bhyverun.c revision 276403 
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 276403 2014-12-30 08:24:14Z neel $
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD: stable/10/usr.sbin/bhyve/bhyverun.c 276403 2014-12-30 08:24:14Z neel $");
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_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
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 [-abehwxACHPWY] [-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: uuid\n"
141		"       -w: ignore unimplemented MSRs\n"
142		"       -W: force virtio to use single-vector MSI\n"
143		"       -x: local apic is in x2APIC mode\n"
144		"       -Y: disable MPtable generation\n",
145		progname, (int)strlen(progname), "");
146
147	exit(code);
148}
149
150static int
151pincpu_parse(const char *opt)
152{
153	int vcpu, pcpu;
154
155	if (sscanf(opt, "%d:%d", &vcpu, &pcpu) != 2) {
156		fprintf(stderr, "invalid format: %s\n", opt);
157		return (-1);
158	}
159
160	if (vcpu < 0 || vcpu >= VM_MAXCPU) {
161		fprintf(stderr, "vcpu '%d' outside valid range from 0 to %d\n",
162		    vcpu, VM_MAXCPU - 1);
163		return (-1);
164	}
165
166	if (pcpu < 0 || pcpu >= CPU_SETSIZE) {
167		fprintf(stderr, "hostcpu '%d' outside valid range from "
168		    "0 to %d\n", pcpu, CPU_SETSIZE - 1);
169		return (-1);
170	}
171
172	if (vcpumap[vcpu] == NULL) {
173		if ((vcpumap[vcpu] = malloc(sizeof(cpuset_t))) == NULL) {
174			perror("malloc");
175			return (-1);
176		}
177		CPU_ZERO(vcpumap[vcpu]);
178	}
179	CPU_SET(pcpu, vcpumap[vcpu]);
180	return (0);
181}
182
183void
184vm_inject_fault(void *arg, int vcpu, int vector, int errcode_valid,
185    int errcode)
186{
187	struct vmctx *ctx;
188	int error;
189
190	ctx = arg;
191	if (errcode_valid)
192		error = vm_inject_exception2(ctx, vcpu, vector, errcode);
193	else
194		error = vm_inject_exception(ctx, vcpu, vector);
195	assert(error == 0);
196
197	/*
198	 * Set the instruction length to 0 to ensure that the instruction is
199	 * restarted when the fault handler returns.
200	 */
201	vmexit[vcpu].inst_length = 0;
202}
203
204void *
205paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
206{
207
208	return (vm_map_gpa(ctx, gaddr, len));
209}
210
211int
212fbsdrun_vmexit_on_pause(void)
213{
214
215	return (guest_vmexit_on_pause);
216}
217
218int
219fbsdrun_vmexit_on_hlt(void)
220{
221
222	return (guest_vmexit_on_hlt);
223}
224
225int
226fbsdrun_virtio_msix(void)
227{
228
229	return (virtio_msix);
230}
231
232static void *
233fbsdrun_start_thread(void *param)
234{
235	char tname[MAXCOMLEN + 1];
236	struct mt_vmm_info *mtp;
237	int vcpu;
238
239	mtp = param;
240	vcpu = mtp->mt_vcpu;
241
242	snprintf(tname, sizeof(tname), "vcpu %d", vcpu);
243	pthread_set_name_np(mtp->mt_thr, tname);
244
245	vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
246
247	/* not reached */
248	exit(1);
249	return (NULL);
250}
251
252void
253fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip)
254{
255	int error;
256
257	assert(fromcpu == BSP);
258
259	/*
260	 * The 'newcpu' must be activated in the context of 'fromcpu'. If
261	 * vm_activate_cpu() is delayed until newcpu's pthread starts running
262	 * then vmm.ko is out-of-sync with bhyve and this can create a race
263	 * with vm_suspend().
