xref: /freebsd/usr.sbin/bhyve/bhyverun.c (revision 8a166cafe0965f6bd72cd3d2f5372704f05cb5e8)
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/mman.h>
34 #include <sys/time.h>
35 
36 #include <machine/segments.h>
37 
38 #include <stdio.h>
39 #include <stdlib.h>
40 #include <libgen.h>
41 #include <unistd.h>
42 #include <assert.h>
43 #include <errno.h>
44 #include <signal.h>
45 #include <pthread.h>
46 #include <pthread_np.h>
47 
48 #include <machine/vmm.h>
49 #include <vmmapi.h>
50 
51 #include "bhyverun.h"
52 #include "acpi.h"
53 #include "inout.h"
54 #include "dbgport.h"
55 #include "mem.h"
56 #include "mevent.h"
57 #include "mptbl.h"
58 #include "pci_emul.h"
59 #include "xmsr.h"
60 #include "ioapic.h"
61 #include "spinup_ap.h"
62 
63 #define	DEFAULT_GUEST_HZ	100
64 #define	DEFAULT_GUEST_TSLICE	200
65 
66 #define GUEST_NIO_PORT		0x488	/* guest upcalls via i/o port */
67 
68 #define	VMEXIT_SWITCH		0	/* force vcpu switch in mux mode */
69 #define	VMEXIT_CONTINUE		1	/* continue from next instruction */
70 #define	VMEXIT_RESTART		2	/* restart current instruction */
71 #define	VMEXIT_ABORT		3	/* abort the vm run loop */
72 #define	VMEXIT_RESET		4	/* guest machine has reset */
73 
74 #define MB		(1024UL * 1024)
75 #define GB		(1024UL * MB)
76 
77 typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
78 
79 int guest_tslice = DEFAULT_GUEST_TSLICE;
80 int guest_hz = DEFAULT_GUEST_HZ;
81 char *vmname;
82 
83 u_long lomem_sz;
84 u_long himem_sz;
85 
86 int guest_ncpus;
87 
88 static int pincpu = -1;
89 static int guest_vcpu_mux;
90 static int guest_vmexit_on_hlt, guest_vmexit_on_pause, disable_x2apic;
91 
92 static int foundcpus;
93 
94 static int strictio;
95 
96 static int acpi;
97 
98 static char *lomem_addr;
99 static char *himem_addr;
100 
101 static char *progname;
102 static const int BSP = 0;
103 
104 static int cpumask;
105 
106 static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
107 
108 struct vm_exit vmexit[VM_MAXCPU];
109 
110 struct fbsdstats {
111         uint64_t        vmexit_bogus;
112         uint64_t        vmexit_bogus_switch;
113         uint64_t        vmexit_hlt;
114         uint64_t        vmexit_pause;
115         uint64_t        vmexit_mtrap;
116         uint64_t        vmexit_paging;
117         uint64_t        cpu_switch_rotate;
118         uint64_t        cpu_switch_direct;
119         int             io_reset;
120 } stats;
121 
122 struct mt_vmm_info {
123 	pthread_t	mt_thr;
124 	struct vmctx	*mt_ctx;
125 	int		mt_vcpu;
126 } mt_vmm_info[VM_MAXCPU];
127 
128 static void
129 usage(int code)
130 {
131 
132         fprintf(stderr,
133                 "Usage: %s [-aehABHIP][-g <gdb port>][-z <hz>][-s <pci>]"
134 		"[-S <pci>][-p pincpu][-n <pci>][-m lowmem][-M highmem]"
135 		" <vmname>\n"
136 		"       -a: local apic is in XAPIC mode (default is X2APIC)\n"
137 		"       -A: create an ACPI table\n"
138 		"       -g: gdb port (default is %d and 0 means don't open)\n"
