qemu

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 "qemu/help-texts.h"
25
#include "qapi/error.h"
26
#include "qemu/cutils.h"
27
#include "sysemu/block-backend.h"
28
#include "sysemu/runstate.h" /* for qemu_system_killed() prototype */
29
#include "block/block_int.h"
30
#include "block/nbd.h"
31
#include "qemu/main-loop.h"
32
#include "qemu/module.h"
33
#include "qemu/option.h"
34
#include "qemu/error-report.h"
35
#include "qemu/config-file.h"
36
#include "qemu/bswap.h"
37
#include "qemu/log.h"
38
#include "qemu/systemd.h"
39
#include "block/snapshot.h"
40
#include "qapi/qmp/qdict.h"
41
#include "qapi/qmp/qstring.h"
42
#include "qom/object_interfaces.h"
43
#include "io/channel-socket.h"
44
#include "io/net-listener.h"
45
#include "crypto/init.h"
46
#include "crypto/tlscreds.h"
47
#include "trace/control.h"
48
#include "qemu-version.h"
49

50
#ifdef CONFIG_SELINUX
51
#include <selinux/selinux.h>
52
#endif
53

54
#ifdef __linux__
55
#define HAVE_NBD_DEVICE 1
56
#else
57
#define HAVE_NBD_DEVICE 0
58
#endif
59

60
#define SOCKET_PATH                "/var/lock/qemu-nbd-%s"
61
#define QEMU_NBD_OPT_CACHE         256
62
#define QEMU_NBD_OPT_AIO           257
63
#define QEMU_NBD_OPT_DISCARD       258
64
#define QEMU_NBD_OPT_DETECT_ZEROES 259
65
#define QEMU_NBD_OPT_OBJECT        260
66
#define QEMU_NBD_OPT_TLSCREDS      261
67
#define QEMU_NBD_OPT_IMAGE_OPTS    262
68
#define QEMU_NBD_OPT_FORK          263
69
#define QEMU_NBD_OPT_TLSAUTHZ      264
70
#define QEMU_NBD_OPT_PID_FILE      265
71
#define QEMU_NBD_OPT_SELINUX_LABEL 266
72
#define QEMU_NBD_OPT_TLSHOSTNAME   267
73

74
#define MBR_SIZE 512
75

76
static int persistent = 0;
77
static enum { RUNNING, TERMINATE, TERMINATED } state;
78
static int shared = 1;
79
static int nb_fds;
80
static QIONetListener *server;
81
static QCryptoTLSCreds *tlscreds;
82
static const char *tlsauthz;
83

84
static void usage(const char *name)
85
{
86
    (printf) (
87
"Usage: %s [OPTIONS] FILE\n"
88
"  or:  %s -L [OPTIONS]\n"
89
"QEMU Disk Network Block Device Utility\n"
90
"\n"
91
"  -h, --help                display this help and exit\n"
92
"  -V, --version             output version information and exit\n"
93
"\n"
94
"Connection properties:\n"
95
"  -p, --port=PORT           port to listen on (default `%d')\n"
96
"  -b, --bind=IFACE          interface to bind to (default `0.0.0.0')\n"
97
"  -k, --socket=PATH         path to the unix socket\n"
98
"                            (default '"SOCKET_PATH"')\n"
99
"  -e, --shared=NUM          device can be shared by NUM clients (default '1')\n"
100
"  -t, --persistent          don't exit on the last connection\n"
101
"  -v, --verbose             display extra debugging information\n"
102
"  -x, --export-name=NAME    expose export by name (default is empty string)\n"
103
"  -D, --description=TEXT    export a human-readable description\n"
104
"\n"
105
"Exposing part of the image:\n"
106
"  -o, --offset=OFFSET       offset into the image\n"
107
"  -A, --allocation-depth    expose the allocation depth\n"
108
"  -B, --bitmap=NAME         expose a persistent dirty bitmap\n"
109
"\n"
110
"General purpose options:\n"
111
"  -L, --list                list exports available from another NBD server\n"
112
"  --object type,id=ID,...   define an object such as 'secret' for providing\n"
113
"                            passwords and/or encryption keys\n"
114
"  --tls-creds=ID            use id of an earlier --object to provide TLS\n"
115
"  --tls-authz=ID            use id of an earlier --object to provide\n"
116
"                            authorization\n"
117
"  --tls-hostname=HOSTNAME   override hostname used to check x509 certificate\n"
118
"  -T, --trace [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
119
"                            specify tracing options\n"
120
"  --fork                    fork off the server process and exit the parent\n"
121
"                            once the server is running\n"
122
"  --pid-file=PATH           store the server's process ID in the given file\n"
123
#ifdef CONFIG_SELINUX
124
"  --selinux-label=LABEL     set SELinux process label on listening socket\n"
125
#endif
126
#if HAVE_NBD_DEVICE
127
"\n"
128
"Kernel NBD client support:\n"
129
"  -c, --connect=DEV         connect FILE to the local NBD device DEV\n"
130
"  -d, --disconnect          disconnect the specified device\n"
131
#endif
132
"\n"
133
"Block device options:\n"
134
"  -f, --format=FORMAT       set image format (raw, qcow2, ...)\n"
135
"  -r, --read-only           export read-only\n"
136
"  -s, --snapshot            use FILE as an external snapshot, create a temporary\n"
137
"                            file with backing_file=FILE, redirect the write to\n"
138
"                            the temporary one\n"
139
"  -l, --load-snapshot=SNAPSHOT_PARAM\n"
140
"                            load an internal snapshot inside FILE and export it\n"
141
"                            as an read-only device, SNAPSHOT_PARAM format is\n"
142
"                            'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
143
"                            '[ID_OR_NAME]'\n"
144
"  -n, --nocache             disable host cache\n"
145
"      --cache=MODE          set cache mode used to access the disk image, the\n"
146
"                            valid options are: 'none', 'writeback' (default),\n"
147
"                            'writethrough', 'directsync' and 'unsafe'\n"
148
"      --aio=MODE            set AIO mode (native, io_uring or threads)\n"
149
"      --discard=MODE        set discard mode (ignore, unmap)\n"
150
"      --detect-zeroes=MODE  set detect-zeroes mode (off, on, unmap)\n"
151
"      --image-opts          treat FILE as a full set of image options\n"
152
"\n"
153
QEMU_HELP_BOTTOM "\n"
154
    , name, name, NBD_DEFAULT_PORT, "DEVICE");
155
}
156

