xref: /freebsd/usr.sbin/bhyve/bhyverun.c (revision b24e6ac8b78ebf628872c821b7eb4f7ed008498c)
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 "bootrom.h"
73 #include "inout.h"
74 #include "dbgport.h"
75 #include "debug.h"
76 #include "fwctl.h"
77 #include "gdb.h"
78 #include "ioapic.h"
79 #include "mem.h"
80 #include "mevent.h"
81 #include "mptbl.h"
82 #include "pci_emul.h"
83 #include "pci_irq.h"
84 #include "pci_lpc.h"
85 #include "smbiostbl.h"
86 #include "xmsr.h"
87 #include "spinup_ap.h"
88 #include "rtc.h"
89 #include "vmgenc.h"
90 
91 #define GUEST_NIO_PORT		0x488	/* guest upcalls via i/o port */
92 
93 #define MB		(1024UL * 1024)
94 #define GB		(1024UL * MB)
95 
96 static const char * const vmx_exit_reason_desc[] = {
97 	[EXIT_REASON_EXCEPTION] = "Exception or non-maskable interrupt (NMI)",
98 	[EXIT_REASON_EXT_INTR] = "External interrupt",
99 	[EXIT_REASON_TRIPLE_FAULT] = "Triple fault",
100 	[EXIT_REASON_INIT] = "INIT signal",
101 	[EXIT_REASON_SIPI] = "Start-up IPI (SIPI)",
102 	[EXIT_REASON_IO_SMI] = "I/O system-management interrupt (SMI)",
103 	[EXIT_REASON_SMI] = "Other SMI",
104 	[EXIT_REASON_INTR_WINDOW] = "Interrupt window",
105 	[EXIT_REASON_NMI_WINDOW] = "NMI window",
106 	[EXIT_REASON_TASK_SWITCH] = "Task switch",
107 	[EXIT_REASON_CPUID] = "CPUID",
108 	[EXIT_REASON_GETSEC] = "GETSEC",
109 	[EXIT_REASON_HLT] = "HLT",
110 	[EXIT_REASON_INVD] = "INVD",
111 	[EXIT_REASON_INVLPG] = "INVLPG",
112 	[EXIT_REASON_RDPMC] = "RDPMC",
113 	[EXIT_REASON_RDTSC] = "RDTSC",
114 	[EXIT_REASON_RSM] = "RSM",
115 	[EXIT_REASON_VMCALL] = "VMCALL",
116 	[EXIT_REASON_VMCLEAR] = "VMCLEAR",
117 	[EXIT_REASON_VMLAUNCH] = "VMLAUNCH",
118 	[EXIT_REASON_VMPTRLD] = "VMPTRLD",
119 	[EXIT_REASON_VMPTRST] = "VMPTRST",
120 	[EXIT_REASON_VMREAD] = "VMREAD",
121 	[EXIT_REASON_VMRESUME] = "VMRESUME",
122 	[EXIT_REASON_VMWRITE] = "VMWRITE",
123 	[EXIT_REASON_VMXOFF] = "VMXOFF",
124 	[EXIT_REASON_VMXON] = "VMXON",
125 	[EXIT_REASON_CR_ACCESS] = "Control-register accesses",
126 	[EXIT_REASON_DR_ACCESS] = "MOV DR",
127 	[EXIT_REASON_INOUT] = "I/O instruction",
128 	[EXIT_REASON_RDMSR] = "RDMSR",
129 	[EXIT_REASON_WRMSR] = "WRMSR",
130 	[EXIT_REASON_INVAL_VMCS] =
131 	    "VM-entry failure due to invalid guest state",
132 	[EXIT_REASON_INVAL_MSR] = "VM-entry failure due to MSR loading",
133 	[EXIT_REASON_MWAIT] = "MWAIT",
134 	[EXIT_REASON_MTF] = "Monitor trap flag",
135 	[EXIT_REASON_MONITOR] = "MONITOR",
136 	[EXIT_REASON_PAUSE] = "PAUSE",
137 	[EXIT_REASON_MCE_DURING_ENTRY] =
138 	    "VM-entry failure due to machine-check event",
139 	[EXIT_REASON_TPR] = "TPR below threshold",
140 	[EXIT_REASON_APIC_ACCESS] = "APIC access",
141 	[EXIT_REASON_VIRTUALIZED_EOI] = "Virtualized EOI",
142 	[EXIT_REASON_GDTR_IDTR] = "Access to GDTR or IDTR",
143 	[EXIT_REASON_LDTR_TR] = "Access to LDTR or TR",
144 	[EXIT_REASON_EPT_FAULT] = "EPT violation",
145 	[EXIT_REASON_EPT_MISCONFIG] = "EPT misconfiguration",
146 	[EXIT_REASON_INVEPT] = "INVEPT",
147 	[EXIT_REASON_RDTSCP] = "RDTSCP",
148 	[EXIT_REASON_VMX_PREEMPT] = "VMX-preemption timer expired",
149 	[EXIT_REASON_INVVPID] = "INVVPID",
150 	[EXIT_REASON_WBINVD] = "WBINVD",
151 	[EXIT_REASON_XSETBV] = "XSETBV",
152 	[EXIT_REASON_APIC_WRITE] = "APIC write",
153 	[EXIT_REASON_RDRAND] = "RDRAND",
154 	[EXIT_REASON_INVPCID] = "INVPCID",
155 	[EXIT_REASON_VMFUNC] = "VMFUNC",
156 	[EXIT_REASON_ENCLS] = "ENCLS",
