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.
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.
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/>.
19 #include "qemu/osdep.h"
20 #include "qapi/error.h"
21 #include "qemu-common.h"
22 #include "qemu/cutils.h"
23 #include "sysemu/block-backend.h"
24 #include "block/block_int.h"
25 #include "block/nbd.h"
26 #include "qemu/main-loop.h"
27 #include "qemu/error-report.h"
28 #include "qemu/config-file.h"
29 #include "block/snapshot.h"
30 #include "qapi/util.h"
31 #include "qapi/qmp/qstring.h"
32 #include "qom/object_interfaces.h"
33 #include "io/channel-socket.h"
39 #define SOCKET_PATH "/var/lock/qemu-nbd-%s"
40 #define QEMU_NBD_OPT_CACHE 256
41 #define QEMU_NBD_OPT_AIO 257
42 #define QEMU_NBD_OPT_DISCARD 258
43 #define QEMU_NBD_OPT_DETECT_ZEROES 259
44 #define QEMU_NBD_OPT_OBJECT 260
45 #define QEMU_NBD_OPT_TLSCREDS 261
46 #define QEMU_NBD_OPT_IMAGE_OPTS 262
48 static NBDExport *exp;
52 static SocketAddress *saddr;
53 static int persistent = 0;
54 static enum { RUNNING, TERMINATE, TERMINATING, TERMINATED } state;
55 static int shared = 1;
57 static QIOChannelSocket *server_ioc;
58 static int server_watch = -1;
59 static QCryptoTLSCreds *tlscreds;
61 static void usage(const char *name)
64 "Usage: %s [OPTIONS] FILE\n"
65 "QEMU Disk Network Block Device Server\n"
67 " -h, --help display this help and exit\n"
68 " -V, --version output version information and exit\n"
70 "Connection properties:\n"
71 " -p, --port=PORT port to listen on (default `%d')\n"
72 " -b, --bind=IFACE interface to bind to (default `0.0.0.0')\n"
73 " -k, --socket=PATH path to the unix socket\n"
74 " (default '"SOCKET_PATH"')\n"
75 " -e, --shared=NUM device can be shared by NUM clients (default '1')\n"
76 " -t, --persistent don't exit on the last connection\n"
77 " -v, --verbose display extra debugging information\n"
78 " -x, --export-name=NAME expose export by name\n"
80 "Exposing part of the image:\n"
81 " -o, --offset=OFFSET offset into the image\n"
82 " -P, --partition=NUM only expose partition NUM\n"
84 "General purpose options:\n"
85 " --object type,id=ID,... define an object such as 'secret' for providing\n"
86 " passwords and/or encryption keys\n"
88 "Kernel NBD client support:\n"
89 " -c, --connect=DEV connect FILE to the local NBD device DEV\n"
90 " -d, --disconnect disconnect the specified device\n"
94 "Block device options:\n"
95 " -f, --format=FORMAT set image format (raw, qcow2, ...)\n"
96 " -r, --read-only export read-only\n"
97 " -s, --snapshot use FILE as an external snapshot, create a temporary\n"
98 " file with backing_file=FILE, redirect the write to\n"
99 " the temporary one\n"
100 " -l, --load-snapshot=SNAPSHOT_PARAM\n"
101 " load an internal snapshot inside FILE and export it\n"
102 " as an read-only device, SNAPSHOT_PARAM format is\n"
103 " 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
105 " -n, --nocache disable host cache\n"
106 " --cache=MODE set cache mode (none, writeback, ...)\n"
107 " --aio=MODE set AIO mode (native or threads)\n"
108 " --discard=MODE set discard mode (ignore, unmap)\n"
109 " --detect-zeroes=MODE set detect-zeroes mode (off, on, unmap)\n"
110 " --image-opts treat FILE as a full set of image options\n"
113 , name, NBD_DEFAULT_PORT, "DEVICE");
116 static void version(const char *name)
120 "Written by Anthony Liguori.\n"
123 "This is free software; see the source for copying conditions. There is NO\n"
124 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
128 struct partition_record
132 uint32_t start_cylinder;
133 uint8_t start_sector;
136 uint8_t end_cylinder;
138 uint32_t start_sector_abs;
139 uint32_t nb_sectors_abs;
142 static void read_partition(uint8_t *p, struct partition_record *r)
