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 * This file and its contents are supplied under the terms of the 32 * Common Development and Distribution License ("CDDL"), version 1.0. 33 * You may only use this file in accordance with the terms of version 34 * 1.0 of the CDDL. 35 * 36 * A full copy of the text of the CDDL should have accompanied this 37 * source. A copy of the CDDL is also available via the Internet at 38 * http://www.illumos.org/license/CDDL. 39 * 40 * Copyright 2015 Pluribus Networks Inc. 41 * Copyright 2018 Joyent, Inc. 42 * Copyright 2020 Oxide Computer Company 43 */ 44 45 #include <sys/cdefs.h> 46 __FBSDID("$FreeBSD$"); 47 48 #include <sys/types.h> 49 #ifndef WITHOUT_CAPSICUM 50 #include <sys/capsicum.h> 51 #endif 52 #include <sys/mman.h> 53 #include <sys/time.h> 54 #include <sys/cpuset.h> 55 56 #ifdef __FreeBSD__ 57 #include <amd64/vmm/intel/vmcs.h> 58 #else 59 #include <intel/vmcs.h> 60 #endif 61 62 #include <machine/atomic.h> 63 #include <machine/segments.h> 64 65 #ifndef WITHOUT_CAPSICUM 66 #include <capsicum_helpers.h> 67 #endif 68 #include <stdio.h> 69 #include <stdlib.h> 70 #include <string.h> 71 #include <err.h> 72 #include <errno.h> 73 #include <libgen.h> 74 #include <unistd.h> 75 #include <assert.h> 76 #include <pthread.h> 77 #include <pthread_np.h> 78 #include <sysexits.h> 79 #include <stdbool.h> 80 #include <stdint.h> 81 82 #include <machine/vmm.h> 83 #ifndef WITHOUT_CAPSICUM 84 #include <machine/vmm_dev.h> 85 #endif 86 #include <vmmapi.h> 87 88 #include "bhyverun.h" 89 #include "acpi.h" 90 #include "atkbdc.h" 91 #include "console.h" 92 #include "bootrom.h" 93 #include "inout.h" 94 #include "dbgport.h" 95 #include "debug.h" 96 #include "fwctl.h" 97 #include "gdb.h" 98 #include "ioapic.h" 99 #include "kernemu_dev.h" 100 #include "mem.h" 101 #include "mevent.h" 102 #include "mptbl.h" 103 #include "pci_emul.h" 104 #include "pci_irq.h" 105 #include "pci_lpc.h" 106 #include "smbiostbl.h" 107 #include "xmsr.h" 108 #include "spinup_ap.h" 109 #include "rfb.h" 110 #include "rtc.h" 111 #include "vga.h" 112 #include "vmgenc.h" 113 114 #define GUEST_NIO_PORT 0x488 /* guest upcalls via i/o port */ 115 116 #define MB (1024UL * 1024) 117 #define GB (1024UL * MB) 118 119 static const char * const vmx_exit_reason_desc[] = { 120 [EXIT_REASON_EXCEPTION] = "Exception or non-maskable interrupt (NMI)", 121 [EXIT_REASON_EXT_INTR] = "External interrupt", 122 [EXIT_REASON_TRIPLE_FAULT] = "Triple fault", 123 [EXIT_REASON_INIT] = "INIT signal", 124 [EXIT_REASON_SIPI] = "Start-up IPI (SIPI)", 125 [EXIT_REASON_IO_SMI] = "I/O system-management interrupt (SMI)", 126 [EXIT_REASON_SMI] = "Other SMI", 127 [EXIT_REASON_INTR_WINDOW] = "Interrupt window", 128 [EXIT_REASON_NMI_WINDOW] = "NMI window", 129 [EXIT_REASON_TASK_SWITCH] = "Task switch", 130 [EXIT_REASON_CPUID] = "CPUID", 131 [EXIT_REASON_GETSEC] = "GETSEC", 132 [EXIT_REASON_HLT] = "HLT", 133 [EXIT_REASON_INVD] = "INVD", 134 [EXIT_REASON_INVLPG] = "INVLPG", 135 [EXIT_REASON_RDPMC] = "RDPMC", 136 [EXIT_REASON_RDTSC] = "RDTSC", 137 [EXIT_REASON_RSM] = "RSM", 138 [EXIT_REASON_VMCALL] = "VMCALL", 139 [EXIT_REASON_VMCLEAR] = "VMCLEAR", 140 [EXIT_REASON_VMLAUNCH] = "VMLAUNCH", 141 [EXIT_REASON_VMPTRLD] = "VMPTRLD", 142 [EXIT_REASON_VMPTRST] = "VMPTRST", 143 [EXIT_REASON_VMREAD] = "VMREAD", 144 [EXIT_REASON_VMRESUME] = "VMRESUME", 145 [EXIT_REASON_VMWRITE] = "VMWRITE", 146 [EXIT_REASON_VMXOFF] = "VMXOFF", 147 [EXIT_REASON_VMXON] = "VMXON", 148 [EXIT_REASON_CR_ACCESS] = "Control-register accesses", 149 [EXIT_REASON_DR_ACCESS] = "MOV DR", 150 [EXIT_REASON_INOUT] = "I/O instruction", 151 [EXIT_REASON_RDMSR] = "RDMSR", 152 [EXIT_REASON_WRMSR] = "WRMSR", 153 [EXIT_REASON_INVAL_VMCS] = 154 "VM-entry failure due to invalid guest state", 155 [EXIT_REASON_INVAL_MSR] = "VM-entry failure due to MSR loading", 156 [EXIT_REASON_MWAIT] = "MWAIT", 157 [EXIT_REASON_MTF] = "Monitor trap flag", 158 [EXIT_REASON_MONITOR] = "MONITOR", 159 [EXIT_REASON_PAUSE] = "PAUSE", 160 [EXIT_REASON_MCE_DURING_ENTRY] = 161 "VM-entry failure due to machine-check event", 162 [EXIT_REASON_TPR] = "TPR below threshold", 163 [EXIT_REASON_APIC_ACCESS] = "APIC access", 164 [EXIT_REASON_VIRTUALIZED_EOI] = "Virtualized EOI", 165 [EXIT_REASON_GDTR_IDTR] = "Access to GDTR or IDTR", 166 [EXIT_REASON_LDTR_TR] = "Access to LDTR or TR", 167 [EXIT_REASON_EPT_FAULT] = "EPT violation", 168 [EXIT_REASON_EPT_MISCONFIG] = "EPT misconfiguration", 169 [EXIT_REASON_INVEPT] = "INVEPT", 170 [EXIT_REASON_RDTSCP] = "RDTSCP", 171 [EXIT_REASON_VMX_PREEMPT] = "VMX-preemption timer expired", 172 [EXIT_REASON_INVVPID] = "INVVPID", 173 [EXIT_REASON_WBINVD] = "WBINVD", 174 [EXIT_REASON_XSETBV] = "XSETBV", 175 [EXIT_REASON_APIC_WRITE] = "APIC write", 176 [EXIT_REASON_RDRAND] = "RDRAND", 177 [EXIT_REASON_INVPCID] = "INVPCID", 178 [EXIT_REASON_VMFUNC] = "VMFUNC", 179 [EXIT_REASON_ENCLS] = "ENCLS", 180 [EXIT_REASON_RDSEED] = "RDSEED", 181 [EXIT_REASON_PM_LOG_FULL] = "Page-modification