xref: /freebsd/lib/libvmmapi/vmmapi.c (revision 2e5b60079b7d8c3ca68f1390cd90f305e651f8d3)
1 /*-
2  * Copyright (c) 2011 NetApp, Inc.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  *
26  * $FreeBSD$
27  */
28 
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31 
32 #include <sys/param.h>
33 #include <sys/sysctl.h>
34 #include <sys/ioctl.h>
35 #include <sys/mman.h>
36 #include <sys/_iovec.h>
37 #include <sys/cpuset.h>
38 
39 #include <x86/segments.h>
40 #include <machine/specialreg.h>
41 #include <machine/param.h>
42 
43 #include <stdio.h>
44 #include <stdlib.h>
45 #include <assert.h>
46 #include <string.h>
47 #include <fcntl.h>
48 #include <unistd.h>
49 
50 #include <libutil.h>
51 
52 #include <machine/vmm.h>
53 #include <machine/vmm_dev.h>
54 
55 #include "vmmapi.h"
56 
57 #define	MB	(1024 * 1024UL)
58 #define	GB	(1024 * 1024 * 1024UL)
59 
60 struct vmctx {
61 	int	fd;
62 	uint32_t lowmem_limit;
63 	enum vm_mmap_style vms;
64 	int	memflags;
65 	size_t	lowmem;
66 	char	*lowmem_addr;
67 	size_t	highmem;
68 	char	*highmem_addr;
69 	char	*name;
70 };
71 
72 #define	CREATE(x)  sysctlbyname("hw.vmm.create", NULL, NULL, (x), strlen((x)))
73 #define	DESTROY(x) sysctlbyname("hw.vmm.destroy", NULL, NULL, (x), strlen((x)))
74 
75 static int
76 vm_device_open(const char *name)
77 {
78         int fd, len;
79         char *vmfile;
80 
81 	len = strlen("/dev/vmm/") + strlen(name) + 1;
82 	vmfile = malloc(len);
83 	assert(vmfile != NULL);
84 	snprintf(vmfile, len, "/dev/vmm/%s", name);
85 
86         /* Open the device file */
87         fd = open(vmfile, O_RDWR, 0);
88 
89 	free(vmfile);
90         return (fd);
91 }
92 
93 int
94 vm_create(const char *name)
95 {
96 
97 	return (CREATE((char *)name));
98 }
99 
100 struct vmctx *
101 vm_open(const char *name)
102 {
103 	struct vmctx *vm;
104 
105 	vm = malloc(sizeof(struct vmctx) + strlen(name) + 1);
106 	assert(vm != NULL);
107 
108 	vm->fd = -1;
109 	vm->memflags = 0;
110 	vm->lowmem_limit = 3 * GB;
111 	vm->name = (char *)(vm + 1);
112 	strcpy(vm->name, name);
113 
114 	if ((vm->fd = vm_device_open(vm->name)) < 0)
115 		goto err;
116 
117 	return (vm);
118 err:
119 	vm_destroy(vm);
120 	return (NULL);
121 }
122 
123 void
124 vm_destroy(struct vmctx *vm)
125 {
126 	assert(vm != NULL);
127 
128 	if (vm->fd >= 0)
129 		close(vm->fd);
130 	DESTROY(vm->name);
131 
132 	free(vm);
133 }
134 
135 int
136 vm_parse_memsize(const char *optarg, size_t *ret_memsize)
137 {
138 	char *endptr;
139 	size_t optval;
140 	int error;
141 
142 	optval = strtoul(optarg, &endptr, 0);
143 	if (*optarg != '\0' && *endptr == '\0') {
144 		/*
145 		 * For the sake of backward compatibility if the memory size
146 		 * specified on the command line is less than a megabyte then
147 		 * it is interpreted as being in units of MB.
