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
  27 #include "net/net.h"
  28 #include "clients.h"
  29 #include "hub.h"
  30 #include "hw/qdev-properties.h"
  31 #include "net/slirp.h"
  32 #include "net/eth.h"
  33 #include "util.h"
  34
  35 #include "monitor/monitor.h"
  36 #include "qemu/help_option.h"
  37 #include "qapi/qapi-commands-net.h"
  38 #include "qapi/qapi-visit-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 "qemu/keyval.h"
  52 #include "qapi/error.h"
  53 #include "qapi/opts-visitor.h"
  54 #include "sysemu/runstate.h"
  55 #include "net/colo-compare.h"
  56 #include "net/filter.h"
  57 #include "qapi/string-output-visitor.h"
  58 #include "qapi/qobject-input-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 NetClientStateList net_clients;
  67
  68 typedef struct NetdevQueueEntry {
  69     Netdev *nd;
  70     Location loc;
  71     QSIMPLEQ_ENTRY(NetdevQueueEntry) entry;
  72 } NetdevQueueEntry;
  73
  74 typedef QSIMPLEQ_HEAD(, NetdevQueueEntry) NetdevQueue;
  75
  76 static NetdevQueue nd_queue = QSIMPLEQ_HEAD_INITIALIZER(nd_queue);
  77
  78 /***********************************************************/
  79 /* network device redirectors */
  80
  81 int convert_host_port(struct sockaddr_in *saddr, const char *host,
  82                       const char *port, Error **errp)
  83 {
  84     struct hostent *he;
  85     const char *r;
  86     long p;
  87
  88     memset(saddr, 0, sizeof(*saddr));
  89
  90     saddr->sin_family = AF_INET;
  91     if (host[0] == '\0') {
  92         saddr->sin_addr.s_addr = 0;
  93     } else {
  94         if (qemu_isdigit(host[0])) {
  95             if (!inet_aton(host, &saddr->sin_addr)) {
  96                 error_setg(errp, "host address '%s' is not a valid "
  97                            "IPv4 address", host);
  98                 return -1;
  99             }
 100         } else {
 101             he = gethostbyname(host);
 102             if (he == NULL) {
 103                 error_setg(errp, "can't resolve host address '%s'", host);
 104                 return -1;
 105             }
 106             saddr->sin_addr = *(struct in_addr *)he->h_addr;
 107         }
 108     }
 109     if (qemu_strtol(port, &r, 0, &p) != 0) {
 110         error_setg(errp, "port number '%s' is invalid", port);
 111         return -1;
 112     }
 113     saddr->sin_port = htons(p);
 114     return 0;
 115 }
 116
 117 int parse_host_port(struct sockaddr_in *saddr, const char *str,
 118                     Error **errp)
 119 {
 120     gchar **substrings;
 121     int ret;
 122
 123     substrings = g_strsplit(str, ":", 2);
 124     if (!substrings || !substrings[0] || !substrings[1]) {
 125         error_setg(errp, "host address '%s' doesn't contain ':' "
 126                    "separating host from port", str);
 127         ret = -1;
 128         goto out;
 129     }
 130
 131     ret = convert_host_port(saddr, substrings[0], substrings[1], errp);
 132
 133 out:
 134     g_strfreev(substrings);
 135     return ret;
 136 }
 137
 138 char *qemu_mac_strdup_printf(const uint8_t *macaddr)
 139 {
 140     return g_strdup_printf("%.2x:%.2x:%.2x:%.2x:%.2x:%.2x",
 141                            macaddr[0], macaddr[1], macaddr[2],
 142                            macaddr[3], macaddr[4], macaddr[5]);
 143 }
 144
 145 void qemu_set_info_str(NetClientState *nc, const char *fmt, ...)
 146 {
 147     va_list ap;
 148
 149     va_start(ap, fmt);
 150     vsnprintf(nc->info_str, sizeof(nc->info_str), fmt, ap);
 151     va_end(ap);
 152 }
 153
 154 void qemu_format_nic_info_str(NetClientState *nc, uint8_t macaddr[6])
 155 {
 156     qemu_set_info_str(nc, "model=%s,macaddr=%02x:%02x:%02x:%02x:%02x:%02x",
 157                       nc->model, macaddr[0], macaddr[1], macaddr[2],
 158                       macaddr[3], macaddr[4], macaddr[5]);
 159 }
 160
 161 static int mac_table[256] = {0};
 162
 163 static void qemu_macaddr_set_used(MACAddr *macaddr)
 164 {
 165     int index;
 166
 167     for (index = 0x56; index < 0xFF; index++) {
 168         if (macaddr->a[5] == index) {
 169             mac_table[index]++;
 170         }
 171     }
 172 }
 173
 174 static void qemu_macaddr_set_free(MACAddr *macaddr)
 175 {
 176     int index;
 177     static const MACAddr base = { .a = { 0x52, 0x54, 0x00, 0x12, 0x34, 0 } };
 178
 179     if (memcmp(macaddr->a, &base.a, (sizeof(base.a) - 1)) != 0) {
 180         return;
 181     }
 182     for (index = 0x56; index < 0xFF; index++) {
 183         if (macaddr->a[5] == index) {
 184             mac_table[index]--;
 185         }
 186     }
 187 }
 188
 189 static int qemu_macaddr_get_free(void)
 190 {
 191     int index;
 192
 193     for (index = 0x56; index < 0xFF; index++) {
 194         if (mac_table[index] == 0) {
 195             return index;
 196         }
 197     }
 198
 199     return -1;
 200 }
 201
 202 void qemu_macaddr_default_if_unset(MACAddr *macaddr)
 203 {
 204     static const MACAddr zero = { .a = { 0,0,0,0,0,0 } };
 205     static const MACAddr base = { .a = { 0x52, 0x54, 0x00, 0x12, 0x34, 0 } };
 206
 207     if (memcmp(macaddr, &zero, sizeof(zero)) != 0) {
 208         if (memcmp(macaddr->a, &base.a, (sizeof(base.a) - 1)) != 0) {
 209             return;
 210         } else {
 211             qemu_macaddr_set_used(macaddr);
 212             return;
 213         }
 214     }
 215
 216     macaddr->a[0] = 0x52;
 217     macaddr->a[1] = 0x54;
 218     macaddr->a[2] = 0x00;
 219     macaddr->a[3] = 0x12;
 220     macaddr->a[4] = 0x34;
 221     macaddr->a[5] = qemu_macaddr_get_free();
 222     qemu_macaddr_set_used(macaddr);
 223 }
 224
 225 /**
 226  * Generate a name for net client
 227  *
 228  * Only net clients created with the legacy -net option and NICs need this.