139 		"       -c: # cpus (default 1)\n"
140 		"       -p: pin vcpu 'n' to host cpu 'pincpu + n'\n"
141 		"       -B: inject breakpoint exception on vm entry\n"
142 		"       -H: vmexit from the guest on hlt\n"
143 		"       -I: present an ioapic to the guest\n"
144 		"       -P: vmexit from the guest on pause\n"
145 		"	-e: exit on unhandled i/o access\n"
146 		"       -h: help\n"
147 		"       -z: guest hz (default is %d)\n"
148 		"       -s: <slot,driver,configinfo> PCI slot config\n"
149 		"       -S: <slot,driver,configinfo> legacy PCI slot config\n"
150 		"       -m: lowmem in MB\n"
151 		"       -M: highmem in MB\n"
152 		"       -x: mux vcpus to 1 hcpu\n"
153 		"       -t: mux vcpu timeslice hz (default %d)\n",
154 		progname, DEFAULT_GDB_PORT, DEFAULT_GUEST_HZ,
155 		DEFAULT_GUEST_TSLICE);
156 	exit(code);
157 }
158 
159 void *
160 paddr_guest2host(uintptr_t gaddr)
161 {
162 	if (lomem_sz == 0)
163 		return (NULL);
164 
165 	if (gaddr < lomem_sz) {
166 		return ((void *)(lomem_addr + gaddr));
167 	} else if (gaddr >= 4*GB && gaddr < (4*GB + himem_sz)) {
168 		return ((void *)(himem_addr + gaddr - 4*GB));
169 	} else
170 		return (NULL);
171 }
172 
173 int
174 fbsdrun_disable_x2apic(void)
175 {
176 
177 	return (disable_x2apic);
178 }
179 
180 int
181 fbsdrun_vmexit_on_pause(void)
182 {
183 
184 	return (guest_vmexit_on_pause);
185 }
186 
187 int
188 fbsdrun_vmexit_on_hlt(void)
189 {
190 
191 	return (guest_vmexit_on_hlt);
192 }
193 
194 int
195 fbsdrun_muxed(void)
196 {
197 
198 	return (guest_vcpu_mux);
199 }
200 
201 static void *
202 fbsdrun_start_thread(void *param)
203 {
204 	char tname[MAXCOMLEN + 1];
205 	struct mt_vmm_info *mtp;
206 	int vcpu;
207 
208 	mtp = param;
209 	vcpu = mtp->mt_vcpu;
210 
211 	snprintf(tname, sizeof(tname), "%s vcpu %d", vmname, vcpu);
212 	pthread_set_name_np(mtp->mt_thr, tname);
213 
214 	vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
215 
216 	/* not reached */
217 	exit(1);
218 	return (NULL);
219 }
220 
221 void
222 fbsdrun_addcpu(struct vmctx *ctx, int vcpu, uint64_t rip)
223 {
224 	int error;
225 
226 	if (cpumask & (1 << vcpu)) {
227 		fprintf(stderr, "addcpu: attempting to add existing cpu %d\n",
228 		    vcpu);
229 		exit(1);
230 	}
231 
232 	cpumask |= 1 << vcpu;
233 	foundcpus++;
234 
235 	/*
236 	 * Set up the vmexit struct to allow execution to start
237 	 * at the given RIP
238 	 */
239 	vmexit[vcpu].rip = rip;
240 	vmexit[vcpu].inst_length = 0;
241 
242 	if (vcpu == BSP || !guest_vcpu_mux){
243 		mt_vmm_info[vcpu].mt_ctx = ctx;
244 		mt_vmm_info[vcpu].mt_vcpu = vcpu;
245 
246 		error = pthread_create(&mt_vmm_info[vcpu].mt_thr, NULL,
247 				fbsdrun_start_thread, &mt_vmm_info[vcpu]);
248 		assert(error == 0);
249 	}
250 }
251 
252 static int
253 fbsdrun_get_next_cpu(int curcpu)
254 {
255 
256 	/*
257 	 * Get the next available CPU. Assumes they arrive
258 	 * in ascending order with no gaps.
