qemu-nbd.c 24.4 KB
Newer Older
1
/*
B
bellard 已提交
2 3 4 5 6 7 8 9 10 11 12 13 14 15
 *  Copyright (C) 2005  Anthony Liguori <anthony@codemonkey.ws>
 *
 *  Network Block Device
 *
 *  This program is free software; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation; under version 2 of the License.
 *
 *  This program is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  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/>.
B
bellard 已提交
17 18
 */

P
Peter Maydell 已提交
19
#include "qemu/osdep.h"
20
#include "qemu-common.h"
M
Markus Armbruster 已提交
21
#include "sysemu/block-backend.h"
22
#include "block/block_int.h"
23
#include "block/nbd.h"
24
#include "qemu/main-loop.h"
25 26
#include "qemu/sockets.h"
#include "qemu/error-report.h"
27
#include "qemu/config-file.h"
28
#include "block/snapshot.h"
29
#include "qapi/util.h"
30
#include "qapi/qmp/qstring.h"
31
#include "qom/object_interfaces.h"
B
bellard 已提交
32 33 34 35 36 37

#include <getopt.h>
#include <sys/socket.h>
#include <netinet/in.h>
#include <netinet/tcp.h>
#include <arpa/inet.h>
38
#include <libgen.h>
P
Paolo Bonzini 已提交
39
#include <pthread.h>
40

41 42 43 44
#define SOCKET_PATH                "/var/lock/qemu-nbd-%s"
#define QEMU_NBD_OPT_CACHE         1
#define QEMU_NBD_OPT_AIO           2
#define QEMU_NBD_OPT_DISCARD       3
45
#define QEMU_NBD_OPT_DETECT_ZEROES 4
46
#define QEMU_NBD_OPT_OBJECT        5
B
bellard 已提交
47

P
Paolo Bonzini 已提交
48
static NBDExport *exp;
49
static int verbose;
P
Paolo Bonzini 已提交
50
static char *srcpath;
51
static SocketAddress *saddr;
52 53
static int persistent = 0;
static enum { RUNNING, TERMINATE, TERMINATING, TERMINATED } state;
P
Paolo Bonzini 已提交
54 55
static int shared = 1;
static int nb_fds;
56
static int server_fd;
B
bellard 已提交
57 58 59

static void usage(const char *name)
{
60
    (printf) (
B
bellard 已提交
61 62 63
"Usage: %s [OPTIONS] FILE\n"
"QEMU Disk Network Block Device Server\n"
"\n"
64 65
"  -h, --help                display this help and exit\n"
"  -V, --version             output version information and exit\n"
66 67
"\n"
"Connection properties:\n"
68 69 70 71 72 73 74
"  -p, --port=PORT           port to listen on (default `%d')\n"
"  -b, --bind=IFACE          interface to bind to (default `0.0.0.0')\n"
"  -k, --socket=PATH         path to the unix socket\n"
"                            (default '"SOCKET_PATH"')\n"
"  -e, --shared=NUM          device can be shared by NUM clients (default '1')\n"
"  -t, --persistent          don't exit on the last connection\n"
"  -v, --verbose             display extra debugging information\n"
B
bellard 已提交
75
"\n"
76
"Exposing part of the image:\n"
77 78
"  -o, --offset=OFFSET       offset into the image\n"
"  -P, --partition=NUM       only expose partition NUM\n"
79
"\n"
80 81 82
"General purpose options:\n"
"  --object type,id=ID,...   define an object such as 'secret' for providing\n"
"                            passwords and/or encryption keys\n"
83 84
#ifdef __linux__
"Kernel NBD client support:\n"
85 86
"  -c, --connect=DEV         connect FILE to the local NBD device DEV\n"
"  -d, --disconnect          disconnect the specified device\n"
87 88 89 90
"\n"
#endif
"\n"
"Block device options:\n"
91 92 93 94 95
"  -f, --format=FORMAT       set image format (raw, qcow2, ...)\n"
"  -r, --read-only           export read-only\n"
"  -s, --snapshot            use FILE as an external snapshot, create a temporary\n"
"                            file with backing_file=FILE, redirect the write to\n"
"                            the temporary one\n"
96
"  -l, --load-snapshot=SNAPSHOT_PARAM\n"
97 98 99 100 101 102 103
"                            load an internal snapshot inside FILE and export it\n"
"                            as an read-only device, SNAPSHOT_PARAM format is\n"
"                            'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
"                            '[ID_OR_NAME]'\n"
"  -n, --nocache             disable host cache\n"
"      --cache=MODE          set cache mode (none, writeback, ...)\n"
"      --aio=MODE            set AIO mode (native or threads)\n"
104
"      --discard=MODE        set discard mode (ignore, unmap)\n"
105
"      --detect-zeroes=MODE  set detect-zeroes mode (off, on, unmap)\n"
106 107
"\n"
"Report bugs to <qemu-devel@nongnu.org>\n"
108
    , name, NBD_DEFAULT_PORT, "DEVICE");
B
bellard 已提交
109 110 111 112 113
}

