qemu-nbd.c 23.7 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
 */

19
#include "qemu-common.h"
M
Markus Armbruster 已提交
20
#include "sysemu/block-backend.h"
21
#include "block/block_int.h"
22
#include "block/nbd.h"
23
#include "qemu/main-loop.h"
24 25
#include "qemu/sockets.h"
#include "qemu/error-report.h"
26
#include "block/snapshot.h"
27
#include "qapi/util.h"
28
#include "qapi/qmp/qstring.h"
B
bellard 已提交
29 30 31 32

#include <stdarg.h>
#include <stdio.h>
#include <getopt.h>
33
#include <sys/types.h>
B
bellard 已提交
34 35 36 37
#include <sys/socket.h>
#include <netinet/in.h>
#include <netinet/tcp.h>
#include <arpa/inet.h>
38
#include <signal.h>
39
#include <libgen.h>
P
Paolo Bonzini 已提交
40
#include <pthread.h>
41

42 43 44 45
#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
46
#define QEMU_NBD_OPT_DETECT_ZEROES 4
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 80 81
"\n"
#ifdef __linux__
"Kernel NBD client support:\n"
82 83
"  -c, --connect=DEV         connect FILE to the local NBD device DEV\n"
"  -d, --disconnect          disconnect the specified device\n"
84 85 86 87
"\n"
#endif
"\n"
"Block device options:\n"
88 89 90 91 92
"  -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"
93
"  -l, --load-snapshot=SNAPSHOT_PARAM\n"
94 95 96 97 98 99 100
"                            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"
101
"      --discard=MODE        set discard mode (ignore, unmap)\n"
102
"      --detect-zeroes=MODE  set detect-zeroes mode (off, on, unmap)\n"
103 104
"\n"
"Report bugs to <qemu-devel@nongnu.org>\n"
105
    , name, NBD_DEFAULT_PORT, "DEVICE");
B
bellard 已提交
106 107 108 109 110
}

static void version(const char *name)
{
    printf(
111
"%s version 0.0.1\n"
B
bellard 已提交
112 113 114 115 116
"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"
117
    , name);
B
bellard 已提交
118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143
}

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;
144 145 146

    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
147 148
}

149
static int find_partition(BlockBackend *blk, int partition,
B
bellard 已提交
150 151 152 153 154 155
                          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
156
    int ret;
B
bellard 已提交
157

158
    if ((ret = blk_read(blk, 0, data, 1)) < 0) {
R
Ryota Ozaki 已提交
159
        errno = -ret;
160 161
        error_report("error while reading: %s", strerror(errno));
        exit(EXIT_FAILURE);
R
Ryota Ozaki 已提交
162
    }
B
bellard 已提交
163 164

    if (data[510] != 0x55 || data[511] != 0xaa) {
165
        return -EINVAL;
B
bellard 已提交
166 167 168 169 170
    }

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

171
        if (!mbr[i].system || !mbr[i].nb_sectors_abs) {
B
bellard 已提交
172
            continue;
173
        }
B
bellard 已提交
174 175 176 177 178 179

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

180
            if ((ret = blk_read(blk, mbr[i].start_sector_abs, data1, 1)) < 0) {
R
Ryota Ozaki 已提交
181
                errno = -ret;
182 183
                error_report("error while reading: %s", strerror(errno));
                exit(EXIT_FAILURE);
R
Ryota Ozaki 已提交
184
            }
B
bellard 已提交
185 186 187

            for (j = 0; j < 4; j++) {
                read_partition(&data1[446 + 16 * j], &ext[j]);
188
                if (!ext[j].system || !ext[j].nb_sectors_abs) {
B
bellard 已提交
189
                    continue;
190
                }
B
bellard 已提交
191 192 193 194 195 196 197 198 199 200 201 202 203 204 205

                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;
        }
    }

206
    return -ENOENT;
B
bellard 已提交
207 208
}

P
Paolo Bonzini 已提交
209 210
static void termsig_handler(int signum)
{
211
    state = TERMINATE;
P
Paolo Bonzini 已提交
212
    qemu_notify_event();
P
Paolo Bonzini 已提交
213 214
}

215

P
Paolo Bonzini 已提交
216
static void *show_parts(void *arg)
217
{
218
    char *device = arg;
P
Paolo Bonzini 已提交
219 220 221 222 223 224 225 226
    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);
227
    if (nbd >= 0) {
P
Paolo Bonzini 已提交
228 229 230 231
        close(nbd);
    }
    return NULL;
}
232

