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