bhyverun.c revision 330449
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: stable/11/usr.sbin/bhyve/bhyverun.c 330449 2018-03-05 07:26:05Z eadler $ 29 */ 30 31#include <sys/cdefs.h> 32__FBSDID("$FreeBSD: stable/11/usr.sbin/bhyve/bhyverun.c 330449 2018-03-05 07:26:05Z eadler $"); 33 34#include <sys/types.h> 35#ifndef WITHOUT_CAPSICUM 36#include <sys/capsicum.h> 37#endif 38#include <sys/mman.h> 39#include <sys/time.h> 40 41#include <machine/atomic.h> 42#include <machine/segments.h> 43 44#include <stdio.h> 45#include <stdlib.h> 46#include <string.h> 47#include <err.h> 48#include <errno.h> 49#include <libgen.h> 50#include <unistd.h> 51#include <assert.h> 52#include <errno.h> 53#include <pthread.h> 54#include <pthread_np.h> 55#include <sysexits.h> 56#include <stdbool.h> 57#ifndef WITHOUT_CAPSICUM 58#include <nl_types.h> 59#include <termios.h> 60#endif 61 62#include <machine/vmm.h> 63#ifndef WITHOUT_CAPSICUM 64#include <machine/vmm_dev.h> 65#endif 66#include <vmmapi.h> 67 68#include "bhyverun.h" 69#include "acpi.h" 70#include "atkbdc.h" 71#include "inout.h" 72#include "dbgport.h" 73#include "fwctl.h" 74#include "ioapic.h" 75#include "mem.h" 76#include "mevent.h" 77#include "mptbl.h" 78#include "pci_emul.h" 79#include "pci_irq.h" 80#include "pci_lpc.h" 81#include "smbiostbl.h" 82#include "xmsr.h" 83#include "spinup_ap.h" 84#include "rtc.h" 85 86#define GUEST_NIO_PORT 0x488 /* guest upcalls via i/o port */ 87 88#define MB (1024UL * 1024) 89#define GB (1024UL * MB) 90 91typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu); 92extern int vmexit_task_switch(struct vmctx *, struct vm_exit *, int *vcpu); 93 94char *vmname; 95 96int guest_ncpus; 97char *guest_uuid_str; 98 99static int guest_vmexit_on_hlt, guest_vmexit_on_pause; 100static int virtio_msix = 1; 101static int x2apic_mode = 0; /* default is xAPIC */ 102 103static int strictio; 104static int strictmsr = 1; 105 106static int acpi; 107 108static char *progname; 109static const int BSP = 0; 110 111static cpuset_t cpumask; 112 113static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip); 114 115static struct vm_exit vmexit[VM_MAXCPU]; 116 117struct bhyvestats { 118 uint64_t vmexit_bogus; 119 uint64_t vmexit_reqidle; 120 uint64_t vmexit_hlt; 121 uint64_t vmexit_pause; 122 uint64_t vmexit_mtrap; 123 uint64_t vmexit_inst_emul; 124 uint64_t cpu_switch_rotate; 125 uint64_t cpu_switch_direct; 126} stats; 127 128struct mt_vmm_info { 129 pthread_t mt_thr; 130 struct vmctx *mt_ctx; 131 int mt_vcpu; 132} mt_vmm_info[VM_MAXCPU]; 133 134static cpuset_t *vcpumap[VM_MAXCPU] = { NULL }; 135 136static void 137usage(int code) 138{ 139 140 fprintf(stderr, 141 "Usage: %s [-abehuwxACHPSWY] [-c vcpus] [-g <gdb port>] [-l <lpc>]\n" 142 " %*s [-m mem] [-p vcpu:hostcpu] [-s <pci>] [-U uuid] <vm>\n" 143 " -a: local apic is in xAPIC mode (deprecated)\n" 144 " -A: create ACPI tables\n" 145 " -c: # cpus (default 1)\n" 146 " -C: include guest memory in core file\n" 147 " -e: exit on unhandled I/O access\n" 148 " -g: gdb port\n" 149 " -h: help\n" 150 " -H: vmexit from the guest on hlt\n" 151 " -l: LPC device configuration\n" 152 " -m: memory size in MB\n" 153 " -p: pin 'vcpu' to 'hostcpu'\n" 154 " -P: vmexit from the guest on pause\n" 155 " -s: <slot,driver,configinfo> PCI slot config\n" 156 " -S: guest memory cannot be swapped\n" 157 " -u: RTC keeps UTC time\n" 158 " -U: uuid\n" 159 " -w: ignore unimplemented MSRs\n" 160 " -W: force virtio to use single-vector MSI\n" 161 " -x: local apic is