xref: /freebsd/usr.sbin/bhyve/bhyverun.c (revision 95d45410b5100e07f6f98450bcd841a8945d4726)
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/atomic.h>
37 #include <machine/segments.h>
38 
39 #include <stdio.h>
40 #include <stdlib.h>
41 #include <string.h>
42 #include <err.h>
43 #include <libgen.h>
44 #include <unistd.h>
45 #include <assert.h>
46 #include <errno.h>
47 #include <pthread.h>
48 #include <pthread_np.h>
49 #include <sysexits.h>
50 
51 #include <machine/vmm.h>
52 #include <vmmapi.h>
53 
54 #include "bhyverun.h"
55 #include "acpi.h"
56 #include "inout.h"
57 #include "dbgport.h"
58 #include "ioapic.h"
59 #include "mem.h"
60 #include "mevent.h"
61 #include "mptbl.h"
62 #include "pci_emul.h"
63 #include "pci_irq.h"
64 #include "pci_lpc.h"
65 #include "smbiostbl.h"
66 #include "xmsr.h"
67 #include "spinup_ap.h"
68 #include "rtc.h"
69 
70 #define GUEST_NIO_PORT		0x488	/* guest upcalls via i/o port */
71 
72 #define MB		(1024UL * 1024)
73 #define GB		(1024UL * MB)
74 
75 typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
76 extern int vmexit_task_switch(struct vmctx *, struct vm_exit *, int *vcpu);
77 
78 char *vmname;
79 
80 int guest_ncpus;
81 char *guest_uuid_str;
82 
83 static int guest_vmexit_on_hlt, guest_vmexit_on_pause;
84 static int virtio_msix = 1;
85 static int x2apic_mode = 0;	/* default is xAPIC */
86 
87 static int strictio;
88 static int strictmsr = 1;
89 
90 static int acpi;
91 
92 static char *progname;
93 static const int BSP = 0;
94 
95 static cpuset_t cpumask;
96 
97 static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
98 
99 struct vm_exit vmexit[VM_MAXCPU];
100 
101 struct bhyvestats {
102         uint64_t        vmexit_bogus;
103         uint64_t        vmexit_bogus_switch;
104         uint64_t        vmexit_hlt;
105         uint64_t        vmexit_pause;
106         uint64_t        vmexit_mtrap;
107         uint64_t        vmexit_inst_emul;
108         uint64_t        cpu_switch_rotate;
109         uint64_t        cpu_switch_direct;
110         int             io_reset;
111 	int		io_poweroff;
112 } stats;
113 
114 struct mt_vmm_info {
115 	pthread_t	mt_thr;
116 	struct vmctx	*mt_ctx;
117 	int		mt_vcpu;
118 } mt_vmm_info[VM_MAXCPU];
119 
120 static cpuset_t *vcpumap[VM_MAXCPU] = { NULL };
121 
122 static void
123 usage(int code)
124 {
125 
126         fprintf(stderr,
127                 "Usage: %s [-abehwxACHPWY] [-c vcpus] [-g <gdb port>] [-l <lpc>]\n"
128 		"       %*s [-m mem] [-p vcpu:hostcpu] [-s <pci>] [-U uuid] <vm>\n"
129 		"       -a: local apic is in xAPIC mode (deprecated)\n"
130 		"       -A: create ACPI tables\n"
131 		"       -c: # cpus (default 1)\n"
132 		"       -C: include guest memory in core file\n"
133 		"       -e: exit on unhandled I/O access\n"
134 		"       -g: gdb port\n"
135 		"       -h: help\n"
136 		"       -H: vmexit from the guest on hlt\n"
137 		"       -l: LPC device configuration\n"
138 		"       -m: memory size in MB\n"
139 		"       -p: pin 'vcpu' to 'hostcpu'\n"
140 		"       -P: vmexit from the guest on pause\n"
141 		"       -s: <slot,driver,configinfo> PCI slot config\n"
142 		"       -U: uuid\n"
143 		"       -w: ignore unimplemented MSRs\n"
144 		"       -W: force virtio to use single-vector MSI\n"
145 		"       -x: local apic is in x2APIC mode\n"
146 		"       -Y: disable MPtable generation\n",
147 		progname, (int)strlen(progname), "");
148 
149 	exit(code);
150 }
151 
152 static int
153 pincpu_parse(const char *opt)
154 {
155 	int vcpu, pcpu;
156 
157 	if (sscanf(opt, "%d:%d", &vcpu, &pcpu) != 2) {
158 		fprintf(stderr, "invalid format: %s\n", opt);
159 		return (-1);
160 	}
161 
162 	if (vcpu < 0 || vcpu >= VM_MAXCPU) {
163 		fprintf(stderr, "vcpu '%d' outside valid range from 0 to %d\n",
164 		    vcpu, VM_MAXCPU - 1);
165 		return (-1);
166 	}
