net/net.c

   1 /*
   2  * QEMU System Emulator
   3  *
   4  * Copyright (c) 2003-2008 Fabrice Bellard
   5  *
   6  * Permission is hereby granted, free of charge, to any person obtaining a copy
   7  * of this software and associated documentation files (the "Software"), to deal
   8  * in the Software without restriction, including without limitation the rights
   9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  10  * copies of the Software, and to permit persons to whom the Software is
  11  * furnished to do so, subject to the following conditions:
  12  *
  13  * The above copyright notice and this permission notice shall be included in
  14  * all copies or substantial portions of the Software.
  15  *
  16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  22  * THE SOFTWARE.
  23  */
  24
  25 #include "qemu/osdep.h"
  26
  27 #include "net/net.h"
  28 #include "clients.h"
  29 #include "hub.h"
  30 #include "hw/qdev-properties.h"
  31 #include "net/slirp.h"
  32 #include "net/eth.h"
  33 #include "util.h"
  34
  35 #include "monitor/monitor.h"
  36 #include "qemu/help_option.h"
  37 #include "qapi/qapi-commands-net.h"
  38 #include "qapi/qapi-visit-net.h"
  39 #include "qapi/qmp/qdict.h"
  40 #include "qapi/qmp/qerror.h"
  41 #include "qemu/error-report.h"
  42 #include "qemu/sockets.h"
  43 #include "qemu/cutils.h"
  44 #include "qemu/config-file.h"
  45 #include "qemu/ctype.h"
  46 #include "qemu/id.h"
  47 #include "qemu/iov.h"
  48 #include "qemu/qemu-print.h"
  49 #include "qemu/main-loop.h"
  50 #include "qemu/option.h"
  51 #include "qemu/keyval.h"
  52 #include "qapi/error.h"
  53 #include "qapi/opts-visitor.h"
  54 #include "sysemu/runstate.h"
  55 #include "net/colo-compare.h"
  56 #include "net/filter.h"
  57 #include "qapi/string-output-visitor.h"
  58 #include "qapi/qobject-input-visitor.h"
  59
  60 /* Net bridge is currently not supported for W32. */
  61 #if !defined(_WIN32)
  62 # define CONFIG_NET_BRIDGE
  63 #endif
  64
  65 static VMChangeStateEntry *net_change_state_entry;
  66 NetClientStateList net_clients;
  67
  68 typedef struct NetdevQueueEntry {
  69     Netdev *nd;
  70     Location loc;
  71     QSIMPLEQ_ENTRY(NetdevQueueEntry) entry;
  72 } NetdevQueueEntry;
  73
  74 typedef QSIMPLEQ_HEAD(, NetdevQueueEntry) NetdevQueue;
  75
  76 static NetdevQueue nd_queue = QSIMPLEQ_HEAD_INITIALIZER(nd_queue);
  77
  78 static GHashTable *nic_model_help;
  79
  80 static int nb_nics;
  81 static NICInfo nd_table[MAX_NICS];
  82
  83 /***********************************************************/
  84 /* network device redirectors */
  85
  86 int convert_host_port(struct sockaddr_in *saddr, const char *host,
  87                       const char *port, Error **errp)
  88 {
  89     struct hostent *he;
  90     const char *r;
  91     long p;
  92
  93     memset(saddr, 0, sizeof(*saddr));
  94
  95     saddr->sin_family = AF_INET;
  96     if (host[0] == '\0') {
  97         saddr->sin_addr.s_addr = 0;
  98     } else {
  99         if (qemu_isdigit(host[0])) {
 100             if (!inet_aton(host, &saddr->sin_addr)) {
 101                 error_setg(errp, "host address '%s' is not a valid "
 102                            "IPv4 address", host);
 103                 return -1;
 104             }
 105         } else {
 106             he = gethostbyname(host);
 107             if (he == NULL) {
 108                 error_setg(errp, "can't resolve host address '%s'", host);
 109                 return -1;
 110             }
 111             saddr->sin_addr = *(struct in_addr *)he->h_addr;
 112         }
 113     }
 114     if (qemu_strtol(port, &r, 0, &p) != 0) {
 115         error_setg(errp, "port number '%s' is invalid", port);
 116         return -1;
 117     }
 118     saddr->sin_port = htons(p);
 119     return 0;
 120 }
 121
 122 int parse_host_port(struct sockaddr_in *saddr, const char *str,
 123                     Error **errp)
 124 {
 125     gchar **substrings;
 126     int ret;
 127
 128     substrings = g_strsplit(str, ":", 2);
 129     if (!substrings || !substrings[0] || !substrings[1]) {
 130         error_setg(errp, "host address '%s' doesn't contain ':' "
 131                    "separating host from port", str);
 132         ret = -1;
 133         goto out;
 134     }
 135
 136     ret = convert_host_port(saddr, substrings[0], substrings[1], errp);
 137
 138 out:
 139     g_strfreev(substrings);
 140     return ret;
 141 }
 142
 143 char *qemu_mac_strdup_printf(const uint8_t *macaddr)
 144 {
 145     return g_strdup_printf("%.2x:%.2x:%.2x:%.2x:%.2x:%.2x",
 146                            macaddr[0], macaddr[1], macaddr[2],
 147                            macaddr[3], macaddr[4], macaddr[5]);
 148 }
 149
 150 void qemu_set_info_str(NetClientState *nc, const char *fmt, ...)
 151 {
 152     va_list ap;
 153
 154     va_start(ap, fmt);
 155     vsnprintf(nc->info_str, sizeof(nc->info_str), fmt, ap);
 156     va_end(ap);
 157 }
 158
 159 void qemu_format_nic_info_str(NetClientState *nc, uint8_t macaddr[6])
 160 {
 161     qemu_set_info_str(nc, "model=%s,macaddr=%02x:%02x:%02x:%02x:%02x:%02x",
 162                       nc->model, macaddr[0], macaddr[1], macaddr[2],
 163                       macaddr[3], macaddr[4], macaddr[5]);
 164 }
 165
 166 static int mac_table[256] = {0};
 167
 168 static void qemu_macaddr_set_used(MACAddr *macaddr)
 169 {
 170     int index;
 171
 172     for (index = 0x56; index < 0xFF; index++) {
 173         if (macaddr->a[5] == index) {
 174             mac_table[index]++;
 175         }
 176     }
 177 }
 178
 179 static void qemu_macaddr_set_free(MACAddr *macaddr)
 180 {
 181     int index;
 182     static const MACAddr base = { .a = { 0x52, 0x54, 0x00, 0x12, 0x34, 0 } };
 183
 184     if (memcmp(macaddr->a, &base.a, (sizeof(base.a) - 1)) != 0) {
 185         return;
 186     }
 187     for (index = 0x56; index < 0xFF; index++) {
 188         if (macaddr->a[5] == index) {
 189             mac_table[index]--;
 190         }
 191     }
 192 }
 193
 194 static int qemu_macaddr_get_free(void)
 195 {
 196     int index;
 197
 198     for (index = 0x56; index < 0xFF; index++) {
 199         if (mac_table[index] == 0) {
 200             return index;
 201         }
 202     }
 203
 204     return -1;
 205 }
 206
 207 void qemu_macaddr_default_if_unset(MACAddr *macaddr)
 208 {
 209     static const MACAddr zero = { .a = { 0,0,0,0,0,0 } };
 210     static const MACAddr base = { .a = { 0x52, 0x54, 0x00, 0x12, 0x34, 0 } };
 211
 212     if (memcmp(macaddr, &zero, sizeof(zero)) != 0) {
 213         if (memcmp(macaddr->a, &base.a, (sizeof(base.a) - 1)) != 0) {
 214             return;
 215         } else {
 216             qemu_macaddr_set_used(macaddr);
 217             return;
 218         }
 219     }
 220
 221     macaddr->a[0] = 0x52;
 222     macaddr->a[1] = 0x54;
 223     macaddr->a[2] = 0x00;
 224     macaddr->a[3] = 0x12;
 225     macaddr->a[4] = 0x34;
 226     macaddr->a[5] = qemu_macaddr_get_free();
 227     qemu_macaddr_set_used(macaddr);
 228 }
 229
 230 /**
 231  * Generate a name for net client
 232  *
 233  * Only net clients created with the legacy -net option and NICs need this.
