1 /* 2 * Copyright (C) 2005 Anthony Liguori <anthony@codemonkey.ws> 3 * 4 * Network Block Device 5 * 6 * This program is free software; you can redistribute it and/or modify 7 * it under the terms of the GNU General Public License as published by 8 * the Free Software Foundation; under version 2 of the License. 9 * 10 * This program is distributed in the hope that it will be useful, 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 13 * GNU General Public License for more details. 14 * 15 * You should have received a copy of the GNU General Public License 16 * along with this program; if not, see <http://www.gnu.org/licenses/>. 17 */ 18 19 #include "qemu/osdep.h" 20 #include <getopt.h> 21 #include <libgen.h> 22 #include <pthread.h> 23 24 #include "qemu-common.h" 25 #include "qapi/error.h" 26 #include "qemu/cutils.h" 27 #include "sysemu/block-backend.h" 28 #include "block/block_int.h" 29 #include "block/nbd.h" 30 #include "qemu/main-loop.h" 31 #include "qemu/module.h" 32 #include "qemu/option.h" 33 #include "qemu/error-report.h" 34 #include "qemu/config-file.h" 35 #include "qemu/bswap.h" 36 #include "qemu/log.h" 37 #include "qemu/systemd.h" 38 #include "block/snapshot.h" 39 #include "qapi/qmp/qdict.h" 40 #include "qapi/qmp/qstring.h" 41 #include "qom/object_interfaces.h" 42 #include "io/channel-socket.h" 43 #include "io/net-listener.h" 44 #include "crypto/init.h" 45 #include "trace/control.h" 46 #include "qemu-version.h" 47 48 #ifdef __linux__ 49 #define HAVE_NBD_DEVICE 1 50 #else 51 #define HAVE_NBD_DEVICE 0 52 #endif 53 54 #define SOCKET_PATH "/var/lock/qemu-nbd-%s" 55 #define QEMU_NBD_OPT_CACHE 256 56 #define QEMU_NBD_OPT_AIO 257 57 #define QEMU_NBD_OPT_DISCARD 258 58 #define QEMU_NBD_OPT_DETECT_ZEROES 259 59 #define QEMU_NBD_OPT_OBJECT 260 60 #define QEMU_NBD_OPT_TLSCREDS 261 61 #define QEMU_NBD_OPT_IMAGE_OPTS 262 62 #define QEMU_NBD_OPT_FORK 263 63 #define QEMU_NBD_OPT_TLSAUTHZ 264 64 65 #define MBR_SIZE 512 66 67 static NBDExport *export; 68 static int verbose; 69 static char *srcpath; 70 static SocketAddress *saddr; 71 static int persistent = 0; 72 static enum { RUNNING, TERMINATE, TERMINATING, TERMINATED } state; 73 static int shared = 1; 74 static int nb_fds; 75 static QIONetListener *server; 76 static QCryptoTLSCreds *tlscreds; 77 static const char *tlsauthz; 78 79 static void usage(const char *name) 80 { 81 (printf) ( 82 "Usage: %s [OPTIONS] FILE\n" 83 " or: %s -L [OPTIONS]\n" 84 "QEMU Disk Network Block Device Utility\n" 85 "\n" 86 " -h, --help display this help and exit\n" 87 " -V, --version output version information and exit\n" 88 "\n" 89 "Connection properties:\n" 90 " -p, --port=PORT port to listen on (default `%d')\n" 91 " -b, --bind=IFACE interface to bind to (default `0.0.0.0')\n" 92 " -k, --socket=PATH path to the unix socket\n" 93 " (default '"SOCKET_PATH"')\n" 94 " -e, --shared=NUM device can be shared by NUM clients (default '1')\n" 95 " -t, --persistent don't exit on the last connection\n" 96 " -v, --verbose display extra debugging information\n" 97 " -x, --export-name=NAME expose export by name (default is empty string)\n" 98 " -D, --description=TEXT export a human-readable description\n" 99 "\n" 100 "Exposing part of the image:\n" 101 " -o, --offset=OFFSET offset into the image\n" 102 " -P, --partition=NUM only expose partition NUM\n" 103 " -B, --bitmap=NAME expose a persistent dirty bitmap\n" 104 "\n" 105 "General purpose options:\n" 106 " -L, --list list exports available from another NBD server\n" 107 " --object type,id=ID,... define an object such as 'secret' for providing\n" 108 " passwords and/or encryption keys\n" 109 " --tls-creds=ID use id of an earlier --object to provide TLS\n" 110 " --tls-authz=ID use id of an earlier --object to provide\n" 111 " authorization\n" 112 " -T, --trace [[enable=]<pattern>][,events=<file>][,file=<file>]\n" 113 " specify tracing options\n" 114 " --fork fork off the server process and exit the parent\n" 115 " once the server is running\n" 116 #if HAVE_NBD_DEVICE 117 "\n" 118 "Kernel NBD client support:\n" 119 " -c, --connect=DEV connect FILE to the local NBD device DEV\n" 120 " -d, --disconnect disconnect the specified device\n" 121 #endif 122 "\n" 123 "Block device options:\n" 124 " -f, --format=FORMAT set image format (raw, qcow2, ...)