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 void tap_set_offload(NetClientState *nc, int csum, int tso4,
 270                      int tso6, int ecn, int ufo)
 271 {
 272     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 273     if (s->fd < 0) {
 274         return;
 275     }
 276
 277     tap_fd_set_offload(s->fd, csum, tso4, tso6, ecn, ufo);
 278 }
 279
 280 static void tap_cleanup(NetClientState *nc)
 281 {
 282     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 283     Error *err = NULL;
 284
 285     if (s->vhost_net) {
 286         vhost_net_cleanup(s->vhost_net);
 287         s->vhost_net = NULL;
 288     }
 289
 290     qemu_purge_queued_packets(nc);
 291
 292     if (s->down_script[0]) {
 293         launch_script(s->down_script, s->down_script_arg, s->fd, &err);
 294         if (err) {
 295             error_report_err(err);
 296         }
 297     }
 298
 299     tap_read_poll(s, false);
 300     tap_write_poll(s, false);
 301     close(s->fd);
 302     s->fd = -1;
 303 }
 304
 305 static void tap_poll(NetClientState *nc, bool enable)
 306 {
 307     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 308     tap_read_poll(s, enable);
 309     tap_write_poll(s, enable);
 310 }
 311
 312 int tap_get_fd(NetClientState *nc)
 313 {
 314     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 315     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 316     return s->fd;
 317 }
 318
 319 /* fd support */
 320
 321 static NetClientInfo net_tap_info = {
 322     .type = NET_CLIENT_OPTIONS_KIND_TAP,
 323     .size = sizeof(TAPState),
 324     .receive = tap_receive,
 325     .receive_raw = tap_receive_raw,
 326     .receive_iov = tap_receive_iov,
 327     .poll = tap_poll,
 328     .cleanup = tap_cleanup,
 329     .has_ufo = tap_has_ufo,
 330     .has_vnet_hdr = tap_has_vnet_hdr,
 331     .has_vnet_hdr_len = tap_has_vnet_hdr_len,
 332     .using_vnet_hdr = tap_using_vnet_hdr,
 333     .set_offload = tap_set_offload,
 334     .set_vnet_hdr_len = tap_set_vnet_hdr_len,
 335 };
 336
 337 static TAPState *net_tap_fd_init(NetClientState *peer,
 338                                  const char *model,
 339                                  const char *name,
 340                                  int fd,
 341                                  int vnet_hdr)
 342 {
 343     NetClientState *nc;
 344     TAPState *s;
 345
 346     nc = qemu_new_net_client(&net_tap_info, peer, model, name);
 347
 348     s = DO_UPCAST(TAPState, nc, nc);
 349
 350     s->fd = fd;
 351     s->host_vnet_hdr_len = vnet_hdr ? sizeof(struct virtio_net_hdr) : 0;
 352     s->using_vnet_hdr = false;
 353     s->has_ufo = tap_probe_has_ufo(s->fd);
 354     s->enabled = true;
 355     tap_set_offload(&s->nc, 0, 0, 0, 0, 0);
 356     /*
 357      * Make sure host header length is set correctly in tap:
 358      * it might have been modified by another instance of qemu.
 359      */
 360     if (tap_probe_vnet_hdr_len(s->fd, s->host_vnet_hdr_len)) {
 361         tap_fd_set_vnet_hdr_len(s->fd, s->host_vnet_hdr_len);
 362     }
 363     tap_read_poll(s, true);
 364     s->vhost_net = NULL;
 365     return s;
 366 }
 367
 368 static void launch_script(const char *setup_script, const char *ifname,
 369                           int fd, Error **errp)
 370 {
 371     int pid, status;
 372     char *args[3];
 373     char **parg;
 374
 375     /* try to launch network script */
 376     pid = fork();
 377     if (pid < 0) {
 378         error_setg_errno(errp, errno, "could not launch network script %s",
 379                          setup_script);
 380         return;
 381     }
 382     if (pid == 0) {
 383         int open_max = sysconf(_SC_OPEN_MAX), i;
 384
 385         for (i = 3; i < open_max; i++) {
 386             if (i != fd) {
 387                 close(i);
 388             }
 389         }
 390         parg = args;
 391         *parg++ = (char *)setup_script;
 392         *parg++ = (char *)ifname;
 393         *parg = NULL;
 394         execv(setup_script, args);
 395         _exit(1);
 396     } else {
 397         while (waitpid(pid, &status, 0) != pid) {
 398             /* loop */
 399         }
 400
 401         if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
 402             return;
 403         }
 404         error_setg(errp, "network script %s failed with status %d",
