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 #define QEMU_NBD_OPT_PID_FILE 265 65 66 #define MBR_SIZE 512 67 68 static NBDExport *export; 69 static int verbose; 70 static char *srcpath; 71 static SocketAddress *saddr; 72 static int persistent = 0; 73 static enum { RUNNING, TERMINATE, TERMINATING, TERMINATED } state; 74 static int shared = 1; 75 static int nb_fds; 76 static QIONetListener *server; 77 static QCryptoTLSCreds *tlscreds; 78 static const char *tlsauthz; 79 80 static void usage(const char *name) 81 { 82 (printf) ( 83 "Usage: %s [OPTIONS] FILE\n" 84 " or: %s -L [OPTIONS]\n" 85 "QEMU Disk Network Block Device Utility\n" 86 "\n" 87 " -h, --help display this help and exit\n" 88 " -V, --version output version information and exit\n" 89 "\n" 90 "Connection properties:\n" 91 " -p, --port=PORT port to listen on (default `%d')\n" 92 " -b, --bind=IFACE interface to bind to (default `0.0.0.0')\n" 93 " -k, --socket=PATH path to the unix socket\n" 94 " (default '"SOCKET_PATH"')\n" 95 " -e, --shared=NUM device can be shared by NUM clients (default '1')\n" 96 " -t, --persistent don't exit on the last connection\n" 97 " -v, --verbose display extra debugging information\n" 98 " -x, --export-name=NAME expose export by name (default is empty string)\n" 99 " -D, --description=TEXT export a human-readable description\n" 100 "\n" 101 "Exposing part of the image:\n" 102 " -o, --offset=OFFSET offset into the image\n" 103 " -P, --partition=NUM only expose partition NUM\n" 104 " -B, --bitmap=NAME expose a persistent dirty bitmap\n" 105 "\n" 106 "General purpose options:\n" 107 " -L, --list list exports available from another NBD server\n" 108 " --object type,id=ID,... define an object such as 'secret' for providing\n" 109 " passwords and/or encryption keys\n" 110 " --tls-creds=ID use id of an earlier --object to provide TLS\n" 111 " --tls-authz=ID use id of an earlier --object to provide\n" 112 " authorization\n" 113 " -T, --trace [[enable=]<pattern>][,events=<file>][,file=<file>]\n" 114 " specify tracing options\n" 115 " --fork fork off the server process and exit the parent\n" 116 " once the server is running\n" 117 " --pid-file=PATH store the server's process ID in the given file\n" 118 #if HAVE_NBD_DEVICE 119 "\n" 120 "Kernel NBD client support:\n" 121 " -c, --connect=DEV connect FILE to the local NBD device DEV\n" 122 " -d, --disconnect disconnect the specified device\n" 123 #endif 124 "\n" 125 "Block device options:\n" 126 " -f, --format=FORMAT set image format (raw, qcow2, ...)\n" 127 " -r, --read-only export read-only\n" 128 " -s, --snapshot use FILE as an external snapshot, create a temporary\n" 129 " file with backing_file=FILE, redirect the write to\n" 130 " the temporary one\n" 131 " -l, --load-snapshot=SNAPSHOT_PARAM\n" 132 " load an internal snapshot inside FILE and export it\n" 133 " as an read-only device, SNAPSHOT_PARAM format is\n" 134 " 'snapshot.id=[ID],snapshot.name=[NAME]', or\n" 135 " '[ID_OR_NAME]'\n" 136 " -n, --nocache disable host cache\n" 137 " --cache=MODE set cache mode (none, writeback, ...)\n" 138 " --aio=MODE set AIO mode (native or threads)\n" 139 " --discard=MODE set discard mode (ignore, unmap)\n" 140 " --detect-zeroes=MODE set detect-zeroes mode (off, on, unmap)\n" 141 " --image-opts treat FILE as a full set of image options\n" 142 "\n" 143 QEMU_HELP_BOTTOM "\n" 144 , name, name, NBD_DEFAULT_PORT, "DEVICE"); 145 } 146 147 static void version(const char *name) 148 { 149 printf( 150 "%s " QEMU_FULL_VERSION "\n" 151 "Written by Anthony Liguori.