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 if (error != 0) 265 err(EX_OSERR, "could not activate CPU %d", newcpu); 266 267 CPU_SET_ATOMIC(newcpu, &cpumask); 268 269 /* 270 * Set up the vmexit struct to allow execution to start 271 * at the given RIP 272 */ 273 vmexit[newcpu].rip = rip; 274 vmexit[newcpu].inst_length = 0; 275 276 mt_vmm_info[newcpu].mt_ctx = ctx; 277 mt_vmm_info[newcpu].mt_vcpu = newcpu; 278 279 error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL, 280 fbsdrun_start_thread, &mt_vmm_info[newcpu]); 281 assert(error == 0); 282 } 283 284 static int 285 fbsdrun_deletecpu(struct vmctx *ctx, int vcpu) 286 { 287 288 if (!CPU_ISSET(vcpu, &cpumask)) { 289 fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu); 290 exit(1); 291 } 292 293 CPU_CLR_ATOMIC(vcpu, &cpumask); 294 return (CPU_EMPTY(&cpumask)); 295 } 296 297 static int 298 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu, 299 uint32_t eax) 300 { 301 #if BHYVE_DEBUG 302 /* 303 * put guest-driven debug here 304 */ 305 #endif 306 return (VMEXIT_CONTINUE); 307 } 308 309 static int 310 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 311 { 312 int error; 313 int bytes, port, in, out, string; 314 int vcpu; 315 316 vcpu = *pvcpu; 317 318 port = vme->u.inout.port; 319 bytes = vme->u.inout.bytes; 320 string = vme->u.inout.string; 321 in = vme->u.inout.in; 322 out = !in; 323 324 /* Extra-special case of host notifications */ 325 if (out && port == GUEST_NIO_PORT) { 326 error = vmexit_handle_notify(ctx, vme, pvcpu, vme->u.inout.eax); 327 return (error); 328 } 329 330 error = emulate_inout(ctx, vcpu, vme, strictio); 331 if (error) { 332 fprintf(stderr, "Unhandled %s%c 0x%04x at 0x%lx\n", 333 in ? "in" : "out", 334 bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), 335 port, vmexit->rip); 336 return (VMEXIT_ABORT); 337 } else { 338 return (VMEXIT_CONTINUE); 339 } 340 } 341 342 static int 343 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 344 { 345 uint64_t val; 346 uint32_t eax, edx; 347 int error; 348 349 val = 0; 350 error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val); 351 if (error != 0) { 352 fprintf(stderr, "rdmsr to register %#x on vcpu %d\n", 353 vme->u.msr.code, *pvcpu); 354 if (strictmsr) { 355 vm_inject_gp(ctx, *pvcpu); 356 return (VMEXIT_CONTINUE); 357 } 358 } 359 360 eax = val; 361 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax); 362 assert(error == 0); 363 364 edx = val >> 32; 365 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx); 366 assert(error == 0); 367 368 return (VMEXIT_CONTINUE); 369 } 370 371 static int 372 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 373 { 374 int error; 375 376 error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval); 377 if (error != 0) { 378 fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n", 379 vme->u.msr.code, vme->u.msr.wval, *pvcpu); 380 if (strictmsr) { 381 vm_inject_gp(ctx, *pvcpu); 382 return (VMEXIT_CONTINUE); 383 } 384 } 385 return (VMEXIT_CONTINUE); 386 } 387 388 static int 389 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 390 { 391 int newcpu; 392 int retval = VMEXIT_CONTINUE; 393 394 newcpu = spinup_ap(ctx, *pvcpu, 395 vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip); 396 397 return (retval); 398 } 399 400 #define DEBUG_EPT_MISCONFIG 401 #ifdef DEBUG_EPT_MISCONFIG 402 #define EXIT_REASON_EPT_MISCONFIG 49 403 #define VMCS_GUEST_PHYSICAL_ADDRESS 0x00002400 404 #define VMCS_IDENT(x) ((x) | 0x80000000) 405 406 static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4]; 407 static int ept_misconfig_ptenum; 408 #endif 409 410 static int 411 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 412 { 413 414 fprintf(stderr, "vm exit[%d]\n", *pvcpu); 415 fprintf(stderr, "\treason\t\tVMX\n"); 416 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip); 417 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length); 418 fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status); 419 fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason); 420 fprintf(stderr, "\tqualification\t0x%016lx\n", 421 vmexit->u.vmx.exit_qualification); 422 fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type); 423 fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error); 424 #ifdef DEBUG_EPT_MISCONFIG 425 if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) { 426 vm_get_register(ctx, *pvcpu, 427 VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS), 428 &ept_misconfig_gpa); 429 vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte, 430 &ept_misconfig_ptenum); 431 fprintf(stderr, "\tEPT misconfiguration:\n"); 432 fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa); 433 fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n", 434 ept_misconfig_ptenum, ept_misconfig_pte[0], 435 ept_misconfig_pte[1], ept_misconfig_pte[2], 436 ept_misconfig_pte[3]); 437 } 438 #endif /* DEBUG_EPT_MISCONFIG */ 439 return (VMEXIT_ABORT); 440 } 441 442 static int 443 vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 444 { 445 446 fprintf(stderr, "vm exit[%d]\n", *pvcpu); 447 fprintf(stderr, "\treason\t\tSVM\n"); 448 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip); 449 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length); 450 fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode); 451 fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1); 452 fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2); 453 return (VMEXIT_ABORT); 454 } 455 456 static int 457 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 458 { 459 460 assert(vmexit->inst_length == 0); 461 462 stats.vmexit_bogus++; 463 464 return (VMEXIT_CONTINUE); 465 } 466 467 static int 468 vmexit_reqidle(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 469 { 470 471 assert(vmexit->inst_length == 0); 472 473 stats.vmexit_reqidle++; 474 475 return (VMEXIT_CONTINUE); 476 } 477 478 static int 479 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 480 { 481 482 stats.vmexit_hlt++; 483 484 /* 485 * Just continue execution with the next instruction. We use 486 * the HLT VM exit as a way to be friendly with the host 487 * scheduler. 488 */ 489 return (VMEXIT_CONTINUE); 490 } 491 492 static int 493 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 494 { 495 496 stats.vmexit_pause++; 497 498 return (VMEXIT_CONTINUE); 499 } 500 501 static int 502 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 503 { 504 505 assert(vmexit->inst_length == 0); 506 507 stats.vmexit_mtrap++; 508 509 return (VMEXIT_CONTINUE); 510 } 511 512 static int 513 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 514 { 515 int err, i; 516 struct vie *vie; 517 518 stats.vmexit_inst_emul++; 519 520 vie = &vmexit->u.inst_emul.vie; 521 err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa, 522 vie, &vmexit->u.inst_emul.paging); 523 524 if (err) { 525 if (err == ESRCH) { 526 fprintf(stderr, "Unhandled memory access to 0x%lx\n", 527 vmexit->u.inst_emul.gpa); 528 } 529 530 fprintf(stderr, "Failed to emulate instruction ["); 531 for (i = 0; i < vie->num_valid; i++) { 532 fprintf(stderr, "0x%02x%s", vie->inst[i], 533 i != (vie->num_valid - 1) ? " " : ""); 534 } 535 fprintf(stderr, "] at 0x%lx\n", vmexit->rip); 536 return (VMEXIT_ABORT); 537 } 538 539 return (VMEXIT_CONTINUE); 540 } 541 542 static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER; 543 static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER; 544 545 static int 546 vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 547 { 548 enum vm_suspend_how how; 549 550 how = vmexit->u.suspended.how; 551 552 fbsdrun_deletecpu(ctx, *pvcpu); 553 554 if (*pvcpu != BSP) { 555 pthread_mutex_lock(&resetcpu_mtx); 556 pthread_cond_signal(&resetcpu_cond); 557 pthread_mutex_unlock(&resetcpu_mtx); 558 pthread_exit(NULL); 559 } 560 561 pthread_mutex_lock(&resetcpu_mtx); 562 while (!CPU_EMPTY(&cpumask)) { 563 pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx); 564 } 565 pthread_mutex_unlock(&resetcpu_mtx); 566 567 switch (how) { 568 case VM_SUSPEND_RESET: 569 exit(0); 570 case VM_SUSPEND_POWEROFF: 571 exit(1); 572 case VM_SUSPEND_HALT: 573 exit(2); 574 case VM_SUSPEND_TRIPLEFAULT: 575 exit(3); 576 default: 577 fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how); 578 exit(100); 579 } 580 return (0); /* NOTREACHED */ 581 } 582 583 static vmexit_handler_t handler[VM_EXITCODE_MAX] = { 584 [VM_EXITCODE_INOUT] = vmexit_inout, 585 [VM_EXITCODE_INOUT_STR] = vmexit_inout, 