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 "block/nbd.h"
  20 #include "block/block.h"
  21 #include "block/block_int.h"
  22
  23 #include "block/coroutine.h"
  24
  25 #include <errno.h>
  26 #include <string.h>
  27 #ifndef _WIN32
  28 #include <sys/ioctl.h>
  29 #endif
  30 #if defined(__sun__) || defined(__HAIKU__)
  31 #include <sys/ioccom.h>
  32 #endif
  33 #include <ctype.h>
  34 #include <inttypes.h>
  35
  36 #ifdef __linux__
  37 #include <linux/fs.h>
  38 #endif
  39
  40 #include "qemu/sockets.h"
  41 #include "qemu/queue.h"
  42 #include "qemu/main-loop.h"
  43
  44 //#define DEBUG_NBD
  45
  46 #ifdef DEBUG_NBD
  47 #define TRACE(msg, ...) do { \
  48     LOG(msg, ## __VA_ARGS__); \
  49 } while(0)
  50 #else
  51 #define TRACE(msg, ...) \
  52     do { } while (0)
  53 #endif
  54
  55 #define LOG(msg, ...) do { \
  56     fprintf(stderr, "%s:%s():L%d: " msg "\n", \
  57             __FILE__, __FUNCTION__, __LINE__, ## __VA_ARGS__); \
  58 } while(0)
  59
  60 /* This is all part of the "official" NBD API.
  61  *
  62  * The most up-to-date documentation is available at:
  63  * https://github.com/yoe/nbd/blob/master/doc/proto.txt
  64  */
  65
  66 #define NBD_REQUEST_SIZE        (4 + 4 + 8 + 8 + 4)
  67 #define NBD_REPLY_SIZE          (4 + 4 + 8)
  68 #define NBD_REQUEST_MAGIC       0x25609513
  69 #define NBD_REPLY_MAGIC         0x67446698
  70 #define NBD_OPTS_MAGIC          0x49484156454F5054LL
  71 #define NBD_CLIENT_MAGIC        0x0000420281861253LL
  72 #define NBD_REP_MAGIC           0x3e889045565a9LL
  73
  74 #define NBD_SET_SOCK            _IO(0xab, 0)
  75 #define NBD_SET_BLKSIZE         _IO(0xab, 1)
  76 #define NBD_SET_SIZE            _IO(0xab, 2)
  77 #define NBD_DO_IT               _IO(0xab, 3)
  78 #define NBD_CLEAR_SOCK          _IO(0xab, 4)
  79 #define NBD_CLEAR_QUE           _IO(0xab, 5)
  80 #define NBD_PRINT_DEBUG         _IO(0xab, 6)
  81 #define NBD_SET_SIZE_BLOCKS     _IO(0xab, 7)
  82 #define NBD_DISCONNECT          _IO(0xab, 8)
  83 #define NBD_SET_TIMEOUT         _IO(0xab, 9)
  84 #define NBD_SET_FLAGS           _IO(0xab, 10)
  85
  86 #define NBD_OPT_EXPORT_NAME     (1)
  87 #define NBD_OPT_ABORT           (2)
  88 #define NBD_OPT_LIST            (3)
  89
  90 /* Definitions for opaque data types */
  91
  92 typedef struct NBDRequest NBDRequest;
  93
  94 struct NBDRequest {
  95     QSIMPLEQ_ENTRY(NBDRequest) entry;
  96     NBDClient *client;
  97     uint8_t *data;
  98 };
  99
 100 struct NBDExport {
 101     int refcount;
 102     void (*close)(NBDExport *exp);
 103
 104     BlockDriverState *bs;
 105     char *name;
 106     off_t dev_offset;
 107     off_t size;
 108     uint32_t nbdflags;
 109     QTAILQ_HEAD(, NBDClient) clients;
 110     QTAILQ_ENTRY(NBDExport) next;
 111
 112     AioContext *ctx;
 113 };
 114
 115 static QTAILQ_HEAD(, NBDExport) exports = QTAILQ_HEAD_INITIALIZER(exports);
 116
 117 struct NBDClient {
 118     int refcount;
 119     void (*close)(NBDClient *client);
 120
 121     NBDExport *exp;
 122     int sock;
 123
 124     Coroutine *recv_coroutine;
 125
 126     CoMutex send_lock;
 127     Coroutine *send_coroutine;
 128
 129     bool can_read;
 130
 131     QTAILQ_ENTRY(NBDClient) next;
 132     int nb_requests;
 133     bool closing;
 134 };
 135
 136 /* That's all folks */
 137
 138 static void nbd_set_handlers(NBDClient *client);
 139 static void nbd_unset_handlers(NBDClient *client);
 140 static void nbd_update_can_read(NBDClient *client);
 141
 142 ssize_t nbd_wr_sync(int fd, void *buffer, size_t size, bool do_read)
 143 {
 144     size_t offset = 0;
 145     int err;
 146
 147     if (qemu_in_coroutine()) {
 148         if (do_read) {
 149             return qemu_co_recv(fd, buffer, size);
 150         } else {
 151             return qemu_co_send(fd, buffer, size);
 152         }
 153     }
 154
 155     while (offset < size) {
 156         ssize_t len;
 157
 158         if (do_read) {
 159             len = qemu_recv(fd, buffer + offset, size - offset, 0);
 160         } else {
 161             len = send(fd, buffer + offset, size - offset, 0);
 162         }
 163
 164         if (len < 0) {
 165             err = socket_error();
 166
 167             /* recoverable error */
 168             if (err == EINTR || (offset > 0 && (err == EAGAIN || err == EWOULDBLOCK))) {
 169                 continue;
 170             }
 171
 172             /* unrecoverable error */
 173             return -err;
 174         }
 175
 176         /* eof */
 177         if (len == 0) {
 178             break;
 179         }
 180
 181         offset += len;
 182     }
 183
 184     return offset;
 185 }
 186
 187 static ssize_t read_sync(int fd, void *buffer, size_t size)
 188 {
 189     /* Sockets are kept in blocking mode in the negotiation phase.  After
 190      * that, a non-readable socket simply means that another thread stole
 191      * our request/reply.  Synchronization is done with recv_coroutine, so
 192      * that this is coroutine-safe.