P
Paolo Bonzini 已提交
233 234
static void *nbd_client_thread(void *arg)
{
235
    char *device = arg;
P
Paolo Bonzini 已提交
236 237
    off_t size;
    uint32_t nbdflags;
238
    int fd, sock;
P
Paolo Bonzini 已提交
239 240
    int ret;
    pthread_t show_parts_thread;
M
Max Reitz 已提交
241
    Error *local_error = NULL;
P
Paolo Bonzini 已提交
242

243 244

    sock = socket_connect(saddr, &local_error, NULL, NULL);
245
    if (sock < 0) {
246
        error_report_err(local_error);
247 248
        goto out;
    }
P
Paolo Bonzini 已提交
249 250

    ret = nbd_receive_negotiate(sock, NULL, &nbdflags,
251
                                &size, &local_error);
252
    if (ret < 0) {
M
Max Reitz 已提交
253 254 255 256
        if (local_error) {
            fprintf(stderr, "%s\n", error_get_pretty(local_error));
            error_free(local_error);
        }
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.  */
H
Hani Benhabiles 已提交
263
        fprintf(stderr, "Failed to open %s: %m\n", 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
    if (nbd_client_new(exp, fd, nbd_client_closed)) {
P
Paolo Bonzini 已提交
340
        nb_fds++;
341
        nbd_update_server_fd_handler(server_fd);
342
    } else {
343
        shutdown(fd, 2);
344
        close(fd);
P
Paolo Bonzini 已提交
345 346 347
    }
}

348 349 350 351 352 353 354 355 356
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);
    }
}

357 358 359 360 361 362 363 364 365

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

    saddr = g_new0(SocketAddress, 1);
    if (sockpath) {
366 367 368
        saddr->type = SOCKET_ADDRESS_KIND_UNIX;
        saddr->u.q_unix = g_new0(UnixSocketAddress, 1);
        saddr->u.q_unix->path = g_strdup(sockpath);
369
    } else {
370 371 372
        saddr->type = SOCKET_ADDRESS_KIND_INET;
        saddr->u.inet = g_new0(InetSocketAddress, 1);
        saddr->u.inet->host = g_strdup(bindto);
373
        if (port) {
374
            saddr->u.inet->port = g_strdup(port);
375
        } else  {
376
            saddr->u.inet->port = g_strdup_printf("%d", NBD_DEFAULT_PORT);
377 378 379 380 381 382 383
        }
    }

    return saddr;
}


B
bellard 已提交
384 385
int main(int argc, char **argv)
{
M
Markus Armbruster 已提交
386
    BlockBackend *blk;
B
bellard 已提交
387 388
    BlockDriverState *bs;
    off_t dev_offset = 0;
P
Paolo Bonzini 已提交
389
    uint32_t nbdflags = 0;
390
    bool disconnect = false;
B
bellard 已提交
391
    const char *bindto = "0.0.0.0";
392 393
    const char *port = NULL;
    char *sockpath = NULL;
394
    char *device = NULL;
B
bellard 已提交
395
    off_t fd_size;
396 397 398
    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
399
    struct option lopt[] = {
400 401 402 403 404 405 406 407 408 409 410
        { "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' },
411
        { "load-snapshot", 1, NULL, 'l' },
412
        { "nocache", 0, NULL, 'n' },
413 414
        { "cache", 1, NULL, QEMU_NBD_OPT_CACHE },
        { "aio", 1, NULL, QEMU_NBD_OPT_AIO },
P
Paolo Bonzini 已提交
415
        { "discard", 1, NULL, QEMU_NBD_OPT_DISCARD },
416
        { "detect-zeroes", 1, NULL, QEMU_NBD_OPT_DETECT_ZEROES },
417
        { "shared", 1, NULL, 'e' },
418
        { "format", 1, NULL, 'f' },
419 420 421
        { "persistent", 0, NULL, 't' },
        { "verbose", 0, NULL, 'v' },
        { NULL, 0, NULL, 0 }
B
bellard 已提交
422 423 424 425
    };
    int ch;
    int opt_ind = 0;
    char *end;
426
    int flags = BDRV_O_RDWR;
B
bellard 已提交
427
    int partition = -1;
428
    int ret = 0;
429
    int fd;
430
    bool seen_cache = false;
P
Paolo Bonzini 已提交
431
    bool seen_discard = false;
432
    bool seen_aio = false;
P
Paolo Bonzini 已提交
433
    pthread_t client_thread;
434
    const char *fmt = NULL;
435
    Error *local_err = NULL;
436
    BlockdevDetectZeroesOptions detect_zeroes = BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
437
    QDict *options = NULL;
B
bellard 已提交
438