 229  */
 230 static char *assign_name(NetClientState *nc1, const char *model)
 231 {
 232     NetClientState *nc;
 233     int id = 0;
 234
 235     QTAILQ_FOREACH(nc, &net_clients, next) {
 236         if (nc == nc1) {
 237             continue;
 238         }
 239         if (strcmp(nc->model, model) == 0) {
 240             id++;
 241         }
 242     }
 243
 244     return g_strdup_printf("%s.%d", model, id);
 245 }
 246
 247 static void qemu_net_client_destructor(NetClientState *nc)
 248 {
 249     g_free(nc);
 250 }
 251 static ssize_t qemu_deliver_packet_iov(NetClientState *sender,
 252                                        unsigned flags,
 253                                        const struct iovec *iov,
 254                                        int iovcnt,
 255                                        void *opaque);
 256
 257 static void qemu_net_client_setup(NetClientState *nc,
 258                                   NetClientInfo *info,
 259                                   NetClientState *peer,
 260                                   const char *model,
 261                                   const char *name,
 262                                   NetClientDestructor *destructor,
 263                                   bool is_datapath)
 264 {
 265     nc->info = info;
 266     nc->model = g_strdup(model);
 267     if (name) {
 268         nc->name = g_strdup(name);
 269     } else {
 270         nc->name = assign_name(nc, model);
 271     }
 272
 273     if (peer) {
 274         assert(!peer->peer);
 275         nc->peer = peer;
 276         peer->peer = nc;
 277     }
 278     QTAILQ_INSERT_TAIL(&net_clients, nc, next);
 279
 280     nc->incoming_queue = qemu_new_net_queue(qemu_deliver_packet_iov, nc);
 281     nc->destructor = destructor;
 282     nc->is_datapath = is_datapath;
 283     QTAILQ_INIT(&nc->filters);
 284 }
 285
 286 NetClientState *qemu_new_net_client(NetClientInfo *info,
 287                                     NetClientState *peer,
 288                                     const char *model,
 289                                     const char *name)
 290 {
 291     NetClientState *nc;
 292
 293     assert(info->size >= sizeof(NetClientState));
 294
 295     nc = g_malloc0(info->size);
 296     qemu_net_client_setup(nc, info, peer, model, name,
 297                           qemu_net_client_destructor, true);
 298
 299     return nc;
 300 }
 301
 302 NetClientState *qemu_new_net_control_client(NetClientInfo *info,
 303                                             NetClientState *peer,
 304                                             const char *model,
 305                                             const char *name)
 306 {
 307     NetClientState *nc;
 308
 309     assert(info->size >= sizeof(NetClientState));
 310
 311     nc = g_malloc0(info->size);
 312     qemu_net_client_setup(nc, info, peer, model, name,
 313                           qemu_net_client_destructor, false);
 314
 315     return nc;
 316 }
 317
 318 NICState *qemu_new_nic(NetClientInfo *info,
 319                        NICConf *conf,
 320                        const char *model,
 321                        const char *name,
 322                        void *opaque)
 323 {
 324     NetClientState **peers = conf->peers.ncs;
 325     NICState *nic;
 326     int i, queues = MAX(1, conf->peers.queues);
 327
 328     assert(info->type == NET_CLIENT_DRIVER_NIC);
 329     assert(info->size >= sizeof(NICState));
 330
 331     nic = g_malloc0(info->size + sizeof(NetClientState) * queues);
 332     nic->ncs = (void *)nic + info->size;
 333     nic->conf = conf;
 334     nic->opaque = opaque;
 335
 336     for (i = 0; i < queues; i++) {
 337         qemu_net_client_setup(&nic->ncs[i], info, peers[i], model, name,
 338                               NULL, true);
 339         nic->ncs[i].queue_index = i;
 340     }
 341
 342     return nic;
 343 }
 344
 345 NetClientState *qemu_get_subqueue(NICState *nic, int queue_index)
 346 {
 347     return nic->ncs + queue_index;
 348 }
 349
 350 NetClientState *qemu_get_queue(NICState *nic)
 351 {
 352     return qemu_get_subqueue(nic, 0);
 353 }
 354
 355 NICState *qemu_get_nic(NetClientState *nc)
 356 {
 357     NetClientState *nc0 = nc - nc->queue_index;
 358
 359     return (NICState *)((void *)nc0 - nc->info->size);
 360 }
 361
 362 void *qemu_get_nic_opaque(NetClientState *nc)
 363 {
 364     NICState *nic = qemu_get_nic(nc);
 365
 366     return nic->opaque;
 367 }
 368
 369 NetClientState *qemu_get_peer(NetClientState *nc, int queue_index)
 370 {
 371     assert(nc != NULL);
 372     NetClientState *ncs = nc + queue_index;
 373     return ncs->peer;
 374 }
 375
 376 static void qemu_cleanup_net_client(NetClientState *nc)
 377 {
 378     QTAILQ_REMOVE(&net_clients, nc, next);
 379
 380     if (nc->info->cleanup) {
 381         nc->info->cleanup(nc);
 382     }
 383 }
 384
 385 static void qemu_free_net_client(NetClientState *nc)
 386 {
 387     if (nc->incoming_queue) {
 388         qemu_del_net_queue(nc->incoming_queue);
 389     }
 390     if (nc->peer) {
 391         nc->peer->peer = NULL;
 392     }
 393     g_free(nc->name);
 394     g_free(nc->model);
 395     if (nc->destructor) {
 396         nc->destructor(nc);
 397     }
 398 }
 399
 400 void qemu_del_net_client(NetClientState *nc)
 401 {
 402     NetClientState *ncs[MAX_QUEUE_NUM];
 403     int queues, i;
 404     NetFilterState *nf, *next;
 405
 406     assert(nc->info->type != NET_CLIENT_DRIVER_NIC);
 407
 408     /* If the NetClientState belongs to a multiqueue backend, we will change all
 409      * other NetClientStates also.
 410      */
 411     queues = qemu_find_net_clients_except(nc->name, ncs,
 412                                           NET_CLIENT_DRIVER_NIC,
 413                                           MAX_QUEUE_NUM);
 414     assert(queues != 0);
 415
 416     QTAILQ_FOREACH_SAFE(nf, &nc->filters, next, next) {
 417         object_unparent(OBJECT(nf));
 418     }
 419
 420     /* If there is a peer NIC, delete and cleanup client, but do not free. */
 421     if (nc->peer && nc->peer->info->type == NET_CLIENT_DRIVER_NIC) {
 422         NICState *nic = qemu_get_nic(nc->peer);
 423         if (nic->peer_deleted) {
 424             return;
 425         }
 426         nic->peer_deleted = true;
 427
 428         for (i = 0; i < queues; i++) {
 429             ncs[i]->peer->link_down = true;
 430         }
 431
 432         if (nc->peer->info->link_status_changed) {
 433             nc->peer->info->link_status_changed(nc->peer);
 434         }
 435
 436         for (i = 0; i < queues; i++) {
 437             qemu_cleanup_net_client(ncs[i]);
 438         }
 439
 440         return;
 441     }
 442
 443     for (i = 0; i < queues; i++) {
 444         qemu_cleanup_net_client(ncs[i]);
 445         qemu_free_net_client(ncs[i]);
 446     }
 447 }
 448
 449 void qemu_del_nic(NICState *nic)
 450 {
 451     int i, queues = MAX(nic->conf->peers.queues, 1);
 452
 453     qemu_macaddr_set_free(&nic->conf->macaddr);
 454
 455     for (i = 0; i < queues; i++) {
 456         NetClientState *nc = qemu_get_subqueue(nic, i);
