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 "nbd.h"
  20 #include "block.h"
  21
  22 #include "qemu-coroutine.h"
  23
  24 #include <errno.h>
  25 #include <string.h>
  26 #ifndef _WIN32
  27 #include <sys/ioctl.h>
  28 #endif
  29 #if defined(__sun__) || defined(__HAIKU__)
  30 #include <sys/ioccom.h>
  31 #endif
  32 #include <ctype.h>
  33 #include <inttypes.h>
  34
  35 #ifdef __linux__
  36 #include <linux/fs.h>
  37 #endif
  38
  39 #include "qemu_socket.h"
  40 #include "qemu-queue.h"
  41
  42 //#define DEBUG_NBD
  43
  44 #ifdef DEBUG_NBD
  45 #define TRACE(msg, ...) do { \
  46     LOG(msg, ## __VA_ARGS__); \
  47 } while(0)
  48 #else
  49 #define TRACE(msg, ...) \
  50     do { } while (0)
  51 #endif
  52
  53 #define LOG(msg, ...) do { \
  54     fprintf(stderr, "%s:%s():L%d: " msg "\n", \
  55             __FILE__, __FUNCTION__, __LINE__, ## __VA_ARGS__); \
  56 } while(0)
  57
  58 /* This is all part of the "official" NBD API */
  59
  60 #define NBD_REQUEST_SIZE        (4 + 4 + 8 + 8 + 4)
  61 #define NBD_REPLY_SIZE          (4 + 4 + 8)
  62 #define NBD_REQUEST_MAGIC       0x25609513
  63 #define NBD_REPLY_MAGIC         0x67446698
  64 #define NBD_OPTS_MAGIC          0x49484156454F5054LL
  65 #define NBD_CLIENT_MAGIC        0x0000420281861253LL
  66
  67 #define NBD_SET_SOCK            _IO(0xab, 0)
  68 #define NBD_SET_BLKSIZE         _IO(0xab, 1)
  69 #define NBD_SET_SIZE            _IO(0xab, 2)
  70 #define NBD_DO_IT               _IO(0xab, 3)
  71 #define NBD_CLEAR_SOCK          _IO(0xab, 4)
  72 #define NBD_CLEAR_QUE           _IO(0xab, 5)
  73 #define NBD_PRINT_DEBUG         _IO(0xab, 6)
  74 #define NBD_SET_SIZE_BLOCKS     _IO(0xab, 7)
  75 #define NBD_DISCONNECT          _IO(0xab, 8)
  76 #define NBD_SET_TIMEOUT         _IO(0xab, 9)
  77 #define NBD_SET_FLAGS           _IO(0xab, 10)
  78
  79 #define NBD_OPT_EXPORT_NAME     (1 << 0)
  80
  81 /* Definitions for opaque data types */
  82
  83 typedef struct NBDRequest NBDRequest;
  84
  85 struct NBDRequest {
  86     QSIMPLEQ_ENTRY(NBDRequest) entry;
  87     NBDClient *client;
  88     uint8_t *data;
  89 };
  90
  91 struct NBDExport {
  92     int refcount;
  93     void (*close)(NBDExport *exp);
  94
  95     BlockDriverState *bs;
  96     char *name;
  97     off_t dev_offset;
  98     off_t size;
  99     uint32_t nbdflags;
 100     QTAILQ_HEAD(, NBDClient) clients;
 101     QSIMPLEQ_HEAD(, NBDRequest) requests;
 102     QTAILQ_ENTRY(NBDExport) next;
 103 };
 104
 105 static QTAILQ_HEAD(, NBDExport) exports = QTAILQ_HEAD_INITIALIZER(exports);
 106
 107 struct NBDClient {
 108     int refcount;
 109     void (*close)(NBDClient *client);
 110
 111     NBDExport *exp;
 112     int sock;
 113
 114     Coroutine *recv_coroutine;
 115
 116     CoMutex send_lock;
 117     Coroutine *send_coroutine;
 118
 119     QTAILQ_ENTRY(NBDClient) next;
 120     int nb_requests;
 121     bool closing;
 122 };
 123
 124 /* That's all folks */
 125
 126 ssize_t nbd_wr_sync(int fd, void *buffer, size_t size, bool do_read)
 127 {
 128     size_t offset = 0;
 129     int err;
 130
 131     if (qemu_in_coroutine()) {
 132         if (do_read) {
 133             return qemu_co_recv(fd, buffer, size);
 134         } else {
 135             return qemu_co_send(fd, buffer, size);
 136         }
 137     }
 138
 139     while (offset < size) {
 140         ssize_t len;
 141
 142         if (do_read) {
 143             len = qemu_recv(fd, buffer + offset, size - offset, 0);
 144         } else {
 145             len = send(fd, buffer + offset, size - offset, 0);
 146         }
 147
 148         if (len < 0) {
 149             err = socket_error();
 150
 151             /* recoverable error */
 152             if (err == EINTR || (offset > 0 && err == EAGAIN)) {
 153                 continue;
 154             }
 155
 156             /* unrecoverable error */
 157             return -err;
 158         }
 159
 160         /* eof */
 161         if (len == 0) {
 162             break;
 163         }
 164
 165         offset += len;
 166     }
 167
 168     return offset;
 169 }
 170
 171 static ssize_t read_sync(int fd, void *buffer, size_t size)
 172 {
 173     /* Sockets are kept in blocking mode in the negotiation phase.  After
 174      * that, a non-readable socket simply means that another thread stole
 175      * our request/reply.  Synchronization is done with recv_coroutine, so
 176      * that this is coroutine-safe.
