFAULT, &gg);
1320 	if (error == 0) {
1321 		*fault = gg.fault;
1322 		*gpa = gg.gpa;
1323 	}
1324 	return (error);
1325 }
1326 
1327 #ifndef min
1328 #define	min(a,b)	(((a) < (b)) ? (a) : (b))
1329 #endif
1330 
1331 int
1332 vm_copy_setup(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1333     uint64_t gla, size_t len, int prot, struct iovec *iov, int iovcnt,
1334     int *fault)
1335 {
1336 	void *va;
1337 	uint64_t gpa;
1338 	int error, i, n, off;
1339 
1340 	for (i = 0; i < iovcnt; i++) {
1341 		iov[i].iov_base = 0;
1342 		iov[i].iov_len = 0;
1343 	}
1344 
1345 	while (len) {
1346 		assert(iovcnt > 0);
1347 		error = vm_gla2gpa(ctx, vcpu, paging, gla, prot, &gpa, fault);
1348 		if (error || *fault)
1349 			return (error);
1350 
1351 		off = gpa & PAGE_MASK;
1352 		n = min(len, PAGE_SIZE - off);
1353 
1354 		va = vm_map_gpa(ctx, gpa, n);
1355 		if (va == NULL)
1356 			return (EFAULT);
1357 
1358 		iov->iov_base = va;
1359 		iov->iov_len = n;
1360 		iov++;
1361 		iovcnt--;
1362 
1363 		gla += n;
1364 		len -= n;
1365 	}
1366 	return (0);
1367 }
1368 
1369 void
1370 vm_copy_teardown(struct vmctx *ctx, int vcpu, struct iovec *iov, int iovcnt)
1371 {
1372 
1373 	return;
1374 }
1375 
1376 void
1377 vm_copyin(struct vmctx *ctx, int vcpu, struct iovec *iov, void *vp, size_t len)
1378 {
1379 	const char *src;
1380 	char *dst;
1381 	size_t n;
1382 
1383 	dst = vp;
1384 	while (len) {
1385 		assert(iov->iov_len);
1386 		n = min(len, iov->iov_len);
1387 		src = iov->iov_base;
1388 		bcopy(src, dst, n);
1389 
1390 		iov++;
1391 		dst += n;
1392 		len -= n;
1393 	}
1394 }
1395 
1396 void
1397 vm_copyout(struct vmctx *ctx, int vcpu, const void *vp, struct iovec *iov,
1398     size_t len)
1399 {
1400 	const char *src;
1401 	char *dst;
1402 	size_t n;
1403 
1404 	src = vp;
1405 	while (len) {
1406 		assert(iov->iov_len);
1407 		n = min(len, iov->iov_len);
1408 		dst = iov->iov_base;
1409 		bcopy(src, dst, n);
1410 
1411 		iov++;
1412 		src += n;
1413 		len -= n;
1414 	}
1415 }
1416 
1417 static int
1418 vm_get_cpus(struct vmctx *ctx, int which, cpuset_t *cpus)
1419 {
1420 	struct vm_cpuset vm_cpuset;
1421 	int error;
1422 
1423 	bzero(&vm_cpuset, sizeof(struct vm_cpuset));
1424 	vm_cpuset.which = which;
1425 	vm_cpuset.cpusetsize = sizeof(cpuset_t);
1426 	vm_cpuset.cpus = cpus;
1427 
1428 	error = ioctl(ctx->fd, VM_GET_CPUS, &vm_cpuset);
1429 	return (error);
1430 }
1431 
1432 int
1433 vm_active_cpus(struct vmctx *ctx, cpuset_t *cpus)
1434 {
1435 
1436 	return (vm_get_cpus(ctx, VM_ACTIVE_CPUS, cpus));
1437 }
1438 
1439 int
1440 vm_suspended_cpus(struct vmctx *ctx, cpuset_t *cpus)
1441 {
1442 
1443 	return (vm_get_cpus(ctx, VM_SUSPENDED_CPUS, cpus));
1444 }
1445 
1446 int
1447 vm_debug_cpus(struct vmctx *ctx, cpuset_t *cpus)
1448 {
1449 
1450 	return (vm_get_cpus(ctx, VM_DEBUG_CPUS, cpus));
1451 }
1452 
1453 int
1454 vm_activate_cpu(struct vmctx *ctx, int vcpu)
1455 {
1456 	struct vm_activate_cpu ac;
1457 	int error;
1458 
1459 	bzero(&ac, sizeof(struct vm_activate_cpu));
1460 	ac.vcpuid = vcpu;
