1 #include <stdio.h> 2 #include <string.h> 3 #include <signal.h> 4 #include <unistd.h> 5 #include <stdlib.h> 6 #include <termios.h> 7 #include <sys/utsname.h> 8 #include <sys/types.h> 9 #include <sys/stat.h> 10 #include <ctype.h> 11 12 /* user defined header files */ 13 #include <linux/types.h> 14 #include <kvm/kvm.h> 15 #include <kvm/kvm-cpu.h> 16 #include <kvm/8250-serial.h> 17 #include <kvm/virtio-blk.h> 18 #include <kvm/virtio-net.h> 19 #include <kvm/virtio-console.h> 20 #include <kvm/virtio-rng.h> 21 #include <kvm/disk-image.h> 22 #include <kvm/util.h> 23 #include <kvm/pci.h> 24 #include <kvm/rtc.h> 25 #include <kvm/term.h> 26 #include <kvm/ioport.h> 27 #include <kvm/threadpool.h> 28 #include <kvm/barrier.h> 29 #include <kvm/symbol.h> 30 #include <kvm/virtio-9p.h> 31 #include <kvm/vesa.h> 32 #include <kvm/ioeventfd.h> 33 #include <kvm/i8042.h> 34 #include <kvm/vnc.h> 35 #include <kvm/sdl.h> 36 #include <kvm/framebuffer.h> 37 38 /* header files for gitish interface */ 39 #include <kvm/kvm-run.h> 40 #include <kvm/parse-options.h> 41 #include <kvm/mutex.h> 42 43 #define DEFAULT_KVM_DEV "/dev/kvm" 44 #define DEFAULT_CONSOLE "serial" 45 #define DEFAULT_NETWORK "user" 46 #define DEFAULT_HOST_ADDR "192.168.33.1" 47 #define DEFAULT_GUEST_MAC "00:15:15:15:15:15" 48 #define DEFAULT_SCRIPT "none" 49 50 #define MB_SHIFT (20) 51 #define MIN_RAM_SIZE_MB (64ULL) 52 #define MIN_RAM_SIZE_BYTE (MIN_RAM_SIZE_MB << MB_SHIFT) 53 54 struct kvm *kvm; 55 struct kvm_cpu *kvm_cpus[KVM_NR_CPUS]; 56 __thread struct kvm_cpu *current_kvm_cpu; 57 58 static u64 ram_size; 59 static u8 image_count; 60 static int virtio_rng; 61 static const char *kernel_cmdline; 62 static const char *kernel_filename; 63 static const char *vmlinux_filename; 64 static const char *initrd_filename; 65 static const char *image_filename[MAX_DISK_IMAGES]; 66 static const char *console; 67 static const char *kvm_dev; 68 static const char *network; 69 static const char *host_ip_addr; 70 static const char *guest_mac; 71 static const char *script; 72 static const char *guest_name; 73 static bool single_step; 74 static bool readonly_image[MAX_DISK_IMAGES]; 75 static bool vnc; 76 static bool sdl; 77 extern bool ioport_debug; 78 extern int active_console; 79 extern int debug_iodelay; 80 81 bool do_debug_print = false; 82 83 static int nrcpus; 84 static int vidmode = -1; 85 86 static const char * const run_usage[] = { 87 "kvm run [<options>] [<kernel image>]", 88 NULL 89 }; 90 91 static int img_name_parser(const struct option *opt, const char *arg, int unset) 92 { 93 char *sep; 94 95 if (image_count >= MAX_DISK_IMAGES) 96 die("Currently only 4 images are supported"); 97 98 image_filename[image_count] = arg; 99 sep = strstr(arg, ","); 100 if (sep) { 101 if (strcmp(sep + 1, "ro") == 0) 102 readonly_image[image_count] = 1; 103 *sep = 0; 104 } 105 106 image_count++; 107 108 return 0; 109 } 110 111 static int virtio_9p_rootdir_parser(const struct option *opt, const char *arg, int unset) 112 { 113 char *tag_name; 114 char tmp[PATH_MAX]; 115 116 /* 117 * 9p dir can be of the form dirname,tag_name or 118 * just dirname. In the later case we use the 119 * default tag name 120 */ 121 tag_name = strstr(arg, ","); 122 if (tag_name) { 123 *tag_name = '\0'; 124 tag_name++; 125 } 126 if (realpath(arg, tmp)) 127 virtio_9p__init(kvm, tmp, tag_name); 128 else 129 die("Failed resolving 9p path"); 130 return 0; 131 } 132 133 134 static const struct option options[] = { 135 OPT_GROUP("Basic options:"), 136 OPT_STRING('\0', "name", &guest_name, "guest name", 137 "A name for the guest"), 138 OPT_INTEGER('c', "cpus", &nrcpus, "Number of CPUs"), 139 OPT_U64('m', "mem", &ram_size, "Virtual machine memory size in MiB."), 140 OPT_CALLBACK('d', "disk", NULL, "image", "Disk image", img_name_parser), 141 OPT_STRING('\0', "console", &console, "serial or virtio", 142 "Console to use"), 143 OPT_INCR('\0', "rng", &virtio_rng, 144 "Enable virtio Random Number Generator"), 145 OPT_STRING('\0', "kvm-dev", &kvm_dev, "kvm-dev", "KVM device file"), 146 OPT_CALLBACK('\0', "virtio-9p", NULL, "dirname,tag_name", 147 "Enable 9p over virtio", virtio_9p_rootdir_parser), 148 OPT_BOOLEAN('\0', "vnc", &vnc, "Enable VNC framebuffer"), 149 OPT_BOOLEAN('\0', "sdl", &sdl, "Enable SDL framebuffer"), 150 151 OPT_GROUP("Kernel options:"), 152 OPT_STRING('k', "kernel", &kernel_filename, "kernel", 153 "Kernel to boot in virtual machine"), 154 OPT_STRING('i', "initrd", &initrd_filename, "initrd", 155 "Initial RAM disk image"), 156 OPT_STRING('p', "params", &kernel_cmdline, "params", 157 "Kernel command line arguments"), 158 159 OPT_GROUP("Networking options:"), 160 OPT_STRING('n', "network", &network, "user, tap, none", 161 "Network to use"), 162 OPT_STRING('\0', "host-ip-addr", &host_ip_addr, "a.b.c.d", 163 "Assign this address to the host side networking"), 164 OPT_STRING('\0', "guest-mac", &guest_mac, "aa:bb:cc:dd:ee:ff", 165 "Assign this address to the guest side NIC"), 166 OPT_STRING('\0', "tapscript", &script, "Script path", 167 "Assign a script to process created tap device"), 168 169 OPT_GROUP("BIOS options:"), 170 OPT_INTEGER('\0', "vidmode", &vidmode, 171 "Video mode"), 172 173 OPT_GROUP("Debug options:"), 174 OPT_BOOLEAN('\0', "debug", &do_debug_print, 175 "Enable debug messages"), 176 OPT_BOOLEAN('\0', "debug-single-step", &single_step, 177 "Enable single stepping"), 178 OPT_BOOLEAN('\0', "debug-ioport-debug", &ioport_debug, 179 "Enable ioport debugging"), 180 OPT_INTEGER('\0', "debug_iodelay", &debug_iodelay, 181 "Delay IO by millisecond"), 182 OPT_END() 183 }; 184 185 /* 186 * Serialize debug printout so that the output of multiple vcpus does not 187 * get mixed up: 188 */ 189 static int printout_done; 190 191 static void handle_sigusr1(int sig) 192 { 193 struct kvm_cpu *cpu = current_kvm_cpu; 194 195 if (!cpu) 196 return; 197 198 printf("\n #\n # vCPU #%ld's dump:\n #\n", cpu->cpu_id); 199 kvm_cpu__show_registers(cpu); 