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 static QTAILQ_HEAD(, NetClientState) 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 int qemu_show_nic_models(const char *arg, const char *const *models)
 903 {
 904     int i;
 905
 906     if (!arg || !is_help_option(arg)) {
 907         return 0;
 908     }
 909
 910     printf("Supported NIC models:\n");
 911     for (i = 0 ; models[i]; i++) {
 912         printf("%s\n", models[i]);
 913     }
 914     return 1;
 915 }
 916
 917 void qemu_check_nic_model(NICInfo *nd, const char *model)
 918 {
 919     const char *models[2];
 920
 921     models[0] = model;
 922     models[1] = NULL;
 923
 924     if (qemu_show_nic_models(nd->model, models))
 925         exit(0);
 926     if (qemu_find_nic_model(nd, models, model) < 0)
 927         exit(1);
 928 }
 929
 930 int qemu_find_nic_model(NICInfo *nd, const char * const *models,
 931                         const char *default_model)
 932 {
 933     int i;
 934
 935     if (!nd->model)
 936         nd->model = g_strdup(default_model);
 937
 938     for (i = 0 ; models[i]; i++) {
 939         if (strcmp(nd->model, models[i]) == 0)
 940             return i;
 941     }
 942
 943     error_report("Unsupported NIC model: %s", nd->model);
 944     return -1;
 945 }
 946
 947 static int net_init_nic(const Netdev *netdev, const char *name,
 948                         NetClientState *peer, Error **errp)
 949 {
 950     int idx;
 951     NICInfo *nd;
 952     const NetLegacyNicOptions *nic;
 953
 954     assert(netdev->type == NET_CLIENT_DRIVER_NIC);
 955     nic = &netdev->u.nic;
 956
 957     idx = nic_get_free_idx();
 958     if (idx == -1 || nb_nics >= MAX_NICS) {
 959         error_setg(errp, "too many NICs");
 960         return -1;
 961     }
 962
 963     nd = &nd_table[idx];
 964
 965     memset(nd, 0, sizeof(*nd));
 966
 967     if (nic->has_netdev) {
 968         nd->netdev = qemu_find_netdev(nic->netdev);
 969         if (!nd->netdev) {
 970             error_setg(errp, "netdev '%s' not found", nic->netdev);
 971             return -1;
 972         }
 973     } else {
 974         assert(peer);
 975         nd->netdev = peer;
 976     }
 977     nd->name = g_strdup(name);
 978     if (nic->has_model) {
 979         nd->model = g_strdup(nic->model);
 980     }
 981     if (nic->has_addr) {
 982         nd->devaddr = g_strdup(nic->addr);
 983     }
 984
 985     if (nic->has_macaddr &&
 986         net_parse_macaddr(nd->macaddr.a, nic->macaddr) < 0) {
 987         error_setg(errp, "invalid syntax for ethernet address");
 988         return -1;
 989     }
 990     if (nic->has_macaddr &&
 991         is_multicast_ether_addr(nd->macaddr.a)) {
 992         error_setg(errp,
 993                    "NIC cannot have multicast MAC address (odd 1st byte)");
 994         return -1;
 995     }
 996     qemu_macaddr_default_if_unset(&nd->macaddr);
 997
 998     if (nic->has_vectors) {
 999         if (nic->vectors > 0x7ffffff) {
1000             error_setg(errp, "invalid # of vectors: %"PRIu32, nic->vectors);
1001             return -1;
1002         }
1003         nd->nvectors = nic->vectors;
1004     } else {
1005         nd->nvectors = DEV_NVECTORS_UNSPECIFIED;
1006     }
1007
1008     nd->used = 1;
1009     nb_nics++;
1010
1011     return idx;
1012 }
1013
1014
1015 static int (* const net_client_init_fun[NET_CLIENT_DRIVER__MAX])(
1016     const Netdev *netdev,
1017     const char *name,
1018     NetClientState *peer, Error **errp) = {
1019         [NET_CLIENT_DRIVER_NIC]       = net_init_nic,
1020 #ifdef CONFIG_SLIRP
1021         [NET_CLIENT_DRIVER_USER]      = net_init_slirp,
1022 #endif
