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 <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.h"
  37 #include "sysemu.h"
  38 #include "qemu-char.h"
  39 #include "qemu-common.h"
  40 #include "qemu-error.h"
  41
  42 #include "net/tap-linux.h"
  43
  44 #include "hw/vhost_net.h"
  45
  46 /* Maximum GSO packet size (64k) plus plenty of room for
  47  * the ethernet and virtio_net headers
  48  */
  49 #define TAP_BUFSIZE (4096 + 65536)
  50
  51 typedef struct TAPState {
  52     VLANClientState nc;
  53     int fd;
  54     char down_script[1024];
  55     char down_script_arg[128];
  56     uint8_t buf[TAP_BUFSIZE];
  57     unsigned int read_poll : 1;
  58     unsigned int write_poll : 1;
  59     unsigned int using_vnet_hdr : 1;
  60     unsigned int has_ufo: 1;
  61     VHostNetState *vhost_net;
  62     unsigned host_vnet_hdr_len;
  63 } TAPState;
  64
  65 static int launch_script(const char *setup_script, const char *ifname, int fd);
  66
  67 static int tap_can_send(void *opaque);
  68 static void tap_send(void *opaque);
  69 static void tap_writable(void *opaque);
  70
  71 static void tap_update_fd_handler(TAPState *s)
  72 {
  73     qemu_set_fd_handler2(s->fd,
  74                          s->read_poll  ? tap_can_send : NULL,
  75                          s->read_poll  ? tap_send     : NULL,
  76                          s->write_poll ? tap_writable : NULL,
  77                          s);
  78 }
  79
  80 static void tap_read_poll(TAPState *s, int enable)
  81 {
  82     s->read_poll = !!enable;
  83     tap_update_fd_handler(s);
  84 }
  85
  86 static void tap_write_poll(TAPState *s, int enable)
  87 {
  88     s->write_poll = !!enable;
  89     tap_update_fd_handler(s);
  90 }
  91
  92 static void tap_writable(void *opaque)
  93 {
  94     TAPState *s = opaque;
  95
  96     tap_write_poll(s, 0);
  97
  98     qemu_flush_queued_packets(&s->nc);
  99 }
 100
 101 static ssize_t tap_write_packet(TAPState *s, const struct iovec *iov, int iovcnt)
 102 {
 103     ssize_t len;
 104
 105     do {
 106         len = writev(s->fd, iov, iovcnt);
 107     } while (len == -1 && errno == EINTR);
 108
 109     if (len == -1 && errno == EAGAIN) {
 110         tap_write_poll(s, 1);
 111         return 0;
 112     }
 113
 114     return len;
 115 }
 116
 117 static ssize_t tap_receive_iov(VLANClientState *nc, const struct iovec *iov,
 118                                int iovcnt)
 119 {
 120     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 121     const struct iovec *iovp = iov;
 122     struct iovec iov_copy[iovcnt + 1];
 123     struct virtio_net_hdr_mrg_rxbuf hdr = { };
 124
 125     if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
 126         iov_copy[0].iov_base = &hdr;
 127         iov_copy[0].iov_len =  s->host_vnet_hdr_len;
 128         memcpy(&iov_copy[1], iov, iovcnt * sizeof(*iov));
 129         iovp = iov_copy;
 130         iovcnt++;
 131     }
 132
 133     return tap_write_packet(s, iovp, iovcnt);
 134 }
 135
 136 static ssize_t tap_receive_raw(VLANClientState *nc, const uint8_t *buf, size_t size)
 137 {
 138     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 139     struct iovec iov[2];
 140     int iovcnt = 0;
 141     struct virtio_net_hdr_mrg_rxbuf hdr = { };
 142
 143     if (s->host_vnet_hdr_len) {
 144         iov[iovcnt].iov_base = &hdr;
 145         iov[iovcnt].iov_len  = s->host_vnet_hdr_len;
 146         iovcnt++;
 147     }
 148
 149     iov[iovcnt].iov_base = (char *)buf;
 150     iov[iovcnt].iov_len  = size;
 151     iovcnt++;
 152
 153     return tap_write_packet(s, iov, iovcnt);
 154 }
 155
 156 static ssize_t tap_receive(VLANClientState *nc, const uint8_t *buf, size_t size)
 157 {
