xref: /freebsd/lib/libvmmapi/vmmapi.c (revision 583d27945d443d7b7df576f8a28557f894268ef7)
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/types.h>
33 #include <sys/sysctl.h>
34 #include <sys/ioctl.h>
35 #include <sys/mman.h>
36 #include <sys/_iovec.h>
37 
38 #include <machine/specialreg.h>
39 #include <machine/param.h>
40 
41 #include <stdio.h>
42 #include <stdlib.h>
43 #include <assert.h>
44 #include <string.h>
45 #include <fcntl.h>
46 #include <unistd.h>
47 
48 #include <libutil.h>
49 
50 #include <machine/vmm.h>
51 #include <machine/vmm_dev.h>
52 
53 #include "vmmapi.h"
54 
55 #define	MB	(1024 * 1024UL)
56 #define	GB	(1024 * 1024 * 1024UL)
57 
58 struct vmctx {
59 	int	fd;
60 	uint32_t lowmem_limit;
61 	enum vm_mmap_style vms;
62 	int	memflags;
63 	size_t	lowmem;
64 	char	*lowmem_addr;
65 	size_t	highmem;
66 	char	*highmem_addr;
67 	char	*name;
68 };
69 
70 #define	CREATE(x)  sysctlbyname("hw.vmm.create", NULL, NULL, (x), strlen((x)))
71 #define	DESTROY(x) sysctlbyname("hw.vmm.destroy", NULL, NULL, (x), strlen((x)))
72 
73 static int
74 vm_device_open(const char *name)
75 {
76         int fd, len;
77         char *vmfile;
78 
79 	len = strlen("/dev/vmm/") + strlen(name) + 1;
80 	vmfile = malloc(len);
81 	assert(vmfile != NULL);
82 	snprintf(vmfile, len, "/dev/vmm/%s", name);
83 
84         /* Open the device file */
85         fd = open(vmfile, O_RDWR, 0);
86 
87 	free(vmfile);
88         return (fd);
89 }
90 
91 int
92 vm_create(const char *name)
93 {
94 
95 	return (CREATE((char *)name));
96 }
97 
98 struct vmctx *
99 vm_open(const char *name)
100 {
101 	struct vmctx *vm;
102 
103 	vm = malloc(sizeof(struct vmctx) + strlen(name) + 1);
104 	assert(vm != NULL);
105 
106 	vm->fd = -1;
107 	vm->memflags = 0;
108 	vm->lowmem_limit = 3 * GB;
109 	vm->name = (char *)(vm + 1);
110 	strcpy(vm->name, name);
111 
112 	if ((vm->fd = vm_device_open(vm->name)) < 0)
113 		goto err;
114 
115 	return (vm);
116 err:
117 	vm_destroy(vm);
118 	return (NULL);
119 }
120 
121 void
122 vm_destroy(struct vmctx *vm)
123 {
124 	assert(vm != NULL);
125 
126 	if (vm->fd >= 0)
127 		close(vm->fd);
128 	DESTROY(vm->name);
129 
130 	free(vm);
131 }
132 
133 int
134 vm_parse_memsize(const char *optarg, size_t *ret_memsize)
135 {
136 	char *endptr;
137 	size_t optval;
138 	int error;
139 
140 	optval = strtoul(optarg, &endptr, 0);
141 	if (*optarg != '\0' && *endptr == '\0') {
142 		/*
143 		 * For the sake of backward compatibility if the memory size
144 		 * specified on the command line is less than a megabyte then
145 		 * it is interpreted as being in units of MB.
