xref: /openbmc/qemu/qemu-nbd.c (revision 53d28455)
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 int verbose;
60 static char *srcpath;
61 static SocketAddress *saddr;
62 static int persistent = 0;
63 static enum { RUNNING, TERMINATE, TERMINATING, TERMINATED } state;
64 static int shared = 1;
65 static int nb_fds;
66 static QIONetListener *server;
67 static QCryptoTLSCreds *tlscreds;
68 
69 static void usage(const char *name)
70 {
71     (printf) (
72 "Usage: %s [OPTIONS] FILE\n"
73 "QEMU Disk Network Block Device Server\n"
74 "\n"
75 "  -h, --help                display this help and exit\n"
76 "  -V, --version             output version information and exit\n"
77 "\n"
78 "Connection properties:\n"
79 "  -p, --port=PORT           port to listen on (default `%d')\n"
80 "  -b, --bind=IFACE          interface to bind to (default `0.0.0.0')\n"
81 "  -k, --socket=PATH         path to the unix socket\n"
82 "                            (default '"SOCKET_PATH"')\n"
83 "  -e, --shared=NUM          device can be shared by NUM clients (default '1')\n"
84 "  -t, --persistent          don't exit on the last connection\n"
85 "  -v, --verbose             display extra debugging information\n"
86 "  -x, --export-name=NAME    expose export by name (default is empty string)\n"
87 "  -D, --description=TEXT    export a human-readable description\n"
88 "\n"
89 "Exposing part of the image:\n"
90 "  -o, --offset=OFFSET       offset into the image\n"
91 "  -P, --partition=NUM       only expose partition NUM\n"
92 "\n"
93 "General purpose options:\n"
94 "  --object type,id=ID,...   define an object such as 'secret' for providing\n"
95 "                            passwords and/or encryption keys\n"
96 "  --tls-creds=ID            use id of an earlier --object to provide TLS\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(cioc, tlscreds, NULL, nbd_client_closed);
358 }
359 
360 static void nbd_update_server_watch(void)
361 {
362     if (nbd_can_accept()) {
363         qio_net_listener_set_client_func(server, nbd_accept, NULL, NULL);
364     } else {
365         qio_net_listener_set_client_func(server, NULL, NULL, NULL);
366     }
367 }
368 
369 
370 static SocketAddress *nbd_build_socket_address(const char *sockpath,
371                                                const char *bindto,
372                                                const char *port)
373 {
374     SocketAddress *saddr;
375 
376     saddr = g_new0(SocketAddress, 1);
377     if (sockpath) {
378         saddr->type = SOCKET_ADDRESS_TYPE_UNIX;
379         saddr->u.q_unix.path = g_strdup(sockpath);
380     } else {
381         InetSocketAddress *inet;
382         saddr->type = SOCKET_ADDRESS_TYPE_INET;
383         inet = &saddr->u.inet;
384         inet->host = g_strdup(bindto);
385         if (port) {
386             inet->port = g_strdup(port);
387         } else  {
388             inet->port = g_strdup_printf("%d", NBD_DEFAULT_PORT);
389         }
390     }
391 
392     return saddr;
393 }
394 
395 
396 static QemuOptsList file_opts = {
397     .name = "file",
398     .implied_opt_name = "file",
399     .head = QTAILQ_HEAD_INITIALIZER(file_opts.head),
400     .desc = {
401         /* no elements => accept any params */
402         { /* end of list */ }
403     },
404 };
405 
406 static QemuOptsList qemu_object_opts = {
407     .name = "object",
408     .implied_opt_name = "qom-type",
409     .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head),
410     .desc = {
411         { }
412     },
413 };
414 
415 
416 
417 static QCryptoTLSCreds *nbd_get_tls_creds(const char *id, Error **errp)
418 {
419     Object *obj;
420     QCryptoTLSCreds *creds;
421 
422     obj = object_resolve_path_component(
423         object_get_objects_root(), id);
424     if (!obj) {
425         error_setg(errp, "No TLS credentials with id '%s'",
426                    id);
427         return NULL;
428     }
429     creds = (QCryptoTLSCreds *)
430         object_dynamic_cast(obj, TYPE_QCRYPTO_TLS_CREDS);
431     if (!creds) {
432         error_setg(errp, "Object with id '%s' is not TLS credentials",
433                    id);
434         return NULL;
435     }
436 
437     if (creds->endpoint != QCRYPTO_TLS_CREDS_ENDPOINT_SERVER) {
438         error_setg(errp,
439                    "Expecting TLS credentials with a server endpoint");
440         return NULL;
441     }
442     object_ref(obj);
443     return creds;
444 }
445 
446 static void setup_address_and_port(const char **address, const char **port)
447 {
448     if (*address == NULL) {
449         *address = "0.0.0.0";
450     }
451 
452     if (*port == NULL) {
453         *port = stringify(NBD_DEFAULT_PORT);
454     }
455 }
456 
457 /*
458  * Check socket parameters compatibility when socket activation is used.
