xref: /freebsd/usr.sbin/bhyve/bhyverun.c (revision 23c30549affbe0343c01fd78998248956a246252)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 2011 NetApp, Inc.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  *
28  * $FreeBSD$
29  */
30 
31 #include <sys/cdefs.h>
32 __FBSDID("$FreeBSD$");
33 
34 #include <sys/types.h>
35 #ifndef WITHOUT_CAPSICUM
36 #include <sys/capsicum.h>
37 #endif
38 #include <sys/mman.h>
39 #include <sys/time.h>
40 
41 #include <amd64/vmm/intel/vmcs.h>
42 
43 #include <machine/atomic.h>
44 #include <machine/segments.h>
45 
46 #ifndef WITHOUT_CAPSICUM
47 #include <capsicum_helpers.h>
48 #endif
49 #include <stdio.h>
50 #include <stdlib.h>
51 #include <string.h>
52 #include <err.h>
53 #include <errno.h>
54 #include <libgen.h>
55 #include <unistd.h>
56 #include <assert.h>
57 #include <pthread.h>
58 #include <pthread_np.h>
59 #include <sysexits.h>
60 #include <stdbool.h>
61 #include <stdint.h>
62 
63 #include <machine/vmm.h>
64 #ifndef WITHOUT_CAPSICUM
65 #include <machine/vmm_dev.h>
66 #endif
67 #include <vmmapi.h>
68 
69 #include "bhyverun.h"
70 #include "acpi.h"
71 #include "atkbdc.h"
72 #include "inout.h"
73 #include "dbgport.h"
74 #include "fwctl.h"
75 #include "gdb.h"
76 #include "ioapic.h"
77 #include "mem.h"
78 #include "mevent.h"
79 #include "mptbl.h"
80 #include "pci_emul.h"
81 #include "pci_irq.h"
82 #include "pci_lpc.h"
83 #include "smbiostbl.h"
84 #include "xmsr.h"
85 #include "spinup_ap.h"
86 #include "rtc.h"
87 
88 #define GUEST_NIO_PORT		0x488	/* guest upcalls via i/o port */
89 
90 #define MB		(1024UL * 1024)
91 #define GB		(1024UL * MB)
92 
93 static const char * const vmx_exit_reason_desc[] = {
94 	[EXIT_REASON_EXCEPTION] = "Exception or non-maskable interrupt (NMI)",
95 	[EXIT_REASON_EXT_INTR] = "External interrupt",
96 	[EXIT_REASON_TRIPLE_FAULT] = "Triple fault",
97 	[EXIT_REASON_INIT] = "INIT signal",
98 	[EXIT_REASON_SIPI] = "Start-up IPI (SIPI)",
99 	[EXIT_REASON_IO_SMI] = "I/O system-management interrupt (SMI)",
100 	[EXIT_REASON_SMI] = "Other SMI",
101 	[EXIT_REASON_INTR_WINDOW] = "Interrupt window",
102 	[EXIT_REASON_NMI_WINDOW] = "NMI window",
103 	[EXIT_REASON_TASK_SWITCH] = "Task switch",
104 	[EXIT_REASON_CPUID] = "CPUID",
105 	[EXIT_REASON_GETSEC] = "GETSEC",
106 	[EXIT_REASON_HLT] = "HLT",
107 	[EXIT_REASON_INVD] = "INVD",
108 	[EXIT_REASON_INVLPG] = "INVLPG",
109 	[EXIT_REASON_RDPMC] = "RDPMC",
110 	[EXIT_REASON_RDTSC] = "RDTSC",
111 	[EXIT_REASON_RSM] = "RSM",
112 	[EXIT_REASON_VMCALL] = "VMCALL",
113 	[EXIT_REASON_VMCLEAR] = "VMCLEAR",
114 	[EXIT_REASON_VMLAUNCH] = "VMLAUNCH",
115 	[EXIT_REASON_VMPTRLD] = "VMPTRLD",
116 	[EXIT_REASON_VMPTRST] = "VMPTRST",
117 	[EXIT_REASON_VMREAD] = "VMREAD",
118 	[EXIT_REASON_VMRESUME] = "VMRESUME",
119 	[EXIT_REASON_VMWRITE] = "VMWRITE",
120 	[EXIT_REASON_VMXOFF] = "VMXOFF",
121 	[EXIT_REASON_VMXON] = "VMXON",
122 	[EXIT_REASON_CR_ACCESS] = "Control-register accesses",
123 	[EXIT_REASON_DR_ACCESS] = "MOV DR",
124 	[EXIT_REASON_INOUT] = "I/O instruction",
125 	[EXIT_REASON_RDMSR] = "RDMSR",
126 	[EXIT_REASON_WRMSR] = "WRMSR",
127 	[EXIT_REASON_INVAL_VMCS] =
128 	    "VM-entry failure due to invalid guest state",
129 	[EXIT_REASON_INVAL_MSR] = "VM-entry failure due to MSR loading",
130 	[EXIT_REASON_MWAIT] = "MWAIT",
131 	[EXIT_REASON_MTF] = "Monitor trap flag",
132 	[EXIT_REASON_MONITOR] = "MONITOR",
133 	[EXIT_REASON_PAUSE] = "PAUSE",
134 	[EXIT_REASON_MCE_DURING_ENTRY] =
135 	    "VM-entry failure due to machine-check event",
136 	[EXIT_REASON_TPR] = "TPR below threshold",
137 	[EXIT_REASON_APIC_ACCESS] = "APIC access",
138 	[EXIT_REASON_VIRTUALIZED_EOI] = "Virtualized EOI",
