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 void launch_script(const char *setup_script, const char *ifname,
  63                           int fd, Error **errp);
  64
  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_handler(s->fd,
  71                         s->read_poll && s->enabled ? tap_send : NULL,
  72                         s->write_poll && s->enabled ? tap_writable : NULL,
  73                         s);
  74 }
  75
  76 static void tap_read_poll(TAPState *s, bool enable)
  77 {
  78     s->read_poll = enable;
  79     tap_update_fd_handler(s);
  80 }
  81
  82 static void tap_write_poll(TAPState *s, bool enable)
  83 {
  84     s->write_poll = enable;
  85     tap_update_fd_handler(s);
  86 }
  87
  88 static void tap_writable(void *opaque)
  89 {
  90     TAPState *s = opaque;
  91
  92     tap_write_poll(s, false);
  93
  94     qemu_flush_queued_packets(&s->nc);
  95 }
  96
  97 static ssize_t tap_write_packet(TAPState *s, const struct iovec *iov, int iovcnt)
  98 {
  99     ssize_t len;
 100
 101     do {
 102         len = writev(s->fd, iov, iovcnt);
 103     } while (len == -1 && errno == EINTR);
 104
 105     if (len == -1 && errno == EAGAIN) {
 106         tap_write_poll(s, true);
 107         return 0;
 108     }
 109
 110     return len;
 111 }
 112
 113 static ssize_t tap_receive_iov(NetClientState *nc, const struct iovec *iov,
 114                                int iovcnt)
 115 {
 116     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 117     const struct iovec *iovp = iov;
 118     struct iovec iov_copy[iovcnt + 1];
 119     struct virtio_net_hdr_mrg_rxbuf hdr = { };
 120
 121     if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
 122         iov_copy[0].iov_base = &hdr;
 123         iov_copy[0].iov_len =  s->host_vnet_hdr_len;
 124         memcpy(&iov_copy[1], iov, iovcnt * sizeof(*iov));
 125         iovp = iov_copy;
 126         iovcnt++;
 127     }
 128
 129     return tap_write_packet(s, iovp, iovcnt);
 130 }
 131
 132 static ssize_t tap_receive_raw(NetClientState *nc, const uint8_t *buf, size_t size)
 133 {
 134     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 135     struct iovec iov[2];
 136     int iovcnt = 0;
 137     struct virtio_net_hdr_mrg_rxbuf hdr = { };
 138
 139     if (s->host_vnet_hdr_len) {
 140         iov[iovcnt].iov_base = &hdr;
 141         iov[iovcnt].iov_len  = s->host_vnet_hdr_len;
 142         iovcnt++;
 143     }
 144
 145     iov[iovcnt].iov_base = (char *)buf;
 146     iov[iovcnt].iov_len  = size;
 147     iovcnt++;
 148
 149     return tap_write_packet(s, iov, iovcnt);
 150 }
 151
 152 static ssize_t tap_receive(NetClientState *nc, const uint8_t *buf, size_t size)
 153 {
 154     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 155     struct iovec iov[1];
 156
 157     if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
 158         return tap_receive_raw(nc, buf, size);
 159     }
 160
 161     iov[0].iov_base = (char *)buf;
 162     iov[0].iov_len  = size;
 163
 164     return tap_write_packet(s, iov, 1);
 165 }
 166
 167 #ifndef __sun__
 168 ssize_t tap_read_packet(int tapfd, uint8_t *buf, int maxlen)
 169 {
 170     return read(tapfd, buf, maxlen);
 171 }
 172 #endif
 173
 174 static void tap_send_completed(NetClientState *nc, ssize_t len)
 175 {
 176     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 177     tap_read_poll(s, true);
 178 }
 179
 180 static void tap_send(void *opaque)
 181 {
 182     TAPState *s = opaque;
 183     int size;
 184     int packets = 0;
 185
 186     while (true) {
 187         uint8_t *buf = s->buf;
 188
 189         size = tap_read_packet(s->fd, s->buf, sizeof(s->buf));
 190         if (size <= 0) {
 191             break;
 192         }
 193
 194         if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
 195             buf  += s->host_vnet_hdr_len;
 196             size -= s->host_vnet_hdr_len;
 197         }
 198
 199         size = qemu_send_packet_async(&s->nc, buf, size, tap_send_completed);
 200         if (size == 0) {
 201             tap_read_poll(s, false);
 202             break;
 203         } else if (size < 0) {
 204             break;
 205         }
 206
 207         /*
 208          * When the host keeps receiving more packets while tap_send() is
 209          * running we can hog the QEMU global mutex.  Limit the number of
 210          * packets that are processed per tap_send() callback to prevent
 211          * stalling the guest.