 158     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 159     struct iovec iov[1];
 160
 161     if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
 162         return tap_receive_raw(nc, buf, size);
 163     }
 164
 165     iov[0].iov_base = (char *)buf;
 166     iov[0].iov_len  = size;
 167
 168     return tap_write_packet(s, iov, 1);
 169 }
 170
 171 static int tap_can_send(void *opaque)
 172 {
 173     TAPState *s = opaque;
 174
 175     return qemu_can_send_packet(&s->nc);
 176 }
 177
 178 #ifndef __sun__
 179 ssize_t tap_read_packet(int tapfd, uint8_t *buf, int maxlen)
 180 {
 181     return read(tapfd, buf, maxlen);
 182 }
 183 #endif
 184
 185 static void tap_send_completed(VLANClientState *nc, ssize_t len)
 186 {
 187     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 188     tap_read_poll(s, 1);
 189 }
 190
 191 static void tap_send(void *opaque)
 192 {
 193     TAPState *s = opaque;
 194     int size;
 195
 196     do {
 197         uint8_t *buf = s->buf;
 198
 199         size = tap_read_packet(s->fd, s->buf, sizeof(s->buf));
 200         if (size <= 0) {
 201             break;
 202         }
 203
 204         if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
 205             buf  += s->host_vnet_hdr_len;
 206             size -= s->host_vnet_hdr_len;
 207         }
 208
 209         size = qemu_send_packet_async(&s->nc, buf, size, tap_send_completed);
 210         if (size == 0) {
 211             tap_read_poll(s, 0);
 212         }
 213     } while (size > 0 && qemu_can_send_packet(&s->nc));
 214 }
 215
 216 int tap_has_ufo(VLANClientState *nc)
 217 {
 218     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 219
 220     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 221
 222     return s->has_ufo;
 223 }
 224
 225 int tap_has_vnet_hdr(VLANClientState *nc)
 226 {
 227     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 228
 229     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 230
 231     return !!s->host_vnet_hdr_len;
 232 }
 233
 234 int tap_has_vnet_hdr_len(VLANClientState *nc, int len)
 235 {
 236     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 237
 238     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 239
 240     return tap_probe_vnet_hdr_len(s->fd, len);
 241 }
 242
 243 void tap_set_vnet_hdr_len(VLANClientState *nc, int len)
 244 {
 245     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 246
 247     assert(nc->info->type == NET_CLIENT_TYPE_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 void tap_using_vnet_hdr(VLANClientState *nc, int using_vnet_hdr)
 256 {
 257     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 258
 259     using_vnet_hdr = using_vnet_hdr != 0;
 260
 261     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 262     assert(!!s->host_vnet_hdr_len == using_vnet_hdr);
 263
 264     s->using_vnet_hdr = using_vnet_hdr;
 265 }
 266
 267 void tap_set_offload(VLANClientState *nc, int csum, int tso4,
 268                      int tso6, int ecn, int ufo)
 269 {
 270     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 271     if (s->fd < 0) {
 272         return;
 273     }
 274
 275     tap_fd_set_offload(s->fd, csum, tso4, tso6, ecn, ufo);
 276 }
 277
 278 static void tap_cleanup(VLANClientState *nc)
 279 {
 280     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 281
 282     if (s->vhost_net) {
 283         vhost_net_cleanup(s->vhost_net);
 284         s->vhost_net = NULL;
 285     }
 286
 287     qemu_purge_queued_packets(nc);
 288
 289     if (s->down_script[0])
 290         launch_script(s->down_script, s->down_script_arg, s->fd);
 291
 292     tap_read_poll(s, 0);
 293     tap_write_poll(s, 0);
 294     close(s->fd);
 295     s->fd = -1;
