xref: /freebsd/usr.sbin/bhyve/bhyverun.c (revision 193d9e768ba63fcfb187cfd17f461f7d41345048)
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 #ifndef WITHOUT_CAPSICUM
34 #include <sys/capsicum.h>
35 #endif
36 #include <sys/mman.h>
37 #include <sys/time.h>
38 
39 #include <machine/atomic.h>
40 #include <machine/segments.h>
41 
42 #ifndef WITHOUT_CAPSICUM
43 #include <capsicum_helpers.h>
44 #endif
45 #include <stdio.h>
46 #include <stdlib.h>
47 #include <string.h>
48 #include <err.h>
49 #include <errno.h>
50 #include <libgen.h>
51 #include <unistd.h>
52 #include <assert.h>
53 #include <errno.h>
54 #include <pthread.h>
55 #include <pthread_np.h>
56 #include <sysexits.h>
57 #include <stdbool.h>
58 
59 #include <machine/vmm.h>
60 #ifndef WITHOUT_CAPSICUM
61 #include <machine/vmm_dev.h>
62 #endif
63 #include <vmmapi.h>
64 
65 #include "bhyverun.h"
66 #include "acpi.h"
67 #include "atkbdc.h"
68 #include "inout.h"
69 #include "dbgport.h"
70 #include "fwctl.h"
71 #include "ioapic.h"
72 #include "mem.h"
73 #include "mevent.h"
74 #include "mptbl.h"
75 #include "pci_emul.h"
76 #include "pci_irq.h"
77 #include "pci_lpc.h"
78 #include "smbiostbl.h"
79 #include "xmsr.h"
80 #include "spinup_ap.h"
81 #include "rtc.h"
82 
83 #define GUEST_NIO_PORT		0x488	/* guest upcalls via i/o port */
84 
85 #define MB		(1024UL * 1024)
86 #define GB		(1024UL * MB)
87 
88 typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
89 extern int vmexit_task_switch(struct vmctx *, struct vm_exit *, int *vcpu);
90 
91 char *vmname;
92 
93 int guest_ncpus;
94 char *guest_uuid_str;
95 
96 static int guest_vmexit_on_hlt, guest_vmexit_on_pause;
97 static int virtio_msix = 1;
98 static int x2apic_mode = 0;	/* default is xAPIC */
99 
100 static int strictio;
101 static int strictmsr = 1;
102 
103 static int acpi;
104 
105 static char *progname;
106 static const int BSP = 0;
107 
108 static cpuset_t cpumask;
109 
110 static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
111 
112 static struct vm_exit vmexit[VM_MAXCPU];
113 
114 struct bhyvestats {
115         uint64_t        vmexit_bogus;
116 	uint64_t	vmexit_reqidle;
117         uint64_t        vmexit_hlt;
118         uint64_t        vmexit_pause;
119         uint64_t        vmexit_mtrap;
120         uint64_t        vmexit_inst_emul;
121         uint64_t        cpu_switch_rotate;
122         uint64_t        cpu_switch_direct;
123 } stats;
124 
125 struct mt_vmm_info {
126 	pthread_t	mt_thr;
127 	struct vmctx	*mt_ctx;
128 	int		mt_vcpu;
129 } mt_vmm_info[VM_MAXCPU];
130 
131 static cpuset_t *vcpumap[VM_MAXCPU] = { NULL };
132 
133 static void
134 usage(int code)
135 {
136 
137         fprintf(stderr,
138                 "Usage: %s [-abehuwxACHPSWY] [-c vcpus] [-g <gdb port>] [-l <lpc>]\n"
139 		"       %*s [-m mem] [-p vcpu:hostcpu] [-s <pci>] [-U uuid] <vm>\n"
140 		"       -a: local apic is in xAPIC mode (deprecated)\n"
141 		"       -A: create ACPI tables\n"
142 		"       -c: # cpus (default 1)\n"
143 		"       -C: include guest memory in core file\n"
144 		"       -e: exit on unhandled I/O access\n"
145 		"       -g: gdb port\n"
146 		"       -h: help\n"