1023         [NET_CLIENT_DRIVER_TAP]       = net_init_tap,
1024         [NET_CLIENT_DRIVER_SOCKET]    = net_init_socket,
1025         [NET_CLIENT_DRIVER_STREAM]    = net_init_stream,
1026         [NET_CLIENT_DRIVER_DGRAM]     = net_init_dgram,
1027 #ifdef CONFIG_VDE
1028         [NET_CLIENT_DRIVER_VDE]       = net_init_vde,
1029 #endif
1030 #ifdef CONFIG_NETMAP
1031         [NET_CLIENT_DRIVER_NETMAP]    = net_init_netmap,
1032 #endif
1033 #ifdef CONFIG_NET_BRIDGE
1034         [NET_CLIENT_DRIVER_BRIDGE]    = net_init_bridge,
1035 #endif
1036         [NET_CLIENT_DRIVER_HUBPORT]   = net_init_hubport,
1037 #ifdef CONFIG_VHOST_NET_USER
1038         [NET_CLIENT_DRIVER_VHOST_USER] = net_init_vhost_user,
1039 #endif
1040 #ifdef CONFIG_VHOST_NET_VDPA
1041         [NET_CLIENT_DRIVER_VHOST_VDPA] = net_init_vhost_vdpa,
1042 #endif
1043 #ifdef CONFIG_L2TPV3
1044         [NET_CLIENT_DRIVER_L2TPV3]    = net_init_l2tpv3,
1045 #endif
1046 #ifdef CONFIG_VMNET
1047         [NET_CLIENT_DRIVER_VMNET_HOST] = net_init_vmnet_host,
1048         [NET_CLIENT_DRIVER_VMNET_SHARED] = net_init_vmnet_shared,
1049         [NET_CLIENT_DRIVER_VMNET_BRIDGED] = net_init_vmnet_bridged,
1050 #endif /* CONFIG_VMNET */
1051 };
1052
1053
1054 static int net_client_init1(const Netdev *netdev, bool is_netdev, Error **errp)
1055 {
1056     NetClientState *peer = NULL;
1057     NetClientState *nc;
1058
1059     if (is_netdev) {
1060         if (netdev->type == NET_CLIENT_DRIVER_NIC ||
1061             !net_client_init_fun[netdev->type]) {
1062             error_setg(errp, "network backend '%s' is not compiled into this binary",
1063                        NetClientDriver_str(netdev->type));
1064             return -1;
1065         }
1066     } else {
1067         if (netdev->type == NET_CLIENT_DRIVER_NONE) {
1068             return 0; /* nothing to do */
1069         }
1070         if (netdev->type == NET_CLIENT_DRIVER_HUBPORT) {
1071             error_setg(errp, "network backend '%s' is only supported with -netdev/-nic",
1072                        NetClientDriver_str(netdev->type));
1073             return -1;
1074         }
1075
1076         if (!net_client_init_fun[netdev->type]) {
1077             error_setg(errp, "network backend '%s' is not compiled into this binary",
1078                        NetClientDriver_str(netdev->type));
1079             return -1;
1080         }
1081
1082         /* Do not add to a hub if it's a nic with a netdev= parameter. */
1083         if (netdev->type != NET_CLIENT_DRIVER_NIC ||
1084             !netdev->u.nic.has_netdev) {
1085             peer = net_hub_add_port(0, NULL, NULL);
1086         }
1087     }
1088
1089     nc = qemu_find_netdev(netdev->id);
1090     if (nc) {
1091         error_setg(errp, "Duplicate ID '%s'", netdev->id);
1092         return -1;
1093     }
1094
1095     if (net_client_init_fun[netdev->type](netdev, netdev->id, peer, errp) < 0) {
1096         /* FIXME drop when all init functions store an Error */
1097         if (errp && !*errp) {
1098             error_setg(errp, "Device '%s' could not be initialized",
1099                        NetClientDriver_str(netdev->type));
1100         }
1101         return -1;
1102     }
1103
1104     if (is_netdev) {
1105         nc = qemu_find_netdev(netdev->id);
1106         assert(nc);
1107         nc->is_netdev = true;
1108     }
1109
1110     return 0;
1111 }
1112
1113 void show_netdevs(void)
1114 {
1115     int idx;
1116     const char *available_netdevs[] = {
1117         "socket",
1118         "stream",
1119         "dgram",
1120         "hubport",
1121         "tap",
1122 #ifdef CONFIG_SLIRP
1123         "user",
1124 #endif
1125 #ifdef CONFIG_L2TPV3
1126         "l2tpv3",
