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: releng/11.0/usr.sbin/bhyve/bhyverun.c 302362 2016-07-06 04:56:45Z ngie $
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD: releng/11.0/usr.sbin/bhyve/bhyverun.c 302362 2016-07-06 04:56:45Z 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#include <stdbool.h>
51
52#include <machine/vmm.h>
53#include <vmmapi.h>
54
55#include "bhyverun.h"
56#include "acpi.h"
57#include "atkbdc.h"
58#include "inout.h"
59#include "dbgport.h"
60#include "fwctl.h"
61#include "ioapic.h"
62#include "mem.h"
63#include "mevent.h"
64#include "mptbl.h"
65#include "pci_emul.h"
66#include "pci_irq.h"
67#include "pci_lpc.h"
68#include "smbiostbl.h"
69#include "xmsr.h"
70#include "spinup_ap.h"
71#include "rtc.h"
72
73#define GUEST_NIO_PORT		0x488	/* guest upcalls via i/o port */
74
75#define MB		(1024UL * 1024)
76#define GB		(1024UL * MB)
77
78typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
79extern int vmexit_task_switch(struct vmctx *, struct vm_exit *, int *vcpu);
80
81char *vmname;
82
83int guest_ncpus;
84char *guest_uuid_str;
85
86static int guest_vmexit_on_hlt, guest_vmexit_on_pause;
87static int virtio_msix = 1;
88static int x2apic_mode = 0;	/* default is xAPIC */
89
90static int strictio;
91static int strictmsr = 1;
92
93static int acpi;
94
95static char *progname;
96static const int BSP = 0;
97
98static cpuset_t cpumask;
99
100static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
101
102static struct vm_exit vmexit[VM_MAXCPU];
103
104struct bhyvestats {
105        uint64_t        vmexit_bogus;
106	uint64_t	vmexit_reqidle;
107        uint64_t        vmexit_hlt;
108        uint64_t        vmexit_pause;
109        uint64_t        vmexit_mtrap;
110        uint64_t        vmexit_inst_emul;
111        uint64_t        cpu_switch_rotate;
112        uint64_t        cpu_switch_direct;
113} stats;
114
115struct mt_vmm_info {
116	pthread_t	mt_thr;
117	struct vmctx	*mt_ctx;
118	int		mt_vcpu;
119} mt_vmm_info[VM_MAXCPU];
120
121static cpuset_t *vcpumap[VM_MAXCPU] = { NULL };
122
123static void
124usage(int code)
125{
126
127        fprintf(stderr,
128                "Usage: %s [-abehuwxACHPSWY] [-c vcpus] [-g <gdb port>] [-l <lpc>]\n"
129		"       %*s [-m mem] [-p vcpu:hostcpu] [-s <pci>] [-U uuid] <vm>\n"
130		"       -a: local apic is in xAPIC mode (deprecated)\n"
131		"       -A: create ACPI tables\n"
132		"       -c: # cpus (default 1)\n"
133		"       -C: include guest memory in core file\n"
134		"       -e: exit on unhandled I/O access\n"
135		"       -g: gdb port\n"
136		"       -h: help\n"
137		"       -H: vmexit from the guest on hlt\n"
138		"       -l: LPC device configuration\n"
139		"       -m: memory size in MB\n"
140		"       -p: pin 'vcpu' to 'hostcpu'\n"
141		"       -P: vmexit from the guest on pause\n"
