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 35 /* header files for gitish interface */ 36 #include <kvm/kvm-run.h> 37 #include <kvm/parse-options.h> 38 #include <kvm/mutex.h> 39 40 #define DEFAULT_KVM_DEV "/dev/kvm" 41 #define DEFAULT_CONSOLE "serial" 42 #define DEFAULT_NETWORK "virtio" 43 #define DEFAULT_HOST_ADDR "192.168.33.2" 44 #define DEFAULT_GUEST_MAC "00:11:22:33:44:55" 45 #define DEFAULT_SCRIPT "none" 46 47 #define MB_SHIFT (20) 48 #define MIN_RAM_SIZE_MB (64ULL) 49 #define MIN_RAM_SIZE_BYTE (MIN_RAM_SIZE_MB << MB_SHIFT) 50 51 struct kvm *kvm; 52 struct kvm_cpu *kvm_cpus[KVM_NR_CPUS]; 53 __thread struct kvm_cpu *current_kvm_cpu; 54 55 static u64 ram_size; 56 static u8 image_count; 57 static int virtio_rng; 58 static const char *kernel_cmdline; 59 static const char *kernel_filename; 60 static const char *vmlinux_filename; 61 static const char *initrd_filename; 62 static const char *image_filename[MAX_DISK_IMAGES]; 63 static const char *console; 64 static const char *kvm_dev; 65 static const char *network; 66 static const char *host_ip_addr; 67 static const char *guest_mac; 68 static const char *script; 69 static const char *virtio_9p_dir; 70 static bool single_step; 71 static bool readonly_image[MAX_DISK_IMAGES]; 72 static bool vnc; 73 extern bool ioport_debug; 74 extern int active_console; 75 76 bool do_debug_print = false; 77 78 static int nrcpus; 79 80 static const char * const run_usage[] = { 81 "kvm run [<options>] [<kernel image>]", 82 NULL 83 }; 84 85 static int img_name_parser(const struct option *opt, const char *arg, int unset) 86 { 87 char *sep; 88 89 if (image_count >= MAX_DISK_IMAGES) 90 die("Currently only 4 images are supported"); 91 92 image_filename[image_count] = arg; 93 sep = strstr(arg, ","); 94 if (sep) { 95 if (strcmp(sep + 1, "ro") == 0) 96 readonly_image[image_count] = 1; 97 *sep = 0; 98 } 99 100 image_count++; 101 102 return 0; 103 } 104 105 static const struct option options[] = { 106 OPT_GROUP("Basic options:"), 107 OPT_INTEGER('c', "cpus", &nrcpus, "Number of CPUs"), 108 OPT_U64('m', "mem", &ram_size, "Virtual machine memory size in MiB."), 109 OPT_CALLBACK('d', "disk", NULL, "image", "Disk image", img_name_parser), 110 OPT_STRING('\0', "console", &console, "serial or virtio", 111 "Console to use"), 112 OPT_INCR('\0', "rng", &virtio_rng, 113 "Enable virtio Random Number Generator"), 114 OPT_STRING('\0', "kvm-dev", &kvm_dev, "kvm-dev", "KVM device file"), 115 OPT_STRING('\0', "virtio-9p", &virtio_9p_dir, "root dir", 116 "Enable 9p over virtio"), 117 OPT_BOOLEAN('\0', "vnc", &vnc, "Enable VNC framebuffer"), 118 119 OPT_GROUP("Kernel options:"), 120 OPT_STRING('k', "kernel", &kernel_filename, "kernel", 121 "Kernel to boot in virtual machine"), 122 OPT_STRING('i', "initrd", &initrd_filename, "initrd", 123 "Initial RAM disk image"), 124 OPT_STRING('p', "params", &kernel_cmdline, "params", 125 "Kernel command line arguments"), 126 127 OPT_GROUP("Networking options:"), 128 OPT_STRING('n', "network", &network, "virtio", 129 "Network to use"), 130 OPT_STRING('\0', "host-ip-addr", &host_ip_addr, "a.b.c.d", 131 "Assign this address to the host side networking"), 132 OPT_STRING('\0', "guest-mac", &guest_mac, "aa:bb:cc:dd:ee:ff", 133 "Assign this address to the guest side NIC"), 134 OPT_STRING('\0', "tapscript", &script, "Script path", 135 "Assign a script to process created tap device"), 136 137 OPT_GROUP("Debug options:"), 138 OPT_BOOLEAN('\0', "debug", &do_debug_print, 139 "Enable debug messages"), 140 OPT_BOOLEAN('\0', "debug-single-step", &single_step, 141 "Enable single stepping"), 142 OPT_BOOLEAN('\0', "debug-ioport-debug", &ioport_debug, 143 "Enable ioport debugging"), 144 OPT_END() 145 }; 146 147 /* 148 * Serialize debug printout so that the output of multiple vcpus does not 149 * get mixed up: 150 */ 151 static int printout_done; 152 153 static void handle_sigusr1(int sig) 154 { 155 struct kvm_cpu *cpu = current_kvm_cpu; 156 157 if (!cpu) 158 return; 159 160 printf("\n #\n # vCPU #%ld's dump:\n #\n", cpu->cpu_id); 161 kvm_cpu__show_registers(cpu); 162 kvm_cpu__show_code(cpu); 163 kvm_cpu__show_page_tables(cpu); 164 fflush(stdout); 165 printout_done = 1; 166 mb(); 167 } 168 169 /* Pause/resume the guest using SIGUSR2 */ 170 static int is_paused; 171 172 static void handle_sigusr2(int sig) 173 { 174 if (is_paused) 175 kvm__continue(); 176 else 177 kvm__pause(); 178 179 is_paused = !is_paused; 180 pr_info("Guest %s\n", is_paused ? "paused" : "resumed"); 181 } 182 183 static void handle_sigquit(int sig) 184 { 185 int i; 186 187 for (i = 0; i < nrcpus; i++) { 188 struct kvm_cpu *cpu = kvm_cpus[i]; 189 190 if (!cpu) 191 continue; 192 193 printout_done = 0; 194 pthread_kill(cpu->thread, SIGUSR1); 195 /* 196 * Wait for the vCPU to dump state before signalling 197 * the next thread. Since this is debug code it does 198 * not matter that we are burning CPU time a bit: 199 */ 200 while (!printout_done) 201 mb(); 202 } 203 204 serial8250__inject_sysrq(kvm); 205 } 206 207 static void handle_sigalrm(int sig) 208 { 209 serial8250__inject_interrupt(kvm); 210 virtio_console__inject_interrupt(kvm); 211 } 212 213 static void *kvm_cpu_thread(void *arg) 214 { 215 current_kvm_cpu = arg; 216 217 if (kvm_cpu__start(current_kvm_cpu)) 218 goto panic_kvm; 219 220 kvm_cpu__delete(current_kvm_cpu); 221 222 return (void *) (intptr_t) 0; 223 224 panic_kvm: 225 fprintf(stderr, "KVM exit reason: %u (\"%s\")\n", 226 current_kvm_cpu->kvm_run->exit_reason, 227 kvm_exit_reasons[current_kvm_cpu->kvm_run->exit_reason]); 228 if (current_kvm_cpu->kvm_run->exit_reason == KVM_EXIT_UNKNOWN) 229 fprintf(stderr, "KVM exit code: 0x%Lu\n", 230 current_kvm_cpu->kvm_run->hw.hardware_exit_reason); 231 232 kvm_cpu__show_registers(current_kvm_cpu); 233 kvm_cpu__show_code(current_kvm_cpu); 234 kvm_cpu__show_page_tables(current_kvm_cpu); 235 236 kvm_cpu__delete(current_kvm_cpu); 237 238 return (void *) (intptr_t) 1; 239 } 240 241 static char kernel[PATH_MAX]; 242 243 static const char *host_kernels[] = { 244 "/boot/vmlinuz", 245 "/boot/bzImage", 246 NULL 247 }; 248 249 static const char *default_kernels[] = { 250 "./bzImage", 251 "../../arch/x86/boot/bzImage", 252 NULL 253 }; 254 255 static const