xref: /qemu/qemu-nbd.c (revision d8b20291)
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 #define SOCKET_PATH                "/var/lock/qemu-nbd-%s"
47 #define QEMU_NBD_OPT_CACHE         256
48 #define QEMU_NBD_OPT_AIO           257
49 #define QEMU_NBD_OPT_DISCARD       258
50 #define QEMU_NBD_OPT_DETECT_ZEROES 259
51 #define QEMU_NBD_OPT_OBJECT        260
52 #define QEMU_NBD_OPT_TLSCREDS      261
53 #define QEMU_NBD_OPT_IMAGE_OPTS    262
54 #define QEMU_NBD_OPT_FORK          263
55 
56 #define MBR_SIZE 512
57 
58 static NBDExport *exp;
59 static bool newproto;
60 static int verbose;
61 static char *srcpath;
62 static SocketAddress *saddr;
63 static int persistent = 0;
64 static enum { RUNNING, TERMINATE, TERMINATING, TERMINATED } state;
65 static int shared = 1;
66 static int nb_fds;
67 static QIONetListener *server;
68 static QCryptoTLSCreds *tlscreds;
69 
70 static void usage(const char *name)
71 {
72     (printf) (
73 "Usage: %s [OPTIONS] FILE\n"
74 "QEMU Disk Network Block Device Server\n"
75 "\n"
76 "  -h, --help                display this help and exit\n"
77 "  -V, --version             output version information and exit\n"
78 "\n"
79 "Connection properties:\n"
80 "  -p, --port=PORT           port to listen on (default `%d')\n"
81 "  -b, --bind=IFACE          interface to bind to (default `0.0.0.0')\n"
82 "  -k, --socket=PATH         path to the unix socket\n"
83 "                            (default '"SOCKET_PATH"')\n"
84 "  -e, --shared=NUM          device can be shared by NUM clients (default '1')\n"
85 "  -t, --persistent          don't exit on the last connection\n"
86 "  -v, --verbose             display extra debugging information\n"
87 "  -x, --export-name=NAME    expose export by name\n"
88 "  -D, --description=TEXT    with -x, also export a human-readable description\n"
89 "\n"
90 "Exposing part of the image:\n"
91 "  -o, --offset=OFFSET       offset into the image\n"
92 "  -P, --partition=NUM       only expose partition NUM\n"
93 "\n"
94 "General purpose options:\n"
95 "  --object type,id=ID,...   define an object such as 'secret' for providing\n"
96 "                            passwords and/or encryption keys\n"
97 "  -T, --trace [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
98 "                            specify tracing options\n"
99 "  --fork                    fork off the server process and exit the parent\n"
100 "                            once the server is running\n"
101 #ifdef __linux__
102 "Kernel NBD client support:\n"
103 "  -c, --connect=DEV         connect FILE to the local NBD device DEV\n"
104 "  -d, --disconnect          disconnect the specified device\n"
105 "\n"
106 #endif
107 "\n"
108 "Block device options:\n"
109 "  -f, --format=FORMAT       set image format (raw, qcow2, ...)\n"
110 "  -r, --read-only           export read-only\n"
111 "  -s, --snapshot            use FILE as an external snapshot, create a temporary\n"
112 "                            file with backing_file=FILE, redirect the write to\n"
113 "                            the temporary one\n"
114 "  -l, --load-snapshot=SNAPSHOT_PARAM\n"
115 "                            load an internal snapshot inside FILE and export it\n"
116 "                            as an read-only device, SNAPSHOT_PARAM format is\n"
117 "                            'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
118 "                            '[ID_OR_NAME]'\n"
119 "  -n, --nocache             disable host cache\n"
120 "      --cache=MODE          set cache mode (none, writeback, ...)\n"
121 "      --aio=MODE            set AIO mode (native or threads)\n"
122 "      --discard=MODE        set discard mode (ignore, unmap)\n"
123 "      --detect-zeroes=MODE  set detect-zeroes mode (off, on, unmap)\n"
124 "      --image-opts          treat FILE as a full set of image options\n"
125 "\n"
126 QEMU_HELP_BOTTOM "\n"
127     , name, NBD_DEFAULT_PORT, "DEVICE");
128 }
129 
130 static void version(const char *name)
131 {
132     printf(
133 "%s " QEMU_FULL_VERSION "\n"
134 "Written by Anthony Liguori.\n"
135 "\n"
136 QEMU_COPYRIGHT "\n"
137 "This is free software; see the source for copying conditions.  There is NO\n"
138 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
139     , name);
140 }
141 
142 struct partition_record
143 {
144     uint8_t bootable;
145     uint8_t start_head;
146     uint32_t start_cylinder;
147     uint8_t start_sector;
148     uint8_t system;
149     uint8_t end_head;
150     uint8_t end_cylinder;
151     uint8_t end_sector;