1127 #endif
1128 #ifdef CONFIG_VDE
1129         "vde",
1130 #endif
1131 #ifdef CONFIG_NET_BRIDGE
1132         "bridge",
1133 #endif
1134 #ifdef CONFIG_NETMAP
1135         "netmap",
1136 #endif
1137 #ifdef CONFIG_POSIX
1138         "vhost-user",
1139 #endif
1140 #ifdef CONFIG_VHOST_VDPA
1141         "vhost-vdpa",
1142 #endif
1143 #ifdef CONFIG_VMNET
1144         "vmnet-host",
1145         "vmnet-shared",
1146         "vmnet-bridged",
1147 #endif
1148     };
1149
1150     qemu_printf("Available netdev backend types:\n");
1151     for (idx = 0; idx < ARRAY_SIZE(available_netdevs); idx++) {
1152         qemu_printf("%s\n", available_netdevs[idx]);
1153     }
1154 }
1155
1156 static int net_client_init(QemuOpts *opts, bool is_netdev, Error **errp)
1157 {
1158     gchar **substrings = NULL;
1159     Netdev *object = NULL;
1160     int ret = -1;
1161     Visitor *v = opts_visitor_new(opts);
1162
1163     /* Parse convenience option format ip6-net=fec0::0[/64] */
1164     const char *ip6_net = qemu_opt_get(opts, "ipv6-net");
1165
1166     if (ip6_net) {
1167         char *prefix_addr;
1168         unsigned long prefix_len = 64; /* Default 64bit prefix length. */
1169
1170         substrings = g_strsplit(ip6_net, "/", 2);
1171         if (!substrings || !substrings[0]) {
1172             error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "ipv6-net",
1173                        "a valid IPv6 prefix");
1174             goto out;
1175         }
1176
1177         prefix_addr = substrings[0];
1178
1179         /* Handle user-specified prefix length. */
1180         if (substrings[1] &&
1181             qemu_strtoul(substrings[1], NULL, 10, &prefix_len))
1182         {
1183             error_setg(errp, QERR_INVALID_PARAMETER_VALUE,
1184                        "ipv6-prefixlen", "a number");
1185             goto out;
1186         }
1187
1188         qemu_opt_set(opts, "ipv6-prefix", prefix_addr, &error_abort);
1189         qemu_opt_set_number(opts, "ipv6-prefixlen", prefix_len,
1190                             &error_abort);
1191         qemu_opt_unset(opts, "ipv6-net");
1192     }
1193
1194     /* Create an ID for -net if the user did not specify one */
1195     if (!is_netdev && !qemu_opts_id(opts)) {
1196         qemu_opts_set_id(opts, id_generate(ID_NET));
1197     }
1198
1199     if (visit_type_Netdev(v, NULL, &object, errp)) {
1200         ret = net_client_init1(object, is_netdev, errp);
1201     }
1202
1203     qapi_free_Netdev(object);
1204
1205 out:
1206     g_strfreev(substrings);
1207     visit_free(v);
1208     return ret;
1209 }
1210
1211 void netdev_add(QemuOpts *opts, Error **errp)
1212 {
1213     net_client_init(opts, true, errp);
1214 }
1215
1216 void qmp_netdev_add(Netdev *netdev, Error **errp)
1217 {
1218     if (!id_wellformed(netdev->id)) {
1219         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "id", "an identifier");
1220         return;
1221     }
1222
1223     net_client_init1(netdev, true, errp);
1224 }
1225
1226 void qmp_netdev_del(const char *id, Error **errp)
1227 {
1228     NetClientState *nc;
1229     QemuOpts *opts;
1230
1231     nc = qemu_find_netdev(id);
1232     if (!nc) {
1233         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1234                   "Device '%s' not found", id);
1235         return;
1236     }
1237
1238     if (!nc->is_netdev) {
1239         error_setg(errp, "Device '%s' is not a netdev", id);
1240         return;
1241     }
1242
1243     qemu_del_net_client(nc);
1244
1245     /*
1246      * Wart: we need to delete the QemuOpts associated with netdevs
1247      * created via CLI or HMP, to avoid bogus "Duplicate ID" errors in
1248      * HMP netdev_add.
