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 has_vnet_hdr : 1;
  61     unsigned int using_vnet_hdr : 1;
  62     unsigned int has_ufo: 1;
  63     VHostNetState *vhost_net;
  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 hdr = { 0, };
 125
 126     if (s->has_vnet_hdr && !s->using_vnet_hdr) {
 127         iov_copy[0].iov_base = &hdr;
 128         iov_copy[0].iov_len =  sizeof(hdr);
 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 hdr = { 0, };
 143
 144     if (s->has_vnet_hdr) {
 145         iov[iovcnt].iov_base = &hdr;
 146         iov[iovcnt].iov_len  = sizeof(hdr);
 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->has_vnet_hdr && !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->has_vnet_hdr && !s->using_vnet_hdr) {
 206             buf  += sizeof(struct virtio_net_hdr);
 207             size -= sizeof(struct virtio_net_hdr);
 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->has_vnet_hdr;
 233 }
 234
 235 void tap_using_vnet_hdr(VLANClientState *nc, int using_vnet_hdr)
 236 {
 237     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 238
 239     using_vnet_hdr = using_vnet_hdr != 0;
 240
 241     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 242     assert(s->has_vnet_hdr == using_vnet_hdr);
 243
 244     s->using_vnet_hdr = using_vnet_hdr;
 245 }
 246
 247 void tap_set_offload(VLANClientState *nc, int csum, int tso4,
 248                      int tso6, int ecn, int ufo)
 249 {
 250     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 251
 252     return tap_fd_set_offload(s->fd, csum, tso4, tso6, ecn, ufo);
 253 }
 254
 255 static void tap_cleanup(VLANClientState *nc)
 256 {
 257     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 258
 259     if (s->vhost_net) {
 260         vhost_net_cleanup(s->vhost_net);
 261     }
 262
 263     qemu_purge_queued_packets(nc);
 264
 265     if (s->down_script[0])
 266         launch_script(s->down_script, s->down_script_arg, s->fd);
 267
 268     tap_read_poll(s, 0);
 269     tap_write_poll(s, 0);
 270     close(s->fd);
 271 }
 272
 273 static void tap_poll(VLANClientState *nc, bool enable)
 274 {
 275     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 276     tap_read_poll(s, enable);
 277     tap_write_poll(s, enable);
 278 }
 279
 280 int tap_get_fd(VLANClientState *nc)
 281 {
 282     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 283     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 284     return s->fd;
 285 }
 286
 287 /* fd support */
 288
 289 static NetClientInfo net_tap_info = {
 290     .type = NET_CLIENT_TYPE_TAP,
 291     .size = sizeof(TAPState),
 292     .receive = tap_receive,
 293     .receive_raw = tap_receive_raw,
 294     .receive_iov = tap_receive_iov,
 295     .poll = tap_poll,
 296     .cleanup = tap_cleanup,
 297 };
 298
 299 static TAPState *net_tap_fd_init(VLANState *vlan,
 300                                  const char *model,
 301                                  const char *name,
 302                                  int fd,
 303                                  int vnet_hdr)
 304 {
 305     VLANClientState *nc;
 306     TAPState *s;
 307
 308     nc = qemu_new_net_client(&net_tap_info, vlan, NULL, model, name);
 309
 310     s = DO_UPCAST(TAPState, nc, nc);
 311
 312     s->fd = fd;
 313     s->has_vnet_hdr = vnet_hdr != 0;
 314     s->using_vnet_hdr = 0;
 315     s->has_ufo = tap_probe_has_ufo(s->fd);
 316     tap_set_offload(&s->nc, 0, 0, 0, 0, 0);
 317     tap_read_poll(s, 1);
 318     s->vhost_net = NULL;
 319     return s;
 320 }
 321
 322 static int launch_script(const char *setup_script, const char *ifname, int fd)
 323 {
 324     sigset_t oldmask, mask;
 325     int pid, status;
 326     char *args[3];
 327     char **parg;
 328
 329     sigemptyset(&mask);
 330     sigaddset(&mask, SIGCHLD);
 331     sigprocmask(SIG_BLOCK, &mask, &oldmask);
 332
 333     /* try to launch network script */
 334     pid = fork();
 335     if (pid == 0) {
 336         int open_max = sysconf(_SC_OPEN_MAX), i;
 337
 338         for (i = 0; i < open_max; i++) {
 339             if (i != STDIN_FILENO &&
 340                 i != STDOUT_FILENO &&
 341                 i != STDERR_FILENO &&
 342                 i != fd) {
 343                 close(i);
 344             }
 345         }
 346         parg = args;
 347         *parg++ = (char *)setup_script;
 348         *parg++ = (char *)ifname;
 349         *parg = NULL;
 350         execv(setup_script, args);
 351         _exit(1);
 352     } else if (pid > 0) {
 353         while (waitpid(pid, &status, 0) != pid) {
 354             /* loop */
 355         }