157 	[EXIT_REASON_RDSEED] = "RDSEED",
158 	[EXIT_REASON_PM_LOG_FULL] = "Page-modification log full",
159 	[EXIT_REASON_XSAVES] = "XSAVES",
160 	[EXIT_REASON_XRSTORS] = "XRSTORS"
161 };
162 
163 typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
164 extern int vmexit_task_switch(struct vmctx *, struct vm_exit *, int *vcpu);
165 
166 char *vmname;
167 
168 int guest_ncpus;
169 uint16_t cores, maxcpus, sockets, threads;
170 
171 char *guest_uuid_str;
172 
173 int raw_stdio = 0;
174 
175 static int gdb_port = 0;
176 static int guest_vmexit_on_hlt, guest_vmexit_on_pause;
177 static int virtio_msix = 1;
178 static int x2apic_mode = 0;	/* default is xAPIC */
179 
180 static int strictio;
181 static int strictmsr = 1;
182 
183 static int acpi;
184 
185 static char *progname;
186 static const int BSP = 0;
187 
188 static cpuset_t cpumask;
189 
190 static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
191 
192 static struct vm_exit vmexit[VM_MAXCPU];
193 
194 struct bhyvestats {
195 	uint64_t	vmexit_bogus;
196 	uint64_t	vmexit_reqidle;
197 	uint64_t	vmexit_hlt;
198 	uint64_t	vmexit_pause;
199 	uint64_t	vmexit_mtrap;
200 	uint64_t	vmexit_inst_emul;
201 	uint64_t	cpu_switch_rotate;
202 	uint64_t	cpu_switch_direct;
203 } stats;
204 
205 struct mt_vmm_info {
206 	pthread_t	mt_thr;
207 	struct vmctx	*mt_ctx;
208 	int		mt_vcpu;
209 } mt_vmm_info[VM_MAXCPU];
210 
211 static cpuset_t *vcpumap[VM_MAXCPU] = { NULL };
212 
213 static void
214 usage(int code)
215 {
216 
217         fprintf(stderr,
218 		"Usage: %s [-abehuwxACHPSWY]\n"
219 		"       %*s [-c [[cpus=]numcpus][,sockets=n][,cores=n][,threads=n]]\n"
220 		"       %*s [-g <gdb port>] [-l <lpc>]\n"
221 		"       %*s [-m mem] [-p vcpu:hostcpu] [-s <pci>] [-U uuid] <vm>\n"
222 		"       -a: local apic is in xAPIC mode (deprecated)\n"
223 		"       -A: create ACPI tables\n"
224 		"       -c: number of cpus and/or topology specification\n"
225 		"       -C: include guest memory in core file\n"
226 		"       -e: exit on unhandled I/O access\n"
227 		"       -g: gdb port\n"
228 		"       -h: help\n"
229 		"       -H: vmexit from the guest on hlt\n"
230 		"       -l: LPC device configuration\n"
231 		"       -m: memory size in MB\n"
232 		"       -p: pin 'vcpu' to 'hostcpu'\n"
233 		"       -P: vmexit from the guest on pause\n"
234 		"       -s: <slot,driver,configinfo> PCI slot config\n"
235 		"       -S: guest memory cannot be swapped\n"
236 		"       -u: RTC keeps UTC time\n"
237 		"       -U: uuid\n"
238 		"       -w: ignore unimplemented MSRs\n"
239 		"       -W: force virtio to use single-vector MSI\n"
240 		"       -x: local apic is in x2APIC mode\n"
241 		"       -Y: disable MPtable generation\n",
242 		progname, (int)strlen(progname), "", (int)strlen(progname), "",
243 		(int)strlen(progname), "");
244 
245 	exit(code);
246 }
247 
248 /*
249  * XXX This parser is known to have the following issues:
250  * 1.  It accepts null key=value tokens ",,".
251  * 2.  It accepts whitespace after = and before value.
252  * 3.  Values out of range of INT are silently wrapped.
253  * 4.  It doesn't check non-final values.
254  * 5.  The apparently bogus limits of UINT16_MAX are for future expansion.
255  *
256  * The acceptance of a null specification ('-c ""') is by design to match the
257  * manual page syntax specification, this results in a topology of 1 vCPU.