459  */
460 static const char *socket_activation_validate_opts(const char *device,
461                                                    const char *sockpath,
462                                                    const char *address,
463                                                    const char *port)
464 {
465     if (device != NULL) {
466         return "NBD device can't be set when using socket activation";
467     }
468 
469     if (sockpath != NULL) {
470         return "Unix socket can't be set when using socket activation";
471     }
472 
473     if (address != NULL) {
474         return "The interface can't be set when using socket activation";
475     }
476 
477     if (port != NULL) {
478         return "TCP port number can't be set when using socket activation";
479     }
480 
481     return NULL;
482 }
483 
484 static void qemu_nbd_shutdown(void)
485 {
486     job_cancel_sync_all();
487     bdrv_close_all();
488 }
489 
490 int main(int argc, char **argv)
491 {
492     BlockBackend *blk;
493     BlockDriverState *bs;
494     off_t dev_offset = 0;
495     uint16_t nbdflags = 0;
496     bool disconnect = false;
497     const char *bindto = NULL;
498     const char *port = NULL;
499     char *sockpath = NULL;
500     char *device = NULL;
501     off_t fd_size;
502     QemuOpts *sn_opts = NULL;
503     const char *sn_id_or_name = NULL;
504     const char *sopt = "hVb:o:p:rsnP:c:dvk:e:f:tl:x:T:D:";
505     struct option lopt[] = {
506         { "help", no_argument, NULL, 'h' },
507         { "version", no_argument, NULL, 'V' },
508         { "bind", required_argument, NULL, 'b' },
509         { "port", required_argument, NULL, 'p' },
510         { "socket", required_argument, NULL, 'k' },
511         { "offset", required_argument, NULL, 'o' },
512         { "read-only", no_argument, NULL, 'r' },
513         { "partition", required_argument, NULL, 'P' },
514         { "connect", required_argument, NULL, 'c' },
515         { "disconnect", no_argument, NULL, 'd' },
516         { "snapshot", no_argument, NULL, 's' },
517         { "load-snapshot", required_argument, NULL, 'l' },
518         { "nocache", no_argument, NULL, 'n' },
519         { "cache", required_argument, NULL, QEMU_NBD_OPT_CACHE },
520         { "aio", required_argument, NULL, QEMU_NBD_OPT_AIO },
521         { "discard", required_argument, NULL, QEMU_NBD_OPT_DISCARD },
522         { "detect-zeroes", required_argument, NULL,
523           QEMU_NBD_OPT_DETECT_ZEROES },
524         { "shared", required_argument, NULL, 'e' },
525         { "format", required_argument, NULL, 'f' },
526         { "persistent", no_argument, NULL, 't' },
527         { "verbose", no_argument, NULL, 'v' },
528         { "object", required_argument, NULL, QEMU_NBD_OPT_OBJECT },
529         { "export-name", required_argument, NULL, 'x' },
530         { "description", required_argument, NULL, 'D' },
531         { "tls-creds", required_argument, NULL, QEMU_NBD_OPT_TLSCREDS },
532         { "image-opts", no_argument, NULL, QEMU_NBD_OPT_IMAGE_OPTS },
533         { "trace", required_argument, NULL, 'T' },
534         { "fork", no_argument, NULL, QEMU_NBD_OPT_FORK },
535         { NULL, 0, NULL, 0 }
536     };
537     int ch;
538     int opt_ind = 0;
539     char *end;
540     int flags = BDRV_O_RDWR;
541     int partition = -1;
542     int ret = 0;
543     bool seen_cache = false;
544     bool seen_discard = false;
545     bool seen_aio = false;
546     pthread_t client_thread;
547     const char *fmt = NULL;
548     Error *local_err = NULL;
549     BlockdevDetectZeroesOptions detect_zeroes = BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
550     QDict *options = NULL;
551     const char *export_name = ""; /* Default export name */
552     const char *export_description = NULL;
553     const char *tlscredsid = NULL;
554     bool imageOpts = false;
555     bool writethrough = true;
556     char *trace_file = NULL;
557     bool fork_process = false;
558     int old_stderr = -1;
559     unsigned socket_activation;
560 
561     /* The client thread uses SIGTERM to interrupt the server.  A signal
562      * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
563      */
564     struct sigaction sa_sigterm;
565     memset(&sa_sigterm, 0, sizeof(sa_sigterm));
566     sa_sigterm.sa_handler = termsig_handler;
567     sigaction(SIGTERM, &sa_sigterm, NULL);
568 
569 #ifdef CONFIG_POSIX
570     signal(SIGPIPE, SIG_IGN);
571 #endif
572 
573     module_call_init(MODULE_INIT_TRACE);
574     qcrypto_init(&error_fatal);
575 
576     module_call_init(MODULE_INIT_QOM);
577     qemu_add_opts(&qemu_object_opts);
578     qemu_add_opts(&qemu_trace_opts);
579     qemu_init_exec_dir(argv[0]);
580 
581     while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
582         switch (ch) {
