net/net.c

   1 /*
   2  * QEMU System Emulator
   3  *
   4  * Copyright (c) 2003-2008 Fabrice Bellard
   5  *
   6  * Permission is hereby granted, free of charge, to any person obtaining a copy
   7  * of this software and associated documentation files (the "Software"), to deal
   8  * in the Software without restriction, including without limitation the rights
   9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  10  * copies of the Software, and to permit persons to whom the Software is
  11  * furnished to do so, subject to the following conditions:
  12  *
  13  * The above copyright notice and this permission notice shall be included in
  14  * all copies or substantial portions of the Software.
  15  *
  16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  22  * THE SOFTWARE.
  23  */
  24
  25 #include "qemu/osdep.h"
  26 #include "qemu-common.h"
  27
  28 #include "net/net.h"
  29 #include "clients.h"
  30 #include "hub.h"
  31 #include "hw/qdev-properties.h"
  32 #include "net/slirp.h"
  33 #include "net/eth.h"
  34 #include "util.h"
  35
  36 #include "monitor/monitor.h"
  37 #include "qemu/help_option.h"
  38 #include "qapi/qapi-commands-net.h"
  39 #include "qapi/qmp/qdict.h"
  40 #include "qapi/qmp/qerror.h"
  41 #include "qemu/error-report.h"
  42 #include "qemu/sockets.h"
  43 #include "qemu/cutils.h"
  44 #include "qemu/config-file.h"
  45 #include "qemu/ctype.h"
  46 #include "qemu/id.h"
  47 #include "qemu/iov.h"
  48 #include "qemu/qemu-print.h"
  49 #include "qemu/main-loop.h"
  50 #include "qemu/option.h"
  51 #include "qapi/error.h"
  52 #include "qapi/opts-visitor.h"
  53 #include "sysemu/sysemu.h"
  54 #include "sysemu/runstate.h"
  55 #include "sysemu/sysemu.h"
  56 #include "net/filter.h"
  57 #include "qapi/string-output-visitor.h"
  58 #include "qapi/hmp-output-visitor.h"
  59
  60 /* Net bridge is currently not supported for W32. */
  61 #if !defined(_WIN32)
  62 # define CONFIG_NET_BRIDGE
  63 #endif
  64
  65 static VMChangeStateEntry *net_change_state_entry;
  66 static QTAILQ_HEAD(, NetClientState) net_clients;
  67
  68 /***********************************************************/
  69 /* network device redirectors */
  70
  71 int parse_host_port(struct sockaddr_in *saddr, const char *str,
  72                     Error **errp)
  73 {
  74     gchar **substrings;
  75     struct hostent *he;
  76     const char *addr, *p, *r;
  77     int port, ret = 0;
  78
  79     substrings = g_strsplit(str, ":", 2);
  80     if (!substrings || !substrings[0] || !substrings[1]) {
  81         error_setg(errp, "host address '%s' doesn't contain ':' "
  82                    "separating host from port", str);
  83         ret = -1;
  84         goto out;
  85     }
  86
  87     addr = substrings[0];
  88     p = substrings[1];
  89
  90     saddr->sin_family = AF_INET;
  91     if (addr[0] == '\0') {
  92         saddr->sin_addr.s_addr = 0;
  93     } else {
  94         if (qemu_isdigit(addr[0])) {
  95             if (!inet_aton(addr, &saddr->sin_addr)) {
  96                 error_setg(errp, "host address '%s' is not a valid "
  97                            "IPv4 address", addr);
  98                 ret = -1;
  99                 goto out;
 100             }
 101         } else {
 102             he = gethostbyname(addr);
 103             if (he == NULL) {
 104                 error_setg(errp, "can't resolve host address '%s'", addr);
 105                 ret = -1;
 106                 goto out;
 107             }
 108             saddr->sin_addr = *(struct in_addr *)he->h_addr;
 109         }
 110     }
 111     port = strtol(p, (char **)&r, 0);
 112     if (r == p) {
 113         error_setg(errp, "port number '%s' is invalid", p);
 114         ret = -1;
 115         goto out;
 116     }
 117     saddr->sin_port = htons(port);
 118
 119 out:
 120     g_strfreev(substrings);
 121     return ret;
 122 }
 123
 124 char *qemu_mac_strdup_printf(const uint8_t *macaddr)
 125 {
 126     return g_strdup_printf("%.2x:%.2x:%.2x:%.2x:%.2x:%.2x",
 127                            macaddr[0], macaddr[1], macaddr[2],
 128                            macaddr[3], macaddr[4], macaddr[5]);
 129 }
 130
 131 void qemu_format_nic_info_str(NetClientState *nc, uint8_t macaddr[6])
 132 {
 133     g_free(nc->info_str);
 134     nc->info_str = g_strdup_printf(
 135         "model=%s,macaddr=%02x:%02x:%02x:%02x:%02x:%02x",
 136         nc->model,
 137         macaddr[0], macaddr[1], macaddr[2],
 138         macaddr[3], macaddr[4], macaddr[5]);
 139 }
 140
 141 static int mac_table[256] = {0};
 142
 143 static void qemu_macaddr_set_used(MACAddr *macaddr)
 144 {
 145     int index;
 146
 147     for (index = 0x56; index < 0xFF; index++) {
 148         if (macaddr->a[5] == index) {
 149             mac_table[index]++;
 150         }
 151     }
 152 }
 153
 154 static void qemu_macaddr_set_free(MACAddr *macaddr)
 155 {
 156     int index;
 157     static const MACAddr base = { .a = { 0x52, 0x54, 0x00, 0x12, 0x34, 0 } };
 158
 159     if (memcmp(macaddr->a, &base.a, (sizeof(base.a) - 1)) != 0) {
 160         return;
 161     }
 162     for (index = 0x56; index < 0xFF; index++) {
 163         if (macaddr->a[5] == index) {
 164             mac_table[index]--;
 165         }
 166     }
 167 }
 168
 169 static int qemu_macaddr_get_free(void)
 170 {
 171     int index;
 172
 173     for (index = 0x56; index < 0xFF; index++) {
 174         if (mac_table[index] == 0) {
 175             return index;
 176         }
 177     }
 178
 179     return -1;
 180 }
 181
 182 void qemu_macaddr_default_if_unset(MACAddr *macaddr)
 183 {
 184     static const MACAddr zero = { .a = { 0,0,0,0,0,0 } };
 185     static const MACAddr base = { .a = { 0x52, 0x54, 0x00, 0x12, 0x34, 0 } };
 186
 187     if (memcmp(macaddr, &zero, sizeof(zero)) != 0) {
 188         if (memcmp(macaddr->a, &base.a, (sizeof(base.a) - 1)) != 0) {
 189             return;
 190         } else {
 191             qemu_macaddr_set_used(macaddr);
 192             return;
 193         }
 194     }
 195
 196     macaddr->a[0] = 0x52;
 197     macaddr->a[1] = 0x54;
 198     macaddr->a[2] = 0x00;
 199     macaddr->a[3] = 0x12;
 200     macaddr->a[4] = 0x34;
 201     macaddr->a[5] = qemu_macaddr_get_free();
 202     qemu_macaddr_set_used(macaddr);
 203 }
 204
 205 /**
 206  * Generate a name for net client
 207  *
 208  * Only net clients created with the legacy -net option and NICs need this.