258  */
259 static int
260 topology_parse(const char *opt)
261 {
262 	uint64_t ncpus;
263 	int c, chk, n, s, t, tmp;
264 	char *cp, *str;
265 	bool ns, scts;
266 
267 	c = 1, n = 1, s = 1, t = 1;
268 	ns = false, scts = false;
269 	str = strdup(opt);
270 	if (str == NULL)
271 		goto out;
272 
273 	while ((cp = strsep(&str, ",")) != NULL) {
274 		if (sscanf(cp, "%i%n", &tmp, &chk) == 1) {
275 			n = tmp;
276 			ns = true;
277 		} else if (sscanf(cp, "cpus=%i%n", &tmp, &chk) == 1) {
278 			n = tmp;
279 			ns = true;
280 		} else if (sscanf(cp, "sockets=%i%n", &tmp, &chk) == 1) {
281 			s = tmp;
282 			scts = true;
283 		} else if (sscanf(cp, "cores=%i%n", &tmp, &chk) == 1) {
284 			c = tmp;
285 			scts = true;
286 		} else if (sscanf(cp, "threads=%i%n", &tmp, &chk) == 1) {
287 			t = tmp;
288 			scts = true;
289 #ifdef notyet  /* Do not expose this until vmm.ko implements it */
290 		} else if (sscanf(cp, "maxcpus=%i%n", &tmp, &chk) == 1) {
291 			m = tmp;
292 #endif
293 		/* Skip the empty argument case from -c "" */
294 		} else if (cp[0] == '\0')
295 			continue;
296 		else
297 			goto out;
298 		/* Any trailing garbage causes an error */
299 		if (cp[chk] != '\0')
300 			goto out;
301 	}
302 	free(str);
303 	str = NULL;
304 
305 	/*
306 	 * Range check 1 <= n <= UINT16_MAX all values
307 	 */
308 	if (n < 1 || s < 1 || c < 1 || t < 1 ||
309 	    n > UINT16_MAX || s > UINT16_MAX || c > UINT16_MAX  ||
310 	    t > UINT16_MAX)
311 		return (-1);
312 
313 	/* If only the cpus was specified, use that as sockets */
314 	if (!scts)
315 		s = n;
316 	/*
317 	 * Compute sockets * cores * threads avoiding overflow
318 	 * The range check above insures these are 16 bit values
319 	 * If n was specified check it against computed ncpus
320 	 */
321 	ncpus = (uint64_t)s * c * t;
322 	if (ncpus > UINT16_MAX || (ns && n != ncpus))
323 		return (-1);
324 
325 	guest_ncpus = ncpus;
326 	sockets = s;
327 	cores = c;
328 	threads = t;
329 	return(0);
330 
331 out:
332 	free(str);
333 	return (-1);
334 }
335 
336 static int
337 pincpu_parse(const char *opt)
338 {
339 	int vcpu, pcpu;
340 
341 	if (sscanf(opt, "%d:%d", &vcpu, &pcpu) != 2) {
342 		fprintf(stderr, "invalid format: %s\n", opt);
343 		return (-1);
344 	}
345 
346 	if (vcpu < 0 || vcpu >= VM_MAXCPU) {
347 		fprintf(stderr, "vcpu '%d' outside valid range from 0 to %d\n",
348 		    vcpu, VM_MAXCPU - 1);
349 		return (-1);
350 	}
351 
352 	if (pcpu < 0 || pcpu >= CPU_SETSIZE) {
353 		fprintf(stderr, "hostcpu '%d' outside valid range from "
354 		    "0 to %d\n", pcpu, CPU_SETSIZE - 1);
355 		return (-1);
356 	}
357 
358 	if (vcpumap[vcpu] == NULL) {
359 		if ((vcpumap[vcpu] = malloc(sizeof(cpuset_t))) == NULL) {
360 			perror("malloc");
361 			return (-1);
362 		}
363 		CPU_ZERO(vcpumap[vcpu]);
364 	}
365 	CPU_SET(pcpu, vcpumap[vcpu]);
366 	return (0);
367 }
368 
369 void
370 vm_inject_fault(void *arg, int vcpu, int vector, int errcode_valid,
371     int errcode)
372 {
373 	struct vmctx *ctx;
374 	int error, restart_instruction;
375 
376 	ctx = arg;
377 	restart_instruction = 1;
378 
379 	error = vm_inject_exception(ctx, vcpu, vector, errcode_valid, errcode,
380 	    restart_instruction);
381 	assert(error == 0);
382 }
383 
384 void *
385 paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
386 {
387 
388 	return (vm_map_gpa(ctx, gaddr, len));
389 }
390 
391 int
392 fbsdrun_vmexit_on_pause(void)
393 {
394 
395 	return (guest_vmexit_on_pause);
396 }
397 
398 int
399 fbsdrun_vmexit_on_hlt(void)
400 {
401 
402 	return (guest_vmexit_on_hlt);
403 }
404 
405 int
406 fbsdrun_virtio_msix(void)
407 {
408 
409 	return (virtio_msix);
410 }
411 
412 static void *
413 fbsdrun_start_thread(void *param)
414 {
415 	char tname[MAXCOMLEN + 1];
416 	struct mt_vmm_info *mtp;
417 	int vcpu;
418 
419 	mtp = param;
420 	vcpu = mtp->mt_vcpu;
421 
422 	snprintf(tname, sizeof(tname), "vcpu %d", vcpu);
423 	pthread_set_name_np(mtp->mt_thr, tname);
424 
425 	if (gdb_port != 0)
426 		gdb_cpu_add(vcpu);
427 
428 	vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
429 
430 	/* not reached */
431 	exit(1);
432 	return (NULL);
433 }
434 
435 void
436 fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip)
437 {
438 	int error;
439 
440 	assert(fromcpu == BSP);
441 
442 	/*
443 	 * The 'newcpu' must be activated in the context of 'fromcpu'. If
444 	 * vm_activate_cpu() is delayed until newcpu's pthread starts running
445 	 * then vmm.ko is out-of-sync with bhyve and this can create a race
446 	 * with vm_suspend().