152     uint32_t start_sector_abs;
153     uint32_t nb_sectors_abs;
154 };
155 
156 static void read_partition(uint8_t *p, struct partition_record *r)
157 {
158     r->bootable = p[0];
159     r->start_head = p[1];
160     r->start_cylinder = p[3] | ((p[2] << 2) & 0x0300);
161     r->start_sector = p[2] & 0x3f;
162     r->system = p[4];
163     r->end_head = p[5];
164     r->end_cylinder = p[7] | ((p[6] << 2) & 0x300);
165     r->end_sector = p[6] & 0x3f;
166 
167     r->start_sector_abs = ldl_le_p(p + 8);
168     r->nb_sectors_abs   = ldl_le_p(p + 12);
169 }
170 
171 static int find_partition(BlockBackend *blk, int partition,
172                           off_t *offset, off_t *size)
173 {
174     struct partition_record mbr[4];
175     uint8_t data[MBR_SIZE];
176     int i;
177     int ext_partnum = 4;
178     int ret;
179 
180     ret = blk_pread(blk, 0, data, sizeof(data));
181     if (ret < 0) {
182         error_report("error while reading: %s", strerror(-ret));
183         exit(EXIT_FAILURE);
184     }
185 
186     if (data[510] != 0x55 || data[511] != 0xaa) {
187         return -EINVAL;
188     }
189 
190     for (i = 0; i < 4; i++) {
191         read_partition(&data[446 + 16 * i], &mbr[i]);
192 
193         if (!mbr[i].system || !mbr[i].nb_sectors_abs) {
194             continue;
195         }
196 
197         if (mbr[i].system == 0xF || mbr[i].system == 0x5) {
198             struct partition_record ext[4];
199             uint8_t data1[MBR_SIZE];
200             int j;
201 
202             ret = blk_pread(blk, mbr[i].start_sector_abs * MBR_SIZE,
203                             data1, sizeof(data1));
204             if (ret < 0) {
205                 error_report("error while reading: %s", strerror(-ret));
206                 exit(EXIT_FAILURE);
207             }
208 
209             for (j = 0; j < 4; j++) {
210                 read_partition(&data1[446 + 16 * j], &ext[j]);
211                 if (!ext[j].system || !ext[j].nb_sectors_abs) {
212                     continue;
213                 }
214 
215                 if ((ext_partnum + j + 1) == partition) {
216                     *offset = (uint64_t)ext[j].start_sector_abs << 9;
217                     *size = (uint64_t)ext[j].nb_sectors_abs << 9;
218                     return 0;
219                 }
220             }
221             ext_partnum += 4;
222         } else if ((i + 1) == partition) {
223             *offset = (uint64_t)mbr[i].start_sector_abs << 9;
224             *size = (uint64_t)mbr[i].nb_sectors_abs << 9;
225             return 0;
226         }
227     }
228 
229     return -ENOENT;
230 }
231 
232 static void termsig_handler(int signum)
233 {
234     atomic_cmpxchg(&state, RUNNING, TERMINATE);
235     qemu_notify_event();
236 }
237 
238 
239 static void *show_parts(void *arg)
240 {
241     char *device = arg;
242     int nbd;
243 
244     /* linux just needs an open() to trigger
245      * the partition table update
246      * but remember to load the module with max_part != 0 :
247      *     modprobe nbd max_part=63
248      */
249     nbd = open(device, O_RDWR);
250     if (nbd >= 0) {
251         close(nbd);
252     }
253     return NULL;
254 }
255 
256 static void *nbd_client_thread(void *arg)
257 {
258     char *device = arg;
259     NBDExportInfo info = { .request_sizes = false, };
260     QIOChannelSocket *sioc;
261     int fd;
262     int ret;
263     pthread_t show_parts_thread;
264     Error *local_error = NULL;
265 
266     sioc = qio_channel_socket_new();
267     if (qio_channel_socket_connect_sync(sioc,
268                                         saddr,
269                                         &local_error) < 0) {
270         error_report_err(local_error);
271         goto out;
272     }
273 
274     ret = nbd_receive_negotiate(QIO_CHANNEL(sioc), NULL,
275                                 NULL, NULL, NULL, &info, &local_error);
276     if (ret < 0) {
277         if (local_error) {
278             error_report_err(local_error);
279         }
280         goto out_socket;
281     }
282 
283     fd = open(device, O_RDWR);
284     if (fd < 0) {
285         /* Linux-only, we can use %m in printf.  */
286         error_report("Failed to open %s: %m", device);
287         goto out_socket;
288     }
289 
290     ret = nbd_init(fd, sioc, &info, &local_error);
291     if (ret < 0) {
292         error_report_err(local_error);
293         goto out_fd;
294     }
295 
296     /* update partition table */
297     pthread_create(&show_parts_thread, NULL, show_parts, device);
