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 "net/slirp.h"
  31 #include "net/eth.h"
  32 #include "util.h"
  33
  34 #include "monitor/monitor.h"
  35 #include "qemu/help_option.h"
  36 #include "qapi/qapi-commands-net.h"
  37 #include "qapi/qapi-visit-net.h"
  38 #include "qapi/qmp/qdict.h"
  39 #include "qapi/qmp/qerror.h"
  40 #include "qemu/error-report.h"
  41 #include "qemu/sockets.h"
  42 #include "qemu/cutils.h"
  43 #include "qemu/config-file.h"
  44 #include "qemu/ctype.h"
  45 #include "hw/qdev.h"
  46 #include "qemu/iov.h"
  47 #include "qemu/main-loop.h"
  48 #include "qemu/option.h"
  49 #include "qapi/error.h"
  50 #include "qapi/opts-visitor.h"
  51 #include "sysemu/sysemu.h"
  52 #include "sysemu/qtest.h"
  53 #include "net/filter.h"
  54 #include "qapi/string-output-visitor.h"
  55
  56 /* Net bridge is currently not supported for W32. */
  57 #if !defined(_WIN32)
  58 # define CONFIG_NET_BRIDGE
  59 #endif
  60
  61 static VMChangeStateEntry *net_change_state_entry;
  62 static QTAILQ_HEAD(, NetClientState) net_clients;
  63
  64 /***********************************************************/
  65 /* network device redirectors */
  66
  67 static int get_str_sep(char *buf, int buf_size, const char **pp, int sep)
  68 {
  69     const char *p, *p1;
  70     int len;
  71     p = *pp;
  72     p1 = strchr(p, sep);
  73     if (!p1)
  74         return -1;
  75     len = p1 - p;
  76     p1++;
  77     if (buf_size > 0) {
  78         if (len > buf_size - 1)
  79             len = buf_size - 1;
  80         memcpy(buf, p, len);
  81         buf[len] = '\0';
  82     }
  83     *pp = p1;
  84     return 0;
  85 }
  86
  87 int parse_host_port(struct sockaddr_in *saddr, const char *str,
  88                     Error **errp)
  89 {
  90     gchar **substrings;
  91     struct hostent *he;
  92     const char *addr, *p, *r;
  93     int port, ret = 0;
  94
  95     substrings = g_strsplit(str, ":", 2);
  96     if (!substrings || !substrings[0] || !substrings[1]) {
  97         error_setg(errp, "host address '%s' doesn't contain ':' "
  98                    "separating host from port", str);
  99         ret = -1;
 100         goto out;
 101     }
 102
 103     addr = substrings[0];
 104     p = substrings[1];
 105
 106     saddr->sin_family = AF_INET;
 107     if (addr[0] == '\0') {
 108         saddr->sin_addr.s_addr = 0;
 109     } else {
 110         if (qemu_isdigit(addr[0])) {
 111             if (!inet_aton(addr, &saddr->sin_addr)) {
 112                 error_setg(errp, "host address '%s' is not a valid "
 113                            "IPv4 address", addr);
 114                 ret = -1;
 115                 goto out;
 116             }
 117         } else {
 118             he = gethostbyname(addr);
 119             if (he == NULL) {
 120                 error_setg(errp, "can't resolve host address '%s'", addr);
 121                 ret = -1;
 122                 goto out;
 123             }
 124             saddr->sin_addr = *(struct in_addr *)he->h_addr;
 125         }
 126     }
 127     port = strtol(p, (char **)&r, 0);
 128     if (r == p) {
 129         error_setg(errp, "port number '%s' is invalid", p);
 130         ret = -1;
 131         goto out;
 132     }
 133     saddr->sin_port = htons(port);
 134
 135 out:
 136     g_strfreev(substrings);
 137     return ret;
 138 }
 139
 140 char *qemu_mac_strdup_printf(const uint8_t *macaddr)
 141 {
 142     return g_strdup_printf("%.2x:%.2x:%.2x:%.2x:%.2x:%.2x",
 143                            macaddr[0], macaddr[1], macaddr[2],
 144                            macaddr[3], macaddr[4], macaddr[5]);
 145 }
 146
 147 void qemu_format_nic_info_str(NetClientState *nc, uint8_t macaddr[6])
 148 {
 149     snprintf(nc->info_str, sizeof(nc->info_str),
 150              "model=%s,macaddr=%02x:%02x:%02x:%02x:%02x:%02x",
 151              nc->model,
 152              macaddr[0], macaddr[1], macaddr[2],
 153              macaddr[3], macaddr[4], macaddr[5]);
 154 }
 155
 156 static int mac_table[256] = {0};
 157
 158 static void qemu_macaddr_set_used(MACAddr *macaddr)
 159 {
 160     int index;
 161
 162     for (index = 0x56; index < 0xFF; index++) {
 163         if (macaddr->a[5] == index) {
 164             mac_table[index]++;
 165         }
 166     }
 167 }
 168
 169 static void qemu_macaddr_set_free(MACAddr *macaddr)
 170 {
 171     int index;
 172     static const MACAddr base = { .a = { 0x52, 0x54, 0x00, 0x12, 0x34, 0 } };
 173
 174     if (memcmp(macaddr->a, &base.a, (sizeof(base.a) - 1)) != 0) {
 175         return;
 176     }
 177     for (index = 0x56; index < 0xFF; index++) {
 178         if (macaddr->a[5] == index) {
 179             mac_table[index]--;
 180         }
 181     }
 182 }
 183
 184 static int qemu_macaddr_get_free(void)
 185 {
 186     int index;
 187
 188     for (index = 0x56; index < 0xFF; index++) {
 189         if (mac_table[index] == 0) {
 190             return index;
 191         }
 192     }
 193
 194     return -1;
 195 }
 196
 197 void qemu_macaddr_default_if_unset(MACAddr *macaddr)
 198 {
 199     static const MACAddr zero = { .a = { 0,0,0,0,0,0 } };
 200     static const MACAddr base = { .a = { 0x52, 0x54, 0x00, 0x12, 0x34, 0 } };
 201
 202     if (memcmp(macaddr, &zero, sizeof(zero)) != 0) {
 203         if (memcmp(macaddr->a, &base.a, (sizeof(base.a) - 1)) != 0) {
 204             return;
 205         } else {
 206             qemu_macaddr_set_used(macaddr);
 207             return;
 208         }
 209     }
 210
 211     macaddr->a[0] = 0x52;
 212     macaddr->a[1] = 0x54;
 213     macaddr->a[2] = 0x00;
 214     macaddr->a[3] = 0x12;
 215     macaddr->a[4] = 0x34;
 216     macaddr->a[5] = qemu_macaddr_get_free();
 217     qemu_macaddr_set_used(macaddr);
 218 }
 219
 220 /**
 221  * Generate a name for net client
 222  *
 223  * Only net clients created with the legacy -net option and NICs need this.
