xref: /freebsd/usr.sbin/bhyve/bhyverun.c (revision f305e6eaf8d98d9c2ca7ca97791e84521f25b801)
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 static 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 } stats;
111 
112 struct mt_vmm_info {
113 	pthread_t	mt_thr;
114 	struct vmctx	*mt_ctx;
115 	int		mt_vcpu;
116 } mt_vmm_info[VM_MAXCPU];
117 
118 static cpuset_t *vcpumap[VM_MAXCPU] = { NULL };
119 
120 static void
121 usage(int code)
122 {
123 
124         fprintf(stderr,
125                 "Usage: %s [-abehuwxACHPWY] [-c vcpus] [-g <gdb port>] [-l <lpc>]\n"
126 		"       %*s [-m mem] [-p vcpu:hostcpu] [-s <pci>] [-U uuid] <vm>\n"
127 		"       -a: local apic is in xAPIC mode (deprecated)\n"
128 		"       -A: create ACPI tables\n"
129 		"       -c: # cpus (default 1)\n"
130 		"       -C: include guest memory in core file\n"
131 		"       -e: exit on unhandled I/O access\n"
132 		"       -g: gdb port\n"
133 		"       -h: help\n"
134 		"       -H: vmexit from the guest on hlt\n"
135 		"       -l: LPC device configuration\n"
136 		"       -m: memory size in MB\n"
137 		"       -p: pin 'vcpu' to 'hostcpu'\n"
138 		"       -P: vmexit from the guest on pause\n"
139 		"       -s: <slot,driver,configinfo> PCI slot config\n"
140 		"       -u: RTC keeps UTC time\n"
141 		"       -U: uuid\n"
142 		"       -w: ignore unimplemented MSRs\n"
143 		"       -W: force virtio to use single-vector MSI\n"
144 		"       -x: local apic is in x2APIC mode\n"
145 		"       -Y: disable MPtable generation\n",
146 		progname, (int)strlen(progname), "");
147 
148 	exit(code);
149 }
150 
151 static int
152 pincpu_parse(const char *opt)
153 {
154 	int vcpu, pcpu;
155 
156 	if (sscanf(opt, "%d:%d", &vcpu, &pcpu) != 2) {
157 		fprintf(stderr, "invalid format: %s\n", opt);
158 		return (-1);
159 	}
160 
161 	if (vcpu < 0 || vcpu >= VM_MAXCPU) {
162 		fprintf(stderr, "vcpu '%d' outside valid range from 0 to %d\n",
163 		    vcpu, VM_MAXCPU - 1);
164 		return (-1);
165 	}
166 
167 	if (pcpu < 0 || pcpu >= CPU_SETSIZE) {
168 		fprintf(stderr, "hostcpu '%d' outside valid range from "
169 		    "0 to %d\n", pcpu, CPU_SETSIZE - 1);
170 		return (-1);
171 	}
172 
173 	if (vcpumap[vcpu] == NULL) {
174 		if ((vcpumap[vcpu] = malloc(sizeof(cpuset_t))) == NULL) {
175 			perror("malloc");
176 			return (-1);
177 		}
178 		CPU_ZERO(vcpumap[vcpu]);
179 	}
180 	CPU_SET(pcpu, vcpumap[vcpu]);
181 	return (0);
182 }
183 
184 void
185 vm_inject_fault(void *arg, int vcpu, int vector, int errcode_valid,
186     int errcode)
187 {
188 	struct vmctx *ctx;
189 	int error, restart_instruction;
190 
191 	ctx = arg;
192 	restart_instruction = 1;
193 
194 	error = vm_inject_exception(ctx, vcpu, vector, errcode_valid, errcode,
195 	    restart_instruction);
196 	assert(error == 0);
197 }
198 
199 void *
200 paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
201 {
202 
203 	return (vm_map_gpa(ctx, gaddr, len));
204 }
205 
206 int
207 fbsdrun_vmexit_on_pause(void)
208 {
209 
210 	return (guest_vmexit_on_pause);
211 }
212 
213 int
214 fbsdrun_vmexit_on_hlt(void)
215 {
216 
217 	return (guest_vmexit_on_hlt);
218 }
219 
220 int
221 fbsdrun_virtio_msix(void)
222 {
223 
224 	return (virtio_msix);
225 }
226 
227 static void *
228 fbsdrun_start_thread(void *param)
229 {
230 	char tname[MAXCOMLEN + 1];
231 	struct mt_vmm_info *mtp;
232 	int vcpu;
233 
234 	mtp = param;
