migration/postcopy: discard_length must not be 0
[qemu/ar7.git] / qemu-nbd.c
bloba8cb39e51043a9f2731e1c8df26cfe1d7a3835ee
1 /*
2 * Copyright (C) 2005 Anthony Liguori <anthony@codemonkey.ws>
4 * Network Block Device
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 <getopt.h>
21 #include <libgen.h>
22 #include <pthread.h>
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"
48 #ifdef __linux__
49 #define HAVE_NBD_DEVICE 1
50 #else
51 #define HAVE_NBD_DEVICE 0
52 #endif
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
66 #define MBR_SIZE 512
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;
80 static void usage(const char *name)
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");
147 static void version(const char *name)
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);
159 struct partition_record
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;
173 static void read_partition(uint8_t *p, struct partition_record *r)
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;
184 r->start_sector_abs = ldl_le_p(p + 8);
185 r->nb_sectors_abs = ldl_le_p(p + 12);
188 static int find_partition(BlockBackend *blk, int partition,
189 uint64_t *offset, uint64_t *size)
191 struct partition_record mbr[4];
192 uint8_t data[MBR_SIZE];
193 int i;
194 int ext_partnum = 4;
195 int ret;
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);
203 if (data[510] != 0x55 || data[511] != 0xaa) {
204 return -EINVAL;
207 for (i = 0; i < 4; i++) {
208 read_partition(&data[446 + 16 * i], &mbr[i]);
210 if (!mbr[i].system || !mbr[i].nb_sectors_abs) {
211 continue;
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;
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);
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;
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;
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;
246 return -ENOENT;
249 static void termsig_handler(int signum)
251 atomic_cmpxchg(&state, RUNNING, TERMINATE);
252 qemu_notify_event();
256 static int qemu_nbd_client_list(SocketAddress *saddr, QCryptoTLSCreds *tls,
257 const char *hostname)
259 int ret = EXIT_FAILURE;
260 int rc;
261 Error *err = NULL;
262 QIOChannelSocket *sioc;
263 NBDExportInfo *list;
264 int i, j;
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;
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);
277 goto out;
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);
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",
299 printf(" size: %" PRIu64 "\n", list[i].size);
300 printf(" flags: 0x%x (", list[i].flags);
301 for (size_t bit = 0; bit < ARRAY_SIZE(flag_names); bit++) {
302 if (flag_names[bit] && (list[i].flags & (1 << bit))) {
303 printf(" %s", flag_names[bit]);
306 printf(" )\n");
308 if (list[i].min_block) {
309 printf(" min block: %u\n", list[i].min_block);
310 printf(" opt block: %u\n", list[i].opt_block);
311 printf(" max block: %u\n", list[i].max_block);
313 if (list[i].n_contexts) {
314 printf(" available meta contexts: %d\n", list[i].n_contexts);
315 for (j = 0; j < list[i].n_contexts; j++) {
316 printf(" %s\n", list[i].contexts[j]);
320 nbd_free_export_list(list, rc);
322 ret = EXIT_SUCCESS;
323 out:
324 object_unref(OBJECT(sioc));
325 return ret;
329 #if HAVE_NBD_DEVICE
330 static void *show_parts(void *arg)
332 char *device = arg;
333 int nbd;
335 /* linux just needs an open() to trigger