 193      */
 194     return nbd_wr_sync(fd, buffer, size, true);
 195 }
 196
 197 static ssize_t write_sync(int fd, void *buffer, size_t size)
 198 {
 199     int ret;
 200     do {
 201         /* For writes, we do expect the socket to be writable.  */
 202         ret = nbd_wr_sync(fd, buffer, size, false);
 203     } while (ret == -EAGAIN);
 204     return ret;
 205 }
 206
 207 /* Basic flow for negotiation
 208
 209    Server         Client
 210    Negotiate
 211
 212    or
 213
 214    Server         Client
 215    Negotiate #1
 216                   Option
 217    Negotiate #2
 218
 219    ----
 220
 221    followed by
 222
 223    Server         Client
 224                   Request
 225    Response
 226                   Request
 227    Response
 228                   ...
 229    ...
 230                   Request (type == 2)
 231
 232 */
 233
 234 static int nbd_send_rep(int csock, uint32_t type, uint32_t opt)
 235 {
 236     uint64_t magic;
 237     uint32_t len;
 238
 239     magic = cpu_to_be64(NBD_REP_MAGIC);
 240     if (write_sync(csock, &magic, sizeof(magic)) != sizeof(magic)) {
 241         LOG("write failed (rep magic)");
 242         return -EINVAL;
 243     }
 244     opt = cpu_to_be32(opt);
 245     if (write_sync(csock, &opt, sizeof(opt)) != sizeof(opt)) {
 246         LOG("write failed (rep opt)");
 247         return -EINVAL;
 248     }
 249     type = cpu_to_be32(type);
 250     if (write_sync(csock, &type, sizeof(type)) != sizeof(type)) {
 251         LOG("write failed (rep type)");
 252         return -EINVAL;
 253     }
 254     len = cpu_to_be32(0);
 255     if (write_sync(csock, &len, sizeof(len)) != sizeof(len)) {
 256         LOG("write failed (rep data length)");
 257         return -EINVAL;
 258     }
 259     return 0;
 260 }
 261
 262 static int nbd_send_rep_list(int csock, NBDExport *exp)
 263 {
 264     uint64_t magic, name_len;
 265     uint32_t opt, type, len;
 266
 267     name_len = strlen(exp->name);
 268     magic = cpu_to_be64(NBD_REP_MAGIC);
 269     if (write_sync(csock, &magic, sizeof(magic)) != sizeof(magic)) {
 270         LOG("write failed (magic)");
 271         return -EINVAL;
 272      }
 273     opt = cpu_to_be32(NBD_OPT_LIST);
 274     if (write_sync(csock, &opt, sizeof(opt)) != sizeof(opt)) {
 275         LOG("write failed (opt)");
 276         return -EINVAL;
 277     }
 278     type = cpu_to_be32(NBD_REP_SERVER);
 279     if (write_sync(csock, &type, sizeof(type)) != sizeof(type)) {
 280         LOG("write failed (reply type)");
 281         return -EINVAL;
 282     }
 283     len = cpu_to_be32(name_len + sizeof(len));
 284     if (write_sync(csock, &len, sizeof(len)) != sizeof(len)) {
 285         LOG("write failed (length)");
 286         return -EINVAL;
 287     }
 288     len = cpu_to_be32(name_len);
 289     if (write_sync(csock, &len, sizeof(len)) != sizeof(len)) {
 290         LOG("write failed (length)");
 291         return -EINVAL;
 292     }
 293     if (write_sync(csock, exp->name, name_len) != name_len) {
 294         LOG("write failed (buffer)");
 295         return -EINVAL;
 296     }
 297     return 0;
 298 }
 299
 300 static int nbd_handle_list(NBDClient *client, uint32_t length)
 301 {
 302     int csock;
 303     NBDExport *exp;
 304
 305     csock = client->sock;
 306     if (length) {
 307         return nbd_send_rep(csock, NBD_REP_ERR_INVALID, NBD_OPT_LIST);
 308     }
 309
 310     /* For each export, send a NBD_REP_SERVER reply. */
 311     QTAILQ_FOREACH(exp, &exports, next) {
 312         if (nbd_send_rep_list(csock, exp)) {
 313             return -EINVAL;
 314         }
 315     }
 316     /* Finish with a NBD_REP_ACK. */
 317     return nbd_send_rep(csock, NBD_REP_ACK, NBD_OPT_LIST);
 318 }
 319
 320 static int nbd_handle_export_name(NBDClient *client, uint32_t length)
 321 {
 322     int rc = -EINVAL, csock = client->sock;
 323     char name[256];
 324
 325     /* Client sends:
 326         [20 ..  xx]   export name (length bytes)
 327      */
 328     TRACE("Checking length");
 329     if (length > 255) {
 330         LOG("Bad length received");
 331         goto fail;
 332     }
 333     if (read_sync(csock, name, length) != length) {
 334         LOG("read failed");
 335         goto fail;
 336     }
 337     name[length] = '\0';
 338
 339     client->exp = nbd_export_find(name);
 340     if (!client->exp) {
 341         LOG("export not found");
 342         goto fail;
 343     }
 344
 345     QTAILQ_INSERT_TAIL(&client->exp->clients, client, next);
 346     nbd_export_get(client->exp);
 347     rc = 0;
 348 fail:
 349     return rc;
 350 }
 351
 352 static int nbd_receive_options(NBDClient *client)
 353 {
 354     while (1) {
 355         int csock = client->sock;
 356         uint32_t tmp, length;
 357         uint64_t magic;
 358
 359         /* Client sends:
 360             [ 0 ..   3]   client flags
 361             [ 4 ..  11]   NBD_OPTS_MAGIC
 362             [12 ..  15]   NBD option
 363             [16 ..  19]   length
 364             ...           Rest of request
 365         */
 366
 367         if (read_sync(csock, &tmp, sizeof(tmp)) != sizeof(tmp)) {
 368             LOG("read failed");
 369             return -EINVAL;
 370         }
 371         TRACE("Checking client flags");
