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 51 #include <machine/vmm.h> 52 #include <vmmapi.h> 53 54 #include "bhyverun.h" 55 #include "acpi.h" 56 #include "inout.h" 57 #include "dbgport.h" 58 #include "ioapic.h" 59 #include "mem.h" 60 #include "mevent.h" 61 #include "mptbl.h" 62 #include "pci_emul.h" 63 #include "pci_lpc.h" 64 #include "xmsr.h" 65 #include "spinup_ap.h" 66 #include "rtc.h" 67 68 #define GUEST_NIO_PORT 0x488 /* guest upcalls via i/o port */ 69 70 #define VMEXIT_SWITCH 0 /* force vcpu switch in mux mode */ 71 #define VMEXIT_CONTINUE 1 /* continue from next instruction */ 72 #define VMEXIT_RESTART 2 /* restart current instruction */ 73 #define VMEXIT_ABORT 3 /* abort the vm run loop */ 74 #define VMEXIT_RESET 4 /* guest machine has reset */ 75 #define VMEXIT_POWEROFF 5 /* guest machine has powered off */ 76 77 #define MB (1024UL * 1024) 78 #define GB (1024UL * MB) 79 80 typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu); 81 82 char *vmname; 83 84 int guest_ncpus; 85 86 static int pincpu = -1; 87 static int guest_vmexit_on_hlt, guest_vmexit_on_pause; 88 static int virtio_msix = 1; 89 static int x2apic_mode = 0; /* default is xAPIC */ 90 91 static int strictio; 92 static int strictmsr = 1; 93 94 static int acpi; 95 96 static char *progname; 97 static const int BSP = 0; 98 99 static int cpumask; 100 101 static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip); 102 103 struct vm_exit vmexit[VM_MAXCPU]; 104 105 struct bhyvestats { 106 uint64_t vmexit_bogus; 107 uint64_t vmexit_bogus_switch; 108 uint64_t vmexit_hlt; 109 uint64_t vmexit_pause; 110 uint64_t vmexit_mtrap; 111 uint64_t vmexit_inst_emul; 112 uint64_t cpu_switch_rotate; 113 uint64_t cpu_switch_direct; 114 int io_reset; 115 } stats; 116 117 struct mt_vmm_info { 118 pthread_t mt_thr; 119 struct vmctx *mt_ctx; 120 int mt_vcpu; 121 } mt_vmm_info[VM_MAXCPU]; 122 123 static void 124 usage(int code) 125 { 126 127 fprintf(stderr, 128 "Usage: %s [-aehwAHIPW] [-g <gdb port>] [-s <pci>]\n" 129 " %*s [-c vcpus] [-p pincpu] [-m mem] [-l <lpc>] <vm>\n" 130 " -a: local apic is in xAPIC mode (deprecated)\n" 131 " -A: create an ACPI table\n" 132 " -g: gdb port\n" 133 " -c: # cpus (default 1)\n" 134 " -p: pin vcpu 'n' to host cpu 'pincpu + n'\n" 135 " -H: vmexit from the guest on hlt\n" 136 " -P: vmexit from the guest on pause\n" 137 " -W: force virtio to use single-vector MSI\n" 138 " -e: exit on unhandled I/O access\n" 139 " -h: help\n" 140 " -s: <slot,driver,configinfo> PCI slot config\n" 141 " -l: LPC device configuration\n" 142 " -m: memory size in MB\n" 143 " -w: ignore unimplemented MSRs\n" 144 " -x: local apic is in x2APIC mode\n", 145 progname, (int)strlen(progname), ""); 146 147 exit(code); 148 } 149 150 void * 151 paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len) 152 { 153 154 return (vm_map_gpa(ctx, gaddr, len)); 155 } 156 157 int 158 fbsdrun_vmexit_on_pause(void) 159 { 160 161 return (guest_vmexit_on_pause); 162 } 163 164 int 165 fbsdrun_vmexit_on_hlt(void) 166 { 167 168 return (guest_vmexit_on_hlt); 169 } 170 171 int 172 fbsdrun_virtio_msix(void) 173 { 174 175 return (virtio_msix); 176 } 177 178 static void * 179 fbsdrun_start_thread(void *param) 180 { 181 char tname[MAXCOMLEN + 1]; 182 struct mt_vmm_info *mtp; 183 int vcpu; 184 185 mtp = param; 186 vcpu = mtp->mt_vcpu; 187 188 snprintf(tname, sizeof(tname), "vcpu %d", vcpu); 189 pthread_set_name_np(mtp->mt_thr, tname); 190 191 vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip); 192 193 /* not reached */ 194 exit(1); 195 return (NULL); 196 } 197 198 void 199 fbsdrun_addcpu(struct vmctx *ctx, int vcpu, uint64_t rip) 200 { 201 int error; 202 203 if (cpumask & (1 << vcpu)) { 204 fprintf(stderr, "addcpu: attempting to add existing cpu %d\n", 205 vcpu); 206 exit(1); 207 } 208 209 atomic_set_int(&cpumask, 1 << vcpu); 210 211 /* 212 * Set up the vmexit struct to allow execution to start 213 * at the given RIP 214 */ 215 vmexit[vcpu].rip = rip; 216 vmexit[vcpu].inst_length = 0; 217 218 mt_vmm_info[vcpu].mt_ctx = ctx; 219 mt_vmm_info[vcpu].mt_vcpu = vcpu; 220 221 error = pthread_create(&mt_vmm_info[vcpu].mt_thr, NULL, 222 fbsdrun_start_thread, &mt_vmm_info[vcpu]); 223 assert(error == 0); 224 } 225 226 static int 227 fbsdrun_deletecpu(struct vmctx *ctx, int vcpu) 228 { 229 230 if ((cpumask & (1 << vcpu)) == 0) { 231 fprintf(stderr, "addcpu: attempting to delete unknown cpu %d\n", 232 vcpu); 233 exit(1); 234 } 235 236 atomic_clear_int(&cpumask, 1 << vcpu); 237 return (cpumask == 0); 238 } 239 240 static int 241 vmexit_catch_reset(void) 242 { 243 stats.io_reset++; 244 return (VMEXIT_RESET); 245 } 246 247 static int 248 vmexit_catch_inout(void) 249 { 250 return (VMEXIT_ABORT); 251 } 252 253 static int 254 vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu, 255 uint32_t eax) 256 { 257 #if BHYVE_DEBUG 258 /* 259 * put guest-driven debug here 260 */ 261 #endif 262 return (VMEXIT_CONTINUE); 263 } 264 265 static int 266 vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 267 { 268 int error; 269 int bytes, port, in, out; 270 uint32_t eax; 271 int vcpu; 272 273 vcpu = *pvcpu; 274 275 port = vme->u.inout.port; 276 bytes = vme->u.inout.bytes; 277 eax = vme->u.inout.eax; 278 in = vme->u.inout.in; 279 out = !in; 280 281 /* We don't deal with these */ 282 if (vme->u.inout.string || vme->u.inout.rep) 283 return (VMEXIT_ABORT); 284 285 /* Special case of guest reset */ 286 if (out && port == 0x64 && (uint8_t)eax == 0xFE) 287 return (vmexit_catch_reset()); 288 289 /* Extra-special case of host notifications */ 290 if (out && port == GUEST_NIO_PORT) 291 return (vmexit_handle_notify(ctx, vme, pvcpu, eax)); 292 293 error = emulate_inout(ctx, vcpu, in, port, bytes, &eax, strictio); 294 if (error == INOUT_OK && in) 295 error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RAX, eax); 296 297 switch (error) { 298 case INOUT_OK: 299 return (VMEXIT_CONTINUE); 300 case INOUT_RESET: 301 return (VMEXIT_RESET); 302 case INOUT_POWEROFF: 303 return (VMEXIT_POWEROFF); 304 default: 305 fprintf(stderr, "Unhandled %s%c 0x%04x\n", 306 in ? "in" : "out", 307 bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), port); 308 return (vmexit_catch_inout()); 309 } 310 } 311 312 static int 313 vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 314 { 315 uint64_t val; 316 uint32_t eax, edx; 317 int error; 318 319 val = 0; 320 error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val); 321 if (error != 0) { 322 fprintf(stderr, "rdmsr to register %#x on vcpu %d\n", 323 vme->u.msr.code, *pvcpu); 324 if (strictmsr) 325 return (VMEXIT_ABORT); 326 } 327 328 eax = val; 329 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax); 330 assert(error == 0); 331 332 edx = val >> 32; 333 error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx); 334 assert(error == 0); 335 336 return (VMEXIT_CONTINUE); 337 } 338 339 static int 340 vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 341 { 342 int error; 343 344 error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval); 345 if (error != 0) { 346 fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n", 347 vme->u.msr.code, vme->u.msr.wval, *pvcpu); 348 if (strictmsr) 349 return (VMEXIT_ABORT); 350 } 351 return (VMEXIT_CONTINUE); 352 } 353 354 static int 355 vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 356 { 357 int newcpu; 358 int retval = VMEXIT_CONTINUE; 359 360 newcpu = spinup_ap(ctx, *pvcpu, 361 vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip); 362 363 return (retval); 364 } 365 366 static int 367 vmexit_spindown_cpu(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) 368 { 369 int lastcpu; 370 371 lastcpu = fbsdrun_deletecpu(ctx, *pvcpu); 372 if (!lastcpu) 373 pthread_exit(NULL); 374 return (vmexit_catch_reset()); 375 } 376 377 static int 378 vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 379 { 380 381 fprintf(stderr, "vm exit[%d]\n", *pvcpu); 382 fprintf(stderr, "\treason\t\tVMX\n"); 383 fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip); 