259 	 */
260 	return ((curcpu + 1) % foundcpus);
261 }
262 
263 static int
264 vmexit_catch_reset(void)
265 {
266         stats.io_reset++;
267         return (VMEXIT_RESET);
268 }
269 
270 static int
271 vmexit_catch_inout(void)
272 {
273 	return (VMEXIT_ABORT);
274 }
275 
276 static int
277 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
278 		     uint32_t eax)
279 {
280 #if PG_DEBUG /* put all types of debug here */
281         if (eax == 0) {
282 		pause_noswitch = 1;
283 	} else if (eax == 1) {
284 		pause_noswitch = 0;
285 	} else {
286 		pause_noswitch = 0;
287 		if (eax == 5) {
288 			vm_set_capability(ctx, *pvcpu, VM_CAP_MTRAP_EXIT, 1);
289 		}
290 	}
291 #endif
292         return (VMEXIT_CONTINUE);
293 }
294 
295 static int
296 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
297 {
298 	int error;
299 	int bytes, port, in, out;
300 	uint32_t eax;
301 	int vcpu;
302 
303 	vcpu = *pvcpu;
304 
305 	port = vme->u.inout.port;
306 	bytes = vme->u.inout.bytes;
307 	eax = vme->u.inout.eax;
308 	in = vme->u.inout.in;
309 	out = !in;
310 
311 	/* We don't deal with these */
312 	if (vme->u.inout.string || vme->u.inout.rep)
313 		return (VMEXIT_ABORT);
314 
315 	/* Special case of guest reset */
316 	if (out && port == 0x64 && (uint8_t)eax == 0xFE)
317 		return (vmexit_catch_reset());
318 
319         /* Extra-special case of host notifications */
320         if (out && port == GUEST_NIO_PORT)
321                 return (vmexit_handle_notify(ctx, vme, pvcpu, eax));
322 
323 	error = emulate_inout(ctx, vcpu, in, port, bytes, &eax, strictio);
324 	if (error == 0 && in)
325 		error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RAX, eax);
326 
327 	if (error == 0)
328 		return (VMEXIT_CONTINUE);
329 	else {
330 		fprintf(stderr, "Unhandled %s%c 0x%04x\n",
331 			in ? "in" : "out",
332 			bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), port);
333 		return (vmexit_catch_inout());
334 	}
335 }
336 
337 static int
338 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
339 {
340 	fprintf(stderr, "vm exit rdmsr 0x%x, cpu %d\n", vme->u.msr.code,
341 	    *pvcpu);
342 	return (VMEXIT_ABORT);
343 }
344 
345 static int
346 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
347 {
348 	int newcpu;
349 	int retval = VMEXIT_CONTINUE;
350 
351 	newcpu = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code,vme->u.msr.wval);
352 
353 	if (guest_vcpu_mux && *pvcpu != newcpu) {
354                 retval = VMEXIT_SWITCH;
355                 *pvcpu = newcpu;
356         }
357 
358         return (retval);
359 }
360 
361 static int
362 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
363 {
364 	int newcpu;
365 	int retval = VMEXIT_CONTINUE;
366 
367 	newcpu = spinup_ap(ctx, *pvcpu,
368 			   vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
369 
370 	if (guest_vcpu_mux && *pvcpu != newcpu) {
371 		retval = VMEXIT_SWITCH;
372 		*pvcpu = newcpu;
373 	}
374 
375 	return (retval);
376 }
377 
378 static int
379 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
380 {
381 
382 	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
383 	fprintf(stderr, "\treason\t\tVMX\n");
384 	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
385 	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
386 	fprintf(stderr, "\terror\t\t%d\n", vmexit->u.vmx.error);
387 	fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
388 	fprintf(stderr, "\tqualification\t0x%016lx\n",
389 	    vmexit->u.vmx.exit_qualification);
390 
391 	return (VMEXIT_ABORT);
392 }
393 
394 static int bogus_noswitch = 1;
395 
396 static int
397 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
398 {
399 	stats.vmexit_bogus++;
400 
401 	if (!guest_vcpu_mux || guest_ncpus == 1 || bogus_noswitch) {
402 		return (VMEXIT_RESTART);
403 	} else {
404 		stats.vmexit_bogus_switch++;
405 		vmexit->inst_length = 0;
406 		*pvcpu = -1;
407 		return (VMEXIT_SWITCH);
408 	}
409 }
410 
411 static int
412 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
413 {
414 	stats.vmexit_hlt++;
415 	if (fbsdrun_muxed()) {
416 		*pvcpu = -1;
417 		return (VMEXIT_SWITCH);
418 	} else {
419 		/*
420 		 * Just continue execution with the next instruction. We use
421 		 * the HLT VM exit as a way to be friendly with the host
422 		 * scheduler.
