bhyverun.c revision 290386 
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 290386 2015-11-05 08:27:45Z ngie $
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD: stable/10/usr.sbin/bhyve/bhyverun.c 290386 2015-11-05 08:27:45Z ngie $");
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, string;
311	int vcpu;
312
313	vcpu = *pvcpu;
314
315	port = vme->u.inout.port;
316	bytes = vme->u.inout.bytes;
317	string = vme->u.inout.string;
318	in = vme->u.inout.in;
319	out = !in;
320
321        /* Extra-special case of host notifications */
322        if (out && port == GUEST_NIO_PORT) {
323                error = vmexit_handle_notify(ctx, vme, pvcpu, vme->u.inout.eax);
324		return (error);
325	}
326
327	error = emulate_inout(ctx, vcpu, vme, strictio);
328	if (error) {
329		fprintf(stderr, "Unhandled %s%c 0x%04x at 0x%lx\n",
330		    in ? "in" : "out",
331		    bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'),
332		    port, vmexit->rip);
333		return (VMEXIT_ABORT);
334	} else {
335		return (VMEXIT_CONTINUE);
336	}
337}
338
339static int
340vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
341{
342	uint64_t val;
343	uint32_t eax, edx;
344	int error;
345
346	val = 0;
347	error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val);
348	if (error != 0) {
349		fprintf(stderr, "rdmsr to register %#x on vcpu %d\n",
350		    vme->u.msr.code, *pvcpu);
351		if (strictmsr) {
352			vm_inject_gp(ctx, *pvcpu);
353			return (VMEXIT_CONTINUE);
354		}
355	}
356
357	eax = val;
358	error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax);
359	assert(error == 0);
360
361	edx = val >> 32;
362	error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx);
363	assert(error == 0);
364
365	return (VMEXIT_CONTINUE);
366}
367
368static int
369vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
370{
371	int error;
372
373	error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval);
374	if (error != 0) {
375		fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n",
376		    vme->u.msr.code, vme->u.msr.wval, *pvcpu);
377		if (strictmsr) {
378			vm_inject_gp(ctx, *pvcpu);
379			return (VMEXIT_CONTINUE);
380		}
381	}
382	return (VMEXIT_CONTINUE);
383}
384
385static int
386vmexit_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
403static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4];
404static int ept_misconfig_ptenum;
405#endif
406
407static int
408vmexit_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
439static int
440vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
441{
442
443	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
444	fprintf(stderr, "\treason\t\tSVM\n");
445	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
446	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
447	fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode);
448	fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1);
449	fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2);
450	return (VMEXIT_ABORT);
451}
452
453static int
454vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
455{
456
457	assert(vmexit->inst_length == 0);
458
459	stats.vmexit_bogus++;
460
461	return (VMEXIT_CONTINUE);
462}
463
464static int
465vmexit_reqidle(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
466{
467
468	assert(vmexit->inst_length == 0);
469
470	stats.vmexit_reqidle++;
471
472	return (VMEXIT_CONTINUE);
473}
474
475static int
476vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
477{
478
479	stats.vmexit_hlt++;
480
481	/*
482	 * Just continue execution with the next instruction. We use
483	 * the HLT VM exit as a way to be friendly with the host
484	 * scheduler.
485	 */
486	return (VMEXIT_CONTINUE);
487}
488
489static int
490vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
491{
492
493	stats.vmexit_pause++;
494
495	return (VMEXIT_CONTINUE);
496}
497
498static int
499vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
500{
501
502	assert(vmexit->inst_length == 0);
503
504	stats.vmexit_mtrap++;
505
506	return (VMEXIT_CONTINUE);
507}
508
509static int
510vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
511{
512	int err, i;
513	struct vie *vie;
514
515	stats.vmexit_inst_emul++;
516
517	vie = &vmexit->u.inst_emul.vie;
518	err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
519	    vie, &vmexit->u.inst_emul.paging);
520
521	if (err) {
522		if (err == ESRCH) {
523			fprintf(stderr, "Unhandled memory access to 0x%lx\n",
524			    vmexit->u.inst_emul.gpa);
525		}
526
