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 #include "standard-headers/linux/virtio_net.h"
  60
  61 /* Net bridge is currently not supported for W32. */
  62 #if !defined(_WIN32)
  63 # define CONFIG_NET_BRIDGE
  64 #endif
  65
  66 static VMChangeStateEntry *net_change_state_entry;
  67 NetClientStateList net_clients;
  68
  69 typedef struct NetdevQueueEntry {
  70     Netdev *nd;
  71     Location loc;
  72     QSIMPLEQ_ENTRY(NetdevQueueEntry) entry;
  73 } NetdevQueueEntry;
  74
  75 typedef QSIMPLEQ_HEAD(, NetdevQueueEntry) NetdevQueue;
  76
  77 static NetdevQueue nd_queue = QSIMPLEQ_HEAD_INITIALIZER(nd_queue);
  78
  79 static GHashTable *nic_model_help;
  80
  81 static int nb_nics;
  82 static NICInfo nd_table[MAX_NICS];
  83
  84 /***********************************************************/
  85 /* network device redirectors */
  86
  87 int convert_host_port(struct sockaddr_in *saddr, const char *host,
  88                       const char *port, Error **errp)
  89 {
  90     struct hostent *he;
  91     const char *r;
  92     long p;
  93
  94     memset(saddr, 0, sizeof(*saddr));
  95
  96     saddr->sin_family = AF_INET;
  97     if (host[0] == '\0') {
  98         saddr->sin_addr.s_addr = 0;
  99     } else {
 100         if (qemu_isdigit(host[0])) {
 101             if (!inet_aton(host, &saddr->sin_addr)) {
 102                 error_setg(errp, "host address '%s' is not a valid "
 103                            "IPv4 address", host);
 104                 return -1;
 105             }
 106         } else {
 107             he = gethostbyname(host);
 108             if (he == NULL) {
 109                 error_setg(errp, "can't resolve host address '%s'", host);
 110                 return -1;
 111             }
 112             saddr->sin_addr = *(struct in_addr *)he->h_addr;
 113         }
 114     }
 115     if (qemu_strtol(port, &r, 0, &p) != 0) {
 116         error_setg(errp, "port number '%s' is invalid", port);
 117         return -1;
 118     }
 119     saddr->sin_port = htons(p);
 120     return 0;
 121 }
 122
 123 int parse_host_port(struct sockaddr_in *saddr, const char *str,
 124                     Error **errp)
 125 {
 126     gchar **substrings;
 127     int ret;
 128
 129     substrings = g_strsplit(str, ":", 2);
 130     if (!substrings || !substrings[0] || !substrings[1]) {
 131         error_setg(errp, "host address '%s' doesn't contain ':' "
 132                    "separating host from port", str);
 133         ret = -1;
 134         goto out;
 135     }
 136
 137     ret = convert_host_port(saddr, substrings[0], substrings[1], errp);
 138
 139 out:
 140     g_strfreev(substrings);
 141     return ret;
 142 }
 143
 144 char *qemu_mac_strdup_printf(const uint8_t *macaddr)
 145 {
 146     return g_strdup_printf("%.2x:%.2x:%.2x:%.2x:%.2x:%.2x",
 147                            macaddr[0], macaddr[1], macaddr[2],
 148                            macaddr[3], macaddr[4], macaddr[5]);
 149 }
 150
 151 void qemu_set_info_str(NetClientState *nc, const char *fmt, ...)
 152 {
 153     va_list ap;
 154
 155     va_start(ap, fmt);
 156     vsnprintf(nc->info_str, sizeof(nc->info_str), fmt, ap);
 157     va_end(ap);
 158 }
 159
 160 void qemu_format_nic_info_str(NetClientState *nc, uint8_t macaddr[6])
 161 {
 162     qemu_set_info_str(nc, "model=%s,macaddr=%02x:%02x:%02x:%02x:%02x:%02x",
 163                       nc->model, macaddr[0], macaddr[1], macaddr[2],
 164                       macaddr[3], macaddr[4], macaddr[5]);
 165 }
 166
 167 static int mac_table[256] = {0};
 168
 169 static void qemu_macaddr_set_used(MACAddr *macaddr)
 170 {
 171     int index;
 172
 173     for (index = 0x56; index < 0xFF; index++) {
 174         if (macaddr->a[5] == index) {
 175             mac_table[index]++;
 176         }
 177     }
 178 }
 179
 180 static void qemu_macaddr_set_free(MACAddr *macaddr)
 181 {
 182     int index;
 183     static const MACAddr base = { .a = { 0x52, 0x54, 0x00, 0x12, 0x34, 0 } };
 184
 185     if (memcmp(macaddr->a, &base.a, (sizeof(base.a) - 1)) != 0) {
 186         return;
 187     }
 188     for (index = 0x56; index < 0xFF; index++) {
 189         if (macaddr->a[5] == index) {
 190             mac_table[index]--;
 191         }
 192     }
 193 }
 194
 195 static int qemu_macaddr_get_free(void)
 196 {
 197     int index;
 198
 199     for (index = 0x56; index < 0xFF; index++) {
 200         if (mac_table[index] == 0) {
 201             return index;
 202         }
 203     }
 204
 205     return -1;
 206 }
 207
 208 void qemu_macaddr_default_if_unset(MACAddr *macaddr)
 209 {
 210     static const MACAddr zero = { .a = { 0,0,0,0,0,0 } };
 211     static const MACAddr base = { .a = { 0x52, 0x54, 0x00, 0x12, 0x34, 0 } };
 212
 213     if (memcmp(macaddr, &zero, sizeof(zero)) != 0) {
 214         if (memcmp(macaddr->a, &base.a, (sizeof(base.a) - 1)) != 0) {
 215             return;
 216         } else {
 217             qemu_macaddr_set_used(macaddr);
 218             return;
 219         }
 220     }
 221
 222     macaddr->a[0] = 0x52;
 223     macaddr->a[1] = 0x54;
 224     macaddr->a[2] = 0x00;
 225     macaddr->a[3] = 0x12;
 226     macaddr->a[4] = 0x34;
 227     macaddr->a[5] = qemu_macaddr_get_free();
 228     qemu_macaddr_set_used(macaddr);
 229 }
 230
 231 /**
 232  * Generate a name for net client
 233  *
 234  * Only net clients created with the legacy -net option and NICs need this.