447 	 */
448 	error = vm_activate_cpu(ctx, newcpu);
449 	if (error != 0)
450 		err(EX_OSERR, "could not activate CPU %d", newcpu);
451 
452 	CPU_SET_ATOMIC(newcpu, &cpumask);
453 
454 	/*
455 	 * Set up the vmexit struct to allow execution to start
456 	 * at the given RIP
457 	 */
458 	vmexit[newcpu].rip = rip;
459 	vmexit[newcpu].inst_length = 0;
460 
461 	mt_vmm_info[newcpu].mt_ctx = ctx;
462 	mt_vmm_info[newcpu].mt_vcpu = newcpu;
463 
464 	error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL,
465 	    fbsdrun_start_thread, &mt_vmm_info[newcpu]);
466 	assert(error == 0);
467 }
468 
469 static int
470 fbsdrun_deletecpu(struct vmctx *ctx, int vcpu)
471 {
472 
473 	if (!CPU_ISSET(vcpu, &cpumask)) {
474 		fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu);
475 		exit(4);
476 	}
477 
478 	CPU_CLR_ATOMIC(vcpu, &cpumask);
479 	return (CPU_EMPTY(&cpumask));
480 }
481 
482 static int
483 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
484 		     uint32_t eax)
485 {
486 #if BHYVE_DEBUG
487 	/*
488 	 * put guest-driven debug here
489 	 */
490 #endif
491 	return (VMEXIT_CONTINUE);
492 }
493 
494 static int
495 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
496 {
497 	int error;
498 	int bytes, port, in, out;
499 	int vcpu;
500 
501 	vcpu = *pvcpu;
502 
503 	port = vme->u.inout.port;
504 	bytes = vme->u.inout.bytes;
505 	in = vme->u.inout.in;
506 	out = !in;
507 
508         /* Extra-special case of host notifications */
509         if (out && port == GUEST_NIO_PORT) {
510                 error = vmexit_handle_notify(ctx, vme, pvcpu, vme->u.inout.eax);
511 		return (error);
512 	}
513 
514 	error = emulate_inout(ctx, vcpu, vme, strictio);
515 	if (error) {
516 		fprintf(stderr, "Unhandled %s%c 0x%04x at 0x%lx\n",
517 		    in ? "in" : "out",
518 		    bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'),
519 		    port, vmexit->rip);
520 		return (VMEXIT_ABORT);
521 	} else {
522 		return (VMEXIT_CONTINUE);
523 	}
524 }
525 
526 static int
527 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
528 {
529 	uint64_t val;
530 	uint32_t eax, edx;
531 	int error;
532 
533 	val = 0;
534 	error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val);
535 	if (error != 0) {
536 		fprintf(stderr, "rdmsr to register %#x on vcpu %d\n",
537 		    vme->u.msr.code, *pvcpu);
538 		if (strictmsr) {
539 			vm_inject_gp(ctx, *pvcpu);
540 			return (VMEXIT_CONTINUE);
541 		}
542 	}
543 
544 	eax = val;
545 	error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax);
546 	assert(error == 0);
547 
548 	edx = val >> 32;
549 	error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx);
550 	assert(error == 0);
551 
552 	return (VMEXIT_CONTINUE);
553 }
554 
555 static int
556 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
557 {
558 	int error;
559 
560 	error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval);
561 	if (error != 0) {
562 		fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n",
563 		    vme->u.msr.code, vme->u.msr.wval, *pvcpu);
564 		if (strictmsr) {
565 			vm_inject_gp(ctx, *pvcpu);
566 			return (VMEXIT_CONTINUE);
567 		}
568 	}
569 	return (VMEXIT_CONTINUE);
570 }
571 
572 static int
573 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
574 {
575 
576 	(void)spinup_ap(ctx, *pvcpu,
577 		    vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
578 
579 	return (VMEXIT_CONTINUE);
580 }
581 
582 #define	DEBUG_EPT_MISCONFIG
583 #ifdef DEBUG_EPT_MISCONFIG
584 #define	VMCS_GUEST_PHYSICAL_ADDRESS	0x00002400
585 
586 static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4];
587 static int ept_misconfig_ptenum;
588 #endif
589 
590 static const char *
591 vmexit_vmx_desc(uint32_t exit_reason)
592 {
593 
594 	if (exit_reason >= nitems(vmx_exit_reason_desc) ||
595 	    vmx_exit_reason_desc[exit_reason] == NULL)
596 		return ("Unknown");
597 	return (vmx_exit_reason_desc[exit_reason]);
598 }
599 
600 static int
601 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
602 {
603 
604 	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
605 	fprintf(stderr, "\treason\t\tVMX\n");
606 	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
607 	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
608 	fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status);
609 	fprintf(stderr, "\texit_reason\t%u (%s)\n", vmexit->u.vmx.exit_reason,
610 	    vmexit_vmx_desc(vmexit->u.vmx.exit_reason));
611 	fprintf(stderr, "\tqualification\t0x%016lx\n",
612 	    vmexit->u.vmx.exit_qualification);
613 	fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type);