527		fprintf(stderr, "Failed to emulate instruction [");
528		for (i = 0; i < vie->num_valid; i++) {
529			fprintf(stderr, "0x%02x%s", vie->inst[i],
530			    i != (vie->num_valid - 1) ? " " : "");
531		}
532		fprintf(stderr, "] at 0x%lx\n", vmexit->rip);
533		return (VMEXIT_ABORT);
534	}
535
536	return (VMEXIT_CONTINUE);
537}
538
539static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER;
540static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER;
541
542static int
543vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
544{
545	enum vm_suspend_how how;
546
547	how = vmexit->u.suspended.how;
548
549	fbsdrun_deletecpu(ctx, *pvcpu);
550
551	if (*pvcpu != BSP) {
552		pthread_mutex_lock(&resetcpu_mtx);
553		pthread_cond_signal(&resetcpu_cond);
554		pthread_mutex_unlock(&resetcpu_mtx);
555		pthread_exit(NULL);
556	}
557
558	pthread_mutex_lock(&resetcpu_mtx);
559	while (!CPU_EMPTY(&cpumask)) {
560		pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx);
561	}
562	pthread_mutex_unlock(&resetcpu_mtx);
563
564	switch (how) {
565	case VM_SUSPEND_RESET:
566		exit(0);
567	case VM_SUSPEND_POWEROFF:
568		exit(1);
569	case VM_SUSPEND_HALT:
570		exit(2);
571	case VM_SUSPEND_TRIPLEFAULT:
572		exit(3);
573	default:
574		fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how);
575		exit(100);
576	}
577	return (0);	/* NOTREACHED */
578}
579
580static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
581	[VM_EXITCODE_INOUT]  = vmexit_inout,
582	[VM_EXITCODE_INOUT_STR]  = vmexit_inout,
583	[VM_EXITCODE_VMX]    = vmexit_vmx,
584	[VM_EXITCODE_SVM]    = vmexit_svm,
585	[VM_EXITCODE_BOGUS]  = vmexit_bogus,
586	[VM_EXITCODE_REQIDLE] = vmexit_reqidle,
587	[VM_EXITCODE_RDMSR]  = vmexit_rdmsr,
588	[VM_EXITCODE_WRMSR]  = vmexit_wrmsr,
589	[VM_EXITCODE_MTRAP]  = vmexit_mtrap,
590	[VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
591	[VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
592	[VM_EXITCODE_SUSPENDED] = vmexit_suspend,
593	[VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch,
594};
595
596static void
597vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip)
598{
599	int error, rc, prevcpu;
600	enum vm_exitcode exitcode;
601	cpuset_t active_cpus;
602
603	if (vcpumap[vcpu] != NULL) {
604		error = pthread_setaffinity_np(pthread_self(),
605		    sizeof(cpuset_t), vcpumap[vcpu]);
606		assert(error == 0);
607	}
608
609	error = vm_active_cpus(ctx, &active_cpus);
610	assert(CPU_ISSET(vcpu, &active_cpus));
611
612	error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip);
613	assert(error == 0);
614
615	while (1) {
616		error = vm_run(ctx, vcpu, &vmexit[vcpu]);
617		if (error != 0)
618			break;
619
620		prevcpu = vcpu;
621
622		exitcode = vmexit[vcpu].exitcode;
623		if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
624			fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
625			    exitcode);
626			exit(1);
627		}
628
629                rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
630
631		switch (rc) {
632		case VMEXIT_CONTINUE:
633			break;
634		case VMEXIT_ABORT:
635			abort();
636		default:
637			exit(1);
638		}
639	}
640	fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
641}
642
643static int
644num_vcpus_allowed(struct vmctx *ctx)
645{
646	int tmp, error;
647
648	error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
649
650	/*
651	 * The guest is allowed to spinup more than one processor only if the
652	 * UNRESTRICTED_GUEST capability is available.
653	 */
654	if (error == 0)
655		return (VM_MAXCPU);
656	else
657		return (1);
658}
659
660void
661fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
662{
663	int err, tmp;
664
665	if (fbsdrun_vmexit_on_hlt()) {
666		err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
667		if (err < 0) {
668			fprintf(stderr, "VM exit on HLT not supported\n");
669			exit(1);
670		}
671		vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
672		if (cpu == BSP)
673			handler[VM_EXITCODE_HLT] = vmexit_hlt;
674	}
675
676        if (fbsdrun_vmexit_on_pause()) {
677		/*
678		 * pause exit support required for this mode
679		 */
680		err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
681		if (err < 0) {
682			fprintf(stderr,
683			    "SMP mux requested, no pause support\n");
684			exit(1);
685		}
686		vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
687		if (cpu == BSP)
688			handler[VM_EXITCODE_PAUSE] = vmexit_pause;
689        }
690
691	if (x2apic_mode)
692		err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
693	else
