2 * Copyright (C) 2005 Anthony Liguori <anthony@codemonkey.ws>
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"
34 #include "crypto/init.h"
40 #define SOCKET_PATH "/var/lock/qemu-nbd-%s"
41 #define QEMU_NBD_OPT_CACHE 256
42 #define QEMU_NBD_OPT_AIO 257
43 #define QEMU_NBD_OPT_DISCARD 258
44 #define QEMU_NBD_OPT_DETECT_ZEROES 259
45 #define QEMU_NBD_OPT_OBJECT 260
46 #define QEMU_NBD_OPT_TLSCREDS 261
47 #define QEMU_NBD_OPT_IMAGE_OPTS 262
51 static NBDExport *exp;
55 static SocketAddress *saddr;
56 static int persistent = 0;
57 static enum { RUNNING, TERMINATE, TERMINATING, TERMINATED } state;
58 static int shared = 1;
60 static QIOChannelSocket *server_ioc;
61 static int server_watch = -1;
62 static QCryptoTLSCreds *tlscreds;
64 static void usage(const char *name)
67 "Usage: %s [OPTIONS] FILE\n"
68 "QEMU Disk Network Block Device Server\n"
70 " -h, --help display this help and exit\n"
71 " -V, --version output version information and exit\n"
73 "Connection properties:\n"
74 " -p, --port=PORT port to listen on (default `%d')\n"
75 " -b, --bind=IFACE interface to bind to (default `0.0.0.0')\n"
76 " -k, --socket=PATH path to the unix socket\n"
77 " (default '"SOCKET_PATH"')\n"
78 " -e, --shared=NUM device can be shared by NUM clients (default '1')\n"
79 " -t, --persistent don't exit on the last connection\n"
80 " -v, --verbose display extra debugging information\n"
81 " -x, --export-name=NAME expose export by name\n"
83 "Exposing part of the image:\n"
84 " -o, --offset=OFFSET offset into the image\n"
85 " -P, --partition=NUM only expose partition NUM\n"
87 "General purpose options:\n"
88 " --object type,id=ID,... define an object such as 'secret' for providing\n"
89 " passwords and/or encryption keys\n"
91 "Kernel NBD client support:\n"
92 " -c, --connect=DEV connect FILE to the local NBD device DEV\n"
93 " -d, --disconnect disconnect the specified device\n"
97 "Block device options:\n"
98 " -f, --format=FORMAT set image format (raw, qcow2, ...)\n"
99 " -r, --read-only export read-only\n"
100 " -s, --snapshot use FILE as an external snapshot, create a temporary\n"
101 " file with backing_file=FILE, redirect the write to\n"
102 " the temporary one\n"
103 " -l, --load-snapshot=SNAPSHOT_PARAM\n"
104 " load an internal snapshot inside FILE and export it\n"
105 " as an read-only device, SNAPSHOT_PARAM format is\n"
106 " 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
108 " -n, --nocache disable host cache\n"
109 " --cache=MODE set cache mode (none, writeback, ...)\n"
110 " --aio=MODE set AIO mode (native or threads)\n"
111 " --discard=MODE set discard mode (ignore, unmap)\n"
112 " --detect-zeroes=MODE set detect-zeroes mode (off, on, unmap)\n"
113 " --image-opts treat FILE as a full set of image options\n"
115 "Report bugs to <qemu-devel@nongnu.org>\n"
116 , name, NBD_DEFAULT_PORT, "DEVICE");
119 static void version(const char *name)
123 "Written by Anthony Liguori.\n"
125 "Copyright (C) 2006 Anthony Liguori <anthony@codemonkey.ws>.\n"
126 "This is free software; see the source for copying conditions. There is NO\n"
127 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
131 struct partition_record
135 uint32_t start_cylinder;
136 uint8_t start_sector;
139 uint8_t end_cylinder;
141 uint32_t start_sector_abs;
142 uint32_t nb_sectors_abs;
145 static void read_partition(uint8_t *p, struct partition_record *r)
148 r->start_head = p[1];
149 r->start_cylinder = p[3] | ((p[2] << 2) & 0x0300);
150 r->start_sector = p[2] & 0x3f;
153 r->end_cylinder = p[7] | ((p[6] << 2) & 0x300);
154 r->end_sector = p[6] & 0x3f;
156 r->start_sector_abs = le32_to_cpup((uint32_t *)(p + 8));
157 r->nb_sectors_abs = le32_to_cpup((uint32_t *)(p + 12));