log full", 182 [EXIT_REASON_XSAVES] = "XSAVES", 183 [EXIT_REASON_XRSTORS] = "XRSTORS" 184 }; 185 186 typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu); 187 extern int vmexit_task_switch(struct vmctx *, struct vm_exit *, int *vcpu); 188 189 char *vmname; 190 191 int guest_ncpus; 192 uint16_t cores, maxcpus, sockets, threads; 193 194 char *guest_uuid_str; 195 196 int raw_stdio = 0; 197 198 static int gdb_port = 0; 199 static int guest_vmexit_on_hlt, guest_vmexit_on_pause; 200 static int virtio_msix = 1; 201 static int x2apic_mode = 0; /* default is xAPIC */ 202 static int destroy_on_poweroff = 0; 203 204 static int strictio; 205 static int strictmsr = 1; 206 207 static int acpi; 208 209 static char *progname; 210 static const int BSP = 0; 211 212 static cpuset_t cpumask; 213 214 static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip); 215 216 static struct vm_exit vmexit[VM_MAXCPU]; 217 static struct vm_entry vmentry[VM_MAXCPU]; 218 219 struct bhyvestats { 220 uint64_t vmexit_bogus; 221 uint64_t vmexit_reqidle; 222 uint64_t vmexit_hlt; 223 uint64_t vmexit_pause; 224 uint64_t vmexit_mtrap; 225 uint64_t vmexit_mmio; 226 uint64_t vmexit_inout; 227 uint64_t cpu_switch_rotate; 228 uint64_t cpu_switch_direct; 229 uint64_t mmio_unhandled; 230 } stats; 231 232 struct mt_vmm_info { 233 pthread_t mt_thr; 234 struct vmctx *mt_ctx; 235 int mt_vcpu; 236 uint64_t mt_startrip; 237 } mt_vmm_info[VM_MAXCPU]; 238 239 #ifdef __FreeBSD__ 240 static cpuset_t *vcpumap[VM_MAXCPU] = { NULL }; 241 #endif 242 243 static void 244 usage(int code) 245 { 246 247 fprintf(stderr, 248 #ifdef __FreeBSD__ 249 "Usage: %s [-abehuwxACDHPSWY]\n" 250 #else 251 "Usage: %s [-abdehuwxACDHPSWY]\n" 252 #endif 253 " %*s [-c [[cpus=]numcpus][,sockets=n][,cores=n][,threads=n]]\n" 254 " %*s [-g <gdb port>] [-l <lpc>]\n" 255 #ifdef __FreeBSD__ 256 " %*s [-m mem] [-p vcpu:hostcpu] [-s <pci>] [-U uuid] <vm>\n" 257 #else 258 " %*s [-m mem] [-s <pci>] [-U uuid] <vm>\n" 259 #endif 260 " -a: local apic is in xAPIC mode (deprecated)\n" 261 " -A: create ACPI tables\n" 262 " -c: number of cpus and/or topology specification\n" 263 " -C: include guest memory in core file\n" 264 #ifndef __FreeBSD__ 265 " -d: suspend cpu at boot\n" 266 #endif 267 " -D: destroy on power-off\n" 268 " -e: exit on unhandled I/O access\n" 269 " -g: gdb port\n" 270 " -h: help\n" 271 " -H: vmexit from the guest on hlt\n" 272 " -l: LPC device configuration\n" 273 " -m: memory size\n" 274 #ifdef __FreeBSD__ 275 " -p: pin 'vcpu' to 'hostcpu'\n" 276 #endif 277 " -P: vmexit from the guest on pause\n" 278 " -s: <slot,driver,configinfo> PCI slot config\n" 279 " -S: guest memory cannot be swapped\n" 280 " -u: RTC keeps UTC time\n" 281 " -U: uuid\n" 282 " -w: ignore unimplemented MSRs\n" 283 " -W: force virtio to use single-vector MSI\n" 284 " -x: local apic is in x2APIC mode\n" 285 " -Y: disable MPtable generation\n", 286 progname, (int)strlen(progname), "", (int)strlen(progname), "", 287 (int)strlen(progname), ""); 288 289 exit(code); 290 } 291 292 /* 293 * XXX This parser is known to have the following issues: 294 * 1. It accepts null key=value tokens ",,". 295 * 2. It accepts whitespace after = and before value. 296 * 3. Values out of range of INT are silently wrapped. 297 * 4. It doesn't check non-final values. 298 * 5. The apparently bogus limits of UINT16_MAX are for future expansion. 299 * 300 * The acceptance of a null specification ('-c ""') is by design to match the 301 * manual page syntax specification, this results in a topology of 1 vCPU. 302 */ 303 static int 304 topology_parse(const char *opt) 305 { 306 uint64_t ncpus; 307 int c, chk, n, s, t, tmp; 308 char *cp, *str; 309 bool ns, scts; 310 311 c = 1, n = 1, s = 1, t = 1; 312 ns = false, scts = false; 313 str = strdup(opt); 314 if (str == NULL) 315 goto out; 316 317 while ((cp = strsep(&str, ",")) != NULL) { 318 if (sscanf(cp, "%i%n", &tmp, &chk) == 1) { 319 n = tmp; 320 ns = true; 321 } else if (sscanf(cp, "cpus=%i%n", &tmp, &chk) == 1) { 322 n = tmp; 323 ns = true; 324 } else if (sscanf(cp, "sockets=%i%n", &tmp, &chk) == 1) { 325 s = tmp; 326 scts = true; 327 } else if (sscanf(cp, "cores=%i%n", &tmp, &chk) == 1) { 328 c = tmp; 329 scts = true; 330 } else if (sscanf(cp, "threads=%i%n", &tmp, &chk) == 1) { 331 t = tmp; 332 scts = true; 333 #ifdef notyet /* Do not expose this until vmm.ko implements it */ 334 } else if (sscanf(cp, "maxcpus=%i%n", &tmp, &chk) == 1) { 335 m = tmp; 336 #endif 337 /* Skip the empty argument case from -c "" */ 338 } else if (cp[0] == '\0') 339 continue; 340 else 341 goto out; 342 /* Any trailing garbage causes an error */ 343 if (cp[chk] != '\0') 344 goto out; 345 } 346 free(str); 347 str = NULL; 348 349 /* 350 * Range check 1 <= n <= UINT16_MAX all values 351 */ 352 if (n < 1 || s < 1 || c < 1 || t < 1 || 353 n > UINT16_MAX || s > UINT16_MAX || c > UINT16_MAX || 354 t > UINT16_MAX) 355 return (-1); 356 357 /* If only the cpus was