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     if (s->fd < 0) {
 273         return;
 274     }
 275
 276     tap_fd_set_offload(s->fd, csum, tso4, tso6, ecn, ufo);
 277 }
 278
 279 static void tap_cleanup(VLANClientState *nc)
 280 {
 281     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 282
 283     if (s->vhost_net) {
 284         vhost_net_cleanup(s->vhost_net);
 285         s->vhost_net = NULL;
 286     }
 287
 288     qemu_purge_queued_packets(nc);
 289
 290     if (s->down_script[0])
 291         launch_script(s->down_script, s->down_script_arg, s->fd);
 292
 293     tap_read_poll(s, 0);
 294     tap_write_poll(s, 0);
 295     close(s->fd);
 296     s->fd = -1;
 297 }
 298
 299 static void tap_poll(VLANClientState *nc, bool enable)
 300 {
 301     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 302     tap_read_poll(s, enable);
 303     tap_write_poll(s, enable);
 304 }
 305
 306 int tap_get_fd(VLANClientState *nc)
 307 {
 308     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 309     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 310     return s->fd;
 311 }
 312
 313 /* fd support */
 314
 315 static NetClientInfo net_tap_info = {
 316     .type = NET_CLIENT_TYPE_TAP,
 317     .size = sizeof(TAPState),
 318     .receive = tap_receive,
 319     .receive_raw = tap_receive_raw,
 320     .receive_iov = tap_receive_iov,
 321     .poll = tap_poll,
 322     .cleanup = tap_cleanup,
 323 };
 324
 325 static TAPState *net_tap_fd_init(VLANState *vlan,
 326                                  const char *model,
 327                                  const char *name,
 328                                  int fd,
 329                                  int vnet_hdr)
 330 {
 331     VLANClientState *nc;
 332     TAPState *s;
 333
 334     nc = qemu_new_net_client(&net_tap_info, vlan, NULL, model, name);
 335
 336     s = DO_UPCAST(TAPState, nc, nc);
 337
 338     s->fd = fd;
 339     s->host_vnet_hdr_len = vnet_hdr ? sizeof(struct virtio_net_hdr) : 0;
 340     s->using_vnet_hdr = 0;
 341     s->has_ufo = tap_probe_has_ufo(s->fd);
 342     tap_set_offload(&s->nc, 0, 0, 0, 0, 0);
 343     tap_read_poll(s, 1);
 344     s->vhost_net = NULL;
 345     return s;
 346 }
 347
 348 static int launch_script(const char *setup_script, const char *ifname, int fd)
 349 {
 350     sigset_t oldmask, mask;
 351     int pid, status;
 352     char *args[3];
 353     char **parg;
 354
 355     sigemptyset(&mask);
 356     sigaddset(&mask, SIGCHLD);
 357     sigprocmask(SIG_BLOCK, &mask, &oldmask);
 358
 359     /* try to launch network script */
 360     pid = fork();
 361     if (pid == 0) {
 362         int open_max = sysconf(_SC_OPEN_MAX), i;
 363
 364         for (i = 0; i < open_max; i++) {
 365             if (i != STDIN_FILENO &&
 366                 i != STDOUT_FILENO &&
 367                 i != STDERR_FILENO &&
 368                 i != fd) {
 369                 close(i);
 370             }
 371         }
 372         parg = args;
 373         *parg++ = (char *)setup_script;
 374         *parg++ = (char *)ifname;
 375         *parg = NULL;
 376         execv(setup_script, args);
 377         _exit(1);
 378     } else if (pid > 0) {
 379         while (waitpid(pid, &status, 0) != pid) {
 380             /* loop */
 381         }
 382         sigprocmask(SIG_SETMASK, &oldmask, NULL);
 383
 384         if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
 385             return 0;
 386         }
 387     }
 388     fprintf(stderr, "%s: could not launch network script\n", setup_script);
 389     return -1;
 390 }
 391
 392 static int net_tap_init(QemuOpts *opts, int *vnet_hdr)
 393 {
 394     int fd, vnet_hdr_required;
 395     char ifname[128] = {0,};
 396     const char *setup_script;
 397
 398     if (qemu_opt_get(opts, "ifname")) {
 399         pstrcpy(ifname, sizeof(ifname), qemu_opt_get(opts, "ifname"));
 400     }
 401
 402     *vnet_hdr = qemu_opt_get_bool(opts, "vnet_hdr", 1);
