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 newcpu, uint64_t rip, bool suspend) 518 #endif 519 { 520 int error; 521 522 #ifdef __FreeBSD__ 523 assert(fromcpu == BSP); 524 #endif 525 526 /* 527 * The 'newcpu' must be activated in the context of 'fromcpu'. If 528 * vm_activate_cpu() is delayed until newcpu's pthread starts running 529 * then vmm.ko is out-of-sync with bhyve and this can create a race 530 * with vm_suspend(). 531 */ 532 error = vm_activate_cpu(ctx, newcpu); 533 if (error != 0) 534 err(EX_OSERR, "could not activate CPU %d", newcpu); 535 536 CPU_SET_ATOMIC(newcpu, &cpumask); 537 538 #ifndef __FreeBSD__ 539 if (suspend) 540 (void) vm_suspend_cpu(ctx, newcpu); 541 #endif 542 543 /* 544 * Set up the vmexit struct to allow execution to start 545 * at the given RIP 546 */ 547 mt_vmm_info[newcpu].mt_ctx = ctx; 548 mt_vmm_info[newcpu].mt_vcpu = newcpu; 549 mt_vmm_info[newcpu].mt_startrip = rip; 550 551 error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL, 552 fbsdrun_start_thread, &mt_vmm_info[newcpu]); 553 assert(error == 0); 554 } 555 556 static int 557 fbsdrun_deletecpu(struct vmctx *ctx, int vcpu) 558 { 559 560 if (!CPU_ISSET(vcpu, &cpumask)) { 561 fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu); 562 exit(4); 563 } 564 565 CPU_CLR_ATOMIC(vcpu, &cpumask); 566 return (CPU_EMPTY(&cpumask)); 567 } 568 569 static void 570 vmentry_mmio_read(int vcpu, uint64_t gpa, uint8_t bytes, uint64_t data) 571 { 572 struct vm_entry *entry = &vmentry[vcpu]; 573 struct vm_mmio *mmio = &entry->u.mmio; 574 575 assert(entry->cmd == VEC_DEFAULT); 576 577 entry->cmd = VEC_FULFILL_MMIO; 578 mmio->bytes = bytes; 579 mmio->read = 1; 580 mmio->gpa = gpa; 581 mmio->data = data; 582 } 583 584 static void 585 vmentry_mmio_write(int vcpu, uint64_t gpa, uint8_t bytes) 586 { 587 struct vm_entry *entry = &vmentry[vcpu]; 588 struct vm_mmio *mmio = &entry->u.mmio; 589 590 assert(entry->cmd == VEC_DEFAULT); 591 592 entry->cmd = VEC_FULFILL_MMIO; 593 mmio->bytes = bytes; 594 mmio->read = 0; 595 mmio->gpa = gpa; 596 mmio->data = 0; 597 } 598 599 static void 600 vmentry_inout_read(int vcpu, uint16_t port, uint8_t bytes, uint32_t data) 601 { 602 struct vm_entry *entry = &vmentry[vcpu]; 603 struct vm_inout *inout = &entry->u.inout; 604 605 assert(entry->cmd == VEC_DEFAULT); 606 607 entry->cmd = VEC_FULFILL_INOUT; 608 inout->bytes = bytes; 609 inout->flags = INOUT_IN; 610 inout->port = port; 611 inout->eax = data; 612 } 613 614 static void 615 vmentry_inout_write(int vcpu, uint16_t port, uint8_t bytes) 616 { 617 struct vm_entry *entry = &vmentry[vcpu]; 618 struct vm_inout *inout = &entry->u.inout; 619 620 assert(entry->cmd == VEC_DEFAULT); 621 622 entry->cmd = VEC_FULFILL_INOUT; 623 inout->bytes = bytes; 624 inout->flags = 0; 625 inout->port = port; 626 inout->eax = 0; 627 } 628 629 static int 630 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu, 631 uint32_t eax) 632 { 633 #if BHYVE_DEBUG 634 /* 635 * put guest-driven debug here 636 */ 637 #endif 638 return (VMEXIT_CONTINUE); 639 } 640 641 static int 642 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 643 { 644 int error; 645 int vcpu; 646 struct vm_inout inout; 647 bool in; 648 uint8_t bytes; 649 650 stats.vmexit_inout++; 651 652 vcpu = *pvcpu; 653 