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 "net/tap.h"
  27
  28 #include "config-host.h"
  29
  30 #include <signal.h>
  31 #include <sys/ioctl.h>
  32 #include <sys/stat.h>
  33 #include <sys/wait.h>
  34 #include <sys/socket.h>
  35 #include <net/if.h>
  36
  37 #include "net.h"
  38 #include "sysemu.h"
  39 #include "qemu-char.h"
  40 #include "qemu-common.h"
  41 #include "qemu-error.h"
  42
  43 #include "net/tap-linux.h"
  44
  45 #include "hw/vhost_net.h"
  46
  47 /* Maximum GSO packet size (64k) plus plenty of room for
  48  * the ethernet and virtio_net headers
  49  */
  50 #define TAP_BUFSIZE (4096 + 65536)
  51
  52 typedef struct TAPState {
  53     VLANClientState nc;
  54     int fd;
  55     char down_script[1024];
  56     char down_script_arg[128];
  57     uint8_t buf[TAP_BUFSIZE];
  58     unsigned int read_poll : 1;
  59     unsigned int write_poll : 1;
  60     unsigned int using_vnet_hdr : 1;
  61     unsigned int has_ufo: 1;
  62     VHostNetState *vhost_net;
  63     unsigned host_vnet_hdr_len;
  64 } TAPState;
  65
  66 static int launch_script(const char *setup_script, const char *ifname, int fd);
  67
  68 static int tap_can_send(void *opaque);
  69 static void tap_send(void *opaque);
  70 static void tap_writable(void *opaque);
  71
  72 static void tap_update_fd_handler(TAPState *s)
  73 {
  74     qemu_set_fd_handler2(s->fd,
  75                          s->read_poll  ? tap_can_send : NULL,
  76                          s->read_poll  ? tap_send     : NULL,
  77                          s->write_poll ? tap_writable : NULL,
  78                          s);
  79 }
  80
  81 static void tap_read_poll(TAPState *s, int enable)
  82 {
  83     s->read_poll = !!enable;
  84     tap_update_fd_handler(s);
  85 }
  86
  87 static void tap_write_poll(TAPState *s, int enable)
  88 {
  89     s->write_poll = !!enable;
  90     tap_update_fd_handler(s);
  91 }
  92
  93 static void tap_writable(void *opaque)
  94 {
  95     TAPState *s = opaque;
  96
  97     tap_write_poll(s, 0);
  98
  99     qemu_flush_queued_packets(&s->nc);
 100 }
 101
 102 static ssize_t tap_write_packet(TAPState *s, const struct iovec *iov, int iovcnt)
 103 {
 104     ssize_t len;
 105
 106     do {
 107         len = writev(s->fd, iov, iovcnt);
 108     } while (len == -1 && errno == EINTR);
 109
 110     if (len == -1 && errno == EAGAIN) {
 111         tap_write_poll(s, 1);
 112         return 0;
 113     }
 114
 115     return len;
 116 }
 117
 118 static ssize_t tap_receive_iov(VLANClientState *nc, const struct iovec *iov,
 119                                int iovcnt)
 120 {
 121     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 122     const struct iovec *iovp = iov;
 123     struct iovec iov_copy[iovcnt + 1];
 124     struct virtio_net_hdr_mrg_rxbuf hdr = { };
 125
 126     if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
 127         iov_copy[0].iov_base = &hdr;
 128         iov_copy[0].iov_len =  s->host_vnet_hdr_len;
 129         memcpy(&iov_copy[1], iov, iovcnt * sizeof(*iov));
 130         iovp = iov_copy;
 131         iovcnt++;
 132     }
 133
 134     return tap_write_packet(s, iovp, iovcnt);
 135 }
 136
 137 static ssize_t tap_receive_raw(VLANClientState *nc, const uint8_t *buf, size_t size)
 138 {
 139     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 140     struct iovec iov[2];
 141     int iovcnt = 0;
 142     struct virtio_net_hdr_mrg_rxbuf hdr = { };
 143
 144     if (s->host_vnet_hdr_len) {
 145         iov[iovcnt].iov_base = &hdr;
 146         iov[iovcnt].iov_len  = s->host_vnet_hdr_len;
 147         iovcnt++;
 148     }
 149
 150     iov[iovcnt].iov_base = (char *)buf;