in x2APIC mode\n" 162 " -Y: disable MPtable generation\n", 163 progname, (int)strlen(progname), ""); 164 165 exit(code); 166} 167 168#ifndef WITHOUT_CAPSICUM 169/* 170 * 11-stable capsicum helpers 171 */ 172static void 173bhyve_caph_cache_catpages(void) 174{ 175 176 (void)catopen("libc", NL_CAT_LOCALE); 177} 178 179static int 180bhyve_caph_limit_stdoe(void) 181{ 182 cap_rights_t rights; 183 unsigned long cmds[] = { TIOCGETA, TIOCGWINSZ }; 184 int i, fds[] = { STDOUT_FILENO, STDERR_FILENO }; 185 186 cap_rights_init(&rights, CAP_FCNTL, CAP_FSTAT, CAP_IOCTL); 187 cap_rights_set(&rights, CAP_WRITE); 188 189 for (i = 0; i < nitems(fds); i++) { 190 if (cap_rights_limit(fds[i], &rights) < 0 && errno != ENOSYS) 191 return (-1); 192 193 if (cap_ioctls_limit(fds[i], cmds, nitems(cmds)) < 0 && errno != ENOSYS) 194 return (-1); 195 196 if (cap_fcntls_limit(fds[i], CAP_FCNTL_GETFL) < 0 && errno != ENOSYS) 197 return (-1); 198 } 199 200 return (0); 201} 202 203#endif 204 205static int 206pincpu_parse(const char *opt) 207{ 208 int vcpu, pcpu; 209 210 if (sscanf(opt, "%d:%d", &vcpu, &pcpu) != 2) { 211 fprintf(stderr, "invalid format: %s\n", opt); 212 return (-1); 213 } 214 215 if (vcpu < 0 || vcpu >= VM_MAXCPU) { 216 fprintf(stderr, "vcpu '%d' outside valid range from 0 to %d\n", 217 vcpu, VM_MAXCPU - 1); 218 return (-1); 219 } 220 221 if (pcpu < 0 || pcpu >= CPU_SETSIZE) { 222 fprintf(stderr, "hostcpu '%d' outside valid range from " 223 "0 to %d\n", pcpu, CPU_SETSIZE - 1); 224 return (-1); 225 } 226 227 if (vcpumap[vcpu] == NULL) { 228 if ((vcpumap[vcpu] = malloc(sizeof(cpuset_t))) == NULL) { 229 perror("malloc"); 230 return (-1); 231 } 232 CPU_ZERO(vcpumap[vcpu]); 233 } 234 CPU_SET(pcpu, vcpumap[vcpu]); 235 return (0); 236} 237 238void 239vm_inject_fault(void *arg, int vcpu, int vector, int errcode_valid, 240 int errcode) 241{ 242 struct vmctx *ctx; 243 int error, restart_instruction; 244 245 ctx = arg; 246 restart_instruction = 1; 247 248 error = vm_inject_exception(ctx, vcpu, vector, errcode_valid, errcode, 249 restart_instruction); 250 assert(error == 0); 251} 252 253void * 254paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len) 255{ 256 257 return (vm_map_gpa(ctx, gaddr, len)); 258} 259 260int 261fbsdrun_vmexit_on_pause(void) 262{ 263 264 return (guest_vmexit_on_pause); 265} 266 267int 268fbsdrun_vmexit_on_hlt(void) 269{ 270 271 return (guest_vmexit_on_hlt); 272} 273 274int 275fbsdrun_virtio_msix(void) 276{ 277 278 return (virtio_msix); 279} 280 281static void * 282fbsdrun_start_thread(void *param) 283{ 284 char tname[MAXCOMLEN + 1]; 285 struct mt_vmm_info *mtp; 286 int vcpu; 287 288 mtp = param; 289 vcpu = mtp->mt_vcpu; 290 291 snprintf(tname, sizeof(tname), "vcpu %d", vcpu); 292 pthread_set_name_np(mtp->mt_thr, tname); 293 294 vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip); 295 296 /* not reached */ 297 exit(1); 298 return (NULL); 299} 300 301void 302fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip) 303{ 304 int error; 305 306 assert(fromcpu == BSP); 307 308 /* 309 * The 'newcpu' must be activated in the context of 'fromcpu'. If 310 * vm_activate_cpu() is delayed until newcpu's pthread starts running 311 * then vmm.ko is out-of-sync with bhyve and this can create a race 312 * with vm_suspend(). 