139 	[EXIT_REASON_GDTR_IDTR] = "Access to GDTR or IDTR",
140 	[EXIT_REASON_LDTR_TR] = "Access to LDTR or TR",
141 	[EXIT_REASON_EPT_FAULT] = "EPT violation",
142 	[EXIT_REASON_EPT_MISCONFIG] = "EPT misconfiguration",
143 	[EXIT_REASON_INVEPT] = "INVEPT",
144 	[EXIT_REASON_RDTSCP] = "RDTSCP",
145 	[EXIT_REASON_VMX_PREEMPT] = "VMX-preemption timer expired",
146 	[EXIT_REASON_INVVPID] = "INVVPID",
147 	[EXIT_REASON_WBINVD] = "WBINVD",
148 	[EXIT_REASON_XSETBV] = "XSETBV",
149 	[EXIT_REASON_APIC_WRITE] = "APIC write",
150 	[EXIT_REASON_RDRAND] = "RDRAND",
151 	[EXIT_REASON_INVPCID] = "INVPCID",
152 	[EXIT_REASON_VMFUNC] = "VMFUNC",
153 	[EXIT_REASON_ENCLS] = "ENCLS",
154 	[EXIT_REASON_RDSEED] = "RDSEED",
155 	[EXIT_REASON_PM_LOG_FULL] = "Page-modification log full",
156 	[EXIT_REASON_XSAVES] = "XSAVES",
157 	[EXIT_REASON_XRSTORS] = "XRSTORS"
158 };
159 
160 typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
161 extern int vmexit_task_switch(struct vmctx *, struct vm_exit *, int *vcpu);
162 
163 char *vmname;
164 
165 int guest_ncpus;
166 uint16_t cores, maxcpus, sockets, threads;
167 
168 char *guest_uuid_str;
169 
170 static int gdb_port = 0;
171 static int guest_vmexit_on_hlt, guest_vmexit_on_pause;
172 static int virtio_msix = 1;
173 static int x2apic_mode = 0;	/* default is xAPIC */
174 
175 static int strictio;
176 static int strictmsr = 1;
177 
178 static int acpi;
179 
180 static char *progname;
181 static const int BSP = 0;
182 
183 static cpuset_t cpumask;
184 
185 static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
186 
187 static struct vm_exit vmexit[VM_MAXCPU];
188 
189 struct bhyvestats {
190 	uint64_t	vmexit_bogus;
191 	uint64_t	vmexit_reqidle;
192 	uint64_t	vmexit_hlt;
193 	uint64_t	vmexit_pause;
194 	uint64_t	vmexit_mtrap;
195 	uint64_t	vmexit_inst_emul;
196 	uint64_t	cpu_switch_rotate;
197 	uint64_t	cpu_switch_direct;
198 } stats;
199 
200 struct mt_vmm_info {
201 	pthread_t	mt_thr;
202 	struct vmctx	*mt_ctx;
203 	int		mt_vcpu;
204 } mt_vmm_info[VM_MAXCPU];
205 
206 static cpuset_t *vcpumap[VM_MAXCPU] = { NULL };
207 
208 static void
209 usage(int code)
210 {
211 
212         fprintf(stderr,
213 		"Usage: %s [-abehuwxACHPSWY]\n"
214 		"       %*s [-c [[cpus=]numcpus][,sockets=n][,cores=n][,threads=n]]\n"
215 		"       %*s [-g <gdb port>] [-l <lpc>]\n"
216 		"       %*s [-m mem] [-p vcpu:hostcpu] [-s <pci>] [-U uuid] <vm>\n"
217 		"       -a: local apic is in xAPIC mode (deprecated)\n"
218 		"       -A: create ACPI tables\n"
219 		"       -c: number of cpus and/or topology specification\n"
220 		"       -C: include guest memory in core file\n"
221 		"       -e: exit on unhandled I/O access\n"
222 		"       -g: gdb port\n"
223 		"       -h: help\n"
224 		"       -H: vmexit from the guest on hlt\n"
225 		"       -l: LPC device configuration\n"
226 		"       -m: memory size in MB\n"
227 		"       -p: pin 'vcpu' to 'hostcpu'\n"
228 		"       -P: vmexit from the guest on pause\n"
229 		"       -s: <slot,driver,configinfo> PCI slot config\n"
230 		"       -S: guest memory cannot be swapped\n"
231 		"       -u: RTC keeps UTC time\n"
232 		"       -U: uuid\n"
233 		"       -w: ignore unimplemented MSRs\n"
234 		"       -W: force virtio to use single-vector MSI\n"
235 		"       -x: local apic is in x2APIC mode\n"
236 		"       -Y: disable MPtable generation\n",
237 		progname, (int)strlen(progname), "", (int)strlen(progname), "",
238 		(int)strlen(progname), "");
239 
240 	exit(code);
241 }
242 
243 /*
244  * XXX This parser is known to have the following issues:
245  * 1.  It accepts null key=value tokens ",,".