200 kvm_cpu__show_code(cpu); 201 kvm_cpu__show_page_tables(cpu); 202 fflush(stdout); 203 printout_done = 1; 204 mb(); 205 } 206 207 /* Pause/resume the guest using SIGUSR2 */ 208 static int is_paused; 209 210 static void handle_sigusr2(int sig) 211 { 212 if (is_paused) 213 kvm__continue(); 214 else 215 kvm__pause(); 216 217 is_paused = !is_paused; 218 pr_info("Guest %s\n", is_paused ? "paused" : "resumed"); 219 } 220 221 static void handle_sigquit(int sig) 222 { 223 int i; 224 225 for (i = 0; i < nrcpus; i++) { 226 struct kvm_cpu *cpu = kvm_cpus[i]; 227 228 if (!cpu) 229 continue; 230 231 printout_done = 0; 232 pthread_kill(cpu->thread, SIGUSR1); 233 /* 234 * Wait for the vCPU to dump state before signalling 235 * the next thread. Since this is debug code it does 236 * not matter that we are burning CPU time a bit: 237 */ 238 while (!printout_done) 239 mb(); 240 } 241 242 serial8250__inject_sysrq(kvm); 243 } 244 245 static void handle_sigalrm(int sig) 246 { 247 serial8250__inject_interrupt(kvm); 248 virtio_console__inject_interrupt(kvm); 249 } 250 251 static void *kvm_cpu_thread(void *arg) 252 { 253 current_kvm_cpu = arg; 254 255 if (kvm_cpu__start(current_kvm_cpu)) 256 goto panic_kvm; 257 258 kvm_cpu__delete(current_kvm_cpu); 259 260 return (void *) (intptr_t) 0; 261 262 panic_kvm: 263 fprintf(stderr, "KVM exit reason: %u (\"%s\")\n", 264 current_kvm_cpu->kvm_run->exit_reason, 265 kvm_exit_reasons[current_kvm_cpu->kvm_run->exit_reason]); 266 if (current_kvm_cpu->kvm_run->exit_reason == KVM_EXIT_UNKNOWN) 267 fprintf(stderr, "KVM exit code: 0x%Lu\n", 268 current_kvm_cpu->kvm_run->hw.hardware_exit_reason); 269 270 kvm_cpu__show_registers(current_kvm_cpu); 271 kvm_cpu__show_code(current_kvm_cpu); 272 kvm_cpu__show_page_tables(current_kvm_cpu); 273 274 kvm_cpu__delete(current_kvm_cpu); 275 276 return (void *) (intptr_t) 1; 277 } 278 279 static char kernel[PATH_MAX]; 280 281 static const char *host_kernels[] = { 282 "/boot/vmlinuz", 283 "/boot/bzImage", 284 NULL 285 }; 286 287 static const char *default_kernels[] = { 288 "./bzImage", 289 "../../arch/x86/boot/bzImage", 290 NULL 291 }; 292 293 static const char *default_vmlinux[] = { 294 "../../../vmlinux", 295 "../../vmlinux", 296 NULL 297 }; 298 299 static void kernel_usage_with_options(void) 300 { 301 const char **k; 302 struct utsname uts; 303 304 fprintf(stderr, "Fatal: could not find default kernel image in:\n"); 305 k = &default_kernels[0]; 306 while (*k) { 307 fprintf(stderr, "\t%s\n", *k); 308 k++; 309 } 310 311 if (uname(&uts) < 0) 312 return; 313 314 k = &host_kernels[0]; 315 while (*k) { 316 if (snprintf(kernel, PATH_MAX, "%s-%s", *k, uts.release) < 0) 317 return; 318 fprintf(stderr, "\t%s\n", kernel); 319 k++; 320 } 321 fprintf(stderr, "\nPlease see 'kvm run --help' for more options.