1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause 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 #ifdef BHYVE_SNAPSHOT 40 #include <sys/socket.h> 41 #include <sys/stat.h> 42 #endif 43 #include <sys/time.h> 44 #ifdef BHYVE_SNAPSHOT 45 #include <sys/un.h> 46 #endif 47 48 #include <amd64/vmm/intel/vmcs.h> 49 #include <x86/apicreg.h> 50 51 #include <machine/atomic.h> 52 #include <machine/segments.h> 53 54 #ifndef WITHOUT_CAPSICUM 55 #include <capsicum_helpers.h> 56 #endif 57 #include <stdio.h> 58 #include <stdlib.h> 59 #include <string.h> 60 #include <err.h> 61 #include <errno.h> 62 #ifdef BHYVE_SNAPSHOT 63 #include <fcntl.h> 64 #endif 65 #include <libgen.h> 66 #include <unistd.h> 67 #include <assert.h> 68 #include <pthread.h> 69 #include <pthread_np.h> 70 #include <sysexits.h> 71 #include <stdbool.h> 72 #include <stdint.h> 73 #ifdef BHYVE_SNAPSHOT 74 #include <ucl.h> 75 #include <unistd.h> 76 77 #include <libxo/xo.h> 78 #endif 79 80 #include <machine/vmm.h> 81 #ifndef WITHOUT_CAPSICUM 82 #include <machine/vmm_dev.h> 83 #endif 84 #include <machine/vmm_instruction_emul.h> 85 #include <vmmapi.h> 86 87 #include "bhyverun.h" 88 #include "acpi.h" 89 #include "atkbdc.h" 90 #include "bootrom.h" 91 #include "config.h" 92 #include "inout.h" 93 #include "debug.h" 94 #include "e820.h" 95 #include "fwctl.h" 96 #include "gdb.h" 97 #include "ioapic.h" 98 #include "kernemu_dev.h" 99 #include "mem.h" 100 #include "mevent.h" 101 #include "mptbl.h" 102 #include "pci_emul.h" 103 #include "pci_irq.h" 104 #include "pci_lpc.h" 105 #include "qemu_fwcfg.h" 106 #include "smbiostbl.h" 107 #ifdef BHYVE_SNAPSHOT 108 #include "snapshot.h" 109 #endif 110 #include "xmsr.h" 111 #include "spinup_ap.h" 112 #include "rtc.h" 113 #include "vmgenc.h" 114 115 #define GUEST_NIO_PORT 0x488 /* guest upcalls via i/o port */ 116 117 #define MB (1024UL * 1024) 118 #define GB (1024UL * MB) 119 120 static const char * const vmx_exit_reason_desc[] = { 121 [EXIT_REASON_EXCEPTION] = "Exception or non-maskable interrupt (NMI)", 122 [EXIT_REASON_EXT_INTR] = "External interrupt", 123 [EXIT_REASON_TRIPLE_FAULT] = "Triple fault", 124 [EXIT_REASON_INIT] = "INIT signal", 125 [EXIT_REASON_SIPI] = "Start-up IPI (SIPI)", 126 [EXIT_REASON_IO_SMI] = "I/O system-management interrupt (SMI)", 127 [EXIT_REASON_SMI] = "Other SMI", 128 [EXIT_REASON_INTR_WINDOW] = "Interrupt window", 129 [EXIT_REASON_NMI_WINDOW] = "NMI window", 130 [EXIT_REASON_TASK_SWITCH] = "Task switch", 131 [EXIT_REASON_CPUID] = "CPUID", 132 [EXIT_REASON_GETSEC] = "GETSEC", 133 [EXIT_REASON_HLT] = "HLT", 134 [EXIT_REASON_INVD] = "INVD", 135 [EXIT_REASON_INVLPG] = "INVLPG", 136 [EXIT_REASON_RDPMC] = "RDPMC", 137 [EXIT_REASON_RDTSC] = "RDTSC", 138 [EXIT_REASON_RSM] = "RSM", 139 [EXIT_REASON_VMCALL] = "VMCALL", 140 [EXIT_REASON_VMCLEAR] = "VMCLEAR", 141 [EXIT_REASON_VMLAUNCH] = "VMLAUNCH", 142 [EXIT_REASON_VMPTRLD] = "VMPTRLD", 143 [EXIT_REASON_VMPTRST] = "VMPTRST", 144 [EXIT_REASON_VMREAD] = "VMREAD", 145 [EXIT_REASON_VMRESUME] = "VMRESUME", 146 [EXIT_REASON_VMWRITE] = "VMWRITE", 147 [EXIT_REASON_VMXOFF] = "VMXOFF", 148 [EXIT_REASON_VMXON] = "VMXON", 149 [EXIT_REASON_CR_ACCESS] = "Control-register accesses", 150 [EXIT_REASON_DR_ACCESS] = "MOV DR", 151 [EXIT_REASON_INOUT] = "I/O instruction", 152 [EXIT_REASON_RDMSR] = "RDMSR", 153 [EXIT_REASON_WRMSR] = "WRMSR", 154 [EXIT_REASON_INVAL_VMCS] = 155 "VM-entry failure due to invalid guest state", 156 [EXIT_REASON_INVAL_MSR] = "VM-entry failure due to MSR loading", 157 [EXIT_REASON_MWAIT] = "MWAIT", 158 [EXIT_REASON_MTF] = "Monitor trap flag", 159 [EXIT_REASON_MONITOR] = "MONITOR", 160 [EXIT_REASON_PAUSE] = "PAUSE", 161 [EXIT_REASON_MCE_DURING_ENTRY] = 162 "VM-entry failure due to machine-check event", 163 [EXIT_REASON_TPR] = "TPR below threshold", 164 [EXIT_REASON_APIC_ACCESS] = "APIC access", 165 [EXIT_REASON_VIRTUALIZED_EOI] = "Virtualized EOI", 166 [EXIT_REASON_GDTR_IDTR] = "Access to GDTR or IDTR", 167 [EXIT_REASON_LDTR_TR] = "Access to LDTR or TR", 168 [EXIT_REASON_EPT_FAULT] = "EPT violation", 169 [EXIT_REASON_EPT_MISCONFIG] = "EPT misconfiguration", 170 [EXIT_REASON_INVEPT] = "INVEPT", 171 [EXIT_REASON_RDTSCP] = "RDTSCP", 172 [EXIT_REASON_VMX_PREEMPT] = "VMX-preemption timer expired", 173 [EXIT_REASON_INVVPID] = "INVVPID", 174 [EXIT_REASON_WBINVD] = "WBINVD", 175 [EXIT_REASON_XSETBV] = "XSETBV", 176 [EXIT_REASON_APIC_WRITE] = "APIC write", 177 [EXIT_REASON_RDRAND] = "RDRAND", 178 [EXIT_REASON_INVPCID] = "INVPCID", 179 [EXIT_REASON_VMFUNC] = "VMFUNC", 180 [EXIT_REASON_ENCLS] = "ENCLS", 181 [EXIT_REASON_RDSEED] = "RDSEED", 182 [EXIT_REASON_PM_LOG_FULL] = "Page-modification log full", 183 [EXIT_REASON_XSAVES] = "XSAVES", 184 [EXIT_REASON_XRSTORS] = "XRSTORS" 185 }; 186 187 typedef int (*vmexit_handler_t)(struct vmctx *, struct vcpu *, struct vm_exit *); 188 189 int guest_ncpus; 190 uint16_t cpu_cores, cpu_sockets, cpu_threads; 191 192 int raw_stdio = 0; 193 194 static char *progname; 195 static const int BSP = 0; 196 197 static cpuset_t cpumask; 198 199 static void vm_loop(struct vmctx *ctx, struct vcpu *vcpu); 200 201 static struct bhyvestats { 202 uint64_t vmexit_bogus; 203 uint64_t vmexit_reqidle; 204 uint64_t vmexit_hlt; 205 uint64_t vmexit_pause; 206 uint64_t vmexit_mtrap; 207 uint64_t vmexit_inst_emul; 208 uint64_t cpu_switch_rotate; 