xref: /freebsd/lib/libvmmapi/vmmapi.c (revision a812392203d7c4c3f0db9d8a0f3391374c49c71f)
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, uint64_t rip, struct vm_exit *vmexit)
372 {
373 	int error;
374 	struct vm_run vmrun;
375 
376 	bzero(&vmrun, sizeof(vmrun));
377 	vmrun.cpuid = vcpu;
378 	vmrun.rip = rip;
379 
380 	error = ioctl(ctx->fd, VM_RUN, &vmrun);
381 	bcopy(&vmrun.vm_exit, vmexit, sizeof(struct vm_exit));
382 	return (error);
383 }
384 
385 int
386 vm_suspend(struct vmctx *ctx, enum vm_suspend_how how)
387 {
388 	struct vm_suspend vmsuspend;
389 
390 	bzero(&vmsuspend, sizeof(vmsuspend));
391 	vmsuspend.how = how;
392 	return (ioctl(ctx->fd, VM_SUSPEND, &vmsuspend));
393 }
394 
395 int
396 vm_reinit(struct vmctx *ctx)
397 {
398 
399 	return (ioctl(ctx->fd, VM_REINIT, 0));
400 }
401 
402 static int
403 vm_inject_exception_real(struct vmctx *ctx, int vcpu, int vector,
404     int error_code, int error_code_valid)
405 {
406 	struct vm_exception exc;
407 
408 	bzero(&exc, sizeof(exc));
409 	exc.cpuid = vcpu;
410 	exc.vector = vector;
411 	exc.error_code = error_code;
412 	exc.error_code_valid = error_code_valid;
413 
414 	return (ioctl(ctx->fd, VM_INJECT_EXCEPTION, &exc));
415 }
416 
417 int
418 vm_inject_exception(struct vmctx *ctx, int vcpu, int vector)
419 {
420 
421 	return (vm_inject_exception_real(ctx, vcpu, vector, 0, 0));
422 }
423 
424 int
425 vm_inject_exception2(struct vmctx *ctx, int vcpu, int vector, int errcode)
426 {
427 
428 	return (vm_inject_exception_real(ctx, vcpu, vector, errcode, 1));
429 }
430 
431 int
432 vm_apicid2vcpu(struct vmctx *ctx, int apicid)
433 {
434 	/*
435 	 * The apic id associated with the 'vcpu' has the same numerical value
436 	 * as the 'vcpu' itself.
437 	 */
438 	return (apicid);
439 }
440 
441 int
442 vm_lapic_irq(struct vmctx *ctx, int vcpu, int vector)
443 {
444 	struct vm_lapic_irq vmirq;
445 
446 	bzero(&vmirq, sizeof(vmirq));
447 	vmirq.cpuid = vcpu;
448 	vmirq.vector = vector;
449 
450 	return (ioctl(ctx->fd, VM_LAPIC_IRQ, &vmirq));
451 }
452 
453 int
454 vm_lapic_local_irq(struct vmctx *ctx, int vcpu, int vector)
455 {
456 	struct vm_lapic_irq vmirq;
457 
458 	bzero(&vmirq, sizeof(vmirq));
459 	vmirq.cpuid = vcpu;
460 	vmirq.vector = vector;
461 
462 	return (ioctl(ctx->fd, VM_LAPIC_LOCAL_IRQ, &vmirq));
463 }
464 
465 int
466 vm_lapic_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg)
467 {
468 	struct vm_lapic_msi vmmsi;
469 
470 	bzero(&vmmsi, sizeof(vmmsi));
471 	vmmsi.addr = addr;
472 	vmmsi.msg = msg;
473 
474 	return (ioctl(ctx->fd, VM_LAPIC_MSI, &vmmsi));
475 }
476 
477 int
478 vm_ioapic_assert_irq(struct vmctx *ctx, int irq)
479 {
480 	struct vm_ioapic_irq ioapic_irq;
481 
482 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
483 	ioapic_irq.irq = irq;
484 
485 	return (ioctl(ctx->fd, VM_IOAPIC_ASSERT_IRQ, &ioapic_irq));
486 }
487 
488 int
489 vm_ioapic_deassert_irq(struct vmctx *ctx, int irq)
490 {
491 	struct vm_ioapic_irq ioapic_irq;
492 
