xref: /openbmc/qemu/qemu-nbd.c (revision 9ab5a24b5298ae8272607a8ffc9ba20c15acefaa)
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/osdep.h"
20 #include <getopt.h>
21 #include <libgen.h>
22 #include <pthread.h>
23 
24 #include "qapi/error.h"
25 #include "qemu/cutils.h"
26 #include "sysemu/block-backend.h"
27 #include "block/block_int.h"
28 #include "block/nbd.h"
29 #include "qemu/main-loop.h"
30 #include "qemu/option.h"
31 #include "qemu/error-report.h"
32 #include "qemu/config-file.h"
33 #include "qemu/bswap.h"
34 #include "qemu/log.h"
35 #include "qemu/systemd.h"
36 #include "block/snapshot.h"
37 #include "qapi/qmp/qdict.h"
38 #include "qapi/qmp/qstring.h"
39 #include "qom/object_interfaces.h"
40 #include "io/channel-socket.h"
41 #include "io/net-listener.h"
42 #include "crypto/init.h"
43 #include "trace/control.h"
44 #include "qemu-version.h"
45 
46 #ifdef __linux__
47 #define HAVE_NBD_DEVICE 1
48 #else
49 #define HAVE_NBD_DEVICE 0
50 #endif
51 
52 #define SOCKET_PATH                "/var/lock/qemu-nbd-%s"
53 #define QEMU_NBD_OPT_CACHE         256
54 #define QEMU_NBD_OPT_AIO           257
55 #define QEMU_NBD_OPT_DISCARD       258
56 #define QEMU_NBD_OPT_DETECT_ZEROES 259
57 #define QEMU_NBD_OPT_OBJECT        260
58 #define QEMU_NBD_OPT_TLSCREDS      261
59 #define QEMU_NBD_OPT_IMAGE_OPTS    262
60 #define QEMU_NBD_OPT_FORK          263
61 
62 #define MBR_SIZE 512
63 
64 static NBDExport *export;
65 static int verbose;
66 static char *srcpath;
67 static SocketAddress *saddr;
68 static int persistent = 0;
69 static enum { RUNNING, TERMINATE, TERMINATING, TERMINATED } state;
70 static int shared = 1;
71 static int nb_fds;
72 static QIONetListener *server;
73 static QCryptoTLSCreds *tlscreds;
74 
75 static void usage(const char *name)
76 {
77     (printf) (
78 "Usage: %s [OPTIONS] FILE\n"
79 "QEMU Disk Network Block Device Server\n"
80 "\n"
81 "  -h, --help                display this help and exit\n"
82 "  -V, --version             output version information and exit\n"
83 "\n"
84 "Connection properties:\n"
85 "  -p, --port=PORT           port to listen on (default `%d')\n"
86 "  -b, --bind=IFACE          interface to bind to (default `0.0.0.0')\n"
87 "  -k, --socket=PATH         path to the unix socket\n"
88 "                            (default '"SOCKET_PATH"')\n"
89 "  -e, --shared=NUM          device can be shared by NUM clients (default '1')\n"
90 "  -t, --persistent          don't exit on the last connection\n"
91 "  -v, --verbose             display extra debugging information\n"
92 "  -x, --export-name=NAME    expose export by name (default is empty string)\n"
93 "  -D, --description=TEXT    export a human-readable description\n"
94 "\n"
95 "Exposing part of the image:\n"
96 "  -o, --offset=OFFSET       offset into the image\n"
97 "  -P, --partition=NUM       only expose partition NUM\n"
98 "  -B, --bitmap=NAME         expose a persistent dirty bitmap\n"
99 "\n"
100 "General purpose options:\n"
101 "  --object type,id=ID,...   define an object such as 'secret' for providing\n"
102 "                            passwords and/or encryption keys\n"
103 "  --tls-creds=ID            use id of an earlier --object to provide TLS\n"
104 "  -T, --trace [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
105 "                            specify tracing options\n"
106 "  --fork                    fork off the server process and exit the parent\n"
107 "                            once the server is running\n"
108 #if HAVE_NBD_DEVICE
109 "\n"
110 "Kernel NBD client support:\n"
111 "  -c, --connect=DEV         connect FILE to the local NBD device DEV\n"
112 "  -d, --disconnect          disconnect the specified device\n"
113 #endif
114 "\n"
115 "Block device options:\n"
116 "  -f, --format=FORMAT       set image format (raw, qcow2, ...)\n"
117 "  -r, --read-only           export read-only\n"
118 "  -s, --snapshot            use FILE as an external snapshot, create a temporary\n"
119 "                            file with backing_file=FILE, redirect the write to\n"
120 "                            the temporary one\n"
121 "  -l, --load-snapshot=SNAPSHOT_PARAM\n"
122 "                            load an internal snapshot inside FILE and export it\n"
123 "                            as an read-only device, SNAPSHOT_PARAM format is\n"
124 "                            'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
125 "                            '[ID_OR_NAME]'\n"