 235  */
 236 static char *assign_name(NetClientState *nc1, const char *model)
 237 {
 238     NetClientState *nc;
 239     int id = 0;
 240
 241     QTAILQ_FOREACH(nc, &net_clients, next) {
 242         if (nc == nc1) {
 243             continue;
 244         }
 245         if (strcmp(nc->model, model) == 0) {
 246             id++;
 247         }
 248     }
 249
 250     return g_strdup_printf("%s.%d", model, id);
 251 }
 252
 253 static void qemu_net_client_destructor(NetClientState *nc)
 254 {
 255     g_free(nc);
 256 }
 257 static ssize_t qemu_deliver_packet_iov(NetClientState *sender,
 258                                        unsigned flags,
 259                                        const struct iovec *iov,
 260                                        int iovcnt,
 261                                        void *opaque);
 262
 263 static void qemu_net_client_setup(NetClientState *nc,
 264                                   NetClientInfo *info,
 265                                   NetClientState *peer,
 266                                   const char *model,
 267                                   const char *name,
 268                                   NetClientDestructor *destructor,
 269                                   bool is_datapath)
 270 {
 271     nc->info = info;
 272     nc->model = g_strdup(model);
 273     if (name) {
 274         nc->name = g_strdup(name);
 275     } else {
 276         nc->name = assign_name(nc, model);
 277     }
 278
 279     if (peer) {
 280         assert(!peer->peer);
 281         nc->peer = peer;
 282         peer->peer = nc;
 283     }
 284     QTAILQ_INSERT_TAIL(&net_clients, nc, next);
 285
 286     nc->incoming_queue = qemu_new_net_queue(qemu_deliver_packet_iov, nc);
 287     nc->destructor = destructor;
 288     nc->is_datapath = is_datapath;
 289     QTAILQ_INIT(&nc->filters);
 290 }
 291
 292 NetClientState *qemu_new_net_client(NetClientInfo *info,
 293                                     NetClientState *peer,
 294                                     const char *model,
 295                                     const char *name)
 296 {
 297     NetClientState *nc;
 298
 299     assert(info->size >= sizeof(NetClientState));
 300
 301     nc = g_malloc0(info->size);
 302     qemu_net_client_setup(nc, info, peer, model, name,
 303                           qemu_net_client_destructor, true);
 304
 305     return nc;
 306 }
 307
 308 NetClientState *qemu_new_net_control_client(NetClientInfo *info,
 309                                             NetClientState *peer,
 310                                             const char *model,
 311                                             const char *name)
 312 {
 313     NetClientState *nc;
 314
 315     assert(info->size >= sizeof(NetClientState));
 316
 317     nc = g_malloc0(info->size);
 318     qemu_net_client_setup(nc, info, peer, model, name,
 319                           qemu_net_client_destructor, false);
 320
 321     return nc;
 322 }
 323
 324 NICState *qemu_new_nic(NetClientInfo *info,
 325                        NICConf *conf,
 326                        const char *model,
 327                        const char *name,
 328                        MemReentrancyGuard *reentrancy_guard,
 329                        void *opaque)
 330 {
 331     NetClientState **peers = conf->peers.ncs;
 332     NICState *nic;
 333     int i, queues = MAX(1, conf->peers.queues);
 334
 335     assert(info->type == NET_CLIENT_DRIVER_NIC);
 336     assert(info->size >= sizeof(NICState));
 337
 338     nic = g_malloc0(info->size + sizeof(NetClientState) * queues);
 339     nic->ncs = (void *)nic + info->size;
 340     nic->conf = conf;
 341     nic->reentrancy_guard = reentrancy_guard,
 342     nic->opaque = opaque;
 343
 344     for (i = 0; i < queues; i++) {
 345         qemu_net_client_setup(&nic->ncs[i], info, peers[i], model, name,
 346                               NULL, true);
 347         nic->ncs[i].queue_index = i;
 348     }
 349
 350     return nic;
 351 }
 352
 353 NetClientState *qemu_get_subqueue(NICState *nic, int queue_index)
 354 {
 355     return nic->ncs + queue_index;
 356 }
 357
 358 NetClientState *qemu_get_queue(NICState *nic)
 359 {
 360     return qemu_get_subqueue(nic, 0);
 361 }
 362
 363 NICState *qemu_get_nic(NetClientState *nc)
 364 {
 365     NetClientState *nc0 = nc - nc->queue_index;
 366
 367     return (NICState *)((void *)nc0 - nc->info->size);
 368 }
 369
 370 void *qemu_get_nic_opaque(NetClientState *nc)
 371 {
 372     NICState *nic = qemu_get_nic(nc);
 373
 374     return nic->opaque;
 375 }
 376
 377 NetClientState *qemu_get_peer(NetClientState *nc, int queue_index)
 378 {
 379     assert(nc != NULL);
 380     NetClientState *ncs = nc + queue_index;
 381     return ncs->peer;
 382 }
 383
 384 static void qemu_cleanup_net_client(NetClientState *nc)
 385 {
 386     QTAILQ_REMOVE(&net_clients, nc, next);
 387
 388     if (nc->info->cleanup) {
 389         nc->info->cleanup(nc);
 390     }
 391 }
 392
 393 static void qemu_free_net_client(NetClientState *nc)
 394 {
 395     if (nc->incoming_queue) {
 396         qemu_del_net_queue(nc->incoming_queue);
 397     }
 398     if (nc->peer) {
 399         nc->peer->peer = NULL;
 400     }
 401     g_free(nc->name);
 402     g_free(nc->model);
 403     if (nc->destructor) {
 404         nc->destructor(nc);
 405     }
 406 }
 407
 408 void qemu_del_net_client(NetClientState *nc)
 409 {
 410     NetClientState *ncs[MAX_QUEUE_NUM];
 411     int queues, i;
 412     NetFilterState *nf, *next;
 413
 414     assert(nc->info->type != NET_CLIENT_DRIVER_NIC);
 415
 416     /* If the NetClientState belongs to a multiqueue backend, we will change all
 417      * other NetClientStates also.
 418      */
 419     queues = qemu_find_net_clients_except(nc->name, ncs,
 420                                           NET_CLIENT_DRIVER_NIC,
 421                                           MAX_QUEUE_NUM);
 422     assert(queues != 0);
 423
 424     QTAILQ_FOREACH_SAFE(nf, &nc->filters, next, next) {
 425         object_unparent(OBJECT(nf));
 426     }
 427
 428     /* If there is a peer NIC, delete and cleanup client, but do not free. */
 429     if (nc->peer && nc->peer->info->type == NET_CLIENT_DRIVER_NIC) {
 430         NICState *nic = qemu_get_nic(nc->peer);
 431         if (nic->peer_deleted) {
 432             return;
 433         }
 434         nic->peer_deleted = true;
 435
 436         for (i = 0; i < queues; i++) {
 437             ncs[i]->peer->link_down = true;
 438         }
 439
 440         if (nc->peer->info->link_status_changed) {
 441             nc->peer->info->link_status_changed(nc->peer);
 442         }
 443
 444         for (i = 0; i < queues; i++) {
 445             qemu_cleanup_net_client(ncs[i]);
 446         }
 447
 448         return;
 449     }
 450
 451     for (i = 0; i < queues; i++) {
 452         qemu_cleanup_net_client(ncs[i]);
 453         qemu_free_net_client(ncs[i]);
 454     }
 455 }
 456
 457 void qemu_del_nic(NICState *nic)
 458 {
 459     int i, queues = MAX(nic->conf->peers.queues, 1);
 460
 461     qemu_macaddr_set_free(&nic->conf->macaddr);
 462
 463     for (i = 0; i < queues; i++) {
 464         NetClientState *nc = qemu_get_subqueue(nic, i);
 465         /* If this is a peer NIC and peer has already been deleted, free it now. */
 466         if (nic->peer_deleted) {
 467             qemu_free_net_client(nc->peer);
 468         } else if (nc->peer) {
 469             /* if there are RX packets pending, complete them */
 470             qemu_purge_queued_packets(nc->peer);
 471         }
 472     }
 473
 474     for (i = queues - 1; i >= 0; i--) {
 475         NetClientState *nc = qemu_get_subqueue(nic, i);
 476
 477         qemu_cleanup_net_client(nc);
 478         qemu_free_net_client(nc);
 479     }
 480
 481     g_free(nic);
 482 }
 483
 484 void qemu_foreach_nic(qemu_nic_foreach func, void *opaque)
 485 {
 486     NetClientState *nc;
 487
 488     QTAILQ_FOREACH(nc, &net_clients, next) {
 489         if (nc->info->type == NET_CLIENT_DRIVER_NIC) {
 490             if (nc->queue_index == 0) {
 491                 func(qemu_get_nic(nc), opaque);
 492             }
 493         }
 494     }
 495 }
 496
 497 bool qemu_has_ufo(NetClientState *nc)
 498 {
 499     if (!nc || !nc->info->has_ufo) {
 500         return false;
 501     }
 502
 503     return nc->info->has_ufo(nc);
 504 }
 505
 506 bool qemu_has_uso(NetClientState *nc)
 507 {
 508     if (!nc || !nc->info->has_uso) {
 509         return false;
 510     }
 511
 512     return nc->info->has_uso(nc);
 513 }
 514
 515 bool qemu_has_vnet_hdr(NetClientState *nc)
 516 {
 517     if (!nc || !nc->info->has_vnet_hdr) {
 518         return false;
 519     }
 520
 521     return nc->info->has_vnet_hdr(nc);
 522 }
 523
 524 bool qemu_has_vnet_hdr_len(NetClientState *nc, int len)
 525 {
 526     if (!nc || !nc->info->has_vnet_hdr_len) {
 527         return false;
 528     }
 529
 530     return nc->info->has_vnet_hdr_len(nc, len);
 531 }
 532
 533 void qemu_set_offload(NetClientState *nc, int csum, int tso4, int tso6,
 534                           int ecn, int ufo, int uso4, int uso6)
 535 {
 536     if (!nc || !nc->info->set_offload) {
 537         return;
 538     }
 539
 540     nc->info->set_offload(nc, csum, tso4, tso6, ecn, ufo, uso4, uso6);
 541 }
 542
 543 int qemu_get_vnet_hdr_len(NetClientState *nc)
 544 {
 545     return nc->vnet_hdr_len;
 546 }
 547