148 		 */
149 		if (optval < MB)
150 			optval *= MB;
151 		*ret_memsize = optval;
152 		error = 0;
153 	} else
154 		error = expand_number(optarg, ret_memsize);
155 
156 	return (error);
157 }
158 
159 int
160 vm_get_memory_seg(struct vmctx *ctx, vm_paddr_t gpa, size_t *ret_len,
161 		  int *wired)
162 {
163 	int error;
164 	struct vm_memory_segment seg;
165 
166 	bzero(&seg, sizeof(seg));
167 	seg.gpa = gpa;
168 	error = ioctl(ctx->fd, VM_GET_MEMORY_SEG, &seg);
169 	*ret_len = seg.len;
170 	if (wired != NULL)
171 		*wired = seg.wired;
172 	return (error);
173 }
174 
175 uint32_t
176 vm_get_lowmem_limit(struct vmctx *ctx)
177 {
178 
179 	return (ctx->lowmem_limit);
180 }
181 
182 void
183 vm_set_lowmem_limit(struct vmctx *ctx, uint32_t limit)
184 {
185 
186 	ctx->lowmem_limit = limit;
187 }
188 
189 void
190 vm_set_memflags(struct vmctx *ctx, int flags)
191 {
192 
193 	ctx->memflags = flags;
194 }
195 
196 static int
197 setup_memory_segment(struct vmctx *ctx, vm_paddr_t gpa, size_t len, char **addr)
198 {
199 	int error, mmap_flags;
200 	struct vm_memory_segment seg;
201 
202 	/*
203 	 * Create and optionally map 'len' bytes of memory at guest
204 	 * physical address 'gpa'
205 	 */
206 	bzero(&seg, sizeof(seg));
207 	seg.gpa = gpa;
208 	seg.len = len;
209 	error = ioctl(ctx->fd, VM_MAP_MEMORY, &seg);
210 	if (error == 0 && addr != NULL) {
211 		mmap_flags = MAP_SHARED;
212 		if ((ctx->memflags & VM_MEM_F_INCORE) == 0)
213 			mmap_flags |= MAP_NOCORE;
214 		*addr = mmap(NULL, len, PROT_READ | PROT_WRITE, mmap_flags,
215 		    ctx->fd, gpa);
216 	}
217 	return (error);
218 }
219 
220 int
221 vm_setup_memory(struct vmctx *ctx, size_t memsize, enum vm_mmap_style vms)
222 {
223 	char **addr;
224 	int error;
225 
226 	/* XXX VM_MMAP_SPARSE not implemented yet */
227 	assert(vms == VM_MMAP_NONE || vms == VM_MMAP_ALL);
228 	ctx->vms = vms;
229 
230 	/*
231 	 * If 'memsize' cannot fit entirely in the 'lowmem' segment then
232 	 * create another 'highmem' segment above 4GB for the remainder.
233 	 */
234 	if (memsize > ctx->lowmem_limit) {
235 		ctx->lowmem = ctx->lowmem_limit;
236 		ctx->highmem = memsize - ctx->lowmem;
237 	} else {
238 		ctx->lowmem = memsize;
239 		ctx->highmem = 0;
240 	}
241 
242 	if (ctx->lowmem > 0) {
243 		addr = (vms == VM_MMAP_ALL) ? &ctx->lowmem_addr : NULL;
244 		error = setup_memory_segment(ctx, 0, ctx->lowmem, addr);
245 		if (error)
246 			return (error);
247 	}
248 
249 	if (ctx->highmem > 0) {
250 		addr = (vms == VM_MMAP_ALL) ? &ctx->highmem_addr : NULL;
251 		error = setup_memory_segment(ctx, 4*GB, ctx->highmem, addr);
252 		if (error)
253 			return (error);
254 	}
255 
256 	return (0);
257 }
258 
259 void *
260 vm_map_gpa(struct vmctx *ctx, vm_paddr_t gaddr, size_t len)
261 {
262 
263 	/* XXX VM_MMAP_SPARSE not implemented yet */
264 	assert(ctx->vms == VM_MMAP_ALL);
265 
266 	if (gaddr < ctx->lowmem && gaddr + len <= ctx->lowmem)
267 		return ((void *)(ctx->lowmem_addr + gaddr));
268 
269 	if (gaddr >= 4*GB) {
270 		gaddr -= 4*GB;
271 		if (gaddr < ctx->highmem && gaddr + len <= ctx->highmem)
272 			return ((void *)(ctx->highmem_addr + gaddr));
273 	}
274 
275 	return (NULL);
276 }
277 
278 size_t
279 vm_get_lowmem_size(struct vmctx *ctx)
280 {
281 
282 	return (ctx->lowmem);
283 }
284 
285 size_t
286 vm_get_highmem_size(struct vmctx *ctx)
