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
 193     while (qemu_can_send_packet(&s->nc)) {
 194         uint8_t *buf = s->buf;
 195
 196         size = tap_read_packet(s->fd, s->buf, sizeof(s->buf));
 197         if (size <= 0) {
 198             break;
 199         }
 200
 201         if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
 202             buf  += s->host_vnet_hdr_len;
 203             size -= s->host_vnet_hdr_len;
 204         }
 205
 206         size = qemu_send_packet_async(&s->nc, buf, size, tap_send_completed);
 207         if (size == 0) {
 208             tap_read_poll(s, false);
 209             break;
 210         } else if (size < 0) {
 211             break;
 212         }
 213     }
 214 }
 215
 216 static bool tap_has_ufo(NetClientState *nc)
 217 {
 218     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 219
 220     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 221
 222     return s->has_ufo;
 223 }
 224
 225 static bool tap_has_vnet_hdr(NetClientState *nc)
 226 {
 227     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 228
 229     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 230
 231     return !!s->host_vnet_hdr_len;
 232 }
 233
 234 static bool tap_has_vnet_hdr_len(NetClientState *nc, int len)
 235 {
 236     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 237
 238     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 239
 240     return !!tap_probe_vnet_hdr_len(s->fd, len);
 241 }
 242
 243 static void tap_set_vnet_hdr_len(NetClientState *nc, int len)
 244 {
 245     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 246
 247     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 248     assert(len == sizeof(struct virtio_net_hdr_mrg_rxbuf) ||
 249            len == sizeof(struct virtio_net_hdr));
 250
 251     tap_fd_set_vnet_hdr_len(s->fd, len);
 252     s->host_vnet_hdr_len = len;
 253 }
 254
 255 static void tap_using_vnet_hdr(NetClientState *nc, bool using_vnet_hdr)
 256 {
 257     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 258
 259     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 260     assert(!!s->host_vnet_hdr_len == using_vnet_hdr);
 261
 262     s->using_vnet_hdr = using_vnet_hdr;
 263 }
 264
 265 static void tap_set_offload(NetClientState *nc, int csum, int tso4,
 266                      int tso6, int ecn, int ufo)
 267 {
 268     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 269     if (s->fd < 0) {
 270         return;
 271     }
 272
 273     tap_fd_set_offload(s->fd, csum, tso4, tso6, ecn, ufo);
 274 }
 275
 276 static void tap_cleanup(NetClientState *nc)
 277 {
 278     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 279
 280     if (s->vhost_net) {
 281         vhost_net_cleanup(s->vhost_net);
 282         s->vhost_net = NULL;
 283     }
 284
 285     qemu_purge_queued_packets(nc);
 286
 287     if (s->down_script[0])
 288         launch_script(s->down_script, s->down_script_arg, s->fd);
 289
 290     tap_read_poll(s, false);
 291     tap_write_poll(s, false);
 292     close(s->fd);
 293     s->fd = -1;
 294 }
 295
 296 static void tap_poll(NetClientState *nc, bool enable)
 297 {
 298     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 299     tap_read_poll(s, enable);
 300     tap_write_poll(s, enable);
 301 }
 302
 303 int tap_get_fd(NetClientState *nc)
 304 {
 305     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 306     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 307     return s->fd;
 308 }
 309
 310 /* fd support */
 311
 312 static NetClientInfo net_tap_info = {
 313     .type = NET_CLIENT_OPTIONS_KIND_TAP,
 314     .size = sizeof(TAPState),
 315     .receive = tap_receive,
 316     .receive_raw = tap_receive_raw,
 317     .receive_iov = tap_receive_iov,
 318     .poll = tap_poll,
 319     .cleanup = tap_cleanup,
 320     .has_ufo = tap_has_ufo,
 321     .has_vnet_hdr = tap_has_vnet_hdr,
 322     .has_vnet_hdr_len = tap_has_vnet_hdr_len,
 323     .using_vnet_hdr = tap_using_vnet_hdr,
 324     .set_offload = tap_set_offload,
 325     .set_vnet_hdr_len = tap_set_vnet_hdr_len,
 326 };
 327
 328 static TAPState *net_tap_fd_init(NetClientState *peer,
 329                                  const char *model,
 330                                  const char *name,
 331                                  int fd,
 332                                  int vnet_hdr)
 333 {
 334     NetClientState *nc;
 335     TAPState *s;
 336
 337     nc = qemu_new_net_client(&net_tap_info, peer, model, name);
 338
 339     s = DO_UPCAST(TAPState, nc, nc);
 340
 341     s->fd = fd;
 342     s->host_vnet_hdr_len = vnet_hdr ? sizeof(struct virtio_net_hdr) : 0;
 343     s->using_vnet_hdr = false;
 344     s->has_ufo = tap_probe_has_ufo(s->fd);
 345     s->enabled = true;
 346     tap_set_offload(&s->nc, 0, 0, 0, 0, 0);
 347     /*
 348      * Make sure host header length is set correctly in tap:
 349      * it might have been modified by another instance of qemu.