235 	vcpu = mtp->mt_vcpu;
236 
237 	snprintf(tname, sizeof(tname), "vcpu %d", vcpu);
238 	pthread_set_name_np(mtp->mt_thr, tname);
239 
240 	vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
241 
242 	/* not reached */
243 	exit(1);
244 	return (NULL);
245 }
246 
247 void
248 fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip)
249 {
250 	int error;
251 
252 	assert(fromcpu == BSP);
253 
254 	/*
255 	 * The 'newcpu' must be activated in the context of 'fromcpu'. If
256 	 * vm_activate_cpu() is delayed until newcpu's pthread starts running
257 	 * then vmm.ko is out-of-sync with bhyve and this can create a race
258 	 * with vm_suspend().
259 	 */
260 	error = vm_activate_cpu(ctx, newcpu);
261 	assert(error == 0);
262 
263 	CPU_SET_ATOMIC(newcpu, &cpumask);
264 
265 	/*
266 	 * Set up the vmexit struct to allow execution to start
267 	 * at the given RIP
268 	 */
269 	vmexit[newcpu].rip = rip;
270 	vmexit[newcpu].inst_length = 0;
271 
272 	mt_vmm_info[newcpu].mt_ctx = ctx;
273 	mt_vmm_info[newcpu].mt_vcpu = newcpu;
274 
275 	error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL,
276 	    fbsdrun_start_thread, &mt_vmm_info[newcpu]);
277 	assert(error == 0);
278 }
279 
280 static int
281 fbsdrun_deletecpu(struct vmctx *ctx, int vcpu)
282 {
283 
284 	if (!CPU_ISSET(vcpu, &cpumask)) {
285 		fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu);
286 		exit(1);
287 	}
288 
289 	CPU_CLR_ATOMIC(vcpu, &cpumask);
290 	return (CPU_EMPTY(&cpumask));
291 }
292 
293 static int
294 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
295 		     uint32_t eax)
296 {
297 #if BHYVE_DEBUG
298 	/*
299 	 * put guest-driven debug here
300 	 */
301 #endif
302         return (VMEXIT_CONTINUE);
303 }
304 
305 static int
306 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
307 {
308 	int error;
309 	int bytes, port, in, out, string;
310 	int vcpu;
311 
312 	vcpu = *pvcpu;
313 
314 	port = vme->u.inout.port;
315 	bytes = vme->u.inout.bytes;
316 	string = vme->u.inout.string;
317 	in = vme->u.inout.in;
318 	out = !in;
319 
320         /* Extra-special case of host notifications */
321         if (out && port == GUEST_NIO_PORT) {
322                 error = vmexit_handle_notify(ctx, vme, pvcpu, vme->u.inout.eax);
323 		return (error);
324 	}
325 
326 	error = emulate_inout(ctx, vcpu, vme, strictio);
327 	if (error) {
328 		fprintf(stderr, "Unhandled %s%c 0x%04x\n", in ? "in" : "out",
329 		    bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), port);
330 		return (VMEXIT_ABORT);
331 	} else {
332 		return (VMEXIT_CONTINUE);
333 	}
334 }
335 
336 static int
337 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
338 {
339 	uint64_t val;
340 	uint32_t eax, edx;
341 	int error;
342 
343 	val = 0;
344 	error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val);
345 	if (error != 0) {
346 		fprintf(stderr, "rdmsr to register %#x on vcpu %d\n",
347 		    vme->u.msr.code, *pvcpu);
348 		if (strictmsr) {
349 			vm_inject_gp(ctx, *pvcpu);
350 			return (VMEXIT_CONTINUE);
351 		}
352 	}
353 
354 	eax = val;
355 	error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax);
356 	assert(error == 0);
357 
358 	edx = val >> 32;
359 	error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx);
360 	assert(error == 0);
361 
362 	return (VMEXIT_CONTINUE);
363 }
364 
365 static int
366 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