167 
168 	if (pcpu < 0 || pcpu >= CPU_SETSIZE) {
169 		fprintf(stderr, "hostcpu '%d' outside valid range from "
170 		    "0 to %d\n", pcpu, CPU_SETSIZE - 1);
171 		return (-1);
172 	}
173 
174 	if (vcpumap[vcpu] == NULL) {
175 		if ((vcpumap[vcpu] = malloc(sizeof(cpuset_t))) == NULL) {
176 			perror("malloc");
177 			return (-1);
178 		}
179 		CPU_ZERO(vcpumap[vcpu]);
180 	}
181 	CPU_SET(pcpu, vcpumap[vcpu]);
182 	return (0);
183 }
184 
185 void *
186 paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
187 {
188 
189 	return (vm_map_gpa(ctx, gaddr, len));
190 }
191 
192 int
193 fbsdrun_vmexit_on_pause(void)
194 {
195 
196 	return (guest_vmexit_on_pause);
197 }
198 
199 int
200 fbsdrun_vmexit_on_hlt(void)
201 {
202 
203 	return (guest_vmexit_on_hlt);
204 }
205 
206 int
207 fbsdrun_virtio_msix(void)
208 {
209 
210 	return (virtio_msix);
211 }
212 
213 static void *
214 fbsdrun_start_thread(void *param)
215 {
216 	char tname[MAXCOMLEN + 1];
217 	struct mt_vmm_info *mtp;
218 	int vcpu;
219 
220 	mtp = param;
221 	vcpu = mtp->mt_vcpu;
222 
223 	snprintf(tname, sizeof(tname), "vcpu %d", vcpu);
224 	pthread_set_name_np(mtp->mt_thr, tname);
225 
226 	vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
227 
228 	/* not reached */
229 	exit(1);
230 	return (NULL);
231 }
232 
233 void
234 fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip)
235 {
236 	int error;
237 
238 	assert(fromcpu == BSP);
239 
240 	/*
241 	 * The 'newcpu' must be activated in the context of 'fromcpu'. If
242 	 * vm_activate_cpu() is delayed until newcpu's pthread starts running
243 	 * then vmm.ko is out-of-sync with bhyve and this can create a race
244 	 * with vm_suspend().
245 	 */
246 	error = vm_activate_cpu(ctx, newcpu);
247 	assert(error == 0);
248 
249 	CPU_SET_ATOMIC(newcpu, &cpumask);
250 
251 	/*
252 	 * Set up the vmexit struct to allow execution to start
253 	 * at the given RIP
254 	 */
255 	vmexit[newcpu].rip = rip;
256 	vmexit[newcpu].inst_length = 0;
257 
258 	mt_vmm_info[newcpu].mt_ctx = ctx;
259 	mt_vmm_info[newcpu].mt_vcpu = newcpu;
260 
261 	error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL,
262 	    fbsdrun_start_thread, &mt_vmm_info[newcpu]);
263 	assert(error == 0);
264 }
265 
266 static int
267 fbsdrun_deletecpu(struct vmctx *ctx, int vcpu)
268 {
269 
270 	if (!CPU_ISSET(vcpu, &cpumask)) {
271 		fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu);
272 		exit(1);
273 	}
274 
275 	CPU_CLR_ATOMIC(vcpu, &cpumask);
276 	return (CPU_EMPTY(&cpumask));
277 }
278 
279 static int
280 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
281 		     uint32_t eax)
282 {
283 #if BHYVE_DEBUG
284 	/*
285 	 * put guest-driven debug here
286 	 */
287 #endif
288         return (VMEXIT_CONTINUE);
289 }
290 
291 static int
292 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
293 {
294 	int error;
295 	int bytes, port, in, out, string;
296 	int vcpu;
297 
298 	vcpu = *pvcpu;
299 
300 	port = vme->u.inout.port;
301 	bytes = vme->u.inout.bytes;
302 	string = vme->u.inout.string;
303 	in = vme->u.inout.in;
304 	out = !in;
305 
306         /* Extra-special case of host notifications */
307         if (out && port == GUEST_NIO_PORT) {
308                 error = vmexit_handle_notify(ctx, vme, pvcpu, vme->u.inout.eax);
309 		return (error);
310 	}
311 
312 	error = emulate_inout(ctx, vcpu, vme, strictio);
313 	if (error == INOUT_OK && in && !string) {
314 		error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RAX,
315 		    vme->u.inout.eax);
316 	}
317 
318 	switch (error) {
319 	case INOUT_OK:
320 		return (VMEXIT_CONTINUE);
321 	case INOUT_RESTART:
322 		return (VMEXIT_RESTART);
323 	case INOUT_RESET:
324 		stats.io_reset++;
325 		return (VMEXIT_RESET);
326 	case INOUT_POWEROFF:
327 		stats.io_poweroff++;
328 		return (VMEXIT_POWEROFF);