 372         tmp = be32_to_cpu(tmp);
 373         if (tmp != 0 && tmp != NBD_FLAG_C_FIXED_NEWSTYLE) {
 374             LOG("Bad client flags received");
 375             return -EINVAL;
 376         }
 377
 378         if (read_sync(csock, &magic, sizeof(magic)) != sizeof(magic)) {
 379             LOG("read failed");
 380             return -EINVAL;
 381         }
 382         TRACE("Checking opts magic");
 383         if (magic != be64_to_cpu(NBD_OPTS_MAGIC)) {
 384             LOG("Bad magic received");
 385             return -EINVAL;
 386         }
 387
 388         if (read_sync(csock, &tmp, sizeof(tmp)) != sizeof(tmp)) {
 389             LOG("read failed");
 390             return -EINVAL;
 391         }
 392
 393         if (read_sync(csock, &length, sizeof(length)) != sizeof(length)) {
 394             LOG("read failed");
 395             return -EINVAL;
 396         }
 397         length = be32_to_cpu(length);
 398
 399         TRACE("Checking option");
 400         switch (be32_to_cpu(tmp)) {
 401         case NBD_OPT_LIST:
 402             if (nbd_handle_list(client, length) < 0) {
 403                 return 1;
 404             }
 405             break;
 406
 407         case NBD_OPT_ABORT:
 408             return -EINVAL;
 409
 410         case NBD_OPT_EXPORT_NAME:
 411             return nbd_handle_export_name(client, length);
 412
 413         default:
 414             tmp = be32_to_cpu(tmp);
 415             LOG("Unsupported option 0x%x", tmp);
 416             nbd_send_rep(client->sock, NBD_REP_ERR_UNSUP, tmp);
 417             return -EINVAL;
 418         }
 419     }
 420 }
 421
 422 static int nbd_send_negotiate(NBDClient *client)
 423 {
 424     int csock = client->sock;
 425     char buf[8 + 8 + 8 + 128];
 426     int rc;
 427     const int myflags = (NBD_FLAG_HAS_FLAGS | NBD_FLAG_SEND_TRIM |
 428                          NBD_FLAG_SEND_FLUSH | NBD_FLAG_SEND_FUA);
 429
 430     /* Negotiation header without options:
 431         [ 0 ..   7]   passwd       ("NBDMAGIC")
 432         [ 8 ..  15]   magic        (NBD_CLIENT_MAGIC)
 433         [16 ..  23]   size
 434         [24 ..  25]   server flags (0)
 435         [26 ..  27]   export flags
 436         [28 .. 151]   reserved     (0)
 437
 438        Negotiation header with options, part 1:
 439         [ 0 ..   7]   passwd       ("NBDMAGIC")
 440         [ 8 ..  15]   magic        (NBD_OPTS_MAGIC)
 441         [16 ..  17]   server flags (0)
 442
 443        part 2 (after options are sent):
 444         [18 ..  25]   size
 445         [26 ..  27]   export flags
 446         [28 .. 151]   reserved     (0)
 447      */
 448
 449     qemu_set_block(csock);
 450     rc = -EINVAL;
 451
 452     TRACE("Beginning negotiation.");
 453     memset(buf, 0, sizeof(buf));
 454     memcpy(buf, "NBDMAGIC", 8);
 455     if (client->exp) {
 456         assert ((client->exp->nbdflags & ~65535) == 0);
 457         cpu_to_be64w((uint64_t*)(buf + 8), NBD_CLIENT_MAGIC);
 458         cpu_to_be64w((uint64_t*)(buf + 16), client->exp->size);
 459         cpu_to_be16w((uint16_t*)(buf + 26), client->exp->nbdflags | myflags);
 460     } else {
 461         cpu_to_be64w((uint64_t*)(buf + 8), NBD_OPTS_MAGIC);
 462         cpu_to_be16w((uint16_t *)(buf + 16), NBD_FLAG_FIXED_NEWSTYLE);
 463     }
 464
 465     if (client->exp) {
 466         if (write_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
 467             LOG("write failed");
 468             goto fail;
 469         }
 470     } else {
 471         if (write_sync(csock, buf, 18) != 18) {
 472             LOG("write failed");
 473             goto fail;
 474         }
 475         rc = nbd_receive_options(client);
 476         if (rc != 0) {
 477             LOG("option negotiation failed");
 478             goto fail;
 479         }
 480
 481         assert ((client->exp->nbdflags & ~65535) == 0);
 482         cpu_to_be64w((uint64_t*)(buf + 18), client->exp->size);
 483         cpu_to_be16w((uint16_t*)(buf + 26), client->exp->nbdflags | myflags);
 484         if (write_sync(csock, buf + 18, sizeof(buf) - 18) != sizeof(buf) - 18) {
 485             LOG("write failed");
 486             goto fail;
 487         }
 488     }
 489
 490     TRACE("Negotiation succeeded.");
 491     rc = 0;
 492 fail:
 493     qemu_set_nonblock(csock);
 494     return rc;
 495 }
 496
 497 int nbd_receive_negotiate(int csock, const char *name, uint32_t *flags,
 498                           off_t *size, size_t *blocksize)
 499 {
 500     char buf[256];
 501     uint64_t magic, s;
 502     uint16_t tmp;
 503     int rc;
 504
 505     TRACE("Receiving negotiation.");
 506
 507     rc = -EINVAL;
 508
 509     if (read_sync(csock, buf, 8) != 8) {
 510         LOG("read failed");
 511         goto fail;
 512     }
 513
 514     buf[8] = '\0';
 515     if (strlen(buf) == 0) {
 516         LOG("server connection closed");
 517         goto fail;
 518     }
 519
 520     TRACE("Magic is %c%c%c%c%c%c%c%c",
 521           qemu_isprint(buf[0]) ? buf[0] : '.',
 522           qemu_isprint(buf[1]) ? buf[1] : '.',
 523           qemu_isprint(buf[2]) ? buf[2] : '.',
 524           qemu_isprint(buf[3]) ? buf[3] : '.',
 525           qemu_isprint(buf[4]) ? buf[4] : '.',
 526           qemu_isprint(buf[5]) ? buf[5] : '.',
 527           qemu_isprint(buf[6]) ? buf[6] : '.',