160 static int find_partition(BlockBackend *blk, int partition,
161 off_t *offset, off_t *size)
163 struct partition_record mbr[4];
164 uint8_t data[MBR_SIZE];
169 ret = blk_pread(blk, 0, data, sizeof(data));
171 error_report("error while reading: %s", strerror(-ret));
175 if (data[510] != 0x55 || data[511] != 0xaa) {
179 for (i = 0; i < 4; i++) {
180 read_partition(&data[446 + 16 * i], &mbr[i]);
182 if (!mbr[i].system || !mbr[i].nb_sectors_abs) {
186 if (mbr[i].system == 0xF || mbr[i].system == 0x5) {
187 struct partition_record ext[4];
188 uint8_t data1[MBR_SIZE];
191 ret = blk_pread(blk, mbr[i].start_sector_abs * MBR_SIZE,
192 data1, sizeof(data1));
194 error_report("error while reading: %s", strerror(-ret));
198 for (j = 0; j < 4; j++) {
199 read_partition(&data1[446 + 16 * j], &ext[j]);
200 if (!ext[j].system || !ext[j].nb_sectors_abs) {
204 if ((ext_partnum + j + 1) == partition) {
205 *offset = (uint64_t)ext[j].start_sector_abs << 9;
206 *size = (uint64_t)ext[j].nb_sectors_abs << 9;
211 } else if ((i + 1) == partition) {
212 *offset = (uint64_t)mbr[i].start_sector_abs << 9;
213 *size = (uint64_t)mbr[i].nb_sectors_abs << 9;
221 static void termsig_handler(int signum)
223 atomic_cmpxchg(&state, RUNNING, TERMINATE);
228 static void *show_parts(void *arg)
233 /* linux just needs an open() to trigger
234 * the partition table update
235 * but remember to load the module with max_part != 0 :
236 * modprobe nbd max_part=63
238 nbd = open(device, O_RDWR);
245 static void *nbd_client_thread(void *arg)
250 QIOChannelSocket *sioc;
253 pthread_t show_parts_thread;
254 Error *local_error = NULL;
256 sioc = qio_channel_socket_new();
257 if (qio_channel_socket_connect_sync(sioc,
260 error_report_err(local_error);
264 ret = nbd_receive_negotiate(QIO_CHANNEL(sioc), NULL, &nbdflags,
266 &size, &local_error);
269 error_report_err(local_error);
274 fd = open(device, O_RDWR);
276 /* Linux-only, we can use %m in printf. */
277 error_report("Failed to open %s: %m", device);
281 ret = nbd_init(fd, sioc, nbdflags, size);
286 /* update partition table */
287 pthread_create(&show_parts_thread, NULL, show_parts, device);
290 fprintf(stderr, "NBD device %s is now connected to %s\n",
293 /* Close stderr so that the qemu-nbd process exits. */
294 dup2(STDOUT_FILENO, STDERR_FILENO);
297 ret = nbd_client(fd);
302 object_unref(OBJECT(sioc));
303 kill(getpid(), SIGTERM);
304 return (void *) EXIT_SUCCESS;
309 object_unref(OBJECT(sioc));
311 kill(getpid(), SIGTERM);
312 return (void *) EXIT_FAILURE;
315 static int nbd_can_accept(void)
317 return nb_fds < shared;
320 static void nbd_export_closed(NBDExport *exp)
322 assert(state == TERMINATING);
326 static void nbd_update_server_watch(void);
328 static void nbd_client_closed(NBDClient *client)
331 if (nb_fds == 0 && !persistent && state == RUNNING) {
334 nbd_update_server_watch();
335 nbd_client_put(client);
338 static gboolean nbd_accept(QIOChannel *ioc, GIOCondition cond, gpointer opaque)
340 QIOChannelSocket *cioc;
342 cioc = qio_channel_socket_accept(QIO_CHANNEL_SOCKET(ioc),
348 if (state >= TERMINATE) {
349 object_unref(OBJECT(cioc));
354 nbd_update_server_watch();
355 nbd_client_new(newproto ? NULL : exp, cioc,
356 tlscreds, NULL, nbd_client_closed);
357 object_unref(OBJECT(cioc));
362 static void nbd_update_server_watch(void)
364 if (nbd_can_accept()) {
365 if (server_watch == -1) {
366 server_watch = qio_channel_add_watch(QIO_CHANNEL(server_ioc),
372 if (server_watch != -1) {
373 g_source_remove(server_watch);
380 static SocketAddress *nbd_build_socket_address(const char *sockpath,
384 SocketAddress *saddr;
386 saddr = g_new0(SocketAddress, 1);