246  * 2.  It accepts whitespace after = and before value.
247  * 3.  Values out of range of INT are silently wrapped.
248  * 4.  It doesn't check non-final values.
249  * 5.  The apparently bogus limits of UINT16_MAX are for future expansion.
250  *
251  * The acceptance of a null specification ('-c ""') is by design to match the
252  * manual page syntax specification, this results in a topology of 1 vCPU.
253  */
254 static int
255 topology_parse(const char *opt)
256 {
257 	uint64_t ncpus;
258 	int c, chk, n, s, t, tmp;
259 	char *cp, *str;
260 	bool ns, scts;
261 
262 	c = 1, n = 1, s = 1, t = 1;
263 	ns = false, scts = false;
264 	str = strdup(opt);
265 	if (str == NULL)
266 		goto out;
267 
268 	while ((cp = strsep(&str, ",")) != NULL) {
269 		if (sscanf(cp, "%i%n", &tmp, &chk) == 1) {
270 			n = tmp;
271 			ns = true;
272 		} else if (sscanf(cp, "cpus=%i%n", &tmp, &chk) == 1) {
273 			n = tmp;
274 			ns = true;
275 		} else if (sscanf(cp, "sockets=%i%n", &tmp, &chk) == 1) {
276 			s = tmp;
277 			scts = true;
278 		} else if (sscanf(cp, "cores=%i%n", &tmp, &chk) == 1) {
279 			c = tmp;
280 			scts = true;
281 		} else if (sscanf(cp, "threads=%i%n", &tmp, &chk) == 1) {
282 			t = tmp;
283 			scts = true;
284 #ifdef notyet  /* Do not expose this until vmm.ko implements it */
285 		} else if (sscanf(cp, "maxcpus=%i%n", &tmp, &chk) == 1) {
286 			m = tmp;
287 #endif
288 		/* Skip the empty argument case from -c "" */
289 		} else if (cp[0] == '\0')
290 			continue;
291 		else
292 			goto out;
293 		/* Any trailing garbage causes an error */
294 		if (cp[chk] != '\0')
295 			goto out;
296 	}
297 	free(str);
298 	str = NULL;
299 
300 	/*
301 	 * Range check 1 <= n <= UINT16_MAX all values
302 	 */
303 	if (n < 1 || s < 1 || c < 1 || t < 1 ||
304 	    n > UINT16_MAX || s > UINT16_MAX || c > UINT16_MAX  ||
305 	    t > UINT16_MAX)
306 		return (-1);
307 
308 	/* If only the cpus was specified, use that as sockets */
309 	if (!scts)
310 		s = n;
311 	/*
312 	 * Compute sockets * cores * threads avoiding overflow
313 	 * The range check above insures these are 16 bit values
314 	 * If n was specified check it against computed ncpus
315 	 */
316 	ncpus = (uint64_t)s * c * t;
317 	if (ncpus > UINT16_MAX || (ns && n != ncpus))
318 		return (-1);
319 
320 	guest_ncpus = ncpus;
321 	sockets = s;
322 	cores = c;
323 	threads = t;
324 	return(0);
325 
326 out:
327 	free(str);
328 	return (-1);
329 }
330 
331 static int
332 pincpu_parse(const char *opt)
333 {
334 	int vcpu, pcpu;
335 
336 	if (sscanf(opt, "%d:%d", &vcpu, &pcpu) != 2) {
337 		fprintf(stderr, "invalid format: %s\n", opt);
338 		return (-1);
339 	}
340 
341 	if (vcpu < 0 || vcpu >= VM_MAXCPU) {
342 		fprintf(stderr, "vcpu '%d' outside valid range from 0 to %d\n",
343 		    vcpu, VM_MAXCPU - 1);
344 		return (-1);
345 	}
346 
347 	if (pcpu < 0 || pcpu >= CPU_SETSIZE) {
348 		fprintf(stderr, "hostcpu '%d' outside valid range from "
349 		    "0 to %d\n", pcpu, CPU_SETSIZE - 1);
350 		return (-1);
351 	}
352 
353 	if (vcpumap[vcpu] == NULL) {
354 		if ((vcpumap[vcpu] = malloc(sizeof(cpuset_t))) == NULL) {
355 			perror("malloc");
356 			return (-1);
357 		}
358 		CPU_ZERO(vcpumap[vcpu]);
359 	}
360 	CPU_SET(pcpu, vcpumap[vcpu]);
361 	return (0);
362 }
363 
364 void
365 vm_inject_fault(void *arg, int vcpu, int vector, int errcode_valid,
366     int errcode)
367 {
368 	struct vmctx *ctx;
369 	int error, restart_instruction;
370 
371 	ctx = arg;
372 	restart_instruction = 1;
373 
374 	error = vm_inject_exception(ctx, vcpu, vector, errcode_valid, errcode,
375 	    restart_instruction);
376 	assert(error == 0);
377 }
378 
379 void *
380 paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
381 {
382 
383 	return (vm_map_gpa(ctx, gaddr, len));
384 }
385 
386 int
387 fbsdrun_vmexit_on_pause(void)
388 {
389 
390 	return (guest_vmexit_on_pause);
391 }
392 
393 int
394 fbsdrun_vmexit_on_hlt(void)
395 {
396 
397 	return (guest_vmexit_on_hlt);
398 }
399 
400 int
401 fbsdrun_virtio_msix(void)
402 {
403 
404 	return (virtio_msix);
405 }
406 
407 static void *
408 fbsdrun_start_thread(void *param)
409 {
410 	char tname[MAXCOMLEN + 1];
411 	struct mt_vmm_info *mtp;
412 	int vcpu;
413 
414 	mtp = param;
415 	vcpu = mtp->mt_vcpu;
416 
417 	snprintf(tname, sizeof(tname), "vcpu %d", vcpu);
418 	pthread_set_name_np(mtp->mt_thr, tname);
419 
420 	if (gdb_port != 0)
421 		gdb_cpu_add(vcpu);
422 
423 	vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
424 
425 	/* not reached */
426 	exit(1);
427 	return (NULL);
428 }
429 
430 void
