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-common.h"
20 #include "sysemu/block-backend.h"
21 #include "block/block_int.h"
22 #include "block/nbd.h"
23 #include "qemu/main-loop.h"
24 #include "qemu/sockets.h"
25 #include "qemu/error-report.h"
26 #include "block/snapshot.h"
27 #include "qapi/util.h"
33 #include <sys/types.h>
34 #include <sys/socket.h>
35 #include <netinet/in.h>
36 #include <netinet/tcp.h>
37 #include <arpa/inet.h>
42 #define SOCKET_PATH "/var/lock/qemu-nbd-%s"
43 #define QEMU_NBD_OPT_CACHE 1
44 #define QEMU_NBD_OPT_AIO 2
45 #define QEMU_NBD_OPT_DISCARD 3
46 #define QEMU_NBD_OPT_DETECT_ZEROES 4
48 static NBDExport *exp;
51 static char *sockpath;
52 static int persistent = 0;
53 static enum { RUNNING, TERMINATE, TERMINATING, TERMINATED } state;
54 static int shared = 1;
57 static void usage(const char *name)
60 "Usage: %s [OPTIONS] FILE\n"
61 "QEMU Disk Network Block Device Server\n"
63 " -h, --help display this help and exit\n"
64 " -V, --version output version information and exit\n"
66 "Connection properties:\n"
67 " -p, --port=PORT port to listen on (default `%d')\n"
68 " -b, --bind=IFACE interface to bind to (default `0.0.0.0')\n"
69 " -k, --socket=PATH path to the unix socket\n"
70 " (default '"SOCKET_PATH"')\n"
71 " -e, --shared=NUM device can be shared by NUM clients (default '1')\n"
72 " -t, --persistent don't exit on the last connection\n"
73 " -v, --verbose display extra debugging information\n"
75 "Exposing part of the image:\n"
76 " -o, --offset=OFFSET offset into the image\n"
77 " -P, --partition=NUM only expose partition NUM\n"
80 "Kernel NBD client support:\n"
81 " -c, --connect=DEV connect FILE to the local NBD device DEV\n"
82 " -d, --disconnect disconnect the specified device\n"
86 "Block device options:\n"
87 " -f, --format=FORMAT set image format (raw, qcow2, ...)\n"
88 " -r, --read-only export read-only\n"
89 " -s, --snapshot use FILE as an external snapshot, create a temporary\n"
90 " file with backing_file=FILE, redirect the write to\n"
91 " the temporary one\n"
92 " -l, --load-snapshot=SNAPSHOT_PARAM\n"
93 " load an internal snapshot inside FILE and export it\n"
94 " as an read-only device, SNAPSHOT_PARAM format is\n"
95 " 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
97 " -n, --nocache disable host cache\n"
98 " --cache=MODE set cache mode (none, writeback, ...)\n"
99 #ifdef CONFIG_LINUX_AIO
100 " --aio=MODE set AIO mode (native or threads)\n"
102 " --discard=MODE set discard mode (ignore, unmap)\n"
103 " --detect-zeroes=MODE set detect-zeroes mode (off, on, discard)\n"
105 "Report bugs to <qemu-devel@nongnu.org>\n"
106 , name, NBD_DEFAULT_PORT, "DEVICE");
109 static void version(const char *name)
113 "Written by Anthony Liguori.\n"
115 "Copyright (C) 2006 Anthony Liguori <anthony@codemonkey.ws>.\n"
116 "This is free software; see the source for copying conditions. There is NO\n"
117 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
121 struct partition_record
125 uint32_t start_cylinder;
126 uint8_t start_sector;
129 uint8_t end_cylinder;
131 uint32_t start_sector_abs;
132 uint32_t nb_sectors_abs;
135 static void read_partition(uint8_t *p, struct partition_record *r)
138 r->start_head = p[1];
139 r->start_cylinder = p[3] | ((p[2] << 2) & 0x0300);
140 r->start_sector = p[2] & 0x3f;
143 r->end_cylinder = p[7] | ((p[6] << 2) & 0x300);
144 r->end_sector = p[6] & 0x3f;
145 r->start_sector_abs = p[8] | p[9] << 8 | p[10] << 16 | p[11] << 24;
146 r->nb_sectors_abs = p[12] | p[13] << 8 | p[14] << 16 | p[15] << 24;
149 static int find_partition(BlockBackend *blk, int partition,
150 off_t *offset, off_t *size)
152 struct partition_record mbr[4];
158 if ((ret = blk_read(blk, 0, data, 1)) < 0) {
160 err(EXIT_FAILURE, "error while reading");
