bhyverun.c revision 293412 
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: stable/10/usr.sbin/bhyve/bhyverun.c 293412 2016-01-08 03:02:19Z araujo $
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD: stable/10/usr.sbin/bhyve/bhyverun.c 293412 2016-01-08 03:02:19Z araujo $");
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
75typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
76extern int vmexit_task_switch(struct vmctx *, struct vm_exit *, int *vcpu);
77
78char *vmname;
79
80int guest_ncpus;
81char *guest_uuid_str;
82
83static int guest_vmexit_on_hlt, guest_vmexit_on_pause;
84static int virtio_msix = 1;
85static int x2apic_mode = 0;	/* default is xAPIC */
86
87static int strictio;
88static int strictmsr = 1;
89
90static int acpi;
91
92static char *progname;
93static const int BSP = 0;
94
95static cpuset_t cpumask;
96
97static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
98
99static struct vm_exit vmexit[VM_MAXCPU];
100
101struct bhyvestats {
102        uint64_t        vmexit_bogus;
103	uint64_t	vmexit_reqidle;
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
112struct mt_vmm_info {
113	pthread_t	mt_thr;
114	struct vmctx	*mt_ctx;
115	int		mt_vcpu;
116} mt_vmm_info[VM_MAXCPU];
117
118static cpuset_t *vcpumap[VM_MAXCPU] = { NULL };
119
120static void
121usage(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
151static int
152pincpu_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
184void
185vm_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
199void *
200paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
201{
202
203	return (vm_map_gpa(ctx, gaddr, len));
204}
205
206int
207fbsdrun_vmexit_on_pause(void)
208{
209
210	return (guest_vmexit_on_pause);
211}
212
213int
214fbsdrun_vmexit_on_hlt(void)
215{
216
217	return (guest_vmexit_on_hlt);
218}
219
220int
221fbsdrun_virtio_msix(void)
222{
223
224	return (virtio_msix);
225}
226
227static void *
228fbsdrun_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
247void
248fbsdrun_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	if (error != 0)
262		err(EX_OSERR, "could not activate CPU %d", newcpu);
263
264	CPU_SET_ATOMIC(newcpu, &cpumask);
265
266	/*
267	 * Set up the vmexit struct to allow execution to start
268	 * at the given RIP
269	 */
270	vmexit[newcpu].rip = rip;
271	vmexit[newcpu].inst_length = 0;
272
273	mt_vmm_info[newcpu].mt_ctx = ctx;
274	mt_vmm_info[newcpu].mt_vcpu = newcpu;
275
276	error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL,
277	    fbsdrun_start_thread, &mt_vmm_info[newcpu]);
278	assert(error == 0);
279}
280
281static int
282fbsdrun_deletecpu(struct vmctx *ctx, int vcpu)
283{
284
285	if (!CPU_ISSET(vcpu, &cpumask)) {
286		fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu);
287		exit(1);
288	}
289
290	CPU_CLR_ATOMIC(vcpu, &cpumask);
291	return (CPU_EMPTY(&cpumask));
292}
293
294static int
295vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
296		     uint32_t eax)
297{
298#if BHYVE_DEBUG
299	/*
300	 * put guest-driven debug here
301	 */
302#endif
303        return (VMEXIT_CONTINUE);
304}
305
306static int
307vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
308{
309	int error;
310	int bytes, port, in, out;
311	int vcpu;
312
313	vcpu = *pvcpu;
314
315	port = vme->u.inout.port;
316	bytes = vme->u.inout.bytes;
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 at 0x%lx\n",
329		    in ? "in" : "out",
330		    bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'),
331		    port, vmexit->rip);
332		return (VMEXIT_ABORT);
333	} else {
334		return (VMEXIT_CONTINUE);
335	}
336}
337
338static int
339vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
340{
341	uint64_t val;
342	uint32_t eax, edx;
343	int error;
344
345	val = 0;
346	error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val);
347	if (error != 0) {
348		fprintf(stderr, "rdmsr to register %#x on vcpu %d\n",
349		    vme->u.msr.code, *pvcpu);
350		if (strictmsr) {
351			vm_inject_gp(ctx, *pvcpu);
352			return (VMEXIT_CONTINUE);
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
367static int
368vmexit_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			vm_inject_gp(ctx, *pvcpu);
378			return (VMEXIT_CONTINUE);
379		}
380	}
381	return (VMEXIT_CONTINUE);
382}
383
384static int
385vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
386{
387	int newcpu;
388	int retval = VMEXIT_CONTINUE;
389
390	newcpu = spinup_ap(ctx, *pvcpu,
391			   vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
392
393	return (retval);
394}
395
396#define	DEBUG_EPT_MISCONFIG
397#ifdef DEBUG_EPT_MISCONFIG
398#define	EXIT_REASON_EPT_MISCONFIG	49
399#define	VMCS_GUEST_PHYSICAL_ADDRESS	0x00002400
400#define	VMCS_IDENT(x)			((x) | 0x80000000)
401
402static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4];
403static int ept_misconfig_ptenum;
404#endif
405
406static int
407vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
408{
409
410	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
411	fprintf(stderr, "\treason\t\tVMX\n");
412	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
413	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
414	fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status);
415	fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
416	fprintf(stderr, "\tqualification\t0x%016lx\n",
417	    vmexit->u.vmx.exit_qualification);
418	fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type);
419	fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error);
420#ifdef DEBUG_EPT_MISCONFIG
421	if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) {
422		vm_get_register(ctx, *pvcpu,
423		    VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS),
424		    &ept_misconfig_gpa);
425		vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte,
426		    &ept_misconfig_ptenum);
427		fprintf(stderr, "\tEPT misconfiguration:\n");
428		fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa);
429		fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n",
430		    ept_misconfig_ptenum, ept_misconfig_pte[0],
431		    ept_misconfig_pte[1], ept_misconfig_pte[2],
432		    ept_misconfig_pte[3]);
433	}
434#endif	/* DEBUG_EPT_MISCONFIG */
435	return (VMEXIT_ABORT);
436}
437
438static int
439vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
440{
441
442	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
443	fprintf(stderr, "\treason\t\tSVM\n");
444	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
445	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
446	fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode);
447	fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1);
448	fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2);
449	return (VMEXIT_ABORT);
450}
451
452static int
453vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
454{
455
456	assert(vmexit->inst_length == 0);
457
458	stats.vmexit_bogus++;
459
460	return (VMEXIT_CONTINUE);
461}
462
463static int
464vmexit_reqidle(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
465{
466
467	assert(vmexit->inst_length == 0);
468
469	stats.vmexit_reqidle++;
470
471	return (VMEXIT_CONTINUE);
472}
473
474static int
475vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
476{
477
478	stats.vmexit_hlt++;
479
480	/*
481	 * Just continue execution with the next instruction. We use
482	 * the HLT VM exit as a way to be friendly with the host
483	 * scheduler.
484	 */
485	return (VMEXIT_CONTINUE);
486}
487
488static int
489vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
490{
491
492	stats.vmexit_pause++;
493
494	return (VMEXIT_CONTINUE);
495}
496
497static int
498vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
499{
500
501	assert(vmexit->inst_length == 0);
502
503	stats.vmexit_mtrap++;
504
505	return (VMEXIT_CONTINUE);
506}
507
508static int
509vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
510{
511	int err, i;
512	struct vie *vie;
513
514	stats.vmexit_inst_emul++;
515
516	vie = &vmexit->u.inst_emul.vie;
517	err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
518	    vie, &vmexit->u.inst_emul.paging);
519
520	if (err) {
521		if (err == ESRCH) {
522			fprintf(stderr, "Unhandled memory access to 0x%lx\n",
523			    vmexit->u.inst_emul.gpa);
524		}
525
526		fprintf(stderr, "Failed to emulate instruction [");
527		for (i = 0; i < vie->num_valid; i++) {
528			fprintf(stderr, "0x%02x%s", vie->inst[i],
529			    i != (vie->num_valid - 1) ? " " : "");
530		}
531		fprintf(stderr, "] at 0x%lx\n", vmexit->rip);
532		return (VMEXIT_ABORT);
533	}
534
535	return (VMEXIT_CONTINUE);
536}
537
538static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER;
539static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER;
540
541static int
542vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
543{
544	enum vm_suspend_how how;
545
546	how = vmexit->u.suspended.how;
547
548	fbsdrun_deletecpu(ctx, *pvcpu);
549
550	if (*pvcpu != BSP) {
551		pthread_mutex_lock(&resetcpu_mtx);
552		pthread_cond_signal(&resetcpu_cond);
553		pthread_mutex_unlock(&resetcpu_mtx);
554		pthread_exit(NULL);
555	}
556
557	pthread_mutex_lock(&resetcpu_mtx);
558	while (!CPU_EMPTY(&cpumask)) {
559		pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx);
560	}
561	pthread_mutex_unlock(&resetcpu_mtx);
562
563	switch (how) {
564	case VM_SUSPEND_RESET:
565		exit(0);
566	case VM_SUSPEND_POWEROFF:
567		exit(1);
568	case VM_SUSPEND_HALT:
569		exit(2);
570	case VM_SUSPEND_TRIPLEFAULT:
571		exit(3);
572	default:
573		fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how);
574		exit(100);
575	}
576	return (0);	/* NOTREACHED */
577}
578
579static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
580	[VM_EXITCODE_INOUT]  = vmexit_inout,
581	[VM_EXITCODE_INOUT_STR]  = vmexit_inout,
