1 /*- 2 * Copyright (c) 2011 NetApp, Inc. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND 15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17 * ARE DISCLAIMED. IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE 18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24 * SUCH DAMAGE. 25 * 26 * $FreeBSD$ 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 31 32 #include <sys/types.h> 33 #include <sys/mman.h> 34 #include <sys/time.h> 35 36 #include <machine/atomic.h> 37 #include <machine/segments.h> 38 39 #include <stdio.h> 40 #include <stdlib.h> 41 #include <string.h> 42 #include <err.h> 43 #include <libgen.h> 44 #include <unistd.h> 45 #include <assert.h> 46 #include <errno.h> 47 #include <pthread.h> 48 #include <pthread_np.h> 49 #include <sysexits.h> 50 #include <stdbool.h> 51 52 #include <machine/vmm.h> 53 #include <vmmapi.h> 54 55 #include "bhyverun.h" 56 #include "acpi.h" 57 #include "inout.h" 58 #include "dbgport.h" 59 #include "fwctl.h" 60 #include "ioapic.h" 61 #include "mem.h" 62 #include "mevent.h" 63 #include "mptbl.h" 64 #include "pci_emul.h" 65 #include "pci_irq.h" 66 #include "pci_lpc.h" 67 #include "smbiostbl.h" 68 #include "xmsr.h" 69 #include "spinup_ap.h" 70 #include "rtc.h" 71 72 #define GUEST_NIO_PORT 0x488 /* guest upcalls via i/o port */ 73 74 #define MB (1024UL * 1024) 75 #define GB (1024UL * MB) 76 77 typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu); 78 extern int vmexit_task_switch(struct vmctx *, struct vm_exit *, int *vcpu); 79 80 char *vmname; 81 82 int guest_ncpus; 83 char *guest_uuid_str; 84 85 static int guest_vmexit_on_hlt, guest_vmexit_on_pause; 86 static int virtio_msix = 1; 87 static int x2apic_mode = 0; /* default is xAPIC */ 88 89 static int strictio; 90 static int strictmsr = 1; 91 92 static int acpi; 93 94 static char *progname; 95 static const int BSP = 0; 96 97 static cpuset_t cpumask; 98 99 static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip); 100 101 static struct vm_exit vmexit[VM_MAXCPU]; 102 103 struct bhyvestats { 104 uint64_t vmexit_bogus; 105 uint64_t vmexit_reqidle; 106 uint64_t vmexit_hlt; 107 uint64_t vmexit_pause; 108 uint64_t vmexit_mtrap; 109 uint64_t vmexit_inst_emul; 110 uint64_t cpu_switch_rotate; 111 uint64_t cpu_switch_direct; 112 } stats; 113 114 struct mt_vmm_info { 115 pthread_t mt_thr; 116 struct vmctx *mt_ctx; 117 int mt_vcpu; 118 } mt_vmm_info[VM_MAXCPU]; 119 120 static cpuset_t *vcpumap[VM_MAXCPU] = { NULL }; 121 122 static void 123 usage(int code) 124 { 125 126 fprintf(stderr, 127 "Usage: %s [-abehuwxACHPSWY] [-c vcpus] [-g <gdb port>] [-l <lpc>]\n" 128 " %*s [-m mem] [-p vcpu:hostcpu] [-s <pci>] [-U uuid] <vm>\n" 129 " -a: local apic is in xAPIC mode (deprecated)\n" 130 " -A: create ACPI tables\n" 131 " -c: # cpus (default 1)\n" 132 " -C: include guest memory in core file\n" 133 " -e: exit on unhandled I/O access\n" 134 " -g: gdb port\n" 135 " -h: help\n" 136 " -H: vmexit from the guest on hlt\n" 137 " -l: LPC device configuration\n" 138 " -m: memory size in MB\n" 139 " -p: pin 'vcpu' to 'hostcpu'\n" 140 " -P: vmexit from the guest on pause\n" 141 " -s: <slot,driver,configinfo> PCI slot config\n" 142 " -S: guest memory cannot be swapped\n" 143 " -u: RTC keeps UTC time\n" 144 " -U: uuid\n" 145 " -w: ignore unimplemented MSRs\n" 146 " -W: force virtio to use single-vector MSI\n" 147 " -x: local apic is in x2APIC mode\n" 148 " -Y: disable MPtable generation\n", 149 progname, (int)strlen(progname), ""); 150 151 exit(code); 152 } 153 154 static int 155 pincpu_parse(const char *opt) 156 { 157 int vcpu, pcpu; 158 159 if (sscanf(opt, "%d:%d", &vcpu, &pcpu) != 2) { 160 fprintf(stderr, "invalid format: %s\n", opt); 161 return (-1); 162 } 163 164 if (vcpu < 0 || vcpu >= VM_MAXCPU) { 165 fprintf(stderr, "vcpu '%d' outside valid range from 0 to %d\n", 166 vcpu, VM_MAXCPU - 1); 167 return (-1); 168 } 169 170 if (pcpu < 0 || pcpu >= CPU_SETSIZE) { 171 fprintf(stderr, "hostcpu '%d' outside valid range from " 172 "0 to %d\n", pcpu, CPU_SETSIZE - 1); 173 return (-1); 174 } 175 176 if (vcpumap[vcpu] == NULL) { 177 if ((vcpumap[vcpu] = malloc(sizeof(cpuset_t))) == NULL) { 178 perror("malloc"); 179 return (-1); 180 } 181 CPU_ZERO(vcpumap[vcpu]); 182 } 183 CPU_SET(pcpu, vcpumap[vcpu]); 184 return (0); 185 } 186 187 void 188 vm_inject_fault(void *arg, int vcpu, int vector, int errcode_valid, 189 int errcode) 190 { 191 struct vmctx *ctx; 192 int error, restart_instruction; 193 194 ctx = arg; 195 restart_instruction = 1; 196 197 error = vm_inject_exception(ctx, vcpu, vector, errcode_valid, errcode, 198 restart_instruction); 199 assert(error == 0); 200 } 201 202 void * 203 paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len) 204 { 205 206 return (vm_map_gpa(ctx, gaddr, len)); 207 } 208 209 int 210 fbsdrun_vmexit_on_pause(void) 211 { 212 213 return (guest_vmexit_on_pause); 214 } 215 216 int 217 fbsdrun_vmexit_on_hlt(void) 218 { 219 220 return (guest_vmexit_on_hlt); 221 } 222 223 int 224 fbsdrun_virtio_msix(void) 225 { 226 227 return (virtio_msix); 228 } 229 230 static void * 231 fbsdrun_start_thread(void *param) 232 { 233 char tname[MAXCOMLEN + 1]; 234 struct mt_vmm_info *mtp; 235 int vcpu; 236 237 mtp = param; 238 vcpu = mtp->mt_vcpu; 239 240 snprintf(tname, sizeof(tname), "vcpu %d", vcpu); 241 pthread_set_name_np(mtp->mt_thr, tname); 242 243 vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip); 244 245 /* not reached */ 246 exit(1); 247 return (NULL); 248 } 249 250 void 251 fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip) 252 { 253 int error; 254 255 assert(fromcpu == BSP); 256 257 /* 258 * The 'newcpu' must be activated in the context of 'fromcpu'. If 259 * vm_activate_cpu() is delayed until newcpu's pthread starts running 260 * then vmm.ko is out-of-sync with bhyve and this can create a race 261 * with vm_suspend(). 262 */ 263 error = vm_activate_cpu(ctx, newcpu); 264 assert(error == 0); 265 266 CPU_SET_ATOMIC(newcpu, &cpumask); 267 268 /* 269 * Set up the vmexit struct to allow execution to start 270 * at the given RIP 271 */ 272 vmexit[newcpu].rip = rip; 273 vmexit[newcpu].inst_length = 0; 274 275 mt_vmm_info[newcpu].mt_ctx = ctx; 276 mt_vmm_info[newcpu].mt_vcpu = newcpu; 277 278 error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL, 279 fbsdrun_start_thread, &mt_vmm_info[newcpu]); 280 assert(error == 0); 281 } 282 283 static int 284 fbsdrun_deletecpu(struct vmctx *ctx, int vcpu) 285 { 286 287 if (!CPU_ISSET(vcpu, &cpumask)) { 288 fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu); 289 exit(1); 290 } 291 292 CPU_CLR_ATOMIC(vcpu, &cpumask); 293 return (CPU_EMPTY(&cpumask)); 294 } 295 296 static int 297 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu, 298 uint32_t eax) 299 { 300 #if BHYVE_DEBUG 301 /* 302 * put guest-driven debug here 303 */ 304 #endif 305 return (VMEXIT_CONTINUE); 306 } 307 308 static int 309 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 310 { 311 int error; 312 int bytes, port, in, out, string; 313 int vcpu; 314 315 vcpu = *pvcpu; 316 317 port = vme->u.inout.port; 318 bytes = vme->u.inout.bytes; 319 string = vme->u.inout.string; 320 in = vme->u.inout.in; 321 out = !in; 322 323 /* Extra-special case of host notifications */ 324 if (out && port == GUEST_NIO_PORT) { 325 error = vmexit_handle_notify(ctx, vme, pvcpu, vme->u.inout.eax); 326 return (error); 327 } 328 329 error = emulate_inout(ctx, vcpu, vme, strictio); 330 if (error) { 331 fprintf(stderr, "Unhandled %s%c 0x%04x at 0x%lx\n", 332 in ? "in" : "out", 333 bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), 