493 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
494 	ioapic_irq.irq = irq;
495 
496 	return (ioctl(ctx->fd, VM_IOAPIC_DEASSERT_IRQ, &ioapic_irq));
497 }
498 
499 int
500 vm_ioapic_pulse_irq(struct vmctx *ctx, int irq)
501 {
502 	struct vm_ioapic_irq ioapic_irq;
503 
504 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
505 	ioapic_irq.irq = irq;
506 
507 	return (ioctl(ctx->fd, VM_IOAPIC_PULSE_IRQ, &ioapic_irq));
508 }
509 
510 int
511 vm_ioapic_pincount(struct vmctx *ctx, int *pincount)
512 {
513 
514 	return (ioctl(ctx->fd, VM_IOAPIC_PINCOUNT, pincount));
515 }
516 
517 int
518 vm_isa_assert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
519 {
520 	struct vm_isa_irq isa_irq;
521 
522 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
523 	isa_irq.atpic_irq = atpic_irq;
524 	isa_irq.ioapic_irq = ioapic_irq;
525 
526 	return (ioctl(ctx->fd, VM_ISA_ASSERT_IRQ, &isa_irq));
527 }
528 
529 int
530 vm_isa_deassert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
531 {
532 	struct vm_isa_irq isa_irq;
533 
534 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
535 	isa_irq.atpic_irq = atpic_irq;
536 	isa_irq.ioapic_irq = ioapic_irq;
537 
538 	return (ioctl(ctx->fd, VM_ISA_DEASSERT_IRQ, &isa_irq));
539 }
540 
541 int
542 vm_isa_pulse_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
543 {
544 	struct vm_isa_irq isa_irq;
545 
546 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
547 	isa_irq.atpic_irq = atpic_irq;
548 	isa_irq.ioapic_irq = ioapic_irq;
549 
550 	return (ioctl(ctx->fd, VM_ISA_PULSE_IRQ, &isa_irq));
551 }
552 
553 int
554 vm_isa_set_irq_trigger(struct vmctx *ctx, int atpic_irq,
555     enum vm_intr_trigger trigger)
556 {
557 	struct vm_isa_irq_trigger isa_irq_trigger;
558 
559 	bzero(&isa_irq_trigger, sizeof(struct vm_isa_irq_trigger));
560 	isa_irq_trigger.atpic_irq = atpic_irq;
561 	isa_irq_trigger.trigger = trigger;
562 
563 	return (ioctl(ctx->fd, VM_ISA_SET_IRQ_TRIGGER, &isa_irq_trigger));
564 }
565 
566 int
567 vm_inject_nmi(struct vmctx *ctx, int vcpu)
568 {
569 	struct vm_nmi vmnmi;
570 
571 	bzero(&vmnmi, sizeof(vmnmi));
572 	vmnmi.cpuid = vcpu;
573 
574 	return (ioctl(ctx->fd, VM_INJECT_NMI, &vmnmi));
575 }
576 
577 static struct {
578 	const char	*name;
579 	int		type;
580 } capstrmap[] = {
581 	{ "hlt_exit",		VM_CAP_HALT_EXIT },
582 	{ "mtrap_exit",		VM_CAP_MTRAP_EXIT },
583 	{ "pause_exit",		VM_CAP_PAUSE_EXIT },
584 	{ "unrestricted_guest",	VM_CAP_UNRESTRICTED_GUEST },
585 	{ "enable_invpcid",	VM_CAP_ENABLE_INVPCID },
586 	{ 0 }
587 };
588 
589 int
590 vm_capability_name2type(const char *capname)
591 {
592 	int i;
593 
594 	for (i = 0; capstrmap[i].name != NULL && capname != NULL; i++) {
595 		if (strcmp(capstrmap[i].name, capname) == 0)
596 			return (capstrmap[i].type);
597 	}
598 
599 	return (-1);
600 }
601 
602 const char *
603 vm_capability_type2name(int type)
604 {
605 	int i;
606 
607 	for (i = 0; capstrmap[i].name != NULL; i++) {
608 		if (capstrmap[i].type == type)