582	[VM_EXITCODE_VMX]    = vmexit_vmx,
583	[VM_EXITCODE_SVM]    = vmexit_svm,
584	[VM_EXITCODE_BOGUS]  = vmexit_bogus,
585	[VM_EXITCODE_REQIDLE] = vmexit_reqidle,
586	[VM_EXITCODE_RDMSR]  = vmexit_rdmsr,
587	[VM_EXITCODE_WRMSR]  = vmexit_wrmsr,
588	[VM_EXITCODE_MTRAP]  = vmexit_mtrap,
589	[VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
590	[VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
591	[VM_EXITCODE_SUSPENDED] = vmexit_suspend,
592	[VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch,
593};
594
595static void
596vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip)
597{
598	int error, rc;
599	enum vm_exitcode exitcode;
600	cpuset_t active_cpus;
601
602	if (vcpumap[vcpu] != NULL) {
603		error = pthread_setaffinity_np(pthread_self(),
604		    sizeof(cpuset_t), vcpumap[vcpu]);
605		assert(error == 0);
606	}
607
608	error = vm_active_cpus(ctx, &active_cpus);
609	assert(CPU_ISSET(vcpu, &active_cpus));
610
611	error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip);
612	assert(error == 0);
613
614	while (1) {
615		error = vm_run(ctx, vcpu, &vmexit[vcpu]);
616		if (error != 0)
617			break;
618
619		exitcode = vmexit[vcpu].exitcode;
620		if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
621			fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
622			    exitcode);
623			exit(1);
624		}
625
626                rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
627
628		switch (rc) {
629		case VMEXIT_CONTINUE:
630			break;
631		case VMEXIT_ABORT:
632			abort();
633		default:
634			exit(1);
635		}
636	}
637	fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
638}
639
640static int
641num_vcpus_allowed(struct vmctx *ctx)
642{
643	int tmp, error;
644
645	error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
646
647	/*
648	 * The guest is allowed to spinup more than one processor only if the
649	 * UNRESTRICTED_GUEST capability is available.
650	 */
651	if (error == 0)
652		return (VM_MAXCPU);
653	else
654		return (1);
655}
656
657void
658fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
659{
660	int err, tmp;
661
662	if (fbsdrun_vmexit_on_hlt()) {
663		err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
664		if (err < 0) {
665			fprintf(stderr, "VM exit on HLT not supported\n");
666			exit(1);
667		}
668		vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
669		if (cpu == BSP)
670			handler[VM_EXITCODE_HLT] = vmexit_hlt;
671	}
672
673        if (fbsdrun_vmexit_on_pause()) {
674		/*
675		 * pause exit support required for this mode
676		 */
677		err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
678		if (err < 0) {
679			fprintf(stderr,
680			    "SMP mux requested, no pause support\n");
681			exit(1);
682		}
683		vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
684		if (cpu == BSP)
685			handler[VM_EXITCODE_PAUSE] = vmexit_pause;
686        }
687
688	if (x2apic_mode)
689		err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
690	else
691		err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
692
693	if (err) {
694		fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
695		exit(1);
696	}
697
698	vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
699}
700
701int
702main(int argc, char *argv[])
703{
704	int c, error, gdb_port, err, bvmcons;
705	int dump_guest_memory, max_vcpus, mptgen;
706	int rtc_localtime;
707	struct vmctx *ctx;
708	uint64_t rip;
709	size_t memsize;
710
711	bvmcons = 0;
712	dump_guest_memory = 0;
713	progname = basename(argv[0]);
714	gdb_port = 0;
715	guest_ncpus = 1;
716	memsize = 256 * MB;
717	mptgen = 1;
718	rtc_localtime = 1;
719
720	while ((c = getopt(argc, argv, "abehuwxACHIPWYp:g:c:s:m:l:U:")) != -1) {
721		switch (c) {
722		case 'a':
723			x2apic_mode = 0;
724			break;
725		case 'A':
726			acpi = 1;
727			break;
728		case 'b':
729			bvmcons = 1;
730			break;
731		case 'p':
732                        if (pincpu_parse(optarg) != 0) {
733                            errx(EX_USAGE, "invalid vcpu pinning "
734                                 "configuration '%s'", optarg);
735                        }
736			break;
737                case 'c':
738			guest_ncpus = atoi(optarg);
739			break;
740		case 'C':
741			dump_guest_memory = 1;
742			break;
743		case 'g':
744			gdb_port = atoi(optarg);
745			break;
746		case 'l':
747			if (lpc_device_parse(optarg) != 0) {
748				errx(EX_USAGE, "invalid lpc device "
749				    "configuration '%s'", optarg);
750			}
751			break;
752		case 's':
753			if (pci_parse_slot(optarg) != 0)
754				exit(1);
755			else
756				break;
757                case 'm':
758			error = vm_parse_memsize(optarg, &memsize);
759			if (error)
760				errx(EX_USAGE, "invalid memsize '%s'", optarg);
761			break;
762		case 'H':
763			guest_vmexit_on_hlt = 1;
764			break;
765		case 'I':
766			/*
767			 * The "-I" option was used to add an ioapic to the
768			 * virtual machine.