209 uint64_t cpu_switch_direct; 210 } stats; 211 212 static struct vcpu_info { 213 struct vmctx *ctx; 214 struct vcpu *vcpu; 215 int vcpuid; 216 } *vcpu_info; 217 218 static cpuset_t **vcpumap; 219 220 static void 221 usage(int code) 222 { 223 224 fprintf(stderr, 225 "Usage: %s [-AaCDeHhPSuWwxY]\n" 226 " %*s [-c [[cpus=]numcpus][,sockets=n][,cores=n][,threads=n]]\n" 227 " %*s [-G port] [-k config_file] [-l lpc] [-m mem] [-o var=value]\n" 228 " %*s [-p vcpu:hostcpu] [-r file] [-s pci] [-U uuid] vmname\n" 229 " -A: create ACPI tables\n" 230 " -a: local apic is in xAPIC mode (deprecated)\n" 231 " -C: include guest memory in core file\n" 232 " -c: number of CPUs and/or topology specification\n" 233 " -D: destroy on power-off\n" 234 " -e: exit on unhandled I/O access\n" 235 " -G: start a debug server\n" 236 " -H: vmexit from the guest on HLT\n" 237 " -h: help\n" 238 " -k: key=value flat config file\n" 239 " -K: PS2 keyboard layout\n" 240 " -l: LPC device configuration\n" 241 " -m: memory size\n" 242 " -o: set config 'var' to 'value'\n" 243 " -P: vmexit from the guest on pause\n" 244 " -p: pin 'vcpu' to 'hostcpu'\n" 245 #ifdef BHYVE_SNAPSHOT 246 " -r: path to checkpoint file\n" 247 #endif 248 " -S: guest memory cannot be swapped\n" 249 " -s: <slot,driver,configinfo> PCI slot config\n" 250 " -U: UUID\n" 251 " -u: RTC keeps UTC time\n" 252 " -W: force virtio to use single-vector MSI\n" 253 " -w: ignore unimplemented MSRs\n" 254 " -x: local APIC is in x2APIC mode\n" 255 " -Y: disable MPtable generation\n", 256 progname, (int)strlen(progname), "", (int)strlen(progname), "", 257 (int)strlen(progname), ""); 258 259 exit(code); 260 } 261 262 /* 263 * XXX This parser is known to have the following issues: 264 * 1. It accepts null key=value tokens ",," as setting "cpus" to an 265 * empty string. 266 * 267 * The acceptance of a null specification ('-c ""') is by design to match the 268 * manual page syntax specification, this results in a topology of 1 vCPU. 269 */ 270 static int 271 topology_parse(const char *opt) 272 { 273 char *cp, *str, *tofree; 274 275 if (*opt == '\0') { 276 set_config_value("sockets", "1"); 277 set_config_value("cores", "1"); 278 set_config_value("threads", "1"); 279 set_config_value("cpus", "1"); 280 return (0); 281 } 282 283 tofree = str = strdup(opt); 284 if (str == NULL) 285 errx(4, "Failed to allocate memory"); 286 287 while ((cp = strsep(&str, ",")) != NULL) { 288 if (strncmp(cp, "cpus=", strlen("cpus=")) == 0) 289 set_config_value("cpus", cp + strlen("cpus=")); 290 else if (strncmp(cp, "sockets=", strlen("sockets=")) == 0) 291 set_config_value("sockets", cp + strlen("sockets=")); 292 else if (strncmp(cp, "cores=", strlen("cores=")) == 0) 293 set_config_value("cores", cp + strlen("cores=")); 294 else if (strncmp(cp, "threads=", strlen("threads=")) == 0) 295 set_config_value("threads", cp + strlen("threads=")); 296 #ifdef notyet /* Do not expose this until vmm.ko implements it */ 297 else if (strncmp(cp, "maxcpus=", strlen("maxcpus=")) == 0) 298 set_config_value("maxcpus", cp + strlen("maxcpus=")); 299 #endif 300 else if (strchr(cp, '=') != NULL) 301 goto out; 302 else 303 set_config_value("cpus", cp); 304 } 305 free(tofree); 306 return (0); 307 308 out: 309 free(tofree); 310 return (-1); 311 } 312 313 static int 314 parse_int_value(const char *key, const char *value, int minval, int maxval) 315 { 316 char *cp; 317 long lval; 318 319 errno = 0; 320 lval = strtol(value, &cp, 0); 321 if (errno != 0 || *cp != '\0' || cp == value || lval < minval || 322 lval > maxval) 323 errx(4, "Invalid value for %s: '%s'", key, value); 324 return (lval); 325 } 326 327 /* 328 * Set the sockets, cores, threads, and guest_cpus variables based on 329 * the configured topology. 330 * 331 * The limits of UINT16_MAX are due to the types passed to 332 * vm_set_topology(). vmm.ko may enforce tighter limits. 333 */ 334 static void 335 calc_topology(void) 336 { 337 const char *value; 338 bool explicit_cpus; 339 uint64_t ncpus; 340 341 value = get_config_value("cpus"); 342 if (value != NULL) { 343 guest_ncpus = parse_int_value("cpus", value, 1, UINT16_MAX); 344 explicit_cpus = true; 345 } else { 346 guest_ncpus = 1; 347 explicit_cpus = false; 348 } 349 value = get_config_value("cores"); 350 if (value != NULL) 351 cpu_cores = parse_int_value("cores", value, 1, UINT16_MAX); 352 else 353 cpu_cores = 1; 354 value = get_config_value("threads"); 355 if (value != NULL) 356 cpu_threads = parse_int_value("threads", value, 1, UINT16_MAX); 357 else 358 cpu_threads = 1; 359 value = get_config_value("sockets"); 360 if (value != NULL) 361 cpu_sockets = parse_int_value("sockets", value, 1, UINT16_MAX); 362 else 363 cpu_sockets = guest_ncpus; 364 365 /* 366 * Compute sockets * cores * threads avoiding overflow. The 367 * range check above insures these are 16 bit values. 