384 fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length); 385 fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status); 386 fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason); 387 fprintf(stderr, "\tqualification\t0x%016lx\n", 388 vmexit->u.vmx.exit_qualification); 389 fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type); 390 fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error); 391 392 return (VMEXIT_ABORT); 393 } 394 395 static int 396 vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 397 { 398 399 stats.vmexit_bogus++; 400 401 return (VMEXIT_RESTART); 402 } 403 404 static int 405 vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 406 { 407 408 stats.vmexit_hlt++; 409 410 /* 411 * Just continue execution with the next instruction. We use 412 * the HLT VM exit as a way to be friendly with the host 413 * scheduler. 414 */ 415 return (VMEXIT_CONTINUE); 416 } 417 418 static int 419 vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 420 { 421 422 stats.vmexit_pause++; 423 424 return (VMEXIT_CONTINUE); 425 } 426 427 static int 428 vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 429 { 430 431 stats.vmexit_mtrap++; 432 433 return (VMEXIT_RESTART); 434 } 435 436 static int 437 vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) 438 { 439 int err; 440 stats.vmexit_inst_emul++; 441 442 err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa, 443 &vmexit->u.inst_emul.vie); 444 445 if (err) { 446 if (err == EINVAL) { 447 fprintf(stderr, 448 "Failed to emulate instruction at 0x%lx\n", 449 vmexit->rip); 450 } else if (err == ESRCH) { 451 fprintf(stderr, "Unhandled memory access to 0x%lx\n", 452 vmexit->u.inst_emul.gpa); 453 } 454 455 return (VMEXIT_ABORT); 456 } 457 458 return (VMEXIT_CONTINUE); 459 } 460 461 static vmexit_handler_t handler[VM_EXITCODE_MAX] = { 462 [VM_EXITCODE_INOUT] = vmexit_inout, 463 [VM_EXITCODE_VMX] = vmexit_vmx, 464 [VM_EXITCODE_BOGUS] = vmexit_bogus, 465 [VM_EXITCODE_RDMSR] = vmexit_rdmsr, 466 [VM_EXITCODE_WRMSR] = vmexit_wrmsr, 467 [VM_EXITCODE_MTRAP] = vmexit_mtrap, 468 [VM_EXITCODE_INST_EMUL] = vmexit_inst_emul, 469 [VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap, 470 [VM_EXITCODE_SPINDOWN_CPU] = vmexit_spindown_cpu, 471 }; 472 473 static void 474 vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip) 475 { 476 cpuset_t mask; 477 int error, rc, prevcpu; 478 enum vm_exitcode exitcode; 479 480 if (pincpu >= 0) { 481 CPU_ZERO(&mask); 482 CPU_SET(pincpu + vcpu, &mask); 483 error = pthread_setaffinity_np(pthread_self(), 484 sizeof(mask), &mask); 485 assert(error == 0); 486 } 487 488 while (1) { 489 error = vm_run(ctx, vcpu, rip, &vmexit[vcpu]); 490 if (error != 0) 491 break; 492 493 prevcpu = vcpu; 494 495 exitcode = vmexit[vcpu].exitcode; 496 if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) { 497 fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n", 498 exitcode); 499 exit(1); 500 } 501 502 rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu); 503 504 switch (rc) { 505 case VMEXIT_CONTINUE: 506 rip = vmexit[vcpu].rip + vmexit[vcpu].inst_length; 507 break; 508 case VMEXIT_RESTART: 509 rip = vmexit[vcpu].rip; 510 break; 511 case VMEXIT_RESET: 512 exit(0); 513 default: 514 exit(1); 515 } 516 } 517 fprintf(stderr, "vm_run error %d, errno %d\n", error, errno); 518 } 519 520 static int 521 num_vcpus_allowed(struct vmctx *ctx) 522 { 523 int tmp, error; 524 525 error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp); 526 527 /* 528 * The guest is allowed to spinup more than one processor only if the 529 * UNRESTRICTED_GUEST capability is available. 