298 
299     if (verbose) {
300         fprintf(stderr, "NBD device %s is now connected to %s\n",
301                 device, srcpath);
302     } else {
303         /* Close stderr so that the qemu-nbd process exits.  */
304         dup2(STDOUT_FILENO, STDERR_FILENO);
305     }
306 
307     ret = nbd_client(fd);
308     if (ret) {
309         goto out_fd;
310     }
311     close(fd);
312     object_unref(OBJECT(sioc));
313     kill(getpid(), SIGTERM);
314     return (void *) EXIT_SUCCESS;
315 
316 out_fd:
317     close(fd);
318 out_socket:
319     object_unref(OBJECT(sioc));
320 out:
321     kill(getpid(), SIGTERM);
322     return (void *) EXIT_FAILURE;
323 }
324 
325 static int nbd_can_accept(void)
326 {
327     return state == RUNNING && nb_fds < shared;
328 }
329 
330 static void nbd_export_closed(NBDExport *exp)
331 {
332     assert(state == TERMINATING);
333     state = TERMINATED;
334 }
335 
336 static void nbd_update_server_watch(void);
337 
338 static void nbd_client_closed(NBDClient *client, bool negotiated)
339 {
340     nb_fds--;
341     if (negotiated && nb_fds == 0 && !persistent && state == RUNNING) {
342         state = TERMINATE;
343     }
344     nbd_update_server_watch();
345     nbd_client_put(client);
346 }
347 
348 static void nbd_accept(QIONetListener *listener, QIOChannelSocket *cioc,
349                        gpointer opaque)
350 {
351     if (state >= TERMINATE) {
352         return;
353     }
354 
355     nb_fds++;
356     nbd_update_server_watch();
357     nbd_client_new(newproto ? NULL : exp, cioc,
358                    tlscreds, NULL, nbd_client_closed);
359 }
360 
361 static void nbd_update_server_watch(void)
362 {
363     if (nbd_can_accept()) {
364         qio_net_listener_set_client_func(server, nbd_accept, NULL, NULL);
365     } else {
366         qio_net_listener_set_client_func(server, NULL, NULL, NULL);
367     }
368 }
369 
370 
371 static SocketAddress *nbd_build_socket_address(const char *sockpath,
372                                                const char *bindto,
373                                                const char *port)
374 {
375     SocketAddress *saddr;
376 
377     saddr = g_new0(SocketAddress, 1);
378     if (sockpath) {
379         saddr->type = SOCKET_ADDRESS_TYPE_UNIX;
380         saddr->u.q_unix.path = g_strdup(sockpath);
381     } else {
382         InetSocketAddress *inet;
383         saddr->type = SOCKET_ADDRESS_TYPE_INET;
384         inet = &saddr->u.inet;
385         inet->host = g_strdup(bindto);
386         if (port) {
387             inet->port = g_strdup(port);
388         } else  {
389             inet->port = g_strdup_printf("%d", NBD_DEFAULT_PORT);
390         }
391     }
392 
393     return saddr;
394 }
395 
396 
397 static QemuOptsList file_opts = {
398     .name = "file",
399     .implied_opt_name = "file",
400     .head = QTAILQ_HEAD_INITIALIZER(file_opts.head),
401     .desc = {
402         /* no elements => accept any params */
403         { /* end of list */ }
404     },
405 };
406 
407 static QemuOptsList qemu_object_opts = {
408     .name = "object",
409     .implied_opt_name = "qom-type",
410     .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head),
411     .desc = {
412         { }
413     },
414 };
415 
416 
417 
418 static QCryptoTLSCreds *nbd_get_tls_creds(const char *id, Error **errp)
419 {
420     Object *obj;
421     QCryptoTLSCreds *creds;
422 
423     obj = object_resolve_path_component(
424         object_get_objects_root(), id);
425     if (!obj) {
426         error_setg(errp, "No TLS credentials with id '%s'",
427                    id);
428         return NULL;
429     }
430     creds = (QCryptoTLSCreds *)
431         object_dynamic_cast(obj, TYPE_QCRYPTO_TLS_CREDS);
432     if (!creds) {
433         error_setg(errp, "Object with id '%s' is not TLS credentials",
434                    id);
435         return NULL;
436     }
437 
438     if (creds->endpoint != QCRYPTO_TLS_CREDS_ENDPOINT_SERVER) {
439         error_setg(errp,
440                    "Expecting TLS credentials with a server endpoint");
441         return NULL;
442     }
443     object_ref(obj);
444     return creds;
445 }
446 
447 static void setup_address_and_port(const char **address, const char **port)
448 {
449     if (*address == NULL) {
450         *address = "0.0.0.0";
451     }
452 
453     if (*port == NULL) {
454         *port = stringify(NBD_DEFAULT_PORT);
455     }
456 }
457 
458 /*
459  * Check socket parameters compatibility when socket activation is used.