 548 void qemu_set_vnet_hdr_len(NetClientState *nc, int len)
 549 {
 550     if (!nc || !nc->info->set_vnet_hdr_len) {
 551         return;
 552     }
 553
 554     assert(len == sizeof(struct virtio_net_hdr_mrg_rxbuf) ||
 555            len == sizeof(struct virtio_net_hdr) ||
 556            len == sizeof(struct virtio_net_hdr_v1_hash));
 557
 558     nc->vnet_hdr_len = len;
 559     nc->info->set_vnet_hdr_len(nc, len);
 560 }
 561
 562 int qemu_set_vnet_le(NetClientState *nc, bool is_le)
 563 {
 564 #if HOST_BIG_ENDIAN
 565     if (!nc || !nc->info->set_vnet_le) {
 566         return -ENOSYS;
 567     }
 568
 569     return nc->info->set_vnet_le(nc, is_le);
 570 #else
 571     return 0;
 572 #endif
 573 }
 574
 575 int qemu_set_vnet_be(NetClientState *nc, bool is_be)
 576 {
 577 #if HOST_BIG_ENDIAN
 578     return 0;
 579 #else
 580     if (!nc || !nc->info->set_vnet_be) {
 581         return -ENOSYS;
 582     }
 583
 584     return nc->info->set_vnet_be(nc, is_be);
 585 #endif
 586 }
 587
 588 int qemu_can_receive_packet(NetClientState *nc)
 589 {
 590     if (nc->receive_disabled) {
 591         return 0;
 592     } else if (nc->info->can_receive &&
 593                !nc->info->can_receive(nc)) {
 594         return 0;
 595     }
 596     return 1;
 597 }
 598
 599 int qemu_can_send_packet(NetClientState *sender)
 600 {
 601     int vm_running = runstate_is_running();
 602
 603     if (!vm_running) {
 604         return 0;
 605     }
 606
 607     if (!sender->peer) {
 608         return 1;
 609     }
 610
 611     return qemu_can_receive_packet(sender->peer);
 612 }
 613
 614 static ssize_t filter_receive_iov(NetClientState *nc,
 615                                   NetFilterDirection direction,
 616                                   NetClientState *sender,
 617                                   unsigned flags,
 618                                   const struct iovec *iov,
 619                                   int iovcnt,
 620                                   NetPacketSent *sent_cb)
 621 {
 622     ssize_t ret = 0;
 623     NetFilterState *nf = NULL;
 624
 625     if (direction == NET_FILTER_DIRECTION_TX) {
 626         QTAILQ_FOREACH(nf, &nc->filters, next) {
 627             ret = qemu_netfilter_receive(nf, direction, sender, flags, iov,
 628                                          iovcnt, sent_cb);
 629             if (ret) {
 630                 return ret;
 631             }
 632         }
 633     } else {
 634         QTAILQ_FOREACH_REVERSE(nf, &nc->filters, next) {
 635             ret = qemu_netfilter_receive(nf, direction, sender, flags, iov,
 636                                          iovcnt, sent_cb);
 637             if (ret) {
 638                 return ret;
 639             }
 640         }
 641     }
 642
 643     return ret;
 644 }
 645
 646 static ssize_t filter_receive(NetClientState *nc,
 647                               NetFilterDirection direction,
 648                               NetClientState *sender,
 649                               unsigned flags,
 650                               const uint8_t *data,
 651                               size_t size,
 652                               NetPacketSent *sent_cb)
 653 {
 654     struct iovec iov = {
 655         .iov_base = (void *)data,
 656         .iov_len = size
 657     };
 658
 659     return filter_receive_iov(nc, direction, sender, flags, &iov, 1, sent_cb);
 660 }
 661
 662 void qemu_purge_queued_packets(NetClientState *nc)
 663 {
 664     if (!nc->peer) {
 665         return;
 666     }
 667
 668     qemu_net_queue_purge(nc->peer->incoming_queue, nc);
 669 }
 670
 671 void qemu_flush_or_purge_queued_packets(NetClientState *nc, bool purge)
 672 {
 673     nc->receive_disabled = 0;
 674
 675     if (nc->peer && nc->peer->info->type == NET_CLIENT_DRIVER_HUBPORT) {
 676         if (net_hub_flush(nc->peer)) {
 677             qemu_notify_event();
 678         }
 679     }
 680     if (qemu_net_queue_flush(nc->incoming_queue)) {
 681         /* We emptied the queue successfully, signal to the IO thread to repoll
 682          * the file descriptor (for tap, for example).
 683          */
 684         qemu_notify_event();
 685     } else if (purge) {
 686         /* Unable to empty the queue, purge remaining packets */
 687         qemu_net_queue_purge(nc->incoming_queue, nc->peer);
 688     }
 689 }
 690
 691 void qemu_flush_queued_packets(NetClientState *nc)
 692 {
 693     qemu_flush_or_purge_queued_packets(nc, false);
 694 }
 695
 696 static ssize_t qemu_send_packet_async_with_flags(NetClientState *sender,
 697                                                  unsigned flags,
 698                                                  const uint8_t *buf, int size,
 699                                                  NetPacketSent *sent_cb)
 700 {
 701     NetQueue *queue;
 702     int ret;
 703
 704 #ifdef DEBUG_NET
 705     printf("qemu_send_packet_async:\n");
 706     qemu_hexdump(stdout, "net", buf, size);
 707 #endif
 708
 709     if (sender->link_down || !sender->peer) {
 710         return size;
 711     }
 712
 713     /* Let filters handle the packet first */
 714     ret = filter_receive(sender, NET_FILTER_DIRECTION_TX,
 715                          sender, flags, buf, size, sent_cb);
 716     if (ret) {
 717         return ret;
 718     }
 719
 720     ret = filter_receive(sender->peer, NET_FILTER_DIRECTION_RX,
 721                          sender, flags, buf, size, sent_cb);
 722     if (ret) {
 723         return ret;
 724     }
 725
 726     queue = sender->peer->incoming_queue;
 727
 728     return qemu_net_queue_send(queue, sender, flags, buf, size, sent_cb);
 729 }
 730
 731 ssize_t qemu_send_packet_async(NetClientState *sender,
 732                                const uint8_t *buf, int size,
 733                                NetPacketSent *sent_cb)
 734 {
 735     return qemu_send_packet_async_with_flags(sender, QEMU_NET_PACKET_FLAG_NONE,
 736                                              buf, size, sent_cb);
 737 }
 738
 739 ssize_t qemu_send_packet(NetClientState *nc, const uint8_t *buf, int size)
 740 {
 741     return qemu_send_packet_async(nc, buf, size, NULL);
 742 }
 743
 744 ssize_t qemu_receive_packet(NetClientState *nc, const uint8_t *buf, int size)
 745 {
 746     if (!qemu_can_receive_packet(nc)) {
 747         return 0;
 748     }
 749
 750     return qemu_net_queue_receive(nc->incoming_queue, buf, size);
 751 }
 752
 753 ssize_t qemu_send_packet_raw(NetClientState *nc, const uint8_t *buf, int size)
 754 {
 755     return qemu_send_packet_async_with_flags(nc, QEMU_NET_PACKET_FLAG_RAW,
 756                                              buf, size, NULL);
 757 }
 758
 759 static ssize_t nc_sendv_compat(NetClientState *nc, const struct iovec *iov,
 760                                int iovcnt, unsigned flags)
 761 {
 762     uint8_t *buf = NULL;
 763     uint8_t *buffer;
 764     size_t offset;
 765     ssize_t ret;
 766
 767     if (iovcnt == 1) {
 768         buffer = iov[0].iov_base;
 769         offset = iov[0].iov_len;
 770     } else {
 771         offset = iov_size(iov, iovcnt);
 772         if (offset > NET_BUFSIZE) {
 773             return -1;
 774         }
 775         buf = g_malloc(offset);
 776         buffer = buf;
 777         offset = iov_to_buf(iov, iovcnt, 0, buf, offset);
 778     }
 779
 780     ret = nc->info->receive(nc, buffer, offset);
 781
 782     g_free(buf);
 783     return ret;
 784 }
 785
 786 static ssize_t qemu_deliver_packet_iov(NetClientState *sender,
 787                                        unsigned flags,
 788                                        const struct iovec *iov,
 789                                        int iovcnt,
 790                                        void *opaque)
 791 {
 792     MemReentrancyGuard *owned_reentrancy_guard;
 793     NetClientState *nc = opaque;
 794     int ret;
 795     struct virtio_net_hdr_v1_hash vnet_hdr = { };
 796     g_autofree struct iovec *iov_copy = NULL;
 797
 798
 799     if (nc->link_down) {
 800         return iov_size(iov, iovcnt);
 801     }
 802
 803     if (nc->receive_disabled) {
 804         return 0;
 805     }
 806
 807     if (nc->info->type != NET_CLIENT_DRIVER_NIC ||
 808         qemu_get_nic(nc)->reentrancy_guard->engaged_in_io) {
 809         owned_reentrancy_guard = NULL;
 810     } else {
 811         owned_reentrancy_guard = qemu_get_nic(nc)->reentrancy_guard;
 812         owned_reentrancy_guard->engaged_in_io = true;
 813     }
 814
 815     if ((flags & QEMU_NET_PACKET_FLAG_RAW) && nc->vnet_hdr_len) {
 816         iov_copy = g_new(struct iovec, iovcnt + 1);
 817         iov_copy[0].iov_base = &vnet_hdr;
 818         iov_copy[0].iov_len =  nc->vnet_hdr_len;
 819         memcpy(&iov_copy[1], iov, iovcnt * sizeof(*iov));
 820         iov = iov_copy;
 821     }
 822
 823     if (nc->info->receive_iov) {
 824         ret = nc->info->receive_iov(nc, iov, iovcnt);
 825     } else {
 826         ret = nc_sendv_compat(nc, iov, iovcnt, flags);
 827     }
 828
 829     if (owned_reentrancy_guard) {
 830         owned_reentrancy_guard->engaged_in_io = false;
 831     }
 832
 833     if (ret == 0) {
 834         nc->receive_disabled = 1;
 835     }
 836
 837     return ret;
 838 }
 839
 840 ssize_t qemu_sendv_packet_async(NetClientState *sender,
 841                                 const struct iovec *iov, int iovcnt,
 842                                 NetPacketSent *sent_cb)
 843 {
 844     NetQueue *queue;