\n" 125 " -r, --read-only export read-only\n" 126 " -s, --snapshot use FILE as an external snapshot, create a temporary\n" 127 " file with backing_file=FILE, redirect the write to\n" 128 " the temporary one\n" 129 " -l, --load-snapshot=SNAPSHOT_PARAM\n" 130 " load an internal snapshot inside FILE and export it\n" 131 " as an read-only device, SNAPSHOT_PARAM format is\n" 132 " 'snapshot.id=[ID],snapshot.name=[NAME]', or\n" 133 " '[ID_OR_NAME]'\n" 134 " -n, --nocache disable host cache\n" 135 " --cache=MODE set cache mode (none, writeback, ...)\n" 136 " --aio=MODE set AIO mode (native or threads)\n" 137 " --discard=MODE set discard mode (ignore, unmap)\n" 138 " --detect-zeroes=MODE set detect-zeroes mode (off, on, unmap)\n" 139 " --image-opts treat FILE as a full set of image options\n" 140 "\n" 141 QEMU_HELP_BOTTOM "\n" 142 , name, name, NBD_DEFAULT_PORT, "DEVICE"); 143 } 144 145 static void version(const char *name) 146 { 147 printf( 148 "%s " QEMU_FULL_VERSION "\n" 149 "Written by Anthony Liguori.\n" 150 "\n" 151 QEMU_COPYRIGHT "\n" 152 "This is free software; see the source for copying conditions. There is NO\n" 153 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n" 154 , name); 155 } 156 157 struct partition_record 158 { 159 uint8_t bootable; 160 uint8_t start_head; 161 uint32_t start_cylinder; 162 uint8_t start_sector; 163 uint8_t system; 164 uint8_t end_head; 165 uint8_t end_cylinder; 166 uint8_t end_sector; 167 uint32_t start_sector_abs; 168 uint32_t nb_sectors_abs; 169 }; 170 171 static void read_partition(uint8_t *p, struct partition_record *r) 172 { 173 r->bootable = p[0]; 174 r->start_head = p[1]; 175 r->start_cylinder = p[3] | ((p[2] << 2) & 0x0300); 176 r->start_sector = p[2] & 0x3f; 177 r->system = p[4]; 178 r->end_head = p[5]; 179 r->end_cylinder = p[7] | ((p[6] << 2) & 0x300); 180 r->end_sector = p[6] & 0x3f; 181 182 r->start_sector_abs = ldl_le_p(p + 8); 183 r->nb_sectors_abs = ldl_le_p(p + 12); 184 } 185 186 static int find_partition(BlockBackend *blk, int partition, 187 uint64_t *offset, uint64_t *size) 188 { 189 struct partition_record mbr[4]; 190 uint8_t data[MBR_SIZE]; 191 int i; 192 int ext_partnum = 4; 193 int ret; 194 195 ret = blk_pread(blk, 0, data, sizeof(data)); 196 if (ret < 0) { 197 error_report("error while reading: %s", strerror(-ret)); 198 exit(EXIT_FAILURE); 199 } 200 201 if (data[510] != 0x55 || data[511] != 0xaa) { 202 return -EINVAL; 203 } 204 205 for (i = 0; i < 4; i++) { 206 read_partition(&data[446 + 16 * i], &mbr[i]); 207 208 if (!mbr[i].system || !mbr[i].nb_sectors_abs) { 209 continue; 210 } 211 212 if (mbr[i].system == 0xF || mbr[i].system == 0x5) { 213 struct partition_record ext[4]; 214 uint8_t data1[MBR_SIZE]; 215 int j; 216 217 ret = blk_pread(blk, mbr[i].start_sector_abs * MBR_SIZE, 218 data1, sizeof(data1)); 219 if (ret < 0) { 220 error_report("error while reading: %s", strerror(-ret)); 221 exit(EXIT_FAILURE); 222 } 223 224 for (j = 0; j < 4; j++) { 225 read_partition(&data1[446 + 16 * j], &ext[j]); 226 if (!ext[j].system || !ext[j].nb_sectors_abs) { 227 continue; 228 } 229 230 if ((ext_partnum + j + 1) == partition) { 231 *offset = (uint64_t)ext[j].start_sector_abs << 9; 232 *size = (uint64_t)ext[j].nb_sectors_abs << 9; 233 return 0; 234 } 235 } 236 ext_partnum += 4; 237 } else if ((i + 1) == partition) { 238 *offset = (uint64_t)mbr[i].start_sector_abs << 9; 239 *size = (uint64_t)mbr[i].nb_sectors_abs << 9; 240 return 0; 241 } 242 } 243 244 return -ENOENT; 245 } 246 247 static void termsig_handler(int signum) 248 { 249 atomic_cmpxchg(&state, RUNNING, TERMINATE); 250 qemu_notify_event(); 251 } 252 253 254 static int qemu_nbd_client_list(SocketAddress *saddr, QCryptoTLSCreds *tls, 255 const char *hostname) 256 { 257 int ret = EXIT_FAILURE; 258 int rc; 259 Error *err = NULL; 260 QIOChannelSocket *sioc; 261 NBDExportInfo *list; 262 int i, j; 263 264 sioc = qio_channel_socket_new(); 265 if (qio_channel_socket_connect_sync(sioc, saddr, &err) < 0) { 266 error_report_err(err); 267 return EXIT_FAILURE; 268 } 269 rc = nbd_receive_export_list(QIO_CHANNEL(sioc), tls, hostname, &list, 270 &err); 271 if (rc < 0) { 272 if (err) { 273 error_report_err(err); 274 } 275 goto out; 276 } 277 printf("exports available: %d\n", rc); 278 for (i = 0; i < rc; i++) { 