287 {
288 
289 	return (ctx->highmem);
290 }
291 
292 int
293 vm_set_desc(struct vmctx *ctx, int vcpu, int reg,
294 	    uint64_t base, uint32_t limit, uint32_t access)
295 {
296 	int error;
297 	struct vm_seg_desc vmsegdesc;
298 
299 	bzero(&vmsegdesc, sizeof(vmsegdesc));
300 	vmsegdesc.cpuid = vcpu;
301 	vmsegdesc.regnum = reg;
302 	vmsegdesc.desc.base = base;
303 	vmsegdesc.desc.limit = limit;
304 	vmsegdesc.desc.access = access;
305 
306 	error = ioctl(ctx->fd, VM_SET_SEGMENT_DESCRIPTOR, &vmsegdesc);
307 	return (error);
308 }
309 
310 int
311 vm_get_desc(struct vmctx *ctx, int vcpu, int reg,
312 	    uint64_t *base, uint32_t *limit, uint32_t *access)
313 {
314 	int error;
315 	struct vm_seg_desc vmsegdesc;
316 
317 	bzero(&vmsegdesc, sizeof(vmsegdesc));
318 	vmsegdesc.cpuid = vcpu;
319 	vmsegdesc.regnum = reg;
320 
321 	error = ioctl(ctx->fd, VM_GET_SEGMENT_DESCRIPTOR, &vmsegdesc);
322 	if (error == 0) {
323 		*base = vmsegdesc.desc.base;
324 		*limit = vmsegdesc.desc.limit;
325 		*access = vmsegdesc.desc.access;
326 	}
327 	return (error);
328 }
329 
330 int
331 vm_get_seg_desc(struct vmctx *ctx, int vcpu, int reg, struct seg_desc *seg_desc)
332 {
333 	int error;
334 
335 	error = vm_get_desc(ctx, vcpu, reg, &seg_desc->base, &seg_desc->limit,
336 	    &seg_desc->access);
337 	return (error);
338 }
339 
340 int
341 vm_set_register(struct vmctx *ctx, int vcpu, int reg, uint64_t val)
342 {
343 	int error;
344 	struct vm_register vmreg;
345 
346 	bzero(&vmreg, sizeof(vmreg));
347 	vmreg.cpuid = vcpu;
348 	vmreg.regnum = reg;
349 	vmreg.regval = val;
350 
351 	error = ioctl(ctx->fd, VM_SET_REGISTER, &vmreg);
352 	return (error);
353 }
354 
355 int
356 vm_get_register(struct vmctx *ctx, int vcpu, int reg, uint64_t *ret_val)
357 {
358 	int error;
359 	struct vm_register vmreg;
360 
361 	bzero(&vmreg, sizeof(vmreg));
362 	vmreg.cpuid = vcpu;
363 	vmreg.regnum = reg;
364 
365 	error = ioctl(ctx->fd, VM_GET_REGISTER, &vmreg);
366 	*ret_val = vmreg.regval;
367 	return (error);
368 }
369 
370 int
371 vm_run(struct vmctx *ctx, int vcpu, struct vm_exit *vmexit)
372 {
373 	int error;
374 	struct vm_run vmrun;
375 
376 	bzero(&vmrun, sizeof(vmrun));
377 	vmrun.cpuid = vcpu;
378 
379 	error = ioctl(ctx->fd, VM_RUN, &vmrun);
380 	bcopy(&vmrun.vm_exit, vmexit, sizeof(struct vm_exit));
381 	return (error);
382 }
383 
384 int
385 vm_suspend(struct vmctx *ctx, enum vm_suspend_how how)
386 {
387 	struct vm_suspend vmsuspend;
388 
389 	bzero(&vmsuspend, sizeof(vmsuspend));
390 	vmsuspend.how = how;
391 	return (ioctl(ctx->fd, VM_SUSPEND, &vmsuspend));
392 }
393 
394 int
395 vm_reinit(struct vmctx *ctx)
396 {
397 
398 	return (ioctl(ctx->fd, VM_REINIT, 0));
399 }
400 
401 int
402 vm_inject_exception(struct vmctx *ctx, int vcpu, int vector, int errcode_valid,
403     uint32_t errcode, int restart_instruction)
404 {
405 	struct vm_exception exc;
406 
407 	exc.cpuid = vcpu;
408 	exc.vector = vector;
409 	exc.error_code = errcode;
410 	exc.error_code_valid = errcode_valid;
411 	exc.restart_instruction = restart_instruction;
412 
413 	return (ioctl(ctx->fd, VM_INJECT_EXCEPTION, &exc));
414 }
415 
416 int
417 vm_apicid2vcpu(struct vmctx *ctx, int apicid)
418 {
419 	/*
420 	 * The apic id associated with the 'vcpu' has the same numerical value
421 	 * as the 'vcpu' itself.