1249      */
1250     opts = qemu_opts_find(qemu_find_opts("netdev"), id);
1251     if (opts) {
1252         qemu_opts_del(opts);
1253     }
1254 }
1255
1256 static void netfilter_print_info(Monitor *mon, NetFilterState *nf)
1257 {
1258     char *str;
1259     ObjectProperty *prop;
1260     ObjectPropertyIterator iter;
1261     Visitor *v;
1262
1263     /* generate info str */
1264     object_property_iter_init(&iter, OBJECT(nf));
1265     while ((prop = object_property_iter_next(&iter))) {
1266         if (!strcmp(prop->name, "type")) {
1267             continue;
1268         }
1269         v = string_output_visitor_new(false, &str);
1270         object_property_get(OBJECT(nf), prop->name, v, NULL);
1271         visit_complete(v, &str);
1272         visit_free(v);
1273         monitor_printf(mon, ",%s=%s", prop->name, str);
1274         g_free(str);
1275     }
1276     monitor_printf(mon, "\n");
1277 }
1278
1279 void print_net_client(Monitor *mon, NetClientState *nc)
1280 {
1281     NetFilterState *nf;
1282
1283     monitor_printf(mon, "%s: index=%d,type=%s,%s\n", nc->name,
1284                    nc->queue_index,
1285                    NetClientDriver_str(nc->info->type),
1286                    nc->info_str);
1287     if (!QTAILQ_EMPTY(&nc->filters)) {
1288         monitor_printf(mon, "filters:\n");
1289     }
1290     QTAILQ_FOREACH(nf, &nc->filters, next) {
1291         monitor_printf(mon, "  - %s: type=%s",
1292                        object_get_canonical_path_component(OBJECT(nf)),
1293                        object_get_typename(OBJECT(nf)));
1294         netfilter_print_info(mon, nf);
1295     }
1296 }
1297
1298 RxFilterInfoList *qmp_query_rx_filter(bool has_name, const char *name,
1299                                       Error **errp)
1300 {
1301     NetClientState *nc;
1302     RxFilterInfoList *filter_list = NULL, **tail = &filter_list;
1303
1304     QTAILQ_FOREACH(nc, &net_clients, next) {
1305         RxFilterInfo *info;
1306
1307         if (has_name && strcmp(nc->name, name) != 0) {
1308             continue;
1309         }
1310
1311         /* only query rx-filter information of NIC */
1312         if (nc->info->type != NET_CLIENT_DRIVER_NIC) {
1313             if (has_name) {
1314                 error_setg(errp, "net client(%s) isn't a NIC", name);
1315                 assert(!filter_list);
1316                 return NULL;
1317             }
1318             continue;
1319         }
1320
1321         /* only query information on queue 0 since the info is per nic,
1322          * not per queue
1323          */
1324         if (nc->queue_index != 0)
1325             continue;
1326
1327         if (nc->info->query_rx_filter) {
1328             info = nc->info->query_rx_filter(nc);
1329             QAPI_LIST_APPEND(tail, info);
1330         } else if (has_name) {
1331             error_setg(errp, "net client(%s) doesn't support"
1332                        " rx-filter querying", name);
1333             assert(!filter_list);
1334             return NULL;
1335         }
1336
1337         if (has_name) {
1338             break;
1339         }
1340     }
1341
1342     if (filter_list == NULL && has_name) {
1343         error_setg(errp, "invalid net client name: %s", name);
1344     }
1345
1346     return filter_list;
1347 }
1348
1349 void hmp_info_network(Monitor *mon, const QDict *qdict)
1350 {
1351     NetClientState *nc, *peer;
1352     NetClientDriver type;
1353
1354     net_hub_info(mon);
1355
1356     QTAILQ_FOREACH(nc, &net_clients, next) {
1357         peer = nc->peer;
1358         type = nc->info->type;
1359
1360         /* Skip if already printed in hub info */
1361         if (net_hub_id_for_client(nc, NULL) == 0) {
1362             continue;
1363         }
1364
1365         if (!peer || type == NET_CLIENT_DRIVER_NIC) {
1366             print_net_client(mon, nc);
1367         } /* else it's a netdev connected to a NIC, printed with the NIC */
1368         if (peer && type == NET_CLIENT_DRIVER_NIC) {
1369             monitor_printf(mon, " \\ ");
1370             print_net_client(mon, peer);
1371         }
1372     }
1373 }
1374