431 fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip)
432 {
433 	int error;
434 
435 	assert(fromcpu == BSP);
436 
437 	/*
438 	 * The 'newcpu' must be activated in the context of 'fromcpu'. If
439 	 * vm_activate_cpu() is delayed until newcpu's pthread starts running
440 	 * then vmm.ko is out-of-sync with bhyve and this can create a race
441 	 * with vm_suspend().
442 	 */
443 	error = vm_activate_cpu(ctx, newcpu);
444 	if (error != 0)
445 		err(EX_OSERR, "could not activate CPU %d", newcpu);
446 
447 	CPU_SET_ATOMIC(newcpu, &cpumask);
448 
449 	/*
450 	 * Set up the vmexit struct to allow execution to start
451 	 * at the given RIP
452 	 */
453 	vmexit[newcpu].rip = rip;
454 	vmexit[newcpu].inst_length = 0;
455 
456 	mt_vmm_info[newcpu].mt_ctx = ctx;
457 	mt_vmm_info[newcpu].mt_vcpu = newcpu;
458 
459 	error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL,
460 	    fbsdrun_start_thread, &mt_vmm_info[newcpu]);
461 	assert(error == 0);
462 }
463 
464 static int
465 fbsdrun_deletecpu(struct vmctx *ctx, int vcpu)
466 {
467 
468 	if (!CPU_ISSET(vcpu, &cpumask)) {
469 		fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu);
470 		exit(4);
471 	}
472 
473 	CPU_CLR_ATOMIC(vcpu, &cpumask);
474 	return (CPU_EMPTY(&cpumask));
475 }
476 
477 static int
478 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
479 		     uint32_t eax)
480 {
481 #if BHYVE_DEBUG
482 	/*
483 	 * put guest-driven debug here
484 	 */
485 #endif
486 	return (VMEXIT_CONTINUE);
487 }
488 
489 static int
490 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
491 {
492 	int error;
493 	int bytes, port, in, out;
494 	int vcpu;
495 
496 	vcpu = *pvcpu;
497 
498 	port = vme->u.inout.port;
499 	bytes = vme->u.inout.bytes;
500 	in = vme->u.inout.in;
501 	out = !in;
502 
503         /* Extra-special case of host notifications */
504         if (out && port == GUEST_NIO_PORT) {
505                 error = vmexit_handle_notify(ctx, vme, pvcpu, vme->u.inout.eax);
506 		return (error);
507 	}
508 
509 	error = emulate_inout(ctx, vcpu, vme, strictio);
510 	if (error) {
511 		fprintf(stderr, "Unhandled %s%c 0x%04x at 0x%lx\n",
512 		    in ? "in" : "out",
513 		    bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'),
514 		    port, vmexit->rip);
515 		return (VMEXIT_ABORT);
516 	} else {
517 		return (VMEXIT_CONTINUE);
518 	}
519 }
520 
521 static int
522 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
523 {
524 	uint64_t val;
525 	uint32_t eax, edx;
526 	int error;
527 
528 	val = 0;
529 	error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val);
530 	if (error != 0) {
531 		fprintf(stderr, "rdmsr to register %#x on vcpu %d\n",
532 		    vme->u.msr.code, *pvcpu);
533 		if (strictmsr) {
534 			vm_inject_gp(ctx, *pvcpu);
535 			return (VMEXIT_CONTINUE);
536 		}
537 	}
538 
539 	eax = val;
540 	error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax);
541 	assert(error == 0);
542 
543 	edx = val >> 32;
544 	error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx);
545 	assert(error == 0);
546 
547 	return (VMEXIT_CONTINUE);
548 }
549 
550 static int
551 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
552 {
553 	int error;
554 
555 	error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval);
556 	if (error != 0) {
557 		fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n",
558 		    vme->u.msr.code, vme->u.msr.wval, *pvcpu);
559 		if (strictmsr) {
560 			vm_inject_gp(ctx, *pvcpu);
561 			return (VMEXIT_CONTINUE);
562 		}
563 	}
564 	return (VMEXIT_CONTINUE);
565 }
566 
567 static int
568 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
569 {
570 
571 	(void)spinup_ap(ctx, *pvcpu,
572 		    vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
573 
574 	return (VMEXIT_CONTINUE);
575 }
576 
577 #define	DEBUG_EPT_MISCONFIG
578 #ifdef DEBUG_EPT_MISCONFIG
579 #define	VMCS_GUEST_PHYSICAL_ADDRESS	0x00002400
580 
581 static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4];
582 static int ept_misconfig_ptenum;
583 #endif
584 
585 static const char *
586 vmexit_vmx_desc(uint32_t exit_reason)
587 {
588 
589 	if (exit_reason >= nitems(vmx_exit_reason_desc) ||
590 	    vmx_exit_reason_desc[exit_reason] == NULL)
591 		return ("Unknown");
592 	return (vmx_exit_reason_desc[exit_reason]);
593 }
594 
595 static int
596 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
597 {
598 
599 	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
600 	fprintf(stderr, "\treason\t\tVMX\n");
601 	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
602 	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