specified, use that as sockets */ 358 if (!scts) 359 s = n; 360 /* 361 * Compute sockets * cores * threads avoiding overflow 362 * The range check above insures these are 16 bit values 363 * If n was specified check it against computed ncpus 364 */ 365 ncpus = (uint64_t)s * c * t; 366 if (ncpus > UINT16_MAX || (ns && n != ncpus)) 367 return (-1); 368 369 guest_ncpus = ncpus; 370 sockets = s; 371 cores = c; 372 threads = t; 373 return(0); 374 375 out: 376 free(str); 377 return (-1); 378 } 379 380 #ifndef WITHOUT_CAPSICUM 381 /* 382 * 11-stable capsicum helpers 383 */ 384 static void 385 bhyve_caph_cache_catpages(void) 386 { 387 388 (void)catopen("libc", NL_CAT_LOCALE); 389 } 390 391 static int 392 bhyve_caph_limit_stdoe(void) 393 { 394 cap_rights_t rights; 395 unsigned long cmds[] = { TIOCGETA, TIOCGWINSZ }; 396 int i, fds[] = { STDOUT_FILENO, STDERR_FILENO }; 397 398 cap_rights_init(&rights, CAP_FCNTL, CAP_FSTAT, CAP_IOCTL); 399 cap_rights_set(&rights, CAP_WRITE); 400 401 for (i = 0; i < nitems(fds); i++) { 402 if (cap_rights_limit(fds[i], &rights) < 0 && errno != ENOSYS) 403 return (-1); 404 405 if (cap_ioctls_limit(fds[i], cmds, nitems(cmds)) < 0 && errno != ENOSYS) 406 return (-1); 407 408 if (cap_fcntls_limit(fds[i], CAP_FCNTL_GETFL) < 0 && errno != ENOSYS) 409 return (-1); 410 } 411 412 return (0); 413 } 414 415 #endif 416 417 #ifdef __FreeBSD__ 418 static int 419 pincpu_parse(const char *opt) 420 { 421 int vcpu, pcpu; 422 423 if (sscanf(opt, "%d:%d", &vcpu, &pcpu) != 2) { 424 fprintf(stderr, "invalid format: %s\n", opt); 425 return (-1); 426 } 427 428 if (vcpu < 0 || vcpu >= VM_MAXCPU) { 429 fprintf(stderr, "vcpu '%d' outside valid range from 0 to %d\n", 430 vcpu, VM_MAXCPU - 1); 431 return (-1); 432 } 433 434 if (pcpu < 0 || pcpu >= CPU_SETSIZE) { 435 fprintf(stderr, "hostcpu '%d' outside valid range from " 436 "0 to %d\n", pcpu, CPU_SETSIZE - 1); 437 return (-1); 438 } 439 440 if (vcpumap[vcpu] == NULL) { 441 if ((vcpumap[vcpu] = malloc(sizeof(cpuset_t))) == NULL) { 442 perror("malloc"); 443 return (-1); 444 } 445 CPU_ZERO(vcpumap[vcpu]); 446 } 447 CPU_SET(pcpu, vcpumap[vcpu]); 448 return (0); 449 } 450 451 void 452 vm_inject_fault(void *arg, int vcpu, int vector, int errcode_valid, 453 int errcode) 454 { 455 struct vmctx *ctx; 456 int error, restart_instruction; 457 458 ctx = arg; 459 restart_instruction = 1; 460 461 error = vm_inject_exception(ctx, vcpu, vector, errcode_valid, errcode, 462 restart_instruction); 463 assert(error == 0); 464 } 465 #endif /* __FreeBSD__ */ 466 467 void * 468 paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len) 469 { 470 471 return (vm_map_gpa(ctx, gaddr, len)); 472 } 473 474 int 475 fbsdrun_vmexit_on_pause(void) 476 { 477 478 return (guest_vmexit_on_pause); 479 } 480 481 int 482 fbsdrun_vmexit_on_hlt(void) 483 { 484 485 return (guest_vmexit_on_hlt); 486 } 487 488 int 489 fbsdrun_virtio_msix(void) 490 { 491 492 return (virtio_msix); 493 } 494 495 static void * 496 fbsdrun_start_thread(void *param) 497 { 498 char tname[MAXCOMLEN + 1]; 499 struct mt_vmm_info *mtp; 500 int vcpu; 501 502 mtp = param; 503 vcpu = mtp->mt_vcpu; 504 505 snprintf(tname, sizeof(tname), "vcpu %d", vcpu); 506 pthread_set_name_np(mtp->mt_thr, tname); 507 508 if (gdb_port != 0) 509 gdb_cpu_add(vcpu); 510 511 vm_loop(mtp->mt_ctx, vcpu, mtp->mt_startrip); 512 513 /* not reached */ 514 exit(1); 515 return (NULL); 516 } 517 518 #ifdef __FreeBSD__ 519 void 520 fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip) 521 #else 522 void 523 fbsdrun_addcpu(struct vmctx *ctx, int newcpu, uint64_t rip, bool suspend) 524 #endif 525 { 526 int error; 527 528 #ifdef __FreeBSD__ 529 assert(fromcpu == BSP); 530 #endif 531 532 /* 533 * The 'newcpu' must be activated in the context of 'fromcpu'. If 534 * vm_activate_cpu() is delayed until newcpu's pthread starts running 535 * then vmm.ko is out-of-sync with bhyve and this can create a race 536 * with vm_suspend(). 537 */ 538 error = vm_activate_cpu(ctx, newcpu); 539 if (error != 0) 540 err(EX_OSERR, "could not activate CPU %d", newcpu); 541 542 CPU_SET_ATOMIC(newcpu, &cpumask); 543 544 #ifndef __FreeBSD__ 545 if (suspend) 546 (void) vm_suspend_cpu(ctx, newcpu); 547 #endif 548 549 /* 550 * Set up the vmexit struct to allow execution to start 551 * at the given RIP 552 */ 553 mt_vmm_info[newcpu].mt_ctx = ctx; 554 mt_vmm_info[newcpu].mt_vcpu = newcpu; 555 mt_vmm_info[newcpu].mt_startrip = rip; 556 557 error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL, 558 fbsdrun_start_thread, &mt_vmm_info[newcpu]); 559 assert(error == 0); 560 } 561 562 static int 563 fbsdrun_deletecpu(struct vmctx *ctx, int vcpu) 564 { 565 566 if (!CPU_ISSET(vcpu, &cpumask)) { 567 fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu); 568 exit(4); 569 } 570 571 CPU_CLR_ATOMIC(vcpu, &cpumask); 572 return (CPU_EMPTY(&cpumask)); 573 } 574 575 static void 576 vmentry_mmio_read(int vcpu, uint64_t gpa, uint8_t bytes, uint64_t data) 577 { 