1375 void colo_notify_filters_event(int event, Error **errp)
1376 {
1377     NetClientState *nc;
1378     NetFilterState *nf;
1379     NetFilterClass *nfc = NULL;
1380     Error *local_err = NULL;
1381
1382     QTAILQ_FOREACH(nc, &net_clients, next) {
1383         QTAILQ_FOREACH(nf, &nc->filters, next) {
1384             nfc = NETFILTER_GET_CLASS(OBJECT(nf));
1385             nfc->handle_event(nf, event, &local_err);
1386             if (local_err) {
1387                 error_propagate(errp, local_err);
1388                 return;
1389             }
1390         }
1391     }
1392 }
1393
1394 void qmp_set_link(const char *name, bool up, Error **errp)
1395 {
1396     NetClientState *ncs[MAX_QUEUE_NUM];
1397     NetClientState *nc;
1398     int queues, i;
1399
1400     queues = qemu_find_net_clients_except(name, ncs,
1401                                           NET_CLIENT_DRIVER__MAX,
1402                                           MAX_QUEUE_NUM);
1403
1404     if (queues == 0) {
1405         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1406                   "Device '%s' not found", name);
1407         return;
1408     }
1409     nc = ncs[0];
1410
1411     for (i = 0; i < queues; i++) {
1412         ncs[i]->link_down = !up;
1413     }
1414
1415     if (nc->info->link_status_changed) {
1416         nc->info->link_status_changed(nc);
1417     }
1418
1419     if (nc->peer) {
1420         /* Change peer link only if the peer is NIC and then notify peer.
1421          * If the peer is a HUBPORT or a backend, we do not change the
1422          * link status.
1423          *
1424          * This behavior is compatible with qemu hubs where there could be
1425          * multiple clients that can still communicate with each other in
1426          * disconnected mode. For now maintain this compatibility.
1427          */
1428         if (nc->peer->info->type == NET_CLIENT_DRIVER_NIC) {
1429             for (i = 0; i < queues; i++) {
1430                 ncs[i]->peer->link_down = !up;
1431             }
1432         }
1433         if (nc->peer->info->link_status_changed) {
1434             nc->peer->info->link_status_changed(nc->peer);
1435         }
1436     }
1437 }
1438
1439 static void net_vm_change_state_handler(void *opaque, bool running,
1440                                         RunState state)
1441 {
1442     NetClientState *nc;
1443     NetClientState *tmp;
1444
1445     QTAILQ_FOREACH_SAFE(nc, &net_clients, next, tmp) {
1446         if (running) {
1447             /* Flush queued packets and wake up backends. */
1448             if (nc->peer && qemu_can_send_packet(nc)) {
1449                 qemu_flush_queued_packets(nc->peer);
1450             }
1451         } else {
1452             /* Complete all queued packets, to guarantee we don't modify
1453              * state later when VM is not running.
1454              */
1455             qemu_flush_or_purge_queued_packets(nc, true);
1456         }
1457     }
1458 }
1459
1460 void net_cleanup(void)
1461 {
1462     NetClientState *nc;
1463
1464     /*cleanup colo compare module for COLO*/
1465     colo_compare_cleanup();
1466
1467     /* We may del multiple entries during qemu_del_net_client(),
1468      * so QTAILQ_FOREACH_SAFE() is also not safe here.
1469      */
1470     while (!QTAILQ_EMPTY(&net_clients)) {
1471         nc = QTAILQ_FIRST(&net_clients);
1472         if (nc->info->type == NET_CLIENT_DRIVER_NIC) {
1473             qemu_del_nic(qemu_get_nic(nc));
1474         } else {
1475             qemu_del_net_client(nc);
1476         }
1477     }
1478
1479     qemu_del_vm_change_state_handler(net_change_state_entry);
1480 }
1481
1482 void net_check_clients(void)
1483 {
1484     NetClientState *nc;
1485     int i;
1486
1487     net_hub_check_clients();
1488
1489     QTAILQ_FOREACH(nc, &net_clients, next) {
1490         if (!nc->peer) {
1491             warn_report("%s %s has no peer",
1492                         nc->info->type == NET_CLIENT_DRIVER_NIC
1493                         ? "nic" : "netdev",
1494                         nc->name);
1495         }
1496     }
1497
1498     /* Check that all NICs requested via -net nic actually got created.