 209  */
 210 static char *assign_name(NetClientState *nc1, const char *model)
 211 {
 212     NetClientState *nc;
 213     int id = 0;
 214
 215     QTAILQ_FOREACH(nc, &net_clients, next) {
 216         if (nc == nc1) {
 217             continue;
 218         }
 219         if (strcmp(nc->model, model) == 0) {
 220             id++;
 221         }
 222     }
 223
 224     return g_strdup_printf("%s.%d", model, id);
 225 }
 226
 227 static void qemu_net_client_destructor(NetClientState *nc)
 228 {
 229     g_free(nc);
 230 }
 231 static ssize_t qemu_deliver_packet_iov(NetClientState *sender,
 232                                        unsigned flags,
 233                                        const struct iovec *iov,
 234                                        int iovcnt,
 235                                        void *opaque);
 236
 237 static void qemu_net_client_setup(NetClientState *nc,
 238                                   NetClientInfo *info,
 239                                   NetClientState *peer,
 240                                   const char *model,
 241                                   const char *name,
 242                                   NetClientDestructor *destructor)
 243 {
 244     nc->info = info;
 245     nc->model = g_strdup(model);
 246     if (name) {
 247         nc->name = g_strdup(name);
 248     } else {
 249         nc->name = assign_name(nc, model);
 250     }
 251
 252     if (peer) {
 253         assert(!peer->peer);
 254         nc->peer = peer;
 255         peer->peer = nc;
 256     }
 257     QTAILQ_INSERT_TAIL(&net_clients, nc, next);
 258
 259     nc->incoming_queue = qemu_new_net_queue(qemu_deliver_packet_iov, nc);
 260     nc->destructor = destructor;
 261     QTAILQ_INIT(&nc->filters);
 262 }
 263
 264 NetClientState *qemu_new_net_client(NetClientInfo *info,
 265                                     NetClientState *peer,
 266                                     const char *model,
 267                                     const char *name)
 268 {
 269     NetClientState *nc;
 270
 271     assert(info->size >= sizeof(NetClientState));
 272
 273     nc = g_malloc0(info->size);
 274     qemu_net_client_setup(nc, info, peer, model, name,
 275                           qemu_net_client_destructor);
 276
 277     return nc;
 278 }
 279
 280 NICState *qemu_new_nic(NetClientInfo *info,
 281                        NICConf *conf,
 282                        const char *model,
 283                        const char *name,
 284                        void *opaque)
 285 {
 286     NetClientState **peers = conf->peers.ncs;
 287     NICState *nic;
 288     int i, queues = MAX(1, conf->peers.queues);
 289
 290     assert(info->type == NET_CLIENT_DRIVER_NIC);
 291     assert(info->size >= sizeof(NICState));
 292
 293     nic = g_malloc0(info->size + sizeof(NetClientState) * queues);
 294     nic->ncs = (void *)nic + info->size;
 295     nic->conf = conf;
 296     nic->opaque = opaque;
 297
 298     for (i = 0; i < queues; i++) {
 299         qemu_net_client_setup(&nic->ncs[i], info, peers[i], model, name,
 300                               NULL);
 301         nic->ncs[i].queue_index = i;
 302     }
 303
 304     return nic;
 305 }
 306
 307 NetClientState *qemu_get_subqueue(NICState *nic, int queue_index)
 308 {
 309     return nic->ncs + queue_index;
 310 }
 311
 312 NetClientState *qemu_get_queue(NICState *nic)
 313 {
 314     return qemu_get_subqueue(nic, 0);
 315 }
 316
 317 NICState *qemu_get_nic(NetClientState *nc)
 318 {
 319     NetClientState *nc0 = nc - nc->queue_index;
 320
 321     return (NICState *)((void *)nc0 - nc->info->size);
 322 }
 323
 324 void *qemu_get_nic_opaque(NetClientState *nc)
 325 {
 326     NICState *nic = qemu_get_nic(nc);
 327
 328     return nic->opaque;
 329 }
 330
 331 NetClientState *qemu_get_peer(NetClientState *nc, int queue_index)
 332 {
 333     assert(nc != NULL);
 334     NetClientState *ncs = nc + queue_index;
 335     return ncs->peer;
 336 }
 337
 338 static void qemu_cleanup_net_client(NetClientState *nc)
 339 {
 340     QTAILQ_REMOVE(&net_clients, nc, next);
 341
 342     if (nc->info->cleanup) {
 343         nc->info->cleanup(nc);
 344     }
 345 }
 346
 347 static void qemu_free_net_client(NetClientState *nc)
 348 {
 349     if (nc->incoming_queue) {
 350         qemu_del_net_queue(nc->incoming_queue);
 351     }
 352     if (nc->peer) {
 353         nc->peer->peer = NULL;
 354     }
 355     g_free(nc->name);
 356     g_free(nc->model);
 357     g_free(nc->info_str);
 358     qapi_free_NetdevInfo(nc->stored_config);
 359     if (nc->destructor) {
 360         nc->destructor(nc);
 361     }
 362 }
 363
 364 void qemu_del_net_client(NetClientState *nc)
 365 {
 366     NetClientState *ncs[MAX_QUEUE_NUM];
 367     int queues, i;
 368     NetFilterState *nf, *next;
 369
 370     assert(nc->info->type != NET_CLIENT_DRIVER_NIC);
 371
 372     /* If the NetClientState belongs to a multiqueue backend, we will change all
 373      * other NetClientStates also.
 374      */
 375     queues = qemu_find_net_clients_except(nc->name, ncs,
 376                                           NET_CLIENT_DRIVER_NIC,
 377                                           MAX_QUEUE_NUM);
 378     assert(queues != 0);
 379
 380     QTAILQ_FOREACH_SAFE(nf, &nc->filters, next, next) {
 381         object_unparent(OBJECT(nf));
 382     }
 383
 384     /* If there is a peer NIC, delete and cleanup client, but do not free. */
 385     if (nc->peer && nc->peer->info->type == NET_CLIENT_DRIVER_NIC) {
 386         NICState *nic = qemu_get_nic(nc->peer);
 387         if (nic->peer_deleted) {
 388             return;
 389         }
 390         nic->peer_deleted = true;
 391
 392         for (i = 0; i < queues; i++) {
 393             ncs[i]->peer->link_down = true;
 394         }
 395
 396         if (nc->peer->info->link_status_changed) {
 397             nc->peer->info->link_status_changed(nc->peer);
 398         }
 399
 400         for (i = 0; i < queues; i++) {
 401             qemu_cleanup_net_client(ncs[i]);
 402         }
 403
 404         return;
 405     }
 406
 407     for (i = 0; i < queues; i++) {
 408         qemu_cleanup_net_client(ncs[i]);
 409         qemu_free_net_client(ncs[i]);
 410     }
 411 }
 412
 413 void qemu_del_nic(NICState *nic)