578 struct vm_entry *entry = &vmentry[vcpu]; 579 struct vm_mmio *mmio = &entry->u.mmio; 580 581 assert(entry->cmd == VEC_DEFAULT); 582 583 entry->cmd = VEC_FULFILL_MMIO; 584 mmio->bytes = bytes; 585 mmio->read = 1; 586 mmio->gpa = gpa; 587 mmio->data = data; 588 } 589 590 static void 591 vmentry_mmio_write(int vcpu, uint64_t gpa, uint8_t bytes) 592 { 593 struct vm_entry *entry = &vmentry[vcpu]; 594 struct vm_mmio *mmio = &entry->u.mmio; 595 596 assert(entry->cmd == VEC_DEFAULT); 597 598 entry->cmd = VEC_FULFILL_MMIO; 599 mmio->bytes = bytes; 600 mmio->read = 0; 601 mmio->gpa = gpa; 602 mmio->data = 0; 603 } 604 605 static void 606 vmentry_inout_read(int vcpu, uint16_t port, uint8_t bytes, uint32_t data) 607 { 608 struct vm_entry *entry = &vmentry[vcpu]; 609 struct vm_inout *inout = &entry->u.inout; 610 611 assert(entry->cmd == VEC_DEFAULT); 612 613 entry->cmd = VEC_FULFILL_INOUT; 614 inout->bytes = bytes; 615 inout->flags = INOUT_IN; 616 inout->port = port; 617 inout->eax = data; 618 } 619 620 static void 621 vmentry_inout_write(int vcpu, uint16_t port, uint8_t bytes) 622 { 623 struct vm_entry *entry = &vmentry[vcpu]; 624 struct vm_inout *inout = &entry->u.inout; 625 626 assert(entry->cmd == VEC_DEFAULT); 627 628 entry->cmd = VEC_FULFILL_INOUT; 629 inout->bytes = bytes; 630 inout->flags = 0; 631 inout->port = port; 632 inout->eax = 0; 633 } 634 635 static int 636 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu, 637 uint32_t eax) 638 { 639 #if BHYVE_DEBUG 640 /* 641 * put guest-driven debug here 642 */ 643 #endif 644 return (VMEXIT_CONTINUE); 645 } 646 647 static int 648 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 649 { 650 int error; 651 int vcpu; 652 struct vm_inout inout; 653 bool in; 654 uint8_t bytes; 655 656 stats.vmexit_inout++; 657 658 vcpu = *pvcpu; 659 inout = vme->u.inout; 660 in = (inout.flags & INOUT_IN) != 0; 661 bytes = inout.bytes; 662 663 /* Extra-special case of host notifications */ 664 if (!in && inout.port == GUEST_NIO_PORT) { 665 error = vmexit_handle_notify(ctx, vme, pvcpu, inout.eax); 666 vmentry_inout_write(vcpu, inout.port, bytes); 667 return (error); 668 } 669 670 error = emulate_inout(ctx, vcpu, &inout, strictio != 0); 671 if (error) { 672 fprintf(stderr, "Unhandled %s%c 0x%04x at 0x%lx\n", 673 in ? "in" : "out", 674 bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), 675 inout.port, vmexit->rip); 676 return (VMEXIT_ABORT); 677 } else { 678 /* 679 * Communicate the status of the inout operation back to the 680 * in-kernel instruction emulation. 681 */ 682 if (in) { 683 vmentry_inout_read(vcpu, inout.port, bytes, inout.eax); 684 } else { 685 vmentry_inout_write(vcpu, inout.port, bytes); 686 } 687 return (VMEXIT_CONTINUE); 688 } 689 } 690 691 static int 692 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 693 { 694 uint64_t val; 695 uint32_t eax, edx; 696 int error; 697 698 val = 0; 699 error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val); 700 if (error != 0) { 701 fprintf(stderr, "rdmsr to register %#x on vcpu %d\n", 702 vme->u.msr.code, *pvcpu); 703 if (strictmsr) { 704 vm_inject_gp(ctx, *pvcpu); 705 return (VMEXIT_CONTINUE); 706 } 707 } 708 709 eax = val; 710 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax); 711 assert(error == 0); 712 713 edx = val >> 32; 714 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx); 715 assert(error == 0); 716 717 return (VMEXIT_CONTINUE); 718 } 719 720 static int 721 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 722 { 723 int error; 724 725 error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval); 726 if (error != 0) { 727 fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n", 728 vme->u.msr.code, vme->u.msr.wval, *pvcpu); 729 if (strictmsr) { 730 vm_inject_gp(ctx, *pvcpu); 731 return (VMEXIT_CONTINUE); 732 } 733 } 734 return (VMEXIT_CONTINUE); 735 } 736 737 #ifdef __FreeBSD__ 738 static int 739 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 740 { 741 742 (void)spinup_ap(ctx, *pvcpu, 743 vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip); 744 745 return (VMEXIT_CONTINUE); 746 } 747 #else 748 static int 749 vmexit_run_state(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 750 { 751 /* 752 * Run-state transitions (INIT, SIPI, etc) are handled in-kernel, so an 753 * exit to userspace with that code is not expected. 