609 			return (capstrmap[i].name);
610 	}
611 
612 	return (NULL);
613 }
614 
615 int
616 vm_get_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap,
617 		  int *retval)
618 {
619 	int error;
620 	struct vm_capability vmcap;
621 
622 	bzero(&vmcap, sizeof(vmcap));
623 	vmcap.cpuid = vcpu;
624 	vmcap.captype = cap;
625 
626 	error = ioctl(ctx->fd, VM_GET_CAPABILITY, &vmcap);
627 	*retval = vmcap.capval;
628 	return (error);
629 }
630 
631 int
632 vm_set_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap, int val)
633 {
634 	struct vm_capability vmcap;
635 
636 	bzero(&vmcap, sizeof(vmcap));
637 	vmcap.cpuid = vcpu;
638 	vmcap.captype = cap;
639 	vmcap.capval = val;
640 
641 	return (ioctl(ctx->fd, VM_SET_CAPABILITY, &vmcap));
642 }
643 
644 int
645 vm_assign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
646 {
647 	struct vm_pptdev pptdev;
648 
649 	bzero(&pptdev, sizeof(pptdev));
650 	pptdev.bus = bus;
651 	pptdev.slot = slot;
652 	pptdev.func = func;
653 
654 	return (ioctl(ctx->fd, VM_BIND_PPTDEV, &pptdev));
655 }
656 
657 int
658 vm_unassign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
659 {
660 	struct vm_pptdev pptdev;
661 
662 	bzero(&pptdev, sizeof(pptdev));
663 	pptdev.bus = bus;
664 	pptdev.slot = slot;
665 	pptdev.func = func;
666 
667 	return (ioctl(ctx->fd, VM_UNBIND_PPTDEV, &pptdev));
668 }
669 
670 int
671 vm_map_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func,
672 		   vm_paddr_t gpa, size_t len, vm_paddr_t hpa)
673 {
674 	struct vm_pptdev_mmio pptmmio;
675 
676 	bzero(&pptmmio, sizeof(pptmmio));
677 	pptmmio.bus = bus;
678 	pptmmio.slot = slot;
679 	pptmmio.func = func;
680 	pptmmio.gpa = gpa;
681 	pptmmio.len = len;
682 	pptmmio.hpa = hpa;
683 
684 	return (ioctl(ctx->fd, VM_MAP_PPTDEV_MMIO, &pptmmio));
685 }
686 
687 int
688 vm_setup_pptdev_msi(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
689     uint64_t addr, uint64_t msg, int numvec)
690 {
691 	struct vm_pptdev_msi pptmsi;
692 
693 	bzero(&pptmsi, sizeof(pptmsi));
694 	pptmsi.vcpu = vcpu;
695 	pptmsi.bus = bus;
696 	pptmsi.slot = slot;
697 	pptmsi.func = func;
698 	pptmsi.msg = msg;
699 	pptmsi.addr = addr;
700 	pptmsi.numvec = numvec;
701 
702 	return (ioctl(ctx->fd, VM_PPTDEV_MSI, &pptmsi));
703 }
704 
705 int
706 vm_setup_pptdev_msix(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
707     int idx, uint64_t addr, uint64_t msg, uint32_t vector_control)
708 {
709 	struct vm_pptdev_msix pptmsix;
710 
711 	bzero(&pptmsix, sizeof(pptmsix));
712 	pptmsix.vcpu = vcpu;
713 	pptmsix.bus = bus;
714 	pptmsix.slot = slot;
715 	pptmsix.func = func;
716 	pptmsix.idx = idx;
717 	pptmsix.msg = msg;
718 	pptmsix.addr = addr;
719 	pptmsix.vector_control = vector_control;
720 
721 	return ioctl(ctx->fd, VM_PPTDEV_MSIX, &pptmsix);
722 }
723 
724 uint64_t *
725 vm_get_stats(struct vmctx *ctx, int vcpu, struct timeval *ret_tv,
726 	     int *ret_entries)
727 {
728 	int error;
729 
730 	static struct vm_stats vmstats;
731 
732 	vmstats.cpuid = vcpu;