 177      */
 178     return nbd_wr_sync(fd, buffer, size, true);
 179 }
 180
 181 static ssize_t write_sync(int fd, void *buffer, size_t size)
 182 {
 183     int ret;
 184     do {
 185         /* For writes, we do expect the socket to be writable.  */
 186         ret = nbd_wr_sync(fd, buffer, size, false);
 187     } while (ret == -EAGAIN);
 188     return ret;
 189 }
 190
 191 static void combine_addr(char *buf, size_t len, const char* address,
 192                          uint16_t port)
 193 {
 194     /* If the address-part contains a colon, it's an IPv6 IP so needs [] */
 195     if (strstr(address, ":")) {
 196         snprintf(buf, len, "[%s]:%u", address, port);
 197     } else {
 198         snprintf(buf, len, "%s:%u", address, port);
 199     }
 200 }
 201
 202 int tcp_socket_outgoing(const char *address, uint16_t port)
 203 {
 204     char address_and_port[128];
 205     combine_addr(address_and_port, 128, address, port);
 206     return tcp_socket_outgoing_spec(address_and_port);
 207 }
 208
 209 int tcp_socket_outgoing_spec(const char *address_and_port)
 210 {
 211     return inet_connect(address_and_port, true, NULL, NULL);
 212 }
 213
 214 int tcp_socket_incoming(const char *address, uint16_t port)
 215 {
 216     char address_and_port[128];
 217     combine_addr(address_and_port, 128, address, port);
 218     return tcp_socket_incoming_spec(address_and_port);
 219 }
 220
 221 int tcp_socket_incoming_spec(const char *address_and_port)
 222 {
 223     char *ostr  = NULL;
 224     int olen = 0;
 225     return inet_listen(address_and_port, ostr, olen, SOCK_STREAM, 0, NULL);
 226 }
 227
 228 int unix_socket_incoming(const char *path)
 229 {
 230     char *ostr = NULL;
 231     int olen = 0;
 232
 233     return unix_listen(path, ostr, olen);
 234 }
 235
 236 int unix_socket_outgoing(const char *path)
 237 {
 238     return unix_connect(path);
 239 }
 240
 241 /* Basic flow for negotiation
 242
 243    Server         Client
 244    Negotiate
 245
 246    or
 247
 248    Server         Client
 249    Negotiate #1
 250                   Option
 251    Negotiate #2
 252
 253    ----
 254
 255    followed by
 256
 257    Server         Client
 258                   Request
 259    Response
 260                   Request
 261    Response
 262                   ...
 263    ...
 264                   Request (type == 2)
 265
 266 */
 267
 268 static int nbd_receive_options(NBDClient *client)
 269 {
 270     int csock = client->sock;
 271     char name[256];
 272     uint32_t tmp, length;
 273     uint64_t magic;
 274     int rc;
 275
 276     /* Client sends:
 277         [ 0 ..   3]   reserved (0)
 278         [ 4 ..  11]   NBD_OPTS_MAGIC
 279         [12 ..  15]   NBD_OPT_EXPORT_NAME
 280         [16 ..  19]   length
 281         [20 ..  xx]   export name (length bytes)
 282      */
 283
 284     rc = -EINVAL;
 285     if (read_sync(csock, &tmp, sizeof(tmp)) != sizeof(tmp)) {
 286         LOG("read failed");
 287         goto fail;
 288     }
 289     TRACE("Checking reserved");
 290     if (tmp != 0) {
 291         LOG("Bad reserved received");
 292         goto fail;
 293     }
 294
 295     if (read_sync(csock, &magic, sizeof(magic)) != sizeof(magic)) {
 296         LOG("read failed");
 297         goto fail;
 298     }
 299     TRACE("Checking reserved");
 300     if (magic != be64_to_cpu(NBD_OPTS_MAGIC)) {
 301         LOG("Bad magic received");
 302         goto fail;
 303     }
 304
 305     if (read_sync(csock, &tmp, sizeof(tmp)) != sizeof(tmp)) {
 306         LOG("read failed");
 307         goto fail;
 308     }
 309     TRACE("Checking option");
 310     if (tmp != be32_to_cpu(NBD_OPT_EXPORT_NAME)) {
 311         LOG("Bad option received");
 312         goto fail;
 313     }
 314
 315     if (read_sync(csock, &length, sizeof(length)) != sizeof(length)) {
 316         LOG("read failed");
 317         goto fail;
 318     }
 319     TRACE("Checking length");
 320     length = be32_to_cpu(length);
 321     if (length > 255) {
 322         LOG("Bad length received");
 323         goto fail;
 324     }
 325     if (read_sync(csock, name, length) != length) {
 326         LOG("read failed");
 327         goto fail;
 328     }
 329     name[length] = '\0';
 330
 331     client->exp = nbd_export_find(name);
 332     if (!client->exp) {
 333         LOG("export not found");
 334         goto fail;
 335     }
 336
 337     QTAILQ_INSERT_TAIL(&client->exp->clients, client, next);
 338     nbd_export_get(client->exp);
 339