329 	default:
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_ABORT);
334 	}
335 }
336 
337 static int
338 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
339 {
340 	uint64_t val;
341 	uint32_t eax, edx;
342 	int error;
343 
344 	val = 0;
345 	error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val);
346 	if (error != 0) {
347 		fprintf(stderr, "rdmsr to register %#x on vcpu %d\n",
348 		    vme->u.msr.code, *pvcpu);
349 		if (strictmsr) {
350 			error = vm_inject_exception2(ctx, *pvcpu, IDT_GP, 0);
351 			assert(error == 0);
352 			return (VMEXIT_RESTART);
353 		}
354 	}
355 
356 	eax = val;
357 	error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax);
358 	assert(error == 0);
359 
360 	edx = val >> 32;
361 	error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx);
362 	assert(error == 0);
363 
364 	return (VMEXIT_CONTINUE);
365 }
366 
367 static int
368 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
369 {
370 	int error;
371 
372 	error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval);
373 	if (error != 0) {
374 		fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n",
375 		    vme->u.msr.code, vme->u.msr.wval, *pvcpu);
376 		if (strictmsr) {
377 			error = vm_inject_exception2(ctx, *pvcpu, IDT_GP, 0);
378 			assert(error == 0);
379 			return (VMEXIT_RESTART);
380 		}
381 	}
382 	return (VMEXIT_CONTINUE);
383 }
384 
385 static int
386 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
387 {
388 	int newcpu;
389 	int retval = VMEXIT_CONTINUE;
390 
391 	newcpu = spinup_ap(ctx, *pvcpu,
392 			   vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
393 
394 	return (retval);
395 }
396 
397 #define	DEBUG_EPT_MISCONFIG
398 #ifdef DEBUG_EPT_MISCONFIG
399 #define	EXIT_REASON_EPT_MISCONFIG	49
400 #define	VMCS_GUEST_PHYSICAL_ADDRESS	0x00002400
401 #define	VMCS_IDENT(x)			((x) | 0x80000000)
402 
403 static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4];
404 static int ept_misconfig_ptenum;
405 #endif
406 
407 static int
408 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
409 {
410 
411 	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
412 	fprintf(stderr, "\treason\t\tVMX\n");
413 	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
414 	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
415 	fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status);
416 	fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
417 	fprintf(stderr, "\tqualification\t0x%016lx\n",
418 	    vmexit->u.vmx.exit_qualification);
419 	fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type);
420 	fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error);
421 #ifdef DEBUG_EPT_MISCONFIG
422 	if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) {
423 		vm_get_register(ctx, *pvcpu,
424 		    VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS),
425 		    &ept_misconfig_gpa);
426 		vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte,
427 		    &ept_misconfig_ptenum);
428 		fprintf(stderr, "\tEPT misconfiguration:\n");
429 		fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa);
430 		fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n",
431 		    ept_misconfig_ptenum, ept_misconfig_pte[0],
432 		    ept_misconfig_pte[1], ept_misconfig_pte[2],
433 		    ept_misconfig_pte[3]);
434 	}
435 #endif	/* DEBUG_EPT_MISCONFIG */
436 	return (VMEXIT_ABORT);
437 }
438 
439 static int
440 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
441 {
442 
443 	stats.vmexit_bogus++;
444 
445 	return (VMEXIT_RESTART);
446 }
447 
448 static int
449 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
450 {
451 
452 	stats.vmexit_hlt++;
453 
454 	/*
455 	 * Just continue execution with the next instruction. We use
456 	 * the HLT VM exit as a way to be friendly with the host
457 	 * scheduler.