336 * the partition table update
337 * but remember to load the module with max_part != 0 :
338 * modprobe nbd max_part=63
340 nbd = open(device, O_RDWR);
341 if (nbd >= 0) {
342 close(nbd);
344 return NULL;
347 static void *nbd_client_thread(void *arg)
349 char *device = arg;
350 NBDExportInfo info = { .request_sizes = false, .name = g_strdup("") };
351 QIOChannelSocket *sioc;
352 int fd;
353 int ret;
354 pthread_t show_parts_thread;
355 Error *local_error = NULL;
357 sioc = qio_channel_socket_new();
358 if (qio_channel_socket_connect_sync(sioc,
359 saddr,
360 &local_error) < 0) {
361 error_report_err(local_error);
362 goto out;
365 ret = nbd_receive_negotiate(QIO_CHANNEL(sioc),
366 NULL, NULL, NULL, &info, &local_error);
367 if (ret < 0) {
368 if (local_error) {
369 error_report_err(local_error);
371 goto out_socket;
374 fd = open(device, O_RDWR);
375 if (fd < 0) {
376 /* Linux-only, we can use %m in printf. */
377 error_report("Failed to open %s: %m", device);
378 goto out_socket;
381 ret = nbd_init(fd, sioc, &info, &local_error);
382 if (ret < 0) {
383 error_report_err(local_error);
384 goto out_fd;
387 /* update partition table */
388 pthread_create(&show_parts_thread, NULL, show_parts, device);
390 if (verbose) {
391 fprintf(stderr, "NBD device %s is now connected to %s\n",
392 device, srcpath);
393 } else {
394 /* Close stderr so that the qemu-nbd process exits. */
395 dup2(STDOUT_FILENO, STDERR_FILENO);
398 ret = nbd_client(fd);
399 if (ret) {
400 goto out_fd;
402 close(fd);
403 object_unref(OBJECT(sioc));
404 g_free(info.name);
405 kill(getpid(), SIGTERM);
406 return (void *) EXIT_SUCCESS;
408 out_fd:
409 close(fd);
410 out_socket:
411 object_unref(OBJECT(sioc));
412 out:
413 g_free(info.name);
414 kill(getpid(), SIGTERM);
415 return (void *) EXIT_FAILURE;
417 #endif /* HAVE_NBD_DEVICE */
419 static int nbd_can_accept(void)
421 return state == RUNNING && nb_fds < shared;
424 static void nbd_export_closed(NBDExport *export)
426 assert(state == TERMINATING);
427 state = TERMINATED;
430 static void nbd_update_server_watch(void);
432 static void nbd_client_closed(NBDClient *client, bool negotiated)
434 nb_fds--;
435 if (negotiated && nb_fds == 0 && !persistent && state == RUNNING) {
436 state = TERMINATE;
438 nbd_update_server_watch();
439 nbd_client_put(client);
442 static void nbd_accept(QIONetListener *listener, QIOChannelSocket *cioc,
443 gpointer opaque)
445 if (state >= TERMINATE) {
446 return;
449 nb_fds++;
450 nbd_update_server_watch();
451 nbd_client_new(cioc, tlscreds, tlsauthz, nbd_client_closed);
454 static void nbd_update_server_watch(void)
456 if (nbd_can_accept()) {
457 qio_net_listener_set_client_func(server, nbd_accept, NULL, NULL);
458 } else {
459 qio_net_listener_set_client_func(server, NULL, NULL, NULL);
464 static SocketAddress *nbd_build_socket_address(const char *sockpath,
465 const char *bindto,
466 const char *port)
468 SocketAddress *saddr;
470 saddr = g_new0(SocketAddress, 1);
471 if (sockpath) {
472 saddr->type = SOCKET_ADDRESS_TYPE_UNIX;
473 saddr->u.q_unix.path = g_strdup(sockpath);
474 } else {
475 InetSocketAddress *inet;
476 saddr->type = SOCKET_ADDRESS_TYPE_INET;
477 inet = &saddr->u.inet;
478 inet->host = g_strdup(bindto);
479 if (port) {
480 inet->port = g_strdup(port);
481 } else {
482 inet->port = g_strdup_printf("%d", NBD_DEFAULT_PORT);
486 return saddr;