733 
734 	error = ioctl(ctx->fd, VM_STATS, &vmstats);
735 	if (error == 0) {
736 		if (ret_entries)
737 			*ret_entries = vmstats.num_entries;
738 		if (ret_tv)
739 			*ret_tv = vmstats.tv;
740 		return (vmstats.statbuf);
741 	} else
742 		return (NULL);
743 }
744 
745 const char *
746 vm_get_stat_desc(struct vmctx *ctx, int index)
747 {
748 	static struct vm_stat_desc statdesc;
749 
750 	statdesc.index = index;
751 	if (ioctl(ctx->fd, VM_STAT_DESC, &statdesc) == 0)
752 		return (statdesc.desc);
753 	else
754 		return (NULL);
755 }
756 
757 int
758 vm_get_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state *state)
759 {
760 	int error;
761 	struct vm_x2apic x2apic;
762 
763 	bzero(&x2apic, sizeof(x2apic));
764 	x2apic.cpuid = vcpu;
765 
766 	error = ioctl(ctx->fd, VM_GET_X2APIC_STATE, &x2apic);
767 	*state = x2apic.state;
768 	return (error);
769 }
770 
771 int
772 vm_set_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state state)
773 {
774 	int error;
775 	struct vm_x2apic x2apic;
776 
777 	bzero(&x2apic, sizeof(x2apic));
778 	x2apic.cpuid = vcpu;
779 	x2apic.state = state;
780 
781 	error = ioctl(ctx->fd, VM_SET_X2APIC_STATE, &x2apic);
782 
783 	return (error);
784 }
785 
786 /*
787  * From Intel Vol 3a:
788  * Table 9-1. IA-32 Processor States Following Power-up, Reset or INIT
789  */
790 int
791 vcpu_reset(struct vmctx *vmctx, int vcpu)
792 {
793 	int error;
794 	uint64_t rflags, rip, cr0, cr4, zero, desc_base, rdx;
795 	uint32_t desc_access, desc_limit;
796 	uint16_t sel;
797 
798 	zero = 0;
799 
800 	rflags = 0x2;
801 	error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RFLAGS, rflags);
802 	if (error)
803 		goto done;
804 
805 	rip = 0xfff0;
806 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RIP, rip)) != 0)
807 		goto done;
808 
809 	cr0 = CR0_NE;
810 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR0, cr0)) != 0)
811 		goto done;
812 
813 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR3, zero)) != 0)
814 		goto done;
815 
816 	cr4 = 0;
817 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR4, cr4)) != 0)
818 		goto done;
819 
820 	/*
821 	 * CS: present, r/w, accessed, 16-bit, byte granularity, usable
822 	 */
823 	desc_base = 0xffff0000;
824 	desc_limit = 0xffff;
825 	desc_access = 0x0093;
826 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_CS,
827 			    desc_base, desc_limit, desc_access);
828 	if (error)
829 		goto done;
830 
831 	sel = 0xf000;
832 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CS, sel)) != 0)
833 		goto done;
834 
835 	/*
836 	 * SS,DS,ES,FS,GS: present, r/w, accessed, 16-bit, byte granularity
837 	 */
838 	desc_base = 0;
839 	desc_limit = 0xffff;
840 	desc_access = 0x0093;
841 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_SS,
842 			    desc_base, desc_limit, desc_access);
843 	if (error)
844 		goto done;
845 
846 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_DS,
847 			    desc_base, desc_limit, desc_access);
848 	if (error)
849 		goto done;
