net/tap.c

   1 /*
   2  * QEMU System Emulator
   3  *
   4  * Copyright (c) 2003-2008 Fabrice Bellard
   5  * Copyright (c) 2009 Red Hat, Inc.
   6  *
   7  * Permission is hereby granted, free of charge, to any person obtaining a copy
   8  * of this software and associated documentation files (the "Software"), to deal
   9  * in the Software without restriction, including without limitation the rights
  10  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  11  * copies of the Software, and to permit persons to whom the Software is
  12  * furnished to do so, subject to the following conditions:
  13  *
  14  * The above copyright notice and this permission notice shall be included in
  15  * all copies or substantial portions of the Software.
  16  *
  17  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  18  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  19  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  20  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  21  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  22  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  23  * THE SOFTWARE.
  24  */
  25
  26 #include "tap_int.h"
  27
  28 #include "config-host.h"
  29
  30 #include <sys/ioctl.h>
  31 #include <sys/stat.h>
  32 #include <sys/wait.h>
  33 #include <sys/socket.h>
  34 #include <net/if.h>
  35
  36 #include "net/net.h"
  37 #include "clients.h"
  38 #include "monitor/monitor.h"
  39 #include "sysemu/sysemu.h"
  40 #include "qemu-common.h"
  41 #include "qemu/error-report.h"
  42
  43 #include "net/tap.h"
  44
  45 #include "net/vhost_net.h"
  46
  47 typedef struct TAPState {
  48     NetClientState nc;
  49     int fd;
  50     char down_script[1024];
  51     char down_script_arg[128];
  52     uint8_t buf[NET_BUFSIZE];
  53     bool read_poll;
  54     bool write_poll;
  55     bool using_vnet_hdr;
  56     bool has_ufo;
  57     bool enabled;
  58     VHostNetState *vhost_net;
  59     unsigned host_vnet_hdr_len;
  60 } TAPState;
  61
  62 static int launch_script(const char *setup_script, const char *ifname, int fd);
  63
  64 static int tap_can_send(void *opaque);
  65 static void tap_send(void *opaque);
  66 static void tap_writable(void *opaque);
  67
  68 static void tap_update_fd_handler(TAPState *s)
  69 {
  70     qemu_set_fd_handler2(s->fd,
  71                          s->read_poll && s->enabled ? tap_can_send : NULL,
  72                          s->read_poll && s->enabled ? tap_send     : NULL,
  73                          s->write_poll && s->enabled ? tap_writable : NULL,
  74                          s);
  75 }
  76
  77 static void tap_read_poll(TAPState *s, bool enable)
  78 {
  79     s->read_poll = enable;
  80     tap_update_fd_handler(s);
  81 }
  82
  83 static void tap_write_poll(TAPState *s, bool enable)
  84 {
  85     s->write_poll = enable;
  86     tap_update_fd_handler(s);
  87 }
  88
  89 static void tap_writable(void *opaque)
  90 {
  91     TAPState *s = opaque;
  92
  93     tap_write_poll(s, false);
  94
  95     qemu_flush_queued_packets(&s->nc);
  96 }
  97
  98 static ssize_t tap_write_packet(TAPState *s, const struct iovec *iov, int iovcnt)
  99 {
 100     ssize_t len;
 101
 102     do {
 103         len = writev(s->fd, iov, iovcnt);
 104     } while (len == -1 && errno == EINTR);
 105
 106     if (len == -1 && errno == EAGAIN) {
 107         tap_write_poll(s, true);
 108         return 0;
 109     }
 110
 111     return len;
 112 }
 113
 114 static ssize_t tap_receive_iov(NetClientState *nc, const struct iovec *iov,
 115                                int iovcnt)
 116 {
 117     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 118     const struct iovec *iovp = iov;
 119     struct iovec iov_copy[iovcnt + 1];
 120     struct virtio_net_hdr_mrg_rxbuf hdr = { };
 121
 122     if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
 123         iov_copy[0].iov_base = &hdr;
 124         iov_copy[0].iov_len =  s->host_vnet_hdr_len;
 125         memcpy(&iov_copy[1], iov, iovcnt * sizeof(*iov));
 126         iovp = iov_copy;
 127         iovcnt++;
 128     }
 129
