ssi: Convert legacy SSI_SLAVE -> DEVICE casts
[qemu.git] / qemu-nbd.c
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-common.h"
20 #include "block/block.h"
21 #include "block/nbd.h"
22 #include "qemu/main-loop.h"
23 #include "qemu/sockets.h"
24 #include "qemu/error-report.h"
25 #include "block/snapshot.h"
26
27 #include <stdarg.h>
28 #include <stdio.h>
29 #include <getopt.h>
30 #include <err.h>
31 #include <sys/types.h>
32 #include <sys/socket.h>
33 #include <netinet/in.h>
34 #include <netinet/tcp.h>
35 #include <arpa/inet.h>
36 #include <signal.h>
37 #include <libgen.h>
38 #include <pthread.h>
39
40 #define SOCKET_PATH "/var/lock/qemu-nbd-%s"
41 #define QEMU_NBD_OPT_CACHE 1
42 #define QEMU_NBD_OPT_AIO 2
43 #define QEMU_NBD_OPT_DISCARD 3
44
45 static NBDExport *exp;
46 static int verbose;
47 static char *srcpath;
48 static char *sockpath;
49 static int persistent = 0;
50 static enum { RUNNING, TERMINATE, TERMINATING, TERMINATED } state;
51 static int shared = 1;
52 static int nb_fds;
53
54 static void usage(const char *name)
55 {
56 (printf) (
57 "Usage: %s [OPTIONS] FILE\n"
58 "QEMU Disk Network Block Device Server\n"
59 "\n"
60 " -h, --help display this help and exit\n"
61 " -V, --version output version information and exit\n"
62 "\n"
63 "Connection properties:\n"
64 " -p, --port=PORT port to listen on (default `%d')\n"
65 " -b, --bind=IFACE interface to bind to (default `0.0.0.0')\n"
66 " -k, --socket=PATH path to the unix socket\n"
67 " (default '"SOCKET_PATH"')\n"
68 " -e, --shared=NUM device can be shared by NUM clients (default '1')\n"
69 " -t, --persistent don't exit on the last connection\n"
70 " -v, --verbose display extra debugging information\n"
71 "\n"
72 "Exposing part of the image:\n"
73 " -o, --offset=OFFSET offset into the image\n"
74 " -P, --partition=NUM only expose partition NUM\n"
75 "\n"
76 #ifdef __linux__
77 "Kernel NBD client support:\n"
78 " -c, --connect=DEV connect FILE to the local NBD device DEV\n"
79 " -d, --disconnect disconnect the specified device\n"
80 "\n"
81 #endif
82 "\n"
83 "Block device options:\n"
84 " -f, --format=FORMAT set image format (raw, qcow2, ...)\n"
85 " -r, --read-only export read-only\n"
86 " -s, --snapshot use FILE as an external snapshot, create a temporary\n"
87 " file with backing_file=FILE, redirect the write to\n"
88 " the temporary one\n"
89 " -l, --load-snapshot=SNAPSHOT_PARAM\n"
90 " load an internal snapshot inside FILE and export it\n"
91 " as an read-only device, SNAPSHOT_PARAM format is\n"
92 " 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
93 " '[ID_OR_NAME]'\n"
94 " -n, --nocache disable host cache\n"
95 " --cache=MODE set cache mode (none, writeback, ...)\n"
96 #ifdef CONFIG_LINUX_AIO
97 " --aio=MODE set AIO mode (native or threads)\n"
98 #endif
99 "\n"
100 "Report bugs to <qemu-devel@nongnu.org>\n"
101 , name, NBD_DEFAULT_PORT, "DEVICE");
102 }
103
104 static void version(const char *name)
105 {
106 printf(
107 "%s version 0.0.1\n"
108 "Written by Anthony Liguori.\n"
109 "\n"
110 "Copyright (C) 2006 Anthony Liguori <anthony@codemonkey.ws>.\n"