 212          */
 213         packets++;
 214         if (packets >= 50) {
 215             break;
 216         }
 217     }
 218 }
 219
 220 static bool tap_has_ufo(NetClientState *nc)
 221 {
 222     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 223
 224     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 225
 226     return s->has_ufo;
 227 }
 228
 229 static bool tap_has_vnet_hdr(NetClientState *nc)
 230 {
 231     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 232
 233     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 234
 235     return !!s->host_vnet_hdr_len;
 236 }
 237
 238 static bool tap_has_vnet_hdr_len(NetClientState *nc, int len)
 239 {
 240     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 241
 242     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 243
 244     return !!tap_probe_vnet_hdr_len(s->fd, len);
 245 }
 246
 247 static void tap_set_vnet_hdr_len(NetClientState *nc, int len)
 248 {
 249     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 250
 251     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 252     assert(len == sizeof(struct virtio_net_hdr_mrg_rxbuf) ||
 253            len == sizeof(struct virtio_net_hdr));
 254
 255     tap_fd_set_vnet_hdr_len(s->fd, len);
 256     s->host_vnet_hdr_len = len;
 257 }
 258
 259 static void tap_using_vnet_hdr(NetClientState *nc, bool using_vnet_hdr)
 260 {
 261     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 262
 263     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 264     assert(!!s->host_vnet_hdr_len == using_vnet_hdr);
 265
 266     s->using_vnet_hdr = using_vnet_hdr;
 267 }
 268
 269 static int tap_set_vnet_le(NetClientState *nc, bool is_le)
 270 {
 271     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 272
 273     return tap_fd_set_vnet_le(s->fd, is_le);
 274 }
 275
 276 static int tap_set_vnet_be(NetClientState *nc, bool is_be)
 277 {
 278     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 279
 280     return tap_fd_set_vnet_be(s->fd, is_be);
 281 }
 282
 283 static void tap_set_offload(NetClientState *nc, int csum, int tso4,
 284                      int tso6, int ecn, int ufo)
 285 {
 286     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 287     if (s->fd < 0) {
 288         return;
 289     }
 290
 291     tap_fd_set_offload(s->fd, csum, tso4, tso6, ecn, ufo);
 292 }
 293
 294 static void tap_cleanup(NetClientState *nc)
 295 {
 296     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 297     Error *err = NULL;
 298
 299     if (s->vhost_net) {
 300         vhost_net_cleanup(s->vhost_net);
 301         s->vhost_net = NULL;
 302     }
 303
 304     qemu_purge_queued_packets(nc);
 305
 306     if (s->down_script[0]) {
 307         launch_script(s->down_script, s->down_script_arg, s->fd, &err);
 308         if (err) {
 309             error_report_err(err);
 310         }
 311     }
 312
 313     tap_read_poll(s, false);
 314     tap_write_poll(s, false);
 315     close(s->fd);
 316     s->fd = -1;
 317 }
 318
 319 static void tap_poll(NetClientState *nc, bool enable)
 320 {
 321     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 322     tap_read_poll(s, enable);
 323     tap_write_poll(s, enable);
 324 }
 325
 326 int tap_get_fd(NetClientState *nc)
 327 {
 328     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 329     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 330     return s->fd;
 331 }
 332
 333 /* fd support */
 334
 335 static NetClientInfo net_tap_info = {
 336     .type = NET_CLIENT_OPTIONS_KIND_TAP,
 337     .size = sizeof(TAPState),
 338     .receive = tap_receive,
 339     .receive_raw = tap_receive_raw,
 340     .receive_iov = tap_receive_iov,
 341     .poll = tap_poll,
 342     .cleanup = tap_cleanup,
 343     .has_ufo = tap_has_ufo,
 344     .has_vnet_hdr = tap_has_vnet_hdr,
 345     .has_vnet_hdr_len = tap_has_vnet_hdr_len,
 346     .using_vnet_hdr = tap_using_vnet_hdr,
 347     .set_offload = tap_set_offload,