 130     return tap_write_packet(s, iovp, iovcnt);
 131 }
 132
 133 static ssize_t tap_receive_raw(NetClientState *nc, const uint8_t *buf, size_t size)
 134 {
 135     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 136     struct iovec iov[2];
 137     int iovcnt = 0;
 138     struct virtio_net_hdr_mrg_rxbuf hdr = { };
 139
 140     if (s->host_vnet_hdr_len) {
 141         iov[iovcnt].iov_base = &hdr;
 142         iov[iovcnt].iov_len  = s->host_vnet_hdr_len;
 143         iovcnt++;
 144     }
 145
 146     iov[iovcnt].iov_base = (char *)buf;
 147     iov[iovcnt].iov_len  = size;
 148     iovcnt++;
 149
 150     return tap_write_packet(s, iov, iovcnt);
 151 }
 152
 153 static ssize_t tap_receive(NetClientState *nc, const uint8_t *buf, size_t size)
 154 {
 155     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 156     struct iovec iov[1];
 157
 158     if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
 159         return tap_receive_raw(nc, buf, size);
 160     }
 161
 162     iov[0].iov_base = (char *)buf;
 163     iov[0].iov_len  = size;
 164
 165     return tap_write_packet(s, iov, 1);
 166 }
 167
 168 static int tap_can_send(void *opaque)
 169 {
 170     TAPState *s = opaque;
 171
 172     return qemu_can_send_packet(&s->nc);
 173 }
 174
 175 #ifndef __sun__
 176 ssize_t tap_read_packet(int tapfd, uint8_t *buf, int maxlen)
 177 {
 178     return read(tapfd, buf, maxlen);
 179 }
 180 #endif
 181
 182 static void tap_send_completed(NetClientState *nc, ssize_t len)
 183 {
 184     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 185     tap_read_poll(s, true);
 186 }
 187
 188 static void tap_send(void *opaque)
 189 {
 190     TAPState *s = opaque;
 191     int size;
 192     int packets = 0;
 193
 194     while (qemu_can_send_packet(&s->nc)) {
 195         uint8_t *buf = s->buf;
 196
 197         size = tap_read_packet(s->fd, s->buf, sizeof(s->buf));
 198         if (size <= 0) {
 199             break;
 200         }
 201
 202         if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
 203             buf  += s->host_vnet_hdr_len;
 204             size -= s->host_vnet_hdr_len;
 205         }
 206
 207         size = qemu_send_packet_async(&s->nc, buf, size, tap_send_completed);
 208         if (size == 0) {
 209             tap_read_poll(s, false);
 210             break;
 211         } else if (size < 0) {
 212             break;
 213         }
 214
 215         /*
 216          * When the host keeps receiving more packets while tap_send() is
 217          * running we can hog the QEMU global mutex.  Limit the number of
 218          * packets that are processed per tap_send() callback to prevent
 219          * stalling the guest.
 220          */
 221         packets++;
 222         if (packets >= 50) {
 223             break;
 224         }
 225     }
 226 }
 227
 228 static bool tap_has_ufo(NetClientState *nc)
 229 {
 230     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 231
 232     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 233
 234     return s->has_ufo;
 235 }
 236
 237 static bool tap_has_vnet_hdr(NetClientState *nc)
 238 {
 239     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 240
 241     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 242
 243     return !!s->host_vnet_hdr_len;
 244 }
 245
 246 static bool tap_has_vnet_hdr_len(NetClientState *nc, int len)
 247 {
 248     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 249
 250     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 251
 252     return !!tap_probe_vnet_hdr_len(s->fd, len);
 253 }
 254
 255 static void tap_set_vnet_hdr_len(NetClientState *nc, int len)
 256 {
 257     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 258
 259     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 260     assert(len == sizeof(struct virtio_net_hdr_mrg_rxbuf) ||
 261            len == sizeof(struct virtio_net_hdr));
 262
 263     tap_fd_set_vnet_hdr_len(s->fd, len);
 264     s->host_vnet_hdr_len = len;
 265 }
 266
 267 static void tap_using_vnet_hdr(NetClientState *nc, bool using_vnet_hdr)
 268 {
 269     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 270