126 "  -n, --nocache             disable host cache\n"
127 "      --cache=MODE          set cache mode (none, writeback, ...)\n"
128 "      --aio=MODE            set AIO mode (native or threads)\n"
129 "      --discard=MODE        set discard mode (ignore, unmap)\n"
130 "      --detect-zeroes=MODE  set detect-zeroes mode (off, on, unmap)\n"
131 "      --image-opts          treat FILE as a full set of image options\n"
132 "\n"
133 QEMU_HELP_BOTTOM "\n"
134     , name, NBD_DEFAULT_PORT, "DEVICE");
135 }
136 
137 static void version(const char *name)
138 {
139     printf(
140 "%s " QEMU_FULL_VERSION "\n"
141 "Written by Anthony Liguori.\n"
142 "\n"
143 QEMU_COPYRIGHT "\n"
144 "This is free software; see the source for copying conditions.  There is NO\n"
145 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
146     , name);
147 }
148 
149 struct partition_record
150 {
151     uint8_t bootable;
152     uint8_t start_head;
153     uint32_t start_cylinder;
154     uint8_t start_sector;
155     uint8_t system;
156     uint8_t end_head;
157     uint8_t end_cylinder;
158     uint8_t end_sector;
159     uint32_t start_sector_abs;
160     uint32_t nb_sectors_abs;
161 };
162 
163 static void read_partition(uint8_t *p, struct partition_record *r)
164 {
165     r->bootable = p[0];
166     r->start_head = p[1];
167     r->start_cylinder = p[3] | ((p[2] << 2) & 0x0300);
168     r->start_sector = p[2] & 0x3f;
169     r->system = p[4];
170     r->end_head = p[5];
171     r->end_cylinder = p[7] | ((p[6] << 2) & 0x300);
172     r->end_sector = p[6] & 0x3f;
173 
174     r->start_sector_abs = ldl_le_p(p + 8);
175     r->nb_sectors_abs   = ldl_le_p(p + 12);
176 }
177 
178 static int find_partition(BlockBackend *blk, int partition,
179                           off_t *offset, off_t *size)
180 {
181     struct partition_record mbr[4];
182     uint8_t data[MBR_SIZE];
183     int i;
184     int ext_partnum = 4;
185     int ret;
186 
187     ret = blk_pread(blk, 0, data, sizeof(data));
188     if (ret < 0) {
189         error_report("error while reading: %s", strerror(-ret));
190         exit(EXIT_FAILURE);
191     }
192 
193     if (data[510] != 0x55 || data[511] != 0xaa) {
194         return -EINVAL;
195     }
196 
197     for (i = 0; i < 4; i++) {
198         read_partition(&data[446 + 16 * i], &mbr[i]);
199 
200         if (!mbr[i].system || !mbr[i].nb_sectors_abs) {
201             continue;
202         }
203 
204         if (mbr[i].system == 0xF || mbr[i].system == 0x5) {
205             struct partition_record ext[4];
206             uint8_t data1[MBR_SIZE];
207             int j;
208 
209             ret = blk_pread(blk, mbr[i].start_sector_abs * MBR_SIZE,
210                             data1, sizeof(data1));
211             if (ret < 0) {
212                 error_report("error while reading: %s", strerror(-ret));
213                 exit(EXIT_FAILURE);
214             }
215 
216             for (j = 0; j < 4; j++) {
217                 read_partition(&data1[446 + 16 * j], &ext[j]);
218                 if (!ext[j].system || !ext[j].nb_sectors_abs) {
219                     continue;
220                 }
221 
222                 if ((ext_partnum + j + 1) == partition) {
223                     *offset = (uint64_t)ext[j].start_sector_abs << 9;
224                     *size = (uint64_t)ext[j].nb_sectors_abs << 9;
225                     return 0;
226                 }
227             }
228             ext_partnum += 4;
229         } else if ((i + 1) == partition) {
230             *offset = (uint64_t)mbr[i].start_sector_abs << 9;
231             *size = (uint64_t)mbr[i].nb_sectors_abs << 9;
232             return 0;
233         }
234     }
235 
236     return -ENOENT;
237 }
238 
239 static void termsig_handler(int signum)
240 {
241     atomic_cmpxchg(&state, RUNNING, TERMINATE);
242     qemu_notify_event();
243 }
244 
245 
246 #if HAVE_NBD_DEVICE
247 static void *show_parts(void *arg)
248 {
249     char *device = arg;
250     int nbd;
251 
252     /* linux just needs an open() to trigger
253      * the partition table update
254      * but remember to load the module with max_part != 0 :
255      *     modprobe nbd max_part=63
256      */
257     nbd = open(device, O_RDWR);
258     if (nbd >= 0) {
259         close(nbd);
260     }
261     return NULL;
262 }
263 
264 static void *nbd_client_thread(void *arg)
265 {
266     char *device = arg;
267     NBDExportInfo info = { .request_sizes = false, };