 457         /* If this is a peer NIC and peer has already been deleted, free it now. */
 458         if (nic->peer_deleted) {
 459             qemu_free_net_client(nc->peer);
 460         } else if (nc->peer) {
 461             /* if there are RX packets pending, complete them */
 462             qemu_purge_queued_packets(nc->peer);
 463         }
 464     }
 465
 466     for (i = queues - 1; i >= 0; i--) {
 467         NetClientState *nc = qemu_get_subqueue(nic, i);
 468
 469         qemu_cleanup_net_client(nc);
 470         qemu_free_net_client(nc);
 471     }
 472
 473     g_free(nic);
 474 }
 475
 476 void qemu_foreach_nic(qemu_nic_foreach func, void *opaque)
 477 {
 478     NetClientState *nc;
 479
 480     QTAILQ_FOREACH(nc, &net_clients, next) {
 481         if (nc->info->type == NET_CLIENT_DRIVER_NIC) {
 482             if (nc->queue_index == 0) {
 483                 func(qemu_get_nic(nc), opaque);
 484             }
 485         }
 486     }
 487 }
 488
 489 bool qemu_has_ufo(NetClientState *nc)
 490 {
 491     if (!nc || !nc->info->has_ufo) {
 492         return false;
 493     }
 494
 495     return nc->info->has_ufo(nc);
 496 }
 497
 498 bool qemu_has_vnet_hdr(NetClientState *nc)
 499 {
 500     if (!nc || !nc->info->has_vnet_hdr) {
 501         return false;
 502     }
 503
 504     return nc->info->has_vnet_hdr(nc);
 505 }
 506
 507 bool qemu_has_vnet_hdr_len(NetClientState *nc, int len)
 508 {
 509     if (!nc || !nc->info->has_vnet_hdr_len) {
 510         return false;
 511     }
 512
 513     return nc->info->has_vnet_hdr_len(nc, len);
 514 }
 515
 516 void qemu_using_vnet_hdr(NetClientState *nc, bool enable)
 517 {
 518     if (!nc || !nc->info->using_vnet_hdr) {
 519         return;
 520     }
 521
 522     nc->info->using_vnet_hdr(nc, enable);
 523 }
 524
 525 void qemu_set_offload(NetClientState *nc, int csum, int tso4, int tso6,
 526                           int ecn, int ufo)
 527 {
 528     if (!nc || !nc->info->set_offload) {
 529         return;
 530     }
 531
 532     nc->info->set_offload(nc, csum, tso4, tso6, ecn, ufo);
 533 }
 534
 535 void qemu_set_vnet_hdr_len(NetClientState *nc, int len)
 536 {
 537     if (!nc || !nc->info->set_vnet_hdr_len) {
 538         return;
 539     }
 540
 541     nc->vnet_hdr_len = len;
 542     nc->info->set_vnet_hdr_len(nc, len);
 543 }
 544
 545 int qemu_set_vnet_le(NetClientState *nc, bool is_le)
 546 {
 547 #if HOST_BIG_ENDIAN
 548     if (!nc || !nc->info->set_vnet_le) {
 549         return -ENOSYS;
 550     }
 551
 552     return nc->info->set_vnet_le(nc, is_le);
 553 #else
 554     return 0;
 555 #endif
 556 }
 557
 558 int qemu_set_vnet_be(NetClientState *nc, bool is_be)
 559 {
 560 #if HOST_BIG_ENDIAN
 561     return 0;
 562 #else
 563     if (!nc || !nc->info->set_vnet_be) {
 564         return -ENOSYS;
 565     }
 566
 567     return nc->info->set_vnet_be(nc, is_be);
 568 #endif
 569 }
 570
 571 int qemu_can_receive_packet(NetClientState *nc)
 572 {
 573     if (nc->receive_disabled) {
 574         return 0;
 575     } else if (nc->info->can_receive &&
 576                !nc->info->can_receive(nc)) {
 577         return 0;
 578     }
 579     return 1;
 580 }
 581
 582 int qemu_can_send_packet(NetClientState *sender)
 583 {
 584     int vm_running = runstate_is_running();
 585
 586     if (!vm_running) {
 587         return 0;
 588     }
 589
 590     if (!sender->peer) {
 591         return 1;
 592     }
 593
 594     return qemu_can_receive_packet(sender->peer);
 595 }
 596
 597 static ssize_t filter_receive_iov(NetClientState *nc,
 598                                   NetFilterDirection direction,
 599                                   NetClientState *sender,
 600                                   unsigned flags,
 601                                   const struct iovec *iov,
 602                                   int iovcnt,
 603                                   NetPacketSent *sent_cb)
 604 {
 605     ssize_t ret = 0;
 606     NetFilterState *nf = NULL;
 607
 608     if (direction == NET_FILTER_DIRECTION_TX) {
 609         QTAILQ_FOREACH(nf, &nc->filters, next) {
 610             ret = qemu_netfilter_receive(nf, direction, sender, flags, iov,
 611                                          iovcnt, sent_cb);
 612             if (ret) {
 613                 return ret;
 614             }
 615         }
 616     } else {
 617         QTAILQ_FOREACH_REVERSE(nf, &nc->filters, next) {
 618             ret = qemu_netfilter_receive(nf, direction, sender, flags, iov,
 619                                          iovcnt, sent_cb);
 620             if (ret) {
 621                 return ret;
 622             }
 623         }
 624     }
 625
 626     return ret;
 627 }
 628
 629 static ssize_t filter_receive(NetClientState *nc,
 630                               NetFilterDirection direction,
 631                               NetClientState *sender,
 632                               unsigned flags,
 633                               const uint8_t *data,
 634                               size_t size,
 635                               NetPacketSent *sent_cb)
 636 {
 637     struct iovec iov = {
 638         .iov_base = (void *)data,
 639         .iov_len = size
 640     };
 641
 642     return filter_receive_iov(nc, direction, sender, flags, &iov, 1, sent_cb);
 643 }
 644
 645 void qemu_purge_queued_packets(NetClientState *nc)
 646 {
 647     if (!nc->peer) {
 648         return;
 649     }
 650
 651     qemu_net_queue_purge(nc->peer->incoming_queue, nc);
 652 }
 653
 654 void qemu_flush_or_purge_queued_packets(NetClientState *nc, bool purge)
 655 {
 656     nc->receive_disabled = 0;
 657
 658     if (nc->peer && nc->peer->info->type == NET_CLIENT_DRIVER_HUBPORT) {
 659         if (net_hub_flush(nc->peer)) {
 660             qemu_notify_event();
 661         }
 662     }
 663     if (qemu_net_queue_flush(nc->incoming_queue)) {
 664         /* We emptied the queue successfully, signal to the IO thread to repoll
 665          * the file descriptor (for tap, for example).
 666          */
 667         qemu_notify_event();
 668     } else if (purge) {
 669         /* Unable to empty the queue, purge remaining packets */
 670         qemu_net_queue_purge(nc->incoming_queue, nc->peer);
 671     }
 672 }
 673
 674 void qemu_flush_queued_packets(NetClientState *nc)
 675 {
 676     qemu_flush_or_purge_queued_packets(nc, false);
 677 }
 678
 679 static ssize_t qemu_send_packet_async_with_flags(NetClientState *sender,
 680                                                  unsigned flags,
 681                                                  const uint8_t *buf, int size,
 682                                                  NetPacketSent *sent_cb)
 683 {
 684     NetQueue *queue;
 685     int ret;
 686
 687 #ifdef DEBUG_NET
 688     printf("qemu_send_packet_async:\n");
 689     qemu_hexdump(stdout, "net", buf, size);
 690 #endif
 691
 692     if (sender->link_down || !sender->peer) {
 693         return size;
 694     }
 695
 696     /* Let filters handle the packet first */
 697     ret = filter_receive(sender, NET_FILTER_DIRECTION_TX,