586 [VM_EXITCODE_VMX] = vmexit_vmx, 587 [VM_EXITCODE_SVM] = vmexit_svm, 588 [VM_EXITCODE_BOGUS] = vmexit_bogus, 589 [VM_EXITCODE_REQIDLE] = vmexit_reqidle, 590 [VM_EXITCODE_RDMSR] = vmexit_rdmsr, 591 [VM_EXITCODE_WRMSR] = vmexit_wrmsr, 592 [VM_EXITCODE_MTRAP] = vmexit_mtrap, 593 [VM_EXITCODE_INST_EMUL] = vmexit_inst_emul, 594 [VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap, 595 [VM_EXITCODE_SUSPENDED] = vmexit_suspend, 596 [VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch, 597 }; 598 599 static void 600 vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip) 601 { 602 int error, rc, prevcpu; 603 enum vm_exitcode exitcode; 604 cpuset_t active_cpus; 605 606 if (vcpumap[vcpu] != NULL) { 607 error = pthread_setaffinity_np(pthread_self(), 608 sizeof(cpuset_t), vcpumap[vcpu]); 609 assert(error == 0); 610 } 611 612 error = vm_active_cpus(ctx, &active_cpus); 613 assert(CPU_ISSET(vcpu, &active_cpus)); 614 615 error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip); 616 assert(error == 0); 617 618 while (1) { 619 error = vm_run(ctx, vcpu, &vmexit[vcpu]); 620 if (error != 0) 621 break; 622 623 prevcpu = vcpu; 624 625 exitcode = vmexit[vcpu].exitcode; 626 if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) { 627 fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n", 628 exitcode); 629 exit(1); 630 } 631 632 rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu); 633 634 switch (rc) { 635 case VMEXIT_CONTINUE: 636 break; 637 case VMEXIT_ABORT: 638 abort(); 639 default: 640 exit(1); 641 } 642 } 643 fprintf(stderr, "vm_run error %d, errno %d\n", error, errno); 644 } 645 646 static int 647 num_vcpus_allowed(struct vmctx *ctx) 648 { 649 int tmp, error; 650 651 error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp); 652 653 /* 654 * The guest is allowed to spinup more than one processor only if the 655 * UNRESTRICTED_GUEST capability is available. 656 */ 657 if (error == 0) 658 return (VM_MAXCPU); 659 else 660 return (1); 661 } 662 663 void 664 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu) 665 { 666 int err, tmp; 667 668 if (fbsdrun_vmexit_on_hlt()) { 669 err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp); 670 if (err < 0) { 671 fprintf(stderr, "VM exit on HLT not supported\n"); 672 exit(1); 673 } 674 vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1); 675 if (cpu == BSP) 676 handler[VM_EXITCODE_HLT] = vmexit_hlt; 677 } 678 679 if (fbsdrun_vmexit_on_pause()) { 680 /* 681 * pause exit support required for this mode 682 */ 683 err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp); 684 if (err < 0) { 685 fprintf(stderr, 686 "SMP mux requested, no pause support\n"); 687 exit(1); 688 } 689 vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1); 690 if (cpu == BSP) 691 handler[VM_EXITCODE_PAUSE] = vmexit_pause; 692 } 693 694 if (x2apic_mode) 695 err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED); 696 else 697 err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED); 698 699 if (err) { 700 fprintf(stderr, "Unable to set x2apic state (%d)\n", err); 701 exit(1); 702 } 703 704 vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1); 705 } 706 707 static struct vmctx * 708 do_open(const char *vmname) 709 { 710 struct vmctx *ctx; 711 int error; 712 bool reinit, romboot; 713 714 reinit = romboot = false; 715 716 if (lpc_bootrom()) 717 romboot = true; 718 719 error = vm_create(vmname); 720 if (error) { 721 if (errno == EEXIST) { 722 if (romboot) { 723 reinit = true; 724 } else { 725 /* 726 * The virtual machine has been setup by the 727 * userspace bootloader. 