694		err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
695
696	if (err) {
697		fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
698		exit(1);
699	}
700
701	vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
702}
703
704int
705main(int argc, char *argv[])
706{
707	int c, error, gdb_port, err, bvmcons;
708	int dump_guest_memory, max_vcpus, mptgen;
709	int rtc_localtime;
710	struct vmctx *ctx;
711	uint64_t rip;
712	size_t memsize;
713
714	bvmcons = 0;
715	dump_guest_memory = 0;
716	progname = basename(argv[0]);
717	gdb_port = 0;
718	guest_ncpus = 1;
719	memsize = 256 * MB;
720	mptgen = 1;
721	rtc_localtime = 1;
722
723	while ((c = getopt(argc, argv, "abehuwxACHIPWYp:g:c:s:m:l:U:")) != -1) {
724		switch (c) {
725		case 'a':
726			x2apic_mode = 0;
727			break;
728		case 'A':
729			acpi = 1;
730			break;
731		case 'b':
732			bvmcons = 1;
733			break;
734		case 'p':
735                        if (pincpu_parse(optarg) != 0) {
736                            errx(EX_USAGE, "invalid vcpu pinning "
737                                 "configuration '%s'", optarg);
738                        }
739			break;
740                case 'c':
741			guest_ncpus = atoi(optarg);
742			break;
743		case 'C':
744			dump_guest_memory = 1;
745			break;
746		case 'g':
747			gdb_port = atoi(optarg);
748			break;
749		case 'l':
750			if (lpc_device_parse(optarg) != 0) {
751				errx(EX_USAGE, "invalid lpc device "
752				    "configuration '%s'", optarg);
753			}
754			break;
755		case 's':
756			if (pci_parse_slot(optarg) != 0)
757				exit(1);
758			else
759				break;
760                case 'm':
761			error = vm_parse_memsize(optarg, &memsize);
762			if (error)
763				errx(EX_USAGE, "invalid memsize '%s'", optarg);
764			break;
765		case 'H':
766			guest_vmexit_on_hlt = 1;
767			break;
768		case 'I':
769			/*
770			 * The "-I" option was used to add an ioapic to the
771			 * virtual machine.
772			 *
773			 * An ioapic is now provided unconditionally for each
774			 * virtual machine and this option is now deprecated.
775			 */
776			break;
777		case 'P':
778			guest_vmexit_on_pause = 1;
779			break;
780		case 'e':
781			strictio = 1;
782			break;
783		case 'u':
784			rtc_localtime = 0;
785			break;
786		case 'U':
787			guest_uuid_str = optarg;
788			break;
789		case 'w':
790			strictmsr = 0;
791			break;
792		case 'W':
793			virtio_msix = 0;
794			break;
795		case 'x':
796			x2apic_mode = 1;
797			break;
798		case 'Y':
799			mptgen = 0;
800			break;
801		case 'h':
802			usage(0);
803		default:
804			usage(1);
805		}
806	}
807	argc -= optind;
808	argv += optind;
809
810	if (argc != 1)
811		usage(1);
812
813	vmname = argv[0];
814
815	ctx = vm_open(vmname);
816	if (ctx == NULL) {
817		perror("vm_open");
818		exit(1);
819	}
820
821	if (guest_ncpus < 1) {
822		fprintf(stderr, "Invalid guest vCPUs (%d)\n", guest_ncpus);
823		exit(1);
824	}
825
826	max_vcpus = num_vcpus_allowed(ctx);
827	if (guest_ncpus > max_vcpus) {
828		fprintf(stderr, "%d vCPUs requested but only %d available\n",
829			guest_ncpus, max_vcpus);
830		exit(1);
831	}
832
833	fbsdrun_set_capabilities(ctx, BSP);
834
835	if (dump_guest_memory)
836		vm_set_memflags(ctx, VM_MEM_F_INCORE);
837	err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
838	if (err) {
839		fprintf(stderr, "Unable to setup memory (%d)\n", err);
840		exit(1);
841	}
842
843	error = init_msr();
844	if (error) {
845		fprintf(stderr, "init_msr error %d", error);
846		exit(1);
847	}
848
849	init_mem();
850	init_inout();
851	pci_irq_init(ctx);
852	ioapic_init(ctx);
853
854	rtc_init(ctx, rtc_localtime);
855	sci_init(ctx);
856
857	/*
858	 * Exit if a device emulation finds an error in it's initilization
859	 */
860	if (init_pci(ctx) != 0)
861		exit(1);
862
863	if (gdb_port != 0)
864		init_dbgport(gdb_port);
865
866	if (bvmcons)
867		init_bvmcons();
868
869	error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
870	assert(error == 0);
871
872	/*
873	 * build the guest tables, MP etc.
874	 */
875	if (mptgen) {
876		error = mptable_build(ctx, guest_ncpus);
877		if (error)
878			exit(1);
879	}
880
881	error = smbios_build(ctx);
882	assert(error == 0);
883
884	if (acpi) {
885		error = acpi_build(ctx, guest_ncpus);
886		assert(error == 0);
887	}
888
889	/*
890	 * Change the proc title to include the VM name.
891	 */
892	setproctitle("%s", vmname);
893
894	/*
895	 * Add CPU 0
896	 */
897	fbsdrun_addcpu(ctx, BSP, BSP, rip);
898
899	/*
900	 * Head off to the main event dispatch loop
901	 */
902	mevent_dispatch();
903
904	exit(1);
905}
906