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 <net/if.h>
  35
  36 #include "net.h"
  37 #include "sysemu.h"
  38 #include "qemu-char.h"
  39 #include "qemu-common.h"
  40
  41 #include "net/tap-linux.h"
  42
  43 /* Maximum GSO packet size (64k) plus plenty of room for
  44  * the ethernet and virtio_net headers
  45  */
  46 #define TAP_BUFSIZE (4096 + 65536)
  47
  48 typedef struct TAPState {
  49     VLANClientState *vc;
  50     int fd;
  51     char down_script[1024];
  52     char down_script_arg[128];
  53     uint8_t buf[TAP_BUFSIZE];
  54     unsigned int read_poll : 1;
  55     unsigned int write_poll : 1;
  56     unsigned int has_vnet_hdr : 1;
  57     unsigned int using_vnet_hdr : 1;
  58     unsigned int has_ufo: 1;
  59 } TAPState;
  60
  61 static int launch_script(const char *setup_script, const char *ifname, int fd);
  62
  63 static int tap_can_send(void *opaque);
  64 static void tap_send(void *opaque);
  65 static void tap_writable(void *opaque);
  66
  67 static void tap_update_fd_handler(TAPState *s)
  68 {
  69     qemu_set_fd_handler2(s->fd,
  70                          s->read_poll  ? tap_can_send : NULL,
  71                          s->read_poll  ? tap_send     : NULL,
  72                          s->write_poll ? tap_writable : NULL,
  73                          s);
  74 }
  75
  76 static void tap_read_poll(TAPState *s, int enable)
  77 {
  78     s->read_poll = !!enable;
  79     tap_update_fd_handler(s);
  80 }
  81
  82 static void tap_write_poll(TAPState *s, int enable)
  83 {
  84     s->write_poll = !!enable;
  85     tap_update_fd_handler(s);
  86 }
  87
  88 static void tap_writable(void *opaque)
  89 {
  90     TAPState *s = opaque;
  91
  92     tap_write_poll(s, 0);
  93
  94     qemu_flush_queued_packets(s->vc);
  95 }
  96
  97 static ssize_t tap_write_packet(TAPState *s, const struct iovec *iov, int iovcnt)
  98 {
  99     ssize_t len;
 100
 101     do {
 102         len = writev(s->fd, iov, iovcnt);
 103     } while (len == -1 && errno == EINTR);
 104
 105     if (len == -1 && errno == EAGAIN) {
 106         tap_write_poll(s, 1);
 107         return 0;
 108     }
 109
 110     return len;
 111 }
 112
 113 static ssize_t tap_receive_iov(VLANClientState *vc, const struct iovec *iov,
 114                                int iovcnt)
 115 {
 116     TAPState *s = vc->opaque;
 117     const struct iovec *iovp = iov;
 118     struct iovec iov_copy[iovcnt + 1];
 119     struct virtio_net_hdr hdr = { 0, };
 120
 121     if (s->has_vnet_hdr && !s->using_vnet_hdr) {
 122         iov_copy[0].iov_base = &hdr;
 123         iov_copy[0].iov_len =  sizeof(hdr);
 124         memcpy(&iov_copy[1], iov, iovcnt * sizeof(*iov));
 125         iovp = iov_copy;
 126         iovcnt++;
 127     }
 128
 129     return tap_write_packet(s, iovp, iovcnt);
 130 }
 131
 132 static ssize_t tap_receive_raw(VLANClientState *vc, const uint8_t *buf, size_t size)
 133 {
 134     TAPState *s = vc->opaque;
 135     struct iovec iov[2];
 136     int iovcnt = 0;
 137     struct virtio_net_hdr hdr = { 0, };
 138
 139     if (s->has_vnet_hdr) {
 140         iov[iovcnt].iov_base = &hdr;
 141         iov[iovcnt].iov_len  = sizeof(hdr);
 142         iovcnt++;
 143     }
 144
 145     iov[iovcnt].iov_base = (char *)buf;
 146     iov[iovcnt].iov_len  = size;
 147     iovcnt++;
 148
 149     return tap_write_packet(s, iov, iovcnt);
 150 }
 151
 152 static ssize_t tap_receive(VLANClientState *vc, const uint8_t *buf, size_t size)
 153 {
 154     TAPState *s = vc->opaque;
 155     struct iovec iov[1];
 156
 157     if (s->has_vnet_hdr && !s->using_vnet_hdr) {
 158         return tap_receive_raw(vc, buf, size);
 159     }
 160
 161     iov[0].iov_base = (char *)buf;
 162     iov[0].iov_len  = size;
 163
 164     return tap_write_packet(s, iov, 1);
 165 }
 166
 167 static int tap_can_send(void *opaque)
 168 {
 169     TAPState *s = opaque;
 170
 171     return qemu_can_send_packet(s->vc);
 172 }
 173