inout = vme->u.inout; 654 in = (inout.flags & INOUT_IN) != 0; 655 bytes = inout.bytes; 656 657 /* Extra-special case of host notifications */ 658 if (!in && inout.port == GUEST_NIO_PORT) { 659 error = vmexit_handle_notify(ctx, vme, pvcpu, inout.eax); 660 vmentry_inout_write(vcpu, inout.port, bytes); 661 return (error); 662 } 663 664 error = emulate_inout(ctx, vcpu, &inout, strictio != 0); 665 if (error) { 666 fprintf(stderr, "Unhandled %s%c 0x%04x at 0x%lx\n", 667 in ? "in" : "out", 668 bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), 669 inout.port, vmexit->rip); 670 return (VMEXIT_ABORT); 671 } else { 672 /* 673 * Communicate the status of the inout operation back to the 674 * in-kernel instruction emulation. 675 */ 676 if (in) { 677 vmentry_inout_read(vcpu, inout.port, bytes, inout.eax); 678 } else { 679 vmentry_inout_write(vcpu, inout.port, bytes); 680 } 681 return (VMEXIT_CONTINUE); 682 } 683 } 684 685 static int 686 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 687 { 688 uint64_t val; 689 uint32_t eax, edx; 690 int error; 691 692 val = 0; 693 error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val); 694 if (error != 0) { 695 fprintf(stderr, "rdmsr to register %#x on vcpu %d\n", 696 vme->u.msr.code, *pvcpu); 697 if (strictmsr) { 698 vm_inject_gp(ctx, *pvcpu); 699 return (VMEXIT_CONTINUE); 700 } 701 } 702 703 eax = val; 704 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax); 705 assert(error == 0); 706 707 edx = val >> 32; 708 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx); 709 assert(error == 0); 710 711 return (VMEXIT_CONTINUE); 712 } 713 714 static int 715 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 716 { 717 int error; 718 719 error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval); 720 if (error != 0) { 721 fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n", 722 vme->u.msr.code, vme->u.msr.wval, *pvcpu); 723 if (strictmsr) { 724 vm_inject_gp(ctx, *pvcpu); 725 return (VMEXIT_CONTINUE); 726 } 727 } 728 return (VMEXIT_CONTINUE); 729 } 730 731 #ifdef __FreeBSD__ 732 static int 733 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 734 { 735 736 (void)spinup_ap(ctx, *pvcpu, 737 vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip); 738 739 return (VMEXIT_CONTINUE); 740 } 741 #else 742 static int 743 vmexit_run_state(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 744 { 745 /* 746 * Run-state transitions (INIT, SIPI, etc) are handled in-kernel, so an 747 * exit to userspace with that code is not expected. 748 */ 749 fprintf(stderr, "unexpected run-state VM exit"); 750 return (VMEXIT_ABORT); 751 } 752 #endif /* __FreeBSD__ */ 753 754 #ifdef __FreeBSD__ 755 #define DEBUG_EPT_MISCONFIG 756 #else 757 /* EPT misconfig debugging not possible now that raw VMCS access is gone */ 758 #endif 759 760 #ifdef DEBUG_EPT_MISCONFIG 761 #define VMCS_GUEST_PHYSICAL_ADDRESS 0x00002400 762 763 static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4]; 764 static int ept_misconfig_ptenum; 765 #endif 766 767 static const char * 768 vmexit_vmx_desc(uint32_t exit_reason) 769 { 770 771 if (exit_reason >= nitems(vmx_exit_reason_desc) || 772 vmx_exit_reason_desc[exit_reason] == NULL) 773 return ("Unknown"); 774 return (vmx_exit_reason_desc[exit_reason]); 