264	 */
265	error = vm_activate_cpu(ctx, newcpu);
266	assert(error == 0);
267
268	CPU_SET_ATOMIC(newcpu, &cpumask);
269
270	/*
271	 * Set up the vmexit struct to allow execution to start
272	 * at the given RIP
273	 */
274	vmexit[newcpu].rip = rip;
275	vmexit[newcpu].inst_length = 0;
276
277	mt_vmm_info[newcpu].mt_ctx = ctx;
278	mt_vmm_info[newcpu].mt_vcpu = newcpu;
279
280	error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL,
281	    fbsdrun_start_thread, &mt_vmm_info[newcpu]);
282	assert(error == 0);
283}
284
285static int
286fbsdrun_deletecpu(struct vmctx *ctx, int vcpu)
287{
288
289	if (!CPU_ISSET(vcpu, &cpumask)) {
290		fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu);
291		exit(1);
292	}
293
294	CPU_CLR_ATOMIC(vcpu, &cpumask);
295	return (CPU_EMPTY(&cpumask));
296}
297
298static int
299vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
300		     uint32_t eax)
301{
302#if BHYVE_DEBUG
303	/*
304	 * put guest-driven debug here
305	 */
306#endif
307        return (VMEXIT_CONTINUE);
308}
309
310static int
311vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
312{
313	int error;
314	int bytes, port, in, out, string;
315	int vcpu;
316
317	vcpu = *pvcpu;
318
319	port = vme->u.inout.port;
320	bytes = vme->u.inout.bytes;
321	string = vme->u.inout.string;
322	in = vme->u.inout.in;
323	out = !in;
324
325        /* Extra-special case of host notifications */
326        if (out && port == GUEST_NIO_PORT) {
327                error = vmexit_handle_notify(ctx, vme, pvcpu, vme->u.inout.eax);
328		return (error);
329	}
330
331	error = emulate_inout(ctx, vcpu, vme, strictio);
332	if (!error && in && !string) {
333		error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RAX,
334		    vme->u.inout.eax);
335		assert(error == 0);
336	}
337
338	if (error) {
339		fprintf(stderr, "Unhandled %s%c 0x%04x\n", in ? "in" : "out",
340		    bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), port);
341		return (VMEXIT_ABORT);
342	} else {
343		return (VMEXIT_CONTINUE);
344	}
345}
346
347static int
348vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
349{
350	uint64_t val;
351	uint32_t eax, edx;
352	int error;
353
354	val = 0;
355	error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val);
356	if (error != 0) {
357		fprintf(stderr, "rdmsr to register %#x on vcpu %d\n",
358		    vme->u.msr.code, *pvcpu);
359		if (strictmsr) {
360			vm_inject_gp(ctx, *pvcpu);
361			return (VMEXIT_RESTART);
362		}
363	}
364
365	eax = val;
366	error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax);
367	assert(error == 0);
368
369	edx = val >> 32;
370	error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx);
371	assert(error == 0);
372
373	return (VMEXIT_CONTINUE);
374}
375
376static int
377vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
378{
379	int error;
380
381	error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval);
382	if (error != 0) {
383		fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n",
384		    vme->u.msr.code, vme->u.msr.wval, *pvcpu);
385		if (strictmsr) {
386			vm_inject_gp(ctx, *pvcpu);
387			return (VMEXIT_RESTART);
388		}
389	}
390	return (VMEXIT_CONTINUE);
391}
392
393static int
394vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
395{
396	int newcpu;
397	int retval = VMEXIT_CONTINUE;
398
399	newcpu = spinup_ap(ctx, *pvcpu,
400			   vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
401
402	return (retval);
403}
404
405#define	DEBUG_EPT_MISCONFIG
406#ifdef DEBUG_EPT_MISCONFIG
407#define	EXIT_REASON_EPT_MISCONFIG	49
408#define	VMCS_GUEST_PHYSICAL_ADDRESS	0x00002400
409#define	VMCS_IDENT(x)			((x) | 0x80000000)
410
411static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4];
412static int ept_misconfig_ptenum;
413#endif
414
415static int
416vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
417{
418
419	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
420	fprintf(stderr, "\treason\t\tVMX\n");
421	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
422	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
423	fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status);
424	fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
425	fprintf(stderr, "\tqualification\t0x%016lx\n",
426	    vmexit->u.vmx.exit_qualification);
427	fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type);