\n" 152 "\n" 153 QEMU_COPYRIGHT "\n" 154 "This is free software; see the source for copying conditions. There is NO\n" 155 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n" 156 , name); 157 } 158 159 struct partition_record 160 { 161 uint8_t bootable; 162 uint8_t start_head; 163 uint32_t start_cylinder; 164 uint8_t start_sector; 165 uint8_t system; 166 uint8_t end_head; 167 uint8_t end_cylinder; 168 uint8_t end_sector; 169 uint32_t start_sector_abs; 170 uint32_t nb_sectors_abs; 171 }; 172 173 static void read_partition(uint8_t *p, struct partition_record *r) 174 { 175 r->bootable = p[0]; 176 r->start_head = p[1]; 177 r->start_cylinder = p[3] | ((p[2] << 2) & 0x0300); 178 r->start_sector = p[2] & 0x3f; 179 r->system = p[4]; 180 r->end_head = p[5]; 181 r->end_cylinder = p[7] | ((p[6] << 2) & 0x300); 182 r->end_sector = p[6] & 0x3f; 183 184 r->start_sector_abs = ldl_le_p(p + 8); 185 r->nb_sectors_abs = ldl_le_p(p + 12); 186 } 187 188 static int find_partition(BlockBackend *blk, int partition, 189 uint64_t *offset, uint64_t *size) 190 { 191 struct partition_record mbr[4]; 192 uint8_t data[MBR_SIZE]; 193 int i; 194 int ext_partnum = 4; 195 int ret; 196 197 ret = blk_pread(blk, 0, data, sizeof(data)); 198 if (ret < 0) { 199 error_report("error while reading: %s", strerror(-ret)); 200 exit(EXIT_FAILURE); 201 } 202 203 if (data[510] != 0x55 || data[511] != 0xaa) { 204 return -EINVAL; 205 } 206 207 for (i = 0; i < 4; i++) { 208 read_partition(&data[446 + 16 * i], &mbr[i]); 209 210 if (!mbr[i].system || !mbr[i].nb_sectors_abs) { 211 continue; 212 } 213 214 if (mbr[i].system == 0xF || mbr[i].system == 0x5) { 215 struct partition_record ext[4]; 216 uint8_t data1[MBR_SIZE]; 217 int j; 218 219 ret = blk_pread(blk, mbr[i].start_sector_abs * MBR_SIZE, 220 data1, sizeof(data1)); 221 if (ret < 0) { 222 error_report("error while reading: %s", strerror(-ret)); 223 exit(EXIT_FAILURE); 224 } 225 226 for (j = 0; j < 4; j++) { 227 read_partition(&data1[446 + 16 * j], &ext[j]); 228 if (!ext[j].system || !ext[j].nb_sectors_abs) { 229 continue; 230 } 231 232 if ((ext_partnum + j + 1) == partition) { 233 *offset = (uint64_t)ext[j].start_sector_abs << 9; 234 *size = (uint64_t)ext[j].nb_sectors_abs << 9; 235 return 0; 236 } 237 } 238 ext_partnum += 4; 239 } else if ((i + 1) == partition) { 240 *offset = (uint64_t)mbr[i].start_sector_abs << 9; 241 *size = (uint64_t)mbr[i].nb_sectors_abs << 9; 242 return 0; 243 } 244 } 245 246 return -ENOENT; 247 } 248 249 static void termsig_handler(int signum) 250 { 251 atomic_cmpxchg(&state, RUNNING, TERMINATE); 252 qemu_notify_event(); 253 } 254 255 256 static int qemu_nbd_client_list(SocketAddress *saddr, QCryptoTLSCreds *tls, 257 const char *hostname) 258 { 259 int ret = EXIT_FAILURE; 260 int rc; 261 Error *err = NULL; 262 QIOChannelSocket *sioc; 263 NBDExportInfo *list; 264 int i, j; 265 266 sioc = qio_channel_socket_new(); 267 if (qio_channel_socket_connect_sync(sioc, saddr, &err) < 0) { 268 error_report_err(err); 269 return EXIT_FAILURE; 270 } 271 rc = nbd_receive_export_list(QIO_CHANNEL(sioc), tls, hostname, &list, 272 &err); 273 if (rc < 0) { 274 if (err) { 275 error_report_err(err); 276 } 277 goto out; 278 } 279 printf("exports available: %d\n", rc); 280 for (i = 0; i < rc; i++) { 281 printf(" export: '%s'\n", list[i].name); 282 if (list[i].description && *list[i].description) { 283 printf(" description: %s\n", list[i].description); 284 } 285 if (list[i].flags & NBD_FLAG_HAS_FLAGS) { 286 static const char *const