 845     size_t size = iov_size(iov, iovcnt);
 846     int ret;
 847
 848     if (size > NET_BUFSIZE) {
 849         return size;
 850     }
 851
 852     if (sender->link_down || !sender->peer) {
 853         return size;
 854     }
 855
 856     /* Let filters handle the packet first */
 857     ret = filter_receive_iov(sender, NET_FILTER_DIRECTION_TX, sender,
 858                              QEMU_NET_PACKET_FLAG_NONE, iov, iovcnt, sent_cb);
 859     if (ret) {
 860         return ret;
 861     }
 862
 863     ret = filter_receive_iov(sender->peer, NET_FILTER_DIRECTION_RX, sender,
 864                              QEMU_NET_PACKET_FLAG_NONE, iov, iovcnt, sent_cb);
 865     if (ret) {
 866         return ret;
 867     }
 868
 869     queue = sender->peer->incoming_queue;
 870
 871     return qemu_net_queue_send_iov(queue, sender,
 872                                    QEMU_NET_PACKET_FLAG_NONE,
 873                                    iov, iovcnt, sent_cb);
 874 }
 875
 876 ssize_t
 877 qemu_sendv_packet(NetClientState *nc, const struct iovec *iov, int iovcnt)
 878 {
 879     return qemu_sendv_packet_async(nc, iov, iovcnt, NULL);
 880 }
 881
 882 NetClientState *qemu_find_netdev(const char *id)
 883 {
 884     NetClientState *nc;
 885
 886     QTAILQ_FOREACH(nc, &net_clients, next) {
 887         if (nc->info->type == NET_CLIENT_DRIVER_NIC)
 888             continue;
 889         if (!strcmp(nc->name, id)) {
 890             return nc;
 891         }
 892     }
 893
 894     return NULL;
 895 }
 896
 897 int qemu_find_net_clients_except(const char *id, NetClientState **ncs,
 898                                  NetClientDriver type, int max)
 899 {
 900     NetClientState *nc;
 901     int ret = 0;
 902
 903     QTAILQ_FOREACH(nc, &net_clients, next) {
 904         if (nc->info->type == type) {
 905             continue;
 906         }
 907         if (!id || !strcmp(nc->name, id)) {
 908             if (ret < max) {
 909                 ncs[ret] = nc;
 910             }
 911             ret++;
 912         }
 913     }
 914
 915     return ret;
 916 }
 917
 918 static int nic_get_free_idx(void)
 919 {
 920     int index;
 921
 922     for (index = 0; index < MAX_NICS; index++)
 923         if (!nd_table[index].used)
 924             return index;
 925     return -1;
 926 }
 927
 928 GPtrArray *qemu_get_nic_models(const char *device_type)
 929 {
 930     GPtrArray *nic_models = g_ptr_array_new();
 931     GSList *list = object_class_get_list_sorted(device_type, false);
 932
 933     while (list) {
 934         DeviceClass *dc = OBJECT_CLASS_CHECK(DeviceClass, list->data,
 935                                              TYPE_DEVICE);
 936         GSList *next;
 937         if (test_bit(DEVICE_CATEGORY_NETWORK, dc->categories) &&
 938             dc->user_creatable) {
 939             const char *name = object_class_get_name(list->data);
 940             /*
 941              * A network device might also be something else than a NIC, see
 942              * e.g. the "rocker" device. Thus we have to look for the "netdev"
 943              * property, too. Unfortunately, some devices like virtio-net only
 944              * create this property during instance_init, so we have to create
 945              * a temporary instance here to be able to check it.
 946              */
 947             Object *obj = object_new_with_class(OBJECT_CLASS(dc));
 948             if (object_property_find(obj, "netdev")) {
 949                 g_ptr_array_add(nic_models, (gpointer)name);
 950             }
 951             object_unref(obj);
 952         }
 953         next = list->next;
 954         g_slist_free_1(list);
 955         list = next;
 956     }
 957     g_ptr_array_add(nic_models, NULL);
 958
 959     return nic_models;
 960 }
 961
 962 static int net_init_nic(const Netdev *netdev, const char *name,
 963                         NetClientState *peer, Error **errp)
 964 {
 965     int idx;
 966     NICInfo *nd;
 967     const NetLegacyNicOptions *nic;
 968
 969     assert(netdev->type == NET_CLIENT_DRIVER_NIC);
 970     nic = &netdev->u.nic;
 971
 972     idx = nic_get_free_idx();
 973     if (idx == -1 || nb_nics >= MAX_NICS) {
 974         error_setg(errp, "too many NICs");
 975         return -1;
 976     }
 977
 978     nd = &nd_table[idx];
 979
 980     memset(nd, 0, sizeof(*nd));
 981
 982     if (nic->netdev) {
 983         nd->netdev = qemu_find_netdev(nic->netdev);
 984         if (!nd->netdev) {
 985             error_setg(errp, "netdev '%s' not found", nic->netdev);
 986             return -1;
 987         }
 988     } else {
 989         assert(peer);
 990         nd->netdev = peer;
 991     }
 992     nd->name = g_strdup(name);
 993     if (nic->model) {
 994         nd->model = g_strdup(nic->model);
 995     }
 996     if (nic->addr) {
 997         nd->devaddr = g_strdup(nic->addr);
 998     }
 999
1000     if (nic->macaddr &&
1001         net_parse_macaddr(nd->macaddr.a, nic->macaddr) < 0) {
1002         error_setg(errp, "invalid syntax for ethernet address");
1003         return -1;
1004     }
1005     if (nic->macaddr &&
1006         is_multicast_ether_addr(nd->macaddr.a)) {
1007         error_setg(errp,
1008                    "NIC cannot have multicast MAC address (odd 1st byte)");
1009         return -1;
1010     }
1011     qemu_macaddr_default_if_unset(&nd->macaddr);
1012
1013     if (nic->has_vectors) {
1014         if (nic->vectors > 0x7ffffff) {
1015             error_setg(errp, "invalid # of vectors: %"PRIu32, nic->vectors);
1016             return -1;
1017         }
1018         nd->nvectors = nic->vectors;
1019     } else {
1020         nd->nvectors = DEV_NVECTORS_UNSPECIFIED;
1021     }
1022
1023     nd->used = 1;
1024     nb_nics++;
1025
1026     return idx;
1027 }
1028
1029 static gboolean add_nic_result(gpointer key, gpointer value, gpointer user_data)
1030 {
1031     GPtrArray *results = user_data;
1032     GPtrArray *alias_list = value;
1033     const char *model = key;
1034     char *result;
1035
1036     if (!alias_list) {
1037         result = g_strdup(model);
1038     } else {
1039         GString *result_str = g_string_new(model);
1040         int i;
1041
1042         g_string_append(result_str, " (aka ");
1043         for (i = 0; i < alias_list->len; i++) {
1044             if (i) {
1045                 g_string_append(result_str, ", ");
1046             }
1047             g_string_append(result_str, alias_list->pdata[i]);
1048         }
1049         g_string_append(result_str, ")");
1050         result = result_str->str;
1051         g_string_free(result_str, false);
1052         g_ptr_array_unref(alias_list);
1053     }
1054     g_ptr_array_add(results, result);
1055     return true;
1056 }
1057
1058 static int model_cmp(char **a, char **b)
1059 {
1060     return strcmp(*a, *b);
1061 }
1062
1063 static void show_nic_models(void)
1064 {
1065     GPtrArray *results = g_ptr_array_new();
1066     int i;
1067
1068     g_hash_table_foreach_remove(nic_model_help, add_nic_result, results);
1069     g_ptr_array_sort(results, (GCompareFunc)model_cmp);
1070
1071     printf("Available NIC models for this configuration:\n");
1072     for (i = 0 ; i < results->len; i++) {
1073         printf("%s\n", (char *)results->pdata[i]);
1074     }
1075     g_hash_table_unref(nic_model_help);
1076     nic_model_help = NULL;
1077 }
1078
1079 static void add_nic_model_help(const char *model, const char *alias)
1080 {
1081     GPtrArray *alias_list = NULL;
1082
1083     if (g_hash_table_lookup_extended(nic_model_help, model, NULL,
1084                                      (gpointer *)&alias_list)) {
1085         /* Already exists, no alias to add: return */
1086         if (!alias) {
1087             return;
1088         }
1089         if (alias_list) {
1090             /* Check if this alias is already in the list. Add if not. */
1091             if (!g_ptr_array_find_with_equal_func(alias_list, alias,
1092                                                   g_str_equal, NULL)) {
1093                 g_ptr_array_add(alias_list, g_strdup(alias));
1094             }
1095             return;
1096         }
1097     }
1098     /* Either this model wasn't in the list already, or a first alias added */
1099     if (alias) {
1100         alias_list = g_ptr_array_new();
1101         g_ptr_array_set_free_func(alias_list, g_free);
1102         g_ptr_array_add(alias_list, g_strdup(alias));
1103     }
1104     g_hash_table_replace(nic_model_help, g_strdup(model), alias_list);
1105 }
1106
1107 NICInfo *qemu_find_nic_info(const char *typename, bool match_default,
1108                             const char *alias)
1109 {
1110     NICInfo *nd;
1111     int i;
1112
1113     if (nic_model_help) {
1114         add_nic_model_help(typename, alias);
1115     }
1116
1117     for (i = 0; i < nb_nics; i++) {
1118         nd = &nd_table[i];
1119
1120         if (!nd->used || nd->instantiated) {
1121             continue;
1122         }
1123
1124         if ((match_default && !nd->model) || !g_strcmp0(nd->model, typename)
1125             || (alias && !g_strcmp0(nd->model, alias))) {
1126             return nd;
1127         }
1128     }
1129     return NULL;
1130 }
1131
1132 static bool is_nic_model_help_option(const char *model)
1133 {
1134     if (model && is_help_option(model)) {
1135         /*
1136          * Trigger the help output by instantiating the hash table which
1137          * will gather tha available models as they get registered.