490 static QemuOptsList file_opts = {
491 .name = "file",
492 .implied_opt_name = "file",
493 .head = QTAILQ_HEAD_INITIALIZER(file_opts.head),
494 .desc = {
495 /* no elements => accept any params */
496 { /* end of list */ }
500 static QemuOptsList qemu_object_opts = {
501 .name = "object",
502 .implied_opt_name = "qom-type",
503 .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head),
504 .desc = {
511 static QCryptoTLSCreds *nbd_get_tls_creds(const char *id, bool list,
512 Error **errp)
514 Object *obj;
515 QCryptoTLSCreds *creds;
517 obj = object_resolve_path_component(
518 object_get_objects_root(), id);
519 if (!obj) {
520 error_setg(errp, "No TLS credentials with id '%s'",
521 id);
522 return NULL;
524 creds = (QCryptoTLSCreds *)
525 object_dynamic_cast(obj, TYPE_QCRYPTO_TLS_CREDS);
526 if (!creds) {
527 error_setg(errp, "Object with id '%s' is not TLS credentials",
528 id);
529 return NULL;
532 if (list) {
533 if (creds->endpoint != QCRYPTO_TLS_CREDS_ENDPOINT_CLIENT) {
534 error_setg(errp,
535 "Expecting TLS credentials with a client endpoint");
536 return NULL;
538 } else {
539 if (creds->endpoint != QCRYPTO_TLS_CREDS_ENDPOINT_SERVER) {
540 error_setg(errp,
541 "Expecting TLS credentials with a server endpoint");
542 return NULL;
545 object_ref(obj);
546 return creds;
549 static void setup_address_and_port(const char **address, const char **port)
551 if (*address == NULL) {
552 *address = "0.0.0.0";
555 if (*port == NULL) {
556 *port = stringify(NBD_DEFAULT_PORT);
561 * Check socket parameters compatibility when socket activation is used.
563 static const char *socket_activation_validate_opts(const char *device,
564 const char *sockpath,
565 const char *address,
566 const char *port,
567 bool list)
569 if (device != NULL) {
570 return "NBD device can't be set when using socket activation";
573 if (sockpath != NULL) {
574 return "Unix socket can't be set when using socket activation";
577 if (address != NULL) {
578 return "The interface can't be set when using socket activation";
581 if (port != NULL) {
582 return "TCP port number can't be set when using socket activation";
585 if (list) {
586 return "List mode is incompatible with socket activation";
589 return NULL;
592 static void qemu_nbd_shutdown(void)
594 job_cancel_sync_all();
595 bdrv_close_all();
598 int main(int argc, char **argv)
600 BlockBackend *blk;
601 BlockDriverState *bs;
602 uint64_t dev_offset = 0;
603 uint16_t nbdflags = 0;
604 bool disconnect = false;
605 const char *bindto = NULL;
606 const char *port = NULL;
607 char *sockpath = NULL;
608 char *device = NULL;
609 int64_t fd_size;
610 QemuOpts *sn_opts = NULL;
611 const char *sn_id_or_name = NULL;
612 const char *sopt = "hVb:o:p:rsnP:c:dvk:e:f:tl:x:T:D:B:L";
613 struct option lopt[] = {
614 { "help", no_argument, NULL, 'h' },
615 { "version", no_argument, NULL, 'V' },
616 { "bind", required_argument, NULL, 'b' },
617 { "port", required_argument, NULL, 'p' },
618 { "socket", required_argument, NULL, 'k' },
619 { "offset", required_argument, NULL, 'o' },
620 { "read-only", no_argument, NULL, 'r' },
621 { "partition", required_argument, NULL, 'P' },
622 { "bitmap", required_argument, NULL, 'B' },
623 { "connect", required_argument, NULL, 'c' },
624 { "disconnect", no_argument, NULL, 'd' },
625 { "list", no_argument, NULL, 'L' },
626 { "snapshot", no_argument, NULL, 's' },