163 if (data[510] != 0x55 || data[511] != 0xaa) {
167 for (i = 0; i < 4; i++) {
168 read_partition(&data[446 + 16 * i], &mbr[i]);
170 if (!mbr[i].nb_sectors_abs)
173 if (mbr[i].system == 0xF || mbr[i].system == 0x5) {
174 struct partition_record ext[4];
178 if ((ret = blk_read(blk, mbr[i].start_sector_abs, data1, 1)) < 0) {
180 err(EXIT_FAILURE, "error while reading");
183 for (j = 0; j < 4; j++) {
184 read_partition(&data1[446 + 16 * j], &ext[j]);
185 if (!ext[j].nb_sectors_abs)
188 if ((ext_partnum + j + 1) == partition) {
189 *offset = (uint64_t)ext[j].start_sector_abs << 9;
190 *size = (uint64_t)ext[j].nb_sectors_abs << 9;
195 } else if ((i + 1) == partition) {
196 *offset = (uint64_t)mbr[i].start_sector_abs << 9;
197 *size = (uint64_t)mbr[i].nb_sectors_abs << 9;
205 static void termsig_handler(int signum)
211 static void combine_addr(char *buf, size_t len, const char* address,
214 /* If the address-part contains a colon, it's an IPv6 IP so needs [] */
215 if (strstr(address, ":")) {
216 snprintf(buf, len, "[%s]:%u", address, port);
218 snprintf(buf, len, "%s:%u", address, port);
222 static int tcp_socket_incoming(const char *address, uint16_t port)
224 char address_and_port[128];
225 Error *local_err = NULL;
227 combine_addr(address_and_port, 128, address, port);
228 int fd = inet_listen(address_and_port, NULL, 0, SOCK_STREAM, 0, &local_err);
230 if (local_err != NULL) {
231 error_report("%s", error_get_pretty(local_err));
232 error_free(local_err);
237 static int unix_socket_incoming(const char *path)
239 Error *local_err = NULL;
240 int fd = unix_listen(path, NULL, 0, &local_err);
242 if (local_err != NULL) {
243 error_report("%s", error_get_pretty(local_err));
244 error_free(local_err);
249 static int unix_socket_outgoing(const char *path)
251 Error *local_err = NULL;
252 int fd = unix_connect(path, &local_err);
254 if (local_err != NULL) {
255 error_report("%s", error_get_pretty(local_err));
256 error_free(local_err);
261 static void *show_parts(void *arg)
266 /* linux just needs an open() to trigger
267 * the partition table update
268 * but remember to load the module with max_part != 0 :
269 * modprobe nbd max_part=63
271 nbd = open(device, O_RDWR);
278 static void *nbd_client_thread(void *arg)
286 pthread_t show_parts_thread;
287 Error *local_error = NULL;
289 sock = unix_socket_outgoing(sockpath);
294 ret = nbd_receive_negotiate(sock, NULL, &nbdflags,
295 &size, &blocksize, &local_error);
298 fprintf(stderr, "%s\n", error_get_pretty(local_error));
299 error_free(local_error);
304 fd = open(device, O_RDWR);
306 /* Linux-only, we can use %m in printf. */
307 fprintf(stderr, "Failed to open %s: %m\n", device);
311 ret = nbd_init(fd, sock, nbdflags, size, blocksize);
316 /* update partition table */
317 pthread_create(&show_parts_thread, NULL, show_parts, device);
320 fprintf(stderr, "NBD device %s is now connected to %s\n",
323 /* Close stderr so that the qemu-nbd process exits. */
324 dup2(STDOUT_FILENO, STDERR_FILENO);
327 ret = nbd_client(fd);
332 kill(getpid(), SIGTERM);
333 return (void *) EXIT_SUCCESS;
340 kill(getpid(), SIGTERM);
341 return (void *) EXIT_FAILURE;
344 static int nbd_can_accept(void *opaque)
346 return nb_fds < shared;
349 static void nbd_export_closed(NBDExport *exp)
351 assert(state == TERMINATING);
355 static void nbd_client_closed(NBDClient *client)
358 if (nb_fds == 0 && !persistent && state == RUNNING) {
362 nbd_client_put(client);
365 static void nbd_accept(void *opaque)
367 int server_fd = (uintptr_t) opaque;
368 struct sockaddr_in addr;
369 socklen_t addr_len = sizeof(addr);
371 int fd = accept(server_fd, (struct sockaddr *)&addr, &addr_len);
377 if (state >= TERMINATE) {
382 if (nbd_client_new(exp, fd, nbd_client_closed)) {