603 	fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status);
604 	fprintf(stderr, "\texit_reason\t%u (%s)\n", vmexit->u.vmx.exit_reason,
605 	    vmexit_vmx_desc(vmexit->u.vmx.exit_reason));
606 	fprintf(stderr, "\tqualification\t0x%016lx\n",
607 	    vmexit->u.vmx.exit_qualification);
608 	fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type);
609 	fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error);
610 #ifdef DEBUG_EPT_MISCONFIG
611 	if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) {
612 		vm_get_register(ctx, *pvcpu,
613 		    VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS),
614 		    &ept_misconfig_gpa);
615 		vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte,
616 		    &ept_misconfig_ptenum);
617 		fprintf(stderr, "\tEPT misconfiguration:\n");
618 		fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa);
619 		fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n",
620 		    ept_misconfig_ptenum, ept_misconfig_pte[0],
621 		    ept_misconfig_pte[1], ept_misconfig_pte[2],
622 		    ept_misconfig_pte[3]);
623 	}
624 #endif	/* DEBUG_EPT_MISCONFIG */
625 	return (VMEXIT_ABORT);
626 }
627 
628 static int
629 vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
630 {
631 
632 	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
633 	fprintf(stderr, "\treason\t\tSVM\n");
634 	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
635 	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
636 	fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode);
637 	fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1);
638 	fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2);
639 	return (VMEXIT_ABORT);
640 }
641 
642 static int
643 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
644 {
645 
646 	assert(vmexit->inst_length == 0);
647 
648 	stats.vmexit_bogus++;
649 
650 	return (VMEXIT_CONTINUE);
651 }
652 
653 static int
654 vmexit_reqidle(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
655 {
656 
657 	assert(vmexit->inst_length == 0);
658 
659 	stats.vmexit_reqidle++;
660 
661 	return (VMEXIT_CONTINUE);
662 }
663 
664 static int
665 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
666 {
667 
668 	stats.vmexit_hlt++;
669 
670 	/*
671 	 * Just continue execution with the next instruction. We use
672 	 * the HLT VM exit as a way to be friendly with the host
673 	 * scheduler.
674 	 */
675 	return (VMEXIT_CONTINUE);
676 }
677 
678 static int
679 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
680 {
681 
682 	stats.vmexit_pause++;
683 
684 	return (VMEXIT_CONTINUE);
685 }
686 
687 static int
688 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
689 {
690 
691 	assert(vmexit->inst_length == 0);
692 
693 	stats.vmexit_mtrap++;
694 
695 	if (gdb_port == 0) {
696 		fprintf(stderr, "vm_loop: unexpected VMEXIT_MTRAP\n");
697 		exit(4);
698 	}
699 	gdb_cpu_mtrap(*pvcpu);
700 	return (VMEXIT_CONTINUE);
701 }
702 
703 static int
704 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
705 {
706 	int err, i;
707 	struct vie *vie;
708 
709 	stats.vmexit_inst_emul++;
710 
711 	vie = &vmexit->u.inst_emul.vie;
712 	err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
713 	    vie, &vmexit->u.inst_emul.paging);
714 
715 	if (err) {
716 		if (err == ESRCH) {
717 			fprintf(stderr, "Unhandled memory access to 0x%lx\n",
718 			    vmexit->u.inst_emul.gpa);
719 		}
720 
721 		fprintf(stderr, "Failed to emulate instruction [");
722 		for (i = 0; i < vie->num_valid; i++) {
723 			fprintf(stderr, "0x%02x%s", vie->inst[i],
724 			    i != (vie->num_valid - 1) ? " " : "");
725 		}
726 		fprintf(stderr, "] at 0x%lx\n", vmexit->rip);
727 		return (VMEXIT_ABORT);
728 	}
729 
730 	return (VMEXIT_CONTINUE);
731 }
732 
733 static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER;
734 static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER;
735 
736 static int
737 vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
738 {
739 	enum vm_suspend_how how;
740 
741 	how = vmexit->u.suspended.how;
742 
743 	fbsdrun_deletecpu(ctx, *pvcpu);
744 
745 	if (*pvcpu != BSP) {
746 		pthread_mutex_lock(&resetcpu_mtx);
747 		pthread_cond_signal(&resetcpu_cond);
748 		pthread_mutex_unlock(&resetcpu_mtx);
749 		pthread_exit(NULL);
750 	}
751 
752 	pthread_mutex_lock(&resetcpu_mtx);
753 	while (!CPU_EMPTY(&cpumask)) {
754 		pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx);
755 	}
756 	pthread_mutex_unlock(&resetcpu_mtx);
757 
758 	switch (how) {
759 	case VM_SUSPEND_RESET:
760 		exit(0);
761 	case VM_SUSPEND_POWEROFF:
762 		exit(1);
