qemu-nbd.c

   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-common.h"
  20 #include "block_int.h"
  21 #include "nbd.h"
  22
  23 #include <stdarg.h>
  24 #include <stdio.h>
  25 #include <getopt.h>
  26 #include <err.h>
  27 #include <sys/types.h>
  28 #include <sys/socket.h>
  29 #include <netinet/in.h>
  30 #include <netinet/tcp.h>
  31 #include <arpa/inet.h>
  32 #include <signal.h>
  33 #include <libgen.h>
  34 #include <pthread.h>
  35
  36 #define SOCKET_PATH    "/var/lock/qemu-nbd-%s"
  37
  38 static NBDExport *exp;
  39 static int verbose;
  40 static char *device;
  41 static char *srcpath;
  42 static char *sockpath;
  43 static bool sigterm_reported;
  44 static bool nbd_started;
  45 static int shared = 1;
  46 static int nb_fds;
  47
  48 static void usage(const char *name)
  49 {
  50     printf(
  51 "Usage: %s [OPTIONS] FILE\n"
  52 "QEMU Disk Network Block Device Server\n"
  53 "\n"
  54 "  -p, --port=PORT      port to listen on (default `%d')\n"
  55 "  -o, --offset=OFFSET  offset into the image\n"
  56 "  -b, --bind=IFACE     interface to bind to (default `0.0.0.0')\n"
  57 "  -k, --socket=PATH    path to the unix socket\n"
  58 "                       (default '"SOCKET_PATH"')\n"
  59 "  -r, --read-only      export read-only\n"
  60 "  -P, --partition=NUM  only expose partition NUM\n"
  61 "  -s, --snapshot       use snapshot file\n"
  62 "  -n, --nocache        disable host cache\n"
  63 "  -c, --connect=DEV    connect FILE to the local NBD device DEV\n"
  64 "  -d, --disconnect     disconnect the specified device\n"
  65 "  -e, --shared=NUM     device can be shared by NUM clients (default '1')\n"
  66 "  -t, --persistent     don't exit on the last connection\n"
  67 "  -v, --verbose        display extra debugging information\n"
  68 "  -h, --help           display this help and exit\n"
  69 "  -V, --version        output version information and exit\n"
  70 "\n"
  71 "Report bugs to <anthony@codemonkey.ws>\n"
  72     , name, NBD_DEFAULT_PORT, "DEVICE");
  73 }
  74
  75 static void version(const char *name)
  76 {
  77     printf(
  78 "%s version 0.0.1\n"
  79 "Written by Anthony Liguori.\n"
  80 "\n"
  81 "Copyright (C) 2006 Anthony Liguori <anthony@codemonkey.ws>.\n"
  82 "This is free software; see the source for copying conditions.  There is NO\n"
  83 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
  84     , name);
  85 }
  86
  87 struct partition_record
  88 {
  89     uint8_t bootable;
  90     uint8_t start_head;
  91     uint32_t start_cylinder;
  92     uint8_t start_sector;
  93     uint8_t system;
  94     uint8_t end_head;
  95     uint8_t end_cylinder;
  96     uint8_t end_sector;
  97     uint32_t start_sector_abs;
  98     uint32_t nb_sectors_abs;
  99 };
 100
 101 static void read_partition(uint8_t *p, struct partition_record *r)
 102 {
 103     r->bootable = p[0];
 104     r->start_head = p[1];
 105     r->start_cylinder = p[3] | ((p[2] << 2) & 0x0300);
 106     r->start_sector = p[2] & 0x3f;
 107     r->system = p[4];
 108     r->end_head = p[5];
 109     r->end_cylinder = p[7] | ((p[6] << 2) & 0x300);
 110     r->end_sector = p[6] & 0x3f;
 111     r->start_sector_abs = p[8] | p[9] << 8 | p[10] << 16 | p[11] << 24;
 112     r->nb_sectors_abs = p[12] | p[13] << 8 | p[14] << 16 | p[15] << 24;
 113 }
 114
 115 static int find_partition(BlockDriverState *bs, int partition,
 116                           off_t *offset, off_t *size)
 117 {
 118     struct partition_record mbr[4];
 119     uint8_t data[512];
 120     int i;
 121     int ext_partnum = 4;
 122     int ret;
 123
 124     if ((ret = bdrv_read(bs, 0, data, 1)) < 0) {
 125         errno = -ret;
 126         err(EXIT_FAILURE, "error while reading");
 127     }
 128
 129     if (data[510] != 0x55 || data[511] != 0xaa) {
 130         errno = -EINVAL;
 131         return -1;
 132     }
 133
 134     for (i = 0; i < 4; i++) {
 135         read_partition(&data[446 + 16 * i], &mbr[i]);
 136
 137         if (!mbr[i].nb_sectors_abs)
 138             continue;
 139
 140         if (mbr[i].system == 0xF || mbr[i].system == 0x5) {
 141             struct partition_record ext[4];
 142             uint8_t data1[512];
 143             int j;
 144
 145             if ((ret = bdrv_read(bs, mbr[i].start_sector_abs, data1, 1)) < 0) {
 146                 errno = -ret;
 147                 err(EXIT_FAILURE, "error while reading");
 148             }
 149
 150             for (j = 0; j < 4; j++) {