146 		 */
147 		if (optval < MB)
148 			optval *= MB;
149 		*ret_memsize = optval;
150 		error = 0;
151 	} else
152 		error = expand_number(optarg, ret_memsize);
153 
154 	return (error);
155 }
156 
157 int
158 vm_get_memory_seg(struct vmctx *ctx, vm_paddr_t gpa, size_t *ret_len,
159 		  int *wired)
160 {
161 	int error;
162 	struct vm_memory_segment seg;
163 
164 	bzero(&seg, sizeof(seg));
165 	seg.gpa = gpa;
166 	error = ioctl(ctx->fd, VM_GET_MEMORY_SEG, &seg);
167 	*ret_len = seg.len;
168 	if (wired != NULL)
169 		*wired = seg.wired;
170 	return (error);
171 }
172 
173 uint32_t
174 vm_get_lowmem_limit(struct vmctx *ctx)
175 {
176 
177 	return (ctx->lowmem_limit);
178 }
179 
180 void
181 vm_set_lowmem_limit(struct vmctx *ctx, uint32_t limit)
182 {
183 
184 	ctx->lowmem_limit = limit;
185 }
186 
187 void
188 vm_set_memflags(struct vmctx *ctx, int flags)
189 {
190 
191 	ctx->memflags = flags;
192 }
193 
194 static int
195 setup_memory_segment(struct vmctx *ctx, vm_paddr_t gpa, size_t len, char **addr)
196 {
197 	int error, mmap_flags;
198 	struct vm_memory_segment seg;
199 
200 	/*
201 	 * Create and optionally map 'len' bytes of memory at guest
202 	 * physical address 'gpa'
203 	 */
204 	bzero(&seg, sizeof(seg));
205 	seg.gpa = gpa;
206 	seg.len = len;
207 	error = ioctl(ctx->fd, VM_MAP_MEMORY, &seg);
208 	if (error == 0 && addr != NULL) {
209 		mmap_flags = MAP_SHARED;
210 		if ((ctx->memflags & VM_MEM_F_INCORE) == 0)
211 			mmap_flags |= MAP_NOCORE;
212 		*addr = mmap(NULL, len, PROT_READ | PROT_WRITE, mmap_flags,
213 		    ctx->fd, gpa);
214 	}
215 	return (error);
216 }
217 
218 int
219 vm_setup_memory(struct vmctx *ctx, size_t memsize, enum vm_mmap_style vms)
220 {
221 	char **addr;
222 	int error;
223 
224 	/* XXX VM_MMAP_SPARSE not implemented yet */
225 	assert(vms == VM_MMAP_NONE || vms == VM_MMAP_ALL);
226 	ctx->vms = vms;
227 
228 	/*
229 	 * If 'memsize' cannot fit entirely in the 'lowmem' segment then
230 	 * create another 'highmem' segment above 4GB for the remainder.
231 	 */
232 	if (memsize > ctx->lowmem_limit) {
233 		ctx->lowmem = ctx->lowmem_limit;
234 		ctx->highmem = memsize - ctx->lowmem;
235 	} else {
236 		ctx->lowmem = memsize;
237 		ctx->highmem = 0;
238 	}
239 
240 	if (ctx->lowmem > 0) {
241 		addr = (vms == VM_MMAP_ALL) ? &ctx->lowmem_addr : NULL;
242 		error = setup_memory_segment(ctx, 0, ctx->lowmem, addr);
243 		if (error)
244 			return (error);
245 	}
246 
247 	if (ctx->highmem > 0) {
248 		addr = (vms == VM_MMAP_ALL) ? &ctx->highmem_addr : NULL;
249 		error = setup_memory_segment(ctx, 4*GB, ctx->highmem, addr);
250 		if (error)
251 			return (error);
252 	}
253 
254 	return (0);
255 }
256 
257 void *
258 vm_map_gpa(struct vmctx *ctx, vm_paddr_t gaddr, size_t len)
259 {
260 
261 	/* XXX VM_MMAP_SPARSE not implemented yet */
262 	assert(ctx->vms == VM_MMAP_ALL);
263 
264 	if (gaddr < ctx->lowmem && gaddr + len <= ctx->lowmem)
265 		return ((void *)(ctx->lowmem_addr + gaddr));
266 
267 	if (gaddr >= 4*GB) {
268 		gaddr -= 4*GB;
269 		if (gaddr < ctx->highmem && gaddr + len <= ctx->highmem)
270 			return ((void *)(ctx->highmem_addr + gaddr));
271 	}
272 
273 	return (NULL);
274 }
275 
276 int