P
Paolo Bonzini 已提交
439 440 441
    /* 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
442 443 444 445
    struct sigaction sa_sigterm;
    memset(&sa_sigterm, 0, sizeof(sa_sigterm));
    sa_sigterm.sa_handler = termsig_handler;
    sigaction(SIGTERM, &sa_sigterm, NULL);
446
    qemu_init_exec_dir(argv[0]);
P
Paolo Bonzini 已提交
447

B
bellard 已提交
448 449 450
    while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
        switch (ch) {
        case 's':
451 452 453
            flags |= BDRV_O_SNAPSHOT;
            break;
        case 'n':
454 455 456 457
            optarg = (char *) "none";
            /* fallthrough */
        case QEMU_NBD_OPT_CACHE:
            if (seen_cache) {
458 459
                error_report("-n and --cache can only be specified once");
                exit(EXIT_FAILURE);
460 461 462
            }
            seen_cache = true;
            if (bdrv_parse_cache_flags(optarg, &flags) == -1) {
463 464
                error_report("Invalid cache mode `%s'", optarg);
                exit(EXIT_FAILURE);
465
            }
B
bellard 已提交
466
            break;
467 468
        case QEMU_NBD_OPT_AIO:
            if (seen_aio) {
469 470
                error_report("--aio can only be specified once");
                exit(EXIT_FAILURE);
471 472 473 474 475 476 477
            }
            seen_aio = true;
            if (!strcmp(optarg, "native")) {
                flags |= BDRV_O_NATIVE_AIO;
            } else if (!strcmp(optarg, "threads")) {
                /* this is the default */
            } else {
478 479
               error_report("invalid aio mode `%s'", optarg);
               exit(EXIT_FAILURE);
480 481
            }
            break;
P
Paolo Bonzini 已提交
482 483
        case QEMU_NBD_OPT_DISCARD:
            if (seen_discard) {
484 485
                error_report("--discard can only be specified once");
                exit(EXIT_FAILURE);
P
Paolo Bonzini 已提交
486 487 488
            }
            seen_discard = true;
            if (bdrv_parse_discard_flags(optarg, &flags) == -1) {
489 490
                error_report("Invalid discard mode `%s'", optarg);
                exit(EXIT_FAILURE);
P
Paolo Bonzini 已提交
491 492
            }
            break;
493 494 495 496
        case QEMU_NBD_OPT_DETECT_ZEROES:
            detect_zeroes =
                qapi_enum_parse(BlockdevDetectZeroesOptions_lookup,
                                optarg,
497
                                BLOCKDEV_DETECT_ZEROES_OPTIONS__MAX,
498 499 500
                                BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
                                &local_err);
            if (local_err) {
501 502 503
                error_report("Failed to parse detect_zeroes mode: %s",
                             error_get_pretty(local_err));
                exit(EXIT_FAILURE);
504 505 506
            }
            if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
                !(flags & BDRV_O_UNMAP)) {
507 508 509
                error_report("setting detect-zeroes to unmap is not allowed "
                             "without setting discard operation to unmap");
                exit(EXIT_FAILURE);
510 511
            }
            break;
B
bellard 已提交
512 513 514 515
        case 'b':
            bindto = optarg;
            break;
        case 'p':
516
            port = optarg;
B
bellard 已提交
517 518 519 520
            break;
        case 'o':
                dev_offset = strtoll (optarg, &end, 0);
            if (*end) {
521 522
                error_report("Invalid offset `%s'", optarg);
                exit(EXIT_FAILURE);
B
bellard 已提交
523 524
            }
            if (dev_offset < 0) {
525 526
                error_report("Offset must be positive `%s'", optarg);
                exit(EXIT_FAILURE);
B
bellard 已提交
527 528
            }
            break;
529 530
        case 'l':
            if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
531 532
                sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
                                                  optarg, false);
533
                if (!sn_opts) {
534 535 536
                    error_report("Failed in parsing snapshot param `%s'",
                                 optarg);
                    exit(EXIT_FAILURE);
537 538 539 540 541
                }
            } else {
                sn_id_or_name = optarg;
            }
            /* fall through */
B
bellard 已提交
542
        case 'r':
P
Paolo Bonzini 已提交
543
            nbdflags |= NBD_FLAG_READ_ONLY;
N
Naphtali Sprei 已提交
544
            flags &= ~BDRV_O_RDWR;
B
bellard 已提交
545 546 547
            break;
        case 'P':
            partition = strtol(optarg, &end, 0);
548
            if (*end) {
549 550
                error_report("Invalid partition `%s'", optarg);
                exit(EXIT_FAILURE);
551 552
            }
            if (partition < 1 || partition > 8) {
553 554
                error_report("Invalid partition %d", partition);
                exit(EXIT_FAILURE);
555
            }
B
bellard 已提交
556
            break;
557
        case 'k':
P
Paolo Bonzini 已提交
558
            sockpath = optarg;
559
            if (sockpath[0] != '/') {
560 561
                error_report("socket path must be absolute\n");
                exit(EXIT_FAILURE);
562
            }
563 564 565 566 567 568 569
            break;
        case 'd':
            disconnect = true;
            break;
        case 'c':
            device = optarg;
            break;
570 571 572
        case 'e':
            shared = strtol(optarg, &end, 0);
            if (*end) {
573 574
                error_report("Invalid shared device number '%s'", optarg);
                exit(EXIT_FAILURE);
575 576
            }
            if (shared < 1) {
577 578
                error_report("Shared device number must be greater than 0\n");
                exit(EXIT_FAILURE);
579 580
            }
            break;
581 582 583
        case 'f':
            fmt = optarg;
            break;
584 585 586
        case 't':
            persistent = 1;
            break;
B
bellard 已提交
587 588 589 590 591 592 593 594 595 596 597 598
        case 'v':
            verbose = 1;
            break;
        case 'V':
            version(argv[0]);
            exit(0);
            break;
        case 'h':
            usage(argv[0]);
            exit(0);
            break;
        case '?':
599 600
            error_report("Try `%s --help' for more information.", argv[0]);
            exit(EXIT_FAILURE);
B
bellard 已提交
601 602 603 604
        }
    }