 340     TRACE("Option negotiation succeeded.");
 341     rc = 0;
 342 fail:
 343     return rc;
 344 }
 345
 346 static int nbd_send_negotiate(NBDClient *client)
 347 {
 348     int csock = client->sock;
 349     char buf[8 + 8 + 8 + 128];
 350     int rc;
 351     const int myflags = (NBD_FLAG_HAS_FLAGS | NBD_FLAG_SEND_TRIM |
 352                          NBD_FLAG_SEND_FLUSH | NBD_FLAG_SEND_FUA);
 353
 354     /* Negotiation header without options:
 355         [ 0 ..   7]   passwd       ("NBDMAGIC")
 356         [ 8 ..  15]   magic        (NBD_CLIENT_MAGIC)
 357         [16 ..  23]   size
 358         [24 ..  25]   server flags (0)
 359         [24 ..  27]   export flags
 360         [28 .. 151]   reserved     (0)
 361
 362        Negotiation header with options, part 1:
 363         [ 0 ..   7]   passwd       ("NBDMAGIC")
 364         [ 8 ..  15]   magic        (NBD_OPTS_MAGIC)
 365         [16 ..  17]   server flags (0)
 366
 367        part 2 (after options are sent):
 368         [18 ..  25]   size
 369         [26 ..  27]   export flags
 370         [28 .. 151]   reserved     (0)
 371      */
 372
 373     socket_set_block(csock);
 374     rc = -EINVAL;
 375
 376     TRACE("Beginning negotiation.");
 377     memcpy(buf, "NBDMAGIC", 8);
 378     if (client->exp) {
 379         assert ((client->exp->nbdflags & ~65535) == 0);
 380         cpu_to_be64w((uint64_t*)(buf + 8), NBD_CLIENT_MAGIC);
 381         cpu_to_be64w((uint64_t*)(buf + 16), client->exp->size);
 382         cpu_to_be16w((uint16_t*)(buf + 26), client->exp->nbdflags | myflags);
 383     } else {
 384         cpu_to_be64w((uint64_t*)(buf + 8), NBD_OPTS_MAGIC);
 385     }
 386     memset(buf + 28, 0, 124);
 387
 388     if (client->exp) {
 389         if (write_sync(csock, buf, sizeof(buf)) != sizeof(buf)) {
 390             LOG("write failed");
 391             goto fail;
 392         }
 393     } else {
 394         if (write_sync(csock, buf, 18) != 18) {
 395             LOG("write failed");
 396             goto fail;
 397         }
 398         rc = nbd_receive_options(client);
 399         if (rc < 0) {
 400             LOG("option negotiation failed");
 401             goto fail;
 402         }
 403
 404         assert ((client->exp->nbdflags & ~65535) == 0);
 405         cpu_to_be64w((uint64_t*)(buf + 18), client->exp->size);
 406         cpu_to_be16w((uint16_t*)(buf + 26), client->exp->nbdflags | myflags);
 407         if (write_sync(csock, buf + 18, sizeof(buf) - 18) != sizeof(buf) - 18) {
 408             LOG("write failed");
 409             goto fail;
 410         }
 411     }
 412
 413     TRACE("Negotiation succeeded.");
 414     rc = 0;
 415 fail:
 416     socket_set_nonblock(csock);
 417     return rc;
 418 }
 419
 420 int nbd_receive_negotiate(int csock, const char *name, uint32_t *flags,
 421                           off_t *size, size_t *blocksize)
 422 {
 423     char buf[256];
 424     uint64_t magic, s;
 425     uint16_t tmp;
 426     int rc;
 427
 428     TRACE("Receiving negotiation.");
 429
 430     socket_set_block(csock);
 431     rc = -EINVAL;
 432
 433     if (read_sync(csock, buf, 8) != 8) {
 434         LOG("read failed");
 435         goto fail;
 436     }
 437
 438     buf[8] = '\0';
 439     if (strlen(buf) == 0) {
 440         LOG("server connection closed");
 441         goto fail;
 442     }
 443
 444     TRACE("Magic is %c%c%c%c%c%c%c%c",
 445           qemu_isprint(buf[0]) ? buf[0] : '.',
 446           qemu_isprint(buf[1]) ? buf[1] : '.',
 447           qemu_isprint(buf[2]) ? buf[2] : '.',
 448           qemu_isprint(buf[3]) ? buf[3] : '.',
 449           qemu_isprint(buf[4]) ? buf[4] : '.',
 450           qemu_isprint(buf[5]) ? buf[5] : '.',
 451           qemu_isprint(buf[6]) ? buf[6] : '.',
 452           qemu_isprint(buf[7]) ? buf[7] : '.');
 453
 454     if (memcmp(buf, "NBDMAGIC", 8) != 0) {
 455         LOG("Invalid magic received");
 456         goto fail;
 457     }
 458
 459     if (read_sync(csock, &magic, sizeof(magic)) != sizeof(magic)) {
 460         LOG("read failed");
 461         goto fail;
 462     }
 463     magic = be64_to_cpu(magic);
 464     TRACE("Magic is 0x%" PRIx64, magic);
 465
 466     if (name) {
 467         uint32_t reserved = 0;
 468         uint32_t opt;
 469         uint32_t namesize;
 470
 471         TRACE("Checking magic (opts_magic)");
 472         if (magic != NBD_OPTS_MAGIC) {
 473             LOG("Bad magic received");
 474             goto fail;
 475         }