 234  */
 235 static char *assign_name(NetClientState *nc1, const char *model)
 236 {
 237     NetClientState *nc;
 238     int id = 0;
 239
 240     QTAILQ_FOREACH(nc, &net_clients, next) {
 241         if (nc == nc1) {
 242             continue;
 243         }
 244         if (strcmp(nc->model, model) == 0) {
 245             id++;
 246         }
 247     }
 248
 249     return g_strdup_printf("%s.%d", model, id);
 250 }
 251
 252 static void qemu_net_client_destructor(NetClientState *nc)
 253 {
 254     g_free(nc);
 255 }
 256 static ssize_t qemu_deliver_packet_iov(NetClientState *sender,
 257                                        unsigned flags,
 258                                        const struct iovec *iov,
 259                                        int iovcnt,
 260                                        void *opaque);
 261
 262 static void qemu_net_client_setup(NetClientState *nc,
 263                                   NetClientInfo *info,
 264                                   NetClientState *peer,
 265                                   const char *model,
 266                                   const char *name,
 267                                   NetClientDestructor *destructor,
 268                                   bool is_datapath)
 269 {
 270     nc->info = info;
 271     nc->model = g_strdup(model);
 272     if (name) {
 273         nc->name = g_strdup(name);
 274     } else {
 275         nc->name = assign_name(nc, model);
 276     }
 277
 278     if (peer) {
 279         assert(!peer->peer);
 280         nc->peer = peer;
 281         peer->peer = nc;
 282     }
 283     QTAILQ_INSERT_TAIL(&net_clients, nc, next);
 284
 285     nc->incoming_queue = qemu_new_net_queue(qemu_deliver_packet_iov, nc);
 286     nc->destructor = destructor;
 287     nc->is_datapath = is_datapath;
 288     QTAILQ_INIT(&nc->filters);
 289 }
 290
 291 NetClientState *qemu_new_net_client(NetClientInfo *info,
 292                                     NetClientState *peer,
 293                                     const char *model,
 294                                     const char *name)
 295 {
 296     NetClientState *nc;
 297
 298     assert(info->size >= sizeof(NetClientState));
 299
 300     nc = g_malloc0(info->size);
 301     qemu_net_client_setup(nc, info, peer, model, name,
 302                           qemu_net_client_destructor, true);
 303
 304     return nc;
 305 }
 306
 307 NetClientState *qemu_new_net_control_client(NetClientInfo *info,
 308                                             NetClientState *peer,
 309                                             const char *model,
 310                                             const char *name)
 311 {
 312     NetClientState *nc;
 313
 314     assert(info->size >= sizeof(NetClientState));
 315
 316     nc = g_malloc0(info->size);
 317     qemu_net_client_setup(nc, info, peer, model, name,
 318                           qemu_net_client_destructor, false);
 319
 320     return nc;
 321 }
 322
 323 NICState *qemu_new_nic(NetClientInfo *info,
 324                        NICConf *conf,
 325                        const char *model,
 326                        const char *name,
 327                        MemReentrancyGuard *reentrancy_guard,
 328                        void *opaque)
 329 {
 330     NetClientState **peers = conf->peers.ncs;
 331     NICState *nic;
 332     int i, queues = MAX(1, conf->peers.queues);
 333
 334     assert(info->type == NET_CLIENT_DRIVER_NIC);
 335     assert(info->size >= sizeof(NICState));
 336
 337     nic = g_malloc0(info->size + sizeof(NetClientState) * queues);
 338     nic->ncs = (void *)nic + info->size;
 339     nic->conf = conf;
 340     nic->reentrancy_guard = reentrancy_guard,
 341     nic->opaque = opaque;
 342
 343     for (i = 0; i < queues; i++) {
 344         qemu_net_client_setup(&nic->ncs[i], info, peers[i], model, name,
 345                               NULL, true);
 346         nic->ncs[i].queue_index = i;
 347     }
 348
 349     return nic;
 350 }
 351
 352 NetClientState *qemu_get_subqueue(NICState *nic, int queue_index)
 353 {
 354     return nic->ncs + queue_index;
 355 }
 356
 357 NetClientState *qemu_get_queue(NICState *nic)
 358 {
 359     return qemu_get_subqueue(nic, 0);
 360 }
 361
 362 NICState *qemu_get_nic(NetClientState *nc)
 363 {
 364     NetClientState *nc0 = nc - nc->queue_index;
 365
 366     return (NICState *)((void *)nc0 - nc->info->size);
 367 }
 368
 369 void *qemu_get_nic_opaque(NetClientState *nc)
 370 {
 371     NICState *nic = qemu_get_nic(nc);
 372
 373     return nic->opaque;
 374 }
 375
 376 NetClientState *qemu_get_peer(NetClientState *nc, int queue_index)
 377 {
 378     assert(nc != NULL);
 379     NetClientState *ncs = nc + queue_index;
 380     return ncs->peer;
 381 }
 382
 383 static void qemu_cleanup_net_client(NetClientState *nc)
 384 {
 385     QTAILQ_REMOVE(&net_clients, nc, next);
 386
 387     if (nc->info->cleanup) {
 388         nc->info->cleanup(nc);
 389     }
 390 }
 391
 392 static void qemu_free_net_client(NetClientState *nc)
 393 {
 394     if (nc->incoming_queue) {
 395         qemu_del_net_queue(nc->incoming_queue);
 396     }
 397     if (nc->peer) {
 398         nc->peer->peer = NULL;
 399     }
 400     g_free(nc->name);
 401     g_free(nc->model);
 402     if (nc->destructor) {
 403         nc->destructor(nc);
 404     }
 405 }
 406
 407 void qemu_del_net_client(NetClientState *nc)
 408 {
 409     NetClientState *ncs[MAX_QUEUE_NUM];
 410     int queues, i;
 411     NetFilterState *nf, *next;
 412
 413     assert(nc->info->type != NET_CLIENT_DRIVER_NIC);
 414
 415     /* If the NetClientState belongs to a multiqueue backend, we will change all
 416      * other NetClientStates also.
 417      */
 418     queues = qemu_find_net_clients_except(nc->name, ncs,
 419                                           NET_CLIENT_DRIVER_NIC,
 420                                           MAX_QUEUE_NUM);
 421     assert(queues != 0);
 422
 423     QTAILQ_FOREACH_SAFE(nf, &nc->filters, next, next) {
 424         object_unparent(OBJECT(nf));
 425     }
 426
 427     /* If there is a peer NIC, delete and cleanup client, but do not free. */
 428     if (nc->peer && nc->peer->info->type == NET_CLIENT_DRIVER_NIC) {
 429         NICState *nic = qemu_get_nic(nc->peer);
 430         if (nic->peer_deleted) {
 431             return;
 432         }
 433         nic->peer_deleted = true;
 434
 435         for (i = 0; i < queues; i++) {
 436             ncs[i]->peer->link_down = true;
 437         }
 438
 439         if (nc->peer->info->link_status_changed) {
 440             nc->peer->info->link_status_changed(nc->peer);
 441         }
 442
 443         for (i = 0; i < queues; i++) {
 444             qemu_cleanup_net_client(ncs[i]);
 445         }
 446
 447         return;
 448     }
 449
 450     for (i = 0; i < queues; i++) {
 451         qemu_cleanup_net_client(ncs[i]);
 452         qemu_free_net_client(ncs[i]);
 453     }
 454 }
 455
 456 void qemu_del_nic(NICState *nic)
 457 {
 458     int i, queues = MAX(nic->conf->peers.queues, 1);
 459
 460     qemu_macaddr_set_free(&nic->conf->macaddr);
 461
 462     for (i = 0; i < queues; i++) {
 463         NetClientState *nc = qemu_get_subqueue(nic, i);
 464         /* If this is a peer NIC and peer has already been deleted, free it now. */
 465         if (nic->peer_deleted) {
 466             qemu_free_net_client(nc->peer);
 467         } else if (nc->peer) {
 468             /* if there are RX packets pending, complete them */
 469             qemu_purge_queued_packets(nc->peer);
 470         }
 471     }
 472
 473     for (i = queues - 1; i >= 0; i--) {
 474         NetClientState *nc = qemu_get_subqueue(nic, i);
 475
 476         qemu_cleanup_net_client(nc);
 477         qemu_free_net_client(nc);
 478     }
 479
 480     g_free(nic);
 481 }
 482
 483 void qemu_foreach_nic(qemu_nic_foreach func, void *opaque)
 484 {
 485     NetClientState *nc;
 486
 487     QTAILQ_FOREACH(nc, &net_clients, next) {
 488         if (nc->info->type == NET_CLIENT_DRIVER_NIC) {
 489             if (nc->queue_index == 0) {
 490                 func(qemu_get_nic(nc), opaque);
 491             }
 492         }
 493     }
 494 }
 495
 496 bool qemu_has_ufo(NetClientState *nc)
 497 {
 498     if (!nc || !nc->info->has_ufo) {
 499         return false;
 500     }
 501
 502     return nc->info->has_ufo(nc);
 503 }
 504
 505 bool qemu_has_uso(NetClientState *nc)
 506 {
 507     if (!nc || !nc->info->has_uso) {
 508         return false;
 509     }
 510
 511     return nc->info->has_uso(nc);
 512 }
 513
 514 bool qemu_has_vnet_hdr(NetClientState *nc)
 515 {