    if ((argc - optind) != 1) {
605 606 607 608
        error_report("Invalid number of argument.\n"
                     "Try `%s --help' for more information.",
                     argv[0]);
        exit(EXIT_FAILURE);
B
bellard 已提交
609 610
    }

611 612
    if (disconnect) {
        fd = open(argv[optind], O_RDWR);
613
        if (fd < 0) {
614 615 616
            error_report("Cannot open %s: %s", argv[optind],
                         strerror(errno));
            exit(EXIT_FAILURE);
617
        }
618 619 620 621 622 623
        nbd_disconnect(fd);

        close(fd);

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

624
        return 0;
625 626
    }

627 628 629 630 631
    if (device && !verbose) {
        int stderr_fd[2];
        pid_t pid;
        int ret;

632
        if (qemu_pipe(stderr_fd) < 0) {
633 634 635
            error_report("Error setting up communication pipe: %s",
                         strerror(errno));
            exit(EXIT_FAILURE);
636 637 638 639 640 641
        }

        /* 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
642
        if (pid < 0) {
643 644
            error_report("Failed to fork: %s", strerror(errno));
            exit(EXIT_FAILURE);
M
Max Reitz 已提交
645
        } else if (pid == 0) {
646
            close(stderr_fd[0]);
647
            ret = qemu_daemon(1, 0);
648 649 650

            /* Temporarily redirect stderr to the parent's pipe...  */
            dup2(stderr_fd[1], STDERR_FILENO);
651
            if (ret < 0) {
652 653
                error_report("Failed to daemonize: %s", strerror(errno));
                exit(EXIT_FAILURE);
654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669
            }

            /* ... 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);
670
                if (ret < 0) {
671 672 673
                    exit(EXIT_FAILURE);
                }
            }
674
            if (ret < 0) {
675 676 677
                error_report("Cannot read from daemon: %s",
                             strerror(errno));
                exit(EXIT_FAILURE);
678 679 680 681 682 683 684 685 686
            }

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

687 688 689
    if (device != NULL && sockpath == NULL) {
        sockpath = g_malloc(128);
        snprintf(sockpath, 128, SOCKET_PATH, basename(device));
690 691
    }