 350      */
 351     if (tap_probe_vnet_hdr_len(s->fd, s->host_vnet_hdr_len)) {
 352         tap_fd_set_vnet_hdr_len(s->fd, s->host_vnet_hdr_len);
 353     }
 354     tap_read_poll(s, true);
 355     s->vhost_net = NULL;
 356     return s;
 357 }
 358
 359 static int launch_script(const char *setup_script, const char *ifname, int fd)
 360 {
 361     int pid, status;
 362     char *args[3];
 363     char **parg;
 364
 365     /* try to launch network script */
 366     pid = fork();
 367     if (pid == 0) {
 368         int open_max = sysconf(_SC_OPEN_MAX), i;
 369
 370         for (i = 3; i < open_max; i++) {
 371             if (i != fd) {
 372                 close(i);
 373             }
 374         }
 375         parg = args;
 376         *parg++ = (char *)setup_script;
 377         *parg++ = (char *)ifname;
 378         *parg = NULL;
 379         execv(setup_script, args);
 380         _exit(1);
 381     } else if (pid > 0) {
 382         while (waitpid(pid, &status, 0) != pid) {
 383             /* loop */
 384         }
 385
 386         if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
 387             return 0;
 388         }
 389     }
 390     fprintf(stderr, "%s: could not launch network script\n", setup_script);
 391     return -1;
 392 }
 393
 394 static int recv_fd(int c)
 395 {
 396     int fd;
 397     uint8_t msgbuf[CMSG_SPACE(sizeof(fd))];
 398     struct msghdr msg = {
 399         .msg_control = msgbuf,
 400         .msg_controllen = sizeof(msgbuf),
 401     };
 402     struct cmsghdr *cmsg;
 403     struct iovec iov;
 404     uint8_t req[1];
 405     ssize_t len;
 406
 407     cmsg = CMSG_FIRSTHDR(&msg);
 408     cmsg->cmsg_level = SOL_SOCKET;
 409     cmsg->cmsg_type = SCM_RIGHTS;
 410     cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
 411     msg.msg_controllen = cmsg->cmsg_len;
 412
 413     iov.iov_base = req;
 414     iov.iov_len = sizeof(req);
 415
 416     msg.msg_iov = &iov;
 417     msg.msg_iovlen = 1;
 418
 419     len = recvmsg(c, &msg, 0);
 420     if (len > 0) {
 421         memcpy(&fd, CMSG_DATA(cmsg), sizeof(fd));
 422         return fd;
 423     }
 424
 425     return len;
 426 }
 427
 428 static int net_bridge_run_helper(const char *helper, const char *bridge)
 429 {
 430     sigset_t oldmask, mask;
 431     int pid, status;
 432     char *args[5];
 433     char **parg;
 434     int sv[2];
 435
 436     sigemptyset(&mask);
 437     sigaddset(&mask, SIGCHLD);
 438     sigprocmask(SIG_BLOCK, &mask, &oldmask);
 439
 440     if (socketpair(PF_UNIX, SOCK_STREAM, 0, sv) == -1) {
 441         return -1;
 442     }
 443
 444     /* try to launch bridge helper */
 445     pid = fork();
 446     if (pid == 0) {
 447         int open_max = sysconf(_SC_OPEN_MAX), i;
 448         char fd_buf[6+10];
 449         char br_buf[6+IFNAMSIZ] = {0};
 450         char helper_cmd[PATH_MAX + sizeof(fd_buf) + sizeof(br_buf) + 15];
 451
 452         for (i = 3; i < open_max; i++) {
 453             if (i != sv[1]) {
 454                 close(i);
 455             }
 456         }
 457
 458         snprintf(fd_buf, sizeof(fd_buf), "%s%d", "--fd=", sv[1]);
 459
 460         if (strrchr(helper, ' ') || strrchr(helper, '\t')) {
 461             /* assume helper is a command */
 462
 463             if (strstr(helper, "--br=") == NULL) {
 464                 snprintf(br_buf, sizeof(br_buf), "%s%s", "--br=", bridge);
 465             }
 466
 467             snprintf(helper_cmd, sizeof(helper_cmd), "%s %s %s %s",
 468                      helper, "--use-vnet", fd_buf, br_buf);