 405                    setup_script, status);
 406     }
 407 }
 408
 409 static int recv_fd(int c)
 410 {
 411     int fd;
 412     uint8_t msgbuf[CMSG_SPACE(sizeof(fd))];
 413     struct msghdr msg = {
 414         .msg_control = msgbuf,
 415         .msg_controllen = sizeof(msgbuf),
 416     };
 417     struct cmsghdr *cmsg;
 418     struct iovec iov;
 419     uint8_t req[1];
 420     ssize_t len;
 421
 422     cmsg = CMSG_FIRSTHDR(&msg);
 423     cmsg->cmsg_level = SOL_SOCKET;
 424     cmsg->cmsg_type = SCM_RIGHTS;
 425     cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
 426     msg.msg_controllen = cmsg->cmsg_len;
 427
 428     iov.iov_base = req;
 429     iov.iov_len = sizeof(req);
 430
 431     msg.msg_iov = &iov;
 432     msg.msg_iovlen = 1;
 433
 434     len = recvmsg(c, &msg, 0);
 435     if (len > 0) {
 436         memcpy(&fd, CMSG_DATA(cmsg), sizeof(fd));
 437         return fd;
 438     }
 439
 440     return len;
 441 }
 442
 443 static int net_bridge_run_helper(const char *helper, const char *bridge,
 444                                  Error **errp)
 445 {
 446     sigset_t oldmask, mask;
 447     int pid, status;
 448     char *args[5];
 449     char **parg;
 450     int sv[2];
 451
 452     sigemptyset(&mask);
 453     sigaddset(&mask, SIGCHLD);
 454     sigprocmask(SIG_BLOCK, &mask, &oldmask);
 455
 456     if (socketpair(PF_UNIX, SOCK_STREAM, 0, sv) == -1) {
 457         error_setg_errno(errp, errno, "socketpair() failed");
 458         return -1;
 459     }
 460
 461     /* try to launch bridge helper */
 462     pid = fork();
 463     if (pid < 0) {
 464         error_setg_errno(errp, errno, "Can't fork bridge helper");
 465         return -1;
 466     }
 467     if (pid == 0) {
 468         int open_max = sysconf(_SC_OPEN_MAX), i;
 469         char fd_buf[6+10];
 470         char br_buf[6+IFNAMSIZ] = {0};
 471         char helper_cmd[PATH_MAX + sizeof(fd_buf) + sizeof(br_buf) + 15];
 472
 473         for (i = 3; i < open_max; i++) {
 474             if (i != sv[1]) {
 475                 close(i);
 476             }
 477         }
 478
 479         snprintf(fd_buf, sizeof(fd_buf), "%s%d", "--fd=", sv[1]);
 480
 481         if (strrchr(helper, ' ') || strrchr(helper, '\t')) {
 482             /* assume helper is a command */
 483
 484             if (strstr(helper, "--br=") == NULL) {
 485                 snprintf(br_buf, sizeof(br_buf), "%s%s", "--br=", bridge);
 486             }
 487
 488             snprintf(helper_cmd, sizeof(helper_cmd), "%s %s %s %s",
 489                      helper, "--use-vnet", fd_buf, br_buf);
 490
 491             parg = args;
 492             *parg++ = (char *)"sh";
 493             *parg++ = (char *)"-c";
 494             *parg++ = helper_cmd;
 495             *parg++ = NULL;
 496
 497             execv("/bin/sh", args);
 498         } else {
 499             /* assume helper is just the executable path name */
 500
 501             snprintf(br_buf, sizeof(br_buf), "%s%s", "--br=", bridge);
 502
 503             parg = args;
 504             *parg++ = (char *)helper;
 505             *parg++ = (char *)"--use-vnet";
 506             *parg++ = fd_buf;
 507             *parg++ = br_buf;
 508             *parg++ = NULL;
 509
 510             execv(helper, args);
 511         }
 512         _exit(1);
 513
 514     } else {
 515         int fd;
 516         int saved_errno;
 517
 518         close(sv[1]);
 519
 520         do {
 521             fd = recv_fd(sv[0]);
 522         } while (fd == -1 && errno == EINTR);
 523         saved_errno = errno;
 524
 525         close(sv[0]);
 526
 527         while (waitpid(pid, &status, 0) != pid) {
 528             /* loop */
 529         }
 530         sigprocmask(SIG_SETMASK, &oldmask, NULL);
 531         if (fd < 0) {
 532             error_setg_errno(errp, saved_errno,
 533                              "failed to recv file descriptor");
 534             return -1;
 535         }
 536         if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) {
 537             error_setg(errp, "bridge helper failed");
 538             return -1;
 539         }
 540         return fd;
 541     }
 542 }
 543
 544 int net_init_bridge(const NetClientOptions *opts, const char *name,
 545                     NetClientState *peer, Error **errp)
 546 {
 547     const NetdevBridgeOptions *bridge;
 548     const char *helper, *br;