 528           qemu_isprint(buf[7]) ? buf[7] : '.');
 529
 530     if (memcmp(buf, "NBDMAGIC", 8) != 0) {
 531         LOG("Invalid magic received");
 532         goto fail;
 533     }
 534
 535     if (read_sync(csock, &magic, sizeof(magic)) != sizeof(magic)) {
 536         LOG("read failed");
 537         goto fail;
 538     }
 539     magic = be64_to_cpu(magic);
 540     TRACE("Magic is 0x%" PRIx64, magic);
 541
 542     if (name) {
 543         uint32_t reserved = 0;
 544         uint32_t opt;
 545         uint32_t namesize;
 546
 547         TRACE("Checking magic (opts_magic)");
 548         if (magic != NBD_OPTS_MAGIC) {
 549             LOG("Bad magic received");
 550             goto fail;
 551         }
 552         if (read_sync(csock, &tmp, sizeof(tmp)) != sizeof(tmp)) {
 553             LOG("flags read failed");
 554             goto fail;
 555         }
 556         *flags = be16_to_cpu(tmp) << 16;
 557         /* reserved for future use */
 558         if (write_sync(csock, &reserved, sizeof(reserved)) !=
 559             sizeof(reserved)) {
 560             LOG("write failed (reserved)");
 561             goto fail;
 562         }
 563         /* write the export name */
 564         magic = cpu_to_be64(magic);
 565         if (write_sync(csock, &magic, sizeof(magic)) != sizeof(magic)) {
 566             LOG("write failed (magic)");
 567             goto fail;
 568         }
 569         opt = cpu_to_be32(NBD_OPT_EXPORT_NAME);
 570         if (write_sync(csock, &opt, sizeof(opt)) != sizeof(opt)) {
 571             LOG("write failed (opt)");
 572             goto fail;
 573         }
 574         namesize = cpu_to_be32(strlen(name));
 575         if (write_sync(csock, &namesize, sizeof(namesize)) !=
 576             sizeof(namesize)) {
 577             LOG("write failed (namesize)");
 578             goto fail;
 579         }
 580         if (write_sync(csock, (char*)name, strlen(name)) != strlen(name)) {
 581             LOG("write failed (name)");
 582             goto fail;
 583         }
 584     } else {
 585         TRACE("Checking magic (cli_magic)");
 586
 587         if (magic != NBD_CLIENT_MAGIC) {
 588             LOG("Bad magic received");
 589             goto fail;
 590         }
 591     }
 592
 593     if (read_sync(csock, &s, sizeof(s)) != sizeof(s)) {
 594         LOG("read failed");
 595         goto fail;
 596     }
 597     *size = be64_to_cpu(s);
 598     *blocksize = 1024;
 599     TRACE("Size is %" PRIu64, *size);
 600
 601     if (!name) {
 602         if (read_sync(csock, flags, sizeof(*flags)) != sizeof(*flags)) {
 603             LOG("read failed (flags)");
 604             goto fail;
 605         }
 606         *flags = be32_to_cpup(flags);
 607     } else {
 608         if (read_sync(csock, &tmp, sizeof(tmp)) != sizeof(tmp)) {
 609             LOG("read failed (tmp)");
 610             goto fail;
 611         }
 612         *flags |= be32_to_cpu(tmp);
 613     }
 614     if (read_sync(csock, &buf, 124) != 124) {
 615         LOG("read failed (buf)");
 616         goto fail;
 617     }
 618     rc = 0;
 619
 620 fail:
 621     return rc;
 622 }
 623
 624 #ifdef __linux__
 625 int nbd_init(int fd, int csock, uint32_t flags, off_t size, size_t blocksize)
 626 {
 627     TRACE("Setting NBD socket");
 628
 629     if (ioctl(fd, NBD_SET_SOCK, csock) < 0) {
 630         int serrno = errno;
 631         LOG("Failed to set NBD socket");
 632         return -serrno;
 633     }
 634
 635     TRACE("Setting block size to %lu", (unsigned long)blocksize);
 636
 637     if (ioctl(fd, NBD_SET_BLKSIZE, blocksize) < 0) {
 638         int serrno = errno;
 639         LOG("Failed setting NBD block size");
 640         return -serrno;
 641     }
 642
 643         TRACE("Setting size to %zd block(s)", (size_t)(size / blocksize));
 644
 645     if (ioctl(fd, NBD_SET_SIZE_BLOCKS, size / blocksize) < 0) {
 646         int serrno = errno;
 647         LOG("Failed setting size (in blocks)");
 648         return -serrno;
 649     }
 650
 651     if (ioctl(fd, NBD_SET_FLAGS, flags) < 0) {
 652         if (errno == ENOTTY) {
 653             int read_only = (flags & NBD_FLAG_READ_ONLY) != 0;
 654             TRACE("Setting readonly attribute");
 655
 656             if (ioctl(fd, BLKROSET, (unsigned long) &read_only) < 0) {
 657                 int serrno = errno;
 658                 LOG("Failed setting read-only attribute");
 659                 return -serrno;
 660             }
 661         } else {
 662             int serrno = errno;
 663             LOG("Failed setting flags");
 664             return -serrno;
 665         }
 666     }
 667
 668     TRACE("Negotiation ended");
 669
 670     return 0;
 671 }
 672
 673 int nbd_disconnect(int fd)
 674 {
 675     ioctl(fd, NBD_CLEAR_QUE);
 676     ioctl(fd, NBD_DISCONNECT);
 677     ioctl(fd, NBD_CLEAR_SOCK);
 678     return 0;
 679 }
 680
 681 int nbd_client(int fd)
 682 {
 683     int ret;
 684     int serrno;
 685
 686     TRACE("Doing NBD loop");
 687
 688     ret = ioctl(fd, NBD_DO_IT);
 689     if (ret < 0 && errno == EPIPE) {
 690         /* NBD_DO_IT normally returns EPIPE when someone has disconnected
 691          * the socket via NBD_DISCONNECT.  We do not want to return 1 in
 692          * that case.