 469
 470             parg = args;
 471             *parg++ = (char *)"sh";
 472             *parg++ = (char *)"-c";
 473             *parg++ = helper_cmd;
 474             *parg++ = NULL;
 475
 476             execv("/bin/sh", args);
 477         } else {
 478             /* assume helper is just the executable path name */
 479
 480             snprintf(br_buf, sizeof(br_buf), "%s%s", "--br=", bridge);
 481
 482             parg = args;
 483             *parg++ = (char *)helper;
 484             *parg++ = (char *)"--use-vnet";
 485             *parg++ = fd_buf;
 486             *parg++ = br_buf;
 487             *parg++ = NULL;
 488
 489             execv(helper, args);
 490         }
 491         _exit(1);
 492
 493     } else if (pid > 0) {
 494         int fd;
 495
 496         close(sv[1]);
 497
 498         do {
 499             fd = recv_fd(sv[0]);
 500         } while (fd == -1 && errno == EINTR);
 501
 502         close(sv[0]);
 503
 504         while (waitpid(pid, &status, 0) != pid) {
 505             /* loop */
 506         }
 507         sigprocmask(SIG_SETMASK, &oldmask, NULL);
 508         if (fd < 0) {
 509             fprintf(stderr, "failed to recv file descriptor\n");
 510             return -1;
 511         }
 512
 513         if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
 514             return fd;
 515         }
 516     }
 517     fprintf(stderr, "failed to launch bridge helper\n");
 518     return -1;
 519 }
 520
 521 int net_init_bridge(const NetClientOptions *opts, const char *name,
 522                     NetClientState *peer)
 523 {
 524     const NetdevBridgeOptions *bridge;
 525     const char *helper, *br;
 526
 527     TAPState *s;
 528     int fd, vnet_hdr;
 529
 530     assert(opts->kind == NET_CLIENT_OPTIONS_KIND_BRIDGE);
 531     bridge = opts->bridge;
 532
 533     helper = bridge->has_helper ? bridge->helper : DEFAULT_BRIDGE_HELPER;
 534     br     = bridge->has_br     ? bridge->br     : DEFAULT_BRIDGE_INTERFACE;
 535
 536     fd = net_bridge_run_helper(helper, br);
 537     if (fd == -1) {
 538         return -1;
 539     }
 540
 541     fcntl(fd, F_SETFL, O_NONBLOCK);
 542
 543     vnet_hdr = tap_probe_vnet_hdr(fd);
 544
 545     s = net_tap_fd_init(peer, "bridge", name, fd, vnet_hdr);
 546     if (!s) {
 547         close(fd);
 548         return -1;
 549     }
 550
 551     snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s,br=%s", helper,
 552              br);
 553
 554     return 0;
 555 }
 556
 557 static int net_tap_init(const NetdevTapOptions *tap, int *vnet_hdr,
 558                         const char *setup_script, char *ifname,
 559                         size_t ifname_sz, int mq_required)
 560 {
 561     int fd, vnet_hdr_required;
 562
 563     if (tap->has_vnet_hdr) {
 564         *vnet_hdr = tap->vnet_hdr;
 565         vnet_hdr_required = *vnet_hdr;
 566     } else {
 567         *vnet_hdr = 1;
 568         vnet_hdr_required = 0;
 569     }
 570
 571     TFR(fd = tap_open(ifname, ifname_sz, vnet_hdr, vnet_hdr_required,
 572                       mq_required));
 573     if (fd < 0) {
 574         return -1;
 575     }
 576
 577     if (setup_script &&
 578         setup_script[0] != '\0' &&
 579         strcmp(setup_script, "no") != 0 &&
 580         launch_script(setup_script, ifname, fd)) {
 581         close(fd);
 582         return -1;
 583     }
 584
 585     return fd;
 586 }
 587
 588 #define MAX_TAP_QUEUES 1024
 589
 590 static int net_init_tap_one(const NetdevTapOptions *tap, NetClientState *peer,