268     QIOChannelSocket *sioc;
269     int fd;
270     int ret;
271     pthread_t show_parts_thread;
272     Error *local_error = NULL;
273 
274     sioc = qio_channel_socket_new();
275     if (qio_channel_socket_connect_sync(sioc,
276                                         saddr,
277                                         &local_error) < 0) {
278         error_report_err(local_error);
279         goto out;
280     }
281 
282     ret = nbd_receive_negotiate(QIO_CHANNEL(sioc), NULL,
283                                 NULL, NULL, NULL, &info, &local_error);
284     if (ret < 0) {
285         if (local_error) {
286             error_report_err(local_error);
287         }
288         goto out_socket;
289     }
290 
291     fd = open(device, O_RDWR);
292     if (fd < 0) {
293         /* Linux-only, we can use %m in printf.  */
294         error_report("Failed to open %s: %m", device);
295         goto out_socket;
296     }
297 
298     ret = nbd_init(fd, sioc, &info, &local_error);
299     if (ret < 0) {
300         error_report_err(local_error);
301         goto out_fd;
302     }
303 
304     /* update partition table */
305     pthread_create(&show_parts_thread, NULL, show_parts, device);
306 
307     if (verbose) {
308         fprintf(stderr, "NBD device %s is now connected to %s\n",
309                 device, srcpath);
310     } else {
311         /* Close stderr so that the qemu-nbd process exits.  */
312         dup2(STDOUT_FILENO, STDERR_FILENO);
313     }
314 
315     ret = nbd_client(fd);
316     if (ret) {
317         goto out_fd;
318     }
319     close(fd);
320     object_unref(OBJECT(sioc));
321     kill(getpid(), SIGTERM);
322     return (void *) EXIT_SUCCESS;
323 
324 out_fd:
325     close(fd);
326 out_socket:
327     object_unref(OBJECT(sioc));
328 out:
329     kill(getpid(), SIGTERM);
330     return (void *) EXIT_FAILURE;
331 }
332 #endif /* HAVE_NBD_DEVICE */
333 
334 static int nbd_can_accept(void)
335 {
336     return state == RUNNING && nb_fds < shared;
337 }
338 
339 static void nbd_export_closed(NBDExport *export)
340 {
341     assert(state == TERMINATING);
342     state = TERMINATED;
343 }
344 
345 static void nbd_update_server_watch(void);
346 
347 static void nbd_client_closed(NBDClient *client, bool negotiated)
348 {
349     nb_fds--;
350     if (negotiated && nb_fds == 0 && !persistent && state == RUNNING) {
351         state = TERMINATE;
352     }
353     nbd_update_server_watch();
354     nbd_client_put(client);
355 }
356 
357 static void nbd_accept(QIONetListener *listener, QIOChannelSocket *cioc,
358                        gpointer opaque)
359 {
360     if (state >= TERMINATE) {
361         return;
362     }
363 
364     nb_fds++;
365     nbd_update_server_watch();
366     nbd_client_new(cioc, tlscreds, NULL, nbd_client_closed);
367 }
368 
369 static void nbd_update_server_watch(void)
370 {
371     if (nbd_can_accept()) {
372         qio_net_listener_set_client_func(server, nbd_accept, NULL, NULL);
373     } else {
374         qio_net_listener_set_client_func(server, NULL, NULL, NULL);
375     }
376 }
377 
378 
379 static SocketAddress *nbd_build_socket_address(const char *sockpath,
380                                                const char *bindto,
381                                                const char *port)
382 {
383     SocketAddress *saddr;
384 
385     saddr = g_new0(SocketAddress, 1);
386     if (sockpath) {
387         saddr->type = SOCKET_ADDRESS_TYPE_UNIX;
388         saddr->u.q_unix.path = g_strdup(sockpath);
389     } else {
390         InetSocketAddress *inet;
391         saddr->type = SOCKET_ADDRESS_TYPE_INET;
392         inet = &saddr->u.inet;
393         inet->host = g_strdup(bindto);
394         if (port) {
395             inet->port = g_strdup(port);
396         } else  {
397             inet->port = g_strdup_printf("%d", NBD_DEFAULT_PORT);
398         }
399     }
400 
401     return saddr;
402 }
403 
404 
405 static QemuOptsList file_opts = {
406     .name = "file",
407     .implied_opt_name = "file",
408     .head = QTAILQ_HEAD_INITIALIZER(file_opts.head),
409     .desc = {
410         /* no elements => accept any params */
411         { /* end of list */ }
412     },
413 };
414 
415 static QemuOptsList qemu_object_opts = {
416     .name = "object",
417     .implied_opt_name = "qom-type",
418     .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head),
419     .desc = {
420         { }