 693          */
 694         ret = 0;
 695     }
 696     serrno = errno;
 697
 698     TRACE("NBD loop returned %d: %s", ret, strerror(serrno));
 699
 700     TRACE("Clearing NBD queue");
 701     ioctl(fd, NBD_CLEAR_QUE);
 702
 703     TRACE("Clearing NBD socket");
 704     ioctl(fd, NBD_CLEAR_SOCK);
 705
 706     errno = serrno;
 707     return ret;
 708 }
 709 #else
 710 int nbd_init(int fd, int csock, uint32_t flags, off_t size, size_t blocksize)
 711 {
 712     return -ENOTSUP;
 713 }
 714
 715 int nbd_disconnect(int fd)
 716 {
 717     return -ENOTSUP;
 718 }
 719
 720 int nbd_client(int fd)
 721 {
 722     return -ENOTSUP;
 723 }
 724 #endif
 725
 726 ssize_t nbd_send_request(int csock, struct nbd_request *request)
 727 {
 728     uint8_t buf[NBD_REQUEST_SIZE];
 729     ssize_t ret;
 730
 731     cpu_to_be32w((uint32_t*)buf, NBD_REQUEST_MAGIC);
 732     cpu_to_be32w((uint32_t*)(buf + 4), request->type);
 733     cpu_to_be64w((uint64_t*)(buf + 8), request->handle);
 734     cpu_to_be64w((uint64_t*)(buf + 16), request->from);
 735     cpu_to_be32w((uint32_t*)(buf + 24), request->len);
 736
 737     TRACE("Sending request to client: "
 738           "{ .from = %" PRIu64", .len = %u, .handle = %" PRIu64", .type=%i}",
 739           request->from, request->len, request->handle, request->type);
 740
 741     ret = write_sync(csock, buf, sizeof(buf));
 742     if (ret < 0) {
 743         return ret;
 744     }
 745
 746     if (ret != sizeof(buf)) {
 747         LOG("writing to socket failed");
 748         return -EINVAL;
 749     }
 750     return 0;
 751 }
 752
 753 static ssize_t nbd_receive_request(int csock, struct nbd_request *request)
 754 {
 755     uint8_t buf[NBD_REQUEST_SIZE];
 756     uint32_t magic;
 757     ssize_t ret;
 758
 759     ret = read_sync(csock, buf, sizeof(buf));
 760     if (ret < 0) {
 761         return ret;
 762     }
 763
 764     if (ret != sizeof(buf)) {
 765         LOG("read failed");
 766         return -EINVAL;
 767     }
 768
 769     /* Request
 770        [ 0 ..  3]   magic   (NBD_REQUEST_MAGIC)
 771        [ 4 ..  7]   type    (0 == READ, 1 == WRITE)
 772        [ 8 .. 15]   handle
 773        [16 .. 23]   from
 774        [24 .. 27]   len
 775      */
 776
 777     magic = be32_to_cpup((uint32_t*)buf);
 778     request->type  = be32_to_cpup((uint32_t*)(buf + 4));
 779     request->handle = be64_to_cpup((uint64_t*)(buf + 8));
 780     request->from  = be64_to_cpup((uint64_t*)(buf + 16));
 781     request->len   = be32_to_cpup((uint32_t*)(buf + 24));
 782
 783     TRACE("Got request: "
 784           "{ magic = 0x%x, .type = %d, from = %" PRIu64" , len = %u }",
 785           magic, request->type, request->from, request->len);
 786
 787     if (magic != NBD_REQUEST_MAGIC) {
 788         LOG("invalid magic (got 0x%x)", magic);
 789         return -EINVAL;
 790     }
 791     return 0;
 792 }
 793
 794 ssize_t nbd_receive_reply(int csock, struct nbd_reply *reply)
 795 {
 796     uint8_t buf[NBD_REPLY_SIZE];
 797     uint32_t magic;
 798     ssize_t ret;
 799
 800     ret = read_sync(csock, buf, sizeof(buf));
 801     if (ret < 0) {
 802         return ret;
 803     }
 804
 805     if (ret != sizeof(buf)) {
 806         LOG("read failed");
 807         return -EINVAL;
 808     }
 809
 810     /* Reply
 811        [ 0 ..  3]    magic   (NBD_REPLY_MAGIC)
 812        [ 4 ..  7]    error   (0 == no error)
 813        [ 7 .. 15]    handle
 814      */
 815
 816     magic = be32_to_cpup((uint32_t*)buf);
 817     reply->error  = be32_to_cpup((uint32_t*)(buf + 4));
 818     reply->handle = be64_to_cpup((uint64_t*)(buf + 8));
 819
 820     TRACE("Got reply: "
 821           "{ magic = 0x%x, .error = %d, handle = %" PRIu64" }",
 822           magic, reply->error, reply->handle);
 823
 824     if (magic != NBD_REPLY_MAGIC) {
 825         LOG("invalid magic (got 0x%x)", magic);
 826         return -EINVAL;
 827     }
 828     return 0;
 829 }
 830
 831 static ssize_t nbd_send_reply(int csock, struct nbd_reply *reply)
 832 {
 833     uint8_t buf[NBD_REPLY_SIZE];
 834     ssize_t ret;
 835
 836     /* Reply
 837        [ 0 ..  3]    magic   (NBD_REPLY_MAGIC)
 838        [ 4 ..  7]    error   (0 == no error)
 839        [ 7 .. 15]    handle
 840      */
 841     cpu_to_be32w((uint32_t*)buf, NBD_REPLY_MAGIC);
 842     cpu_to_be32w((uint32_t*)(buf + 4), reply->error);
 843     cpu_to_be64w((uint64_t*)(buf + 8), reply->handle);
 844
 845     TRACE("Sending response to client");
 846
 847     ret = write_sync(csock, buf, sizeof(buf));
 848     if (ret < 0) {
 849         return ret;
 850     }
 851
 852     if (ret != sizeof(buf)) {
 853         LOG("writing to socket failed");
 854         return -EINVAL;
 855     }
 856     return 0;
 857 }
 858
 859 #define MAX_NBD_REQUESTS 16
 860
 861 void nbd_client_get(NBDClient *client)
 862 {
 863     client->refcount++;
 864 }
 865
 866 void nbd_client_put(NBDClient *client)
 867 {
 868     if (--client->refcount == 0) {
 869         /* The last reference should be dropped by client->close,
 870          * which is called by nbd_client_close.