 403     if (qemu_opt_get(opts, "vnet_hdr")) {
 404         vnet_hdr_required = *vnet_hdr;
 405     } else {
 406         vnet_hdr_required = 0;
 407     }
 408
 409     TFR(fd = tap_open(ifname, sizeof(ifname), vnet_hdr, vnet_hdr_required));
 410     if (fd < 0) {
 411         return -1;
 412     }
 413
 414     setup_script = qemu_opt_get(opts, "script");
 415     if (setup_script &&
 416         setup_script[0] != '\0' &&
 417         strcmp(setup_script, "no") != 0 &&
 418         launch_script(setup_script, ifname, fd)) {
 419         close(fd);
 420         return -1;
 421     }
 422
 423     qemu_opt_set(opts, "ifname", ifname);
 424
 425     return fd;
 426 }
 427
 428 int net_init_tap(QemuOpts *opts, Monitor *mon, const char *name, VLANState *vlan)
 429 {
 430     TAPState *s;
 431     int fd, vnet_hdr = 0;
 432
 433     if (qemu_opt_get(opts, "fd")) {
 434         if (qemu_opt_get(opts, "ifname") ||
 435             qemu_opt_get(opts, "script") ||
 436             qemu_opt_get(opts, "downscript") ||
 437             qemu_opt_get(opts, "vnet_hdr")) {
 438             error_report("ifname=, script=, downscript= and vnet_hdr= is invalid with fd=");
 439             return -1;
 440         }
 441
 442         fd = net_handle_fd_param(mon, qemu_opt_get(opts, "fd"));
 443         if (fd == -1) {
 444             return -1;
 445         }
 446
 447         fcntl(fd, F_SETFL, O_NONBLOCK);
 448
 449         vnet_hdr = tap_probe_vnet_hdr(fd);
 450     } else {
 451         if (!qemu_opt_get(opts, "script")) {
 452             qemu_opt_set(opts, "script", DEFAULT_NETWORK_SCRIPT);
 453         }
 454
 455         if (!qemu_opt_get(opts, "downscript")) {
 456             qemu_opt_set(opts, "downscript", DEFAULT_NETWORK_DOWN_SCRIPT);
 457         }
 458
 459         fd = net_tap_init(opts, &vnet_hdr);
 460         if (fd == -1) {
 461             return -1;
 462         }
 463     }
 464
 465     s = net_tap_fd_init(vlan, "tap", name, fd, vnet_hdr);
 466     if (!s) {
 467         close(fd);
 468         return -1;
 469     }
 470
 471     if (tap_set_sndbuf(s->fd, opts) < 0) {
 472         return -1;
 473     }
 474
 475     if (qemu_opt_get(opts, "fd")) {
 476         snprintf(s->nc.info_str, sizeof(s->nc.info_str), "fd=%d", fd);
 477     } else {
 478         const char *ifname, *script, *downscript;
 479
 480         ifname     = qemu_opt_get(opts, "ifname");
 481         script     = qemu_opt_get(opts, "script");
 482         downscript = qemu_opt_get(opts, "downscript");
 483
 484         snprintf(s->nc.info_str, sizeof(s->nc.info_str),
 485                  "ifname=%s,script=%s,downscript=%s",
 486                  ifname, script, downscript);
 487
 488         if (strcmp(downscript, "no") != 0) {
 489             snprintf(s->down_script, sizeof(s->down_script), "%s", downscript);
 490             snprintf(s->down_script_arg, sizeof(s->down_script_arg), "%s", ifname);
 491         }
 492     }
 493
 494     if (qemu_opt_get_bool(opts, "vhost", !!qemu_opt_get(opts, "vhostfd"))) {
 495         int vhostfd, r;
 496         if (qemu_opt_get(opts, "vhostfd")) {
 497             r = net_handle_fd_param(mon, qemu_opt_get(opts, "vhostfd"));
 498             if (r == -1) {
 499                 return -1;
 500             }
 501             vhostfd = r;
 502         } else {
 503             vhostfd = -1;
 504         }
 505         s->vhost_net = vhost_net_init(&s->nc, vhostfd);
 506         if (!s->vhost_net) {
 507             error_report("vhost-net requested but could not be initialized");
 508             return -1;
 509         }
 510     } else if (qemu_opt_get(opts, "vhostfd")) {
 511         error_report("vhostfd= is not valid without vhost");
 512         return -1;
 513     }
 514
 515     return 0;
 516 }
 517
 518 VHostNetState *tap_get_vhost_net(VLANClientState *nc)
 519 {
 520     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 521     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 522     return s->vhost_net;
 523 }