 414 {
 415     int i, queues = MAX(nic->conf->peers.queues, 1);
 416
 417     qemu_macaddr_set_free(&nic->conf->macaddr);
 418
 419     for (i = 0; i < queues; i++) {
 420         NetClientState *nc = qemu_get_subqueue(nic, i);
 421         /* If this is a peer NIC and peer has already been deleted, free it now. */
 422         if (nic->peer_deleted) {
 423             qemu_free_net_client(nc->peer);
 424         } else if (nc->peer) {
 425             /* if there are RX packets pending, complete them */
 426             qemu_purge_queued_packets(nc->peer);
 427         }
 428     }
 429
 430     for (i = queues - 1; i >= 0; i--) {
 431         NetClientState *nc = qemu_get_subqueue(nic, i);
 432
 433         qemu_cleanup_net_client(nc);
 434         qemu_free_net_client(nc);
 435     }
 436
 437     g_free(nic);
 438 }
 439
 440 void qemu_foreach_nic(qemu_nic_foreach func, void *opaque)
 441 {
 442     NetClientState *nc;
 443
 444     QTAILQ_FOREACH(nc, &net_clients, next) {
 445         if (nc->info->type == NET_CLIENT_DRIVER_NIC) {
 446             if (nc->queue_index == 0) {
 447                 func(qemu_get_nic(nc), opaque);
 448             }
 449         }
 450     }
 451 }
 452
 453 bool qemu_has_ufo(NetClientState *nc)
 454 {
 455     if (!nc || !nc->info->has_ufo) {
 456         return false;
 457     }
 458
 459     return nc->info->has_ufo(nc);
 460 }
 461
 462 bool qemu_has_vnet_hdr(NetClientState *nc)
 463 {
 464     if (!nc || !nc->info->has_vnet_hdr) {
 465         return false;
 466     }
 467
 468     return nc->info->has_vnet_hdr(nc);
 469 }
 470
 471 bool qemu_has_vnet_hdr_len(NetClientState *nc, int len)
 472 {
 473     if (!nc || !nc->info->has_vnet_hdr_len) {
 474         return false;
 475     }
 476
 477     return nc->info->has_vnet_hdr_len(nc, len);
 478 }
 479
 480 void qemu_using_vnet_hdr(NetClientState *nc, bool enable)
 481 {
 482     if (!nc || !nc->info->using_vnet_hdr) {
 483         return;
 484     }
 485
 486     nc->info->using_vnet_hdr(nc, enable);
 487 }
 488
 489 void qemu_set_offload(NetClientState *nc, int csum, int tso4, int tso6,
 490                           int ecn, int ufo)
 491 {
 492     if (!nc || !nc->info->set_offload) {
 493         return;
 494     }
 495
 496     nc->info->set_offload(nc, csum, tso4, tso6, ecn, ufo);
 497 }
 498
 499 void qemu_set_vnet_hdr_len(NetClientState *nc, int len)
 500 {
 501     if (!nc || !nc->info->set_vnet_hdr_len) {
 502         return;
 503     }
 504
 505     nc->vnet_hdr_len = len;
 506     nc->info->set_vnet_hdr_len(nc, len);
 507 }
 508
 509 int qemu_set_vnet_le(NetClientState *nc, bool is_le)
 510 {
 511 #ifdef HOST_WORDS_BIGENDIAN
 512     if (!nc || !nc->info->set_vnet_le) {
 513         return -ENOSYS;
 514     }
 515
 516     return nc->info->set_vnet_le(nc, is_le);
 517 #else
 518     return 0;
 519 #endif
 520 }
 521
 522 int qemu_set_vnet_be(NetClientState *nc, bool is_be)
 523 {
 524 #ifdef HOST_WORDS_BIGENDIAN
 525     return 0;
 526 #else
 527     if (!nc || !nc->info->set_vnet_be) {
 528         return -ENOSYS;
 529     }
 530
 531     return nc->info->set_vnet_be(nc, is_be);
 532 #endif
 533 }
 534
 535 int qemu_can_receive_packet(NetClientState *nc)
 536 {
 537     if (nc->receive_disabled) {
 538         return 0;
 539     } else if (nc->info->can_receive &&
 540                !nc->info->can_receive(nc)) {
 541         return 0;
 542     }
 543     return 1;
 544 }
 545
 546 int qemu_can_send_packet(NetClientState *sender)
 547 {
 548     int vm_running = runstate_is_running();
 549
 550     if (!vm_running) {
 551         return 0;
 552     }
 553
 554     if (!sender->peer) {
 555         return 1;
 556     }
 557
 558     return qemu_can_receive_packet(sender->peer);
 559 }
 560
 561 static ssize_t filter_receive_iov(NetClientState *nc,
 562                                   NetFilterDirection direction,
 563                                   NetClientState *sender,
 564                                   unsigned flags,
 565                                   const struct iovec *iov,
 566                                   int iovcnt,
 567                                   NetPacketSent *sent_cb)
 568 {
 569     ssize_t ret = 0;
 570     NetFilterState *nf = NULL;
 571
 572     if (direction == NET_FILTER_DIRECTION_TX) {
 573         QTAILQ_FOREACH(nf, &nc->filters, next) {
 574             ret = qemu_netfilter_receive(nf, direction, sender, flags, iov,
 575                                          iovcnt, sent_cb);
 576             if (ret) {
 577                 return ret;
 578             }
 579         }
 580     } else {
 581         QTAILQ_FOREACH_REVERSE(nf, &nc->filters, next) {
 582             ret = qemu_netfilter_receive(nf, direction, sender, flags, iov,
 583                                          iovcnt, sent_cb);
 584             if (ret) {
 585                 return ret;
 586             }
 587         }
 588     }
 589
 590     return ret;
 591 }
 592
 593 static ssize_t filter_receive(NetClientState *nc,
 594                               NetFilterDirection direction,
 595                               NetClientState *sender,
 596                               unsigned flags,
 597                               const uint8_t *data,
 598                               size_t size,
 599                               NetPacketSent *sent_cb)
 600 {
 601     struct iovec iov = {
 602         .iov_base = (void *)data,
 603         .iov_len = size
 604     };
 605
 606     return filter_receive_iov(nc, direction, sender, flags, &iov, 1, sent_cb);
 607 }
 608
 609 void qemu_purge_queued_packets(NetClientState *nc)
 610 {
 611     if (!nc->peer) {
 612         return;
 613     }
 614
 615     qemu_net_queue_purge(nc->peer->incoming_queue, nc);
 616 }
 617
 618 void qemu_flush_or_purge_queued_packets(NetClientState *nc, bool purge)
 619 {
 620     nc->receive_disabled = 0;
 621
 622     if (nc->peer && nc->peer->info->type == NET_CLIENT_DRIVER_HUBPORT) {
 623         if (net_hub_flush(nc->peer)) {
 624             qemu_notify_event();
 625         }
 626     }
 627     if (qemu_net_queue_flush(nc->incoming_queue)) {
 628         /* We emptied the queue successfully, signal to the IO thread to repoll
 629          * the file descriptor (for tap, for example).
