xref: /freebsd/usr.sbin/bhyve/bhyverun.c (revision 3fc9e2c36555140de248a0b4def91bbfa44d7c2c)
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 #include "rtc.h"
63 
64 #define	DEFAULT_GUEST_HZ	100
65 #define	DEFAULT_GUEST_TSLICE	200
66 
67 #define GUEST_NIO_PORT		0x488	/* guest upcalls via i/o port */
68 
69 #define	VMEXIT_SWITCH		0	/* force vcpu switch in mux mode */
70 #define	VMEXIT_CONTINUE		1	/* continue from next instruction */
71 #define	VMEXIT_RESTART		2	/* restart current instruction */
72 #define	VMEXIT_ABORT		3	/* abort the vm run loop */
73 #define	VMEXIT_RESET		4	/* guest machine has reset */
74 
75 #define MB		(1024UL * 1024)
76 #define GB		(1024UL * MB)
77 
78 typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
79 
80 int guest_tslice = DEFAULT_GUEST_TSLICE;
81 int guest_hz = DEFAULT_GUEST_HZ;
82 char *vmname;
83 
84 int guest_ncpus;
85 
86 static int pincpu = -1;
87 static int guest_vcpu_mux;
88 static int guest_vmexit_on_hlt, guest_vmexit_on_pause, disable_x2apic;
89 
90 static int foundcpus;
91 
92 static int strictio;
93 
94 static int acpi;
95 
96 static char *progname;
97 static const int BSP = 0;
98 
99 static int cpumask;
100 
101 static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
102 
103 struct vm_exit vmexit[VM_MAXCPU];
104 
105 struct fbsdstats {
106         uint64_t        vmexit_bogus;
107         uint64_t        vmexit_bogus_switch;
108         uint64_t        vmexit_hlt;
109         uint64_t        vmexit_pause;
110         uint64_t        vmexit_mtrap;
111         uint64_t        vmexit_paging;
112         uint64_t        cpu_switch_rotate;
113         uint64_t        cpu_switch_direct;
114         int             io_reset;
115 } stats;
116 
117 struct mt_vmm_info {
118 	pthread_t	mt_thr;
119 	struct vmctx	*mt_ctx;
120 	int		mt_vcpu;
121 } mt_vmm_info[VM_MAXCPU];
122 
123 static void
124 usage(int code)
125 {
126 
127         fprintf(stderr,
128                 "Usage: %s [-aehABHIP][-g <gdb port>][-z <hz>][-s <pci>]"
129 		"[-S <pci>][-p pincpu][-n <pci>][-m lowmem][-M highmem]"
130 		" <vmname>\n"
131 		"       -a: local apic is in XAPIC mode (default is X2APIC)\n"
132 		"       -A: create an ACPI table\n"
133 		"       -g: gdb port (default is %d and 0 means don't open)\n"
134 		"       -c: # cpus (default 1)\n"
135 		"       -p: pin vcpu 'n' to host cpu 'pincpu + n'\n"
136 		"       -B: inject breakpoint exception on vm entry\n"
137 		"       -H: vmexit from the guest on hlt\n"
138 		"       -I: present an ioapic to the guest\n"
139 		"       -P: vmexit from the guest on pause\n"
140 		"	-e: exit on unhandled i/o access\n"
141 		"       -h: help\n"
142 		"       -z: guest hz (default is %d)\n"
143 		"       -s: <slot,driver,configinfo> PCI slot config\n"
144 		"       -S: <slot,driver,configinfo> legacy PCI slot config\n"
145 		"       -m: memory size in MB\n"
146 		"       -x: mux vcpus to 1 hcpu\n"
147 		"       -t: mux vcpu timeslice hz (default %d)\n",
148 		progname, DEFAULT_GDB_PORT, DEFAULT_GUEST_HZ,
149 		DEFAULT_GUEST_TSLICE);
150 	exit(code);
151 }
152 
153 void *
154 paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
155 {
156 
157 	return (vm_map_gpa(ctx, gaddr, len));
158 }
159 
160 int
161 fbsdrun_disable_x2apic(void)
162 {
163 
164 	return (disable_x2apic);
165 }
166 
167 int
168 fbsdrun_vmexit_on_pause(void)
169 {
170 
171 	return (guest_vmexit_on_pause);
172 }
173 
174 int
175 fbsdrun_vmexit_on_hlt(void)
176 {
177 
178 	return (guest_vmexit_on_hlt);
179 }
180 
181 int
182 fbsdrun_muxed(void)
183 {
184 