428	fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error);
429#ifdef DEBUG_EPT_MISCONFIG
430	if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) {
431		vm_get_register(ctx, *pvcpu,
432		    VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS),
433		    &ept_misconfig_gpa);
434		vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte,
435		    &ept_misconfig_ptenum);
436		fprintf(stderr, "\tEPT misconfiguration:\n");
437		fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa);
438		fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n",
439		    ept_misconfig_ptenum, ept_misconfig_pte[0],
440		    ept_misconfig_pte[1], ept_misconfig_pte[2],
441		    ept_misconfig_pte[3]);
442	}
443#endif	/* DEBUG_EPT_MISCONFIG */
444	return (VMEXIT_ABORT);
445}
446
447static int
448vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
449{
450
451	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
452	fprintf(stderr, "\treason\t\tSVM\n");
453	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
454	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
455	fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode);
456	fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1);
457	fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2);
458	return (VMEXIT_ABORT);
459}
460
461static int
462vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
463{
464
465	stats.vmexit_bogus++;
466
467	return (VMEXIT_RESTART);
468}
469
470static int
471vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
472{
473
474	stats.vmexit_hlt++;
475
476	/*
477	 * Just continue execution with the next instruction. We use
478	 * the HLT VM exit as a way to be friendly with the host
479	 * scheduler.
480	 */
481	return (VMEXIT_CONTINUE);
482}
483
484static int
485vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
486{
487
488	stats.vmexit_pause++;
489
490	return (VMEXIT_CONTINUE);
491}
492
493static int
494vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
495{
496
497	stats.vmexit_mtrap++;
498
499	return (VMEXIT_RESTART);
500}
501
502static int
503vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
504{
505	int err;
506	stats.vmexit_inst_emul++;
507
508	err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
509	    &vmexit->u.inst_emul.vie, &vmexit->u.inst_emul.paging);
510
511	if (err) {
512		if (err == EINVAL) {
513			fprintf(stderr,
514			    "Failed to emulate instruction at 0x%lx\n",
515			    vmexit->rip);
516		} else if (err == ESRCH) {
517			fprintf(stderr, "Unhandled memory access to 0x%lx\n",
518			    vmexit->u.inst_emul.gpa);
519		}
520
521		return (VMEXIT_ABORT);
522	}
523
524	return (VMEXIT_CONTINUE);
525}
526
527static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER;
528static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER;
529
530static int
531vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
532{
533	enum vm_suspend_how how;
534
535	how = vmexit->u.suspended.how;
536
537	fbsdrun_deletecpu(ctx, *pvcpu);
538
539	if (*pvcpu != BSP) {
540		pthread_mutex_lock(&resetcpu_mtx);
541		pthread_cond_signal(&resetcpu_cond);
542		pthread_mutex_unlock(&resetcpu_mtx);
543		pthread_exit(NULL);
544	}
545
546	pthread_mutex_lock(&resetcpu_mtx);
547	while (!CPU_EMPTY(&cpumask)) {
548		pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx);
549	}
550	pthread_mutex_unlock(&resetcpu_mtx);
551
552	switch (how) {
553	case VM_SUSPEND_RESET:
554		exit(0);
555	case VM_SUSPEND_POWEROFF:
556		exit(1);
557	case VM_SUSPEND_HALT:
558		exit(2);
559	case VM_SUSPEND_TRIPLEFAULT:
560		exit(3);
561	default:
562		fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how);
563		exit(100);
564	}
565	return (0);	/* NOTREACHED */
566}
567
568static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
569	[VM_EXITCODE_INOUT]  = vmexit_inout,
570	[VM_EXITCODE_INOUT_STR]  = vmexit_inout,
571	[VM_EXITCODE_VMX]    = vmexit_vmx,
572	[VM_EXITCODE_SVM]    = vmexit_svm,
573	[VM_EXITCODE_BOGUS]  = vmexit_bogus,
574	[VM_EXITCODE_RDMSR]  = vmexit_rdmsr,
575	[VM_EXITCODE_WRMSR]  = vmexit_wrmsr,
576	[VM_EXITCODE_MTRAP]  = vmexit_mtrap,
577	[VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
578	[VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
579	[VM_EXITCODE_SUSPENDED] = vmexit_suspend,
580	[VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch,
581};
582
583static void
584vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip)