775 } 776 777 static int 778 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 779 { 780 781 fprintf(stderr, "vm exit[%d]\n", *pvcpu); 782 fprintf(stderr, "\treason\t\tVMX\n"); 783 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip); 784 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length); 785 fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status); 786 fprintf(stderr, "\texit_reason\t%u (%s)\n", vmexit->u.vmx.exit_reason, 787 vmexit_vmx_desc(vmexit->u.vmx.exit_reason)); 788 fprintf(stderr, "\tqualification\t0x%016lx\n", 789 vmexit->u.vmx.exit_qualification); 790 fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type); 791 fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error); 792 #ifdef DEBUG_EPT_MISCONFIG 793 if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) { 794 vm_get_register(ctx, *pvcpu, 795 VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS), 796 &ept_misconfig_gpa); 797 vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte, 798 &ept_misconfig_ptenum); 799 fprintf(stderr, "\tEPT misconfiguration:\n"); 800 fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa); 801 fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n", 802 ept_misconfig_ptenum, ept_misconfig_pte[0], 803 ept_misconfig_pte[1], ept_misconfig_pte[2], 804 ept_misconfig_pte[3]); 805 } 806 #endif /* DEBUG_EPT_MISCONFIG */ 807 return (VMEXIT_ABORT); 808 } 809 810 static int 811 vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 812 { 813 814 fprintf(stderr, "vm exit[%d]\n", *pvcpu); 815 fprintf(stderr, "\treason\t\tSVM\n"); 816 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip); 817 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length); 818 fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode); 819 fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1); 820 fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2); 821 return (VMEXIT_ABORT); 822 } 823 824 static int 825 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 826 { 827 828 assert(vmexit->inst_length == 0); 829 830 stats.vmexit_bogus++; 831 832 return (VMEXIT_CONTINUE); 833 } 834 835 static int 836 vmexit_reqidle(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 837 { 838 839 assert(vmexit->inst_length == 0); 840 841 stats.vmexit_reqidle++; 842 843 return (VMEXIT_CONTINUE); 844 } 845 846 static int 847 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 848 { 849 850 stats.vmexit_hlt++; 851 852 /* 853 * Just continue execution with the next instruction. We use 854 * the HLT VM exit as a way to be friendly with the host 855 * scheduler. 856 */ 857 return (VMEXIT_CONTINUE); 858 } 859 860 static int 861 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 862 { 863 864 stats.vmexit_pause++; 865 866 return (VMEXIT_CONTINUE); 867 } 868 869 static int 870 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 871 { 872 873 assert(vmexit->inst_length == 0); 874 875 stats.vmexit_mtrap++; 876 877 if (gdb_port == 0) { 878 fprintf(stderr, "vm_loop: unexpected VMEXIT_MTRAP\n"); 879 exit(4); 880 } 881 gdb_cpu_mtrap(*pvcpu); 882 return (VMEXIT_CONTINUE); 883 } 884 885 static int 886 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 887 { 888 uint8_t i, valid; 889 890 fprintf(stderr, "Failed