 476         if (read_sync(csock, &tmp, sizeof(tmp)) != sizeof(tmp)) {
 477             LOG("flags read failed");
 478             goto fail;
 479         }
 480         *flags = be16_to_cpu(tmp) << 16;
 481         /* reserved for future use */
 482         if (write_sync(csock, &reserved, sizeof(reserved)) !=
 483             sizeof(reserved)) {
 484             LOG("write failed (reserved)");
 485             goto fail;
 486         }
 487         /* write the export name */
 488         magic = cpu_to_be64(magic);
 489         if (write_sync(csock, &magic, sizeof(magic)) != sizeof(magic)) {
 490             LOG("write failed (magic)");
 491             goto fail;
 492         }
 493         opt = cpu_to_be32(NBD_OPT_EXPORT_NAME);
 494         if (write_sync(csock, &opt, sizeof(opt)) != sizeof(opt)) {
 495             LOG("write failed (opt)");
 496             goto fail;
 497         }
 498         namesize = cpu_to_be32(strlen(name));
 499         if (write_sync(csock, &namesize, sizeof(namesize)) !=
 500             sizeof(namesize)) {
 501             LOG("write failed (namesize)");
 502             goto fail;
 503         }
 504         if (write_sync(csock, (char*)name, strlen(name)) != strlen(name)) {
 505             LOG("write failed (name)");
 506             goto fail;
 507         }
 508     } else {
 509         TRACE("Checking magic (cli_magic)");
 510
 511         if (magic != NBD_CLIENT_MAGIC) {
 512             LOG("Bad magic received");
 513             goto fail;
 514         }
 515     }
 516
 517     if (read_sync(csock, &s, sizeof(s)) != sizeof(s)) {
 518         LOG("read failed");
 519         goto fail;
 520     }
 521     *size = be64_to_cpu(s);
 522     *blocksize = 1024;
 523     TRACE("Size is %" PRIu64, *size);
 524
 525     if (!name) {
 526         if (read_sync(csock, flags, sizeof(*flags)) != sizeof(*flags)) {
 527             LOG("read failed (flags)");
 528             goto fail;
 529         }
 530         *flags = be32_to_cpup(flags);
 531     } else {
 532         if (read_sync(csock, &tmp, sizeof(tmp)) != sizeof(tmp)) {
 533             LOG("read failed (tmp)");
 534             goto fail;
 535         }
 536         *flags |= be32_to_cpu(tmp);
 537     }
 538     if (read_sync(csock, &buf, 124) != 124) {
 539         LOG("read failed (buf)");
 540         goto fail;
 541     }
 542     rc = 0;
 543
 544 fail:
 545     socket_set_nonblock(csock);
 546     return rc;
 547 }
 548
 549 #ifdef __linux__
 550 int nbd_init(int fd, int csock, uint32_t flags, off_t size, size_t blocksize)
 551 {
 552     TRACE("Setting NBD socket");
 553
 554     if (ioctl(fd, NBD_SET_SOCK, csock) < 0) {
 555         int serrno = errno;
 556         LOG("Failed to set NBD socket");
 557         return -serrno;
 558     }
 559
 560     TRACE("Setting block size to %lu", (unsigned long)blocksize);
 561
 562     if (ioctl(fd, NBD_SET_BLKSIZE, blocksize) < 0) {
 563         int serrno = errno;
 564         LOG("Failed setting NBD block size");
 565         return -serrno;
 566     }
 567
 568         TRACE("Setting size to %zd block(s)", (size_t)(size / blocksize));
 569
 570     if (ioctl(fd, NBD_SET_SIZE_BLOCKS, size / blocksize) < 0) {
 571         int serrno = errno;
 572         LOG("Failed setting size (in blocks)");
 573         return -serrno;
 574     }
 575
 576     if (flags & NBD_FLAG_READ_ONLY) {
 577         int read_only = 1;
 578         TRACE("Setting readonly attribute");
 579
 580         if (ioctl(fd, BLKROSET, (unsigned long) &read_only) < 0) {
 581             int serrno = errno;
 582             LOG("Failed setting read-only attribute");
 583             return -serrno;
 584         }
 585     }
 586
 587     if (ioctl(fd, NBD_SET_FLAGS, flags) < 0
 588         && errno != ENOTTY) {
 589         int serrno = errno;
 590         LOG("Failed setting flags");
 591         return -serrno;
 592     }
 593
 594     TRACE("Negotiation ended");
 595
 596     return 0;
 597 }
 598
 599 int nbd_disconnect(int fd)
 600 {
 601     ioctl(fd, NBD_CLEAR_QUE);
 602     ioctl(fd, NBD_DISCONNECT);
 603     ioctl(fd, NBD_CLEAR_SOCK);
 604     return 0;
 605 }
 606
 607 int nbd_client(int fd)
 608 {
 609     int ret;
 610     int serrno;
 611
 612     TRACE("Doing NBD loop");
 613
 614     ret = ioctl(fd, NBD_DO_IT);
 615     if (ret < 0 && errno == EPIPE) {
 616         /* NBD_DO_IT normally returns EPIPE when someone has disconnected
 617          * the socket via NBD_DISCONNECT.  We do not want to return 1 in
 618          * that case.