 296 }
 297
 298 static void tap_poll(VLANClientState *nc, bool enable)
 299 {
 300     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 301     tap_read_poll(s, enable);
 302     tap_write_poll(s, enable);
 303 }
 304
 305 int tap_get_fd(VLANClientState *nc)
 306 {
 307     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 308     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 309     return s->fd;
 310 }
 311
 312 /* fd support */
 313
 314 static NetClientInfo net_tap_info = {
 315     .type = NET_CLIENT_TYPE_TAP,
 316     .size = sizeof(TAPState),
 317     .receive = tap_receive,
 318     .receive_raw = tap_receive_raw,
 319     .receive_iov = tap_receive_iov,
 320     .poll = tap_poll,
 321     .cleanup = tap_cleanup,
 322 };
 323
 324 static TAPState *net_tap_fd_init(VLANState *vlan,
 325                                  const char *model,
 326                                  const char *name,
 327                                  int fd,
 328                                  int vnet_hdr)
 329 {
 330     VLANClientState *nc;
 331     TAPState *s;
 332
 333     nc = qemu_new_net_client(&net_tap_info, vlan, NULL, model, name);
 334
 335     s = DO_UPCAST(TAPState, nc, nc);
 336
 337     s->fd = fd;
 338     s->host_vnet_hdr_len = vnet_hdr ? sizeof(struct virtio_net_hdr) : 0;
 339     s->using_vnet_hdr = 0;
 340     s->has_ufo = tap_probe_has_ufo(s->fd);
 341     tap_set_offload(&s->nc, 0, 0, 0, 0, 0);
 342     tap_read_poll(s, 1);
 343     s->vhost_net = NULL;
 344     return s;
 345 }
 346
 347 static int launch_script(const char *setup_script, const char *ifname, int fd)
 348 {
 349     sigset_t oldmask, mask;
 350     int pid, status;
 351     char *args[3];
 352     char **parg;
 353
 354     sigemptyset(&mask);
 355     sigaddset(&mask, SIGCHLD);
 356     sigprocmask(SIG_BLOCK, &mask, &oldmask);
 357
 358     /* try to launch network script */
 359     pid = fork();
 360     if (pid == 0) {
 361         int open_max = sysconf(_SC_OPEN_MAX), i;
 362
 363         for (i = 0; i < open_max; i++) {
 364             if (i != STDIN_FILENO &&
 365                 i != STDOUT_FILENO &&
 366                 i != STDERR_FILENO &&
 367                 i != fd) {
 368                 close(i);
 369             }
 370         }
 371         parg = args;
 372         *parg++ = (char *)setup_script;
 373         *parg++ = (char *)ifname;
 374         *parg = NULL;
 375         execv(setup_script, args);
 376         _exit(1);
 377     } else if (pid > 0) {
 378         while (waitpid(pid, &status, 0) != pid) {
 379             /* loop */
 380         }
 381         sigprocmask(SIG_SETMASK, &oldmask, NULL);
 382
 383         if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
 384             return 0;
 385         }
 386     }
 387     fprintf(stderr, "%s: could not launch network script\n", setup_script);
 388     return -1;
 389 }
 390
 391 static int net_tap_init(QemuOpts *opts, int *vnet_hdr)
 392 {
 393     int fd, vnet_hdr_required;
 394     char ifname[128] = {0,};
 395     const char *setup_script;
 396
 397     if (qemu_opt_get(opts, "ifname")) {
 398         pstrcpy(ifname, sizeof(ifname), qemu_opt_get(opts, "ifname"));
 399     }
 400
 401     *vnet_hdr = qemu_opt_get_bool(opts, "vnet_hdr", 1);
 402     if (qemu_opt_get(opts, "vnet_hdr")) {
 403         vnet_hdr_required = *vnet_hdr;
 404     } else {
 405         vnet_hdr_required = 0;
 406     }
 407
 408     TFR(fd = tap_open(ifname, sizeof(ifname), vnet_hdr, vnet_hdr_required));
 409     if (fd < 0) {
 410         return -1;
 411     }
 412
 413     setup_script = qemu_opt_get(opts, "script");
 414     if (setup_script &&
 415         setup_script[0] != '\0' &&