static void version(const char *name)
{
    printf(
114
"%s version 0.0.1\n"
B
bellard 已提交
115 116 117 118 119
"Written by Anthony Liguori.\n"
"\n"
"Copyright (C) 2006 Anthony Liguori <anthony@codemonkey.ws>.\n"
"This is free software; see the source for copying conditions.  There is NO\n"
"warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
120
    , name);
B
bellard 已提交
121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146
}

struct partition_record
{
    uint8_t bootable;
    uint8_t start_head;
    uint32_t start_cylinder;
    uint8_t start_sector;
    uint8_t system;
    uint8_t end_head;
    uint8_t end_cylinder;
    uint8_t end_sector;
    uint32_t start_sector_abs;
    uint32_t nb_sectors_abs;
};

static void read_partition(uint8_t *p, struct partition_record *r)
{
    r->bootable = p[0];
    r->start_head = p[1];
    r->start_cylinder = p[3] | ((p[2] << 2) & 0x0300);
    r->start_sector = p[2] & 0x3f;
    r->system = p[4];
    r->end_head = p[5];
    r->end_cylinder = p[7] | ((p[6] << 2) & 0x300);
    r->end_sector = p[6] & 0x3f;
147 148 149

    r->start_sector_abs = le32_to_cpup((uint32_t *)(p +  8));
    r->nb_sectors_abs   = le32_to_cpup((uint32_t *)(p + 12));
B
bellard 已提交
150 151
}

152
static int find_partition(BlockBackend *blk, int partition,
B
bellard 已提交
153 154 155 156 157 158
                          off_t *offset, off_t *size)
{
    struct partition_record mbr[4];
    uint8_t data[512];
    int i;
    int ext_partnum = 4;
R
Ryota Ozaki 已提交
159
    int ret;
B
bellard 已提交
160

161
    if ((ret = blk_read(blk, 0, data, 1)) < 0) {
162
        error_report("error while reading: %s", strerror(-ret));
163
        exit(EXIT_FAILURE);
R
Ryota Ozaki 已提交
164
    }
B
bellard 已提交
165 166

    if (data[510] != 0x55 || data[511] != 0xaa) {
167
        return -EINVAL;
B
bellard 已提交
168 169 170 171 172
    }

    for (i = 0; i < 4; i++) {
        read_partition(&data[446 + 16 * i], &mbr[i]);

173
        if (!mbr[i].system || !mbr[i].nb_sectors_abs) {
B
bellard 已提交
174
            continue;
175
        }
B
bellard 已提交
176 177 178 179 180 181

        if (mbr[i].system == 0xF || mbr[i].system == 0x5) {
            struct partition_record ext[4];
            uint8_t data1[512];
            int j;