145 r->start_head = p[1];
146 r->start_cylinder = p[3] | ((p[2] << 2) & 0x0300);
147 r->start_sector = p[2] & 0x3f;
150 r->end_cylinder = p[7] | ((p[6] << 2) & 0x300);
151 r->end_sector = p[6] & 0x3f;
153 r->start_sector_abs = le32_to_cpup((uint32_t *)(p + 8));
154 r->nb_sectors_abs = le32_to_cpup((uint32_t *)(p + 12));
157 static int find_partition(BlockBackend *blk, int partition,
158 off_t *offset, off_t *size)
160 struct partition_record mbr[4];
166 if ((ret = blk_read(blk, 0, data, 1)) < 0) {
167 error_report("error while reading: %s", strerror(-ret));
171 if (data[510] != 0x55 || data[511] != 0xaa) {
175 for (i = 0; i < 4; i++) {
176 read_partition(&data[446 + 16 * i], &mbr[i]);
178 if (!mbr[i].system || !mbr[i].nb_sectors_abs) {
182 if (mbr[i].system == 0xF || mbr[i].system == 0x5) {
183 struct partition_record ext[4];
187 if ((ret = blk_read(blk, mbr[i].start_sector_abs, data1, 1)) < 0) {
188 error_report("error while reading: %s", strerror(-ret));
192 for (j = 0; j < 4; j++) {
193 read_partition(&data1[446 + 16 * j], &ext[j]);
194 if (!ext[j].system || !ext[j].nb_sectors_abs) {
198 if ((ext_partnum + j + 1) == partition) {
199 *offset = (uint64_t)ext[j].start_sector_abs << 9;
200 *size = (uint64_t)ext[j].nb_sectors_abs << 9;
205 } else if ((i + 1) == partition) {
206 *offset = (uint64_t)mbr[i].start_sector_abs << 9;
207 *size = (uint64_t)mbr[i].nb_sectors_abs << 9;
215 static void termsig_handler(int signum)
222 static void *show_parts(void *arg)
227 /* linux just needs an open() to trigger
228 * the partition table update
229 * but remember to load the module with max_part != 0 :
230 * modprobe nbd max_part=63
232 nbd = open(device, O_RDWR);
239 static void *nbd_client_thread(void *arg)
244 QIOChannelSocket *sioc;
247 pthread_t show_parts_thread;
248 Error *local_error = NULL;
250 sioc = qio_channel_socket_new();
251 if (qio_channel_socket_connect_sync(sioc,
254 error_report_err(local_error);
258 ret = nbd_receive_negotiate(QIO_CHANNEL(sioc), NULL, &nbdflags,
260 &size, &local_error);
263 error_report_err(local_error);
268 fd = open(device, O_RDWR);
270 /* Linux-only, we can use %m in printf. */
271 error_report("Failed to open %s: %m", device);
275 ret = nbd_init(fd, sioc, nbdflags, size);
280 /* update partition table */
281 pthread_create(&show_parts_thread, NULL, show_parts, device);
284 fprintf(stderr, "NBD device %s is now connected to %s\n",
287 /* Close stderr so that the qemu-nbd process exits. */
288 dup2(STDOUT_FILENO, STDERR_FILENO);
291 ret = nbd_client(fd);
296 object_unref(OBJECT(sioc));
297 kill(getpid(), SIGTERM);
298 return (void *) EXIT_SUCCESS;
303 object_unref(OBJECT(sioc));
305 kill(getpid(), SIGTERM);
306 return (void *) EXIT_FAILURE;
309 static int nbd_can_accept(void)
311 return nb_fds < shared;
314 static void nbd_export_closed(NBDExport *exp)
316 assert(state == TERMINATING);
320 static void nbd_update_server_watch(void);
322 static void nbd_client_closed(NBDClient *client)
325 if (nb_fds == 0 && !persistent && state == RUNNING) {
328 nbd_update_server_watch();
329 nbd_client_put(client);
332 static gboolean nbd_accept(QIOChannel *ioc, GIOCondition cond, gpointer opaque)
334 QIOChannelSocket *cioc;
336 cioc = qio_channel_socket_accept(QIO_CHANNEL_SOCKET(ioc),
342 if (state >= TERMINATE) {
343 object_unref(OBJECT(cioc));
348 nbd_update_server_watch();
349 nbd_client_new(newproto ? NULL : exp, cioc,
350 tlscreds, NULL, nbd_client_closed);
351 object_unref(OBJECT(cioc));
356 static void nbd_update_server_watch(void)
358 if (nbd_can_accept()) {
359 if (server_watch == -1) {
360 server_watch = qio_channel_add_watch(QIO_CHANNEL(server_ioc),
366 if (server_watch != -1) {
367 g_source_remove(server_watch);