142		"       -s: <slot,driver,configinfo> PCI slot config\n"
143		"       -S: guest memory cannot be swapped\n"
144		"       -u: RTC keeps UTC time\n"
145		"       -U: uuid\n"
146		"       -w: ignore unimplemented MSRs\n"
147		"       -W: force virtio to use single-vector MSI\n"
148		"       -x: local apic is in x2APIC mode\n"
149		"       -Y: disable MPtable generation\n",
150		progname, (int)strlen(progname), "");
151
152	exit(code);
153}
154
155static int
156pincpu_parse(const char *opt)
157{
158	int vcpu, pcpu;
159
160	if (sscanf(opt, "%d:%d", &vcpu, &pcpu) != 2) {
161		fprintf(stderr, "invalid format: %s\n", opt);
162		return (-1);
163	}
164
165	if (vcpu < 0 || vcpu >= VM_MAXCPU) {
166		fprintf(stderr, "vcpu '%d' outside valid range from 0 to %d\n",
167		    vcpu, VM_MAXCPU - 1);
168		return (-1);
169	}
170
171	if (pcpu < 0 || pcpu >= CPU_SETSIZE) {
172		fprintf(stderr, "hostcpu '%d' outside valid range from "
173		    "0 to %d\n", pcpu, CPU_SETSIZE - 1);
174		return (-1);
175	}
176
177	if (vcpumap[vcpu] == NULL) {
178		if ((vcpumap[vcpu] = malloc(sizeof(cpuset_t))) == NULL) {
179			perror("malloc");
180			return (-1);
181		}
182		CPU_ZERO(vcpumap[vcpu]);
183	}
184	CPU_SET(pcpu, vcpumap[vcpu]);
185	return (0);
186}
187
188void
189vm_inject_fault(void *arg, int vcpu, int vector, int errcode_valid,
190    int errcode)
191{
192	struct vmctx *ctx;
193	int error, restart_instruction;
194
195	ctx = arg;
196	restart_instruction = 1;
197
198	error = vm_inject_exception(ctx, vcpu, vector, errcode_valid, errcode,
199	    restart_instruction);
200	assert(error == 0);
201}
202
203void *
204paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
205{
206
207	return (vm_map_gpa(ctx, gaddr, len));
208}
209
210int
211fbsdrun_vmexit_on_pause(void)
212{
213
214	return (guest_vmexit_on_pause);
215}
216
217int
218fbsdrun_vmexit_on_hlt(void)
219{
220
221	return (guest_vmexit_on_hlt);
222}
223
224int
225fbsdrun_virtio_msix(void)
226{
227
228	return (virtio_msix);
229}
230
231static void *
232fbsdrun_start_thread(void *param)
233{
234	char tname[MAXCOMLEN + 1];
235	struct mt_vmm_info *mtp;
236	int vcpu;
237
238	mtp = param;
239	vcpu = mtp->mt_vcpu;
240
241	snprintf(tname, sizeof(tname), "vcpu %d", vcpu);
242	pthread_set_name_np(mtp->mt_thr, tname);
243
244	vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
245
246	/* not reached */
247	exit(1);
248	return (NULL);
249}
250
251void
252fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip)
253{
254	int error;
255
256	assert(fromcpu == BSP);
257
258	/*
259	 * The 'newcpu' must be activated in the context of 'fromcpu'. If
260	 * vm_activate_cpu() is delayed until newcpu's pthread starts running
261	 * then vmm.ko is out-of-sync with bhyve and this can create a race
262	 * with vm_suspend().