763 	case VM_SUSPEND_HALT:
764 		exit(2);
765 	case VM_SUSPEND_TRIPLEFAULT:
766 		exit(3);
767 	default:
768 		fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how);
769 		exit(100);
770 	}
771 	return (0);	/* NOTREACHED */
772 }
773 
774 static int
775 vmexit_debug(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
776 {
777 
778 	if (gdb_port == 0) {
779 		fprintf(stderr, "vm_loop: unexpected VMEXIT_DEBUG\n");
780 		exit(4);
781 	}
782 	gdb_cpu_suspend(*pvcpu);
783 	return (VMEXIT_CONTINUE);
784 }
785 
786 static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
787 	[VM_EXITCODE_INOUT]  = vmexit_inout,
788 	[VM_EXITCODE_INOUT_STR]  = vmexit_inout,
789 	[VM_EXITCODE_VMX]    = vmexit_vmx,
790 	[VM_EXITCODE_SVM]    = vmexit_svm,
791 	[VM_EXITCODE_BOGUS]  = vmexit_bogus,
792 	[VM_EXITCODE_REQIDLE] = vmexit_reqidle,
793 	[VM_EXITCODE_RDMSR]  = vmexit_rdmsr,
794 	[VM_EXITCODE_WRMSR]  = vmexit_wrmsr,
795 	[VM_EXITCODE_MTRAP]  = vmexit_mtrap,
796 	[VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
797 	[VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
798 	[VM_EXITCODE_SUSPENDED] = vmexit_suspend,
799 	[VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch,
800 	[VM_EXITCODE_DEBUG] = vmexit_debug,
801 };
802 
803 static void
804 vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip)
805 {
806 	int error, rc;
807 	enum vm_exitcode exitcode;
808 	cpuset_t active_cpus;
809 
810 	if (vcpumap[vcpu] != NULL) {
811 		error = pthread_setaffinity_np(pthread_self(),
812 		    sizeof(cpuset_t), vcpumap[vcpu]);
813 		assert(error == 0);
814 	}
815 
816 	error = vm_active_cpus(ctx, &active_cpus);
817 	assert(CPU_ISSET(vcpu, &active_cpus));
818 
819 	error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip);
820 	assert(error == 0);
821 
822 	while (1) {
823 		error = vm_run(ctx, vcpu, &vmexit[vcpu]);
824 		if (error != 0)
825 			break;
826 
827 		exitcode = vmexit[vcpu].exitcode;
828 		if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
829 			fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
830 			    exitcode);
831 			exit(4);
832 		}
833 
834 		rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
835 
836 		switch (rc) {
837 		case VMEXIT_CONTINUE:
838 			break;
839 		case VMEXIT_ABORT:
840 			abort();
841 		default:
842 			exit(4);
843 		}
844 	}
845 	fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
846 }
847 
848 static int
849 num_vcpus_allowed(struct vmctx *ctx)
850 {
851 	int tmp, error;
852 
853 	error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
854 
855 	/*
856 	 * The guest is allowed to spinup more than one processor only if the
857 	 * UNRESTRICTED_GUEST capability is available.
858 	 */
859 	if (error == 0)
860 		return (VM_MAXCPU);
861 	else
862 		return (1);
863 }
864 
865 void
866 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
867 {
868 	int err, tmp;
869 
870 	if (fbsdrun_vmexit_on_hlt()) {
871 		err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
872 		if (err < 0) {
873 			fprintf(stderr, "VM exit on HLT not supported\n");
874 			exit(4);
875 		}
876 		vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
877 		if (cpu == BSP)
878 			handler[VM_EXITCODE_HLT] = vmexit_hlt;
879 	}
880 
881         if (fbsdrun_vmexit_on_pause()) {
882 		/*
883 		 * pause exit support required for this mode
884 		 */
885 		err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
886 		if (err < 0) {
887 			fprintf(stderr,
888 			    "SMP mux requested, no pause support\n");
889 			exit(4);
890 		}
891 		vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
892 		if (cpu == BSP)
893 			handler[VM_EXITCODE_PAUSE] = vmexit_pause;
894         }
895 
896 	if (x2apic_mode)
897 		err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
898 	else
899 		err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
900 
901 	if (err) {
902 		fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
903 		exit(4);
904 	}
905 
906 	vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
907 }
908 
909 static struct vmctx *
910 do_open(const char *vmname)
911 {
912 	struct vmctx *ctx;
913 	int error;
914 	bool reinit, romboot;
915 #ifndef WITHOUT_CAPSICUM
916 	cap_rights_t rights;
917 	const cap_ioctl_t *cmds;
918 	size_t ncmds;
919 #endif
920 
921 	reinit = romboot = false;
922 
923 	if (lpc_bootrom())
924 		romboot = true;
925 
926 	error = vm_create(vmname);
927 	if (error) {
928 		if (errno == EEXIST) {
929 			if (romboot) {
930 				reinit = true;
931 			} else {
932 				/*
933 				 * The virtual machine has been setup by the
934 				 * userspace bootloader.