422 	 */
423 	return (apicid);
424 }
425 
426 int
427 vm_lapic_irq(struct vmctx *ctx, int vcpu, int vector)
428 {
429 	struct vm_lapic_irq vmirq;
430 
431 	bzero(&vmirq, sizeof(vmirq));
432 	vmirq.cpuid = vcpu;
433 	vmirq.vector = vector;
434 
435 	return (ioctl(ctx->fd, VM_LAPIC_IRQ, &vmirq));
436 }
437 
438 int
439 vm_lapic_local_irq(struct vmctx *ctx, int vcpu, int vector)
440 {
441 	struct vm_lapic_irq vmirq;
442 
443 	bzero(&vmirq, sizeof(vmirq));
444 	vmirq.cpuid = vcpu;
445 	vmirq.vector = vector;
446 
447 	return (ioctl(ctx->fd, VM_LAPIC_LOCAL_IRQ, &vmirq));
448 }
449 
450 int
451 vm_lapic_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg)
452 {
453 	struct vm_lapic_msi vmmsi;
454 
455 	bzero(&vmmsi, sizeof(vmmsi));
456 	vmmsi.addr = addr;
457 	vmmsi.msg = msg;
458 
459 	return (ioctl(ctx->fd, VM_LAPIC_MSI, &vmmsi));
460 }
461 
462 int
463 vm_ioapic_assert_irq(struct vmctx *ctx, int irq)
464 {
465 	struct vm_ioapic_irq ioapic_irq;
466 
467 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
468 	ioapic_irq.irq = irq;
469 
470 	return (ioctl(ctx->fd, VM_IOAPIC_ASSERT_IRQ, &ioapic_irq));
471 }
472 
473 int
474 vm_ioapic_deassert_irq(struct vmctx *ctx, int irq)
475 {
476 	struct vm_ioapic_irq ioapic_irq;
477 
478 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
479 	ioapic_irq.irq = irq;
480 
481 	return (ioctl(ctx->fd, VM_IOAPIC_DEASSERT_IRQ, &ioapic_irq));
482 }
483 
484 int
485 vm_ioapic_pulse_irq(struct vmctx *ctx, int irq)
486 {
487 	struct vm_ioapic_irq ioapic_irq;
488 
489 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
490 	ioapic_irq.irq = irq;
491 
492 	return (ioctl(ctx->fd, VM_IOAPIC_PULSE_IRQ, &ioapic_irq));
493 }
494 
495 int
496 vm_ioapic_pincount(struct vmctx *ctx, int *pincount)
497 {
498 
499 	return (ioctl(ctx->fd, VM_IOAPIC_PINCOUNT, pincount));
500 }
501 
502 int
503 vm_isa_assert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
504 {
505 	struct vm_isa_irq isa_irq;
506 
507 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
508 	isa_irq.atpic_irq = atpic_irq;
509 	isa_irq.ioapic_irq = ioapic_irq;
510 
511 	return (ioctl(ctx->fd, VM_ISA_ASSERT_IRQ, &isa_irq));
512 }
513 
514 int
515 vm_isa_deassert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
516 {
517 	struct vm_isa_irq isa_irq;
518 
519 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
520 	isa_irq.atpic_irq = atpic_irq;
521 	isa_irq.ioapic_irq = ioapic_irq;
522 
523 	return (ioctl(ctx->fd, VM_ISA_DEASSERT_IRQ, &isa_irq));
524 }
525 
526 int
527 vm_isa_pulse_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
528 {
529 	struct vm_isa_irq isa_irq;
530 
531 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
532 	isa_irq.atpic_irq = atpic_irq;
533 	isa_irq.ioapic_irq = ioapic_irq;
534 
535 	return (ioctl(ctx->fd, VM_ISA_PULSE_IRQ, &isa_irq));
536 }
537 
538 int
539 vm_isa_set_irq_trigger(struct vmctx *ctx, int atpic_irq,
540     enum vm_intr_trigger trigger)
541 {
542 	struct vm_isa_irq_trigger isa_irq_trigger;
543 
544 	bzero(&isa_irq_trigger, sizeof(struct vm_isa_irq_trigger));
545 	isa_irq_trigger.atpic_irq = atpic_irq;
546 	isa_irq_trigger.trigger = trigger;
547 
548 	return (ioctl(ctx->fd, VM_ISA_SET_IRQ_TRIGGER, &isa_irq_trigger));
549 }
550 
551 int
552 vm_inject_nmi(struct vmctx *ctx, int vcpu)
553 {
554 	struct vm_nmi vmnmi;
555 
556 	bzero(&vmnmi, sizeof(vmnmi));
557 	vmnmi.cpuid = vcpu;
558 
559 	return (ioctl(ctx->fd, VM_INJECT_NMI, &vmnmi));
560 }
561 
562 static struct {
563 	const char	*name;
564 	int		type;
565 } capstrmap[] = {
566 	{ "hlt_exit",		VM_CAP_HALT_EXIT },
567 	{ "mtrap_exit",		VM_CAP_MTRAP_EXIT },
568 	{ "pause_exit",		VM_CAP_PAUSE_EXIT },
569 	{ "unrestricted_guest",	VM_CAP_UNRESTRICTED_GUEST },
570 	{ "enable_invpcid",	VM_CAP_ENABLE_INVPCID },
571 	{ 0 }
572 };
573 
574 int
575 vm_capability_name2type(const char *capname)
576 {
577 	int i;
578 