263	 */
264	error = vm_activate_cpu(ctx, newcpu);
265	if (error != 0)
266		err(EX_OSERR, "could not activate CPU %d", newcpu);
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;
315	int vcpu;
316
317	vcpu = *pvcpu;
318
319	port = vme->u.inout.port;
320	bytes = vme->u.inout.bytes;
321	in = vme->u.inout.in;
322	out = !in;
323
324        /* Extra-special case of host notifications */
325        if (out && port == GUEST_NIO_PORT) {
326                error = vmexit_handle_notify(ctx, vme, pvcpu, vme->u.inout.eax);
327		return (error);
328	}
329
330	error = emulate_inout(ctx, vcpu, vme, strictio);
331	if (error) {
332		fprintf(stderr, "Unhandled %s%c 0x%04x at 0x%lx\n",
333		    in ? "in" : "out",
334		    bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'),
335		    port, vmexit->rip);
336		return (VMEXIT_ABORT);
337	} else {
338		return (VMEXIT_CONTINUE);
339	}
340}
341
342static int
343vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
344{
345	uint64_t val;
346	uint32_t eax, edx;
347	int error;
348
349	val = 0;
350	error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val);
351	if (error != 0) {
352		fprintf(stderr, "rdmsr to register %#x on vcpu %d\n",
353		    vme->u.msr.code, *pvcpu);
354		if (strictmsr) {
355			vm_inject_gp(ctx, *pvcpu);
356			return (VMEXIT_CONTINUE);
357		}
358	}
359
360	eax = val;
361	error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax);
362	assert(error == 0);
363
364	edx = val >> 32;
365	error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx);
366	assert(error == 0);
367
368	return (VMEXIT_CONTINUE);
369}
370
371static int
372vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
373{
374	int error;
375
376	error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval);
377	if (error != 0) {
378		fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n",
379		    vme->u.msr.code, vme->u.msr.wval, *pvcpu);
380		if (strictmsr) {
381			vm_inject_gp(ctx, *pvcpu);
382			return (VMEXIT_CONTINUE);
383		}
384	}
385	return (VMEXIT_CONTINUE);
386}
387
388static int
389vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
390{
391
392	(void)spinup_ap(ctx, *pvcpu,
393		    vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
394
395	return (VMEXIT_CONTINUE);
396}
397
398#define	DEBUG_EPT_MISCONFIG
399#ifdef DEBUG_EPT_MISCONFIG
400#define	EXIT_REASON_EPT_MISCONFIG	49
401#define	VMCS_GUEST_PHYSICAL_ADDRESS	0x00002400
402#define	VMCS_IDENT(x)			((x) | 0x80000000)
403
404static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4];
405static int ept_misconfig_ptenum;
406#endif
407
408static int
409vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
410{
411
412	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
413	fprintf(stderr, "\treason\t\tVMX\n");
414	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
415	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
416	fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status);
417	fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
418	fprintf(stderr, "\tqualification\t0x%016lx\n",
419	    vmexit->u.vmx.exit_qualification);
420	fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type);
421	fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error);
422#ifdef DEBUG_EPT_MISCONFIG
423	if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) {
424		vm_get_register(ctx, *pvcpu,
425		    VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS),
426		    &ept_misconfig_gpa);
427		vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte,
428		    &ept_misconfig_ptenum);
429		fprintf(stderr, "\tEPT misconfiguration:\n");
430		fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa);
431		fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n",
432		    ept_misconfig_ptenum, ept_misconfig_pte[0],
433		    ept_misconfig_pte[1], ept_misconfig_pte[2],
434		    ept_misconfig_pte[3]);
435	}
436#endif	/* DEBUG_EPT_MISCONFIG */
437	return (VMEXIT_ABORT);
438}
439
440static int
441vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
442{
443
444	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
445	fprintf(stderr, "\treason\t\tSVM\n");
446	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
447	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
448	fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode);
449	fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1);
450	fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2);
451	return (VMEXIT_ABORT);
452}
453
454static int
455vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
456{
457
458	assert(vmexit->inst_length == 0);
459
460	stats.vmexit_bogus++;
461
462	return (VMEXIT_CONTINUE);
463}
464
465static int
466vmexit_reqidle(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
467{
468
469	assert(vmexit->inst_length == 0);
470
471	stats.vmexit_reqidle++;
472
473	return (VMEXIT_CONTINUE);
474}
475
476static int
477vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
478{
479
480	stats.vmexit_hlt++;
481
482	/*
483	 * Just continue execution with the next instruction. We use
484	 * the HLT VM exit as a way to be friendly with the host
485	 * scheduler.