583         case 's':
584             flags |= BDRV_O_SNAPSHOT;
585             break;
586         case 'n':
587             optarg = (char *) "none";
588             /* fallthrough */
589         case QEMU_NBD_OPT_CACHE:
590             if (seen_cache) {
591                 error_report("-n and --cache can only be specified once");
592                 exit(EXIT_FAILURE);
593             }
594             seen_cache = true;
595             if (bdrv_parse_cache_mode(optarg, &flags, &writethrough) == -1) {
596                 error_report("Invalid cache mode `%s'", optarg);
597                 exit(EXIT_FAILURE);
598             }
599             break;
600         case QEMU_NBD_OPT_AIO:
601             if (seen_aio) {
602                 error_report("--aio can only be specified once");
603                 exit(EXIT_FAILURE);
604             }
605             seen_aio = true;
606             if (!strcmp(optarg, "native")) {
607                 flags |= BDRV_O_NATIVE_AIO;
608             } else if (!strcmp(optarg, "threads")) {
609                 /* this is the default */
610             } else {
611                error_report("invalid aio mode `%s'", optarg);
612                exit(EXIT_FAILURE);
613             }
614             break;
615         case QEMU_NBD_OPT_DISCARD:
616             if (seen_discard) {
617                 error_report("--discard can only be specified once");
618                 exit(EXIT_FAILURE);
619             }
620             seen_discard = true;
621             if (bdrv_parse_discard_flags(optarg, &flags) == -1) {
622                 error_report("Invalid discard mode `%s'", optarg);
623                 exit(EXIT_FAILURE);
624             }
625             break;
626         case QEMU_NBD_OPT_DETECT_ZEROES:
627             detect_zeroes =
628                 qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup,
629                                 optarg,
630                                 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
631                                 &local_err);
632             if (local_err) {
633                 error_reportf_err(local_err,
634                                   "Failed to parse detect_zeroes mode: ");
635                 exit(EXIT_FAILURE);
636             }
637             if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
638                 !(flags & BDRV_O_UNMAP)) {
639                 error_report("setting detect-zeroes to unmap is not allowed "
640                              "without setting discard operation to unmap");
641                 exit(EXIT_FAILURE);
642             }
643             break;
644         case 'b':
645             bindto = optarg;
646             break;
647         case 'p':
648             port = optarg;
649             break;
650         case 'o':
651                 dev_offset = strtoll (optarg, &end, 0);
652             if (*end) {
653                 error_report("Invalid offset `%s'", optarg);
654                 exit(EXIT_FAILURE);
655             }
656             if (dev_offset < 0) {
657                 error_report("Offset must be positive `%s'", optarg);
658                 exit(EXIT_FAILURE);
659             }
660             break;
661         case 'l':
662             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
663                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
664                                                   optarg, false);
665                 if (!sn_opts) {
666                     error_report("Failed in parsing snapshot param `%s'",
667                                  optarg);
668                     exit(EXIT_FAILURE);
669                 }
670             } else {
671                 sn_id_or_name = optarg;
672             }
673             /* fall through */
674         case 'r':
675             nbdflags |= NBD_FLAG_READ_ONLY;
676             flags &= ~BDRV_O_RDWR;
677             break;
678         case 'P':
679             partition = strtol(optarg, &end, 0);
680             if (*end) {
681                 error_report("Invalid partition `%s'", optarg);
682                 exit(EXIT_FAILURE);
683             }
684             if (partition < 1 || partition > 8) {
685                 error_report("Invalid partition %d", partition);
686                 exit(EXIT_FAILURE);
687             }
688             break;
689         case 'k':
690             sockpath = optarg;
691             if (sockpath[0] != '/') {
692                 error_report("socket path must be absolute");
693                 exit(EXIT_FAILURE);
694             }
695             break;
696         case 'd':
697             disconnect = true;
698             break;
699         case 'c':
700             device = optarg;
701             break;
702         case 'e':