850 
851 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_ES,
852 			    desc_base, desc_limit, desc_access);
853 	if (error)
854 		goto done;
855 
856 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_FS,
857 			    desc_base, desc_limit, desc_access);
858 	if (error)
859 		goto done;
860 
861 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GS,
862 			    desc_base, desc_limit, desc_access);
863 	if (error)
864 		goto done;
865 
866 	sel = 0;
867 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_SS, sel)) != 0)
868 		goto done;
869 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_DS, sel)) != 0)
870 		goto done;
871 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_ES, sel)) != 0)
872 		goto done;
873 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_FS, sel)) != 0)
874 		goto done;
875 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_GS, sel)) != 0)
876 		goto done;
877 
878 	/* General purpose registers */
879 	rdx = 0xf00;
880 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RAX, zero)) != 0)
881 		goto done;
882 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBX, zero)) != 0)
883 		goto done;
884 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RCX, zero)) != 0)
885 		goto done;
886 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDX, rdx)) != 0)
887 		goto done;
888 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSI, zero)) != 0)
889 		goto done;
890 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDI, zero)) != 0)
891 		goto done;
892 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBP, zero)) != 0)
893 		goto done;
894 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSP, zero)) != 0)
895 		goto done;
896 
897 	/* GDTR, IDTR */
898 	desc_base = 0;
899 	desc_limit = 0xffff;
900 	desc_access = 0;
901 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GDTR,
902 			    desc_base, desc_limit, desc_access);
903 	if (error != 0)
904 		goto done;
905 
906 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_IDTR,
907 			    desc_base, desc_limit, desc_access);
908 	if (error != 0)
909 		goto done;
910 
911 	/* TR */
912 	desc_base = 0;
913 	desc_limit = 0xffff;
914 	desc_access = 0x0000008b;
915 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_TR, 0, 0, desc_access);
916 	if (error)
917 		goto done;
918 
919 	sel = 0;
920 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_TR, sel)) != 0)
921 		goto done;
922 
923 	/* LDTR */
924 	desc_base = 0;
925 	desc_limit = 0xffff;
926 	desc_access = 0x00000082;
927 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_LDTR, desc_base,
928 			    desc_limit, desc_access);
929 	if (error)
930 		goto done;
931 
932 	sel = 0;
933 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_LDTR, 0)) != 0)
934 		goto done;
935 
936 	/* XXX cr2, debug registers */
937 
938 	error = 0;
939 done:
940 	return (error);
941 }
942 
943 int
944 vm_get_gpa_pmap(struct vmctx *ctx, uint64_t gpa, uint64_t *pte, int *num)
945 {