to emulate instruction sequence "); 891 892 valid = vmexit->u.inst_emul.num_valid; 893 if (valid != 0) { 894 assert(valid <= sizeof (vmexit->u.inst_emul.inst)); 895 fprintf(stderr, "["); 896 for (i = 0; i < valid; i++) { 897 if (i == 0) { 898 fprintf(stderr, "%02x", 899 vmexit->u.inst_emul.inst[i]); 900 } else { 901 fprintf(stderr, ", %02x", 902 vmexit->u.inst_emul.inst[i]); 903 } 904 } 905 fprintf(stderr, "] "); 906 } 907 fprintf(stderr, "@ %rip = %x\n", vmexit->rip); 908 909 return (VMEXIT_ABORT); 910 } 911 912 static int 913 vmexit_mmio(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 914 { 915 int vcpu, err; 916 struct vm_mmio mmio; 917 bool is_read; 918 919 stats.vmexit_mmio++; 920 921 vcpu = *pvcpu; 922 mmio = vmexit->u.mmio; 923 is_read = (mmio.read != 0); 924 925 err = emulate_mem(ctx, vcpu, &mmio); 926 927 if (err == ESRCH) { 928 fprintf(stderr, "Unhandled memory access to 0x%lx\n", mmio.gpa); 929 stats.mmio_unhandled++; 930 931 /* 932 * Access to non-existent physical addresses is not likely to 933 * result in fatal errors on hardware machines, but rather reads 934 * of all-ones or discarded-but-acknowledged writes. 935 */ 936 mmio.data = ~0UL; 937 err = 0; 938 } 939 940 if (err == 0) { 941 if (is_read) { 942 vmentry_mmio_read(vcpu, mmio.gpa, mmio.bytes, 943 mmio.data); 944 } else { 945 vmentry_mmio_write(vcpu, mmio.gpa, mmio.bytes); 946 } 947 return (VMEXIT_CONTINUE); 948 } 949 950 fprintf(stderr, "Unhandled mmio error to 0x%lx: %d\n", mmio.gpa, err); 951 return (VMEXIT_ABORT); 952 } 953 954 static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER; 955 static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER; 956 957 static int 958 vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 959 { 960 enum vm_suspend_how how; 961 962 how = vmexit->u.suspended.how; 963 964 fbsdrun_deletecpu(ctx, *pvcpu); 965 966 if (*pvcpu != BSP) { 967 pthread_mutex_lock(&resetcpu_mtx); 968 pthread_cond_signal(&resetcpu_cond); 969 pthread_mutex_unlock(&resetcpu_mtx); 970 pthread_exit(NULL); 971 } 972 973 pthread_mutex_lock(&resetcpu_mtx); 974 while (!CPU_EMPTY(&cpumask)) { 975 pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx); 976 } 977 pthread_mutex_unlock(&resetcpu_mtx); 978 979 switch (how) { 980 case VM_SUSPEND_RESET: 981 exit(0); 982 case VM_SUSPEND_POWEROFF: 983 exit(1); 984 case VM_SUSPEND_HALT: 985 exit(2); 986 case VM_SUSPEND_TRIPLEFAULT: 987 exit(3); 988 default: 989 fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how); 990 exit(100); 991 } 992 return (0); /* NOTREACHED */ 993 } 994 995 static int 996 vmexit_debug(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 997 { 998 999 if (gdb_port == 0) { 1000 fprintf(stderr, "vm_loop: unexpected VMEXIT_DEBUG\n"); 1001 exit(4); 1002 } 1003 gdb_cpu_suspend(*pvcpu); 1004 return (VMEXIT_CONTINUE); 1005 } 1006 1007 static int 1008 vmexit_breakpoint(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 1009 { 1010 1011 if (gdb_port == 0) { 1012 fprintf(stderr, "vm_loop: unexpected VMEXIT_DEBUG\n"); 1013 exit(4); 1014 } 1015 gdb_cpu_breakpoint(*pvcpu, vmexit); 1016 return (VMEXIT_CONTINUE); 1017 } 1018 1019 static vmexit_handler_t handler[VM_EXITCODE_MAX] = { 1020 [VM_EXITCODE_INOUT] = vmexit_inout, 1021 [VM_EXITCODE_MMIO] = vmexit_mmio, 