1138          */
1139         if (!nic_model_help) {
1140             nic_model_help = g_hash_table_new_full(g_str_hash, g_str_equal,
1141                                                    g_free, NULL);
1142         }
1143         return true;
1144     }
1145     return false;
1146 }
1147
1148 /* "I have created a device. Please configure it if you can" */
1149 bool qemu_configure_nic_device(DeviceState *dev, bool match_default,
1150                                const char *alias)
1151 {
1152     NICInfo *nd = qemu_find_nic_info(object_get_typename(OBJECT(dev)),
1153                                      match_default, alias);
1154
1155     if (nd) {
1156         qdev_set_nic_properties(dev, nd);
1157         return true;
1158     }
1159     return false;
1160 }
1161
1162 /* "Please create a device, if you have a configuration for it" */
1163 DeviceState *qemu_create_nic_device(const char *typename, bool match_default,
1164                                     const char *alias)
1165 {
1166     NICInfo *nd = qemu_find_nic_info(typename, match_default, alias);
1167     DeviceState *dev;
1168
1169     if (!nd) {
1170         return NULL;
1171     }
1172
1173     dev = qdev_new(typename);
1174     qdev_set_nic_properties(dev, nd);
1175     return dev;
1176 }
1177
1178 void qemu_create_nic_bus_devices(BusState *bus, const char *parent_type,
1179                                  const char *default_model,
1180                                  const char *alias, const char *alias_target)
1181 {
1182     GPtrArray *nic_models = qemu_get_nic_models(parent_type);
1183     const char *model;
1184     DeviceState *dev;
1185     NICInfo *nd;
1186     int i;
1187
1188     if (nic_model_help) {
1189         if (alias_target) {
1190             add_nic_model_help(alias_target, alias);
1191         }
1192         for (i = 0; i < nic_models->len - 1; i++) {
1193             add_nic_model_help(nic_models->pdata[i], NULL);
1194         }
1195     }
1196
1197     /* Drop the NULL terminator which would make g_str_equal() unhappy */
1198     nic_models->len--;
1199
1200     for (i = 0; i < nb_nics; i++) {
1201         nd = &nd_table[i];
1202
1203         if (!nd->used || nd->instantiated) {
1204             continue;
1205         }
1206
1207         model = nd->model ? nd->model : default_model;
1208         if (!model) {
1209             continue;
1210         }
1211
1212         /* Each bus type is allowed *one* substitution */
1213         if (g_str_equal(model, alias)) {
1214             model = alias_target;
1215         }
1216
1217         if (!g_ptr_array_find_with_equal_func(nic_models, model,
1218                                               g_str_equal, NULL)) {
1219             /* This NIC does not live on this bus. */
1220             continue;
1221         }
1222
1223         dev = qdev_new(model);
1224         qdev_set_nic_properties(dev, nd);
1225         qdev_realize_and_unref(dev, bus, &error_fatal);
1226     }
1227
1228     g_ptr_array_free(nic_models, true);
1229 }
1230
1231 static int (* const net_client_init_fun[NET_CLIENT_DRIVER__MAX])(
1232     const Netdev *netdev,
1233     const char *name,
1234     NetClientState *peer, Error **errp) = {
1235         [NET_CLIENT_DRIVER_NIC]       = net_init_nic,
1236 #ifdef CONFIG_SLIRP
1237         [NET_CLIENT_DRIVER_USER]      = net_init_slirp,
1238 #endif
1239         [NET_CLIENT_DRIVER_TAP]       = net_init_tap,
1240         [NET_CLIENT_DRIVER_SOCKET]    = net_init_socket,
1241         [NET_CLIENT_DRIVER_STREAM]    = net_init_stream,
1242         [NET_CLIENT_DRIVER_DGRAM]     = net_init_dgram,
1243 #ifdef CONFIG_VDE
1244         [NET_CLIENT_DRIVER_VDE]       = net_init_vde,
1245 #endif
1246 #ifdef CONFIG_NETMAP
1247         [NET_CLIENT_DRIVER_NETMAP]    = net_init_netmap,
1248 #endif
1249 #ifdef CONFIG_AF_XDP
1250         [NET_CLIENT_DRIVER_AF_XDP]    = net_init_af_xdp,
1251 #endif
1252 #ifdef CONFIG_NET_BRIDGE
1253         [NET_CLIENT_DRIVER_BRIDGE]    = net_init_bridge,
1254 #endif
1255         [NET_CLIENT_DRIVER_HUBPORT]   = net_init_hubport,
1256 #ifdef CONFIG_VHOST_NET_USER
1257         [NET_CLIENT_DRIVER_VHOST_USER] = net_init_vhost_user,
1258 #endif
1259 #ifdef CONFIG_VHOST_NET_VDPA
1260         [NET_CLIENT_DRIVER_VHOST_VDPA] = net_init_vhost_vdpa,
1261 #endif
1262 #ifdef CONFIG_L2TPV3
1263         [NET_CLIENT_DRIVER_L2TPV3]    = net_init_l2tpv3,
1264 #endif
1265 #ifdef CONFIG_VMNET
1266         [NET_CLIENT_DRIVER_VMNET_HOST] = net_init_vmnet_host,
1267         [NET_CLIENT_DRIVER_VMNET_SHARED] = net_init_vmnet_shared,
1268         [NET_CLIENT_DRIVER_VMNET_BRIDGED] = net_init_vmnet_bridged,
1269 #endif /* CONFIG_VMNET */
1270 };
1271
1272
1273 static int net_client_init1(const Netdev *netdev, bool is_netdev, Error **errp)
1274 {
1275     NetClientState *peer = NULL;
1276     NetClientState *nc;
1277
1278     if (is_netdev) {
1279         if (netdev->type == NET_CLIENT_DRIVER_NIC ||
1280             !net_client_init_fun[netdev->type]) {
1281             error_setg(errp, "network backend '%s' is not compiled into this binary",
1282                        NetClientDriver_str(netdev->type));
1283             return -1;
1284         }
1285     } else {
1286         if (netdev->type == NET_CLIENT_DRIVER_NONE) {
1287             return 0; /* nothing to do */
1288         }
1289         if (netdev->type == NET_CLIENT_DRIVER_HUBPORT) {
1290             error_setg(errp, "network backend '%s' is only supported with -netdev/-nic",
1291                        NetClientDriver_str(netdev->type));
1292             return -1;
1293         }
1294
1295         if (!net_client_init_fun[netdev->type]) {
1296             error_setg(errp, "network backend '%s' is not compiled into this binary",
1297                        NetClientDriver_str(netdev->type));
1298             return -1;
1299         }
1300
1301         /* Do not add to a hub if it's a nic with a netdev= parameter. */
1302         if (netdev->type != NET_CLIENT_DRIVER_NIC ||
1303             !netdev->u.nic.netdev) {
1304             peer = net_hub_add_port(0, NULL, NULL);
1305         }
1306     }
1307
1308     nc = qemu_find_netdev(netdev->id);
1309     if (nc) {
1310         error_setg(errp, "Duplicate ID '%s'", netdev->id);
1311         return -1;
1312     }
1313
1314     if (net_client_init_fun[netdev->type](netdev, netdev->id, peer, errp) < 0) {
1315         /* FIXME drop when all init functions store an Error */
1316         if (errp && !*errp) {
1317             error_setg(errp, "Device '%s' could not be initialized",
1318                        NetClientDriver_str(netdev->type));
1319         }
1320         return -1;
1321     }
1322
1323     if (is_netdev) {
1324         nc = qemu_find_netdev(netdev->id);
1325         assert(nc);
1326         nc->is_netdev = true;
1327     }
1328
1329     return 0;
1330 }
1331
1332 void show_netdevs(void)
1333 {
1334     int idx;
1335     const char *available_netdevs[] = {
1336         "socket",
1337         "stream",
1338         "dgram",
1339         "hubport",
1340         "tap",
1341 #ifdef CONFIG_SLIRP
1342         "user",
1343 #endif
1344 #ifdef CONFIG_L2TPV3
1345         "l2tpv3",
1346 #endif
1347 #ifdef CONFIG_VDE
1348         "vde",
1349 #endif
1350 #ifdef CONFIG_NET_BRIDGE
1351         "bridge",
1352 #endif
1353 #ifdef CONFIG_NETMAP
1354         "netmap",
1355 #endif
1356 #ifdef CONFIG_AF_XDP
1357         "af-xdp",
1358 #endif
1359 #ifdef CONFIG_POSIX
1360         "vhost-user",
1361 #endif
1362 #ifdef CONFIG_VHOST_VDPA
1363         "vhost-vdpa",
1364 #endif
1365 #ifdef CONFIG_VMNET
1366         "vmnet-host",
1367         "vmnet-shared",
1368         "vmnet-bridged",
1369 #endif
1370     };
1371
1372     qemu_printf("Available netdev backend types:\n");
1373     for (idx = 0; idx < ARRAY_SIZE(available_netdevs); idx++) {