 630          */
 631         qemu_notify_event();
 632     } else if (purge) {
 633         /* Unable to empty the queue, purge remaining packets */
 634         qemu_net_queue_purge(nc->incoming_queue, nc->peer);
 635     }
 636 }
 637
 638 void qemu_flush_queued_packets(NetClientState *nc)
 639 {
 640     qemu_flush_or_purge_queued_packets(nc, false);
 641 }
 642
 643 static ssize_t qemu_send_packet_async_with_flags(NetClientState *sender,
 644                                                  unsigned flags,
 645                                                  const uint8_t *buf, int size,
 646                                                  NetPacketSent *sent_cb)
 647 {
 648     NetQueue *queue;
 649     int ret;
 650
 651 #ifdef DEBUG_NET
 652     printf("qemu_send_packet_async:\n");
 653     qemu_hexdump(stdout, "net", buf, size);
 654 #endif
 655
 656     if (sender->link_down || !sender->peer) {
 657         return size;
 658     }
 659
 660     /* Let filters handle the packet first */
 661     ret = filter_receive(sender, NET_FILTER_DIRECTION_TX,
 662                          sender, flags, buf, size, sent_cb);
 663     if (ret) {
 664         return ret;
 665     }
 666
 667     ret = filter_receive(sender->peer, NET_FILTER_DIRECTION_RX,
 668                          sender, flags, buf, size, sent_cb);
 669     if (ret) {
 670         return ret;
 671     }
 672
 673     queue = sender->peer->incoming_queue;
 674
 675     return qemu_net_queue_send(queue, sender, flags, buf, size, sent_cb);
 676 }
 677
 678 ssize_t qemu_send_packet_async(NetClientState *sender,
 679                                const uint8_t *buf, int size,
 680                                NetPacketSent *sent_cb)
 681 {
 682     return qemu_send_packet_async_with_flags(sender, QEMU_NET_PACKET_FLAG_NONE,
 683                                              buf, size, sent_cb);
 684 }
 685
 686 ssize_t qemu_send_packet(NetClientState *nc, const uint8_t *buf, int size)
 687 {
 688     return qemu_send_packet_async(nc, buf, size, NULL);
 689 }
 690
 691 ssize_t qemu_receive_packet(NetClientState *nc, const uint8_t *buf, int size)
 692 {
 693     if (!qemu_can_receive_packet(nc)) {
 694         return 0;
 695     }
 696
 697     return qemu_net_queue_receive(nc->incoming_queue, buf, size);
 698 }
 699
 700 ssize_t qemu_receive_packet_iov(NetClientState *nc, const struct iovec *iov,
 701                                 int iovcnt)
 702 {
 703     if (!qemu_can_receive_packet(nc)) {
 704         return 0;
 705     }
 706
 707     return qemu_net_queue_receive_iov(nc->incoming_queue, iov, iovcnt);
 708 }
 709
 710 ssize_t qemu_send_packet_raw(NetClientState *nc, const uint8_t *buf, int size)
 711 {
 712     return qemu_send_packet_async_with_flags(nc, QEMU_NET_PACKET_FLAG_RAW,
 713                                              buf, size, NULL);
 714 }
 715
 716 static ssize_t nc_sendv_compat(NetClientState *nc, const struct iovec *iov,
 717                                int iovcnt, unsigned flags)
 718 {
 719     uint8_t *buf = NULL;
 720     uint8_t *buffer;
 721     size_t offset;
 722     ssize_t ret;
 723
 724     if (iovcnt == 1) {
 725         buffer = iov[0].iov_base;
 726         offset = iov[0].iov_len;
 727     } else {
 728         offset = iov_size(iov, iovcnt);
 729         if (offset > NET_BUFSIZE) {
 730             return -1;
 731         }
 732         buf = g_malloc(offset);
 733         buffer = buf;
 734         offset = iov_to_buf(iov, iovcnt, 0, buf, offset);
 735     }
 736
 737     if (flags & QEMU_NET_PACKET_FLAG_RAW && nc->info->receive_raw) {
 738         ret = nc->info->receive_raw(nc, buffer, offset);
 739     } else {
 740         ret = nc->info->receive(nc, buffer, offset);
 741     }
 742
 743     g_free(buf);
 744     return ret;
 745 }
 746
 747 static ssize_t qemu_deliver_packet_iov(NetClientState *sender,
 748                                        unsigned flags,
 749                                        const struct iovec *iov,
 750                                        int iovcnt,
 751                                        void *opaque)
 752 {
 753     NetClientState *nc = opaque;
 754     int ret;
 755
 756
 757     if (nc->link_down) {
 758         return iov_size(iov, iovcnt);
 759     }
 760
 761     if (nc->receive_disabled) {
 762         return 0;
 763     }
 764
 765     if (nc->info->receive_iov && !(flags & QEMU_NET_PACKET_FLAG_RAW)) {
 766         ret = nc->info->receive_iov(nc, iov, iovcnt);
 767     } else {
 768         ret = nc_sendv_compat(nc, iov, iovcnt, flags);
 769     }
 770
 771     if (ret == 0) {
 772         nc->receive_disabled = 1;
 773     }
 774
 775     return ret;
 776 }
 777
 778 ssize_t qemu_sendv_packet_async(NetClientState *sender,
 779                                 const struct iovec *iov, int iovcnt,
 780                                 NetPacketSent *sent_cb)
 781 {
 782     NetQueue *queue;
 783     size_t size = iov_size(iov, iovcnt);
 784     int ret;
 785
 786     if (size > NET_BUFSIZE) {
 787         return size;
 788     }
 789
 790     if (sender->link_down || !sender->peer) {
 791         return size;
 792     }
 793
 794     /* Let filters handle the packet first */
 795     ret = filter_receive_iov(sender, NET_FILTER_DIRECTION_TX, sender,
 796                              QEMU_NET_PACKET_FLAG_NONE, iov, iovcnt, sent_cb);
 797     if (ret) {
 798         return ret;
 799     }
 800
 801     ret = filter_receive_iov(sender->peer, NET_FILTER_DIRECTION_RX, sender,
 802                              QEMU_NET_PACKET_FLAG_NONE, iov, iovcnt, sent_cb);
 803     if (ret) {
 804         return ret;
 805     }
 806
 807     queue = sender->peer->incoming_queue;
 808
 809     return qemu_net_queue_send_iov(queue, sender,
 810                                    QEMU_NET_PACKET_FLAG_NONE,
 811                                    iov, iovcnt, sent_cb);
 812 }
 813
 814 ssize_t
 815 qemu_sendv_packet(NetClientState *nc, const struct iovec *iov, int iovcnt)
 816 {
 817     return qemu_sendv_packet_async(nc, iov, iovcnt, NULL);
 818 }
 819
 820 NetClientState *qemu_find_netdev(const char *id)
 821 {
 822     NetClientState *nc;
 823
 824     QTAILQ_FOREACH(nc, &net_clients, next) {
 825         if (nc->info->type == NET_CLIENT_DRIVER_NIC)
 826             continue;
 827         if (!strcmp(nc->name, id)) {
 828             return nc;
 829         }
 830     }
 831
 832     return NULL;
 833 }
 834
 835 int qemu_find_net_clients_except(const char *id, NetClientState **ncs,
 836                                  NetClientDriver type, int max)
 837 {
 838     NetClientState *nc;
 839     int ret = 0;
 840
 841     QTAILQ_FOREACH(nc, &net_clients, next) {
 842         if (nc->info->type == type) {
 843             continue;
 844         }
 845         if (!id || !strcmp(nc->name, id)) {