374 static SocketAddress *nbd_build_socket_address(const char *sockpath,
378 SocketAddress *saddr;
380 saddr = g_new0(SocketAddress, 1);
382 saddr->type = SOCKET_ADDRESS_KIND_UNIX;
383 saddr->u.q_unix.data = g_new0(UnixSocketAddress, 1);
384 saddr->u.q_unix.data->path = g_strdup(sockpath);
386 InetSocketAddress *inet;
387 saddr->type = SOCKET_ADDRESS_KIND_INET;
388 inet = saddr->u.inet.data = g_new0(InetSocketAddress, 1);
389 inet->host = g_strdup(bindto);
391 inet->port = g_strdup(port);
393 inet->port = g_strdup_printf("%d", NBD_DEFAULT_PORT);
401 static QemuOptsList file_opts = {
403 .implied_opt_name = "file",
404 .head = QTAILQ_HEAD_INITIALIZER(file_opts.head),
406 /* no elements => accept any params */
407 { /* end of list */ }
411 static QemuOptsList qemu_object_opts = {
413 .implied_opt_name = "qom-type",
414 .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head),
422 static QCryptoTLSCreds *nbd_get_tls_creds(const char *id, Error **errp)
425 QCryptoTLSCreds *creds;
427 obj = object_resolve_path_component(
428 object_get_objects_root(), id);
430 error_setg(errp, "No TLS credentials with id '%s'",
434 creds = (QCryptoTLSCreds *)
435 object_dynamic_cast(obj, TYPE_QCRYPTO_TLS_CREDS);
437 error_setg(errp, "Object with id '%s' is not TLS credentials",
442 if (creds->endpoint != QCRYPTO_TLS_CREDS_ENDPOINT_SERVER) {
444 "Expecting TLS credentials with a server endpoint");
452 int main(int argc, char **argv)
455 BlockDriverState *bs;
456 off_t dev_offset = 0;
457 uint32_t nbdflags = 0;
458 bool disconnect = false;
459 const char *bindto = "0.0.0.0";
460 const char *port = NULL;
461 char *sockpath = NULL;
464 QemuOpts *sn_opts = NULL;
465 const char *sn_id_or_name = NULL;
466 const char *sopt = "hVb:o:p:rsnP:c:dvk:e:f:tl:x:";
467 struct option lopt[] = {
468 { "help", no_argument, NULL, 'h' },
469 { "version", no_argument, NULL, 'V' },
470 { "bind", required_argument, NULL, 'b' },
471 { "port", required_argument, NULL, 'p' },
472 { "socket", required_argument, NULL, 'k' },
473 { "offset", required_argument, NULL, 'o' },
474 { "read-only", no_argument, NULL, 'r' },
475 { "partition", required_argument, NULL, 'P' },
476 { "connect", required_argument, NULL, 'c' },
477 { "disconnect", no_argument, NULL, 'd' },
478 { "snapshot", no_argument, NULL, 's' },
479 { "load-snapshot", required_argument, NULL, 'l' },
480 { "nocache", no_argument, NULL, 'n' },
481 { "cache", required_argument, NULL, QEMU_NBD_OPT_CACHE },
482 { "aio", required_argument, NULL, QEMU_NBD_OPT_AIO },
483 { "discard", required_argument, NULL, QEMU_NBD_OPT_DISCARD },
484 { "detect-zeroes", required_argument, NULL,
485 QEMU_NBD_OPT_DETECT_ZEROES },
486 { "shared", required_argument, NULL, 'e' },
487 { "format", required_argument, NULL, 'f' },
488 { "persistent", no_argument, NULL, 't' },
489 { "verbose", no_argument, NULL, 'v' },
490 { "object", required_argument, NULL, QEMU_NBD_OPT_OBJECT },
491 { "export-name", required_argument, NULL, 'x' },
492 { "tls-creds", required_argument, NULL, QEMU_NBD_OPT_TLSCREDS },
493 { "image-opts", no_argument, NULL, QEMU_NBD_OPT_IMAGE_OPTS },
499 int flags = BDRV_O_RDWR;
502 bool seen_cache = false;
503 bool seen_discard = false;
504 bool seen_aio = false;
505 pthread_t client_thread;
506 const char *fmt = NULL;
507 Error *local_err = NULL;
508 BlockdevDetectZeroesOptions detect_zeroes = BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
509 QDict *options = NULL;
510 const char *export_name = NULL;
511 const char *tlscredsid = NULL;
512 bool imageOpts = false;
513 bool writethrough = true;
515 /* The client thread uses SIGTERM to interrupt the server. A signal
516 * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
518 struct sigaction sa_sigterm;