486	 */
487	return (VMEXIT_CONTINUE);
488}
489
490static int
491vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
492{
493
494	stats.vmexit_pause++;
495
496	return (VMEXIT_CONTINUE);
497}
498
499static int
500vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
501{
502
503	assert(vmexit->inst_length == 0);
504
505	stats.vmexit_mtrap++;
506
507	return (VMEXIT_CONTINUE);
508}
509
510static int
511vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
512{
513	int err, i;
514	struct vie *vie;
515
516	stats.vmexit_inst_emul++;
517
518	vie = &vmexit->u.inst_emul.vie;
519	err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
520	    vie, &vmexit->u.inst_emul.paging);
521
522	if (err) {
523		if (err == ESRCH) {
524			fprintf(stderr, "Unhandled memory access to 0x%lx\n",
525			    vmexit->u.inst_emul.gpa);
526		}
527
528		fprintf(stderr, "Failed to emulate instruction [");
529		for (i = 0; i < vie->num_valid; i++) {
530			fprintf(stderr, "0x%02x%s", vie->inst[i],
531			    i != (vie->num_valid - 1) ? " " : "");
532		}
533		fprintf(stderr, "] at 0x%lx\n", vmexit->rip);
534		return (VMEXIT_ABORT);
535	}
536
537	return (VMEXIT_CONTINUE);
538}
539
540static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER;
541static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER;
542
543static int
544vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
545{
546	enum vm_suspend_how how;
547
548	how = vmexit->u.suspended.how;
549
550	fbsdrun_deletecpu(ctx, *pvcpu);
551
552	if (*pvcpu != BSP) {
553		pthread_mutex_lock(&resetcpu_mtx);
554		pthread_cond_signal(&resetcpu_cond);
555		pthread_mutex_unlock(&resetcpu_mtx);
556		pthread_exit(NULL);
557	}
558
559	pthread_mutex_lock(&resetcpu_mtx);
560	while (!CPU_EMPTY(&cpumask)) {
561		pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx);
562	}
563	pthread_mutex_unlock(&resetcpu_mtx);
564
565	switch (how) {
566	case VM_SUSPEND_RESET:
567		exit(0);
568	case VM_SUSPEND_POWEROFF:
569		exit(1);
570	case VM_SUSPEND_HALT:
571		exit(2);
572	case VM_SUSPEND_TRIPLEFAULT:
573		exit(3);
574	default:
575		fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how);
576		exit(100);
577	}
578	return (0);	/* NOTREACHED */
579}
580
581static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
582	[VM_EXITCODE_INOUT]  = vmexit_inout,
583	[VM_EXITCODE_INOUT_STR]  = vmexit_inout,
584	[VM_EXITCODE_VMX]    = vmexit_vmx,
585	[VM_EXITCODE_SVM]    = vmexit_svm,
586	[VM_EXITCODE_BOGUS]  = vmexit_bogus,
587	[VM_EXITCODE_REQIDLE] = vmexit_reqidle,
588	[VM_EXITCODE_RDMSR]  = vmexit_rdmsr,
589	[VM_EXITCODE_WRMSR]  = vmexit_wrmsr,
590	[VM_EXITCODE_MTRAP]  = vmexit_mtrap,
591	[VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
592	[VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
593	[VM_EXITCODE_SUSPENDED] = vmexit_suspend,
594	[VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch,
595};
596
597static void
598vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip)
599{
600	int error, rc;
601	enum vm_exitcode exitcode;
602	cpuset_t active_cpus;
603
604	if (vcpumap[vcpu] != NULL) {
605		error = pthread_setaffinity_np(pthread_self(),
606		    sizeof(cpuset_t), vcpumap[vcpu]);
607		assert(error == 0);
608	}
609
610	error = vm_active_cpus(ctx, &active_cpus);
611	assert(CPU_ISSET(vcpu, &active_cpus));
612
613	error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip);
614	assert(error == 0);
615
616	while (1) {
617		error = vm_run(ctx, vcpu, &vmexit[vcpu]);
618		if (error != 0)
619			break;
620
621		exitcode = vmexit[vcpu].exitcode;
622		if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
623			fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
624			    exitcode);
625			exit(1);
626		}
627
628		rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
629
630		switch (rc) {
631		case VMEXIT_CONTINUE:
632			break;
633		case VMEXIT_ABORT:
634			abort();
635		default:
636			exit(1);
637		}
638	}
639	fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
640}
641
642static int
643num_vcpus_allowed(struct vmctx *ctx)
644{
645	int tmp, error;
646
647	error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
648
649	/*
650	 * The guest is allowed to spinup more than one processor only if the
651	 * UNRESTRICTED_GUEST capability is available.