 591                             const char *model, const char *name,
 592                             const char *ifname, const char *script,
 593                             const char *downscript, const char *vhostfdname,
 594                             int vnet_hdr, int fd)
 595 {
 596     TAPState *s;
 597     int vhostfd;
 598
 599     s = net_tap_fd_init(peer, model, name, fd, vnet_hdr);
 600     if (!s) {
 601         return -1;
 602     }
 603
 604     if (tap_set_sndbuf(s->fd, tap) < 0) {
 605         return -1;
 606     }
 607
 608     if (tap->has_fd || tap->has_fds) {
 609         snprintf(s->nc.info_str, sizeof(s->nc.info_str), "fd=%d", fd);
 610     } else if (tap->has_helper) {
 611         snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s",
 612                  tap->helper);
 613     } else {
 614         snprintf(s->nc.info_str, sizeof(s->nc.info_str),
 615                  "ifname=%s,script=%s,downscript=%s", ifname, script,
 616                  downscript);
 617
 618         if (strcmp(downscript, "no") != 0) {
 619             snprintf(s->down_script, sizeof(s->down_script), "%s", downscript);
 620             snprintf(s->down_script_arg, sizeof(s->down_script_arg),
 621                      "%s", ifname);
 622         }
 623     }
 624
 625     if (tap->has_vhost ? tap->vhost :
 626         vhostfdname || (tap->has_vhostforce && tap->vhostforce)) {
 627         VhostNetOptions options;
 628
 629         options.backend_type = VHOST_BACKEND_TYPE_KERNEL;
 630         options.net_backend = &s->nc;
 631         options.force = tap->has_vhostforce && tap->vhostforce;
 632
 633         if (tap->has_vhostfd || tap->has_vhostfds) {
 634             vhostfd = monitor_handle_fd_param(cur_mon, vhostfdname);
 635             if (vhostfd == -1) {
 636                 return -1;
 637             }
 638         } else {
 639             vhostfd = open("/dev/vhost-net", O_RDWR);
 640             if (vhostfd < 0) {
 641                 error_report("tap: open vhost char device failed: %s",
 642                            strerror(errno));
 643                 return -1;
 644             }
 645         }
 646         options.opaque = (void *)(uintptr_t)vhostfd;
 647
 648         s->vhost_net = vhost_net_init(&options);
 649         if (!s->vhost_net) {
 650             error_report("vhost-net requested but could not be initialized");
 651             return -1;
 652         }
 653     } else if (tap->has_vhostfd || tap->has_vhostfds) {
 654         error_report("vhostfd= is not valid without vhost");
 655         return -1;
 656     }
 657
 658     return 0;
 659 }
 660
 661 static int get_fds(char *str, char *fds[], int max)
 662 {
 663     char *ptr = str, *this;
 664     size_t len = strlen(str);
 665     int i = 0;
 666
 667     while (i < max && ptr < str + len) {
 668         this = strchr(ptr, ':');
 669
 670         if (this == NULL) {
 671             fds[i] = g_strdup(ptr);
 672         } else {
 673             fds[i] = g_strndup(ptr, this - ptr);
 674         }
 675
 676         i++;
 677         if (this == NULL) {
 678             break;
 679         } else {
 680             ptr = this + 1;
 681         }
 682     }
 683
 684     return i;
 685 }
 686
 687 int net_init_tap(const NetClientOptions *opts, const char *name,
 688                  NetClientState *peer)
 689 {
 690     const NetdevTapOptions *tap;
 691     int fd, vnet_hdr = 0, i = 0, queues;
 692     /* for the no-fd, no-helper case */