 416         strcmp(setup_script, "no") != 0 &&
 417         launch_script(setup_script, ifname, fd)) {
 418         close(fd);
 419         return -1;
 420     }
 421
 422     qemu_opt_set(opts, "ifname", ifname);
 423
 424     return fd;
 425 }
 426
 427 int net_init_tap(QemuOpts *opts, Monitor *mon, const char *name, VLANState *vlan)
 428 {
 429     TAPState *s;
 430     int fd, vnet_hdr = 0;
 431
 432     if (qemu_opt_get(opts, "fd")) {
 433         if (qemu_opt_get(opts, "ifname") ||
 434             qemu_opt_get(opts, "script") ||
 435             qemu_opt_get(opts, "downscript") ||
 436             qemu_opt_get(opts, "vnet_hdr")) {
 437             error_report("ifname=, script=, downscript= and vnet_hdr= is invalid with fd=");
 438             return -1;
 439         }
 440
 441         fd = net_handle_fd_param(mon, qemu_opt_get(opts, "fd"));
 442         if (fd == -1) {
 443             return -1;
 444         }
 445
 446         fcntl(fd, F_SETFL, O_NONBLOCK);
 447
 448         vnet_hdr = tap_probe_vnet_hdr(fd);
 449     } else {
 450         if (!qemu_opt_get(opts, "script")) {
 451             qemu_opt_set(opts, "script", DEFAULT_NETWORK_SCRIPT);
 452         }
 453
 454         if (!qemu_opt_get(opts, "downscript")) {
 455             qemu_opt_set(opts, "downscript", DEFAULT_NETWORK_DOWN_SCRIPT);
 456         }
 457
 458         fd = net_tap_init(opts, &vnet_hdr);
 459         if (fd == -1) {
 460             return -1;
 461         }
 462     }
 463
 464     s = net_tap_fd_init(vlan, "tap", name, fd, vnet_hdr);
 465     if (!s) {
 466         close(fd);
 467         return -1;
 468     }
 469
 470     if (tap_set_sndbuf(s->fd, opts) < 0) {
 471         return -1;
 472     }
 473
 474     if (qemu_opt_get(opts, "fd")) {
 475         snprintf(s->nc.info_str, sizeof(s->nc.info_str), "fd=%d", fd);
 476     } else {
 477         const char *ifname, *script, *downscript;
 478
 479         ifname     = qemu_opt_get(opts, "ifname");
 480         script     = qemu_opt_get(opts, "script");
 481         downscript = qemu_opt_get(opts, "downscript");
 482
 483         snprintf(s->nc.info_str, sizeof(s->nc.info_str),
 484                  "ifname=%s,script=%s,downscript=%s",
 485                  ifname, script, downscript);
 486
 487         if (strcmp(downscript, "no") != 0) {
 488             snprintf(s->down_script, sizeof(s->down_script), "%s", downscript);
 489             snprintf(s->down_script_arg, sizeof(s->down_script_arg), "%s", ifname);
 490         }
 491     }
 492
 493     if (qemu_opt_get_bool(opts, "vhost", !!qemu_opt_get(opts, "vhostfd") ||
 494                           qemu_opt_get_bool(opts, "vhostforce", false))) {
 495         int vhostfd, r;
 496         bool force = qemu_opt_get_bool(opts, "vhostforce", false);
 497         if (qemu_opt_get(opts, "vhostfd")) {
 498             r = net_handle_fd_param(mon, qemu_opt_get(opts, "vhostfd"));
 499             if (r == -1) {
 500                 return -1;
 501             }
 502             vhostfd = r;
 503         } else {
 504             vhostfd = -1;
 505         }
 506         s->vhost_net = vhost_net_init(&s->nc, vhostfd, force);
 507         if (!s->vhost_net) {
 508             error_report("vhost-net requested but could not be initialized");
 509             return -1;
 510         }
 511     } else if (qemu_opt_get(opts, "vhostfd")) {
 512         error_report("vhostfd= is not valid without vhost");
 513         return -1;
 514     }
 515
 516     return 0;
 517 }
 518
 519 VHostNetState *tap_get_vhost_net(VLANClientState *nc)
 520 {
 521     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 522     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 523     return s->vhost_net;
 524 }