 871          */
 872         assert(client->closing);
 873
 874         nbd_unset_handlers(client);
 875         close(client->sock);
 876         client->sock = -1;
 877         if (client->exp) {
 878             QTAILQ_REMOVE(&client->exp->clients, client, next);
 879             nbd_export_put(client->exp);
 880         }
 881         g_free(client);
 882     }
 883 }
 884
 885 void nbd_client_close(NBDClient *client)
 886 {
 887     if (client->closing) {
 888         return;
 889     }
 890
 891     client->closing = true;
 892
 893     /* Force requests to finish.  They will drop their own references,
 894      * then we'll close the socket and free the NBDClient.
 895      */
 896     shutdown(client->sock, 2);
 897
 898     /* Also tell the client, so that they release their reference.  */
 899     if (client->close) {
 900         client->close(client);
 901     }
 902 }
 903
 904 static NBDRequest *nbd_request_get(NBDClient *client)
 905 {
 906     NBDRequest *req;
 907
 908     assert(client->nb_requests <= MAX_NBD_REQUESTS - 1);
 909     client->nb_requests++;
 910     nbd_update_can_read(client);
 911
 912     req = g_slice_new0(NBDRequest);
 913     nbd_client_get(client);
 914     req->client = client;
 915     return req;
 916 }
 917
 918 static void nbd_request_put(NBDRequest *req)
 919 {
 920     NBDClient *client = req->client;
 921
 922     if (req->data) {
 923         qemu_vfree(req->data);
 924     }
 925     g_slice_free(NBDRequest, req);
 926
 927     client->nb_requests--;
 928     nbd_update_can_read(client);
 929     nbd_client_put(client);
 930 }
 931
 932 static void bs_aio_attached(AioContext *ctx, void *opaque)
 933 {
 934     NBDExport *exp = opaque;
 935     NBDClient *client;
 936
 937     TRACE("Export %s: Attaching clients to AIO context %p\n", exp->name, ctx);
 938
 939     exp->ctx = ctx;
 940
 941     QTAILQ_FOREACH(client, &exp->clients, next) {
 942         nbd_set_handlers(client);
 943     }
 944 }
 945
 946 static void bs_aio_detach(void *opaque)
 947 {
 948     NBDExport *exp = opaque;
 949     NBDClient *client;
 950
 951     TRACE("Export %s: Detaching clients from AIO context %p\n", exp->name, exp->ctx);
 952
 953     QTAILQ_FOREACH(client, &exp->clients, next) {
 954         nbd_unset_handlers(client);
 955     }
 956
 957     exp->ctx = NULL;
 958 }
 959
 960 NBDExport *nbd_export_new(BlockDriverState *bs, off_t dev_offset,
 961                           off_t size, uint32_t nbdflags,
 962                           void (*close)(NBDExport *))
 963 {
 964     NBDExport *exp = g_malloc0(sizeof(NBDExport));
 965     exp->refcount = 1;
 966     QTAILQ_INIT(&exp->clients);
 967     exp->bs = bs;
 968     exp->dev_offset = dev_offset;
 969     exp->nbdflags = nbdflags;
 970     exp->size = size == -1 ? bdrv_getlength(bs) : size;
 971     exp->close = close;
 972     exp->ctx = bdrv_get_aio_context(bs);
 973     bdrv_ref(bs);
 974     bdrv_add_aio_context_notifier(bs, bs_aio_attached, bs_aio_detach, exp);
 975     /*
 976      * NBD exports are used for non-shared storage migration.  Make sure
 977      * that BDRV_O_INCOMING is cleared and the image is ready for write
 978      * access since the export could be available before migration handover.