 174 #ifndef __sun__
 175 ssize_t tap_read_packet(int tapfd, uint8_t *buf, int maxlen)
 176 {
 177     return read(tapfd, buf, maxlen);
 178 }
 179 #endif
 180
 181 static void tap_send_completed(VLANClientState *vc, ssize_t len)
 182 {
 183     TAPState *s = vc->opaque;
 184     tap_read_poll(s, 1);
 185 }
 186
 187 static void tap_send(void *opaque)
 188 {
 189     TAPState *s = opaque;
 190     int size;
 191
 192     do {
 193         uint8_t *buf = s->buf;
 194
 195         size = tap_read_packet(s->fd, s->buf, sizeof(s->buf));
 196         if (size <= 0) {
 197             break;
 198         }
 199
 200         if (s->has_vnet_hdr && !s->using_vnet_hdr) {
 201             buf  += sizeof(struct virtio_net_hdr);
 202             size -= sizeof(struct virtio_net_hdr);
 203         }
 204
 205         size = qemu_send_packet_async(s->vc, buf, size, tap_send_completed);
 206         if (size == 0) {
 207             tap_read_poll(s, 0);
 208         }
 209     } while (size > 0);
 210 }
 211
 212 int tap_has_ufo(VLANClientState *vc)
 213 {
 214     TAPState *s = vc->opaque;
 215
 216     assert(vc->type == NET_CLIENT_TYPE_TAP);
 217
 218     return s->has_ufo;
 219 }
 220
 221 int tap_has_vnet_hdr(VLANClientState *vc)
 222 {
 223     TAPState *s = vc->opaque;
 224
 225     assert(vc->type == NET_CLIENT_TYPE_TAP);
 226
 227     return s->has_vnet_hdr;
 228 }
 229
 230 void tap_using_vnet_hdr(VLANClientState *vc, int using_vnet_hdr)
 231 {
 232     TAPState *s = vc->opaque;
 233
 234     using_vnet_hdr = using_vnet_hdr != 0;
 235
 236     assert(vc->type == NET_CLIENT_TYPE_TAP);
 237     assert(s->has_vnet_hdr == using_vnet_hdr);
 238
 239     s->using_vnet_hdr = using_vnet_hdr;
 240 }
 241
 242 void tap_set_offload(VLANClientState *vc, int csum, int tso4,
 243                      int tso6, int ecn, int ufo)
 244 {
 245     TAPState *s = vc->opaque;
 246
 247     return tap_fd_set_offload(s->fd, csum, tso4, tso6, ecn, ufo);
 248 }
 249
 250 static void tap_cleanup(VLANClientState *vc)
 251 {
 252     TAPState *s = vc->opaque;
 253
 254     qemu_purge_queued_packets(vc);
 255
 256     if (s->down_script[0])
 257         launch_script(s->down_script, s->down_script_arg, s->fd);
 258
 259     tap_read_poll(s, 0);
 260     tap_write_poll(s, 0);
 261     close(s->fd);
 262     qemu_free(s);
 263 }
 264
 265 /* fd support */
 266
 267 static TAPState *net_tap_fd_init(VLANState *vlan,
 268                                  const char *model,
 269                                  const char *name,
 270                                  int fd,
 271                                  int vnet_hdr)
 272 {
 273     TAPState *s;
 274
 275     s = qemu_mallocz(sizeof(TAPState));
 276     s->fd = fd;
 277     s->has_vnet_hdr = vnet_hdr != 0;
 278     s->using_vnet_hdr = 0;
 279     s->vc = qemu_new_vlan_client(NET_CLIENT_TYPE_TAP,
 280                                  vlan, NULL, model, name, NULL,
 281                                  tap_receive, tap_receive_raw,
 282                                  tap_receive_iov, tap_cleanup, s);
 283     s->has_ufo = tap_probe_has_ufo(s->fd);
 284     tap_set_offload(s->vc, 0, 0, 0, 0, 0);
 285     tap_read_poll(s, 1);
 286     return s;
 287 }
 288
 289 static int launch_script(const char *setup_script, const char *ifname, int fd)
 290 {
 291     sigset_t oldmask, mask;
 292     int pid, status;
 293     char *args[3];
 294     char **parg;
 295
 296     sigemptyset(&mask);
 297     sigaddset(&mask, SIGCHLD);
 298     sigprocmask(SIG_BLOCK, &mask, &oldmask);
 299
 300     /* try to launch network script */
 301     pid = fork();
 302     if (pid == 0) {
 303         int open_max = sysconf(_SC_OPEN_MAX), i;
 304
 305         for (i = 0; i < open_max; i++) {
 306             if (i != STDIN_FILENO &&
 307                 i != STDOUT_FILENO &&
 308                 i != STDERR_FILENO &&
 309                 i != fd) {
 310                 close(i);
 311             }
 312         }