1374         qemu_printf("%s\n", available_netdevs[idx]);
1375     }
1376 }
1377
1378 static int net_client_init(QemuOpts *opts, bool is_netdev, Error **errp)
1379 {
1380     gchar **substrings = NULL;
1381     Netdev *object = NULL;
1382     int ret = -1;
1383     Visitor *v = opts_visitor_new(opts);
1384
1385     /* Parse convenience option format ipv6-net=fec0::0[/64] */
1386     const char *ip6_net = qemu_opt_get(opts, "ipv6-net");
1387
1388     if (ip6_net) {
1389         char *prefix_addr;
1390         unsigned long prefix_len = 64; /* Default 64bit prefix length. */
1391
1392         substrings = g_strsplit(ip6_net, "/", 2);
1393         if (!substrings || !substrings[0]) {
1394             error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "ipv6-net",
1395                        "a valid IPv6 prefix");
1396             goto out;
1397         }
1398
1399         prefix_addr = substrings[0];
1400
1401         /* Handle user-specified prefix length. */
1402         if (substrings[1] &&
1403             qemu_strtoul(substrings[1], NULL, 10, &prefix_len))
1404         {
1405             error_setg(errp,
1406                        "parameter 'ipv6-net' expects a number after '/'");
1407             goto out;
1408         }
1409
1410         qemu_opt_set(opts, "ipv6-prefix", prefix_addr, &error_abort);
1411         qemu_opt_set_number(opts, "ipv6-prefixlen", prefix_len,
1412                             &error_abort);
1413         qemu_opt_unset(opts, "ipv6-net");
1414     }
1415
1416     /* Create an ID for -net if the user did not specify one */
1417     if (!is_netdev && !qemu_opts_id(opts)) {
1418         qemu_opts_set_id(opts, id_generate(ID_NET));
1419     }
1420
1421     if (visit_type_Netdev(v, NULL, &object, errp)) {
1422         ret = net_client_init1(object, is_netdev, errp);
1423     }
1424
1425     qapi_free_Netdev(object);
1426
1427 out:
1428     g_strfreev(substrings);
1429     visit_free(v);
1430     return ret;
1431 }
1432
1433 void netdev_add(QemuOpts *opts, Error **errp)
1434 {
1435     net_client_init(opts, true, errp);
1436 }
1437
1438 void qmp_netdev_add(Netdev *netdev, Error **errp)
1439 {
1440     if (!id_wellformed(netdev->id)) {
1441         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "id", "an identifier");
1442         return;
1443     }
1444
1445     net_client_init1(netdev, true, errp);
1446 }
1447
1448 void qmp_netdev_del(const char *id, Error **errp)
1449 {
1450     NetClientState *nc;
1451     QemuOpts *opts;
1452
1453     nc = qemu_find_netdev(id);
1454     if (!nc) {
1455         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1456                   "Device '%s' not found", id);
1457         return;
1458     }
1459
1460     if (!nc->is_netdev) {
1461         error_setg(errp, "Device '%s' is not a netdev", id);
1462         return;
1463     }
1464
1465     qemu_del_net_client(nc);
1466
1467     /*
1468      * Wart: we need to delete the QemuOpts associated with netdevs
1469      * created via CLI or HMP, to avoid bogus "Duplicate ID" errors in
1470      * HMP netdev_add.
1471      */
1472     opts = qemu_opts_find(qemu_find_opts("netdev"), id);
1473     if (opts) {
1474         qemu_opts_del(opts);
1475     }
1476 }
1477
1478 static void netfilter_print_info(Monitor *mon, NetFilterState *nf)
1479 {
1480     char *str;
1481     ObjectProperty *prop;
1482     ObjectPropertyIterator iter;
1483     Visitor *v;
1484
1485     /* generate info str */
1486     object_property_iter_init(&iter, OBJECT(nf));
1487     while ((prop = object_property_iter_next(&iter))) {
1488         if (!strcmp(prop->name, "type")) {
1489             continue;
1490         }
1491         v = string_output_visitor_new(false, &str);
1492         object_property_get(OBJECT(nf), prop->name, v, NULL);
1493         visit_complete(v, &str);
1494         visit_free(v);
1495         monitor_printf(mon, ",%s=%s", prop->name, str);
1496         g_free(str);
1497     }
1498     monitor_printf(mon, "\n");
1499 }
1500
1501 void print_net_client(Monitor *mon, NetClientState *nc)
1502 {
1503     NetFilterState *nf;
1504
1505     monitor_printf(mon, "%s: index=%d,type=%s,%s\n", nc->name,
1506                    nc->queue_index,
1507                    NetClientDriver_str(nc->info->type),
1508                    nc->info_str);
1509     if (!QTAILQ_EMPTY(&nc->filters)) {
1510         monitor_printf(mon, "filters:\n");
1511     }
1512     QTAILQ_FOREACH(nf, &nc->filters, next) {
1513         monitor_printf(mon, "  - %s: type=%s",
1514                        object_get_canonical_path_component(OBJECT(nf)),
1515                        object_get_typename(OBJECT(nf)));
1516         netfilter_print_info(mon, nf);
1517     }
1518 }
1519
1520 RxFilterInfoList *qmp_query_rx_filter(const char *name, Error **errp)
1521 {
1522     NetClientState *nc;
1523     RxFilterInfoList *filter_list = NULL, **tail = &filter_list;
1524
1525     QTAILQ_FOREACH(nc, &net_clients, next) {
1526         RxFilterInfo *info;
1527
1528         if (name && strcmp(nc->name, name) != 0) {
1529             continue;
1530         }
1531
1532         /* only query rx-filter information of NIC */
1533         if (nc->info->type != NET_CLIENT_DRIVER_NIC) {
1534             if (name) {
1535                 error_setg(errp, "net client(%s) isn't a NIC", name);
1536                 assert(!filter_list);
1537                 return NULL;
1538             }
1539             continue;
1540         }
1541
1542         /* only query information on queue 0 since the info is per nic,
1543          * not per queue
1544          */
1545         if (nc->queue_index != 0)
1546             continue;
1547
1548         if (nc->info->query_rx_filter) {
1549             info = nc->info->query_rx_filter(nc);
1550             QAPI_LIST_APPEND(tail, info);
1551         } else if (name) {
1552             error_setg(errp, "net client(%s) doesn't support"
1553                        " rx-filter querying", name);
1554             assert(!filter_list);
1555             return NULL;
1556         }
1557
1558         if (name) {
1559             break;
1560         }
1561     }
1562
1563     if (filter_list == NULL && name) {
1564         error_setg(errp, "invalid net client name: %s", name);
1565     }
1566
1567     return filter_list;
1568 }
1569
1570 void colo_notify_filters_event(int event, Error **errp)
1571 {
1572     NetClientState *nc;
1573     NetFilterState *nf;
1574     NetFilterClass *nfc = NULL;
1575     Error *local_err = NULL;
1576
1577     QTAILQ_FOREACH(nc, &net_clients, next) {
1578         QTAILQ_FOREACH(nf, &nc->filters, next) {
1579             nfc = NETFILTER_GET_CLASS(OBJECT(nf));
1580             nfc->handle_event(nf, event, &local_err);
1581             if (local_err) {
1582                 error_propagate(errp, local_err);
1583                 return;
1584             }
1585         }
1586     }
1587 }
1588
1589 void qmp_set_link(const char *name, bool up, Error **errp)
1590 {
1591     NetClientState *ncs[MAX_QUEUE_NUM];
1592     NetClientState *nc;
1593     int queues, i;
1594
1595     queues = qemu_find_net_clients_except(name, ncs,
1596                                           NET_CLIENT_DRIVER__MAX,
1597                                           MAX_QUEUE_NUM);
1598
1599     if (queues == 0) {
1600         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1601                   "Device '%s' not found", name);
1602         return;
1603     }
1604     nc = ncs[0];
1605
1606     for (i = 0; i < queues; i++) {
1607         ncs[i]->link_down = !up;
1608     }
1609
1610     if (nc->info->link_status_changed) {
1611         nc->info->link_status_changed(nc);
1612     }
1613
1614     if (nc->peer) {
1615         /* Change peer link only if the peer is NIC and then notify peer.