458 	 */
459 	return (VMEXIT_CONTINUE);
460 }
461 
462 static int
463 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
464 {
465 
466 	stats.vmexit_pause++;
467 
468 	return (VMEXIT_CONTINUE);
469 }
470 
471 static int
472 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
473 {
474 
475 	stats.vmexit_mtrap++;
476 
477 	return (VMEXIT_RESTART);
478 }
479 
480 static int
481 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
482 {
483 	int err;
484 	stats.vmexit_inst_emul++;
485 
486 	err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
487 			  &vmexit->u.inst_emul.vie);
488 
489 	if (err) {
490 		if (err == EINVAL) {
491 			fprintf(stderr,
492 			    "Failed to emulate instruction at 0x%lx\n",
493 			    vmexit->rip);
494 		} else if (err == ESRCH) {
495 			fprintf(stderr, "Unhandled memory access to 0x%lx\n",
496 			    vmexit->u.inst_emul.gpa);
497 		}
498 
499 		return (VMEXIT_ABORT);
500 	}
501 
502 	return (VMEXIT_CONTINUE);
503 }
504 
505 static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER;
506 static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER;
507 
508 static int
509 vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
510 {
511 	enum vm_suspend_how how;
512 
513 	how = vmexit->u.suspended.how;
514 
515 	fbsdrun_deletecpu(ctx, *pvcpu);
516 
517 	if (*pvcpu != BSP) {
518 		pthread_mutex_lock(&resetcpu_mtx);
519 		pthread_cond_signal(&resetcpu_cond);
520 		pthread_mutex_unlock(&resetcpu_mtx);
521 		pthread_exit(NULL);
522 	}
523 
524 	pthread_mutex_lock(&resetcpu_mtx);
525 	while (!CPU_EMPTY(&cpumask)) {
526 		pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx);
527 	}
528 	pthread_mutex_unlock(&resetcpu_mtx);
529 
530 	switch (how) {
531 	case VM_SUSPEND_RESET:
532 		exit(0);
533 	case VM_SUSPEND_POWEROFF:
534 		exit(1);
535 	case VM_SUSPEND_HALT:
536 		exit(2);
537 	case VM_SUSPEND_TRIPLEFAULT:
538 		exit(3);
539 	default:
540 		fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how);
541 		exit(100);
542 	}
543 	return (0);	/* NOTREACHED */
544 }
545 
546 static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
547 	[VM_EXITCODE_INOUT]  = vmexit_inout,
548 	[VM_EXITCODE_INOUT_STR]  = vmexit_inout,
549 	[VM_EXITCODE_VMX]    = vmexit_vmx,
550 	[VM_EXITCODE_BOGUS]  = vmexit_bogus,
551 	[VM_EXITCODE_RDMSR]  = vmexit_rdmsr,
552 	[VM_EXITCODE_WRMSR]  = vmexit_wrmsr,
553 	[VM_EXITCODE_MTRAP]  = vmexit_mtrap,
554 	[VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
555 	[VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
556 	[VM_EXITCODE_SUSPENDED] = vmexit_suspend,
557 	[VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch,
558 };
559 
560 static void
561 vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip)
562 {
563 	int error, rc, prevcpu;
564 	enum vm_exitcode exitcode;
565 	enum vm_suspend_how how;
566 	cpuset_t active_cpus;
567 
568 	if (vcpumap[vcpu] != NULL) {
569 		error = pthread_setaffinity_np(pthread_self(),
570 		    sizeof(cpuset_t), vcpumap[vcpu]);
571 		assert(error == 0);
572 	}
573 
574 	error = vm_active_cpus(ctx, &active_cpus);
575 	assert(CPU_ISSET(vcpu, &active_cpus));
576 
577 	while (1) {
578 		error = vm_run(ctx, vcpu, rip, &vmexit[vcpu]);
579 		if (error != 0)
580 			break;
581 
582 		prevcpu = vcpu;
583 
584 		exitcode = vmexit[vcpu].exitcode;
585 		if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
586 			fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
587 			    exitcode);
588 			exit(1);
589 		}
590 
591                 rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
592 
593 		switch (rc) {
594 		case VMEXIT_CONTINUE:
595                         rip = vmexit[vcpu].rip + vmexit[vcpu].inst_length;
596 			break;
597 		case VMEXIT_RESTART:
598                         rip = vmexit[vcpu].rip;
599 			break;