368 */ 369 ncpus = (uint64_t)cpu_sockets * cpu_cores * cpu_threads; 370 if (ncpus > UINT16_MAX) 371 errx(4, "Computed number of vCPUs too high: %ju", 372 (uintmax_t)ncpus); 373 374 if (explicit_cpus) { 375 if (guest_ncpus != (int)ncpus) 376 errx(4, "Topology (%d sockets, %d cores, %d threads) " 377 "does not match %d vCPUs", 378 cpu_sockets, cpu_cores, cpu_threads, 379 guest_ncpus); 380 } else 381 guest_ncpus = ncpus; 382 } 383 384 static int 385 pincpu_parse(const char *opt) 386 { 387 const char *value; 388 char *newval; 389 char key[16]; 390 int vcpu, pcpu; 391 392 if (sscanf(opt, "%d:%d", &vcpu, &pcpu) != 2) { 393 fprintf(stderr, "invalid format: %s\n", opt); 394 return (-1); 395 } 396 397 if (vcpu < 0) { 398 fprintf(stderr, "invalid vcpu '%d'\n", vcpu); 399 return (-1); 400 } 401 402 if (pcpu < 0 || pcpu >= CPU_SETSIZE) { 403 fprintf(stderr, "hostcpu '%d' outside valid range from " 404 "0 to %d\n", pcpu, CPU_SETSIZE - 1); 405 return (-1); 406 } 407 408 snprintf(key, sizeof(key), "vcpu.%d.cpuset", vcpu); 409 value = get_config_value(key); 410 411 if (asprintf(&newval, "%s%s%d", value != NULL ? value : "", 412 value != NULL ? "," : "", pcpu) == -1) { 413 perror("failed to build new cpuset string"); 414 return (-1); 415 } 416 417 set_config_value(key, newval); 418 free(newval); 419 return (0); 420 } 421 422 static void 423 parse_cpuset(int vcpu, const char *list, cpuset_t *set) 424 { 425 char *cp, *token; 426 int pcpu, start; 427 428 CPU_ZERO(set); 429 start = -1; 430 token = __DECONST(char *, list); 431 for (;;) { 432 pcpu = strtoul(token, &cp, 0); 433 if (cp == token) 434 errx(4, "invalid cpuset for vcpu %d: '%s'", vcpu, list); 435 if (pcpu < 0 || pcpu >= CPU_SETSIZE) 436 errx(4, "hostcpu '%d' outside valid range from 0 to %d", 437 pcpu, CPU_SETSIZE - 1); 438 switch (*cp) { 439 case ',': 440 case '\0': 441 if (start >= 0) { 442 if (start > pcpu) 443 errx(4, "Invalid hostcpu range %d-%d", 444 start, pcpu); 445 while (start < pcpu) { 446 CPU_SET(start, set); 447 start++; 448 } 449 start = -1; 450 } 451 CPU_SET(pcpu, set); 452 break; 453 case '-': 454 if (start >= 0) 455 errx(4, "invalid cpuset for vcpu %d: '%s'", 456 vcpu, list); 457 start = pcpu; 458 break; 459 default: 460 errx(4, "invalid cpuset for vcpu %d: '%s'", vcpu, list); 461 } 462 if (*cp == '\0') 463 break; 464 token = cp + 1; 465 } 466 } 467 468 static void 469 build_vcpumaps(void) 470 { 471 char key[16]; 472 const char *value; 473 int vcpu; 474 475 vcpumap = calloc(guest_ncpus, sizeof(*vcpumap)); 476 for (vcpu = 0; vcpu < guest_ncpus; vcpu++) { 477 snprintf(key, sizeof(key), "vcpu.%d.cpuset", vcpu); 478 value = get_config_value(key); 479 if (value == NULL) 480 continue; 481 vcpumap[vcpu] = malloc(sizeof(cpuset_t)); 482 if (vcpumap[vcpu] == NULL) 483 err(4, "Failed to allocate cpuset for vcpu %d", vcpu); 484 parse_cpuset(vcpu, value, vcpumap[vcpu]); 485 } 486 } 487 488 void 489 vm_inject_fault(struct vcpu *vcpu, int vector, int errcode_valid, 490 int errcode) 491 { 492 int error, restart_instruction; 493 494 restart_instruction = 1; 495 496 error = vm_inject_exception(vcpu, vector, errcode_valid, errcode, 497 restart_instruction); 498 assert(error == 0); 499 } 500 501 void * 502 paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len) 503 { 504 505 return (vm_map_gpa(ctx, gaddr, len)); 506 } 507 508 #ifdef BHYVE_SNAPSHOT 509 uintptr_t 510 paddr_host2guest(struct vmctx *ctx, void *addr) 511 { 512 return (vm_rev_map_gpa(ctx, addr)); 513 } 514 #endif 515 516 int 517 fbsdrun_virtio_msix(void) 518 { 519 520 return (get_config_bool_default("virtio_msix", true)); 521 } 522 523 static void * 524 fbsdrun_start_thread(void *param) 525 { 526 char tname[MAXCOMLEN + 1]; 527 struct vcpu_info *vi = param; 528 int error; 529 530 snprintf(tname, sizeof(tname), "vcpu %d", vi->vcpuid); 531 pthread_set_name_np(pthread_self(), tname); 532 533 if (vcpumap[vi->vcpuid] != NULL) { 534 error = pthread_setaffinity_np(pthread_self(), 535 sizeof(cpuset_t), vcpumap[vi->vcpuid]); 536 assert(error == 0); 537 } 538 539 #ifdef BHYVE_SNAPSHOT 540 checkpoint_cpu_add(vi->vcpuid); 541 #endif 542 gdb_cpu_add(vi->vcpu); 543 544 vm_loop(vi->ctx, vi->vcpu); 545 546 /* not reached */ 547 exit(1); 548 return (NULL); 549 } 550 551 static void 552 fbsdrun_addcpu(struct vcpu_info *vi) 553 { 554 pthread_t thr; 555 int error; 556 557 error = vm_activate_cpu(vi->vcpu); 558 if (error != 0) 559 err(EX_OSERR, "could not activate CPU %d", vi->vcpuid); 560 561 CPU_SET_ATOMIC(vi->vcpuid, &cpumask); 562 563 vm_suspend_cpu(vi->vcpu); 564 565 error = pthread_create(&thr, NULL, fbsdrun_start_thread, vi); 566 assert(error == 0); 567 } 568 569 static int 570 fbsdrun_deletecpu(int vcpu) 571 { 572 573 if (!CPU_ISSET(vcpu, &cpumask)) { 574 fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu); 575 exit(4); 576 } 577 578 CPU_CLR_ATOMIC(vcpu, &cpumask); 579 return (CPU_EMPTY(&cpumask)); 580 } 581 582 static int 583 vmexit_handle_notify(struct vmctx *ctx __unused, struct vcpu *vcpu __unused, 584 struct vm_exit *vme __unused, uint32_t eax __unused) 585 { 586 #if BHYVE_DEBUG 587 /* 588 * put guest-driven debug here 589 */ 590 #endif 591 return (VMEXIT_CONTINUE); 592 } 593 594 static int 595 vmexit_inout(struct vmctx *ctx, struct vcpu *vcpu, struct vm_exit *vme) 596 { 597 int error; 598 int bytes, port, in, out; 599 600 port = vme->u.inout.port; 601 bytes = vme->u.inout.bytes; 602 in = vme->u.inout.in; 603 out = !in; 604 605 /* Extra-special case of host notifications */ 606 if (out && port == GUEST_NIO_PORT) { 607 error = vmexit_handle_notify(ctx, vcpu, vme, vme->u.inout.eax); 608 return (error); 609 } 610 611 error = emulate_inout(ctx, vcpu, vme); 612 if (error) { 613 fprintf(stderr, "Unhandled %s%c 0x%04x at 0x%lx\n", 614 in ? "in" : "out", 615 bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), 616 port, vme->rip); 617 return (VMEXIT_ABORT); 618 } else { 619 return (VMEXIT_CONTINUE); 620 } 621 } 622 623 static int 624 vmexit_rdmsr(struct vmctx *ctx __unused, struct vcpu *vcpu, struct vm_exit *vme) 625 { 626 uint64_t val; 627 uint32_t eax, edx; 628 int error; 629 630 val = 0; 631 error = emulate_rdmsr(vcpu, vme->u.msr.code, &val); 632 if (error != 0) { 633 fprintf(stderr, "rdmsr to register %#x on vcpu %d\n", 634 vme->u.msr.code, vcpu_id(vcpu)); 635 if (get_config_bool("x86.strictmsr")) { 636 vm_inject_gp(vcpu); 637 return (VMEXIT_CONTINUE); 638 } 639 } 640 641 eax = val; 642 error = vm_set_register(vcpu, VM_REG_GUEST_RAX, eax); 643 assert(error == 0); 644 645 edx = val >> 32; 646 error = vm_set_register(vcpu, VM_REG_GUEST_RDX, edx); 647 assert(error == 0); 648 649 return (VMEXIT_CONTINUE); 650 } 651 652 static int 653 vmexit_wrmsr(struct vmctx *ctx __unused, struct vcpu *vcpu, struct vm_exit *vme) 654 { 655 int error; 656 657 error = emulate_wrmsr(vcpu, vme->u.msr.code, vme->u.msr.wval); 658 if (error != 0) { 659 fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n", 660 vme->u.msr.code, vme->u.msr.wval, vcpu_id(vcpu)); 661 if (get_config_bool("x86.strictmsr")) { 662 vm_inject_gp(vcpu); 663 return (VMEXIT_CONTINUE); 664 } 665 } 666 return (VMEXIT_CONTINUE); 667 } 668 669 #define DEBUG_EPT_MISCONFIG 670 #ifdef DEBUG_EPT_MISCONFIG 671 #define VMCS_GUEST_PHYSICAL_ADDRESS 0x00002400 672 673 static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4]; 674 static int ept_misconfig_ptenum; 675 #endif 676 677 static const char * 678 vmexit_vmx_desc(uint32_t exit_reason) 679 { 680 681 if (exit_reason >= nitems(vmx_exit_reason_desc) || 682 vmx_exit_reason_desc[exit_reason] == NULL) 683 return ("Unknown"); 684 return (vmx_exit_reason_desc[exit_reason]); 685 } 686 687 static int 688 vmexit_vmx(struct vmctx *ctx, struct vcpu *vcpu, struct vm_exit *vme) 689 { 690 691 fprintf(stderr, "vm exit[%d]\n", vcpu_id(vcpu)); 692 fprintf(stderr, "\treason\t\tVMX\n"); 693 fprintf(stderr, "\trip\t\t0x%016lx\n", vme->rip); 694 fprintf(stderr, "\tinst_length\t%d\n", vme->inst_length); 695 fprintf(stderr, "\tstatus\t\t%d\n", vme->u.vmx.status); 696 fprintf(stderr, "\texit_reason\t%u (%s)\n", vme->u.vmx.exit_reason, 697 vmexit_vmx_desc(vme->u.vmx.exit_reason)); 698 fprintf(stderr, "\tqualification\t0x%016lx\n", 699 vme->u.vmx.exit_qualification); 700 fprintf(stderr, "\tinst_type\t\t%d\n", vme->u.vmx.inst_type); 701 fprintf(stderr, "\tinst_error\t\t%d\n", vme->u.vmx.inst_error); 702 #ifdef DEBUG_EPT_MISCONFIG 703 if (vme->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) { 704 vm_get_register(vcpu, 705 VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS), 706 &ept_misconfig_gpa); 707 vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte, 708 &ept_misconfig_ptenum); 709 fprintf(stderr, "\tEPT misconfiguration:\n"); 710 fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa); 711 fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n", 712 ept_misconfig_ptenum, ept_misconfig_pte[0], 713 ept_misconfig_pte[1], ept_misconfig_pte[2], 714 ept_misconfig_pte[3]); 715 } 716 #endif /* DEBUG_EPT_MISCONFIG */ 717 return (VMEXIT_ABORT); 718 } 719 720 static int 721 vmexit_svm(struct vmctx *ctx __unused, struct vcpu *vcpu, struct vm_exit *vme) 722 { 723 724 fprintf(stderr, "vm exit[%d]\n", vcpu_id(vcpu)); 725 fprintf(stderr, "\treason\t\tSVM\n"); 726 fprintf(stderr, "\trip\t\t0x%016lx\n", vme->rip); 727 fprintf(stderr, "\tinst_length\t%d\n", vme->inst_length); 728 fprintf(stderr, "\texitcode\t%#lx\n", vme->u.svm.exitcode); 729 fprintf(stderr, "\texitinfo1\t%#lx\n", vme->u.svm.exitinfo1); 730 fprintf(stderr, "\texitinfo2\t%#lx\n", vme->u.svm.exitinfo2); 731 return (VMEXIT_ABORT); 732 } 733 734 static int 735 vmexit_bogus(struct vmctx *ctx __unused, struct vcpu *vcpu __unused, 736 struct vm_exit *vme) 737 { 738 739 assert(vme->inst_length == 0); 740 741 stats.vmexit_bogus++; 742 743 return (VMEXIT_CONTINUE); 744 } 745 746 static int 747 vmexit_reqidle(struct vmctx *ctx __unused, struct vcpu *vcpu __unused, 748 struct vm_exit *vme) 749 { 750 751 assert(vme->inst_length == 0); 752 753 stats.vmexit_reqidle++; 754 755 return (VMEXIT_CONTINUE); 756 } 757 758 static int 759 vmexit_hlt(struct vmctx *ctx __unused, struct vcpu *vcpu __unused, 760 struct vm_exit *vme __unused) 761 { 762 763 stats.vmexit_hlt++; 764 765 /* 766 * Just continue execution with the next instruction. We use 767 * the HLT VM exit as a way to be friendly with the host 768 * scheduler. 769 */ 770 return (VMEXIT_CONTINUE); 771 } 772 773 static int 774 vmexit_pause(struct vmctx *ctx __unused, struct vcpu *vcpu __unused, 775 struct vm_exit *vme __unused) 776 { 777 778 stats.vmexit_pause++; 779 780 return (VMEXIT_CONTINUE); 781 } 782 783 static int 784 vmexit_mtrap(struct vmctx *ctx __unused, struct vcpu *vcpu, 785 struct vm_exit *vme) 786 { 787 788 assert(vme->inst_length == 0); 789 790 stats.vmexit_mtrap++; 791 792 #ifdef BHYVE_SNAPSHOT 793 checkpoint_cpu_suspend(vcpu_id(vcpu)); 794 #endif 795 gdb_cpu_mtrap(vcpu); 796 #ifdef BHYVE_SNAPSHOT 797 checkpoint_cpu_resume(vcpu_id(vcpu)); 798 #endif 799 800 return (VMEXIT_CONTINUE); 801 } 802 803 static int 804 vmexit_inst_emul(struct vmctx *ctx __unused, struct vcpu *vcpu, 805 struct vm_exit *vme) 806 { 807 int err, i, cs_d; 808 struct vie *vie; 809 enum vm_cpu_mode mode; 810 811 stats.vmexit_inst_emul++; 812 813 vie = &vme->u.inst_emul.vie; 814 if (!vie->decoded) { 815 /* 816 * Attempt to decode in userspace as a fallback. This allows 817 * updating instruction decode in bhyve without rebooting the 818 * kernel (rapid prototyping), albeit with much slower 819 * emulation. 