460  */
461 static const char *socket_activation_validate_opts(const char *device,
462                                                    const char *sockpath,
463                                                    const char *address,
464                                                    const char *port)
465 {
466     if (device != NULL) {
467         return "NBD device can't be set when using socket activation";
468     }
469 
470     if (sockpath != NULL) {
471         return "Unix socket can't be set when using socket activation";
472     }
473 
474     if (address != NULL) {
475         return "The interface can't be set when using socket activation";
476     }
477 
478     if (port != NULL) {
479         return "TCP port number can't be set when using socket activation";
480     }
481 
482     return NULL;
483 }
484 
485 static void qemu_nbd_shutdown(void)
486 {
487     job_cancel_sync_all();
488     bdrv_close_all();
489 }
490 
491 int main(int argc, char **argv)
492 {
493     BlockBackend *blk;
494     BlockDriverState *bs;
495     off_t dev_offset = 0;
496     uint16_t nbdflags = 0;
497     bool disconnect = false;
498     const char *bindto = NULL;
499     const char *port = NULL;
500     char *sockpath = NULL;
501     char *device = NULL;
502     off_t fd_size;
503     QemuOpts *sn_opts = NULL;
504     const char *sn_id_or_name = NULL;
505     const char *sopt = "hVb:o:p:rsnP:c:dvk:e:f:tl:x:T:D:";
506     struct option lopt[] = {
507         { "help", no_argument, NULL, 'h' },
508         { "version", no_argument, NULL, 'V' },
509         { "bind", required_argument, NULL, 'b' },
510         { "port", required_argument, NULL, 'p' },
511         { "socket", required_argument, NULL, 'k' },
512         { "offset", required_argument, NULL, 'o' },
513         { "read-only", no_argument, NULL, 'r' },
514         { "partition", required_argument, NULL, 'P' },
515         { "connect", required_argument, NULL, 'c' },
516         { "disconnect", no_argument, NULL, 'd' },
517         { "snapshot", no_argument, NULL, 's' },
518         { "load-snapshot", required_argument, NULL, 'l' },
519         { "nocache", no_argument, NULL, 'n' },
520         { "cache", required_argument, NULL, QEMU_NBD_OPT_CACHE },
521         { "aio", required_argument, NULL, QEMU_NBD_OPT_AIO },
522         { "discard", required_argument, NULL, QEMU_NBD_OPT_DISCARD },
523         { "detect-zeroes", required_argument, NULL,
524           QEMU_NBD_OPT_DETECT_ZEROES },
525         { "shared", required_argument, NULL, 'e' },
526         { "format", required_argument, NULL, 'f' },
527         { "persistent", no_argument, NULL, 't' },
528         { "verbose", no_argument, NULL, 'v' },
529         { "object", required_argument, NULL, QEMU_NBD_OPT_OBJECT },
530         { "export-name", required_argument, NULL, 'x' },
531         { "description", required_argument, NULL, 'D' },
532         { "tls-creds", required_argument, NULL, QEMU_NBD_OPT_TLSCREDS },
533         { "image-opts", no_argument, NULL, QEMU_NBD_OPT_IMAGE_OPTS },
534         { "trace", required_argument, NULL, 'T' },
535         { "fork", no_argument, NULL, QEMU_NBD_OPT_FORK },
536         { NULL, 0, NULL, 0 }
537     };
538     int ch;
539     int opt_ind = 0;
540     char *end;
541     int flags = BDRV_O_RDWR;
542     int partition = -1;
543     int ret = 0;
544     bool seen_cache = false;
545     bool seen_discard = false;
546     bool seen_aio = false;
547     pthread_t client_thread;
548     const char *fmt = NULL;
549     Error *local_err = NULL;
550     BlockdevDetectZeroesOptions detect_zeroes = BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
551     QDict *options = NULL;
552     const char *export_name = NULL;
553     const char *export_description = NULL;
554     const char *tlscredsid = NULL;
555     bool imageOpts = false;
556     bool writethrough = true;
557     char *trace_file = NULL;
558     bool fork_process = false;
559     int old_stderr = -1;
560     unsigned socket_activation;
561 
562     /* The client thread uses SIGTERM to interrupt the server.  A signal
563      * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
564      */
565     struct sigaction sa_sigterm;
566     memset(&sa_sigterm, 0, sizeof(sa_sigterm));
567     sa_sigterm.sa_handler = termsig_handler;
568     sigaction(SIGTERM, &sa_sigterm, NULL);
569 
570 #ifdef CONFIG_POSIX
571     signal(SIGPIPE, SIG_IGN);
572 #endif
573 
574     module_call_init(MODULE_INIT_TRACE);
575     qcrypto_init(&error_fatal);
576 
577     module_call_init(MODULE_INIT_QOM);
578     qemu_add_opts(&qemu_object_opts);
579     qemu_add_opts(&qemu_trace_opts);
580     qemu_init_exec_dir(argv[0]);
581 
582     while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
583         switch (ch) {
584         case 's':
585             flags |= BDRV_O_SNAPSHOT;
586             break;
587         case 'n':