 698                          sender, flags, buf, size, sent_cb);
 699     if (ret) {
 700         return ret;
 701     }
 702
 703     ret = filter_receive(sender->peer, NET_FILTER_DIRECTION_RX,
 704                          sender, flags, buf, size, sent_cb);
 705     if (ret) {
 706         return ret;
 707     }
 708
 709     queue = sender->peer->incoming_queue;
 710
 711     return qemu_net_queue_send(queue, sender, flags, buf, size, sent_cb);
 712 }
 713
 714 ssize_t qemu_send_packet_async(NetClientState *sender,
 715                                const uint8_t *buf, int size,
 716                                NetPacketSent *sent_cb)
 717 {
 718     return qemu_send_packet_async_with_flags(sender, QEMU_NET_PACKET_FLAG_NONE,
 719                                              buf, size, sent_cb);
 720 }
 721
 722 ssize_t qemu_send_packet(NetClientState *nc, const uint8_t *buf, int size)
 723 {
 724     return qemu_send_packet_async(nc, buf, size, NULL);
 725 }
 726
 727 ssize_t qemu_receive_packet(NetClientState *nc, const uint8_t *buf, int size)
 728 {
 729     if (!qemu_can_receive_packet(nc)) {
 730         return 0;
 731     }
 732
 733     return qemu_net_queue_receive(nc->incoming_queue, buf, size);
 734 }
 735
 736 ssize_t qemu_receive_packet_iov(NetClientState *nc, const struct iovec *iov,
 737                                 int iovcnt)
 738 {
 739     if (!qemu_can_receive_packet(nc)) {
 740         return 0;
 741     }
 742
 743     return qemu_net_queue_receive_iov(nc->incoming_queue, iov, iovcnt);
 744 }
 745
 746 ssize_t qemu_send_packet_raw(NetClientState *nc, const uint8_t *buf, int size)
 747 {
 748     return qemu_send_packet_async_with_flags(nc, QEMU_NET_PACKET_FLAG_RAW,
 749                                              buf, size, NULL);
 750 }
 751
 752 static ssize_t nc_sendv_compat(NetClientState *nc, const struct iovec *iov,
 753                                int iovcnt, unsigned flags)
 754 {
 755     uint8_t *buf = NULL;
 756     uint8_t *buffer;
 757     size_t offset;
 758     ssize_t ret;
 759
 760     if (iovcnt == 1) {
 761         buffer = iov[0].iov_base;
 762         offset = iov[0].iov_len;
 763     } else {
 764         offset = iov_size(iov, iovcnt);
 765         if (offset > NET_BUFSIZE) {
 766             return -1;
 767         }
 768         buf = g_malloc(offset);
 769         buffer = buf;
 770         offset = iov_to_buf(iov, iovcnt, 0, buf, offset);
 771     }
 772
 773     if (flags & QEMU_NET_PACKET_FLAG_RAW && nc->info->receive_raw) {
 774         ret = nc->info->receive_raw(nc, buffer, offset);
 775     } else {
 776         ret = nc->info->receive(nc, buffer, offset);
 777     }
 778
 779     g_free(buf);
 780     return ret;
 781 }
 782
 783 static ssize_t qemu_deliver_packet_iov(NetClientState *sender,
 784                                        unsigned flags,
 785                                        const struct iovec *iov,
 786                                        int iovcnt,
 787                                        void *opaque)
 788 {
 789     NetClientState *nc = opaque;
 790     int ret;
 791
 792
 793     if (nc->link_down) {
 794         return iov_size(iov, iovcnt);
 795     }
 796
 797     if (nc->receive_disabled) {
 798         return 0;
 799     }
 800
 801     if (nc->info->receive_iov && !(flags & QEMU_NET_PACKET_FLAG_RAW)) {
 802         ret = nc->info->receive_iov(nc, iov, iovcnt);
 803     } else {
 804         ret = nc_sendv_compat(nc, iov, iovcnt, flags);
 805     }
 806
 807     if (ret == 0) {
 808         nc->receive_disabled = 1;
 809     }
 810
 811     return ret;
 812 }
 813
 814 ssize_t qemu_sendv_packet_async(NetClientState *sender,
 815                                 const struct iovec *iov, int iovcnt,
 816                                 NetPacketSent *sent_cb)
 817 {
 818     NetQueue *queue;
 819     size_t size = iov_size(iov, iovcnt);
 820     int ret;
 821
 822     if (size > NET_BUFSIZE) {
 823         return size;
 824     }
 825
 826     if (sender->link_down || !sender->peer) {
 827         return size;
 828     }
 829
 830     /* Let filters handle the packet first */
 831     ret = filter_receive_iov(sender, NET_FILTER_DIRECTION_TX, sender,
 832                              QEMU_NET_PACKET_FLAG_NONE, iov, iovcnt, sent_cb);
 833     if (ret) {
 834         return ret;
 835     }
 836
 837     ret = filter_receive_iov(sender->peer, NET_FILTER_DIRECTION_RX, sender,
 838                              QEMU_NET_PACKET_FLAG_NONE, iov, iovcnt, sent_cb);
 839     if (ret) {
 840         return ret;
 841     }
 842
 843     queue = sender->peer->incoming_queue;
 844
 845     return qemu_net_queue_send_iov(queue, sender,
 846                                    QEMU_NET_PACKET_FLAG_NONE,
 847                                    iov, iovcnt, sent_cb);
 848 }
 849
 850 ssize_t
 851 qemu_sendv_packet(NetClientState *nc, const struct iovec *iov, int iovcnt)
 852 {
 853     return qemu_sendv_packet_async(nc, iov, iovcnt, NULL);
 854 }
 855
 856 NetClientState *qemu_find_netdev(const char *id)
 857 {
 858     NetClientState *nc;
 859
 860     QTAILQ_FOREACH(nc, &net_clients, next) {
 861         if (nc->info->type == NET_CLIENT_DRIVER_NIC)
 862             continue;
 863         if (!strcmp(nc->name, id)) {
 864             return nc;
 865         }
 866     }
 867
 868     return NULL;
 869 }
 870
 871 int qemu_find_net_clients_except(const char *id, NetClientState **ncs,
 872                                  NetClientDriver type, int max)
 873 {
 874     NetClientState *nc;
 875     int ret = 0;
 876
 877     QTAILQ_FOREACH(nc, &net_clients, next) {
 878         if (nc->info->type == type) {
 879             continue;
 880         }
 881         if (!id || !strcmp(nc->name, id)) {
 882             if (ret < max) {
 883                 ncs[ret] = nc;
 884             }
 885             ret++;
 886         }
 887     }
 888
 889     return ret;
 890 }
 891
 892 static int nic_get_free_idx(void)
 893 {
 894     int index;
 895
 896     for (index = 0; index < MAX_NICS; index++)
 897         if (!nd_table[index].used)
 898             return index;
 899     return -1;
 900 }
 901
 902 GPtrArray *qemu_get_nic_models(const char *device_type)
 903 {
 904     GPtrArray *nic_models = g_ptr_array_new();
 905     GSList *list = object_class_get_list_sorted(device_type, false);
 906
 907     while (list) {
 908         DeviceClass *dc = OBJECT_CLASS_CHECK(DeviceClass, list->data,
 909                                              TYPE_DEVICE);
 910         GSList *next;
 911         if (test_bit(DEVICE_CATEGORY_NETWORK, dc->categories) &&
 912             dc->user_creatable) {
 913             const char *name = object_class_get_name(list->data);
 914             /*
 915              * A network device might also be something else than a NIC, see
 916              * e.g. the "rocker" device. Thus we have to look for the "netdev"
 917              * property, too. Unfortunately, some devices like virtio-net only
 918              * create this property during instance_init, so we have to create
 919              * a temporary instance here to be able to check it.