1022 [VM_EXITCODE_VMX] = vmexit_vmx, 1023 [VM_EXITCODE_SVM] = vmexit_svm, 1024 [VM_EXITCODE_BOGUS] = vmexit_bogus, 1025 [VM_EXITCODE_REQIDLE] = vmexit_reqidle, 1026 [VM_EXITCODE_RDMSR] = vmexit_rdmsr, 1027 [VM_EXITCODE_WRMSR] = vmexit_wrmsr, 1028 [VM_EXITCODE_MTRAP] = vmexit_mtrap, 1029 [VM_EXITCODE_INST_EMUL] = vmexit_inst_emul, 1030 #ifdef __FreeBSD__ 1031 [VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap, 1032 #else 1033 [VM_EXITCODE_RUN_STATE] = vmexit_run_state, 1034 #endif 1035 [VM_EXITCODE_SUSPENDED] = vmexit_suspend, 1036 [VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch, 1037 [VM_EXITCODE_DEBUG] = vmexit_debug, 1038 [VM_EXITCODE_BPT] = vmexit_breakpoint, 1039 }; 1040 1041 static void 1042 vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip) 1043 { 1044 int error, rc; 1045 enum vm_exitcode exitcode; 1046 cpuset_t active_cpus; 1047 struct vm_exit *vexit; 1048 struct vm_entry *ventry; 1049 1050 #ifdef __FreeBSD__ 1051 if (vcpumap[vcpu] != NULL) { 1052 error = pthread_setaffinity_np(pthread_self(), 1053 sizeof(cpuset_t), vcpumap[vcpu]); 1054 assert(error == 0); 1055 } 1056 #endif 1057 error = vm_active_cpus(ctx, &active_cpus); 1058 assert(CPU_ISSET(vcpu, &active_cpus)); 1059 1060 error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip); 1061 assert(error == 0); 1062 1063 ventry = &vmentry[vcpu]; 1064 vexit = &vmexit[vcpu]; 1065 1066 while (1) { 1067 error = vm_run(ctx, vcpu, ventry, vexit); 1068 if (error != 0) 1069 break; 1070 1071 if (ventry->cmd != VEC_DEFAULT) { 1072 /* 1073 * Discard any lingering entry state after it has been 1074 * submitted via vm_run(). 1075 */ 1076 bzero(ventry, sizeof (*ventry)); 1077 } 1078 1079 exitcode = vexit->exitcode; 1080 if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) { 1081 fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n", 1082 exitcode); 1083 exit(4); 1084 } 1085 1086 rc = (*handler[exitcode])(ctx, vexit, &vcpu); 1087 1088 switch (rc) { 1089 case VMEXIT_CONTINUE: 1090 break; 1091 case VMEXIT_ABORT: 1092 abort(); 1093 default: 1094 exit(4); 1095 } 1096 } 1097 fprintf(stderr, "vm_run error %d, errno %d\n", error, errno); 1098 } 1099 1100 static int 1101 num_vcpus_allowed(struct vmctx *ctx) 1102 { 1103 #ifdef __FreeBSD__ 1104 int tmp, error; 1105 1106 error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp); 1107 1108 /* 1109 * The guest is allowed to spinup more than one processor only if the 1110 * UNRESTRICTED_GUEST capability is available. 1111 */ 1112 if (error == 0) 1113 return (VM_MAXCPU); 1114 else 1115 return (1); 1116 #else 1117 /* Unrestricted Guest is always enabled on illumos */ 1118 return (VM_MAXCPU); 1119 #endif /* __FreeBSD__ */ 1120 } 1121 1122 void 1123 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu) 1124 { 1125 int err, tmp; 1126 1127 if (fbsdrun_vmexit_on_hlt()) { 1128 err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp); 1129 if (err < 0) { 1130 fprintf(stderr, "VM exit on HLT not supported\n"); 1131 exit(4); 1132 } 1133 vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1); 1134 if (cpu == BSP) 1135 handler[VM_EXITCODE_HLT] = vmexit_hlt; 1136 } 1137 1138 if (fbsdrun_vmexit_on_pause()) { 1139 /* 1140 * pause exit support required for this mode 1141 */ 1142 