 271     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 272     assert(!!s->host_vnet_hdr_len == using_vnet_hdr);
 273
 274     s->using_vnet_hdr = using_vnet_hdr;
 275 }
 276
 277 static void tap_set_offload(NetClientState *nc, int csum, int tso4,
 278                      int tso6, int ecn, int ufo)
 279 {
 280     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 281     if (s->fd < 0) {
 282         return;
 283     }
 284
 285     tap_fd_set_offload(s->fd, csum, tso4, tso6, ecn, ufo);
 286 }
 287
 288 static void tap_cleanup(NetClientState *nc)
 289 {
 290     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 291
 292     if (s->vhost_net) {
 293         vhost_net_cleanup(s->vhost_net);
 294         s->vhost_net = NULL;
 295     }
 296
 297     qemu_purge_queued_packets(nc);
 298
 299     if (s->down_script[0])
 300         launch_script(s->down_script, s->down_script_arg, s->fd);
 301
 302     tap_read_poll(s, false);
 303     tap_write_poll(s, false);
 304     close(s->fd);
 305     s->fd = -1;
 306 }
 307
 308 static void tap_poll(NetClientState *nc, bool enable)
 309 {
 310     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 311     tap_read_poll(s, enable);
 312     tap_write_poll(s, enable);
 313 }
 314
 315 int tap_get_fd(NetClientState *nc)
 316 {
 317     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 318     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 319     return s->fd;
 320 }
 321
 322 /* fd support */
 323
 324 static NetClientInfo net_tap_info = {
 325     .type = NET_CLIENT_OPTIONS_KIND_TAP,
 326     .size = sizeof(TAPState),
 327     .receive = tap_receive,
 328     .receive_raw = tap_receive_raw,
 329     .receive_iov = tap_receive_iov,
 330     .poll = tap_poll,
 331     .cleanup = tap_cleanup,
 332     .has_ufo = tap_has_ufo,
 333     .has_vnet_hdr = tap_has_vnet_hdr,
 334     .has_vnet_hdr_len = tap_has_vnet_hdr_len,
 335     .using_vnet_hdr = tap_using_vnet_hdr,
 336     .set_offload = tap_set_offload,
 337     .set_vnet_hdr_len = tap_set_vnet_hdr_len,
 338 };
 339
 340 static TAPState *net_tap_fd_init(NetClientState *peer,
 341                                  const char *model,
 342                                  const char *name,
 343                                  int fd,
 344                                  int vnet_hdr)
 345 {
 346     NetClientState *nc;
 347     TAPState *s;
 348
 349     nc = qemu_new_net_client(&net_tap_info, peer, model, name);
 350
 351     s = DO_UPCAST(TAPState, nc, nc);
 352
 353     s->fd = fd;
 354     s->host_vnet_hdr_len = vnet_hdr ? sizeof(struct virtio_net_hdr) : 0;
 355     s->using_vnet_hdr = false;
 356     s->has_ufo = tap_probe_has_ufo(s->fd);
 357     s->enabled = true;
 358     tap_set_offload(&s->nc, 0, 0, 0, 0, 0);
 359     /*
 360      * Make sure host header length is set correctly in tap:
 361      * it might have been modified by another instance of qemu.
 362      */
 363     if (tap_probe_vnet_hdr_len(s->fd, s->host_vnet_hdr_len)) {
 364         tap_fd_set_vnet_hdr_len(s->fd, s->host_vnet_hdr_len);
 365     }
 366     tap_read_poll(s, true);
 367     s->vhost_net = NULL;
 368     return s;
 369 }
 370
 371 static int launch_script(const char *setup_script, const char *ifname, int fd)
 372 {
 373     int pid, status;
 374     char *args[3];
 375     char **parg;
 376
 377     /* try to launch network script */
 378     pid = fork();
 379     if (pid == 0) {
 380         int open_max = sysconf(_SC_OPEN_MAX), i;
 381
 382         for (i = 3; i < open_max; i++) {
 383             if (i != fd) {
 384                 close(i);
 385             }
 386         }
 387         parg = args;
 388         *parg++ = (char *)setup_script;
 389         *parg++ = (char *)ifname;
 390         *parg = NULL;
 391         execv(setup_script, args);
 392         _exit(1);
 393     } else if (pid > 0) {
 394         while (waitpid(pid, &status, 0) != pid) {
 395             /* loop */
 396         }
 397
 398         if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
 399             return 0;
 400         }
 401     }