 920              */
 921             Object *obj = object_new_with_class(OBJECT_CLASS(dc));
 922             if (object_property_find(obj, "netdev")) {
 923                 g_ptr_array_add(nic_models, (gpointer)name);
 924             }
 925             object_unref(obj);
 926         }
 927         next = list->next;
 928         g_slist_free_1(list);
 929         list = next;
 930     }
 931     g_ptr_array_add(nic_models, NULL);
 932
 933     return nic_models;
 934 }
 935
 936 int qemu_show_nic_models(const char *arg, const char *const *models)
 937 {
 938     int i;
 939
 940     if (!arg || !is_help_option(arg)) {
 941         return 0;
 942     }
 943
 944     printf("Available NIC models:\n");
 945     for (i = 0 ; models[i]; i++) {
 946         printf("%s\n", models[i]);
 947     }
 948     return 1;
 949 }
 950
 951 void qemu_check_nic_model(NICInfo *nd, const char *model)
 952 {
 953     const char *models[2];
 954
 955     models[0] = model;
 956     models[1] = NULL;
 957
 958     if (qemu_show_nic_models(nd->model, models))
 959         exit(0);
 960     if (qemu_find_nic_model(nd, models, model) < 0)
 961         exit(1);
 962 }
 963
 964 int qemu_find_nic_model(NICInfo *nd, const char * const *models,
 965                         const char *default_model)
 966 {
 967     int i;
 968
 969     if (!nd->model)
 970         nd->model = g_strdup(default_model);
 971
 972     for (i = 0 ; models[i]; i++) {
 973         if (strcmp(nd->model, models[i]) == 0)
 974             return i;
 975     }
 976
 977     error_report("Unsupported NIC model: %s", nd->model);
 978     return -1;
 979 }
 980
 981 static int net_init_nic(const Netdev *netdev, const char *name,
 982                         NetClientState *peer, Error **errp)
 983 {
 984     int idx;
 985     NICInfo *nd;
 986     const NetLegacyNicOptions *nic;
 987
 988     assert(netdev->type == NET_CLIENT_DRIVER_NIC);
 989     nic = &netdev->u.nic;
 990
 991     idx = nic_get_free_idx();
 992     if (idx == -1 || nb_nics >= MAX_NICS) {
 993         error_setg(errp, "too many NICs");
 994         return -1;
 995     }
 996
 997     nd = &nd_table[idx];
 998
 999     memset(nd, 0, sizeof(*nd));
1000
1001     if (nic->netdev) {
1002         nd->netdev = qemu_find_netdev(nic->netdev);
1003         if (!nd->netdev) {
1004             error_setg(errp, "netdev '%s' not found", nic->netdev);
1005             return -1;
1006         }
1007     } else {
1008         assert(peer);
1009         nd->netdev = peer;
1010     }
1011     nd->name = g_strdup(name);
1012     if (nic->model) {
1013         nd->model = g_strdup(nic->model);
1014     }
1015     if (nic->addr) {
1016         nd->devaddr = g_strdup(nic->addr);
1017     }
1018
1019     if (nic->macaddr &&
1020         net_parse_macaddr(nd->macaddr.a, nic->macaddr) < 0) {
1021         error_setg(errp, "invalid syntax for ethernet address");
1022         return -1;
1023     }
1024     if (nic->macaddr &&
1025         is_multicast_ether_addr(nd->macaddr.a)) {
1026         error_setg(errp,
1027                    "NIC cannot have multicast MAC address (odd 1st byte)");
1028         return -1;
1029     }
1030     qemu_macaddr_default_if_unset(&nd->macaddr);
1031
1032     if (nic->has_vectors) {
1033         if (nic->vectors > 0x7ffffff) {
1034             error_setg(errp, "invalid # of vectors: %"PRIu32, nic->vectors);
1035             return -1;
1036         }
1037         nd->nvectors = nic->vectors;
1038     } else {
1039         nd->nvectors = DEV_NVECTORS_UNSPECIFIED;
1040     }
1041
1042     nd->used = 1;
1043     nb_nics++;
1044
1045     return idx;
1046 }
1047
1048
1049 static int (* const net_client_init_fun[NET_CLIENT_DRIVER__MAX])(
1050     const Netdev *netdev,
1051     const char *name,
1052     NetClientState *peer, Error **errp) = {
1053         [NET_CLIENT_DRIVER_NIC]       = net_init_nic,
1054 #ifdef CONFIG_SLIRP
1055         [NET_CLIENT_DRIVER_USER]      = net_init_slirp,
1056 #endif
1057         [NET_CLIENT_DRIVER_TAP]       = net_init_tap,
1058         [NET_CLIENT_DRIVER_SOCKET]    = net_init_socket,
1059         [NET_CLIENT_DRIVER_STREAM]    = net_init_stream,
1060         [NET_CLIENT_DRIVER_DGRAM]     = net_init_dgram,
1061 #ifdef CONFIG_VDE
1062         [NET_CLIENT_DRIVER_VDE]       = net_init_vde,
1063 #endif
1064 #ifdef CONFIG_NETMAP
1065         [NET_CLIENT_DRIVER_NETMAP]    = net_init_netmap,
1066 #endif
1067 #ifdef CONFIG_NET_BRIDGE
1068         [NET_CLIENT_DRIVER_BRIDGE]    = net_init_bridge,
1069 #endif
1070         [NET_CLIENT_DRIVER_HUBPORT]   = net_init_hubport,
1071 #ifdef CONFIG_VHOST_NET_USER
1072         [NET_CLIENT_DRIVER_VHOST_USER] = net_init_vhost_user,
1073 #endif
1074 #ifdef CONFIG_VHOST_NET_VDPA
1075         [NET_CLIENT_DRIVER_VHOST_VDPA] = net_init_vhost_vdpa,
1076 #endif
1077 #ifdef CONFIG_L2TPV3
1078         [NET_CLIENT_DRIVER_L2TPV3]    = net_init_l2tpv3,
1079 #endif
1080 #ifdef CONFIG_VMNET
1081         [NET_CLIENT_DRIVER_VMNET_HOST] = net_init_vmnet_host,
1082         [NET_CLIENT_DRIVER_VMNET_SHARED] = net_init_vmnet_shared,
1083         [NET_CLIENT_DRIVER_VMNET_BRIDGED] = net_init_vmnet_bridged,
1084 #endif /* CONFIG_VMNET */
1085 };
1086
1087
1088 static int net_client_init1(const Netdev *netdev, bool is_netdev, Error **errp)
1089 {
1090     NetClientState *peer = NULL;
1091     NetClientState *nc;
1092
1093     if (is_netdev) {
1094         if (netdev->type == NET_CLIENT_DRIVER_NIC ||
1095             !net_client_init_fun[netdev->type]) {
1096             error_setg(errp, "network backend '%s' is not compiled into this binary",
1097                        NetClientDriver_str(netdev->type));
1098             return -1;
1099         }
1100     } else {
1101         if (netdev->type == NET_CLIENT_DRIVER_NONE) {
1102             return 0; /* nothing to do */
1103         }
1104         if (netdev->type == NET_CLIENT_DRIVER_HUBPORT) {
1105             error_setg(errp, "network backend '%s' is only supported with -netdev/-nic",
1106                        NetClientDriver_str(netdev->type));
1107             return -1;
1108         }
1109
1110         if (!net_client_init_fun[netdev->type]) {
1111             error_setg(errp, "network backend '%s' is not compiled into this binary",
1112                        NetClientDriver_str(netdev->type));
1113             return -1;
1114         }
1115
1116         /* Do not add to a hub if it's a nic with a netdev= parameter. */
1117         if (netdev->type != NET_CLIENT_DRIVER_NIC ||
1118             !netdev->u.nic.netdev) {
1119             peer = net_hub_add_port(0, NULL, NULL);
1120         }
1121     }
1122
1123     nc = qemu_find_netdev(netdev->id);
1124     if (nc) {
1125         error_setg(errp, "Duplicate ID '%s'", netdev->id);
1126         return -1;
1127     }
1128
1129     if (net_client_init_fun[netdev->type](netdev, netdev->id, peer, errp) < 0) {
1130         /* FIXME drop when all init functions store an Error */
1131         if (errp && !*errp) {
1132             error_setg(errp, "Device '%s' could not be initialized",
1133                        NetClientDriver_str(netdev->type));
1134         }
1135         return -1;
1136     }
1137
1138     if (is_netdev) {
1139         nc = qemu_find_netdev(netdev->id);
1140         assert(nc);
1141         nc->is_netdev = true;
1142     }
1143
1144     return 0;
1145 }
1146
1147 void show_netdevs(void)
1148 {
1149     int idx;
1150     const char *available_netdevs[] = {
1151         "socket",
1152         "stream",
1153         "dgram",
1154         "hubport",
1155         "tap",
1156 #ifdef CONFIG_SLIRP
1157         "user",
1158 #endif
1159 #ifdef CONFIG_L2TPV3
1160         "l2tpv3",
1161 #endif
1162 #ifdef CONFIG_VDE
1163         "vde",
1164 #endif
1165 #ifdef CONFIG_NET_BRIDGE
1166         "bridge",
1167 #endif
1168 #ifdef CONFIG_NETMAP
1169         "netmap",
1170 #endif
1171 #ifdef CONFIG_POSIX
1172         "vhost-user",
1173 #endif
1174 #ifdef CONFIG_VHOST_VDPA
1175         "vhost-vdpa",
1176 #endif
1177 #ifdef CONFIG_VMNET