820 */ 821 vie_restart(vie); 822 mode = vme->u.inst_emul.paging.cpu_mode; 823 cs_d = vme->u.inst_emul.cs_d; 824 if (vmm_decode_instruction(mode, cs_d, vie) != 0) 825 goto fail; 826 if (vm_set_register(vcpu, VM_REG_GUEST_RIP, 827 vme->rip + vie->num_processed) != 0) 828 goto fail; 829 } 830 831 err = emulate_mem(vcpu, vme->u.inst_emul.gpa, vie, 832 &vme->u.inst_emul.paging); 833 if (err) { 834 if (err == ESRCH) { 835 EPRINTLN("Unhandled memory access to 0x%lx\n", 836 vme->u.inst_emul.gpa); 837 } 838 goto fail; 839 } 840 841 return (VMEXIT_CONTINUE); 842 843 fail: 844 fprintf(stderr, "Failed to emulate instruction sequence [ "); 845 for (i = 0; i < vie->num_valid; i++) 846 fprintf(stderr, "%02x", vie->inst[i]); 847 FPRINTLN(stderr, " ] at 0x%lx", vme->rip); 848 return (VMEXIT_ABORT); 849 } 850 851 static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER; 852 static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER; 853 854 static int 855 vmexit_suspend(struct vmctx *ctx, struct vcpu *vcpu, struct vm_exit *vme) 856 { 857 enum vm_suspend_how how; 858 int vcpuid = vcpu_id(vcpu); 859 860 how = vme->u.suspended.how; 861 862 fbsdrun_deletecpu(vcpuid); 863 864 if (vcpuid != BSP) { 865 pthread_mutex_lock(&resetcpu_mtx); 866 pthread_cond_signal(&resetcpu_cond); 867 pthread_mutex_unlock(&resetcpu_mtx); 868 pthread_exit(NULL); 869 } 870 871 pthread_mutex_lock(&resetcpu_mtx); 872 while (!CPU_EMPTY(&cpumask)) { 873 pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx); 874 } 875 pthread_mutex_unlock(&resetcpu_mtx); 876 877 switch (how) { 878 case VM_SUSPEND_RESET: 879 exit(0); 880 case VM_SUSPEND_POWEROFF: 881 if (get_config_bool_default("destroy_on_poweroff", false)) 882 vm_destroy(ctx); 883 exit(1); 884 case VM_SUSPEND_HALT: 885 exit(2); 886 case VM_SUSPEND_TRIPLEFAULT: 887 exit(3); 888 default: 889 fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how); 890 exit(100); 891 } 892 return (0); /* NOTREACHED */ 893 } 894 895 static int 896 vmexit_debug(struct vmctx *ctx __unused, struct vcpu *vcpu, 897 struct vm_exit *vme __unused) 898 { 899 900 #ifdef BHYVE_SNAPSHOT 901 checkpoint_cpu_suspend(vcpu_id(vcpu)); 902 #endif 903 gdb_cpu_suspend(vcpu); 904 #ifdef BHYVE_SNAPSHOT 905 checkpoint_cpu_resume(vcpu_id(vcpu)); 906 #endif 907 /* 908 * XXX-MJ sleep for a short period to avoid chewing up the CPU in the 909 * window between activation of the vCPU thread and the STARTUP IPI. 910 */ 911 usleep(1000); 912 return (VMEXIT_CONTINUE); 913 } 914 915 static int 916 vmexit_breakpoint(struct vmctx *ctx __unused, struct vcpu *vcpu, 917 struct vm_exit *vme) 918 { 919 920 gdb_cpu_breakpoint(vcpu, vme); 921 return (VMEXIT_CONTINUE); 922 } 923 924 static int 925 vmexit_ipi(struct vmctx *ctx __unused, struct vcpu *vcpu __unused, 926 struct vm_exit *vme) 927 { 928 int error = -1; 929 int i; 930 switch (vme->u.ipi.mode) { 931 case APIC_DELMODE_INIT: 932 CPU_FOREACH_ISSET(i, &vme->u.ipi.dmask) { 933 error = vm_suspend_cpu(vcpu_info[i].vcpu); 934 if (error) { 935 warnx("%s: failed to suspend cpu %d\n", 936 __func__, i); 937 break; 938 } 939 } 940 break; 941 case APIC_DELMODE_STARTUP: 942 CPU_FOREACH_ISSET(i, &vme->u.ipi.dmask) { 943 spinup_ap(vcpu_info[i].vcpu, 944 vme->u.ipi.vector << PAGE_SHIFT); 945 } 946 error = 0; 947 break; 948 default: 949 break; 950 } 951 952 return (error); 953 } 954 955 static vmexit_handler_t handler[VM_EXITCODE_MAX] = { 956 [VM_EXITCODE_INOUT] = vmexit_inout, 957 [VM_EXITCODE_INOUT_STR] = vmexit_inout, 958 [VM_EXITCODE_VMX] = vmexit_vmx, 959 [VM_EXITCODE_SVM] = vmexit_svm, 960 [VM_EXITCODE_BOGUS] = vmexit_bogus, 961 [VM_EXITCODE_REQIDLE] = vmexit_reqidle, 962 [VM_EXITCODE_RDMSR] = vmexit_rdmsr, 963 [VM_EXITCODE_WRMSR] = vmexit_wrmsr, 964 [VM_EXITCODE_MTRAP] = vmexit_mtrap, 965 [VM_EXITCODE_INST_EMUL] = vmexit_inst_emul, 966 [VM_EXITCODE_SUSPENDED] = vmexit_suspend, 967 [VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch, 968 [VM_EXITCODE_DEBUG] = vmexit_debug, 969 [VM_EXITCODE_BPT] = vmexit_breakpoint, 970 [VM_EXITCODE_IPI] = vmexit_ipi, 971 }; 972 973 static void 974 vm_loop(struct vmctx *ctx, struct vcpu *vcpu) 975 { 976 struct vm_exit vme; 977 int error, rc; 978 enum vm_exitcode exitcode; 979 cpuset_t active_cpus; 980 981 error = vm_active_cpus(ctx, &active_cpus); 982 assert(CPU_ISSET(vcpu_id(vcpu), &active_cpus)); 983 984 while (1) { 985 error = vm_run(vcpu, &vme); 986 if (error != 0) 987 break; 988 989 exitcode = vme.exitcode; 990 if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) { 991 fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n", 992 exitcode); 993 exit(4); 994 } 995 996 rc = (*handler[exitcode])(ctx, vcpu, &vme); 997 998 switch (rc) { 999 case VMEXIT_CONTINUE: 1000 break; 1001 case VMEXIT_ABORT: 1002 abort(); 1003 default: 1004 exit(4); 1005 } 1006 } 1007 fprintf(stderr, "vm_run error %d, errno %d\n", error, errno); 1008 } 1009 1010 static int 1011 num_vcpus_allowed(struct vmctx *ctx, struct vcpu *vcpu) 1012 { 1013 uint16_t sockets, cores, threads, maxcpus; 1014 int tmp, error; 1015 1016 /* 1017 * The guest is allowed to spinup more than one processor only if the 1018 * UNRESTRICTED_GUEST capability is available. 