111 "This is free software; see the source for copying conditions. There is NO\n"
112 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
113 , name);
114 }
115
116 struct partition_record
117 {
118 uint8_t bootable;
119 uint8_t start_head;
120 uint32_t start_cylinder;
121 uint8_t start_sector;
122 uint8_t system;
123 uint8_t end_head;
124 uint8_t end_cylinder;
125 uint8_t end_sector;
126 uint32_t start_sector_abs;
127 uint32_t nb_sectors_abs;
128 };
129
130 static void read_partition(uint8_t *p, struct partition_record *r)
131 {
132 r->bootable = p[0];
133 r->start_head = p[1];
134 r->start_cylinder = p[3] | ((p[2] << 2) & 0x0300);
135 r->start_sector = p[2] & 0x3f;
136 r->system = p[4];
137 r->end_head = p[5];
138 r->end_cylinder = p[7] | ((p[6] << 2) & 0x300);
139 r->end_sector = p[6] & 0x3f;
140 r->start_sector_abs = p[8] | p[9] << 8 | p[10] << 16 | p[11] << 24;
141 r->nb_sectors_abs = p[12] | p[13] << 8 | p[14] << 16 | p[15] << 24;
142 }
143
144 static int find_partition(BlockDriverState *bs, int partition,
145 off_t *offset, off_t *size)
146 {
147 struct partition_record mbr[4];
148 uint8_t data[512];
149 int i;
150 int ext_partnum = 4;
151 int ret;
152
153 if ((ret = bdrv_read(bs, 0, data, 1)) < 0) {
154 errno = -ret;
155 err(EXIT_FAILURE, "error while reading");
156 }
157
158 if (data[510] != 0x55 || data[511] != 0xaa) {
159 return -EINVAL;
160 }
161
162 for (i = 0; i < 4; i++) {
163 read_partition(&data[446 + 16 * i], &mbr[i]);
164
165 if (!mbr[i].nb_sectors_abs)
166 continue;
167
168 if (mbr[i].system == 0xF || mbr[i].system == 0x5) {
169 struct partition_record ext[4];
170 uint8_t data1[512];
171 int j;
172
173 if ((ret = bdrv_read(bs, mbr[i].start_sector_abs, data1, 1)) < 0) {
174 errno = -ret;
175 err(EXIT_FAILURE, "error while reading");
176 }
177
178 for (j = 0; j < 4; j++) {
179 read_partition(&data1[446 + 16 * j], &ext[j]);
180 if (!ext[j].nb_sectors_abs)
181 continue;
182
183 if ((ext_partnum + j + 1) == partition) {
184 *offset = (uint64_t)ext[j].start_sector_abs << 9;
185 *size = (uint64_t)ext[j].nb_sectors_abs << 9;
186 return 0;
187 }
188 }
189 ext_partnum += 4;
190 } else if ((i + 1) == partition) {
191 *offset = (uint64_t)mbr[i].start_sector_abs << 9;
192 *size = (uint64_t)mbr[i].nb_sectors_abs << 9;
193 return 0;
194 }
195 }
196
197 return -ENOENT;
198 }
199
200 static void termsig_handler(int signum)
201 {
202 state = TERMINATE;
203 qemu_notify_event();
204 }
205
206 static void combine_addr(char *buf, size_t len, const char* address,
207 uint16_t port)
208 {
209 /* If the address-part contains a colon, it's an IPv6 IP so needs [] */
210 if (strstr(address, ":")) {
211 snprintf(buf, len, "[%s]:%u", address, port);
212 } else {
213 snprintf(buf, len, "%s:%u", address, port);
214 }
215 }
216
217 static int tcp_socket_incoming(const char *address, uint16_t port)
218 {
219 char address_and_port[128];
220 Error *local_err = NULL;
221
222 combine_addr(address_and_port, 128, address, port);
223 int fd = inet_listen(address_and_port, NULL, 0, SOCK_STREAM, 0, &local_err);
224
225 if (local_err != NULL) {
226 qerror_report_err(local_err);
227 error_free(local_err);