614 	fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error);
615 #ifdef DEBUG_EPT_MISCONFIG
616 	if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) {
617 		vm_get_register(ctx, *pvcpu,
618 		    VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS),
619 		    &ept_misconfig_gpa);
620 		vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte,
621 		    &ept_misconfig_ptenum);
622 		fprintf(stderr, "\tEPT misconfiguration:\n");
623 		fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa);
624 		fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n",
625 		    ept_misconfig_ptenum, ept_misconfig_pte[0],
626 		    ept_misconfig_pte[1], ept_misconfig_pte[2],
627 		    ept_misconfig_pte[3]);
628 	}
629 #endif	/* DEBUG_EPT_MISCONFIG */
630 	return (VMEXIT_ABORT);
631 }
632 
633 static int
634 vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
635 {
636 
637 	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
638 	fprintf(stderr, "\treason\t\tSVM\n");
639 	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
640 	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
641 	fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode);
642 	fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1);
643 	fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2);
644 	return (VMEXIT_ABORT);
645 }
646 
647 static int
648 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
649 {
650 
651 	assert(vmexit->inst_length == 0);
652 
653 	stats.vmexit_bogus++;
654 
655 	return (VMEXIT_CONTINUE);
656 }
657 
658 static int
659 vmexit_reqidle(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
660 {
661 
662 	assert(vmexit->inst_length == 0);
663 
664 	stats.vmexit_reqidle++;
665 
666 	return (VMEXIT_CONTINUE);
667 }
668 
669 static int
670 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
671 {
672 
673 	stats.vmexit_hlt++;
674 
675 	/*
676 	 * Just continue execution with the next instruction. We use
677 	 * the HLT VM exit as a way to be friendly with the host
678 	 * scheduler.
679 	 */
680 	return (VMEXIT_CONTINUE);
681 }
682 
683 static int
684 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
685 {
686 
687 	stats.vmexit_pause++;
688 
689 	return (VMEXIT_CONTINUE);
690 }
691 
692 static int
693 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
694 {
695 
696 	assert(vmexit->inst_length == 0);
697 
698 	stats.vmexit_mtrap++;
699 
700 	if (gdb_port == 0) {
701 		fprintf(stderr, "vm_loop: unexpected VMEXIT_MTRAP\n");
702 		exit(4);
703 	}
704 	gdb_cpu_mtrap(*pvcpu);
705 	return (VMEXIT_CONTINUE);
706 }
707 
708 static int
709 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
710 {
711 	int err, i;
712 	struct vie *vie;
713 
714 	stats.vmexit_inst_emul++;
715 
716 	vie = &vmexit->u.inst_emul.vie;
717 	err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
718 	    vie, &vmexit->u.inst_emul.paging);
719 
720 	if (err) {
721 		if (err == ESRCH) {
722 			EPRINTLN("Unhandled memory access to 0x%lx\n",
723 			    vmexit->u.inst_emul.gpa);
724 		}
725 
726 		fprintf(stderr, "Failed to emulate instruction sequence [ ");
727 		for (i = 0; i < vie->num_valid; i++)
728 			fprintf(stderr, "%02x", vie->inst[i]);
729 		FPRINTLN(stderr, " ] at 0x%lx", vmexit->rip);
730 		return (VMEXIT_ABORT);
731 	}
732 
733 	return (VMEXIT_CONTINUE);
734 }
735 
736 static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER;
737 static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER;
738 
739 static int
740 vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
741 {
742 	enum vm_suspend_how how;
743 
744 	how = vmexit->u.suspended.how;
745 
746 	fbsdrun_deletecpu(ctx, *pvcpu);
747 
748 	if (*pvcpu != BSP) {
749 		pthread_mutex_lock(&resetcpu_mtx);
750 		pthread_cond_signal(&resetcpu_cond);
751 		pthread_mutex_unlock(&resetcpu_mtx);
752 		pthread_exit(NULL);
753 	}
754 
755 	pthread_mutex_lock(&resetcpu_mtx);
756 	while (!CPU_EMPTY(&cpumask)) {
757 		pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx);
758 	}
759 	pthread_mutex_unlock(&resetcpu_mtx);
760 
761 	switch (how) {
762 	case VM_SUSPEND_RESET:
763 		exit(0);
764 	case VM_SUSPEND_POWEROFF:
765 		exit(1);
766 	case VM_SUSPEND_HALT:
767 		exit(2);
768 	case VM_SUSPEND_TRIPLEFAULT:
769 		exit(3);
770 	default:
771 		fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how);
772 		exit(100);
773 	}
774 	return (0);	/* NOTREACHED */
775 }
776 
777 static int
778 vmexit_debug(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
779 {
780 
781 	if (gdb_port == 0) {