1461 	error = ioctl(ctx->fd, VM_ACTIVATE_CPU, &ac);
1462 	return (error);
1463 }
1464 
1465 int
1466 vm_suspend_cpu(struct vmctx *ctx, int vcpu)
1467 {
1468 	struct vm_activate_cpu ac;
1469 	int error;
1470 
1471 	bzero(&ac, sizeof(struct vm_activate_cpu));
1472 	ac.vcpuid = vcpu;
1473 	error = ioctl(ctx->fd, VM_SUSPEND_CPU, &ac);
1474 	return (error);
1475 }
1476 
1477 int
1478 vm_resume_cpu(struct vmctx *ctx, int vcpu)
1479 {
1480 	struct vm_activate_cpu ac;
1481 	int error;
1482 
1483 	bzero(&ac, sizeof(struct vm_activate_cpu));
1484 	ac.vcpuid = vcpu;
1485 	error = ioctl(ctx->fd, VM_RESUME_CPU, &ac);
1486 	return (error);
1487 }
1488 
1489 int
1490 vm_get_intinfo(struct vmctx *ctx, int vcpu, uint64_t *info1, uint64_t *info2)
1491 {
1492 	struct vm_intinfo vmii;
1493 	int error;
1494 
1495 	bzero(&vmii, sizeof(struct vm_intinfo));
1496 	vmii.vcpuid = vcpu;
1497 	error = ioctl(ctx->fd, VM_GET_INTINFO, &vmii);
1498 	if (error == 0) {
1499 		*info1 = vmii.info1;
1500 		*info2 = vmii.info2;
1501 	}
1502 	return (error);
1503 }
1504 
1505 int
1506 vm_set_intinfo(struct vmctx *ctx, int vcpu, uint64_t info1)
1507 {
1508 	struct vm_intinfo vmii;
1509 	int error;
1510 
1511 	bzero(&vmii, sizeof(struct vm_intinfo));
1512 	vmii.vcpuid = vcpu;
1513 	vmii.info1 = info1;
1514 	error = ioctl(ctx->fd, VM_SET_INTINFO, &vmii);
1515 	return (error);
1516 }
1517 
1518 int
1519 vm_rtc_write(struct vmctx *ctx, int offset, uint8_t value)
1520 {
1521 	struct vm_rtc_data rtcdata;
1522 	int error;
1523 
1524 	bzero(&rtcdata, sizeof(struct vm_rtc_data));
1525 	rtcdata.offset = offset;
1526 	rtcdata.value = value;
1527 	error = ioctl(ctx->fd, VM_RTC_WRITE, &rtcdata);
1528 	return (error);
1529 }
1530 
1531 int
1532 vm_rtc_read(struct vmctx *ctx, int offset, uint8_t *retval)
1533 {
1534 	struct vm_rtc_data rtcdata;
1535 	int error;
1536 
1537 	bzero(&rtcdata, sizeof(struct vm_rtc_data));
1538 	rtcdata.offset = offset;
1539 	error = ioctl(ctx->fd, VM_RTC_READ, &rtcdata);
1540 	if (error == 0)
1541 		*retval = rtcdata.value;
1542 	return (error);
1543 }
1544 
1545 int
1546 vm_rtc_settime(struct vmctx *ctx, time_t secs)
1547 {
1548 	struct vm_rtc_time rtctime;
1549 	int error;
1550 
1551 	bzero(&rtctime, sizeof(struct vm_rtc_time));
1552 	rtctime.secs = secs;
1553 	error = ioctl(ctx->fd, VM_RTC_SETTIME, &rtctime);
1554 	return (error);
1555 }
1556 
1557 int
1558 vm_rtc_gettime(struct vmctx *ctx, time_t *secs)
1559 {
1560 	struct vm_rtc_time rtctime;
1561 	int error;
1562 
1563 	bzero(&rtctime, sizeof(struct vm_rtc_time));
1564 	error = ioctl(ctx->fd, VM_RTC_GETTIME, &rtctime);
1565 	if (error == 0)
1566 		*secs = rtctime.secs;
1567 	return (error);
1568 }
1569 
1570 int
1571 vm_restart_instruction(void *arg, int vcpu)
1572 {
1573 	struct vmctx *ctx = arg;
1574 
1575 	return (ioctl(ctx->fd, VM_RESTART_INSTRUCTION, &vcpu));
1576 }
1577 
1578 int
1579 vm_snapshot_req(struct vm_snapshot_meta *meta)
1580 {
1581 
1582 	if (ioctl(meta->ctx->fd, VM_SNAPSHOT_REQ, meta) == -1) {
1583 #ifdef SNAPSHOT_DEBUG
1584 		fprintf(stderr, "%s: snapshot failed for %s: %d\r\n",