 846             if (ret < max) {
 847                 ncs[ret] = nc;
 848             }
 849             ret++;
 850         }
 851     }
 852
 853     return ret;
 854 }
 855
 856 static int nic_get_free_idx(void)
 857 {
 858     int index;
 859
 860     for (index = 0; index < MAX_NICS; index++)
 861         if (!nd_table[index].used)
 862             return index;
 863     return -1;
 864 }
 865
 866 int qemu_show_nic_models(const char *arg, const char *const *models)
 867 {
 868     int i;
 869
 870     if (!arg || !is_help_option(arg)) {
 871         return 0;
 872     }
 873
 874     printf("Supported NIC models:\n");
 875     for (i = 0 ; models[i]; i++) {
 876         printf("%s\n", models[i]);
 877     }
 878     return 1;
 879 }
 880
 881 void qemu_check_nic_model(NICInfo *nd, const char *model)
 882 {
 883     const char *models[2];
 884
 885     models[0] = model;
 886     models[1] = NULL;
 887
 888     if (qemu_show_nic_models(nd->model, models))
 889         exit(0);
 890     if (qemu_find_nic_model(nd, models, model) < 0)
 891         exit(1);
 892 }
 893
 894 int qemu_find_nic_model(NICInfo *nd, const char * const *models,
 895                         const char *default_model)
 896 {
 897     int i;
 898
 899     if (!nd->model)
 900         nd->model = g_strdup(default_model);
 901
 902     for (i = 0 ; models[i]; i++) {
 903         if (strcmp(nd->model, models[i]) == 0)
 904             return i;
 905     }
 906
 907     error_report("Unsupported NIC model: %s", nd->model);
 908     return -1;
 909 }
 910
 911 static int net_init_nic(const Netdev *netdev, const char *name,
 912                         NetClientState *peer, Error **errp)
 913 {
 914     int idx;
 915     NICInfo *nd;
 916     const NetLegacyNicOptions *nic;
 917
 918     assert(netdev->type == NET_CLIENT_DRIVER_NIC);
 919     nic = &netdev->u.nic;
 920
 921     idx = nic_get_free_idx();
 922     if (idx == -1 || nb_nics >= MAX_NICS) {
 923         error_setg(errp, "too many NICs");
 924         return -1;
 925     }
 926
 927     nd = &nd_table[idx];
 928
 929     memset(nd, 0, sizeof(*nd));
 930
 931     if (nic->has_netdev) {
 932         nd->netdev = qemu_find_netdev(nic->netdev);
 933         if (!nd->netdev) {
 934             error_setg(errp, "netdev '%s' not found", nic->netdev);
 935             return -1;
 936         }
 937     } else {
 938         assert(peer);
 939         nd->netdev = peer;
 940     }
 941     nd->name = g_strdup(name);
 942     if (nic->has_model) {
 943         nd->model = g_strdup(nic->model);
 944     }
 945     if (nic->has_addr) {
 946         nd->devaddr = g_strdup(nic->addr);
 947     }
 948
 949     if (nic->has_macaddr &&
 950         net_parse_macaddr(nd->macaddr.a, nic->macaddr) < 0) {
 951         error_setg(errp, "invalid syntax for ethernet address");
 952         return -1;
 953     }
 954     if (nic->has_macaddr &&
 955         is_multicast_ether_addr(nd->macaddr.a)) {
 956         error_setg(errp,
 957                    "NIC cannot have multicast MAC address (odd 1st byte)");
 958         return -1;
 959     }
 960     qemu_macaddr_default_if_unset(&nd->macaddr);
 961
 962     if (nic->has_vectors) {
 963         if (nic->vectors > 0x7ffffff) {
 964             error_setg(errp, "invalid # of vectors: %"PRIu32, nic->vectors);
 965             return -1;
 966         }
 967         nd->nvectors = nic->vectors;
 968     } else {
 969         nd->nvectors = DEV_NVECTORS_UNSPECIFIED;
 970     }
 971
 972     nd->used = 1;
 973     nb_nics++;
 974
 975     return idx;
 976 }
 977
 978
 979 static int (* const net_client_init_fun[NET_CLIENT_DRIVER__MAX])(
 980     const Netdev *netdev,
 981     const char *name,
 982     NetClientState *peer, Error **errp) = {
 983         [NET_CLIENT_DRIVER_NIC]       = net_init_nic,
 984 #ifdef CONFIG_SLIRP
 985         [NET_CLIENT_DRIVER_USER]      = net_init_slirp,
 986 #endif
 987         [NET_CLIENT_DRIVER_TAP]       = net_init_tap,
 988         [NET_CLIENT_DRIVER_SOCKET]    = net_init_socket,
 989 #ifdef CONFIG_VDE
 990         [NET_CLIENT_DRIVER_VDE]       = net_init_vde,
 991 #endif
 992 #ifdef CONFIG_NETMAP
 993         [NET_CLIENT_DRIVER_NETMAP]    = net_init_netmap,
 994 #endif
 995 #ifdef CONFIG_NET_BRIDGE
 996         [NET_CLIENT_DRIVER_BRIDGE]    = net_init_bridge,
 997 #endif
 998         [NET_CLIENT_DRIVER_HUBPORT]   = net_init_hubport,
 999 #ifdef CONFIG_VHOST_NET_USER
1000         [NET_CLIENT_DRIVER_VHOST_USER] = net_init_vhost_user,
1001 #endif
1002 #ifdef CONFIG_VHOST_NET_VDPA
1003         [NET_CLIENT_DRIVER_VHOST_VDPA] = net_init_vhost_vdpa,
1004 #endif
1005 #ifdef CONFIG_L2TPV3
1006         [NET_CLIENT_DRIVER_L2TPV3]    = net_init_l2tpv3,
1007 #endif
1008 };
1009
1010
1011 static int net_client_init1(const Netdev *netdev, bool is_netdev, Error **errp)
1012 {
1013     NetClientState *peer = NULL;
1014     NetClientState *nc;
1015
1016     if (is_netdev) {
1017         if (netdev->type == NET_CLIENT_DRIVER_NIC ||
1018             !net_client_init_fun[netdev->type]) {
1019             error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "type",
1020                        "a netdev backend type");
1021             return -1;
1022         }
1023     } else {
1024         if (netdev->type == NET_CLIENT_DRIVER_NONE) {
1025             return 0; /* nothing to do */
1026         }
1027         if (netdev->type == NET_CLIENT_DRIVER_HUBPORT ||
1028             !net_client_init_fun[netdev->type]) {
1029             error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "type",
1030                        "a net backend type (maybe it is not compiled "
1031                        "into this binary)");
1032             return -1;
1033         }
1034
1035         /* Do not add to a hub if it's a nic with a netdev= parameter. */
1036         if (netdev->type != NET_CLIENT_DRIVER_NIC ||
1037             !netdev->u.nic.has_netdev) {
1038             peer = net_hub_add_port(0, NULL, NULL);
1039         }
1040     }
1041
1042     nc = qemu_find_netdev(netdev->id);
1043     if (nc) {
1044         error_setg(errp, "Duplicate ID '%s'", netdev->id);
1045         return -1;
1046     }
1047
1048     if (net_client_init_fun[netdev->type](netdev, netdev->id, peer, errp) < 0) {
1049         /* FIXME drop when all init functions store an Error */
1050         if (errp && !*errp) {
1051             error_setg(errp, "Device '%s' could not be initialized",
1052                        NetClientDriver_str(netdev->type));
1053         }
1054         return -1;
1055     }
1056
1057     if (is_netdev) {
1058         nc = qemu_find_netdev(netdev->id);
1059         assert(nc);
1060         nc->is_netdev = true;
1061     }
1062
1063     return 0;
1064 }
1065
1066 void show_netdevs(void)
1067 {
1068     int idx;
1069     const char *available_netdevs[] = {
1070         "socket",
1071         "hubport",