 151     iov[iovcnt].iov_len  = size;
 152     iovcnt++;
 153
 154     return tap_write_packet(s, iov, iovcnt);
 155 }
 156
 157 static ssize_t tap_receive(VLANClientState *nc, const uint8_t *buf, size_t size)
 158 {
 159     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 160     struct iovec iov[1];
 161
 162     if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
 163         return tap_receive_raw(nc, buf, size);
 164     }
 165
 166     iov[0].iov_base = (char *)buf;
 167     iov[0].iov_len  = size;
 168
 169     return tap_write_packet(s, iov, 1);
 170 }
 171
 172 static int tap_can_send(void *opaque)
 173 {
 174     TAPState *s = opaque;
 175
 176     return qemu_can_send_packet(&s->nc);
 177 }
 178
 179 #ifndef __sun__
 180 ssize_t tap_read_packet(int tapfd, uint8_t *buf, int maxlen)
 181 {
 182     return read(tapfd, buf, maxlen);
 183 }
 184 #endif
 185
 186 static void tap_send_completed(VLANClientState *nc, ssize_t len)
 187 {
 188     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 189     tap_read_poll(s, 1);
 190 }
 191
 192 static void tap_send(void *opaque)
 193 {
 194     TAPState *s = opaque;
 195     int size;
 196
 197     do {
 198         uint8_t *buf = s->buf;
 199
 200         size = tap_read_packet(s->fd, s->buf, sizeof(s->buf));
 201         if (size <= 0) {
 202             break;
 203         }
 204
 205         if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
 206             buf  += s->host_vnet_hdr_len;
 207             size -= s->host_vnet_hdr_len;
 208         }
 209
 210         size = qemu_send_packet_async(&s->nc, buf, size, tap_send_completed);
 211         if (size == 0) {
 212             tap_read_poll(s, 0);
 213         }
 214     } while (size > 0 && qemu_can_send_packet(&s->nc));
 215 }
 216
 217 int tap_has_ufo(VLANClientState *nc)
 218 {
 219     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 220
 221     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 222
 223     return s->has_ufo;
 224 }
 225
 226 int tap_has_vnet_hdr(VLANClientState *nc)
 227 {
 228     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 229
 230     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 231
 232     return !!s->host_vnet_hdr_len;
 233 }
 234
 235 int tap_has_vnet_hdr_len(VLANClientState *nc, int len)
 236 {
 237     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 238
 239     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 240
 241     return tap_probe_vnet_hdr_len(s->fd, len);
 242 }
 243
 244 void tap_set_vnet_hdr_len(VLANClientState *nc, int len)
 245 {
 246     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 247
 248     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 249     assert(len == sizeof(struct virtio_net_hdr_mrg_rxbuf) ||
 250            len == sizeof(struct virtio_net_hdr));
 251
 252     tap_fd_set_vnet_hdr_len(s->fd, len);
 253     s->host_vnet_hdr_len = len;
 254 }
 255
 256 void tap_using_vnet_hdr(VLANClientState *nc, int using_vnet_hdr)
 257 {
 258     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 259
 260     using_vnet_hdr = using_vnet_hdr != 0;
 261
 262     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 263     assert(!!s->host_vnet_hdr_len == using_vnet_hdr);
 264
 265     s->using_vnet_hdr = using_vnet_hdr;
 266 }
 267
 268 void tap_set_offload(VLANClientState *nc, int csum, int tso4,
 269                      int tso6, int ecn, int ufo)
 270 {
 271     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 272
 273     return tap_fd_set_offload(s->fd, csum, tso4, tso6, ecn, ufo);
 274 }
 275
 276 static void tap_cleanup(VLANClientState *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, 0);