421     },
422 };
423 
424 
425 
426 static QCryptoTLSCreds *nbd_get_tls_creds(const char *id, Error **errp)
427 {
428     Object *obj;
429     QCryptoTLSCreds *creds;
430 
431     obj = object_resolve_path_component(
432         object_get_objects_root(), id);
433     if (!obj) {
434         error_setg(errp, "No TLS credentials with id '%s'",
435                    id);
436         return NULL;
437     }
438     creds = (QCryptoTLSCreds *)
439         object_dynamic_cast(obj, TYPE_QCRYPTO_TLS_CREDS);
440     if (!creds) {
441         error_setg(errp, "Object with id '%s' is not TLS credentials",
442                    id);
443         return NULL;
444     }
445 
446     if (creds->endpoint != QCRYPTO_TLS_CREDS_ENDPOINT_SERVER) {
447         error_setg(errp,
448                    "Expecting TLS credentials with a server endpoint");
449         return NULL;
450     }
451     object_ref(obj);
452     return creds;
453 }
454 
455 static void setup_address_and_port(const char **address, const char **port)
456 {
457     if (*address == NULL) {
458         *address = "0.0.0.0";
459     }
460 
461     if (*port == NULL) {
462         *port = stringify(NBD_DEFAULT_PORT);
463     }
464 }
465 
466 /*
467  * Check socket parameters compatibility when socket activation is used.
468  */
469 static const char *socket_activation_validate_opts(const char *device,
470                                                    const char *sockpath,
471                                                    const char *address,
472                                                    const char *port)
473 {
474     if (device != NULL) {
475         return "NBD device can't be set when using socket activation";
476     }
477 
478     if (sockpath != NULL) {
479         return "Unix socket can't be set when using socket activation";
480     }
481 
482     if (address != NULL) {
483         return "The interface can't be set when using socket activation";
484     }
485 
486     if (port != NULL) {
487         return "TCP port number can't be set when using socket activation";
488     }
489 
490     return NULL;
491 }
492 
493 static void qemu_nbd_shutdown(void)
494 {
495     job_cancel_sync_all();
496     bdrv_close_all();
497 }
498 
499 int main(int argc, char **argv)
500 {
501     BlockBackend *blk;
502     BlockDriverState *bs;
503     off_t dev_offset = 0;
504     uint16_t nbdflags = 0;
505     bool disconnect = false;
506     const char *bindto = NULL;
507     const char *port = NULL;
508     char *sockpath = NULL;
509     char *device = NULL;
510     off_t fd_size;
511     QemuOpts *sn_opts = NULL;
512     const char *sn_id_or_name = NULL;
513     const char *sopt = "hVb:o:p:rsnP:c:dvk:e:f:tl:x:T:D:B:";
514     struct option lopt[] = {
515         { "help", no_argument, NULL, 'h' },
516         { "version", no_argument, NULL, 'V' },
517         { "bind", required_argument, NULL, 'b' },
518         { "port", required_argument, NULL, 'p' },
519         { "socket", required_argument, NULL, 'k' },
520         { "offset", required_argument, NULL, 'o' },
521         { "read-only", no_argument, NULL, 'r' },
522         { "partition", required_argument, NULL, 'P' },
523         { "bitmap", required_argument, NULL, 'B' },
524         { "connect", required_argument, NULL, 'c' },
525         { "disconnect", no_argument, NULL, 'd' },
526         { "snapshot", no_argument, NULL, 's' },
527         { "load-snapshot", required_argument, NULL, 'l' },
528         { "nocache", no_argument, NULL, 'n' },
529         { "cache", required_argument, NULL, QEMU_NBD_OPT_CACHE },
530         { "aio", required_argument, NULL, QEMU_NBD_OPT_AIO },
531         { "discard", required_argument, NULL, QEMU_NBD_OPT_DISCARD },
532         { "detect-zeroes", required_argument, NULL,
533           QEMU_NBD_OPT_DETECT_ZEROES },
534         { "shared", required_argument, NULL, 'e' },
535         { "format", required_argument, NULL, 'f' },
536         { "persistent", no_argument, NULL, 't' },
537         { "verbose", no_argument, NULL, 'v' },
538         { "object", required_argument, NULL, QEMU_NBD_OPT_OBJECT },
539         { "export-name", required_argument, NULL, 'x' },
540         { "description", required_argument, NULL, 'D' },
541         { "tls-creds", required_argument, NULL, QEMU_NBD_OPT_TLSCREDS },
542         { "image-opts", no_argument, NULL, QEMU_NBD_OPT_IMAGE_OPTS },
543         { "trace", required_argument, NULL, 'T' },