147 		"       -H: vmexit from the guest on hlt\n"
148 		"       -l: LPC device configuration\n"
149 		"       -m: memory size in MB\n"
150 		"       -p: pin 'vcpu' to 'hostcpu'\n"
151 		"       -P: vmexit from the guest on pause\n"
152 		"       -s: <slot,driver,configinfo> PCI slot config\n"
153 		"       -S: guest memory cannot be swapped\n"
154 		"       -u: RTC keeps UTC time\n"
155 		"       -U: uuid\n"
156 		"       -w: ignore unimplemented MSRs\n"
157 		"       -W: force virtio to use single-vector MSI\n"
158 		"       -x: local apic is in x2APIC mode\n"
159 		"       -Y: disable MPtable generation\n",
160 		progname, (int)strlen(progname), "");
161 
162 	exit(code);
163 }
164 
165 static int
166 pincpu_parse(const char *opt)
167 {
168 	int vcpu, pcpu;
169 
170 	if (sscanf(opt, "%d:%d", &vcpu, &pcpu) != 2) {
171 		fprintf(stderr, "invalid format: %s\n", opt);
172 		return (-1);
173 	}
174 
175 	if (vcpu < 0 || vcpu >= VM_MAXCPU) {
176 		fprintf(stderr, "vcpu '%d' outside valid range from 0 to %d\n",
177 		    vcpu, VM_MAXCPU - 1);
178 		return (-1);
179 	}
180 
181 	if (pcpu < 0 || pcpu >= CPU_SETSIZE) {
182 		fprintf(stderr, "hostcpu '%d' outside valid range from "
183 		    "0 to %d\n", pcpu, CPU_SETSIZE - 1);
184 		return (-1);
185 	}
186 
187 	if (vcpumap[vcpu] == NULL) {
188 		if ((vcpumap[vcpu] = malloc(sizeof(cpuset_t))) == NULL) {
189 			perror("malloc");
190 			return (-1);
191 		}
192 		CPU_ZERO(vcpumap[vcpu]);
193 	}
194 	CPU_SET(pcpu, vcpumap[vcpu]);
195 	return (0);
196 }
197 
198 void
199 vm_inject_fault(void *arg, int vcpu, int vector, int errcode_valid,
200     int errcode)
201 {
202 	struct vmctx *ctx;
203 	int error, restart_instruction;
204 
205 	ctx = arg;
206 	restart_instruction = 1;
207 
208 	error = vm_inject_exception(ctx, vcpu, vector, errcode_valid, errcode,
209 	    restart_instruction);
210 	assert(error == 0);
211 }
212 
213 void *
214 paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
215 {
216 
217 	return (vm_map_gpa(ctx, gaddr, len));
218 }
219 
220 int
221 fbsdrun_vmexit_on_pause(void)
222 {
223 
224 	return (guest_vmexit_on_pause);
225 }
226 
227 int
228 fbsdrun_vmexit_on_hlt(void)
229 {
230 
231 	return (guest_vmexit_on_hlt);
232 }
233 
234 int
235 fbsdrun_virtio_msix(void)
236 {
237 
238 	return (virtio_msix);
239 }
240 
241 static void *
242 fbsdrun_start_thread(void *param)
243 {
244 	char tname[MAXCOMLEN + 1];
245 	struct mt_vmm_info *mtp;
246 	int vcpu;
247 
248 	mtp = param;
249 	vcpu = mtp->mt_vcpu;
250 
251 	snprintf(tname, sizeof(tname), "vcpu %d", vcpu);
252 	pthread_set_name_np(mtp->mt_thr, tname);
253 
254 	vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
255 
256 	/* not reached */
257 	exit(1);
258 	return (NULL);
259 }
260 
261 void
262 fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip)
263 {
264 	int error;
265 
266 	assert(fromcpu == BSP);
267 
268 	/*
269 	 * The 'newcpu' must be activated in the context of 'fromcpu'. If
270 	 * vm_activate_cpu() is delayed until newcpu's pthread starts running
271 	 * then vmm.ko is out-of-sync with bhyve and this can create a race
272 	 * with vm_suspend().