char *default_vmlinux[] = { 256 "../../../vmlinux", 257 "../../vmlinux", 258 NULL 259 }; 260 261 static void kernel_usage_with_options(void) 262 { 263 const char **k; 264 struct utsname uts; 265 266 fprintf(stderr, "Fatal: could not find default kernel image in:\n"); 267 k = &default_kernels[0]; 268 while (*k) { 269 fprintf(stderr, "\t%s\n", *k); 270 k++; 271 } 272 273 if (uname(&uts) < 0) 274 return; 275 276 k = &host_kernels[0]; 277 while (*k) { 278 if (snprintf(kernel, PATH_MAX, "%s-%s", *k, uts.release) < 0) 279 return; 280 fprintf(stderr, "\t%s\n", kernel); 281 k++; 282 } 283 fprintf(stderr, "\nPlease see 'kvm run --help' for more options.\n\n"); 284 } 285 286 static u64 host_ram_size(void) 287 { 288 long page_size; 289 long nr_pages; 290 291 nr_pages = sysconf(_SC_PHYS_PAGES); 292 if (nr_pages < 0) { 293 pr_warning("sysconf(_SC_PHYS_PAGES) failed"); 294 return 0; 295 } 296 297 page_size = sysconf(_SC_PAGE_SIZE); 298 if (page_size < 0) { 299 pr_warning("sysconf(_SC_PAGE_SIZE) failed"); 300 return 0; 301 } 302 303 return (nr_pages * page_size) >> MB_SHIFT; 304 } 305 306 /* 307 * If user didn't specify how much memory it wants to allocate for the guest, 308 * avoid filling the whole host RAM. 309 */ 310 #define RAM_SIZE_RATIO 0.8 311 312 static u64 get_ram_size(int nr_cpus) 313 { 314 long available; 315 long ram_size; 316 317 ram_size = 64 * (nr_cpus + 3); 318 319 available = host_ram_size() * RAM_SIZE_RATIO; 320 if (!available) 321 available = MIN_RAM_SIZE_MB; 322 323 if (ram_size > available) 324 ram_size = available; 325 326 return ram_size; 327 } 328 329 static const char *find_kernel(void) 330 { 331 const char **k; 332 struct stat st; 333 struct utsname uts; 334 335 k = &default_kernels[0]; 336 while (*k) { 337 if (stat(*k, &st) < 0 || !S_ISREG(st.st_mode)) { 338 k++; 339 continue; 340 } 341 strncpy(kernel, *k, PATH_MAX); 342 return kernel; 343 } 344 345 if (uname(&uts) < 0) 346 return NULL; 347 348 k = &host_kernels[0]; 349 while (*k) { 350 if (snprintf(kernel, PATH_MAX, "%s-%s", *k, uts.release) < 0) 351 return NULL; 352 353 if (stat(kernel, &st) < 0 || !S_ISREG(st.st_mode)) { 354 k++; 355 continue; 356 } 357 return kernel; 358 359 } 360 return NULL; 361 } 362 363 static const char *find_vmlinux(void) 364 { 365 const char **vmlinux; 366 367 vmlinux = &default_vmlinux[0]; 368 while (*vmlinux) { 369 struct stat st; 370 371 if (stat(*vmlinux, &st) < 0 || !S_ISREG(st.st_mode)) { 372 vmlinux++; 373 continue; 374 } 375 return *vmlinux; 376 } 377 return NULL; 378 } 379 380 static int root_device(char *dev, long *part) 381 { 382 struct stat st; 383 384 if (stat("/", &st) < 0) 385 return -1; 386 387 *part = minor(st.st_dev); 388 389 sprintf(dev, "/dev/block/%u:0", major(st.st_dev)); 390 if (access(dev, R_OK) < 0) 391 return -1; 392 393 return 0; 394 } 395 396 static char *host_image(char *cmd_line, size_t size) 397 { 398 char *t; 399 char device[PATH_MAX]; 400 long part = 0; 401 402 t = malloc(PATH_MAX); 403 if (!t) 404 return NULL; 405 406 /* check for the root file system */ 407 if (root_device(device, &part) < 0) { 408 free(t); 409 return NULL; 