1019 */ 1020 error = vm_get_capability(vcpu, VM_CAP_UNRESTRICTED_GUEST, &tmp); 1021 if (error != 0) 1022 return (1); 1023 1024 error = vm_get_topology(ctx, &sockets, &cores, &threads, &maxcpus); 1025 if (error == 0) 1026 return (maxcpus); 1027 else 1028 return (1); 1029 } 1030 1031 static void 1032 fbsdrun_set_capabilities(struct vcpu *vcpu, bool bsp) 1033 { 1034 int err, tmp; 1035 1036 if (get_config_bool_default("x86.vmexit_on_hlt", false)) { 1037 err = vm_get_capability(vcpu, VM_CAP_HALT_EXIT, &tmp); 1038 if (err < 0) { 1039 fprintf(stderr, "VM exit on HLT not supported\n"); 1040 exit(4); 1041 } 1042 vm_set_capability(vcpu, VM_CAP_HALT_EXIT, 1); 1043 if (bsp) 1044 handler[VM_EXITCODE_HLT] = vmexit_hlt; 1045 } 1046 1047 if (get_config_bool_default("x86.vmexit_on_pause", false)) { 1048 /* 1049 * pause exit support required for this mode 1050 */ 1051 err = vm_get_capability(vcpu, VM_CAP_PAUSE_EXIT, &tmp); 1052 if (err < 0) { 1053 fprintf(stderr, 1054 "SMP mux requested, no pause support\n"); 1055 exit(4); 1056 } 1057 vm_set_capability(vcpu, VM_CAP_PAUSE_EXIT, 1); 1058 if (bsp) 1059 handler[VM_EXITCODE_PAUSE] = vmexit_pause; 1060 } 1061 1062 if (get_config_bool_default("x86.x2apic", false)) 1063 err = vm_set_x2apic_state(vcpu, X2APIC_ENABLED); 1064 else 1065 err = vm_set_x2apic_state(vcpu, X2APIC_DISABLED); 1066 1067 if (err) { 1068 fprintf(stderr, "Unable to set x2apic state (%d)\n", err); 1069 exit(4); 1070 } 1071 1072 vm_set_capability(vcpu, VM_CAP_ENABLE_INVPCID, 1); 1073 1074 err = vm_set_capability(vcpu, VM_CAP_IPI_EXIT, 1); 1075 assert(err == 0); 1076 } 1077 1078 static struct vmctx * 1079 do_open(const char *vmname) 1080 { 1081 struct vmctx *ctx; 1082 int error; 1083 bool reinit, romboot; 1084 1085 reinit = romboot = false; 1086 1087 if (lpc_bootrom()) 1088 romboot = true; 1089 1090 error = vm_create(vmname); 1091 if (error) { 1092 if (errno == EEXIST) { 1093 if (romboot) { 1094 reinit = true; 1095 } else { 1096 /* 1097 * The virtual machine has been setup by the 1098 * userspace bootloader. 1099 */ 1100 } 1101 } else { 1102 perror("vm_create"); 1103 exit(4); 1104 } 1105 } else { 1106 if (!romboot) { 1107 /* 1108 * If the virtual machine was just created then a 1109 * bootrom must be configured to boot it. 1110 */ 1111 fprintf(stderr, "virtual machine cannot be booted\n"); 1112 exit(4); 1113 } 1114 } 1115 1116 ctx = vm_open(vmname); 1117 if (ctx == NULL) { 1118 perror("vm_open"); 1119 exit(4); 1120 } 1121 1122 #ifndef WITHOUT_CAPSICUM 1123 if (vm_limit_rights(ctx) != 0) 1124 err(EX_OSERR, "vm_limit_rights"); 1125 #endif 1126 1127 if (reinit) { 1128 error = vm_reinit(ctx); 1129 if (error) { 1130 perror("vm_reinit"); 1131 exit(4); 1132 } 1133 } 1134 error = vm_set_topology(ctx, cpu_sockets, cpu_cores, cpu_threads, 1135 0 /* maxcpus, unimplemented */); 1136 if (error) 1137 errx(EX_OSERR, "vm_set_topology"); 1138 return (ctx); 1139 } 1140 1141 static void 1142 spinup_vcpu(struct vcpu_info *vi, bool bsp) 1143 { 1144 int error; 1145 1146 if (!bsp) { 1147 fbsdrun_set_capabilities(vi->vcpu, false); 1148 1149 /* 1150 * Enable the 'unrestricted guest' mode for APs. 1151 * 1152 * APs startup in power-on 16-bit mode. 1153 */ 1154 error = vm_set_capability(vi->vcpu, VM_CAP_UNRESTRICTED_GUEST, 1); 1155 assert(error == 0); 1156 } 1157 1158 fbsdrun_addcpu(vi); 1159 } 1160 1161 static bool 1162 parse_config_option(const char *option) 1163 { 1164 const char *value; 1165 char *path; 1166 1167 value = strchr(option, '='); 1168 if (value == NULL || value[1] == '\0') 1169 return (false); 1170 path = strndup(option, value - option); 1171 if (path == NULL) 1172 err(4, "Failed to allocate memory"); 1173 set_config_value(path, value + 1); 1174 return (true); 1175 } 1176 1177 static void 1178 parse_simple_config_file(const char *path) 1179 { 1180 FILE *fp; 1181 char *line, *cp; 1182 size_t linecap; 1183 unsigned int lineno; 1184 1185 fp = fopen(path, "r"); 1186 if (fp == NULL) 1187 err(4, "Failed to open configuration file %s", path); 1188 line = NULL; 1189 linecap = 0; 1190 lineno = 1; 1191 for (lineno = 1; getline(&line, &linecap, fp) > 0; lineno++) { 1192 if (*line == '#' || *line == '\n') 1193 continue; 1194 cp = strchr(line, '\n'); 1195 if (cp != NULL) 1196 *cp = '\0'; 1197 if (!parse_config_option(line)) 1198 errx(4, "%s line %u: invalid config option '%s'", path, 1199 lineno, line); 1200 } 1201 free(line); 1202 fclose(fp); 1203 } 1204 1205 static void 1206 parse_gdb_options(const char *opt) 1207 { 1208 const char *sport; 1209 char *colon; 1210 1211 if (opt[0] == 'w') { 1212 set_config_bool("gdb.wait", true); 1213 opt++; 1214 } 1215 1216 colon = strrchr(opt, ':'); 1217 if (colon == NULL) { 1218 sport = opt; 1219 } else { 1220 *colon = '\0'; 1221 colon++; 1222 sport = colon; 1223 set_config_value("gdb.address", opt); 1224 } 1225 1226 set_config_value("gdb.port", sport); 1227 } 1228 1229 static void 1230 set_defaults(void) 1231 { 1232 1233 set_config_bool("acpi_tables", false); 1234 set_config_value("memory.size", "256M"); 1235 set_config_bool("x86.strictmsr", true); 