388 saddr->type = SOCKET_ADDRESS_KIND_UNIX;
389 saddr->u.q_unix.data = g_new0(UnixSocketAddress, 1);
390 saddr->u.q_unix.data->path = g_strdup(sockpath);
392 InetSocketAddress *inet;
393 saddr->type = SOCKET_ADDRESS_KIND_INET;
394 inet = saddr->u.inet.data = g_new0(InetSocketAddress, 1);
395 inet->host = g_strdup(bindto);
397 inet->port = g_strdup(port);
399 inet->port = g_strdup_printf("%d", NBD_DEFAULT_PORT);
407 static QemuOptsList file_opts = {
409 .implied_opt_name = "file",
410 .head = QTAILQ_HEAD_INITIALIZER(file_opts.head),
412 /* no elements => accept any params */
413 { /* end of list */ }
417 static QemuOptsList qemu_object_opts = {
419 .implied_opt_name = "qom-type",
420 .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head),
428 static QCryptoTLSCreds *nbd_get_tls_creds(const char *id, Error **errp)
431 QCryptoTLSCreds *creds;
433 obj = object_resolve_path_component(
434 object_get_objects_root(), id);
436 error_setg(errp, "No TLS credentials with id '%s'",
440 creds = (QCryptoTLSCreds *)
441 object_dynamic_cast(obj, TYPE_QCRYPTO_TLS_CREDS);
443 error_setg(errp, "Object with id '%s' is not TLS credentials",
448 if (creds->endpoint != QCRYPTO_TLS_CREDS_ENDPOINT_SERVER) {
450 "Expecting TLS credentials with a server endpoint");
458 int main(int argc, char **argv)
461 BlockDriverState *bs;
462 off_t dev_offset = 0;
463 uint32_t nbdflags = 0;
464 bool disconnect = false;
465 const char *bindto = "0.0.0.0";
466 const char *port = NULL;
467 char *sockpath = NULL;
470 QemuOpts *sn_opts = NULL;
471 const char *sn_id_or_name = NULL;
472 const char *sopt = "hVb:o:p:rsnP:c:dvk:e:f:tl:x:";
473 struct option lopt[] = {
474 { "help", no_argument, NULL, 'h' },
475 { "version", no_argument, NULL, 'V' },
476 { "bind", required_argument, NULL, 'b' },
477 { "port", required_argument, NULL, 'p' },
478 { "socket", required_argument, NULL, 'k' },
479 { "offset", required_argument, NULL, 'o' },
480 { "read-only", no_argument, NULL, 'r' },
481 { "partition", required_argument, NULL, 'P' },
482 { "connect", required_argument, NULL, 'c' },
483 { "disconnect", no_argument, NULL, 'd' },
484 { "snapshot", no_argument, NULL, 's' },
485 { "load-snapshot", required_argument, NULL, 'l' },
486 { "nocache", no_argument, NULL, 'n' },
487 { "cache", required_argument, NULL, QEMU_NBD_OPT_CACHE },
488 { "aio", required_argument, NULL, QEMU_NBD_OPT_AIO },
489 { "discard", required_argument, NULL, QEMU_NBD_OPT_DISCARD },
490 { "detect-zeroes", required_argument, NULL,
491 QEMU_NBD_OPT_DETECT_ZEROES },
492 { "shared", required_argument, NULL, 'e' },
493 { "format", required_argument, NULL, 'f' },
494 { "persistent", no_argument, NULL, 't' },
495 { "verbose", no_argument, NULL, 'v' },
496 { "object", required_argument, NULL, QEMU_NBD_OPT_OBJECT },
497 { "export-name", required_argument, NULL, 'x' },
498 { "tls-creds", required_argument, NULL, QEMU_NBD_OPT_TLSCREDS },
499 { "image-opts", no_argument, NULL, QEMU_NBD_OPT_IMAGE_OPTS },
505 int flags = BDRV_O_RDWR;
508 bool seen_cache = false;
509 bool seen_discard = false;
510 bool seen_aio = false;
511 pthread_t client_thread;
512 const char *fmt = NULL;
513 Error *local_err = NULL;
514 BlockdevDetectZeroesOptions detect_zeroes = BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
515 QDict *options = NULL;
516 const char *export_name = NULL;
517 const char *tlscredsid = NULL;
518 bool imageOpts = false;
519 bool writethrough = true;
521 /* The client thread uses SIGTERM to interrupt the server. A signal
522 * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
524 struct sigaction sa_sigterm;
525 memset(&sa_sigterm, 0, sizeof(sa_sigterm));
526 sa_sigterm.sa_handler = termsig_handler;