flag_names[] = { 287 [NBD_FLAG_READ_ONLY_BIT] = "readonly", 288 [NBD_FLAG_SEND_FLUSH_BIT] = "flush", 289 [NBD_FLAG_SEND_FUA_BIT] = "fua", 290 [NBD_FLAG_ROTATIONAL_BIT] = "rotational", 291 [NBD_FLAG_SEND_TRIM_BIT] = "trim", 292 [NBD_FLAG_SEND_WRITE_ZEROES_BIT] = "zeroes", 293 [NBD_FLAG_SEND_DF_BIT] = "df", 294 [NBD_FLAG_CAN_MULTI_CONN_BIT] = "multi", 295 [NBD_FLAG_SEND_RESIZE_BIT] = "resize", 296 [NBD_FLAG_SEND_CACHE_BIT] = "cache", 297 [NBD_FLAG_SEND_FAST_ZERO_BIT] = "fast-zero", 298 }; 299 300 printf(" size: %" PRIu64 "\n", list[i].size); 301 printf(" flags: 0x%x (", list[i].flags); 302 for (size_t bit = 0; bit < ARRAY_SIZE(flag_names); bit++) { 303 if (flag_names[bit] && (list[i].flags & (1 << bit))) { 304 printf(" %s", flag_names[bit]); 305 } 306 } 307 printf(" )\n"); 308 } 309 if (list[i].min_block) { 310 printf(" min block: %u\n", list[i].min_block); 311 printf(" opt block: %u\n", list[i].opt_block); 312 printf(" max block: %u\n", list[i].max_block); 313 } 314 if (list[i].n_contexts) { 315 printf(" available meta contexts: %d\n", list[i].n_contexts); 316 for (j = 0; j < list[i].n_contexts; j++) { 317 printf(" %s\n", list[i].contexts[j]); 318 } 319 } 320 } 321 nbd_free_export_list(list, rc); 322 323 ret = EXIT_SUCCESS; 324 out: 325 object_unref(OBJECT(sioc)); 326 return ret; 327 } 328 329 330 #if HAVE_NBD_DEVICE 331 static void *show_parts(void *arg) 332 { 333 char *device = arg; 334 int nbd; 335 336 /* linux just needs an open() to trigger 337 * the partition table update 338 * but remember to load the module with max_part != 0 : 339 * modprobe nbd max_part=63 340 */ 341 nbd = open(device, O_RDWR); 342 if (nbd >= 0) { 343 close(nbd); 344 } 345 return NULL; 346 } 347 348 static void *nbd_client_thread(void *arg) 349 { 350 char *device = arg; 351 NBDExportInfo info = { .request_sizes = false, .name = g_strdup("") }; 352 QIOChannelSocket *sioc; 353 int fd; 354 int ret; 355 pthread_t show_parts_thread; 356 Error *local_error = NULL; 357 358 sioc = qio_channel_socket_new(); 359 if (qio_channel_socket_connect_sync(sioc, 360 saddr, 361 &local_error) < 0) { 362 error_report_err(local_error); 363 goto out; 364 } 365 366 ret = nbd_receive_negotiate(NULL, QIO_CHANNEL(sioc), 367 NULL, NULL, NULL, &info, &local_error); 368 if (ret < 0) { 369 if (local_error) { 370 error_report_err(local_error); 371 } 372 goto out_socket; 373 } 374 375 fd = open(device, O_RDWR); 376 if (fd < 0) { 377 /* Linux-only, we can use %m in printf. */ 378 error_report("Failed to open %s: %m", device); 379 goto out_socket; 380 } 381 382 ret = nbd_init(fd, sioc, &info, &local_error); 383 if (ret < 0) { 384 error_report_err(local_error); 385 goto out_fd; 386 } 387 388 /* update partition table */ 389 pthread_create(&show_parts_thread, NULL, show_parts, device); 390 391 if (verbose) { 392 fprintf(stderr, "NBD device %s is now connected to %s\n", 393 device, srcpath); 394 } else { 395 /* Close stderr so that the qemu-nbd process exits. */ 396 dup2(STDOUT_FILENO, STDERR_FILENO); 397 } 398 399 ret = nbd_client(fd); 400 if (ret) { 401 goto out_fd; 402 } 403 close(fd); 404 object_unref(OBJECT(sioc)); 405 g_free(info.name); 406 kill(getpid(), SIGTERM); 407 return (void *) EXIT_SUCCESS; 408 409 out_fd: 410 close(fd); 411 out_socket: 412 object_unref(OBJECT(sioc)); 413 out: 414 g_free(info.name); 415 kill(getpid(), SIGTERM); 416 return (void *) EXIT_FAILURE; 417 } 418 #endif /* HAVE_NBD_DEVICE */ 419 420 static int nbd_can_accept(void) 421 { 422 return state == RUNNING && nb_fds < shared; 423 } 424 425 