579 	for (i = 0; capstrmap[i].name != NULL && capname != NULL; i++) {
580 		if (strcmp(capstrmap[i].name, capname) == 0)
581 			return (capstrmap[i].type);
582 	}
583 
584 	return (-1);
585 }
586 
587 const char *
588 vm_capability_type2name(int type)
589 {
590 	int i;
591 
592 	for (i = 0; capstrmap[i].name != NULL; i++) {
593 		if (capstrmap[i].type == type)
594 			return (capstrmap[i].name);
595 	}
596 
597 	return (NULL);
598 }
599 
600 int
601 vm_get_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap,
602 		  int *retval)
603 {
604 	int error;
605 	struct vm_capability vmcap;
606 
607 	bzero(&vmcap, sizeof(vmcap));
608 	vmcap.cpuid = vcpu;
609 	vmcap.captype = cap;
610 
611 	error = ioctl(ctx->fd, VM_GET_CAPABILITY, &vmcap);
612 	*retval = vmcap.capval;
613 	return (error);
614 }
615 
616 int
617 vm_set_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap, int val)
618 {
619 	struct vm_capability vmcap;
620 
621 	bzero(&vmcap, sizeof(vmcap));
622 	vmcap.cpuid = vcpu;
623 	vmcap.captype = cap;
624 	vmcap.capval = val;
625 
626 	return (ioctl(ctx->fd, VM_SET_CAPABILITY, &vmcap));
627 }
628 
629 int
630 vm_assign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
631 {
632 	struct vm_pptdev pptdev;
633 
634 	bzero(&pptdev, sizeof(pptdev));
635 	pptdev.bus = bus;
636 	pptdev.slot = slot;
637 	pptdev.func = func;
638 
639 	return (ioctl(ctx->fd, VM_BIND_PPTDEV, &pptdev));
640 }
641 
642 int
643 vm_unassign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
644 {
645 	struct vm_pptdev pptdev;
646 
647 	bzero(&pptdev, sizeof(pptdev));
648 	pptdev.bus = bus;
649 	pptdev.slot = slot;
650 	pptdev.func = func;
651 
652 	return (ioctl(ctx->fd, VM_UNBIND_PPTDEV, &pptdev));
653 }
654 
655 int
656 vm_map_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func,
657 		   vm_paddr_t gpa, size_t len, vm_paddr_t hpa)
658 {
659 	struct vm_pptdev_mmio pptmmio;
660 
661 	bzero(&pptmmio, sizeof(pptmmio));
662 	pptmmio.bus = bus;
663 	pptmmio.slot = slot;
664 	pptmmio.func = func;
665 	pptmmio.gpa = gpa;
666 	pptmmio.len = len;
667 	pptmmio.hpa = hpa;
668 
669 	return (ioctl(ctx->fd, VM_MAP_PPTDEV_MMIO, &pptmmio));
670 }
671 
672 int
673 vm_setup_pptdev_msi(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
674     uint64_t addr, uint64_t msg, int numvec)
675 {
676 	struct vm_pptdev_msi pptmsi;
677 
678 	bzero(&pptmsi, sizeof(pptmsi));
679 	pptmsi.vcpu = vcpu;
680 	pptmsi.bus = bus;
681 	pptmsi.slot = slot;
682 	pptmsi.func = func;
683 	pptmsi.msg = msg;
684 	pptmsi.addr = addr;
685 	pptmsi.numvec = numvec;
686 
687 	return (ioctl(ctx->fd, VM_PPTDEV_MSI, &pptmsi));
688 }
689 
690 int
691 vm_setup_pptdev_msix(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
692     int idx, uint64_t addr, uint64_t msg, uint32_t vector_control)
693 {
694 	struct vm_pptdev_msix pptmsix;
695 
696 	bzero(&pptmsix, sizeof(pptmsix));
697 	pptmsix.vcpu = vcpu;
698 	pptmsix.bus = bus;
699 	pptmsix.slot = slot;
700 	pptmsix.func = func;
701 	pptmsix.idx = idx;
702 	pptmsix.msg = msg;
703 	pptmsix.addr = addr;
704 	pptmsix.vector_control = vector_control;
705 
706 	return ioctl(ctx->fd, VM_PPTDEV_MSIX, &pptmsix);
707 }
708 
709 uint64_t *
710 vm_get_stats(struct vmctx *ctx, int vcpu, struct timeval *ret_tv,
711 	     int *ret_entries)
712 {
713 	int error;
714 
715 	static struct vm_stats vmstats;
716 
717 	vmstats.cpuid = vcpu;
718 
719 	error = ioctl(ctx->fd, VM_STATS, &vmstats);
720 	if (error == 0) {
721 		if (ret_entries)
722 			*ret_entries = vmstats.num_entries;
723 		if (ret_tv)
724 			*ret_tv = vmstats.tv;
725 		return (vmstats.statbuf);
726 	} else
727 		return (NULL);
728 }
729 
730 const char *
731 vm_get_stat_desc(struct vmctx *ctx, int index)
732 {
733 	static struct vm_stat_desc statdesc;
734 
735 	statdesc.index = index;
736 	if (ioctl(ctx->fd, VM_STAT_DESC, &statdesc) == 0)
737 		return (statdesc.desc);
738 	else