652	 */
653	if (error == 0)
654		return (VM_MAXCPU);
655	else
656		return (1);
657}
658
659void
660fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
661{
662	int err, tmp;
663
664	if (fbsdrun_vmexit_on_hlt()) {
665		err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
666		if (err < 0) {
667			fprintf(stderr, "VM exit on HLT not supported\n");
668			exit(1);
669		}
670		vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
671		if (cpu == BSP)
672			handler[VM_EXITCODE_HLT] = vmexit_hlt;
673	}
674
675        if (fbsdrun_vmexit_on_pause()) {
676		/*
677		 * pause exit support required for this mode
678		 */
679		err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
680		if (err < 0) {
681			fprintf(stderr,
682			    "SMP mux requested, no pause support\n");
683			exit(1);
684		}
685		vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
686		if (cpu == BSP)
687			handler[VM_EXITCODE_PAUSE] = vmexit_pause;
688        }
689
690	if (x2apic_mode)
691		err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
692	else
693		err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
694
695	if (err) {
696		fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
697		exit(1);
698	}
699
700	vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
701}
702
703static struct vmctx *
704do_open(const char *vmname)
705{
706	struct vmctx *ctx;
707	int error;
708	bool reinit, romboot;
709
710	reinit = romboot = false;
711
712	if (lpc_bootrom())
713		romboot = true;
714
715	error = vm_create(vmname);
716	if (error) {
717		if (errno == EEXIST) {
718			if (romboot) {
719				reinit = true;
720			} else {
721				/*
722				 * The virtual machine has been setup by the
723				 * userspace bootloader.
724				 */
725			}
726		} else {
727			perror("vm_create");
728			exit(1);
729		}
730	} else {
731		if (!romboot) {
732			/*
733			 * If the virtual machine was just created then a
734			 * bootrom must be configured to boot it.
735			 */
736			fprintf(stderr, "virtual machine cannot be booted\n");
737			exit(1);
738		}
739	}
740
741	ctx = vm_open(vmname);
742	if (ctx == NULL) {
743		perror("vm_open");
744		exit(1);
745	}
746
747	if (reinit) {
748		error = vm_reinit(ctx);
749		if (error) {
750			perror("vm_reinit");
751			exit(1);
752		}
753	}
754	return (ctx);
755}
756
757int
758main(int argc, char *argv[])
759{
760	int c, error, gdb_port, err, bvmcons;
761	int max_vcpus, mptgen, memflags;
762	int rtc_localtime;
763	struct vmctx *ctx;
764	uint64_t rip;
765	size_t memsize;
766	char *optstr;
767
768	bvmcons = 0;
769	progname = basename(argv[0]);
770	gdb_port = 0;
771	guest_ncpus = 1;
772	memsize = 256 * MB;
773	mptgen = 1;
774	rtc_localtime = 1;
775	memflags = 0;
776
777	optstr = "abehuwxACHIPSWYp:g:c:s:m:l:U:";
778	while ((c = getopt(argc, argv, optstr)) != -1) {
779		switch (c) {
780		case 'a':
781			x2apic_mode = 0;
782			break;
783		case 'A':
784			acpi = 1;
785			break;
786		case 'b':
787			bvmcons = 1;
788			break;
789		case 'p':
790                        if (pincpu_parse(optarg) != 0) {
791                            errx(EX_USAGE, "invalid vcpu pinning "
792                                 "configuration '%s'", optarg);
793                        }
794			break;
795                case 'c':
796			guest_ncpus = atoi(optarg);
797			break;
798		case 'C':
799			memflags |= VM_MEM_F_INCORE;
800			break;
801		case 'g':
802			gdb_port = atoi(optarg);
803			break;
804		case 'l':
805			if (lpc_device_parse(optarg) != 0) {
806				errx(EX_USAGE, "invalid lpc device "
807				    "configuration '%s'", optarg);
808			}
809			break;
810		case 's':
811			if (pci_parse_slot(optarg) != 0)
812				exit(1);
813			else
814				break;
815		case 'S':
816			memflags |= VM_MEM_F_WIRED;
817			break;
818                case 'm':
819			error = vm_parse_memsize(optarg, &memsize);
820			if (error)
821				errx(EX_USAGE, "invalid memsize '%s'", optarg);
822			break;
823		case 'H':
824			guest_vmexit_on_hlt = 1;
825			break;
826		case 'I':
827			/*
828			 * The "-I" option was used to add an ioapic to the
829			 * virtual machine.