279 printf(" export: '%s'\n", list[i].name); 280 if (list[i].description && *list[i].description) { 281 printf(" description: %s\n", list[i].description); 282 } 283 if (list[i].flags & NBD_FLAG_HAS_FLAGS) { 284 static const char *const flag_names[] = { 285 [NBD_FLAG_READ_ONLY_BIT] = "readonly", 286 [NBD_FLAG_SEND_FLUSH_BIT] = "flush", 287 [NBD_FLAG_SEND_FUA_BIT] = "fua", 288 [NBD_FLAG_ROTATIONAL_BIT] = "rotational", 289 [NBD_FLAG_SEND_TRIM_BIT] = "trim", 290 [NBD_FLAG_SEND_WRITE_ZEROES_BIT] = "zeroes", 291 [NBD_FLAG_SEND_DF_BIT] = "df", 292 [NBD_FLAG_CAN_MULTI_CONN_BIT] = "multi", 293 [NBD_FLAG_SEND_RESIZE_BIT] = "resize", 294 [NBD_FLAG_SEND_CACHE_BIT] = "cache", 295 }; 296 297 printf(" size: %" PRIu64 "\n", list[i].size); 298 printf(" flags: 0x%x (", list[i].flags); 299 for (size_t bit = 0; bit < ARRAY_SIZE(flag_names); bit++) { 300 if (flag_names[bit] && (list[i].flags & (1 << bit))) { 301 printf(" %s", flag_names[bit]); 302 } 303 } 304 printf(" )\n"); 305 } 306 if (list[i].min_block) { 307 printf(" min block: %u\n", list[i].min_block); 308 printf(" opt block: %u\n", list[i].opt_block); 309 printf(" max block: %u\n", list[i].max_block); 310 } 311 if (list[i].n_contexts) { 312 printf(" available meta contexts: %d\n", list[i].n_contexts); 313 for (j = 0; j < list[i].n_contexts; j++) { 314 printf(" %s\n", list[i].contexts[j]); 315 } 316 } 317 } 318 nbd_free_export_list(list, rc); 319 320 ret = EXIT_SUCCESS; 321 out: 322 object_unref(OBJECT(sioc)); 323 return ret; 324 } 325 326 327 #if HAVE_NBD_DEVICE 328 static void *show_parts(void *arg) 329 { 330 char *device = arg; 331 int nbd; 332 333 /* linux just needs an open() to trigger 334 * the partition table update 335 * but remember to load the module with max_part != 0 : 336 * modprobe nbd max_part=63 337 */ 338 nbd = open(device, O_RDWR); 339 if (nbd >= 0) { 340 close(nbd); 341 } 342 return NULL; 343 } 344 345 static void *nbd_client_thread(void *arg) 346 { 347 char *device = arg; 348 NBDExportInfo info = { .request_sizes = false, .name = g_strdup("") }; 349 QIOChannelSocket *sioc; 350 int fd; 351 int ret; 352 pthread_t show_parts_thread; 353 Error *local_error = NULL; 354 355 sioc = qio_channel_socket_new(); 356 if (qio_channel_socket_connect_sync(sioc, 357 saddr, 358 &local_error) < 0) { 359 error_report_err(local_error); 360 goto out; 361 } 362 363 ret = nbd_receive_negotiate(QIO_CHANNEL(sioc), 364 NULL, NULL, NULL, &info, &local_error); 365 if (ret < 0) { 366 if (local_error) { 367 error_report_err(local_error); 368 } 369 goto out_socket; 370 } 371 372 fd = open(device, O_RDWR); 373 if (fd < 0) { 374 /* Linux-only, we can use %m in printf. */ 375 error_report("Failed to open %s: %m", device); 376 goto out_socket; 377 } 378 379 ret = nbd_init(fd, sioc, &info, &local_error); 380 if (ret < 0) { 381 error_report_err(local_error); 382 goto out_fd; 383 } 384 385 /* update partition table */ 386 pthread_create(&show_parts_thread, NULL, show_parts, device); 387 388 if (verbose) { 389 fprintf(stderr, "NBD device %s is now connected to %s\n", 390 device, srcpath); 391 } else { 392 /* Close stderr so that the qemu-nbd process exits. */ 393 dup2(STDOUT_FILENO, STDERR_FILENO); 394 } 395 396 ret = nbd_client(fd); 397 if (ret) { 398 goto out_fd; 399 } 400 close(fd); 401 object_unref(OBJECT(sioc)); 402 g_free(info.name); 403 kill(getpid(), SIGTERM); 404 return (void *) EXIT_SUCCESS; 405 406 out_fd: 407 close(fd); 408 out_socket: 409 object_unref(OBJECT(sioc)); 410 out: 411 g_free(info.name); 412 kill(getpid(), SIGTERM); 413 return (void *) EXIT_FAILURE; 414 } 415 #endif /* HAVE_NBD_DEVICE */ 416 417 static int nbd_can_accept(void) 418 { 419 return state == RUNNING && nb_fds < shared; 420 } 421 422 static void nbd_export_closed(NBDExport *export) 423 { 424 assert(state == TERMINATING); 425 state = TERMINATED; 426 } 427 428 static void nbd_update_server_watch(void); 429 430 static void nbd_client_closed(NBDClient *client, bool negotiated) 431 { 432 nb_fds--; 433 if (negotiated && nb_fds == 0 && !persistent && state == RUNNING) { 434 state = TERMINATE; 435 } 436 nbd_update_server_watch(); 437 nbd_client_put(client); 438 } 439 440 static void nbd_accept(QIONetListener *listener, QIOChannelSocket *cioc, 441 gpointer opaque) 442 { 443 if (state >= TERMINATE) { 