588             optarg = (char *) "none";
589             /* fallthrough */
590         case QEMU_NBD_OPT_CACHE:
591             if (seen_cache) {
592                 error_report("-n and --cache can only be specified once");
593                 exit(EXIT_FAILURE);
594             }
595             seen_cache = true;
596             if (bdrv_parse_cache_mode(optarg, &flags, &writethrough) == -1) {
597                 error_report("Invalid cache mode `%s'", optarg);
598                 exit(EXIT_FAILURE);
599             }
600             break;
601         case QEMU_NBD_OPT_AIO:
602             if (seen_aio) {
603                 error_report("--aio can only be specified once");
604                 exit(EXIT_FAILURE);
605             }
606             seen_aio = true;
607             if (!strcmp(optarg, "native")) {
608                 flags |= BDRV_O_NATIVE_AIO;
609             } else if (!strcmp(optarg, "threads")) {
610                 /* this is the default */
611             } else {
612                error_report("invalid aio mode `%s'", optarg);
613                exit(EXIT_FAILURE);
614             }
615             break;
616         case QEMU_NBD_OPT_DISCARD:
617             if (seen_discard) {
618                 error_report("--discard can only be specified once");
619                 exit(EXIT_FAILURE);
620             }
621             seen_discard = true;
622             if (bdrv_parse_discard_flags(optarg, &flags) == -1) {
623                 error_report("Invalid discard mode `%s'", optarg);
624                 exit(EXIT_FAILURE);
625             }
626             break;
627         case QEMU_NBD_OPT_DETECT_ZEROES:
628             detect_zeroes =
629                 qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup,
630                                 optarg,
631                                 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
632                                 &local_err);
633             if (local_err) {
634                 error_reportf_err(local_err,
635                                   "Failed to parse detect_zeroes mode: ");
636                 exit(EXIT_FAILURE);
637             }
638             if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
639                 !(flags & BDRV_O_UNMAP)) {
640                 error_report("setting detect-zeroes to unmap is not allowed "
641                              "without setting discard operation to unmap");
642                 exit(EXIT_FAILURE);
643             }
644             break;
645         case 'b':
646             bindto = optarg;
647             break;
648         case 'p':
649             port = optarg;
650             break;
651         case 'o':
652                 dev_offset = strtoll (optarg, &end, 0);
653             if (*end) {
654                 error_report("Invalid offset `%s'", optarg);
655                 exit(EXIT_FAILURE);
656             }
657             if (dev_offset < 0) {
658                 error_report("Offset must be positive `%s'", optarg);
659                 exit(EXIT_FAILURE);
660             }
661             break;
662         case 'l':
663             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
664                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
665                                                   optarg, false);
666                 if (!sn_opts) {
667                     error_report("Failed in parsing snapshot param `%s'",
668                                  optarg);
669                     exit(EXIT_FAILURE);
670                 }
671             } else {
672                 sn_id_or_name = optarg;
673             }
674             /* fall through */
675         case 'r':
676             nbdflags |= NBD_FLAG_READ_ONLY;
677             flags &= ~BDRV_O_RDWR;
678             break;
679         case 'P':
680             partition = strtol(optarg, &end, 0);
681             if (*end) {
682                 error_report("Invalid partition `%s'", optarg);
683                 exit(EXIT_FAILURE);
684             }
685             if (partition < 1 || partition > 8) {
686                 error_report("Invalid partition %d", partition);
687                 exit(EXIT_FAILURE);
688             }
689             break;
690         case 'k':
691             sockpath = optarg;
692             if (sockpath[0] != '/') {
693                 error_report("socket path must be absolute");
694                 exit(EXIT_FAILURE);
695             }
696             break;
697         case 'd':
698             disconnect = true;
699             break;
700         case 'c':
701             device = optarg;
702             break;
703         case 'e':
704             shared = strtol(optarg, &end, 0);
705             if (*end) {
706                 error_report("Invalid shared device number '%s'", optarg);
707                 exit(EXIT_FAILURE);