1178         "vmnet-host",
1179         "vmnet-shared",
1180         "vmnet-bridged",
1181 #endif
1182     };
1183
1184     qemu_printf("Available netdev backend types:\n");
1185     for (idx = 0; idx < ARRAY_SIZE(available_netdevs); idx++) {
1186         qemu_printf("%s\n", available_netdevs[idx]);
1187     }
1188 }
1189
1190 static int net_client_init(QemuOpts *opts, bool is_netdev, Error **errp)
1191 {
1192     gchar **substrings = NULL;
1193     Netdev *object = NULL;
1194     int ret = -1;
1195     Visitor *v = opts_visitor_new(opts);
1196
1197     /* Parse convenience option format ip6-net=fec0::0[/64] */
1198     const char *ip6_net = qemu_opt_get(opts, "ipv6-net");
1199
1200     if (ip6_net) {
1201         char *prefix_addr;
1202         unsigned long prefix_len = 64; /* Default 64bit prefix length. */
1203
1204         substrings = g_strsplit(ip6_net, "/", 2);
1205         if (!substrings || !substrings[0]) {
1206             error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "ipv6-net",
1207                        "a valid IPv6 prefix");
1208             goto out;
1209         }
1210
1211         prefix_addr = substrings[0];
1212
1213         /* Handle user-specified prefix length. */
1214         if (substrings[1] &&
1215             qemu_strtoul(substrings[1], NULL, 10, &prefix_len))
1216         {
1217             error_setg(errp, QERR_INVALID_PARAMETER_VALUE,
1218                        "ipv6-prefixlen", "a number");
1219             goto out;
1220         }
1221
1222         qemu_opt_set(opts, "ipv6-prefix", prefix_addr, &error_abort);
1223         qemu_opt_set_number(opts, "ipv6-prefixlen", prefix_len,
1224                             &error_abort);
1225         qemu_opt_unset(opts, "ipv6-net");
1226     }
1227
1228     /* Create an ID for -net if the user did not specify one */
1229     if (!is_netdev && !qemu_opts_id(opts)) {
1230         qemu_opts_set_id(opts, id_generate(ID_NET));
1231     }
1232
1233     if (visit_type_Netdev(v, NULL, &object, errp)) {
1234         ret = net_client_init1(object, is_netdev, errp);
1235     }
1236
1237     qapi_free_Netdev(object);
1238
1239 out:
1240     g_strfreev(substrings);
1241     visit_free(v);
1242     return ret;
1243 }
1244
1245 void netdev_add(QemuOpts *opts, Error **errp)
1246 {
1247     net_client_init(opts, true, errp);
1248 }
1249
1250 void qmp_netdev_add(Netdev *netdev, Error **errp)
1251 {
1252     if (!id_wellformed(netdev->id)) {
1253         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "id", "an identifier");
1254         return;
1255     }
1256
1257     net_client_init1(netdev, true, errp);
1258 }
1259
1260 void qmp_netdev_del(const char *id, Error **errp)
1261 {
1262     NetClientState *nc;
1263     QemuOpts *opts;
1264
1265     nc = qemu_find_netdev(id);
1266     if (!nc) {
1267         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1268                   "Device '%s' not found", id);
1269         return;
1270     }
1271
1272     if (!nc->is_netdev) {
1273         error_setg(errp, "Device '%s' is not a netdev", id);
1274         return;
1275     }
1276
1277     qemu_del_net_client(nc);
1278
1279     /*
1280      * Wart: we need to delete the QemuOpts associated with netdevs
1281      * created via CLI or HMP, to avoid bogus "Duplicate ID" errors in
1282      * HMP netdev_add.
1283      */
1284     opts = qemu_opts_find(qemu_find_opts("netdev"), id);
1285     if (opts) {
1286         qemu_opts_del(opts);
1287     }
1288 }
1289
1290 static void netfilter_print_info(Monitor *mon, NetFilterState *nf)
1291 {
1292     char *str;
1293     ObjectProperty *prop;
1294     ObjectPropertyIterator iter;
1295     Visitor *v;
1296
1297     /* generate info str */
1298     object_property_iter_init(&iter, OBJECT(nf));
1299     while ((prop = object_property_iter_next(&iter))) {
1300         if (!strcmp(prop->name, "type")) {
1301             continue;
1302         }
1303         v = string_output_visitor_new(false, &str);
1304         object_property_get(OBJECT(nf), prop->name, v, NULL);
1305         visit_complete(v, &str);
1306         visit_free(v);
1307         monitor_printf(mon, ",%s=%s", prop->name, str);
1308         g_free(str);
1309     }
1310     monitor_printf(mon, "\n");
1311 }
1312
1313 void print_net_client(Monitor *mon, NetClientState *nc)
1314 {
1315     NetFilterState *nf;
1316
1317     monitor_printf(mon, "%s: index=%d,type=%s,%s\n", nc->name,
1318                    nc->queue_index,
1319                    NetClientDriver_str(nc->info->type),
1320                    nc->info_str);
1321     if (!QTAILQ_EMPTY(&nc->filters)) {
1322         monitor_printf(mon, "filters:\n");
1323     }
1324     QTAILQ_FOREACH(nf, &nc->filters, next) {
1325         monitor_printf(mon, "  - %s: type=%s",
1326                        object_get_canonical_path_component(OBJECT(nf)),
1327                        object_get_typename(OBJECT(nf)));
1328         netfilter_print_info(mon, nf);
1329     }
1330 }
1331
1332 RxFilterInfoList *qmp_query_rx_filter(const char *name, Error **errp)
1333 {
1334     NetClientState *nc;
1335     RxFilterInfoList *filter_list = NULL, **tail = &filter_list;
1336
1337     QTAILQ_FOREACH(nc, &net_clients, next) {
1338         RxFilterInfo *info;
1339
1340         if (name && strcmp(nc->name, name) != 0) {
1341             continue;
1342         }
1343
1344         /* only query rx-filter information of NIC */
1345         if (nc->info->type != NET_CLIENT_DRIVER_NIC) {
1346             if (name) {
1347                 error_setg(errp, "net client(%s) isn't a NIC", name);
1348                 assert(!filter_list);
1349                 return NULL;
1350             }
1351             continue;
1352         }
1353
1354         /* only query information on queue 0 since the info is per nic,
1355          * not per queue
1356          */
1357         if (nc->queue_index != 0)
1358             continue;
1359
1360         if (nc->info->query_rx_filter) {
1361             info = nc->info->query_rx_filter(nc);
1362             QAPI_LIST_APPEND(tail, info);
1363         } else if (name) {
1364             error_setg(errp, "net client(%s) doesn't support"
1365                        " rx-filter querying", name);
1366             assert(!filter_list);
1367             return NULL;
1368         }
1369
1370         if (name) {
1371             break;
1372         }
1373     }
1374
1375     if (filter_list == NULL && name) {
1376         error_setg(errp, "invalid net client name: %s", name);
1377     }
1378
1379     return filter_list;
1380 }
1381
1382 void colo_notify_filters_event(int event, Error **errp)
1383 {
1384     NetClientState *nc;
1385     NetFilterState *nf;
1386     NetFilterClass *nfc = NULL;
1387     Error *local_err = NULL;
1388
1389     QTAILQ_FOREACH(nc, &net_clients, next) {
1390         QTAILQ_FOREACH(nf, &nc->filters, next) {
1391             nfc = NETFILTER_GET_CLASS(OBJECT(nf));
1392             nfc->handle_event(nf, event, &local_err);
1393             if (local_err) {
1394                 error_propagate(errp, local_err);
1395                 return;
1396             }
1397         }
1398     }
1399 }
1400
1401 void qmp_set_link(const char *name, bool up, Error **errp)
1402 {
1403     NetClientState *ncs[MAX_QUEUE_NUM];
1404     NetClientState *nc;
1405     int queues, i;
1406
1407     queues = qemu_find_net_clients_except(name, ncs,
1408                                           NET_CLIENT_DRIVER__MAX,
1409                                           MAX_QUEUE_NUM);
1410
1411     if (queues == 0) {
1412         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1413                   "Device '%s' not found", name);
1414         return;
1415     }
1416     nc = ncs[0];
1417
1418     for (i = 0; i < queues; i++) {
1419         ncs[i]->link_down = !up;
1420     }
1421
1422     if (nc->info->link_status_changed) {
1423         nc->info->link_status_changed(nc);
1424     }
1425
1426     if (nc->peer) {
1427         /* Change peer link only if the peer is NIC and then notify peer.