228 }
229 return fd;
230 }
231
232 static int unix_socket_incoming(const char *path)
233 {
234 Error *local_err = NULL;
235 int fd = unix_listen(path, NULL, 0, &local_err);
236
237 if (local_err != NULL) {
238 qerror_report_err(local_err);
239 error_free(local_err);
240 }
241 return fd;
242 }
243
244 static int unix_socket_outgoing(const char *path)
245 {
246 Error *local_err = NULL;
247 int fd = unix_connect(path, &local_err);
248
249 if (local_err != NULL) {
250 qerror_report_err(local_err);
251 error_free(local_err);
252 }
253 return fd;
254 }
255
256 static void *show_parts(void *arg)
257 {
258 char *device = arg;
259 int nbd;
260
261 /* linux just needs an open() to trigger
262 * the partition table update
263 * but remember to load the module with max_part != 0 :
264 * modprobe nbd max_part=63
265 */
266 nbd = open(device, O_RDWR);
267 if (nbd >= 0) {
268 close(nbd);
269 }
270 return NULL;
271 }
272
273 static void *nbd_client_thread(void *arg)
274 {
275 char *device = arg;
276 off_t size;
277 size_t blocksize;
278 uint32_t nbdflags;
279 int fd, sock;
280 int ret;
281 pthread_t show_parts_thread;
282
283 sock = unix_socket_outgoing(sockpath);
284 if (sock < 0) {
285 goto out;
286 }
287
288 ret = nbd_receive_negotiate(sock, NULL, &nbdflags,
289 &size, &blocksize);
290 if (ret < 0) {
291 goto out;
292 }
293
294 fd = open(device, O_RDWR);
295 if (fd < 0) {
296 /* Linux-only, we can use %m in printf. */
297 fprintf(stderr, "Failed to open %s: %m", device);
298 goto out;
299 }
300
301 ret = nbd_init(fd, sock, nbdflags, size, blocksize);
302 if (ret < 0) {
303 goto out;
304 }
305
306 /* update partition table */
307 pthread_create(&show_parts_thread, NULL, show_parts, device);
308
309 if (verbose) {
310 fprintf(stderr, "NBD device %s is now connected to %s\n",
311 device, srcpath);
312 } else {
313 /* Close stderr so that the qemu-nbd process exits. */
314 dup2(STDOUT_FILENO, STDERR_FILENO);
315 }
316
317 ret = nbd_client(fd);
318 if (ret) {
319 goto out;
320 }
321 close(fd);
322 kill(getpid(), SIGTERM);
323 return (void *) EXIT_SUCCESS;
324
325 out:
326 kill(getpid(), SIGTERM);
327 return (void *) EXIT_FAILURE;
328 }
329
330 static int nbd_can_accept(void *opaque)
331 {
332 return nb_fds < shared;
333 }
334
335 static void nbd_export_closed(NBDExport *exp)
336 {
337 assert(state == TERMINATING);
338 state = TERMINATED;
339 }
340
341 static void nbd_client_closed(NBDClient *client)
342 {
343 nb_fds--;
344 if (nb_fds == 0 && !persistent && state == RUNNING) {
345 state = TERMINATE;
346 }
347 qemu_notify_event();
348 nbd_client_put(client);
349 }
350
351 static void nbd_accept(void *opaque)
352 {
353 int server_fd = (uintptr_t) opaque;
354 struct sockaddr_in addr;
355 socklen_t addr_len = sizeof(addr);
356
357 int fd = accept(server_fd, (struct sockaddr *)&addr, &addr_len);
358 if (state >= TERMINATE) {
359 close(fd);
360 return;
361 }
362
363 if (fd >= 0 && nbd_client_new(exp, fd, nbd_client_closed)) {
364 nb_fds++;
365 }
366 }
367
368 int main(int argc, char **argv)
369 {
370 BlockDriverState *bs;
371 BlockDriver *drv;
372 off_t dev_offset = 0;