 979      */
 980     bdrv_invalidate_cache(bs, NULL);
 981     return exp;
 982 }
 983
 984 NBDExport *nbd_export_find(const char *name)
 985 {
 986     NBDExport *exp;
 987     QTAILQ_FOREACH(exp, &exports, next) {
 988         if (strcmp(name, exp->name) == 0) {
 989             return exp;
 990         }
 991     }
 992
 993     return NULL;
 994 }
 995
 996 void nbd_export_set_name(NBDExport *exp, const char *name)
 997 {
 998     if (exp->name == name) {
 999         return;
1000     }
1001
1002     nbd_export_get(exp);
1003     if (exp->name != NULL) {
1004         g_free(exp->name);
1005         exp->name = NULL;
1006         QTAILQ_REMOVE(&exports, exp, next);
1007         nbd_export_put(exp);
1008     }
1009     if (name != NULL) {
1010         nbd_export_get(exp);
1011         exp->name = g_strdup(name);
1012         QTAILQ_INSERT_TAIL(&exports, exp, next);
1013     }
1014     nbd_export_put(exp);
1015 }
1016
1017 void nbd_export_close(NBDExport *exp)
1018 {
1019     NBDClient *client, *next;
1020
1021     nbd_export_get(exp);
1022     QTAILQ_FOREACH_SAFE(client, &exp->clients, next, next) {
1023         nbd_client_close(client);
1024     }
1025     nbd_export_set_name(exp, NULL);
1026     nbd_export_put(exp);
1027     if (exp->bs) {
1028         bdrv_remove_aio_context_notifier(exp->bs, bs_aio_attached,
1029                                          bs_aio_detach, exp);
1030         bdrv_unref(exp->bs);
1031         exp->bs = NULL;
1032     }
1033 }
1034
1035 void nbd_export_get(NBDExport *exp)
1036 {
1037     assert(exp->refcount > 0);
1038     exp->refcount++;
1039 }
1040
1041 void nbd_export_put(NBDExport *exp)
1042 {
1043     assert(exp->refcount > 0);
1044     if (exp->refcount == 1) {
1045         nbd_export_close(exp);
1046     }
1047
1048     if (--exp->refcount == 0) {
1049         assert(exp->name == NULL);
1050
1051         if (exp->close) {
1052             exp->close(exp);
1053         }
1054
1055         g_free(exp);
1056     }
1057 }
1058
1059 BlockDriverState *nbd_export_get_blockdev(NBDExport *exp)
1060 {
1061     return exp->bs;
1062 }
1063
1064 void nbd_export_close_all(void)
1065 {
1066     NBDExport *exp, *next;
1067
1068     QTAILQ_FOREACH_SAFE(exp, &exports, next, next) {
1069         nbd_export_close(exp);
1070     }
1071 }
1072
1073 static ssize_t nbd_co_send_reply(NBDRequest *req, struct nbd_reply *reply,
1074                                  int len)
1075 {
1076     NBDClient *client = req->client;
1077     int csock = client->sock;
1078     ssize_t rc, ret;
1079
1080     qemu_co_mutex_lock(&client->send_lock);
1081     client->send_coroutine = qemu_coroutine_self();
1082     nbd_set_handlers(client);
1083
1084     if (!len) {
1085         rc = nbd_send_reply(csock, reply);
1086     } else {
1087         socket_set_cork(csock, 1);
1088         rc = nbd_send_reply(csock, reply);
1089         if (rc >= 0) {
1090             ret = qemu_co_send(csock, req->data, len);
1091             if (ret != len) {
1092                 rc = -EIO;
1093             }
1094         }
1095         socket_set_cork(csock, 0);
1096     }
1097
1098     client->send_coroutine = NULL;
1099     nbd_set_handlers(client);
1100     qemu_co_mutex_unlock(&client->send_lock);
1101     return rc;
1102 }
1103
1104 static ssize_t nbd_co_receive_request(NBDRequest *req, struct nbd_request *request)
1105 {
1106     NBDClient *client = req->client;
1107     int csock = client->sock;
1108     uint32_t command;
1109     ssize_t rc;
1110
1111     client->recv_coroutine = qemu_coroutine_self();
1112     nbd_update_can_read(client);
1113
1114     rc = nbd_receive_request(csock, request);
1115     if (rc < 0) {
1116         if (rc != -EAGAIN) {
1117             rc = -EIO;
1118         }
1119         goto out;
1120     }
1121
1122     if (request->len > NBD_MAX_BUFFER_SIZE) {
1123         LOG("len (%u) is larger than max len (%u)",
1124             request->len, NBD_MAX_BUFFER_SIZE);
1125         rc = -EINVAL;
1126         goto out;
1127     }
1128
1129     if ((request->from + request->len) < request->from) {
1130         LOG("integer overflow detected! "
1131             "you're probably being attacked");
1132         rc = -EINVAL;
1133         goto out;
1134     }
1135
1136     TRACE("Decoding type");
1137
1138     command = request->type & NBD_CMD_MASK_COMMAND;
1139     if (command == NBD_CMD_READ || command == NBD_CMD_WRITE) {
1140         req->data = qemu_blockalign(client->exp->bs, request->len);
1141     }
1142     if (command == NBD_CMD_WRITE) {
1143         TRACE("Reading %u byte(s)", request->len);
1144
1145         if (qemu_co_recv(csock, req->data, request->len) != request->len) {
1146             LOG("reading from socket failed");
1147             rc = -EIO;
1148             goto out;
1149         }
1150     }
1151     rc = 0;
1152
1153 out:
1154     client->recv_coroutine = NULL;
1155     nbd_update_can_read(client);
1156
1157     return rc;
1158 }
1159
1160 static void nbd_trip(void *opaque)
1161 {
1162     NBDClient *client = opaque;
1163     NBDExport *exp = client->exp;
1164     NBDRequest *req;
1165     struct nbd_request request;
1166     struct nbd_reply reply;
1167     ssize_t ret;
1168     uint32_t command;
1169
1170     TRACE("Reading request.");
1171     if (client->closing) {
1172         return;
1173     }
1174
1175     req = nbd_request_get(client);
1176     ret = nbd_co_receive_request(req, &request);
1177     if (ret == -EAGAIN) {
1178         goto done;
1179     }
1180     if (ret == -EIO) {
1181         goto out;
1182     }
1183
1184     reply.handle = request.handle;
1185     reply.error = 0;
1186
1187     if (ret < 0) {
1188         reply.error = -ret;
1189         goto error_reply;
1190     }
1191     command = request.type & NBD_CMD_MASK_COMMAND;
1192     if (command != NBD_CMD_DISC && (request.from + request.len) > exp->size) {
1193             LOG("From: %" PRIu64 ", Len: %u, Size: %" PRIu64