708             }
709             if (shared < 1) {
710                 error_report("Shared device number must be greater than 0");
711                 exit(EXIT_FAILURE);
712             }
713             break;
714         case 'f':
715             fmt = optarg;
716             break;
717         case 't':
718             persistent = 1;
719             break;
720         case 'x':
721             export_name = optarg;
722             break;
723         case 'D':
724             export_description = optarg;
725             break;
726         case 'v':
727             verbose = 1;
728             break;
729         case 'V':
730             version(argv[0]);
731             exit(0);
732             break;
733         case 'h':
734             usage(argv[0]);
735             exit(0);
736             break;
737         case '?':
738             error_report("Try `%s --help' for more information.", argv[0]);
739             exit(EXIT_FAILURE);
740         case QEMU_NBD_OPT_OBJECT: {
741             QemuOpts *opts;
742             opts = qemu_opts_parse_noisily(&qemu_object_opts,
743                                            optarg, true);
744             if (!opts) {
745                 exit(EXIT_FAILURE);
746             }
747         }   break;
748         case QEMU_NBD_OPT_TLSCREDS:
749             tlscredsid = optarg;
750             break;
751         case QEMU_NBD_OPT_IMAGE_OPTS:
752             imageOpts = true;
753             break;
754         case 'T':
755             g_free(trace_file);
756             trace_file = trace_opt_parse(optarg);
757             break;
758         case QEMU_NBD_OPT_FORK:
759             fork_process = true;
760             break;
761         }
762     }
763 
764     if ((argc - optind) != 1) {
765         error_report("Invalid number of arguments");
766         error_printf("Try `%s --help' for more information.\n", argv[0]);
767         exit(EXIT_FAILURE);
768     }
769 
770     if (qemu_opts_foreach(&qemu_object_opts,
771                           user_creatable_add_opts_foreach,
772                           NULL, NULL)) {
773         exit(EXIT_FAILURE);
774     }
775 
776     if (!trace_init_backends()) {
777         exit(1);
778     }
779     trace_init_file(trace_file);
780     qemu_set_log(LOG_TRACE);
781 
782     socket_activation = check_socket_activation();
783     if (socket_activation == 0) {
784         setup_address_and_port(&bindto, &port);
785     } else {
786         /* Using socket activation - check user didn't use -p etc. */
787         const char *err_msg = socket_activation_validate_opts(device, sockpath,
788                                                               bindto, port);
789         if (err_msg != NULL) {
790             error_report("%s", err_msg);
791             exit(EXIT_FAILURE);
792         }
793 
794         /* qemu-nbd can only listen on a single socket.  */
795         if (socket_activation > 1) {
796             error_report("qemu-nbd does not support socket activation with %s > 1",
797                          "LISTEN_FDS");
798             exit(EXIT_FAILURE);
799         }
800     }
801 
802     if (tlscredsid) {
803         if (sockpath) {
804             error_report("TLS is only supported with IPv4/IPv6");
805             exit(EXIT_FAILURE);
806         }
807         if (device) {
808             error_report("TLS is not supported with a host device");
809             exit(EXIT_FAILURE);
810         }
811         if (!export_name) {
812             /* Set the default NBD protocol export name, since
813              * we *must* use new style protocol for TLS */
814             export_name = "";
815         }
816         tlscreds = nbd_get_tls_creds(tlscredsid, &local_err);
817         if (local_err) {
818             error_report("Failed to get TLS creds %s",
819                          error_get_pretty(local_err));
820             exit(EXIT_FAILURE);
821         }
822     }
823 
824     if (disconnect) {
825         int nbdfd = open(argv[optind], O_RDWR);
826         if (nbdfd < 0) {
827             error_report("Cannot open %s: %s", argv[optind],
828                          strerror(errno));
829             exit(EXIT_FAILURE);
830         }
831         nbd_disconnect(nbdfd);
832 
833         close(nbdfd);
834 
835         printf("%s disconnected\n", argv[optind]);
836 
837         return 0;
838     }
839 
840     if ((device && !verbose) || fork_process) {
841         int stderr_fd[2];
842         pid_t pid;
843         int ret;
844 
845         if (qemu_pipe(stderr_fd) < 0) {
846             error_report("Error setting up communication pipe: %s",
847                          strerror(errno));
848             exit(EXIT_FAILURE);
849         }
850 
851         /* Now daemonize, but keep a communication channel open to
852          * print errors and exit with the proper status code.