 402     fprintf(stderr, "%s: could not launch network script\n", setup_script);
 403     return -1;
 404 }
 405
 406 static int recv_fd(int c)
 407 {
 408     int fd;
 409     uint8_t msgbuf[CMSG_SPACE(sizeof(fd))];
 410     struct msghdr msg = {
 411         .msg_control = msgbuf,
 412         .msg_controllen = sizeof(msgbuf),
 413     };
 414     struct cmsghdr *cmsg;
 415     struct iovec iov;
 416     uint8_t req[1];
 417     ssize_t len;
 418
 419     cmsg = CMSG_FIRSTHDR(&msg);
 420     cmsg->cmsg_level = SOL_SOCKET;
 421     cmsg->cmsg_type = SCM_RIGHTS;
 422     cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
 423     msg.msg_controllen = cmsg->cmsg_len;
 424
 425     iov.iov_base = req;
 426     iov.iov_len = sizeof(req);
 427
 428     msg.msg_iov = &iov;
 429     msg.msg_iovlen = 1;
 430
 431     len = recvmsg(c, &msg, 0);
 432     if (len > 0) {
 433         memcpy(&fd, CMSG_DATA(cmsg), sizeof(fd));
 434         return fd;
 435     }
 436
 437     return len;
 438 }
 439
 440 static int net_bridge_run_helper(const char *helper, const char *bridge)
 441 {
 442     sigset_t oldmask, mask;
 443     int pid, status;
 444     char *args[5];
 445     char **parg;
 446     int sv[2];
 447
 448     sigemptyset(&mask);
 449     sigaddset(&mask, SIGCHLD);
 450     sigprocmask(SIG_BLOCK, &mask, &oldmask);
 451
 452     if (socketpair(PF_UNIX, SOCK_STREAM, 0, sv) == -1) {
 453         return -1;
 454     }
 455
 456     /* try to launch bridge helper */
 457     pid = fork();
 458     if (pid == 0) {
 459         int open_max = sysconf(_SC_OPEN_MAX), i;
 460         char fd_buf[6+10];
 461         char br_buf[6+IFNAMSIZ] = {0};
 462         char helper_cmd[PATH_MAX + sizeof(fd_buf) + sizeof(br_buf) + 15];
 463
 464         for (i = 3; i < open_max; i++) {
 465             if (i != sv[1]) {
 466                 close(i);
 467             }
 468         }
 469
 470         snprintf(fd_buf, sizeof(fd_buf), "%s%d", "--fd=", sv[1]);
 471
 472         if (strrchr(helper, ' ') || strrchr(helper, '\t')) {
 473             /* assume helper is a command */
 474
 475             if (strstr(helper, "--br=") == NULL) {
 476                 snprintf(br_buf, sizeof(br_buf), "%s%s", "--br=", bridge);
 477             }
 478
 479             snprintf(helper_cmd, sizeof(helper_cmd), "%s %s %s %s",
 480                      helper, "--use-vnet", fd_buf, br_buf);
 481
 482             parg = args;
 483             *parg++ = (char *)"sh";
 484             *parg++ = (char *)"-c";
 485             *parg++ = helper_cmd;
 486             *parg++ = NULL;
 487
 488             execv("/bin/sh", args);
 489         } else {
 490             /* assume helper is just the executable path name */
 491
 492             snprintf(br_buf, sizeof(br_buf), "%s%s", "--br=", bridge);
 493
 494             parg = args;
 495             *parg++ = (char *)helper;
 496             *parg++ = (char *)"--use-vnet";
 497             *parg++ = fd_buf;
 498             *parg++ = br_buf;
 499             *parg++ = NULL;
 500
 501             execv(helper, args);
 502         }
 503         _exit(1);
 504
 505     } else if (pid > 0) {
 506         int fd;
 507
 508         close(sv[1]);
 509
 510         do {
 511             fd = recv_fd(sv[0]);
 512         } while (fd == -1 && errno == EINTR);
 513
 514         close(sv[0]);
 515
 516         while (waitpid(pid, &status, 0) != pid) {
 517             /* loop */
 518         }
 519         sigprocmask(SIG_SETMASK, &oldmask, NULL);
 520         if (fd < 0) {
 521             fprintf(stderr, "failed to recv file descriptor\n");
 522             return -1;
 523         }
 524
 525         if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
 526             return fd;
 527         }
 528     }
 529     fprintf(stderr, "failed to launch bridge helper\n");
 530     return -1;
 531 }
 532
 533 int net_init_bridge(const NetClientOptions *opts, const char *name,
 534                     NetClientState *peer)
 535 {
 536     const NetdevBridgeOptions *bridge;
 537     const char *helper, *br;
 538
 539     TAPState *s;
 540     int fd, vnet_hdr;
 541