277 vm_set_desc(struct vmctx *ctx, int vcpu, int reg,
278 	    uint64_t base, uint32_t limit, uint32_t access)
279 {
280 	int error;
281 	struct vm_seg_desc vmsegdesc;
282 
283 	bzero(&vmsegdesc, sizeof(vmsegdesc));
284 	vmsegdesc.cpuid = vcpu;
285 	vmsegdesc.regnum = reg;
286 	vmsegdesc.desc.base = base;
287 	vmsegdesc.desc.limit = limit;
288 	vmsegdesc.desc.access = access;
289 
290 	error = ioctl(ctx->fd, VM_SET_SEGMENT_DESCRIPTOR, &vmsegdesc);
291 	return (error);
292 }
293 
294 int
295 vm_get_desc(struct vmctx *ctx, int vcpu, int reg,
296 	    uint64_t *base, uint32_t *limit, uint32_t *access)
297 {
298 	int error;
299 	struct vm_seg_desc vmsegdesc;
300 
301 	bzero(&vmsegdesc, sizeof(vmsegdesc));
302 	vmsegdesc.cpuid = vcpu;
303 	vmsegdesc.regnum = reg;
304 
305 	error = ioctl(ctx->fd, VM_GET_SEGMENT_DESCRIPTOR, &vmsegdesc);
306 	if (error == 0) {
307 		*base = vmsegdesc.desc.base;
308 		*limit = vmsegdesc.desc.limit;
309 		*access = vmsegdesc.desc.access;
310 	}
311 	return (error);
312 }
313 
314 int
315 vm_set_register(struct vmctx *ctx, int vcpu, int reg, uint64_t val)
316 {
317 	int error;
318 	struct vm_register vmreg;
319 
320 	bzero(&vmreg, sizeof(vmreg));
321 	vmreg.cpuid = vcpu;
322 	vmreg.regnum = reg;
323 	vmreg.regval = val;
324 
325 	error = ioctl(ctx->fd, VM_SET_REGISTER, &vmreg);
326 	return (error);
327 }
328 
329 int
330 vm_get_register(struct vmctx *ctx, int vcpu, int reg, uint64_t *ret_val)
331 {
332 	int error;
333 	struct vm_register vmreg;
334 
335 	bzero(&vmreg, sizeof(vmreg));
336 	vmreg.cpuid = vcpu;
337 	vmreg.regnum = reg;
338 
339 	error = ioctl(ctx->fd, VM_GET_REGISTER, &vmreg);
340 	*ret_val = vmreg.regval;
341 	return (error);
342 }
343 
344 int
345 vm_run(struct vmctx *ctx, int vcpu, uint64_t rip, struct vm_exit *vmexit)
346 {
347 	int error;
348 	struct vm_run vmrun;
349 
350 	bzero(&vmrun, sizeof(vmrun));
351 	vmrun.cpuid = vcpu;
352 	vmrun.rip = rip;
353 
354 	error = ioctl(ctx->fd, VM_RUN, &vmrun);
355 	bcopy(&vmrun.vm_exit, vmexit, sizeof(struct vm_exit));
356 	return (error);
357 }
358 
359 int
360 vm_suspend(struct vmctx *ctx, enum vm_suspend_how how)
361 {
362 	struct vm_suspend vmsuspend;
363 
364 	bzero(&vmsuspend, sizeof(vmsuspend));
365 	vmsuspend.how = how;
366 	return (ioctl(ctx->fd, VM_SUSPEND, &vmsuspend));
367 }
368 
369 static int
370 vm_inject_exception_real(struct vmctx *ctx, int vcpu, int vector,
371     int error_code, int error_code_valid)
372 {
373 	struct vm_exception exc;
374 
375 	bzero(&exc, sizeof(exc));
376 	exc.cpuid = vcpu;
377 	exc.vector = vector;
378 	exc.error_code = error_code;
379 	exc.error_code_valid = error_code_valid;
380 
381 	return (ioctl(ctx->fd, VM_INJECT_EXCEPTION, &exc));
382 }
383 
384 int
385 vm_inject_exception(struct vmctx *ctx, int vcpu, int vector)
386 {
387 
388 	return (vm_inject_exception_real(ctx, vcpu, vector, 0, 0));
389 }
390 
391 int
392 vm_inject_exception2(struct vmctx *ctx, int vcpu, int vector, int errcode)
393 {
394 
395 	return (vm_inject_exception_real(ctx, vcpu, vector, errcode, 1));
396 }
397 
398 int
399 vm_apicid2vcpu(struct vmctx *ctx, int apicid)
400 {
401 	/*
402 	 * The apic id associated with the 'vcpu' has the same numerical value
403 	 * as the 'vcpu' itself.