 224  */
 225 static char *assign_name(NetClientState *nc1, const char *model)
 226 {
 227     NetClientState *nc;
 228     int id = 0;
 229
 230     QTAILQ_FOREACH(nc, &net_clients, next) {
 231         if (nc == nc1) {
 232             continue;
 233         }
 234         if (strcmp(nc->model, model) == 0) {
 235             id++;
 236         }
 237     }
 238
 239     return g_strdup_printf("%s.%d", model, id);
 240 }
 241
 242 static void qemu_net_client_destructor(NetClientState *nc)
 243 {
 244     g_free(nc);
 245 }
 246 static ssize_t qemu_deliver_packet_iov(NetClientState *sender,
 247                                        unsigned flags,
 248                                        const struct iovec *iov,
 249                                        int iovcnt,
 250                                        void *opaque);
 251
 252 static void qemu_net_client_setup(NetClientState *nc,
 253                                   NetClientInfo *info,
 254                                   NetClientState *peer,
 255                                   const char *model,
 256                                   const char *name,
 257                                   NetClientDestructor *destructor)
 258 {
 259     nc->info = info;
 260     nc->model = g_strdup(model);
 261     if (name) {
 262         nc->name = g_strdup(name);
 263     } else {
 264         nc->name = assign_name(nc, model);
 265     }
 266
 267     if (peer) {
 268         assert(!peer->peer);
 269         nc->peer = peer;
 270         peer->peer = nc;
 271     }
 272     QTAILQ_INSERT_TAIL(&net_clients, nc, next);
 273
 274     nc->incoming_queue = qemu_new_net_queue(qemu_deliver_packet_iov, nc);
 275     nc->destructor = destructor;
 276     QTAILQ_INIT(&nc->filters);
 277 }
 278
 279 NetClientState *qemu_new_net_client(NetClientInfo *info,
 280                                     NetClientState *peer,
 281                                     const char *model,
 282                                     const char *name)
 283 {
 284     NetClientState *nc;
 285
 286     assert(info->size >= sizeof(NetClientState));
 287
 288     nc = g_malloc0(info->size);
 289     qemu_net_client_setup(nc, info, peer, model, name,
 290                           qemu_net_client_destructor);
 291
 292     return nc;
 293 }
 294
 295 NICState *qemu_new_nic(NetClientInfo *info,
 296                        NICConf *conf,
 297                        const char *model,
 298                        const char *name,
 299                        void *opaque)
 300 {
 301     NetClientState **peers = conf->peers.ncs;
 302     NICState *nic;
 303     int i, queues = MAX(1, conf->peers.queues);
 304
 305     assert(info->type == NET_CLIENT_DRIVER_NIC);
 306     assert(info->size >= sizeof(NICState));
 307
 308     nic = g_malloc0(info->size + sizeof(NetClientState) * queues);
 309     nic->ncs = (void *)nic + info->size;
 310     nic->conf = conf;
 311     nic->opaque = opaque;
 312
 313     for (i = 0; i < queues; i++) {
 314         qemu_net_client_setup(&nic->ncs[i], info, peers[i], model, name,
 315                               NULL);
 316         nic->ncs[i].queue_index = i;
 317     }
 318
 319     return nic;
 320 }
 321
 322 NetClientState *qemu_get_subqueue(NICState *nic, int queue_index)
 323 {
 324     return nic->ncs + queue_index;
 325 }
 326
 327 NetClientState *qemu_get_queue(NICState *nic)
 328 {
 329     return qemu_get_subqueue(nic, 0);
 330 }
 331
 332 NICState *qemu_get_nic(NetClientState *nc)
 333 {
 334     NetClientState *nc0 = nc - nc->queue_index;
 335
 336     return (NICState *)((void *)nc0 - nc->info->size);
 337 }
 338
 339 void *qemu_get_nic_opaque(NetClientState *nc)
 340 {
 341     NICState *nic = qemu_get_nic(nc);
 342
 343     return nic->opaque;
 344 }
 345
 346 static void qemu_cleanup_net_client(NetClientState *nc)
 347 {
 348     QTAILQ_REMOVE(&net_clients, nc, next);
 349
 350     if (nc->info->cleanup) {
 351         nc->info->cleanup(nc);
 352     }
 353 }
 354
 355 static void qemu_free_net_client(NetClientState *nc)
 356 {
 357     if (nc->incoming_queue) {
 358         qemu_del_net_queue(nc->incoming_queue);
 359     }
 360     if (nc->peer) {
 361         nc->peer->peer = NULL;
 362     }
 363     g_free(nc->name);
 364     g_free(nc->model);
 365     if (nc->destructor) {
 366         nc->destructor(nc);
 367     }
 368 }
 369
 370 void qemu_del_net_client(NetClientState *nc)
 371 {
 372     NetClientState *ncs[MAX_QUEUE_NUM];
 373     int queues, i;
 374     NetFilterState *nf, *next;
 375
 376     assert(nc->info->type != NET_CLIENT_DRIVER_NIC);
 377
 378     /* If the NetClientState belongs to a multiqueue backend, we will change all
 379      * other NetClientStates also.
 380      */
 381     queues = qemu_find_net_clients_except(nc->name, ncs,
 382                                           NET_CLIENT_DRIVER_NIC,
 383                                           MAX_QUEUE_NUM);
 384     assert(queues != 0);
 385
 386     QTAILQ_FOREACH_SAFE(nf, &nc->filters, next, next) {
 387         object_unparent(OBJECT(nf));
 388     }
 389
 390     /* If there is a peer NIC, delete and cleanup client, but do not free. */
 391     if (nc->peer && nc->peer->info->type == NET_CLIENT_DRIVER_NIC) {
 392         NICState *nic = qemu_get_nic(nc->peer);
 393         if (nic->peer_deleted) {
 394             return;
 395         }
 396         nic->peer_deleted = true;
 397
 398         for (i = 0; i < queues; i++) {
 399             ncs[i]->peer->link_down = true;
 400         }
 401
 402         if (nc->peer->info->link_status_changed) {
 403             nc->peer->info->link_status_changed(nc->peer);
 404         }
 405
 406         for (i = 0; i < queues; i++) {
 407             qemu_cleanup_net_client(ncs[i]);
 408         }
 409
 410         return;
 411     }
 412
 413     for (i = 0; i < queues; i++) {
 414         qemu_cleanup_net_client(ncs[i]);
 415         qemu_free_net_client(ncs[i]);
 416     }
 417 }
 418
 419 void qemu_del_nic(NICState *nic)
 420 {
 421     int i, queues = MAX(nic->conf->peers.queues, 1);
 422
 423     qemu_macaddr_set_free(&nic->conf->macaddr);
 424
 425     /* If this is a peer NIC and peer has already been deleted, free it now. */
 426     if (nic->peer_deleted) {
 427         for (i = 0; i < queues; i++) {
 428             qemu_free_net_client(qemu_get_subqueue(nic, i)->peer);
 429         }
 430     }
 431
 432     for (i = queues - 1; i >= 0; i--) {
 433         NetClientState *nc = qemu_get_subqueue(nic, i);
 434
 435         qemu_cleanup_net_client(nc);
 436         qemu_free_net_client(nc);
 437     }
 438