\n\n"); 322 } 323 324 static u64 host_ram_size(void) 325 { 326 long page_size; 327 long nr_pages; 328 329 nr_pages = sysconf(_SC_PHYS_PAGES); 330 if (nr_pages < 0) { 331 pr_warning("sysconf(_SC_PHYS_PAGES) failed"); 332 return 0; 333 } 334 335 page_size = sysconf(_SC_PAGE_SIZE); 336 if (page_size < 0) { 337 pr_warning("sysconf(_SC_PAGE_SIZE) failed"); 338 return 0; 339 } 340 341 return (nr_pages * page_size) >> MB_SHIFT; 342 } 343 344 /* 345 * If user didn't specify how much memory it wants to allocate for the guest, 346 * avoid filling the whole host RAM. 347 */ 348 #define RAM_SIZE_RATIO 0.8 349 350 static u64 get_ram_size(int nr_cpus) 351 { 352 u64 available; 353 u64 ram_size; 354 355 ram_size = 64 * (nr_cpus + 3); 356 357 available = host_ram_size() * RAM_SIZE_RATIO; 358 if (!available) 359 available = MIN_RAM_SIZE_MB; 360 361 if (ram_size > available) 362 ram_size = available; 363 364 return ram_size; 365 } 366 367 static const char *find_kernel(void) 368 { 369 const char **k; 370 struct stat st; 371 struct utsname uts; 372 373 k = &default_kernels[0]; 374 while (*k) { 375 if (stat(*k, &st) < 0 || !S_ISREG(st.st_mode)) { 376 k++; 377 continue; 378 } 379 strncpy(kernel, *k, PATH_MAX); 380 return kernel; 381 } 382 383 if (uname(&uts) < 0) 384 return NULL; 385 386 k = &host_kernels[0]; 387 while (*k) { 388 if (snprintf(kernel, PATH_MAX, "%s-%s", *k, uts.release) < 0) 389 return NULL; 390 391 if (stat(kernel, &st) < 0 || !S_ISREG(st.st_mode)) { 392 k++; 393 continue; 394 } 395 return kernel; 396 397 } 398 return NULL; 399 } 400 401 static const char *find_vmlinux(void) 402 { 403 const char **vmlinux; 404 405 vmlinux = &default_vmlinux[0]; 406 while (*vmlinux) { 407 struct stat st; 408 409 if (stat(*vmlinux, &st) < 0 || !S_ISREG(st.st_mode)) { 410 vmlinux++; 411 continue; 412 } 413 return *vmlinux; 414 } 415 return NULL; 416 } 417 418 static int root_device(char *dev, long *part) 419 { 420 struct stat st; 421 422 if (stat("/", &st) < 0) 423 return -1; 424 425 *part = minor(st.st_dev); 426 427 sprintf(dev, "/dev/block/%u:0", major(st.st_dev)); 428 if (access(dev, R_OK) < 0) 429 return -1; 430 431 return 0; 432 } 433 434 static char *host_image(char *cmd_line, size_t size) 435 { 436 char *t; 437 char device[PATH_MAX]; 438 long part = 0; 439 440 t = malloc(PATH_MAX); 441 if (!t) 442 return NULL; 443 444 /* check for the root file system */ 445 if (root_device(device, &part) < 0) { 446 free(t); 447 return NULL; 448 } 449 strncpy(t, device, PATH_MAX); 450 if (!strstr(cmd_line, "root=")) { 451 char tmp[PATH_MAX]; 452 snprintf(tmp, sizeof(tmp), "root=/dev/vda%ld rw ", part); 453 strlcat(cmd_line, tmp, size); 454 } 455 return t; 456 } 457 458 void kvm_run_help(void) 459 { 460 usage_with_options(run_usage, options); 461 } 462 463 int kvm_cmd_run(int argc, const char **argv, const char *prefix) 464 { 465 struct virtio_net_parameters net_params; 466 static char real_cmdline[2048]; 467 struct framebuffer *fb = NULL; 468 unsigned int nr_online_cpus; 469 int exit_code = 0; 470 int max_cpus; 471 char *hi; 472 int i; 473 void *ret; 474 475 signal(SIGALRM, handle_sigalrm); 476 