404 	 */
405 	return (apicid);
406 }
407 
408 int
409 vm_lapic_irq(struct vmctx *ctx, int vcpu, int vector)
410 {
411 	struct vm_lapic_irq vmirq;
412 
413 	bzero(&vmirq, sizeof(vmirq));
414 	vmirq.cpuid = vcpu;
415 	vmirq.vector = vector;
416 
417 	return (ioctl(ctx->fd, VM_LAPIC_IRQ, &vmirq));
418 }
419 
420 int
421 vm_lapic_local_irq(struct vmctx *ctx, int vcpu, int vector)
422 {
423 	struct vm_lapic_irq vmirq;
424 
425 	bzero(&vmirq, sizeof(vmirq));
426 	vmirq.cpuid = vcpu;
427 	vmirq.vector = vector;
428 
429 	return (ioctl(ctx->fd, VM_LAPIC_LOCAL_IRQ, &vmirq));
430 }
431 
432 int
433 vm_lapic_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg)
434 {
435 	struct vm_lapic_msi vmmsi;
436 
437 	bzero(&vmmsi, sizeof(vmmsi));
438 	vmmsi.addr = addr;
439 	vmmsi.msg = msg;
440 
441 	return (ioctl(ctx->fd, VM_LAPIC_MSI, &vmmsi));
442 }
443 
444 int
445 vm_ioapic_assert_irq(struct vmctx *ctx, int irq)
446 {
447 	struct vm_ioapic_irq ioapic_irq;
448 
449 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
450 	ioapic_irq.irq = irq;
451 
452 	return (ioctl(ctx->fd, VM_IOAPIC_ASSERT_IRQ, &ioapic_irq));
453 }
454 
455 int
456 vm_ioapic_deassert_irq(struct vmctx *ctx, int irq)
457 {
458 	struct vm_ioapic_irq ioapic_irq;
459 
460 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
461 	ioapic_irq.irq = irq;
462 
463 	return (ioctl(ctx->fd, VM_IOAPIC_DEASSERT_IRQ, &ioapic_irq));
464 }
465 
466 int
467 vm_ioapic_pulse_irq(struct vmctx *ctx, int irq)
468 {
469 	struct vm_ioapic_irq ioapic_irq;
470 
471 	bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
472 	ioapic_irq.irq = irq;
473 
474 	return (ioctl(ctx->fd, VM_IOAPIC_PULSE_IRQ, &ioapic_irq));
475 }
476 
477 int
478 vm_ioapic_pincount(struct vmctx *ctx, int *pincount)
479 {
480 
481 	return (ioctl(ctx->fd, VM_IOAPIC_PINCOUNT, pincount));
482 }
483 
484 int
485 vm_isa_assert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
486 {
487 	struct vm_isa_irq isa_irq;
488 
489 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
490 	isa_irq.atpic_irq = atpic_irq;
491 	isa_irq.ioapic_irq = ioapic_irq;
492 
493 	return (ioctl(ctx->fd, VM_ISA_ASSERT_IRQ, &isa_irq));
494 }
495 
496 int
497 vm_isa_deassert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
498 {
499 	struct vm_isa_irq isa_irq;
500 
501 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
502 	isa_irq.atpic_irq = atpic_irq;
503 	isa_irq.ioapic_irq = ioapic_irq;
504 
505 	return (ioctl(ctx->fd, VM_ISA_DEASSERT_IRQ, &isa_irq));
506 }
507 
508 int
509 vm_isa_pulse_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
510 {
511 	struct vm_isa_irq isa_irq;
512 
513 	bzero(&isa_irq, sizeof(struct vm_isa_irq));
514 	isa_irq.atpic_irq = atpic_irq;
515 	isa_irq.ioapic_irq = ioapic_irq;
516 
517 	return (ioctl(ctx->fd, VM_ISA_PULSE_IRQ, &isa_irq));
518 }
519 
520 int
521 vm_isa_set_irq_trigger(struct vmctx *ctx, int atpic_irq,
522     enum vm_intr_trigger trigger)
523 {
524 	struct vm_isa_irq_trigger isa_irq_trigger;
525 
526 	bzero(&isa_irq_trigger, sizeof(struct vm_isa_irq_trigger));
527 	isa_irq_trigger.atpic_irq = atpic_irq;
528 	isa_irq_trigger.trigger = trigger;
529 
530 	return (ioctl(ctx->fd, VM_ISA_SET_IRQ_TRIGGER, &isa_irq_trigger));
531 }
532 
533 int