390 int main(int argc, char **argv)
393 BlockDriverState *bs;
395 off_t dev_offset = 0;
396 uint32_t nbdflags = 0;
397 bool disconnect = false;
398 const char *bindto = "0.0.0.0";
400 int port = NBD_DEFAULT_PORT;
402 QemuOpts *sn_opts = NULL;
403 const char *sn_id_or_name = NULL;
404 const char *sopt = "hVb:o:p:rsnP:c:dvk:e:f:tl:";
405 struct option lopt[] = {
406 { "help", 0, NULL, 'h' },
407 { "version", 0, NULL, 'V' },
408 { "bind", 1, NULL, 'b' },
409 { "port", 1, NULL, 'p' },
410 { "socket", 1, NULL, 'k' },
411 { "offset", 1, NULL, 'o' },
412 { "read-only", 0, NULL, 'r' },
413 { "partition", 1, NULL, 'P' },
414 { "connect", 1, NULL, 'c' },
415 { "disconnect", 0, NULL, 'd' },
416 { "snapshot", 0, NULL, 's' },
417 { "load-snapshot", 1, NULL, 'l' },
418 { "nocache", 0, NULL, 'n' },
419 { "cache", 1, NULL, QEMU_NBD_OPT_CACHE },
420 #ifdef CONFIG_LINUX_AIO
421 { "aio", 1, NULL, QEMU_NBD_OPT_AIO },
423 { "discard", 1, NULL, QEMU_NBD_OPT_DISCARD },
424 { "detect-zeroes", 1, NULL, QEMU_NBD_OPT_DETECT_ZEROES },
425 { "shared", 1, NULL, 'e' },
426 { "format", 1, NULL, 'f' },
427 { "persistent", 0, NULL, 't' },
428 { "verbose", 0, NULL, 'v' },
435 int flags = BDRV_O_RDWR;
439 bool seen_cache = false;
440 bool seen_discard = false;
441 #ifdef CONFIG_LINUX_AIO
442 bool seen_aio = false;
444 pthread_t client_thread;
445 const char *fmt = NULL;
446 Error *local_err = NULL;
447 BlockdevDetectZeroesOptions detect_zeroes = BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
449 /* The client thread uses SIGTERM to interrupt the server. A signal
450 * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
452 struct sigaction sa_sigterm;
453 memset(&sa_sigterm, 0, sizeof(sa_sigterm));
454 sa_sigterm.sa_handler = termsig_handler;
455 sigaction(SIGTERM, &sa_sigterm, NULL);
456 qemu_init_exec_dir(argv[0]);
458 while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
461 flags |= BDRV_O_SNAPSHOT;
464 optarg = (char *) "none";
466 case QEMU_NBD_OPT_CACHE:
468 errx(EXIT_FAILURE, "-n and --cache can only be specified once");
471 if (bdrv_parse_cache_flags(optarg, &flags) == -1) {
472 errx(EXIT_FAILURE, "Invalid cache mode `%s'", optarg);
475 #ifdef CONFIG_LINUX_AIO
476 case QEMU_NBD_OPT_AIO:
478 errx(EXIT_FAILURE, "--aio can only be specified once");
481 if (!strcmp(optarg, "native")) {
482 flags |= BDRV_O_NATIVE_AIO;
483 } else if (!strcmp(optarg, "threads")) {
484 /* this is the default */
486 errx(EXIT_FAILURE, "invalid aio mode `%s'", optarg);
490 case QEMU_NBD_OPT_DISCARD:
492 errx(EXIT_FAILURE, "--discard can only be specified once");
495 if (bdrv_parse_discard_flags(optarg, &flags) == -1) {
496 errx(EXIT_FAILURE, "Invalid discard mode `%s'", optarg);
499 case QEMU_NBD_OPT_DETECT_ZEROES:
501 qapi_enum_parse(BlockdevDetectZeroesOptions_lookup,
503 BLOCKDEV_DETECT_ZEROES_OPTIONS_MAX,
504 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
507 errx(EXIT_FAILURE, "Failed to parse detect_zeroes mode: %s",
508 error_get_pretty(local_err));
510 if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
511 !(flags & BDRV_O_UNMAP)) {
512 errx(EXIT_FAILURE, "setting detect-zeroes to unmap is not allowed "
513 "without setting discard operation to unmap");
520 li = strtol(optarg, &end, 0);
522 errx(EXIT_FAILURE, "Invalid port `%s'", optarg);
524 if (li < 1 || li > 65535) {
525 errx(EXIT_FAILURE, "Port out of range `%s'", optarg);
530 dev_offset = strtoll (optarg, &end, 0);
532 errx(EXIT_FAILURE, "Invalid offset `%s'", optarg);
534 if (dev_offset < 0) {
535 errx(EXIT_FAILURE, "Offset must be positive `%s'", optarg);
539 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
540 sn_opts = qemu_opts_parse(&internal_snapshot_opts, optarg, 0);
542 errx(EXIT_FAILURE, "Failed in parsing snapshot param `%s'",
546 sn_id_or_name = optarg;