754 */ 755 fprintf(stderr, "unexpected run-state VM exit"); 756 return (VMEXIT_ABORT); 757 } 758 #endif /* __FreeBSD__ */ 759 760 #ifdef __FreeBSD__ 761 #define DEBUG_EPT_MISCONFIG 762 #else 763 /* EPT misconfig debugging not possible now that raw VMCS access is gone */ 764 #endif 765 766 #ifdef DEBUG_EPT_MISCONFIG 767 #define VMCS_GUEST_PHYSICAL_ADDRESS 0x00002400 768 769 static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4]; 770 static int ept_misconfig_ptenum; 771 #endif 772 773 static const char * 774 vmexit_vmx_desc(uint32_t exit_reason) 775 { 776 777 if (exit_reason >= nitems(vmx_exit_reason_desc) || 778 vmx_exit_reason_desc[exit_reason] == NULL) 779 return ("Unknown"); 780 return (vmx_exit_reason_desc[exit_reason]); 781 } 782 783 static int 784 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 785 { 786 787 fprintf(stderr, "vm exit[%d]\n", *pvcpu); 788 fprintf(stderr, "\treason\t\tVMX\n"); 789 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip); 790 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length); 791 fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status); 792 fprintf(stderr, "\texit_reason\t%u (%s)\n", vmexit->u.vmx.exit_reason, 793 vmexit_vmx_desc(vmexit->u.vmx.exit_reason)); 794 fprintf(stderr, "\tqualification\t0x%016lx\n", 795 vmexit->u.vmx.exit_qualification); 796 fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type); 797 fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error); 798 #ifdef DEBUG_EPT_MISCONFIG 799 if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) { 800 vm_get_register(ctx, *pvcpu, 801 VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS), 802 &ept_misconfig_gpa); 803 vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte, 804 &ept_misconfig_ptenum); 805 fprintf(stderr, "\tEPT misconfiguration:\n"); 806 fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa); 807 fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n", 808 ept_misconfig_ptenum, ept_misconfig_pte[0], 809 ept_misconfig_pte[1], ept_misconfig_pte[2], 810 ept_misconfig_pte[3]); 811 } 812 #endif /* DEBUG_EPT_MISCONFIG */ 813 return (VMEXIT_ABORT); 814 } 815 816 static int 817 vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 818 { 819 820 fprintf(stderr, "vm exit[%d]\n", *pvcpu); 821 fprintf(stderr, "\treason\t\tSVM\n"); 822 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip); 823 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length); 824 fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode); 825 fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1); 826 fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2); 827 return (VMEXIT_ABORT); 828 } 829 830 static int 831 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 832 { 833 834 assert(vmexit->inst_length == 0); 835 836 stats.vmexit_bogus++; 837 838 return (VMEXIT_CONTINUE); 839 } 840 841 static int 842 vmexit_reqidle(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 843 { 844 845 assert(vmexit->inst_length == 0); 846 847 stats.vmexit_reqidle++; 848 849 return (VMEXIT_CONTINUE); 850 } 851 852 static int 853 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 854 { 855 856 stats.vmexit_hlt++; 857 858 /* 859 * Just continue execution with the next instruction. We use 860 * the HLT VM exit as a way to be friendly with the host 861 * scheduler. 862 */ 863 return (VMEXIT_CONTINUE); 864 } 865 866 static int 867 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 868 { 869 870 stats.vmexit_pause++; 871 872 return (VMEXIT_CONTINUE); 873 } 874 875 static int 876 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 877 { 878 879 assert(vmexit->inst_length == 0); 880 881 stats.vmexit_mtrap++; 882 883 if (gdb_port == 0) { 884 fprintf(stderr, "vm_loop: unexpected VMEXIT_MTRAP\n"); 885 exit(4); 886 } 887 gdb_cpu_mtrap(*pvcpu); 888 return (VMEXIT_CONTINUE); 889 } 890 891 static int 892 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 893 { 894 uint8_t i, valid; 895 896 fprintf(stderr, "Failed to emulate instruction sequence "); 897 898 valid = vmexit->u.inst_emul.num_valid; 899 if (valid != 0) { 900 assert(valid <= sizeof (vmexit->u.inst_emul.inst)); 901 fprintf(stderr, "["); 902 for (i = 0; i < valid; i++) { 903 if (i == 0) { 904 fprintf(stderr, "%02x", 905 vmexit->u.inst_emul.inst[i]); 906 } else { 907 fprintf(stderr, ", %02x", 908 vmexit->u.inst_emul.inst[i]); 909 } 910 } 911 fprintf(stderr, "] "); 912 } 913 fprintf(stderr, "@ %rip = %x\n", vmexit->rip); 914 915 return (VMEXIT_ABORT); 916 } 917 918 static int 919 vmexit_mmio(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 920 { 921 int vcpu, err; 922 struct vm_mmio mmio; 923 bool is_read; 924 925 stats.vmexit_mmio++; 926 927 vcpu = *pvcpu; 928 mmio = vmexit->u.mmio; 929 is_read = (mmio.read != 0); 930 931 err = emulate_mem(ctx, vcpu, &mmio); 932 933 if (err == ESRCH) { 934 fprintf(stderr, "Unhandled memory access to 0x%lx\n", mmio.gpa); 935 stats.mmio_unhandled++; 936 937 /* 938 * Access to non-existent physical addresses is not likely to 939 * result in fatal errors on hardware machines, but rather reads 940 * of all-ones or discarded-but-acknowledged writes. 