544         { "fork", no_argument, NULL, QEMU_NBD_OPT_FORK },
545         { NULL, 0, NULL, 0 }
546     };
547     int ch;
548     int opt_ind = 0;
549     char *end;
550     int flags = BDRV_O_RDWR;
551     int partition = -1;
552     int ret = 0;
553     bool seen_cache = false;
554     bool seen_discard = false;
555     bool seen_aio = false;
556     pthread_t client_thread;
557     const char *fmt = NULL;
558     Error *local_err = NULL;
559     BlockdevDetectZeroesOptions detect_zeroes = BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
560     QDict *options = NULL;
561     const char *export_name = ""; /* Default export name */
562     const char *export_description = NULL;
563     const char *bitmap = NULL;
564     const char *tlscredsid = NULL;
565     bool imageOpts = false;
566     bool writethrough = true;
567     char *trace_file = NULL;
568     bool fork_process = false;
569     int old_stderr = -1;
570     unsigned socket_activation;
571 
572     /* The client thread uses SIGTERM to interrupt the server.  A signal
573      * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
574      */
575     struct sigaction sa_sigterm;
576     memset(&sa_sigterm, 0, sizeof(sa_sigterm));
577     sa_sigterm.sa_handler = termsig_handler;
578     sigaction(SIGTERM, &sa_sigterm, NULL);
579 
580 #ifdef CONFIG_POSIX
581     signal(SIGPIPE, SIG_IGN);
582 #endif
583 
584     module_call_init(MODULE_INIT_TRACE);
585     error_set_progname(argv[0]);
586     qcrypto_init(&error_fatal);
587 
588     module_call_init(MODULE_INIT_QOM);
589     qemu_add_opts(&qemu_object_opts);
590     qemu_add_opts(&qemu_trace_opts);
591     qemu_init_exec_dir(argv[0]);
592 
593     while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
594         switch (ch) {
595         case 's':
596             flags |= BDRV_O_SNAPSHOT;
597             break;
598         case 'n':
599             optarg = (char *) "none";
600             /* fallthrough */
601         case QEMU_NBD_OPT_CACHE:
602             if (seen_cache) {
603                 error_report("-n and --cache can only be specified once");
604                 exit(EXIT_FAILURE);
605             }
606             seen_cache = true;
607             if (bdrv_parse_cache_mode(optarg, &flags, &writethrough) == -1) {
608                 error_report("Invalid cache mode `%s'", optarg);
609                 exit(EXIT_FAILURE);
610             }
611             break;
612         case QEMU_NBD_OPT_AIO:
613             if (seen_aio) {
614                 error_report("--aio can only be specified once");
615                 exit(EXIT_FAILURE);
616             }
617             seen_aio = true;
618             if (!strcmp(optarg, "native")) {
619                 flags |= BDRV_O_NATIVE_AIO;
620             } else if (!strcmp(optarg, "threads")) {
621                 /* this is the default */
622             } else {
623                error_report("invalid aio mode `%s'", optarg);
624                exit(EXIT_FAILURE);
625             }
626             break;
627         case QEMU_NBD_OPT_DISCARD:
628             if (seen_discard) {
629                 error_report("--discard can only be specified once");
630                 exit(EXIT_FAILURE);
631             }
632             seen_discard = true;
633             if (bdrv_parse_discard_flags(optarg, &flags) == -1) {
634                 error_report("Invalid discard mode `%s'", optarg);
635                 exit(EXIT_FAILURE);
636             }
637             break;
638         case QEMU_NBD_OPT_DETECT_ZEROES:
639             detect_zeroes =
640                 qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup,
641                                 optarg,
642                                 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
643                                 &local_err);
644             if (local_err) {
645                 error_reportf_err(local_err,
646                                   "Failed to parse detect_zeroes mode: ");
647                 exit(EXIT_FAILURE);
648             }
649             if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
650                 !(flags & BDRV_O_UNMAP)) {
651                 error_report("setting detect-zeroes to unmap is not allowed "
652                              "without setting discard operation to unmap");
653                 exit(EXIT_FAILURE);
654             }
655             break;
656         case 'b':