782 		fprintf(stderr, "vm_loop: unexpected VMEXIT_DEBUG\n");
783 		exit(4);
784 	}
785 	gdb_cpu_suspend(*pvcpu);
786 	return (VMEXIT_CONTINUE);
787 }
788 
789 static int
790 vmexit_breakpoint(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
791 {
792 
793 	if (gdb_port == 0) {
794 		fprintf(stderr, "vm_loop: unexpected VMEXIT_DEBUG\n");
795 		exit(4);
796 	}
797 	gdb_cpu_breakpoint(*pvcpu, vmexit);
798 	return (VMEXIT_CONTINUE);
799 }
800 
801 static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
802 	[VM_EXITCODE_INOUT]  = vmexit_inout,
803 	[VM_EXITCODE_INOUT_STR]  = vmexit_inout,
804 	[VM_EXITCODE_VMX]    = vmexit_vmx,
805 	[VM_EXITCODE_SVM]    = vmexit_svm,
806 	[VM_EXITCODE_BOGUS]  = vmexit_bogus,
807 	[VM_EXITCODE_REQIDLE] = vmexit_reqidle,
808 	[VM_EXITCODE_RDMSR]  = vmexit_rdmsr,
809 	[VM_EXITCODE_WRMSR]  = vmexit_wrmsr,
810 	[VM_EXITCODE_MTRAP]  = vmexit_mtrap,
811 	[VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
812 	[VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
813 	[VM_EXITCODE_SUSPENDED] = vmexit_suspend,
814 	[VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch,
815 	[VM_EXITCODE_DEBUG] = vmexit_debug,
816 	[VM_EXITCODE_BPT] = vmexit_breakpoint,
817 };
818 
819 static void
820 vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip)
821 {
822 	int error, rc;
823 	enum vm_exitcode exitcode;
824 	cpuset_t active_cpus;
825 
826 	if (vcpumap[vcpu] != NULL) {
827 		error = pthread_setaffinity_np(pthread_self(),
828 		    sizeof(cpuset_t), vcpumap[vcpu]);
829 		assert(error == 0);
830 	}
831 
832 	error = vm_active_cpus(ctx, &active_cpus);
833 	assert(CPU_ISSET(vcpu, &active_cpus));
834 
835 	error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip);
836 	assert(error == 0);
837 
838 	while (1) {
839 		error = vm_run(ctx, vcpu, &vmexit[vcpu]);
840 		if (error != 0)
841 			break;
842 
843 		exitcode = vmexit[vcpu].exitcode;
844 		if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
845 			fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
846 			    exitcode);
847 			exit(4);
848 		}
849 
850 		rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
851 
852 		switch (rc) {
853 		case VMEXIT_CONTINUE:
854 			break;
855 		case VMEXIT_ABORT:
856 			abort();
857 		default:
858 			exit(4);
859 		}
860 	}
861 	fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
862 }
863 
864 static int
865 num_vcpus_allowed(struct vmctx *ctx)
866 {
867 	int tmp, error;
868 
869 	error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
870 
871 	/*
872 	 * The guest is allowed to spinup more than one processor only if the
873 	 * UNRESTRICTED_GUEST capability is available.
874 	 */
875 	if (error == 0)
876 		return (VM_MAXCPU);
877 	else
878 		return (1);
879 }
880 
881 void
882 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
883 {
884 	int err, tmp;
885 
886 	if (fbsdrun_vmexit_on_hlt()) {
887 		err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
888 		if (err < 0) {
889 			fprintf(stderr, "VM exit on HLT not supported\n");
890 			exit(4);
891 		}
892 		vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
893 		if (cpu == BSP)
894 			handler[VM_EXITCODE_HLT] = vmexit_hlt;
895 	}
896 
897         if (fbsdrun_vmexit_on_pause()) {
898 		/*
899 		 * pause exit support required for this mode
900 		 */
901 		err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
902 		if (err < 0) {
903 			fprintf(stderr,
904 			    "SMP mux requested, no pause support\n");
905 			exit(4);
906 		}
907 		vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
908 		if (cpu == BSP)
909 			handler[VM_EXITCODE_PAUSE] = vmexit_pause;
910         }
911 
912 	if (x2apic_mode)
913 		err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
914 	else
915 		err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
916 
917 	if (err) {
918 		fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
919 		exit(4);
920 	}
921 
922 	vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
923 }
924 
925 static struct vmctx *
926 do_open(const char *vmname)
927 {
928 	struct vmctx *ctx;
929 	int error;
930 	bool reinit, romboot;
931 #ifndef WITHOUT_CAPSICUM
932 	cap_rights_t rights;
933 	const cap_ioctl_t *cmds;
934 	size_t ncmds;
935 #endif
936 
937 	reinit = romboot = false;
938 
939 	if (lpc_bootrom())
940 		romboot = true;
941 
942 	error = vm_create(vmname);
943 	if (error) {
944 		if (errno == EEXIST) {
945 			if (romboot) {
946 				reinit = true;
947 			} else {
948 				/*
949 				 * The virtual machine has been setup by the
950 				 * userspace bootloader.