 693     const char *script = NULL; /* suppress wrong "uninit'd use" gcc warning */
 694     const char *downscript = NULL;
 695     const char *vhostfdname;
 696     char ifname[128];
 697
 698     assert(opts->kind == NET_CLIENT_OPTIONS_KIND_TAP);
 699     tap = opts->tap;
 700     queues = tap->has_queues ? tap->queues : 1;
 701     vhostfdname = tap->has_vhostfd ? tap->vhostfd : NULL;
 702
 703     /* QEMU vlans does not support multiqueue tap, in this case peer is set.
 704      * For -netdev, peer is always NULL. */
 705     if (peer && (tap->has_queues || tap->has_fds || tap->has_vhostfds)) {
 706         error_report("Multiqueue tap cannot be used with QEMU vlans");
 707         return -1;
 708     }
 709
 710     if (tap->has_fd) {
 711         if (tap->has_ifname || tap->has_script || tap->has_downscript ||
 712             tap->has_vnet_hdr || tap->has_helper || tap->has_queues ||
 713             tap->has_fds || tap->has_vhostfds) {
 714             error_report("ifname=, script=, downscript=, vnet_hdr=, "
 715                          "helper=, queues=, fds=, and vhostfds= "
 716                          "are invalid with fd=");
 717             return -1;
 718         }
 719
 720         fd = monitor_handle_fd_param(cur_mon, tap->fd);
 721         if (fd == -1) {
 722             return -1;
 723         }
 724
 725         fcntl(fd, F_SETFL, O_NONBLOCK);
 726
 727         vnet_hdr = tap_probe_vnet_hdr(fd);
 728
 729         if (net_init_tap_one(tap, peer, "tap", name, NULL,
 730                              script, downscript,
 731                              vhostfdname, vnet_hdr, fd)) {
 732             return -1;
 733         }
 734     } else if (tap->has_fds) {
 735         char *fds[MAX_TAP_QUEUES];
 736         char *vhost_fds[MAX_TAP_QUEUES];
 737         int nfds, nvhosts;
 738
 739         if (tap->has_ifname || tap->has_script || tap->has_downscript ||
 740             tap->has_vnet_hdr || tap->has_helper || tap->has_queues ||
 741             tap->has_vhostfd) {
 742             error_report("ifname=, script=, downscript=, vnet_hdr=, "
 743                          "helper=, queues=, and vhostfd= "
 744                          "are invalid with fds=");
 745             return -1;
 746         }
 747
 748         nfds = get_fds(tap->fds, fds, MAX_TAP_QUEUES);
 749         if (tap->has_vhostfds) {
 750             nvhosts = get_fds(tap->vhostfds, vhost_fds, MAX_TAP_QUEUES);
 751             if (nfds != nvhosts) {
 752                 error_report("The number of fds passed does not match the "
 753                              "number of vhostfds passed");
 754                 return -1;
 755             }
 756         }
 757
 758         for (i = 0; i < nfds; i++) {
 759             fd = monitor_handle_fd_param(cur_mon, fds[i]);
 760             if (fd == -1) {
 761                 return -1;
 762             }
 763
 764             fcntl(fd, F_SETFL, O_NONBLOCK);
 765
 766             if (i == 0) {
 767                 vnet_hdr = tap_probe_vnet_hdr(fd);
 768             } else if (vnet_hdr != tap_probe_vnet_hdr(fd)) {
 769                 error_report("vnet_hdr not consistent across given tap fds");
 770                 return -1;
 771             }
 772
 773             if (net_init_tap_one(tap, peer, "tap", name, ifname,
 774                                  script, downscript,
 775                                  tap->has_vhostfds ? vhost_fds[i] : NULL,
 776                                  vnet_hdr, fd)) {
 777                 return -1;
 778             }
 779         }