530 */ 531 if (error == 0) 532 return (VM_MAXCPU); 533 else 534 return (1); 535 } 536 537 void 538 fbsdrun_set_capabilities(struct vmctx *ctx, int cpu) 539 { 540 int err, tmp; 541 542 if (fbsdrun_vmexit_on_hlt()) { 543 err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp); 544 if (err < 0) { 545 fprintf(stderr, "VM exit on HLT not supported\n"); 546 exit(1); 547 } 548 vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1); 549 if (cpu == BSP) 550 handler[VM_EXITCODE_HLT] = vmexit_hlt; 551 } 552 553 if (fbsdrun_vmexit_on_pause()) { 554 /* 555 * pause exit support required for this mode 556 */ 557 err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp); 558 if (err < 0) { 559 fprintf(stderr, 560 "SMP mux requested, no pause support\n"); 561 exit(1); 562 } 563 vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1); 564 if (cpu == BSP) 565 handler[VM_EXITCODE_PAUSE] = vmexit_pause; 566 } 567 568 if (x2apic_mode) 569 err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED); 570 else 571 err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED); 572 573 if (err) { 574 fprintf(stderr, "Unable to set x2apic state (%d)\n", err); 575 exit(1); 576 } 577 578 vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1); 579 } 580 581 int 582 main(int argc, char *argv[]) 583 { 584 int c, error, gdb_port, err, bvmcons; 585 int max_vcpus; 586 struct vmctx *ctx; 587 uint64_t rip; 588 size_t memsize; 589 590 bvmcons = 0; 591 progname = basename(argv[0]); 592 gdb_port = 0; 593 guest_ncpus = 1; 594 memsize = 256 * MB; 595 596 while ((c = getopt(argc, argv, "abehwxAHIPWp:g:c:s:m:l:")) != -1) { 597 switch (c) { 598 case 'a': 599 x2apic_mode = 0; 600 break; 601 case 'A': 602 acpi = 1; 603 break; 604 case 'b': 605 bvmcons = 1; 606 break; 607 case 'p': 608 pincpu = atoi(optarg); 609 break; 610 case 'c': 611 guest_ncpus = atoi(optarg); 612 break; 613 case 'g': 614 gdb_port = atoi(optarg); 615 break; 616 case 'l': 617 if (lpc_device_parse(optarg) != 0) { 618 errx(EX_USAGE, "invalid lpc device " 619 "configuration '%s'", optarg); 620 } 621 break; 622 case 's': 623 if (pci_parse_slot(optarg) != 0) 624 exit(1); 625 else 626 break; 627 case 'm': 628 error = vm_parse_memsize(optarg, &memsize); 629 if (error) 630 errx(EX_USAGE, "invalid memsize '%s'", optarg); 631 break; 632 case 'H': 633 guest_vmexit_on_hlt = 1; 634 break; 635 case 'I': 636 /* 637 * The "-I" option was used to add an ioapic to the 638 * virtual machine. 639 * 640 * An ioapic is now provided unconditionally for each 641 * virtual machine and this option is now deprecated. 642 */ 643 break; 644 case 'P': 645 guest_vmexit_on_pause = 1; 646 break; 647 case 'e': 648 strictio = 1; 649 break; 650 case 'w': 651 strictmsr = 0; 652 break; 653 case 'W': 654 virtio_msix = 0; 655 break; 656 case 'x': 657 x2apic_mode = 1; 658 break; 659 case 'h': 660 usage(0); 661 default: 662 usage(1); 663 } 664 } 665 argc -= optind; 666 argv += optind; 667 668 if (argc != 1) 669 usage(1); 670 671 vmname = argv[0]; 672 673 ctx = vm_open(vmname); 674 if (ctx == NULL) { 675 perror("vm_open"); 676 exit(1); 677 } 678 679 max_vcpus = num_vcpus_allowed(ctx); 680 if (guest_ncpus > max_vcpus) { 681 fprintf(stderr, "%d vCPUs requested but only %d available\n", 682 guest_ncpus, max_vcpus); 683 exit(1); 684 } 685 686 fbsdrun_set_capabilities(ctx, BSP); 687 688 err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL); 689 if (err) { 690 fprintf(stderr, "Unable to setup memory (%d)\n", err); 691 exit(1); 692 } 693 694 init_mem(); 695 init_inout(); 696 ioapic_init(ctx); 697 698 rtc_init(ctx); 699 700 /* 701 * Exit if a device emulation finds an error in it's initilization 702 */ 703 if (init_pci(ctx) != 0) 704 exit(1); 705 706 if (gdb_port != 0) 707 init_dbgport(gdb_port); 708 709 if (bvmcons) 710 init_bvmcons(); 711 712 error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip); 713 assert(error == 0); 714 715 /* 716 * build the guest tables, MP etc. 717 */ 718 mptable_build(ctx, guest_ncpus); 719 720 if (acpi) { 721 error = acpi_build(ctx, guest_ncpus); 722 assert(error == 0); 723 } 724 725 /* 726 * Change the proc title to include the VM name. 727 */ 728 setproctitle("%s", vmname); 729 730 /* 731 * Add CPU 0 732 */ 733 fbsdrun_addcpu(ctx, BSP, rip); 734 735 /* 736 * Head off to the main event dispatch loop 737 */ 738 mevent_dispatch(); 739 740 exit(1); 741 } 742