334 port, vmexit->rip); 335 return (VMEXIT_ABORT); 336 } else { 337 return (VMEXIT_CONTINUE); 338 } 339 } 340 341 static int 342 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 343 { 344 uint64_t val; 345 uint32_t eax, edx; 346 int error; 347 348 val = 0; 349 error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val); 350 if (error != 0) { 351 fprintf(stderr, "rdmsr to register %#x on vcpu %d\n", 352 vme->u.msr.code, *pvcpu); 353 if (strictmsr) { 354 vm_inject_gp(ctx, *pvcpu); 355 return (VMEXIT_CONTINUE); 356 } 357 } 358 359 eax = val; 360 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax); 361 assert(error == 0); 362 363 edx = val >> 32; 364 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx); 365 assert(error == 0); 366 367 return (VMEXIT_CONTINUE); 368 } 369 370 static int 371 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 372 { 373 int error; 374 375 error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval); 376 if (error != 0) { 377 fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n", 378 vme->u.msr.code, vme->u.msr.wval, *pvcpu); 379 if (strictmsr) { 380 vm_inject_gp(ctx, *pvcpu); 381 return (VMEXIT_CONTINUE); 382 } 383 } 384 return (VMEXIT_CONTINUE); 385 } 386 387 static int 388 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 389 { 390 int newcpu; 391 int retval = VMEXIT_CONTINUE; 392 393 newcpu = spinup_ap(ctx, *pvcpu, 394 vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip); 395 396 return (retval); 397 } 398 399 #define DEBUG_EPT_MISCONFIG 400 #ifdef DEBUG_EPT_MISCONFIG 401 #define EXIT_REASON_EPT_MISCONFIG 49 402 #define VMCS_GUEST_PHYSICAL_ADDRESS 0x00002400 403 #define VMCS_IDENT(x) ((x) | 0x80000000) 404 405 static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4]; 406 static int ept_misconfig_ptenum; 407 #endif 408 409 static int 410 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 411 { 412 413 fprintf(stderr, "vm exit[%d]\n", *pvcpu); 414 fprintf(stderr, "\treason\t\tVMX\n"); 415 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip); 416 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length); 417 fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status); 418 fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason); 419 fprintf(stderr, "\tqualification\t0x%016lx\n", 420 vmexit->u.vmx.exit_qualification); 421 fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type); 422 fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error); 423 #ifdef DEBUG_EPT_MISCONFIG 424 if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) { 425 vm_get_register(ctx, *pvcpu, 426 VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS), 427 &ept_misconfig_gpa); 428 vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte, 429 &ept_misconfig_ptenum); 430 fprintf(stderr, "\tEPT misconfiguration:\n"); 431 fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa); 432 fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n", 433 ept_misconfig_ptenum, ept_misconfig_pte[0], 434 ept_misconfig_pte[1], ept_misconfig_pte[2], 435 ept_misconfig_pte[3]); 436 } 437 #endif /* DEBUG_EPT_MISCONFIG */ 438 return (VMEXIT_ABORT); 439 } 440 441 static int 442 vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 443 { 444 445 fprintf(stderr, "vm exit[%d]\n", *pvcpu); 446 fprintf(stderr, "\treason\t\tSVM\n"); 447 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip); 448 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length); 449 fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode); 450 fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1); 451 fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2); 452 return (VMEXIT_ABORT); 453 } 454 455 static int 456 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 457 { 458 459 