 439     g_free(nic);
 440 }
 441
 442 void qemu_foreach_nic(qemu_nic_foreach func, void *opaque)
 443 {
 444     NetClientState *nc;
 445
 446     QTAILQ_FOREACH(nc, &net_clients, next) {
 447         if (nc->info->type == NET_CLIENT_DRIVER_NIC) {
 448             if (nc->queue_index == 0) {
 449                 func(qemu_get_nic(nc), opaque);
 450             }
 451         }
 452     }
 453 }
 454
 455 bool qemu_has_ufo(NetClientState *nc)
 456 {
 457     if (!nc || !nc->info->has_ufo) {
 458         return false;
 459     }
 460
 461     return nc->info->has_ufo(nc);
 462 }
 463
 464 bool qemu_has_vnet_hdr(NetClientState *nc)
 465 {
 466     if (!nc || !nc->info->has_vnet_hdr) {
 467         return false;
 468     }
 469
 470     return nc->info->has_vnet_hdr(nc);
 471 }
 472
 473 bool qemu_has_vnet_hdr_len(NetClientState *nc, int len)
 474 {
 475     if (!nc || !nc->info->has_vnet_hdr_len) {
 476         return false;
 477     }
 478
 479     return nc->info->has_vnet_hdr_len(nc, len);
 480 }
 481
 482 void qemu_using_vnet_hdr(NetClientState *nc, bool enable)
 483 {
 484     if (!nc || !nc->info->using_vnet_hdr) {
 485         return;
 486     }
 487
 488     nc->info->using_vnet_hdr(nc, enable);
 489 }
 490
 491 void qemu_set_offload(NetClientState *nc, int csum, int tso4, int tso6,
 492                           int ecn, int ufo)
 493 {
 494     if (!nc || !nc->info->set_offload) {
 495         return;
 496     }
 497
 498     nc->info->set_offload(nc, csum, tso4, tso6, ecn, ufo);
 499 }
 500
 501 void qemu_set_vnet_hdr_len(NetClientState *nc, int len)
 502 {
 503     if (!nc || !nc->info->set_vnet_hdr_len) {
 504         return;
 505     }
 506
 507     nc->vnet_hdr_len = len;
 508     nc->info->set_vnet_hdr_len(nc, len);
 509 }
 510
 511 int qemu_set_vnet_le(NetClientState *nc, bool is_le)
 512 {
 513 #ifdef HOST_WORDS_BIGENDIAN
 514     if (!nc || !nc->info->set_vnet_le) {
 515         return -ENOSYS;
 516     }
 517
 518     return nc->info->set_vnet_le(nc, is_le);
 519 #else
 520     return 0;
 521 #endif
 522 }
 523
 524 int qemu_set_vnet_be(NetClientState *nc, bool is_be)
 525 {
 526 #ifdef HOST_WORDS_BIGENDIAN
 527     return 0;
 528 #else
 529     if (!nc || !nc->info->set_vnet_be) {
 530         return -ENOSYS;
 531     }
 532
 533     return nc->info->set_vnet_be(nc, is_be);
 534 #endif
 535 }
 536
 537 int qemu_can_send_packet(NetClientState *sender)
 538 {
 539     int vm_running = runstate_is_running();
 540
 541     if (!vm_running) {
 542         return 0;
 543     }
 544
 545     if (!sender->peer) {
 546         return 1;
 547     }
 548
 549     if (sender->peer->receive_disabled) {
 550         return 0;
 551     } else if (sender->peer->info->can_receive &&
 552                !sender->peer->info->can_receive(sender->peer)) {
 553         return 0;
 554     }
 555     return 1;
 556 }
 557
 558 static ssize_t filter_receive_iov(NetClientState *nc,
 559                                   NetFilterDirection direction,
 560                                   NetClientState *sender,
 561                                   unsigned flags,
 562                                   const struct iovec *iov,
 563                                   int iovcnt,
 564                                   NetPacketSent *sent_cb)
 565 {
 566     ssize_t ret = 0;
 567     NetFilterState *nf = NULL;
 568
 569     if (direction == NET_FILTER_DIRECTION_TX) {
 570         QTAILQ_FOREACH(nf, &nc->filters, next) {
 571             ret = qemu_netfilter_receive(nf, direction, sender, flags, iov,
 572                                          iovcnt, sent_cb);
 573             if (ret) {
 574                 return ret;
 575             }
 576         }
 577     } else {
 578         QTAILQ_FOREACH_REVERSE(nf, &nc->filters, next) {
 579             ret = qemu_netfilter_receive(nf, direction, sender, flags, iov,
 580                                          iovcnt, sent_cb);
 581             if (ret) {
 582                 return ret;
 583             }
 584         }
 585     }
 586
 587     return ret;
 588 }
 589
 590 static ssize_t filter_receive(NetClientState *nc,
 591                               NetFilterDirection direction,
 592                               NetClientState *sender,
 593                               unsigned flags,
 594                               const uint8_t *data,
 595                               size_t size,
 596                               NetPacketSent *sent_cb)
 597 {
 598     struct iovec iov = {
 599         .iov_base = (void *)data,
 600         .iov_len = size
 601     };
 602
 603     return filter_receive_iov(nc, direction, sender, flags, &iov, 1, sent_cb);
 604 }
 605
 606 void qemu_purge_queued_packets(NetClientState *nc)
 607 {
 608     if (!nc->peer) {
 609         return;
 610     }
 611
 612     qemu_net_queue_purge(nc->peer->incoming_queue, nc);
 613 }
 614
 615 void qemu_flush_or_purge_queued_packets(NetClientState *nc, bool purge)
 616 {
 617     nc->receive_disabled = 0;
 618
 619     if (nc->peer && nc->peer->info->type == NET_CLIENT_DRIVER_HUBPORT) {
 620         if (net_hub_flush(nc->peer)) {
 621             qemu_notify_event();
 622         }
 623     }
 624     if (qemu_net_queue_flush(nc->incoming_queue)) {
 625         /* We emptied the queue successfully, signal to the IO thread to repoll
 626          * the file descriptor (for tap, for example).
 627          */
 628         qemu_notify_event();
 629     } else if (purge) {
 630         /* Unable to empty the queue, purge remaining packets */
 631         qemu_net_queue_purge(nc->incoming_queue, nc);
 632     }
 633 }
 634
 635 void qemu_flush_queued_packets(NetClientState *nc)
 636 {
 637     qemu_flush_or_purge_queued_packets(nc, false);
 638 }
 639
 640 static ssize_t qemu_send_packet_async_with_flags(NetClientState *sender,
 641                                                  unsigned flags,
 642                                                  const uint8_t *buf, int size,
 643                                                  NetPacketSent *sent_cb)
 644 {
 645     NetQueue *queue;
 646     int ret;
 647
 648 #ifdef DEBUG_NET
 649     printf("qemu_send_packet_async:\n");
 650     qemu_hexdump((const char *)buf, stdout, "net", size);
 651 #endif
 652
 653     if (sender->link_down || !sender->peer) {
 654         return size;
 655     }
 656
 657     /* Let filters handle the packet first */
 658     ret = filter_receive(sender, NET_FILTER_DIRECTION_TX,
 659                          sender, flags, buf, size, sent_cb);
 660     if (ret) {
 661         return ret;
 662     }
 663