 619          */
 620         ret = 0;
 621     }
 622     serrno = errno;
 623
 624     TRACE("NBD loop returned %d: %s", ret, strerror(serrno));
 625
 626     TRACE("Clearing NBD queue");
 627     ioctl(fd, NBD_CLEAR_QUE);
 628
 629     TRACE("Clearing NBD socket");
 630     ioctl(fd, NBD_CLEAR_SOCK);
 631
 632     errno = serrno;
 633     return ret;
 634 }
 635 #else
 636 int nbd_init(int fd, int csock, uint32_t flags, off_t size, size_t blocksize)
 637 {
 638     return -ENOTSUP;
 639 }
 640
 641 int nbd_disconnect(int fd)
 642 {
 643     return -ENOTSUP;
 644 }
 645
 646 int nbd_client(int fd)
 647 {
 648     return -ENOTSUP;
 649 }
 650 #endif
 651
 652 ssize_t nbd_send_request(int csock, struct nbd_request *request)
 653 {
 654     uint8_t buf[NBD_REQUEST_SIZE];
 655     ssize_t ret;
 656
 657     cpu_to_be32w((uint32_t*)buf, NBD_REQUEST_MAGIC);
 658     cpu_to_be32w((uint32_t*)(buf + 4), request->type);
 659     cpu_to_be64w((uint64_t*)(buf + 8), request->handle);
 660     cpu_to_be64w((uint64_t*)(buf + 16), request->from);
 661     cpu_to_be32w((uint32_t*)(buf + 24), request->len);
 662
 663     TRACE("Sending request to client: "
 664           "{ .from = %" PRIu64", .len = %u, .handle = %" PRIu64", .type=%i}",
 665           request->from, request->len, request->handle, request->type);
 666
 667     ret = write_sync(csock, buf, sizeof(buf));
 668     if (ret < 0) {
 669         return ret;
 670     }
 671
 672     if (ret != sizeof(buf)) {
 673         LOG("writing to socket failed");
 674         return -EINVAL;
 675     }
 676     return 0;
 677 }
 678
 679 static ssize_t nbd_receive_request(int csock, struct nbd_request *request)
 680 {
 681     uint8_t buf[NBD_REQUEST_SIZE];
 682     uint32_t magic;
 683     ssize_t ret;
 684
 685     ret = read_sync(csock, buf, sizeof(buf));
 686     if (ret < 0) {
 687         return ret;
 688     }
 689
 690     if (ret != sizeof(buf)) {
 691         LOG("read failed");
 692         return -EINVAL;
 693     }
 694
 695     /* Request
 696        [ 0 ..  3]   magic   (NBD_REQUEST_MAGIC)
 697        [ 4 ..  7]   type    (0 == READ, 1 == WRITE)
 698        [ 8 .. 15]   handle
 699        [16 .. 23]   from
 700        [24 .. 27]   len
 701      */
 702
 703     magic = be32_to_cpup((uint32_t*)buf);
 704     request->type  = be32_to_cpup((uint32_t*)(buf + 4));
 705     request->handle = be64_to_cpup((uint64_t*)(buf + 8));
 706     request->from  = be64_to_cpup((uint64_t*)(buf + 16));
 707     request->len   = be32_to_cpup((uint32_t*)(buf + 24));
 708
 709     TRACE("Got request: "
 710           "{ magic = 0x%x, .type = %d, from = %" PRIu64" , len = %u }",
 711           magic, request->type, request->from, request->len);
 712
 713     if (magic != NBD_REQUEST_MAGIC) {
 714         LOG("invalid magic (got 0x%x)", magic);
 715         return -EINVAL;
 716     }
 717     return 0;
 718 }
 719
 720 ssize_t nbd_receive_reply(int csock, struct nbd_reply *reply)
 721 {
 722     uint8_t buf[NBD_REPLY_SIZE];
 723     uint32_t magic;
 724     ssize_t ret;
 725
 726     ret = read_sync(csock, buf, sizeof(buf));
 727     if (ret < 0) {
 728         return ret;
 729     }
 730
 731     if (ret != sizeof(buf)) {
 732         LOG("read failed");
 733         return -EINVAL;
 734     }
 735
 736     /* Reply
 737        [ 0 ..  3]    magic   (NBD_REPLY_MAGIC)
 738        [ 4 ..  7]    error   (0 == no error)
 739        [ 7 .. 15]    handle
 740      */
 741
 742     magic = be32_to_cpup((uint32_t*)buf);
 743     reply->error  = be32_to_cpup((uint32_t*)(buf + 4));
 744     reply->handle = be64_to_cpup((uint64_t*)(buf + 8));
 745
 746     TRACE("Got reply: "
 747           "{ magic = 0x%x, .error = %d, handle = %" PRIu64" }",
 748           magic, reply->error, reply->handle);
 749
 750     if (magic != NBD_REPLY_MAGIC) {
 751         LOG("invalid magic (got 0x%x)", magic);
 752         return -EINVAL;
 753     }
 754     return 0;
 755 }
 756
 757 static ssize_t nbd_send_reply(int csock, struct nbd_reply *reply)
 758 {
 759     uint8_t buf[NBD_REPLY_SIZE];
 760     ssize_t ret;
 761
 762     /* Reply
 763        [ 0 ..  3]    magic   (NBD_REPLY_MAGIC)
 764        [ 4 ..  7]    error   (0 == no error)
 765        [ 7 .. 15]    handle
 766      */
 767     cpu_to_be32w((uint32_t*)buf, NBD_REPLY_MAGIC);
 768     cpu_to_be32w((uint32_t*)(buf + 4), reply->error);
 769     cpu_to_be64w((uint64_t*)(buf + 8), reply->handle);
 770
 771     TRACE("Sending response to client");
 772
 773     ret = write_sync(csock, buf, sizeof(buf));
 774     if (ret < 0) {
 775         return ret;
 776     }
 777
 778     if (ret != sizeof(buf)) {
 779         LOG("writing to socket failed");
 780         return -EINVAL;
 781     }
 782     return 0;
 783 }
 784
 785 #define MAX_NBD_REQUESTS 16
 786
 787 void nbd_client_get(NBDClient *client)
 788 {
 789     client->refcount++;
 790 }
 791
 792 void nbd_client_put(NBDClient *client)
 793 {
 794     if (--client->refcount == 0) {
 795         /* The last reference should be dropped by client->close,
 796          * which is called by nbd_client_close.