assert(vmexit->inst_length == 0); 460 461 stats.vmexit_bogus++; 462 463 return (VMEXIT_CONTINUE); 464 } 465 466 static int 467 vmexit_reqidle(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 468 { 469 470 assert(vmexit->inst_length == 0); 471 472 stats.vmexit_reqidle++; 473 474 return (VMEXIT_CONTINUE); 475 } 476 477 static int 478 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 479 { 480 481 stats.vmexit_hlt++; 482 483 /* 484 * Just continue execution with the next instruction. We use 485 * the HLT VM exit as a way to be friendly with the host 486 * scheduler. 487 */ 488 return (VMEXIT_CONTINUE); 489 } 490 491 static int 492 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 493 { 494 495 stats.vmexit_pause++; 496 497 return (VMEXIT_CONTINUE); 498 } 499 500 static int 501 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 502 { 503 504 assert(vmexit->inst_length == 0); 505 506 stats.vmexit_mtrap++; 507 508 return (VMEXIT_CONTINUE); 509 } 510 511 static int 512 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 513 { 514 int err, i; 515 struct vie *vie; 516 517 stats.vmexit_inst_emul++; 518 519 vie = &vmexit->u.inst_emul.vie; 520 err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa, 521 vie, &vmexit->u.inst_emul.paging); 522 523 if (err) { 524 if (err == ESRCH) { 525 fprintf(stderr, "Unhandled memory access to 0x%lx\n", 526 vmexit->u.inst_emul.gpa); 527 } 528 529 fprintf(stderr, "Failed to emulate instruction ["); 530 for (i = 0; i < vie->num_valid; i++) { 531 fprintf(stderr, "0x%02x%s", vie->inst[i], 532 i != (vie->num_valid - 1) ? " " : ""); 533 } 534 fprintf(stderr, "] at 0x%lx\n", vmexit->rip); 535 return (VMEXIT_ABORT); 536 } 537 538 return (VMEXIT_CONTINUE); 539 } 540 541 static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER; 542 static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER; 543 544 static int 545 vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 546 { 547 enum vm_suspend_how how; 548 549 how = vmexit->u.suspended.how; 550 551 fbsdrun_deletecpu(ctx, *pvcpu); 552 553 if (*pvcpu != BSP) { 554 pthread_mutex_lock(&resetcpu_mtx); 555 pthread_cond_signal(&resetcpu_cond); 556 pthread_mutex_unlock(&resetcpu_mtx); 557 pthread_exit(NULL); 558 } 559 560 pthread_mutex_lock(&resetcpu_mtx); 561 while (!CPU_EMPTY(&cpumask)) { 562 pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx); 563 } 564 pthread_mutex_unlock(&resetcpu_mtx); 565 566 switch (how) { 567 case VM_SUSPEND_RESET: 568 exit(0); 569 case VM_SUSPEND_POWEROFF: 570 exit(1); 571 case VM_SUSPEND_HALT: 572 exit(2); 573 case VM_SUSPEND_TRIPLEFAULT: 574 exit(3); 575 default: 576 fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how); 577 exit(100); 578 } 579 return (0); /* NOTREACHED */ 580 } 581 582 static vmexit_handler_t handler[VM_EXITCODE_MAX] = { 583 [VM_EXITCODE_INOUT] = vmexit_inout, 584 [VM_EXITCODE_INOUT_STR] = vmexit_inout, 585 [VM_EXITCODE_VMX] = vmexit_vmx, 586 [VM_EXITCODE_SVM] = vmexit_svm, 587 [VM_EXITCODE_BOGUS] = vmexit_bogus, 588 [VM_EXITCODE_REQIDLE] = vmexit_reqidle, 589 [VM_EXITCODE_RDMSR] = vmexit_rdmsr, 590 [VM_EXITCODE_WRMSR] = vmexit_wrmsr, 591 [VM_EXITCODE_MTRAP] = vmexit_mtrap, 592 [VM_EXITCODE_INST_EMUL] = vmexit_inst_emul, 593 [VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap, 594 [VM_EXITCODE_SUSPENDED] = vmexit_suspend, 595 [VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch, 596 }; 597 598 static void 599 vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip) 600 { 601 int error, rc, prevcpu; 602 enum vm_exitcode exitcode; 603 cpuset_t active_cpus; 604 605 if (vcpumap[vcpu] != NULL) { 606 error = pthread_setaffinity_np(pthread_self(), 607 sizeof(cpuset_t), vcpumap[vcpu]); 608 assert(error == 0); 609 } 610 611 error = vm_active_cpus(ctx, &active_cpus); 