 549     TAPState *s;
 550     int fd, vnet_hdr;
 551
 552     assert(opts->kind == NET_CLIENT_OPTIONS_KIND_BRIDGE);
 553     bridge = opts->bridge;
 554
 555     helper = bridge->has_helper ? bridge->helper : DEFAULT_BRIDGE_HELPER;
 556     br     = bridge->has_br     ? bridge->br     : DEFAULT_BRIDGE_INTERFACE;
 557
 558     fd = net_bridge_run_helper(helper, br, errp);
 559     if (fd == -1) {
 560         return -1;
 561     }
 562
 563     fcntl(fd, F_SETFL, O_NONBLOCK);
 564     vnet_hdr = tap_probe_vnet_hdr(fd);
 565     s = net_tap_fd_init(peer, "bridge", name, fd, vnet_hdr);
 566
 567     snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s,br=%s", helper,
 568              br);
 569
 570     return 0;
 571 }
 572
 573 static int net_tap_init(const NetdevTapOptions *tap, int *vnet_hdr,
 574                         const char *setup_script, char *ifname,
 575                         size_t ifname_sz, int mq_required, Error **errp)
 576 {
 577     Error *err = NULL;
 578     int fd, vnet_hdr_required;
 579
 580     if (tap->has_vnet_hdr) {
 581         *vnet_hdr = tap->vnet_hdr;
 582         vnet_hdr_required = *vnet_hdr;
 583     } else {
 584         *vnet_hdr = 1;
 585         vnet_hdr_required = 0;
 586     }
 587
 588     TFR(fd = tap_open(ifname, ifname_sz, vnet_hdr, vnet_hdr_required,
 589                       mq_required, errp));
 590     if (fd < 0) {
 591         return -1;
 592     }
 593
 594     if (setup_script &&
 595         setup_script[0] != '\0' &&
 596         strcmp(setup_script, "no") != 0) {
 597         launch_script(setup_script, ifname, fd, &err);
 598         if (err) {
 599             error_propagate(errp, err);
 600             close(fd);
 601             return -1;
 602         }
 603     }
 604
 605     return fd;
 606 }
 607
 608 #define MAX_TAP_QUEUES 1024
 609
 610 static void net_init_tap_one(const NetdevTapOptions *tap, NetClientState *peer,
 611                              const char *model, const char *name,
 612                              const char *ifname, const char *script,
 613                              const char *downscript, const char *vhostfdname,
 614                              int vnet_hdr, int fd, Error **errp)
 615 {
 616     Error *err = NULL;
 617     TAPState *s = net_tap_fd_init(peer, model, name, fd, vnet_hdr);
 618     int vhostfd;
 619
 620     tap_set_sndbuf(s->fd, tap, &err);
 621     if (err) {
 622         error_propagate(errp, err);
 623         return;
 624     }
 625
 626     if (tap->has_fd || tap->has_fds) {
 627         snprintf(s->nc.info_str, sizeof(s->nc.info_str), "fd=%d", fd);
 628     } else if (tap->has_helper) {
 629         snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s",
 630                  tap->helper);
 631     } else {
 632         snprintf(s->nc.info_str, sizeof(s->nc.info_str),
 633                  "ifname=%s,script=%s,downscript=%s", ifname, script,
 634                  downscript);
 635
 636         if (strcmp(downscript, "no") != 0) {
 637             snprintf(s->down_script, sizeof(s->down_script), "%s", downscript);
 638             snprintf(s->down_script_arg, sizeof(s->down_script_arg),
 639                      "%s", ifname);
 640         }
 641     }
 642
 643     if (tap->has_vhost ? tap->vhost :
 644         vhostfdname || (tap->has_vhostforce && tap->vhostforce)) {
 645         VhostNetOptions options;
 646
 647         options.backend_type = VHOST_BACKEND_TYPE_KERNEL;
 648         options.net_backend = &s->nc;
 649         options.force = tap->has_vhostforce && tap->vhostforce;
 650
 651         if (tap->has_vhostfd || tap->has_vhostfds) {
 652             vhostfd = monitor_fd_param(cur_mon, vhostfdname, &err);
 653             if (vhostfd == -1) {
 654                 error_propagate(errp, err);
 655                 return;
 656             }
 657         } else {
 658             vhostfd = open("/dev/vhost-net", O_RDWR);
 659             if (vhostfd < 0) {
 660                 error_setg_errno(errp, errno,
 661                                  "tap: open vhost char device failed");
 662                 return;
 663             }
 664         }
 665         options.opaque = (void *)(uintptr_t)vhostfd;
 666
 667         s->vhost_net = vhost_net_init(&options);
 668         if (!s->vhost_net) {
 669             error_setg(errp,