 313         parg = args;
 314         *parg++ = (char *)setup_script;
 315         *parg++ = (char *)ifname;
 316         *parg++ = NULL;
 317         execv(setup_script, args);
 318         _exit(1);
 319     } else if (pid > 0) {
 320         while (waitpid(pid, &status, 0) != pid) {
 321             /* loop */
 322         }
 323         sigprocmask(SIG_SETMASK, &oldmask, NULL);
 324
 325         if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
 326             return 0;
 327         }
 328     }
 329     fprintf(stderr, "%s: could not launch network script\n", setup_script);
 330     return -1;
 331 }
 332
 333 static int net_tap_init(QemuOpts *opts, int *vnet_hdr)
 334 {
 335     int fd, vnet_hdr_required;
 336     char ifname[128] = {0,};
 337     const char *setup_script;
 338
 339     if (qemu_opt_get(opts, "ifname")) {
 340         pstrcpy(ifname, sizeof(ifname), qemu_opt_get(opts, "ifname"));
 341     }
 342
 343     *vnet_hdr = qemu_opt_get_bool(opts, "vnet_hdr", 1);
 344     if (qemu_opt_get(opts, "vnet_hdr")) {
 345         vnet_hdr_required = *vnet_hdr;
 346     } else {
 347         vnet_hdr_required = 0;
 348     }
 349
 350     TFR(fd = tap_open(ifname, sizeof(ifname), vnet_hdr, vnet_hdr_required));
 351     if (fd < 0) {
 352         return -1;
 353     }
 354
 355     setup_script = qemu_opt_get(opts, "script");
 356     if (setup_script &&
 357         setup_script[0] != '\0' &&
 358         strcmp(setup_script, "no") != 0 &&
 359         launch_script(setup_script, ifname, fd)) {
 360         close(fd);
 361         return -1;
 362     }
 363
 364     qemu_opt_set(opts, "ifname", ifname);
 365
 366     return fd;
 367 }
 368
 369 int net_init_tap(QemuOpts *opts, Monitor *mon, const char *name, VLANState *vlan)
 370 {
 371     TAPState *s;
 372     int fd, vnet_hdr;
 373
 374     if (qemu_opt_get(opts, "fd")) {
 375         if (qemu_opt_get(opts, "ifname") ||
 376             qemu_opt_get(opts, "script") ||
 377             qemu_opt_get(opts, "downscript") ||
 378             qemu_opt_get(opts, "vnet_hdr")) {
 379             qemu_error("ifname=, script=, downscript= and vnet_hdr= is invalid with fd=\n");
 380             return -1;
 381         }
 382
 383         fd = net_handle_fd_param(mon, qemu_opt_get(opts, "fd"));
 384         if (fd == -1) {
 385             return -1;
 386         }
 387
 388         fcntl(fd, F_SETFL, O_NONBLOCK);
 389
 390         vnet_hdr = tap_probe_vnet_hdr(fd);
 391     } else {
 392         if (!qemu_opt_get(opts, "script")) {
 393             qemu_opt_set(opts, "script", DEFAULT_NETWORK_SCRIPT);
 394         }
 395
 396         if (!qemu_opt_get(opts, "downscript")) {
 397             qemu_opt_set(opts, "downscript", DEFAULT_NETWORK_DOWN_SCRIPT);
 398         }
 399
 400         fd = net_tap_init(opts, &vnet_hdr);
 401     }
 402
 403     s = net_tap_fd_init(vlan, "tap", name, fd, vnet_hdr);
 404     if (!s) {
 405         close(fd);
 406         return -1;
 407     }
 408
 409     if (tap_set_sndbuf(s->fd, opts) < 0) {
 410         return -1;
 411     }
 412
 413     if (qemu_opt_get(opts, "fd")) {
 414         snprintf(s->vc->info_str, sizeof(s->vc->info_str), "fd=%d", fd);
 415     } else {
 416         const char *ifname, *script, *downscript;
 417
 418         ifname     = qemu_opt_get(opts, "ifname");
 419         script     = qemu_opt_get(opts, "script");
 420         downscript = qemu_opt_get(opts, "downscript");
 421
 422         snprintf(s->vc->info_str, sizeof(s->vc->info_str),
 423                  "ifname=%s,script=%s,downscript=%s",
 424                  ifname, script, downscript);
 425
 426         if (strcmp(downscript, "no") != 0) {
 427             snprintf(s->down_script, sizeof(s->down_script), "%s", downscript);
 428             snprintf(s->down_script_arg, sizeof(s->down_script_arg), "%s", ifname);
 429         }
 430     }
 431
 432     if (vlan) {
 433         vlan->nb_host_devs++;
 434     }
 435
 436     return 0;
 437 }