 348     .set_vnet_hdr_len = tap_set_vnet_hdr_len,
 349     .set_vnet_le = tap_set_vnet_le,
 350     .set_vnet_be = tap_set_vnet_be,
 351 };
 352
 353 static TAPState *net_tap_fd_init(NetClientState *peer,
 354                                  const char *model,
 355                                  const char *name,
 356                                  int fd,
 357                                  int vnet_hdr)
 358 {
 359     NetClientState *nc;
 360     TAPState *s;
 361
 362     nc = qemu_new_net_client(&net_tap_info, peer, model, name);
 363
 364     s = DO_UPCAST(TAPState, nc, nc);
 365
 366     s->fd = fd;
 367     s->host_vnet_hdr_len = vnet_hdr ? sizeof(struct virtio_net_hdr) : 0;
 368     s->using_vnet_hdr = false;
 369     s->has_ufo = tap_probe_has_ufo(s->fd);
 370     s->enabled = true;
 371     tap_set_offload(&s->nc, 0, 0, 0, 0, 0);
 372     /*
 373      * Make sure host header length is set correctly in tap:
 374      * it might have been modified by another instance of qemu.
 375      */
 376     if (tap_probe_vnet_hdr_len(s->fd, s->host_vnet_hdr_len)) {
 377         tap_fd_set_vnet_hdr_len(s->fd, s->host_vnet_hdr_len);
 378     }
 379     tap_read_poll(s, true);
 380     s->vhost_net = NULL;
 381     return s;
 382 }
 383
 384 static void launch_script(const char *setup_script, const char *ifname,
 385                           int fd, Error **errp)
 386 {
 387     int pid, status;
 388     char *args[3];
 389     char **parg;
 390
 391     /* try to launch network script */
 392     pid = fork();
 393     if (pid < 0) {
 394         error_setg_errno(errp, errno, "could not launch network script %s",
 395                          setup_script);
 396         return;
 397     }
 398     if (pid == 0) {
 399         int open_max = sysconf(_SC_OPEN_MAX), i;
 400
 401         for (i = 3; i < open_max; i++) {
 402             if (i != fd) {
 403                 close(i);
 404             }
 405         }
 406         parg = args;
 407         *parg++ = (char *)setup_script;
 408         *parg++ = (char *)ifname;
 409         *parg = NULL;
 410         execv(setup_script, args);
 411         _exit(1);
 412     } else {
 413         while (waitpid(pid, &status, 0) != pid) {
 414             /* loop */
 415         }
 416
 417         if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
 418             return;
 419         }
 420         error_setg(errp, "network script %s failed with status %d",
 421                    setup_script, status);
 422     }
 423 }
 424
 425 static int recv_fd(int c)
 426 {
 427     int fd;
 428     uint8_t msgbuf[CMSG_SPACE(sizeof(fd))];
 429     struct msghdr msg = {
 430         .msg_control = msgbuf,
 431         .msg_controllen = sizeof(msgbuf),
 432     };
 433     struct cmsghdr *cmsg;
 434     struct iovec iov;
 435     uint8_t req[1];
 436     ssize_t len;
 437
 438     cmsg = CMSG_FIRSTHDR(&msg);
 439     cmsg->cmsg_level = SOL_SOCKET;
 440     cmsg->cmsg_type = SCM_RIGHTS;
 441     cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
 442     msg.msg_controllen = cmsg->cmsg_len;
 443
 444     iov.iov_base = req;
 445     iov.iov_len = sizeof(req);
 446
 447     msg.msg_iov = &iov;
 448     msg.msg_iovlen = 1;
 449
 450     len = recvmsg(c, &msg, 0);
 451     if (len > 0) {
 452         memcpy(&fd, CMSG_DATA(cmsg), sizeof(fd));
 453         return fd;
 454     }
 455
 456     return len;
 457 }
 458
 459 static int net_bridge_run_helper(const char *helper, const char *bridge,
 460                                  Error **errp)
 461 {
 462     sigset_t oldmask, mask;
 463     int pid, status;
 464     char *args[5];
 465     char **parg;
 466     int sv[2];
 467
 468     sigemptyset(&mask);
 469     sigaddset(&mask, SIGCHLD);
 470     sigprocmask(SIG_BLOCK, &mask, &oldmask);
 471
 472     if (socketpair(PF_UNIX, SOCK_STREAM, 0, sv) == -1) {
 473         error_setg_errno(errp, errno, "socketpair() failed");
 474         return -1;