367 {
368 	int error;
369 
370 	error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval);
371 	if (error != 0) {
372 		fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n",
373 		    vme->u.msr.code, vme->u.msr.wval, *pvcpu);
374 		if (strictmsr) {
375 			vm_inject_gp(ctx, *pvcpu);
376 			return (VMEXIT_CONTINUE);
377 		}
378 	}
379 	return (VMEXIT_CONTINUE);
380 }
381 
382 static int
383 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
384 {
385 	int newcpu;
386 	int retval = VMEXIT_CONTINUE;
387 
388 	newcpu = spinup_ap(ctx, *pvcpu,
389 			   vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
390 
391 	return (retval);
392 }
393 
394 #define	DEBUG_EPT_MISCONFIG
395 #ifdef DEBUG_EPT_MISCONFIG
396 #define	EXIT_REASON_EPT_MISCONFIG	49
397 #define	VMCS_GUEST_PHYSICAL_ADDRESS	0x00002400
398 #define	VMCS_IDENT(x)			((x) | 0x80000000)
399 
400 static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4];
401 static int ept_misconfig_ptenum;
402 #endif
403 
404 static int
405 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
406 {
407 
408 	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
409 	fprintf(stderr, "\treason\t\tVMX\n");
410 	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
411 	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
412 	fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status);
413 	fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
414 	fprintf(stderr, "\tqualification\t0x%016lx\n",
415 	    vmexit->u.vmx.exit_qualification);
416 	fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type);
417 	fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error);
418 #ifdef DEBUG_EPT_MISCONFIG
419 	if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) {
420 		vm_get_register(ctx, *pvcpu,
421 		    VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS),
422 		    &ept_misconfig_gpa);
423 		vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte,
424 		    &ept_misconfig_ptenum);
425 		fprintf(stderr, "\tEPT misconfiguration:\n");
426 		fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa);
427 		fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n",
428 		    ept_misconfig_ptenum, ept_misconfig_pte[0],
429 		    ept_misconfig_pte[1], ept_misconfig_pte[2],
430 		    ept_misconfig_pte[3]);
431 	}
432 #endif	/* DEBUG_EPT_MISCONFIG */
433 	return (VMEXIT_ABORT);
434 }
435 
436 static int
437 vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
438 {
439 
440 	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
441 	fprintf(stderr, "\treason\t\tSVM\n");
442 	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
443 	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
444 	fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode);
445 	fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1);
446 	fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2);
447 	return (VMEXIT_ABORT);
448 }
449 
450 static int
451 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
452 {
453 
454 	assert(vmexit->inst_length == 0);
455 
456 	stats.vmexit_bogus++;
457 
458 	return (VMEXIT_CONTINUE);
459 }
460 
461 static int
462 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
463 {
464 
465 	stats.vmexit_hlt++;
466 
467 	/*
468 	 * Just continue execution with the next instruction. We use
469 	 * the HLT VM exit as a way to be friendly with the host
470 	 * scheduler.