157
static void version(const char *name)
158
{
159
    printf(
160
"%s " QEMU_FULL_VERSION "\n"
161
"Written by Anthony Liguori.\n"
162
"\n"
163
QEMU_COPYRIGHT "\n"
164
"This is free software; see the source for copying conditions.  There is NO\n"
165
"warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
166
    , name);
167
}
168

169
#ifdef CONFIG_POSIX
170
/*
171
 * The client thread uses SIGTERM to interrupt the server.  A signal
172
 * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
173
 */
174
void qemu_system_killed(int signum, pid_t pid)
175
{
176
    qatomic_cmpxchg(&state, RUNNING, TERMINATE);
177
    qemu_notify_event();
178
}
179
#endif /* CONFIG_POSIX */
180

181
static int qemu_nbd_client_list(SocketAddress *saddr, QCryptoTLSCreds *tls,
182
                                const char *hostname)
183
{
184
    int ret = EXIT_FAILURE;
185
    int rc;
186
    Error *err = NULL;
187
    QIOChannelSocket *sioc;
188
    NBDExportInfo *list;
189
    int i, j;
190

191
    sioc = qio_channel_socket_new();
192
    if (qio_channel_socket_connect_sync(sioc, saddr, &err) < 0) {
193
        error_report_err(err);
194
        goto out;
195
    }
196
    rc = nbd_receive_export_list(QIO_CHANNEL(sioc), tls, hostname, &list,
197
                                 &err);
198
    if (rc < 0) {
199
        if (err) {
200
            error_report_err(err);
201
        }
202
        goto out;
203
    }
204
    printf("exports available: %d\n", rc);
205
    for (i = 0; i < rc; i++) {
206
        printf(" export: '%s'\n", list[i].name);
207
        if (list[i].description && *list[i].description) {
208
            printf("  description: %s\n", list[i].description);
209
        }
210
        if (list[i].flags & NBD_FLAG_HAS_FLAGS) {
211
            static const char *const flag_names[] = {
212
                [NBD_FLAG_READ_ONLY_BIT]            = "readonly",
213
                [NBD_FLAG_SEND_FLUSH_BIT]           = "flush",
214
                [NBD_FLAG_SEND_FUA_BIT]             = "fua",
215
                [NBD_FLAG_ROTATIONAL_BIT]           = "rotational",
216
                [NBD_FLAG_SEND_TRIM_BIT]            = "trim",
217
                [NBD_FLAG_SEND_WRITE_ZEROES_BIT]    = "zeroes",
218
                [NBD_FLAG_SEND_DF_BIT]              = "df",
219
                [NBD_FLAG_CAN_MULTI_CONN_BIT]       = "multi",
220
                [NBD_FLAG_SEND_RESIZE_BIT]          = "resize",
221
                [NBD_FLAG_SEND_CACHE_BIT]           = "cache",
222
                [NBD_FLAG_SEND_FAST_ZERO_BIT]       = "fast-zero",
223
                [NBD_FLAG_BLOCK_STAT_PAYLOAD_BIT]   = "block-status-payload",
224
            };
225

226
            printf("  size:  %" PRIu64 "\n", list[i].size);
227
            printf("  flags: 0x%x (", list[i].flags);
228
            for (size_t bit = 0; bit < ARRAY_SIZE(flag_names); bit++) {
229
                if (flag_names[bit] && (list[i].flags & (1 << bit))) {
230
                    printf(" %s", flag_names[bit]);
231
                }
232
            }
233
            printf(" )\n");
234
        }
235
        if (list[i].min_block) {
236
            printf("  min block: %u\n", list[i].min_block);
237
            printf("  opt block: %u\n", list[i].opt_block);
238
            printf("  max block: %u\n", list[i].max_block);
239
        }
240
        printf("  transaction size: %s\n",
241
               list[i].mode >= NBD_MODE_EXTENDED ?
242
               "64-bit" : "32-bit");
243
        if (list[i].n_contexts) {
244
            printf("  available meta contexts: %d\n", list[i].n_contexts);
245
            for (j = 0; j < list[i].n_contexts; j++) {
246
                printf("   %s\n", list[i].contexts[j]);
247
            }
248
        }
249
    }
250
    nbd_free_export_list(list, rc);
251

252
    ret = EXIT_SUCCESS;
253
 out:
254
    object_unref(OBJECT(sioc));
255
    return ret;
256
}
257

258

259
struct NbdClientOpts {
260
    char *device;
261
    char *srcpath;
262
    SocketAddress *saddr;
263
    int old_stderr;
264
    bool fork_process;
265
    bool verbose;
266
};
267

268
static void nbd_client_release_pipe(int old_stderr)
269
{
270
    /* Close stderr so that the qemu-nbd process exits.  */
271
    if (dup2(old_stderr, STDERR_FILENO) < 0) {
272
        error_report("Could not release pipe to parent: %s",
273
                     strerror(errno));
274
        exit(EXIT_FAILURE);
275
    }
276
    if (old_stderr != STDOUT_FILENO && close(old_stderr) < 0) {
277
        error_report("Could not release qemu-nbd: %s", strerror(errno));
278
        exit(EXIT_FAILURE);
279
    }
280
}
281