 475     }
 476
 477     /* try to launch bridge helper */
 478     pid = fork();
 479     if (pid < 0) {
 480         error_setg_errno(errp, errno, "Can't fork bridge helper");
 481         return -1;
 482     }
 483     if (pid == 0) {
 484         int open_max = sysconf(_SC_OPEN_MAX), i;
 485         char fd_buf[6+10];
 486         char br_buf[6+IFNAMSIZ] = {0};
 487         char helper_cmd[PATH_MAX + sizeof(fd_buf) + sizeof(br_buf) + 15];
 488
 489         for (i = 3; i < open_max; i++) {
 490             if (i != sv[1]) {
 491                 close(i);
 492             }
 493         }
 494
 495         snprintf(fd_buf, sizeof(fd_buf), "%s%d", "--fd=", sv[1]);
 496
 497         if (strrchr(helper, ' ') || strrchr(helper, '\t')) {
 498             /* assume helper is a command */
 499
 500             if (strstr(helper, "--br=") == NULL) {
 501                 snprintf(br_buf, sizeof(br_buf), "%s%s", "--br=", bridge);
 502             }
 503
 504             snprintf(helper_cmd, sizeof(helper_cmd), "%s %s %s %s",
 505                      helper, "--use-vnet", fd_buf, br_buf);
 506
 507             parg = args;
 508             *parg++ = (char *)"sh";
 509             *parg++ = (char *)"-c";
 510             *parg++ = helper_cmd;
 511             *parg++ = NULL;
 512
 513             execv("/bin/sh", args);
 514         } else {
 515             /* assume helper is just the executable path name */
 516
 517             snprintf(br_buf, sizeof(br_buf), "%s%s", "--br=", bridge);
 518
 519             parg = args;
 520             *parg++ = (char *)helper;
 521             *parg++ = (char *)"--use-vnet";
 522             *parg++ = fd_buf;
 523             *parg++ = br_buf;
 524             *parg++ = NULL;
 525
 526             execv(helper, args);
 527         }
 528         _exit(1);
 529
 530     } else {
 531         int fd;
 532         int saved_errno;
 533
 534         close(sv[1]);
 535
 536         do {
 537             fd = recv_fd(sv[0]);
 538         } while (fd == -1 && errno == EINTR);
 539         saved_errno = errno;
 540
 541         close(sv[0]);
 542
 543         while (waitpid(pid, &status, 0) != pid) {
 544             /* loop */
 545         }
 546         sigprocmask(SIG_SETMASK, &oldmask, NULL);
 547         if (fd < 0) {
 548             error_setg_errno(errp, saved_errno,
 549                              "failed to recv file descriptor");
 550             return -1;
 551         }
 552         if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) {
 553             error_setg(errp, "bridge helper failed");
 554             return -1;
 555         }
 556         return fd;
 557     }
 558 }
 559
 560 int net_init_bridge(const NetClientOptions *opts, const char *name,
 561                     NetClientState *peer, Error **errp)
 562 {
 563     const NetdevBridgeOptions *bridge;
 564     const char *helper, *br;
 565     TAPState *s;
 566     int fd, vnet_hdr;
 567
 568     assert(opts->kind == NET_CLIENT_OPTIONS_KIND_BRIDGE);
 569     bridge = opts->bridge;
 570
 571     helper = bridge->has_helper ? bridge->helper : DEFAULT_BRIDGE_HELPER;
 572     br     = bridge->has_br     ? bridge->br     : DEFAULT_BRIDGE_INTERFACE;
 573
 574     fd = net_bridge_run_helper(helper, br, errp);
 575     if (fd == -1) {
 576         return -1;
 577     }
 578
 579     fcntl(fd, F_SETFL, O_NONBLOCK);
 580     vnet_hdr = tap_probe_vnet_hdr(fd);
 581     s = net_tap_fd_init(peer, "bridge", name, fd, vnet_hdr);
 582
 583     snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s,br=%s", helper,
 584              br);
 585
 586     return 0;
 587 }
 588
 589 static int net_tap_init(const NetdevTapOptions *tap, int *vnet_hdr,
 590                         const char *setup_script, char *ifname,
 591                         size_t ifname_sz, int mq_required, Error **errp)
 592 {
 593     Error *err = NULL;
 594     int fd, vnet_hdr_required;
 595
 596     if (tap->has_vnet_hdr) {
 597         *vnet_hdr = tap->vnet_hdr;