static void nbd_export_closed(NBDExport *export) 426 { 427 assert(state == TERMINATING); 428 state = TERMINATED; 429 } 430 431 static void nbd_update_server_watch(void); 432 433 static void nbd_client_closed(NBDClient *client, bool negotiated) 434 { 435 nb_fds--; 436 if (negotiated && nb_fds == 0 && !persistent && state == RUNNING) { 437 state = TERMINATE; 438 } 439 nbd_update_server_watch(); 440 nbd_client_put(client); 441 } 442 443 static void nbd_accept(QIONetListener *listener, QIOChannelSocket *cioc, 444 gpointer opaque) 445 { 446 if (state >= TERMINATE) { 447 return; 448 } 449 450 nb_fds++; 451 nbd_update_server_watch(); 452 nbd_client_new(cioc, tlscreds, tlsauthz, nbd_client_closed); 453 } 454 455 static void nbd_update_server_watch(void) 456 { 457 if (nbd_can_accept()) { 458 qio_net_listener_set_client_func(server, nbd_accept, NULL, NULL); 459 } else { 460 qio_net_listener_set_client_func(server, NULL, NULL, NULL); 461 } 462 } 463 464 465 static SocketAddress *nbd_build_socket_address(const char *sockpath, 466 const char *bindto, 467 const char *port) 468 { 469 SocketAddress *saddr; 470 471 saddr = g_new0(SocketAddress, 1); 472 if (sockpath) { 473 saddr->type = SOCKET_ADDRESS_TYPE_UNIX; 474 saddr->u.q_unix.path = g_strdup(sockpath); 475 } else { 476 InetSocketAddress *inet; 477 saddr->type = SOCKET_ADDRESS_TYPE_INET; 478 inet = &saddr->u.inet; 479 inet->host = g_strdup(bindto); 480 if (port) { 481 inet->port = g_strdup(port); 482 } else { 483 inet->port = g_strdup_printf("%d", NBD_DEFAULT_PORT); 484 } 485 } 486 487 return saddr; 488 } 489 490 491 static QemuOptsList file_opts = { 492 .name = "file", 493 .implied_opt_name = "file", 494 .head = QTAILQ_HEAD_INITIALIZER(file_opts.head), 495 .desc = { 496 /* no elements => accept any params */ 497 { /* end of list */ } 498 }, 499 }; 500 501 static QemuOptsList qemu_object_opts = { 502 .name = "object", 503 .implied_opt_name = "qom-type", 504 .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head), 505 .desc = { 506 { } 507 }, 508 }; 509 510 511 512 static QCryptoTLSCreds *nbd_get_tls_creds(const char *id, bool list, 513 Error **errp) 514 { 515 Object *obj; 516 QCryptoTLSCreds *creds; 517 518 obj = object_resolve_path_component( 519 object_get_objects_root(), id); 520 if (!obj) { 521 error_setg(errp, "No TLS credentials with id '%s'", 522 id); 523 return NULL; 524 } 525 creds = (QCryptoTLSCreds *) 526 object_dynamic_cast(obj, TYPE_QCRYPTO_TLS_CREDS); 527 if (!creds) { 528 error_setg(errp, "Object with id '%s' is not TLS credentials", 529 id); 530 return NULL; 531 } 532 533 if (list) { 534 if (creds->endpoint != QCRYPTO_TLS_CREDS_ENDPOINT_CLIENT) { 535 error_setg(errp, 536 "Expecting TLS credentials with a client endpoint"); 537 return NULL; 538 } 539 } else { 540 if (creds->endpoint != QCRYPTO_TLS_CREDS_ENDPOINT_SERVER) { 541 error_setg(errp, 542 "Expecting TLS credentials with a server endpoint"); 543 return NULL; 544 } 545 } 546 object_ref(obj); 547 return creds; 548 } 549 550 static void setup_address_and_port(const char **address, const char **port) 551 { 552 if (*address == NULL) { 553 *address = "0.0.0.0"; 554 } 555 556 if (*port == NULL) { 557 *port = stringify(NBD_DEFAULT_PORT); 558 } 559 } 560 561 /* 562 * Check socket parameters compatibility when socket activation is used. 563 */ 564 static const char *socket_activation_validate_opts(const char *device, 565 const char *sockpath, 566 const char *address, 567 const char *port, 568 bool list) 569 { 570 if (device != NULL) { 571 return "NBD