1428          * If the peer is a HUBPORT or a backend, we do not change the
1429          * link status.
1430          *
1431          * This behavior is compatible with qemu hubs where there could be
1432          * multiple clients that can still communicate with each other in
1433          * disconnected mode. For now maintain this compatibility.
1434          */
1435         if (nc->peer->info->type == NET_CLIENT_DRIVER_NIC) {
1436             for (i = 0; i < queues; i++) {
1437                 ncs[i]->peer->link_down = !up;
1438             }
1439         }
1440         if (nc->peer->info->link_status_changed) {
1441             nc->peer->info->link_status_changed(nc->peer);
1442         }
1443     }
1444 }
1445
1446 static void net_vm_change_state_handler(void *opaque, bool running,
1447                                         RunState state)
1448 {
1449     NetClientState *nc;
1450     NetClientState *tmp;
1451
1452     QTAILQ_FOREACH_SAFE(nc, &net_clients, next, tmp) {
1453         if (running) {
1454             /* Flush queued packets and wake up backends. */
1455             if (nc->peer && qemu_can_send_packet(nc)) {
1456                 qemu_flush_queued_packets(nc->peer);
1457             }
1458         } else {
1459             /* Complete all queued packets, to guarantee we don't modify
1460              * state later when VM is not running.
1461              */
1462             qemu_flush_or_purge_queued_packets(nc, true);
1463         }
1464     }
1465 }
1466
1467 void net_cleanup(void)
1468 {
1469     NetClientState *nc;
1470
1471     /*cleanup colo compare module for COLO*/
1472     colo_compare_cleanup();
1473
1474     /* We may del multiple entries during qemu_del_net_client(),
1475      * so QTAILQ_FOREACH_SAFE() is also not safe here.
1476      */
1477     while (!QTAILQ_EMPTY(&net_clients)) {
1478         nc = QTAILQ_FIRST(&net_clients);
1479         if (nc->info->type == NET_CLIENT_DRIVER_NIC) {
1480             qemu_del_nic(qemu_get_nic(nc));
1481         } else {
1482             qemu_del_net_client(nc);
1483         }
1484     }
1485
1486     qemu_del_vm_change_state_handler(net_change_state_entry);
1487 }
1488
1489 void net_check_clients(void)
1490 {
1491     NetClientState *nc;
1492     int i;
1493
1494     net_hub_check_clients();
1495
1496     QTAILQ_FOREACH(nc, &net_clients, next) {
1497         if (!nc->peer) {
1498             warn_report("%s %s has no peer",
1499                         nc->info->type == NET_CLIENT_DRIVER_NIC
1500                         ? "nic" : "netdev",
1501                         nc->name);
1502         }
1503     }
1504
1505     /* Check that all NICs requested via -net nic actually got created.
1506      * NICs created via -device don't need to be checked here because
1507      * they are always instantiated.
1508      */
1509     for (i = 0; i < MAX_NICS; i++) {
1510         NICInfo *nd = &nd_table[i];
1511         if (nd->used && !nd->instantiated) {
1512             warn_report("requested NIC (%s, model %s) "
1513                         "was not created (not supported by this machine?)",
1514                         nd->name ? nd->name : "anonymous",
1515                         nd->model ? nd->model : "unspecified");
1516         }
1517     }
1518 }
1519
1520 static int net_init_client(void *dummy, QemuOpts *opts, Error **errp)
1521 {
1522     return net_client_init(opts, false, errp);
1523 }
1524
1525 static int net_init_netdev(void *dummy, QemuOpts *opts, Error **errp)
1526 {
1527     const char *type = qemu_opt_get(opts, "type");
1528
1529     if (type && is_help_option(type)) {
1530         show_netdevs();
1531         exit(0);
1532     }
1533     return net_client_init(opts, true, errp);
1534 }
1535
1536 /* For the convenience "--nic" parameter */
1537 static int net_param_nic(void *dummy, QemuOpts *opts, Error **errp)
1538 {
1539     char *mac, *nd_id;
1540     int idx, ret;
1541     NICInfo *ni;
1542     const char *type;
1543
1544     type = qemu_opt_get(opts, "type");
1545     if (type) {
1546         if (g_str_equal(type, "none")) {
1547             return 0;    /* Nothing to do, default_net is cleared in vl.c */
1548         }
1549         if (is_help_option(type)) {
1550             GPtrArray *nic_models = qemu_get_nic_models(TYPE_DEVICE);
1551             show_netdevs();
1552             printf("\n");
1553             qemu_show_nic_models(type, (const char **)nic_models->pdata);
1554             g_ptr_array_free(nic_models, true);
1555             exit(0);
1556         }
1557     }
1558
1559     idx = nic_get_free_idx();
1560     if (idx == -1 || nb_nics >= MAX_NICS) {
1561         error_setg(errp, "no more on-board/default NIC slots available");
1562         return -1;
1563     }
1564
1565     if (!type) {
1566         qemu_opt_set(opts, "type", "user", &error_abort);
1567     }
1568
1569     ni = &nd_table[idx];
1570     memset(ni, 0, sizeof(*ni));
1571     ni->model = qemu_opt_get_del(opts, "model");
1572
1573     /* Create an ID if the user did not specify one */
1574     nd_id = g_strdup(qemu_opts_id(opts));
1575     if (!nd_id) {
1576         nd_id = id_generate(ID_NET);
1577         qemu_opts_set_id(opts, nd_id);
1578     }
1579
1580     /* Handle MAC address */
1581     mac = qemu_opt_get_del(opts, "mac");
1582     if (mac) {
1583         ret = net_parse_macaddr(ni->macaddr.a, mac);
1584         g_free(mac);
1585         if (ret) {
1586             error_setg(errp, "invalid syntax for ethernet address");
1587             goto out;
1588         }
1589         if (is_multicast_ether_addr(ni->macaddr.a)) {
1590             error_setg(errp, "NIC cannot have multicast MAC address");
1591             ret = -1;
1592             goto out;
1593         }
1594     }
1595     qemu_macaddr_default_if_unset(&ni->macaddr);
1596
1597     ret = net_client_init(opts, true, errp);
1598     if (ret == 0) {
1599         ni->netdev = qemu_find_netdev(nd_id);
1600         ni->used = true;
1601         nb_nics++;
1602     }
1603
1604 out:
1605     g_free(nd_id);
1606     return ret;
1607 }
1608
1609 static void netdev_init_modern(void)
1610 {
1611     while (!QSIMPLEQ_EMPTY(&nd_queue)) {
1612         NetdevQueueEntry *nd = QSIMPLEQ_FIRST(&nd_queue);
1613
1614         QSIMPLEQ_REMOVE_HEAD(&nd_queue, entry);
1615         loc_push_restore(&nd->loc);
1616         net_client_init1(nd->nd, true, &error_fatal);
1617         loc_pop(&nd->loc);
1618         qapi_free_Netdev(nd->nd);
1619         g_free(nd);
1620     }
1621 }
1622
1623 void net_init_clients(void)
1624 {
1625     net_change_state_entry =
1626         qemu_add_vm_change_state_handler(net_vm_change_state_handler, NULL);
1627
1628     QTAILQ_INIT(&net_clients);
1629
1630     netdev_init_modern();
1631
1632     qemu_opts_foreach(qemu_find_opts("netdev"), net_init_netdev, NULL,
1633                       &error_fatal);
1634
1635     qemu_opts_foreach(qemu_find_opts("nic"), net_param_nic, NULL,
1636                       &error_fatal);
1637
1638     qemu_opts_foreach(qemu_find_opts("net"), net_init_client, NULL,
1639                       &error_fatal);
1640 }
1641
1642 /*
1643  * Does this -netdev argument use modern rather than traditional syntax?