657             bindto = optarg;
658             break;
659         case 'p':
660             port = optarg;
661             break;
662         case 'o':
663                 dev_offset = strtoll (optarg, &end, 0);
664             if (*end) {
665                 error_report("Invalid offset `%s'", optarg);
666                 exit(EXIT_FAILURE);
667             }
668             if (dev_offset < 0) {
669                 error_report("Offset must be positive `%s'", optarg);
670                 exit(EXIT_FAILURE);
671             }
672             break;
673         case 'l':
674             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
675                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
676                                                   optarg, false);
677                 if (!sn_opts) {
678                     error_report("Failed in parsing snapshot param `%s'",
679                                  optarg);
680                     exit(EXIT_FAILURE);
681                 }
682             } else {
683                 sn_id_or_name = optarg;
684             }
685             /* fall through */
686         case 'r':
687             nbdflags |= NBD_FLAG_READ_ONLY;
688             flags &= ~BDRV_O_RDWR;
689             break;
690         case 'P':
691             partition = strtol(optarg, &end, 0);
692             if (*end) {
693                 error_report("Invalid partition `%s'", optarg);
694                 exit(EXIT_FAILURE);
695             }
696             if (partition < 1 || partition > 8) {
697                 error_report("Invalid partition %d", partition);
698                 exit(EXIT_FAILURE);
699             }
700             break;
701         case 'B':
702             bitmap = optarg;
703             break;
704         case 'k':
705             sockpath = optarg;
706             if (sockpath[0] != '/') {
707                 error_report("socket path must be absolute");
708                 exit(EXIT_FAILURE);
709             }
710             break;
711         case 'd':
712             disconnect = true;
713             break;
714         case 'c':
715             device = optarg;
716             break;
717         case 'e':
718             shared = strtol(optarg, &end, 0);
719             if (*end) {
720                 error_report("Invalid shared device number '%s'", optarg);
721                 exit(EXIT_FAILURE);
722             }
723             if (shared < 1) {
724                 error_report("Shared device number must be greater than 0");
725                 exit(EXIT_FAILURE);
726             }
727             break;
728         case 'f':
729             fmt = optarg;
730             break;
731         case 't':
732             persistent = 1;
733             break;
734         case 'x':
735             export_name = optarg;
736             break;
737         case 'D':
738             export_description = optarg;
739             break;
740         case 'v':
741             verbose = 1;
742             break;
743         case 'V':
744             version(argv[0]);
745             exit(0);
746             break;
747         case 'h':
748             usage(argv[0]);
749             exit(0);
750             break;
751         case '?':
752             error_report("Try `%s --help' for more information.", argv[0]);
753             exit(EXIT_FAILURE);
754         case QEMU_NBD_OPT_OBJECT: {
755             QemuOpts *opts;
756             opts = qemu_opts_parse_noisily(&qemu_object_opts,
757                                            optarg, true);
758             if (!opts) {
759                 exit(EXIT_FAILURE);
760             }
761         }   break;
762         case QEMU_NBD_OPT_TLSCREDS:
763             tlscredsid = optarg;
764             break;
765         case QEMU_NBD_OPT_IMAGE_OPTS:
766             imageOpts = true;
767             break;
768         case 'T':
769             g_free(trace_file);
770             trace_file = trace_opt_parse(optarg);
771             break;
772         case QEMU_NBD_OPT_FORK:
773             fork_process = true;
774             break;
775         }
776     }
777 
778     if ((argc - optind) != 1) {
779         error_report("Invalid number of arguments");
780         error_printf("Try `%s --help' for more information.\n", argv[0]);
781         exit(EXIT_FAILURE);
782     }
783 
784     qemu_opts_foreach(&qemu_object_opts,
785                       user_creatable_add_opts_foreach,