1499      * NICs created via -device don't need to be checked here because
1500      * they are always instantiated.
1501      */
1502     for (i = 0; i < MAX_NICS; i++) {
1503         NICInfo *nd = &nd_table[i];
1504         if (nd->used && !nd->instantiated) {
1505             warn_report("requested NIC (%s, model %s) "
1506                         "was not created (not supported by this machine?)",
1507                         nd->name ? nd->name : "anonymous",
1508                         nd->model ? nd->model : "unspecified");
1509         }
1510     }
1511 }
1512
1513 static int net_init_client(void *dummy, QemuOpts *opts, Error **errp)
1514 {
1515     return net_client_init(opts, false, errp);
1516 }
1517
1518 static int net_init_netdev(void *dummy, QemuOpts *opts, Error **errp)
1519 {
1520     const char *type = qemu_opt_get(opts, "type");
1521
1522     if (type && is_help_option(type)) {
1523         show_netdevs();
1524         exit(0);
1525     }
1526     return net_client_init(opts, true, errp);
1527 }
1528
1529 /* For the convenience "--nic" parameter */
1530 static int net_param_nic(void *dummy, QemuOpts *opts, Error **errp)
1531 {
1532     char *mac, *nd_id;
1533     int idx, ret;
1534     NICInfo *ni;
1535     const char *type;
1536
1537     type = qemu_opt_get(opts, "type");
1538     if (type && g_str_equal(type, "none")) {
1539         return 0;    /* Nothing to do, default_net is cleared in vl.c */
1540     }
1541
1542     idx = nic_get_free_idx();
1543     if (idx == -1 || nb_nics >= MAX_NICS) {
1544         error_setg(errp, "no more on-board/default NIC slots available");
1545         return -1;
1546     }
1547
1548     if (!type) {
1549         qemu_opt_set(opts, "type", "user", &error_abort);
1550     }
1551
1552     ni = &nd_table[idx];
1553     memset(ni, 0, sizeof(*ni));
1554     ni->model = qemu_opt_get_del(opts, "model");
1555
1556     /* Create an ID if the user did not specify one */
1557     nd_id = g_strdup(qemu_opts_id(opts));
1558     if (!nd_id) {
1559         nd_id = id_generate(ID_NET);
1560         qemu_opts_set_id(opts, nd_id);
1561     }
1562
1563     /* Handle MAC address */
1564     mac = qemu_opt_get_del(opts, "mac");
1565     if (mac) {
1566         ret = net_parse_macaddr(ni->macaddr.a, mac);
1567         g_free(mac);
1568         if (ret) {
1569             error_setg(errp, "invalid syntax for ethernet address");
1570             goto out;
1571         }
1572         if (is_multicast_ether_addr(ni->macaddr.a)) {
1573             error_setg(errp, "NIC cannot have multicast MAC address");
1574             ret = -1;
1575             goto out;
1576         }
1577     }
1578     qemu_macaddr_default_if_unset(&ni->macaddr);
1579
1580     ret = net_client_init(opts, true, errp);
1581     if (ret == 0) {
1582         ni->netdev = qemu_find_netdev(nd_id);
1583         ni->used = true;
1584         nb_nics++;
1585     }
1586
1587 out:
1588     g_free(nd_id);
1589     return ret;
1590 }
1591
1592 static void netdev_init_modern(void)
1593 {
1594     while (!QSIMPLEQ_EMPTY(&nd_queue)) {
1595         NetdevQueueEntry *nd = QSIMPLEQ_FIRST(&nd_queue);
1596
1597         QSIMPLEQ_REMOVE_HEAD(&nd_queue, entry);
1598         loc_push_restore(&nd->loc);
1599         net_client_init1(nd->nd, true, &error_fatal);
1600         loc_pop(&nd->loc);
1601         qapi_free_Netdev(nd->nd);
1602         g_free(nd);
1603     }
1604 }
1605
1606 void net_init_clients(void)
1607 {
1608     net_change_state_entry =
1609         qemu_add_vm_change_state_handler(net_vm_change_state_handler, NULL);
1610
1611     QTAILQ_INIT(&net_clients);
1612
1613     netdev_init_modern();
1614
1615     qemu_opts_foreach(qemu_find_opts("netdev"), net_init_netdev, NULL,
1616                       &error_fatal);
1617
1618     qemu_opts_foreach(qemu_find_opts("nic"), net_param_nic, NULL,
1619                       &error_fatal);
1620
1621     qemu_opts_foreach(qemu_find_opts("net"), net_init_client, NULL,
1622                       &error_fatal);
1623 }
1624
1625 /*
1626  * Does this -netdev argument use modern rather than traditional syntax?