627 { "load-snapshot", required_argument, NULL, 'l' },
628 { "nocache", no_argument, NULL, 'n' },
629 { "cache", required_argument, NULL, QEMU_NBD_OPT_CACHE },
630 { "aio", required_argument, NULL, QEMU_NBD_OPT_AIO },
631 { "discard", required_argument, NULL, QEMU_NBD_OPT_DISCARD },
632 { "detect-zeroes", required_argument, NULL,
633 QEMU_NBD_OPT_DETECT_ZEROES },
634 { "shared", required_argument, NULL, 'e' },
635 { "format", required_argument, NULL, 'f' },
636 { "persistent", no_argument, NULL, 't' },
637 { "verbose", no_argument, NULL, 'v' },
638 { "object", required_argument, NULL, QEMU_NBD_OPT_OBJECT },
639 { "export-name", required_argument, NULL, 'x' },
640 { "description", required_argument, NULL, 'D' },
641 { "tls-creds", required_argument, NULL, QEMU_NBD_OPT_TLSCREDS },
642 { "tls-authz", required_argument, NULL, QEMU_NBD_OPT_TLSAUTHZ },
643 { "image-opts", no_argument, NULL, QEMU_NBD_OPT_IMAGE_OPTS },
644 { "trace", required_argument, NULL, 'T' },
645 { "fork", no_argument, NULL, QEMU_NBD_OPT_FORK },
646 { "pid-file", required_argument, NULL, QEMU_NBD_OPT_PID_FILE },
647 { NULL, 0, NULL, 0 }
649 int ch;
650 int opt_ind = 0;
651 int flags = BDRV_O_RDWR;
652 int partition = 0;
653 int ret = 0;
654 bool seen_cache = false;
655 bool seen_discard = false;
656 bool seen_aio = false;
657 pthread_t client_thread;
658 const char *fmt = NULL;
659 Error *local_err = NULL;
660 BlockdevDetectZeroesOptions detect_zeroes = BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
661 QDict *options = NULL;
662 const char *export_name = NULL; /* defaults to "" later for server mode */
663 const char *export_description = NULL;
664 const char *bitmap = NULL;
665 const char *tlscredsid = NULL;
666 bool imageOpts = false;
667 bool writethrough = true;
668 char *trace_file = NULL;
669 bool fork_process = false;
670 bool list = false;
671 int old_stderr = -1;
672 unsigned socket_activation;
673 const char *pid_file_name = NULL;
675 /* The client thread uses SIGTERM to interrupt the server. A signal
676 * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
678 struct sigaction sa_sigterm;
679 memset(&sa_sigterm, 0, sizeof(sa_sigterm));
680 sa_sigterm.sa_handler = termsig_handler;
681 sigaction(SIGTERM, &sa_sigterm, NULL);
683 #ifdef CONFIG_POSIX
684 signal(SIGPIPE, SIG_IGN);
685 #endif
687 error_init(argv[0]);
688 module_call_init(MODULE_INIT_TRACE);
689 qcrypto_init(&error_fatal);
691 module_call_init(MODULE_INIT_QOM);
692 qemu_add_opts(&qemu_object_opts);
693 qemu_add_opts(&qemu_trace_opts);
694 qemu_init_exec_dir(argv[0]);
696 while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
697 switch (ch) {
698 case 's':
699 flags |= BDRV_O_SNAPSHOT;
700 break;
701 case 'n':
702 optarg = (char *) "none";
703 /* fallthrough */
704 case QEMU_NBD_OPT_CACHE:
705 if (seen_cache) {
706 error_report("-n and --cache can only be specified once");
707 exit(EXIT_FAILURE);
709 seen_cache = true;
710 if (bdrv_parse_cache_mode(optarg, &flags, &writethrough) == -1) {
711 error_report("Invalid cache mode `%s'", optarg);
712 exit(EXIT_FAILURE);
714 break;
715 case QEMU_NBD_OPT_AIO:
716 if (seen_aio) {
717 error_report("--aio can only be specified once");
718 exit(EXIT_FAILURE);
720 seen_aio = true;
721 if (!strcmp(optarg, "native")) {
722 flags |= BDRV_O_NATIVE_AIO;
723 } else if (!strcmp(optarg, "threads")) {
724 /* this is the default */
725 } else {