692 693
    saddr = nbd_build_socket_address(sockpath, bindto, port);

694
    if (qemu_init_main_loop(&local_err)) {
695
        error_report_err(local_err);
696 697
        exit(EXIT_FAILURE);
    }
698 699 700
    bdrv_init();
    atexit(bdrv_close_all);

701
    if (fmt) {
702 703
        options = qdict_new();
        qdict_put(options, "driver", qstring_from_str(fmt));
704 705
    }

706
    srcpath = argv[optind];
707 708
    blk = blk_new_open("hda", srcpath, NULL, options, flags, &local_err);
    if (!blk) {
709 710 711
        error_report("Failed to blk_new_open '%s': %s", argv[optind],
                     error_get_pretty(local_err));
        exit(EXIT_FAILURE);
712
    }
713
    bs = blk_bs(blk);
714

715 716 717 718 719 720 721 722 723 724 725
    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) {
        errno = -ret;
726 727 728
        error_report("Failed to load snapshot: %s: %s",
                     error_get_pretty(local_err), strerror(errno));
        exit(EXIT_FAILURE);
729 730
    }

731
    bs->detect_zeroes = detect_zeroes;
732
    fd_size = blk_getlength(blk);
M
Max Reitz 已提交
733
    if (fd_size < 0) {
734 735 736
        error_report("Failed to determine the image length: %s",
                     strerror(-fd_size));
        exit(EXIT_FAILURE);
M
Max Reitz 已提交
737
    }
738

739
    if (partition != -1) {
740
        ret = find_partition(blk, partition, &dev_offset, &fd_size);
741 742
        if (ret < 0) {
            errno = -ret;
743 744 745
            error_report("Could not find partition %d: %s", partition,
                         strerror(errno));
            exit(EXIT_FAILURE);
746
        }
747 748
    }

M
Max Reitz 已提交
749 750 751
    exp = nbd_export_new(blk, dev_offset, fd_size, nbdflags, nbd_export_closed,
                         &local_err);
    if (!exp) {
752
        error_report_err(local_err);
753
        exit(EXIT_FAILURE);
M
Max Reitz 已提交
754
    }
755

756
    fd = socket_listen(saddr, &local_err);
757
    if (fd < 0) {
758
        error_report_err(local_err);
B
bellard 已提交
759
        return 1;
P
Paolo Bonzini 已提交
760
    }
P
Paolo Bonzini 已提交
761 762 763 764

    if (device) {
        int ret;

765
        ret = pthread_create(&client_thread, NULL, nbd_client_thread, device);
P
Paolo Bonzini 已提交
766
        if (ret != 0) {
767 768
            error_report("Failed to create client thread: %s", strerror(ret));
            exit(EXIT_FAILURE);
P
Paolo Bonzini 已提交
769 770 771 772 773 774
        }
    } else {
        /* Shut up GCC warnings.  */
        memset(&client_thread, 0, sizeof(client_thread));
    }

775 776
    server_fd = fd;
    nbd_update_server_fd_handler(fd);
B
bellard 已提交
777

778 779 780
    /* now when the initialization is (almost) complete, chdir("/")
     * to free any busy filesystems */
    if (chdir("/") < 0) {
781 782 783
        error_report("Could not chdir to root directory: %s",
                     strerror(errno));
        exit(EXIT_FAILURE);
784 785
    }

786
    state = RUNNING;
787
    do {
P
Paolo Bonzini 已提交
788
        main_loop_wait(false);
789 790 791 792 793 794 795
        if (state == TERMINATE) {
            state = TERMINATING;
            nbd_export_close(exp);
            nbd_export_put(exp);
            exp = NULL;
        }
    } while (state != TERMINATED);
B
bellard 已提交
796

M
Markus Armbruster 已提交
797
    blk_unref(blk);
P
Paolo Bonzini 已提交
798 799 800
    if (sockpath) {
        unlink(sockpath);
    }
B
bellard 已提交
801

802
    qemu_opts_del(sn_opts);
803

P
Paolo Bonzini 已提交
804 805 806 807 808 809 810
    if (device) {
        void *ret;
        pthread_join(client_thread, &ret);
        exit(ret != NULL);
    } else {
        exit(EXIT_SUCCESS);
    }
B
bellard 已提交
811
}