519 memset(&sa_sigterm, 0, sizeof(sa_sigterm));
520 sa_sigterm.sa_handler = termsig_handler;
521 sigaction(SIGTERM, &sa_sigterm, NULL);
522 module_call_init(MODULE_INIT_QOM);
523 qemu_add_opts(&qemu_object_opts);
524 qemu_init_exec_dir(argv[0]);
526 while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
529 flags |= BDRV_O_SNAPSHOT;
532 optarg = (char *) "none";
534 case QEMU_NBD_OPT_CACHE:
536 error_report("-n and --cache can only be specified once");
540 if (bdrv_parse_cache_mode(optarg, &flags, &writethrough) == -1) {
541 error_report("Invalid cache mode `%s'", optarg);
545 case QEMU_NBD_OPT_AIO:
547 error_report("--aio can only be specified once");
551 if (!strcmp(optarg, "native")) {
552 flags |= BDRV_O_NATIVE_AIO;
553 } else if (!strcmp(optarg, "threads")) {
554 /* this is the default */
556 error_report("invalid aio mode `%s'", optarg);
560 case QEMU_NBD_OPT_DISCARD:
562 error_report("--discard can only be specified once");
566 if (bdrv_parse_discard_flags(optarg, &flags) == -1) {
567 error_report("Invalid discard mode `%s'", optarg);
571 case QEMU_NBD_OPT_DETECT_ZEROES:
573 qapi_enum_parse(BlockdevDetectZeroesOptions_lookup,
575 BLOCKDEV_DETECT_ZEROES_OPTIONS__MAX,
576 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
579 error_reportf_err(local_err,
580 "Failed to parse detect_zeroes mode: ");
583 if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
584 !(flags & BDRV_O_UNMAP)) {
585 error_report("setting detect-zeroes to unmap is not allowed "
586 "without setting discard operation to unmap");
597 dev_offset = strtoll (optarg, &end, 0);
599 error_report("Invalid offset `%s'", optarg);
602 if (dev_offset < 0) {
603 error_report("Offset must be positive `%s'", optarg);
608 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
609 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
612 error_report("Failed in parsing snapshot param `%s'",
617 sn_id_or_name = optarg;
621 nbdflags |= NBD_FLAG_READ_ONLY;
622 flags &= ~BDRV_O_RDWR;
625 partition = strtol(optarg, &end, 0);
627 error_report("Invalid partition `%s'", optarg);
630 if (partition < 1 || partition > 8) {
631 error_report("Invalid partition %d", partition);
637 if (sockpath[0] != '/') {
638 error_report("socket path must be absolute");
649 shared = strtol(optarg, &end, 0);
651 error_report("Invalid shared device number '%s'", optarg);
655 error_report("Shared device number must be greater than 0");
666 export_name = optarg;
680 error_report("Try `%s --help' for more information.", argv[0]);
682 case QEMU_NBD_OPT_OBJECT: {
684 opts = qemu_opts_parse_noisily(&qemu_object_opts,
690 case QEMU_NBD_OPT_TLSCREDS:
693 case QEMU_NBD_OPT_IMAGE_OPTS:
699 if ((argc - optind) != 1) {
700 error_report("Invalid number of arguments");
701 error_printf("Try `%s --help' for more information.\n", argv[0]);
705 if (qemu_opts_foreach(&qemu_object_opts,
706 user_creatable_add_opts_foreach,
708 error_report_err(local_err);
714 error_report("TLS is only supported with IPv4/IPv6");
718 error_report("TLS is not supported with a host device");
722 /* Set the default NBD protocol export name, since
723 * we *must* use new style protocol for TLS */
726 tlscreds = nbd_get_tls_creds(tlscredsid, &local_err);
728 error_report("Failed to get TLS creds %s",
729 error_get_pretty(local_err));
735 int nbdfd = open(argv[optind], O_RDWR);
737 error_report("Cannot open %s: %s", argv[optind],
741 nbd_disconnect(nbdfd);
745 printf("%s disconnected\n", argv[optind]);
750 if (device && !verbose) {
755 if (qemu_pipe(stderr_fd) < 0) {
756 error_report("Error setting up communication pipe: %s",
761 /* Now daemonize, but keep a communication channel open to
762 * print errors and exit with the proper status code.