 356         sigprocmask(SIG_SETMASK, &oldmask, NULL);
 357
 358         if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
 359             return 0;
 360         }
 361     }
 362     fprintf(stderr, "%s: could not launch network script\n", setup_script);
 363     return -1;
 364 }
 365
 366 static int net_tap_init(QemuOpts *opts, int *vnet_hdr)
 367 {
 368     int fd, vnet_hdr_required;
 369     char ifname[128] = {0,};
 370     const char *setup_script;
 371
 372     if (qemu_opt_get(opts, "ifname")) {
 373         pstrcpy(ifname, sizeof(ifname), qemu_opt_get(opts, "ifname"));
 374     }
 375
 376     *vnet_hdr = qemu_opt_get_bool(opts, "vnet_hdr", 1);
 377     if (qemu_opt_get(opts, "vnet_hdr")) {
 378         vnet_hdr_required = *vnet_hdr;
 379     } else {
 380         vnet_hdr_required = 0;
 381     }
 382
 383     TFR(fd = tap_open(ifname, sizeof(ifname), vnet_hdr, vnet_hdr_required));
 384     if (fd < 0) {
 385         return -1;
 386     }
 387
 388     setup_script = qemu_opt_get(opts, "script");
 389     if (setup_script &&
 390         setup_script[0] != '\0' &&
 391         strcmp(setup_script, "no") != 0 &&
 392         launch_script(setup_script, ifname, fd)) {
 393         close(fd);
 394         return -1;
 395     }
 396
 397     qemu_opt_set(opts, "ifname", ifname);
 398
 399     return fd;
 400 }
 401
 402 int net_init_tap(QemuOpts *opts, Monitor *mon, const char *name, VLANState *vlan)
 403 {
 404     TAPState *s;
 405     int fd, vnet_hdr = 0;
 406
 407     if (qemu_opt_get(opts, "fd")) {
 408         if (qemu_opt_get(opts, "ifname") ||
 409             qemu_opt_get(opts, "script") ||
 410             qemu_opt_get(opts, "downscript") ||
 411             qemu_opt_get(opts, "vnet_hdr")) {
 412             error_report("ifname=, script=, downscript= and vnet_hdr= is invalid with fd=");
 413             return -1;
 414         }
 415
 416         fd = net_handle_fd_param(mon, qemu_opt_get(opts, "fd"));
 417         if (fd == -1) {
 418             return -1;
 419         }
 420
 421         fcntl(fd, F_SETFL, O_NONBLOCK);
 422
 423         vnet_hdr = tap_probe_vnet_hdr(fd);
 424     } else {
 425         if (!qemu_opt_get(opts, "script")) {
 426             qemu_opt_set(opts, "script", DEFAULT_NETWORK_SCRIPT);
 427         }
 428
 429         if (!qemu_opt_get(opts, "downscript")) {
 430             qemu_opt_set(opts, "downscript", DEFAULT_NETWORK_DOWN_SCRIPT);
 431         }
 432
 433         fd = net_tap_init(opts, &vnet_hdr);
 434         if (fd == -1) {
 435             return -1;
 436         }
 437     }
 438
 439     s = net_tap_fd_init(vlan, "tap", name, fd, vnet_hdr);
 440     if (!s) {
 441         close(fd);
 442         return -1;
 443     }
 444
 445     if (tap_set_sndbuf(s->fd, opts) < 0) {
 446         return -1;
 447     }
 448
 449     if (qemu_opt_get(opts, "fd")) {
 450         snprintf(s->nc.info_str, sizeof(s->nc.info_str), "fd=%d", fd);
 451     } else {
 452         const char *ifname, *script, *downscript;
 453
 454         ifname     = qemu_opt_get(opts, "ifname");
 455         script     = qemu_opt_get(opts, "script");
 456         downscript = qemu_opt_get(opts, "downscript");
 457
 458         snprintf(s->nc.info_str, sizeof(s->nc.info_str),
 459                  "ifname=%s,script=%s,downscript=%s",
 460                  ifname, script, downscript);
 461
 462         if (strcmp(downscript, "no") != 0) {
 463             snprintf(s->down_script, sizeof(s->down_script), "%s", downscript);
 464             snprintf(s->down_script_arg, sizeof(s->down_script_arg), "%s", ifname);
 465         }
 466     }
 467
 468     if (qemu_opt_get_bool(opts, "vhost", !!qemu_opt_get(opts, "vhostfd"))) {
 469         int vhostfd, r;
 470         if (qemu_opt_get(opts, "vhostfd")) {
 471             r = net_handle_fd_param(mon, qemu_opt_get(opts, "vhostfd"));
 472             if (r == -1) {
 473                 return -1;
 474             }
 475             vhostfd = r;
 476         } else {
 477             vhostfd = -1;
 478         }
 479         s->vhost_net = vhost_net_init(&s->nc, vhostfd);
 480         if (!s->vhost_net) {
 481             error_report("vhost-net requested but could not be initialized");
 482             return -1;
 483         }
 484     } else if (qemu_opt_get(opts, "vhostfd")) {
 485         error_report("vhostfd= is not valid without vhost");
 486         return -1;
 487     }
 488
 489     return 0;
 490 }
 491
 492 VHostNetState *tap_get_vhost_net(VLANClientState *nc)
 493 {
 494     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 495     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 496     return s->vhost_net;
 497 }