 797          */
 798         assert(client->closing);
 799
 800         qemu_set_fd_handler2(client->sock, NULL, NULL, NULL, NULL);
 801         close(client->sock);
 802         client->sock = -1;
 803         if (client->exp) {
 804             QTAILQ_REMOVE(&client->exp->clients, client, next);
 805             nbd_export_put(client->exp);
 806         }
 807         g_free(client);
 808     }
 809 }
 810
 811 void nbd_client_close(NBDClient *client)
 812 {
 813     if (client->closing) {
 814         return;
 815     }
 816
 817     client->closing = true;
 818
 819     /* Force requests to finish.  They will drop their own references,
 820      * then we'll close the socket and free the NBDClient.
 821      */
 822     shutdown(client->sock, 2);
 823
 824     /* Also tell the client, so that they release their reference.  */
 825     if (client->close) {
 826         client->close(client);
 827     }
 828 }
 829
 830 static NBDRequest *nbd_request_get(NBDClient *client)
 831 {
 832     NBDRequest *req;
 833     NBDExport *exp = client->exp;
 834
 835     assert(client->nb_requests <= MAX_NBD_REQUESTS - 1);
 836     client->nb_requests++;
 837
 838     if (QSIMPLEQ_EMPTY(&exp->requests)) {
 839         req = g_malloc0(sizeof(NBDRequest));
 840         req->data = qemu_blockalign(exp->bs, NBD_BUFFER_SIZE);
 841     } else {
 842         req = QSIMPLEQ_FIRST(&exp->requests);
 843         QSIMPLEQ_REMOVE_HEAD(&exp->requests, entry);
 844     }
 845     nbd_client_get(client);
 846     req->client = client;
 847     return req;
 848 }
 849
 850 static void nbd_request_put(NBDRequest *req)
 851 {
 852     NBDClient *client = req->client;
 853     QSIMPLEQ_INSERT_HEAD(&client->exp->requests, req, entry);
 854     if (client->nb_requests-- == MAX_NBD_REQUESTS) {
 855         qemu_notify_event();
 856     }
 857     nbd_client_put(client);
 858 }
 859
 860 NBDExport *nbd_export_new(BlockDriverState *bs, off_t dev_offset,
 861                           off_t size, uint32_t nbdflags,
 862                           void (*close)(NBDExport *))
 863 {
 864     NBDExport *exp = g_malloc0(sizeof(NBDExport));
 865     QSIMPLEQ_INIT(&exp->requests);
 866     exp->refcount = 1;
 867     QTAILQ_INIT(&exp->clients);
 868     exp->bs = bs;
 869     exp->dev_offset = dev_offset;
 870     exp->nbdflags = nbdflags;
 871     exp->size = size == -1 ? bdrv_getlength(bs) : size;
 872     exp->close = close;
 873     return exp;
 874 }
 875
 876 NBDExport *nbd_export_find(const char *name)
 877 {
 878     NBDExport *exp;
 879     QTAILQ_FOREACH(exp, &exports, next) {
 880         if (strcmp(name, exp->name) == 0) {
 881             return exp;
 882         }
 883     }
 884
 885     return NULL;
 886 }
 887
 888 void nbd_export_set_name(NBDExport *exp, const char *name)
 889 {
 890     if (exp->name == name) {
 891         return;
 892     }
 893
 894     nbd_export_get(exp);
 895     if (exp->name != NULL) {
 896         g_free(exp->name);
 897         exp->name = NULL;
 898         QTAILQ_REMOVE(&exports, exp, next);
 899         nbd_export_put(exp);
 900     }
 901     if (name != NULL) {
 902         nbd_export_get(exp);
 903         exp->name = g_strdup(name);
 904         QTAILQ_INSERT_TAIL(&exports, exp, next);
 905     }
 906     nbd_export_put(exp);
 907 }
 908
 909 void nbd_export_close(NBDExport *exp)
 910 {
 911     NBDClient *client, *next;
 912
 913     nbd_export_get(exp);
 914     QTAILQ_FOREACH_SAFE(client, &exp->clients, next, next) {
 915         nbd_client_close(client);
 916     }
 917     nbd_export_set_name(exp, NULL);
 918     nbd_export_put(exp);
 919 }
 920
 921 void nbd_export_get(NBDExport *exp)
 922 {
 923     assert(exp->refcount > 0);
 924     exp->refcount++;
 925 }
 926
 927 void nbd_export_put(NBDExport *exp)
 928 {
 929     assert(exp->refcount > 0);
 930     if (exp->refcount == 1) {
 931         nbd_export_close(exp);
 932     }
 933
 934     if (--exp->refcount == 0) {
 935         assert(exp->name == NULL);
 936
 937         if (exp->close) {
 938             exp->close(exp);
 939         }
 940
 941         while (!QSIMPLEQ_EMPTY(&exp->requests)) {
 942             NBDRequest *first = QSIMPLEQ_FIRST(&exp->requests);
 943             QSIMPLEQ_REMOVE_HEAD(&exp->requests, entry);
 944             qemu_vfree(first->data);
 945             g_free(first);
 946         }
 947
 948         g_free(exp);
 949     }
 950 }
 951