313 */ 314 error = vm_activate_cpu(ctx, newcpu); 315 if (error != 0) 316 err(EX_OSERR, "could not activate CPU %d", newcpu); 317 318 CPU_SET_ATOMIC(newcpu, &cpumask); 319 320 /* 321 * Set up the vmexit struct to allow execution to start 322 * at the given RIP 323 */ 324 vmexit[newcpu].rip = rip; 325 vmexit[newcpu].inst_length = 0; 326 327 mt_vmm_info[newcpu].mt_ctx = ctx; 328 mt_vmm_info[newcpu].mt_vcpu = newcpu; 329 330 error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL, 331 fbsdrun_start_thread, &mt_vmm_info[newcpu]); 332 assert(error == 0); 333} 334 335static int 336fbsdrun_deletecpu(struct vmctx *ctx, int vcpu) 337{ 338 339 if (!CPU_ISSET(vcpu, &cpumask)) { 340 fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu); 341 exit(1); 342 } 343 344 CPU_CLR_ATOMIC(vcpu, &cpumask); 345 return (CPU_EMPTY(&cpumask)); 346} 347 348static int 349vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu, 350 uint32_t eax) 351{ 352#if BHYVE_DEBUG 353 /* 354 * put guest-driven debug here 355 */ 356#endif 357 return (VMEXIT_CONTINUE); 358} 359 360static int 361vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 362{ 363 int error; 364 int bytes, port, in, out; 365 int vcpu; 366 367 vcpu = *pvcpu; 368 369 port = vme->u.inout.port; 370 bytes = vme->u.inout.bytes; 371 in = vme->u.inout.in; 372 out = !in; 373 374 /* Extra-special case of host notifications */ 375 if (out && port == GUEST_NIO_PORT) { 376 error = vmexit_handle_notify(ctx, vme, pvcpu, vme->u.inout.eax); 377 return (error); 378 } 379 380 error = emulate_inout(ctx, vcpu, vme, strictio); 381 if (error) { 382 fprintf(stderr, "Unhandled %s%c 0x%04x at 0x%lx\n", 383 in ? "in" : "out", 384 bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), 385 port, vmexit->rip); 386 return (VMEXIT_ABORT); 387 } else { 388 return (VMEXIT_CONTINUE); 389 } 390} 391 392static int 393vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 394{ 395 uint64_t val; 396 uint32_t eax, edx; 397 int error; 398 399 val = 0; 400 error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val); 401 if (error != 0) { 402 fprintf(stderr, "rdmsr to register %#x on vcpu %d\n", 403 vme->u.msr.code, *pvcpu); 404 if (strictmsr) { 405 vm_inject_gp(ctx, *pvcpu); 406 return (VMEXIT_CONTINUE); 407 } 408 } 409 410 eax = val; 411 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax); 412 assert(error == 0); 413 414 edx = val >> 32; 415 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx); 416 assert(error == 0); 417 418 return (VMEXIT_CONTINUE); 419} 420 421static int 422vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 423{ 424 int error; 425 426 error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval); 427 if (error != 0) { 428 fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n", 429 vme->u.msr.code, vme->u.msr.wval, *pvcpu); 430 if (strictmsr) { 431 vm_inject_gp(ctx, *pvcpu); 432 return (VMEXIT_CONTINUE); 433 } 434 } 435 return (VMEXIT_CONTINUE); 436} 437 438static int 439vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 440{ 441 442 (void)spinup_ap(ctx, *pvcpu, 443 vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip); 444 445 return (VMEXIT_CONTINUE); 446} 447 448#define DEBUG_EPT_MISCONFIG 449#ifdef DEBUG_EPT_MISCONFIG 450#define EXIT_REASON_EPT_MISCONFIG 49 451#define VMCS_GUEST_PHYSICAL_ADDRESS 0x00002400 452#define VMCS_IDENT(x) ((x) | 0x80000000) 453 454static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4]; 455static int ept_misconfig_ptenum; 456#endif 457 458static int 459vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 460{ 461 462 fprintf(stderr, "vm exit[%d]\n", *pvcpu); 463 fprintf(stderr, "\treason\t\tVMX\n"); 464 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip); 