600 		case VMEXIT_RESET:
601 		case VMEXIT_POWEROFF:
602 			if (rc == VMEXIT_RESET)
603 				how = VM_SUSPEND_RESET;
604 			else
605 				how = VM_SUSPEND_POWEROFF;
606 			error = vm_suspend(ctx, how);
607 			assert(error == 0 || errno == EALREADY);
608                         rip = vmexit[vcpu].rip + vmexit[vcpu].inst_length;
609 			break;
610 		case VMEXIT_ABORT:
611 			abort();
612 		default:
613 			exit(1);
614 		}
615 	}
616 	fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
617 }
618 
619 static int
620 num_vcpus_allowed(struct vmctx *ctx)
621 {
622 	int tmp, error;
623 
624 	error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
625 
626 	/*
627 	 * The guest is allowed to spinup more than one processor only if the
628 	 * UNRESTRICTED_GUEST capability is available.
629 	 */
630 	if (error == 0)
631 		return (VM_MAXCPU);
632 	else
633 		return (1);
634 }
635 
636 void
637 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
638 {
639 	int err, tmp;
640 
641 	if (fbsdrun_vmexit_on_hlt()) {
642 		err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
643 		if (err < 0) {
644 			fprintf(stderr, "VM exit on HLT not supported\n");
645 			exit(1);
646 		}
647 		vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
648 		if (cpu == BSP)
649 			handler[VM_EXITCODE_HLT] = vmexit_hlt;
650 	}
651 
652         if (fbsdrun_vmexit_on_pause()) {
653 		/*
654 		 * pause exit support required for this mode
655 		 */
656 		err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
657 		if (err < 0) {
658 			fprintf(stderr,
659 			    "SMP mux requested, no pause support\n");
660 			exit(1);
661 		}
662 		vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
663 		if (cpu == BSP)
664 			handler[VM_EXITCODE_PAUSE] = vmexit_pause;
665         }
666 
667 	if (x2apic_mode)
668 		err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
669 	else
670 		err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
671 
672 	if (err) {
673 		fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
674 		exit(1);
675 	}
676 
677 	vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
678 }
679 
680 int
681 main(int argc, char *argv[])
682 {
683 	int c, error, gdb_port, err, bvmcons;
684 	int dump_guest_memory, max_vcpus, mptgen;
685 	struct vmctx *ctx;
686 	uint64_t rip;
687 	size_t memsize;
688 
689 	bvmcons = 0;
690 	dump_guest_memory = 0;
691 	progname = basename(argv[0]);
692 	gdb_port = 0;
693 	guest_ncpus = 1;
694 	memsize = 256 * MB;
695 	mptgen = 1;
696 
697 	while ((c = getopt(argc, argv, "abehwxACHIPWYp:g:c:s:m:l:U:")) != -1) {
698 		switch (c) {
699 		case 'a':
700 			x2apic_mode = 0;
701 			break;
702 		case 'A':
703 			acpi = 1;
704 			break;
705 		case 'b':
706 			bvmcons = 1;
707 			break;
708 		case 'p':
709                         if (pincpu_parse(optarg) != 0) {
710                             errx(EX_USAGE, "invalid vcpu pinning "
711                                  "configuration '%s'", optarg);
712                         }
713 			break;
714                 case 'c':
715 			guest_ncpus = atoi(optarg);
716 			break;
717 		case 'C':
718 			dump_guest_memory = 1;
719 			break;
720 		case 'g':
721 			gdb_port = atoi(optarg);
722 			break;
723 		case 'l':
724 			if (lpc_device_parse(optarg) != 0) {
725 				errx(EX_USAGE, "invalid lpc device "
726 				    "configuration '%s'", optarg);
727 			}
728 			break;
729 		case 's':
730 			if (pci_parse_slot(optarg) != 0)
731 				exit(1);
732 			else
733 				break;
734                 case 'm':
735 			error = vm_parse_memsize(optarg, &memsize);
736 			if (error)
737 				errx(EX_USAGE, "invalid memsize '%s'", optarg);
738 			break;
739 		case 'H':
740 			guest_vmexit_on_hlt = 1;
741 			break;
742 		case 'I':
743 			/*
744 			 * The "-I" option was used to add an ioapic to the
745 			 * virtual machine.