941 */ 942 mmio.data = ~0UL; 943 err = 0; 944 } 945 946 if (err == 0) { 947 if (is_read) { 948 vmentry_mmio_read(vcpu, mmio.gpa, mmio.bytes, 949 mmio.data); 950 } else { 951 vmentry_mmio_write(vcpu, mmio.gpa, mmio.bytes); 952 } 953 return (VMEXIT_CONTINUE); 954 } 955 956 fprintf(stderr, "Unhandled mmio error to 0x%lx: %d\n", mmio.gpa, err); 957 return (VMEXIT_ABORT); 958 } 959 960 static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER; 961 static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER; 962 963 static int 964 vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 965 { 966 enum vm_suspend_how how; 967 968 how = vmexit->u.suspended.how; 969 970 fbsdrun_deletecpu(ctx, *pvcpu); 971 972 if (*pvcpu != BSP) { 973 pthread_mutex_lock(&resetcpu_mtx); 974 pthread_cond_signal(&resetcpu_cond); 975 pthread_mutex_unlock(&resetcpu_mtx); 976 pthread_exit(NULL); 977 } 978 979 pthread_mutex_lock(&resetcpu_mtx); 980 while (!CPU_EMPTY(&cpumask)) { 981 pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx); 982 } 983 pthread_mutex_unlock(&resetcpu_mtx); 984 985 switch (how) { 986 case VM_SUSPEND_RESET: 987 exit(0); 988 case VM_SUSPEND_POWEROFF: 989 if (destroy_on_poweroff) 990 vm_destroy(ctx); 991 exit(1); 992 case VM_SUSPEND_HALT: 993 exit(2); 994 case VM_SUSPEND_TRIPLEFAULT: 995 exit(3); 996 default: 997 fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how); 998 exit(100); 999 } 1000 return (0); /* NOTREACHED */ 1001 } 1002 1003 static int 1004 vmexit_debug(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 1005 { 1006 1007 if (gdb_port == 0) { 1008 fprintf(stderr, "vm_loop: unexpected VMEXIT_DEBUG\n"); 1009 exit(4); 1010 } 1011 gdb_cpu_suspend(*pvcpu); 1012 return (VMEXIT_CONTINUE); 1013 } 1014 1015 static int 1016 vmexit_breakpoint(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 1017 { 1018 1019 if (gdb_port == 0) { 1020 fprintf(stderr, "vm_loop: unexpected VMEXIT_DEBUG\n"); 1021 exit(4); 1022 } 1023 gdb_cpu_breakpoint(*pvcpu, vmexit); 1024 return (VMEXIT_CONTINUE); 1025 } 1026 1027 static vmexit_handler_t handler[VM_EXITCODE_MAX] = { 1028 [VM_EXITCODE_INOUT] = vmexit_inout, 1029 [VM_EXITCODE_MMIO] = vmexit_mmio, 1030 [VM_EXITCODE_VMX] = vmexit_vmx, 1031 [VM_EXITCODE_SVM] = vmexit_svm, 1032 [VM_EXITCODE_BOGUS] = vmexit_bogus, 1033 [VM_EXITCODE_REQIDLE] = vmexit_reqidle, 1034 [VM_EXITCODE_RDMSR] = vmexit_rdmsr, 1035 [VM_EXITCODE_WRMSR] = vmexit_wrmsr, 1036 [VM_EXITCODE_MTRAP] = vmexit_mtrap, 1037 [VM_EXITCODE_INST_EMUL] = vmexit_inst_emul, 1038 #ifdef __FreeBSD__ 1039 [VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap, 1040 #else 1041 [VM_EXITCODE_RUN_STATE] = vmexit_run_state, 1042 #endif 1043 [VM_EXITCODE_SUSPENDED] = vmexit_suspend, 1044 [VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch, 1045 [VM_EXITCODE_DEBUG] = vmexit_debug, 1046 [VM_EXITCODE_BPT] = vmexit_breakpoint, 1047 }; 1048 1049 static void 1050 vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip) 1051 { 1052 int error, rc; 1053 enum vm_exitcode exitcode; 1054 cpuset_t active_cpus; 1055 struct vm_exit *vexit; 1056 struct vm_entry *ventry; 1057 1058 #ifdef __FreeBSD__ 1059 if (vcpumap[vcpu] != NULL) { 1060 error = pthread_setaffinity_np(pthread_self(), 1061 sizeof(cpuset_t), vcpumap[vcpu]); 1062 assert(error == 0); 1063 } 1064 #endif 1065 error = vm_active_cpus(ctx, &active_cpus); 1066 assert(CPU_ISSET(vcpu, &active_cpus)); 1067 1068 error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip); 1069 assert(error == 0); 1070 1071 ventry = &vmentry[vcpu]; 1072 vexit = &vmexit[vcpu]; 1073 1074 while (1) { 1075 error = vm_run(ctx, vcpu, ventry, vexit); 1076 if (error != 0) 1077 break; 1078 1079 if (ventry->cmd != VEC_DEFAULT) { 1080 /* 1081 * Discard any lingering entry state after it has been 1082 * submitted via vm_run(). 1083 */ 1084 bzero(ventry, sizeof (*ventry)); 1085 } 1086 1087 exitcode = vexit->exitcode; 1088 if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) { 1089 fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n", 1090 exitcode); 1091 exit(4); 1092 } 1093 1094 rc = (*handler[exitcode])(ctx, vexit, &vcpu); 1095 1096 switch (rc) { 1097 case VMEXIT_CONTINUE: 1098 break; 1099 case VMEXIT_ABORT: 1100 abort(); 1101 default: 1102 exit(4); 1103 } 1104 } 1105 fprintf(stderr, "vm_run error %d, errno %d\n", error, errno); 1106 } 1107 1108 static int 1109 num_vcpus_allowed(struct vmctx *ctx) 1110 { 1111 #ifdef __FreeBSD__ 1112 int tmp, error; 1113 1114 error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp); 1115 1116 /* 1117 * The guest is allowed to spinup more than one processor only if the 1118 * UNRESTRICTED_GUEST capability is available. 