423 		 */
424 		return (VMEXIT_CONTINUE);
425 	}
426 }
427 
428 static int pause_noswitch;
429 
430 static int
431 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
432 {
433 	stats.vmexit_pause++;
434 
435 	if (fbsdrun_muxed() && !pause_noswitch) {
436 		*pvcpu = -1;
437 		return (VMEXIT_SWITCH);
438         } else {
439 		return (VMEXIT_CONTINUE);
440 	}
441 }
442 
443 static int
444 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
445 {
446 	stats.vmexit_mtrap++;
447 
448 	return (VMEXIT_RESTART);
449 }
450 
451 static int
452 vmexit_paging(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
453 {
454 	int err;
455 	stats.vmexit_paging++;
456 
457 	err = emulate_mem(ctx, *pvcpu, vmexit->u.paging.gpa,
458 			  &vmexit->u.paging.vie);
459 
460 	if (err) {
461 		if (err == EINVAL) {
462 			fprintf(stderr,
463 			    "Failed to emulate instruction at 0x%lx\n",
464 			    vmexit->rip);
465 		} else if (err == ESRCH) {
466 			fprintf(stderr, "Unhandled memory access to 0x%lx\n",
467 			    vmexit->u.paging.gpa);
468 		}
469 
470 		return (VMEXIT_ABORT);
471 	}
472 
473 	return (VMEXIT_CONTINUE);
474 }
475 
476 static void
477 sigalrm(int sig)
478 {
479 	return;
480 }
481 
482 static void
483 setup_timeslice(void)
484 {
485 	struct sigaction sa;
486 	struct itimerval itv;
487 	int error;
488 
489 	/*
490 	 * Setup a realtime timer to generate a SIGALRM at a
491 	 * frequency of 'guest_tslice' ticks per second.
492 	 */
493 	sigemptyset(&sa.sa_mask);
494 	sa.sa_flags = 0;
495 	sa.sa_handler = sigalrm;
496 
497 	error = sigaction(SIGALRM, &sa, NULL);
498 	assert(error == 0);
499 
500 	itv.it_interval.tv_sec = 0;
501 	itv.it_interval.tv_usec = 1000000 / guest_tslice;
502 	itv.it_value.tv_sec = 0;
503 	itv.it_value.tv_usec = 1000000 / guest_tslice;
504 
505 	error = setitimer(ITIMER_REAL, &itv, NULL);
506 	assert(error == 0);
507 }
508 
509 static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
510 	[VM_EXITCODE_INOUT]  = vmexit_inout,
511 	[VM_EXITCODE_VMX]    = vmexit_vmx,
512 	[VM_EXITCODE_BOGUS]  = vmexit_bogus,
513 	[VM_EXITCODE_RDMSR]  = vmexit_rdmsr,
514 	[VM_EXITCODE_WRMSR]  = vmexit_wrmsr,
515 	[VM_EXITCODE_MTRAP]  = vmexit_mtrap,
516 	[VM_EXITCODE_PAGING] = vmexit_paging,
517 	[VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
518 };
519 
520 static void
521 vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip)
522 {
523 	cpuset_t mask;
524 	int error, rc, prevcpu;
525 
526 	if (guest_vcpu_mux)
527 		setup_timeslice();
528 
529 	if (pincpu >= 0) {
530 		CPU_ZERO(&mask);
531 		CPU_SET(pincpu + vcpu, &mask);
532 		error = pthread_setaffinity_np(pthread_self(),
533 					       sizeof(mask), &mask);
534 		assert(error == 0);
535 	}
536 
537 	while (1) {
538 		error = vm_run(ctx, vcpu, rip, &vmexit[vcpu]);
539 		if (error != 0) {
540 			/*
541 			 * It is possible that 'vmmctl' or some other process
542 			 * has transitioned the vcpu to CANNOT_RUN state right
543 			 * before we tried to transition it to RUNNING.