373 uint32_t nbdflags = 0;
374 bool disconnect = false;
375 const char *bindto = "0.0.0.0";
376 char *device = NULL;
377 int port = NBD_DEFAULT_PORT;
378 off_t fd_size;
379 QemuOpts *sn_opts = NULL;
380 const char *sn_id_or_name = NULL;
381 const char *sopt = "hVb:o:p:rsnP:c:dvk:e:f:tl:";
382 struct option lopt[] = {
383 { "help", 0, NULL, 'h' },
384 { "version", 0, NULL, 'V' },
385 { "bind", 1, NULL, 'b' },
386 { "port", 1, NULL, 'p' },
387 { "socket", 1, NULL, 'k' },
388 { "offset", 1, NULL, 'o' },
389 { "read-only", 0, NULL, 'r' },
390 { "partition", 1, NULL, 'P' },
391 { "connect", 1, NULL, 'c' },
392 { "disconnect", 0, NULL, 'd' },
393 { "snapshot", 0, NULL, 's' },
394 { "load-snapshot", 1, NULL, 'l' },
395 { "nocache", 0, NULL, 'n' },
396 { "cache", 1, NULL, QEMU_NBD_OPT_CACHE },
397 #ifdef CONFIG_LINUX_AIO
398 { "aio", 1, NULL, QEMU_NBD_OPT_AIO },
399 #endif
400 { "discard", 1, NULL, QEMU_NBD_OPT_DISCARD },
401 { "shared", 1, NULL, 'e' },
402 { "format", 1, NULL, 'f' },
403 { "persistent", 0, NULL, 't' },
404 { "verbose", 0, NULL, 'v' },
405 { NULL, 0, NULL, 0 }
406 };
407 int ch;
408 int opt_ind = 0;
409 int li;
410 char *end;
411 int flags = BDRV_O_RDWR;
412 int partition = -1;
413 int ret;
414 int fd;
415 bool seen_cache = false;
416 bool seen_discard = false;
417 #ifdef CONFIG_LINUX_AIO
418 bool seen_aio = false;
419 #endif
420 pthread_t client_thread;
421 const char *fmt = NULL;
422 Error *local_err = NULL;
423
424 /* The client thread uses SIGTERM to interrupt the server. A signal
425 * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
426 */
427 struct sigaction sa_sigterm;
428 memset(&sa_sigterm, 0, sizeof(sa_sigterm));
429 sa_sigterm.sa_handler = termsig_handler;
430 sigaction(SIGTERM, &sa_sigterm, NULL);
431 qemu_init_exec_dir(argv[0]);
432
433 while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
434 switch (ch) {
435 case 's':
436 flags |= BDRV_O_SNAPSHOT;
437 break;
438 case 'n':
439 optarg = (char *) "none";
440 /* fallthrough */
441 case QEMU_NBD_OPT_CACHE:
442 if (seen_cache) {
443 errx(EXIT_FAILURE, "-n and --cache can only be specified once");
444 }
445 seen_cache = true;
446 if (bdrv_parse_cache_flags(optarg, &flags) == -1) {
447 errx(EXIT_FAILURE, "Invalid cache mode `%s'", optarg);
448 }
449 break;
450 #ifdef CONFIG_LINUX_AIO
451 case QEMU_NBD_OPT_AIO:
452 if (seen_aio) {
453 errx(EXIT_FAILURE, "--aio can only be specified once");
454 }
455 seen_aio = true;
456 if (!strcmp(optarg, "native")) {
457 flags |= BDRV_O_NATIVE_AIO;
458 } else if (!strcmp(optarg, "threads")) {
459 /* this is the default */
460 } else {
461 errx(EXIT_FAILURE, "invalid aio mode `%s'", optarg);
462 }
463 break;
464 #endif
465 case QEMU_NBD_OPT_DISCARD:
466 if (seen_discard) {
467 errx(EXIT_FAILURE, "--discard can only be specified once");
468 }
469 seen_discard = true;
470 if (bdrv_parse_discard_flags(optarg, &flags) == -1) {
471 errx(EXIT_FAILURE, "Invalid discard mode `%s'", optarg);
472 }
473 break;
474 case 'b':
475 bindto = optarg;
476 break;
477 case 'p':
478 li = strtol(optarg, &end, 0);
479 if (*end) {
480 errx(EXIT_FAILURE, "Invalid port `%s'", optarg);