444 return; 445 } 446 447 nb_fds++; 448 nbd_update_server_watch(); 449 nbd_client_new(cioc, tlscreds, tlsauthz, nbd_client_closed); 450 } 451 452 static void nbd_update_server_watch(void) 453 { 454 if (nbd_can_accept()) { 455 qio_net_listener_set_client_func(server, nbd_accept, NULL, NULL); 456 } else { 457 qio_net_listener_set_client_func(server, NULL, NULL, NULL); 458 } 459 } 460 461 462 static SocketAddress *nbd_build_socket_address(const char *sockpath, 463 const char *bindto, 464 const char *port) 465 { 466 SocketAddress *saddr; 467 468 saddr = g_new0(SocketAddress, 1); 469 if (sockpath) { 470 saddr->type = SOCKET_ADDRESS_TYPE_UNIX; 471 saddr->u.q_unix.path = g_strdup(sockpath); 472 } else { 473 InetSocketAddress *inet; 474 saddr->type = SOCKET_ADDRESS_TYPE_INET; 475 inet = &saddr->u.inet; 476 inet->host = g_strdup(bindto); 477 if (port) { 478 inet->port = g_strdup(port); 479 } else { 480 inet->port = g_strdup_printf("%d", NBD_DEFAULT_PORT); 481 } 482 } 483 484 return saddr; 485 } 486 487 488 static QemuOptsList file_opts = { 489 .name = "file", 490 .implied_opt_name = "file", 491 .head = QTAILQ_HEAD_INITIALIZER(file_opts.head), 492 .desc = { 493 /* no elements => accept any params */ 494 { /* end of list */ } 495 }, 496 }; 497 498 static QemuOptsList qemu_object_opts = { 499 .name = "object", 500 .implied_opt_name = "qom-type", 501 .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head), 502 .desc = { 503 { } 504 }, 505 }; 506 507 508 509 static QCryptoTLSCreds *nbd_get_tls_creds(const char *id, bool list, 510 Error **errp) 511 { 512 Object *obj; 513 QCryptoTLSCreds *creds; 514 515 obj = object_resolve_path_component( 516 object_get_objects_root(), id); 517 if (!obj) { 518 error_setg(errp, "No TLS credentials with id '%s'", 519 id); 520 return NULL; 521 } 522 creds = (QCryptoTLSCreds *) 523 object_dynamic_cast(obj, TYPE_QCRYPTO_TLS_CREDS); 524 if (!creds) { 525 error_setg(errp, "Object with id '%s' is not TLS credentials", 526 id); 527 return NULL; 528 } 529 530 if (list) { 531 if (creds->endpoint != QCRYPTO_TLS_CREDS_ENDPOINT_CLIENT) { 532 error_setg(errp, 533 "Expecting TLS credentials with a client endpoint"); 534 return NULL; 535 } 536 } else { 537 if (creds->endpoint != QCRYPTO_TLS_CREDS_ENDPOINT_SERVER) { 538 error_setg(errp, 539 "Expecting TLS credentials with a server endpoint"); 540 return NULL; 541 } 542 } 543 object_ref(obj); 544 return creds; 545 } 546 547 static void setup_address_and_port(const char **address, const char **port) 548 { 549 if (*address == NULL) { 550 *address = "0.0.0.0"; 551 } 552 553 if (*port == NULL) { 554 *port = stringify(NBD_DEFAULT_PORT); 555 } 556 } 557 558 /* 559 * Check socket parameters compatibility when socket activation is used. 560 */ 561 static const char *socket_activation_validate_opts(const char *device, 562 const char *sockpath, 563 const char *address, 564 const char *port, 565 bool list) 566 { 567 if (device != NULL) { 568 return "NBD device can't be set when using socket activation"; 569 } 570 571 if (sockpath != NULL) { 572 return "Unix socket can't be set when using socket activation"; 573 } 574 575 if (address != NULL) { 576 return "The interface can't be set when using socket activation"; 577 } 578 579 if (port != NULL) { 580 return "TCP port number can't be set when using socket activation"; 581 } 582 583 if (list) { 584 return "List mode is incompatible with socket activation"; 585 } 586 587 return NULL; 588 } 589 590 static void qemu_nbd_shutdown(void) 591 { 592 job_cancel_sync_all(); 593 bdrv_close_all(); 594 } 595 596 int main(int argc, char **argv) 597 { 598 BlockBackend *blk; 599 BlockDriverState *bs; 600 uint64_t dev_offset = 0; 601 uint16_t nbdflags = 0; 602 bool disconnect = false; 603 const char *bindto = NULL; 604 const char *port = NULL; 605 char *sockpath = NULL; 606 char *device = NULL; 607 int64_t fd_size; 608 QemuOpts *sn_opts = NULL; 609 const char *sn_id_or_name = NULL; 610 const char *sopt = "hVb:o:p:rsnP:c:dvk:e:f:tl:x:T:D:B:L"; 611 struct option lopt[] = { 612 { "help", no_argument, NULL, 'h' }, 613 { "version", no_argument, NULL, 'V' }, 614 { "bind", required_argument, NULL, 'b' }, 615 { "port", required_argument, NULL, 'p' }, 616 { "socket", required_argument, NULL, 'k' }, 617 { "offset", required_argument, NULL, 