585{
586	int error, rc, prevcpu;
587	enum vm_exitcode exitcode;
588	cpuset_t active_cpus;
589
590	if (vcpumap[vcpu] != NULL) {
591		error = pthread_setaffinity_np(pthread_self(),
592		    sizeof(cpuset_t), vcpumap[vcpu]);
593		assert(error == 0);
594	}
595
596	error = vm_active_cpus(ctx, &active_cpus);
597	assert(CPU_ISSET(vcpu, &active_cpus));
598
599	while (1) {
600		error = vm_run(ctx, vcpu, rip, &vmexit[vcpu]);
601		if (error != 0)
602			break;
603
604		prevcpu = vcpu;
605
606		exitcode = vmexit[vcpu].exitcode;
607		if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
608			fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
609			    exitcode);
610			exit(1);
611		}
612
613                rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
614
615		switch (rc) {
616		case VMEXIT_CONTINUE:
617                        rip = vmexit[vcpu].rip + vmexit[vcpu].inst_length;
618			break;
619		case VMEXIT_RESTART:
620                        rip = vmexit[vcpu].rip;
621			break;
622		case VMEXIT_ABORT:
623			abort();
624		default:
625			exit(1);
626		}
627	}
628	fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
629}
630
631static int
632num_vcpus_allowed(struct vmctx *ctx)
633{
634	int tmp, error;
635
636	error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
637
638	/*
639	 * The guest is allowed to spinup more than one processor only if the
640	 * UNRESTRICTED_GUEST capability is available.
641	 */
642	if (error == 0)
643		return (VM_MAXCPU);
644	else
645		return (1);
646}
647
648void
649fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
650{
651	int err, tmp;
652
653	if (fbsdrun_vmexit_on_hlt()) {
654		err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
655		if (err < 0) {
656			fprintf(stderr, "VM exit on HLT not supported\n");
657			exit(1);
658		}
659		vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
660		if (cpu == BSP)
661			handler[VM_EXITCODE_HLT] = vmexit_hlt;
662	}
663
664        if (fbsdrun_vmexit_on_pause()) {
665		/*
666		 * pause exit support required for this mode
667		 */
668		err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
669		if (err < 0) {
670			fprintf(stderr,
671			    "SMP mux requested, no pause support\n");
672			exit(1);
673		}
674		vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
675		if (cpu == BSP)
676			handler[VM_EXITCODE_PAUSE] = vmexit_pause;
677        }
678
679	if (x2apic_mode)
680		err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
681	else
682		err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
683
684	if (err) {
685		fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
686		exit(1);
687	}
688
689	vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
690}
691
692int
693main(int argc, char *argv[])
694{
695	int c, error, gdb_port, err, bvmcons;
696	int dump_guest_memory, max_vcpus, mptgen;
697	struct vmctx *ctx;
698	uint64_t rip;
699	size_t memsize;
700
701	bvmcons = 0;
702	dump_guest_memory = 0;
703	progname = basename(argv[0]);
704	gdb_port = 0;
705	guest_ncpus = 1;
706	memsize = 256 * MB;
707	mptgen = 1;
708
709	while ((c = getopt(argc, argv, "abehwxACHIPWYp:g:c:s:m:l:U:")) != -1) {
710		switch (c) {
711		case 'a':
712			x2apic_mode = 0;
713			break;
714		case 'A':
715			acpi = 1;
716			break;
717		case 'b':
718			bvmcons = 1;
719			break;
720		case 'p':
721                        if (pincpu_parse(optarg) != 0) {
722                            errx(EX_USAGE, "invalid vcpu pinning "
723                                 "configuration '%s'", optarg);
724                        }
725			break;
726                case 'c':
727			guest_ncpus = atoi(optarg);
728			break;
729		case 'C':
730			dump_guest_memory = 1;
731			break;
732		case 'g':
733			gdb_port = atoi(optarg);
734			break;
735		case 'l':
736			if (lpc_device_parse(optarg) != 0) {
737				errx(EX_USAGE, "invalid lpc device "
738				    "configuration '%s'", optarg);
739			}
740			break;
741		case 's':
742			if (pci_parse_slot(optarg) != 0)
743				exit(1);
744			else
745				break;
746                case 'm':
747			error = vm_parse_memsize(optarg, &memsize);
748			if (error)
749				errx(EX_USAGE, "invalid memsize '%s'", optarg);
750			break;
751		case 'H':
752			guest_vmexit_on_hlt = 1;
753			break;
754		case 'I':
755			/*
756			 * The "-I" option was used to add an ioapic to the
757			 * virtual machine.