 670                        "vhost-net requested but could not be initialized");
 671             return;
 672         }
 673     } else if (tap->has_vhostfd || tap->has_vhostfds) {
 674         error_setg(errp, "vhostfd= is not valid without vhost");
 675     }
 676 }
 677
 678 static int get_fds(char *str, char *fds[], int max)
 679 {
 680     char *ptr = str, *this;
 681     size_t len = strlen(str);
 682     int i = 0;
 683
 684     while (i < max && ptr < str + len) {
 685         this = strchr(ptr, ':');
 686
 687         if (this == NULL) {
 688             fds[i] = g_strdup(ptr);
 689         } else {
 690             fds[i] = g_strndup(ptr, this - ptr);
 691         }
 692
 693         i++;
 694         if (this == NULL) {
 695             break;
 696         } else {
 697             ptr = this + 1;
 698         }
 699     }
 700
 701     return i;
 702 }
 703
 704 int net_init_tap(const NetClientOptions *opts, const char *name,
 705                  NetClientState *peer, Error **errp)
 706 {
 707     const NetdevTapOptions *tap;
 708     int fd, vnet_hdr = 0, i = 0, queues;
 709     /* for the no-fd, no-helper case */
 710     const char *script = NULL; /* suppress wrong "uninit'd use" gcc warning */
 711     const char *downscript = NULL;
 712     Error *err = NULL;
 713     const char *vhostfdname;
 714     char ifname[128];
 715
 716     assert(opts->kind == NET_CLIENT_OPTIONS_KIND_TAP);
 717     tap = opts->tap;
 718     queues = tap->has_queues ? tap->queues : 1;
 719     vhostfdname = tap->has_vhostfd ? tap->vhostfd : NULL;
 720
 721     /* QEMU vlans does not support multiqueue tap, in this case peer is set.
 722      * For -netdev, peer is always NULL. */
 723     if (peer && (tap->has_queues || tap->has_fds || tap->has_vhostfds)) {
 724         error_setg(errp, "Multiqueue tap cannot be used with QEMU vlans");
 725         return -1;
 726     }
 727
 728     if (tap->has_fd) {
 729         if (tap->has_ifname || tap->has_script || tap->has_downscript ||
 730             tap->has_vnet_hdr || tap->has_helper || tap->has_queues ||
 731             tap->has_fds || tap->has_vhostfds) {
 732             error_setg(errp, "ifname=, script=, downscript=, vnet_hdr=, "
 733                        "helper=, queues=, fds=, and vhostfds= "
 734                        "are invalid with fd=");
 735             return -1;
 736         }
 737
 738         fd = monitor_fd_param(cur_mon, tap->fd, &err);
 739         if (fd == -1) {
 740             error_propagate(errp, err);
 741             return -1;
 742         }
 743
 744         fcntl(fd, F_SETFL, O_NONBLOCK);
 745
 746         vnet_hdr = tap_probe_vnet_hdr(fd);
 747
 748         net_init_tap_one(tap, peer, "tap", name, NULL,
 749                          script, downscript,
 750                          vhostfdname, vnet_hdr, fd, &err);
 751         if (err) {
 752             error_propagate(errp, err);
 753             return -1;
 754         }
 755     } else if (tap->has_fds) {
 756         char *fds[MAX_TAP_QUEUES];
 757         char *vhost_fds[MAX_TAP_QUEUES];
 758         int nfds, nvhosts;
 759
 760         if (tap->has_ifname || tap->has_script || tap->has_downscript ||
 761             tap->has_vnet_hdr || tap->has_helper || tap->has_queues ||
 762             tap->has_vhostfd) {
 763             error_setg(errp, "ifname=, script=, downscript=, vnet_hdr=, "
 764                        "helper=, queues=, and vhostfd= "
 765                        "are invalid with fds=");
 766             return -1;
 767         }
 768
 769         nfds = get_fds(tap->fds, fds, MAX_TAP_QUEUES);
 770         if (tap->has_vhostfds) {
 771             nvhosts = get_fds(tap->vhostfds, vhost_fds, MAX_TAP_QUEUES);
 772             if (nfds != nvhosts) {
 773                 error_setg(errp, "The number of fds passed does not match "
 774                            "the number of vhostfds passed");
 775                 return -1;
 776             }
 777         }
 778
 779         for (i = 0; i < nfds; i++) {
 780             fd = monitor_fd_param(cur_mon, fds[i], &err);
 781             if (fd == -1) {
 782                 error_propagate(errp, err);
 783                 return -1;
 784             }
 785
 786             fcntl(fd, F_SETFL, O_NONBLOCK);
 787
 788             if (i == 0) {
 789                 vnet_hdr = tap_probe_vnet_hdr(fd);