 598         vnet_hdr_required = *vnet_hdr;
 599     } else {
 600         *vnet_hdr = 1;
 601         vnet_hdr_required = 0;
 602     }
 603
 604     TFR(fd = tap_open(ifname, ifname_sz, vnet_hdr, vnet_hdr_required,
 605                       mq_required, errp));
 606     if (fd < 0) {
 607         return -1;
 608     }
 609
 610     if (setup_script &&
 611         setup_script[0] != '\0' &&
 612         strcmp(setup_script, "no") != 0) {
 613         launch_script(setup_script, ifname, fd, &err);
 614         if (err) {
 615             error_propagate(errp, err);
 616             close(fd);
 617             return -1;
 618         }
 619     }
 620
 621     return fd;
 622 }
 623
 624 #define MAX_TAP_QUEUES 1024
 625
 626 static void net_init_tap_one(const NetdevTapOptions *tap, NetClientState *peer,
 627                              const char *model, const char *name,
 628                              const char *ifname, const char *script,
 629                              const char *downscript, const char *vhostfdname,
 630                              int vnet_hdr, int fd, Error **errp)
 631 {
 632     Error *err = NULL;
 633     TAPState *s = net_tap_fd_init(peer, model, name, fd, vnet_hdr);
 634     int vhostfd;
 635
 636     tap_set_sndbuf(s->fd, tap, &err);
 637     if (err) {
 638         error_propagate(errp, err);
 639         return;
 640     }
 641
 642     if (tap->has_fd || tap->has_fds) {
 643         snprintf(s->nc.info_str, sizeof(s->nc.info_str), "fd=%d", fd);
 644     } else if (tap->has_helper) {
 645         snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s",
 646                  tap->helper);
 647     } else {
 648         snprintf(s->nc.info_str, sizeof(s->nc.info_str),
 649                  "ifname=%s,script=%s,downscript=%s", ifname, script,
 650                  downscript);
 651
 652         if (strcmp(downscript, "no") != 0) {
 653             snprintf(s->down_script, sizeof(s->down_script), "%s", downscript);
 654             snprintf(s->down_script_arg, sizeof(s->down_script_arg),
 655                      "%s", ifname);
 656         }
 657     }
 658
 659     if (tap->has_vhost ? tap->vhost :
 660         vhostfdname || (tap->has_vhostforce && tap->vhostforce)) {
 661         VhostNetOptions options;
 662
 663         options.backend_type = VHOST_BACKEND_TYPE_KERNEL;
 664         options.net_backend = &s->nc;
 665
 666         if (tap->has_vhostfd || tap->has_vhostfds) {
 667             vhostfd = monitor_fd_param(cur_mon, vhostfdname, &err);
 668             if (vhostfd == -1) {
 669                 error_propagate(errp, err);
 670                 return;
 671             }
 672         } else {
 673             vhostfd = open("/dev/vhost-net", O_RDWR);
 674             if (vhostfd < 0) {
 675                 error_setg_errno(errp, errno,
 676                                  "tap: open vhost char device failed");
 677                 return;
 678             }
 679         }
 680         options.opaque = (void *)(uintptr_t)vhostfd;
 681
 682         s->vhost_net = vhost_net_init(&options);
 683         if (!s->vhost_net) {
 684             error_setg(errp,
 685                        "vhost-net requested but could not be initialized");
 686             return;
 687         }
 688     } else if (tap->has_vhostfd || tap->has_vhostfds) {
 689         error_setg(errp, "vhostfd= is not valid without vhost");
 690     }
 691 }
 692
 693 static int get_fds(char *str, char *fds[], int max)
 694 {
 695     char *ptr = str, *this;
 696     size_t len = strlen(str);
 697     int i = 0;
 698
 699     while (i < max && ptr < str + len) {
 700         this = strchr(ptr, ':');
 701
 702         if (this == NULL) {
 703             fds[i] = g_strdup(ptr);
 704         } else {
 705             fds[i] = g_strndup(ptr, this - ptr);
 706         }
 707
 708         i++;
 709         if (this == NULL) {
 710             break;
 711         } else {