935 				 */
936 			}
937 		} else {
938 			perror("vm_create");
939 			exit(4);
940 		}
941 	} else {
942 		if (!romboot) {
943 			/*
944 			 * If the virtual machine was just created then a
945 			 * bootrom must be configured to boot it.
946 			 */
947 			fprintf(stderr, "virtual machine cannot be booted\n");
948 			exit(4);
949 		}
950 	}
951 
952 	ctx = vm_open(vmname);
953 	if (ctx == NULL) {
954 		perror("vm_open");
955 		exit(4);
956 	}
957 
958 #ifndef WITHOUT_CAPSICUM
959 	cap_rights_init(&rights, CAP_IOCTL, CAP_MMAP_RW);
960 	if (caph_rights_limit(vm_get_device_fd(ctx), &rights) == -1)
961 		errx(EX_OSERR, "Unable to apply rights for sandbox");
962 	vm_get_ioctls(&ncmds);
963 	cmds = vm_get_ioctls(NULL);
964 	if (cmds == NULL)
965 		errx(EX_OSERR, "out of memory");
966 	if (caph_ioctls_limit(vm_get_device_fd(ctx), cmds, ncmds) == -1)
967 		errx(EX_OSERR, "Unable to apply rights for sandbox");
968 	free((cap_ioctl_t *)cmds);
969 #endif
970 
971 	if (reinit) {
972 		error = vm_reinit(ctx);
973 		if (error) {
974 			perror("vm_reinit");
975 			exit(4);
976 		}
977 	}
978 	error = vm_set_topology(ctx, sockets, cores, threads, maxcpus);
979 	if (error)
980 		errx(EX_OSERR, "vm_set_topology");
981 	return (ctx);
982 }
983 
984 int
985 main(int argc, char *argv[])
986 {
987 	int c, error, dbg_port, err, bvmcons;
988 	int max_vcpus, mptgen, memflags;
989 	int rtc_localtime;
990 	bool gdb_stop;
991 	struct vmctx *ctx;
992 	uint64_t rip;
993 	size_t memsize;
994 	char *optstr;
995 
996 	bvmcons = 0;
997 	progname = basename(argv[0]);
998 	dbg_port = 0;
999 	gdb_stop = false;
1000 	guest_ncpus = 1;
1001 	sockets = cores = threads = 1;
1002 	maxcpus = 0;
1003 	memsize = 256 * MB;
1004 	mptgen = 1;
1005 	rtc_localtime = 1;
1006 	memflags = 0;
1007 
1008 	optstr = "abehuwxACHIPSWYp:g:G:c:s:m:l:U:";
1009 	while ((c = getopt(argc, argv, optstr)) != -1) {
1010 		switch (c) {
1011 		case 'a':
1012 			x2apic_mode = 0;
1013 			break;
1014 		case 'A':
1015 			acpi = 1;
1016 			break;
1017 		case 'b':
1018 			bvmcons = 1;
1019 			break;
1020 		case 'p':
1021                         if (pincpu_parse(optarg) != 0) {
1022                             errx(EX_USAGE, "invalid vcpu pinning "
1023                                  "configuration '%s'", optarg);
1024                         }
1025 			break;
1026                 case 'c':
1027 			if (topology_parse(optarg) != 0) {
1028 			    errx(EX_USAGE, "invalid cpu topology "
1029 				"'%s'", optarg);
1030 			}
1031 			break;
1032 		case 'C':
1033 			memflags |= VM_MEM_F_INCORE;
1034 			break;
1035 		case 'g':
1036 			dbg_port = atoi(optarg);
1037 			break;
1038 		case 'G':
1039 			if (optarg[0] == 'w') {
1040 				gdb_stop = true;
1041 				optarg++;
1042 			}
1043 			gdb_port = atoi(optarg);
1044 			break;
1045 		case 'l':
1046 			if (strncmp(optarg, "help", strlen(optarg)) == 0) {
1047 				lpc_print_supported_devices();
1048 				exit(0);
1049 			} else if (lpc_device_parse(optarg) != 0) {
1050 				errx(EX_USAGE, "invalid lpc device "
1051 				    "configuration '%s'", optarg);
1052 			}
1053 			break;
1054 		case 's':
1055 			if (strncmp(optarg, "help", strlen(optarg)) == 0) {
1056 				pci_print_supported_devices();
1057 				exit(0);
1058 			} else if (pci_parse_slot(optarg) != 0)
1059 				exit(4);
1060 			else
1061 				break;
1062 		case 'S':
1063 			memflags |= VM_MEM_F_WIRED;
1064 			break;
1065                 case 'm':
1066 			error = vm_parse_memsize(optarg, &memsize);
1067 			if (error)
1068 				errx(EX_USAGE, "invalid memsize '%s'", optarg);
1069 			break;
1070 		case 'H':
1071 			guest_vmexit_on_hlt = 1;
1072 			break;
1073 		case 'I':
1074 			/*
1075 			 * The "-I" option was used to add an ioapic to the
1076 			 * virtual machine.