 516     if (!nc || !nc->info->has_vnet_hdr) {
 517         return false;
 518     }
 519
 520     return nc->info->has_vnet_hdr(nc);
 521 }
 522
 523 bool qemu_has_vnet_hdr_len(NetClientState *nc, int len)
 524 {
 525     if (!nc || !nc->info->has_vnet_hdr_len) {
 526         return false;
 527     }
 528
 529     return nc->info->has_vnet_hdr_len(nc, len);
 530 }
 531
 532 void qemu_set_offload(NetClientState *nc, int csum, int tso4, int tso6,
 533                           int ecn, int ufo, int uso4, int uso6)
 534 {
 535     if (!nc || !nc->info->set_offload) {
 536         return;
 537     }
 538
 539     nc->info->set_offload(nc, csum, tso4, tso6, ecn, ufo, uso4, uso6);
 540 }
 541
 542 int qemu_get_vnet_hdr_len(NetClientState *nc)
 543 {
 544     return nc->vnet_hdr_len;
 545 }
 546
 547 void qemu_set_vnet_hdr_len(NetClientState *nc, int len)
 548 {
 549     if (!nc || !nc->info->set_vnet_hdr_len) {
 550         return;
 551     }
 552
 553     nc->vnet_hdr_len = len;
 554     nc->info->set_vnet_hdr_len(nc, len);
 555 }
 556
 557 int qemu_set_vnet_le(NetClientState *nc, bool is_le)
 558 {
 559 #if HOST_BIG_ENDIAN
 560     if (!nc || !nc->info->set_vnet_le) {
 561         return -ENOSYS;
 562     }
 563
 564     return nc->info->set_vnet_le(nc, is_le);
 565 #else
 566     return 0;
 567 #endif
 568 }
 569
 570 int qemu_set_vnet_be(NetClientState *nc, bool is_be)
 571 {
 572 #if HOST_BIG_ENDIAN
 573     return 0;
 574 #else
 575     if (!nc || !nc->info->set_vnet_be) {
 576         return -ENOSYS;
 577     }
 578
 579     return nc->info->set_vnet_be(nc, is_be);
 580 #endif
 581 }
 582
 583 int qemu_can_receive_packet(NetClientState *nc)
 584 {
 585     if (nc->receive_disabled) {
 586         return 0;
 587     } else if (nc->info->can_receive &&
 588                !nc->info->can_receive(nc)) {
 589         return 0;
 590     }
 591     return 1;
 592 }
 593
 594 int qemu_can_send_packet(NetClientState *sender)
 595 {
 596     int vm_running = runstate_is_running();
 597
 598     if (!vm_running) {
 599         return 0;
 600     }
 601
 602     if (!sender->peer) {
 603         return 1;
 604     }
 605
 606     return qemu_can_receive_packet(sender->peer);
 607 }
 608
 609 static ssize_t filter_receive_iov(NetClientState *nc,
 610                                   NetFilterDirection direction,
 611                                   NetClientState *sender,
 612                                   unsigned flags,
 613                                   const struct iovec *iov,
 614                                   int iovcnt,
 615                                   NetPacketSent *sent_cb)
 616 {
 617     ssize_t ret = 0;
 618     NetFilterState *nf = NULL;
 619
 620     if (direction == NET_FILTER_DIRECTION_TX) {
 621         QTAILQ_FOREACH(nf, &nc->filters, next) {
 622             ret = qemu_netfilter_receive(nf, direction, sender, flags, iov,
 623                                          iovcnt, sent_cb);
 624             if (ret) {
 625                 return ret;
 626             }
 627         }
 628     } else {
 629         QTAILQ_FOREACH_REVERSE(nf, &nc->filters, next) {
 630             ret = qemu_netfilter_receive(nf, direction, sender, flags, iov,
 631                                          iovcnt, sent_cb);
 632             if (ret) {
 633                 return ret;
 634             }
 635         }
 636     }
 637
 638     return ret;
 639 }
 640
 641 static ssize_t filter_receive(NetClientState *nc,
 642                               NetFilterDirection direction,
 643                               NetClientState *sender,
 644                               unsigned flags,
 645                               const uint8_t *data,
 646                               size_t size,
 647                               NetPacketSent *sent_cb)
 648 {
 649     struct iovec iov = {
 650         .iov_base = (void *)data,
 651         .iov_len = size
 652     };
 653
 654     return filter_receive_iov(nc, direction, sender, flags, &iov, 1, sent_cb);
 655 }
 656
 657 void qemu_purge_queued_packets(NetClientState *nc)
 658 {
 659     if (!nc->peer) {
 660         return;
 661     }
 662
 663     qemu_net_queue_purge(nc->peer->incoming_queue, nc);
 664 }
 665
 666 void qemu_flush_or_purge_queued_packets(NetClientState *nc, bool purge)
 667 {
 668     nc->receive_disabled = 0;
 669
 670     if (nc->peer && nc->peer->info->type == NET_CLIENT_DRIVER_HUBPORT) {
 671         if (net_hub_flush(nc->peer)) {
 672             qemu_notify_event();
 673         }
 674     }
 675     if (qemu_net_queue_flush(nc->incoming_queue)) {
 676         /* We emptied the queue successfully, signal to the IO thread to repoll
 677          * the file descriptor (for tap, for example).
 678          */
 679         qemu_notify_event();
 680     } else if (purge) {
 681         /* Unable to empty the queue, purge remaining packets */
 682         qemu_net_queue_purge(nc->incoming_queue, nc->peer);
 683     }
 684 }
 685
 686 void qemu_flush_queued_packets(NetClientState *nc)
 687 {
 688     qemu_flush_or_purge_queued_packets(nc, false);
 689 }
 690
 691 static ssize_t qemu_send_packet_async_with_flags(NetClientState *sender,
 692                                                  unsigned flags,
 693                                                  const uint8_t *buf, int size,
 694                                                  NetPacketSent *sent_cb)
 695 {
 696     NetQueue *queue;
 697     int ret;
 698
 699 #ifdef DEBUG_NET
 700     printf("qemu_send_packet_async:\n");
 701     qemu_hexdump(stdout, "net", buf, size);
 702 #endif
 703
 704     if (sender->link_down || !sender->peer) {
 705         return size;
 706     }
 707
 708     /* Let filters handle the packet first */
 709     ret = filter_receive(sender, NET_FILTER_DIRECTION_TX,
 710                          sender, flags, buf, size, sent_cb);
 711     if (ret) {
 712         return ret;
 713     }
 714
 715     ret = filter_receive(sender->peer, NET_FILTER_DIRECTION_RX,
 716                          sender, flags, buf, size, sent_cb);
 717     if (ret) {
 718         return ret;
 719     }
 720
 721     queue = sender->peer->incoming_queue;
 722
 723     return qemu_net_queue_send(queue, sender, flags, buf, size, sent_cb);
 724 }
 725
 726 ssize_t qemu_send_packet_async(NetClientState *sender,
 727                                const uint8_t *buf, int size,
 728                                NetPacketSent *sent_cb)
 729 {
 730     return qemu_send_packet_async_with_flags(sender, QEMU_NET_PACKET_FLAG_NONE,
 731                                              buf, size, sent_cb);
 732 }
 733
 734 ssize_t qemu_send_packet(NetClientState *nc, const uint8_t *buf, int size)
 735 {
 736     return qemu_send_packet_async(nc, buf, size, NULL);
 737 }
 738
 739 ssize_t qemu_receive_packet(NetClientState *nc, const uint8_t *buf, int size)
 740 {
 741     if (!qemu_can_receive_packet(nc)) {
 742         return 0;
 743     }
 744
 745     return qemu_net_queue_receive(nc->incoming_queue, buf, size);
 746 }
 747
 748 ssize_t qemu_receive_packet_iov(NetClientState *nc, const struct iovec *iov,
 749                                 int iovcnt)
 750 {
 751     if (!qemu_can_receive_packet(nc)) {
 752         return 0;
 753     }
 754
 755     return qemu_net_queue_receive_iov(nc->incoming_queue, iov, iovcnt);
 756 }
 757
 758 ssize_t qemu_send_packet_raw(NetClientState *nc, const uint8_t *buf, int size)
 759 {
 760     return qemu_send_packet_async_with_flags(nc, QEMU_NET_PACKET_FLAG_RAW,
 761                                              buf, size, NULL);
 762 }
 763
 764 static ssize_t nc_sendv_compat(NetClientState *nc, const struct iovec *iov,
 765                                int iovcnt, unsigned flags)
 766 {
 767     uint8_t *buf = NULL;
 768     uint8_t *buffer;
 769     size_t offset;
 770     ssize_t ret;
 771
 772     if (iovcnt == 1) {
 773         buffer = iov[0].iov_base;
 774         offset = iov[0].iov_len;
 775     } else {
 776         offset = iov_size(iov, iovcnt);
 777         if (offset > NET_BUFSIZE) {
 778             return -1;
 779         }
 780         buf = g_malloc(offset);
 781         buffer = buf;
 782         offset = iov_to_buf(iov, iovcnt, 0, buf, offset);
 783     }
 784
 785     if (flags & QEMU_NET_PACKET_FLAG_RAW && nc->info->receive_raw) {
 786         ret = nc->info->receive_raw(nc, buffer, offset);
 787     } else {
 788         ret = nc->info->receive(nc, buffer, offset);
 789     }
 790
 791     g_free(buf);
 792     return ret;
 793 }
 794
 795 static ssize_t qemu_deliver_packet_iov(NetClientState *sender,
 796                                        unsigned flags,