544 			 *
545 			 * This is expected to be temporary so just retry.
546 			 */
547 			if (errno == EBUSY)
548 				continue;
549 			else
550 				break;
551 		}
552 
553 		prevcpu = vcpu;
554                 rc = (*handler[vmexit[vcpu].exitcode])(ctx, &vmexit[vcpu],
555                                                        &vcpu);
556 		switch (rc) {
557                 case VMEXIT_SWITCH:
558 			assert(guest_vcpu_mux);
559 			if (vcpu == -1) {
560 				stats.cpu_switch_rotate++;
561 				vcpu = fbsdrun_get_next_cpu(prevcpu);
562 			} else {
563 				stats.cpu_switch_direct++;
564 			}
565 			/* fall through */
566 		case VMEXIT_CONTINUE:
567                         rip = vmexit[vcpu].rip + vmexit[vcpu].inst_length;
568 			break;
569 		case VMEXIT_RESTART:
570                         rip = vmexit[vcpu].rip;
571 			break;
572 		case VMEXIT_RESET:
573 			exit(0);
574 		default:
575 			exit(1);
576 		}
577 	}
578 	fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
579 }
580 
581 static int
582 num_vcpus_allowed(struct vmctx *ctx)
583 {
584 	int tmp, error;
585 
586 	error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
587 
588 	/*
589 	 * The guest is allowed to spinup more than one processor only if the
590 	 * UNRESTRICTED_GUEST capability is available.
591 	 */
592 	if (error == 0)
593 		return (VM_MAXCPU);
594 	else
595 		return (1);
596 }
597 
598 int
599 main(int argc, char *argv[])
600 {
601 	int c, error, gdb_port, inject_bkpt, tmp, err, ioapic, bvmcons;
602 	int max_vcpus;
603 	struct vmctx *ctx;
604 	uint64_t rip;
605 
606 	bvmcons = 0;
607 	inject_bkpt = 0;
608 	progname = basename(argv[0]);
609 	gdb_port = DEFAULT_GDB_PORT;
610 	guest_ncpus = 1;
611 	ioapic = 0;
612 
613 	while ((c = getopt(argc, argv, "abehABHIPxp:g:c:z:s:S:n:m:M:")) != -1) {
614 		switch (c) {
615 		case 'a':
616 			disable_x2apic = 1;
617 			break;
618 		case 'A':
619 			acpi = 1;
620 			break;
621 		case 'b':
622 			bvmcons = 1;
623 			break;
624 		case 'B':
625 			inject_bkpt = 1;
626 			break;
627 		case 'x':
628 			guest_vcpu_mux = 1;
629 			break;
630 		case 'p':
631 			pincpu = atoi(optarg);
632 			break;
633                 case 'c':
634 			guest_ncpus = atoi(optarg);
635 			break;
636 		case 'g':
637 			gdb_port = atoi(optarg);
638 			break;
639 		case 'z':
640 			guest_hz = atoi(optarg);
641 			break;
642 		case 't':
643 			guest_tslice = atoi(optarg);
644 			break;
645 		case 's':
646 			pci_parse_slot(optarg, 0);
647 			break;
648 		case 'S':
649 			pci_parse_slot(optarg, 1);
650 			break;
651                 case 'm':
652 			lomem_sz = strtoul(optarg, NULL, 0) * MB;
653 			break;
654                 case 'M':
655 			himem_sz = strtoul(optarg, NULL, 0) * MB;
656 			break;
657 		case 'H':
658 			guest_vmexit_on_hlt = 1;
659 			break;
660 		case 'I':
661 			ioapic = 1;
662 			break;
663 		case 'P':
664 			guest_vmexit_on_pause = 1;
665 			break;
666 		case 'e':
667 			strictio = 1;
668 			break;