612 assert(CPU_ISSET(vcpu, &active_cpus)); 613 614 error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip); 615 assert(error == 0); 616 617 while (1) { 618 error = vm_run(ctx, vcpu, &vmexit[vcpu]); 619 if (error != 0) 620 break; 621 622 prevcpu = vcpu; 623 624 exitcode = vmexit[vcpu].exitcode; 625 if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) { 626 fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n", 627 exitcode); 628 exit(1); 629 } 630 631 rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu); 632 633 switch (rc) { 634 case VMEXIT_CONTINUE: 635 break; 636 case VMEXIT_ABORT: 637 abort(); 638 default: 639 exit(1); 640 } 641 } 642 fprintf(stderr, "vm_run error %d, errno %d\n", error, errno); 643 } 644 645 static int 646 num_vcpus_allowed(struct vmctx *ctx) 647 { 648 int tmp, error; 649 650 error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp); 651 652 /* 653 * The guest is allowed to spinup more than one processor only if the 654 * UNRESTRICTED_GUEST capability is available. 655 */ 656 if (error == 0) 657 return (VM_MAXCPU); 658 else 659 return (1); 660 } 661 662 void 663 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu) 664 { 665 int err, tmp; 666 667 if (fbsdrun_vmexit_on_hlt()) { 668 err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp); 669 if (err < 0) { 670 fprintf(stderr, "VM exit on HLT not supported\n"); 671 exit(1); 672 } 673 vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1); 674 if (cpu == BSP) 675 handler[VM_EXITCODE_HLT] = vmexit_hlt; 676 } 677 678 if (fbsdrun_vmexit_on_pause()) { 679 /* 680 * pause exit support required for this mode 681 */ 682 err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp); 683 if (err < 0) { 684 fprintf(stderr, 685 "SMP mux requested, no pause support\n"); 686 exit(1); 687 } 688 vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1); 689 if (cpu == BSP) 690 handler[VM_EXITCODE_PAUSE] = vmexit_pause; 691 } 692 693 if (x2apic_mode) 694 err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED); 695 else 696 err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED); 697 698 if (err) { 699 fprintf(stderr, "Unable to set x2apic state (%d)\n", err); 700 exit(1); 701 } 702 703 vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1); 704 } 705 706 static struct vmctx * 707 do_open(const char *vmname) 708 { 709 struct vmctx *ctx; 710 int error; 711 bool reinit, romboot; 712 713 reinit = romboot = false; 714 715 if (lpc_bootrom()) 716 romboot = true; 717 718 error = vm_create(vmname); 719 if (error) { 720 if (errno == EEXIST) { 721 if (romboot) { 722 reinit = true; 723 } else { 724 /* 725 * The virtual machine has been setup by the 726 * userspace bootloader. 727 */ 728 } 729 } else { 730 perror("vm_create"); 731 exit(1); 732 } 733 } else { 734 if (!romboot) { 735 /* 736 * If the virtual machine was just created then a 737 * bootrom must be configured to boot it. 738 */ 739 fprintf(stderr, "virtual machine cannot be booted\n"); 740 exit(1); 741 } 742 } 743 744 ctx = vm_open(vmname); 745 if (ctx == NULL) { 746 perror("vm_open"); 747 exit(1); 748 } 749 750 if (reinit) { 751 error = vm_reinit(ctx); 752 if (error) { 753 perror("vm_reinit"); 754 exit(1); 755 } 756 } 757 return (ctx); 758 } 759 760 int 761 main(int argc, char *argv[]) 762 { 763 int c, error, gdb_port, err, bvmcons; 764 int max_vcpus, mptgen, memflags; 765 int rtc_localtime; 766 struct vmctx *ctx; 767 uint64_t rip; 768 size_t memsize; 769 char *optstr; 770 771 bvmcons = 0; 772 progname = basename(argv[0]); 773 gdb_port = 0; 774 guest_ncpus = 1; 775 memsize = 256 * MB; 776 mptgen = 1; 777 rtc_localtime = 1; 778 memflags = 0; 779 780 optstr = "abehuwxACHIPSWYp:g:c:s:m:l:U:"; 781 while ((c = getopt(argc, argv, optstr)) != -1) { 782 switch (c) { 783 case 