 797                                        const struct iovec *iov,
 798                                        int iovcnt,
 799                                        void *opaque)
 800 {
 801     MemReentrancyGuard *owned_reentrancy_guard;
 802     NetClientState *nc = opaque;
 803     int ret;
 804
 805
 806     if (nc->link_down) {
 807         return iov_size(iov, iovcnt);
 808     }
 809
 810     if (nc->receive_disabled) {
 811         return 0;
 812     }
 813
 814     if (nc->info->type != NET_CLIENT_DRIVER_NIC ||
 815         qemu_get_nic(nc)->reentrancy_guard->engaged_in_io) {
 816         owned_reentrancy_guard = NULL;
 817     } else {
 818         owned_reentrancy_guard = qemu_get_nic(nc)->reentrancy_guard;
 819         owned_reentrancy_guard->engaged_in_io = true;
 820     }
 821
 822     if (nc->info->receive_iov && !(flags & QEMU_NET_PACKET_FLAG_RAW)) {
 823         ret = nc->info->receive_iov(nc, iov, iovcnt);
 824     } else {
 825         ret = nc_sendv_compat(nc, iov, iovcnt, flags);
 826     }
 827
 828     if (owned_reentrancy_guard) {
 829         owned_reentrancy_guard->engaged_in_io = false;
 830     }
 831
 832     if (ret == 0) {
 833         nc->receive_disabled = 1;
 834     }
 835
 836     return ret;
 837 }
 838
 839 ssize_t qemu_sendv_packet_async(NetClientState *sender,
 840                                 const struct iovec *iov, int iovcnt,
 841                                 NetPacketSent *sent_cb)
 842 {
 843     NetQueue *queue;
 844     size_t size = iov_size(iov, iovcnt);
 845     int ret;
 846
 847     if (size > NET_BUFSIZE) {
 848         return size;
 849     }
 850
 851     if (sender->link_down || !sender->peer) {
 852         return size;
 853     }
 854
 855     /* Let filters handle the packet first */
 856     ret = filter_receive_iov(sender, NET_FILTER_DIRECTION_TX, sender,
 857                              QEMU_NET_PACKET_FLAG_NONE, iov, iovcnt, sent_cb);
 858     if (ret) {
 859         return ret;
 860     }
 861
 862     ret = filter_receive_iov(sender->peer, NET_FILTER_DIRECTION_RX, sender,
 863                              QEMU_NET_PACKET_FLAG_NONE, iov, iovcnt, sent_cb);
 864     if (ret) {
 865         return ret;
 866     }
 867
 868     queue = sender->peer->incoming_queue;
 869
 870     return qemu_net_queue_send_iov(queue, sender,
 871                                    QEMU_NET_PACKET_FLAG_NONE,
 872                                    iov, iovcnt, sent_cb);
 873 }
 874
 875 ssize_t
 876 qemu_sendv_packet(NetClientState *nc, const struct iovec *iov, int iovcnt)
 877 {
 878     return qemu_sendv_packet_async(nc, iov, iovcnt, NULL);
 879 }
 880
 881 NetClientState *qemu_find_netdev(const char *id)
 882 {
 883     NetClientState *nc;
 884
 885     QTAILQ_FOREACH(nc, &net_clients, next) {
 886         if (nc->info->type == NET_CLIENT_DRIVER_NIC)
 887             continue;
 888         if (!strcmp(nc->name, id)) {
 889             return nc;
 890         }
 891     }
 892
 893     return NULL;
 894 }
 895
 896 int qemu_find_net_clients_except(const char *id, NetClientState **ncs,
 897                                  NetClientDriver type, int max)
 898 {
 899     NetClientState *nc;
 900     int ret = 0;
 901
 902     QTAILQ_FOREACH(nc, &net_clients, next) {
 903         if (nc->info->type == type) {
 904             continue;
 905         }
 906         if (!id || !strcmp(nc->name, id)) {
 907             if (ret < max) {
 908                 ncs[ret] = nc;
 909             }
 910             ret++;
 911         }
 912     }
 913
 914     return ret;
 915 }
 916
 917 static int nic_get_free_idx(void)
 918 {
 919     int index;
 920
 921     for (index = 0; index < MAX_NICS; index++)
 922         if (!nd_table[index].used)
 923             return index;
 924     return -1;
 925 }
 926
 927 GPtrArray *qemu_get_nic_models(const char *device_type)
 928 {
 929     GPtrArray *nic_models = g_ptr_array_new();
 930     GSList *list = object_class_get_list_sorted(device_type, false);
 931
 932     while (list) {
 933         DeviceClass *dc = OBJECT_CLASS_CHECK(DeviceClass, list->data,
 934                                              TYPE_DEVICE);
 935         GSList *next;
 936         if (test_bit(DEVICE_CATEGORY_NETWORK, dc->categories) &&
 937             dc->user_creatable) {
 938             const char *name = object_class_get_name(list->data);
 939             /*
 940              * A network device might also be something else than a NIC, see
 941              * e.g. the "rocker" device. Thus we have to look for the "netdev"
 942              * property, too. Unfortunately, some devices like virtio-net only
 943              * create this property during instance_init, so we have to create
 944              * a temporary instance here to be able to check it.
 945              */
 946             Object *obj = object_new_with_class(OBJECT_CLASS(dc));
 947             if (object_property_find(obj, "netdev")) {
 948                 g_ptr_array_add(nic_models, (gpointer)name);
 949             }
 950             object_unref(obj);
 951         }
 952         next = list->next;
 953         g_slist_free_1(list);
 954         list = next;
 955     }
 956     g_ptr_array_add(nic_models, NULL);
 957
 958     return nic_models;
 959 }
 960
 961 static int net_init_nic(const Netdev *netdev, const char *name,
 962                         NetClientState *peer, Error **errp)
 963 {
 964     int idx;
 965     NICInfo *nd;
 966     const NetLegacyNicOptions *nic;
 967
 968     assert(netdev->type == NET_CLIENT_DRIVER_NIC);
 969     nic = &netdev->u.nic;
 970
 971     idx = nic_get_free_idx();
 972     if (idx == -1 || nb_nics >= MAX_NICS) {
 973         error_setg(errp, "too many NICs");
 974         return -1;
 975     }
 976
 977     nd = &nd_table[idx];
 978
 979     memset(nd, 0, sizeof(*nd));
 980
 981     if (nic->netdev) {
 982         nd->netdev = qemu_find_netdev(nic->netdev);
 983         if (!nd->netdev) {
 984             error_setg(errp, "netdev '%s' not found", nic->netdev);
 985             return -1;
 986         }
 987     } else {
 988         assert(peer);
 989         nd->netdev = peer;
 990     }
 991     nd->name = g_strdup(name);
 992     if (nic->model) {
 993         nd->model = g_strdup(nic->model);
 994     }
 995     if (nic->addr) {
 996         nd->devaddr = g_strdup(nic->addr);
 997     }
 998
 999     if (nic->macaddr &&
1000         net_parse_macaddr(nd->macaddr.a, nic->macaddr) < 0) {
1001         error_setg(errp, "invalid syntax for ethernet address");
1002         return -1;
1003     }
1004     if (nic->macaddr &&
1005         is_multicast_ether_addr(nd->macaddr.a)) {
1006         error_setg(errp,
1007                    "NIC cannot have multicast MAC address (odd 1st byte)");
1008         return -1;
1009     }
1010     qemu_macaddr_default_if_unset(&nd->macaddr);
1011
1012     if (nic->has_vectors) {
1013         if (nic->vectors > 0x7ffffff) {
1014             error_setg(errp, "invalid # of vectors: %"PRIu32, nic->vectors);
1015             return -1;
1016         }
1017         nd->nvectors = nic->vectors;
1018     } else {
1019         nd->nvectors = DEV_NVECTORS_UNSPECIFIED;
1020     }
1021
1022     nd->used = 1;
1023     nb_nics++;
1024
1025     return idx;
1026 }
1027
1028 static gboolean add_nic_result(gpointer key, gpointer value, gpointer user_data)
1029 {
1030     GPtrArray *results = user_data;
1031     GPtrArray *alias_list = value;
1032     const char *model = key;
1033     char *result;
1034
1035     if (!alias_list) {
1036         result = g_strdup(model);
1037     } else {
1038         GString *result_str = g_string_new(model);
1039         int i;
1040
1041         g_string_append(result_str, " (aka ");
1042         for (i = 0; i < alias_list->len; i++) {
1043             if (i) {
1044                 g_string_append(result_str, ", ");
1045             }
1046             g_string_append(result_str, alias_list->pdata[i]);
1047         }
1048         g_string_append(result_str, ")");
1049         result = result_str->str;
1050         g_string_free(result_str, false);
1051         g_ptr_array_unref(alias_list);
1052     }
1053     g_ptr_array_add(results, result);
1054     return true;
1055 }
1056
1057 static int model_cmp(char **a, char **b)
1058 {
1059     return strcmp(*a, *b);
1060 }
1061
1062 static void show_nic_models(void)
1063 {
1064     GPtrArray *results = g_ptr_array_new();
1065     int i;
1066
1067     g_hash_table_foreach_remove(nic_model_help, add_nic_result, results);
1068     g_ptr_array_sort(results, (GCompareFunc)model_cmp);
1069
1070     printf("Available NIC models for this configuration:\n");