282
#if HAVE_NBD_DEVICE
283
static void *show_parts(void *arg)
284
{
285
    char *device = arg;
286
    int nbd;
287

288
    /* linux just needs an open() to trigger
289
     * the partition table update
290
     * but remember to load the module with max_part != 0 :
291
     *     modprobe nbd max_part=63
292
     */
293
    nbd = open(device, O_RDWR);
294
    if (nbd >= 0) {
295
        close(nbd);
296
    }
297
    return NULL;
298
}
299

300
static void *nbd_client_thread(void *arg)
301
{
302
    struct NbdClientOpts *opts = arg;
303
    /* TODO: Revisit this if nbd.ko ever gains support for structured reply */
304
    NBDExportInfo info = { .request_sizes = false, .name = g_strdup(""),
305
                           .mode = NBD_MODE_SIMPLE };
306
    QIOChannelSocket *sioc;
307
    int fd = -1;
308
    int ret = EXIT_FAILURE;
309
    pthread_t show_parts_thread;
310
    Error *local_error = NULL;
311

312
    sioc = qio_channel_socket_new();
313
    if (qio_channel_socket_connect_sync(sioc,
314
                                        opts->saddr,
315
                                        &local_error) < 0) {
316
        error_report_err(local_error);
317
        goto out;
318
    }
319

320
    if (nbd_receive_negotiate(QIO_CHANNEL(sioc), NULL, NULL, NULL,
321
                              &info, &local_error) < 0) {
322
        if (local_error) {
323
            error_report_err(local_error);
324
        }
325
        goto out;
326
    }
327

328
    fd = open(opts->device, O_RDWR);
329
    if (fd < 0) {
330
        /* Linux-only, we can use %m in printf.  */
331
        error_report("Failed to open %s: %m", opts->device);
332
        goto out;
333
    }
334

335
    if (nbd_init(fd, sioc, &info, &local_error) < 0) {
336
        error_report_err(local_error);
337
        goto out;
338
    }
339

340
    /* update partition table */
341
    pthread_create(&show_parts_thread, NULL, show_parts, opts->device);
342

343
    if (opts->verbose && !opts->fork_process) {
344
        fprintf(stderr, "NBD device %s is now connected to %s\n",
345
                opts->device, opts->srcpath);
346
    } else {
347
        nbd_client_release_pipe(opts->old_stderr);
348
    }
349

350
    if (nbd_client(fd) < 0) {
351
        goto out;
352
    }
353

354
    ret = EXIT_SUCCESS;
355

356
 out:
357
    if (fd >= 0) {
358
        close(fd);
359
    }
360
    object_unref(OBJECT(sioc));
361
    g_free(info.name);
362
    kill(getpid(), SIGTERM);
363
    return (void *) (intptr_t) ret;
364
}
365
#endif /* HAVE_NBD_DEVICE */
366

367
static int nbd_can_accept(void)
368
{
369
    return state == RUNNING && (shared == 0 || nb_fds < shared);
370
}
371

372
static void nbd_update_server_watch(void);
373

374
static void nbd_client_closed(NBDClient *client, bool negotiated)
375
{
376
    nb_fds--;
377
    if (negotiated && nb_fds == 0 && !persistent && state == RUNNING) {
378
        state = TERMINATE;
379
    }
380
    nbd_update_server_watch();
381
    nbd_client_put(client);
382
}
383

384
static void nbd_accept(QIONetListener *listener, QIOChannelSocket *cioc,
385
                       gpointer opaque)
386
{
387
    if (state >= TERMINATE) {
388
        return;
389
    }
390

391
    nb_fds++;
392
    nbd_update_server_watch();
393
    /* TODO - expose handshake timeout as command line option */
394
    nbd_client_new(cioc, NBD_DEFAULT_HANDSHAKE_MAX_SECS,
395
                   tlscreds, tlsauthz, nbd_client_closed, NULL);
396
}
397

398
static void nbd_update_server_watch(void)
399
{
400
    if (nbd_can_accept()) {
401
        qio_net_listener_set_client_func(server, nbd_accept, NULL, NULL);
402
    } else {
403
        qio_net_listener_set_client_func(server, NULL, NULL, NULL);
404
    }
405
}
406

407

408
static SocketAddress *nbd_build_socket_address(const char *sockpath,
409
                                               const char *bindto,
410
                                               const char *port)
411
{
412
    SocketAddress *saddr;
413

414
    saddr = g_new0(SocketAddress, 1);
415
    if (sockpath) {
416
        saddr->type = SOCKET_ADDRESS_TYPE_UNIX;
417
        saddr->u.q_unix.path = g_strdup(sockpath);
418
    } else {
419
        InetSocketAddress *inet;
420
        saddr->type = SOCKET_ADDRESS_TYPE_INET;
421
        inet = &saddr->u.inet;
422
        inet->host = g_strdup(bindto);
423
        if (port) {
424
            inet->port = g_strdup(port);
425
        } else  {
426
            inet->port = g_strdup_printf("%d", NBD_DEFAULT_PORT);
427
        }
428
    }
429

430
    return saddr;
431
}
432

433

434
static QemuOptsList file_opts = {
435
    .name = "file",
436
    .implied_opt_name = "file",
437
    .head = QTAILQ_HEAD_INITIALIZER(file_opts.head),
438
    .desc = {
439
        /* no elements => accept any params */
440
        { /* end of list */ }
441
    },
442
};
443

444
static QCryptoTLSCreds *nbd_get_tls_creds(const char *id, bool list,
445
                                          Error **errp)
446
{
447
    Object *obj;
448
    QCryptoTLSCreds *creds;
449

450
    obj = object_resolve_path_component(
451
        object_get_objects_root(), id);
452
    if (!obj) {
453
        error_setg(errp, "No TLS credentials with id '%s'",
454
                   id);
455
        return NULL;
456
    }
457
    creds = (QCryptoTLSCreds *)
458
        object_dynamic_cast(obj, TYPE_QCRYPTO_TLS_CREDS);
459
    if (!creds) {
460
        error_setg(errp, "Object with id '%s' is not TLS credentials",
461
                   id);
462
        return NULL;
463
    }
464