signal(SIGQUIT, handle_sigquit); 477 signal(SIGUSR1, handle_sigusr1); 478 signal(SIGUSR2, handle_sigusr2); 479 480 nr_online_cpus = sysconf(_SC_NPROCESSORS_ONLN); 481 482 while (argc != 0) { 483 argc = parse_options(argc, argv, options, run_usage, 484 PARSE_OPT_STOP_AT_NON_OPTION); 485 if (argc != 0) { 486 if (kernel_filename) { 487 fprintf(stderr, "Cannot handle parameter: " 488 "%s\n", argv[0]); 489 usage_with_options(run_usage, options); 490 return EINVAL; 491 } 492 /* first unhandled parameter is treated as a kernel 493 image 494 */ 495 kernel_filename = argv[0]; 496 argv++; 497 argc--; 498 } 499 500 } 501 502 if (!kernel_filename) 503 kernel_filename = find_kernel(); 504 505 if (!kernel_filename) { 506 kernel_usage_with_options(); 507 return EINVAL; 508 } 509 510 vmlinux_filename = find_vmlinux(); 511 512 if (nrcpus == 0) 513 nrcpus = nr_online_cpus; 514 else if (nrcpus < 1 || nrcpus > KVM_NR_CPUS) 515 die("Number of CPUs %d is out of [1;%d] range", nrcpus, KVM_NR_CPUS); 516 517 if (!ram_size) 518 ram_size = get_ram_size(nrcpus); 519 520 if (ram_size < MIN_RAM_SIZE_MB) 521 die("Not enough memory specified: %lluMB (min %lluMB)", ram_size, MIN_RAM_SIZE_MB); 522 523 if (ram_size > host_ram_size()) 524 pr_warning("Guest memory size %lluMB exceeds host physical RAM size %lluMB", ram_size, host_ram_size()); 525 526 ram_size <<= MB_SHIFT; 527 528 if (!kvm_dev) 529 kvm_dev = DEFAULT_KVM_DEV; 530 531 if (!console) 532 console = DEFAULT_CONSOLE; 533 534 if (!strncmp(console, "virtio", 6)) 535 active_console = CONSOLE_VIRTIO; 536 else 537 active_console = CONSOLE_8250; 538 539 if (!host_ip_addr) 540 host_ip_addr = DEFAULT_HOST_ADDR; 541 542 if (!guest_mac) 543 guest_mac = DEFAULT_GUEST_MAC; 544 545 if (!script) 546 script = DEFAULT_SCRIPT; 547 548 symbol__init(vmlinux_filename); 549 550 term_init(); 551 552 kvm = kvm__init(kvm_dev, ram_size, guest_name); 553 554 ioeventfd__init(); 555 556 max_cpus = kvm__max_cpus(kvm); 557 558 if (nrcpus > max_cpus) { 559 printf(" # Limit the number of CPUs to %d\n", max_cpus); 560 kvm->nrcpus = max_cpus; 561 } 562 563 kvm->nrcpus = nrcpus; 564 565 /* 566 * vidmode should be either specified 567 * either set by default 568 */ 569 if (vnc || sdl) { 570 if (vidmode == -1) 571 vidmode = 0x312; 572 } else 573 vidmode = 0; 574 575 memset(real_cmdline, 0, sizeof(real_cmdline)); 576 strcpy(real_cmdline, "notsc noapic noacpi pci=conf1 reboot=k panic=1"); 577 if (vnc || sdl) { 578 strcat(real_cmdline, " video=vesafb console=tty0"); 579 } else 580 strcat(real_cmdline, " console=ttyS0 earlyprintk=serial"); 581 strcat(real_cmdline, " "); 582 if (kernel_cmdline) 583 strlcat(real_cmdline, kernel_cmdline, sizeof(real_cmdline)); 584 585 hi = NULL; 586 if (!image_filename[0]) { 587 hi = host_image(real_cmdline, sizeof(real_cmdline)); 588 if (hi) { 589 image_filename[0] = hi; 590 readonly_image[0] = true; 591 image_count++; 592 } 593 } 594 595 if (!strstr(real_cmdline, "root=")) 596 strlcat(real_cmdline, " root=/dev/vda