1194             ", Offset: %" PRIu64 "\n",
1195                     request.from, request.len,
1196                     (uint64_t)exp->size, (uint64_t)exp->dev_offset);
1197         LOG("requested operation past EOF--bad client?");
1198         goto invalid_request;
1199     }
1200
1201     switch (command) {
1202     case NBD_CMD_READ:
1203         TRACE("Request type is READ");
1204
1205         if (request.type & NBD_CMD_FLAG_FUA) {
1206             ret = bdrv_co_flush(exp->bs);
1207             if (ret < 0) {
1208                 LOG("flush failed");
1209                 reply.error = -ret;
1210                 goto error_reply;
1211             }
1212         }
1213
1214         ret = bdrv_read(exp->bs, (request.from + exp->dev_offset) / 512,
1215                         req->data, request.len / 512);
1216         if (ret < 0) {
1217             LOG("reading from file failed");
1218             reply.error = -ret;
1219             goto error_reply;
1220         }
1221
1222         TRACE("Read %u byte(s)", request.len);
1223         if (nbd_co_send_reply(req, &reply, request.len) < 0)
1224             goto out;
1225         break;
1226     case NBD_CMD_WRITE:
1227         TRACE("Request type is WRITE");
1228
1229         if (exp->nbdflags & NBD_FLAG_READ_ONLY) {
1230             TRACE("Server is read-only, return error");
1231             reply.error = EROFS;
1232             goto error_reply;
1233         }
1234
1235         TRACE("Writing to device");
1236
1237         ret = bdrv_write(exp->bs, (request.from + exp->dev_offset) / 512,
1238                          req->data, request.len / 512);
1239         if (ret < 0) {
1240             LOG("writing to file failed");
1241             reply.error = -ret;
1242             goto error_reply;
1243         }
1244
1245         if (request.type & NBD_CMD_FLAG_FUA) {
1246             ret = bdrv_co_flush(exp->bs);
1247             if (ret < 0) {
1248                 LOG("flush failed");
1249                 reply.error = -ret;
1250                 goto error_reply;
1251             }
1252         }
1253
1254         if (nbd_co_send_reply(req, &reply, 0) < 0) {
1255             goto out;
1256         }
1257         break;
1258     case NBD_CMD_DISC:
1259         TRACE("Request type is DISCONNECT");
1260         errno = 0;
1261         goto out;
1262     case NBD_CMD_FLUSH:
1263         TRACE("Request type is FLUSH");
1264
1265         ret = bdrv_co_flush(exp->bs);
1266         if (ret < 0) {
1267             LOG("flush failed");
1268             reply.error = -ret;
1269         }
1270         if (nbd_co_send_reply(req, &reply, 0) < 0) {
1271             goto out;
1272         }
1273         break;
1274     case NBD_CMD_TRIM:
1275         TRACE("Request type is TRIM");
1276         ret = bdrv_co_discard(exp->bs, (request.from + exp->dev_offset) / 512,
1277                               request.len / 512);
1278         if (ret < 0) {
1279             LOG("discard failed");
1280             reply.error = -ret;
1281         }
1282         if (nbd_co_send_reply(req, &reply, 0) < 0) {
1283             goto out;
1284         }
1285         break;
1286     default:
1287         LOG("invalid request type (%u) received", request.type);
1288     invalid_request:
1289         reply.error = -EINVAL;
1290     error_reply:
1291         if (nbd_co_send_reply(req, &reply, 0) < 0) {
1292             goto out;
1293         }
1294         break;
1295     }
1296
1297     TRACE("Request/Reply complete");
1298
1299 done:
1300     nbd_request_put(req);
1301     return;
1302
1303 out:
1304     nbd_request_put(req);
1305     nbd_client_close(client);
1306 }
1307
1308 static void nbd_read(void *opaque)
1309 {
1310     NBDClient *client = opaque;
1311
1312     if (client->recv_coroutine) {
1313         qemu_coroutine_enter(client->recv_coroutine, NULL);
1314     } else {
1315         qemu_coroutine_enter(qemu_coroutine_create(nbd_trip), client);
1316     }
1317 }
1318
1319 static void nbd_restart_write(void *opaque)
1320 {
1321     NBDClient *client = opaque;
1322
1323     qemu_coroutine_enter(client->send_coroutine, NULL);
1324 }
1325
1326 static void nbd_set_handlers(NBDClient *client)
1327 {
1328     if (client->exp && client->exp->ctx) {
1329         aio_set_fd_handler(client->exp->ctx, client->sock,
1330                            client->can_read ? nbd_read : NULL,
1331                            client->send_coroutine ? nbd_restart_write : NULL,
1332                            client);
1333     }
1334 }
1335
1336 static void nbd_unset_handlers(NBDClient *client)
1337 {
1338     if (client->exp && client->exp->ctx) {
1339         aio_set_fd_handler(client->exp->ctx, client->sock, NULL, NULL, NULL);
1340     }
1341 }
1342
1343 static void nbd_update_can_read(NBDClient *client)
1344 {
1345     bool can_read = client->recv_coroutine ||
1346                     client->nb_requests < MAX_NBD_REQUESTS;
1347
1348     if (can_read != client->can_read) {
1349         client->can_read = can_read;
1350         nbd_set_handlers(client);
1351
1352         /* There is no need to invoke aio_notify(), since aio_set_fd_handler()
1353          * in nbd_set_handlers() will have taken care of that */
1354     }
1355 }
1356
1357 NBDClient *nbd_client_new(NBDExport *exp, int csock,
1358                           void (*close)(NBDClient *))
1359 {
1360     NBDClient *client;
1361     client = g_malloc0(sizeof(NBDClient));
1362     client->refcount = 1;
1363     client->exp = exp;
1364     client->sock = csock;
1365     client->can_read = true;
1366     if (nbd_send_negotiate(client)) {
1367         g_free(client);
1368         return NULL;
1369     }
1370     client->close = close;
1371     qemu_co_mutex_init(&client->send_lock);
1372     nbd_set_handlers(client);
1373
1374     if (exp) {
1375         QTAILQ_INSERT_TAIL(&exp->clients, client, next);
1376         nbd_export_get(exp);
1377     }
1378     return client;
1379 }