726 error_report("invalid aio mode `%s'", optarg);
727 exit(EXIT_FAILURE);
729 break;
730 case QEMU_NBD_OPT_DISCARD:
731 if (seen_discard) {
732 error_report("--discard can only be specified once");
733 exit(EXIT_FAILURE);
735 seen_discard = true;
736 if (bdrv_parse_discard_flags(optarg, &flags) == -1) {
737 error_report("Invalid discard mode `%s'", optarg);
738 exit(EXIT_FAILURE);
740 break;
741 case QEMU_NBD_OPT_DETECT_ZEROES:
742 detect_zeroes =
743 qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup,
744 optarg,
745 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
746 &local_err);
747 if (local_err) {
748 error_reportf_err(local_err,
749 "Failed to parse detect_zeroes mode: ");
750 exit(EXIT_FAILURE);
752 if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
753 !(flags & BDRV_O_UNMAP)) {
754 error_report("setting detect-zeroes to unmap is not allowed "
755 "without setting discard operation to unmap");
756 exit(EXIT_FAILURE);
758 break;
759 case 'b':
760 bindto = optarg;
761 break;
762 case 'p':
763 port = optarg;
764 break;
765 case 'o':
766 if (qemu_strtou64(optarg, NULL, 0, &dev_offset) < 0) {
767 error_report("Invalid offset '%s'", optarg);
768 exit(EXIT_FAILURE);
770 break;
771 case 'l':
772 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
773 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
774 optarg, false);
775 if (!sn_opts) {
776 error_report("Failed in parsing snapshot param `%s'",
777 optarg);
778 exit(EXIT_FAILURE);
780 } else {
781 sn_id_or_name = optarg;
783 /* fall through */
784 case 'r':
785 nbdflags |= NBD_FLAG_READ_ONLY;
786 flags &= ~BDRV_O_RDWR;
787 break;
788 case 'P':
789 warn_report("The '-P' option is deprecated; use --image-opts with "
790 "a raw device wrapper for subset exports instead");
791 if (qemu_strtoi(optarg, NULL, 0, &partition) < 0 ||
792 partition < 1 || partition > 8) {
793 error_report("Invalid partition '%s'", optarg);
794 exit(EXIT_FAILURE);
796 break;
797 case 'B':
798 bitmap = optarg;
799 break;
800 case 'k':
801 sockpath = optarg;
802 if (sockpath[0] != '/') {
803 error_report("socket path must be absolute");
804 exit(EXIT_FAILURE);
806 break;
807 case 'd':
808 disconnect = true;
809 break;
810 case 'c':
811 device = optarg;
812 break;
813 case 'e':
814 if (qemu_strtoi(optarg, NULL, 0, &shared) < 0 ||
815 shared < 1) {
816 error_report("Invalid shared device number '%s'", optarg);
817 exit(EXIT_FAILURE);
819 break;
820 case 'f':
821 fmt = optarg;
822 break;
823 case 't':
824 persistent = 1;
825 break;
826 case 'x':
827 export_name = optarg;
828 break;
829 case 'D':
830 export_description = optarg;
831 break;
832 case 'v':
833 verbose = 1;
834 break;
835 case 'V':
836 version(argv[0]);
837 exit(0);
838 break;
839 case 'h':
840 usage(argv[0]);
841 exit(0);
842 break;
843 case '?':
844 error_report("Try `%s --help' for more information.", argv[0]);
845 exit(EXIT_FAILURE);
846 case QEMU_NBD_OPT_OBJECT: {
847 QemuOpts *opts;
848 opts = qemu_opts_parse_noisily(&qemu_object_opts,
849 optarg, true);
850 if (!opts) {
851 exit(EXIT_FAILURE);
853 } break;
854 case QEMU_NBD_OPT_TLSCREDS:
855 tlscredsid = optarg;
856 break;
857 case QEMU_NBD_OPT_IMAGE_OPTS:
858 imageOpts = true;
859 break;
860 case 'T':
861 g_free(trace_file);
862 trace_file = trace_opt_parse(optarg);
863 break;
864 case QEMU_NBD_OPT_TLSAUTHZ:
865 tlsauthz = optarg;
866 break;