1071     for (i = 0 ; i < results->len; i++) {
1072         printf("%s\n", (char *)results->pdata[i]);
1073     }
1074     g_hash_table_unref(nic_model_help);
1075     nic_model_help = NULL;
1076 }
1077
1078 static void add_nic_model_help(const char *model, const char *alias)
1079 {
1080     GPtrArray *alias_list = NULL;
1081
1082     if (g_hash_table_lookup_extended(nic_model_help, model, NULL,
1083                                      (gpointer *)&alias_list)) {
1084         /* Already exists, no alias to add: return */
1085         if (!alias) {
1086             return;
1087         }
1088         if (alias_list) {
1089             /* Check if this alias is already in the list. Add if not. */
1090             if (!g_ptr_array_find_with_equal_func(alias_list, alias,
1091                                                   g_str_equal, NULL)) {
1092                 g_ptr_array_add(alias_list, g_strdup(alias));
1093             }
1094             return;
1095         }
1096     }
1097     /* Either this model wasn't in the list already, or a first alias added */
1098     if (alias) {
1099         alias_list = g_ptr_array_new();
1100         g_ptr_array_set_free_func(alias_list, g_free);
1101         g_ptr_array_add(alias_list, g_strdup(alias));
1102     }
1103     g_hash_table_replace(nic_model_help, g_strdup(model), alias_list);
1104 }
1105
1106 NICInfo *qemu_find_nic_info(const char *typename, bool match_default,
1107                             const char *alias)
1108 {
1109     NICInfo *nd;
1110     int i;
1111
1112     if (nic_model_help) {
1113         add_nic_model_help(typename, alias);
1114     }
1115
1116     for (i = 0; i < nb_nics; i++) {
1117         nd = &nd_table[i];
1118
1119         if (!nd->used || nd->instantiated) {
1120             continue;
1121         }
1122
1123         if ((match_default && !nd->model) || !g_strcmp0(nd->model, typename)
1124             || (alias && !g_strcmp0(nd->model, alias))) {
1125             return nd;
1126         }
1127     }
1128     return NULL;
1129 }
1130
1131
1132 /* "I have created a device. Please configure it if you can" */
1133 bool qemu_configure_nic_device(DeviceState *dev, bool match_default,
1134                                const char *alias)
1135 {
1136     NICInfo *nd = qemu_find_nic_info(object_get_typename(OBJECT(dev)),
1137                                      match_default, alias);
1138
1139     if (nd) {
1140         qdev_set_nic_properties(dev, nd);
1141         return true;
1142     }
1143     return false;
1144 }
1145
1146 /* "Please create a device, if you have a configuration for it" */
1147 DeviceState *qemu_create_nic_device(const char *typename, bool match_default,
1148                                     const char *alias)
1149 {
1150     NICInfo *nd = qemu_find_nic_info(typename, match_default, alias);
1151     DeviceState *dev;
1152
1153     if (!nd) {
1154         return NULL;
1155     }
1156
1157     dev = qdev_new(typename);
1158     qdev_set_nic_properties(dev, nd);
1159     return dev;
1160 }
1161
1162 void qemu_create_nic_bus_devices(BusState *bus, const char *parent_type,
1163                                  const char *default_model,
1164                                  const char *alias, const char *alias_target)
1165 {
1166     GPtrArray *nic_models = qemu_get_nic_models(parent_type);
1167     const char *model;
1168     DeviceState *dev;
1169     NICInfo *nd;
1170     int i;
1171
1172     if (nic_model_help) {
1173         if (alias_target) {
1174             add_nic_model_help(alias_target, alias);
1175         }
1176         for (i = 0; i < nic_models->len - 1; i++) {
1177             add_nic_model_help(nic_models->pdata[i], NULL);
1178         }
1179     }
1180
1181     /* Drop the NULL terminator which would make g_str_equal() unhappy */
1182     nic_models->len--;
1183
1184     for (i = 0; i < nb_nics; i++) {
1185         nd = &nd_table[i];
1186
1187         if (!nd->used || nd->instantiated) {
1188             continue;
1189         }
1190
1191         model = nd->model ? nd->model : default_model;
1192         if (!model) {
1193             continue;
1194         }
1195
1196         /* Each bus type is allowed *one* substitution */
1197         if (g_str_equal(model, alias)) {
1198             model = alias_target;
1199         }
1200
1201         if (!g_ptr_array_find_with_equal_func(nic_models, model,
1202                                               g_str_equal, NULL)) {
1203             /* This NIC does not live on this bus. */
1204             continue;
1205         }
1206
1207         dev = qdev_new(model);
1208         qdev_set_nic_properties(dev, nd);
1209         qdev_realize_and_unref(dev, bus, &error_fatal);
1210     }
1211
1212     g_ptr_array_free(nic_models, true);
1213 }
1214
1215 static int (* const net_client_init_fun[NET_CLIENT_DRIVER__MAX])(
1216     const Netdev *netdev,
1217     const char *name,
1218     NetClientState *peer, Error **errp) = {
1219         [NET_CLIENT_DRIVER_NIC]       = net_init_nic,
1220 #ifdef CONFIG_SLIRP
1221         [NET_CLIENT_DRIVER_USER]      = net_init_slirp,
1222 #endif
1223         [NET_CLIENT_DRIVER_TAP]       = net_init_tap,
1224         [NET_CLIENT_DRIVER_SOCKET]    = net_init_socket,
1225         [NET_CLIENT_DRIVER_STREAM]    = net_init_stream,
1226         [NET_CLIENT_DRIVER_DGRAM]     = net_init_dgram,
1227 #ifdef CONFIG_VDE
1228         [NET_CLIENT_DRIVER_VDE]       = net_init_vde,
1229 #endif
1230 #ifdef CONFIG_NETMAP
1231         [NET_CLIENT_DRIVER_NETMAP]    = net_init_netmap,
1232 #endif
1233 #ifdef CONFIG_AF_XDP
1234         [NET_CLIENT_DRIVER_AF_XDP]    = net_init_af_xdp,
1235 #endif
1236 #ifdef CONFIG_NET_BRIDGE
1237         [NET_CLIENT_DRIVER_BRIDGE]    = net_init_bridge,
1238 #endif
1239         [NET_CLIENT_DRIVER_HUBPORT]   = net_init_hubport,
1240 #ifdef CONFIG_VHOST_NET_USER
1241         [NET_CLIENT_DRIVER_VHOST_USER] = net_init_vhost_user,
1242 #endif
1243 #ifdef CONFIG_VHOST_NET_VDPA
1244         [NET_CLIENT_DRIVER_VHOST_VDPA] = net_init_vhost_vdpa,
1245 #endif
1246 #ifdef CONFIG_L2TPV3
1247         [NET_CLIENT_DRIVER_L2TPV3]    = net_init_l2tpv3,
1248 #endif
1249 #ifdef CONFIG_VMNET
1250         [NET_CLIENT_DRIVER_VMNET_HOST] = net_init_vmnet_host,
1251         [NET_CLIENT_DRIVER_VMNET_SHARED] = net_init_vmnet_shared,
1252         [NET_CLIENT_DRIVER_VMNET_BRIDGED] = net_init_vmnet_bridged,
1253 #endif /* CONFIG_VMNET */
1254 };
1255
1256
1257 static int net_client_init1(const Netdev *netdev, bool is_netdev, Error **errp)
1258 {
1259     NetClientState *peer = NULL;
1260     NetClientState *nc;
1261
1262     if (is_netdev) {
1263         if (netdev->type == NET_CLIENT_DRIVER_NIC ||
1264             !net_client_init_fun[netdev->type]) {
1265             error_setg(errp, "network backend '%s' is not compiled into this binary",
1266                        NetClientDriver_str(netdev->type));
1267             return -1;
1268         }
1269     } else {
1270         if (netdev->type == NET_CLIENT_DRIVER_NONE) {
1271             return 0; /* nothing to do */
1272         }
1273         if (netdev->type == NET_CLIENT_DRIVER_HUBPORT) {
1274             error_setg(errp, "network backend '%s' is only supported with -netdev/-nic",
1275                        NetClientDriver_str(netdev->type));
1276             return -1;
1277         }
1278
1279         if (!net_client_init_fun[netdev->type]) {
1280             error_setg(errp, "network backend '%s' is not compiled into this binary",
1281                        NetClientDriver_str(netdev->type));
1282             return -1;
1283         }
1284
1285         /* Do not add to a hub if it's a nic with a netdev= parameter. */
1286         if (netdev->type != NET_CLIENT_DRIVER_NIC ||
1287             !netdev->u.nic.netdev) {
1288             peer = net_hub_add_port(0, NULL, NULL);
1289         }
1290     }
1291
1292     nc = qemu_find_netdev(netdev->id);
1293     if (nc) {
1294         error_setg(errp, "Duplicate ID '%s'", netdev->id);
1295         return -1;
1296     }
1297
1298     if (net_client_init_fun[netdev->type](netdev, netdev->id, peer, errp) < 0) {
1299         /* FIXME drop when all init functions store an Error */
1300         if (errp && !*errp) {
1301             error_setg(errp, "Device '%s' could not be initialized",
1302                        NetClientDriver_str(netdev->type));
1303         }
1304         return -1;
1305     }
1306
1307     if (is_netdev) {
1308         nc = qemu_find_netdev(netdev->id);
1309         assert(nc);
1310         nc->is_netdev = true;
1311     }
1312
1313     return 0;
1314 }
1315
1316 void show_netdevs(void)
1317 {
1318     int idx;
1319     const char *available_netdevs[] = {