410 } 411 strncpy(t, device, PATH_MAX); 412 if (!strstr(cmd_line, "root=")) { 413 char tmp[PATH_MAX]; 414 snprintf(tmp, sizeof(tmp), "root=/dev/vda%ld rw ", part); 415 strlcat(cmd_line, tmp, size); 416 } 417 return t; 418 } 419 420 void kvm_run_help(void) 421 { 422 usage_with_options(run_usage, options); 423 } 424 425 int kvm_cmd_run(int argc, const char **argv, const char *prefix) 426 { 427 struct virtio_net_parameters net_params; 428 static char real_cmdline[2048]; 429 unsigned int nr_online_cpus; 430 int exit_code = 0; 431 int max_cpus; 432 char *hi; 433 int i; 434 void *ret; 435 u16 vidmode = 0; 436 437 signal(SIGALRM, handle_sigalrm); 438 signal(SIGQUIT, handle_sigquit); 439 signal(SIGUSR1, handle_sigusr1); 440 signal(SIGUSR2, handle_sigusr2); 441 442 nr_online_cpus = sysconf(_SC_NPROCESSORS_ONLN); 443 444 while (argc != 0) { 445 argc = parse_options(argc, argv, options, run_usage, 446 PARSE_OPT_STOP_AT_NON_OPTION); 447 if (argc != 0) { 448 if (kernel_filename) { 449 fprintf(stderr, "Cannot handle parameter: " 450 "%s\n", argv[0]); 451 usage_with_options(run_usage, options); 452 return EINVAL; 453 } 454 /* first unhandled parameter is treated as a kernel 455 image 456 */ 457 kernel_filename = argv[0]; 458 argv++; 459 argc--; 460 } 461 462 } 463 464 if (!kernel_filename) 465 kernel_filename = find_kernel(); 466 467 if (!kernel_filename) { 468 kernel_usage_with_options(); 469 return EINVAL; 470 } 471 472 vmlinux_filename = find_vmlinux(); 473 474 if (nrcpus == 0) 475 nrcpus = nr_online_cpus; 476 else if (nrcpus < 1 || nrcpus > KVM_NR_CPUS) 477 die("Number of CPUs %d is out of [1;%d] range", nrcpus, KVM_NR_CPUS); 478 479 if (!ram_size) 480 ram_size = get_ram_size(nrcpus); 481 482 if (ram_size < MIN_RAM_SIZE_MB) 483 die("Not enough memory specified: %lluMB (min %lluMB)", ram_size, MIN_RAM_SIZE_MB); 484 485 if (ram_size > host_ram_size()) 486 pr_warning("Guest memory size %lluMB exceeds host physical RAM size %lluMB", ram_size, host_ram_size()); 487 488 ram_size <<= MB_SHIFT; 489 490 if (!kvm_dev) 491 kvm_dev = DEFAULT_KVM_DEV; 492 493 if (!console) 494 console = DEFAULT_CONSOLE; 495 496 if (!strncmp(console, "virtio", 6)) 497 active_console = CONSOLE_VIRTIO; 498 else 499 active_console = CONSOLE_8250; 500 501 if (!host_ip_addr) 502 host_ip_addr = DEFAULT_HOST_ADDR; 503 504 if (!guest_mac) 505 guest_mac = DEFAULT_GUEST_MAC; 506 507 if (!script) 508 script = DEFAULT_SCRIPT; 509 510 if (virtio_9p_dir) { 511 char tmp[PATH_MAX]; 512 513 if (realpath(virtio_9p_dir, tmp)) 514 virtio_9p__init(kvm, tmp); 515 else 516 die("Failed resolving 9p path"); 517 } 518 519 symbol__init(vmlinux_filename); 520 521 term_init(); 522 523 kvm = kvm__init(kvm_dev, ram_size); 524 525 ioeventfd__init(); 526 527 max_cpus = kvm__max_cpus(kvm); 528 529 if (nrcpus > max_cpus) { 530 printf(" # Limit the number of CPUs to %d\n", max_cpus); 531 kvm->nrcpus = max_cpus; 532 } 533 534 kvm->nrcpus = nrcpus; 535 536 memset(real_cmdline, 0, sizeof(real_cmdline)); 537 strcpy(real_cmdline, "notsc noapic noacpi pci=conf1"); 538 if (vnc) { 539 strcat(real_cmdline, " video=vesafb console=tty0"); 