786                       NULL, &error_fatal);
787 
788     if (!trace_init_backends()) {
789         exit(1);
790     }
791     trace_init_file(trace_file);
792     qemu_set_log(LOG_TRACE);
793 
794     socket_activation = check_socket_activation();
795     if (socket_activation == 0) {
796         setup_address_and_port(&bindto, &port);
797     } else {
798         /* Using socket activation - check user didn't use -p etc. */
799         const char *err_msg = socket_activation_validate_opts(device, sockpath,
800                                                               bindto, port);
801         if (err_msg != NULL) {
802             error_report("%s", err_msg);
803             exit(EXIT_FAILURE);
804         }
805 
806         /* qemu-nbd can only listen on a single socket.  */
807         if (socket_activation > 1) {
808             error_report("qemu-nbd does not support socket activation with %s > 1",
809                          "LISTEN_FDS");
810             exit(EXIT_FAILURE);
811         }
812     }
813 
814     if (tlscredsid) {
815         if (sockpath) {
816             error_report("TLS is only supported with IPv4/IPv6");
817             exit(EXIT_FAILURE);
818         }
819         if (device) {
820             error_report("TLS is not supported with a host device");
821             exit(EXIT_FAILURE);
822         }
823         tlscreds = nbd_get_tls_creds(tlscredsid, &local_err);
824         if (local_err) {
825             error_report("Failed to get TLS creds %s",
826                          error_get_pretty(local_err));
827             exit(EXIT_FAILURE);
828         }
829     }
830 
831 #if !HAVE_NBD_DEVICE
832     if (disconnect || device) {
833         error_report("Kernel /dev/nbdN support not available");
834         exit(EXIT_FAILURE);
835     }
836 #else /* HAVE_NBD_DEVICE */
837     if (disconnect) {
838         int nbdfd = open(argv[optind], O_RDWR);
839         if (nbdfd < 0) {
840             error_report("Cannot open %s: %s", argv[optind],
841                          strerror(errno));
842             exit(EXIT_FAILURE);
843         }
844         nbd_disconnect(nbdfd);
845 
846         close(nbdfd);
847 
848         printf("%s disconnected\n", argv[optind]);
849 
850         return 0;
851     }
852 #endif
853 
854     if ((device && !verbose) || fork_process) {
855         int stderr_fd[2];
856         pid_t pid;
857         int ret;
858 
859         if (qemu_pipe(stderr_fd) < 0) {
860             error_report("Error setting up communication pipe: %s",
861                          strerror(errno));
862             exit(EXIT_FAILURE);
863         }
864 
865         /* Now daemonize, but keep a communication channel open to
866          * print errors and exit with the proper status code.
867          */
868         pid = fork();
869         if (pid < 0) {
870             error_report("Failed to fork: %s", strerror(errno));
871             exit(EXIT_FAILURE);
872         } else if (pid == 0) {
873             close(stderr_fd[0]);
874             ret = qemu_daemon(1, 0);
875 
876             /* Temporarily redirect stderr to the parent's pipe...  */
877             old_stderr = dup(STDERR_FILENO);
878             dup2(stderr_fd[1], STDERR_FILENO);
879             if (ret < 0) {
880                 error_report("Failed to daemonize: %s", strerror(errno));
881                 exit(EXIT_FAILURE);
882             }
883 
884             /* ... close the descriptor we inherited and go on.  */
885             close(stderr_fd[1]);
886         } else {
887             bool errors = false;
888             char *buf;
889 
890             /* In the parent.  Print error messages from the child until
891              * it closes the pipe.
892              */
893             close(stderr_fd[1]);
894             buf = g_malloc(1024);
895             while ((ret = read(stderr_fd[0], buf, 1024)) > 0) {
896                 errors = true;
897                 ret = qemu_write_full(STDERR_FILENO, buf, ret);
898                 if (ret < 0) {
899                     exit(EXIT_FAILURE);
900                 }
901             }
902             if (ret < 0) {
903                 error_report("Cannot read from daemon: %s",
904                              strerror(errno));
905                 exit(EXIT_FAILURE);
906             }
907 
908             /* Usually the daemon should not print any message.
909              * Exit with zero status in that case.