err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp); 1143 if (err < 0) { 1144 fprintf(stderr, 1145 "SMP mux requested, no pause support\n"); 1146 exit(4); 1147 } 1148 vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1); 1149 if (cpu == BSP) 1150 handler[VM_EXITCODE_PAUSE] = vmexit_pause; 1151 } 1152 1153 if (x2apic_mode) 1154 err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED); 1155 else 1156 err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED); 1157 1158 if (err) { 1159 fprintf(stderr, "Unable to set x2apic state (%d)\n", err); 1160 exit(4); 1161 } 1162 1163 #ifdef __FreeBSD__ 1164 vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1); 1165 #endif 1166 } 1167 1168 static struct vmctx * 1169 do_open(const char *vmname) 1170 { 1171 struct vmctx *ctx; 1172 int error; 1173 bool reinit, romboot; 1174 #ifndef WITHOUT_CAPSICUM 1175 cap_rights_t rights; 1176 const cap_ioctl_t *cmds; 1177 size_t ncmds; 1178 #endif 1179 1180 reinit = romboot = false; 1181 1182 if (lpc_bootrom()) 1183 romboot = true; 1184 1185 error = vm_create(vmname); 1186 if (error) { 1187 if (errno == EEXIST) { 1188 if (romboot) { 1189 reinit = true; 1190 } else { 1191 /* 1192 * The virtual machine has been setup by the 1193 * userspace bootloader. 1194 */ 1195 } 1196 } else { 1197 perror("vm_create"); 1198 exit(4); 1199 } 1200 } else { 1201 if (!romboot) { 1202 /* 1203 * If the virtual machine was just created then a 1204 * bootrom must be configured to boot it. 1205 */ 1206 fprintf(stderr, "virtual machine cannot be booted\n"); 1207 exit(4); 1208 } 1209 } 1210 1211 ctx = vm_open(vmname); 1212 if (ctx == NULL) { 1213 perror("vm_open"); 1214 exit(4); 1215 } 1216 1217 #ifndef WITHOUT_CAPSICUM 1218 cap_rights_init(&rights, CAP_IOCTL, CAP_MMAP_RW); 1219 if (caph_rights_limit(vm_get_device_fd(ctx), &rights) == -1) 1220 errx(EX_OSERR, "Unable to apply rights for sandbox"); 1221 vm_get_ioctls(&ncmds); 1222 cmds = vm_get_ioctls(NULL); 1223 if (cmds == NULL) 1224 errx(EX_OSERR, "out of memory"); 1225 if (caph_ioctls_limit(vm_get_device_fd(ctx), cmds, ncmds) == -1) 1226 errx(EX_OSERR, "Unable to apply rights for sandbox"); 1227 free((cap_ioctl_t *)cmds); 1228 #endif 1229 1230 if (reinit) { 1231 error = vm_reinit(ctx); 1232 if (error) { 1233 perror("vm_reinit"); 1234 exit(4); 1235 } 1236 } 1237 error = vm_set_topology(ctx, sockets, cores, threads, maxcpus); 1238 if (error) 1239 errx(EX_OSERR, "vm_set_topology"); 1240 return (ctx); 1241 } 1242 1243 int 1244 main(int argc, char *argv[]) 1245 { 1246 int c, error, dbg_port, err, bvmcons; 1247 int max_vcpus, mptgen, memflags; 1248 int rtc_localtime; 1249 bool gdb_stop; 1250 #ifndef __FreeBSD__ 1251 bool suspend = false; 1252 #endif 1253 struct vmctx *ctx; 1254 uint64_t rip; 1255 size_t memsize; 1256 char *optstr; 1257 1258 bvmcons = 0; 1259 progname = basename(argv[0]); 1260 dbg_port = 0; 1261 gdb_stop = false; 1262 guest_ncpus = 1; 1263 sockets = cores = threads = 1; 1264 maxcpus = 0; 1265 memsize = 256 * MB; 1266 mptgen = 1; 1267 rtc_localtime = 1; 1268 memflags = 0; 1269 1270 #ifdef __FreeBSD__ 1271 optstr = "abehuwxACHIPSWYp:g:G:c:s:m:l:B:U:"; 1272 #else 1273 optstr = "abdehuwxACHIPSWYg:G:c:s:m:l:B:U:"; 1274 #endif 1275 while ((c = getopt(argc, argv, optstr)) != -1) { 1276 switch (c) { 