669 		case 'h':
670 			usage(0);
671 		default:
672 			usage(1);
673 		}
674 	}
675 	argc -= optind;
676 	argv += optind;
677 
678 	if (argc != 1)
679 		usage(1);
680 
681 	/* No need to mux if guest is uni-processor */
682 	if (guest_ncpus <= 1)
683 		guest_vcpu_mux = 0;
684 
685 	/* vmexit on hlt if guest is muxed */
686 	if (guest_vcpu_mux) {
687 		guest_vmexit_on_hlt = 1;
688 		guest_vmexit_on_pause = 1;
689 	}
690 
691 	vmname = argv[0];
692 
693 	ctx = vm_open(vmname);
694 	if (ctx == NULL) {
695 		perror("vm_open");
696 		exit(1);
697 	}
698 
699 	max_vcpus = num_vcpus_allowed(ctx);
700 	if (guest_ncpus > max_vcpus) {
701 		fprintf(stderr, "%d vCPUs requested but only %d available\n",
702 			guest_ncpus, max_vcpus);
703 		exit(1);
704 	}
705 
706 	if (fbsdrun_vmexit_on_hlt()) {
707 		err = vm_get_capability(ctx, BSP, VM_CAP_HALT_EXIT, &tmp);
708 		if (err < 0) {
709 			fprintf(stderr, "VM exit on HLT not supported\n");
710 			exit(1);
711 		}
712 		vm_set_capability(ctx, BSP, VM_CAP_HALT_EXIT, 1);
713 		handler[VM_EXITCODE_HLT] = vmexit_hlt;
714 	}
715 
716         if (fbsdrun_vmexit_on_pause()) {
717 		/*
718 		 * pause exit support required for this mode
719 		 */
720 		err = vm_get_capability(ctx, BSP, VM_CAP_PAUSE_EXIT, &tmp);
721 		if (err < 0) {
722 			fprintf(stderr,
723 			    "SMP mux requested, no pause support\n");
724 			exit(1);
725 		}
726 		vm_set_capability(ctx, BSP, VM_CAP_PAUSE_EXIT, 1);
727 		handler[VM_EXITCODE_PAUSE] = vmexit_pause;
728         }
729 
730 	if (fbsdrun_disable_x2apic())
731 		err = vm_set_x2apic_state(ctx, BSP, X2APIC_DISABLED);
732 	else
733 		err = vm_set_x2apic_state(ctx, BSP, X2APIC_ENABLED);
734 
735 	if (err) {
736 		fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
737 		exit(1);
738 	}
739 
740 	if (lomem_sz != 0) {
741 		lomem_addr = vm_map_memory(ctx, 0, lomem_sz);
742 		if (lomem_addr == (char *) MAP_FAILED) {
743 			lomem_sz = 0;
744 		} else if (himem_sz != 0) {
745 			himem_addr = vm_map_memory(ctx, 4*GB, himem_sz);
746 			if (himem_addr == (char *) MAP_FAILED) {
747 				lomem_sz = 0;
748 				himem_sz = 0;
749 			}
750 		}
751 	}
752 
753 	init_inout();
754 	init_pci(ctx);
755 	if (ioapic)
756 		ioapic_init(0);
757 
758 	if (gdb_port != 0)
759 		init_dbgport(gdb_port);
760 
761 	if (bvmcons)
762 		init_bvmcons();
763 
764 	error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
765 	assert(error == 0);
766 
767 	if (inject_bkpt) {
768 		error = vm_inject_event(ctx, BSP, VM_HW_EXCEPTION, IDT_BP);
769 		assert(error == 0);
770 	}
771 
772 	/*
773 	 * build the guest tables, MP etc.
774 	 */
775 	mptable_build(ctx, guest_ncpus, ioapic);
776 
777 	if (acpi) {
778 		error = acpi_build(ctx, guest_ncpus, ioapic);
779 		assert(error == 0);
780 	}
781 
782 	/*
783 	 * Add CPU 0
784 	 */
785 	fbsdrun_addcpu(ctx, BSP, rip);
786 
787 	/*
788 	 * Head off to the main event dispatch loop
789 	 */
790 	mevent_dispatch();
791 
792 	exit(1);
793 }
794