182
            if ((ret = blk_read(blk, mbr[i].start_sector_abs, data1, 1)) < 0) {
183
                error_report("error while reading: %s", strerror(-ret));
184
                exit(EXIT_FAILURE);
R
Ryota Ozaki 已提交
185
            }
B
bellard 已提交
186 187 188

            for (j = 0; j < 4; j++) {
                read_partition(&data1[446 + 16 * j], &ext[j]);
189
                if (!ext[j].system || !ext[j].nb_sectors_abs) {
B
bellard 已提交
190
                    continue;
191
                }
B
bellard 已提交
192 193 194 195 196 197 198 199 200 201 202 203 204 205 206

                if ((ext_partnum + j + 1) == partition) {
                    *offset = (uint64_t)ext[j].start_sector_abs << 9;
                    *size = (uint64_t)ext[j].nb_sectors_abs << 9;
                    return 0;
                }
            }
            ext_partnum += 4;
        } else if ((i + 1) == partition) {
            *offset = (uint64_t)mbr[i].start_sector_abs << 9;
            *size = (uint64_t)mbr[i].nb_sectors_abs << 9;
            return 0;
        }
    }

207
    return -ENOENT;
B
bellard 已提交
208 209
}

P
Paolo Bonzini 已提交
210 211
static void termsig_handler(int signum)
{
212
    state = TERMINATE;
P
Paolo Bonzini 已提交
213
    qemu_notify_event();
P
Paolo Bonzini 已提交
214 215
}

216

P
Paolo Bonzini 已提交
217
static void *show_parts(void *arg)
218
{
219
    char *device = arg;
P
Paolo Bonzini 已提交
220 221 222 223 224 225 226 227
    int nbd;

    /* linux just needs an open() to trigger
     * the partition table update
     * but remember to load the module with max_part != 0 :
     *     modprobe nbd max_part=63
     */
    nbd = open(device, O_RDWR);
228
    if (nbd >= 0) {
P
Paolo Bonzini 已提交
229 230 231 232
        close(nbd);
    }
    return NULL;
}
233

P
Paolo Bonzini 已提交
234 235
static void *nbd_client_thread(void *arg)
{
236
    char *device = arg;
P
Paolo Bonzini 已提交
237 238
    off_t size;
    uint32_t nbdflags;
239
    int fd, sock;
P
Paolo Bonzini 已提交
240 241
    int ret;
    pthread_t show_parts_thread;
M
Max Reitz 已提交
242
    Error *local_error = NULL;
P
Paolo Bonzini 已提交
243

244 245

    sock = socket_connect(saddr, &local_error, NULL, NULL);
246
    if (sock < 0) {
247
        error_report_err(local_error);
248 249
        goto out;
    }
P
Paolo Bonzini 已提交
250 251

    ret = nbd_receive_negotiate(sock, NULL, &nbdflags,
252
                                &size, &local_error);
253
    if (ret < 0) {
M
Max Reitz 已提交
254
        if (local_error) {
255
            error_report_err(local_error);
M
Max Reitz 已提交
256
        }
P
Paolo Bonzini 已提交
257
        goto out_socket;
P
Paolo Bonzini 已提交
258 259
    }

260
    fd = open(device, O_RDWR);
261
    if (fd < 0) {
262
        /* Linux-only, we can use %m in printf.  */
263
        error_report("Failed to open %s: %m", device);
P
Paolo Bonzini 已提交
264
        goto out_socket;
265 266
    }

267
    ret = nbd_init(fd, sock, nbdflags, size);
268
    if (ret < 0) {
P
Paolo Bonzini 已提交
269
        goto out_fd;
P
Paolo Bonzini 已提交
270 271 272
    }

    /* update partition table */
273
    pthread_create(&show_parts_thread, NULL, show_parts, device);
P
Paolo Bonzini 已提交
274

275 276 277 278 279 280 281
    if (verbose) {
        fprintf(stderr, "NBD device %s is now connected to %s\n",
                device, srcpath);
    } else {
        /* Close stderr so that the qemu-nbd process exits.  */
        dup2(STDOUT_FILENO, STDERR_FILENO);
    }
P
Paolo Bonzini 已提交
282 283 284

    ret = nbd_client(fd);
    if (ret) {
P
Paolo Bonzini 已提交
285
        goto out_fd;
286
    }
P
Paolo Bonzini 已提交
287 288 289 290
    close(fd);
    kill(getpid(), SIGTERM);
    return (void *) EXIT_SUCCESS;