550 nbdflags |= NBD_FLAG_READ_ONLY;
551 flags &= ~BDRV_O_RDWR;
554 partition = strtol(optarg, &end, 0);
556 errx(EXIT_FAILURE, "Invalid partition `%s'", optarg);
558 if (partition < 1 || partition > 8) {
559 errx(EXIT_FAILURE, "Invalid partition %d", partition);
564 if (sockpath[0] != '/') {
565 errx(EXIT_FAILURE, "socket path must be absolute\n");
575 shared = strtol(optarg, &end, 0);
577 errx(EXIT_FAILURE, "Invalid shared device number '%s'", optarg);
580 errx(EXIT_FAILURE, "Shared device number must be greater than 0\n");
601 errx(EXIT_FAILURE, "Try `%s --help' for more information.",
606 if ((argc - optind) != 1) {
607 errx(EXIT_FAILURE, "Invalid number of argument.\n"
608 "Try `%s --help' for more information.",
613 fd = open(argv[optind], O_RDWR);
615 err(EXIT_FAILURE, "Cannot open %s", argv[optind]);
621 printf("%s disconnected\n", argv[optind]);
626 if (device && !verbose) {
631 if (qemu_pipe(stderr_fd) < 0) {
632 err(EXIT_FAILURE, "Error setting up communication pipe");
635 /* Now daemonize, but keep a communication channel open to
636 * print errors and exit with the proper status code.
641 ret = qemu_daemon(1, 0);
643 /* Temporarily redirect stderr to the parent's pipe... */
644 dup2(stderr_fd[1], STDERR_FILENO);
646 err(EXIT_FAILURE, "Failed to daemonize");
649 /* ... close the descriptor we inherited and go on. */
655 /* In the parent. Print error messages from the child until
656 * it closes the pipe.
659 buf = g_malloc(1024);
660 while ((ret = read(stderr_fd[0], buf, 1024)) > 0) {
662 ret = qemu_write_full(STDERR_FILENO, buf, ret);
668 err(EXIT_FAILURE, "Cannot read from daemon");
671 /* Usually the daemon should not print any message.
672 * Exit with zero status in that case.
678 if (device != NULL && sockpath == NULL) {
679 sockpath = g_malloc(128);
680 snprintf(sockpath, 128, SOCKET_PATH, basename(device));
683 if (qemu_init_main_loop(&local_err)) {
684 error_report("%s", error_get_pretty(local_err));
685 error_free(local_err);
689 atexit(bdrv_close_all);
692 drv = bdrv_find_format(fmt);
694 errx(EXIT_FAILURE, "Unknown file format '%s'", fmt);
700 blk = blk_new_with_bs("hda", &error_abort);
703 srcpath = argv[optind];
704 ret = bdrv_open(&bs, srcpath, NULL, NULL, flags, drv, &local_err);
707 err(EXIT_FAILURE, "Failed to bdrv_open '%s': %s", argv[optind],
708 error_get_pretty(local_err));
712 ret = bdrv_snapshot_load_tmp(bs,
713 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
714 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
716 } else if (sn_id_or_name) {
717 ret = bdrv_snapshot_load_tmp_by_id_or_name(bs, sn_id_or_name,
723 "Failed to load snapshot: %s",
724 error_get_pretty(local_err));
727 bs->detect_zeroes = detect_zeroes;
728 fd_size = blk_getlength(blk);
730 if (partition != -1) {
731 ret = find_partition(blk, partition, &dev_offset, &fd_size);
734 err(EXIT_FAILURE, "Could not find partition %d", partition);
738 exp = nbd_export_new(blk, dev_offset, fd_size, nbdflags, nbd_export_closed);
741 fd = unix_socket_incoming(sockpath);
743 fd = tcp_socket_incoming(bindto, port);
753 ret = pthread_create(&client_thread, NULL, nbd_client_thread, device);
755 errx(EXIT_FAILURE, "Failed to create client thread: %s",
759 /* Shut up GCC warnings. */
760 memset(&client_thread, 0, sizeof(client_thread));
763 qemu_set_fd_handler2(fd, nbd_can_accept, nbd_accept, NULL,
764 (void *)(uintptr_t)fd);
766 /* now when the initialization is (almost) complete, chdir("/")
767 * to free any busy filesystems */
768 if (chdir("/") < 0) {
769 err(EXIT_FAILURE, "Could not chdir to root directory");
774 main_loop_wait(false);
775 if (state == TERMINATE) {
777 nbd_export_close(exp);
781 } while (state != TERMINATED);
788 qemu_opts_del(sn_opts);
792 pthread_join(client_thread, &ret);