739 		return (NULL);
740 }
741 
742 int
743 vm_get_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state *state)
744 {
745 	int error;
746 	struct vm_x2apic x2apic;
747 
748 	bzero(&x2apic, sizeof(x2apic));
749 	x2apic.cpuid = vcpu;
750 
751 	error = ioctl(ctx->fd, VM_GET_X2APIC_STATE, &x2apic);
752 	*state = x2apic.state;
753 	return (error);
754 }
755 
756 int
757 vm_set_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state state)
758 {
759 	int error;
760 	struct vm_x2apic x2apic;
761 
762 	bzero(&x2apic, sizeof(x2apic));
763 	x2apic.cpuid = vcpu;
764 	x2apic.state = state;
765 
766 	error = ioctl(ctx->fd, VM_SET_X2APIC_STATE, &x2apic);
767 
768 	return (error);
769 }
770 
771 /*
772  * From Intel Vol 3a:
773  * Table 9-1. IA-32 Processor States Following Power-up, Reset or INIT
774  */
775 int
776 vcpu_reset(struct vmctx *vmctx, int vcpu)
777 {
778 	int error;
779 	uint64_t rflags, rip, cr0, cr4, zero, desc_base, rdx;
780 	uint32_t desc_access, desc_limit;
781 	uint16_t sel;
782 
783 	zero = 0;
784 
785 	rflags = 0x2;
786 	error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RFLAGS, rflags);
787 	if (error)
788 		goto done;
789 
790 	rip = 0xfff0;
791 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RIP, rip)) != 0)
792 		goto done;
793 
794 	cr0 = CR0_NE;
795 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR0, cr0)) != 0)
796 		goto done;
797 
798 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR3, zero)) != 0)
799 		goto done;
800 
801 	cr4 = 0;
802 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR4, cr4)) != 0)
803 		goto done;
804 
805 	/*
806 	 * CS: present, r/w, accessed, 16-bit, byte granularity, usable
807 	 */
808 	desc_base = 0xffff0000;
809 	desc_limit = 0xffff;
810 	desc_access = 0x0093;
811 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_CS,
812 			    desc_base, desc_limit, desc_access);
813 	if (error)
814 		goto done;
815 
816 	sel = 0xf000;
817 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CS, sel)) != 0)
818 		goto done;
819 
820 	/*
821 	 * SS,DS,ES,FS,GS: present, r/w, accessed, 16-bit, byte granularity
822 	 */
823 	desc_base = 0;
824 	desc_limit = 0xffff;
825 	desc_access = 0x0093;
826 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_SS,
827 			    desc_base, desc_limit, desc_access);
828 	if (error)
829 		goto done;
830 
831 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_DS,
832 			    desc_base, desc_limit, desc_access);
833 	if (error)
834 		goto done;
835 
836 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_ES,
837 			    desc_base, desc_limit, desc_access);
838 	if (error)
839 		goto done;
840 
841 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_FS,
842 			    desc_base, desc_limit, desc_access);
843 	if (error)
844 		goto done;
845 
846 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GS,
847 			    desc_base, desc_limit, desc_access);
848 	if (error)
849 		goto done;
850 
851 	sel = 0;
852 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_SS, sel)) != 0)
853 		goto done;
854 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_DS, sel)) != 0)
855 		goto done;
856 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_ES, sel)) != 0)
857 		goto done;
858 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_FS, sel)) != 0)
859 		goto done;
860 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_GS, sel)) != 0)
861 		goto done;
862 
863 	/* General purpose registers */
864 	rdx = 0xf00;
865 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RAX, zero)) != 0)
866 		goto done;
867 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBX, zero)) != 0)
868 		goto done;
869 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RCX, zero)) != 0)
870 		goto done;
871 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDX, rdx)) != 0)
872 		goto done;
873 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSI, zero)) != 0)