481 }
482 if (li < 1 || li > 65535) {
483 errx(EXIT_FAILURE, "Port out of range `%s'", optarg);
484 }
485 port = (uint16_t)li;
486 break;
487 case 'o':
488 dev_offset = strtoll (optarg, &end, 0);
489 if (*end) {
490 errx(EXIT_FAILURE, "Invalid offset `%s'", optarg);
491 }
492 if (dev_offset < 0) {
493 errx(EXIT_FAILURE, "Offset must be positive `%s'", optarg);
494 }
495 break;
496 case 'l':
497 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
498 sn_opts = qemu_opts_parse(&internal_snapshot_opts, optarg, 0);
499 if (!sn_opts) {
500 errx(EXIT_FAILURE, "Failed in parsing snapshot param `%s'",
501 optarg);
502 }
503 } else {
504 sn_id_or_name = optarg;
505 }
506 /* fall through */
507 case 'r':
508 nbdflags |= NBD_FLAG_READ_ONLY;
509 flags &= ~BDRV_O_RDWR;
510 break;
511 case 'P':
512 partition = strtol(optarg, &end, 0);
513 if (*end)
514 errx(EXIT_FAILURE, "Invalid partition `%s'", optarg);
515 if (partition < 1 || partition > 8)
516 errx(EXIT_FAILURE, "Invalid partition %d", partition);
517 break;
518 case 'k':
519 sockpath = optarg;
520 if (sockpath[0] != '/')
521 errx(EXIT_FAILURE, "socket path must be absolute\n");
522 break;
523 case 'd':
524 disconnect = true;
525 break;
526 case 'c':
527 device = optarg;
528 break;
529 case 'e':
530 shared = strtol(optarg, &end, 0);
531 if (*end) {
532 errx(EXIT_FAILURE, "Invalid shared device number '%s'", optarg);
533 }
534 if (shared < 1) {
535 errx(EXIT_FAILURE, "Shared device number must be greater than 0\n");
536 }
537 break;
538 case 'f':
539 fmt = optarg;
540 break;
541 case 't':
542 persistent = 1;
543 break;
544 case 'v':
545 verbose = 1;
546 break;
547 case 'V':
548 version(argv[0]);
549 exit(0);
550 break;
551 case 'h':
552 usage(argv[0]);
553 exit(0);
554 break;
555 case '?':
556 errx(EXIT_FAILURE, "Try `%s --help' for more information.",
557 argv[0]);
558 }
559 }
560
561 if ((argc - optind) != 1) {
562 errx(EXIT_FAILURE, "Invalid number of argument.\n"
563 "Try `%s --help' for more information.",
564 argv[0]);
565 }
566
567 if (disconnect) {
568 fd = open(argv[optind], O_RDWR);
569 if (fd < 0) {
570 err(EXIT_FAILURE, "Cannot open %s", argv[optind]);
571 }
572 nbd_disconnect(fd);
573
574 close(fd);
575
576 printf("%s disconnected\n", argv[optind]);
577
578 return 0;
579 }
580
581 if (device && !verbose) {
582 int stderr_fd[2];
583 pid_t pid;
584 int ret;
585
586 if (qemu_pipe(stderr_fd) < 0) {
587 err(EXIT_FAILURE, "Error setting up communication pipe");
588 }
589
590 /* Now daemonize, but keep a communication channel open to
591 * print errors and exit with the proper status code.
592 */
593 pid = fork();
594 if (pid == 0) {
595 close(stderr_fd[0]);
596 ret = qemu_daemon(1, 0);
597
598 /* Temporarily redirect stderr to the parent's pipe... */
599 dup2(stderr_fd[1], STDERR_FILENO);
600 if (ret < 0) {
601 err(EXIT_FAILURE, "Failed to daemonize");
602 }
603
604 /* ... close the descriptor we inherited and go on. */
605 close(stderr_fd[1]);
606 } else {
607 bool errors = false;
608 char *buf;
609
610 /* In the parent. Print error messages from the child until
611 * it closes the pipe.