769			 *
770			 * An ioapic is now provided unconditionally for each
771			 * virtual machine and this option is now deprecated.
772			 */
773			break;
774		case 'P':
775			guest_vmexit_on_pause = 1;
776			break;
777		case 'e':
778			strictio = 1;
779			break;
780		case 'u':
781			rtc_localtime = 0;
782			break;
783		case 'U':
784			guest_uuid_str = optarg;
785			break;
786		case 'w':
787			strictmsr = 0;
788			break;
789		case 'W':
790			virtio_msix = 0;
791			break;
792		case 'x':
793			x2apic_mode = 1;
794			break;
795		case 'Y':
796			mptgen = 0;
797			break;
798		case 'h':
799			usage(0);
800		default:
801			usage(1);
802		}
803	}
804	argc -= optind;
805	argv += optind;
806
807	if (argc != 1)
808		usage(1);
809
810	vmname = argv[0];
811
812	ctx = vm_open(vmname);
813	if (ctx == NULL) {
814		perror("vm_open");
815		exit(1);
816	}
817
818	if (guest_ncpus < 1) {
819		fprintf(stderr, "Invalid guest vCPUs (%d)\n", guest_ncpus);
820		exit(1);
821	}
822
823	max_vcpus = num_vcpus_allowed(ctx);
824	if (guest_ncpus > max_vcpus) {
825		fprintf(stderr, "%d vCPUs requested but only %d available\n",
826			guest_ncpus, max_vcpus);
827		exit(1);
828	}
829
830	fbsdrun_set_capabilities(ctx, BSP);
831
832	if (dump_guest_memory)
833		vm_set_memflags(ctx, VM_MEM_F_INCORE);
834	err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
835	if (err) {
836		fprintf(stderr, "Unable to setup memory (%d)\n", err);
837		exit(1);
838	}
839
840	error = init_msr();
841	if (error) {
842		fprintf(stderr, "init_msr error %d", error);
843		exit(1);
844	}
845
846	init_mem();
847	init_inout();
848	pci_irq_init(ctx);
849	ioapic_init(ctx);
850
851	rtc_init(ctx, rtc_localtime);
852	sci_init(ctx);
853
854	/*
855	 * Exit if a device emulation finds an error in it's initilization
856	 */
857	if (init_pci(ctx) != 0)
858		exit(1);
859
860	if (gdb_port != 0)
861		init_dbgport(gdb_port);
862
863	if (bvmcons)
864		init_bvmcons();
865
866	error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
867	assert(error == 0);
868
869	/*
870	 * build the guest tables, MP etc.
871	 */
872	if (mptgen) {
873		error = mptable_build(ctx, guest_ncpus);
874		if (error)
875			exit(1);
876	}
877
878	error = smbios_build(ctx);
879	assert(error == 0);
880
881	if (acpi) {
882		error = acpi_build(ctx, guest_ncpus);
883		assert(error == 0);
884	}
885
886	/*
887	 * Change the proc title to include the VM name.
888	 */
889	setproctitle("%s", vmname);
890
891	/*
892	 * Add CPU 0
893	 */
894	fbsdrun_addcpu(ctx, BSP, BSP, rip);
895
896	/*
897	 * Head off to the main event dispatch loop
898	 */
899	mevent_dispatch();
900
901	exit(1);
902}
903