874 		goto done;
875 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDI, zero)) != 0)
876 		goto done;
877 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBP, zero)) != 0)
878 		goto done;
879 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSP, zero)) != 0)
880 		goto done;
881 
882 	/* GDTR, IDTR */
883 	desc_base = 0;
884 	desc_limit = 0xffff;
885 	desc_access = 0;
886 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GDTR,
887 			    desc_base, desc_limit, desc_access);
888 	if (error != 0)
889 		goto done;
890 
891 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_IDTR,
892 			    desc_base, desc_limit, desc_access);
893 	if (error != 0)
894 		goto done;
895 
896 	/* TR */
897 	desc_base = 0;
898 	desc_limit = 0xffff;
899 	desc_access = 0x0000008b;
900 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_TR, 0, 0, desc_access);
901 	if (error)
902 		goto done;
903 
904 	sel = 0;
905 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_TR, sel)) != 0)
906 		goto done;
907 
908 	/* LDTR */
909 	desc_base = 0;
910 	desc_limit = 0xffff;
911 	desc_access = 0x00000082;
912 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_LDTR, desc_base,
913 			    desc_limit, desc_access);
914 	if (error)
915 		goto done;
916 
917 	sel = 0;
918 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_LDTR, 0)) != 0)
919 		goto done;
920 
921 	/* XXX cr2, debug registers */
922 
923 	error = 0;
924 done:
925 	return (error);
926 }
927 
928 int
929 vm_get_gpa_pmap(struct vmctx *ctx, uint64_t gpa, uint64_t *pte, int *num)
930 {
931 	int error, i;
932 	struct vm_gpa_pte gpapte;
933 
934 	bzero(&gpapte, sizeof(gpapte));
935 	gpapte.gpa = gpa;
936 
937 	error = ioctl(ctx->fd, VM_GET_GPA_PMAP, &gpapte);
938 
939 	if (error == 0) {
940 		*num = gpapte.ptenum;
941 		for (i = 0; i < gpapte.ptenum; i++)
942 			pte[i] = gpapte.pte[i];
943 	}
944 
945 	return (error);
946 }
947 
948 int
949 vm_get_hpet_capabilities(struct vmctx *ctx, uint32_t *capabilities)
950 {
951 	int error;
952 	struct vm_hpet_cap cap;
953 
954 	bzero(&cap, sizeof(struct vm_hpet_cap));
955 	error = ioctl(ctx->fd, VM_GET_HPET_CAPABILITIES, &cap);
956 	if (capabilities != NULL)
957 		*capabilities = cap.capabilities;
958 	return (error);
959 }
960 
961 static int
962 gla2gpa(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
963     uint64_t gla, int prot, int *fault, uint64_t *gpa)
964 {
965 	struct vm_gla2gpa gg;
966 	int error;
967 
968 	bzero(&gg, sizeof(struct vm_gla2gpa));
969 	gg.vcpuid = vcpu;
970 	gg.prot = prot;
971 	gg.gla = gla;
972 	gg.paging = *paging;
973 
974 	error = ioctl(ctx->fd, VM_GLA2GPA, &gg);
975 	if (error == 0) {
976 		*fault = gg.fault;
977 		*gpa = gg.gpa;
978 	}
979 	return (error);
980 }
981 
982 #ifndef min
983 #define	min(a,b)	(((a) < (b)) ? (a) : (b))
984 #endif
985 
986 int
987 vm_copy_setup(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
988     uint64_t gla, size_t len, int prot, struct iovec *iov, int iovcnt)
989 {
990 	uint64_t gpa;
991 	int error, fault, i, n, off;
992 
993 	for (i = 0; i < iovcnt; i++) {
994 		iov[i].iov_base = 0;
995 		iov[i].iov_len = 0;
996 	}
997 
998 	while (len) {
999 		assert(iovcnt > 0);
1000 		error = gla2gpa(ctx, vcpu, paging, gla, prot, &fault, &gpa);
1001 		if (error)
1002 			return (-1);
1003 		if (fault)
1004 			return (1);
1005 
1006 		off = gpa & PAGE_MASK;
1007 		n = min(len, PAGE_SIZE - off);
1008 
1009 		iov->iov_base = (void *)gpa;
1010 		iov->iov_len = n;
1011 		iov++;
1012 		iovcnt--;
1013 
1014 		gla += n;
1015 		len -= n;
1016 	}
1017 	return (0);
1018 }
1019 
1020 void
1021 vm_copyin(struct vmctx *ctx, int vcpu, struct iovec *iov, void *vp, size_t len)
1022 {
1023 	const char *src;
1024 	char *dst;
1025 	uint64_t gpa;
1026 	size_t n;
1027 
1028 	dst = vp;
1029 	while (len) {
1030 		assert(iov->iov_len);