612 */
613 close(stderr_fd[1]);
614 buf = g_malloc(1024);
615 while ((ret = read(stderr_fd[0], buf, 1024)) > 0) {
616 errors = true;
617 ret = qemu_write_full(STDERR_FILENO, buf, ret);
618 if (ret < 0) {
619 exit(EXIT_FAILURE);
620 }
621 }
622 if (ret < 0) {
623 err(EXIT_FAILURE, "Cannot read from daemon");
624 }
625
626 /* Usually the daemon should not print any message.
627 * Exit with zero status in that case.
628 */
629 exit(errors);
630 }
631 }
632
633 if (device != NULL && sockpath == NULL) {
634 sockpath = g_malloc(128);
635 snprintf(sockpath, 128, SOCKET_PATH, basename(device));
636 }
637
638 qemu_init_main_loop();
639 bdrv_init();
640 atexit(bdrv_close_all);
641
642 if (fmt) {
643 drv = bdrv_find_format(fmt);
644 if (!drv) {
645 errx(EXIT_FAILURE, "Unknown file format '%s'", fmt);
646 }
647 } else {
648 drv = NULL;
649 }
650
651 bs = bdrv_new("hda");
652 srcpath = argv[optind];
653 ret = bdrv_open(&bs, srcpath, NULL, NULL, flags, drv, &local_err);
654 if (ret < 0) {
655 errno = -ret;
656 err(EXIT_FAILURE, "Failed to bdrv_open '%s': %s", argv[optind],
657 error_get_pretty(local_err));
658 }
659
660 if (sn_opts) {
661 ret = bdrv_snapshot_load_tmp(bs,
662 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
663 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
664 &local_err);
665 } else if (sn_id_or_name) {
666 ret = bdrv_snapshot_load_tmp_by_id_or_name(bs, sn_id_or_name,
667 &local_err);
668 }
669 if (ret < 0) {
670 errno = -ret;
671 err(EXIT_FAILURE,
672 "Failed to load snapshot: %s",
673 error_get_pretty(local_err));
674 }
675
676 fd_size = bdrv_getlength(bs);
677
678 if (partition != -1) {
679 ret = find_partition(bs, partition, &dev_offset, &fd_size);
680 if (ret < 0) {
681 errno = -ret;
682 err(EXIT_FAILURE, "Could not find partition %d", partition);
683 }
684 }
685
686 exp = nbd_export_new(bs, dev_offset, fd_size, nbdflags, nbd_export_closed);
687
688 if (sockpath) {
689 fd = unix_socket_incoming(sockpath);
690 } else {
691 fd = tcp_socket_incoming(bindto, port);
692 }
693
694 if (fd < 0) {
695 return 1;
696 }
697
698 if (device) {
699 int ret;
700
701 ret = pthread_create(&client_thread, NULL, nbd_client_thread, device);
702 if (ret != 0) {
703 errx(EXIT_FAILURE, "Failed to create client thread: %s",
704 strerror(ret));
705 }
706 } else {
707 /* Shut up GCC warnings. */
708 memset(&client_thread, 0, sizeof(client_thread));
709 }
710
711 qemu_set_fd_handler2(fd, nbd_can_accept, nbd_accept, NULL,
712 (void *)(uintptr_t)fd);
713
714 /* now when the initialization is (almost) complete, chdir("/")
715 * to free any busy filesystems */
716 if (chdir("/") < 0) {
717 err(EXIT_FAILURE, "Could not chdir to root directory");
718 }
719
720 state = RUNNING;
721 do {
722 main_loop_wait(false);
723 if (state == TERMINATE) {
724 state = TERMINATING;
725 nbd_export_close(exp);
726 nbd_export_put(exp);
727 exp = NULL;
728 }
729 } while (state != TERMINATED);
730
731 bdrv_close(bs);
732 if (sockpath) {
733 unlink(sockpath);
734 }
735
736 if (sn_opts) {
737 qemu_opts_del(sn_opts);
738 }
739
740 if (device) {
741 void *ret;
742 pthread_join(client_thread, &ret);
743 exit(ret != NULL);
744 } else {
745 exit(EXIT_SUCCESS);
746 }
747 }