P
Paolo Bonzini 已提交
291 292 293 294
out_fd:
    close(fd);
out_socket:
    closesocket(sock);
P
Paolo Bonzini 已提交
295 296 297
out:
    kill(getpid(), SIGTERM);
    return (void *) EXIT_FAILURE;
298 299
}

300
static int nbd_can_accept(void)
P
Paolo Bonzini 已提交
301 302 303 304
{
    return nb_fds < shared;
}

305 306 307 308 309 310
static void nbd_export_closed(NBDExport *exp)
{
    assert(state == TERMINATING);
    state = TERMINATED;
}

311 312
static void nbd_update_server_fd_handler(int fd);

313
static void nbd_client_closed(NBDClient *client)
P
Paolo Bonzini 已提交
314
{
315
    nb_fds--;
316 317 318
    if (nb_fds == 0 && !persistent && state == RUNNING) {
        state = TERMINATE;
    }
319
    nbd_update_server_fd_handler(server_fd);
320
    nbd_client_put(client);
P
Paolo Bonzini 已提交
321 322 323 324 325 326 327 328
}

static void nbd_accept(void *opaque)
{
    struct sockaddr_in addr;
    socklen_t addr_len = sizeof(addr);

    int fd = accept(server_fd, (struct sockaddr *)&addr, &addr_len);
P
Paolo Bonzini 已提交
329 330 331 332 333
    if (fd < 0) {
        perror("accept");
        return;
    }

334 335 336 337 338
    if (state >= TERMINATE) {
        close(fd);
        return;
    }

339 340 341
    nb_fds++;
    nbd_update_server_fd_handler(server_fd);
    nbd_client_new(exp, fd, nbd_client_closed);
P
Paolo Bonzini 已提交
342 343
}

344 345 346 347 348 349 350 351 352
static void nbd_update_server_fd_handler(int fd)
{
    if (nbd_can_accept()) {
        qemu_set_fd_handler(fd, nbd_accept, NULL, (void *)(uintptr_t)fd);
    } else {
        qemu_set_fd_handler(fd, NULL, NULL, NULL);
    }
}

353 354 355 356 357 358 359 360 361

static SocketAddress *nbd_build_socket_address(const char *sockpath,
                                               const char *bindto,
                                               const char *port)
{
    SocketAddress *saddr;

    saddr = g_new0(SocketAddress, 1);
    if (sockpath) {
362 363 364
        saddr->type = SOCKET_ADDRESS_KIND_UNIX;
        saddr->u.q_unix = g_new0(UnixSocketAddress, 1);
        saddr->u.q_unix->path = g_strdup(sockpath);
365
    } else {
366 367 368
        saddr->type = SOCKET_ADDRESS_KIND_INET;
        saddr->u.inet = g_new0(InetSocketAddress, 1);
        saddr->u.inet->host = g_strdup(bindto);
369
        if (port) {
370
            saddr->u.inet->port = g_strdup(port);
371
        } else  {
372
            saddr->u.inet->port = g_strdup_printf("%d", NBD_DEFAULT_PORT);
373 374 375 376 377 378 379
        }
    }

    return saddr;
}


380 381 382 383 384 385 386 387 388 389
static QemuOptsList qemu_object_opts = {
    .name = "object",
    .implied_opt_name = "qom-type",
    .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head),
    .desc = {
        { }
    },
};