758			 *
759			 * An ioapic is now provided unconditionally for each
760			 * virtual machine and this option is now deprecated.
761			 */
762			break;
763		case 'P':
764			guest_vmexit_on_pause = 1;
765			break;
766		case 'e':
767			strictio = 1;
768			break;
769		case 'U':
770			guest_uuid_str = optarg;
771			break;
772		case 'w':
773			strictmsr = 0;
774			break;
775		case 'W':
776			virtio_msix = 0;
777			break;
778		case 'x':
779			x2apic_mode = 1;
780			break;
781		case 'Y':
782			mptgen = 0;
783			break;
784		case 'h':
785			usage(0);
786		default:
787			usage(1);
788		}
789	}
790	argc -= optind;
791	argv += optind;
792
793	if (argc != 1)
794		usage(1);
795
796	vmname = argv[0];
797
798	ctx = vm_open(vmname);
799	if (ctx == NULL) {
800		perror("vm_open");
801		exit(1);
802	}
803
804	max_vcpus = num_vcpus_allowed(ctx);
805	if (guest_ncpus > max_vcpus) {
806		fprintf(stderr, "%d vCPUs requested but only %d available\n",
807			guest_ncpus, max_vcpus);
808		exit(1);
809	}
810
811	fbsdrun_set_capabilities(ctx, BSP);
812
813	if (dump_guest_memory)
814		vm_set_memflags(ctx, VM_MEM_F_INCORE);
815	err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
816	if (err) {
817		fprintf(stderr, "Unable to setup memory (%d)\n", err);
818		exit(1);
819	}
820
821	error = init_msr();
822	if (error) {
823		fprintf(stderr, "init_msr error %d", error);
824		exit(1);
825	}
826
827	init_mem();
828	init_inout();
829	pci_irq_init(ctx);
830	ioapic_init(ctx);
831
832	rtc_init(ctx);
833	sci_init(ctx);
834
835	/*
836	 * Exit if a device emulation finds an error in it's initilization
837	 */
838	if (init_pci(ctx) != 0)
839		exit(1);
840
841	if (gdb_port != 0)
842		init_dbgport(gdb_port);
843
844	if (bvmcons)
845		init_bvmcons();
846
847	error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
848	assert(error == 0);
849
850	/*
851	 * build the guest tables, MP etc.
852	 */
853	if (mptgen) {
854		error = mptable_build(ctx, guest_ncpus);
855		if (error)
856			exit(1);
857	}
858
859	error = smbios_build(ctx);
860	assert(error == 0);
861
862	if (acpi) {
863		error = acpi_build(ctx, guest_ncpus);
864		assert(error == 0);
865	}
866
867	/*
868	 * Change the proc title to include the VM name.
869	 */
870	setproctitle("%s", vmname);
871
872	/*
873	 * Add CPU 0
874	 */
875	fbsdrun_addcpu(ctx, BSP, BSP, rip);
876
877	/*
878	 * Head off to the main event dispatch loop
879	 */
880	mevent_dispatch();
881
882	exit(1);
883}
884