device can't be set when using socket activation"; 572 } 573 574 if (sockpath != NULL) { 575 return "Unix socket can't be set when using socket activation"; 576 } 577 578 if (address != NULL) { 579 return "The interface can't be set when using socket activation"; 580 } 581 582 if (port != NULL) { 583 return "TCP port number can't be set when using socket activation"; 584 } 585 586 if (list) { 587 return "List mode is incompatible with socket activation"; 588 } 589 590 return NULL; 591 } 592 593 static void qemu_nbd_shutdown(void) 594 { 595 job_cancel_sync_all(); 596 bdrv_close_all(); 597 } 598 599 int main(int argc, char **argv) 600 { 601 BlockBackend *blk; 602 BlockDriverState *bs; 603 uint64_t dev_offset = 0; 604 bool readonly = false; 605 bool disconnect = false; 606 const char *bindto = NULL; 607 const char *port = NULL; 608 char *sockpath = NULL; 609 char *device = NULL; 610 int64_t fd_size; 611 QemuOpts *sn_opts = NULL; 612 const char *sn_id_or_name = NULL; 613 const char *sopt = "hVb:o:p:rsnP:c:dvk:e:f:tl:x:T:D:B:L"; 614 struct option lopt[] = { 615 { "help", no_argument, NULL, 'h' }, 616 { "version", no_argument, NULL, 'V' }, 617 { "bind", required_argument, NULL, 'b' }, 618 { "port", required_argument, NULL, 'p' }, 619 { "socket", required_argument, NULL, 'k' }, 620 { "offset", required_argument, NULL, 'o' }, 621 { "read-only", no_argument, NULL, 'r' }, 622 { "partition", required_argument, NULL, 'P' }, 623 { "bitmap", required_argument, NULL, 'B' }, 624 { "connect", required_argument, NULL, 'c' }, 625 { "disconnect", no_argument, NULL, 'd' }, 626 { "list", no_argument, NULL, 'L' }, 627 { "snapshot", no_argument, NULL, 's' }, 628 { "load-snapshot", required_argument, NULL, 'l' }, 629 { "nocache", no_argument, NULL, 'n' }, 630 { "cache", required_argument, NULL, QEMU_NBD_OPT_CACHE }, 631 { "aio", required_argument, NULL, QEMU_NBD_OPT_AIO }, 632 { "discard", required_argument, NULL, QEMU_NBD_OPT_DISCARD }, 633 { "detect-zeroes", required_argument, NULL, 634 QEMU_NBD_OPT_DETECT_ZEROES }, 635 { "shared", required_argument, NULL, 'e' }, 636 { "format", required_argument, NULL, 'f' }, 637 { "persistent", no_argument, NULL, 't' }, 638 { "verbose", no_argument, NULL, 'v' }, 639 { "object", required_argument, NULL, QEMU_NBD_OPT_OBJECT }, 640 { "export-name", required_argument, NULL, 'x' }, 641 { "description", required_argument, NULL, 'D' }, 642 { "tls-creds", required_argument, NULL, QEMU_NBD_OPT_TLSCREDS }, 643 { "tls-authz", required_argument, NULL, QEMU_NBD_OPT_TLSAUTHZ }, 644 { "image-opts", no_argument, NULL, QEMU_NBD_OPT_IMAGE_OPTS }, 645 { "trace", required_argument, NULL, 'T' }, 646 { "fork", no_argument, NULL, QEMU_NBD_OPT_FORK }, 647 { "pid-file", required_argument, NULL, QEMU_NBD_OPT_PID_FILE }, 648 { NULL, 0, NULL, 0 } 649 }; 650 int ch; 651 int opt_ind = 0; 652 int flags = BDRV_O_RDWR; 653 int partition = 0; 654 int ret = 0; 655 bool seen_cache = false; 656 bool seen_discard = false; 657 bool seen_aio = false; 658 pthread_t client_thread; 659 const char *fmt = NULL; 660 Error *local_err = NULL; 661 BlockdevDetectZeroesOptions detect_zeroes = BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF; 662 QDict *options = NULL; 663 const char *export_name = NULL; /* defaults to "" later for server mode */ 664 const char *export_description = NULL; 665 const char *bitmap = NULL; 666 const char *tlscredsid = NULL; 667 bool imageOpts = false; 668 bool writethrough = true; 669 char *trace_file = NULL; 670 bool fork_process = false; 671 bool list = false; 672 int old_stderr = -1; 673 unsigned socket_activation; 674 const char *pid_file_name = NULL; 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 error_init(argv[0]); 689 module_call_init(MODULE_INIT_TRACE); 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 readonly = true; 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_TLSAUTHZ: 866 tlsauthz = optarg; 867 break; 868 case QEMU_NBD_OPT_FORK: 869 fork_process = true; 870 break; 871 case 'L': 872 list = true; 873 break; 874 case QEMU_NBD_OPT_PID_FILE: 875 pid_file_name = optarg; 876 break; 877 } 878 } 879 880 if (list) { 881 if (argc != optind) { 882 error_report("List mode is incompatible with a file name"); 883 exit(EXIT_FAILURE); 884 } 885 if (export_name || export_description || dev_offset || partition || 886 device || disconnect || fmt || sn_id_or_name || bitmap || 887 seen_aio || seen_discard || seen_cache) { 888 error_report("List mode is incompatible with per-device settings"); 889 exit(EXIT_FAILURE); 890 } 891 if (fork_process) { 892 error_report("List mode is incompatible with forking"); 893 exit(EXIT_FAILURE); 894 } 895 } else if ((argc - optind) != 1) { 896 error_report("Invalid number of arguments"); 897 error_printf("Try `%s --help' for more information.\n", argv[0]); 898 exit(EXIT_FAILURE); 899 } else if (!export_name) { 900 export_name = ""; 901 } 902 903 qemu_opts_foreach(&qemu_object_opts, 904 user_creatable_add_opts_foreach, 905 NULL, &error_fatal); 906 907 if (!trace_init_backends()) { 908 exit(1); 909 } 910 trace_init_file(trace_file); 911 qemu_set_log(LOG_TRACE); 912 913 socket_activation = check_socket_activation(); 914 if (socket_activation == 0) { 915 setup_address_and_port(&bindto, &port); 916 } else { 917 /* Using socket activation - check user didn't use -p etc. */ 918 const char *err_msg = socket_activation_validate_opts(device, sockpath, 919 bindto, port, 920 list); 921 if (err_msg != NULL) { 922 error_report("%s", err_msg); 923 exit(EXIT_FAILURE); 924 } 925 926 /* qemu-nbd can only listen on a single socket. */ 927 if (socket_activation > 1) { 928 error_report("qemu-nbd does not support socket activation with %s > 1", 929 "LISTEN_FDS"); 930 exit(EXIT_FAILURE); 931 } 932 } 933 934 if (tlscredsid) { 935 if (sockpath) { 936 error_report("TLS is only supported with IPv4/IPv6"); 937 exit(EXIT_FAILURE); 938 } 939 if (device) { 940 error_report("TLS is not supported with a host device"); 941 exit(EXIT_FAILURE); 942 } 943 if (tlsauthz && list) { 944 error_report("TLS authorization is incompatible with export list"); 945 exit(EXIT_FAILURE); 946 } 947 tlscreds = nbd_get_tls_creds(tlscredsid, list, &local_err); 948 if (local_err) { 949 error_report("Failed to get TLS creds %s", 950 error_get_pretty(local_err)); 951 exit(EXIT_FAILURE); 952 } 953 } else { 954 if (tlsauthz) { 955 error_report("--tls-authz is not permitted without --tls-creds"); 956 exit(EXIT_FAILURE); 957 } 958 } 959 960 if (list) { 961 saddr = nbd_build_socket_address(sockpath, bindto, port); 962 return qemu_nbd_client_list(saddr, tlscreds, bindto); 963 } 964 965 #if !HAVE_NBD_DEVICE 966 if (disconnect || device) { 967 error_report("Kernel /dev/nbdN support not available"); 968 exit(EXIT_FAILURE); 969 } 970 #else /* HAVE_NBD_DEVICE */ 971 if (disconnect) { 972 int nbdfd = open(argv[optind], O_RDWR); 973 if (nbdfd < 0) { 974 error_report("Cannot open %s: %s", argv[optind], 975 strerror(errno)); 976 exit(EXIT_FAILURE); 977 } 978 nbd_disconnect(nbdfd); 979 980 close(nbdfd); 981 982 printf("%s