471 	 */
472 	return (VMEXIT_CONTINUE);
473 }
474 
475 static int
476 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
477 {
478 
479 	stats.vmexit_pause++;
480 
481 	return (VMEXIT_CONTINUE);
482 }
483 
484 static int
485 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
486 {
487 
488 	assert(vmexit->inst_length == 0);
489 
490 	stats.vmexit_mtrap++;
491 
492 	return (VMEXIT_CONTINUE);
493 }
494 
495 static int
496 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
497 {
498 	int err, i;
499 	struct vie *vie;
500 
501 	stats.vmexit_inst_emul++;
502 
503 	vie = &vmexit->u.inst_emul.vie;
504 	err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
505 	    vie, &vmexit->u.inst_emul.paging);
506 
507 	if (err) {
508 		if (err == ESRCH) {
509 			fprintf(stderr, "Unhandled memory access to 0x%lx\n",
510 			    vmexit->u.inst_emul.gpa);
511 		}
512 
513 		fprintf(stderr, "Failed to emulate instruction [");
514 		for (i = 0; i < vie->num_valid; i++) {
515 			fprintf(stderr, "0x%02x%s", vie->inst[i],
516 			    i != (vie->num_valid - 1) ? " " : "");
517 		}
518 		fprintf(stderr, "] at 0x%lx\n", vmexit->rip);
519 		return (VMEXIT_ABORT);
520 	}
521 
522 	return (VMEXIT_CONTINUE);
523 }
524 
525 static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER;
526 static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER;
527 
528 static int
529 vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
530 {
531 	enum vm_suspend_how how;
532 
533 	how = vmexit->u.suspended.how;
534 
535 	fbsdrun_deletecpu(ctx, *pvcpu);
536 
537 	if (*pvcpu != BSP) {
538 		pthread_mutex_lock(&resetcpu_mtx);
539 		pthread_cond_signal(&resetcpu_cond);
540 		pthread_mutex_unlock(&resetcpu_mtx);
541 		pthread_exit(NULL);
542 	}
543 
544 	pthread_mutex_lock(&resetcpu_mtx);
545 	while (!CPU_EMPTY(&cpumask)) {
546 		pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx);
547 	}
548 	pthread_mutex_unlock(&resetcpu_mtx);
549 
550 	switch (how) {
551 	case VM_SUSPEND_RESET:
552 		exit(0);
553 	case VM_SUSPEND_POWEROFF:
554 		exit(1);
555 	case VM_SUSPEND_HALT:
556 		exit(2);
557 	case VM_SUSPEND_TRIPLEFAULT:
558 		exit(3);
559 	default:
560 		fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how);
561 		exit(100);
562 	}
563 	return (0);	/* NOTREACHED */
564 }
565 
566 static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
567 	[VM_EXITCODE_INOUT]  = vmexit_inout,
568 	[VM_EXITCODE_INOUT_STR]  = vmexit_inout,
569 	[VM_EXITCODE_VMX]    = vmexit_vmx,
570 	[VM_EXITCODE_SVM]    = vmexit_svm,
571 	[VM_EXITCODE_BOGUS]  = vmexit_bogus,
572 	[VM_EXITCODE_RDMSR]  = vmexit_rdmsr,
573 	[VM_EXITCODE_WRMSR]  = vmexit_wrmsr,
574 	[VM_EXITCODE_MTRAP]  = vmexit_mtrap,
575 	[VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
576 	[VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
577 	[VM_EXITCODE_SUSPENDED] = vmexit_suspend,
578 	[VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch,
579 };
580 
581 static void
582 vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip)
583 {
584 	int error, rc, prevcpu;
585 	enum vm_exitcode exitcode;
586 	cpuset_t active_cpus;
587 
588 	if (vcpumap[vcpu] != NULL) {
589 		error = pthread_setaffinity_np(pthread_self(),
590 		    sizeof(cpuset_t), vcpumap[vcpu]);
591 		assert(error == 0);
592 	}
593 
594 	error = vm_active_cpus(ctx, &active_cpus);
595 	assert(CPU_ISSET(vcpu, &active_cpus));
596 
597 	error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip);
598 	assert(error == 0);
599 
600 	while (1) {
601 		error = vm_run(ctx, vcpu, &vmexit[vcpu]);
602 		if (error != 0)
603 			break;
604 
605 		prevcpu = vcpu;
606 
607 		exitcode = vmexit[vcpu].exitcode;
608 		if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
609 			fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
610 			    exitcode);
611 			exit(1);
612 		}
613 
614                 rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
615 
616 		switch (rc) {
617 		case VMEXIT_CONTINUE:
618 			break;
619 		case VMEXIT_ABORT:
620 			abort();
621 		default:
622 			exit(1);
623 		}
624 	}
625 	fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
626 }
627 
628 static int
629 num_vcpus_allowed(struct vmctx *ctx)
630 {
631 	int tmp, error;
632 
633 	error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
634 
635 	/*
636 	 * The guest is allowed to spinup more than one processor only if the
637 	 * UNRESTRICTED_GUEST capability is available.