703             shared = strtol(optarg, &end, 0);
704             if (*end) {
705                 error_report("Invalid shared device number '%s'", optarg);
706                 exit(EXIT_FAILURE);
707             }
708             if (shared < 1) {
709                 error_report("Shared device number must be greater than 0");
710                 exit(EXIT_FAILURE);
711             }
712             break;
713         case 'f':
714             fmt = optarg;
715             break;
716         case 't':
717             persistent = 1;
718             break;
719         case 'x':
720             export_name = optarg;
721             break;
722         case 'D':
723             export_description = optarg;
724             break;
725         case 'v':
726             verbose = 1;
727             break;
728         case 'V':
729             version(argv[0]);
730             exit(0);
731             break;
732         case 'h':
733             usage(argv[0]);
734             exit(0);
735             break;
736         case '?':
737             error_report("Try `%s --help' for more information.", argv[0]);
738             exit(EXIT_FAILURE);
739         case QEMU_NBD_OPT_OBJECT: {
740             QemuOpts *opts;
741             opts = qemu_opts_parse_noisily(&qemu_object_opts,
742                                            optarg, true);
743             if (!opts) {
744                 exit(EXIT_FAILURE);
745             }
746         }   break;
747         case QEMU_NBD_OPT_TLSCREDS:
748             tlscredsid = optarg;
749             break;
750         case QEMU_NBD_OPT_IMAGE_OPTS:
751             imageOpts = true;
752             break;
753         case 'T':
754             g_free(trace_file);
755             trace_file = trace_opt_parse(optarg);
756             break;
757         case QEMU_NBD_OPT_FORK:
758             fork_process = true;
759             break;
760         }
761     }
762 
763     if ((argc - optind) != 1) {
764         error_report("Invalid number of arguments");
765         error_printf("Try `%s --help' for more information.\n", argv[0]);
766         exit(EXIT_FAILURE);
767     }
768 
769     qemu_opts_foreach(&qemu_object_opts,
770                       user_creatable_add_opts_foreach,
771                       NULL, &error_fatal);
772 
773     if (!trace_init_backends()) {
774         exit(1);
775     }
776     trace_init_file(trace_file);
777     qemu_set_log(LOG_TRACE);
778 
779     socket_activation = check_socket_activation();
780     if (socket_activation == 0) {
781         setup_address_and_port(&bindto, &port);
782     } else {
783         /* Using socket activation - check user didn't use -p etc. */
784         const char *err_msg = socket_activation_validate_opts(device, sockpath,
785                                                               bindto, port);
786         if (err_msg != NULL) {
787             error_report("%s", err_msg);
788             exit(EXIT_FAILURE);
789         }
790 
791         /* qemu-nbd can only listen on a single socket.  */
792         if (socket_activation > 1) {
793             error_report("qemu-nbd does not support socket activation with %s > 1",
794                          "LISTEN_FDS");
795             exit(EXIT_FAILURE);
796         }
797     }
798 
799     if (tlscredsid) {
800         if (sockpath) {
801             error_report("TLS is only supported with IPv4/IPv6");
802             exit(EXIT_FAILURE);
803         }
804         if (device) {
805             error_report("TLS is not supported with a host device");
806             exit(EXIT_FAILURE);
807         }
808         tlscreds = nbd_get_tls_creds(tlscredsid, &local_err);
809         if (local_err) {
810             error_report("Failed to get TLS creds %s",
811                          error_get_pretty(local_err));
812             exit(EXIT_FAILURE);
813         }
814     }
815 
816     if (disconnect) {
817         int nbdfd = open(argv[optind], O_RDWR);
818         if (nbdfd < 0) {
819             error_report("Cannot open %s: %s", argv[optind],
820                          strerror(errno));
821             exit(EXIT_FAILURE);
822         }
823         nbd_disconnect(nbdfd);
824 
825         close(nbdfd);
826 
827         printf("%s disconnected\n", argv[optind]);
828 
829         return 0;
830     }
831 
832     if ((device && !verbose) || fork_process) {
833         int stderr_fd[2];
834         pid_t pid;
835         int ret;
836 
837         if (qemu_pipe(stderr_fd) < 0) {
838             error_report("Error setting up communication pipe: %s",
839                          strerror(errno));
840             exit(EXIT_FAILURE);
841         }
842 
843         /* Now daemonize, but keep a communication channel open to
844          * print errors and exit with the proper status code.