273 	 */
274 	error = vm_activate_cpu(ctx, newcpu);
275 	if (error != 0)
276 		err(EX_OSERR, "could not activate CPU %d", newcpu);
277 
278 	CPU_SET_ATOMIC(newcpu, &cpumask);
279 
280 	/*
281 	 * Set up the vmexit struct to allow execution to start
282 	 * at the given RIP
283 	 */
284 	vmexit[newcpu].rip = rip;
285 	vmexit[newcpu].inst_length = 0;
286 
287 	mt_vmm_info[newcpu].mt_ctx = ctx;
288 	mt_vmm_info[newcpu].mt_vcpu = newcpu;
289 
290 	error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL,
291 	    fbsdrun_start_thread, &mt_vmm_info[newcpu]);
292 	assert(error == 0);
293 }
294 
295 static int
296 fbsdrun_deletecpu(struct vmctx *ctx, int vcpu)
297 {
298 
299 	if (!CPU_ISSET(vcpu, &cpumask)) {
300 		fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu);
301 		exit(1);
302 	}
303 
304 	CPU_CLR_ATOMIC(vcpu, &cpumask);
305 	return (CPU_EMPTY(&cpumask));
306 }
307 
308 static int
309 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
310 		     uint32_t eax)
311 {
312 #if BHYVE_DEBUG
313 	/*
314 	 * put guest-driven debug here
315 	 */
316 #endif
317         return (VMEXIT_CONTINUE);
318 }
319 
320 static int
321 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
322 {
323 	int error;
324 	int bytes, port, in, out;
325 	int vcpu;
326 
327 	vcpu = *pvcpu;
328 
329 	port = vme->u.inout.port;
330 	bytes = vme->u.inout.bytes;
331 	in = vme->u.inout.in;
332 	out = !in;
333 
334         /* Extra-special case of host notifications */
335         if (out && port == GUEST_NIO_PORT) {
336                 error = vmexit_handle_notify(ctx, vme, pvcpu, vme->u.inout.eax);
337 		return (error);
338 	}
339 
340 	error = emulate_inout(ctx, vcpu, vme, strictio);
341 	if (error) {
342 		fprintf(stderr, "Unhandled %s%c 0x%04x at 0x%lx\n",
343 		    in ? "in" : "out",
344 		    bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'),
345 		    port, vmexit->rip);
346 		return (VMEXIT_ABORT);
347 	} else {
348 		return (VMEXIT_CONTINUE);
349 	}
350 }
351 
352 static int
353 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
354 {
355 	uint64_t val;
356 	uint32_t eax, edx;
357 	int error;
358 
359 	val = 0;
360 	error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val);
361 	if (error != 0) {
362 		fprintf(stderr, "rdmsr to register %#x on vcpu %d\n",
363 		    vme->u.msr.code, *pvcpu);
364 		if (strictmsr) {
365 			vm_inject_gp(ctx, *pvcpu);
366 			return (VMEXIT_CONTINUE);
367 		}
368 	}
369 
370 	eax = val;
371 	error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax);
372 	assert(error == 0);
373 
374 	edx = val >> 32;
375 	error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx);
376 	assert(error == 0);
377 
378 	return (VMEXIT_CONTINUE);
379 }
380 
381 static int
382 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
383 {
384 	int error;
385 
386 	error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval);
387 	if (error != 0) {
388 		fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n",
389 		    vme->u.msr.code, vme->u.msr.wval, *pvcpu);
390 		if (strictmsr) {
391 			vm_inject_gp(ctx, *pvcpu);
392 			return (VMEXIT_CONTINUE);
393 		}
394 	}
395 	return (VMEXIT_CONTINUE);
396 }
397 
398 static int
399 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
400 {
401 
402 	(void)spinup_ap(ctx, *pvcpu,
403 		    vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
404 
405 	return (VMEXIT_CONTINUE);
406 }
407 
408 #define	DEBUG_EPT_MISCONFIG
409 #ifdef DEBUG_EPT_MISCONFIG
410 #define	EXIT_REASON_EPT_MISCONFIG	49
411 #define	VMCS_GUEST_PHYSICAL_ADDRESS	0x00002400
412 #define	VMCS_IDENT(x)			((x) | 0x80000000)
413 
414 static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4];
415 static int ept_misconfig_ptenum;
416 #endif
417 
418 static int
419 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
420 {
421 
422 	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
423 	fprintf(stderr, "\treason\t\tVMX\n");
424 	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
425 	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
426 	fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status);
427 	fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
428 	fprintf(stderr, "\tqualification\t0x%016lx\n",