728 */ 729 } 730 } else { 731 perror("vm_create"); 732 exit(1); 733 } 734 } else { 735 if (!romboot) { 736 /* 737 * If the virtual machine was just created then a 738 * bootrom must be configured to boot it. 739 */ 740 fprintf(stderr, "virtual machine cannot be booted\n"); 741 exit(1); 742 } 743 } 744 745 ctx = vm_open(vmname); 746 if (ctx == NULL) { 747 perror("vm_open"); 748 exit(1); 749 } 750 751 if (reinit) { 752 error = vm_reinit(ctx); 753 if (error) { 754 perror("vm_reinit"); 755 exit(1); 756 } 757 } 758 return (ctx); 759 } 760 761 int 762 main(int argc, char *argv[]) 763 { 764 int c, error, gdb_port, err, bvmcons; 765 int max_vcpus, mptgen, memflags; 766 int rtc_localtime; 767 struct vmctx *ctx; 768 uint64_t rip; 769 size_t memsize; 770 char *optstr; 771 772 bvmcons = 0; 773 progname = basename(argv[0]); 774 gdb_port = 0; 775 guest_ncpus = 1; 776 memsize = 256 * MB; 777 mptgen = 1; 778 rtc_localtime = 1; 779 memflags = 0; 780 781 optstr = "abehuwxACHIPSWYp:g:c:s:m:l:U:"; 782 while ((c = getopt(argc, argv, optstr)) != -1) { 783 switch (c) { 784 case 'a': 785 x2apic_mode = 0; 786 break; 787 case 'A': 788 acpi = 1; 789 break; 790 case 'b': 791 bvmcons = 1; 792 break; 793 case 'p': 794 if (pincpu_parse(optarg) != 0) { 795 errx(EX_USAGE, "invalid vcpu pinning " 796 "configuration '%s'", optarg); 797 } 798 break; 799 case 'c': 800 guest_ncpus = atoi(optarg); 801 break; 802 case 'C': 803 memflags |= VM_MEM_F_INCORE; 804 break; 805 case 'g': 806 gdb_port = atoi(optarg); 807 break; 808 case 'l': 809 if (lpc_device_parse(optarg) != 0) { 810 errx(EX_USAGE, "invalid lpc device " 811 "configuration '%s'", optarg); 812 } 813 break; 814 case 's': 815 if (pci_parse_slot(optarg) != 0) 816 exit(1); 817 else 818 break; 819 case 'S': 820 memflags |= VM_MEM_F_WIRED; 821 break; 822 case 'm': 823 error = vm_parse_memsize(optarg, &memsize); 824 if (error) 825 errx(EX_USAGE, "invalid memsize '%s'", optarg); 826 break; 827 case 'H': 828 guest_vmexit_on_hlt = 1; 829 break; 830 case 'I': 831 /* 832 * The "-I" option was used to add an ioapic to the 833 * virtual machine. 834 * 835 * An ioapic is now provided unconditionally for each 836 * virtual machine and this option is now deprecated. 837 */ 838 break; 839 case 'P': 840 guest_vmexit_on_pause = 1; 841 break; 842 case 'e': 843 strictio = 1; 844 break; 845 case 'u': 846 rtc_localtime = 0; 847 break; 848 case 'U': 849 guest_uuid_str = optarg; 850 break; 851 case 'w': 852 strictmsr = 0; 853 break; 854 case 'W': 855 virtio_msix = 0; 856 break; 857 case 'x': 858 x2apic_mode = 1; 859 break; 860 case 'Y': 861 mptgen = 0; 862 break; 863 case 'h': 864 usage(0); 865 default: 866 usage(1); 867 } 868 } 869 argc -= optind; 870 argv += optind; 871 872 if (argc != 1) 873 usage(1); 874 875 vmname = argv[0]; 876 ctx = do_open(vmname); 877 878 if (guest_ncpus < 1) { 879 fprintf(stderr, "Invalid guest vCPUs (%d)\n", guest_ncpus); 880 exit(1); 881 } 882 883 max_vcpus = num_vcpus_allowed(ctx); 884 if (guest_ncpus > max_vcpus) { 885 fprintf(stderr, "%d vCPUs requested but only %d available\n", 886 guest_ncpus, max_vcpus); 887 exit(1); 888 } 889 890 fbsdrun_set_capabilities(ctx, BSP); 891 892 vm_set_memflags(ctx, memflags); 893 err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL); 894 if (err) { 895 fprintf(stderr, "Unable to setup memory (%d)\n", errno); 896 exit(1); 897 } 898 899 error = init_msr(); 900 if (error) { 901 fprintf(stderr, "init_msr error %d", error); 902 exit(1); 903 } 904 905 init_mem(); 906 init_inout(); 907 pci_irq_init(ctx); 908 ioapic_init(ctx); 909 910 rtc_init(ctx, rtc_localtime); 911 sci_init(ctx); 912 913 /* 914 * Exit if a device emulation finds an error in it's initilization 915 */ 916 if (init_pci(ctx) != 0) 917 exit(1); 918 919 if (gdb_port != 0) 920 init_dbgport(gdb_port); 921 922 if (bvmcons) 923 init_bvmcons(); 924 925 if (lpc_bootrom()) { 926 if (vm_set_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, 1)) { 927 fprintf(stderr, "ROM boot failed: unrestricted guest " 928 "capability not available\n"); 929 exit(1); 930 } 931 error = vcpu_reset(ctx, BSP); 932 assert(error == 0); 933 } 934 935 error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip); 936 assert(error == 0); 937 938 /* 939 * build the guest tables, MP etc. 940 */ 941 if (mptgen) { 942 error = mptable_build(ctx, guest_ncpus); 943 if (error) 944 exit(1); 945 } 946 947 error = smbios_build(ctx); 948 assert(error == 0); 949 950 if (acpi) { 951 error = acpi_build(ctx, guest_ncpus); 952 assert(error == 0); 953 } 954 955 if (lpc_bootrom()) 956 fwctl_init(); 957 958 /* 959 * Change the proc title to include the VM name. 960 */ 961 setproctitle("%s", vmname); 962 963 /* 964 * Add CPU 0 965 */ 966 fbsdrun_addcpu(ctx, BSP, BSP, rip); 967 968 /* 969 * Head off to the main event dispatch loop 970 */ 971 mevent_dispatch(); 972 973 exit(1); 974 } 975