746 			 *
747 			 * An ioapic is now provided unconditionally for each
748 			 * virtual machine and this option is now deprecated.
749 			 */
750 			break;
751 		case 'P':
752 			guest_vmexit_on_pause = 1;
753 			break;
754 		case 'e':
755 			strictio = 1;
756 			break;
757 		case 'U':
758 			guest_uuid_str = optarg;
759 			break;
760 		case 'w':
761 			strictmsr = 0;
762 			break;
763 		case 'W':
764 			virtio_msix = 0;
765 			break;
766 		case 'x':
767 			x2apic_mode = 1;
768 			break;
769 		case 'Y':
770 			mptgen = 0;
771 			break;
772 		case 'h':
773 			usage(0);
774 		default:
775 			usage(1);
776 		}
777 	}
778 	argc -= optind;
779 	argv += optind;
780 
781 	if (argc != 1)
782 		usage(1);
783 
784 	vmname = argv[0];
785 
786 	ctx = vm_open(vmname);
787 	if (ctx == NULL) {
788 		perror("vm_open");
789 		exit(1);
790 	}
791 
792 	max_vcpus = num_vcpus_allowed(ctx);
793 	if (guest_ncpus > max_vcpus) {
794 		fprintf(stderr, "%d vCPUs requested but only %d available\n",
795 			guest_ncpus, max_vcpus);
796 		exit(1);
797 	}
798 
799 	fbsdrun_set_capabilities(ctx, BSP);
800 
801 	if (dump_guest_memory)
802 		vm_set_memflags(ctx, VM_MEM_F_INCORE);
803 	err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
804 	if (err) {
805 		fprintf(stderr, "Unable to setup memory (%d)\n", err);
806 		exit(1);
807 	}
808 
809 	init_mem();
810 	init_inout();
811 	pci_irq_init(ctx);
812 	ioapic_init(ctx);
813 
814 	rtc_init(ctx);
815 	sci_init(ctx);
816 
817 	/*
818 	 * Exit if a device emulation finds an error in it's initilization
819 	 */
820 	if (init_pci(ctx) != 0)
821 		exit(1);
822 
823 	if (gdb_port != 0)
824 		init_dbgport(gdb_port);
825 
826 	if (bvmcons)
827 		init_bvmcons();
828 
829 	error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
830 	assert(error == 0);
831 
832 	/*
833 	 * build the guest tables, MP etc.
834 	 */
835 	if (mptgen) {
836 		error = mptable_build(ctx, guest_ncpus);
837 		if (error)
838 			exit(1);
839 	}
840 
841 	error = smbios_build(ctx);
842 	assert(error == 0);
843 
844 	if (acpi) {
845 		error = acpi_build(ctx, guest_ncpus);
846 		assert(error == 0);
847 	}
848 
849 	/*
850 	 * Change the proc title to include the VM name.
851 	 */
852 	setproctitle("%s", vmname);
853 
854 	/*
855 	 * Add CPU 0
856 	 */
857 	fbsdrun_addcpu(ctx, BSP, BSP, rip);
858 
859 	/*
860 	 * Head off to the main event dispatch loop
861 	 */
862 	mevent_dispatch();
863 
864 	exit(1);
865 }
866