 291     tap_write_poll(s, 0);
 292     close(s->fd);
 293 }
 294
 295 static void tap_poll(VLANClientState *nc, bool enable)
 296 {
 297     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 298     tap_read_poll(s, enable);
 299     tap_write_poll(s, enable);
 300 }
 301
 302 int tap_get_fd(VLANClientState *nc)
 303 {
 304     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 305     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 306     return s->fd;
 307 }
 308
 309 /* fd support */
 310
 311 static NetClientInfo net_tap_info = {
 312     .type = NET_CLIENT_TYPE_TAP,
 313     .size = sizeof(TAPState),
 314     .receive = tap_receive,
 315     .receive_raw = tap_receive_raw,
 316     .receive_iov = tap_receive_iov,
 317     .poll = tap_poll,
 318     .cleanup = tap_cleanup,
 319 };
 320
 321 static TAPState *net_tap_fd_init(VLANState *vlan,
 322                                  const char *model,
 323                                  const char *name,
 324                                  int fd,
 325                                  int vnet_hdr)
 326 {
 327     VLANClientState *nc;
 328     TAPState *s;
 329
 330     nc = qemu_new_net_client(&net_tap_info, vlan, NULL, model, name);
 331
 332     s = DO_UPCAST(TAPState, nc, nc);
 333
 334     s->fd = fd;
 335     s->host_vnet_hdr_len = vnet_hdr ? sizeof(struct virtio_net_hdr) : 0;
 336     s->using_vnet_hdr = 0;
 337     s->has_ufo = tap_probe_has_ufo(s->fd);
 338     tap_set_offload(&s->nc, 0, 0, 0, 0, 0);
 339     tap_read_poll(s, 1);
 340     s->vhost_net = NULL;
 341     return s;
 342 }
 343
 344 static int launch_script(const char *setup_script, const char *ifname, int fd)
 345 {
 346     sigset_t oldmask, mask;
 347     int pid, status;
 348     char *args[3];
 349     char **parg;
 350
 351     sigemptyset(&mask);
 352     sigaddset(&mask, SIGCHLD);
 353     sigprocmask(SIG_BLOCK, &mask, &oldmask);
 354
 355     /* try to launch network script */
 356     pid = fork();
 357     if (pid == 0) {
 358         int open_max = sysconf(_SC_OPEN_MAX), i;
 359
 360         for (i = 0; i < open_max; i++) {
 361             if (i != STDIN_FILENO &&
 362                 i != STDOUT_FILENO &&
 363                 i != STDERR_FILENO &&
 364                 i != fd) {
 365                 close(i);
 366             }
 367         }
 368         parg = args;
 369         *parg++ = (char *)setup_script;
 370         *parg++ = (char *)ifname;
 371         *parg = NULL;
 372         execv(setup_script, args);
 373         _exit(1);
 374     } else if (pid > 0) {
 375         while (waitpid(pid, &status, 0) != pid) {
 376             /* loop */
 377         }
 378         sigprocmask(SIG_SETMASK, &oldmask, NULL);
 379
 380         if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
 381             return 0;
 382         }
 383     }
 384     fprintf(stderr, "%s: could not launch network script\n", setup_script);
 385     return -1;
 386 }
 387
 388 static int net_tap_init(QemuOpts *opts, int *vnet_hdr)
 389 {
 390     int fd, vnet_hdr_required;
 391     char ifname[128] = {0,};
 392     const char *setup_script;
 393
 394     if (qemu_opt_get(opts, "ifname")) {
 395         pstrcpy(ifname, sizeof(ifname), qemu_opt_get(opts, "ifname"));
 396     }
 397
 398     *vnet_hdr = qemu_opt_get_bool(opts, "vnet_hdr", 1);
 399     if (qemu_opt_get(opts, "vnet_hdr")) {
 400         vnet_hdr_required = *vnet_hdr;
 401     } else {
 402         vnet_hdr_required = 0;
 403     }
 404
 405     TFR(fd = tap_open(ifname, sizeof(ifname), vnet_hdr, vnet_hdr_required));
 406     if (fd < 0) {
 407         return -1;
 408     }
 409