946 	int error, i;
947 	struct vm_gpa_pte gpapte;
948 
949 	bzero(&gpapte, sizeof(gpapte));
950 	gpapte.gpa = gpa;
951 
952 	error = ioctl(ctx->fd, VM_GET_GPA_PMAP, &gpapte);
953 
954 	if (error == 0) {
955 		*num = gpapte.ptenum;
956 		for (i = 0; i < gpapte.ptenum; i++)
957 			pte[i] = gpapte.pte[i];
958 	}
959 
960 	return (error);
961 }
962 
963 int
964 vm_get_hpet_capabilities(struct vmctx *ctx, uint32_t *capabilities)
965 {
966 	int error;
967 	struct vm_hpet_cap cap;
968 
969 	bzero(&cap, sizeof(struct vm_hpet_cap));
970 	error = ioctl(ctx->fd, VM_GET_HPET_CAPABILITIES, &cap);
971 	if (capabilities != NULL)
972 		*capabilities = cap.capabilities;
973 	return (error);
974 }
975 
976 static int
977 gla2gpa(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
978     uint64_t gla, int prot, int *fault, uint64_t *gpa)
979 {
980 	struct vm_gla2gpa gg;
981 	int error;
982 
983 	bzero(&gg, sizeof(struct vm_gla2gpa));
984 	gg.vcpuid = vcpu;
985 	gg.prot = prot;
986 	gg.gla = gla;
987 	gg.paging = *paging;
988 
989 	error = ioctl(ctx->fd, VM_GLA2GPA, &gg);
990 	if (error == 0) {
991 		*fault = gg.fault;
992 		*gpa = gg.gpa;
993 	}
994 	return (error);
995 }
996 
997 #ifndef min
998 #define	min(a,b)	(((a) < (b)) ? (a) : (b))
999 #endif
1000 
1001 int
1002 vm_copy_setup(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1003     uint64_t gla, size_t len, int prot, struct iovec *iov, int iovcnt)
1004 {
1005 	uint64_t gpa;
1006 	int error, fault, i, n, off;
1007 
1008 	for (i = 0; i < iovcnt; i++) {
1009 		iov[i].iov_base = 0;
1010 		iov[i].iov_len = 0;
1011 	}
1012 
1013 	while (len) {
1014 		assert(iovcnt > 0);
1015 		error = gla2gpa(ctx, vcpu, paging, gla, prot, &fault, &gpa);
1016 		if (error)
1017 			return (-1);
1018 		if (fault)
1019 			return (1);
1020 
1021 		off = gpa & PAGE_MASK;
1022 		n = min(len, PAGE_SIZE - off);
1023 
1024 		iov->iov_base = (void *)gpa;
1025 		iov->iov_len = n;
1026 		iov++;
1027 		iovcnt--;
1028 
1029 		gla += n;
1030 		len -= n;
1031 	}
1032 	return (0);
1033 }
1034 
1035 void
1036 vm_copyin(struct vmctx *ctx, int vcpu, struct iovec *iov, void *vp, size_t len)
1037 {
1038 	const char *src;
1039 	char *dst;
1040 	uint64_t gpa;
1041 	size_t n;
1042 
1043 	dst = vp;
1044 	while (len) {
1045 		assert(iov->iov_len);
1046 		gpa = (uint64_t)iov->iov_base;
1047 		n = min(len, iov->iov_len);
1048 		src = vm_map_gpa(ctx, gpa, n);
1049 		bcopy(src, dst, n);
1050 
1051 		iov++;
1052 		dst += n;
1053 		len -= n;
1054 	}
1055 }
1056 
1057 void
1058 vm_copyout(struct vmctx *ctx, int vcpu, const void *vp, struct iovec *iov,
1059     size_t len)
1060 {
1061 	const char *src;
1062 	char *dst;
1063 	uint64_t gpa;
1064 	size_t n;
1065 
1066 	src = vp;
1067 	while (len) {
1068 		assert(iov->iov_len);
1069 		gpa = (uint64_t)iov->iov_base;
1070 		n = min(len, iov->iov_len);
1071 		dst = vm_map_gpa(ctx, gpa, n);
1072 		bcopy(src, dst, n);
1073 
1074 		iov++;
1075 		src += n;
1076 		len -= n;
1077 	}
1078 }
1079 
1080 static int