534 vm_inject_nmi(struct vmctx *ctx, int vcpu)
535 {
536 	struct vm_nmi vmnmi;
537 
538 	bzero(&vmnmi, sizeof(vmnmi));
539 	vmnmi.cpuid = vcpu;
540 
541 	return (ioctl(ctx->fd, VM_INJECT_NMI, &vmnmi));
542 }
543 
544 static struct {
545 	const char	*name;
546 	int		type;
547 } capstrmap[] = {
548 	{ "hlt_exit",		VM_CAP_HALT_EXIT },
549 	{ "mtrap_exit",		VM_CAP_MTRAP_EXIT },
550 	{ "pause_exit",		VM_CAP_PAUSE_EXIT },
551 	{ "unrestricted_guest",	VM_CAP_UNRESTRICTED_GUEST },
552 	{ "enable_invpcid",	VM_CAP_ENABLE_INVPCID },
553 	{ 0 }
554 };
555 
556 int
557 vm_capability_name2type(const char *capname)
558 {
559 	int i;
560 
561 	for (i = 0; capstrmap[i].name != NULL && capname != NULL; i++) {
562 		if (strcmp(capstrmap[i].name, capname) == 0)
563 			return (capstrmap[i].type);
564 	}
565 
566 	return (-1);
567 }
568 
569 const char *
570 vm_capability_type2name(int type)
571 {
572 	int i;
573 
574 	for (i = 0; capstrmap[i].name != NULL; i++) {
575 		if (capstrmap[i].type == type)
576 			return (capstrmap[i].name);
577 	}
578 
579 	return (NULL);
580 }
581 
582 int
583 vm_get_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap,
584 		  int *retval)
585 {
586 	int error;
587 	struct vm_capability vmcap;
588 
589 	bzero(&vmcap, sizeof(vmcap));
590 	vmcap.cpuid = vcpu;
591 	vmcap.captype = cap;
592 
593 	error = ioctl(ctx->fd, VM_GET_CAPABILITY, &vmcap);
594 	*retval = vmcap.capval;
595 	return (error);
596 }
597 
598 int
599 vm_set_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap, int val)
600 {
601 	struct vm_capability vmcap;
602 
603 	bzero(&vmcap, sizeof(vmcap));
604 	vmcap.cpuid = vcpu;
605 	vmcap.captype = cap;
606 	vmcap.capval = val;
607 
608 	return (ioctl(ctx->fd, VM_SET_CAPABILITY, &vmcap));
609 }
610 
611 int
612 vm_assign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
613 {
614 	struct vm_pptdev pptdev;
615 
616 	bzero(&pptdev, sizeof(pptdev));
617 	pptdev.bus = bus;
618 	pptdev.slot = slot;
619 	pptdev.func = func;
620 
621 	return (ioctl(ctx->fd, VM_BIND_PPTDEV, &pptdev));
622 }
623 
624 int
625 vm_unassign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
626 {
627 	struct vm_pptdev pptdev;
628 
629 	bzero(&pptdev, sizeof(pptdev));
630 	pptdev.bus = bus;
631 	pptdev.slot = slot;
632 	pptdev.func = func;
633 
634 	return (ioctl(ctx->fd, VM_UNBIND_PPTDEV, &pptdev));
635 }
636 
637 int
638 vm_map_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func,
639 		   vm_paddr_t gpa, size_t len, vm_paddr_t hpa)
640 {
641 	struct vm_pptdev_mmio pptmmio;
642 
643 	bzero(&pptmmio, sizeof(pptmmio));
644 	pptmmio.bus = bus;
645 	pptmmio.slot = slot;
646 	pptmmio.func = func;
647 	pptmmio.gpa = gpa;
648 	pptmmio.len = len;
649 	pptmmio.hpa = hpa;
650 
651 	return (ioctl(ctx->fd, VM_MAP_PPTDEV_MMIO, &pptmmio));
652 }
653 
654 int
655 vm_setup_pptdev_msi(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
656     uint64_t addr, uint64_t msg, int numvec)
657 {
658 	struct vm_pptdev_msi pptmsi;
659 
660 	bzero(&pptmsi, sizeof(pptmsi));
661 	pptmsi.vcpu = vcpu;
662 	pptmsi.bus = bus;
663 	pptmsi.slot = slot;
664 	pptmsi.func = func;
665 	pptmsi.msg = msg;
666 	pptmsi.addr = addr;
667 	pptmsi.numvec = numvec;
668 
669 	return (ioctl(ctx->fd, VM_PPTDEV_MSI, &pptmsi));