1277 case 'a': 1278 x2apic_mode = 0; 1279 break; 1280 case 'A': 1281 acpi = 1; 1282 break; 1283 case 'b': 1284 bvmcons = 1; 1285 break; 1286 case 'B': 1287 if (smbios_parse(optarg) != 0) { 1288 errx(EX_USAGE, "invalid SMBIOS " 1289 "configuration '%s'", optarg); 1290 } 1291 break; 1292 #ifndef __FreeBSD__ 1293 case 'd': 1294 suspend = true; 1295 break; 1296 #else 1297 case 'p': 1298 if (pincpu_parse(optarg) != 0) { 1299 errx(EX_USAGE, "invalid vcpu pinning " 1300 "configuration '%s'", optarg); 1301 } 1302 break; 1303 #endif 1304 case 'c': 1305 if (topology_parse(optarg) != 0) { 1306 errx(EX_USAGE, "invalid cpu topology " 1307 "'%s'", optarg); 1308 } 1309 break; 1310 case 'C': 1311 memflags |= VM_MEM_F_INCORE; 1312 break; 1313 case 'g': 1314 dbg_port = atoi(optarg); 1315 break; 1316 case 'G': 1317 if (optarg[0] == 'w') { 1318 gdb_stop = true; 1319 optarg++; 1320 } 1321 gdb_port = atoi(optarg); 1322 break; 1323 case 'l': 1324 if (strncmp(optarg, "help", strlen(optarg)) == 0) { 1325 lpc_print_supported_devices(); 1326 exit(0); 1327 } else if (lpc_device_parse(optarg) != 0) { 1328 errx(EX_USAGE, "invalid lpc device " 1329 "configuration '%s'", optarg); 1330 } 1331 break; 1332 case 's': 1333 if (strncmp(optarg, "help", strlen(optarg)) == 0) { 1334 pci_print_supported_devices(); 1335 exit(0); 1336 } else if (pci_parse_slot(optarg) != 0) 1337 exit(4); 1338 else 1339 break; 1340 case 'S': 1341 memflags |= VM_MEM_F_WIRED; 1342 break; 1343 case 'm': 1344 error = vm_parse_memsize(optarg, &memsize); 1345 if (error) 1346 errx(EX_USAGE, "invalid memsize '%s'", optarg); 1347 break; 1348 case 'H': 1349 guest_vmexit_on_hlt = 1; 1350 break; 1351 case 'I': 1352 /* 1353 * The "-I" option was used to add an ioapic to the 1354 * virtual machine. 1355 * 1356 * An ioapic is now provided unconditionally for each 1357 * virtual machine and this option is now deprecated. 1358 */ 1359 break; 1360 case 'P': 1361 guest_vmexit_on_pause = 1; 1362 break; 1363 case 'e': 1364 strictio = 1; 1365 break; 1366 case 'u': 1367 rtc_localtime = 0; 1368 break; 1369 case 'U': 1370 guest_uuid_str = optarg; 1371 break; 1372 case 'w': 1373 strictmsr = 0; 1374 break; 1375 case 'W': 1376 virtio_msix = 0; 1377 break; 1378 case 'x': 1379 x2apic_mode = 1; 1380 break; 1381 case 'Y': 1382 mptgen = 0; 1383 break; 1384 case 'h': 1385 usage(0); 1386 default: 1387 usage(1); 1388 } 1389 } 1390 argc -= optind; 1391 argv += optind; 1392 1393 if (argc != 1) 1394 usage(1); 1395 1396 vmname = argv[0]; 1397 ctx = do_open(vmname); 1398 1399 max_vcpus = num_vcpus_allowed(ctx); 1400 if (guest_ncpus > max_vcpus) { 1401 fprintf(stderr, "%d vCPUs requested but only %d available\n", 1402 guest_ncpus, max_vcpus); 1403 exit(4); 1404 } 1405 1406 fbsdrun_set_capabilities(ctx, BSP); 1407 1408 vm_set_memflags(ctx, memflags); 1409 #ifdef __FreeBSD__ 1410 err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL); 1411 #else 1412 do { 1413 errno = 0; 1414 err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL); 1415 error = errno; 1416 if (err != 0 && error == ENOMEM) { 1417 (void) fprintf(stderr, "Unable to allocate memory " 1418 "(%llu), retrying in 1 second\n", memsize); 1419 sleep(1); 1420 } 1421 } while (error == ENOMEM); 1422 #endif 1423 if (err) { 1424 fprintf(stderr, "Unable