1119 */ 1120 if (error == 0) 1121 return (VM_MAXCPU); 1122 else 1123 return (1); 1124 #else 1125 /* Unrestricted Guest is always enabled on illumos */ 1126 return (VM_MAXCPU); 1127 #endif /* __FreeBSD__ */ 1128 } 1129 1130 void 1131 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu) 1132 { 1133 int err, tmp; 1134 1135 if (fbsdrun_vmexit_on_hlt()) { 1136 err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp); 1137 if (err < 0) { 1138 fprintf(stderr, "VM exit on HLT not supported\n"); 1139 exit(4); 1140 } 1141 vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1); 1142 if (cpu == BSP) 1143 handler[VM_EXITCODE_HLT] = vmexit_hlt; 1144 } 1145 1146 if (fbsdrun_vmexit_on_pause()) { 1147 /* 1148 * pause exit support required for this mode 1149 */ 1150 err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp); 1151 if (err < 0) { 1152 fprintf(stderr, 1153 "SMP mux requested, no pause support\n"); 1154 exit(4); 1155 } 1156 vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1); 1157 if (cpu == BSP) 1158 handler[VM_EXITCODE_PAUSE] = vmexit_pause; 1159 } 1160 1161 if (x2apic_mode) 1162 err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED); 1163 else 1164 err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED); 1165 1166 if (err) { 1167 fprintf(stderr, "Unable to set x2apic state (%d)\n", err); 1168 exit(4); 1169 } 1170 1171 #ifdef __FreeBSD__ 1172 vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1); 1173 #endif 1174 } 1175 1176 static struct vmctx * 1177 do_open(const char *vmname) 1178 { 1179 struct vmctx *ctx; 1180 int error; 1181 bool reinit, romboot; 1182 #ifndef WITHOUT_CAPSICUM 1183 cap_rights_t rights; 1184 const cap_ioctl_t *cmds; 1185 size_t ncmds; 1186 #endif 1187 1188 reinit = romboot = false; 1189 1190 if (lpc_bootrom()) 1191 romboot = true; 1192 1193 error = vm_create(vmname); 1194 if (error) { 1195 if (errno == EEXIST) { 1196 if (romboot) { 1197 reinit = true; 1198 } else { 1199 /* 1200 * The virtual machine has been setup by the 1201 * userspace bootloader. 1202 */ 1203 } 1204 } else { 1205 perror("vm_create"); 1206 exit(4); 1207 } 1208 } else { 1209 if (!romboot) { 1210 /* 1211 * If the virtual machine was just created then a 1212 * bootrom must be configured to boot it. 1213 */ 1214 fprintf(stderr, "virtual machine cannot be booted\n"); 1215 exit(4); 1216 } 1217 } 1218 1219 ctx = vm_open(vmname); 1220 if (ctx == NULL) { 1221 perror("vm_open"); 1222 exit(4); 1223 } 1224 1225 #ifndef WITHOUT_CAPSICUM 1226 cap_rights_init(&rights, CAP_IOCTL, CAP_MMAP_RW); 1227 if (caph_rights_limit(vm_get_device_fd(ctx), &rights) == -1) 1228 errx(EX_OSERR, "Unable to apply rights for sandbox"); 1229 vm_get_ioctls(&ncmds); 1230 cmds = vm_get_ioctls(NULL); 1231 if (cmds == NULL) 1232 errx(EX_OSERR, "out of memory"); 1233 if (caph_ioctls_limit(vm_get_device_fd(ctx), cmds, ncmds) == -1) 1234 errx(EX_OSERR, "Unable to apply rights for sandbox"); 1235 free((cap_ioctl_t *)cmds); 1236 #endif 1237 1238 if (reinit) { 1239 error = vm_reinit(ctx); 1240 if (error) { 1241 perror("vm_reinit"); 1242 exit(4); 1243 } 1244 } 1245 error = vm_set_topology(ctx, sockets, cores, threads, maxcpus); 1246 if (error) 1247 errx(EX_OSERR, "vm_set_topology"); 1248 return (ctx); 1249 } 1250 1251 int 1252 main(int argc, char *argv[]) 1253 { 1254 int c, error, dbg_port, err, bvmcons; 1255 int max_vcpus, mptgen, memflags; 1256 int rtc_localtime; 1257 bool gdb_stop; 1258 #ifndef __FreeBSD__ 1259 bool suspend = false; 1260 #endif 1261 struct vmctx *ctx; 1262 uint64_t rip; 1263 size_t memsize; 1264 char *optstr; 1265 1266 bvmcons = 0; 1267 progname = basename(argv[0]); 1268 dbg_port = 0; 1269 gdb_stop = false; 1270 guest_ncpus = 1; 1271 sockets = cores = threads = 1; 1272 maxcpus = 0; 1273 memsize = 256 * MB; 1274 mptgen = 1; 1275 rtc_localtime = 1; 1276 memflags = 0; 1277 1278 #ifdef __FreeBSD__ 1279 optstr = "abehuwxACDHIPSWYp:g:G:c:s:m:l:B:U:"; 1280 #else 1281 optstr = "abdehuwxACDHIPSWYg:G:c:s:m:l:B:U:"; 1282 #endif 1283 while ((c = getopt(argc, argv, optstr)) != -1) { 1284 switch (c) { 1285 case 'a': 1286 x2apic_mode = 0; 1287 break; 1288 case 'A': 1289 acpi = 1; 1290 break; 1291 case 'b': 1292 bvmcons = 1; 1293 break; 1294 case 'D': 1295 destroy_on_poweroff = 1; 1296 break; 1297 case 'B': 1298 if (smbios_parse(optarg) != 0) { 1299 errx(EX_USAGE, "invalid SMBIOS " 1300 "configuration '%s'", optarg); 1301 } 1302 break; 1303 #ifndef __FreeBSD__ 1304 case 'd': 1305 suspend = true; 1306 break; 1307 #else 1308 case 'p': 1309 if (pincpu_parse(optarg) != 0) { 1310 errx(EX_USAGE, "invalid vcpu pinning " 1311 "configuration '%s'", optarg); 1312 } 1313 break; 1314 #endif 1315 case 'c': 1316 if (topology_parse(optarg) != 0) { 1317 errx(EX_USAGE, "invalid cpu topology " 1318 "'%s'", optarg); 1319 } 1320 break; 1321 case 'C': 1322 memflags |= VM_MEM_F_INCORE; 1323 break; 1324 case 'g': 1325 dbg_port = atoi(optarg); 1326 break; 1327 case 'G': 1328 if (optarg[0] == 'w') { 1329 gdb_stop = true; 1330 optarg++; 1331 } 1332 gdb_port = atoi(optarg); 1333 break; 1334 case 'l': 1335 if (strncmp(optarg, "help", strlen(optarg)) == 0) { 1336 lpc_print_supported_devices(); 1337 exit(0); 1338 } else if (lpc_device_parse(optarg) != 0) { 1339 errx(EX_USAGE, "invalid lpc device " 1340 "configuration '%s'", optarg); 1341 } 1342 break; 1343 case 's': 1344 if (strncmp(optarg, "help", strlen(optarg)) == 0) { 1345 pci_print_supported_devices(); 1346 exit(0); 1347 } else if (pci_parse_slot(optarg) != 0) 1348 exit(4); 1349 else 1350 break; 1351 case 'S': 1352 memflags |= VM_MEM_F_WIRED; 1353 break; 1354 case 'm': 1355 error = vm_parse_memsize(optarg, &memsize); 1356 if (error) 1357 errx(EX_USAGE, "invalid memsize '%s'", optarg); 1358 break; 1359 case 'H': 1360 guest_vmexit_on_hlt = 1; 1361 break; 1362 case 'I': 1363 /* 1364 * The "-I" option was used to add an ioapic to the 1365 * virtual machine. 