 151                 read_partition(&data1[446 + 16 * j], &ext[j]);
 152                 if (!ext[j].nb_sectors_abs)
 153                     continue;
 154
 155                 if ((ext_partnum + j + 1) == partition) {
 156                     *offset = (uint64_t)ext[j].start_sector_abs << 9;
 157                     *size = (uint64_t)ext[j].nb_sectors_abs << 9;
 158                     return 0;
 159                 }
 160             }
 161             ext_partnum += 4;
 162         } else if ((i + 1) == partition) {
 163             *offset = (uint64_t)mbr[i].start_sector_abs << 9;
 164             *size = (uint64_t)mbr[i].nb_sectors_abs << 9;
 165             return 0;
 166         }
 167     }
 168
 169     errno = -ENOENT;
 170     return -1;
 171 }
 172
 173 static void termsig_handler(int signum)
 174 {
 175     sigterm_reported = true;
 176     qemu_notify_event();
 177 }
 178
 179 static void *show_parts(void *arg)
 180 {
 181     int nbd;
 182
 183     /* linux just needs an open() to trigger
 184      * the partition table update
 185      * but remember to load the module with max_part != 0 :
 186      *     modprobe nbd max_part=63
 187      */
 188     nbd = open(device, O_RDWR);
 189     if (nbd != -1) {
 190         close(nbd);
 191     }
 192     return NULL;
 193 }
 194
 195 static void *nbd_client_thread(void *arg)
 196 {
 197     int fd = *(int *)arg;
 198     off_t size;
 199     size_t blocksize;
 200     uint32_t nbdflags;
 201     int sock;
 202     int ret;
 203     pthread_t show_parts_thread;
 204
 205     do {
 206         sock = unix_socket_outgoing(sockpath);
 207         if (sock == -1) {
 208             goto out;
 209         }
 210     } while (sock == -1);
 211
 212     ret = nbd_receive_negotiate(sock, NULL, &nbdflags,
 213                                 &size, &blocksize);
 214     if (ret == -1) {
 215         goto out;
 216     }
 217
 218     ret = nbd_init(fd, sock, nbdflags, size, blocksize);
 219     if (ret == -1) {
 220         goto out;
 221     }
 222
 223     /* update partition table */
 224     pthread_create(&show_parts_thread, NULL, show_parts, NULL);
 225
 226     if (verbose) {
 227         fprintf(stderr, "NBD device %s is now connected to %s\n",
 228                 device, srcpath);
 229     } else {
 230         /* Close stderr so that the qemu-nbd process exits.  */
 231         dup2(STDOUT_FILENO, STDERR_FILENO);
 232     }
 233
 234     ret = nbd_client(fd);
 235     if (ret) {
 236         goto out;
 237     }
 238     close(fd);
 239     kill(getpid(), SIGTERM);
 240     return (void *) EXIT_SUCCESS;
 241
 242 out:
 243     kill(getpid(), SIGTERM);
 244     return (void *) EXIT_FAILURE;
 245 }
 246
 247 static int nbd_can_accept(void *opaque)
 248 {
 249     return nb_fds < shared;
 250 }
 251
 252 static void nbd_read(void *opaque)
 253 {
 254     int fd = (uintptr_t) opaque;
 255
 256     if (nbd_trip(exp, fd) != 0) {
 257         qemu_set_fd_handler2(fd, NULL, NULL, NULL, NULL);
 258         close(fd);
 259         nb_fds--;
 260     }
 261 }
 262
 263 static void nbd_accept(void *opaque)
 264 {
 265     int server_fd = (uintptr_t) opaque;
 266     struct sockaddr_in addr;
 267     socklen_t addr_len = sizeof(addr);
 268
 269     int fd = accept(server_fd, (struct sockaddr *)&addr, &addr_len);
 270     nbd_started = true;
 271     if (fd != -1 && nbd_negotiate(exp, fd) != -1) {
 272         qemu_set_fd_handler2(fd, NULL, nbd_read, NULL, (void *) (intptr_t) fd);
 273         nb_fds++;
 274     }
 275 }
 276
 277 int main(int argc, char **argv)
 278 {
 279     BlockDriverState *bs;
 280     off_t dev_offset = 0;
 281     uint32_t nbdflags = 0;
 282     bool disconnect = false;
 283     const char *bindto = "0.0.0.0";
 284     int port = NBD_DEFAULT_PORT;
 285     off_t fd_size;
 286     const char *sopt = "hVb:o:p:rsnP:c:dvk:e:t";
 287     struct option lopt[] = {
 288         { "help", 0, NULL, 'h' },
 289         { "version", 0, NULL, 'V' },
 290         { "bind", 1, NULL, 'b' },
 291         { "port", 1, NULL, 'p' },
 292         { "socket", 1, NULL, 'k' },
 293         { "offset", 1, NULL, 'o' },
 294         { "read-only", 0, NULL, 'r' },
 295         { "partition", 1, NULL, 'P' },
 296         { "connect", 1, NULL, 'c' },
 297         { "disconnect", 0, NULL, 'd' },
 298         { "snapshot", 0, NULL, 's' },
 299         { "nocache", 0, NULL, 'n' },
 300         { "shared", 1, NULL, 'e' },
 301         { "persistent", 0, NULL, 't' },