 712             ptr = this + 1;
 713         }
 714     }
 715
 716     return i;
 717 }
 718
 719 int net_init_tap(const NetClientOptions *opts, const char *name,
 720                  NetClientState *peer, Error **errp)
 721 {
 722     const NetdevTapOptions *tap;
 723     int fd, vnet_hdr = 0, i = 0, queues;
 724     /* for the no-fd, no-helper case */
 725     const char *script = NULL; /* suppress wrong "uninit'd use" gcc warning */
 726     const char *downscript = NULL;
 727     Error *err = NULL;
 728     const char *vhostfdname;
 729     char ifname[128];
 730
 731     assert(opts->kind == NET_CLIENT_OPTIONS_KIND_TAP);
 732     tap = opts->tap;
 733     queues = tap->has_queues ? tap->queues : 1;
 734     vhostfdname = tap->has_vhostfd ? tap->vhostfd : NULL;
 735
 736     /* QEMU vlans does not support multiqueue tap, in this case peer is set.
 737      * For -netdev, peer is always NULL. */
 738     if (peer && (tap->has_queues || tap->has_fds || tap->has_vhostfds)) {
 739         error_setg(errp, "Multiqueue tap cannot be used with QEMU vlans");
 740         return -1;
 741     }
 742
 743     if (tap->has_fd) {
 744         if (tap->has_ifname || tap->has_script || tap->has_downscript ||
 745             tap->has_vnet_hdr || tap->has_helper || tap->has_queues ||
 746             tap->has_fds || tap->has_vhostfds) {
 747             error_setg(errp, "ifname=, script=, downscript=, vnet_hdr=, "
 748                        "helper=, queues=, fds=, and vhostfds= "
 749                        "are invalid with fd=");
 750             return -1;
 751         }
 752
 753         fd = monitor_fd_param(cur_mon, tap->fd, &err);
 754         if (fd == -1) {
 755             error_propagate(errp, err);
 756             return -1;
 757         }
 758
 759         fcntl(fd, F_SETFL, O_NONBLOCK);
 760
 761         vnet_hdr = tap_probe_vnet_hdr(fd);
 762
 763         net_init_tap_one(tap, peer, "tap", name, NULL,
 764                          script, downscript,
 765                          vhostfdname, vnet_hdr, fd, &err);
 766         if (err) {
 767             error_propagate(errp, err);
 768             return -1;
 769         }
 770     } else if (tap->has_fds) {
 771         char *fds[MAX_TAP_QUEUES];
 772         char *vhost_fds[MAX_TAP_QUEUES];
 773         int nfds, nvhosts;
 774
 775         if (tap->has_ifname || tap->has_script || tap->has_downscript ||
 776             tap->has_vnet_hdr || tap->has_helper || tap->has_queues ||
 777             tap->has_vhostfd) {
 778             error_setg(errp, "ifname=, script=, downscript=, vnet_hdr=, "
 779                        "helper=, queues=, and vhostfd= "
 780                        "are invalid with fds=");
 781             return -1;
 782         }
 783
 784         nfds = get_fds(tap->fds, fds, MAX_TAP_QUEUES);
 785         if (tap->has_vhostfds) {
 786             nvhosts = get_fds(tap->vhostfds, vhost_fds, MAX_TAP_QUEUES);
 787             if (nfds != nvhosts) {
 788                 error_setg(errp, "The number of fds passed does not match "
 789                            "the number of vhostfds passed");
 790                 return -1;
 791             }
 792         }
 793
 794         for (i = 0; i < nfds; i++) {
 795             fd = monitor_fd_param(cur_mon, fds[i], &err);
 796             if (fd == -1) {
 797                 error_propagate(errp, err);
 798                 return -1;
 799             }
 800
 801             fcntl(fd, F_SETFL, O_NONBLOCK);
 802
 803             if (i == 0) {
 804                 vnet_hdr = tap_probe_vnet_hdr(fd);
 805             } else if (vnet_hdr != tap_probe_vnet_hdr(fd)) {
 806                 error_setg(errp,
 807                            "vnet_hdr not consistent across given tap fds");
 808                 return -1;
 809             }
 810
 811             net_init_tap_one(tap, peer, "tap", name, ifname,
 812                              script, downscript,
 813                              tap->has_vhostfds ? vhost_fds[i] : NULL,