'o' }, 618 { "read-only", no_argument, NULL, 'r' }, 619 { "partition", required_argument, NULL, 'P' }, 620 { "bitmap", required_argument, NULL, 'B' }, 621 { "connect", required_argument, NULL, 'c' }, 622 { "disconnect", no_argument, NULL, 'd' }, 623 { "list", no_argument, NULL, 'L' }, 624 { "snapshot", no_argument, NULL, 's' }, 625 { "load-snapshot", required_argument, NULL, 'l' }, 626 { "nocache", no_argument, NULL, 'n' }, 627 { "cache", required_argument, NULL, QEMU_NBD_OPT_CACHE }, 628 { "aio", required_argument, NULL, QEMU_NBD_OPT_AIO }, 629 { "discard", required_argument, NULL, QEMU_NBD_OPT_DISCARD }, 630 { "detect-zeroes", required_argument, NULL, 631 QEMU_NBD_OPT_DETECT_ZEROES }, 632 { "shared", required_argument, NULL, 'e' }, 633 { "format", required_argument, NULL, 'f' }, 634 { "persistent", no_argument, NULL, 't' }, 635 { "verbose", no_argument, NULL, 'v' }, 636 { "object", required_argument, NULL, QEMU_NBD_OPT_OBJECT }, 637 { "export-name", required_argument, NULL, 'x' }, 638 { "description", required_argument, NULL, 'D' }, 639 { "tls-creds", required_argument, NULL, QEMU_NBD_OPT_TLSCREDS }, 640 { "tls-authz", required_argument, NULL, QEMU_NBD_OPT_TLSAUTHZ }, 641 { "image-opts", no_argument, NULL, QEMU_NBD_OPT_IMAGE_OPTS }, 642 { "trace", required_argument, NULL, 'T' }, 643 { "fork", no_argument, NULL, QEMU_NBD_OPT_FORK }, 644 { NULL, 0, NULL, 0 } 645 }; 646 int ch; 647 int opt_ind = 0; 648 int flags = BDRV_O_RDWR; 649 int partition = 0; 650 int ret = 0; 651 bool seen_cache = false; 652 bool seen_discard = false; 653 bool seen_aio = false; 654 pthread_t client_thread; 655 const char *fmt = NULL; 656 Error *local_err = NULL; 657 BlockdevDetectZeroesOptions detect_zeroes = BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF; 658 QDict *options = NULL; 659 const char *export_name = NULL; /* defaults to "" later for server mode */ 660 const char *export_description = NULL; 661 const char *bitmap = NULL; 662 const char *tlscredsid = NULL; 663 bool imageOpts = false; 664 bool writethrough = true; 665 char *trace_file = NULL; 666 bool fork_process = false; 667 bool list = false; 668 int old_stderr = -1; 669 unsigned socket_activation; 670 671 /* The client thread uses SIGTERM to interrupt the server. A signal 672 * handler ensures that "qemu-nbd -v -c" exits with a nice status code. 673 */ 674 struct sigaction sa_sigterm; 675 memset(&sa_sigterm, 0, sizeof(sa_sigterm)); 676 sa_sigterm.sa_handler = termsig_handler; 677 sigaction(SIGTERM, &sa_sigterm, NULL); 678 679 #ifdef CONFIG_POSIX 680 signal(SIGPIPE, SIG_IGN); 681 #endif 682 683 error_init(argv[0]); 684 module_call_init(MODULE_INIT_TRACE); 685 qcrypto_init(&error_fatal); 686 687 module_call_init(MODULE_INIT_QOM); 688 qemu_add_opts(&qemu_object_opts); 689 qemu_add_opts(&qemu_trace_opts); 690 qemu_init_exec_dir(argv[0]); 691 692 while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) { 693 switch (ch) { 694 case 's': 695 flags |= BDRV_O_SNAPSHOT; 696 break; 697 case 'n': 698 optarg = (char *) "none"; 699 /* fallthrough */ 700 case QEMU_NBD_OPT_CACHE: 701 if (seen_cache) { 702 error_report("-n and --cache can only be specified once"); 703 exit(EXIT_FAILURE); 704 } 705 seen_cache = true; 706 if (bdrv_parse_cache_mode(optarg, &flags, &writethrough) == -1) { 707 error_report("Invalid cache mode `%s'", optarg); 708 exit(EXIT_FAILURE); 709 } 710 break; 711 case QEMU_NBD_OPT_AIO: 712 if (seen_aio) { 713 error_report("--aio can only be specified once"); 714 exit(EXIT_FAILURE); 715 } 716 seen_aio = true; 717 if (!strcmp(optarg, "native")) { 718 flags |= BDRV_O_NATIVE_AIO; 719 } else if (!strcmp(optarg, "threads")) { 720 /* this is the default */ 721 } else { 722 error_report("invalid aio mode `%s'", optarg); 723 exit(EXIT_FAILURE); 724 } 725 break; 726 case QEMU_NBD_OPT_DISCARD: 727 if (seen_discard) { 728 error_report("--discard can only be specified once"); 729 exit(EXIT_FAILURE); 730 } 731 seen_discard = true; 732 if (bdrv_parse_discard_flags(optarg, &flags) == -1) { 733 error_report("Invalid discard mode `%s'", optarg); 734 exit(EXIT_FAILURE); 735 } 736 break; 737 case QEMU_NBD_OPT_DETECT_ZEROES: 738 detect_zeroes = 739 qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup, 740 optarg, 741 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF, 742 &local_err); 743 if (local_err) { 744 error_reportf_err(local_err, 745 "Failed to parse detect_zeroes mode: "); 746 exit(EXIT_FAILURE); 747 } 748 if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP && 749 !