 952 BlockDriverState *nbd_export_get_blockdev(NBDExport *exp)
 953 {
 954     return exp->bs;
 955 }
 956
 957 void nbd_export_close_all(void)
 958 {
 959     NBDExport *exp, *next;
 960
 961     QTAILQ_FOREACH_SAFE(exp, &exports, next, next) {
 962         nbd_export_close(exp);
 963     }
 964 }
 965
 966 static int nbd_can_read(void *opaque);
 967 static void nbd_read(void *opaque);
 968 static void nbd_restart_write(void *opaque);
 969
 970 static ssize_t nbd_co_send_reply(NBDRequest *req, struct nbd_reply *reply,
 971                                  int len)
 972 {
 973     NBDClient *client = req->client;
 974     int csock = client->sock;
 975     ssize_t rc, ret;
 976
 977     qemu_co_mutex_lock(&client->send_lock);
 978     qemu_set_fd_handler2(csock, nbd_can_read, nbd_read,
 979                          nbd_restart_write, client);
 980     client->send_coroutine = qemu_coroutine_self();
 981
 982     if (!len) {
 983         rc = nbd_send_reply(csock, reply);
 984     } else {
 985         socket_set_cork(csock, 1);
 986         rc = nbd_send_reply(csock, reply);
 987         if (rc >= 0) {
 988             ret = qemu_co_send(csock, req->data, len);
 989             if (ret != len) {
 990                 rc = -EIO;
 991             }
 992         }
 993         socket_set_cork(csock, 0);
 994     }
 995
 996     client->send_coroutine = NULL;
 997     qemu_set_fd_handler2(csock, nbd_can_read, nbd_read, NULL, client);
 998     qemu_co_mutex_unlock(&client->send_lock);
 999     return rc;
1000 }
1001
1002 static ssize_t nbd_co_receive_request(NBDRequest *req, struct nbd_request *request)
1003 {
1004     NBDClient *client = req->client;
1005     int csock = client->sock;
1006     ssize_t rc;
1007
1008     client->recv_coroutine = qemu_coroutine_self();
1009     rc = nbd_receive_request(csock, request);
1010     if (rc < 0) {
1011         if (rc != -EAGAIN) {
1012             rc = -EIO;
1013         }
1014         goto out;
1015     }
1016
1017     if (request->len > NBD_BUFFER_SIZE) {
1018         LOG("len (%u) is larger than max len (%u)",
1019             request->len, NBD_BUFFER_SIZE);
1020         rc = -EINVAL;
1021         goto out;
1022     }
1023
1024     if ((request->from + request->len) < request->from) {
1025         LOG("integer overflow detected! "
1026             "you're probably being attacked");
1027         rc = -EINVAL;
1028         goto out;
1029     }
1030
1031     TRACE("Decoding type");
1032
1033     if ((request->type & NBD_CMD_MASK_COMMAND) == NBD_CMD_WRITE) {
1034         TRACE("Reading %u byte(s)", request->len);
1035
1036         if (qemu_co_recv(csock, req->data, request->len) != request->len) {
1037             LOG("reading from socket failed");
1038             rc = -EIO;
1039             goto out;
1040         }
1041     }
1042     rc = 0;
1043
1044 out:
1045     client->recv_coroutine = NULL;
1046     return rc;
1047 }
1048
1049 static void nbd_trip(void *opaque)
1050 {
1051     NBDClient *client = opaque;
1052     NBDExport *exp = client->exp;
1053     NBDRequest *req;
1054     struct nbd_request request;
1055     struct nbd_reply reply;
1056     ssize_t ret;
1057
1058     TRACE("Reading request.");
1059     if (client->closing) {
1060         return;
1061     }
1062
1063     req = nbd_request_get(client);
1064     ret = nbd_co_receive_request(req, &request);
1065     if (ret == -EAGAIN) {
1066         goto done;
1067     }
1068     if (ret == -EIO) {
1069         goto out;
1070     }
1071
1072     reply.handle = request.handle;
1073     reply.error = 0;
1074
1075     if (ret < 0) {
1076         reply.error = -ret;
1077         goto error_reply;
1078     }
1079
1080     if ((request.from + request.len) > exp->size) {
1081             LOG("From: %" PRIu64 ", Len: %u, Size: %" PRIu64
1082             ", Offset: %" PRIu64 "\n",
1083                     request.from, request.len,
1084                     (uint64_t)exp->size, (uint64_t)exp->dev_offset);
1085         LOG("requested operation past EOF--bad client?");
1086         goto invalid_request;
1087     }
1088
1089     switch (request.type & NBD_CMD_MASK_COMMAND) {
1090     case NBD_CMD_READ:
1091         TRACE("Request type is READ");
1092
1093         if (request.type & NBD_CMD_FLAG_FUA) {
1094             ret = bdrv_co_flush(exp->bs);
1095             if (ret < 0) {
1096                 LOG("flush failed");
1097                 reply.error = -ret;