1644  * Modern syntax is to be parsed with netdev_parse_modern().
1645  * Traditional syntax is to be parsed with net_client_parse().
1646  */
1647 bool netdev_is_modern(const char *optarg)
1648 {
1649     QemuOpts *opts;
1650     bool is_modern;
1651     const char *type;
1652     static QemuOptsList dummy_opts = {
1653         .name = "netdev",
1654         .implied_opt_name = "type",
1655         .head = QTAILQ_HEAD_INITIALIZER(dummy_opts.head),
1656         .desc = { { } },
1657     };
1658
1659     if (optarg[0] == '{') {
1660         /* This is JSON, which means it's modern syntax */
1661         return true;
1662     }
1663
1664     opts = qemu_opts_create(&dummy_opts, NULL, false, &error_abort);
1665     qemu_opts_do_parse(opts, optarg, dummy_opts.implied_opt_name,
1666                        &error_abort);
1667     type = qemu_opt_get(opts, "type");
1668     is_modern = !g_strcmp0(type, "stream") || !g_strcmp0(type, "dgram");
1669
1670     qemu_opts_reset(&dummy_opts);
1671
1672     return is_modern;
1673 }
1674
1675 /*
1676  * netdev_parse_modern() uses modern, more expressive syntax than
1677  * net_client_parse(), but supports only the -netdev option.
1678  * netdev_parse_modern() appends to @nd_queue, whereas net_client_parse()
1679  * appends to @qemu_netdev_opts.
1680  */
1681 void netdev_parse_modern(const char *optarg)
1682 {
1683     Visitor *v;
1684     NetdevQueueEntry *nd;
1685
1686     v = qobject_input_visitor_new_str(optarg, "type", &error_fatal);
1687     nd = g_new(NetdevQueueEntry, 1);
1688     visit_type_Netdev(v, NULL, &nd->nd, &error_fatal);
1689     visit_free(v);
1690     loc_save(&nd->loc);
1691
1692     QSIMPLEQ_INSERT_TAIL(&nd_queue, nd, entry);
1693 }
1694
1695 void net_client_parse(QemuOptsList *opts_list, const char *optarg)
1696 {
1697     if (!qemu_opts_parse_noisily(opts_list, optarg, true)) {
1698         exit(1);
1699     }
1700 }
1701
1702 /* From FreeBSD */
1703 /* XXX: optimize */
1704 uint32_t net_crc32(const uint8_t *p, int len)
1705 {
1706     uint32_t crc;
1707     int carry, i, j;
1708     uint8_t b;
1709
1710     crc = 0xffffffff;
1711     for (i = 0; i < len; i++) {
1712         b = *p++;
1713         for (j = 0; j < 8; j++) {
1714             carry = ((crc & 0x80000000L) ? 1 : 0) ^ (b & 0x01);
1715             crc <<= 1;
1716             b >>= 1;
1717             if (carry) {
1718                 crc = ((crc ^ POLYNOMIAL_BE) | carry);
1719             }
1720         }
1721     }
1722
1723     return crc;
1724 }
1725
1726 uint32_t net_crc32_le(const uint8_t *p, int len)
1727 {
1728     uint32_t crc;
1729     int carry, i, j;
1730     uint8_t b;
1731
1732     crc = 0xffffffff;
1733     for (i = 0; i < len; i++) {
1734         b = *p++;
1735         for (j = 0; j < 8; j++) {
1736             carry = (crc & 0x1) ^ (b & 0x01);
1737             crc >>= 1;
1738             b >>= 1;
1739             if (carry) {
1740                 crc ^= POLYNOMIAL_LE;
1741             }
1742         }
1743     }
1744
1745     return crc;
1746 }
1747
1748 QemuOptsList qemu_netdev_opts = {
1749     .name = "netdev",
1750     .implied_opt_name = "type",
1751     .head = QTAILQ_HEAD_INITIALIZER(qemu_netdev_opts.head),
1752     .desc = {
1753         /*
1754          * no elements => accept any params
1755          * validation will happen later
1756          */
1757         { /* end of list */ }
1758     },
1759 };
1760
1761 QemuOptsList qemu_nic_opts = {
1762     .name = "nic",
1763     .implied_opt_name = "type",
1764     .head = QTAILQ_HEAD_INITIALIZER(qemu_nic_opts.head),
1765     .desc = {
1766         /*
1767          * no elements => accept any params
1768          * validation will happen later
1769          */
1770         { /* end of list */ }
1771     },
1772 };
1773
1774 QemuOptsList qemu_net_opts = {
1775     .name = "net",
1776     .implied_opt_name = "type",
1777     .head = QTAILQ_HEAD_INITIALIZER(qemu_net_opts.head),
1778     .desc = {
1779         /*
1780          * no elements => accept any params
1781          * validation will happen later
1782          */
1783         { /* end of list */ }
1784     },
1785 };
1786
1787 void net_socket_rs_init(SocketReadState *rs,
1788                         SocketReadStateFinalize *finalize,
1789                         bool vnet_hdr)
1790 {
1791     rs->state = 0;
1792     rs->vnet_hdr = vnet_hdr;
1793     rs->index = 0;
1794     rs->packet_len = 0;
1795     rs->vnet_hdr_len = 0;
1796     memset(rs->buf, 0, sizeof(rs->buf));
1797     rs->finalize = finalize;
1798 }
1799
1800 /*
1801  * Returns
1802  * 0: success
1803  * -1: error occurs
1804  */
1805 int net_fill_rstate(SocketReadState *rs, const uint8_t *buf, int size)
1806 {
1807     unsigned int l;
1808
1809     while (size > 0) {
1810         /* Reassemble a packet from the network.
1811          * 0 = getting length.
1812          * 1 = getting vnet header length.
1813          * 2 = getting data.
1814          */
1815         switch (rs->state) {
1816         case 0:
1817             l = 4 - rs->index;
1818             if (l > size) {
1819                 l = size;
1820             }
1821             memcpy(rs->buf + rs->index, buf, l);
1822             buf += l;
1823             size -= l;
1824             rs->index += l;
1825             if (rs->index == 4) {
1826                 /* got length */
1827                 rs->packet_len = ntohl(*(uint32_t *)rs->buf);
1828                 rs->index = 0;
1829                 if (rs->vnet_hdr) {
1830                     rs->state = 1;
1831                 } else {
1832                     rs->state = 2;
1833                     rs->vnet_hdr_len = 0;
1834                 }
1835             }
1836             break;
1837         case 1:
1838             l = 4 - rs->index;
1839             if (l > size) {
1840                 l = size;
1841             }
1842             memcpy(rs->buf + rs->index, buf, l);
1843             buf += l;
1844             size -= l;
1845             rs->index += l;
1846             if (rs->index == 4) {
1847                 /* got vnet header length */
1848                 rs->vnet_hdr_len = ntohl(*(uint32_t *)rs->buf);
1849                 rs->index = 0;
1850                 rs->state = 2;
1851             }
1852             break;
1853         case 2:
1854             l = rs->packet_len - rs->index;
1855             if (l > size) {
1856                 l = size;
1857             }
1858             if (rs->index + l <= sizeof(rs->buf)) {
1859                 memcpy(rs->buf + rs->index, buf, l);
1860             } else {
1861                 fprintf(stderr, "serious error: oversized packet received,"
1862                     "connection terminated.\n");
1863                 rs->index = rs->state = 0;
1864                 return -1;
1865             }
1866
1867             rs->index += l;
1868             buf += l;
1869             size -= l;
1870             if (rs->index >= rs->packet_len) {
1871                 rs->index = 0;
1872                 rs->state = 0;
1873                 assert(rs->finalize);
1874                 rs->finalize(rs);
1875             }
1876             break;
1877         }
1878     }
1879
1880     assert(size == 0);
1881     return 0;
1882 }