B
bellard 已提交
390 391
int main(int argc, char **argv)
{
M
Markus Armbruster 已提交
392
    BlockBackend *blk;
B
bellard 已提交
393 394
    BlockDriverState *bs;
    off_t dev_offset = 0;
P
Paolo Bonzini 已提交
395
    uint32_t nbdflags = 0;
396
    bool disconnect = false;
B
bellard 已提交
397
    const char *bindto = "0.0.0.0";
398 399
    const char *port = NULL;
    char *sockpath = NULL;
400
    char *device = NULL;
B
bellard 已提交
401
    off_t fd_size;
402 403 404
    QemuOpts *sn_opts = NULL;
    const char *sn_id_or_name = NULL;
    const char *sopt = "hVb:o:p:rsnP:c:dvk:e:f:tl:";
B
bellard 已提交
405
    struct option lopt[] = {
406 407 408 409 410 411 412 413 414 415 416
        { "help", 0, NULL, 'h' },
        { "version", 0, NULL, 'V' },
        { "bind", 1, NULL, 'b' },
        { "port", 1, NULL, 'p' },
        { "socket", 1, NULL, 'k' },
        { "offset", 1, NULL, 'o' },
        { "read-only", 0, NULL, 'r' },
        { "partition", 1, NULL, 'P' },
        { "connect", 1, NULL, 'c' },
        { "disconnect", 0, NULL, 'd' },
        { "snapshot", 0, NULL, 's' },
417
        { "load-snapshot", 1, NULL, 'l' },
418
        { "nocache", 0, NULL, 'n' },
419 420
        { "cache", 1, NULL, QEMU_NBD_OPT_CACHE },
        { "aio", 1, NULL, QEMU_NBD_OPT_AIO },
P
Paolo Bonzini 已提交
421
        { "discard", 1, NULL, QEMU_NBD_OPT_DISCARD },
422
        { "detect-zeroes", 1, NULL, QEMU_NBD_OPT_DETECT_ZEROES },
423
        { "shared", 1, NULL, 'e' },
424
        { "format", 1, NULL, 'f' },
425 426
        { "persistent", 0, NULL, 't' },
        { "verbose", 0, NULL, 'v' },
427
        { "object", 1, NULL, QEMU_NBD_OPT_OBJECT },
428
        { NULL, 0, NULL, 0 }
B
bellard 已提交
429 430 431 432
    };
    int ch;
    int opt_ind = 0;
    char *end;
433
    int flags = BDRV_O_RDWR;
B
bellard 已提交
434
    int partition = -1;
435
    int ret = 0;
436
    int fd;
437
    bool seen_cache = false;
P
Paolo Bonzini 已提交
438
    bool seen_discard = false;
439
    bool seen_aio = false;
P
Paolo Bonzini 已提交
440
    pthread_t client_thread;
441
    const char *fmt = NULL;
442
    Error *local_err = NULL;
443
    BlockdevDetectZeroesOptions detect_zeroes = BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
444
    QDict *options = NULL;
B
bellard 已提交
445

P
Paolo Bonzini 已提交
446 447 448
    /* The client thread uses SIGTERM to interrupt the server.  A signal
     * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
     */
P
Paolo Bonzini 已提交
449 450 451 452
    struct sigaction sa_sigterm;
    memset(&sa_sigterm, 0, sizeof(sa_sigterm));
    sa_sigterm.sa_handler = termsig_handler;
    sigaction(SIGTERM, &sa_sigterm, NULL);
453 454
    module_call_init(MODULE_INIT_QOM);
    qemu_add_opts(&qemu_object_opts);
455
    qemu_init_exec_dir(argv[0]);
P
Paolo Bonzini 已提交
456