670 }
671 
672 int
673 vm_setup_pptdev_msix(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
674     int idx, uint64_t addr, uint64_t msg, uint32_t vector_control)
675 {
676 	struct vm_pptdev_msix pptmsix;
677 
678 	bzero(&pptmsix, sizeof(pptmsix));
679 	pptmsix.vcpu = vcpu;
680 	pptmsix.bus = bus;
681 	pptmsix.slot = slot;
682 	pptmsix.func = func;
683 	pptmsix.idx = idx;
684 	pptmsix.msg = msg;
685 	pptmsix.addr = addr;
686 	pptmsix.vector_control = vector_control;
687 
688 	return ioctl(ctx->fd, VM_PPTDEV_MSIX, &pptmsix);
689 }
690 
691 uint64_t *
692 vm_get_stats(struct vmctx *ctx, int vcpu, struct timeval *ret_tv,
693 	     int *ret_entries)
694 {
695 	int error;
696 
697 	static struct vm_stats vmstats;
698 
699 	vmstats.cpuid = vcpu;
700 
701 	error = ioctl(ctx->fd, VM_STATS, &vmstats);
702 	if (error == 0) {
703 		if (ret_entries)
704 			*ret_entries = vmstats.num_entries;
705 		if (ret_tv)
706 			*ret_tv = vmstats.tv;
707 		return (vmstats.statbuf);
708 	} else
709 		return (NULL);
710 }
711 
712 const char *
713 vm_get_stat_desc(struct vmctx *ctx, int index)
714 {
715 	static struct vm_stat_desc statdesc;
716 
717 	statdesc.index = index;
718 	if (ioctl(ctx->fd, VM_STAT_DESC, &statdesc) == 0)
719 		return (statdesc.desc);
720 	else
721 		return (NULL);
722 }
723 
724 int
725 vm_get_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state *state)
726 {
727 	int error;
728 	struct vm_x2apic x2apic;
729 
730 	bzero(&x2apic, sizeof(x2apic));
731 	x2apic.cpuid = vcpu;
732 
733 	error = ioctl(ctx->fd, VM_GET_X2APIC_STATE, &x2apic);
734 	*state = x2apic.state;
735 	return (error);
736 }
737 
738 int
739 vm_set_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state state)
740 {
741 	int error;
742 	struct vm_x2apic x2apic;
743 
744 	bzero(&x2apic, sizeof(x2apic));
745 	x2apic.cpuid = vcpu;
746 	x2apic.state = state;
747 
748 	error = ioctl(ctx->fd, VM_SET_X2APIC_STATE, &x2apic);
749 
750 	return (error);
751 }
752 
753 /*
754  * From Intel Vol 3a:
755  * Table 9-1. IA-32 Processor States Following Power-up, Reset or INIT
756  */
757 int
758 vcpu_reset(struct vmctx *vmctx, int vcpu)
759 {
760 	int error;
761 	uint64_t rflags, rip, cr0, cr4, zero, desc_base, rdx;
762 	uint32_t desc_access, desc_limit;
763 	uint16_t sel;
764 
765 	zero = 0;
766 
767 	rflags = 0x2;
768 	error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RFLAGS, rflags);
769 	if (error)
770 		goto done;
771 
772 	rip = 0xfff0;
773 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RIP, rip)) != 0)
774 		goto done;
775 
776 	cr0 = CR0_NE;
777 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR0, cr0)) != 0)
778 		goto done;
779 
780 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR3, zero)) != 0)
781 		goto done;
782 
783 	cr4 = 0;
784 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR4, cr4)) != 0)
785 		goto done;
786 
787 	/*
788 	 * CS: present, r/w, accessed, 16-bit, byte granularity, usable
789 	 */
790 	desc_base = 0xffff0000;
791 	desc_limit = 0xffff;
792 	desc_access = 0x0093;
793 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_CS,
794 			    desc_base, desc_limit, desc_access);
795 	if (error)
796 		goto done;
797 
798 	sel = 0xf000;