 664     ret = filter_receive(sender->peer, NET_FILTER_DIRECTION_RX,
 665                          sender, flags, buf, size, sent_cb);
 666     if (ret) {
 667         return ret;
 668     }
 669
 670     queue = sender->peer->incoming_queue;
 671
 672     return qemu_net_queue_send(queue, sender, flags, buf, size, sent_cb);
 673 }
 674
 675 ssize_t qemu_send_packet_async(NetClientState *sender,
 676                                const uint8_t *buf, int size,
 677                                NetPacketSent *sent_cb)
 678 {
 679     return qemu_send_packet_async_with_flags(sender, QEMU_NET_PACKET_FLAG_NONE,
 680                                              buf, size, sent_cb);
 681 }
 682
 683 ssize_t qemu_send_packet(NetClientState *nc, const uint8_t *buf, int size)
 684 {
 685     return qemu_send_packet_async(nc, buf, size, NULL);
 686 }
 687
 688 ssize_t qemu_send_packet_raw(NetClientState *nc, const uint8_t *buf, int size)
 689 {
 690     return qemu_send_packet_async_with_flags(nc, QEMU_NET_PACKET_FLAG_RAW,
 691                                              buf, size, NULL);
 692 }
 693
 694 static ssize_t nc_sendv_compat(NetClientState *nc, const struct iovec *iov,
 695                                int iovcnt, unsigned flags)
 696 {
 697     uint8_t *buf = NULL;
 698     uint8_t *buffer;
 699     size_t offset;
 700     ssize_t ret;
 701
 702     if (iovcnt == 1) {
 703         buffer = iov[0].iov_base;
 704         offset = iov[0].iov_len;
 705     } else {
 706         offset = iov_size(iov, iovcnt);
 707         if (offset > NET_BUFSIZE) {
 708             return -1;
 709         }
 710         buf = g_malloc(offset);
 711         buffer = buf;
 712         offset = iov_to_buf(iov, iovcnt, 0, buf, offset);
 713     }
 714
 715     if (flags & QEMU_NET_PACKET_FLAG_RAW && nc->info->receive_raw) {
 716         ret = nc->info->receive_raw(nc, buffer, offset);
 717     } else {
 718         ret = nc->info->receive(nc, buffer, offset);
 719     }
 720
 721     g_free(buf);
 722     return ret;
 723 }
 724
 725 static ssize_t qemu_deliver_packet_iov(NetClientState *sender,
 726                                        unsigned flags,
 727                                        const struct iovec *iov,
 728                                        int iovcnt,
 729                                        void *opaque)
 730 {
 731     NetClientState *nc = opaque;
 732     int ret;
 733
 734
 735     if (nc->link_down) {
 736         return iov_size(iov, iovcnt);
 737     }
 738
 739     if (nc->receive_disabled) {
 740         return 0;
 741     }
 742
 743     if (nc->info->receive_iov && !(flags & QEMU_NET_PACKET_FLAG_RAW)) {
 744         ret = nc->info->receive_iov(nc, iov, iovcnt);
 745     } else {
 746         ret = nc_sendv_compat(nc, iov, iovcnt, flags);
 747     }
 748
 749     if (ret == 0) {
 750         nc->receive_disabled = 1;
 751     }
 752
 753     return ret;
 754 }
 755
 756 ssize_t qemu_sendv_packet_async(NetClientState *sender,
 757                                 const struct iovec *iov, int iovcnt,
 758                                 NetPacketSent *sent_cb)
 759 {
 760     NetQueue *queue;
 761     size_t size = iov_size(iov, iovcnt);
 762     int ret;
 763
 764     if (size > NET_BUFSIZE) {
 765         return size;
 766     }
 767
 768     if (sender->link_down || !sender->peer) {
 769         return size;
 770     }
 771
 772     /* Let filters handle the packet first */
 773     ret = filter_receive_iov(sender, NET_FILTER_DIRECTION_TX, sender,
 774                              QEMU_NET_PACKET_FLAG_NONE, iov, iovcnt, sent_cb);
 775     if (ret) {
 776         return ret;
 777     }
 778
 779     ret = filter_receive_iov(sender->peer, NET_FILTER_DIRECTION_RX, sender,
 780                              QEMU_NET_PACKET_FLAG_NONE, iov, iovcnt, sent_cb);
 781     if (ret) {
 782         return ret;
 783     }
 784
 785     queue = sender->peer->incoming_queue;
 786
 787     return qemu_net_queue_send_iov(queue, sender,
 788                                    QEMU_NET_PACKET_FLAG_NONE,
 789                                    iov, iovcnt, sent_cb);
 790 }
 791
 792 ssize_t
 793 qemu_sendv_packet(NetClientState *nc, const struct iovec *iov, int iovcnt)
 794 {
 795     return qemu_sendv_packet_async(nc, iov, iovcnt, NULL);
 796 }
 797
 798 NetClientState *qemu_find_netdev(const char *id)
 799 {
 800     NetClientState *nc;
 801
 802     QTAILQ_FOREACH(nc, &net_clients, next) {
 803         if (nc->info->type == NET_CLIENT_DRIVER_NIC)
 804             continue;
 805         if (!strcmp(nc->name, id)) {
 806             return nc;
 807         }
 808     }
 809
 810     return NULL;
 811 }
 812
 813 int qemu_find_net_clients_except(const char *id, NetClientState **ncs,
 814                                  NetClientDriver type, int max)
 815 {
 816     NetClientState *nc;
 817     int ret = 0;
 818
 819     QTAILQ_FOREACH(nc, &net_clients, next) {
 820         if (nc->info->type == type) {
 821             continue;
 822         }
 823         if (!id || !strcmp(nc->name, id)) {
 824             if (ret < max) {
 825                 ncs[ret] = nc;
 826             }
 827             ret++;
 828         }
 829     }
 830
 831     return ret;
 832 }
 833
 834 static int nic_get_free_idx(void)
 835 {
 836     int index;
 837
 838     for (index = 0; index < MAX_NICS; index++)
 839         if (!nd_table[index].used)
 840             return index;
 841     return -1;
 842 }
 843
 844 int qemu_show_nic_models(const char *arg, const char *const *models)
 845 {
 846     int i;
 847
 848     if (!arg || !is_help_option(arg)) {
 849         return 0;
 850     }
 851
 852     printf("Supported NIC models:\n");
 853     for (i = 0 ; models[i]; i++) {
 854         printf("%s\n", models[i]);
 855     }
 856     return 1;
 857 }
 858
 859 void qemu_check_nic_model(NICInfo *nd, const char *model)
 860 {
 861     const char *models[2];
 862
 863     models[0] = model;
 864     models[1] = NULL;
 865
 866     if (qemu_show_nic_models(nd->model, models))
 867         exit(0);
 868     if (qemu_find_nic_model(nd, models, model) < 0)
 869         exit(1);
 870 }
 871
 872 int qemu_find_nic_model(NICInfo *nd, const char * const *models,
 873                         const char *default_model)
 874 {
 875     int i;
 876
 877     if (!nd->model)
 878         nd->model = g_strdup(default_model);
 879
 880     for (i = 0 ; models[i]; i++) {
 881         if (strcmp(nd->model, models[i]) == 0)
 882             return i;
 883     }
 884
 885     error_report("Unsupported NIC model: %s", nd->model);
 886     return -1;
 887 }
 888