1627  * Modern syntax is to be parsed with netdev_parse_modern().
1628  * Traditional syntax is to be parsed with net_client_parse().
1629  */
1630 bool netdev_is_modern(const char *optarg)
1631 {
1632     QemuOpts *opts;
1633     bool is_modern;
1634     const char *type;
1635     static QemuOptsList dummy_opts = {
1636         .name = "netdev",
1637         .implied_opt_name = "type",
1638         .head = QTAILQ_HEAD_INITIALIZER(dummy_opts.head),
1639         .desc = { { } },
1640     };
1641
1642     if (optarg[0] == '{') {
1643         /* This is JSON, which means it's modern syntax */
1644         return true;
1645     }
1646
1647     opts = qemu_opts_create(&dummy_opts, NULL, false, &error_abort);
1648     qemu_opts_do_parse(opts, optarg, dummy_opts.implied_opt_name,
1649                        &error_abort);
1650     type = qemu_opt_get(opts, "type");
1651     is_modern = !g_strcmp0(type, "stream") || !g_strcmp0(type, "dgram");
1652
1653     qemu_opts_reset(&dummy_opts);
1654
1655     return is_modern;
1656 }
1657
1658 /*
1659  * netdev_parse_modern() uses modern, more expressive syntax than
1660  * net_client_parse(), but supports only the -netdev option.
1661  * netdev_parse_modern() appends to @nd_queue, whereas net_client_parse()
1662  * appends to @qemu_netdev_opts.
1663  */
1664 void netdev_parse_modern(const char *optarg)
1665 {
1666     Visitor *v;
1667     NetdevQueueEntry *nd;
1668
1669     v = qobject_input_visitor_new_str(optarg, "type", &error_fatal);
1670     nd = g_new(NetdevQueueEntry, 1);
1671     visit_type_Netdev(v, NULL, &nd->nd, &error_fatal);
1672     visit_free(v);
1673     loc_save(&nd->loc);
1674
1675     QSIMPLEQ_INSERT_TAIL(&nd_queue, nd, entry);
1676 }
1677
1678 void net_client_parse(QemuOptsList *opts_list, const char *optarg)
1679 {
1680     if (!qemu_opts_parse_noisily(opts_list, optarg, true)) {
1681         exit(1);
1682     }
1683 }
1684
1685 /* From FreeBSD */
1686 /* XXX: optimize */
1687 uint32_t net_crc32(const uint8_t *p, int len)
1688 {
1689     uint32_t crc;
1690     int carry, i, j;
1691     uint8_t b;
1692
1693     crc = 0xffffffff;
1694     for (i = 0; i < len; i++) {
1695         b = *p++;
1696         for (j = 0; j < 8; j++) {
1697             carry = ((crc & 0x80000000L) ? 1 : 0) ^ (b & 0x01);
1698             crc <<= 1;
1699             b >>= 1;
1700             if (carry) {
1701                 crc = ((crc ^ POLYNOMIAL_BE) | carry);
1702             }
1703         }
1704     }
1705
1706     return crc;
1707 }
1708
1709 uint32_t net_crc32_le(const uint8_t *p, int len)
1710 {
1711     uint32_t crc;
1712     int carry, i, j;
1713     uint8_t b;
1714
1715     crc = 0xffffffff;
1716     for (i = 0; i < len; i++) {
1717         b = *p++;
1718         for (j = 0; j < 8; j++) {
1719             carry = (crc & 0x1) ^ (b & 0x01);
1720             crc >>= 1;
1721             b >>= 1;
1722             if (carry) {
1723                 crc ^= POLYNOMIAL_LE;
1724             }
1725         }
1726     }
1727
1728     return crc;
1729 }
1730
1731 QemuOptsList qemu_netdev_opts = {
1732     .name = "netdev",
1733     .implied_opt_name = "type",
1734     .head = QTAILQ_HEAD_INITIALIZER(qemu_netdev_opts.head),
1735     .desc = {
1736         /*
1737          * no elements => accept any params
1738          * validation will happen later
1739          */
1740         { /* end of list */ }
1741     },
1742 };
1743
1744 QemuOptsList qemu_nic_opts = {
1745     .name = "nic",
1746     .implied_opt_name = "type",
1747     .head = QTAILQ_HEAD_INITIALIZER(qemu_nic_opts.head),
1748     .desc = {
1749         /*