(flags & BDRV_O_UNMAP)) { 750 error_report("setting detect-zeroes to unmap is not allowed " 751 "without setting discard operation to unmap"); 752 exit(EXIT_FAILURE); 753 } 754 break; 755 case 'b': 756 bindto = optarg; 757 break; 758 case 'p': 759 port = optarg; 760 break; 761 case 'o': 762 if (qemu_strtou64(optarg, NULL, 0, &dev_offset) < 0) { 763 error_report("Invalid offset '%s'", optarg); 764 exit(EXIT_FAILURE); 765 } 766 break; 767 case 'l': 768 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) { 769 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts, 770 optarg, false); 771 if (!sn_opts) { 772 error_report("Failed in parsing snapshot param `%s'", 773 optarg); 774 exit(EXIT_FAILURE); 775 } 776 } else { 777 sn_id_or_name = optarg; 778 } 779 /* fall through */ 780 case 'r': 781 nbdflags |= NBD_FLAG_READ_ONLY; 782 flags &= ~BDRV_O_RDWR; 783 break; 784 case 'P': 785 warn_report("The '-P' option is deprecated; use --image-opts with " 786 "a raw device wrapper for subset exports instead"); 787 if (qemu_strtoi(optarg, NULL, 0, &partition) < 0 || 788 partition < 1 || partition > 8) { 789 error_report("Invalid partition '%s'", optarg); 790 exit(EXIT_FAILURE); 791 } 792 break; 793 case 'B': 794 bitmap = optarg; 795 break; 796 case 'k': 797 sockpath = optarg; 798 if (sockpath[0] != '/') { 799 error_report("socket path must be absolute"); 800 exit(EXIT_FAILURE); 801 } 802 break; 803 case 'd': 804 disconnect = true; 805 break; 806 case 'c': 807 device = optarg; 808 break; 809 case 'e': 810 if (qemu_strtoi(optarg, NULL, 0, &shared) < 0 || 811 shared < 1) { 812 error_report("Invalid shared device number '%s'", optarg); 813 exit(EXIT_FAILURE); 814 } 815 break; 816 case 'f': 817 fmt = optarg; 818 break; 819 case 't': 820 persistent = 1; 821 break; 822 case 'x': 823 export_name = optarg; 824 break; 825 case 'D': 826 export_description = optarg; 827 break; 828 case 'v': 829 verbose = 1; 830 break; 831 case 'V': 832 version(argv[0]); 833 exit(0); 834 break; 835 case 'h': 836 usage(argv[0]); 837 exit(0); 838 break; 839 case '?': 840 error_report("Try `%s --help' for more information.", argv[0]); 841 exit(EXIT_FAILURE); 842 case QEMU_NBD_OPT_OBJECT: { 843 QemuOpts *opts; 844 opts = qemu_opts_parse_noisily(&qemu_object_opts, 845 optarg, true); 846 if (!opts) { 847 exit(EXIT_FAILURE); 848 } 849 } break; 850 case QEMU_NBD_OPT_TLSCREDS: 851 tlscredsid = optarg; 852 break; 853 case QEMU_NBD_OPT_IMAGE_OPTS: 854 imageOpts = true; 855 break; 856 case 'T': 857 g_free(trace_file); 858 trace_file = trace_opt_parse(optarg); 859 break; 860 case QEMU_NBD_OPT_TLSAUTHZ: 861 tlsauthz = optarg; 862 break; 863 case QEMU_NBD_OPT_FORK: 864 fork_process = true; 865 break; 866 case 'L': 867 list = true; 868 break; 869 } 870 } 871 872 if (list) { 873 if (argc != optind) { 874 error_report("List mode is incompatible with a file name"); 875 exit(EXIT_FAILURE); 876 } 877 if (export_name || export_description || dev_offset || partition || 878 device || disconnect || fmt || sn_id_or_name || bitmap || 879 seen_aio || seen_discard || seen_cache) { 880 error_report("List mode is incompatible with per-device settings"); 881 exit(EXIT_FAILURE); 882 } 883 if (fork_process) { 884 error_report("List mode is incompatible with forking"); 885 exit(EXIT_FAILURE); 886 } 887 } else if ((argc - optind) != 1) { 888 error_report("Invalid number of arguments"); 889 error_printf("Try `%s --help' for more information.