1320         "socket",
1321         "stream",
1322         "dgram",
1323         "hubport",
1324         "tap",
1325 #ifdef CONFIG_SLIRP
1326         "user",
1327 #endif
1328 #ifdef CONFIG_L2TPV3
1329         "l2tpv3",
1330 #endif
1331 #ifdef CONFIG_VDE
1332         "vde",
1333 #endif
1334 #ifdef CONFIG_NET_BRIDGE
1335         "bridge",
1336 #endif
1337 #ifdef CONFIG_NETMAP
1338         "netmap",
1339 #endif
1340 #ifdef CONFIG_AF_XDP
1341         "af-xdp",
1342 #endif
1343 #ifdef CONFIG_POSIX
1344         "vhost-user",
1345 #endif
1346 #ifdef CONFIG_VHOST_VDPA
1347         "vhost-vdpa",
1348 #endif
1349 #ifdef CONFIG_VMNET
1350         "vmnet-host",
1351         "vmnet-shared",
1352         "vmnet-bridged",
1353 #endif
1354     };
1355
1356     qemu_printf("Available netdev backend types:\n");
1357     for (idx = 0; idx < ARRAY_SIZE(available_netdevs); idx++) {
1358         qemu_printf("%s\n", available_netdevs[idx]);
1359     }
1360 }
1361
1362 static int net_client_init(QemuOpts *opts, bool is_netdev, Error **errp)
1363 {
1364     gchar **substrings = NULL;
1365     Netdev *object = NULL;
1366     int ret = -1;
1367     Visitor *v = opts_visitor_new(opts);
1368
1369     /* Parse convenience option format ipv6-net=fec0::0[/64] */
1370     const char *ip6_net = qemu_opt_get(opts, "ipv6-net");
1371
1372     if (ip6_net) {
1373         char *prefix_addr;
1374         unsigned long prefix_len = 64; /* Default 64bit prefix length. */
1375
1376         substrings = g_strsplit(ip6_net, "/", 2);
1377         if (!substrings || !substrings[0]) {
1378             error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "ipv6-net",
1379                        "a valid IPv6 prefix");
1380             goto out;
1381         }
1382
1383         prefix_addr = substrings[0];
1384
1385         /* Handle user-specified prefix length. */
1386         if (substrings[1] &&
1387             qemu_strtoul(substrings[1], NULL, 10, &prefix_len))
1388         {
1389             error_setg(errp,
1390                        "parameter 'ipv6-net' expects a number after '/'");
1391             goto out;
1392         }
1393
1394         qemu_opt_set(opts, "ipv6-prefix", prefix_addr, &error_abort);
1395         qemu_opt_set_number(opts, "ipv6-prefixlen", prefix_len,
1396                             &error_abort);
1397         qemu_opt_unset(opts, "ipv6-net");
1398     }
1399
1400     /* Create an ID for -net if the user did not specify one */
1401     if (!is_netdev && !qemu_opts_id(opts)) {
1402         qemu_opts_set_id(opts, id_generate(ID_NET));
1403     }
1404
1405     if (visit_type_Netdev(v, NULL, &object, errp)) {
1406         ret = net_client_init1(object, is_netdev, errp);
1407     }
1408
1409     qapi_free_Netdev(object);
1410
1411 out:
1412     g_strfreev(substrings);
1413     visit_free(v);
1414     return ret;
1415 }
1416
1417 void netdev_add(QemuOpts *opts, Error **errp)
1418 {
1419     net_client_init(opts, true, errp);
1420 }
1421
1422 void qmp_netdev_add(Netdev *netdev, Error **errp)
1423 {
1424     if (!id_wellformed(netdev->id)) {
1425         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "id", "an identifier");
1426         return;
1427     }
1428
1429     net_client_init1(netdev, true, errp);
1430 }
1431
1432 void qmp_netdev_del(const char *id, Error **errp)
1433 {
1434     NetClientState *nc;
1435     QemuOpts *opts;
1436
1437     nc = qemu_find_netdev(id);
1438     if (!nc) {
1439         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1440                   "Device '%s' not found", id);
1441         return;
1442     }
1443
1444     if (!nc->is_netdev) {
1445         error_setg(errp, "Device '%s' is not a netdev", id);
1446         return;
1447     }
1448
1449     qemu_del_net_client(nc);
1450
1451     /*
1452      * Wart: we need to delete the QemuOpts associated with netdevs
1453      * created via CLI or HMP, to avoid bogus "Duplicate ID" errors in
1454      * HMP netdev_add.
1455      */
1456     opts = qemu_opts_find(qemu_find_opts("netdev"), id);
1457     if (opts) {
1458         qemu_opts_del(opts);
1459     }
1460 }
1461
1462 static void netfilter_print_info(Monitor *mon, NetFilterState *nf)
1463 {
1464     char *str;
1465     ObjectProperty *prop;
1466     ObjectPropertyIterator iter;
1467     Visitor *v;
1468
1469     /* generate info str */
1470     object_property_iter_init(&iter, OBJECT(nf));
1471     while ((prop = object_property_iter_next(&iter))) {
1472         if (!strcmp(prop->name, "type")) {
1473             continue;
1474         }
1475         v = string_output_visitor_new(false, &str);
1476         object_property_get(OBJECT(nf), prop->name, v, NULL);
1477         visit_complete(v, &str);
1478         visit_free(v);
1479         monitor_printf(mon, ",%s=%s", prop->name, str);
1480         g_free(str);
1481     }
1482     monitor_printf(mon, "\n");
1483 }
1484
1485 void print_net_client(Monitor *mon, NetClientState *nc)
1486 {
1487     NetFilterState *nf;
1488
1489     monitor_printf(mon, "%s: index=%d,type=%s,%s\n", nc->name,
1490                    nc->queue_index,
1491                    NetClientDriver_str(nc->info->type),
1492                    nc->info_str);
1493     if (!QTAILQ_EMPTY(&nc->filters)) {
1494         monitor_printf(mon, "filters:\n");
1495     }
1496     QTAILQ_FOREACH(nf, &nc->filters, next) {
1497         monitor_printf(mon, "  - %s: type=%s",
1498                        object_get_canonical_path_component(OBJECT(nf)),
1499                        object_get_typename(OBJECT(nf)));
1500         netfilter_print_info(mon, nf);
1501     }
1502 }
1503
1504 RxFilterInfoList *qmp_query_rx_filter(const char *name, Error **errp)
1505 {
1506     NetClientState *nc;
1507     RxFilterInfoList *filter_list = NULL, **tail = &filter_list;
1508
1509     QTAILQ_FOREACH(nc, &net_clients, next) {
1510         RxFilterInfo *info;
1511
1512         if (name && strcmp(nc->name, name) != 0) {
1513             continue;
1514         }
1515
1516         /* only query rx-filter information of NIC */
1517         if (nc->info->type != NET_CLIENT_DRIVER_NIC) {
1518             if (name) {
1519                 error_setg(errp, "net client(%s) isn't a NIC", name);
1520                 assert(!filter_list);
1521                 return NULL;
1522             }
1523             continue;
1524         }
1525
1526         /* only query information on queue 0 since the info is per nic,
1527          * not per queue
1528          */
1529         if (nc->queue_index != 0)
1530             continue;
1531
1532         if (nc->info->query_rx_filter) {
1533             info = nc->info->query_rx_filter(nc);
1534             QAPI_LIST_APPEND(tail, info);
1535         } else if (name) {
1536             error_setg(errp, "net client(%s) doesn't support"
1537                        " rx-filter querying", name);
1538             assert(!filter_list);
1539             return NULL;
1540         }
1541
1542         if (name) {
1543             break;
1544         }
1545     }
1546
1547     if (filter_list == NULL && name) {
1548         error_setg(errp, "invalid net client name: %s", name);
1549     }
1550
1551     return filter_list;
1552 }
1553
1554 void colo_notify_filters_event(int event, Error **errp)
1555 {
1556     NetClientState *nc;
1557     NetFilterState *nf;
1558     NetFilterClass *nfc = NULL;
1559     Error *local_err = NULL;
1560
1561     QTAILQ_FOREACH(nc, &net_clients, next) {
1562         QTAILQ_FOREACH(nf, &nc->filters, next) {
1563             nfc = NETFILTER_GET_CLASS(OBJECT(nf));
1564             nfc->handle_event(nf, event, &local_err);
1565             if (local_err) {
1566                 error_propagate(errp, local_err);
1567                 return;
1568             }
1569         }
1570     }
1571 }
1572
1573 void qmp_set_link(const char *name, bool up, Error **errp)
1574 {
1575     NetClientState *ncs[MAX_QUEUE_NUM];
1576     NetClientState *nc;
1577     int queues, i;
1578
1579     queues = qemu_find_net_clients_except(name, ncs,
1580                                           NET_CLIENT_DRIVER__MAX,
1581                                           MAX_QUEUE_NUM);
1582
1583     if (queues == 0) {
1584         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1585                   "Device '%s' not found", name);
1586         return;
1587     }
1588     nc = ncs[0];
1589
1590     for (i = 0; i < queues; i++) {
1591         ncs[i]->link_down = !up;
1592     }
1593
1594     if (nc->info->link_status_changed) {
1595         nc->info->link_status_changed(nc);
1596     }
1597
1598     if (nc->peer) {
1599         /* Change peer link only if the peer is NIC and then notify peer.