429 	    vmexit->u.vmx.exit_qualification);
430 	fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type);
431 	fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error);
432 #ifdef DEBUG_EPT_MISCONFIG
433 	if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) {
434 		vm_get_register(ctx, *pvcpu,
435 		    VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS),
436 		    &ept_misconfig_gpa);
437 		vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte,
438 		    &ept_misconfig_ptenum);
439 		fprintf(stderr, "\tEPT misconfiguration:\n");
440 		fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa);
441 		fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n",
442 		    ept_misconfig_ptenum, ept_misconfig_pte[0],
443 		    ept_misconfig_pte[1], ept_misconfig_pte[2],
444 		    ept_misconfig_pte[3]);
445 	}
446 #endif	/* DEBUG_EPT_MISCONFIG */
447 	return (VMEXIT_ABORT);
448 }
449 
450 static int
451 vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
452 {
453 
454 	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
455 	fprintf(stderr, "\treason\t\tSVM\n");
456 	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
457 	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
458 	fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode);
459 	fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1);
460 	fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2);
461 	return (VMEXIT_ABORT);
462 }
463 
464 static int
465 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
466 {
467 
468 	assert(vmexit->inst_length == 0);
469 
470 	stats.vmexit_bogus++;
471 
472 	return (VMEXIT_CONTINUE);
473 }
474 
475 static int
476 vmexit_reqidle(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
477 {
478 
479 	assert(vmexit->inst_length == 0);
480 
481 	stats.vmexit_reqidle++;
482 
483 	return (VMEXIT_CONTINUE);
484 }
485 
486 static int
487 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
488 {
489 
490 	stats.vmexit_hlt++;
491 
492 	/*
493 	 * Just continue execution with the next instruction. We use
494 	 * the HLT VM exit as a way to be friendly with the host
495 	 * scheduler.
496 	 */
497 	return (VMEXIT_CONTINUE);
498 }
499 
500 static int
501 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
502 {
503 
504 	stats.vmexit_pause++;
505 
506 	return (VMEXIT_CONTINUE);
507 }
508 
509 static int
510 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
511 {
512 
513 	assert(vmexit->inst_length == 0);
514 
515 	stats.vmexit_mtrap++;
516 
517 	return (VMEXIT_CONTINUE);
518 }
519 
520 static int
521 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
522 {
523 	int err, i;
524 	struct vie *vie;
525 
526 	stats.vmexit_inst_emul++;
527 
528 	vie = &vmexit->u.inst_emul.vie;
529 	err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
530 	    vie, &vmexit->u.inst_emul.paging);
531 
532 	if (err) {
533 		if (err == ESRCH) {
534 			fprintf(stderr, "Unhandled memory access to 0x%lx\n",
535 			    vmexit->u.inst_emul.gpa);
536 		}
537 
538 		fprintf(stderr, "Failed to emulate instruction [");
539 		for (i = 0; i < vie->num_valid; i++) {
540 			fprintf(stderr, "0x%02x%s", vie->inst[i],
541 			    i != (vie->num_valid - 1) ? " " : "");
542 		}
543 		fprintf(stderr, "] at 0x%lx\n", vmexit->rip);
544 		return (VMEXIT_ABORT);
545 	}
546 
547 	return (VMEXIT_CONTINUE);
548 }
549 
550 static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER;
551 static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER;
552 
553 static int
554 vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
555 {
556 	enum vm_suspend_how how;
557 
558 	how = vmexit->u.suspended.how;
559 
560 	fbsdrun_deletecpu(ctx, *pvcpu);
561 
562 	if (*pvcpu != BSP) {
563 		pthread_mutex_lock(&resetcpu_mtx);
564 		pthread_cond_signal(&resetcpu_cond);
565 		pthread_mutex_unlock(&resetcpu_mtx);
566 		pthread_exit(NULL);
567 	}
568 
569 	pthread_mutex_lock(&resetcpu_mtx);
570 	while (!CPU_EMPTY(&cpumask)) {
571 		pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx);
572 	}
573 	pthread_mutex_unlock(&resetcpu_mtx);
574 
575 	switch (how) {
576 	case VM_SUSPEND_RESET:
577 		exit(0);
578 	case VM_SUSPEND_POWEROFF:
579 		exit(1);
580 	case VM_SUSPEND_HALT:
581 		exit(2);