\n", argv[0]); 890 exit(EXIT_FAILURE); 891 } else if (!export_name) { 892 export_name = ""; 893 } 894 895 qemu_opts_foreach(&qemu_object_opts, 896 user_creatable_add_opts_foreach, 897 NULL, &error_fatal); 898 899 if (!trace_init_backends()) { 900 exit(1); 901 } 902 trace_init_file(trace_file); 903 qemu_set_log(LOG_TRACE); 904 905 socket_activation = check_socket_activation(); 906 if (socket_activation == 0) { 907 setup_address_and_port(&bindto, &port); 908 } else { 909 /* Using socket activation - check user didn't use -p etc. */ 910 const char *err_msg = socket_activation_validate_opts(device, sockpath, 911 bindto, port, 912 list); 913 if (err_msg != NULL) { 914 error_report("%s", err_msg); 915 exit(EXIT_FAILURE); 916 } 917 918 /* qemu-nbd can only listen on a single socket. */ 919 if (socket_activation > 1) { 920 error_report("qemu-nbd does not support socket activation with %s > 1", 921 "LISTEN_FDS"); 922 exit(EXIT_FAILURE); 923 } 924 } 925 926 if (tlscredsid) { 927 if (sockpath) { 928 error_report("TLS is only supported with IPv4/IPv6"); 929 exit(EXIT_FAILURE); 930 } 931 if (device) { 932 error_report("TLS is not supported with a host device"); 933 exit(EXIT_FAILURE); 934 } 935 if (tlsauthz && list) { 936 error_report("TLS authorization is incompatible with export list"); 937 exit(EXIT_FAILURE); 938 } 939 tlscreds = nbd_get_tls_creds(tlscredsid, list, &local_err); 940 if (local_err) { 941 error_report("Failed to get TLS creds %s", 942 error_get_pretty(local_err)); 943 exit(EXIT_FAILURE); 944 } 945 } else { 946 if (tlsauthz) { 947 error_report("--tls-authz is not permitted without --tls-creds"); 948 exit(EXIT_FAILURE); 949 } 950 } 951 952 if (list) { 953 saddr = nbd_build_socket_address(sockpath, bindto, port); 954 return qemu_nbd_client_list(saddr, tlscreds, bindto); 955 } 956 957 #if !HAVE_NBD_DEVICE 958 if (disconnect || device) { 959 error_report("Kernel /dev/nbdN support not available"); 960 exit(EXIT_FAILURE); 961 } 962 #else /* HAVE_NBD_DEVICE */ 963 if (disconnect) { 964 int nbdfd = open(argv[optind], O_RDWR); 965 if (nbdfd < 0) { 966 error_report("Cannot open %s: %s", argv[optind], 967 strerror(errno)); 968 exit(EXIT_FAILURE); 969 } 970 nbd_disconnect(nbdfd); 971 972 close(nbdfd); 973 974 printf("%s disconnected\n", argv[optind]); 975 976 return 0; 977 } 978 #endif 979 980 if ((device && !verbose) || fork_process) { 981 int stderr_fd[2]; 982 pid_t pid; 983 int ret; 984 985 if (qemu_pipe(stderr_fd) < 0) { 986 error_report("Error setting up communication pipe: %s", 987 strerror(errno)); 988 exit(EXIT_FAILURE); 989 } 990 991 /* Now daemonize, but keep a communication channel open to 992 * print errors and exit with the proper status code. 993 */ 994 pid = fork(); 995 if (pid < 0) { 996 error_report("Failed to fork: %s", strerror(errno)); 997 exit(EXIT_FAILURE); 998 } else if (pid == 0) { 999 close(stderr_fd[0]); 1000 ret = qemu_daemon(1, 0); 1001 1002 /* Temporarily redirect stderr to the parent's pipe... */ 1003 old_stderr = dup(STDERR_FILENO); 1004 dup2(stderr_fd[1], STDERR_FILENO); 1005 if (ret < 0) { 1006 error_report("Failed to daemonize: %s", strerror(errno)); 1007 exit(EXIT_FAILURE); 1008 } 1009 1010 /* ... close the descriptor we inherited and go on. */ 1011 close(stderr_fd[1]); 1012 } else { 1013 bool errors = false; 1014 char *buf; 1015 1016 /* In the parent. Print error messages from the child until 1017 * it closes the pipe. 1018 */ 1019 close(stderr_fd[1]); 1020 buf = g_malloc(1024); 1021 while ((ret = read(stderr_fd[0], buf, 1024)) > 0) { 1022 errors = true; 1023 ret = qemu_write_full(STDERR_FILENO, buf, ret); 1024 if (ret < 0) { 1025 exit(EXIT_FAILURE); 1026 } 1027 } 1028 if (ret < 0) { 1029 error_report("Cannot read from daemon: %s", 1030 strerror(errno)); 1031 exit(EXIT_FAILURE); 1032 } 1033 1034 /* Usually the daemon should not print any message. 1035 * Exit with zero status in that case. 1036 */ 1037 exit(errors); 1038 } 1039 } 1040 1041 if (device != NULL && sockpath == NULL) { 1042 sockpath = g_malloc(128); 1043 snprintf(sockpath, 128, SOCKET_PATH, basename(device)); 1044 } 1045 1046 server = qio_net_listener_new(); 1047 if (socket_activation == 0) { 1048 saddr = nbd_build_socket_address(sockpath, bindto, port); 1049 if (qio_net_listener_open_sync(server, saddr, &local_err) < 0) { 1050 object_unref(OBJECT(server)); 1051 error_report_err(local_err); 1052 exit(EXIT_FAILURE); 1053 } 1054 } else { 1055 size_t i; 1056 /* See comment in check_socket_activation above. */ 1057 for (i = 0; i < socket_activation; i++) { 1058 QIOChannelSocket *sioc; 1059 sioc = qio_channel_socket_new_fd(FIRST_SOCKET_ACTIVATION_FD + i, 1060 &local_err); 1061 if (sioc == NULL) { 1062 object_unref(OBJECT(server)); 1063 error_report("Failed to use socket activation: %s", 1064 error_get_pretty(local_err)); 1065 exit(EXIT_FAILURE); 1066 } 1067 qio_net_listener_add(server, sioc); 1068 object_unref(OBJECT(sioc)); 