B
bellard 已提交
457 458 459
    while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
        switch (ch) {
        case 's':
460 461 462
            flags |= BDRV_O_SNAPSHOT;
            break;
        case 'n':
463 464 465 466
            optarg = (char *) "none";
            /* fallthrough */
        case QEMU_NBD_OPT_CACHE:
            if (seen_cache) {
467 468
                error_report("-n and --cache can only be specified once");
                exit(EXIT_FAILURE);
469 470 471
            }
            seen_cache = true;
            if (bdrv_parse_cache_flags(optarg, &flags) == -1) {
472 473
                error_report("Invalid cache mode `%s'", optarg);
                exit(EXIT_FAILURE);
474
            }
B
bellard 已提交
475
            break;
476 477
        case QEMU_NBD_OPT_AIO:
            if (seen_aio) {
478 479
                error_report("--aio can only be specified once");
                exit(EXIT_FAILURE);
480 481 482 483 484 485 486
            }
            seen_aio = true;
            if (!strcmp(optarg, "native")) {
                flags |= BDRV_O_NATIVE_AIO;
            } else if (!strcmp(optarg, "threads")) {
                /* this is the default */
            } else {
487 488
               error_report("invalid aio mode `%s'", optarg);
               exit(EXIT_FAILURE);
489 490
            }
            break;
P
Paolo Bonzini 已提交
491 492
        case QEMU_NBD_OPT_DISCARD:
            if (seen_discard) {
493 494
                error_report("--discard can only be specified once");
                exit(EXIT_FAILURE);
P
Paolo Bonzini 已提交
495 496 497
            }
            seen_discard = true;
            if (bdrv_parse_discard_flags(optarg, &flags) == -1) {
498 499
                error_report("Invalid discard mode `%s'", optarg);
                exit(EXIT_FAILURE);
P
Paolo Bonzini 已提交
500 501
            }
            break;
502 503 504 505
        case QEMU_NBD_OPT_DETECT_ZEROES:
            detect_zeroes =
                qapi_enum_parse(BlockdevDetectZeroesOptions_lookup,
                                optarg,
506
                                BLOCKDEV_DETECT_ZEROES_OPTIONS__MAX,
507 508 509
                                BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
                                &local_err);
            if (local_err) {
510 511
                error_reportf_err(local_err,
                                  "Failed to parse detect_zeroes mode: ");
512
                exit(EXIT_FAILURE);
513 514 515
            }
            if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
                !(flags & BDRV_O_UNMAP)) {
516 517 518
                error_report("setting detect-zeroes to unmap is not allowed "
                             "without setting discard operation to unmap");
                exit(EXIT_FAILURE);
519 520
            }
            break;
B
bellard 已提交
521 522 523 524
        case 'b':
            bindto = optarg;
            break;
        case 'p':
525
            port = optarg;
B
bellard 已提交
526 527 528 529
            break;
        case 'o':
                dev_offset = strtoll (optarg, &end, 0);
            if (*end) {
530 531
                error_report("Invalid offset `%s'", optarg);
                exit(EXIT_FAILURE);
B
bellard 已提交
532 533
            }
            if (dev_offset < 0) {
534 535
                error_report("Offset must be positive `%s'", optarg);
                exit(EXIT_FAILURE);
B
bellard 已提交
536 537
            }
            break;
538 539
        case 'l':
            if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
540 541
                sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
                                                  optarg, false);
542
                if (!sn_opts) {
543 544 545
                    error_report("Failed in parsing snapshot param `%s'",
                                 optarg);
                    exit(EXIT_FAILURE);
546 547 548 549 550
                }
            } else {
                sn_id_or_name = optarg;
            }
            /* fall through */
B
bellard 已提交
551
        case 'r':
P
Paolo Bonzini 已提交
552
            nbdflags |= NBD_FLAG_READ_ONLY;
N
Naphtali Sprei 已提交
553
            flags &= ~BDRV_O_RDWR;
B
bellard 已提交
554 555 556
            break;
        case 'P':
            partition = strtol(optarg, &end, 0);
557
            if (*end) {
558 559
                error_report("Invalid partition `%s'", optarg);
                exit(EXIT_FAILURE);
560 561
            }
            if (partition < 1 || partition > 8) {
562 563
                error_report("Invalid partition %d", partition);
                exit(EXIT_FAILURE);
564
            }
B
bellard 已提交
565
            break;
566
        case 'k':
P
Paolo Bonzini 已提交
567
            sockpath = optarg;
568
            if (sockpath[0] != '/') {
569
                error_report("socket path must be absolute");
570
                exit(EXIT_FAILURE);
571
            }
572 573 574 575 576 577 578
            break;
        case 'd':
            disconnect = true;
            break;
        case 'c':
            device = optarg;
            break;
579 580 581
        case 'e':
            shared = strtol(optarg, &end, 0);
            if (*end) {
582 583
                error_report("Invalid shared device number '%s'", optarg);
                exit(EXIT_FAILURE);
584 585
            }
            if (shared < 1) {
586
                error_report("Shared device number must be greater than 0");
587
                exit(EXIT_FAILURE);
588 589
            }
            break;
590 591 592
        case 'f':
            fmt = optarg;
            break;
593 594 595
        case 't':
            persistent = 1;
            break;
B
bellard 已提交
596 597 598 599 600 601 602 603 604 605 606 607
        case 'v':
            verbose = 1;
            break;
        case 'V':
            version(argv[0]);
            exit(0);
            break;
        case 'h':
            usage(argv[0]);
            exit(0);
            break;
        case '?':
608 609
            error_report("Try `%s --help' for more information.", argv[0]);
            exit(EXIT_FAILURE);
610 611 612 613 614 615 616 617
        case QEMU_NBD_OPT_OBJECT: {
            QemuOpts *opts;
            opts = qemu_opts_parse_noisily(&qemu_object_opts,
                                           optarg, true);
            if (!opts) {
                exit(EXIT_FAILURE);
            }
        }   break;
B
bellard 已提交
618 619 620 621
        }
    }