 302         { "verbose", 0, NULL, 'v' },
 303         { NULL, 0, NULL, 0 }
 304     };
 305     int ch;
 306     int opt_ind = 0;
 307     int li;
 308     char *end;
 309     int flags = BDRV_O_RDWR;
 310     int partition = -1;
 311     int ret;
 312     int fd;
 313     int persistent = 0;
 314     pthread_t client_thread;
 315
 316     /* The client thread uses SIGTERM to interrupt the server.  A signal
 317      * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
 318      */
 319     struct sigaction sa_sigterm;
 320     memset(&sa_sigterm, 0, sizeof(sa_sigterm));
 321     sa_sigterm.sa_handler = termsig_handler;
 322     sigaction(SIGTERM, &sa_sigterm, NULL);
 323
 324     while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
 325         switch (ch) {
 326         case 's':
 327             flags |= BDRV_O_SNAPSHOT;
 328             break;
 329         case 'n':
 330             flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB;
 331             break;
 332         case 'b':
 333             bindto = optarg;
 334             break;
 335         case 'p':
 336             li = strtol(optarg, &end, 0);
 337             if (*end) {
 338                 errx(EXIT_FAILURE, "Invalid port `%s'", optarg);
 339             }
 340             if (li < 1 || li > 65535) {
 341                 errx(EXIT_FAILURE, "Port out of range `%s'", optarg);
 342             }
 343             port = (uint16_t)li;
 344             break;
 345         case 'o':
 346                 dev_offset = strtoll (optarg, &end, 0);
 347             if (*end) {
 348                 errx(EXIT_FAILURE, "Invalid offset `%s'", optarg);
 349             }
 350             if (dev_offset < 0) {
 351                 errx(EXIT_FAILURE, "Offset must be positive `%s'", optarg);
 352             }
 353             break;
 354         case 'r':
 355             nbdflags |= NBD_FLAG_READ_ONLY;
 356             flags &= ~BDRV_O_RDWR;
 357             break;
 358         case 'P':
 359             partition = strtol(optarg, &end, 0);
 360             if (*end)
 361                 errx(EXIT_FAILURE, "Invalid partition `%s'", optarg);
 362             if (partition < 1 || partition > 8)
 363                 errx(EXIT_FAILURE, "Invalid partition %d", partition);
 364             break;
 365         case 'k':
 366             sockpath = optarg;
 367             if (sockpath[0] != '/')
 368                 errx(EXIT_FAILURE, "socket path must be absolute\n");
 369             break;
 370         case 'd':
 371             disconnect = true;
 372             break;
 373         case 'c':
 374             device = optarg;
 375             break;
 376         case 'e':
 377             shared = strtol(optarg, &end, 0);
 378             if (*end) {
 379                 errx(EXIT_FAILURE, "Invalid shared device number '%s'", optarg);
 380             }
 381             if (shared < 1) {
 382                 errx(EXIT_FAILURE, "Shared device number must be greater than 0\n");
 383             }
 384             break;
 385         case 't':
 386             persistent = 1;
 387             break;
 388         case 'v':
 389             verbose = 1;
 390             break;
 391         case 'V':
 392             version(argv[0]);
 393             exit(0);
 394             break;
 395         case 'h':
 396             usage(argv[0]);
 397             exit(0);
 398             break;
 399         case '?':
 400             errx(EXIT_FAILURE, "Try `%s --help' for more information.",
 401                  argv[0]);
 402         }
 403     }
 404
 405     if ((argc - optind) != 1) {
 406         errx(EXIT_FAILURE, "Invalid number of argument.\n"
 407              "Try `%s --help' for more information.",
 408              argv[0]);
 409     }
 410
 411     if (disconnect) {
 412         fd = open(argv[optind], O_RDWR);
 413         if (fd == -1)
 414             err(EXIT_FAILURE, "Cannot open %s", argv[optind]);
 415
 416         nbd_disconnect(fd);
 417
 418         close(fd);
 419
 420         printf("%s disconnected\n", argv[optind]);
 421
 422         return 0;
 423     }
 424
 425     if (device && !verbose) {
 426         int stderr_fd[2];
 427         pid_t pid;
 428         int ret;
 429
 430         if (qemu_pipe(stderr_fd) == -1) {
 431             err(EXIT_FAILURE, "Error setting up communication pipe");
 432         }
 433
 434         /* Now daemonize, but keep a communication channel open to
 435          * print errors and exit with the proper status code.