1072         "tap",
1073 #ifdef CONFIG_SLIRP
1074         "user",
1075 #endif
1076 #ifdef CONFIG_L2TPV3
1077         "l2tpv3",
1078 #endif
1079 #ifdef CONFIG_VDE
1080         "vde",
1081 #endif
1082 #ifdef CONFIG_NET_BRIDGE
1083         "bridge",
1084 #endif
1085 #ifdef CONFIG_NETMAP
1086         "netmap",
1087 #endif
1088 #ifdef CONFIG_POSIX
1089         "vhost-user",
1090 #endif
1091 #ifdef CONFIG_VHOST_VDPA
1092         "vhost-vdpa",
1093 #endif
1094     };
1095
1096     qemu_printf("Available netdev backend types:\n");
1097     for (idx = 0; idx < ARRAY_SIZE(available_netdevs); idx++) {
1098         qemu_printf("%s\n", available_netdevs[idx]);
1099     }
1100 }
1101
1102 static int net_client_init(QemuOpts *opts, bool is_netdev, Error **errp)
1103 {
1104     gchar **substrings = NULL;
1105     Netdev *object = NULL;
1106     int ret = -1;
1107     Visitor *v = opts_visitor_new(opts);
1108
1109     /* Parse convenience option format ip6-net=fec0::0[/64] */
1110     const char *ip6_net = qemu_opt_get(opts, "ipv6-net");
1111
1112     if (ip6_net) {
1113         char *prefix_addr;
1114         unsigned long prefix_len = 64; /* Default 64bit prefix length. */
1115
1116         substrings = g_strsplit(ip6_net, "/", 2);
1117         if (!substrings || !substrings[0]) {
1118             error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "ipv6-net",
1119                        "a valid IPv6 prefix");
1120             goto out;
1121         }
1122
1123         prefix_addr = substrings[0];
1124
1125         /* Handle user-specified prefix length. */
1126         if (substrings[1] &&
1127             qemu_strtoul(substrings[1], NULL, 10, &prefix_len))
1128         {
1129             error_setg(errp, QERR_INVALID_PARAMETER_VALUE,
1130                        "ipv6-prefixlen", "a number");
1131             goto out;
1132         }
1133
1134         qemu_opt_set(opts, "ipv6-prefix", prefix_addr, &error_abort);
1135         qemu_opt_set_number(opts, "ipv6-prefixlen", prefix_len,
1136                             &error_abort);
1137         qemu_opt_unset(opts, "ipv6-net");
1138     }
1139
1140     /* Create an ID for -net if the user did not specify one */
1141     if (!is_netdev && !qemu_opts_id(opts)) {
1142         qemu_opts_set_id(opts, id_generate(ID_NET));
1143     }
1144
1145     if (visit_type_Netdev(v, NULL, &object, errp)) {
1146         ret = net_client_init1(object, is_netdev, errp);
1147     }
1148
1149     qapi_free_Netdev(object);
1150
1151 out:
1152     g_strfreev(substrings);
1153     visit_free(v);
1154     return ret;
1155 }
1156
1157 void netdev_add(QemuOpts *opts, Error **errp)
1158 {
1159     net_client_init(opts, true, errp);
1160 }
1161
1162 void qmp_netdev_add(Netdev *netdev, Error **errp)
1163 {
1164     if (!id_wellformed(netdev->id)) {
1165         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "id", "an identifier");
1166         return;
1167     }
1168
1169     net_client_init1(netdev, true, errp);
1170 }
1171
1172 void qmp_netdev_del(const char *id, Error **errp)
1173 {
1174     NetClientState *nc;
1175     QemuOpts *opts;
1176
1177     nc = qemu_find_netdev(id);
1178     if (!nc) {
1179         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1180                   "Device '%s' not found", id);
1181         return;
1182     }
1183
1184     if (!nc->is_netdev) {
1185         error_setg(errp, "Device '%s' is not a netdev", id);
1186         return;
1187     }
1188
1189     qemu_del_net_client(nc);
1190
1191     /*
1192      * Wart: we need to delete the QemuOpts associated with netdevs
1193      * created via CLI or HMP, to avoid bogus "Duplicate ID" errors in
1194      * HMP netdev_add.
1195      */
1196     opts = qemu_opts_find(qemu_find_opts("netdev"), id);
1197     if (opts) {
1198         qemu_opts_del(opts);
1199     }
1200 }
1201
1202 static void netfilter_print_info(Monitor *mon, NetFilterState *nf)
1203 {
1204     char *str;
1205     ObjectProperty *prop;
1206     ObjectPropertyIterator iter;
1207     Visitor *v;
1208
1209     /* generate info str */
1210     object_property_iter_init(&iter, OBJECT(nf));
1211     while ((prop = object_property_iter_next(&iter))) {
1212         if (!strcmp(prop->name, "type")) {
1213             continue;
1214         }
1215         v = string_output_visitor_new(false, &str);
1216         object_property_get(OBJECT(nf), prop->name, v, NULL);
1217         visit_complete(v, &str);
1218         visit_free(v);
1219         monitor_printf(mon, ",%s=%s", prop->name, str);
1220         g_free(str);
1221     }
1222     monitor_printf(mon, "\n");
1223 }
1224
1225 static char *generate_info_str(NetClientState *nc)
1226 {
1227     NetdevInfo *ni = nc->stored_config;
1228     char *ret_out = NULL;
1229     Visitor *v;
1230
1231     /* Use legacy field info_str for NIC and hubports */
1232     if ((nc->info->type == NET_CLIENT_DRIVER_NIC) ||
1233         (nc->info->type == NET_CLIENT_DRIVER_HUBPORT)) {
1234         return g_strdup(nc->info_str ? nc->info_str : "");
1235     }
1236
1237     if (!ni) {
1238         return g_malloc0(1);
1239     }
1240
1241     v = hmp_output_visitor_new(&ret_out);
1242     if (visit_type_NetdevInfo(v, "", &ni, NULL)) {
1243         visit_complete(v, &ret_out);
1244     }
1245     visit_free(v);
1246
1247     return ret_out;
1248 }
1249
1250 void print_net_client(Monitor *mon, NetClientState *nc)
1251 {
1252     NetFilterState *nf;
1253     char *info_str = generate_info_str(nc);
1254
1255     monitor_printf(mon, "%s: index=%d,type=%s,%s\n", nc->name,
1256                    nc->queue_index,
1257                    NetClientDriver_str(nc->info->type),
1258                    info_str);
1259     g_free(info_str);
1260
1261     if (!QTAILQ_EMPTY(&nc->filters)) {
1262         monitor_printf(mon, "filters:\n");
1263     }
1264     QTAILQ_FOREACH(nf, &nc->filters, next) {
1265         monitor_printf(mon, "  - %s: type=%s",
1266                        object_get_canonical_path_component(OBJECT(nf)),
1267                        object_get_typename(OBJECT(nf)));
1268         netfilter_print_info(mon, nf);
1269     }
1270 }
1271
1272 RxFilterInfoList *qmp_query_rx_filter(bool has_name, const char *name,
1273                                       Error **errp)
1274 {
1275     NetClientState *nc;
1276     RxFilterInfoList *filter_list = NULL, **tail = &filter_list;
1277
1278     QTAILQ_FOREACH(nc, &net_clients, next) {
1279         RxFilterInfo *info;
1280
1281         if (has_name && strcmp(nc->name, name) != 0) {
1282             continue;
1283         }
1284
1285         /* only query rx-filter information of NIC */
1286         if (nc->info->type != NET_CLIENT_DRIVER_NIC) {
1287             if (has_name) {
1288                 error_setg(errp, "net client(%s) isn't a NIC", name);
1289                 assert(!filter_list);
1290                 return NULL;