527 sigaction(SIGTERM, &sa_sigterm, NULL);
529 if (qcrypto_init(&local_err) < 0) {
530 error_reportf_err(local_err, "cannot initialize crypto: ");
534 module_call_init(MODULE_INIT_QOM);
535 qemu_add_opts(&qemu_object_opts);
536 qemu_init_exec_dir(argv[0]);
538 while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
541 flags |= BDRV_O_SNAPSHOT;
544 optarg = (char *) "none";
546 case QEMU_NBD_OPT_CACHE:
548 error_report("-n and --cache can only be specified once");
552 if (bdrv_parse_cache_mode(optarg, &flags, &writethrough) == -1) {
553 error_report("Invalid cache mode `%s'", optarg);
557 case QEMU_NBD_OPT_AIO:
559 error_report("--aio can only be specified once");
563 if (!strcmp(optarg, "native")) {
564 flags |= BDRV_O_NATIVE_AIO;
565 } else if (!strcmp(optarg, "threads")) {
566 /* this is the default */
568 error_report("invalid aio mode `%s'", optarg);
572 case QEMU_NBD_OPT_DISCARD:
574 error_report("--discard can only be specified once");
578 if (bdrv_parse_discard_flags(optarg, &flags) == -1) {
579 error_report("Invalid discard mode `%s'", optarg);
583 case QEMU_NBD_OPT_DETECT_ZEROES:
585 qapi_enum_parse(BlockdevDetectZeroesOptions_lookup,
587 BLOCKDEV_DETECT_ZEROES_OPTIONS__MAX,
588 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
591 error_reportf_err(local_err,
592 "Failed to parse detect_zeroes mode: ");
595 if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
596 !(flags & BDRV_O_UNMAP)) {
597 error_report("setting detect-zeroes to unmap is not allowed "
598 "without setting discard operation to unmap");
609 dev_offset = strtoll (optarg, &end, 0);
611 error_report("Invalid offset `%s'", optarg);
614 if (dev_offset < 0) {
615 error_report("Offset must be positive `%s'", optarg);
620 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
621 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
624 error_report("Failed in parsing snapshot param `%s'",
629 sn_id_or_name = optarg;
633 nbdflags |= NBD_FLAG_READ_ONLY;
634 flags &= ~BDRV_O_RDWR;
637 partition = strtol(optarg, &end, 0);
639 error_report("Invalid partition `%s'", optarg);
642 if (partition < 1 || partition > 8) {
643 error_report("Invalid partition %d", partition);
649 if (sockpath[0] != '/') {
650 error_report("socket path must be absolute");
661 shared = strtol(optarg, &end, 0);
663 error_report("Invalid shared device number '%s'", optarg);
667 error_report("Shared device number must be greater than 0");
678 export_name = optarg;
692 error_report("Try `%s --help' for more information.", argv[0]);
694 case QEMU_NBD_OPT_OBJECT: {
696 opts = qemu_opts_parse_noisily(&qemu_object_opts,
702 case QEMU_NBD_OPT_TLSCREDS:
705 case QEMU_NBD_OPT_IMAGE_OPTS:
711 if ((argc - optind) != 1) {
712 error_report("Invalid number of arguments");
713 error_printf("Try `%s --help' for more information.\n", argv[0]);
717 if (qemu_opts_foreach(&qemu_object_opts,
718 user_creatable_add_opts_foreach,
725 error_report("TLS is only supported with IPv4/IPv6");
729 error_report("TLS is not supported with a host device");
733 /* Set the default NBD protocol export name, since
734 * we *must* use new style protocol for TLS */
737 tlscreds = nbd_get_tls_creds(tlscredsid, &local_err);
739 error_report("Failed to get TLS creds %s",
740 error_get_pretty(local_err));
746 int nbdfd = open(argv[optind], O_RDWR);
748 error_report("Cannot open %s: %s", argv[optind],
752 nbd_disconnect(nbdfd);
756 printf("%s disconnected\n", argv[optind]);
761 if (device && !verbose) {
766 if (qemu_pipe(stderr_fd) < 0) {
767 error_report("Error setting up communication pipe: %s",
772 /* Now daemonize, but keep a communication channel open to
773 * print errors and exit with the proper status code.