 814                              vnet_hdr, fd, &err);
 815             if (err) {
 816                 error_propagate(errp, err);
 817                 return -1;
 818             }
 819         }
 820     } else if (tap->has_helper) {
 821         if (tap->has_ifname || tap->has_script || tap->has_downscript ||
 822             tap->has_vnet_hdr || tap->has_queues || tap->has_vhostfds) {
 823             error_setg(errp, "ifname=, script=, downscript=, vnet_hdr=, "
 824                        "queues=, and vhostfds= are invalid with helper=");
 825             return -1;
 826         }
 827
 828         fd = net_bridge_run_helper(tap->helper, DEFAULT_BRIDGE_INTERFACE,
 829                                    errp);
 830         if (fd == -1) {
 831             return -1;
 832         }
 833
 834         fcntl(fd, F_SETFL, O_NONBLOCK);
 835         vnet_hdr = tap_probe_vnet_hdr(fd);
 836
 837         net_init_tap_one(tap, peer, "bridge", name, ifname,
 838                          script, downscript, vhostfdname,
 839                          vnet_hdr, fd, &err);
 840         if (err) {
 841             error_propagate(errp, err);
 842             close(fd);
 843             return -1;
 844         }
 845     } else {
 846         if (tap->has_vhostfds) {
 847             error_setg(errp, "vhostfds= is invalid if fds= wasn't specified");
 848             return -1;
 849         }
 850         script = tap->has_script ? tap->script : DEFAULT_NETWORK_SCRIPT;
 851         downscript = tap->has_downscript ? tap->downscript :
 852             DEFAULT_NETWORK_DOWN_SCRIPT;
 853
 854         if (tap->has_ifname) {
 855             pstrcpy(ifname, sizeof ifname, tap->ifname);
 856         } else {
 857             ifname[0] = '\0';
 858         }
 859
 860         for (i = 0; i < queues; i++) {
 861             fd = net_tap_init(tap, &vnet_hdr, i >= 1 ? "no" : script,
 862                               ifname, sizeof ifname, queues > 1, errp);
 863             if (fd == -1) {
 864                 return -1;
 865             }
 866
 867             if (queues > 1 && i == 0 && !tap->has_ifname) {
 868                 if (tap_fd_get_ifname(fd, ifname)) {
 869                     error_setg(errp, "Fail to get ifname");
 870                     close(fd);
 871                     return -1;
 872                 }
 873             }
 874
 875             net_init_tap_one(tap, peer, "tap", name, ifname,
 876                              i >= 1 ? "no" : script,
 877                              i >= 1 ? "no" : downscript,
 878                              vhostfdname, vnet_hdr, fd, &err);
 879             if (err) {
 880                 error_propagate(errp, err);
 881                 close(fd);
 882                 return -1;
 883             }
 884         }
 885     }
 886
 887     return 0;
 888 }
 889
 890 VHostNetState *tap_get_vhost_net(NetClientState *nc)
 891 {
 892     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 893     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 894     return s->vhost_net;
 895 }
 896
 897 int tap_enable(NetClientState *nc)
 898 {
 899     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 900     int ret;
 901
 902     if (s->enabled) {
 903         return 0;
 904     } else {
 905         ret = tap_fd_enable(s->fd);
 906         if (ret == 0) {
 907             s->enabled = true;
 908             tap_update_fd_handler(s);
 909         }
 910         return ret;
 911     }
 912 }
 913
 914 int tap_disable(NetClientState *nc)
 915 {
 916     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 917     int ret;
 918
 919     if (s->enabled == 0) {
 920         return 0;
 921     } else {
 922         ret = tap_fd_disable(s->fd);
 923         if (ret == 0) {
 924             qemu_purge_queued_packets(nc);
 925             s->enabled = false;
 926             tap_update_fd_handler(s);
 927         }
 928         return ret;
 929     }
 930 }