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
  42 #include "net/tap-linux.h"
  43
  44 /* Maximum GSO packet size (64k) plus plenty of room for
  45  * the ethernet and virtio_net headers
  46  */
  47 #define TAP_BUFSIZE (4096 + 65536)
  48
  49 typedef struct TAPState {
  50     VLANClientState nc;
  51     int fd;
  52     char down_script[1024];
  53     char down_script_arg[128];
  54     uint8_t buf[TAP_BUFSIZE];
  55     unsigned int read_poll : 1;
  56     unsigned int write_poll : 1;
  57     unsigned int has_vnet_hdr : 1;
  58     unsigned int using_vnet_hdr : 1;
  59     unsigned int has_ufo: 1;
  60 } TAPState;
  61
  62 static int launch_script(const char *setup_script, const char *ifname, int fd);
  63
  64 static int tap_can_send(void *opaque);
  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_handler2(s->fd,
  71                          s->read_poll  ? tap_can_send : NULL,
  72                          s->read_poll  ? tap_send     : NULL,
  73                          s->write_poll ? tap_writable : NULL,
  74                          s);
  75 }
  76
  77 static void tap_read_poll(TAPState *s, int enable)
  78 {
  79     s->read_poll = !!enable;
  80     tap_update_fd_handler(s);
  81 }
  82
  83 static void tap_write_poll(TAPState *s, int enable)
  84 {
  85     s->write_poll = !!enable;
  86     tap_update_fd_handler(s);
  87 }
  88
  89 static void tap_writable(void *opaque)
  90 {
  91     TAPState *s = opaque;
  92
  93     tap_write_poll(s, 0);
  94
  95     qemu_flush_queued_packets(&s->nc);
  96 }
  97
  98 static ssize_t tap_write_packet(TAPState *s, const struct iovec *iov, int iovcnt)
  99 {
 100     ssize_t len;
 101
 102     do {
 103         len = writev(s->fd, iov, iovcnt);
 104     } while (len == -1 && errno == EINTR);
 105
 106     if (len == -1 && errno == EAGAIN) {
 107         tap_write_poll(s, 1);
 108         return 0;
 109     }
 110
 111     return len;
 112 }
 113
 114 static ssize_t tap_receive_iov(VLANClientState *nc, const struct iovec *iov,
 115                                int iovcnt)
 116 {
 117     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 118     const struct iovec *iovp = iov;
 119     struct iovec iov_copy[iovcnt + 1];
 120     struct virtio_net_hdr hdr = { 0, };
 121
 122     if (s->has_vnet_hdr && !s->using_vnet_hdr) {
 123         iov_copy[0].iov_base = &hdr;
 124         iov_copy[0].iov_len =  sizeof(hdr);
 125         memcpy(&iov_copy[1], iov, iovcnt * sizeof(*iov));
 126         iovp = iov_copy;
 127         iovcnt++;
 128     }
 129
 130     return tap_write_packet(s, iovp, iovcnt);
 131 }
 132
 133 static ssize_t tap_receive_raw(VLANClientState *nc, const uint8_t *buf, size_t size)
 134 {
 135     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 136     struct iovec iov[2];
 137     int iovcnt = 0;
 138     struct virtio_net_hdr hdr = { 0, };
 139
 140     if (s->has_vnet_hdr) {
 141         iov[iovcnt].iov_base = &hdr;
 142         iov[iovcnt].iov_len  = sizeof(hdr);
 143         iovcnt++;
 144     }
 145
 146     iov[iovcnt].iov_base = (char *)buf;
 147     iov[iovcnt].iov_len  = size;
 148     iovcnt++;
 149
 150     return tap_write_packet(s, iov, iovcnt);
 151 }
 152
 153 static ssize_t tap_receive(VLANClientState *nc, const uint8_t *buf, size_t size)
 154 {
 155     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 156     struct iovec iov[1];
 157
 158     if (s->has_vnet_hdr && !s->using_vnet_hdr) {
 159         return tap_receive_raw(nc, buf, size);
 160     }
 161
 162     iov[0].iov_base = (char *)buf;
 163     iov[0].iov_len  = size;
 164
 165     return tap_write_packet(s, iov, 1);
 166 }
 167
 168 static int tap_can_send(void *opaque)
 169 {
 170     TAPState *s = opaque;
 171
 172     return qemu_can_send_packet(&s->nc);
 173 }
 174
 175 #ifndef __sun__
 176 ssize_t tap_read_packet(int tapfd, uint8_t *buf, int maxlen)
 177 {
 178     return read(tapfd, buf, maxlen);
 179 }
 180 #endif
 181
 182 static void tap_send_completed(VLANClientState *nc, ssize_t len)
 183 {
 184     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 185     tap_read_poll(s, 1);