1291             }
1292             continue;
1293         }
1294
1295         /* only query information on queue 0 since the info is per nic,
1296          * not per queue
1297          */
1298         if (nc->queue_index != 0)
1299             continue;
1300
1301         if (nc->info->query_rx_filter) {
1302             info = nc->info->query_rx_filter(nc);
1303             QAPI_LIST_APPEND(tail, info);
1304         } else if (has_name) {
1305             error_setg(errp, "net client(%s) doesn't support"
1306                        " rx-filter querying", name);
1307             assert(!filter_list);
1308             return NULL;
1309         }
1310
1311         if (has_name) {
1312             break;
1313         }
1314     }
1315
1316     if (filter_list == NULL && has_name) {
1317         error_setg(errp, "invalid net client name: %s", name);
1318     }
1319
1320     return filter_list;
1321 }
1322
1323 NetdevInfoList *qmp_query_netdev(Error **errp)
1324 {
1325     NetdevInfoList *list = NULL;
1326     NetClientState *nc;
1327
1328     QTAILQ_FOREACH(nc, &net_clients, next) {
1329         /*
1330          * Only look at netdevs (backend network devices), not for each queue
1331          * or NIC / hubport
1332          */
1333         if (nc->stored_config) {
1334             NetdevInfo *element = QAPI_CLONE(NetdevInfo, nc->stored_config);
1335
1336             g_free(element->id); /* Need to dealloc empty id after clone */
1337             element->id = g_strdup(nc->name);
1338
1339             element->has_peer_id = nc->peer != NULL;
1340             if (element->has_peer_id) {
1341                 element->peer_id = g_strdup(nc->peer->name);
1342             }
1343
1344             QAPI_LIST_PREPEND(list, element);
1345         }
1346     }
1347
1348     return list;
1349 }
1350
1351 void hmp_info_network(Monitor *mon, const QDict *qdict)
1352 {
1353     NetClientState *nc, *peer;
1354     NetClientDriver type;
1355
1356     net_hub_info(mon);
1357
1358     QTAILQ_FOREACH(nc, &net_clients, next) {
1359         peer = nc->peer;
1360         type = nc->info->type;
1361
1362         /* Skip if already printed in hub info */
1363         if (net_hub_id_for_client(nc, NULL) == 0) {
1364             continue;
1365         }
1366
1367         if (!peer || type == NET_CLIENT_DRIVER_NIC) {
1368             print_net_client(mon, nc);
1369         } /* else it's a netdev connected to a NIC, printed with the NIC */
1370         if (peer && type == NET_CLIENT_DRIVER_NIC) {
1371             monitor_printf(mon, " \\ ");
1372             print_net_client(mon, peer);
1373         }
1374     }
1375 }
1376
1377 void colo_notify_filters_event(int event, Error **errp)
1378 {
1379     NetClientState *nc;
1380     NetFilterState *nf;
1381     NetFilterClass *nfc = NULL;
1382     Error *local_err = NULL;
1383
1384     QTAILQ_FOREACH(nc, &net_clients, next) {
1385         QTAILQ_FOREACH(nf, &nc->filters, next) {
1386             nfc = NETFILTER_GET_CLASS(OBJECT(nf));
1387             nfc->handle_event(nf, event, &local_err);
1388             if (local_err) {
1389                 error_propagate(errp, local_err);
1390                 return;
1391             }
1392         }
1393     }
1394 }
1395
1396 void qmp_set_link(const char *name, bool up, Error **errp)
1397 {
1398     NetClientState *ncs[MAX_QUEUE_NUM];
1399     NetClientState *nc;
1400     int queues, i;
1401
1402     queues = qemu_find_net_clients_except(name, ncs,
1403                                           NET_CLIENT_DRIVER__MAX,
1404                                           MAX_QUEUE_NUM);
1405
1406     if (queues == 0) {
1407         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1408                   "Device '%s' not found", name);
1409         return;
1410     }
1411     nc = ncs[0];
1412
1413     for (i = 0; i < queues; i++) {
1414         ncs[i]->link_down = !up;
1415     }
1416
1417     if (nc->info->link_status_changed) {
1418         nc->info->link_status_changed(nc);
1419     }
1420
1421     if (nc->peer) {
1422         /* Change peer link only if the peer is NIC and then notify peer.
1423          * If the peer is a HUBPORT or a backend, we do not change the
1424          * link status.
1425          *
1426          * This behavior is compatible with qemu hubs where there could be
1427          * multiple clients that can still communicate with each other in
1428          * disconnected mode. For now maintain this compatibility.
1429          */
1430         if (nc->peer->info->type == NET_CLIENT_DRIVER_NIC) {
1431             for (i = 0; i < queues; i++) {
1432                 ncs[i]->peer->link_down = !up;
1433             }
1434         }
1435         if (nc->peer->info->link_status_changed) {
1436             nc->peer->info->link_status_changed(nc->peer);
1437         }
1438     }
1439 }
1440
1441 static void net_vm_change_state_handler(void *opaque, bool running,
1442                                         RunState state)
1443 {
1444     NetClientState *nc;
1445     NetClientState *tmp;
1446
1447     QTAILQ_FOREACH_SAFE(nc, &net_clients, next, tmp) {
1448         if (running) {
1449             /* Flush queued packets and wake up backends. */
1450             if (nc->peer && qemu_can_send_packet(nc)) {
1451                 qemu_flush_queued_packets(nc->peer);
1452             }
1453         } else {
1454             /* Complete all queued packets, to guarantee we don't modify
1455              * state later when VM is not running.
1456              */
1457             qemu_flush_or_purge_queued_packets(nc, true);
1458         }
1459     }
1460 }
1461
1462 void net_cleanup(void)
1463 {
1464     NetClientState *nc;
1465
1466     /* We may del multiple entries during qemu_del_net_client(),
1467      * so QTAILQ_FOREACH_SAFE() is also not safe here.
1468      */
1469     while (!QTAILQ_EMPTY(&net_clients)) {
1470         nc = QTAILQ_FIRST(&net_clients);
1471         if (nc->info->type == NET_CLIENT_DRIVER_NIC) {
1472             qemu_del_nic(qemu_get_nic(nc));
1473         } else {
1474             qemu_del_net_client(nc);
1475         }
1476     }
1477
1478     qemu_del_vm_change_state_handler(net_change_state_entry);
1479 }
1480
1481 void net_check_clients(void)
1482 {
1483     NetClientState *nc;
1484     int i;
1485
1486     net_hub_check_clients();
1487
1488     QTAILQ_FOREACH(nc, &net_clients, next) {
1489         if (!nc->peer) {
1490             warn_report("%s %s has no peer",
1491                         nc->info->type == NET_CLIENT_DRIVER_NIC
1492                         ? "nic" : "netdev",
1493                         nc->name);
1494         }
1495     }
1496
1497     /* Check that all NICs requested via -net nic actually got created.
1498      * NICs created via -device don't need to be checked here because
1499      * they are always instantiated.