465
    if (!qcrypto_tls_creds_check_endpoint(creds,
466
                                          list
467
                                          ? QCRYPTO_TLS_CREDS_ENDPOINT_CLIENT
468
                                          : QCRYPTO_TLS_CREDS_ENDPOINT_SERVER,
469
                                          errp)) {
470
        return NULL;
471
    }
472
    object_ref(obj);
473
    return creds;
474
}
475

476
static void setup_address_and_port(const char **address, const char **port)
477
{
478
    if (*address == NULL) {
479
        *address = "0.0.0.0";
480
    }
481

482
    if (*port == NULL) {
483
        *port = stringify(NBD_DEFAULT_PORT);
484
    }
485
}
486

487
/*
488
 * Check socket parameters compatibility when socket activation is used.
489
 */
490
static const char *socket_activation_validate_opts(const char *device,
491
                                                   const char *sockpath,
492
                                                   const char *address,
493
                                                   const char *port,
494
                                                   const char *selinux,
495
                                                   bool list)
496
{
497
    if (device != NULL) {
498
        return "NBD device can't be set when using socket activation";
499
    }
500

501
    if (sockpath != NULL) {
502
        return "Unix socket can't be set when using socket activation";
503
    }
504

505
    if (address != NULL) {
506
        return "The interface can't be set when using socket activation";
507
    }
508

509
    if (port != NULL) {
510
        return "TCP port number can't be set when using socket activation";
511
    }
512

513
    if (selinux != NULL) {
514
        return "SELinux label can't be set when using socket activation";
515
    }
516

517
    if (list) {
518
        return "List mode is incompatible with socket activation";
519
    }
520

521
    return NULL;
522
}
523

524
static void qemu_nbd_shutdown(void)
525
{
526
    job_cancel_sync_all();
527
    blk_exp_close_all();
528
    bdrv_close_all();
529
}
530

531
int main(int argc, char **argv)
532
{
533
    BlockBackend *blk;
534
    BlockDriverState *bs;
535
    uint64_t dev_offset = 0;
536
    bool readonly = false;
537
    bool disconnect = false;
538
    const char *bindto = NULL;
539
    const char *port = NULL;
540
    char *sockpath = NULL;
541
    QemuOpts *sn_opts = NULL;
542
    const char *sn_id_or_name = NULL;
543
    const char *sopt = "hVb:o:p:rsnc:dvk:e:f:tl:x:T:D:AB:L";
544
    struct option lopt[] = {
545
        { "help", no_argument, NULL, 'h' },
546
        { "version", no_argument, NULL, 'V' },
547
        { "bind", required_argument, NULL, 'b' },
548
        { "port", required_argument, NULL, 'p' },
549
        { "socket", required_argument, NULL, 'k' },
550
        { "offset", required_argument, NULL, 'o' },
551
        { "read-only", no_argument, NULL, 'r' },
552
        { "allocation-depth", no_argument, NULL, 'A' },
553
        { "bitmap", required_argument, NULL, 'B' },
554
        { "connect", required_argument, NULL, 'c' },
555
        { "disconnect", no_argument, NULL, 'd' },
556
        { "list", no_argument, NULL, 'L' },
557
        { "snapshot", no_argument, NULL, 's' },
558
        { "load-snapshot", required_argument, NULL, 'l' },
559
        { "nocache", no_argument, NULL, 'n' },
560
        { "cache", required_argument, NULL, QEMU_NBD_OPT_CACHE },
561
        { "aio", required_argument, NULL, QEMU_NBD_OPT_AIO },
562
        { "discard", required_argument, NULL, QEMU_NBD_OPT_DISCARD },
563
        { "detect-zeroes", required_argument, NULL,
564
          QEMU_NBD_OPT_DETECT_ZEROES },
565
        { "shared", required_argument, NULL, 'e' },
566
        { "format", required_argument, NULL, 'f' },
567
        { "persistent", no_argument, NULL, 't' },
568
        { "verbose", no_argument, NULL, 'v' },
569
        { "object", required_argument, NULL, QEMU_NBD_OPT_OBJECT },
570
        { "export-name", required_argument, NULL, 'x' },
571
        { "description", required_argument, NULL, 'D' },
572
        { "tls-creds", required_argument, NULL, QEMU_NBD_OPT_TLSCREDS },
573
        { "tls-hostname", required_argument, NULL, QEMU_NBD_OPT_TLSHOSTNAME },
574
        { "tls-authz", required_argument, NULL, QEMU_NBD_OPT_TLSAUTHZ },
575
        { "image-opts", no_argument, NULL, QEMU_NBD_OPT_IMAGE_OPTS },
576
        { "trace", required_argument, NULL, 'T' },
577
        { "fork", no_argument, NULL, QEMU_NBD_OPT_FORK },
578
        { "pid-file", required_argument, NULL, QEMU_NBD_OPT_PID_FILE },
579
        { "selinux-label", required_argument, NULL,
580
          QEMU_NBD_OPT_SELINUX_LABEL },
581
        { NULL, 0, NULL, 0 }
582
    };
583
    int ch;
584
    int opt_ind = 0;
585
    int flags = BDRV_O_RDWR;
586
    int ret = 0;
587
    bool seen_cache = false;
588
    bool seen_discard = false;
589
    bool seen_aio = false;
590
    pthread_t client_thread;
591
    const char *fmt = NULL;
592
    Error *local_err = NULL;
593
    BlockdevDetectZeroesOptions detect_zeroes =
594
        BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
595
    QDict *options = NULL;
596
    const char *export_name = NULL; /* defaults to "" later for server mode */
597
    const char *export_description = NULL;
598
    BlockDirtyBitmapOrStrList *bitmaps = NULL;
599
    bool alloc_depth = false;
600
    const char *tlscredsid = NULL;
601
    const char *tlshostname = NULL;
602
    bool imageOpts = false;
603
    bool writethrough = false; /* Client will flush as needed. */
604
    bool list = false;
605
    unsigned socket_activation;
606
    const char *pid_file_name = NULL;
607
    const char *selinux_label = NULL;
608
    BlockExportOptions *export_opts;
609
    struct NbdClientOpts opts = {
610
        .fork_process = false,
611
        .verbose = false,
612
        .device = NULL,
613
        .srcpath = NULL,
614
        .saddr = NULL,
615
        .old_stderr = STDOUT_FILENO,
616
    };
617