 889 static int net_init_nic(const Netdev *netdev, const char *name,
 890                         NetClientState *peer, Error **errp)
 891 {
 892     int idx;
 893     NICInfo *nd;
 894     const NetLegacyNicOptions *nic;
 895
 896     assert(netdev->type == NET_CLIENT_DRIVER_NIC);
 897     nic = &netdev->u.nic;
 898
 899     idx = nic_get_free_idx();
 900     if (idx == -1 || nb_nics >= MAX_NICS) {
 901         error_setg(errp, "too many NICs");
 902         return -1;
 903     }
 904
 905     nd = &nd_table[idx];
 906
 907     memset(nd, 0, sizeof(*nd));
 908
 909     if (nic->has_netdev) {
 910         nd->netdev = qemu_find_netdev(nic->netdev);
 911         if (!nd->netdev) {
 912             error_setg(errp, "netdev '%s' not found", nic->netdev);
 913             return -1;
 914         }
 915     } else {
 916         assert(peer);
 917         nd->netdev = peer;
 918     }
 919     nd->name = g_strdup(name);
 920     if (nic->has_model) {
 921         nd->model = g_strdup(nic->model);
 922     }
 923     if (nic->has_addr) {
 924         nd->devaddr = g_strdup(nic->addr);
 925     }
 926
 927     if (nic->has_macaddr &&
 928         net_parse_macaddr(nd->macaddr.a, nic->macaddr) < 0) {
 929         error_setg(errp, "invalid syntax for ethernet address");
 930         return -1;
 931     }
 932     if (nic->has_macaddr &&
 933         is_multicast_ether_addr(nd->macaddr.a)) {
 934         error_setg(errp,
 935                    "NIC cannot have multicast MAC address (odd 1st byte)");
 936         return -1;
 937     }
 938     qemu_macaddr_default_if_unset(&nd->macaddr);
 939
 940     if (nic->has_vectors) {
 941         if (nic->vectors > 0x7ffffff) {
 942             error_setg(errp, "invalid # of vectors: %"PRIu32, nic->vectors);
 943             return -1;
 944         }
 945         nd->nvectors = nic->vectors;
 946     } else {
 947         nd->nvectors = DEV_NVECTORS_UNSPECIFIED;
 948     }
 949
 950     nd->used = 1;
 951     nb_nics++;
 952
 953     return idx;
 954 }
 955
 956
 957 static int (* const net_client_init_fun[NET_CLIENT_DRIVER__MAX])(
 958     const Netdev *netdev,
 959     const char *name,
 960     NetClientState *peer, Error **errp) = {
 961         [NET_CLIENT_DRIVER_NIC]       = net_init_nic,
 962 #ifdef CONFIG_SLIRP
 963         [NET_CLIENT_DRIVER_USER]      = net_init_slirp,
 964 #endif
 965         [NET_CLIENT_DRIVER_TAP]       = net_init_tap,
 966         [NET_CLIENT_DRIVER_SOCKET]    = net_init_socket,
 967 #ifdef CONFIG_VDE
 968         [NET_CLIENT_DRIVER_VDE]       = net_init_vde,
 969 #endif
 970 #ifdef CONFIG_NETMAP
 971         [NET_CLIENT_DRIVER_NETMAP]    = net_init_netmap,
 972 #endif
 973 #ifdef CONFIG_NET_BRIDGE
 974         [NET_CLIENT_DRIVER_BRIDGE]    = net_init_bridge,
 975 #endif
 976         [NET_CLIENT_DRIVER_HUBPORT]   = net_init_hubport,
 977 #ifdef CONFIG_VHOST_NET_USER
 978         [NET_CLIENT_DRIVER_VHOST_USER] = net_init_vhost_user,
 979 #endif
 980 #ifdef CONFIG_L2TPV3
 981         [NET_CLIENT_DRIVER_L2TPV3]    = net_init_l2tpv3,
 982 #endif
 983 };
 984
 985
 986 static int net_client_init1(const void *object, bool is_netdev, Error **errp)
 987 {
 988     Netdev legacy = {0};
 989     const Netdev *netdev;
 990     const char *name;
 991     NetClientState *peer = NULL;
 992
 993     if (is_netdev) {
 994         netdev = object;
 995         name = netdev->id;
 996
 997         if (netdev->type == NET_CLIENT_DRIVER_NIC ||
 998             !net_client_init_fun[netdev->type]) {
 999             error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "type",
1000                        "a netdev backend type");
1001             return -1;
1002         }
1003     } else {
1004         const NetLegacy *net = object;
1005         const NetLegacyOptions *opts = net->opts;
1006         legacy.id = net->id;
1007         netdev = &legacy;
1008         /* missing optional values have been initialized to "all bits zero" */
1009         name = net->has_id ? net->id : net->name;
1010
1011         if (net->has_name) {
1012             warn_report("The 'name' parameter is deprecated, use 'id' instead");
1013         }
1014
1015         /* Map the old options to the new flat type */
1016         switch (opts->type) {
1017         case NET_LEGACY_OPTIONS_TYPE_NONE:
1018             return 0; /* nothing to do */
1019         case NET_LEGACY_OPTIONS_TYPE_NIC:
1020             legacy.type = NET_CLIENT_DRIVER_NIC;
1021             legacy.u.nic = opts->u.nic;
1022             break;
1023         case NET_LEGACY_OPTIONS_TYPE_USER:
1024             legacy.type = NET_CLIENT_DRIVER_USER;
1025             legacy.u.user = opts->u.user;
1026             break;
1027         case NET_LEGACY_OPTIONS_TYPE_TAP:
1028             legacy.type = NET_CLIENT_DRIVER_TAP;
1029             legacy.u.tap = opts->u.tap;
1030             break;
1031         case NET_LEGACY_OPTIONS_TYPE_L2TPV3:
1032             legacy.type = NET_CLIENT_DRIVER_L2TPV3;
1033             legacy.u.l2tpv3 = opts->u.l2tpv3;
1034             break;
1035         case NET_LEGACY_OPTIONS_TYPE_SOCKET:
1036             legacy.type = NET_CLIENT_DRIVER_SOCKET;
1037             legacy.u.socket = opts->u.socket;
1038             break;
1039         case NET_LEGACY_OPTIONS_TYPE_VDE:
1040             legacy.type = NET_CLIENT_DRIVER_VDE;
1041             legacy.u.vde = opts->u.vde;
1042             break;
1043         case NET_LEGACY_OPTIONS_TYPE_BRIDGE:
1044             legacy.type = NET_CLIENT_DRIVER_BRIDGE;
1045             legacy.u.bridge = opts->u.bridge;
1046             break;
1047         case NET_LEGACY_OPTIONS_TYPE_NETMAP:
1048             legacy.type = NET_CLIENT_DRIVER_NETMAP;
1049             legacy.u.netmap = opts->u.netmap;
1050             break;
1051         case NET_LEGACY_OPTIONS_TYPE_VHOST_USER:
1052             legacy.type = NET_CLIENT_DRIVER_VHOST_USER;
1053             legacy.u.vhost_user = opts->u.vhost_user;
1054             break;
1055         default:
1056             abort();
1057         }
1058
1059         if (!net_client_init_fun[netdev->type]) {
1060             error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "type",
1061                        "a net backend type (maybe it is not compiled "
1062                        "into this binary)");
1063             return -1;
1064         }
1065
1066         /* Do not add to a hub if it's a nic with a netdev= parameter. */
1067         if (netdev->type != NET_CLIENT_DRIVER_NIC ||
1068             !opts->u.nic.has_netdev) {
1069             peer = net_hub_add_port(0, NULL, NULL);
1070         }
1071     }
1072
1073     if (net_client_init_fun[netdev->type](netdev, name, peer, errp) < 0) {