185 	return (guest_vcpu_mux);
186 }
187 
188 static void *
189 fbsdrun_start_thread(void *param)
190 {
191 	char tname[MAXCOMLEN + 1];
192 	struct mt_vmm_info *mtp;
193 	int vcpu;
194 
195 	mtp = param;
196 	vcpu = mtp->mt_vcpu;
197 
198 	snprintf(tname, sizeof(tname), "%s vcpu %d", vmname, vcpu);
199 	pthread_set_name_np(mtp->mt_thr, tname);
200 
201 	vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
202 
203 	/* not reached */
204 	exit(1);
205 	return (NULL);
206 }
207 
208 void
209 fbsdrun_addcpu(struct vmctx *ctx, int vcpu, uint64_t rip)
210 {
211 	int error;
212 
213 	if (cpumask & (1 << vcpu)) {
214 		fprintf(stderr, "addcpu: attempting to add existing cpu %d\n",
215 		    vcpu);
216 		exit(1);
217 	}
218 
219 	cpumask |= 1 << vcpu;
220 	foundcpus++;
221 
222 	/*
223 	 * Set up the vmexit struct to allow execution to start
224 	 * at the given RIP
225 	 */
226 	vmexit[vcpu].rip = rip;
227 	vmexit[vcpu].inst_length = 0;
228 
229 	if (vcpu == BSP || !guest_vcpu_mux){
230 		mt_vmm_info[vcpu].mt_ctx = ctx;
231 		mt_vmm_info[vcpu].mt_vcpu = vcpu;
232 
233 		error = pthread_create(&mt_vmm_info[vcpu].mt_thr, NULL,
234 				fbsdrun_start_thread, &mt_vmm_info[vcpu]);
235 		assert(error == 0);
236 	}
237 }
238 
239 static int
240 fbsdrun_get_next_cpu(int curcpu)
241 {
242 
243 	/*
244 	 * Get the next available CPU. Assumes they arrive
245 	 * in ascending order with no gaps.
246 	 */
247 	return ((curcpu + 1) % foundcpus);
248 }
249 
250 static int
251 vmexit_catch_reset(void)
252 {
253         stats.io_reset++;
254         return (VMEXIT_RESET);
255 }
256 
257 static int
258 vmexit_catch_inout(void)
259 {
260 	return (VMEXIT_ABORT);
261 }
262 
263 static int
264 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
265 		     uint32_t eax)
266 {
267 #if PG_DEBUG /* put all types of debug here */
268         if (eax == 0) {
269 		pause_noswitch = 1;
270 	} else if (eax == 1) {
271 		pause_noswitch = 0;
272 	} else {
273 		pause_noswitch = 0;
274 		if (eax == 5) {
275 			vm_set_capability(ctx, *pvcpu, VM_CAP_MTRAP_EXIT, 1);
276 		}
277 	}
278 #endif
279         return (VMEXIT_CONTINUE);
280 }
281 
282 static int
283 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
284 {
285 	int error;
286 	int bytes, port, in, out;
287 	uint32_t eax;
288 	int vcpu;
289 
290 	vcpu = *pvcpu;
291 
292 	port = vme->u.inout.port;
293 	bytes = vme->u.inout.bytes;
294 	eax = vme->u.inout.eax;
295 	in = vme->u.inout.in;
296 	out = !in;
297 
298 	/* We don't deal with these */
299 	if (vme->u.inout.string || vme->u.inout.rep)
300 		return (VMEXIT_ABORT);
301 
302 	/* Special case of guest reset */
303 	if (out && port == 0x64 && (uint8_t)eax == 0xFE)
304 		return (vmexit_catch_reset());
305 
306         /* Extra-special case of host notifications */
307         if (out && port == GUEST_NIO_PORT)
308                 return (vmexit_handle_notify(ctx, vme, pvcpu, eax));
309 
310 	error = emulate_inout(ctx, vcpu, in, port, bytes, &eax, strictio);
311 	if (error == 0 && in)
312 		error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RAX, eax);
313 
314 	if (error == 0)
315 		return (VMEXIT_CONTINUE);
316 	else {
317 		fprintf(stderr, "Unhandled %s%c 0x%04x\n",
318 			in ? "in" : "out",
319 			bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), port);
320 		return (vmexit_catch_inout());
321 	}
322 }
323 
324 static int
325 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
326 {
327 	fprintf(stderr, "vm exit rdmsr 0x%x, cpu %d\n", vme->u.msr.code,
328 	    *pvcpu);
329 	return (VMEXIT_ABORT);
330 }
331 
332 static int