 542     assert(opts->kind == NET_CLIENT_OPTIONS_KIND_BRIDGE);
 543     bridge = opts->bridge;
 544
 545     helper = bridge->has_helper ? bridge->helper : DEFAULT_BRIDGE_HELPER;
 546     br     = bridge->has_br     ? bridge->br     : DEFAULT_BRIDGE_INTERFACE;
 547
 548     fd = net_bridge_run_helper(helper, br);
 549     if (fd == -1) {
 550         return -1;
 551     }
 552
 553     fcntl(fd, F_SETFL, O_NONBLOCK);
 554
 555     vnet_hdr = tap_probe_vnet_hdr(fd);
 556
 557     s = net_tap_fd_init(peer, "bridge", name, fd, vnet_hdr);
 558     if (!s) {
 559         close(fd);
 560         return -1;
 561     }
 562
 563     snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s,br=%s", helper,
 564              br);
 565
 566     return 0;
 567 }
 568
 569 static int net_tap_init(const NetdevTapOptions *tap, int *vnet_hdr,
 570                         const char *setup_script, char *ifname,
 571                         size_t ifname_sz, int mq_required)
 572 {
 573     int fd, vnet_hdr_required;
 574
 575     if (tap->has_vnet_hdr) {
 576         *vnet_hdr = tap->vnet_hdr;
 577         vnet_hdr_required = *vnet_hdr;
 578     } else {
 579         *vnet_hdr = 1;
 580         vnet_hdr_required = 0;
 581     }
 582
 583     TFR(fd = tap_open(ifname, ifname_sz, vnet_hdr, vnet_hdr_required,
 584                       mq_required));
 585     if (fd < 0) {
 586         return -1;
 587     }
 588
 589     if (setup_script &&
 590         setup_script[0] != '\0' &&
 591         strcmp(setup_script, "no") != 0 &&
 592         launch_script(setup_script, ifname, fd)) {
 593         close(fd);
 594         return -1;
 595     }
 596
 597     return fd;
 598 }
 599
 600 #define MAX_TAP_QUEUES 1024
 601
 602 static int net_init_tap_one(const NetdevTapOptions *tap, NetClientState *peer,
 603                             const char *model, const char *name,
 604                             const char *ifname, const char *script,
 605                             const char *downscript, const char *vhostfdname,
 606                             int vnet_hdr, int fd)
 607 {
 608     TAPState *s;
 609     int vhostfd;
 610
 611     s = net_tap_fd_init(peer, model, name, fd, vnet_hdr);
 612     if (!s) {
 613         return -1;
 614     }
 615
 616     if (tap_set_sndbuf(s->fd, tap) < 0) {
 617         return -1;
 618     }
 619
 620     if (tap->has_fd || tap->has_fds) {
 621         snprintf(s->nc.info_str, sizeof(s->nc.info_str), "fd=%d", fd);
 622     } else if (tap->has_helper) {
 623         snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s",
 624                  tap->helper);
 625     } else {
 626         snprintf(s->nc.info_str, sizeof(s->nc.info_str),
 627                  "ifname=%s,script=%s,downscript=%s", ifname, script,
 628                  downscript);
 629
 630         if (strcmp(downscript, "no") != 0) {
 631             snprintf(s->down_script, sizeof(s->down_script), "%s", downscript);
 632             snprintf(s->down_script_arg, sizeof(s->down_script_arg),
 633                      "%s", ifname);
 634         }
 635     }
 636
 637     if (tap->has_vhost ? tap->vhost :
 638         vhostfdname || (tap->has_vhostforce && tap->vhostforce)) {
 639         VhostNetOptions options;
 640
 641         options.backend_type = VHOST_BACKEND_TYPE_KERNEL;
 642         options.net_backend = &s->nc;
 643         options.force = tap->has_vhostforce && tap->vhostforce;
 644
 645         if (tap->has_vhostfd || tap->has_vhostfds) {
 646             vhostfd = monitor_handle_fd_param(cur_mon, vhostfdname);
 647             if (vhostfd == -1) {
 648                 return -1;
 649             }
 650         } else {
 651             vhostfd = open("/dev/vhost-net", O_RDWR);
 652             if (vhostfd < 0) {
 653                 error_report("tap: open vhost char device failed: %s",
 654                            strerror(errno));
 655                 return -1;
 656             }
 657         }
 658         options.opaque = (void *)(uintptr_t)vhostfd;
 659
 660         s->vhost_net = vhost_net_init(&options);
 661         if (!s->vhost_net) {