618
#ifdef CONFIG_POSIX
619
    os_setup_early_signal_handling();
620
    os_setup_signal_handling();
621
#endif
622

623
    socket_init();
624
    error_init(argv[0]);
625
    module_call_init(MODULE_INIT_TRACE);
626
    qcrypto_init(&error_fatal);
627

628
    module_call_init(MODULE_INIT_QOM);
629
    qemu_add_opts(&qemu_trace_opts);
630
    qemu_init_exec_dir(argv[0]);
631

632
    while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
633
        switch (ch) {
634
        case 's':
635
            flags |= BDRV_O_SNAPSHOT;
636
            break;
637
        case 'n':
638
            optarg = (char *) "none";
639
            /* fallthrough */
640
        case QEMU_NBD_OPT_CACHE:
641
            if (seen_cache) {
642
                error_report("-n and --cache can only be specified once");
643
                exit(EXIT_FAILURE);
644
            }
645
            seen_cache = true;
646
            if (bdrv_parse_cache_mode(optarg, &flags, &writethrough) == -1) {
647
                error_report("Invalid cache mode `%s'", optarg);
648
                exit(EXIT_FAILURE);
649
            }
650
            break;
651
        case QEMU_NBD_OPT_AIO:
652
            if (seen_aio) {
653
                error_report("--aio can only be specified once");
654
                exit(EXIT_FAILURE);
655
            }
656
            seen_aio = true;
657
            if (bdrv_parse_aio(optarg, &flags) < 0) {
658
                error_report("Invalid aio mode '%s'", optarg);
659
                exit(EXIT_FAILURE);
660
            }
661
            break;
662
        case QEMU_NBD_OPT_DISCARD:
663
            if (seen_discard) {
664
                error_report("--discard can only be specified once");
665
                exit(EXIT_FAILURE);
666
            }
667
            seen_discard = true;
668
            if (bdrv_parse_discard_flags(optarg, &flags) == -1) {
669
                error_report("Invalid discard mode `%s'", optarg);
670
                exit(EXIT_FAILURE);
671
            }
672
            break;
673
        case QEMU_NBD_OPT_DETECT_ZEROES:
674
            detect_zeroes =
675
                qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup,
676
                                optarg,
677
                                BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
678
                                &local_err);
679
            if (local_err) {
680
                error_reportf_err(local_err,
681
                                  "Failed to parse detect_zeroes mode: ");
682
                exit(EXIT_FAILURE);
683
            }
684
            if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
685
                !(flags & BDRV_O_UNMAP)) {
686
                error_report("setting detect-zeroes to unmap is not allowed "
687
                             "without setting discard operation to unmap");
688
                exit(EXIT_FAILURE);
689
            }
690
            break;
691
        case 'b':
692
            bindto = optarg;
693
            break;
694
        case 'p':
695
            port = optarg;
696
            break;
697
        case 'o':
698
            if (qemu_strtou64(optarg, NULL, 0, &dev_offset) < 0) {
699
                error_report("Invalid offset '%s'", optarg);
700
                exit(EXIT_FAILURE);
701
            }
702
            break;
703
        case 'l':
704
            if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
705
                sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
706
                                                  optarg, false);
707
                if (!sn_opts) {
708
                    error_report("Failed in parsing snapshot param `%s'",
709
                                 optarg);
710
                    exit(EXIT_FAILURE);
711
                }
712
            } else {
713
                sn_id_or_name = optarg;
714
            }
715
            /* fall through */
716
        case 'r':
717
            readonly = true;
718
            flags &= ~BDRV_O_RDWR;
719
            break;
720
        case 'A':
721
            alloc_depth = true;
722
            break;
723
        case 'B':
724
            {
725
                BlockDirtyBitmapOrStr *el = g_new(BlockDirtyBitmapOrStr, 1);
726
                *el = (BlockDirtyBitmapOrStr) {
727
                    .type = QTYPE_QSTRING,
728
                    .u.local = g_strdup(optarg),
729
                };
730
                QAPI_LIST_PREPEND(bitmaps, el);
731
            }
732
            break;
733
        case 'k':
734
            sockpath = optarg;
735
            if (sockpath[0] != '/') {
736
                error_report("socket path must be absolute");
737
                exit(EXIT_FAILURE);
738
            }
739
            break;
740
        case 'd':
741
            disconnect = true;
742
            break;
743
        case 'c':
744
            opts.device = optarg;
745
            break;
746
        case 'e':
747
            if (qemu_strtoi(optarg, NULL, 0, &shared) < 0 ||
748
                shared < 0) {
749
                error_report("Invalid shared device number '%s'", optarg);
750
                exit(EXIT_FAILURE);
751
            }
752
            break;
753
        case 'f':
754
            fmt = optarg;
755
            break;
756
        case 't':
757
            persistent = 1;
758
            break;
759
        case 'x':
760
            export_name = optarg;
761
            if (strlen(export_name) > NBD_MAX_STRING_SIZE) {
762
                error_report("export name '%s' too long", export_name);
763
                exit(EXIT_FAILURE);
764
            }
765
            break;
766
        case 'D':
767
            export_description = optarg;
768
            if (strlen(export_description) > NBD_MAX_STRING_SIZE) {
769
                error_report("export description '%s' too long",
770
                             export_description);
771
                exit(EXIT_FAILURE);
772
            }
773
            break;
774
        case 'v':
775
            opts.verbose = true;
776
            break;
777
        case 'V':
778
            version(argv[0]);
779
            exit(0);
780
            break;
781
        case 'h':
782
            usage(argv[0]);
783
            exit(0);
784
            break;
785
        case '?':
786
            error_report("Try `%s --help' for more information.", argv[0]);
787
            exit(EXIT_FAILURE);
788
        case QEMU_NBD_OPT_OBJECT:
789
            user_creatable_process_cmdline(optarg);
790
            break;
791
        case QEMU_NBD_OPT_TLSCREDS:
792
            tlscredsid = optarg;
793
            break;
794
        case QEMU_NBD_OPT_TLSHOSTNAME:
795
            tlshostname = optarg;
796
            break;
797
        case QEMU_NBD_OPT_IMAGE_OPTS:
798
            imageOpts = true;
799
            break;
800
        case 'T':
801
            trace_opt_parse(optarg);
802
            break;
803
        case QEMU_NBD_OPT_TLSAUTHZ:
804
            tlsauthz = optarg;
805
            break;
806
        case QEMU_NBD_OPT_FORK:
807
            opts.fork_process = true;
808
            break;
809
        case 'L':
810
            list = true;
811
            break;
812
        case QEMU_NBD_OPT_PID_FILE:
813
            pid_file_name = optarg;
814
            break;
815
        case QEMU_NBD_OPT_SELINUX_LABEL:
816
            selinux_label = optarg;
817
            break;
818
        }
819
    }
820