1098                 goto error_reply;
1099             }
1100         }
1101
1102         ret = bdrv_read(exp->bs, (request.from + exp->dev_offset) / 512,
1103                         req->data, request.len / 512);
1104         if (ret < 0) {
1105             LOG("reading from file failed");
1106             reply.error = -ret;
1107             goto error_reply;
1108         }
1109
1110         TRACE("Read %u byte(s)", request.len);
1111         if (nbd_co_send_reply(req, &reply, request.len) < 0)
1112             goto out;
1113         break;
1114     case NBD_CMD_WRITE:
1115         TRACE("Request type is WRITE");
1116
1117         if (exp->nbdflags & NBD_FLAG_READ_ONLY) {
1118             TRACE("Server is read-only, return error");
1119             reply.error = EROFS;
1120             goto error_reply;
1121         }
1122
1123         TRACE("Writing to device");
1124
1125         ret = bdrv_write(exp->bs, (request.from + exp->dev_offset) / 512,
1126                          req->data, request.len / 512);
1127         if (ret < 0) {
1128             LOG("writing to file failed");
1129             reply.error = -ret;
1130             goto error_reply;
1131         }
1132
1133         if (request.type & NBD_CMD_FLAG_FUA) {
1134             ret = bdrv_co_flush(exp->bs);
1135             if (ret < 0) {
1136                 LOG("flush failed");
1137                 reply.error = -ret;
1138                 goto error_reply;
1139             }
1140         }
1141
1142         if (nbd_co_send_reply(req, &reply, 0) < 0) {
1143             goto out;
1144         }
1145         break;
1146     case NBD_CMD_DISC:
1147         TRACE("Request type is DISCONNECT");
1148         errno = 0;
1149         goto out;
1150     case NBD_CMD_FLUSH:
1151         TRACE("Request type is FLUSH");
1152
1153         ret = bdrv_co_flush(exp->bs);
1154         if (ret < 0) {
1155             LOG("flush failed");
1156             reply.error = -ret;
1157         }
1158         if (nbd_co_send_reply(req, &reply, 0) < 0) {
1159             goto out;
1160         }
1161         break;
1162     case NBD_CMD_TRIM:
1163         TRACE("Request type is TRIM");
1164         ret = bdrv_co_discard(exp->bs, (request.from + exp->dev_offset) / 512,
1165                               request.len / 512);
1166         if (ret < 0) {
1167             LOG("discard failed");
1168             reply.error = -ret;
1169         }
1170         if (nbd_co_send_reply(req, &reply, 0) < 0) {
1171             goto out;
1172         }
1173         break;
1174     default:
1175         LOG("invalid request type (%u) received", request.type);
1176     invalid_request:
1177         reply.error = -EINVAL;
1178     error_reply:
1179         if (nbd_co_send_reply(req, &reply, 0) < 0) {
1180             goto out;
1181         }
1182         break;
1183     }
1184
1185     TRACE("Request/Reply complete");
1186
1187 done:
1188     nbd_request_put(req);
1189     return;
1190
1191 out:
1192     nbd_request_put(req);
1193     nbd_client_close(client);
1194 }
1195
1196 static int nbd_can_read(void *opaque)
1197 {
1198     NBDClient *client = opaque;
1199
1200     return client->recv_coroutine || client->nb_requests < MAX_NBD_REQUESTS;
1201 }
1202
1203 static void nbd_read(void *opaque)
1204 {
1205     NBDClient *client = opaque;
1206
1207     if (client->recv_coroutine) {
1208         qemu_coroutine_enter(client->recv_coroutine, NULL);
1209     } else {
1210         qemu_coroutine_enter(qemu_coroutine_create(nbd_trip), client);
1211     }
1212 }
1213
1214 static void nbd_restart_write(void *opaque)
1215 {
1216     NBDClient *client = opaque;
1217
1218     qemu_coroutine_enter(client->send_coroutine, NULL);
1219 }
1220
1221 NBDClient *nbd_client_new(NBDExport *exp, int csock,
1222                           void (*close)(NBDClient *))
1223 {
1224     NBDClient *client;
1225     client = g_malloc0(sizeof(NBDClient));
1226     client->refcount = 1;
1227     client->exp = exp;
1228     client->sock = csock;
1229     if (nbd_send_negotiate(client) < 0) {
1230         g_free(client);
1231         return NULL;
1232     }
1233     client->close = close;
1234     qemu_co_mutex_init(&client->send_lock);
1235     qemu_set_fd_handler2(csock, nbd_can_read, nbd_read, NULL, client);
1236
1237     if (exp) {
1238         QTAILQ_INSERT_TAIL(&exp->clients, client, next);
1239         nbd_export_get(exp);
1240     }
1241     return client;
1242 }