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