821
    if (list) {
822
        if (argc != optind) {
823
            error_report("List mode is incompatible with a file name");
824
            exit(EXIT_FAILURE);
825
        }
826
        if (export_name || export_description || dev_offset ||
827
            opts.device || disconnect || fmt || sn_id_or_name || bitmaps ||
828
            alloc_depth || seen_aio || seen_discard || seen_cache) {
829
            error_report("List mode is incompatible with per-device settings");
830
            exit(EXIT_FAILURE);
831
        }
832
        if (opts.fork_process) {
833
            error_report("List mode is incompatible with forking");
834
            exit(EXIT_FAILURE);
835
        }
836
    } else if ((argc - optind) != 1) {
837
        error_report("Invalid number of arguments");
838
        error_printf("Try `%s --help' for more information.\n", argv[0]);
839
        exit(EXIT_FAILURE);
840
    } else if (!export_name) {
841
        export_name = "";
842
    }
843

844
    if (!trace_init_backends()) {
845
        exit(1);
846
    }
847
    trace_init_file();
848
    qemu_set_log(LOG_TRACE, &error_fatal);
849

850
    socket_activation = check_socket_activation();
851
    if (socket_activation == 0) {
852
        if (!sockpath) {
853
            setup_address_and_port(&bindto, &port);
854
        }
855
    } else {
856
        /* Using socket activation - check user didn't use -p etc. */
857
        const char *err_msg = socket_activation_validate_opts(opts.device,
858
                                                              sockpath,
859
                                                              bindto, port,
860
                                                              selinux_label,
861
                                                              list);
862
        if (err_msg != NULL) {
863
            error_report("%s", err_msg);
864
            exit(EXIT_FAILURE);
865
        }
866

867
        /* qemu-nbd can only listen on a single socket.  */
868
        if (socket_activation > 1) {
869
            error_report("qemu-nbd does not support socket activation with %s > 1",
870
                         "LISTEN_FDS");
871
            exit(EXIT_FAILURE);
872
        }
873
    }
874

875
    if (tlscredsid) {
876
        if (opts.device) {
877
            error_report("TLS is not supported with a host device");
878
            exit(EXIT_FAILURE);
879
        }
880
        if (tlsauthz && list) {
881
            error_report("TLS authorization is incompatible with export list");
882
            exit(EXIT_FAILURE);
883
        }
884
        if (tlshostname && !list) {
885
            error_report("TLS hostname is only supported with export list");
886
            exit(EXIT_FAILURE);
887
        }
888
        tlscreds = nbd_get_tls_creds(tlscredsid, list, &local_err);
889
        if (local_err) {
890
            error_reportf_err(local_err, "Failed to get TLS creds: ");
891
            exit(EXIT_FAILURE);
892
        }
893
    } else {
894
        if (tlsauthz) {
895
            error_report("--tls-authz is not permitted without --tls-creds");
896
            exit(EXIT_FAILURE);
897
        }
898
        if (tlshostname) {
899
            error_report("--tls-hostname is not permitted without --tls-creds");
900
            exit(EXIT_FAILURE);
901
        }
902
    }
903

904
    if (selinux_label) {
905
#ifdef CONFIG_SELINUX
906
        if (sockpath == NULL && opts.device == NULL) {
907
            error_report("--selinux-label is not permitted without --socket");
908
            exit(EXIT_FAILURE);
909
        }
910
#else
911
        error_report("SELinux support not enabled in this binary");
912
        exit(EXIT_FAILURE);
913
#endif
914
    }
915

916
    if (list) {
917
        opts.saddr = nbd_build_socket_address(sockpath, bindto, port);
918
        return qemu_nbd_client_list(opts.saddr, tlscreds,
919
                                    tlshostname ? tlshostname : bindto);
920
    }
921

922
#if !HAVE_NBD_DEVICE
923
    if (disconnect || opts.device) {
924
        error_report("Kernel /dev/nbdN support not available");
925
        exit(EXIT_FAILURE);
926
    }
927
#else /* HAVE_NBD_DEVICE */
928
    if (disconnect) {
929
        int nbdfd = open(argv[optind], O_RDWR);
930
        if (nbdfd < 0) {
931
            error_report("Cannot open %s: %s", argv[optind],
932
                         strerror(errno));
933
            exit(EXIT_FAILURE);
934
        }
935
        nbd_disconnect(nbdfd);
936