1600          * If the peer is a HUBPORT or a backend, we do not change the
1601          * link status.
1602          *
1603          * This behavior is compatible with qemu hubs where there could be
1604          * multiple clients that can still communicate with each other in
1605          * disconnected mode. For now maintain this compatibility.
1606          */
1607         if (nc->peer->info->type == NET_CLIENT_DRIVER_NIC) {
1608             for (i = 0; i < queues; i++) {
1609                 ncs[i]->peer->link_down = !up;
1610             }
1611         }
1612         if (nc->peer->info->link_status_changed) {
1613             nc->peer->info->link_status_changed(nc->peer);
1614         }
1615     }
1616 }
1617
1618 static void net_vm_change_state_handler(void *opaque, bool running,
1619                                         RunState state)
1620 {
1621     NetClientState *nc;
1622     NetClientState *tmp;
1623
1624     QTAILQ_FOREACH_SAFE(nc, &net_clients, next, tmp) {
1625         if (running) {
1626             /* Flush queued packets and wake up backends. */
1627             if (nc->peer && qemu_can_send_packet(nc)) {
1628                 qemu_flush_queued_packets(nc->peer);
1629             }
1630         } else {
1631             /* Complete all queued packets, to guarantee we don't modify
1632              * state later when VM is not running.
1633              */
1634             qemu_flush_or_purge_queued_packets(nc, true);
1635         }
1636     }
1637 }
1638
1639 void net_cleanup(void)
1640 {
1641     NetClientState *nc, **p = &QTAILQ_FIRST(&net_clients);
1642
1643     /*cleanup colo compare module for COLO*/
1644     colo_compare_cleanup();
1645
1646     /*
1647      * Walk the net_clients list and remove the netdevs but *not* any
1648      * NET_CLIENT_DRIVER_NIC entries. The latter are owned by the device
1649      * model which created them, and in some cases (e.g. xen-net-device)
1650      * the device itself may do cleanup at exit and will be upset if we
1651      * just delete its NIC from underneath it.
1652      *
1653      * Since qemu_del_net_client() may delete multiple entries, using
1654      * QTAILQ_FOREACH_SAFE() is not safe here. The only safe pointer
1655      * to keep as a bookmark is a NET_CLIENT_DRIVER_NIC entry, so keep
1656      * 'p' pointing to either the head of the list, or the 'next' field
1657      * of the latest NET_CLIENT_DRIVER_NIC, and operate on *p as we walk
1658      * the list.
1659      *
1660      * The 'nc' variable isn't part of the list traversal; it's purely
1661      * for convenience as too much '(*p)->' has a tendency to make the
1662      * readers' eyes bleed.
1663      */
1664     while (*p) {
1665         nc = *p;
1666         if (nc->info->type == NET_CLIENT_DRIVER_NIC) {
1667             /* Skip NET_CLIENT_DRIVER_NIC entries */
1668             p = &QTAILQ_NEXT(nc, next);
1669         } else {
1670             qemu_del_net_client(nc);
1671         }
1672     }
1673
1674     qemu_del_vm_change_state_handler(net_change_state_entry);
1675 }
1676
1677 void net_check_clients(void)
1678 {
1679     NetClientState *nc;
1680     int i;
1681
1682     if (nic_model_help) {
1683         show_nic_models();
1684         exit(0);
1685     }
1686     net_hub_check_clients();
1687
1688     QTAILQ_FOREACH(nc, &net_clients, next) {
1689         if (!nc->peer) {
1690             warn_report("%s %s has no peer",
1691                         nc->info->type == NET_CLIENT_DRIVER_NIC
1692                         ? "nic" : "netdev",
1693                         nc->name);
1694         }
1695     }
1696
1697     /* Check that all NICs requested via -net nic actually got created.
1698      * NICs created via -device don't need to be checked here because
1699      * they are always instantiated.
1700      */
1701     for (i = 0; i < MAX_NICS; i++) {
1702         NICInfo *nd = &nd_table[i];
1703         if (nd->used && !nd->instantiated) {
1704             warn_report("requested NIC (%s, model %s) "
1705                         "was not created (not supported by this machine?)",
1706                         nd->name ? nd->name : "anonymous",
1707                         nd->model ? nd->model : "unspecified");
1708         }
1709     }
1710 }
1711
1712 static int net_init_client(void *dummy, QemuOpts *opts, Error **errp)
1713 {
1714     return net_client_init(opts, false, errp);
1715 }
1716
1717 static int net_init_netdev(void *dummy, QemuOpts *opts, Error **errp)
1718 {
1719     const char *type = qemu_opt_get(opts, "type");
1720
1721     if (type && is_help_option(type)) {
1722         show_netdevs();
1723         exit(0);
1724     }
1725     return net_client_init(opts, true, errp);
1726 }
1727
1728 /* For the convenience "--nic" parameter */
1729 static int net_param_nic(void *dummy, QemuOpts *opts, Error **errp)
1730 {
1731     char *mac, *nd_id;
1732     int idx, ret;
1733     NICInfo *ni;
1734     const char *type;
1735
1736     type = qemu_opt_get(opts, "type");
1737     if (type) {
1738         if (g_str_equal(type, "none")) {
1739             return 0;    /* Nothing to do, default_net is cleared in vl.c */
1740         }
1741         if (is_help_option(type)) {
1742             GPtrArray *nic_models = qemu_get_nic_models(TYPE_DEVICE);
1743             int i;
1744             show_netdevs();
1745             printf("\n");
1746             printf("Available NIC models "
1747                    "(use -nic model=help for a filtered list):\n");
1748             for (i = 0 ; nic_models->pdata[i]; i++) {
1749                 printf("%s\n", (char *)nic_models->pdata[i]);
1750             }
1751             g_ptr_array_free(nic_models, true);
1752             exit(0);
1753         }
1754     }
1755
1756     idx = nic_get_free_idx();
1757     if (idx == -1 || nb_nics >= MAX_NICS) {
1758         error_setg(errp, "no more on-board/default NIC slots available");
1759         return -1;
1760     }
1761
1762     if (!type) {
1763         qemu_opt_set(opts, "type", "user", &error_abort);
1764     }
1765
1766     ni = &nd_table[idx];
1767     memset(ni, 0, sizeof(*ni));
1768     ni->model = qemu_opt_get_del(opts, "model");
1769
1770     if (!nic_model_help && !g_strcmp0(ni->model, "help")) {
1771         nic_model_help = g_hash_table_new_full(g_str_hash, g_str_equal,
1772                                                g_free, NULL);
1773         return 0;
1774     }
1775
1776     /* Create an ID if the user did not specify one */
1777     nd_id = g_strdup(qemu_opts_id(opts));
1778     if (!nd_id) {
1779         nd_id = id_generate(ID_NET);
1780         qemu_opts_set_id(opts, nd_id);
1781     }
1782
1783     /* Handle MAC address */
1784     mac = qemu_opt_get_del(opts, "mac");
1785     if (mac) {
1786         ret = net_parse_macaddr(ni->macaddr.a, mac);
1787         g_free(mac);
1788         if (ret) {
1789             error_setg(errp, "invalid syntax for ethernet address");
1790             goto out;
1791         }
1792         if (is_multicast_ether_addr(ni->macaddr.a)) {
1793             error_setg(errp, "NIC cannot have multicast MAC address");
1794             ret = -1;
1795             goto out;
1796         }
1797     }
1798     qemu_macaddr_default_if_unset(&ni->macaddr);
1799
1800     ret = net_client_init(opts, true, errp);
1801     if (ret == 0) {
1802         ni->netdev = qemu_find_netdev(nd_id);
1803         ni->used = true;
1804         nb_nics++;
1805     }
1806
1807 out:
1808     g_free(nd_id);
1809     return ret;
1810 }
1811
1812 static void netdev_init_modern(void)
1813 {
1814     while (!QSIMPLEQ_EMPTY(&nd_queue)) {
1815         NetdevQueueEntry *nd = QSIMPLEQ_FIRST(&nd_queue);
1816
1817         QSIMPLEQ_REMOVE_HEAD(&nd_queue, entry);
1818         loc_push_restore(&nd->loc);
1819         net_client_init1(nd->nd, true, &error_fatal);
1820         loc_pop(&nd->loc);
1821         qapi_free_Netdev(nd->nd);
1822         g_free(nd);
1823     }
1824 }
1825
1826 void net_init_clients(void)
1827 {
1828     net_change_state_entry =
1829         qemu_add_vm_change_state_handler(net_vm_change_state_handler, NULL);
1830
1831     QTAILQ_INIT(&net_clients);
1832
1833     netdev_init_modern();
1834
1835     qemu_opts_foreach(qemu_find_opts("netdev"), net_init_netdev, NULL,
1836                       &error_fatal);
1837
1838     qemu_opts_foreach(qemu_find_opts("nic"), net_param_nic, NULL,
1839                       &error_fatal);
1840
1841     qemu_opts_foreach(qemu_find_opts("net"), net_init_client, NULL,
1842                       &error_fatal);
1843 }
1844
1845 /*
1846  * Does this -netdev argument use modern rather than traditional syntax?