845          */
846         pid = fork();
847         if (pid < 0) {
848             error_report("Failed to fork: %s", strerror(errno));
849             exit(EXIT_FAILURE);
850         } else if (pid == 0) {
851             close(stderr_fd[0]);
852             ret = qemu_daemon(1, 0);
853 
854             /* Temporarily redirect stderr to the parent's pipe...  */
855             old_stderr = dup(STDERR_FILENO);
856             dup2(stderr_fd[1], STDERR_FILENO);
857             if (ret < 0) {
858                 error_report("Failed to daemonize: %s", strerror(errno));
859                 exit(EXIT_FAILURE);
860             }
861 
862             /* ... close the descriptor we inherited and go on.  */
863             close(stderr_fd[1]);
864         } else {
865             bool errors = false;
866             char *buf;
867 
868             /* In the parent.  Print error messages from the child until
869              * it closes the pipe.
870              */
871             close(stderr_fd[1]);
872             buf = g_malloc(1024);
873             while ((ret = read(stderr_fd[0], buf, 1024)) > 0) {
874                 errors = true;
875                 ret = qemu_write_full(STDERR_FILENO, buf, ret);
876                 if (ret < 0) {
877                     exit(EXIT_FAILURE);
878                 }
879             }
880             if (ret < 0) {
881                 error_report("Cannot read from daemon: %s",
882                              strerror(errno));
883                 exit(EXIT_FAILURE);
884             }
885 
886             /* Usually the daemon should not print any message.
887              * Exit with zero status in that case.
888              */
889             exit(errors);
890         }
891     }
892 
893     if (device != NULL && sockpath == NULL) {
894         sockpath = g_malloc(128);
895         snprintf(sockpath, 128, SOCKET_PATH, basename(device));
896     }
897 
898     server = qio_net_listener_new();
899     if (socket_activation == 0) {
900         saddr = nbd_build_socket_address(sockpath, bindto, port);
901         if (qio_net_listener_open_sync(server, saddr, &local_err) < 0) {
902             object_unref(OBJECT(server));
903             error_report_err(local_err);
904             exit(EXIT_FAILURE);
905         }
906     } else {
907         size_t i;
908         /* See comment in check_socket_activation above. */
909         for (i = 0; i < socket_activation; i++) {
910             QIOChannelSocket *sioc;
911             sioc = qio_channel_socket_new_fd(FIRST_SOCKET_ACTIVATION_FD + i,
912                                              &local_err);
913             if (sioc == NULL) {
914                 object_unref(OBJECT(server));
915                 error_report("Failed to use socket activation: %s",
916                              error_get_pretty(local_err));
917                 exit(EXIT_FAILURE);
918             }
919             qio_net_listener_add(server, sioc);
920             object_unref(OBJECT(sioc));
921         }
922     }
923 
924     if (qemu_init_main_loop(&local_err)) {
925         error_report_err(local_err);
926         exit(EXIT_FAILURE);
927     }
928     bdrv_init();
929     atexit(qemu_nbd_shutdown);
930 
931     srcpath = argv[optind];
932     if (imageOpts) {
933         QemuOpts *opts;
934         if (fmt) {
935             error_report("--image-opts and -f are mutually exclusive");
936             exit(EXIT_FAILURE);
937         }
938         opts = qemu_opts_parse_noisily(&file_opts, srcpath, true);
939         if (!opts) {
940             qemu_opts_reset(&file_opts);
941             exit(EXIT_FAILURE);
942         }
943         options = qemu_opts_to_qdict(opts, NULL);
944         qemu_opts_reset(&file_opts);
945         blk = blk_new_open(NULL, NULL, options, flags, &local_err);
946     } else {
947         if (fmt) {