799 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CS, sel)) != 0)
800 		goto done;
801 
802 	/*
803 	 * SS,DS,ES,FS,GS: present, r/w, accessed, 16-bit, byte granularity
804 	 */
805 	desc_base = 0;
806 	desc_limit = 0xffff;
807 	desc_access = 0x0093;
808 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_SS,
809 			    desc_base, desc_limit, desc_access);
810 	if (error)
811 		goto done;
812 
813 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_DS,
814 			    desc_base, desc_limit, desc_access);
815 	if (error)
816 		goto done;
817 
818 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_ES,
819 			    desc_base, desc_limit, desc_access);
820 	if (error)
821 		goto done;
822 
823 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_FS,
824 			    desc_base, desc_limit, desc_access);
825 	if (error)
826 		goto done;
827 
828 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GS,
829 			    desc_base, desc_limit, desc_access);
830 	if (error)
831 		goto done;
832 
833 	sel = 0;
834 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_SS, sel)) != 0)
835 		goto done;
836 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_DS, sel)) != 0)
837 		goto done;
838 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_ES, sel)) != 0)
839 		goto done;
840 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_FS, sel)) != 0)
841 		goto done;
842 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_GS, sel)) != 0)
843 		goto done;
844 
845 	/* General purpose registers */
846 	rdx = 0xf00;
847 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RAX, zero)) != 0)
848 		goto done;
849 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBX, zero)) != 0)
850 		goto done;
851 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RCX, zero)) != 0)
852 		goto done;
853 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDX, rdx)) != 0)
854 		goto done;
855 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSI, zero)) != 0)
856 		goto done;
857 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDI, zero)) != 0)
858 		goto done;
859 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBP, zero)) != 0)
860 		goto done;
861 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSP, zero)) != 0)
862 		goto done;
863 
864 	/* GDTR, IDTR */
865 	desc_base = 0;
866 	desc_limit = 0xffff;
867 	desc_access = 0;
868 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GDTR,
869 			    desc_base, desc_limit, desc_access);
870 	if (error != 0)
871 		goto done;
872 
873 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_IDTR,
874 			    desc_base, desc_limit, desc_access);
875 	if (error != 0)
876 		goto done;
877 
878 	/* TR */
879 	desc_base = 0;
880 	desc_limit = 0xffff;
881 	desc_access = 0x0000008b;
882 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_TR, 0, 0, desc_access);
883 	if (error)
884 		goto done;
885 
886 	sel = 0;
887 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_TR, sel)) != 0)
888 		goto done;
889 
890 	/* LDTR */
891 	desc_base = 0;
892 	desc_limit = 0xffff;
893 	desc_access = 0x00000082;
894 	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_LDTR, desc_base,
895 			    desc_limit, desc_access);
896 	if (error)
897 		goto done;
898 
899 	sel = 0;