867 case QEMU_NBD_OPT_FORK:
868 fork_process = true;
869 break;
870 case 'L':
871 list = true;
872 break;
873 case QEMU_NBD_OPT_PID_FILE:
874 pid_file_name = optarg;
875 break;
879 if (list) {
880 if (argc != optind) {
881 error_report("List mode is incompatible with a file name");
882 exit(EXIT_FAILURE);
884 if (export_name || export_description || dev_offset || partition ||
885 device || disconnect || fmt || sn_id_or_name || bitmap ||
886 seen_aio || seen_discard || seen_cache) {
887 error_report("List mode is incompatible with per-device settings");
888 exit(EXIT_FAILURE);
890 if (fork_process) {
891 error_report("List mode is incompatible with forking");
892 exit(EXIT_FAILURE);
894 } else if ((argc - optind) != 1) {
895 error_report("Invalid number of arguments");
896 error_printf("Try `%s --help' for more information.\n", argv[0]);
897 exit(EXIT_FAILURE);
898 } else if (!export_name) {
899 export_name = "";
902 qemu_opts_foreach(&qemu_object_opts,
903 user_creatable_add_opts_foreach,
904 NULL, &error_fatal);
906 if (!trace_init_backends()) {
907 exit(1);
909 trace_init_file(trace_file);
910 qemu_set_log(LOG_TRACE);
912 socket_activation = check_socket_activation();
913 if (socket_activation == 0) {
914 setup_address_and_port(&bindto, &port);
915 } else {
916 /* Using socket activation - check user didn't use -p etc. */
917 const char *err_msg = socket_activation_validate_opts(device, sockpath,
918 bindto, port,
919 list);
920 if (err_msg != NULL) {
921 error_report("%s", err_msg);
922 exit(EXIT_FAILURE);
925 /* qemu-nbd can only listen on a single socket. */
926 if (socket_activation > 1) {
927 error_report("qemu-nbd does not support socket activation with %s > 1",
928 "LISTEN_FDS");
929 exit(EXIT_FAILURE);
933 if (tlscredsid) {
934 if (sockpath) {
935 error_report("TLS is only supported with IPv4/IPv6");
936 exit(EXIT_FAILURE);
938 if (device) {
939 error_report("TLS is not supported with a host device");
940 exit(EXIT_FAILURE);
942 if (tlsauthz && list) {
943 error_report("TLS authorization is incompatible with export list");
944 exit(EXIT_FAILURE);
946 tlscreds = nbd_get_tls_creds(tlscredsid, list, &local_err);
947 if (local_err) {
948 error_report("Failed to get TLS creds %s",
949 error_get_pretty(local_err));
950 exit(EXIT_FAILURE);
952 } else {
953 if (tlsauthz) {
954 error_report("--tls-authz is not permitted without --tls-creds");
955 exit(EXIT_FAILURE);
959 if (list) {
960 saddr = nbd_build_socket_address(sockpath, bindto, port);
961 return qemu_nbd_client_list(saddr, tlscreds, bindto);
964 #if !HAVE_NBD_DEVICE
965 if (disconnect || device) {
966 error_report("Kernel /dev/nbdN support not available");
967 exit(EXIT_FAILURE);
969 #else /* HAVE_NBD_DEVICE */
970 if (disconnect) {
971 int nbdfd = open(argv[optind], O_RDWR);
972 if (nbdfd < 0) {
973 error_report("Cannot open %s: %s", argv[optind],
974 strerror(errno));
975 exit(EXIT_FAILURE);
977 nbd_disconnect(nbdfd);
979 close(nbdfd);
981 printf("%s disconnected\n", argv[optind]);
983 return 0;
985 #endif
987 if ((device && !verbose) || fork_process) {
988 int stderr_fd[2];
989 pid_t pid;
990 int ret;
992 if (qemu_pipe(stderr_fd) < 0) {
993 error_report("Error setting up communication pipe: %s",
994 strerror(errno));
995 exit(EXIT_FAILURE);
998 /* Now daemonize, but keep a communication channel open to
999 * print errors and exit with the proper status code.