540 vidmode = 0x312; 541 } else { 542 strcat(real_cmdline, " console=ttyS0 earlyprintk=serial"); 543 } 544 strcat(real_cmdline, " "); 545 if (kernel_cmdline) 546 strlcat(real_cmdline, kernel_cmdline, sizeof(real_cmdline)); 547 548 hi = NULL; 549 if (!image_filename[0]) { 550 hi = host_image(real_cmdline, sizeof(real_cmdline)); 551 if (hi) { 552 image_filename[0] = hi; 553 readonly_image[0] = true; 554 image_count++; 555 } 556 } 557 558 if (!strstr(real_cmdline, "root=")) 559 strlcat(real_cmdline, " root=/dev/vda rw ", sizeof(real_cmdline)); 560 561 if (image_count) { 562 kvm->nr_disks = image_count; 563 kvm->disks = disk_image__open_all(image_filename, readonly_image, image_count); 564 if (!kvm->disks) 565 die("Unable to load all disk images."); 566 567 virtio_blk__init_all(kvm); 568 } 569 570 free(hi); 571 572 printf(" # kvm run -k %s -m %Lu -c %d\n", kernel_filename, ram_size / 1024 / 1024, nrcpus); 573 574 if (!kvm__load_kernel(kvm, kernel_filename, initrd_filename, 575 real_cmdline, vidmode)) 576 die("unable to load kernel %s", kernel_filename); 577 578 kvm->vmlinux = vmlinux_filename; 579 580 ioport__setup_legacy(); 581 582 rtc__init(); 583 584 serial8250__init(kvm); 585 586 pci__init(); 587 588 if (active_console == CONSOLE_VIRTIO) 589 virtio_console__init(kvm); 590 591 if (virtio_rng) 592 while (virtio_rng--) 593 virtio_rng__init(kvm); 594 595 if (!network) 596 network = DEFAULT_NETWORK; 597 598 if (!strncmp(network, "virtio", 6)) { 599 net_params = (struct virtio_net_parameters) { 600 .host_ip = host_ip_addr, 601 .kvm = kvm, 602 .script = script 603 }; 604 sscanf(guest_mac, "%hhx:%hhx:%hhx:%hhx:%hhx:%hhx", 605 net_params.guest_mac, 606 net_params.guest_mac+1, 607 net_params.guest_mac+2, 608 net_params.guest_mac+3, 609 net_params.guest_mac+4, 610 net_params.guest_mac+5); 611 612 virtio_net__init(&net_params); 613 } 614 615 kvm__start_timer(kvm); 616 617 kvm__setup_bios(kvm); 618 619 for (i = 0; i < nrcpus; i++) { 620 kvm_cpus[i] = kvm_cpu__init(kvm, i); 621 if (!kvm_cpus[i]) 622 die("unable to initialize KVM VCPU"); 623 624 if (single_step) 625 kvm_cpu__enable_singlestep(kvm_cpus[i]); 626 } 627 628 kvm__init_ram(kvm); 629 630 if (vnc) { 631 kbd__init(kvm); 632 vesa__init(kvm); 633 } 634 635 thread_pool__init(nr_online_cpus); 636 ioeventfd__start(); 637 638 for (i = 0; i < nrcpus; i++) { 639 if (pthread_create(&kvm_cpus[i]->thread, NULL, kvm_cpu_thread, kvm_cpus[i]) != 0) 640 die("unable to create KVM VCPU thread"); 641 } 642 643 /* Only VCPU #0 is going to exit by itself when shutting down */ 644 if (pthread_join(kvm_cpus[0]->thread, &ret) != 0) 645 exit_code = 1; 646 647 for (i = 1; i < nrcpus; i++) { 648 pthread_kill(kvm_cpus[i]->thread, SIGKVMEXIT); 649 if (pthread_join(kvm_cpus[i]->thread, &ret) != 0) 650 die("pthread_join"); 651 652 if (ret != NULL) 653 exit_code = 1; 654 } 655 656 virtio_blk__delete_all(kvm); 657 virtio_rng__delete_all(kvm); 658 659 disk_image__close_all(kvm->disks, image_count); 660 kvm__delete(kvm); 661 662 if (!exit_code) 663 printf("\n # KVM session ended normally.\n"); 664 665 return exit_code; 666 } 667