 436          */
 437         pid = fork();
 438         if (pid == 0) {
 439             close(stderr_fd[0]);
 440             ret = qemu_daemon(0, 0);
 441
 442             /* Temporarily redirect stderr to the parent's pipe...  */
 443             dup2(stderr_fd[1], STDERR_FILENO);
 444             if (ret == -1) {
 445                 err(EXIT_FAILURE, "Failed to daemonize");
 446             }
 447
 448             /* ... close the descriptor we inherited and go on.  */
 449             close(stderr_fd[1]);
 450         } else {
 451             bool errors = false;
 452             char *buf;
 453
 454             /* In the parent.  Print error messages from the child until
 455              * it closes the pipe.
 456              */
 457             close(stderr_fd[1]);
 458             buf = g_malloc(1024);
 459             while ((ret = read(stderr_fd[0], buf, 1024)) > 0) {
 460                 errors = true;
 461                 ret = qemu_write_full(STDERR_FILENO, buf, ret);
 462                 if (ret == -1) {
 463                     exit(EXIT_FAILURE);
 464                 }
 465             }
 466             if (ret == -1) {
 467                 err(EXIT_FAILURE, "Cannot read from daemon");
 468             }
 469
 470             /* Usually the daemon should not print any message.
 471              * Exit with zero status in that case.
 472              */
 473             exit(errors);
 474         }
 475     }
 476
 477     if (device) {
 478         /* Open before spawning new threads.  In the future, we may
 479          * drop privileges after opening.
 480          */
 481         fd = open(device, O_RDWR);
 482         if (fd == -1) {
 483             err(EXIT_FAILURE, "Failed to open %s", device);
 484         }
 485
 486         if (sockpath == NULL) {
 487             sockpath = g_malloc(128);
 488             snprintf(sockpath, 128, SOCKET_PATH, basename(device));
 489         }
 490     }
 491
 492     bdrv_init();
 493     atexit(bdrv_close_all);
 494
 495     bs = bdrv_new("hda");
 496     srcpath = argv[optind];
 497     if ((ret = bdrv_open(bs, srcpath, flags, NULL)) < 0) {
 498         errno = -ret;
 499         err(EXIT_FAILURE, "Failed to bdrv_open '%s'", argv[optind]);
 500     }
 501
 502     fd_size = bs->total_sectors * 512;
 503
 504     if (partition != -1 &&
 505         find_partition(bs, partition, &dev_offset, &fd_size)) {
 506         err(EXIT_FAILURE, "Could not find partition %d", partition);
 507     }
 508
 509     exp = nbd_export_new(bs, dev_offset, fd_size, nbdflags);
 510
 511     if (sockpath) {
 512         fd = unix_socket_incoming(sockpath);
 513     } else {
 514         fd = tcp_socket_incoming(bindto, port);
 515     }
 516
 517     if (fd == -1) {
 518         return 1;
 519     }
 520
 521     if (device) {
 522         int ret;
 523
 524         ret = pthread_create(&client_thread, NULL, nbd_client_thread, &fd);
 525         if (ret != 0) {
 526             errx(EXIT_FAILURE, "Failed to create client thread: %s",
 527                  strerror(ret));
 528         }
 529     } else {
 530         /* Shut up GCC warnings.  */
 531         memset(&client_thread, 0, sizeof(client_thread));
 532     }
 533
 534     qemu_init_main_loop();
 535     qemu_set_fd_handler2(fd, nbd_can_accept, nbd_accept, NULL,
 536                          (void *)(uintptr_t)fd);
 537
 538     do {
 539         main_loop_wait(false);
 540     } while (!sigterm_reported && (persistent || !nbd_started || nb_fds > 0));
 541
 542     nbd_export_close(exp);
 543     if (sockpath) {
 544         unlink(sockpath);
 545     }
 546
 547     if (device) {
 548         void *ret;
 549         pthread_join(client_thread, &ret);
 550         exit(ret != NULL);
 551     } else {
 552         exit(EXIT_SUCCESS);
 553     }
 554 }