1074         /* FIXME drop when all init functions store an Error */
1075         if (errp && !*errp) {
1076             error_setg(errp, QERR_DEVICE_INIT_FAILED,
1077                        NetClientDriver_str(netdev->type));
1078         }
1079         return -1;
1080     }
1081     return 0;
1082 }
1083
1084 static void show_netdevs(void)
1085 {
1086     int idx;
1087     const char *available_netdevs[] = {
1088         "socket",
1089         "hubport",
1090         "tap",
1091 #ifdef CONFIG_SLIRP
1092         "user",
1093 #endif
1094 #ifdef CONFIG_L2TPV3
1095         "l2tpv3",
1096 #endif
1097 #ifdef CONFIG_VDE
1098         "vde",
1099 #endif
1100 #ifdef CONFIG_NET_BRIDGE
1101         "bridge",
1102 #endif
1103 #ifdef CONFIG_NETMAP
1104         "netmap",
1105 #endif
1106 #ifdef CONFIG_POSIX
1107         "vhost-user",
1108 #endif
1109     };
1110
1111     printf("Available netdev backend types:\n");
1112     for (idx = 0; idx < ARRAY_SIZE(available_netdevs); idx++) {
1113         puts(available_netdevs[idx]);
1114     }
1115 }
1116
1117 static int net_client_init(QemuOpts *opts, bool is_netdev, Error **errp)
1118 {
1119     gchar **substrings = NULL;
1120     void *object = NULL;
1121     Error *err = NULL;
1122     int ret = -1;
1123     Visitor *v = opts_visitor_new(opts);
1124
1125     const char *type = qemu_opt_get(opts, "type");
1126
1127     if (is_netdev && type && is_help_option(type)) {
1128         show_netdevs();
1129         exit(0);
1130     } else {
1131         /* Parse convenience option format ip6-net=fec0::0[/64] */
1132         const char *ip6_net = qemu_opt_get(opts, "ipv6-net");
1133
1134         if (ip6_net) {
1135             char *prefix_addr;
1136             unsigned long prefix_len = 64; /* Default 64bit prefix length. */
1137
1138             substrings = g_strsplit(ip6_net, "/", 2);
1139             if (!substrings || !substrings[0]) {
1140                 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "ipv6-net",
1141                            "a valid IPv6 prefix");
1142                 goto out;
1143             }
1144
1145             prefix_addr = substrings[0];
1146
1147             if (substrings[1]) {
1148                 /* User-specified prefix length.  */
1149                 int err;
1150
1151                 err = qemu_strtoul(substrings[1], NULL, 10, &prefix_len);
1152                 if (err) {
1153                     error_setg(errp, QERR_INVALID_PARAMETER_VALUE,
1154                                "ipv6-prefixlen", "a number");
1155                     goto out;
1156                 }
1157             }
1158
1159             qemu_opt_set(opts, "ipv6-prefix", prefix_addr, &error_abort);
1160             qemu_opt_set_number(opts, "ipv6-prefixlen", prefix_len,
1161                                 &error_abort);
1162             qemu_opt_unset(opts, "ipv6-net");
1163         }
1164     }
1165
1166     if (is_netdev) {
1167         visit_type_Netdev(v, NULL, (Netdev **)&object, &err);
1168     } else {
1169         visit_type_NetLegacy(v, NULL, (NetLegacy **)&object, &err);
1170     }
1171
1172     if (!err) {
1173         ret = net_client_init1(object, is_netdev, &err);
1174     }
1175
1176     if (is_netdev) {
1177         qapi_free_Netdev(object);
1178     } else {
1179         qapi_free_NetLegacy(object);
1180     }
1181
1182 out:
1183     error_propagate(errp, err);
1184     g_strfreev(substrings);
1185     visit_free(v);
1186     return ret;
1187 }
1188
1189 void netdev_add(QemuOpts *opts, Error **errp)
1190 {
1191     net_client_init(opts, true, errp);
1192 }
1193
1194 void qmp_netdev_add(QDict *qdict, QObject **ret, Error **errp)
1195 {
1196     Error *local_err = NULL;
1197     QemuOptsList *opts_list;
1198     QemuOpts *opts;
1199
1200     opts_list = qemu_find_opts_err("netdev", &local_err);
1201     if (local_err) {
1202         goto out;
1203     }
1204
1205     opts = qemu_opts_from_qdict(opts_list, qdict, &local_err);
1206     if (local_err) {
1207         goto out;
1208     }
1209
1210     netdev_add(opts, &local_err);
1211     if (local_err) {
1212         qemu_opts_del(opts);
1213         goto out;
1214     }
1215
1216 out:
1217     error_propagate(errp, local_err);
1218 }
1219
1220 void qmp_netdev_del(const char *id, Error **errp)
1221 {
1222     NetClientState *nc;
1223     QemuOpts *opts;
1224
1225     nc = qemu_find_netdev(id);
1226     if (!nc) {
1227         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1228                   "Device '%s' not found", id);
1229         return;
1230     }
1231
1232     opts = qemu_opts_find(qemu_find_opts_err("netdev", NULL), id);
1233     if (!opts) {
1234         error_setg(errp, "Device '%s' is not a netdev", id);
1235         return;
1236     }
1237
1238     qemu_del_net_client(nc);
1239     qemu_opts_del(opts);
1240 }
1241
1242 static void netfilter_print_info(Monitor *mon, NetFilterState *nf)
1243 {
1244     char *str;
1245     ObjectProperty *prop;
1246     ObjectPropertyIterator iter;
1247     Visitor *v;
1248
1249     /* generate info str */
1250     object_property_iter_init(&iter, OBJECT(nf));
1251     while ((prop = object_property_iter_next(&iter))) {
1252         if (!strcmp(prop->name, "type")) {
1253             continue;
1254         }
1255         v = string_output_visitor_new(false, &str);
1256         object_property_get(OBJECT(nf), v, prop->name, NULL);
1257         visit_complete(v, &str);
1258         visit_free(v);
1259         monitor_printf(mon, ",%s=%s", prop->name, str);
1260         g_free(str);
1261     }
1262     monitor_printf(mon, "\n");
1263 }
1264
1265 void print_net_client(Monitor *mon, NetClientState *nc)
1266 {
1267     NetFilterState *nf;
1268
1269     monitor_printf(mon, "%s: index=%d,type=%s,%s\n", nc->name,
1270                    nc->queue_index,
1271                    NetClientDriver_str(nc->info->type),
1272                    nc->info_str);
1273     if (!QTAILQ_EMPTY(&nc->filters)) {
1274         monitor_printf(mon, "filters:\n");
1275     }
1276     QTAILQ_FOREACH(nf, &nc->filters, next) {
1277         char *path = object_get_canonical_path_component(OBJECT(nf));
1278
1279         monitor_printf(mon, "  - %s: type=%s", path,
1280                        object_get_typename(OBJECT(nf)));
1281         netfilter_print_info(mon, nf);
1282         g_free(path);
1283     }
1284 }
1285
1286 RxFilterInfoList *qmp_query_rx_filter(bool has_name, const char *name,
1287                                       Error **errp)
1288 {
1289     NetClientState *nc;
1290     RxFilterInfoList *filter_list = NULL, *last_entry = NULL;
1291
1292     QTAILQ_FOREACH(nc, &net_clients, next) {