1081 vm_get_cpus(struct vmctx *ctx, int which, cpuset_t *cpus)
1082 {
1083 	struct vm_cpuset vm_cpuset;
1084 	int error;
1085 
1086 	bzero(&vm_cpuset, sizeof(struct vm_cpuset));
1087 	vm_cpuset.which = which;
1088 	vm_cpuset.cpusetsize = sizeof(cpuset_t);
1089 	vm_cpuset.cpus = cpus;
1090 
1091 	error = ioctl(ctx->fd, VM_GET_CPUS, &vm_cpuset);
1092 	return (error);
1093 }
1094 
1095 int
1096 vm_active_cpus(struct vmctx *ctx, cpuset_t *cpus)
1097 {
1098 
1099 	return (vm_get_cpus(ctx, VM_ACTIVE_CPUS, cpus));
1100 }
1101 
1102 int
1103 vm_suspended_cpus(struct vmctx *ctx, cpuset_t *cpus)
1104 {
1105 
1106 	return (vm_get_cpus(ctx, VM_SUSPENDED_CPUS, cpus));
1107 }
1108 
1109 int
1110 vm_activate_cpu(struct vmctx *ctx, int vcpu)
1111 {
1112 	struct vm_activate_cpu ac;
1113 	int error;
1114 
1115 	bzero(&ac, sizeof(struct vm_activate_cpu));
1116 	ac.vcpuid = vcpu;
1117 	error = ioctl(ctx->fd, VM_ACTIVATE_CPU, &ac);
1118 	return (error);
1119 }
1120 
1121 int
1122 vm_get_intinfo(struct vmctx *ctx, int vcpu, uint64_t *info1, uint64_t *info2)
1123 {
1124 	struct vm_intinfo vmii;
1125 	int error;
1126 
1127 	bzero(&vmii, sizeof(struct vm_intinfo));
1128 	vmii.vcpuid = vcpu;
1129 	error = ioctl(ctx->fd, VM_GET_INTINFO, &vmii);
1130 	if (error == 0) {
1131 		*info1 = vmii.info1;
1132 		*info2 = vmii.info2;
1133 	}
1134 	return (error);
1135 }
1136 
1137 int
1138 vm_set_intinfo(struct vmctx *ctx, int vcpu, uint64_t info1)
1139 {
1140 	struct vm_intinfo vmii;
1141 	int error;
1142 
1143 	bzero(&vmii, sizeof(struct vm_intinfo));
1144 	vmii.vcpuid = vcpu;
1145 	vmii.info1 = info1;
1146 	error = ioctl(ctx->fd, VM_SET_INTINFO, &vmii);
1147 	return (error);
1148 }
1149 
1150 int
1151 vm_rtc_write(struct vmctx *ctx, int offset, uint8_t value)
1152 {
1153 	struct vm_rtc_data rtcdata;
1154 	int error;
1155 
1156 	bzero(&rtcdata, sizeof(struct vm_rtc_data));
1157 	rtcdata.offset = offset;
1158 	rtcdata.value = value;
1159 	error = ioctl(ctx->fd, VM_RTC_WRITE, &rtcdata);
1160 	return (error);
1161 }
1162 
1163 int
1164 vm_rtc_read(struct vmctx *ctx, int offset, uint8_t *retval)
1165 {
1166 	struct vm_rtc_data rtcdata;
1167 	int error;
1168 
1169 	bzero(&rtcdata, sizeof(struct vm_rtc_data));
1170 	rtcdata.offset = offset;
1171 	error = ioctl(ctx->fd, VM_RTC_READ, &rtcdata);
1172 	if (error == 0)
1173 		*retval = rtcdata.value;
1174 	return (error);
1175 }
1176 
1177 int
1178 vm_rtc_settime(struct vmctx *ctx, time_t secs)
1179 {
1180 	struct vm_rtc_time rtctime;
1181 	int error;
1182 
1183 	bzero(&rtctime, sizeof(struct vm_rtc_time));
1184 	rtctime.secs = secs;
1185 	error = ioctl(ctx->fd, VM_RTC_SETTIME, &rtctime);
1186 	return (error);
1187 }
1188 
1189 int
1190 vm_rtc_gettime(struct vmctx *ctx, time_t *secs)
1191 {
1192 	struct vm_rtc_time rtctime;
1193 	int error;
1194 
1195 	bzero(&rtctime, sizeof(struct vm_rtc_time));
1196 	error = ioctl(ctx->fd, VM_RTC_GETTIME, &rtctime);
1197 	if (error == 0)
1198 		*secs = rtctime.secs;
1199 	return (error);
1200 }
1201