465 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length); 466 fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status); 467 fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason); 468 fprintf(stderr, "\tqualification\t0x%016lx\n", 469 vmexit->u.vmx.exit_qualification); 470 fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type); 471 fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error); 472#ifdef DEBUG_EPT_MISCONFIG 473 if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) { 474 vm_get_register(ctx, *pvcpu, 475 VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS), 476 &ept_misconfig_gpa); 477 vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte, 478 &ept_misconfig_ptenum); 479 fprintf(stderr, "\tEPT misconfiguration:\n"); 480 fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa); 481 fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n", 482 ept_misconfig_ptenum, ept_misconfig_pte[0], 483 ept_misconfig_pte[1], ept_misconfig_pte[2], 484 ept_misconfig_pte[3]); 485 } 486#endif /* DEBUG_EPT_MISCONFIG */ 487 return (VMEXIT_ABORT); 488} 489 490static int 491vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 492{ 493 494 fprintf(stderr, "vm exit[%d]\n", *pvcpu); 495 fprintf(stderr, "\treason\t\tSVM\n"); 496 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip); 497 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length); 498 fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode); 499 fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1); 500 fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2); 501 return (VMEXIT_ABORT); 502} 503 504static int 505vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 506{ 507 508 assert(vmexit->inst_length == 0); 509 510 stats.vmexit_bogus++; 511 512 return (VMEXIT_CONTINUE); 513} 514 515static int 516vmexit_reqidle(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 517{ 518 519 assert(vmexit->inst_length == 0); 520 521 stats.vmexit_reqidle++; 522 523 return (VMEXIT_CONTINUE); 524} 525 526static int 527vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 528{ 529 530 stats.vmexit_hlt++; 531 532 /* 533 * Just continue execution with the next instruction. We use 534 * the HLT VM exit as a way to be friendly with the host 535 * scheduler. 536 */ 537 return (VMEXIT_CONTINUE); 538} 539 540static int 541vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 542{ 543 544 stats.vmexit_pause++; 545 546 return (VMEXIT_CONTINUE); 547} 548 549static int 550vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 551{ 552 553 assert(vmexit->inst_length == 0); 554 555 stats.vmexit_mtrap++; 556 557 return (VMEXIT_CONTINUE); 558} 559 560static int 561vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 562{ 563 int err, i; 564 struct vie *vie; 565 566 stats.vmexit_inst_emul++; 567 568 vie = &vmexit->u.inst_emul.vie; 569 err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa, 570 vie, &vmexit->u.inst_emul.paging); 571 572 if (err) { 573 if (err == ESRCH) { 574 fprintf(stderr, "Unhandled memory access to 0x%lx\n", 575 vmexit->u.inst_emul.gpa); 576 } 577 578 fprintf(stderr, "Failed to emulate instruction ["); 579 for (i = 0; i < vie->num_valid; i++) { 580 fprintf(stderr, "0x%02x%s", vie->inst[i], 581 i != (vie->num_valid - 1) ? " " : ""); 582 } 583 fprintf(stderr, "] at 0x%lx\n", vmexit->rip); 584 return (VMEXIT_ABORT); 585 } 586 587 return (VMEXIT_CONTINUE); 588} 589 590static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER; 