1616          * If the peer is a HUBPORT or a backend, we do not change the
1617          * link status.
1618          *
1619          * This behavior is compatible with qemu hubs where there could be
1620          * multiple clients that can still communicate with each other in
1621          * disconnected mode. For now maintain this compatibility.
1622          */
1623         if (nc->peer->info->type == NET_CLIENT_DRIVER_NIC) {
1624             for (i = 0; i < queues; i++) {
1625                 ncs[i]->peer->link_down = !up;
1626             }
1627         }
1628         if (nc->peer->info->link_status_changed) {
1629             nc->peer->info->link_status_changed(nc->peer);
1630         }
1631     }
1632 }
1633
1634 static void net_vm_change_state_handler(void *opaque, bool running,
1635                                         RunState state)
1636 {
1637     NetClientState *nc;
1638     NetClientState *tmp;
1639
1640     QTAILQ_FOREACH_SAFE(nc, &net_clients, next, tmp) {
1641         if (running) {
1642             /* Flush queued packets and wake up backends. */
1643             if (nc->peer && qemu_can_send_packet(nc)) {
1644                 qemu_flush_queued_packets(nc->peer);
1645             }
1646         } else {
1647             /* Complete all queued packets, to guarantee we don't modify
1648              * state later when VM is not running.
1649              */
1650             qemu_flush_or_purge_queued_packets(nc, true);
1651         }
1652     }
1653 }
1654
1655 void net_cleanup(void)
1656 {
1657     NetClientState *nc, **p = &QTAILQ_FIRST(&net_clients);
1658
1659     /*cleanup colo compare module for COLO*/
1660     colo_compare_cleanup();
1661
1662     /*
1663      * Walk the net_clients list and remove the netdevs but *not* any
1664      * NET_CLIENT_DRIVER_NIC entries. The latter are owned by the device
1665      * model which created them, and in some cases (e.g. xen-net-device)
1666      * the device itself may do cleanup at exit and will be upset if we
1667      * just delete its NIC from underneath it.
1668      *
1669      * Since qemu_del_net_client() may delete multiple entries, using
1670      * QTAILQ_FOREACH_SAFE() is not safe here. The only safe pointer
1671      * to keep as a bookmark is a NET_CLIENT_DRIVER_NIC entry, so keep
1672      * 'p' pointing to either the head of the list, or the 'next' field
1673      * of the latest NET_CLIENT_DRIVER_NIC, and operate on *p as we walk
1674      * the list.
1675      *
1676      * The 'nc' variable isn't part of the list traversal; it's purely
1677      * for convenience as too much '(*p)->' has a tendency to make the
1678      * readers' eyes bleed.
1679      */
1680     while (*p) {
1681         nc = *p;
1682         if (nc->info->type == NET_CLIENT_DRIVER_NIC) {
1683             /* Skip NET_CLIENT_DRIVER_NIC entries */
1684             p = &QTAILQ_NEXT(nc, next);
1685         } else {
1686             qemu_del_net_client(nc);
1687         }
1688     }
1689
1690     qemu_del_vm_change_state_handler(net_change_state_entry);
1691 }
1692
1693 void net_check_clients(void)
1694 {
1695     NetClientState *nc;
1696     int i;
1697
1698     if (nic_model_help) {
1699         show_nic_models();
1700         exit(0);
1701     }
1702     net_hub_check_clients();
1703
1704     QTAILQ_FOREACH(nc, &net_clients, next) {
1705         if (!nc->peer) {
1706             warn_report("%s %s has no peer",
1707                         nc->info->type == NET_CLIENT_DRIVER_NIC
1708                         ? "nic" : "netdev",
1709                         nc->name);
1710         }
1711     }
1712
1713     /* Check that all NICs requested via -net nic actually got created.
1714      * NICs created via -device don't need to be checked here because
1715      * they are always instantiated.
1716      */
1717     for (i = 0; i < MAX_NICS; i++) {
1718         NICInfo *nd = &nd_table[i];
1719         if (nd->used && !nd->instantiated) {
1720             warn_report("requested NIC (%s, model %s) "
1721                         "was not created (not supported by this machine?)",
1722                         nd->name ? nd->name : "anonymous",
1723                         nd->model ? nd->model : "unspecified");
1724         }
1725     }
1726 }
1727
1728 static int net_init_client(void *dummy, QemuOpts *opts, Error **errp)
1729 {
1730     const char *model = qemu_opt_get(opts, "model");
1731
1732     if (is_nic_model_help_option(model)) {
1733         return 0;
1734     }
1735
1736     return net_client_init(opts, false, errp);
1737 }
1738
1739 static int net_init_netdev(void *dummy, QemuOpts *opts, Error **errp)
1740 {
1741     const char *type = qemu_opt_get(opts, "type");
1742
1743     if (type && is_help_option(type)) {
1744         show_netdevs();
1745         exit(0);
1746     }
1747     return net_client_init(opts, true, errp);
1748 }
1749
1750 /* For the convenience "--nic" parameter */
1751 static int net_param_nic(void *dummy, QemuOpts *opts, Error **errp)
1752 {
1753     char *mac, *nd_id;
1754     int idx, ret;
1755     NICInfo *ni;
1756     const char *type;
1757
1758     type = qemu_opt_get(opts, "type");
1759     if (type) {
1760         if (g_str_equal(type, "none")) {
1761             return 0;    /* Nothing to do, default_net is cleared in vl.c */
1762         }
1763         if (is_help_option(type)) {
1764             GPtrArray *nic_models = qemu_get_nic_models(TYPE_DEVICE);
1765             int i;
1766             show_netdevs();
1767             printf("\n");
1768             printf("Available NIC models "
1769                    "(use -nic model=help for a filtered list):\n");
1770             for (i = 0 ; nic_models->pdata[i]; i++) {
1771                 printf("%s\n", (char *)nic_models->pdata[i]);
1772             }
1773             g_ptr_array_free(nic_models, true);
1774             exit(0);
1775         }
1776     }
1777
1778     idx = nic_get_free_idx();
1779     if (idx == -1 || nb_nics >= MAX_NICS) {
1780         error_setg(errp, "no more on-board/default NIC slots available");
1781         return -1;
1782     }
1783
1784     if (!type) {
1785         qemu_opt_set(opts, "type", "user", &error_abort);
1786     }
1787
1788     ni = &nd_table[idx];
1789     memset(ni, 0, sizeof(*ni));
1790     ni->model = qemu_opt_get_del(opts, "model");
1791
1792     if (is_nic_model_help_option(ni->model)) {
1793         return 0;
1794     }
1795
1796     /* Create an ID if the user did not specify one */
1797     nd_id = g_strdup(qemu_opts_id(opts));
1798     if (!nd_id) {
1799         nd_id = id_generate(ID_NET);
1800         qemu_opts_set_id(opts, nd_id);
1801     }
1802
1803     /* Handle MAC address */
1804     mac = qemu_opt_get_del(opts, "mac");
1805     if (mac) {
1806         ret = net_parse_macaddr(ni->macaddr.a, mac);
1807         g_free(mac);
1808         if (ret) {
1809             error_setg(errp, "invalid syntax for ethernet address");
1810             goto out;
1811         }
1812         if (is_multicast_ether_addr(ni->macaddr.a)) {
1813             error_setg(errp, "NIC cannot have multicast MAC address");
1814             ret = -1;
1815             goto out;
1816         }
1817     }
1818     qemu_macaddr_default_if_unset(&ni->macaddr);
1819
1820     ret = net_client_init(opts, true, errp);
1821     if (ret == 0) {
1822         ni->netdev = qemu_find_netdev(nd_id);
1823         ni->used = true;
1824         nb_nics++;
1825     }
1826
1827 out:
1828     g_free(nd_id);
1829     return ret;
1830 }
1831
1832 static void netdev_init_modern(void)
1833 {
1834     while (!QSIMPLEQ_EMPTY(&nd_queue)) {
1835         NetdevQueueEntry *nd = QSIMPLEQ_FIRST(&nd_queue);
1836
1837         QSIMPLEQ_REMOVE_HEAD(&nd_queue, entry);
1838         loc_push_restore(&nd->loc);
1839         net_client_init1(nd->nd, true, &error_fatal);
1840         loc_pop(&nd->loc);
1841         qapi_free_Netdev(nd->nd);
1842         g_free(nd);
1843     }
1844 }
1845
1846 void net_init_clients(void)
1847 {
1848     net_change_state_entry =
1849         qemu_add_vm_change_state_handler(net_vm_change_state_handler, NULL);
1850
1851     QTAILQ_INIT(&net_clients);
1852
1853     netdev_init_modern();
1854
1855     qemu_opts_foreach(qemu_find_opts("netdev"), net_init_netdev, NULL,
1856                       &error_fatal);
1857
1858     qemu_opts_foreach(qemu_find_opts("nic"), net_param_nic, NULL,
1859                       &error_fatal);
1860
1861     qemu_opts_foreach(qemu_find_opts("net"), net_init_client, NULL,
1862                       &error_fatal);
1863 }
1864
1865 /*
1866  * Does this -netdev argument use modern rather than traditional syntax?