'a': 784 x2apic_mode = 0; 785 break; 786 case 'A': 787 acpi = 1; 788 break; 789 case 'b': 790 bvmcons = 1; 791 break; 792 case 'p': 793 if (pincpu_parse(optarg) != 0) { 794 errx(EX_USAGE, "invalid vcpu pinning " 795 "configuration '%s'", optarg); 796 } 797 break; 798 case 'c': 799 guest_ncpus = atoi(optarg); 800 break; 801 case 'C': 802 memflags |= VM_MEM_F_INCORE; 803 break; 804 case 'g': 805 gdb_port = atoi(optarg); 806 break; 807 case 'l': 808 if (lpc_device_parse(optarg) != 0) { 809 errx(EX_USAGE, "invalid lpc device " 810 "configuration '%s'", optarg); 811 } 812 break; 813 case 's': 814 if (pci_parse_slot(optarg) != 0) 815 exit(1); 816 else 817 break; 818 case 'S': 819 memflags |= VM_MEM_F_WIRED; 820 break; 821 case 'm': 822 error = vm_parse_memsize(optarg, &memsize); 823 if (error) 824 errx(EX_USAGE, "invalid memsize '%s'", optarg); 825 break; 826 case 'H': 827 guest_vmexit_on_hlt = 1; 828 break; 829 case 'I': 830 /* 831 * The "-I" option was used to add an ioapic to the 832 * virtual machine. 833 * 834 * An ioapic is now provided unconditionally for each 835 * virtual machine and this option is now deprecated. 836 */ 837 break; 838 case 'P': 839 guest_vmexit_on_pause = 1; 840 break; 841 case 'e': 842 strictio = 1; 843 break; 844 case 'u': 845 rtc_localtime = 0; 846 break; 847 case 'U': 848 guest_uuid_str = optarg; 849 break; 850 case 'w': 851 strictmsr = 0; 852 break; 853 case 'W': 854 virtio_msix = 0; 855 break; 856 case 'x': 857 x2apic_mode = 1; 858 break; 859 case 'Y': 860 mptgen = 0; 861 break; 862 case 'h': 863 usage(0); 864 default: 865 usage(1); 866 } 867 } 868 argc -= optind; 869 argv += optind; 870 871 if (argc != 1) 872 usage(1); 873 874 vmname = argv[0]; 875 ctx = do_open(vmname); 876 877 if (guest_ncpus < 1) { 878 fprintf(stderr, "Invalid guest vCPUs (%d)\n", guest_ncpus); 879 exit(1); 880 } 881 882 max_vcpus = num_vcpus_allowed(ctx); 883 if (guest_ncpus > max_vcpus) { 884 fprintf(stderr, "%d vCPUs requested but only %d available\n", 885 guest_ncpus, max_vcpus); 886 exit(1); 887 } 888 889 fbsdrun_set_capabilities(ctx, BSP); 890 891 vm_set_memflags(ctx, memflags); 892 err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL); 893 if (err) { 894 fprintf(stderr, "Unable to setup memory (%d)\n", errno); 895 exit(1); 896 } 897 898 error = init_msr(); 899 if (error) { 900 fprintf(stderr, "init_msr error %d", error); 901 exit(1); 902 } 903 904 init_mem(); 905 init_inout(); 906 pci_irq_init(ctx); 907 ioapic_init(ctx); 908 909 rtc_init(ctx, rtc_localtime); 910 sci_init(ctx); 911 912 /* 913 * Exit if a device emulation finds an error in it's initilization 914 */ 915 if (init_pci(ctx) != 0) 916 exit(1); 917 918 if (gdb_port != 0) 919 init_dbgport(gdb_port); 920 921 if (bvmcons) 922 init_bvmcons(); 923 924 if (lpc_bootrom()) { 925 if (vm_set_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, 1)) { 926 fprintf(stderr, "ROM boot failed: unrestricted guest " 927 "capability not available\n"); 928 exit(1); 929 } 930 error = vcpu_reset(ctx, BSP); 931 assert(error == 0); 932 } 933 934 error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip); 935 assert(error == 0); 936 937 /* 938 * build the guest tables, MP etc. 939 */ 940 if (mptgen) { 941 error = mptable_build(ctx, guest_ncpus); 942 if (error) 943 exit(1); 944 } 945 946 error = smbios_build(ctx); 947 assert(error == 0); 948 949 if (acpi) { 950 error = acpi_build(ctx, guest_ncpus); 951 assert(error == 0); 952 } 953 954 if (lpc_bootrom()) 955 fwctl_init(); 956 957 /* 958 * Change the proc title to include the VM name. 959 */ 960 setproctitle("%s", vmname); 961 962 /* 963 * Add CPU 0 964 */ 965 fbsdrun_addcpu(ctx, BSP, BSP, rip); 966 967 /* 968 * Head off to the main event dispatch loop 969 */ 970 mevent_dispatch(); 971 972 exit(1); 973 } 974