333 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
334 {
335 	int newcpu;
336 	int retval = VMEXIT_CONTINUE;
337 
338 	newcpu = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code,vme->u.msr.wval);
339 
340 	if (guest_vcpu_mux && *pvcpu != newcpu) {
341                 retval = VMEXIT_SWITCH;
342                 *pvcpu = newcpu;
343         }
344 
345         return (retval);
346 }
347 
348 static int
349 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
350 {
351 	int newcpu;
352 	int retval = VMEXIT_CONTINUE;
353 
354 	newcpu = spinup_ap(ctx, *pvcpu,
355 			   vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
356 
357 	if (guest_vcpu_mux && *pvcpu != newcpu) {
358 		retval = VMEXIT_SWITCH;
359 		*pvcpu = newcpu;
360 	}
361 
362 	return (retval);
363 }
364 
365 static int
366 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
367 {
368 
369 	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
370 	fprintf(stderr, "\treason\t\tVMX\n");
371 	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
372 	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
373 	fprintf(stderr, "\terror\t\t%d\n", vmexit->u.vmx.error);
374 	fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
375 	fprintf(stderr, "\tqualification\t0x%016lx\n",
376 	    vmexit->u.vmx.exit_qualification);
377 
378 	return (VMEXIT_ABORT);
379 }
380 
381 static int bogus_noswitch = 1;
382 
383 static int
384 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
385 {
386 	stats.vmexit_bogus++;
387 
388 	if (!guest_vcpu_mux || guest_ncpus == 1 || bogus_noswitch) {
389 		return (VMEXIT_RESTART);
390 	} else {
391 		stats.vmexit_bogus_switch++;
392 		vmexit->inst_length = 0;
393 		*pvcpu = -1;
394 		return (VMEXIT_SWITCH);
395 	}
396 }
397 
398 static int
399 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
400 {
401 	stats.vmexit_hlt++;
402 	if (fbsdrun_muxed()) {
403 		*pvcpu = -1;
404 		return (VMEXIT_SWITCH);
405 	} else {
406 		/*
407 		 * Just continue execution with the next instruction. We use
408 		 * the HLT VM exit as a way to be friendly with the host
409 		 * scheduler.
410 		 */
411 		return (VMEXIT_CONTINUE);
412 	}
413 }
414 
415 static int pause_noswitch;
416 
417 static int
418 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
419 {
420 	stats.vmexit_pause++;
421 
422 	if (fbsdrun_muxed() && !pause_noswitch) {
423 		*pvcpu = -1;
424 		return (VMEXIT_SWITCH);
425         } else {
426 		return (VMEXIT_CONTINUE);
427 	}
428 }
429 
430 static int
431 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
432 {
433 	stats.vmexit_mtrap++;
434 
435 	return (VMEXIT_RESTART);
436 }
437 
438 static int
439 vmexit_paging(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
440 {
441 	int err;
442 	stats.vmexit_paging++;
443 
444 	err = emulate_mem(ctx, *pvcpu, vmexit->u.paging.gpa,
445 			  &vmexit->u.paging.vie);
446 
447 	if (err) {
448 		if (err == EINVAL) {
449 			fprintf(stderr,
450 			    "Failed to emulate instruction at 0x%lx\n",
451 			    vmexit->rip);
452 		} else if (err == ESRCH) {
453 			fprintf(stderr, "Unhandled memory access to 0x%lx\n",
454 			    vmexit->u.paging.gpa);
455 		}
456 
457 		return (VMEXIT_ABORT);
458 	}
459 
460 	return (VMEXIT_CONTINUE);
461 }
462 
463 static void
464 sigalrm(int sig)
465 {
466 	return;
467 }
468 
469 static void
470 setup_timeslice(void)
471 {
472 	struct sigaction sa;
473 	struct itimerval itv;
474 	int error;
475 
476 	/*
477 	 * Setup a realtime timer to generate a SIGALRM at a
478 	 * frequency of 'guest_tslice' ticks per second.