rw ", sizeof(real_cmdline)); 597 598 if (image_count) { 599 kvm->nr_disks = image_count; 600 kvm->disks = disk_image__open_all(image_filename, readonly_image, image_count); 601 if (!kvm->disks) 602 die("Unable to load all disk images."); 603 604 virtio_blk__init_all(kvm); 605 } 606 607 free(hi); 608 609 printf(" # kvm run -k %s -m %Lu -c %d\n", kernel_filename, ram_size / 1024 / 1024, nrcpus); 610 611 if (!kvm__load_kernel(kvm, kernel_filename, initrd_filename, 612 real_cmdline, vidmode)) 613 die("unable to load kernel %s", kernel_filename); 614 615 kvm->vmlinux = vmlinux_filename; 616 617 ioport__setup_legacy(); 618 619 rtc__init(); 620 621 serial8250__init(kvm); 622 623 pci__init(); 624 625 if (active_console == CONSOLE_VIRTIO) 626 virtio_console__init(kvm); 627 628 if (virtio_rng) 629 while (virtio_rng--) 630 virtio_rng__init(kvm); 631 632 if (!network) 633 network = DEFAULT_NETWORK; 634 635 if (strncmp(network, "none", 4)) { 636 net_params.host_ip = host_ip_addr; 637 net_params.kvm = kvm; 638 net_params.script = script; 639 sscanf(guest_mac, "%hhx:%hhx:%hhx:%hhx:%hhx:%hhx", 640 net_params.guest_mac, 641 net_params.guest_mac+1, 642 net_params.guest_mac+2, 643 net_params.guest_mac+3, 644 net_params.guest_mac+4, 645 net_params.guest_mac+5); 646 647 if (!strncmp(network, "user", 4)) 648 net_params.mode = NET_MODE_USER; 649 else if (!strncmp(network, "tap", 3)) 650 net_params.mode = NET_MODE_TAP; 651 else 652 die("Unkown network mode %s, please use -network user, tap, none", network); 653 virtio_net__init(&net_params); 654 } 655 656 kvm__start_timer(kvm); 657 658 kvm__setup_bios(kvm); 659 660 for (i = 0; i < nrcpus; i++) { 661 kvm_cpus[i] = kvm_cpu__init(kvm, i); 662 if (!kvm_cpus[i]) 663 die("unable to initialize KVM VCPU"); 664 665 if (single_step) 666 kvm_cpu__enable_singlestep(kvm_cpus[i]); 667 } 668 669 kvm__init_ram(kvm); 670 671 kbd__init(kvm); 672 673 if (vnc || sdl) 674 fb = vesa__init(kvm); 675 676 if (vnc) { 677 if (fb) 678 vnc__init(fb); 679 } 680 681 if (sdl) { 682 if (fb) 683 sdl__init(fb); 684 } 685 686 fb__start(); 687 688 thread_pool__init(nr_online_cpus); 689 ioeventfd__start(); 690 691 for (i = 0; i < nrcpus; i++) { 692 if (pthread_create(&kvm_cpus[i]->thread, NULL, kvm_cpu_thread, kvm_cpus[i]) != 0) 693 die("unable to create KVM VCPU thread"); 694 } 695 696 /* Only VCPU #0 is going to exit by itself when shutting down */ 697 if (pthread_join(kvm_cpus[0]->thread, &ret) != 0) 698 exit_code = 1; 699 700 for (i = 1; i < nrcpus; i++) { 701 pthread_kill(kvm_cpus[i]->thread, SIGKVMEXIT); 702 if (pthread_join(kvm_cpus[i]->thread, &ret) != 0) 703 die("pthread_join"); 704 705 if (ret != NULL) 706 exit_code = 1; 707 } 708 709 fb__stop(); 710 711 virtio_blk__delete_all(kvm); 712 virtio_rng__delete_all(kvm); 713 714 disk_image__close_all(kvm->disks, image_count); 715 kvm__delete(kvm); 716 717 if (!exit_code) 718 printf("\n # KVM session ended normally.\n"); 719 720 return exit_code; 721 } 722