 410     setup_script = qemu_opt_get(opts, "script");
 411     if (setup_script &&
 412         setup_script[0] != '\0' &&
 413         strcmp(setup_script, "no") != 0 &&
 414         launch_script(setup_script, ifname, fd)) {
 415         close(fd);
 416         return -1;
 417     }
 418
 419     qemu_opt_set(opts, "ifname", ifname);
 420
 421     return fd;
 422 }
 423
 424 int net_init_tap(QemuOpts *opts, Monitor *mon, const char *name, VLANState *vlan)
 425 {
 426     TAPState *s;
 427     int fd, vnet_hdr = 0;
 428
 429     if (qemu_opt_get(opts, "fd")) {
 430         if (qemu_opt_get(opts, "ifname") ||
 431             qemu_opt_get(opts, "script") ||
 432             qemu_opt_get(opts, "downscript") ||
 433             qemu_opt_get(opts, "vnet_hdr")) {
 434             error_report("ifname=, script=, downscript= and vnet_hdr= is invalid with fd=");
 435             return -1;
 436         }
 437
 438         fd = net_handle_fd_param(mon, qemu_opt_get(opts, "fd"));
 439         if (fd == -1) {
 440             return -1;
 441         }
 442
 443         fcntl(fd, F_SETFL, O_NONBLOCK);
 444
 445         vnet_hdr = tap_probe_vnet_hdr(fd);
 446     } else {
 447         if (!qemu_opt_get(opts, "script")) {
 448             qemu_opt_set(opts, "script", DEFAULT_NETWORK_SCRIPT);
 449         }
 450
 451         if (!qemu_opt_get(opts, "downscript")) {
 452             qemu_opt_set(opts, "downscript", DEFAULT_NETWORK_DOWN_SCRIPT);
 453         }
 454
 455         fd = net_tap_init(opts, &vnet_hdr);
 456         if (fd == -1) {
 457             return -1;
 458         }
 459     }
 460
 461     s = net_tap_fd_init(vlan, "tap", name, fd, vnet_hdr);
 462     if (!s) {
 463         close(fd);
 464         return -1;
 465     }
 466
 467     if (tap_set_sndbuf(s->fd, opts) < 0) {
 468         return -1;
 469     }
 470
 471     if (qemu_opt_get(opts, "fd")) {
 472         snprintf(s->nc.info_str, sizeof(s->nc.info_str), "fd=%d", fd);
 473     } else {
 474         const char *ifname, *script, *downscript;
 475
 476         ifname     = qemu_opt_get(opts, "ifname");
 477         script     = qemu_opt_get(opts, "script");
 478         downscript = qemu_opt_get(opts, "downscript");
 479
 480         snprintf(s->nc.info_str, sizeof(s->nc.info_str),
 481                  "ifname=%s,script=%s,downscript=%s",
 482                  ifname, script, downscript);
 483
 484         if (strcmp(downscript, "no") != 0) {
 485             snprintf(s->down_script, sizeof(s->down_script), "%s", downscript);
 486             snprintf(s->down_script_arg, sizeof(s->down_script_arg), "%s", ifname);
 487         }
 488     }
 489
 490     if (qemu_opt_get_bool(opts, "vhost", !!qemu_opt_get(opts, "vhostfd"))) {
 491         int vhostfd, r;
 492         if (qemu_opt_get(opts, "vhostfd")) {
 493             r = net_handle_fd_param(mon, qemu_opt_get(opts, "vhostfd"));
 494             if (r == -1) {
 495                 return -1;
 496             }
 497             vhostfd = r;
 498         } else {
 499             vhostfd = -1;
 500         }
 501         s->vhost_net = vhost_net_init(&s->nc, vhostfd);
 502         if (!s->vhost_net) {
 503             error_report("vhost-net requested but could not be initialized");
 504             return -1;
 505         }
 506     } else if (qemu_opt_get(opts, "vhostfd")) {
 507         error_report("vhostfd= is not valid without vhost");
 508         return -1;
 509     }
 510
 511     return 0;
 512 }
 513
 514 VHostNetState *tap_get_vhost_net(VLANClientState *nc)
 515 {
 516     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 517     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 518     return s->vhost_net;
 519 }