 780     } else if (tap->has_helper) {
 781         if (tap->has_ifname || tap->has_script || tap->has_downscript ||
 782             tap->has_vnet_hdr || tap->has_queues || tap->has_vhostfds) {
 783             error_report("ifname=, script=, downscript=, and vnet_hdr= "
 784                          "queues=, and vhostfds= are invalid with helper=");
 785             return -1;
 786         }
 787
 788         fd = net_bridge_run_helper(tap->helper, DEFAULT_BRIDGE_INTERFACE);
 789         if (fd == -1) {
 790             return -1;
 791         }
 792
 793         fcntl(fd, F_SETFL, O_NONBLOCK);
 794         vnet_hdr = tap_probe_vnet_hdr(fd);
 795
 796         if (net_init_tap_one(tap, peer, "bridge", name, ifname,
 797                              script, downscript, vhostfdname,
 798                              vnet_hdr, fd)) {
 799             close(fd);
 800             return -1;
 801         }
 802     } else {
 803         if (tap->has_vhostfds) {
 804             error_report("vhostfds= is invalid if fds= wasn't specified");
 805             return -1;
 806         }
 807         script = tap->has_script ? tap->script : DEFAULT_NETWORK_SCRIPT;
 808         downscript = tap->has_downscript ? tap->downscript :
 809             DEFAULT_NETWORK_DOWN_SCRIPT;
 810
 811         if (tap->has_ifname) {
 812             pstrcpy(ifname, sizeof ifname, tap->ifname);
 813         } else {
 814             ifname[0] = '\0';
 815         }
 816
 817         for (i = 0; i < queues; i++) {
 818             fd = net_tap_init(tap, &vnet_hdr, i >= 1 ? "no" : script,
 819                               ifname, sizeof ifname, queues > 1);
 820             if (fd == -1) {
 821                 return -1;
 822             }
 823
 824             if (queues > 1 && i == 0 && !tap->has_ifname) {
 825                 if (tap_fd_get_ifname(fd, ifname)) {
 826                     error_report("Fail to get ifname");
 827                     close(fd);
 828                     return -1;
 829                 }
 830             }
 831
 832             if (net_init_tap_one(tap, peer, "tap", name, ifname,
 833                                  i >= 1 ? "no" : script,
 834                                  i >= 1 ? "no" : downscript,
 835                                  vhostfdname, vnet_hdr, fd)) {
 836                 close(fd);
 837                 return -1;
 838             }
 839         }
 840     }
 841
 842     return 0;
 843 }
 844
 845 VHostNetState *tap_get_vhost_net(NetClientState *nc)
 846 {
 847     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 848     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 849     return s->vhost_net;
 850 }
 851
 852 int tap_enable(NetClientState *nc)
 853 {
 854     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 855     int ret;
 856
 857     if (s->enabled) {
 858         return 0;
 859     } else {
 860         ret = tap_fd_enable(s->fd);
 861         if (ret == 0) {
 862             s->enabled = true;
 863             tap_update_fd_handler(s);
 864         }
 865         return ret;
 866     }
 867 }
 868
 869 int tap_disable(NetClientState *nc)
 870 {
 871     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 872     int ret;
 873
 874     if (s->enabled == 0) {
 875         return 0;
 876     } else {
 877         ret = tap_fd_disable(s->fd);
 878         if (ret == 0) {
 879             qemu_purge_queued_packets(nc);
 880             s->enabled = false;
 881             tap_update_fd_handler(s);
 882         }
 883         return ret;
 884     }
 885 }