1867  * Modern syntax is to be parsed with netdev_parse_modern().
1868  * Traditional syntax is to be parsed with net_client_parse().
1869  */
1870 bool netdev_is_modern(const char *optstr)
1871 {
1872     QemuOpts *opts;
1873     bool is_modern;
1874     const char *type;
1875     static QemuOptsList dummy_opts = {
1876         .name = "netdev",
1877         .implied_opt_name = "type",
1878         .head = QTAILQ_HEAD_INITIALIZER(dummy_opts.head),
1879         .desc = { { } },
1880     };
1881
1882     if (optstr[0] == '{') {
1883         /* This is JSON, which means it's modern syntax */
1884         return true;
1885     }
1886
1887     opts = qemu_opts_create(&dummy_opts, NULL, false, &error_abort);
1888     qemu_opts_do_parse(opts, optstr, dummy_opts.implied_opt_name,
1889                        &error_abort);
1890     type = qemu_opt_get(opts, "type");
1891     is_modern = !g_strcmp0(type, "stream") || !g_strcmp0(type, "dgram");
1892
1893     qemu_opts_reset(&dummy_opts);
1894
1895     return is_modern;
1896 }
1897
1898 /*
1899  * netdev_parse_modern() uses modern, more expressive syntax than
1900  * net_client_parse(), but supports only the -netdev option.
1901  * netdev_parse_modern() appends to @nd_queue, whereas net_client_parse()
1902  * appends to @qemu_netdev_opts.
1903  */
1904 void netdev_parse_modern(const char *optstr)
1905 {
1906     Visitor *v;
1907     NetdevQueueEntry *nd;
1908
1909     v = qobject_input_visitor_new_str(optstr, "type", &error_fatal);
1910     nd = g_new(NetdevQueueEntry, 1);
1911     visit_type_Netdev(v, NULL, &nd->nd, &error_fatal);
1912     visit_free(v);
1913     loc_save(&nd->loc);
1914
1915     QSIMPLEQ_INSERT_TAIL(&nd_queue, nd, entry);
1916 }
1917
1918 void net_client_parse(QemuOptsList *opts_list, const char *optstr)
1919 {
1920     if (!qemu_opts_parse_noisily(opts_list, optstr, true)) {
1921         exit(1);
1922     }
1923 }
1924
1925 /* From FreeBSD */
1926 /* XXX: optimize */
1927 uint32_t net_crc32(const uint8_t *p, int len)
1928 {
1929     uint32_t crc;
1930     int carry, i, j;
1931     uint8_t b;
1932
1933     crc = 0xffffffff;
1934     for (i = 0; i < len; i++) {
1935         b = *p++;
1936         for (j = 0; j < 8; j++) {
1937             carry = ((crc & 0x80000000L) ? 1 : 0) ^ (b & 0x01);
1938             crc <<= 1;
1939             b >>= 1;
1940             if (carry) {
1941                 crc = ((crc ^ POLYNOMIAL_BE) | carry);
1942             }
1943         }
1944     }
1945
1946     return crc;
1947 }
1948
1949 uint32_t net_crc32_le(const uint8_t *p, int len)
1950 {
1951     uint32_t crc;
1952     int carry, i, j;
1953     uint8_t b;
1954
1955     crc = 0xffffffff;
1956     for (i = 0; i < len; i++) {
1957         b = *p++;
1958         for (j = 0; j < 8; j++) {
1959             carry = (crc & 0x1) ^ (b & 0x01);
1960             crc >>= 1;
1961             b >>= 1;
1962             if (carry) {
1963                 crc ^= POLYNOMIAL_LE;
1964             }
1965         }
1966     }
1967
1968     return crc;
1969 }
1970
1971 QemuOptsList qemu_netdev_opts = {
1972     .name = "netdev",
1973     .implied_opt_name = "type",
1974     .head = QTAILQ_HEAD_INITIALIZER(qemu_netdev_opts.head),
1975     .desc = {
1976         /*
1977          * no elements => accept any params
1978          * validation will happen later
1979          */
1980         { /* end of list */ }
1981     },
1982 };
1983
1984 QemuOptsList qemu_nic_opts = {
1985     .name = "nic",
1986     .implied_opt_name = "type",
1987     .head = QTAILQ_HEAD_INITIALIZER(qemu_nic_opts.head),
1988     .desc = {
1989         /*
1990          * no elements => accept any params
1991          * validation will happen later
1992          */
1993         { /* end of list */ }
1994     },
1995 };
1996
1997 QemuOptsList qemu_net_opts = {
1998     .name = "net",
1999     .implied_opt_name = "type",
2000     .head = QTAILQ_HEAD_INITIALIZER(qemu_net_opts.head),
2001     .desc = {
2002         /*
2003          * no elements => accept any params
2004          * validation will happen later
2005          */
2006         { /* end of list */ }
2007     },
2008 };
2009
2010 void net_socket_rs_init(SocketReadState *rs,
2011                         SocketReadStateFinalize *finalize,
2012                         bool vnet_hdr)
2013 {
2014     rs->state = 0;
2015     rs->vnet_hdr = vnet_hdr;
2016     rs->index = 0;
2017     rs->packet_len = 0;
2018     rs->vnet_hdr_len = 0;
2019     memset(rs->buf, 0, sizeof(rs->buf));
2020     rs->finalize = finalize;
2021 }
2022
2023 /*
2024  * Returns
2025  * 0: success
2026  * -1: error occurs
2027  */
2028 int net_fill_rstate(SocketReadState *rs, const uint8_t *buf, int size)
2029 {
2030     unsigned int l;
2031
2032     while (size > 0) {
2033         /* Reassemble a packet from the network.
2034          * 0 = getting length.
2035          * 1 = getting vnet header length.
2036          * 2 = getting data.
2037          */
2038         switch (rs->state) {
2039         case 0:
2040             l = 4 - rs->index;
2041             if (l > size) {
2042                 l = size;
2043             }
2044             memcpy(rs->buf + rs->index, buf, l);
2045             buf += l;
2046             size -= l;
2047             rs->index += l;
2048             if (rs->index == 4) {
2049                 /* got length */
2050                 rs->packet_len = ntohl(*(uint32_t *)rs->buf);
2051                 rs->index = 0;
2052                 if (rs->vnet_hdr) {
2053                     rs->state = 1;
2054                 } else {
2055                     rs->state = 2;
2056                     rs->vnet_hdr_len = 0;
2057                 }
2058             }
2059             break;
2060         case 1:
2061             l = 4 - rs->index;
2062             if (l > size) {
2063                 l = size;
2064             }
2065             memcpy(rs->buf + rs->index, buf, l);
2066             buf += l;
2067             size -= l;
2068             rs->index += l;
2069             if (rs->index == 4) {
2070                 /* got vnet header length */
2071                 rs->vnet_hdr_len = ntohl(*(uint32_t *)rs->buf);
2072                 rs->index = 0;
2073                 rs->state = 2;
2074             }
2075             break;
2076         case 2:
2077             l = rs->packet_len - rs->index;
2078             if (l > size) {
2079                 l = size;
2080             }
2081             if (rs->index + l <= sizeof(rs->buf)) {
2082                 memcpy(rs->buf + rs->index, buf, l);
2083             } else {
2084                 fprintf(stderr, "serious error: oversized packet received,"
2085                     "connection terminated.\n");
2086                 rs->index = rs->state = 0;
2087                 return -1;
2088             }
2089
2090             rs->index += l;
2091             buf += l;
2092             size -= l;
2093             if (rs->index >= rs->packet_len) {
2094                 rs->index = 0;
2095                 rs->state = 0;
2096                 assert(rs->finalize);
2097                 rs->finalize(rs);
2098             }
2099             break;
2100         }
2101     }
2102
2103     assert(size == 0);
2104     return 0;
2105 }