479 	 */
480 	sigemptyset(&sa.sa_mask);
481 	sa.sa_flags = 0;
482 	sa.sa_handler = sigalrm;
483 
484 	error = sigaction(SIGALRM, &sa, NULL);
485 	assert(error == 0);
486 
487 	itv.it_interval.tv_sec = 0;
488 	itv.it_interval.tv_usec = 1000000 / guest_tslice;
489 	itv.it_value.tv_sec = 0;
490 	itv.it_value.tv_usec = 1000000 / guest_tslice;
491 
492 	error = setitimer(ITIMER_REAL, &itv, NULL);
493 	assert(error == 0);
494 }
495 
496 static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
497 	[VM_EXITCODE_INOUT]  = vmexit_inout,
498 	[VM_EXITCODE_VMX]    = vmexit_vmx,
499 	[VM_EXITCODE_BOGUS]  = vmexit_bogus,
500 	[VM_EXITCODE_RDMSR]  = vmexit_rdmsr,
501 	[VM_EXITCODE_WRMSR]  = vmexit_wrmsr,
502 	[VM_EXITCODE_MTRAP]  = vmexit_mtrap,
503 	[VM_EXITCODE_PAGING] = vmexit_paging,
504 	[VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
505 };
506 
507 static void
508 vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip)
509 {
510 	cpuset_t mask;
511 	int error, rc, prevcpu;
512 	enum vm_exitcode exitcode;
513 
514 	if (guest_vcpu_mux)
515 		setup_timeslice();
516 
517 	if (pincpu >= 0) {
518 		CPU_ZERO(&mask);
519 		CPU_SET(pincpu + vcpu, &mask);
520 		error = pthread_setaffinity_np(pthread_self(),
521 					       sizeof(mask), &mask);
522 		assert(error == 0);
523 	}
524 
525 	while (1) {
526 		error = vm_run(ctx, vcpu, rip, &vmexit[vcpu]);
527 		if (error != 0) {
528 			/*
529 			 * It is possible that 'vmmctl' or some other process
530 			 * has transitioned the vcpu to CANNOT_RUN state right
531 			 * before we tried to transition it to RUNNING.
532 			 *
533 			 * This is expected to be temporary so just retry.
534 			 */
535 			if (errno == EBUSY)
536 				continue;
537 			else
538 				break;
539 		}
540 
541 		prevcpu = vcpu;
542 
543 		exitcode = vmexit[vcpu].exitcode;
544 		if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
545 			fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
546 			    exitcode);
547 			exit(1);
548 		}
549 
550                 rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
551 
552 		switch (rc) {
553                 case VMEXIT_SWITCH:
554 			assert(guest_vcpu_mux);
555 			if (vcpu == -1) {
556 				stats.cpu_switch_rotate++;
557 				vcpu = fbsdrun_get_next_cpu(prevcpu);
558 			} else {
559 				stats.cpu_switch_direct++;
560 			}
561 			/* fall through */
562 		case VMEXIT_CONTINUE:
563                         rip = vmexit[vcpu].rip + vmexit[vcpu].inst_length;
564 			break;
565 		case VMEXIT_RESTART:
566                         rip = vmexit[vcpu].rip;
567 			break;
568 		case VMEXIT_RESET:
569 			exit(0);
570 		default:
571 			exit(1);
572 		}
573 	}
574 	fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
575 }
576 
577 static int
578 num_vcpus_allowed(struct vmctx *ctx)
579 {
580 	int tmp, error;
581 
582 	error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
583 
584 	/*
585 	 * The guest is allowed to spinup more than one processor only if the
586 	 * UNRESTRICTED_GUEST capability is available.
587 	 */
588 	if (error == 0)
589 		return (VM_MAXCPU);
590 	else
591 		return (1);
592 }
593 
594 int
595 main(int argc, char *argv[])
596 {
597 	int c, error, gdb_port, inject_bkpt, tmp, err, ioapic, bvmcons;
598 	int max_vcpus;
599 	struct vmctx *ctx;
600 	uint64_t rip;
601 	size_t memsize;
602 
603 	bvmcons = 0;
604 	inject_bkpt = 0;
605 	progname = basename(argv[0]);
606 	gdb_port = DEFAULT_GDB_PORT;
607 	guest_ncpus = 1;
608 	ioapic = 0;
609 	memsize = 256 * MB;
610 
611 	while ((c = getopt(argc, argv, "abehABHIPxp:g:c:z:s:S:n:m:")) != -1) {
612 		switch (c) {
613 		case 'a':
614 			disable_x2apic = 1;
615 			break;
616 		case 'A':
617 			acpi = 1;
618 			break;
619 		case 'b':
620 			bvmcons = 1;
621 			break;
622 		case 'B':
623 			inject_bkpt = 1;
624 			break;
625 		case 'x':