830			 *
831			 * An ioapic is now provided unconditionally for each
832			 * virtual machine and this option is now deprecated.
833			 */
834			break;
835		case 'P':
836			guest_vmexit_on_pause = 1;
837			break;
838		case 'e':
839			strictio = 1;
840			break;
841		case 'u':
842			rtc_localtime = 0;
843			break;
844		case 'U':
845			guest_uuid_str = optarg;
846			break;
847		case 'w':
848			strictmsr = 0;
849			break;
850		case 'W':
851			virtio_msix = 0;
852			break;
853		case 'x':
854			x2apic_mode = 1;
855			break;
856		case 'Y':
857			mptgen = 0;
858			break;
859		case 'h':
860			usage(0);
861		default:
862			usage(1);
863		}
864	}
865	argc -= optind;
866	argv += optind;
867
868	if (argc != 1)
869		usage(1);
870
871	vmname = argv[0];
872	ctx = do_open(vmname);
873
874	if (guest_ncpus < 1) {
875		fprintf(stderr, "Invalid guest vCPUs (%d)\n", guest_ncpus);
876		exit(1);
877	}
878
879	max_vcpus = num_vcpus_allowed(ctx);
880	if (guest_ncpus > max_vcpus) {
881		fprintf(stderr, "%d vCPUs requested but only %d available\n",
882			guest_ncpus, max_vcpus);
883		exit(1);
884	}
885
886	fbsdrun_set_capabilities(ctx, BSP);
887
888	vm_set_memflags(ctx, memflags);
889	err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
890	if (err) {
891		fprintf(stderr, "Unable to setup memory (%d)\n", errno);
892		exit(1);
893	}
894
895	error = init_msr();
896	if (error) {
897		fprintf(stderr, "init_msr error %d", error);
898		exit(1);
899	}
900
901	init_mem();
902	init_inout();
903	atkbdc_init(ctx);
904	pci_irq_init(ctx);
905	ioapic_init(ctx);
906
907	rtc_init(ctx, rtc_localtime);
908	sci_init(ctx);
909
910	/*
911	 * Exit if a device emulation finds an error in it's initilization
912	 */
913	if (init_pci(ctx) != 0)
914		exit(1);
915
916	if (gdb_port != 0)
917		init_dbgport(gdb_port);
918
919	if (bvmcons)
920		init_bvmcons();
921
922	if (lpc_bootrom()) {
923		if (vm_set_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, 1)) {
924			fprintf(stderr, "ROM boot failed: unrestricted guest "
925			    "capability not available\n");
926			exit(1);
927		}
928		error = vcpu_reset(ctx, BSP);
929		assert(error == 0);
930	}
931
932	error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
933	assert(error == 0);
934
935	/*
936	 * build the guest tables, MP etc.
937	 */
938	if (mptgen) {
939		error = mptable_build(ctx, guest_ncpus);
940		if (error)
941			exit(1);
942	}
943
944	error = smbios_build(ctx);
945	assert(error == 0);
946
947	if (acpi) {
948		error = acpi_build(ctx, guest_ncpus);
949		assert(error == 0);
950	}
951
952	if (lpc_bootrom())
953		fwctl_init();
954
955	/*
956	 * Change the proc title to include the VM name.
957	 */
958	setproctitle("%s", vmname);
959
960	/*
961	 * Add CPU 0
962	 */
963	fbsdrun_addcpu(ctx, BSP, BSP, rip);
964
965	/*
966	 * Head off to the main event dispatch loop
967	 */
968	mevent_dispatch();
969
970	exit(1);
971}
972