disconnected\n", argv[optind]); 983 984 return 0; 985 } 986 #endif 987 988 if ((device && !verbose) || fork_process) { 989 int stderr_fd[2]; 990 pid_t pid; 991 int ret; 992 993 if (qemu_pipe(stderr_fd) < 0) { 994 error_report("Error setting up communication pipe: %s", 995 strerror(errno)); 996 exit(EXIT_FAILURE); 997 } 998 999 /* Now daemonize, but keep a communication channel open to 1000 * print errors and exit with the proper status code. 1001 */ 1002 pid = fork(); 1003 if (pid < 0) { 1004 error_report("Failed to fork: %s", strerror(errno)); 1005 exit(EXIT_FAILURE); 1006 } else if (pid == 0) { 1007 close(stderr_fd[0]); 1008 1009 old_stderr = dup(STDERR_FILENO); 1010 ret = qemu_daemon(1, 0); 1011 1012 /* Temporarily redirect stderr to the parent's pipe... */ 1013 dup2(stderr_fd[1], STDERR_FILENO); 1014 if (ret < 0) { 1015 error_report("Failed to daemonize: %s", strerror(errno)); 1016 exit(EXIT_FAILURE); 1017 } 1018 1019 /* ... close the descriptor we inherited and go on. */ 1020 close(stderr_fd[1]); 1021 } else { 1022 bool errors = false; 1023 char *buf; 1024 1025 /* In the parent. Print error messages from the child until 1026 * it closes the pipe. 1027 */ 1028 close(stderr_fd[1]); 1029 buf = g_malloc(1024); 1030 while ((ret = read(stderr_fd[0], buf, 1024)) > 0) { 1031 errors = true; 1032 ret = qemu_write_full(STDERR_FILENO, buf, ret); 1033 if (ret < 0) { 1034 exit(EXIT_FAILURE); 1035 } 1036 } 1037 if (ret < 0) { 1038 error_report("Cannot read from daemon: %s", 1039 strerror(errno)); 1040 exit(EXIT_FAILURE); 1041 } 1042 1043 /* Usually the daemon should not print any message. 1044 * Exit with zero status in that case. 1045 */ 1046 exit(errors); 1047 } 1048 } 1049 1050 if (device != NULL && sockpath == NULL) { 1051 sockpath = g_malloc(128); 1052 snprintf(sockpath, 128, SOCKET_PATH, basename(device)); 1053 } 1054 1055 server = qio_net_listener_new(); 1056 if (socket_activation == 0) { 1057 saddr = nbd_build_socket_address(sockpath, bindto, port); 1058 if (qio_net_listener_open_sync(server, saddr, 1, &local_err) < 0) { 1059 object_unref(OBJECT(server)); 1060 error_report_err(local_err); 1061 exit(EXIT_FAILURE); 1062 } 1063 } else { 1064 size_t i; 1065 /* See comment in check_socket_activation above. */ 1066 for (i = 0; i < socket_activation; i++) { 1067 QIOChannelSocket *sioc; 1068 sioc = qio_channel_socket_new_fd(FIRST_SOCKET_ACTIVATION_FD + i, 1069 &local_err); 1070 if (sioc == NULL) { 1071 object_unref(OBJECT(server)); 1072 error_report("Failed to use socket activation: %s", 1073 error_get_pretty(local_err)); 1074 exit(EXIT_FAILURE); 1075 } 1076 qio_net_listener_add(server, sioc); 1077 object_unref(OBJECT(sioc)); 1078 } 1079 } 1080 1081 if (qemu_init_main_loop(&local_err)) { 1082 error_report_err(local_err); 1083 exit(EXIT_FAILURE); 1084 } 1085 bdrv_init(); 1086 atexit(qemu_nbd_shutdown); 1087 1088 srcpath = argv[optind]; 1089 if (imageOpts) { 1090 QemuOpts *opts; 1091 if (fmt) { 1092 error_report("--image-opts and -f are mutually exclusive"); 1093 exit(EXIT_FAILURE); 1094 } 1095 opts = qemu_opts_parse_noisily(&file_opts, srcpath, true); 1096 if (!opts) { 1097 qemu_opts_reset(&file_opts); 1098 exit(EXIT_FAILURE); 1099 } 1100 options = qemu_opts_to_qdict(opts, NULL); 1101 qemu_opts_reset(&file_opts); 1102 blk = blk_new_open(NULL, NULL, options, flags, &local_err); 1103 } else { 1104 if (fmt) { 1105 options = qdict_new(); 1106 qdict_put_str(options, "driver", fmt); 1107 } 1108 blk = blk_new_open(srcpath, NULL, options, flags, &local_err); 1109 } 1110 1111 if (!blk) { 1112 error_reportf_err(local_err, "Failed to blk_new_open '%s': ", 1113 argv[optind]); 1114 exit(EXIT_FAILURE); 1115 } 1116 bs = blk_bs(blk); 1117 1118 blk_set_enable_write_cache(blk, !writethrough); 1119 1120 if (sn_opts) { 1121 ret = bdrv_snapshot_load_tmp(bs, 1122 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID), 1123 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME), 1124 &local_err); 1125 } else if (sn_id_or_name) { 1126 ret = bdrv_snapshot_load_tmp_by_id_or_name(bs, sn_id_or_name, 1127 &local_err); 1128 } 1129 if (ret < 0) { 1130 error_reportf_err(local_err, "Failed to load snapshot: "); 1131 exit(EXIT_FAILURE); 1132 } 1133 1134 bs->detect_zeroes = detect_zeroes; 1135 fd_size = blk_getlength(blk); 1136 if (fd_size < 0) { 1137 error_report("Failed to determine the image length: %s", 1138 strerror(-fd_size)); 1139 exit(EXIT_FAILURE); 1140 } 1141 1142 if (dev_offset >= fd_size) { 1143 error_report("Offset (%" PRIu64 ") has to be smaller than the image " 1144 "size (%" PRId64 ")", dev_offset, fd_size); 1145 exit(EXIT_FAILURE); 1146 } 1147 fd_size -= dev_offset; 1148 1149 if (partition) { 1150 uint64_t limit; 1151 1152 if (dev_offset) { 1153 error_report("Cannot request partition and offset together"); 1154 exit(EXIT_FAILURE); 1155 } 1156 ret = find_partition(blk, partition, &dev_offset, &limit); 1157 if (ret < 0) { 1158 error_report("Could not find partition %d: %s", partition, 1159 strerror(-ret)); 1160 exit(EXIT_FAILURE); 1161 } 1162 /* 1163 * MBR partition limits are (32-bit << 9); this assert lets 1164 * the compiler know that we can't overflow 64 bits. 1165 */ 1166 assert(dev_offset + limit >= dev_offset); 1167 if (dev_offset + limit > fd_size) { 1168 error_report("Discovered partition %d at offset %" PRIu64 1169 " size %" PRIu64 ", but size exceeds file length %" 1170 PRId64, partition, dev_offset, limit, fd_size); 1171 exit(EXIT_FAILURE); 1172 } 1173 fd_size = limit; 1174 } 1175 1176 export = nbd_export_new(bs, dev_offset, fd_size, export_name, 1177 export_description, bitmap, readonly, shared > 1, 1178 nbd_export_closed, writethrough, NULL, 1179 &error_fatal); 1180 1181 if (device) { 1182 #if HAVE_NBD_DEVICE 1183 int ret; 1184 1185 ret = pthread_create(&client_thread, NULL, nbd_client_thread, device); 1186 if (ret != 0) { 1187 error_report("Failed to create client thread: %s", strerror(ret)); 1188 exit(EXIT_FAILURE); 1189 } 1190 #endif 1191 } else { 1192 /* Shut up GCC warnings. */ 1193 memset(&client_thread, 0, sizeof(client_thread)); 1194 } 1195 1196 nbd_update_server_watch(); 1197 1198 if (pid_file_name) { 1199 qemu_write_pidfile(pid_file_name, &error_fatal); 1200 } 1201 1202 /* now when the initialization is (almost) complete, chdir("/") 1203 * to free any busy filesystems */ 1204 if (chdir("/") < 0) { 1205 error_report("Could not chdir to root directory: %s", 1206 strerror(errno)); 1207 exit(EXIT_FAILURE); 1208 } 1209 1210 if (fork_process) { 1211 dup2(old_stderr, STDERR_FILENO); 1212 close(old_stderr); 1213 } 1214 1215 state = RUNNING; 1216 do { 1217 main_loop_wait(false); 1218 if (state == TERMINATE) { 1219 state = TERMINATING; 1220 nbd_export_close(export); 1221 nbd_export_put(export); 1222 export = NULL; 1223 } 1224 } while (state != TERMINATED); 1225 1226 blk_unref(blk); 1227 if (sockpath) { 1228 unlink(sockpath); 1229 } 1230 1231 qemu_opts_del(sn_opts); 1232 1233 if (device) { 1234 void *ret; 1235 pthread_join(client_thread, &ret); 1236 exit(ret != NULL); 1237 } else { 1238 exit(EXIT_SUCCESS); 1239 } 1240 } 1241