638 	 */
639 	if (error == 0)
640 		return (VM_MAXCPU);
641 	else
642 		return (1);
643 }
644 
645 void
646 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
647 {
648 	int err, tmp;
649 
650 	if (fbsdrun_vmexit_on_hlt()) {
651 		err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
652 		if (err < 0) {
653 			fprintf(stderr, "VM exit on HLT not supported\n");
654 			exit(1);
655 		}
656 		vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
657 		if (cpu == BSP)
658 			handler[VM_EXITCODE_HLT] = vmexit_hlt;
659 	}
660 
661         if (fbsdrun_vmexit_on_pause()) {
662 		/*
663 		 * pause exit support required for this mode
664 		 */
665 		err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
666 		if (err < 0) {
667 			fprintf(stderr,
668 			    "SMP mux requested, no pause support\n");
669 			exit(1);
670 		}
671 		vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
672 		if (cpu == BSP)
673 			handler[VM_EXITCODE_PAUSE] = vmexit_pause;
674         }
675 
676 	if (x2apic_mode)
677 		err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
678 	else
679 		err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
680 
681 	if (err) {
682 		fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
683 		exit(1);
684 	}
685 
686 	vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
687 }
688 
689 int
690 main(int argc, char *argv[])
691 {
692 	int c, error, gdb_port, err, bvmcons;
693 	int dump_guest_memory, max_vcpus, mptgen;
694 	int rtc_localtime;
695 	struct vmctx *ctx;
696 	uint64_t rip;
697 	size_t memsize;
698 
699 	bvmcons = 0;
700 	dump_guest_memory = 0;
701 	progname = basename(argv[0]);
702 	gdb_port = 0;
703 	guest_ncpus = 1;
704 	memsize = 256 * MB;
705 	mptgen = 1;
706 	rtc_localtime = 1;
707 
708 	while ((c = getopt(argc, argv, "abehuwxACHIPWYp:g:c:s:m:l:U:")) != -1) {
709 		switch (c) {
710 		case 'a':
711 			x2apic_mode = 0;
712 			break;
713 		case 'A':
714 			acpi = 1;
715 			break;
716 		case 'b':
717 			bvmcons = 1;
718 			break;
719 		case 'p':
720                         if (pincpu_parse(optarg) != 0) {
721                             errx(EX_USAGE, "invalid vcpu pinning "
722                                  "configuration '%s'", optarg);
723                         }
724 			break;
725                 case 'c':
726 			guest_ncpus = atoi(optarg);
727 			break;
728 		case 'C':
729 			dump_guest_memory = 1;
730 			break;
731 		case 'g':
732 			gdb_port = atoi(optarg);
733 			break;
734 		case 'l':
735 			if (lpc_device_parse(optarg) != 0) {
736 				errx(EX_USAGE, "invalid lpc device "
737 				    "configuration '%s'", optarg);
738 			}
739 			break;
740 		case 's':
741 			if (pci_parse_slot(optarg) != 0)
742 				exit(1);
743 			else
744 				break;
745                 case 'm':
746 			error = vm_parse_memsize(optarg, &memsize);
747 			if (error)
748 				errx(EX_USAGE, "invalid memsize '%s'", optarg);
749 			break;
750 		case 'H':
751 			guest_vmexit_on_hlt = 1;
752 			break;
753 		case 'I':
754 			/*
755 			 * The "-I" option was used to add an ioapic to the
756 			 * virtual machine.