948             options = qdict_new();
949             qdict_put_str(options, "driver", fmt);
950         }
951         blk = blk_new_open(srcpath, NULL, options, flags, &local_err);
952     }
953 
954     if (!blk) {
955         error_reportf_err(local_err, "Failed to blk_new_open '%s': ",
956                           argv[optind]);
957         exit(EXIT_FAILURE);
958     }
959     bs = blk_bs(blk);
960 
961     blk_set_enable_write_cache(blk, !writethrough);
962 
963     if (sn_opts) {
964         ret = bdrv_snapshot_load_tmp(bs,
965                                      qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
966                                      qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
967                                      &local_err);
968     } else if (sn_id_or_name) {
969         ret = bdrv_snapshot_load_tmp_by_id_or_name(bs, sn_id_or_name,
970                                                    &local_err);
971     }
972     if (ret < 0) {
973         error_reportf_err(local_err, "Failed to load snapshot: ");
974         exit(EXIT_FAILURE);
975     }
976 
977     bs->detect_zeroes = detect_zeroes;
978     fd_size = blk_getlength(blk);
979     if (fd_size < 0) {
980         error_report("Failed to determine the image length: %s",
981                      strerror(-fd_size));
982         exit(EXIT_FAILURE);
983     }
984 
985     if (dev_offset >= fd_size) {
986         error_report("Offset (%lld) has to be smaller than the image size "
987                      "(%lld)",
988                      (long long int)dev_offset, (long long int)fd_size);
989         exit(EXIT_FAILURE);
990     }
991     fd_size -= dev_offset;
992 
993     if (partition != -1) {
994         ret = find_partition(blk, partition, &dev_offset, &fd_size);
995         if (ret < 0) {
996             error_report("Could not find partition %d: %s", partition,
997                          strerror(-ret));
998             exit(EXIT_FAILURE);
999         }
1000     }
1001 
1002     exp = nbd_export_new(bs, dev_offset, fd_size, nbdflags, nbd_export_closed,
1003                          writethrough, NULL, &error_fatal);
1004     nbd_export_set_name(exp, export_name);
1005     nbd_export_set_description(exp, export_description);
1006 
1007     if (device) {
1008         int ret;
1009 
1010         ret = pthread_create(&client_thread, NULL, nbd_client_thread, device);
1011         if (ret != 0) {
1012             error_report("Failed to create client thread: %s", strerror(ret));
1013             exit(EXIT_FAILURE);
1014         }
1015     } else {
1016         /* Shut up GCC warnings.  */
1017         memset(&client_thread, 0, sizeof(client_thread));
1018     }
1019 
1020     nbd_update_server_watch();
1021 
1022     /* now when the initialization is (almost) complete, chdir("/")
1023      * to free any busy filesystems */
1024     if (chdir("/") < 0) {
1025         error_report("Could not chdir to root directory: %s",
1026                      strerror(errno));
1027         exit(EXIT_FAILURE);
1028     }
1029 
1030     if (fork_process) {
1031         dup2(old_stderr, STDERR_FILENO);
1032         close(old_stderr);
1033     }
1034 
1035     state = RUNNING;
1036     do {
1037         main_loop_wait(false);
1038         if (state == TERMINATE) {
1039             state = TERMINATING;
1040             nbd_export_close(exp);
1041             nbd_export_put(exp);
1042             exp = NULL;
1043         }
1044     } while (state != TERMINATED);
1045 
1046     blk_unref(blk);
1047     if (sockpath) {
1048         unlink(sockpath);
1049     }
1050 
1051     qemu_opts_del(sn_opts);
1052 
1053     if (device) {
1054         void *ret;
1055         pthread_join(client_thread, &ret);
1056         exit(ret != NULL);
1057     } else {
1058         exit(EXIT_SUCCESS);
1059     }
1060 }
1061