1001 pid = fork();
1002 if (pid < 0) {
1003 error_report("Failed to fork: %s", strerror(errno));
1004 exit(EXIT_FAILURE);
1005 } else if (pid == 0) {
1006 close(stderr_fd[0]);
1008 old_stderr = dup(STDERR_FILENO);
1009 ret = qemu_daemon(1, 0);
1011 /* Temporarily redirect stderr to the parent's pipe... */
1012 dup2(stderr_fd[1], STDERR_FILENO);
1013 if (ret < 0) {
1014 error_report("Failed to daemonize: %s", strerror(errno));
1015 exit(EXIT_FAILURE);
1018 /* ... close the descriptor we inherited and go on. */
1019 close(stderr_fd[1]);
1020 } else {
1021 bool errors = false;
1022 char *buf;
1024 /* In the parent. Print error messages from the child until
1025 * it closes the pipe.
1027 close(stderr_fd[1]);
1028 buf = g_malloc(1024);
1029 while ((ret = read(stderr_fd[0], buf, 1024)) > 0) {
1030 errors = true;
1031 ret = qemu_write_full(STDERR_FILENO, buf, ret);
1032 if (ret < 0) {
1033 exit(EXIT_FAILURE);
1036 if (ret < 0) {
1037 error_report("Cannot read from daemon: %s",
1038 strerror(errno));
1039 exit(EXIT_FAILURE);
1042 /* Usually the daemon should not print any message.
1043 * Exit with zero status in that case.
1045 exit(errors);
1049 if (device != NULL && sockpath == NULL) {
1050 sockpath = g_malloc(128);
1051 snprintf(sockpath, 128, SOCKET_PATH, basename(device));
1054 server = qio_net_listener_new();
1055 if (socket_activation == 0) {
1056 saddr = nbd_build_socket_address(sockpath, bindto, port);
1057 if (qio_net_listener_open_sync(server, saddr, &local_err) < 0) {
1058 object_unref(OBJECT(server));
1059 error_report_err(local_err);
1060 exit(EXIT_FAILURE);
1062 } else {
1063 size_t i;
1064 /* See comment in check_socket_activation above. */
1065 for (i = 0; i < socket_activation; i++) {
1066 QIOChannelSocket *sioc;
1067 sioc = qio_channel_socket_new_fd(FIRST_SOCKET_ACTIVATION_FD + i,
1068 &local_err);
1069 if (sioc == NULL) {
1070 object_unref(OBJECT(server));
1071 error_report("Failed to use socket activation: %s",
1072 error_get_pretty(local_err));
1073 exit(EXIT_FAILURE);
1075 qio_net_listener_add(server, sioc);
1076 object_unref(OBJECT(sioc));
1080 if (qemu_init_main_loop(&local_err)) {
1081 error_report_err(local_err);
1082 exit(EXIT_FAILURE);
1084 bdrv_init();
1085 atexit(qemu_nbd_shutdown);
1087 srcpath = argv[optind];
1088 if (imageOpts) {
1089 QemuOpts *opts;
1090 if (fmt) {
1091 error_report("--image-opts and -f are mutually exclusive");
1092 exit(EXIT_FAILURE);
1094 opts = qemu_opts_parse_noisily(&file_opts, srcpath, true);
1095 if (!opts) {
1096 qemu_opts_reset(&file_opts);
1097 exit(EXIT_FAILURE);
1099 options = qemu_opts_to_qdict(opts, NULL);
1100 qemu_opts_reset(&file_opts);
1101 blk = blk_new_open(NULL, NULL, options, flags, &local_err);
1102 } else {
1103 if (fmt) {
1104 options = qdict_new();
1105 qdict_put_str(options, "driver", fmt);
1107 blk = blk_new_open(srcpath, NULL, options, flags, &local_err);
1110 if (!blk) {
1111 error_reportf_err(local_err, "Failed to blk_new_open '%s': ",
1112 argv[optind]);
1113 exit(EXIT_FAILURE);
1115 bs = blk_bs(blk);
1117 blk_set_enable_write_cache(blk, !writethrough);