910              */
911             exit(errors);
912         }
913     }
914 
915     if (device != NULL && sockpath == NULL) {
916         sockpath = g_malloc(128);
917         snprintf(sockpath, 128, SOCKET_PATH, basename(device));
918     }
919 
920     server = qio_net_listener_new();
921     if (socket_activation == 0) {
922         saddr = nbd_build_socket_address(sockpath, bindto, port);
923         if (qio_net_listener_open_sync(server, saddr, &local_err) < 0) {
924             object_unref(OBJECT(server));
925             error_report_err(local_err);
926             exit(EXIT_FAILURE);
927         }
928     } else {
929         size_t i;
930         /* See comment in check_socket_activation above. */
931         for (i = 0; i < socket_activation; i++) {
932             QIOChannelSocket *sioc;
933             sioc = qio_channel_socket_new_fd(FIRST_SOCKET_ACTIVATION_FD + i,
934                                              &local_err);
935             if (sioc == NULL) {
936                 object_unref(OBJECT(server));
937                 error_report("Failed to use socket activation: %s",
938                              error_get_pretty(local_err));
939                 exit(EXIT_FAILURE);
940             }
941             qio_net_listener_add(server, sioc);
942             object_unref(OBJECT(sioc));
943         }
944     }
945 
946     if (qemu_init_main_loop(&local_err)) {
947         error_report_err(local_err);
948         exit(EXIT_FAILURE);
949     }
950     bdrv_init();
951     atexit(qemu_nbd_shutdown);
952 
953     srcpath = argv[optind];
954     if (imageOpts) {
955         QemuOpts *opts;
956         if (fmt) {
957             error_report("--image-opts and -f are mutually exclusive");
958             exit(EXIT_FAILURE);
959         }
960         opts = qemu_opts_parse_noisily(&file_opts, srcpath, true);
961         if (!opts) {
962             qemu_opts_reset(&file_opts);
963             exit(EXIT_FAILURE);
964         }
965         options = qemu_opts_to_qdict(opts, NULL);
966         qemu_opts_reset(&file_opts);
967         blk = blk_new_open(NULL, NULL, options, flags, &local_err);
968     } else {
969         if (fmt) {
970             options = qdict_new();
971             qdict_put_str(options, "driver", fmt);
972         }
973         blk = blk_new_open(srcpath, NULL, options, flags, &local_err);
974     }
975 
976     if (!blk) {
977         error_reportf_err(local_err, "Failed to blk_new_open '%s': ",
978                           argv[optind]);
979         exit(EXIT_FAILURE);
980     }
981     bs = blk_bs(blk);
982 
983     blk_set_enable_write_cache(blk, !writethrough);
984 
985     if (sn_opts) {
986         ret = bdrv_snapshot_load_tmp(bs,
987                                      qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
988                                      qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
989                                      &local_err);
990     } else if (sn_id_or_name) {
991         ret = bdrv_snapshot_load_tmp_by_id_or_name(bs, sn_id_or_name,
992                                                    &local_err);
993     }
994     if (ret < 0) {
995         error_reportf_err(local_err, "Failed to load snapshot: ");
996         exit(EXIT_FAILURE);
997     }
998 
999     bs->detect_zeroes = detect_zeroes;
1000     fd_size = blk_getlength(blk);
1001     if (fd_size < 0) {
1002         error_report("Failed to determine the image length: %s",
1003                      strerror(-fd_size));
1004         exit(EXIT_FAILURE);
1005     }
1006 
1007     if (dev_offset >= fd_size) {
1008         error_report("Offset (%lld) has to be smaller than the image size "
1009                      "(%lld)",
1010                      (long long int)dev_offset, (long long int)fd_size);
1011         exit(EXIT_FAILURE);
1012     }
1013     fd_size -= dev_offset;
1014 
1015     if (partition != -1) {
1016         ret = find_partition(blk, partition, &dev_offset, &fd_size);
1017         if (ret < 0) {
1018             error_report("Could not find partition %d: %s", partition,
1019                          strerror(-ret));
1020             exit(EXIT_FAILURE);
1021         }
1022     }
1023 
1024     export = nbd_export_new(bs, dev_offset, fd_size, export_name,
1025                             export_description, bitmap, nbdflags,
1026                             nbd_export_closed, writethrough, NULL,
1027                             &error_fatal);
1028 
1029     if (device) {
1030 #if HAVE_NBD_DEVICE
1031         int ret;
1032 
1033         ret = pthread_create(&client_thread, NULL, nbd_client_thread, device);
1034         if (ret != 0) {
1035             error_report("Failed to create client thread: %s", strerror(ret));
1036             exit(EXIT_FAILURE);
1037         }
1038 #endif
1039     } else {
1040         /* Shut up GCC warnings.  */
1041         memset(&client_thread, 0, sizeof(client_thread));
1042     }
1043 
1044     nbd_update_server_watch();
1045 
1046     /* now when the initialization is (almost) complete, chdir("/")
1047      * to free any busy filesystems */
1048     if (chdir("/") < 0) {
1049         error_report("Could not chdir to root directory: %s",
1050                      strerror(errno));
1051         exit(EXIT_FAILURE);
1052     }
1053 
1054     if (fork_process) {
1055         dup2(old_stderr, STDERR_FILENO);
1056         close(old_stderr);
1057     }
1058 
1059     state = RUNNING;
1060     do {
1061         main_loop_wait(false);
1062         if (state == TERMINATE) {
1063             state = TERMINATING;
1064             nbd_export_close(export);
1065             nbd_export_put(export);
1066             export = NULL;
1067         }
1068     } while (state != TERMINATED);
1069 
1070     blk_unref(blk);
1071     if (sockpath) {
1072         unlink(sockpath);
1073     }
1074 
1075     qemu_opts_del(sn_opts);
1076 
1077     if (device) {
1078         void *ret;
1079         pthread_join(client_thread, &ret);
1080         exit(ret != NULL);
1081     } else {
1082         exit(EXIT_SUCCESS);
1083     }
1084 }
1085