    if ((argc - optind) != 1) {
622 623
        error_report("Invalid number of arguments");
        error_printf("Try `%s --help' for more information.\n", argv[0]);
624
        exit(EXIT_FAILURE);
B
bellard 已提交
625 626
    }

627 628 629 630 631 632 633
    if (qemu_opts_foreach(&qemu_object_opts,
                          user_creatable_add_opts_foreach,
                          NULL, &local_err)) {
        error_report_err(local_err);
        exit(EXIT_FAILURE);
    }

634 635
    if (disconnect) {
        fd = open(argv[optind], O_RDWR);
636
        if (fd < 0) {
637 638 639
            error_report("Cannot open %s: %s", argv[optind],
                         strerror(errno));
            exit(EXIT_FAILURE);
640
        }
641 642 643 644 645 646
        nbd_disconnect(fd);

        close(fd);

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

647
        return 0;
648 649
    }

650 651 652 653 654
    if (device && !verbose) {
        int stderr_fd[2];
        pid_t pid;
        int ret;

655
        if (qemu_pipe(stderr_fd) < 0) {
656 657 658
            error_report("Error setting up communication pipe: %s",
                         strerror(errno));
            exit(EXIT_FAILURE);
659 660 661 662 663 664
        }

        /* Now daemonize, but keep a communication channel open to
         * print errors and exit with the proper status code.
         */
        pid = fork();
M
Max Reitz 已提交
665
        if (pid < 0) {
666 667
            error_report("Failed to fork: %s", strerror(errno));
            exit(EXIT_FAILURE);
M
Max Reitz 已提交
668
        } else if (pid == 0) {
669
            close(stderr_fd[0]);
670
            ret = qemu_daemon(1, 0);
671 672 673

            /* Temporarily redirect stderr to the parent's pipe...  */
            dup2(stderr_fd[1], STDERR_FILENO);
674
            if (ret < 0) {
675 676
                error_report("Failed to daemonize: %s", strerror(errno));
                exit(EXIT_FAILURE);
677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692
            }

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

            /* In the parent.  Print error messages from the child until
             * it closes the pipe.
             */
            close(stderr_fd[1]);
            buf = g_malloc(1024);
            while ((ret = read(stderr_fd[0], buf, 1024)) > 0) {
                errors = true;
                ret = qemu_write_full(STDERR_FILENO, buf, ret);
693
                if (ret < 0) {
694 695 696
                    exit(EXIT_FAILURE);
                }
            }
697
            if (ret < 0) {
698 699 700
                error_report("Cannot read from daemon: %s",
                             strerror(errno));
                exit(EXIT_FAILURE);
701 702 703 704 705 706 707 708 709
            }

            /* Usually the daemon should not print any message.
             * Exit with zero status in that case.
             */
            exit(errors);
        }
    }

710 711 712
    if (device != NULL && sockpath == NULL) {
        sockpath = g_malloc(128);
        snprintf(sockpath, 128, SOCKET_PATH, basename(device));
713 714
    }

715 716
    saddr = nbd_build_socket_address(sockpath, bindto, port);

717
    if (qemu_init_main_loop(&local_err)) {
718
        error_report_err(local_err);
719 720
        exit(EXIT_FAILURE);
    }
721 722 723
    bdrv_init();
    atexit(bdrv_close_all);

724
    if (fmt) {
725 726
        options = qdict_new();
        qdict_put(options, "driver", qstring_from_str(fmt));
727 728
    }