 662             error_report("vhost-net requested but could not be initialized");
 663             return -1;
 664         }
 665     } else if (tap->has_vhostfd || tap->has_vhostfds) {
 666         error_report("vhostfd= is not valid without vhost");
 667         return -1;
 668     }
 669
 670     return 0;
 671 }
 672
 673 static int get_fds(char *str, char *fds[], int max)
 674 {
 675     char *ptr = str, *this;
 676     size_t len = strlen(str);
 677     int i = 0;
 678
 679     while (i < max && ptr < str + len) {
 680         this = strchr(ptr, ':');
 681
 682         if (this == NULL) {
 683             fds[i] = g_strdup(ptr);
 684         } else {
 685             fds[i] = g_strndup(ptr, this - ptr);
 686         }
 687
 688         i++;
 689         if (this == NULL) {
 690             break;
 691         } else {
 692             ptr = this + 1;
 693         }
 694     }
 695
 696     return i;
 697 }
 698
 699 int net_init_tap(const NetClientOptions *opts, const char *name,
 700                  NetClientState *peer)
 701 {
 702     const NetdevTapOptions *tap;
 703     int fd, vnet_hdr = 0, i = 0, queues;
 704     /* for the no-fd, no-helper case */
 705     const char *script = NULL; /* suppress wrong "uninit'd use" gcc warning */
 706     const char *downscript = NULL;
 707     const char *vhostfdname;
 708     char ifname[128];
 709
 710     assert(opts->kind == NET_CLIENT_OPTIONS_KIND_TAP);
 711     tap = opts->tap;
 712     queues = tap->has_queues ? tap->queues : 1;
 713     vhostfdname = tap->has_vhostfd ? tap->vhostfd : NULL;
 714
 715     /* QEMU vlans does not support multiqueue tap, in this case peer is set.
 716      * For -netdev, peer is always NULL. */
 717     if (peer && (tap->has_queues || tap->has_fds || tap->has_vhostfds)) {
 718         error_report("Multiqueue tap cannot be used with QEMU vlans");
 719         return -1;
 720     }
 721
 722     if (tap->has_fd) {
 723         if (tap->has_ifname || tap->has_script || tap->has_downscript ||
 724             tap->has_vnet_hdr || tap->has_helper || tap->has_queues ||
 725             tap->has_fds || tap->has_vhostfds) {
 726             error_report("ifname=, script=, downscript=, vnet_hdr=, "
 727                          "helper=, queues=, fds=, and vhostfds= "
 728                          "are invalid with fd=");
 729             return -1;
 730         }
 731
 732         fd = monitor_handle_fd_param(cur_mon, tap->fd);
 733         if (fd == -1) {
 734             return -1;
 735         }
 736
 737         fcntl(fd, F_SETFL, O_NONBLOCK);
 738
 739         vnet_hdr = tap_probe_vnet_hdr(fd);
 740
 741         if (net_init_tap_one(tap, peer, "tap", name, NULL,
 742                              script, downscript,
 743                              vhostfdname, vnet_hdr, fd)) {
 744             return -1;
 745         }
 746     } else if (tap->has_fds) {
 747         char *fds[MAX_TAP_QUEUES];
 748         char *vhost_fds[MAX_TAP_QUEUES];
 749         int nfds, nvhosts;
 750
 751         if (tap->has_ifname || tap->has_script || tap->has_downscript ||
 752             tap->has_vnet_hdr || tap->has_helper || tap->has_queues ||
 753             tap->has_vhostfd) {
 754             error_report("ifname=, script=, downscript=, vnet_hdr=, "
 755                          "helper=, queues=, and vhostfd= "
 756                          "are invalid with fds=");
 757             return -1;
 758         }
 759
 760         nfds = get_fds(tap->fds, fds, MAX_TAP_QUEUES);
 761         if (tap->has_vhostfds) {
 762             nvhosts = get_fds(tap->vhostfds, vhost_fds, MAX_TAP_QUEUES);
 763             if (nfds != nvhosts) {
 764                 error_report("The number of fds passed does not match the "
 765                              "number of vhostfds passed");
 766                 return -1;
 767             }
 768         }
 769
 770         for (i = 0; i < nfds; i++) {
 771             fd = monitor_handle_fd_param(cur_mon, fds[i]);
 772             if (fd == -1) {
 773                 return -1;
 774             }
 775
 776             fcntl(fd, F_SETFL, O_NONBLOCK);
 777
 778             if (i == 0) {