1366 * 1367 * An ioapic is now provided unconditionally for each 1368 * virtual machine and this option is now deprecated. 1369 */ 1370 break; 1371 case 'P': 1372 guest_vmexit_on_pause = 1; 1373 break; 1374 case 'e': 1375 strictio = 1; 1376 break; 1377 case 'u': 1378 rtc_localtime = 0; 1379 break; 1380 case 'U': 1381 guest_uuid_str = optarg; 1382 break; 1383 case 'w': 1384 strictmsr = 0; 1385 break; 1386 case 'W': 1387 virtio_msix = 0; 1388 break; 1389 case 'x': 1390 x2apic_mode = 1; 1391 break; 1392 case 'Y': 1393 mptgen = 0; 1394 break; 1395 case 'h': 1396 usage(0); 1397 default: 1398 usage(1); 1399 } 1400 } 1401 argc -= optind; 1402 argv += optind; 1403 1404 if (argc != 1) 1405 usage(1); 1406 1407 vmname = argv[0]; 1408 ctx = do_open(vmname); 1409 1410 max_vcpus = num_vcpus_allowed(ctx); 1411 if (guest_ncpus > max_vcpus) { 1412 fprintf(stderr, "%d vCPUs requested but only %d available\n", 1413 guest_ncpus, max_vcpus); 1414 exit(4); 1415 } 1416 1417 fbsdrun_set_capabilities(ctx, BSP); 1418 1419 vm_set_memflags(ctx, memflags); 1420 #ifdef __FreeBSD__ 1421 err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL); 1422 #else 1423 do { 1424 errno = 0; 1425 err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL); 1426 error = errno; 1427 if (err != 0 && error == ENOMEM) { 1428 (void) fprintf(stderr, "Unable to allocate memory " 1429 "(%llu), retrying in 1 second\n", memsize); 1430 sleep(1); 1431 } 1432 } while (error == ENOMEM); 1433 #endif 1434 if (err) { 1435 fprintf(stderr, "Unable to setup memory (%d)\n", errno); 1436 exit(4); 1437 } 1438 1439 error = init_msr(); 1440 if (error) { 1441 fprintf(stderr, "init_msr error %d", error); 1442 exit(4); 1443 } 1444 1445 init_mem(); 1446 init_inout(); 1447 #ifdef __FreeBSD__ 1448 kernemu_dev_init(); 1449 #endif 1450 init_bootrom(ctx); 1451 atkbdc_init(ctx); 1452 pci_irq_init(ctx); 1453 ioapic_init(ctx); 1454 1455 rtc_init(ctx, rtc_localtime); 1456 sci_init(ctx); 1457 #ifndef __FreeBSD__ 1458 pmtmr_init(ctx); 1459 #endif 1460 1461 /* 1462 * Exit if a device emulation finds an error in its initilization 1463 */ 1464 if (init_pci(ctx) != 0) { 1465 perror("device emulation initialization error"); 1466 exit(4); 1467 } 1468 1469 /* 1470 * Initialize after PCI, to allow a bootrom file to reserve the high 1471 * region. 1472 */ 1473 if (acpi) 1474 vmgenc_init(ctx); 1475 1476 if (dbg_port != 0) 1477 init_dbgport(dbg_port); 1478 1479 #ifdef __FreeBSD__ 1480 if (gdb_port != 0) 1481 init_gdb(ctx, gdb_port, gdb_stop); 1482 #else 1483 if (gdb_port < 0) { 1484 /* 1485 * Set up the internal gdb state needed for basic debugging, but 1486 * skip the step of listening on a port for the GDB server. 1487 */ 1488 init_mdb(ctx, gdb_stop); 1489 } else if (gdb_port != 0) { 1490 init_gdb(ctx, gdb_port, gdb_stop); 1491 } 1492 #endif 1493 1494 if (bvmcons) 1495 init_bvmcons(); 1496 1497 vga_init(1); 1498 1499 if (lpc_bootrom()) { 1500 #ifdef __FreeBSD__ 1501 if (vm_set_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, 1)) { 1502 fprintf(stderr, "ROM boot failed: unrestricted guest " 1503 "capability not available\n"); 1504 exit(4); 1505 } 1506 #else 1507 /* Unrestricted Guest is always enabled on illumos */ 1508 #endif 1509 error = vcpu_reset(ctx, BSP); 1510 assert(error == 0); 1511 } 1512 1513 error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip); 1514 assert(error == 0); 1515 1516 /* 1517 * build the guest tables, MP etc. 1518 */ 1519 if (mptgen) { 1520 error = mptable_build(ctx, guest_ncpus); 1521 if (error) { 1522 perror("error to build the guest tables"); 1523 exit(4); 1524 } 1525 } 1526 1527 error = smbios_build(ctx); 1528 assert(error == 0); 1529 1530 if (acpi) { 1531 error = acpi_build(ctx, guest_ncpus); 1532 assert(error == 0); 1533 } 1534 1535 if (lpc_bootrom()) 1536 fwctl_init(); 1537 1538 /* 1539 * Change the proc title to include the VM name. 1540 */ 1541 setproctitle("%s", vmname); 1542 1543 #ifndef WITHOUT_CAPSICUM 1544 caph_cache_catpages(); 1545 1546 if (caph_limit_stdout() == -1 || caph_limit_stderr() == -1) 1547 errx(EX_OSERR, "Unable to apply rights for sandbox"); 1548 1549 if (caph_enter() == -1) 1550 errx(EX_OSERR, "cap_enter() failed"); 1551 #endif 1552 1553 #ifdef __FreeBSD__ 1554 /* 1555 * Add CPU 0 1556 */ 1557 fbsdrun_addcpu(ctx, BSP, BSP, rip); 1558 #else 1559 /* Set BSP to run (unlike the APs which wait for INIT) */ 1560 error = vm_set_run_state(ctx, BSP, VRS_RUN, 0); 1561 assert(error == 0); 1562 fbsdrun_addcpu(ctx, BSP, rip, suspend); 1563 1564 /* Add subsequent CPUs, which will wait until INIT/SIPI-ed */ 1565 for (uint_t i = 1; i < guest_ncpus; i++) { 1566 spinup_halted_ap(ctx, i); 1567 } 1568 #endif 1569 /* 1570 * Head off to the main event dispatch loop 1571 */ 1572 mevent_dispatch(); 1573 1574 exit(4); 1575 } 1576