582 	case VM_SUSPEND_TRIPLEFAULT:
583 		exit(3);
584 	default:
585 		fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how);
586 		exit(100);
587 	}
588 	return (0);	/* NOTREACHED */
589 }
590 
591 static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
592 	[VM_EXITCODE_INOUT]  = vmexit_inout,
593 	[VM_EXITCODE_INOUT_STR]  = vmexit_inout,
594 	[VM_EXITCODE_VMX]    = vmexit_vmx,
595 	[VM_EXITCODE_SVM]    = vmexit_svm,
596 	[VM_EXITCODE_BOGUS]  = vmexit_bogus,
597 	[VM_EXITCODE_REQIDLE] = vmexit_reqidle,
598 	[VM_EXITCODE_RDMSR]  = vmexit_rdmsr,
599 	[VM_EXITCODE_WRMSR]  = vmexit_wrmsr,
600 	[VM_EXITCODE_MTRAP]  = vmexit_mtrap,
601 	[VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
602 	[VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
603 	[VM_EXITCODE_SUSPENDED] = vmexit_suspend,
604 	[VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch,
605 };
606 
607 static void
608 vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip)
609 {
610 	int error, rc;
611 	enum vm_exitcode exitcode;
612 	cpuset_t active_cpus;
613 
614 	if (vcpumap[vcpu] != NULL) {
615 		error = pthread_setaffinity_np(pthread_self(),
616 		    sizeof(cpuset_t), vcpumap[vcpu]);
617 		assert(error == 0);
618 	}
619 
620 	error = vm_active_cpus(ctx, &active_cpus);
621 	assert(CPU_ISSET(vcpu, &active_cpus));
622 
623 	error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip);
624 	assert(error == 0);
625 
626 	while (1) {
627 		error = vm_run(ctx, vcpu, &vmexit[vcpu]);
628 		if (error != 0)
629 			break;
630 
631 		exitcode = vmexit[vcpu].exitcode;
632 		if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
633 			fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
634 			    exitcode);
635 			exit(1);
636 		}
637 
638 		rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
639 
640 		switch (rc) {
641 		case VMEXIT_CONTINUE:
642 			break;
643 		case VMEXIT_ABORT:
644 			abort();
645 		default:
646 			exit(1);
647 		}
648 	}
649 	fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
650 }
651 
652 static int
653 num_vcpus_allowed(struct vmctx *ctx)
654 {
655 	int tmp, error;
656 
657 	error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
658 
659 	/*
660 	 * The guest is allowed to spinup more than one processor only if the
661 	 * UNRESTRICTED_GUEST capability is available.
662 	 */
663 	if (error == 0)
664 		return (VM_MAXCPU);
665 	else
666 		return (1);
667 }
668 
669 void
670 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
671 {
672 	int err, tmp;
673 
674 	if (fbsdrun_vmexit_on_hlt()) {
675 		err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
676 		if (err < 0) {
677 			fprintf(stderr, "VM exit on HLT not supported\n");
678 			exit(1);
679 		}
680 		vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
681 		if (cpu == BSP)
682 			handler[VM_EXITCODE_HLT] = vmexit_hlt;
683 	}
684 
685         if (fbsdrun_vmexit_on_pause()) {
686 		/*
687 		 * pause exit support required for this mode
688 		 */
689 		err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
690 		if (err < 0) {
691 			fprintf(stderr,
692 			    "SMP mux requested, no pause support\n");
693 			exit(1);
694 		}
695 		vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
696 		if (cpu == BSP)
697 			handler[VM_EXITCODE_PAUSE] = vmexit_pause;
698         }
699 
700 	if (x2apic_mode)
701 		err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
702 	else
703 		err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
704 
705 	if (err) {
706 		fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
707 		exit(1);
708 	}
709 
710 	vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
711 }
712 
713 static struct vmctx *
714 do_open(const char *vmname)
715 {
716 	struct vmctx *ctx;
717 	int error;
718 	bool reinit, romboot;
719 #ifndef WITHOUT_CAPSICUM
720 	cap_rights_t rights;
721 	const cap_ioctl_t *cmds;
722 	size_t ncmds;
723 #endif
724 
725 	reinit = romboot = false;
726 
727 	if (lpc_bootrom())
728 		romboot = true;
729 
730 	error = vm_create(vmname);
731 	if (error) {
732 		if (errno == EEXIST) {
733 			if (romboot) {
734 				reinit = true;
735 			} else {
736 				/*
737 				 * The virtual machine has been setup by the
738 				 * userspace bootloader.