1500      */
1501     for (i = 0; i < MAX_NICS; i++) {
1502         NICInfo *nd = &nd_table[i];
1503         if (nd->used && !nd->instantiated) {
1504             warn_report("requested NIC (%s, model %s) "
1505                         "was not created (not supported by this machine?)",
1506                         nd->name ? nd->name : "anonymous",
1507                         nd->model ? nd->model : "unspecified");
1508         }
1509     }
1510 }
1511
1512 static int net_init_client(void *dummy, QemuOpts *opts, Error **errp)
1513 {
1514     return net_client_init(opts, false, errp);
1515 }
1516
1517 static int net_init_netdev(void *dummy, QemuOpts *opts, Error **errp)
1518 {
1519     const char *type = qemu_opt_get(opts, "type");
1520
1521     if (type && is_help_option(type)) {
1522         show_netdevs();
1523         exit(0);
1524     }
1525     return net_client_init(opts, true, errp);
1526 }
1527
1528 /* For the convenience "--nic" parameter */
1529 static int net_param_nic(void *dummy, QemuOpts *opts, Error **errp)
1530 {
1531     char *mac, *nd_id;
1532     int idx, ret;
1533     NICInfo *ni;
1534     const char *type;
1535
1536     type = qemu_opt_get(opts, "type");
1537     if (type && g_str_equal(type, "none")) {
1538         return 0;    /* Nothing to do, default_net is cleared in vl.c */
1539     }
1540
1541     idx = nic_get_free_idx();
1542     if (idx == -1 || nb_nics >= MAX_NICS) {
1543         error_setg(errp, "no more on-board/default NIC slots available");
1544         return -1;
1545     }
1546
1547     if (!type) {
1548         qemu_opt_set(opts, "type", "user", &error_abort);
1549     }
1550
1551     ni = &nd_table[idx];
1552     memset(ni, 0, sizeof(*ni));
1553     ni->model = qemu_opt_get_del(opts, "model");
1554
1555     /* Create an ID if the user did not specify one */
1556     nd_id = g_strdup(qemu_opts_id(opts));
1557     if (!nd_id) {
1558         nd_id = id_generate(ID_NET);
1559         qemu_opts_set_id(opts, nd_id);
1560     }
1561
1562     /* Handle MAC address */
1563     mac = qemu_opt_get_del(opts, "mac");
1564     if (mac) {
1565         ret = net_parse_macaddr(ni->macaddr.a, mac);
1566         g_free(mac);
1567         if (ret) {
1568             error_setg(errp, "invalid syntax for ethernet address");
1569             goto out;
1570         }
1571         if (is_multicast_ether_addr(ni->macaddr.a)) {
1572             error_setg(errp, "NIC cannot have multicast MAC address");
1573             ret = -1;
1574             goto out;
1575         }
1576     }
1577     qemu_macaddr_default_if_unset(&ni->macaddr);
1578
1579     ret = net_client_init(opts, true, errp);
1580     if (ret == 0) {
1581         ni->netdev = qemu_find_netdev(nd_id);
1582         ni->used = true;
1583         nb_nics++;
1584     }
1585
1586 out:
1587     g_free(nd_id);
1588     return ret;
1589 }
1590
1591 int net_init_clients(Error **errp)
1592 {
1593     net_change_state_entry =
1594         qemu_add_vm_change_state_handler(net_vm_change_state_handler, NULL);
1595
1596     QTAILQ_INIT(&net_clients);
1597
1598     if (qemu_opts_foreach(qemu_find_opts("netdev"),
1599                           net_init_netdev, NULL, errp)) {
1600         return -1;
1601     }
1602
1603     if (qemu_opts_foreach(qemu_find_opts("nic"), net_param_nic, NULL, errp)) {
1604         return -1;
1605     }
1606
1607     if (qemu_opts_foreach(qemu_find_opts("net"), net_init_client, NULL, errp)) {
1608         return -1;
1609     }
1610
1611     return 0;
1612 }
1613
1614 int net_client_parse(QemuOptsList *opts_list, const char *optarg)
1615 {
1616     if (!qemu_opts_parse_noisily(opts_list, optarg, true)) {
1617         return -1;
1618     }
1619
1620     return 0;
1621 }
1622
1623 /* From FreeBSD */
1624 /* XXX: optimize */
1625 uint32_t net_crc32(const uint8_t *p, int len)
1626 {
1627     uint32_t crc;
1628     int carry, i, j;
1629     uint8_t b;
1630
1631     crc = 0xffffffff;
1632     for (i = 0; i < len; i++) {
1633         b = *p++;
1634         for (j = 0; j < 8; j++) {
1635             carry = ((crc & 0x80000000L) ? 1 : 0) ^ (b & 0x01);
1636             crc <<= 1;
1637             b >>= 1;
1638             if (carry) {
1639                 crc = ((crc ^ POLYNOMIAL_BE) | carry);
1640             }
1641         }
1642     }
1643
1644     return crc;
1645 }
1646
1647 uint32_t net_crc32_le(const uint8_t *p, int len)
1648 {
1649     uint32_t crc;
1650     int carry, i, j;
1651     uint8_t b;
1652
1653     crc = 0xffffffff;
1654     for (i = 0; i < len; i++) {
1655         b = *p++;
1656         for (j = 0; j < 8; j++) {
1657             carry = (crc & 0x1) ^ (b & 0x01);
1658             crc >>= 1;
1659             b >>= 1;
1660             if (carry) {
1661                 crc ^= POLYNOMIAL_LE;
1662             }
1663         }
1664     }
1665
1666     return crc;
1667 }
1668
1669 QemuOptsList qemu_netdev_opts = {
1670     .name = "netdev",
1671     .implied_opt_name = "type",
1672     .head = QTAILQ_HEAD_INITIALIZER(qemu_netdev_opts.head),
1673     .desc = {
1674         /*
1675          * no elements => accept any params
1676          * validation will happen later
1677          */
1678         { /* end of list */ }
1679     },
1680 };
1681
1682 QemuOptsList qemu_nic_opts = {
1683     .name = "nic",
1684     .implied_opt_name = "type",
1685     .head = QTAILQ_HEAD_INITIALIZER(qemu_nic_opts.head),
1686     .desc = {
1687         /*
1688          * no elements => accept any params
1689          * validation will happen later
1690          */
1691         { /* end of list */ }
1692     },
1693 };
1694
1695 QemuOptsList qemu_net_opts = {
1696     .name = "net",
1697     .implied_opt_name = "type",
1698     .head = QTAILQ_HEAD_INITIALIZER(qemu_net_opts.head),
1699     .desc = {
1700         /*
1701          * no elements => accept any params
1702          * validation will happen later
1703          */
1704         { /* end of list */ }
1705     },
1706 };
1707
1708 void net_socket_rs_init(SocketReadState *rs,
1709                         SocketReadStateFinalize *finalize,
1710                         bool vnet_hdr)
1711 {
1712     rs->state = 0;
1713     rs->vnet_hdr = vnet_hdr;
1714     rs->index = 0;
1715     rs->packet_len = 0;
1716     rs->vnet_hdr_len = 0;
1717     memset(rs->buf, 0, sizeof(rs->buf));
1718     rs->finalize = finalize;
1719 }
1720
1721 /*
1722  * Returns
1723  * 0: success
1724  * -1: error occurs
1725  */
1726 int net_fill_rstate(SocketReadState *rs, const uint8_t *buf, int size)
1727 {
1728     unsigned int l;
1729
1730     while (size > 0) {
1731         /* Reassemble a packet from the network.
1732          * 0 = getting length.
1733          * 1 = getting vnet header length.
1734          * 2 = getting data.
1735          */
1736         switch (rs->state) {
1737         case 0:
1738             l = 4 - rs->index;
1739             if (l > size) {
1740                 l = size;
1741             }
1742             memcpy(rs->buf + rs->index, buf, l);
1743             buf += l;
1744             size -= l;
1745             rs->index += l;
1746             if (rs->index == 4) {
1747                 /* got length */
1748                 rs->packet_len = ntohl(*(uint32_t *)rs->buf);
1749                 rs->index = 0;
1750                 if (rs->vnet_hdr) {
1751                     rs->state = 1;
1752                 } else {
1753                     rs->state = 2;
1754                     rs->vnet_hdr_len = 0;
1755                 }
1756             }
1757             break;
1758         case 1:
1759             l = 4 - rs->index;
1760             if (l > size) {
1761                 l = size;
1762             }
1763             memcpy(rs->buf + rs->index, buf, l);
1764             buf += l;
1765             size -= l;
1766             rs->index += l;
1767             if (rs->index == 4) {
1768                 /* got vnet header length */
1769                 rs->vnet_hdr_len = ntohl(*(uint32_t *)rs->buf);
1770                 rs->index = 0;
1771                 rs->state = 2;
1772             }
1773             break;
1774         case 2:
1775             l = rs->packet_len - rs->index;
1776             if (l > size) {
1777                 l = size;
1778             }
1779             if (rs->index + l <= sizeof(rs->buf)) {
1780                 memcpy(rs->buf + rs->index, buf, l);
1781             } else {
1782                 fprintf(stderr, "serious error: oversized packet received,"
1783                     "connection terminated.\n");
1784                 rs->index = rs->state = 0;
1785                 return -1;
1786             }
1787
1788             rs->index += l;
1789             buf += l;
1790             size -= l;
1791             if (rs->index >= rs->packet_len) {
1792                 rs->index = 0;
1793                 rs->state = 0;
1794                 assert(rs->finalize);
1795                 rs->finalize(rs);
1796             }
1797             break;
1798         }
1799     }
1800
1801     assert(size == 0);
1802     return 0;
1803 }