1293         RxFilterInfoList *entry;
1294         RxFilterInfo *info;
1295
1296         if (has_name && strcmp(nc->name, name) != 0) {
1297             continue;
1298         }
1299
1300         /* only query rx-filter information of NIC */
1301         if (nc->info->type != NET_CLIENT_DRIVER_NIC) {
1302             if (has_name) {
1303                 error_setg(errp, "net client(%s) isn't a NIC", name);
1304                 return NULL;
1305             }
1306             continue;
1307         }
1308
1309         /* only query information on queue 0 since the info is per nic,
1310          * not per queue
1311          */
1312         if (nc->queue_index != 0)
1313             continue;
1314
1315         if (nc->info->query_rx_filter) {
1316             info = nc->info->query_rx_filter(nc);
1317             entry = g_malloc0(sizeof(*entry));
1318             entry->value = info;
1319
1320             if (!filter_list) {
1321                 filter_list = entry;
1322             } else {
1323                 last_entry->next = entry;
1324             }
1325             last_entry = entry;
1326         } else if (has_name) {
1327             error_setg(errp, "net client(%s) doesn't support"
1328                        " rx-filter querying", name);
1329             return NULL;
1330         }
1331
1332         if (has_name) {
1333             break;
1334         }
1335     }
1336
1337     if (filter_list == NULL && has_name) {
1338         error_setg(errp, "invalid net client name: %s", name);
1339     }
1340
1341     return filter_list;
1342 }
1343
1344 void hmp_info_network(Monitor *mon, const QDict *qdict)
1345 {
1346     NetClientState *nc, *peer;
1347     NetClientDriver type;
1348
1349     net_hub_info(mon);
1350
1351     QTAILQ_FOREACH(nc, &net_clients, next) {
1352         peer = nc->peer;
1353         type = nc->info->type;
1354
1355         /* Skip if already printed in hub info */
1356         if (net_hub_id_for_client(nc, NULL) == 0) {
1357             continue;
1358         }
1359
1360         if (!peer || type == NET_CLIENT_DRIVER_NIC) {
1361             print_net_client(mon, nc);
1362         } /* else it's a netdev connected to a NIC, printed with the NIC */
1363         if (peer && type == NET_CLIENT_DRIVER_NIC) {
1364             monitor_printf(mon, " \\ ");
1365             print_net_client(mon, peer);
1366         }
1367     }
1368 }
1369
1370 void colo_notify_filters_event(int event, Error **errp)
1371 {
1372     NetClientState *nc;
1373     NetFilterState *nf;
1374     NetFilterClass *nfc = NULL;
1375     Error *local_err = NULL;
1376
1377     QTAILQ_FOREACH(nc, &net_clients, next) {
1378         QTAILQ_FOREACH(nf, &nc->filters, next) {
1379             nfc = NETFILTER_GET_CLASS(OBJECT(nf));
1380             nfc->handle_event(nf, event, &local_err);
1381             if (local_err) {
1382                 error_propagate(errp, local_err);
1383                 return;
1384             }
1385         }
1386     }
1387 }
1388
1389 void qmp_set_link(const char *name, bool up, Error **errp)
1390 {
1391     NetClientState *ncs[MAX_QUEUE_NUM];
1392     NetClientState *nc;
1393     int queues, i;
1394
1395     queues = qemu_find_net_clients_except(name, ncs,
1396                                           NET_CLIENT_DRIVER__MAX,
1397                                           MAX_QUEUE_NUM);
1398
1399     if (queues == 0) {
1400         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1401                   "Device '%s' not found", name);
1402         return;
1403     }
1404     nc = ncs[0];
1405
1406     for (i = 0; i < queues; i++) {
1407         ncs[i]->link_down = !up;
1408     }
1409
1410     if (nc->info->link_status_changed) {
1411         nc->info->link_status_changed(nc);
1412     }
1413
1414     if (nc->peer) {
1415         /* Change peer link only if the peer is NIC and then notify peer.
1416          * If the peer is a HUBPORT or a backend, we do not change the
1417          * link status.
1418          *
1419          * This behavior is compatible with qemu hubs where there could be
1420          * multiple clients that can still communicate with each other in
1421          * disconnected mode. For now maintain this compatibility.
1422          */
1423         if (nc->peer->info->type == NET_CLIENT_DRIVER_NIC) {
1424             for (i = 0; i < queues; i++) {
1425                 ncs[i]->peer->link_down = !up;
1426             }
1427         }
1428         if (nc->peer->info->link_status_changed) {
1429             nc->peer->info->link_status_changed(nc->peer);
1430         }
1431     }
1432 }
1433
1434 static void net_vm_change_state_handler(void *opaque, int running,
1435                                         RunState state)
1436 {
1437     NetClientState *nc;
1438     NetClientState *tmp;
1439
1440     QTAILQ_FOREACH_SAFE(nc, &net_clients, next, tmp) {
1441         if (running) {
1442             /* Flush queued packets and wake up backends. */
1443             if (nc->peer && qemu_can_send_packet(nc)) {
1444                 qemu_flush_queued_packets(nc->peer);
1445             }
1446         } else {
1447             /* Complete all queued packets, to guarantee we don't modify
1448              * state later when VM is not running.
1449              */
1450             qemu_flush_or_purge_queued_packets(nc, true);
1451         }
1452     }
1453 }
1454
1455 void net_cleanup(void)
1456 {
1457     NetClientState *nc;
1458
1459     /* We may del multiple entries during qemu_del_net_client(),
1460      * so QTAILQ_FOREACH_SAFE() is also not safe here.
1461      */
1462     while (!QTAILQ_EMPTY(&net_clients)) {
1463         nc = QTAILQ_FIRST(&net_clients);
1464         if (nc->info->type == NET_CLIENT_DRIVER_NIC) {
1465             qemu_del_nic(qemu_get_nic(nc));
1466         } else {
1467             qemu_del_net_client(nc);
1468         }
1469     }
1470
1471     qemu_del_vm_change_state_handler(net_change_state_entry);
1472 }
1473
1474 void net_check_clients(void)
1475 {
1476     NetClientState *nc;
1477     int i;
1478
1479     net_hub_check_clients();
1480
1481     QTAILQ_FOREACH(nc, &net_clients, next) {
1482         if (!nc->peer) {
1483             warn_report("%s %s has no peer",
1484                         nc->info->type == NET_CLIENT_DRIVER_NIC
1485                         ? "nic" : "netdev",
1486                         nc->name);
1487         }
1488     }
1489
1490     /* Check that all NICs requested via -net nic actually got created.
1491      * NICs created via -device don't need to be checked here because
1492      * they are always instantiated.