739 				 */
740 			}
741 		} else {
742 			perror("vm_create");
743 			exit(1);
744 		}
745 	} else {
746 		if (!romboot) {
747 			/*
748 			 * If the virtual machine was just created then a
749 			 * bootrom must be configured to boot it.
750 			 */
751 			fprintf(stderr, "virtual machine cannot be booted\n");
752 			exit(1);
753 		}
754 	}
755 
756 	ctx = vm_open(vmname);
757 	if (ctx == NULL) {
758 		perror("vm_open");
759 		exit(1);
760 	}
761 
762 #ifndef WITHOUT_CAPSICUM
763 	cap_rights_init(&rights, CAP_IOCTL, CAP_MMAP_RW);
764 	if (cap_rights_limit(vm_get_device_fd(ctx), &rights) == -1 &&
765 	    errno != ENOSYS)
766 		errx(EX_OSERR, "Unable to apply rights for sandbox");
767 	vm_get_ioctls(&ncmds);
768 	cmds = vm_get_ioctls(NULL);
769 	if (cmds == NULL)
770 		errx(EX_OSERR, "out of memory");
771 	if (cap_ioctls_limit(vm_get_device_fd(ctx), cmds, ncmds) == -1 &&
772 	    errno != ENOSYS)
773 		errx(EX_OSERR, "Unable to apply rights for sandbox");
774 	free((cap_ioctl_t *)cmds);
775 #endif
776 
777 	if (reinit) {
778 		error = vm_reinit(ctx);
779 		if (error) {
780 			perror("vm_reinit");
781 			exit(1);
782 		}
783 	}
784 	return (ctx);
785 }
786 
787 int
788 main(int argc, char *argv[])
789 {
790 	int c, error, gdb_port, err, bvmcons;
791 	int max_vcpus, mptgen, memflags;
792 	int rtc_localtime;
793 	struct vmctx *ctx;
794 	uint64_t rip;
795 	size_t memsize;
796 	char *optstr;
797 
798 	bvmcons = 0;
799 	progname = basename(argv[0]);
800 	gdb_port = 0;
801 	guest_ncpus = 1;
802 	memsize = 256 * MB;
803 	mptgen = 1;
804 	rtc_localtime = 1;
805 	memflags = 0;
806 
807 	optstr = "abehuwxACHIPSWYp:g:c:s:m:l:U:";
808 	while ((c = getopt(argc, argv, optstr)) != -1) {
809 		switch (c) {
810 		case 'a':
811 			x2apic_mode = 0;
812 			break;
813 		case 'A':
814 			acpi = 1;
815 			break;
816 		case 'b':
817 			bvmcons = 1;
818 			break;
819 		case 'p':
820                         if (pincpu_parse(optarg) != 0) {
821                             errx(EX_USAGE, "invalid vcpu pinning "
822                                  "configuration '%s'", optarg);
823                         }
824 			break;
825                 case 'c':
826 			guest_ncpus = atoi(optarg);
827 			break;
828 		case 'C':
829 			memflags |= VM_MEM_F_INCORE;
830 			break;
831 		case 'g':
832 			gdb_port = atoi(optarg);
833 			break;
834 		case 'l':
835 			if (lpc_device_parse(optarg) != 0) {
836 				errx(EX_USAGE, "invalid lpc device "
837 				    "configuration '%s'", optarg);
838 			}
839 			break;
840 		case 's':
841 			if (pci_parse_slot(optarg) != 0)
842 				exit(1);
843 			else
844 				break;
845 		case 'S':
846 			memflags |= VM_MEM_F_WIRED;
847 			break;
848                 case 'm':
849 			error = vm_parse_memsize(optarg, &memsize);
850 			if (error)
851 				errx(EX_USAGE, "invalid memsize '%s'", optarg);
852 			break;
853 		case 'H':
854 			guest_vmexit_on_hlt = 1;
855 			break;
856 		case 'I':
857 			/*
858 			 * The "-I" option was used to add an ioapic to the
859 			 * virtual machine.
860 			 *
861 			 * An ioapic is now provided unconditionally for each
862 			 * virtual machine and this option is now deprecated.