853          */
854         pid = fork();
855         if (pid < 0) {
856             error_report("Failed to fork: %s", strerror(errno));
857             exit(EXIT_FAILURE);
858         } else if (pid == 0) {
859             close(stderr_fd[0]);
860             ret = qemu_daemon(1, 0);
861 
862             /* Temporarily redirect stderr to the parent's pipe...  */
863             old_stderr = dup(STDERR_FILENO);
864             dup2(stderr_fd[1], STDERR_FILENO);
865             if (ret < 0) {
866                 error_report("Failed to daemonize: %s", strerror(errno));
867                 exit(EXIT_FAILURE);
868             }
869 
870             /* ... close the descriptor we inherited and go on.  */
871             close(stderr_fd[1]);
872         } else {
873             bool errors = false;
874             char *buf;
875 
876             /* In the parent.  Print error messages from the child until
877              * it closes the pipe.
878              */
879             close(stderr_fd[1]);
880             buf = g_malloc(1024);
881             while ((ret = read(stderr_fd[0], buf, 1024)) > 0) {
882                 errors = true;
883                 ret = qemu_write_full(STDERR_FILENO, buf, ret);
884                 if (ret < 0) {
885                     exit(EXIT_FAILURE);
886                 }
887             }
888             if (ret < 0) {
889                 error_report("Cannot read from daemon: %s",
890                              strerror(errno));
891                 exit(EXIT_FAILURE);
892             }
893 
894             /* Usually the daemon should not print any message.
895              * Exit with zero status in that case.
896              */
897             exit(errors);
898         }
899     }
900 
901     if (device != NULL && sockpath == NULL) {
902         sockpath = g_malloc(128);
903         snprintf(sockpath, 128, SOCKET_PATH, basename(device));
904     }
905 
906     server = qio_net_listener_new();
907     if (socket_activation == 0) {
908         saddr = nbd_build_socket_address(sockpath, bindto, port);
909         if (qio_net_listener_open_sync(server, saddr, &local_err) < 0) {
910             object_unref(OBJECT(server));
911             error_report_err(local_err);
912             exit(EXIT_FAILURE);
913         }
914     } else {
915         size_t i;
916         /* See comment in check_socket_activation above. */
917         for (i = 0; i < socket_activation; i++) {
918             QIOChannelSocket *sioc;
919             sioc = qio_channel_socket_new_fd(FIRST_SOCKET_ACTIVATION_FD + i,
920                                              &local_err);
921             if (sioc == NULL) {
922                 object_unref(OBJECT(server));
923                 error_report("Failed to use socket activation: %s",
924                              error_get_pretty(local_err));
925                 exit(EXIT_FAILURE);
926             }
927             qio_net_listener_add(server, sioc);
928             object_unref(OBJECT(sioc));
929         }
930     }
931 
932     if (qemu_init_main_loop(&local_err)) {
933         error_report_err(local_err);
934         exit(EXIT_FAILURE);
935     }
936     bdrv_init();
937     atexit(qemu_nbd_shutdown);
938 
939     srcpath = argv[optind];
940     if (imageOpts) {
941         QemuOpts *opts;
942         if (fmt) {
943             error_report("--image-opts and -f are mutually exclusive");
944             exit(EXIT_FAILURE);
945         }
946         opts = qemu_opts_parse_noisily(&file_opts, srcpath, true);
947         if (!opts) {
948             qemu_opts_reset(&file_opts);
949             exit(EXIT_FAILURE);
950         }
951         options = qemu_opts_to_qdict(opts, NULL);
952         qemu_opts_reset(&file_opts);
953         blk = blk_new_open(NULL, NULL, options, flags, &local_err);
954     } else {
955         if (fmt) {
956             options = qdict_new();
957             qdict_put_str(options, "driver", fmt);
958         }
959         blk = blk_new_open(srcpath, NULL, options, flags, &local_err);