 186 }
 187
 188 static void tap_send(void *opaque)
 189 {
 190     TAPState *s = opaque;
 191     int size;
 192
 193     do {
 194         uint8_t *buf = s->buf;
 195
 196         size = tap_read_packet(s->fd, s->buf, sizeof(s->buf));
 197         if (size <= 0) {
 198             break;
 199         }
 200
 201         if (s->has_vnet_hdr && !s->using_vnet_hdr) {
 202             buf  += sizeof(struct virtio_net_hdr);
 203             size -= sizeof(struct virtio_net_hdr);
 204         }
 205
 206         size = qemu_send_packet_async(&s->nc, buf, size, tap_send_completed);
 207         if (size == 0) {
 208             tap_read_poll(s, 0);
 209         }
 210     } while (size > 0 && qemu_can_send_packet(&s->nc));
 211 }
 212
 213 int tap_has_ufo(VLANClientState *nc)
 214 {
 215     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 216
 217     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 218
 219     return s->has_ufo;
 220 }
 221
 222 int tap_has_vnet_hdr(VLANClientState *nc)
 223 {
 224     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 225
 226     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 227
 228     return s->has_vnet_hdr;
 229 }
 230
 231 void tap_using_vnet_hdr(VLANClientState *nc, int using_vnet_hdr)
 232 {
 233     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 234
 235     using_vnet_hdr = using_vnet_hdr != 0;
 236
 237     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 238     assert(s->has_vnet_hdr == using_vnet_hdr);
 239
 240     s->using_vnet_hdr = using_vnet_hdr;
 241 }
 242
 243 void tap_set_offload(VLANClientState *nc, int csum, int tso4,
 244                      int tso6, int ecn, int ufo)
 245 {
 246     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 247
 248     return tap_fd_set_offload(s->fd, csum, tso4, tso6, ecn, ufo);
 249 }
 250
 251 static void tap_cleanup(VLANClientState *nc)
 252 {
 253     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 254
 255     qemu_purge_queued_packets(nc);
 256
 257     if (s->down_script[0])
 258         launch_script(s->down_script, s->down_script_arg, s->fd);
 259
 260     tap_read_poll(s, 0);
 261     tap_write_poll(s, 0);
 262     close(s->fd);
 263 }
 264
 265 static void tap_poll(VLANClientState *nc, bool enable)
 266 {
 267     TAPState *s = DO_UPCAST(TAPState, nc, nc);
 268     tap_read_poll(s, enable);
 269     tap_write_poll(s, enable);
 270 }
 271
 272 /* fd support */
 273
 274 static NetClientInfo net_tap_info = {
 275     .type = NET_CLIENT_TYPE_TAP,
 276     .size = sizeof(TAPState),
 277     .receive = tap_receive,
 278     .receive_raw = tap_receive_raw,
 279     .receive_iov = tap_receive_iov,
 280     .poll = tap_poll,
 281     .cleanup = tap_cleanup,
 282 };
 283
 284 static TAPState *net_tap_fd_init(VLANState *vlan,
 285                                  const char *model,
 286                                  const char *name,
 287                                  int fd,
 288                                  int vnet_hdr)
 289 {
 290     VLANClientState *nc;
 291     TAPState *s;
 292
 293     nc = qemu_new_net_client(&net_tap_info, vlan, NULL, model, name);
 294
 295     s = DO_UPCAST(TAPState, nc, nc);
 296
 297     s->fd = fd;
 298     s->has_vnet_hdr = vnet_hdr != 0;
 299     s->using_vnet_hdr = 0;
 300     s->has_ufo = tap_probe_has_ufo(s->fd);
 301     tap_set_offload(&s->nc, 0, 0, 0, 0, 0);
 302     tap_read_poll(s, 1);
 303     return s;
 304 }
 305
 306 static int launch_script(const char *setup_script, const char *ifname, int fd)
 307 {
 308     sigset_t oldmask, mask;
 309     int pid, status;
 310     char *args[3];
 311     char **parg;
 312
 313     sigemptyset(&mask);
 314     sigaddset(&mask, SIGCHLD);
 315     sigprocmask(SIG_BLOCK, &mask, &oldmask);
 316
 317     /* try to launch network script */
 318     pid = fork();
 319     if (pid == 0) {
 320         int open_max = sysconf(_SC_OPEN_MAX), i;
 321
 322         for (i = 0; i < open_max; i++) {
 323             if (i != STDIN_FILENO &&
 324                 i != STDOUT_FILENO &&
 325                 i != STDERR_FILENO &&
 326                 i != fd) {