757 			 *
758 			 * An ioapic is now provided unconditionally for each
759 			 * virtual machine and this option is now deprecated.
760 			 */
761 			break;
762 		case 'P':
763 			guest_vmexit_on_pause = 1;
764 			break;
765 		case 'e':
766 			strictio = 1;
767 			break;
768 		case 'u':
769 			rtc_localtime = 0;
770 			break;
771 		case 'U':
772 			guest_uuid_str = optarg;
773 			break;
774 		case 'w':
775 			strictmsr = 0;
776 			break;
777 		case 'W':
778 			virtio_msix = 0;
779 			break;
780 		case 'x':
781 			x2apic_mode = 1;
782 			break;
783 		case 'Y':
784 			mptgen = 0;
785 			break;
786 		case 'h':
787 			usage(0);
788 		default:
789 			usage(1);
790 		}
791 	}
792 	argc -= optind;
793 	argv += optind;
794 
795 	if (argc != 1)
796 		usage(1);
797 
798 	vmname = argv[0];
799 
800 	ctx = vm_open(vmname);
801 	if (ctx == NULL) {
802 		perror("vm_open");
803 		exit(1);
804 	}
805 
806 	max_vcpus = num_vcpus_allowed(ctx);
807 	if (guest_ncpus > max_vcpus) {
808 		fprintf(stderr, "%d vCPUs requested but only %d available\n",
809 			guest_ncpus, max_vcpus);
810 		exit(1);
811 	}
812 
813 	fbsdrun_set_capabilities(ctx, BSP);
814 
815 	if (dump_guest_memory)
816 		vm_set_memflags(ctx, VM_MEM_F_INCORE);
817 	err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
818 	if (err) {
819 		fprintf(stderr, "Unable to setup memory (%d)\n", err);
820 		exit(1);
821 	}
822 
823 	error = init_msr();
824 	if (error) {
825 		fprintf(stderr, "init_msr error %d", error);
826 		exit(1);
827 	}
828 
829 	init_mem();
830 	init_inout();
831 	pci_irq_init(ctx);
832 	ioapic_init(ctx);
833 
834 	rtc_init(ctx, rtc_localtime);
835 	sci_init(ctx);
836 
837 	/*
838 	 * Exit if a device emulation finds an error in it's initilization
839 	 */
840 	if (init_pci(ctx) != 0)
841 		exit(1);
842 
843 	if (gdb_port != 0)
844 		init_dbgport(gdb_port);
845 
846 	if (bvmcons)
847 		init_bvmcons();
848 
849 	error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
850 	assert(error == 0);
851 
852 	/*
853 	 * build the guest tables, MP etc.
854 	 */
855 	if (mptgen) {
856 		error = mptable_build(ctx, guest_ncpus);
857 		if (error)
858 			exit(1);
859 	}
860 
861 	error = smbios_build(ctx);
862 	assert(error == 0);
863 
864 	if (acpi) {
865 		error = acpi_build(ctx, guest_ncpus);
866 		assert(error == 0);
867 	}
868 
869 	/*
870 	 * Change the proc title to include the VM name.
871 	 */
872 	setproctitle("%s", vmname);
873 
874 	/*
875 	 * Add CPU 0
876 	 */
877 	fbsdrun_addcpu(ctx, BSP, BSP, rip);
878 
879 	/*
880 	 * Head off to the main event dispatch loop
881 	 */
882 	mevent_dispatch();
883 
884 	exit(1);
885 }
886