729
    srcpath = argv[optind];
730 731
    blk = blk_new_open("hda", srcpath, NULL, options, flags, &local_err);
    if (!blk) {
732 733
        error_reportf_err(local_err, "Failed to blk_new_open '%s': ",
                          argv[optind]);
734
        exit(EXIT_FAILURE);
735
    }
736
    bs = blk_bs(blk);
737

738 739 740 741 742 743 744 745 746 747
    if (sn_opts) {
        ret = bdrv_snapshot_load_tmp(bs,
                                     qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
                                     qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
                                     &local_err);
    } else if (sn_id_or_name) {
        ret = bdrv_snapshot_load_tmp_by_id_or_name(bs, sn_id_or_name,
                                                   &local_err);
    }
    if (ret < 0) {
748
        error_reportf_err(local_err, "Failed to load snapshot: ");
749
        exit(EXIT_FAILURE);
750 751
    }

752
    bs->detect_zeroes = detect_zeroes;
753
    fd_size = blk_getlength(blk);
M
Max Reitz 已提交
754
    if (fd_size < 0) {
755 756 757
        error_report("Failed to determine the image length: %s",
                     strerror(-fd_size));
        exit(EXIT_FAILURE);
M
Max Reitz 已提交
758
    }
759

760
    if (partition != -1) {
761
        ret = find_partition(blk, partition, &dev_offset, &fd_size);
762
        if (ret < 0) {
763
            error_report("Could not find partition %d: %s", partition,
764
                         strerror(-ret));
765
            exit(EXIT_FAILURE);
766
        }
767 768
    }

M
Max Reitz 已提交
769 770 771
    exp = nbd_export_new(blk, dev_offset, fd_size, nbdflags, nbd_export_closed,
                         &local_err);
    if (!exp) {
772
        error_report_err(local_err);
773
        exit(EXIT_FAILURE);
M
Max Reitz 已提交
774
    }
775

776
    fd = socket_listen(saddr, &local_err);
777
    if (fd < 0) {
778
        error_report_err(local_err);
B
bellard 已提交
779
        return 1;
P
Paolo Bonzini 已提交
780
    }
P
Paolo Bonzini 已提交
781 782 783 784

    if (device) {
        int ret;

785
        ret = pthread_create(&client_thread, NULL, nbd_client_thread, device);
P
Paolo Bonzini 已提交
786
        if (ret != 0) {
787 788
            error_report("Failed to create client thread: %s", strerror(ret));
            exit(EXIT_FAILURE);
P
Paolo Bonzini 已提交
789 790 791 792 793 794
        }
    } else {
        /* Shut up GCC warnings.  */
        memset(&client_thread, 0, sizeof(client_thread));
    }

795 796
    server_fd = fd;
    nbd_update_server_fd_handler(fd);
B
bellard 已提交
797

798 799 800
    /* now when the initialization is (almost) complete, chdir("/")
     * to free any busy filesystems */
    if (chdir("/") < 0) {
801 802 803
        error_report("Could not chdir to root directory: %s",
                     strerror(errno));
        exit(EXIT_FAILURE);
804 805
    }

806
    state = RUNNING;
807
    do {
P
Paolo Bonzini 已提交
808
        main_loop_wait(false);
809 810 811 812 813 814 815
        if (state == TERMINATE) {
            state = TERMINATING;
            nbd_export_close(exp);
            nbd_export_put(exp);
            exp = NULL;
        }
    } while (state != TERMINATED);
B
bellard 已提交
816

M
Markus Armbruster 已提交
817
    blk_unref(blk);
P
Paolo Bonzini 已提交
818 819 820
    if (sockpath) {
        unlink(sockpath);
    }
B
bellard 已提交
821

822
    qemu_opts_del(sn_opts);
823

P
Paolo Bonzini 已提交
824 825 826 827 828 829 830
    if (device) {
        void *ret;
        pthread_join(client_thread, &ret);
        exit(ret != NULL);
    } else {
        exit(EXIT_SUCCESS);
    }
B
bellard 已提交
831
}