1847  * Modern syntax is to be parsed with netdev_parse_modern().
1848  * Traditional syntax is to be parsed with net_client_parse().
1849  */
1850 bool netdev_is_modern(const char *optstr)
1851 {
1852     QemuOpts *opts;
1853     bool is_modern;
1854     const char *type;
1855     static QemuOptsList dummy_opts = {
1856         .name = "netdev",
1857         .implied_opt_name = "type",
1858         .head = QTAILQ_HEAD_INITIALIZER(dummy_opts.head),
1859         .desc = { { } },
1860     };
1861
1862     if (optstr[0] == '{') {
1863         /* This is JSON, which means it's modern syntax */
1864         return true;
1865     }
1866
1867     opts = qemu_opts_create(&dummy_opts, NULL, false, &error_abort);
1868     qemu_opts_do_parse(opts, optstr, dummy_opts.implied_opt_name,
1869                        &error_abort);
1870     type = qemu_opt_get(opts, "type");
1871     is_modern = !g_strcmp0(type, "stream") || !g_strcmp0(type, "dgram");
1872
1873     qemu_opts_reset(&dummy_opts);
1874
1875     return is_modern;
1876 }
1877
1878 /*
1879  * netdev_parse_modern() uses modern, more expressive syntax than
1880  * net_client_parse(), but supports only the -netdev option.
1881  * netdev_parse_modern() appends to @nd_queue, whereas net_client_parse()
1882  * appends to @qemu_netdev_opts.
1883  */
1884 void netdev_parse_modern(const char *optstr)
1885 {
1886     Visitor *v;
1887     NetdevQueueEntry *nd;
1888
1889     v = qobject_input_visitor_new_str(optstr, "type", &error_fatal);
1890     nd = g_new(NetdevQueueEntry, 1);
1891     visit_type_Netdev(v, NULL, &nd->nd, &error_fatal);
1892     visit_free(v);
1893     loc_save(&nd->loc);
1894
1895     QSIMPLEQ_INSERT_TAIL(&nd_queue, nd, entry);
1896 }
1897
1898 void net_client_parse(QemuOptsList *opts_list, const char *optstr)
1899 {
1900     if (!qemu_opts_parse_noisily(opts_list, optstr, true)) {
1901         exit(1);
1902     }
1903 }
1904
1905 /* From FreeBSD */
1906 /* XXX: optimize */
1907 uint32_t net_crc32(const uint8_t *p, int len)
1908 {
1909     uint32_t crc;
1910     int carry, i, j;
1911     uint8_t b;
1912
1913     crc = 0xffffffff;
1914     for (i = 0; i < len; i++) {
1915         b = *p++;
1916         for (j = 0; j < 8; j++) {
1917             carry = ((crc & 0x80000000L) ? 1 : 0) ^ (b & 0x01);
1918             crc <<= 1;
1919             b >>= 1;
1920             if (carry) {
1921                 crc = ((crc ^ POLYNOMIAL_BE) | carry);
1922             }
1923         }
1924     }
1925
1926     return crc;
1927 }
1928
1929 uint32_t net_crc32_le(const uint8_t *p, int len)
1930 {
1931     uint32_t crc;
1932     int carry, i, j;
1933     uint8_t b;
1934
1935     crc = 0xffffffff;
1936     for (i = 0; i < len; i++) {
1937         b = *p++;
1938         for (j = 0; j < 8; j++) {
1939             carry = (crc & 0x1) ^ (b & 0x01);
1940             crc >>= 1;
1941             b >>= 1;
1942             if (carry) {
1943                 crc ^= POLYNOMIAL_LE;
1944             }
1945         }
1946     }
1947
1948     return crc;
1949 }
1950
1951 QemuOptsList qemu_netdev_opts = {
1952     .name = "netdev",
1953     .implied_opt_name = "type",
1954     .head = QTAILQ_HEAD_INITIALIZER(qemu_netdev_opts.head),
1955     .desc = {
1956         /*
1957          * no elements => accept any params
1958          * validation will happen later
1959          */
1960         { /* end of list */ }
1961     },
1962 };
1963
1964 QemuOptsList qemu_nic_opts = {
1965     .name = "nic",
1966     .implied_opt_name = "type",
1967     .head = QTAILQ_HEAD_INITIALIZER(qemu_nic_opts.head),
1968     .desc = {
1969         /*
1970          * no elements => accept any params
1971          * validation will happen later
1972          */
1973         { /* end of list */ }
1974     },
1975 };
1976
1977 QemuOptsList qemu_net_opts = {
1978     .name = "net",
1979     .implied_opt_name = "type",
1980     .head = QTAILQ_HEAD_INITIALIZER(qemu_net_opts.head),
1981     .desc = {
1982         /*
1983          * no elements => accept any params
1984          * validation will happen later
1985          */
1986         { /* end of list */ }
1987     },
1988 };
1989
1990 void net_socket_rs_init(SocketReadState *rs,
1991                         SocketReadStateFinalize *finalize,
1992                         bool vnet_hdr)
1993 {
1994     rs->state = 0;
1995     rs->vnet_hdr = vnet_hdr;
1996     rs->index = 0;
1997     rs->packet_len = 0;
1998     rs->vnet_hdr_len = 0;
1999     memset(rs->buf, 0, sizeof(rs->buf));
2000     rs->finalize = finalize;
2001 }
2002
2003 /*
2004  * Returns
2005  * 0: success
2006  * -1: error occurs
2007  */
2008 int net_fill_rstate(SocketReadState *rs, const uint8_t *buf, int size)
2009 {
2010     unsigned int l;
2011
2012     while (size > 0) {
2013         /* Reassemble a packet from the network.
2014          * 0 = getting length.
2015          * 1 = getting vnet header length.
2016          * 2 = getting data.
2017          */
2018         switch (rs->state) {
2019         case 0:
2020             l = 4 - rs->index;
2021             if (l > size) {
2022                 l = size;
2023             }
2024             memcpy(rs->buf + rs->index, buf, l);
2025             buf += l;
2026             size -= l;
2027             rs->index += l;
2028             if (rs->index == 4) {
2029                 /* got length */
2030                 rs->packet_len = ntohl(*(uint32_t *)rs->buf);
2031                 rs->index = 0;
2032                 if (rs->vnet_hdr) {
2033                     rs->state = 1;
2034                 } else {
2035                     rs->state = 2;
2036                     rs->vnet_hdr_len = 0;
2037                 }
2038             }
2039             break;
2040         case 1:
2041             l = 4 - rs->index;
2042             if (l > size) {
2043                 l = size;
2044             }
2045             memcpy(rs->buf + rs->index, buf, l);
2046             buf += l;
2047             size -= l;
2048             rs->index += l;
2049             if (rs->index == 4) {
2050                 /* got vnet header length */
2051                 rs->vnet_hdr_len = ntohl(*(uint32_t *)rs->buf);
2052                 rs->index = 0;
2053                 rs->state = 2;
2054             }
2055             break;
2056         case 2:
2057             l = rs->packet_len - rs->index;
2058             if (l > size) {
2059                 l = size;
2060             }
2061             if (rs->index + l <= sizeof(rs->buf)) {
2062                 memcpy(rs->buf + rs->index, buf, l);
2063             } else {
2064                 fprintf(stderr, "serious error: oversized packet received,"
2065                     "connection terminated.\n");
2066                 rs->index = rs->state = 0;
2067                 return -1;
2068             }
2069
2070             rs->index += l;
2071             buf += l;
2072             size -= l;
2073             if (rs->index >= rs->packet_len) {
2074                 rs->index = 0;
2075                 rs->state = 0;
2076                 assert(rs->finalize);
2077                 rs->finalize(rs);
2078             }
2079             break;
2080         }
2081     }
2082
2083     assert(size == 0);
2084     return 0;
2085 }