777 error_report("Failed to fork: %s", strerror(errno));
779 } else if (pid == 0) {
781 ret = qemu_daemon(1, 0);
783 /* Temporarily redirect stderr to the parent's pipe... */
784 dup2(stderr_fd[1], STDERR_FILENO);
786 error_report("Failed to daemonize: %s", strerror(errno));
790 /* ... close the descriptor we inherited and go on. */
796 /* In the parent. Print error messages from the child until
797 * it closes the pipe.
800 buf = g_malloc(1024);
801 while ((ret = read(stderr_fd[0], buf, 1024)) > 0) {
803 ret = qemu_write_full(STDERR_FILENO, buf, ret);
809 error_report("Cannot read from daemon: %s",
814 /* Usually the daemon should not print any message.
815 * Exit with zero status in that case.
821 if (device != NULL && sockpath == NULL) {
822 sockpath = g_malloc(128);
823 snprintf(sockpath, 128, SOCKET_PATH, basename(device));
826 saddr = nbd_build_socket_address(sockpath, bindto, port);
828 if (qemu_init_main_loop(&local_err)) {
829 error_report_err(local_err);
833 atexit(bdrv_close_all);
835 srcpath = argv[optind];
839 error_report("--image-opts and -f are mutually exclusive");
842 opts = qemu_opts_parse_noisily(&file_opts, srcpath, true);
844 qemu_opts_reset(&file_opts);
847 options = qemu_opts_to_qdict(opts, NULL);
848 qemu_opts_reset(&file_opts);
849 blk = blk_new_open(NULL, NULL, options, flags, &local_err);
852 options = qdict_new();
853 qdict_put(options, "driver", qstring_from_str(fmt));
855 blk = blk_new_open(srcpath, NULL, options, flags, &local_err);
859 error_reportf_err(local_err, "Failed to blk_new_open '%s': ",
865 blk_set_enable_write_cache(blk, !writethrough);
868 ret = bdrv_snapshot_load_tmp(bs,
869 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
870 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
872 } else if (sn_id_or_name) {
873 ret = bdrv_snapshot_load_tmp_by_id_or_name(bs, sn_id_or_name,
877 error_reportf_err(local_err, "Failed to load snapshot: ");
881 bs->detect_zeroes = detect_zeroes;
882 fd_size = blk_getlength(blk);
884 error_report("Failed to determine the image length: %s",
889 if (partition != -1) {
890 ret = find_partition(blk, partition, &dev_offset, &fd_size);
892 error_report("Could not find partition %d: %s", partition,
898 exp = nbd_export_new(blk, dev_offset, fd_size, nbdflags, nbd_export_closed,
901 error_report_err(local_err);
905 nbd_export_set_name(exp, export_name);
909 server_ioc = qio_channel_socket_new();
910 if (qio_channel_socket_listen_sync(server_ioc, saddr, &local_err) < 0) {
911 object_unref(OBJECT(server_ioc));
912 error_report_err(local_err);
919 ret = pthread_create(&client_thread, NULL, nbd_client_thread, device);
921 error_report("Failed to create client thread: %s", strerror(ret));
925 /* Shut up GCC warnings. */
926 memset(&client_thread, 0, sizeof(client_thread));
929 nbd_update_server_watch();
931 /* now when the initialization is (almost) complete, chdir("/")
932 * to free any busy filesystems */
933 if (chdir("/") < 0) {
934 error_report("Could not chdir to root directory: %s",
941 main_loop_wait(false);
942 if (state == TERMINATE) {
944 nbd_export_close(exp);
948 } while (state != TERMINATED);
955 qemu_opts_del(sn_opts);
959 pthread_join(client_thread, &ret);