1585 		    __func__, meta->dev_name, errno);
1586 #endif
1587 		return (-1);
1588 	}
1589 	return (0);
1590 }
1591 
1592 int
1593 vm_restore_time(struct vmctx *ctx)
1594 {
1595 	int dummy;
1596 
1597 	dummy = 0;
1598 	return (ioctl(ctx->fd, VM_RESTORE_TIME, &dummy));
1599 }
1600 
1601 int
1602 vm_set_topology(struct vmctx *ctx,
1603     uint16_t sockets, uint16_t cores, uint16_t threads, uint16_t maxcpus)
1604 {
1605 	struct vm_cpu_topology topology;
1606 
1607 	bzero(&topology, sizeof (struct vm_cpu_topology));
1608 	topology.sockets = sockets;
1609 	topology.cores = cores;
1610 	topology.threads = threads;
1611 	topology.maxcpus = maxcpus;
1612 	return (ioctl(ctx->fd, VM_SET_TOPOLOGY, &topology));
1613 }
1614 
1615 int
1616 vm_get_topology(struct vmctx *ctx,
1617     uint16_t *sockets, uint16_t *cores, uint16_t *threads, uint16_t *maxcpus)
1618 {
1619 	struct vm_cpu_topology topology;
1620 	int error;
1621 
1622 	bzero(&topology, sizeof (struct vm_cpu_topology));
1623 	error = ioctl(ctx->fd, VM_GET_TOPOLOGY, &topology);
1624 	if (error == 0) {
1625 		*sockets = topology.sockets;
1626 		*cores = topology.cores;
1627 		*threads = topology.threads;
1628 		*maxcpus = topology.maxcpus;
1629 	}
1630 	return (error);
1631 }
1632 
1633 int
1634 vm_get_device_fd(struct vmctx *ctx)
1635 {
1636 
1637 	return (ctx->fd);
1638 }
1639 
1640 const cap_ioctl_t *
1641 vm_get_ioctls(size_t *len)
1642 {
1643 	cap_ioctl_t *cmds;
1644 	/* keep in sync with machine/vmm_dev.h */
1645 	static const cap_ioctl_t vm_ioctl_cmds[] = { VM_RUN, VM_SUSPEND, VM_REINIT,
1646 	    VM_ALLOC_MEMSEG, VM_GET_MEMSEG, VM_MMAP_MEMSEG, VM_MMAP_MEMSEG,
1647 	    VM_MMAP_GETNEXT, VM_SET_REGISTER, VM_GET_REGISTER,
1648 	    VM_SET_SEGMENT_DESCRIPTOR, VM_GET_SEGMENT_DESCRIPTOR,
1649 	    VM_SET_REGISTER_SET, VM_GET_REGISTER_SET,
1650 	    VM_SET_KERNEMU_DEV, VM_GET_KERNEMU_DEV,
1651 	    VM_INJECT_EXCEPTION, VM_LAPIC_IRQ, VM_LAPIC_LOCAL_IRQ,
1652 	    VM_LAPIC_MSI, VM_IOAPIC_ASSERT_IRQ, VM_IOAPIC_DEASSERT_IRQ,
1653 	    VM_IOAPIC_PULSE_IRQ, VM_IOAPIC_PINCOUNT, VM_ISA_ASSERT_IRQ,
1654 	    VM_ISA_DEASSERT_IRQ, VM_ISA_PULSE_IRQ, VM_ISA_SET_IRQ_TRIGGER,
1655 	    VM_SET_CAPABILITY, VM_GET_CAPABILITY, VM_BIND_PPTDEV,
1656 	    VM_UNBIND_PPTDEV, VM_MAP_PPTDEV_MMIO, VM_PPTDEV_MSI,
1657 	    VM_PPTDEV_MSIX, VM_PPTDEV_DISABLE_MSIX,
1658 	    VM_INJECT_NMI, VM_STATS, VM_STAT_DESC,
1659 	    VM_SET_X2APIC_STATE, VM_GET_X2APIC_STATE,
1660 	    VM_GET_HPET_CAPABILITIES, VM_GET_GPA_PMAP, VM_GLA2GPA,
1661 	    VM_GLA2GPA_NOFAULT,
1662 	    VM_ACTIVATE_CPU, VM_GET_CPUS, VM_SUSPEND_CPU, VM_RESUME_CPU,
1663 	    VM_SET_INTINFO, VM_GET_INTINFO,
1664 	    VM_RTC_WRITE, VM_RTC_READ, VM_RTC_SETTIME, VM_RTC_GETTIME,
1665 	    VM_RESTART_INSTRUCTION, VM_SET_TOPOLOGY, VM_GET_TOPOLOGY };
1666 
1667 	if (len == NULL) {
1668 		cmds = malloc(sizeof(vm_ioctl_cmds));
1669 		if (cmds == NULL)
1670 			return (NULL);
1671 		bcopy(vm_ioctl_cmds, cmds, sizeof(vm_ioctl_cmds));
1672 		return (cmds);
1673 	}
1674 
1675 	*len = nitems(vm_ioctl_cmds);
1676 	return (NULL);
1677 }
1678