1077 			 *
1078 			 * An ioapic is now provided unconditionally for each
1079 			 * virtual machine and this option is now deprecated.
1080 			 */
1081 			break;
1082 		case 'P':
1083 			guest_vmexit_on_pause = 1;
1084 			break;
1085 		case 'e':
1086 			strictio = 1;
1087 			break;
1088 		case 'u':
1089 			rtc_localtime = 0;
1090 			break;
1091 		case 'U':
1092 			guest_uuid_str = optarg;
1093 			break;
1094 		case 'w':
1095 			strictmsr = 0;
1096 			break;
1097 		case 'W':
1098 			virtio_msix = 0;
1099 			break;
1100 		case 'x':
1101 			x2apic_mode = 1;
1102 			break;
1103 		case 'Y':
1104 			mptgen = 0;
1105 			break;
1106 		case 'h':
1107 			usage(0);
1108 		default:
1109 			usage(1);
1110 		}
1111 	}
1112 	argc -= optind;
1113 	argv += optind;
1114 
1115 	if (argc != 1)
1116 		usage(1);
1117 
1118 	vmname = argv[0];
1119 	ctx = do_open(vmname);
1120 
1121 	max_vcpus = num_vcpus_allowed(ctx);
1122 	if (guest_ncpus > max_vcpus) {
1123 		fprintf(stderr, "%d vCPUs requested but only %d available\n",
1124 			guest_ncpus, max_vcpus);
1125 		exit(4);
1126 	}
1127 
1128 	fbsdrun_set_capabilities(ctx, BSP);
1129 
1130 	vm_set_memflags(ctx, memflags);
1131 	err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
1132 	if (err) {
1133 		fprintf(stderr, "Unable to setup memory (%d)\n", errno);
1134 		exit(4);
1135 	}
1136 
1137 	error = init_msr();
1138 	if (error) {
1139 		fprintf(stderr, "init_msr error %d", error);
1140 		exit(4);
1141 	}
1142 
1143 	init_mem();
1144 	init_inout();
1145 	atkbdc_init(ctx);
1146 	pci_irq_init(ctx);
1147 	ioapic_init(ctx);
1148 
1149 	rtc_init(ctx, rtc_localtime);
1150 	sci_init(ctx);
1151 
1152 	/*
1153 	 * Exit if a device emulation finds an error in its initilization
1154 	 */
1155 	if (init_pci(ctx) != 0) {
1156 		perror("device emulation initialization error");
1157 		exit(4);
1158 	}
1159 
1160 	if (dbg_port != 0)
1161 		init_dbgport(dbg_port);
1162 
1163 	if (gdb_port != 0)
1164 		init_gdb(ctx, gdb_port, gdb_stop);
1165 
1166 	if (bvmcons)
1167 		init_bvmcons();
1168 
1169 	if (lpc_bootrom()) {
1170 		if (vm_set_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, 1)) {
1171 			fprintf(stderr, "ROM boot failed: unrestricted guest "
1172 			    "capability not available\n");
1173 			exit(4);
1174 		}
1175 		error = vcpu_reset(ctx, BSP);
1176 		assert(error == 0);
1177 	}
1178 
1179 	error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
1180 	assert(error == 0);
1181 
1182 	/*
1183 	 * build the guest tables, MP etc.
1184 	 */
1185 	if (mptgen) {
1186 		error = mptable_build(ctx, guest_ncpus);
1187 		if (error) {
1188 			perror("error to build the guest tables");
1189 			exit(4);
1190 		}
1191 	}
1192 
1193 	error = smbios_build(ctx);
1194 	assert(error == 0);
1195 
1196 	if (acpi) {
1197 		error = acpi_build(ctx, guest_ncpus);
1198 		assert(error == 0);
1199 	}
1200 
1201 	if (lpc_bootrom())
1202 		fwctl_init();
1203 
1204 	/*
1205 	 * Change the proc title to include the VM name.
1206 	 */
1207 	setproctitle("%s", vmname);
1208 
1209 #ifndef WITHOUT_CAPSICUM
1210 	caph_cache_catpages();
1211 
1212 	if (caph_limit_stdout() == -1 || caph_limit_stderr() == -1)
1213 		errx(EX_OSERR, "Unable to apply rights for sandbox");
1214 
1215 	if (caph_enter() == -1)
1216 		errx(EX_OSERR, "cap_enter() failed");
1217 #endif
1218 
1219 	/*
1220 	 * Add CPU 0
1221 	 */
1222 	fbsdrun_addcpu(ctx, BSP, BSP, rip);
1223 
1224 	/*
1225 	 * Head off to the main event dispatch loop
1226 	 */
1227 	mevent_dispatch();
1228 
1229 	exit(4);
1230 }
1231