591static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER; 592 593static int 594vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 595{ 596 enum vm_suspend_how how; 597 598 how = vmexit->u.suspended.how; 599 600 fbsdrun_deletecpu(ctx, *pvcpu); 601 602 if (*pvcpu != BSP) { 603 pthread_mutex_lock(&resetcpu_mtx); 604 pthread_cond_signal(&resetcpu_cond); 605 pthread_mutex_unlock(&resetcpu_mtx); 606 pthread_exit(NULL); 607 } 608 609 pthread_mutex_lock(&resetcpu_mtx); 610 while (!CPU_EMPTY(&cpumask)) { 611 pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx); 612 } 613 pthread_mutex_unlock(&resetcpu_mtx); 614 615 switch (how) { 616 case VM_SUSPEND_RESET: 617 exit(0); 618 case VM_SUSPEND_POWEROFF: 619 exit(1); 620 case VM_SUSPEND_HALT: 621 exit(2); 622 case VM_SUSPEND_TRIPLEFAULT: 623 exit(3); 624 default: 625 fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how); 626 exit(100); 627 } 628 return (0); /* NOTREACHED */ 629} 630 631static vmexit_handler_t handler[VM_EXITCODE_MAX] = { 632 [VM_EXITCODE_INOUT] = vmexit_inout, 633 [VM_EXITCODE_INOUT_STR] = vmexit_inout, 634 [VM_EXITCODE_VMX] = vmexit_vmx, 635 [VM_EXITCODE_SVM] = vmexit_svm, 636 [VM_EXITCODE_BOGUS] = vmexit_bogus, 637 [VM_EXITCODE_REQIDLE] = vmexit_reqidle, 638 [VM_EXITCODE_RDMSR] = vmexit_rdmsr, 639 [VM_EXITCODE_WRMSR] = vmexit_wrmsr, 640 [VM_EXITCODE_MTRAP] = vmexit_mtrap, 641 [VM_EXITCODE_INST_EMUL] = vmexit_inst_emul, 642 [VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap, 643 [VM_EXITCODE_SUSPENDED] = vmexit_suspend, 644 [VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch, 645}; 646 647static void 648vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip) 649{ 650 int error, rc; 651 enum vm_exitcode exitcode; 652 cpuset_t active_cpus; 653 654 if (vcpumap[vcpu] != NULL) { 655 error = pthread_setaffinity_np(pthread_self(), 656 sizeof(cpuset_t), vcpumap[vcpu]); 657 assert(error == 0); 658 } 659 660 error = vm_active_cpus(ctx, &active_cpus); 661 assert(CPU_ISSET(vcpu, &active_cpus)); 662 663 error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip); 664 assert(error == 0); 665 666 while (1) { 667 error = vm_run(ctx, vcpu, &vmexit[vcpu]); 668 if (error != 0) 669 break; 670 671 exitcode = vmexit[vcpu].exitcode; 672 if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) { 673 fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n", 674 exitcode); 675 exit(1); 676 } 677 678 rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu); 679 680 switch (rc) { 681 case VMEXIT_CONTINUE: 682 break; 683 case VMEXIT_ABORT: 684 abort(); 685 default: 686 exit(1); 687 } 688 } 689 fprintf(stderr, "vm_run error %d, errno %d\n", error, errno); 690} 691 692static int 693num_vcpus_allowed(struct vmctx *ctx) 694{ 695 int tmp, error; 696 697 error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp); 698 699 /* 700 * The guest is allowed to spinup more than one processor only if the 701 * UNRESTRICTED_GUEST capability is available. 702 */ 703 if (error == 0) 704 return (VM_MAXCPU); 705 else 706 return (1); 707} 708 709void 710fbsdrun_set_capabilities(struct vmctx *ctx, int cpu) 711{ 712 int err, tmp; 713 714 if (fbsdrun_vmexit_on_hlt()) { 715 err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp); 716 if (err < 0) { 717 fprintf(stderr, "VM exit on HLT not supported\n"); 718 exit(1); 719 } 720 vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1); 721 if (cpu == BSP) 722 handler[VM_EXITCODE_HLT] = vmexit_hlt; 723 } 724 725 if (fbsdrun_vmexit_on_pause()) { 726 /* 727 * pause exit support required for this mode 728 */ 729 err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp); 730 if (err < 0) { 731 fprintf(stderr, 732 "SMP mux requested, no pause support\n"); 733 exit(1); 734 } 735 vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1); 736 if (cpu == BSP) 737 handler[VM_EXITCODE_PAUSE] = vmexit_pause; 738 } 739 740 if (x2apic_mode) 741 err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED); 742 else 743 err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED); 744 745 if (err) { 746 fprintf(stderr, "Unable to set x2apic state (%d)\n", err); 747 exit(1); 748 } 749 750 vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1); 751} 752 753static struct vmctx * 754do_open(const char *vmname) 755{ 756 struct vmctx *ctx; 757 int error; 758 bool reinit, romboot; 759#ifndef WITHOUT_CAPSICUM 760 cap_rights_t rights; 761 const cap_ioctl_t *cmds; 762 size_t ncmds; 763#endif 764 765 reinit = romboot = false; 766 767 if (lpc_bootrom()) 768 romboot = true; 769 770 error = vm_create(vmname); 771 if (error) { 772 if (errno == EEXIST) { 773 if (romboot) { 774 reinit = true; 775 } else { 776 /* 777 * The virtual machine has been setup by the 778 * userspace bootloader. 779 */ 780 } 781 } else { 782 perror("vm_create"); 783 exit(1); 784 } 785 } else { 786 if (!romboot) { 787 /* 788 * If the virtual machine was just created then a 789 * bootrom must be configured to boot it. 790 */ 791 fprintf(stderr, "virtual machine cannot be booted\n"); 792 exit(1); 793 } 794 } 795 796 ctx = vm_open(vmname); 797 if (ctx == NULL) { 798 perror("vm_open"); 799 exit(1); 800 } 801 802#ifndef WITHOUT_CAPSICUM 803 cap_rights_init(&rights, CAP_IOCTL, CAP_MMAP_RW); 804 if (cap_rights_limit(vm_get_device_fd(ctx), &rights) == -1 && 805 errno != ENOSYS) 806 errx(EX_OSERR, "Unable to apply rights for sandbox"); 807 vm_get_ioctls(&ncmds); 808 cmds = vm_get_ioctls(NULL); 809 if (cmds == NULL) 810 errx(EX_OSERR, "out of memory"); 811 if (cap_ioctls_limit(vm_get_device_fd(ctx), cmds, ncmds) == -1 && 812 errno != ENOSYS) 813 errx(EX_OSERR, "Unable to apply rights for sandbox"); 814 free((cap_ioctl_t *)cmds); 815#endif 816 817 if (reinit) { 818 error = vm_reinit(ctx); 819 if (error) { 820 perror("vm_reinit"); 821 exit(1); 822 } 823 } 824 return (ctx); 825} 826 827int 828main(int argc, char *argv[]) 829{ 830 int c, error, gdb_port, err, bvmcons; 831 int max_vcpus, mptgen, memflags; 832 int rtc_localtime; 833 struct vmctx *ctx; 834 uint64_t rip; 835 size_t memsize; 836 char *optstr; 837 838 bvmcons = 0; 839 progname = basename(argv[0]); 840 gdb_port = 0; 841 guest_ncpus = 1; 842 memsize = 256 * MB; 843 mptgen = 1; 844 rtc_localtime = 1; 845 memflags = 0; 846 847 optstr = "abehuwxACHIPSWYp:g:c:s:m:l:U:"; 848 while ((c = getopt(argc, argv, optstr)) != -1) { 849 switch (c) { 850 case 'a': 851 x2apic_mode = 0; 852 break; 853 case 'A': 854 acpi = 1; 855 break; 856 case 'b': 857 bvmcons = 1; 858 break; 859 case 'p': 860 if (pincpu_parse(optarg) != 0) { 861 errx(EX_USAGE, "invalid vcpu pinning " 862 "configuration '%s'", optarg); 863 } 864 break; 865 case 'c': 866 guest_ncpus = atoi(optarg); 867 break; 868 case 'C': 869 memflags |= VM_MEM_F_INCORE; 870 break; 871 case 'g': 872 gdb_port = atoi(optarg); 873 break; 874 case 'l': 875 if (lpc_device_parse(optarg) != 0) { 876 errx(EX_USAGE, "invalid lpc device " 877 "configuration '%s'", optarg); 878 } 879 break; 880 case 's': 881 if (pci_parse_slot(optarg) != 0) 882 exit(1); 883 else 884 break; 885 case 'S': 886 memflags |= VM_MEM_F_WIRED; 887 break; 888 case 'm': 889 error = vm_parse_memsize(optarg, &memsize); 890 if (error) 891 errx(EX_USAGE, "invalid memsize '%s'", optarg); 892 break; 893 case 'H': 894 guest_vmexit_on_hlt = 1; 895 break; 896 case 'I': 897 /* 898 * The "-I" option was used to add an ioapic to the 899 * virtual machine. 