937
        close(nbdfd);
938

939
        printf("%s disconnected\n", argv[optind]);
940

941
        return 0;
942
    }
943
#endif
944

945
    if ((opts.device && !opts.verbose) || opts.fork_process) {
946
#ifndef WIN32
947
        g_autoptr(GError) err = NULL;
948
        int stderr_fd[2];
949
        pid_t pid;
950

951
        if (!g_unix_open_pipe(stderr_fd, FD_CLOEXEC, &err)) {
952
            error_report("Error setting up communication pipe: %s",
953
                         err->message);
954
            exit(EXIT_FAILURE);
955
        }
956

957
        /* Now daemonize, but keep a communication channel open to
958
         * print errors and exit with the proper status code.
959
         */
960
        pid = fork();
961
        if (pid < 0) {
962
            error_report("Failed to fork: %s", strerror(errno));
963
            exit(EXIT_FAILURE);
964
        } else if (pid == 0) {
965
            int saved_errno;
966

967
            close(stderr_fd[0]);
968

969
            /* Remember parent's stderr if we will be restoring it. */
970
            if (opts.verbose /* fork_process is set */) {
971
                opts.old_stderr = dup(STDERR_FILENO);
972
                if (opts.old_stderr < 0) {
973
                    error_report("Could not dup original stderr: %s",
974
                                 strerror(errno));
975
                    exit(EXIT_FAILURE);
976
                }
977
            }
978

979
            ret = qemu_daemon(1, 0);
980
            saved_errno = errno;    /* dup2 will overwrite error below */
981

982
            /* Temporarily redirect stderr to the parent's pipe...  */
983
            if (dup2(stderr_fd[1], STDERR_FILENO) < 0) {
984
                char str[256];
985
                snprintf(str, sizeof(str),
986
                         "%s: Failed to link stderr to the pipe: %s\n",
987
                         g_get_prgname(), strerror(errno));
988
                /*
989
                 * We are unable to use error_report() here as we need to get
990
                 * stderr pointed to the parent's pipe. Write to that pipe
991
                 * manually.
992
                 */
993
                ret = write(stderr_fd[1], str, strlen(str));
994
                exit(EXIT_FAILURE);
995
            }
996

997
            if (ret < 0) {
998
                error_report("Failed to daemonize: %s", strerror(saved_errno));
999
                exit(EXIT_FAILURE);
1000
            }
1001

1002
            /* ... close the descriptor we inherited and go on.  */
1003
            close(stderr_fd[1]);
1004
        } else {
1005
            bool errors = false;
1006
            char *buf;
1007

1008
            /* In the parent.  Print error messages from the child until
1009
             * it closes the pipe.
1010
             */
1011
            close(stderr_fd[1]);
1012
            buf = g_malloc(1024);
1013
            while ((ret = read(stderr_fd[0], buf, 1024)) > 0) {
1014
                errors = true;
1015
                ret = qemu_write_full(STDERR_FILENO, buf, ret);
1016
                if (ret < 0) {
1017
                    exit(EXIT_FAILURE);
1018
                }
1019
            }
1020
            if (ret < 0) {
1021
                error_report("Cannot read from daemon: %s",
1022
                             strerror(errno));
1023
                exit(EXIT_FAILURE);
1024
            }
1025

1026
            /* Usually the daemon should not print any message.
1027
             * Exit with zero status in that case.
1028
             */
1029
            exit(errors);
1030
        }
1031
#else /* WIN32 */
1032
        error_report("Unable to fork into background on Windows hosts");
1033
        exit(EXIT_FAILURE);
1034
#endif /* WIN32 */
1035
    }
1036

1037
    if (opts.device != NULL && sockpath == NULL) {
1038
        sockpath = g_malloc(128);
1039
        snprintf(sockpath, 128, SOCKET_PATH, basename(opts.device));
1040
    }
1041

1042
    server = qio_net_listener_new();
1043
    if (socket_activation == 0) {
1044
        int backlog;
1045

1046
        if (persistent || shared == 0) {
1047
            backlog = SOMAXCONN;
1048
        } else {
1049
            backlog = MIN(shared, SOMAXCONN);
1050
        }
1051
#ifdef CONFIG_SELINUX
1052
        if (selinux_label && setsockcreatecon_raw(selinux_label) == -1) {
1053
            error_report("Cannot set SELinux socket create context to %s: %s",
1054
                         selinux_label, strerror(errno));
1055
            exit(EXIT_FAILURE);
1056
        }
1057
#endif
1058
        opts.saddr = nbd_build_socket_address(sockpath, bindto, port);
1059
        if (qio_net_listener_open_sync(server, opts.saddr, backlog,
1060
                                       &local_err) < 0) {
1061
            object_unref(OBJECT(server));
1062
            error_report_err(local_err);
1063
            exit(EXIT_FAILURE);
1064
        }
1065
#ifdef CONFIG_SELINUX
1066
        if (selinux_label && setsockcreatecon_raw(NULL) == -1) {
1067
            error_report("Cannot clear SELinux socket create context: %s",
1068
                         strerror(errno));
1069
            exit(EXIT_FAILURE);
1070
        }
1071
#endif
1072
    } else {
1073
        size_t i;
1074
        /* See comment in check_socket_activation above. */
1075
        for (i = 0; i < socket_activation; i++) {
1076
            QIOChannelSocket *sioc;
1077
            sioc = qio_channel_socket_new_fd(FIRST_SOCKET_ACTIVATION_FD + i,
1078
                                             &local_err);
1079
            if (sioc == NULL) {
1080
                object_unref(OBJECT(server));
1081
                error_reportf_err(local_err,
1082
                                  "Failed to use socket activation: ");
1083
                exit(EXIT_FAILURE);
1084
            }
1085
            qio_net_listener_add(server, sioc);
1086
            object_unref(OBJECT(sioc));
1087
        }
1088
    }
1089