 790             } else if (vnet_hdr != tap_probe_vnet_hdr(fd)) {
 791                 error_setg(errp,
 792                            "vnet_hdr not consistent across given tap fds");
 793                 return -1;
 794             }
 795
 796             net_init_tap_one(tap, peer, "tap", name, ifname,
 797                              script, downscript,
 798                              tap->has_vhostfds ? vhost_fds[i] : NULL,
 799                              vnet_hdr, fd, &err);
 800             if (err) {
 801                 error_propagate(errp, err);
 802                 return -1;
 803             }
 804         }
 805     } else if (tap->has_helper) {
 806         if (tap->has_ifname || tap->has_script || tap->has_downscript ||
 807             tap->has_vnet_hdr || tap->has_queues || tap->has_vhostfds) {
 808             error_setg(errp, "ifname=, script=, downscript=, vnet_hdr=, "
 809                        "queues=, and vhostfds= are invalid with helper=");
 810             return -1;
 811         }
 812
 813         fd = net_bridge_run_helper(tap->helper, DEFAULT_BRIDGE_INTERFACE,
 814                                    errp);
 815         if (fd == -1) {
 816             return -1;
 817         }
 818
 819         fcntl(fd, F_SETFL, O_NONBLOCK);
 820         vnet_hdr = tap_probe_vnet_hdr(fd);
 821
 822         net_init_tap_one(tap, peer, "bridge", name, ifname,
 823                          script, downscript, vhostfdname,
 824                          vnet_hdr, fd, &err);
 825         if (err) {
 826             error_propagate(errp, err);
 827             close(fd);
 828             return -1;
 829         }
 830     } else {
 831         if (tap->has_vhostfds) {
 832             error_setg(errp, "vhostfds= is invalid if fds= wasn't specified");
 833             return -1;
 834         }
 835         script = tap->has_script ? tap->script : DEFAULT_NETWORK_SCRIPT;
 836         downscript = tap->has_downscript ? tap->downscript :
 837             DEFAULT_NETWORK_DOWN_SCRIPT;
 838
 839         if (tap->has_ifname) {
 840             pstrcpy(ifname, sizeof ifname, tap->ifname);
 841         } else {
 842             ifname[0] = '\0';
 843         }
 844
 845         for (i = 0; i < queues; i++) {
 846             fd = net_tap_init(tap, &vnet_hdr, i >= 1 ? "no" : script,
 847                               ifname, sizeof ifname, queues > 1, errp);
 848             if (fd == -1) {
 849                 return -1;
 850             }
 851
 852             if (queues > 1 && i == 0 && !tap->has_ifname) {
 853                 if (tap_fd_get_ifname(fd, ifname)) {
 854                     error_setg(errp, "Fail to get ifname");
 855                     close(fd);
 856                     return -1;
 857                 }
 858             }
 859
 860             net_init_tap_one(tap, peer, "tap", name, ifname,
 861                              i >= 1 ? "no" : script,
 862                              i >= 1 ? "no" : downscript,
 863                              vhostfdname, vnet_hdr, fd, &err);
 864             if (err) {
 865                 error_propagate(errp, err);
 866                 close(fd);
 867                 return -1;
 868             }
 869         }
 870     }
 871
 872     return 0;
 873 }
 874
 875 VHostNetState *tap_get_vhost_net(NetClientState *nc)
 876 {
 877     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 878     assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
 879     return s->vhost_net;
 880 }
 881
 882 int tap_enable(NetClientState *nc)
 883 {
 884     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 885     int ret;
 886
 887     if (s->enabled) {
 888         return 0;
 889     } else {
 890         ret = tap_fd_enable(s->fd);
 891         if (ret == 0) {
 892             s->enabled = true;
 893             tap_update_fd_handler(s);
 894         }
 895         return ret;
 896     }
 897 }
 898
 899 int tap_disable(NetClientState *nc)
 900 {
 901     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 902     int ret;
 903
 904     if (s->enabled == 0) {
 905         return 0;
 906     } else {
 907         ret = tap_fd_disable(s->fd);
 908         if (ret == 0) {
 909             qemu_purge_queued_packets(nc);
 910             s->enabled = false;
 911             tap_update_fd_handler(s);
 912         }
 913         return ret;
 914     }
 915 }