951 				 */
952 			}
953 		} else {
954 			perror("vm_create");
955 			exit(4);
956 		}
957 	} else {
958 		if (!romboot) {
959 			/*
960 			 * If the virtual machine was just created then a
961 			 * bootrom must be configured to boot it.
962 			 */
963 			fprintf(stderr, "virtual machine cannot be booted\n");
964 			exit(4);
965 		}
966 	}
967 
968 	ctx = vm_open(vmname);
969 	if (ctx == NULL) {
970 		perror("vm_open");
971 		exit(4);
972 	}
973 
974 #ifndef WITHOUT_CAPSICUM
975 	cap_rights_init(&rights, CAP_IOCTL, CAP_MMAP_RW);
976 	if (caph_rights_limit(vm_get_device_fd(ctx), &rights) == -1)
977 		errx(EX_OSERR, "Unable to apply rights for sandbox");
978 	vm_get_ioctls(&ncmds);
979 	cmds = vm_get_ioctls(NULL);
980 	if (cmds == NULL)
981 		errx(EX_OSERR, "out of memory");
982 	if (caph_ioctls_limit(vm_get_device_fd(ctx), cmds, ncmds) == -1)
983 		errx(EX_OSERR, "Unable to apply rights for sandbox");
984 	free((cap_ioctl_t *)cmds);
985 #endif
986 
987 	if (reinit) {
988 		error = vm_reinit(ctx);
989 		if (error) {
990 			perror("vm_reinit");
991 			exit(4);
992 		}
993 	}
994 	error = vm_set_topology(ctx, sockets, cores, threads, maxcpus);
995 	if (error)
996 		errx(EX_OSERR, "vm_set_topology");
997 	return (ctx);
998 }
999 
1000 int
1001 main(int argc, char *argv[])
1002 {
1003 	int c, error, dbg_port, err, bvmcons;
1004 	int max_vcpus, mptgen, memflags;
1005 	int rtc_localtime;
1006 	bool gdb_stop;
1007 	struct vmctx *ctx;
1008 	uint64_t rip;
1009 	size_t memsize;
1010 	char *optstr;
1011 
1012 	bvmcons = 0;
1013 	progname = basename(argv[0]);
1014 	dbg_port = 0;
1015 	gdb_stop = false;
1016 	guest_ncpus = 1;
1017 	sockets = cores = threads = 1;
1018 	maxcpus = 0;
1019 	memsize = 256 * MB;
1020 	mptgen = 1;
1021 	rtc_localtime = 1;
1022 	memflags = 0;
1023 
1024 	optstr = "abehuwxACHIPSWYp:g:G:c:s:m:l:U:";
1025 	while ((c = getopt(argc, argv, optstr)) != -1) {
1026 		switch (c) {
1027 		case 'a':
1028 			x2apic_mode = 0;
1029 			break;
1030 		case 'A':
1031 			acpi = 1;
1032 			break;
1033 		case 'b':
1034 			bvmcons = 1;
1035 			break;
1036 		case 'p':
1037                         if (pincpu_parse(optarg) != 0) {
1038                             errx(EX_USAGE, "invalid vcpu pinning "
1039                                  "configuration '%s'", optarg);
1040                         }
1041 			break;
1042                 case 'c':
1043 			if (topology_parse(optarg) != 0) {
1044 			    errx(EX_USAGE, "invalid cpu topology "
1045 				"'%s'", optarg);
1046 			}
1047 			break;
1048 		case 'C':
1049 			memflags |= VM_MEM_F_INCORE;
1050 			break;
1051 		case 'g':
1052 			dbg_port = atoi(optarg);
1053 			break;
1054 		case 'G':
1055 			if (optarg[0] == 'w') {
1056 				gdb_stop = true;
1057 				optarg++;
1058 			}
1059 			gdb_port = atoi(optarg);
1060 			break;
1061 		case 'l':
1062 			if (strncmp(optarg, "help", strlen(optarg)) == 0) {
1063 				lpc_print_supported_devices();
1064 				exit(0);
1065 			} else if (lpc_device_parse(optarg) != 0) {
1066 				errx(EX_USAGE, "invalid lpc device "
1067 				    "configuration '%s'", optarg);
1068 			}
1069 			break;
1070 		case 's':
1071 			if (strncmp(optarg, "help", strlen(optarg)) == 0) {
1072 				pci_print_supported_devices();
1073 				exit(0);
1074 			} else if (pci_parse_slot(optarg) != 0)
1075 				exit(4);
1076 			else
1077 				break;
1078 		case 'S':
1079 			memflags |= VM_MEM_F_WIRED;
1080 			break;
1081                 case 'm':
1082 			error = vm_parse_memsize(optarg, &memsize);
1083 			if (error)
1084 				errx(EX_USAGE, "invalid memsize '%s'", optarg);
1085 			break;
1086 		case 'H':
1087 			guest_vmexit_on_hlt = 1;
1088 			break;
1089 		case 'I':
1090 			/*
1091 			 * The "-I" option was used to add an ioapic to the
1092 			 * virtual machine.