626 			guest_vcpu_mux = 1;
627 			break;
628 		case 'p':
629 			pincpu = atoi(optarg);
630 			break;
631                 case 'c':
632 			guest_ncpus = atoi(optarg);
633 			break;
634 		case 'g':
635 			gdb_port = atoi(optarg);
636 			break;
637 		case 'z':
638 			guest_hz = atoi(optarg);
639 			break;
640 		case 't':
641 			guest_tslice = atoi(optarg);
642 			break;
643 		case 's':
644 			if (pci_parse_slot(optarg, 0) != 0)
645 				exit(1);
646 			else
647 				break;
648 		case 'S':
649 			if (pci_parse_slot(optarg, 1) != 0)
650 				exit(1);
651 			else
652 				break;
653                 case 'm':
654 			memsize = strtoul(optarg, NULL, 0) * MB;
655 			break;
656 		case 'H':
657 			guest_vmexit_on_hlt = 1;
658 			break;
659 		case 'I':
660 			ioapic = 1;
661 			break;
662 		case 'P':
663 			guest_vmexit_on_pause = 1;
664 			break;
665 		case 'e':
666 			strictio = 1;
667 			break;
668 		case 'h':
669 			usage(0);
670 		default:
671 			usage(1);
672 		}
673 	}
674 	argc -= optind;
675 	argv += optind;
676 
677 	if (argc != 1)
678 		usage(1);
679 
680 	/* No need to mux if guest is uni-processor */
681 	if (guest_ncpus <= 1)
682 		guest_vcpu_mux = 0;
683 
684 	/* vmexit on hlt if guest is muxed */
685 	if (guest_vcpu_mux) {
686 		guest_vmexit_on_hlt = 1;
687 		guest_vmexit_on_pause = 1;
688 	}
689 
690 	vmname = argv[0];
691 
692 	ctx = vm_open(vmname);
693 	if (ctx == NULL) {
694 		perror("vm_open");
695 		exit(1);
696 	}
697 
698 	max_vcpus = num_vcpus_allowed(ctx);
699 	if (guest_ncpus > max_vcpus) {
700 		fprintf(stderr, "%d vCPUs requested but only %d available\n",
701 			guest_ncpus, max_vcpus);
702 		exit(1);
703 	}
704 
705 	if (fbsdrun_vmexit_on_hlt()) {
706 		err = vm_get_capability(ctx, BSP, VM_CAP_HALT_EXIT, &tmp);
707 		if (err < 0) {
708 			fprintf(stderr, "VM exit on HLT not supported\n");
709 			exit(1);
710 		}
711 		vm_set_capability(ctx, BSP, VM_CAP_HALT_EXIT, 1);
712 		handler[VM_EXITCODE_HLT] = vmexit_hlt;
713 	}
714 
715         if (fbsdrun_vmexit_on_pause()) {
716 		/*
717 		 * pause exit support required for this mode
718 		 */
719 		err = vm_get_capability(ctx, BSP, VM_CAP_PAUSE_EXIT, &tmp);
720 		if (err < 0) {
721 			fprintf(stderr,
722 			    "SMP mux requested, no pause support\n");
723 			exit(1);
724 		}
725 		vm_set_capability(ctx, BSP, VM_CAP_PAUSE_EXIT, 1);
726 		handler[VM_EXITCODE_PAUSE] = vmexit_pause;
727         }
728 
729 	if (fbsdrun_disable_x2apic())
730 		err = vm_set_x2apic_state(ctx, BSP, X2APIC_DISABLED);
731 	else
732 		err = vm_set_x2apic_state(ctx, BSP, X2APIC_ENABLED);
733 
734 	if (err) {
735 		fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
736 		exit(1);
737 	}
738 
739 	err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
740 	if (err) {
741 		fprintf(stderr, "Unable to setup memory (%d)\n", err);
742 		exit(1);
743 	}
744 
745 	init_mem();
746 	init_inout();
747 
748 	rtc_init(ctx);
749 
750 	/*
751 	 * Exit if a device emulation finds an error in it's initilization
752 	 */
753 	if (init_pci(ctx) != 0)
754 		exit(1);
755 
756 	if (ioapic)
757 		ioapic_init(0);
758 
759 	if (gdb_port != 0)
760 		init_dbgport(gdb_port);
761 
762 	if (bvmcons)
763 		init_bvmcons();
764 
765 	error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
766 	assert(error == 0);
767 
768 	if (inject_bkpt) {
769 		error = vm_inject_event(ctx, BSP, VM_HW_EXCEPTION, IDT_BP);
770 		assert(error == 0);
771 	}
772 
773 	/*
774 	 * build the guest tables, MP etc.
775 	 */
776 	mptable_build(ctx, guest_ncpus, ioapic);
777 
778 	if (acpi) {
779 		error = acpi_build(ctx, guest_ncpus, ioapic);
780 		assert(error == 0);
781 	}
782 
783 	/*
784 	 * Add CPU 0
785 	 */
786 	fbsdrun_addcpu(ctx, BSP, rip);
787 
788 	/*
789 	 * Head off to the main event dispatch loop
790 	 */
791 	mevent_dispatch();
792 
793 	exit(1);
794 }
795