1031 		gpa = (uint64_t)iov->iov_base;
1032 		n = min(len, iov->iov_len);
1033 		src = vm_map_gpa(ctx, gpa, n);
1034 		bcopy(src, dst, n);
1035 
1036 		iov++;
1037 		dst += n;
1038 		len -= n;
1039 	}
1040 }
1041 
1042 void
1043 vm_copyout(struct vmctx *ctx, int vcpu, const void *vp, struct iovec *iov,
1044     size_t len)
1045 {
1046 	const char *src;
1047 	char *dst;
1048 	uint64_t gpa;
1049 	size_t n;
1050 
1051 	src = vp;
1052 	while (len) {
1053 		assert(iov->iov_len);
1054 		gpa = (uint64_t)iov->iov_base;
1055 		n = min(len, iov->iov_len);
1056 		dst = vm_map_gpa(ctx, gpa, n);
1057 		bcopy(src, dst, n);
1058 
1059 		iov++;
1060 		src += n;
1061 		len -= n;
1062 	}
1063 }
1064 
1065 static int
1066 vm_get_cpus(struct vmctx *ctx, int which, cpuset_t *cpus)
1067 {
1068 	struct vm_cpuset vm_cpuset;
1069 	int error;
1070 
1071 	bzero(&vm_cpuset, sizeof(struct vm_cpuset));
1072 	vm_cpuset.which = which;
1073 	vm_cpuset.cpusetsize = sizeof(cpuset_t);
1074 	vm_cpuset.cpus = cpus;
1075 
1076 	error = ioctl(ctx->fd, VM_GET_CPUS, &vm_cpuset);
1077 	return (error);
1078 }
1079 
1080 int
1081 vm_active_cpus(struct vmctx *ctx, cpuset_t *cpus)
1082 {
1083 
1084 	return (vm_get_cpus(ctx, VM_ACTIVE_CPUS, cpus));
1085 }
1086 
1087 int
1088 vm_suspended_cpus(struct vmctx *ctx, cpuset_t *cpus)
1089 {
1090 
1091 	return (vm_get_cpus(ctx, VM_SUSPENDED_CPUS, cpus));
1092 }
1093 
1094 int
1095 vm_activate_cpu(struct vmctx *ctx, int vcpu)
1096 {
1097 	struct vm_activate_cpu ac;
1098 	int error;
1099 
1100 	bzero(&ac, sizeof(struct vm_activate_cpu));
1101 	ac.vcpuid = vcpu;
1102 	error = ioctl(ctx->fd, VM_ACTIVATE_CPU, &ac);
1103 	return (error);
1104 }
1105 
1106 int
1107 vm_get_intinfo(struct vmctx *ctx, int vcpu, uint64_t *info1, uint64_t *info2)
1108 {
1109 	struct vm_intinfo vmii;
1110 	int error;
1111 
1112 	bzero(&vmii, sizeof(struct vm_intinfo));
1113 	vmii.vcpuid = vcpu;
1114 	error = ioctl(ctx->fd, VM_GET_INTINFO, &vmii);
1115 	if (error == 0) {
1116 		*info1 = vmii.info1;
1117 		*info2 = vmii.info2;
1118 	}
1119 	return (error);
1120 }
1121 
1122 int
1123 vm_set_intinfo(struct vmctx *ctx, int vcpu, uint64_t info1)
1124 {
1125 	struct vm_intinfo vmii;
1126 	int error;
1127 
1128 	bzero(&vmii, sizeof(struct vm_intinfo));
1129 	vmii.vcpuid = vcpu;
1130 	vmii.info1 = info1;
1131 	error = ioctl(ctx->fd, VM_SET_INTINFO, &vmii);
1132 	return (error);
1133 }
1134 
1135 int
1136 vm_rtc_write(struct vmctx *ctx, int offset, uint8_t value)
1137 {
1138 	struct vm_rtc_data rtcdata;
1139 	int error;
1140 
1141 	bzero(&rtcdata, sizeof(struct vm_rtc_data));
1142 	rtcdata.offset = offset;
1143 	rtcdata.value = value;
1144 	error = ioctl(ctx->fd, VM_RTC_WRITE, &rtcdata);
1145 	return (error);
1146 }
1147 
1148 int
1149 vm_rtc_read(struct vmctx *ctx, int offset, uint8_t *retval)
1150 {
1151 	struct vm_rtc_data rtcdata;
1152 	int error;
1153 
1154 	bzero(&rtcdata, sizeof(struct vm_rtc_data));
1155 	rtcdata.offset = offset;
1156 	error = ioctl(ctx->fd, VM_RTC_READ, &rtcdata);
1157 	if (error == 0)
1158 		*retval = rtcdata.value;
1159 	return (error);
1160 }
1161 
1162 int
1163 vm_rtc_settime(struct vmctx *ctx, time_t secs)
1164 {
1165 	struct vm_rtc_time rtctime;
1166 	int error;
1167 
1168 	bzero(&rtctime, sizeof(struct vm_rtc_time));
1169 	rtctime.secs = secs;
1170 	error = ioctl(ctx->fd, VM_RTC_SETTIME, &rtctime);
1171 	return (error);
1172 }
1173 
1174 int
1175 vm_rtc_gettime(struct vmctx *ctx, time_t *secs)
1176 {
1177 	struct vm_rtc_time rtctime;
1178 	int error;
1179 
1180 	bzero(&rtctime, sizeof(struct vm_rtc_time));
1181 	error = ioctl(ctx->fd, VM_RTC_GETTIME, &rtctime);
1182 	if (error == 0)
1183 		*secs = rtctime.secs;
1184 	return (error);
1185 }
1186 
1187 int
1188 vm_restart_instruction(void *arg, int vcpu)
1189 {
1190 	struct vmctx *ctx = arg;
1191 
1192 	return (ioctl(ctx->fd, VM_RESTART_INSTRUCTION, &vcpu));
1193 }
1194