766 error_report("Failed to fork: %s", strerror(errno));
768 } else if (pid == 0) {
770 ret = qemu_daemon(1, 0);
772 /* Temporarily redirect stderr to the parent's pipe... */
773 dup2(stderr_fd[1], STDERR_FILENO);
775 error_report("Failed to daemonize: %s", strerror(errno));
779 /* ... close the descriptor we inherited and go on. */
785 /* In the parent. Print error messages from the child until
786 * it closes the pipe.
789 buf = g_malloc(1024);
790 while ((ret = read(stderr_fd[0], buf, 1024)) > 0) {
792 ret = qemu_write_full(STDERR_FILENO, buf, ret);
798 error_report("Cannot read from daemon: %s",
803 /* Usually the daemon should not print any message.
804 * Exit with zero status in that case.
810 if (device != NULL && sockpath == NULL) {
811 sockpath = g_malloc(128);
812 snprintf(sockpath, 128, SOCKET_PATH, basename(device));
815 saddr = nbd_build_socket_address(sockpath, bindto, port);
817 if (qemu_init_main_loop(&local_err)) {
818 error_report_err(local_err);
822 atexit(bdrv_close_all);
824 srcpath = argv[optind];
828 error_report("--image-opts and -f are mutually exclusive");
831 opts = qemu_opts_parse_noisily(&file_opts, srcpath, true);
833 qemu_opts_reset(&file_opts);
836 options = qemu_opts_to_qdict(opts, NULL);
837 qemu_opts_reset(&file_opts);
838 blk = blk_new_open(NULL, NULL, options, flags, &local_err);
841 options = qdict_new();
842 qdict_put(options, "driver", qstring_from_str(fmt));
844 blk = blk_new_open(srcpath, NULL, options, flags, &local_err);
848 error_reportf_err(local_err, "Failed to blk_new_open '%s': ",
854 blk_set_enable_write_cache(blk, !writethrough);
857 ret = bdrv_snapshot_load_tmp(bs,
858 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
859 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
861 } else if (sn_id_or_name) {
862 ret = bdrv_snapshot_load_tmp_by_id_or_name(bs, sn_id_or_name,
866 error_reportf_err(local_err, "Failed to load snapshot: ");
870 bs->detect_zeroes = detect_zeroes;
871 fd_size = blk_getlength(blk);
873 error_report("Failed to determine the image length: %s",
878 if (partition != -1) {
879 ret = find_partition(blk, partition, &dev_offset, &fd_size);
881 error_report("Could not find partition %d: %s", partition,
887 exp = nbd_export_new(blk, dev_offset, fd_size, nbdflags, nbd_export_closed,
890 error_report_err(local_err);
894 nbd_export_set_name(exp, export_name);
898 server_ioc = qio_channel_socket_new();
899 if (qio_channel_socket_listen_sync(server_ioc, saddr, &local_err) < 0) {
900 object_unref(OBJECT(server_ioc));
901 error_report_err(local_err);
908 ret = pthread_create(&client_thread, NULL, nbd_client_thread, device);
910 error_report("Failed to create client thread: %s", strerror(ret));
914 /* Shut up GCC warnings. */
915 memset(&client_thread, 0, sizeof(client_thread));
918 nbd_update_server_watch();
920 /* now when the initialization is (almost) complete, chdir("/")
921 * to free any busy filesystems */
922 if (chdir("/") < 0) {
923 error_report("Could not chdir to root directory: %s",
930 main_loop_wait(false);
931 if (state == TERMINATE) {
933 nbd_export_close(exp);
937 } while (state != TERMINATED);
944 qemu_opts_del(sn_opts);
948 pthread_join(client_thread, &ret);