 327                 close(i);
 328             }
 329         }
 330         parg = args;
 331         *parg++ = (char *)setup_script;
 332         *parg++ = (char *)ifname;
 333         *parg++ = NULL;
 334         execv(setup_script, args);
 335         _exit(1);
 336     } else if (pid > 0) {
 337         while (waitpid(pid, &status, 0) != pid) {
 338             /* loop */
 339         }
 340         sigprocmask(SIG_SETMASK, &oldmask, NULL);
 341
 342         if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
 343             return 0;
 344         }
 345     }
 346     fprintf(stderr, "%s: could not launch network script\n", setup_script);
 347     return -1;
 348 }
 349
 350 static int net_tap_init(QemuOpts *opts, int *vnet_hdr)
 351 {
 352     int fd, vnet_hdr_required;
 353     char ifname[128] = {0,};
 354     const char *setup_script;
 355
 356     if (qemu_opt_get(opts, "ifname")) {
 357         pstrcpy(ifname, sizeof(ifname), qemu_opt_get(opts, "ifname"));
 358     }
 359
 360     *vnet_hdr = qemu_opt_get_bool(opts, "vnet_hdr", 1);
 361     if (qemu_opt_get(opts, "vnet_hdr")) {
 362         vnet_hdr_required = *vnet_hdr;
 363     } else {
 364         vnet_hdr_required = 0;
 365     }
 366
 367     TFR(fd = tap_open(ifname, sizeof(ifname), vnet_hdr, vnet_hdr_required));
 368     if (fd < 0) {
 369         return -1;
 370     }
 371
 372     setup_script = qemu_opt_get(opts, "script");
 373     if (setup_script &&
 374         setup_script[0] != '\0' &&
 375         strcmp(setup_script, "no") != 0 &&
 376         launch_script(setup_script, ifname, fd)) {
 377         close(fd);
 378         return -1;
 379     }
 380
 381     qemu_opt_set(opts, "ifname", ifname);
 382
 383     return fd;
 384 }
 385
 386 int net_init_tap(QemuOpts *opts, Monitor *mon, const char *name, VLANState *vlan)
 387 {
 388     TAPState *s;
 389     int fd, vnet_hdr = 0;
 390
 391     if (qemu_opt_get(opts, "fd")) {
 392         if (qemu_opt_get(opts, "ifname") ||
 393             qemu_opt_get(opts, "script") ||
 394             qemu_opt_get(opts, "downscript") ||
 395             qemu_opt_get(opts, "vnet_hdr")) {
 396             qemu_error("ifname=, script=, downscript= and vnet_hdr= is invalid with fd=\n");
 397             return -1;
 398         }
 399
 400         fd = net_handle_fd_param(mon, qemu_opt_get(opts, "fd"));
 401         if (fd == -1) {
 402             return -1;
 403         }
 404
 405         fcntl(fd, F_SETFL, O_NONBLOCK);
 406
 407         vnet_hdr = tap_probe_vnet_hdr(fd);
 408     } else {
 409         if (!qemu_opt_get(opts, "script")) {
 410             qemu_opt_set(opts, "script", DEFAULT_NETWORK_SCRIPT);
 411         }
 412
 413         if (!qemu_opt_get(opts, "downscript")) {
 414             qemu_opt_set(opts, "downscript", DEFAULT_NETWORK_DOWN_SCRIPT);
 415         }
 416
 417         fd = net_tap_init(opts, &vnet_hdr);
 418         if (fd == -1) {
 419             return -1;
 420         }
 421     }
 422
 423     s = net_tap_fd_init(vlan, "tap", name, fd, vnet_hdr);
 424     if (!s) {
 425         close(fd);
 426         return -1;
 427     }
 428
 429     if (tap_set_sndbuf(s->fd, opts) < 0) {
 430         return -1;
 431     }
 432
 433     if (qemu_opt_get(opts, "fd")) {
 434         snprintf(s->nc.info_str, sizeof(s->nc.info_str), "fd=%d", fd);
 435     } else {
 436         const char *ifname, *script, *downscript;
 437
 438         ifname     = qemu_opt_get(opts, "ifname");
 439         script     = qemu_opt_get(opts, "script");
 440         downscript = qemu_opt_get(opts, "downscript");
 441
 442         snprintf(s->nc.info_str, sizeof(s->nc.info_str),
 443                  "ifname=%s,script=%s,downscript=%s",
 444                  ifname, script, downscript);
 445
 446         if (strcmp(downscript, "no") != 0) {
 447             snprintf(s->down_script, sizeof(s->down_script), "%s", downscript);
 448             snprintf(s->down_script_arg, sizeof(s->down_script_arg), "%s", ifname);
 449         }
 450     }
 451
 452     if (vlan) {
 453         vlan->nb_host_devs++;
 454     }
 455
 456     return 0;
 457 }