to setup memory (%d)\n", errno); 1425 exit(4); 1426 } 1427 1428 error = init_msr(); 1429 if (error) { 1430 fprintf(stderr, "init_msr error %d", error); 1431 exit(4); 1432 } 1433 1434 init_mem(); 1435 init_inout(); 1436 #ifdef __FreeBSD__ 1437 kernemu_dev_init(); 1438 #endif 1439 init_bootrom(ctx); 1440 atkbdc_init(ctx); 1441 pci_irq_init(ctx); 1442 ioapic_init(ctx); 1443 1444 rtc_init(ctx, rtc_localtime); 1445 sci_init(ctx); 1446 #ifndef __FreeBSD__ 1447 pmtmr_init(ctx); 1448 #endif 1449 1450 /* 1451 * Exit if a device emulation finds an error in its initilization 1452 */ 1453 if (init_pci(ctx) != 0) { 1454 perror("device emulation initialization error"); 1455 exit(4); 1456 } 1457 1458 /* 1459 * Initialize after PCI, to allow a bootrom file to reserve the high 1460 * region. 1461 */ 1462 if (acpi) 1463 vmgenc_init(ctx); 1464 1465 if (dbg_port != 0) 1466 init_dbgport(dbg_port); 1467 1468 #ifdef __FreeBSD__ 1469 if (gdb_port != 0) 1470 init_gdb(ctx, gdb_port, gdb_stop); 1471 #else 1472 if (gdb_port < 0) { 1473 /* 1474 * Set up the internal gdb state needed for basic debugging, but 1475 * skip the step of listening on a port for the GDB server. 1476 */ 1477 init_mdb(ctx, gdb_stop); 1478 } else if (gdb_port != 0) { 1479 init_gdb(ctx, gdb_port, gdb_stop); 1480 } 1481 #endif 1482 1483 if (bvmcons) 1484 init_bvmcons(); 1485 1486 vga_init(1); 1487 1488 if (lpc_bootrom()) { 1489 #ifdef __FreeBSD__ 1490 if (vm_set_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, 1)) { 1491 fprintf(stderr, "ROM boot failed: unrestricted guest " 1492 "capability not available\n"); 1493 exit(4); 1494 } 1495 #else 1496 /* Unrestricted Guest is always enabled on illumos */ 1497 #endif 1498 error = vcpu_reset(ctx, BSP); 1499 assert(error == 0); 1500 } 1501 1502 error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip); 1503 assert(error == 0); 1504 1505 /* 1506 * build the guest tables, MP etc. 1507 */ 1508 if (mptgen) { 1509 error = mptable_build(ctx, guest_ncpus); 1510 if (error) { 1511 perror("error to build the guest tables"); 1512 exit(4); 1513 } 1514 } 1515 1516 error = smbios_build(ctx); 1517 assert(error == 0); 1518 1519 if (acpi) { 1520 error = acpi_build(ctx, guest_ncpus); 1521 assert(error == 0); 1522 } 1523 1524 if (lpc_bootrom()) 1525 fwctl_init(); 1526 1527 /* 1528 * Change the proc title to include the VM name. 1529 */ 1530 setproctitle("%s", vmname); 1531 1532 #ifndef WITHOUT_CAPSICUM 1533 caph_cache_catpages(); 1534 1535 if (caph_limit_stdout() == -1 || caph_limit_stderr() == -1) 1536 errx(EX_OSERR, "Unable to apply rights for sandbox"); 1537 1538 if (caph_enter() == -1) 1539 errx(EX_OSERR, "cap_enter() failed"); 1540 #endif 1541 1542 #ifdef __FreeBSD__ 1543 /* 1544 * Add CPU 0 1545 */ 1546 fbsdrun_addcpu(ctx, BSP, BSP, rip); 1547 #else 1548 /* Set BSP to run (unlike the APs which wait for INIT) */ 1549 error = vm_set_run_state(ctx, BSP, VRS_RUN, 0); 1550 assert(error == 0); 1551 fbsdrun_addcpu(ctx, BSP, rip, suspend); 1552 1553 /* Add subsequent CPUs, which will wait until INIT/SIPI-ed */ 1554 for (uint_t i = 1; i < guest_ncpus; i++) { 1555 spinup_halted_ap(ctx, i); 1556 } 1557 #endif 1558 /* 1559 * Head off to the main event dispatch loop 1560 */ 1561 mevent_dispatch(); 1562 1563 exit(4); 1564 } 1565