1750          * no elements => accept any params
1751          * validation will happen later
1752          */
1753         { /* end of list */ }
1754     },
1755 };
1756
1757 QemuOptsList qemu_net_opts = {
1758     .name = "net",
1759     .implied_opt_name = "type",
1760     .head = QTAILQ_HEAD_INITIALIZER(qemu_net_opts.head),
1761     .desc = {
1762         /*
1763          * no elements => accept any params
1764          * validation will happen later
1765          */
1766         { /* end of list */ }
1767     },
1768 };
1769
1770 void net_socket_rs_init(SocketReadState *rs,
1771                         SocketReadStateFinalize *finalize,
1772                         bool vnet_hdr)
1773 {
1774     rs->state = 0;
1775     rs->vnet_hdr = vnet_hdr;
1776     rs->index = 0;
1777     rs->packet_len = 0;
1778     rs->vnet_hdr_len = 0;
1779     memset(rs->buf, 0, sizeof(rs->buf));
1780     rs->finalize = finalize;
1781 }
1782
1783 /*
1784  * Returns
1785  * 0: success
1786  * -1: error occurs
1787  */
1788 int net_fill_rstate(SocketReadState *rs, const uint8_t *buf, int size)
1789 {
1790     unsigned int l;
1791
1792     while (size > 0) {
1793         /* Reassemble a packet from the network.
1794          * 0 = getting length.
1795          * 1 = getting vnet header length.
1796          * 2 = getting data.
1797          */
1798         switch (rs->state) {
1799         case 0:
1800             l = 4 - rs->index;
1801             if (l > size) {
1802                 l = size;
1803             }
1804             memcpy(rs->buf + rs->index, buf, l);
1805             buf += l;
1806             size -= l;
1807             rs->index += l;
1808             if (rs->index == 4) {
1809                 /* got length */
1810                 rs->packet_len = ntohl(*(uint32_t *)rs->buf);
1811                 rs->index = 0;
1812                 if (rs->vnet_hdr) {
1813                     rs->state = 1;
1814                 } else {
1815                     rs->state = 2;
1816                     rs->vnet_hdr_len = 0;
1817                 }
1818             }
1819             break;
1820         case 1:
1821             l = 4 - rs->index;
1822             if (l > size) {
1823                 l = size;
1824             }
1825             memcpy(rs->buf + rs->index, buf, l);
1826             buf += l;
1827             size -= l;
1828             rs->index += l;
1829             if (rs->index == 4) {
1830                 /* got vnet header length */
1831                 rs->vnet_hdr_len = ntohl(*(uint32_t *)rs->buf);
1832                 rs->index = 0;
1833                 rs->state = 2;
1834             }
1835             break;
1836         case 2:
1837             l = rs->packet_len - rs->index;
1838             if (l > size) {
1839                 l = size;
1840             }
1841             if (rs->index + l <= sizeof(rs->buf)) {
1842                 memcpy(rs->buf + rs->index, buf, l);
1843             } else {
1844                 fprintf(stderr, "serious error: oversized packet received,"
1845                     "connection terminated.\n");
1846                 rs->index = rs->state = 0;
1847                 return -1;
1848             }
1849
1850             rs->index += l;
1851             buf += l;
1852             size -= l;
1853             if (rs->index >= rs->packet_len) {
1854                 rs->index = 0;
1855                 rs->state = 0;
1856                 assert(rs->finalize);
1857                 rs->finalize(rs);
1858             }
1859             break;
1860         }
1861     }
1862
1863     assert(size == 0);
1864     return 0;
1865 }