1069 } 1070 } 1071 1072 if (qemu_init_main_loop(&local_err)) { 1073 error_report_err(local_err); 1074 exit(EXIT_FAILURE); 1075 } 1076 bdrv_init(); 1077 atexit(qemu_nbd_shutdown); 1078 1079 srcpath = argv[optind]; 1080 if (imageOpts) { 1081 QemuOpts *opts; 1082 if (fmt) { 1083 error_report("--image-opts and -f are mutually exclusive"); 1084 exit(EXIT_FAILURE); 1085 } 1086 opts = qemu_opts_parse_noisily(&file_opts, srcpath, true); 1087 if (!opts) { 1088 qemu_opts_reset(&file_opts); 1089 exit(EXIT_FAILURE); 1090 } 1091 options = qemu_opts_to_qdict(opts, NULL); 1092 qemu_opts_reset(&file_opts); 1093 blk = blk_new_open(NULL, NULL, options, flags, &local_err); 1094 } else { 1095 if (fmt) { 1096 options = qdict_new(); 1097 qdict_put_str(options, "driver", fmt); 1098 } 1099 blk = blk_new_open(srcpath, NULL, options, flags, &local_err); 1100 } 1101 1102 if (!blk) { 1103 error_reportf_err(local_err, "Failed to blk_new_open '%s': ", 1104 argv[optind]); 1105 exit(EXIT_FAILURE); 1106 } 1107 bs = blk_bs(blk); 1108 1109 blk_set_enable_write_cache(blk, !writethrough); 1110 1111 if (sn_opts) { 1112 ret = bdrv_snapshot_load_tmp(bs, 1113 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID), 1114 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME), 1115 &local_err); 1116 } else if (sn_id_or_name) { 1117 ret = bdrv_snapshot_load_tmp_by_id_or_name(bs, sn_id_or_name, 1118 &local_err); 1119 } 1120 if (ret < 0) { 1121 error_reportf_err(local_err, "Failed to load snapshot: "); 1122 exit(EXIT_FAILURE); 1123 } 1124 1125 bs->detect_zeroes = detect_zeroes; 1126 fd_size = blk_getlength(blk); 1127 if (fd_size < 0) { 1128 error_report("Failed to determine the image length: %s", 1129 strerror(-fd_size)); 1130 exit(EXIT_FAILURE); 1131 } 1132 1133 if (dev_offset >= fd_size) { 1134 error_report("Offset (%" PRIu64 ") has to be smaller than the image " 1135 "size (%" PRId64 ")", dev_offset, fd_size); 1136 exit(EXIT_FAILURE); 1137 } 1138 fd_size -= dev_offset; 1139 1140 if (partition) { 1141 uint64_t limit; 1142 1143 if (dev_offset) { 1144 error_report("Cannot request partition and offset together"); 1145 exit(EXIT_FAILURE); 1146 } 1147 ret = find_partition(blk, partition, &dev_offset, &limit); 1148 if (ret < 0) { 1149 error_report("Could not find partition %d: %s", partition, 1150 strerror(-ret)); 1151 exit(EXIT_FAILURE); 1152 } 1153 /* 1154 * MBR partition limits are (32-bit << 9); this assert lets 1155 * the compiler know that we can't overflow 64 bits. 1156 */ 1157 assert(dev_offset + limit >= dev_offset); 1158 if (dev_offset + limit > fd_size) { 1159 error_report("Discovered partition %d at offset %" PRIu64 1160 " size %" PRIu64 ", but size exceeds file length %" 1161 PRId64, partition, dev_offset, limit, fd_size); 1162 exit(EXIT_FAILURE); 1163 } 1164 fd_size = limit; 1165 } 1166 1167 export = nbd_export_new(bs, dev_offset, fd_size, export_name, 1168 export_description, bitmap, nbdflags, 1169 nbd_export_closed, writethrough, NULL, 1170 &error_fatal); 1171 1172 if (device) { 1173 #if HAVE_NBD_DEVICE 1174 int ret; 1175 1176 ret = pthread_create(&client_thread, NULL, nbd_client_thread, device); 1177 if (ret != 0) { 1178 error_report("Failed to create client thread: %s", strerror(ret)); 1179 exit(EXIT_FAILURE); 1180 } 1181 #endif 1182 } else { 1183 /* Shut up GCC warnings. */ 1184 memset(&client_thread, 0, sizeof(client_thread)); 1185 } 1186 1187 nbd_update_server_watch(); 1188 1189 /* now when the initialization is (almost) complete, chdir("/") 1190 * to free any busy filesystems */ 1191 if (chdir("/") < 0) { 1192 error_report("Could not chdir to root directory: %s", 1193 strerror(errno)); 1194 exit(EXIT_FAILURE); 1195 } 1196 1197 if (fork_process) { 1198 dup2(old_stderr, STDERR_FILENO); 1199 close(old_stderr); 1200 } 1201 1202 state = RUNNING; 1203 do { 1204 main_loop_wait(false); 1205 if (state == TERMINATE) { 1206 state = TERMINATING; 1207 nbd_export_close(export); 1208 nbd_export_put(export); 1209 export = NULL; 1210 } 1211 } while (state != TERMINATED); 1212 1213 blk_unref(blk); 1214 if (sockpath) { 1215 unlink(sockpath); 1216 } 1217 1218 qemu_opts_del(sn_opts); 1219 1220 if (device) { 1221 void *ret; 1222 pthread_join(client_thread, &ret); 1223 exit(ret != NULL); 1224 } else { 1225 exit(EXIT_SUCCESS); 1226 } 1227 } 1228