1119 if (sn_opts) {
1120 ret = bdrv_snapshot_load_tmp(bs,
1121 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
1122 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
1123 &local_err);
1124 } else if (sn_id_or_name) {
1125 ret = bdrv_snapshot_load_tmp_by_id_or_name(bs, sn_id_or_name,
1126 &local_err);
1128 if (ret < 0) {
1129 error_reportf_err(local_err, "Failed to load snapshot: ");
1130 exit(EXIT_FAILURE);
1133 bs->detect_zeroes = detect_zeroes;
1134 fd_size = blk_getlength(blk);
1135 if (fd_size < 0) {
1136 error_report("Failed to determine the image length: %s",
1137 strerror(-fd_size));
1138 exit(EXIT_FAILURE);
1141 if (dev_offset >= fd_size) {
1142 error_report("Offset (%" PRIu64 ") has to be smaller than the image "
1143 "size (%" PRId64 ")", dev_offset, fd_size);
1144 exit(EXIT_FAILURE);
1146 fd_size -= dev_offset;
1148 if (partition) {
1149 uint64_t limit;
1151 if (dev_offset) {
1152 error_report("Cannot request partition and offset together");
1153 exit(EXIT_FAILURE);
1155 ret = find_partition(blk, partition, &dev_offset, &limit);
1156 if (ret < 0) {
1157 error_report("Could not find partition %d: %s", partition,
1158 strerror(-ret));
1159 exit(EXIT_FAILURE);
1162 * MBR partition limits are (32-bit << 9); this assert lets
1163 * the compiler know that we can't overflow 64 bits.
1165 assert(dev_offset + limit >= dev_offset);
1166 if (dev_offset + limit > fd_size) {
1167 error_report("Discovered partition %d at offset %" PRIu64
1168 " size %" PRIu64 ", but size exceeds file length %"
1169 PRId64, partition, dev_offset, limit, fd_size);
1170 exit(EXIT_FAILURE);
1172 fd_size = limit;
1175 export = nbd_export_new(bs, dev_offset, fd_size, export_name,
1176 export_description, bitmap, nbdflags,
1177 nbd_export_closed, writethrough, NULL,
1178 &error_fatal);
1180 if (device) {
1181 #if HAVE_NBD_DEVICE
1182 int ret;
1184 ret = pthread_create(&client_thread, NULL, nbd_client_thread, device);
1185 if (ret != 0) {
1186 error_report("Failed to create client thread: %s", strerror(ret));
1187 exit(EXIT_FAILURE);
1189 #endif
1190 } else {
1191 /* Shut up GCC warnings. */
1192 memset(&client_thread, 0, sizeof(client_thread));
1195 nbd_update_server_watch();
1197 if (pid_file_name) {
1198 qemu_write_pidfile(pid_file_name, &error_fatal);
1201 /* now when the initialization is (almost) complete, chdir("/")
1202 * to free any busy filesystems */
1203 if (chdir("/") < 0) {
1204 error_report("Could not chdir to root directory: %s",
1205 strerror(errno));
1206 exit(EXIT_FAILURE);
1209 if (fork_process) {
1210 dup2(old_stderr, STDERR_FILENO);
1211 close(old_stderr);
1214 state = RUNNING;
1215 do {
1216 main_loop_wait(false);
1217 if (state == TERMINATE) {
1218 state = TERMINATING;
1219 nbd_export_close(export);
1220 nbd_export_put(export);
1221 export = NULL;
1223 } while (state != TERMINATED);
1225 blk_unref(blk);
1226 if (sockpath) {
1227 unlink(sockpath);
1230 qemu_opts_del(sn_opts);
1232 if (device) {
1233 void *ret;
1234 pthread_join(client_thread, &ret);
1235 exit(ret != NULL);
1236 } else {
1237 exit(EXIT_SUCCESS);