1090
    qemu_init_main_loop(&error_fatal);
1091
    bdrv_init();
1092
    atexit(qemu_nbd_shutdown);
1093

1094
    opts.srcpath = argv[optind];
1095
    if (imageOpts) {
1096
        QemuOpts *o;
1097
        if (fmt) {
1098
            error_report("--image-opts and -f are mutually exclusive");
1099
            exit(EXIT_FAILURE);
1100
        }
1101
        o = qemu_opts_parse_noisily(&file_opts, opts.srcpath, true);
1102
        if (!o) {
1103
            qemu_opts_reset(&file_opts);
1104
            exit(EXIT_FAILURE);
1105
        }
1106
        options = qemu_opts_to_qdict(o, NULL);
1107
        qemu_opts_reset(&file_opts);
1108
        blk = blk_new_open(NULL, NULL, options, flags, &local_err);
1109
    } else {
1110
        if (fmt) {
1111
            options = qdict_new();
1112
            qdict_put_str(options, "driver", fmt);
1113
        }
1114
        blk = blk_new_open(opts.srcpath, NULL, options, flags, &local_err);
1115
    }
1116

1117
    if (!blk) {
1118
        error_reportf_err(local_err, "Failed to blk_new_open '%s': ",
1119
                          argv[optind]);
1120
        exit(EXIT_FAILURE);
1121
    }
1122
    bs = blk_bs(blk);
1123

1124
    if (dev_offset) {
1125
        QDict *raw_opts = qdict_new();
1126
        qdict_put_str(raw_opts, "driver", "raw");
1127
        qdict_put_str(raw_opts, "file", bs->node_name);
1128
        qdict_put_int(raw_opts, "offset", dev_offset);
1129

1130
        bs = bdrv_open(NULL, NULL, raw_opts, flags, &error_fatal);
1131

1132
        blk_remove_bs(blk);
1133
        blk_insert_bs(blk, bs, &error_fatal);
1134
        bdrv_unref(bs);
1135
    }
1136

1137
    blk_set_enable_write_cache(blk, !writethrough);
1138

1139
    if (sn_opts) {
1140
        ret = bdrv_snapshot_load_tmp(bs,
1141
                                     qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
1142
                                     qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
1143
                                     &local_err);
1144
    } else if (sn_id_or_name) {
1145
        ret = bdrv_snapshot_load_tmp_by_id_or_name(bs, sn_id_or_name,
1146
                                                   &local_err);
1147
    }
1148
    if (ret < 0) {
1149
        error_reportf_err(local_err, "Failed to load snapshot: ");
1150
        exit(EXIT_FAILURE);
1151
    }
1152

1153
    bs->detect_zeroes = detect_zeroes;
1154

1155
    nbd_server_is_qemu_nbd(shared);
1156

1157
    export_opts = g_new(BlockExportOptions, 1);
1158
    *export_opts = (BlockExportOptions) {
1159
        .type               = BLOCK_EXPORT_TYPE_NBD,
1160
        .id                 = g_strdup("qemu-nbd-export"),
1161
        .node_name          = g_strdup(bdrv_get_node_name(bs)),
1162
        .has_writethrough   = true,
1163
        .writethrough       = writethrough,
1164
        .has_writable       = true,
1165
        .writable           = !readonly,
1166
        .u.nbd = {
1167
            .name                 = g_strdup(export_name),
1168
            .description          = g_strdup(export_description),
1169
            .has_bitmaps          = !!bitmaps,
1170
            .bitmaps              = bitmaps,
1171
            .has_allocation_depth = alloc_depth,
1172
            .allocation_depth     = alloc_depth,
1173
        },
1174
    };
1175
    blk_exp_add(export_opts, &error_fatal);
1176
    qapi_free_BlockExportOptions(export_opts);
1177

1178
    if (opts.device) {
1179
#if HAVE_NBD_DEVICE
1180
        ret = pthread_create(&client_thread, NULL, nbd_client_thread, &opts);
1181
        if (ret != 0) {
1182
            error_report("Failed to create client thread: %s", strerror(ret));
1183
            exit(EXIT_FAILURE);
1184
        }
1185
#endif
1186
    } else {
1187
        /* Shut up GCC warnings.  */
1188
        memset(&client_thread, 0, sizeof(client_thread));
1189
    }
1190

1191
    nbd_update_server_watch();
1192

1193
    if (pid_file_name) {
1194
        qemu_write_pidfile(pid_file_name, &error_fatal);
1195
    }
1196

1197
    /* now when the initialization is (almost) complete, chdir("/")
1198
     * to free any busy filesystems */
1199
    if (chdir("/") < 0) {
1200
        error_report("Could not chdir to root directory: %s",
1201
                     strerror(errno));
1202
        exit(EXIT_FAILURE);
1203
    }
1204

1205
    if (opts.fork_process) {
1206
        nbd_client_release_pipe(opts.old_stderr);
1207
    }
1208

1209
    state = RUNNING;
1210
    do {
1211
        main_loop_wait(false);
1212
        if (state == TERMINATE) {
1213
            blk_exp_close_all();
1214
            state = TERMINATED;
1215
        }
1216
    } while (state != TERMINATED);
1217

1218
    blk_unref(blk);
1219
    if (sockpath) {
1220
        unlink(sockpath);
1221
    }
1222

1223
    qemu_opts_del(sn_opts);
1224

1225
    if (opts.device) {
1226
        void *result;
1227
        pthread_join(client_thread, &result);
1228
        ret = (intptr_t)result;
1229
        exit(ret);
1230
    } else {
1231
        exit(EXIT_SUCCESS);
1232
    }
1233
}
1234