863 			 */
864 			break;
865 		case 'P':
866 			guest_vmexit_on_pause = 1;
867 			break;
868 		case 'e':
869 			strictio = 1;
870 			break;
871 		case 'u':
872 			rtc_localtime = 0;
873 			break;
874 		case 'U':
875 			guest_uuid_str = optarg;
876 			break;
877 		case 'w':
878 			strictmsr = 0;
879 			break;
880 		case 'W':
881 			virtio_msix = 0;
882 			break;
883 		case 'x':
884 			x2apic_mode = 1;
885 			break;
886 		case 'Y':
887 			mptgen = 0;
888 			break;
889 		case 'h':
890 			usage(0);
891 		default:
892 			usage(1);
893 		}
894 	}
895 	argc -= optind;
896 	argv += optind;
897 
898 	if (argc != 1)
899 		usage(1);
900 
901 	vmname = argv[0];
902 	ctx = do_open(vmname);
903 
904 	if (guest_ncpus < 1) {
905 		fprintf(stderr, "Invalid guest vCPUs (%d)\n", guest_ncpus);
906 		exit(1);
907 	}
908 
909 	max_vcpus = num_vcpus_allowed(ctx);
910 	if (guest_ncpus > max_vcpus) {
911 		fprintf(stderr, "%d vCPUs requested but only %d available\n",
912 			guest_ncpus, max_vcpus);
913 		exit(1);
914 	}
915 
916 	fbsdrun_set_capabilities(ctx, BSP);
917 
918 	vm_set_memflags(ctx, memflags);
919 	err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
920 	if (err) {
921 		fprintf(stderr, "Unable to setup memory (%d)\n", errno);
922 		exit(1);
923 	}
924 
925 	error = init_msr();
926 	if (error) {
927 		fprintf(stderr, "init_msr error %d", error);
928 		exit(1);
929 	}
930 
931 	init_mem();
932 	init_inout();
933 	atkbdc_init(ctx);
934 	pci_irq_init(ctx);
935 	ioapic_init(ctx);
936 
937 	rtc_init(ctx, rtc_localtime);
938 	sci_init(ctx);
939 
940 	/*
941 	 * Exit if a device emulation finds an error in its initilization
942 	 */
943 	if (init_pci(ctx) != 0)
944 		exit(1);
945 
946 	if (gdb_port != 0)
947 		init_dbgport(gdb_port);
948 
949 	if (bvmcons)
950 		init_bvmcons();
951 
952 	if (lpc_bootrom()) {
953 		if (vm_set_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, 1)) {
954 			fprintf(stderr, "ROM boot failed: unrestricted guest "
955 			    "capability not available\n");
956 			exit(1);
957 		}
958 		error = vcpu_reset(ctx, BSP);
959 		assert(error == 0);
960 	}
961 
962 	error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
963 	assert(error == 0);
964 
965 	/*
966 	 * build the guest tables, MP etc.
967 	 */
968 	if (mptgen) {
969 		error = mptable_build(ctx, guest_ncpus);
970 		if (error)
971 			exit(1);
972 	}
973 
974 	error = smbios_build(ctx);
975 	assert(error == 0);
976 
977 	if (acpi) {
978 		error = acpi_build(ctx, guest_ncpus);
979 		assert(error == 0);
980 	}
981 
982 	if (lpc_bootrom())
983 		fwctl_init();
984 
985 #ifndef WITHOUT_CAPSICUM
986 	caph_cache_catpages();
987 
988 	if (caph_limit_stdout() == -1 || caph_limit_stderr() == -1)
989 		errx(EX_OSERR, "Unable to apply rights for sandbox");
990 
991 	if (cap_enter() == -1 && errno != ENOSYS)
992 		errx(EX_OSERR, "cap_enter() failed");
993 #endif
994 
995 	/*
996 	 * Change the proc title to include the VM name.
997 	 */
998 	setproctitle("%s", vmname);
999 
1000 	/*
1001 	 * Add CPU 0
1002 	 */
1003 	fbsdrun_addcpu(ctx, BSP, BSP, rip);
1004 
1005 	/*
1006 	 * Head off to the main event dispatch loop
1007 	 */
1008 	mevent_dispatch();
1009 
1010 	exit(1);
1011 }
1012