 779                 vnet_hdr = tap_probe_vnet_hdr(fd);
 780             } else if (vnet_hdr != tap_probe_vnet_hdr(fd)) {
 781                 error_report("vnet_hdr not consistent across given tap fds");
 782                 return -1;
 783             }
 784
 785             if (net_init_tap_one(tap, peer, "tap", name, ifname,
 786                                  script, downscript,
 787                                  tap->has_vhostfds ? vhost_fds[i] : NULL,
 788                                  vnet_hdr, fd)) {
 789                 return -1;
 790             }
 791         }
 792     } else if (tap->has_helper) {
 793         if (tap->has_ifname || tap->has_script || tap->has_downscript ||
 794             tap->has_vnet_hdr || tap->has_queues || tap->has_vhostfds) {
 795             error_report("ifname=, script=, downscript=, and vnet_hdr= "
 796                          "queues=, and vhostfds= are invalid with helper=");
 797             return -1;
 798         }
 799
 800         fd = net_bridge_run_helper(tap->helper, DEFAULT_BRIDGE_INTERFACE);
 801         if (fd == -1) {
 802             return -1;
 803         }
 804
 805         fcntl(fd, F_SETFL, O_NONBLOCK);
 806         vnet_hdr = tap_probe_vnet_hdr(fd);
 807
 808         if (net_init_tap_one(tap, peer, "bridge", name, ifname,
 809                              script, downscript, vhostfdname,
 810                              vnet_hdr, fd)) {
 811             close(fd);
 812             return -1;
 813         }
 814     } else {
 815         if (tap->has_vhostfds) {
 816             error_report("vhostfds= is invalid if fds= wasn't specified");
 817             return -1;
 818         }
 819         script = tap->has_script ? tap->script : DEFAULT_NETWORK_SCRIPT;
 820         downscript = tap->has_downscript ? tap->downscript :
 821             DEFAULT_NETWORK_DOWN_SCRIPT;
 822
 823         if (tap->has_ifname) {
 824             pstrcpy(ifname, sizeof ifname, tap->ifname);
 825         } else {
 826             ifname[0] = '\0';
 827         }
 828
 829         for (i = 0; i < queues; i++) {
 830             fd = net_tap_init(tap, &vnet_hdr, i >= 1 ? "no" : script,
 831                               ifname, sizeof ifname, queues > 1);
 832             if (fd == -1) {
 833                 return -1;
 834             }
 835
 836             if (queues > 1 && i == 0 && !tap->has_ifname) {
 837                 if (tap_fd_get_ifname(fd, ifname)) {
 838                     error_report("Fail to get ifname");
 839                     close(fd);
 840                     return -1;
 841                 }
 842             }
 843
 844             if (net_init_tap_one(tap, peer, "tap", name, ifname,
 845                                  i >= 1 ? "no" : script,
 846                                  i >= 1 ? "no" : downscript,
 847                                  vhostfdname, vnet_hdr, fd)) {
 848                 close(fd);
 849                 return -1;
 850             }
 851         }
 852     }
 853
 854     return 0;
 855 }
 856
 857 VHostNetState *tap_get_vhost_net(NetClientState *nc)
 858 {
 859     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 860     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 861     return s->vhost_net;
 862 }
 863
 864 int tap_enable(NetClientState *nc)
 865 {
 866     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 867     int ret;
 868
 869     if (s->enabled) {
 870         return 0;
 871     } else {
 872         ret = tap_fd_enable(s->fd);
 873         if (ret == 0) {
 874             s->enabled = true;
 875             tap_update_fd_handler(s);
 876         }
 877         return ret;
 878     }
 879 }
 880
 881 int tap_disable(NetClientState *nc)
 882 {
 883     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 884     int ret;
 885
 886     if (s->enabled == 0) {
 887         return 0;
 888     } else {
 889         ret = tap_fd_disable(s->fd);
 890         if (ret == 0) {
 891             qemu_purge_queued_packets(nc);
 892             s->enabled = false;
 893             tap_update_fd_handler(s);
 894         }
 895         return ret;
 896     }
 897 }