1493      */
1494     for (i = 0; i < MAX_NICS; i++) {
1495         NICInfo *nd = &nd_table[i];
1496         if (nd->used && !nd->instantiated) {
1497             warn_report("requested NIC (%s, model %s) "
1498                         "was not created (not supported by this machine?)",
1499                         nd->name ? nd->name : "anonymous",
1500                         nd->model ? nd->model : "unspecified");
1501         }
1502     }
1503 }
1504
1505 static int net_init_client(void *dummy, QemuOpts *opts, Error **errp)
1506 {
1507     return net_client_init(opts, false, errp);
1508 }
1509
1510 static int net_init_netdev(void *dummy, QemuOpts *opts, Error **errp)
1511 {
1512     return net_client_init(opts, true, errp);
1513 }
1514
1515 /* For the convenience "--nic" parameter */
1516 static int net_param_nic(void *dummy, QemuOpts *opts, Error **errp)
1517 {
1518     char *mac, *nd_id;
1519     int idx, ret;
1520     NICInfo *ni;
1521     const char *type;
1522
1523     type = qemu_opt_get(opts, "type");
1524     if (type && g_str_equal(type, "none")) {
1525         return 0;    /* Nothing to do, default_net is cleared in vl.c */
1526     }
1527
1528     idx = nic_get_free_idx();
1529     if (idx == -1 || nb_nics >= MAX_NICS) {
1530         error_setg(errp, "no more on-board/default NIC slots available");
1531         return -1;
1532     }
1533
1534     if (!type) {
1535         qemu_opt_set(opts, "type", "user", &error_abort);
1536     }
1537
1538     ni = &nd_table[idx];
1539     memset(ni, 0, sizeof(*ni));
1540     ni->model = qemu_opt_get_del(opts, "model");
1541
1542     /* Create an ID if the user did not specify one */
1543     nd_id = g_strdup(qemu_opts_id(opts));
1544     if (!nd_id) {
1545         nd_id = g_strdup_printf("__org.qemu.nic%i\n", idx);
1546         qemu_opts_set_id(opts, nd_id);
1547     }
1548
1549     /* Handle MAC address */
1550     mac = qemu_opt_get_del(opts, "mac");
1551     if (mac) {
1552         ret = net_parse_macaddr(ni->macaddr.a, mac);
1553         g_free(mac);
1554         if (ret) {
1555             error_setg(errp, "invalid syntax for ethernet address");
1556             goto out;
1557         }
1558         if (is_multicast_ether_addr(ni->macaddr.a)) {
1559             error_setg(errp, "NIC cannot have multicast MAC address");
1560             ret = -1;
1561             goto out;
1562         }
1563     }
1564     qemu_macaddr_default_if_unset(&ni->macaddr);
1565
1566     ret = net_client_init(opts, true, errp);
1567     if (ret == 0) {
1568         ni->netdev = qemu_find_netdev(nd_id);
1569         ni->used = true;
1570         nb_nics++;
1571     }
1572
1573 out:
1574     g_free(nd_id);
1575     return ret;
1576 }
1577
1578 int net_init_clients(Error **errp)
1579 {
1580     net_change_state_entry =
1581         qemu_add_vm_change_state_handler(net_vm_change_state_handler, NULL);
1582
1583     QTAILQ_INIT(&net_clients);
1584
1585     if (qemu_opts_foreach(qemu_find_opts("netdev"),
1586                           net_init_netdev, NULL, errp)) {
1587         return -1;
1588     }
1589
1590     if (qemu_opts_foreach(qemu_find_opts("nic"), net_param_nic, NULL, errp)) {
1591         return -1;
1592     }
1593
1594     if (qemu_opts_foreach(qemu_find_opts("net"), net_init_client, NULL, errp)) {
1595         return -1;
1596     }
1597
1598     return 0;
1599 }
1600
1601 int net_client_parse(QemuOptsList *opts_list, const char *optarg)
1602 {
1603     if (!qemu_opts_parse_noisily(opts_list, optarg, true)) {
1604         return -1;
1605     }
1606
1607     return 0;
1608 }
1609
1610 /* From FreeBSD */
1611 /* XXX: optimize */
1612 uint32_t net_crc32(const uint8_t *p, int len)
1613 {
1614     uint32_t crc;
1615     int carry, i, j;
1616     uint8_t b;
1617
1618     crc = 0xffffffff;
1619     for (i = 0; i < len; i++) {
1620         b = *p++;
1621         for (j = 0; j < 8; j++) {
1622             carry = ((crc & 0x80000000L) ? 1 : 0) ^ (b & 0x01);
1623             crc <<= 1;
1624             b >>= 1;
1625             if (carry) {
1626                 crc = ((crc ^ POLYNOMIAL_BE) | carry);
1627             }
1628         }
1629     }
1630
1631     return crc;
1632 }
1633
1634 uint32_t net_crc32_le(const uint8_t *p, int len)
1635 {
1636     uint32_t crc;
1637     int carry, i, j;
1638     uint8_t b;
1639
1640     crc = 0xffffffff;
1641     for (i = 0; i < len; i++) {
1642         b = *p++;
1643         for (j = 0; j < 8; j++) {
1644             carry = (crc & 0x1) ^ (b & 0x01);
1645             crc >>= 1;
1646             b >>= 1;
1647             if (carry) {
1648                 crc ^= POLYNOMIAL_LE;
1649             }
1650         }
1651     }
1652
1653     return crc;
1654 }
1655
1656 QemuOptsList qemu_netdev_opts = {
1657     .name = "netdev",
1658     .implied_opt_name = "type",
1659     .head = QTAILQ_HEAD_INITIALIZER(qemu_netdev_opts.head),
1660     .desc = {
1661         /*
1662          * no elements => accept any params
1663          * validation will happen later
1664          */
1665         { /* end of list */ }
1666     },
1667 };
1668
1669 QemuOptsList qemu_nic_opts = {
1670     .name = "nic",
1671     .implied_opt_name = "type",
1672     .head = QTAILQ_HEAD_INITIALIZER(qemu_nic_opts.head),
1673     .desc = {
1674         /*
1675          * no elements => accept any params
1676          * validation will happen later
1677          */
1678         { /* end of list */ }
1679     },
1680 };
1681
1682 QemuOptsList qemu_net_opts = {
1683     .name = "net",
1684     .implied_opt_name = "type",
1685     .head = QTAILQ_HEAD_INITIALIZER(qemu_net_opts.head),
1686     .desc = {
1687         /*
1688          * no elements => accept any params
1689          * validation will happen later
1690          */
1691         { /* end of list */ }
1692     },
1693 };
1694
1695 void net_socket_rs_init(SocketReadState *rs,
1696                         SocketReadStateFinalize *finalize,
1697                         bool vnet_hdr)
1698 {
1699     rs->state = 0;
1700     rs->vnet_hdr = vnet_hdr;
1701     rs->index = 0;
1702     rs->packet_len = 0;
1703     rs->vnet_hdr_len = 0;
1704     memset(rs->buf, 0, sizeof(rs->buf));
1705     rs->finalize = finalize;
1706 }
1707
1708 /*
1709  * Returns
1710  * 0: success
1711  * -1: error occurs
1712  */
1713 int net_fill_rstate(SocketReadState *rs, const uint8_t *buf, int size)
1714 {
1715     unsigned int l;
1716
1717     while (size > 0) {
1718         /* Reassemble a packet from the network.
1719          * 0 = getting length.
1720          * 1 = getting vnet header length.
1721          * 2 = getting data.
1722          */
1723         switch (rs->state) {
1724         case 0:
1725             l = 4 - rs->index;
1726             if (l > size) {
1727                 l = size;
1728             }
1729             memcpy(rs->buf + rs->index, buf, l);
1730             buf += l;
1731             size -= l;
1732             rs->index += l;
1733             if (rs->index == 4) {
1734                 /* got length */
1735                 rs->packet_len = ntohl(*(uint32_t *)rs->buf);
1736                 rs->index = 0;
1737                 if (rs->vnet_hdr) {
1738                     rs->state = 1;
1739                 } else {
1740                     rs->state = 2;
1741                     rs->vnet_hdr_len = 0;
1742                 }
1743             }
1744             break;
1745         case 1:
1746             l = 4 - rs->index;
1747             if (l > size) {
1748                 l = size;
1749             }
1750             memcpy(rs->buf + rs->index, buf, l);
1751             buf += l;
1752             size -= l;
1753             rs->index += l;
1754             if (rs->index == 4) {
1755                 /* got vnet header length */
1756                 rs->vnet_hdr_len = ntohl(*(uint32_t *)rs->buf);
1757                 rs->index = 0;
1758                 rs->state = 2;
1759             }
1760             break;
1761         case 2:
1762             l = rs->packet_len - rs->index;
1763             if (l > size) {
1764                 l = size;
1765             }
1766             if (rs->index + l <= sizeof(rs->buf)) {
1767                 memcpy(rs->buf + rs->index, buf, l);
1768             } else {
1769                 fprintf(stderr, "serious error: oversized packet received,"
1770                     "connection terminated.\n");
1771                 rs->index = rs->state = 0;
1772                 return -1;
1773             }
1774
1775             rs->index += l;
1776             buf += l;
1777             size -= l;
1778             if (rs->index >= rs->packet_len) {
1779                 rs->index = 0;
1780                 rs->state = 0;
1781                 assert(rs->finalize);
1782                 rs->finalize(rs);
1783             }
1784             break;
1785         }
1786     }
1787
1788     assert(size == 0);
1789     return 0;
1790 }