960     }
961 
962     if (!blk) {
963         error_reportf_err(local_err, "Failed to blk_new_open '%s': ",
964                           argv[optind]);
965         exit(EXIT_FAILURE);
966     }
967     bs = blk_bs(blk);
968 
969     blk_set_enable_write_cache(blk, !writethrough);
970 
971     if (sn_opts) {
972         ret = bdrv_snapshot_load_tmp(bs,
973                                      qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
974                                      qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
975                                      &local_err);
976     } else if (sn_id_or_name) {
977         ret = bdrv_snapshot_load_tmp_by_id_or_name(bs, sn_id_or_name,
978                                                    &local_err);
979     }
980     if (ret < 0) {
981         error_reportf_err(local_err, "Failed to load snapshot: ");
982         exit(EXIT_FAILURE);
983     }
984 
985     bs->detect_zeroes = detect_zeroes;
986     fd_size = blk_getlength(blk);
987     if (fd_size < 0) {
988         error_report("Failed to determine the image length: %s",
989                      strerror(-fd_size));
990         exit(EXIT_FAILURE);
991     }
992 
993     if (dev_offset >= fd_size) {
994         error_report("Offset (%lld) has to be smaller than the image size "
995                      "(%lld)",
996                      (long long int)dev_offset, (long long int)fd_size);
997         exit(EXIT_FAILURE);
998     }
999     fd_size -= dev_offset;
1000 
1001     if (partition != -1) {
1002         ret = find_partition(blk, partition, &dev_offset, &fd_size);
1003         if (ret < 0) {
1004             error_report("Could not find partition %d: %s", partition,
1005                          strerror(-ret));
1006             exit(EXIT_FAILURE);
1007         }
1008     }
1009 
1010     exp = nbd_export_new(bs, dev_offset, fd_size, nbdflags, nbd_export_closed,
1011                          writethrough, NULL, &local_err);
1012     if (!exp) {
1013         error_report_err(local_err);
1014         exit(EXIT_FAILURE);
1015     }
1016     if (export_name) {
1017         nbd_export_set_name(exp, export_name);
1018         nbd_export_set_description(exp, export_description);
1019         newproto = true;
1020     } else if (export_description) {
1021         error_report("Export description requires an export name");
1022         exit(EXIT_FAILURE);
1023     }
1024 
1025     if (device) {
1026         int ret;
1027 
1028         ret = pthread_create(&client_thread, NULL, nbd_client_thread, device);
1029         if (ret != 0) {
1030             error_report("Failed to create client thread: %s", strerror(ret));
1031             exit(EXIT_FAILURE);
1032         }
1033     } else {
1034         /* Shut up GCC warnings.  */
1035         memset(&client_thread, 0, sizeof(client_thread));
1036     }
1037 
1038     nbd_update_server_watch();
1039 
1040     /* now when the initialization is (almost) complete, chdir("/")
1041      * to free any busy filesystems */
1042     if (chdir("/") < 0) {
1043         error_report("Could not chdir to root directory: %s",
1044                      strerror(errno));
1045         exit(EXIT_FAILURE);
1046     }
1047 
1048     if (fork_process) {
1049         dup2(old_stderr, STDERR_FILENO);
1050         close(old_stderr);
1051     }
1052 
1053     state = RUNNING;
1054     do {
1055         main_loop_wait(false);
1056         if (state == TERMINATE) {
1057             state = TERMINATING;
1058             nbd_export_close(exp);
1059             nbd_export_put(exp);
1060             exp = NULL;
1061         }
1062     } while (state != TERMINATED);
1063 
1064     blk_unref(blk);
1065     if (sockpath) {
1066         unlink(sockpath);
1067     }
1068 
1069     qemu_opts_del(sn_opts);
1070 
1071     if (device) {
1072         void *ret;
1073         pthread_join(client_thread, &ret);
1074         exit(ret != NULL);
1075     } else {
1076         exit(EXIT_SUCCESS);
1077     }
1078 }
1079