900 * 901 * An ioapic is now provided unconditionally for each 902 * virtual machine and this option is now deprecated. 903 */ 904 break; 905 case 'P': 906 guest_vmexit_on_pause = 1; 907 break; 908 case 'e': 909 strictio = 1; 910 break; 911 case 'u': 912 rtc_localtime = 0; 913 break; 914 case 'U': 915 guest_uuid_str = optarg; 916 break; 917 case 'w': 918 strictmsr = 0; 919 break; 920 case 'W': 921 virtio_msix = 0; 922 break; 923 case 'x': 924 x2apic_mode = 1; 925 break; 926 case 'Y': 927 mptgen = 0; 928 break; 929 case 'h': 930 usage(0); 931 default: 932 usage(1); 933 } 934 } 935 argc -= optind; 936 argv += optind; 937 938 if (argc != 1) 939 usage(1); 940 941 vmname = argv[0]; 942 ctx = do_open(vmname); 943 944 if (guest_ncpus < 1) { 945 fprintf(stderr, "Invalid guest vCPUs (%d)\n", guest_ncpus); 946 exit(1); 947 } 948 949 max_vcpus = num_vcpus_allowed(ctx); 950 if (guest_ncpus > max_vcpus) { 951 fprintf(stderr, "%d vCPUs requested but only %d available\n", 952 guest_ncpus, max_vcpus); 953 exit(1); 954 } 955 956 fbsdrun_set_capabilities(ctx, BSP); 957 958 vm_set_memflags(ctx, memflags); 959 err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL); 960 if (err) { 961 fprintf(stderr, "Unable to setup memory (%d)\n", errno); 962 exit(1); 963 } 964 965 error = init_msr(); 966 if (error) { 967 fprintf(stderr, "init_msr error %d", error); 968 exit(1); 969 } 970 971 init_mem(); 972 init_inout(); 973 atkbdc_init(ctx); 974 pci_irq_init(ctx); 975 ioapic_init(ctx); 976 977 rtc_init(ctx, rtc_localtime); 978 sci_init(ctx); 979 980 /* 981 * Exit if a device emulation finds an error in it's initilization 982 */ 983 if (init_pci(ctx) != 0) 984 exit(1); 985 986 if (gdb_port != 0) 987 init_dbgport(gdb_port); 988 989 if (bvmcons) 990 init_bvmcons(); 991 992 if (lpc_bootrom()) { 993 if (vm_set_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, 1)) { 994 fprintf(stderr, "ROM boot failed: unrestricted guest " 995 "capability not available\n"); 996 exit(1); 997 } 998 error = vcpu_reset(ctx, BSP); 999 assert(error == 0); 1000 } 1001 1002 error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip); 1003 assert(error == 0); 1004 1005 /* 1006 * build the guest tables, MP etc. 1007 */ 1008 if (mptgen) { 1009 error = mptable_build(ctx, guest_ncpus); 1010 if (error) 1011 exit(1); 1012 } 1013 1014 error = smbios_build(ctx); 1015 assert(error == 0); 1016 1017 if (acpi) { 1018 error = acpi_build(ctx, guest_ncpus); 1019 assert(error == 0); 1020 } 1021 1022 if (lpc_bootrom()) 1023 fwctl_init(); 1024 1025#ifndef WITHOUT_CAPSICUM 1026 bhyve_caph_cache_catpages(); 1027 1028 if (bhyve_caph_limit_stdoe() == -1) 1029 errx(EX_OSERR, "Unable to apply rights for sandbox"); 1030 1031 if (cap_enter() == -1 && errno != ENOSYS) 1032 errx(EX_OSERR, "cap_enter() failed"); 1033#endif 1034 1035 /* 1036 * Change the proc title to include the VM name. 1037 */ 1038 setproctitle("%s", vmname); 1039 1040 /* 1041 * Add CPU 0 1042 */ 1043 fbsdrun_addcpu(ctx, BSP, BSP, rip); 1044 1045 /* 1046 * Head off to the main event dispatch loop 1047 */ 1048 mevent_dispatch(); 1049 1050 exit(1); 1051} 1052