900 	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_LDTR, 0)) != 0)
901 		goto done;
902 
903 	/* XXX cr2, debug registers */
904 
905 	error = 0;
906 done:
907 	return (error);
908 }
909 
910 int
911 vm_get_gpa_pmap(struct vmctx *ctx, uint64_t gpa, uint64_t *pte, int *num)
912 {
913 	int error, i;
914 	struct vm_gpa_pte gpapte;
915 
916 	bzero(&gpapte, sizeof(gpapte));
917 	gpapte.gpa = gpa;
918 
919 	error = ioctl(ctx->fd, VM_GET_GPA_PMAP, &gpapte);
920 
921 	if (error == 0) {
922 		*num = gpapte.ptenum;
923 		for (i = 0; i < gpapte.ptenum; i++)
924 			pte[i] = gpapte.pte[i];
925 	}
926 
927 	return (error);
928 }
929 
930 int
931 vm_get_hpet_capabilities(struct vmctx *ctx, uint32_t *capabilities)
932 {
933 	int error;
934 	struct vm_hpet_cap cap;
935 
936 	bzero(&cap, sizeof(struct vm_hpet_cap));
937 	error = ioctl(ctx->fd, VM_GET_HPET_CAPABILITIES, &cap);
938 	if (capabilities != NULL)
939 		*capabilities = cap.capabilities;
940 	return (error);
941 }
942 
943 static int
944 gla2gpa(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
945     uint64_t gla, int prot, int *fault, uint64_t *gpa)
946 {
947 	struct vm_gla2gpa gg;
948 	int error;
949 
950 	bzero(&gg, sizeof(struct vm_gla2gpa));
951 	gg.vcpuid = vcpu;
952 	gg.prot = prot;
953 	gg.gla = gla;
954 	gg.paging = *paging;
955 
956 	error = ioctl(ctx->fd, VM_GLA2GPA, &gg);
957 	if (error == 0) {
958 		*fault = gg.fault;
959 		*gpa = gg.gpa;
960 	}
961 	return (error);
962 }
963 
964 #ifndef min
965 #define	min(a,b)	(((a) < (b)) ? (a) : (b))
966 #endif
967 
968 int
969 vm_gla2gpa(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
970     uint64_t gla, size_t len, int prot, struct iovec *iov, int iovcnt)
971 {
972 	uint64_t gpa;
973 	int error, fault, i, n, off;
974 
975 	for (i = 0; i < iovcnt; i++) {
976 		iov[i].iov_base = 0;
977 		iov[i].iov_len = 0;
978 	}
979 
980 	while (len) {
981 		assert(iovcnt > 0);
982 		error = gla2gpa(ctx, vcpu, paging, gla, prot, &fault, &gpa);
983 		if (error)
984 			return (-1);
985 		if (fault)
986 			return (1);
987 
988 		off = gpa & PAGE_MASK;
989 		n = min(len, PAGE_SIZE - off);
990 
991 		iov->iov_base = (void *)gpa;
992 		iov->iov_len = n;
993 		iov++;
994 		iovcnt--;
995 
996 		gla += n;
997 		len -= n;
998 	}
999 	return (0);
1000 }
1001 
1002 void
1003 vm_copyin(struct vmctx *ctx, int vcpu, struct iovec *iov, void *vp, size_t len)
1004 {
1005 	const char *src;
1006 	char *dst;
1007 	uint64_t gpa;
1008 	size_t n;
1009 
1010 	dst = vp;
1011 	while (len) {
1012 		assert(iov->iov_len);
1013 		gpa = (uint64_t)iov->iov_base;
1014 		n = min(len, iov->iov_len);
1015 		src = vm_map_gpa(ctx, gpa, n);
1016 		bcopy(src, dst, n);
1017 
1018 		iov++;
1019 		dst += n;
1020 		len -= n;
1021 	}
1022 }
1023 
1024 void
1025 vm_copyout(struct vmctx *ctx, int vcpu, const void *vp, struct iovec *iov,
1026     size_t len)
1027 {
1028 	const char *src;
1029 	char *dst;
1030 	uint64_t gpa;
1031 	size_t n;
1032 
1033 	src = vp;
1034 	while (len) {
1035 		assert(iov->iov_len);
1036 		gpa = (uint64_t)iov->iov_base;
1037 		n = min(len, iov->iov_len);
1038 		dst = vm_map_gpa(ctx, gpa, n);
1039 		bcopy(src, dst, n);
1040 
1041 		iov++;
1042 		src += n;
1043 		len -= n;
1044 	}
1045 }
1046