1093 			 *
1094 			 * An ioapic is now provided unconditionally for each
1095 			 * virtual machine and this option is now deprecated.
1096 			 */
1097 			break;
1098 		case 'P':
1099 			guest_vmexit_on_pause = 1;
1100 			break;
1101 		case 'e':
1102 			strictio = 1;
1103 			break;
1104 		case 'u':
1105 			rtc_localtime = 0;
1106 			break;
1107 		case 'U':
1108 			guest_uuid_str = optarg;
1109 			break;
1110 		case 'w':
1111 			strictmsr = 0;
1112 			break;
1113 		case 'W':
1114 			virtio_msix = 0;
1115 			break;
1116 		case 'x':
1117 			x2apic_mode = 1;
1118 			break;
1119 		case 'Y':
1120 			mptgen = 0;
1121 			break;
1122 		case 'h':
1123 			usage(0);
1124 		default:
1125 			usage(1);
1126 		}
1127 	}
1128 	argc -= optind;
1129 	argv += optind;
1130 
1131 	if (argc != 1)
1132 		usage(1);
1133 
1134 	vmname = argv[0];
1135 	ctx = do_open(vmname);
1136 
1137 	max_vcpus = num_vcpus_allowed(ctx);
1138 	if (guest_ncpus > max_vcpus) {
1139 		fprintf(stderr, "%d vCPUs requested but only %d available\n",
1140 			guest_ncpus, max_vcpus);
1141 		exit(4);
1142 	}
1143 
1144 	fbsdrun_set_capabilities(ctx, BSP);
1145 
1146 	vm_set_memflags(ctx, memflags);
1147 	err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
1148 	if (err) {
1149 		fprintf(stderr, "Unable to setup memory (%d)\n", errno);
1150 		exit(4);
1151 	}
1152 
1153 	error = init_msr();
1154 	if (error) {
1155 		fprintf(stderr, "init_msr error %d", error);
1156 		exit(4);
1157 	}
1158 
1159 	init_mem();
1160 	init_inout();
1161 	init_bootrom(ctx);
1162 	atkbdc_init(ctx);
1163 	pci_irq_init(ctx);
1164 	ioapic_init(ctx);
1165 
1166 	rtc_init(ctx, rtc_localtime);
1167 	sci_init(ctx);
1168 
1169 	/*
1170 	 * Exit if a device emulation finds an error in its initilization
1171 	 */
1172 	if (init_pci(ctx) != 0) {
1173 		perror("device emulation initialization error");
1174 		exit(4);
1175 	}
1176 
1177 	/*
1178 	 * Initialize after PCI, to allow a bootrom file to reserve the high
1179 	 * region.
1180 	 */
1181 	if (acpi)
1182 		vmgenc_init(ctx);
1183 
1184 	if (dbg_port != 0)
1185 		init_dbgport(dbg_port);
1186 
1187 	if (gdb_port != 0)
1188 		init_gdb(ctx, gdb_port, gdb_stop);
1189 
1190 	if (bvmcons)
1191 		init_bvmcons();
1192 
1193 	if (lpc_bootrom()) {
1194 		if (vm_set_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, 1)) {
1195 			fprintf(stderr, "ROM boot failed: unrestricted guest "
1196 			    "capability not available\n");
1197 			exit(4);
1198 		}
1199 		error = vcpu_reset(ctx, BSP);
1200 		assert(error == 0);
1201 	}
1202 
1203 	error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
1204 	assert(error == 0);
1205 
1206 	/*
1207 	 * build the guest tables, MP etc.
1208 	 */
1209 	if (mptgen) {
1210 		error = mptable_build(ctx, guest_ncpus);
1211 		if (error) {
1212 			perror("error to build the guest tables");
1213 			exit(4);
1214 		}
1215 	}
1216 
1217 	error = smbios_build(ctx);
1218 	assert(error == 0);
1219 
1220 	if (acpi) {
1221 		error = acpi_build(ctx, guest_ncpus);
1222 		assert(error == 0);
1223 	}
1224 
1225 	if (lpc_bootrom())
1226 		fwctl_init();
1227 
1228 	/*
1229 	 * Change the proc title to include the VM name.
1230 	 */
1231 	setproctitle("%s", vmname);
1232 
1233 #ifndef WITHOUT_CAPSICUM
1234 	caph_cache_catpages();
1235 
1236 	if (caph_limit_stdout() == -1 || caph_limit_stderr() == -1)
1237 		errx(EX_OSERR, "Unable to apply rights for sandbox");
1238 
1239 	if (caph_enter() == -1)
1240 		errx(EX_OSERR, "cap_enter() failed");
1241 #endif
1242 
1243 	/*
1244 	 * Add CPU 0
1245 	 */
1246 	fbsdrun_addcpu(ctx, BSP, BSP, rip);
1247 
1248 	/*
1249 	 * Head off to the main event dispatch loop
1250 	 */
1251 	mevent_dispatch();
1252 
1253 	exit(4);
1254 }
1255