1236 set_config_value("lpc.fwcfg", "bhyve"); 1237 } 1238 1239 int 1240 main(int argc, char *argv[]) 1241 { 1242 int c, error; 1243 int max_vcpus, memflags; 1244 struct vcpu *bsp; 1245 struct vmctx *ctx; 1246 struct qemu_fwcfg_item *e820_fwcfg_item; 1247 uint64_t rip; 1248 size_t memsize; 1249 const char *optstr, *value, *vmname; 1250 #ifdef BHYVE_SNAPSHOT 1251 char *restore_file; 1252 struct restore_state rstate; 1253 1254 restore_file = NULL; 1255 #endif 1256 1257 init_config(); 1258 set_defaults(); 1259 progname = basename(argv[0]); 1260 1261 #ifdef BHYVE_SNAPSHOT 1262 optstr = "aehuwxACDHIPSWYk:f:o:p:G:c:s:m:l:K:U:r:"; 1263 #else 1264 optstr = "aehuwxACDHIPSWYk:f:o:p:G:c:s:m:l:K:U:"; 1265 #endif 1266 while ((c = getopt(argc, argv, optstr)) != -1) { 1267 switch (c) { 1268 case 'a': 1269 set_config_bool("x86.x2apic", false); 1270 break; 1271 case 'A': 1272 set_config_bool("acpi_tables", true); 1273 break; 1274 case 'D': 1275 set_config_bool("destroy_on_poweroff", true); 1276 break; 1277 case 'p': 1278 if (pincpu_parse(optarg) != 0) { 1279 errx(EX_USAGE, "invalid vcpu pinning " 1280 "configuration '%s'", optarg); 1281 } 1282 break; 1283 case 'c': 1284 if (topology_parse(optarg) != 0) { 1285 errx(EX_USAGE, "invalid cpu topology " 1286 "'%s'", optarg); 1287 } 1288 break; 1289 case 'C': 1290 set_config_bool("memory.guest_in_core", true); 1291 break; 1292 case 'f': 1293 if (qemu_fwcfg_parse_cmdline_arg(optarg) != 0) { 1294 errx(EX_USAGE, "invalid fwcfg item '%s'", optarg); 1295 } 1296 break; 1297 case 'G': 1298 parse_gdb_options(optarg); 1299 break; 1300 case 'k': 1301 parse_simple_config_file(optarg); 1302 break; 1303 case 'K': 1304 set_config_value("keyboard.layout", optarg); 1305 break; 1306 case 'l': 1307 if (strncmp(optarg, "help", strlen(optarg)) == 0) { 1308 lpc_print_supported_devices(); 1309 exit(0); 1310 } else if (lpc_device_parse(optarg) != 0) { 1311 errx(EX_USAGE, "invalid lpc device " 1312 "configuration '%s'", optarg); 1313 } 1314 break; 1315 #ifdef BHYVE_SNAPSHOT 1316 case 'r': 1317 restore_file = optarg; 1318 break; 1319 #endif 1320 case 's': 1321 if (strncmp(optarg, "help", strlen(optarg)) == 0) { 1322 pci_print_supported_devices(); 1323 exit(0); 1324 } else if (pci_parse_slot(optarg) != 0) 1325 exit(4); 1326 else 1327 break; 1328 case 'S': 1329 set_config_bool("memory.wired", true); 1330 break; 1331 case 'm': 1332 set_config_value("memory.size", optarg); 1333 break; 1334 case 'o': 1335 if (!parse_config_option(optarg)) 1336 errx(EX_USAGE, "invalid configuration option '%s'", optarg); 1337 break; 1338 case 'H': 1339 set_config_bool("x86.vmexit_on_hlt", true); 1340 break; 1341 case 'I': 1342 /* 1343 * The "-I" option was used to add an ioapic to the 1344 * virtual machine. 1345 * 1346 * An ioapic is now provided unconditionally for each 1347 * virtual machine and this option is now deprecated. 1348 */ 1349 break; 1350 case 'P': 1351 set_config_bool("x86.vmexit_on_pause", true); 1352 break; 1353 case 'e': 1354 set_config_bool("x86.strictio", true); 1355 break; 1356 case 'u': 1357 set_config_bool("rtc.use_localtime", false); 1358 break; 1359 case 'U': 1360 set_config_value("uuid", optarg); 1361 break; 1362 case 'w': 1363 set_config_bool("x86.strictmsr", false); 1364 break; 1365 case 'W': 1366 set_config_bool("virtio_msix", false); 1367 break; 1368 case 'x': 1369 set_config_bool("x86.x2apic", true); 1370 break; 1371 case 'Y': 1372 set_config_bool("x86.mptable", false); 1373 break; 1374 case 'h': 1375 usage(0); 1376 default: 1377 usage(1); 1378 } 1379 } 1380 argc -= optind; 1381 argv += optind; 1382 1383 if (argc > 1) 1384 usage(1); 1385 1386 #ifdef BHYVE_SNAPSHOT 1387 if (restore_file != NULL) { 1388 error = load_restore_file(restore_file, &rstate); 1389 if (error) { 1390 fprintf(stderr, "Failed to read checkpoint info from " 1391 "file: '%s'.\n", restore_file); 1392 exit(1); 1393 } 1394 vmname = lookup_vmname(&rstate); 1395 if (vmname != NULL) 1396 set_config_value("name", vmname); 1397 } 1398 #endif 1399 1400 if (argc == 1) 1401 set_config_value("name", argv[0]); 1402 1403 vmname = get_config_value("name"); 1404 if (vmname == NULL) 1405 usage(1); 1406 1407 if (get_config_bool_default("config.dump", false)) { 1408 dump_config(); 1409 exit(1); 1410 } 1411 1412 calc_topology(); 1413 build_vcpumaps(); 1414 1415 value = get_config_value("memory.size"); 1416 error = vm_parse_memsize(value, &memsize); 1417 if (error) 1418 errx(EX_USAGE, "invalid memsize '%s'", value); 1419 1420 ctx = do_open(vmname); 1421 1422 #ifdef BHYVE_SNAPSHOT 1423 if (restore_file != NULL) { 1424 guest_ncpus = lookup_guest_ncpus(&rstate); 1425 memflags = lookup_memflags(&rstate); 1426 memsize = lookup_memsize(&rstate); 1427 } 1428 1429 if (guest_ncpus < 1) { 1430 fprintf(stderr, "Invalid guest vCPUs (%d)\n", guest_ncpus); 1431 exit(1); 1432 } 1433 #endif 1434 1435 bsp = vm_vcpu_open(ctx, BSP); 1436 max_vcpus = num_vcpus_allowed(ctx, bsp); 1437 if (guest_ncpus > max_vcpus) { 1438 fprintf(stderr, "%d vCPUs requested but only %d available\n", 1439 guest_ncpus, max_vcpus); 1440 exit(4); 1441 } 1442 1443 fbsdrun_set_capabilities(bsp, true); 1444 1445 /* Allocate per-VCPU resources. */ 1446 vcpu_info = calloc(guest_ncpus, sizeof(*vcpu_info)); 1447 for (int vcpuid = 0; vcpuid < guest_ncpus; vcpuid++) { 1448 vcpu_info[vcpuid].ctx = ctx; 1449 vcpu_info[vcpuid].vcpuid = vcpuid; 1450 if (vcpuid == BSP) 1451 vcpu_info[vcpuid].vcpu = bsp; 1452 else 1453 vcpu_info[vcpuid].vcpu = vm_vcpu_open(ctx, vcpuid); 1454 } 1455 1456 memflags = 0; 1457 if (get_config_bool_default("memory.wired", false)) 1458 memflags |= VM_MEM_F_WIRED; 1459 if (get_config_bool_default("memory.guest_in_core", false)) 1460 memflags |= VM_MEM_F_INCORE; 1461 vm_set_memflags(ctx, memflags); 1462 error = vm_setup_memory(ctx, memsize, VM_MMAP_ALL); 1463 if (error) { 1464 fprintf(stderr, "Unable to setup memory (%d)\n", errno); 1465 exit(4); 1466 } 1467 1468 error = init_msr(); 1469 if (error) { 1470 fprintf(stderr, "init_msr error %d", error); 1471 exit(4); 1472 } 1473 1474 init_mem(guest_ncpus); 1475 init_inout(); 1476 kernemu_dev_init(); 1477 init_bootrom(ctx); 1478 atkbdc_init(ctx); 1479 pci_irq_init(ctx); 1480 ioapic_init(ctx); 1481 1482 rtc_init(ctx); 1483 sci_init(ctx); 1484 1485 if (qemu_fwcfg_init(ctx) != 0) { 1486 fprintf(stderr, "qemu fwcfg initialization error"); 1487 exit(4); 1488 } 1489 1490 if (qemu_fwcfg_add_file("opt/bhyve/hw.ncpu", sizeof(guest_ncpus), 1491 &guest_ncpus) != 0) { 1492 fprintf(stderr, "Could not add qemu fwcfg opt/bhyve/hw.ncpu"); 1493 exit(4); 1494 } 1495 1496 if (e820_init(ctx) != 0) { 1497 fprintf(stderr, "Unable to setup E820"); 1498 exit(4); 1499 } 1500 1501 /* 1502 * Exit if a device emulation finds an error in its initilization 1503 */ 1504 if (init_pci(ctx) != 0) { 1505 perror("device emulation initialization error"); 1506 exit(4); 1507 } 1508 1509 /* 1510 * Initialize after PCI, to allow a bootrom file to reserve the high 1511 * region. 1512 */ 1513 if (get_config_bool("acpi_tables")) 1514 vmgenc_init(ctx); 1515 1516 init_gdb(ctx); 1517 1518 if (lpc_bootrom()) { 1519 if (vm_set_capability(bsp, VM_CAP_UNRESTRICTED_GUEST, 1)) { 1520 fprintf(stderr, "ROM boot failed: unrestricted guest " 1521 "capability not available\n"); 1522 exit(4); 1523 } 1524 error = vcpu_reset(bsp); 1525 assert(error == 0); 1526 } 1527 1528 /* 1529 * Add all vCPUs. 1530 */ 1531 for (int vcpuid = 0; vcpuid < guest_ncpus; vcpuid++) 1532 spinup_vcpu(&vcpu_info[vcpuid], vcpuid == BSP); 1533 1534 #ifdef BHYVE_SNAPSHOT 1535 if (restore_file != NULL) { 1536 fprintf(stdout, "Pausing pci devs...\r\n"); 1537 if (vm_pause_user_devs() != 0) { 1538 fprintf(stderr, "Failed to pause PCI device state.\n"); 1539 exit(1); 1540 } 1541 1542 fprintf(stdout, "Restoring vm mem...\r\n"); 1543 if (restore_vm_mem(ctx, &rstate) != 0) { 1544 fprintf(stderr, "Failed to restore VM memory.\n"); 1545 exit(1); 1546 } 1547 1548 fprintf(stdout, "Restoring pci devs...\r\n"); 1549 if (vm_restore_user_devs(&rstate) != 0) { 1550 fprintf(stderr, "Failed to restore PCI device state.\n"); 1551 exit(1); 1552 } 1553 1554 fprintf(stdout, "Restoring kernel structs...\r\n"); 1555 if (vm_restore_kern_structs(ctx, &rstate) != 0) { 1556 fprintf(stderr, "Failed to restore kernel structs.\n"); 1557 exit(1); 1558 } 1559 1560 fprintf(stdout, "Resuming pci devs...\r\n"); 1561 if (vm_resume_user_devs() != 0) { 1562 fprintf(stderr, "Failed to resume PCI device state.\n"); 1563 exit(1); 1564 } 1565 } 1566 #endif 1567 1568 error = vm_get_register(bsp, VM_REG_GUEST_RIP, &rip); 1569 assert(error == 0); 1570 1571 /* 1572 * build the guest tables, MP etc. 1573 */ 1574 if (get_config_bool_default("x86.mptable", true)) { 1575 error = mptable_build(ctx, guest_ncpus); 1576 if (error) { 1577 perror("error to build the guest tables"); 1578 exit(4); 1579 } 1580 } 1581 1582 error = smbios_build(ctx); 1583 if (error != 0) 1584 exit(4); 1585 1586 if (get_config_bool("acpi_tables")) { 1587 error = acpi_build(ctx, guest_ncpus); 1588 assert(error == 0); 1589 } 1590 1591 e820_fwcfg_item = e820_get_fwcfg_item(); 1592 if (e820_fwcfg_item == NULL) { 1593 fprintf(stderr, "invalid e820 table"); 1594 exit(4); 1595 } 1596 if (qemu_fwcfg_add_file("etc/e820", e820_fwcfg_item->size, 1597 e820_fwcfg_item->data) != 0) { 1598 fprintf(stderr, "could not add qemu fwcfg etc/e820"); 1599 exit(4); 1600 } 1601 free(e820_fwcfg_item); 1602 1603 if (lpc_bootrom() && strcmp(lpc_fwcfg(), "bhyve") == 0) { 1604 fwctl_init(); 1605 } 1606 1607 /* 1608 * Change the proc title to include the VM name. 1609 */ 1610 setproctitle("%s", vmname); 1611 1612 #ifdef BHYVE_SNAPSHOT 1613 /* initialize mutex/cond variables */ 1614 init_snapshot(); 1615 1616 /* 1617 * checkpointing thread for communication with bhyvectl 1618 */ 1619 if (init_checkpoint_thread(ctx) != 0) 1620 errx(EX_OSERR, "Failed to start checkpoint thread"); 1621 #endif 1622 1623 #ifndef WITHOUT_CAPSICUM 1624 caph_cache_catpages(); 1625 1626 if (caph_limit_stdout() == -1 || caph_limit_stderr() == -1) 1627 errx(EX_OSERR, "Unable to apply rights for sandbox"); 1628 1629 if (caph_enter() == -1) 1630 errx(EX_OSERR, "cap_enter() failed"); 1631 #endif 1632 1633 #ifdef BHYVE_SNAPSHOT 1634 if (restore_file != NULL) { 1635 destroy_restore_state(&rstate); 1636 if (vm_restore_time(ctx) < 0) 1637 err(EX_OSERR, "Unable to restore time"); 1638 1639 for (int vcpuid = 0; vcpuid < guest_ncpus; vcpuid++) 1640 vm_resume_cpu(vcpu_info[vcpuid].vcpu); 1641 } else 1642 #endif 1643 vm_resume_cpu(bsp); 1644 1645 /* 1646 * Head off to the main event dispatch loop 1647 */ 1648 mevent_dispatch(); 1649 1650 exit(4); 1651 } 1652