net.c

   1 /*
   2  * QEMU System Emulator
   3  *
   4  * Copyright (c) 2003-2008 Fabrice Bellard
   5  *
   6  * Permission is hereby granted, free of charge, to any person obtaining a copy
   7  * of this software and associated documentation files (the "Software"), to deal
   8  * in the Software without restriction, including without limitation the rights
   9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  10  * copies of the Software, and to permit persons to whom the Software is
  11  * furnished to do so, subject to the following conditions:
  12  *
  13  * The above copyright notice and this permission notice shall be included in
  14  * all copies or substantial portions of the Software.
  15  *
  16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  22  * THE SOFTWARE.
  23  */
  24 #include "qemu-common.h"
  25 #include "net.h"
  26 #include "console.h"
  27 #include "sysemu.h"
  28 #include "qemu-timer.h"
  29 #include "qemu-char.h"
  30 #include "audio/audio.h"
  31
  32 #include <unistd.h>
  33 #include <fcntl.h>
  34 #include <signal.h>
  35 #include <time.h>
  36 #include <errno.h>
  37 #include <sys/time.h>
  38 #include <zlib.h>
  39
  40 #ifndef _WIN32
  41 #include <sys/times.h>
  42 #include <sys/wait.h>
  43 #include <termios.h>
  44 #include <sys/mman.h>
  45 #include <sys/ioctl.h>
  46 #include <sys/resource.h>
  47 #include <sys/socket.h>
  48 #include <netinet/in.h>
  49 #include <net/if.h>
  50 #ifdef __NetBSD__
  51 #include <net/if_tap.h>
  52 #endif
  53 #ifdef __linux__
  54 #include <linux/if_tun.h>
  55 #endif
  56 #include <arpa/inet.h>
  57 #include <dirent.h>
  58 #include <netdb.h>
  59 #include <sys/select.h>
  60 #ifdef _BSD
  61 #include <sys/stat.h>
  62 #ifdef __FreeBSD__
  63 #include <libutil.h>
  64 #else
  65 #include <util.h>
  66 #endif
  67 #elif defined (__GLIBC__) && defined (__FreeBSD_kernel__)
  68 #include <freebsd/stdlib.h>
  69 #else
  70 #ifdef __linux__
  71 #include <pty.h>
  72 #include <malloc.h>
  73 #include <linux/rtc.h>
  74
  75 /* For the benefit of older linux systems which don't supply it,
  76    we use a local copy of hpet.h. */
  77 /* #include <linux/hpet.h> */
  78 #include "hpet.h"
  79
  80 #include <linux/ppdev.h>
  81 #include <linux/parport.h>
  82 #endif
  83 #ifdef __sun__
  84 #include <sys/stat.h>
  85 #include <sys/ethernet.h>
  86 #include <sys/sockio.h>
  87 #include <netinet/arp.h>
  88 #include <netinet/in.h>
  89 #include <netinet/in_systm.h>
  90 #include <netinet/ip.h>
  91 #include <netinet/ip_icmp.h> // must come after ip.h
  92 #include <netinet/udp.h>
  93 #include <netinet/tcp.h>
  94 #include <net/if.h>
  95 #include <syslog.h>
  96 #include <stropts.h>
  97 #endif
  98 #endif
  99 #endif
 100
 101 #include "qemu_socket.h"
 102
 103 #if defined(CONFIG_SLIRP)
 104 #include "libslirp.h"
 105 #endif
 106
 107 #if defined(__OpenBSD__)
 108 #include <util.h>
 109 #endif
 110
 111 #if defined(CONFIG_VDE)
 112 #include <libvdeplug.h>
 113 #endif
 114
 115 #ifdef _WIN32
 116 #include <malloc.h>
 117 #include <sys/timeb.h>
 118 #include <mmsystem.h>
 119 #define getopt_long_only getopt_long
 120 #define memalign(align, size) malloc(size)
 121 #endif
 122
 123 // FIXME: #include "qemu-kvm.h"
 124
 125 static VLANState *first_vlan;
 126
 127 /***********************************************************/
 128 /* network device redirectors */
 129
 130 #if defined(DEBUG_NET) || defined(DEBUG_SLIRP)
 131 static void hex_dump(FILE *f, const uint8_t *buf, int size)
 132 {
 133     int len, i, j, c;
 134
 135     for(i=0;i<size;i+=16) {
 136         len = size - i;
 137         if (len > 16)
 138             len = 16;
 139         fprintf(f, "%08x ", i);
 140         for(j=0;j<16;j++) {
 141             if (j < len)
 142                 fprintf(f, " %02x", buf[i+j]);
 143             else
 144                 fprintf(f, "   ");
 145         }
 146         fprintf(f, " ");
 147         for(j=0;j<len;j++) {
 148             c = buf[i+j];
 149             if (c < ' ' || c > '~')
 150                 c = '.';
 151             fprintf(f, "%c", c);
 152         }
 153         fprintf(f, "\n");
 154     }
 155 }
 156 #endif
 157
 158 static int parse_macaddr(uint8_t *macaddr, const char *p)
 159 {
 160     int i;
 161     char *last_char;
 162     long int offset;
 163
 164     errno = 0;
 165     offset = strtol(p, &last_char, 0);
 166     if (0 == errno && '\0' == *last_char &&
 167             offset >= 0 && offset <= 0xFFFFFF) {
 168         macaddr[3] = (offset & 0xFF0000) >> 16;
 169         macaddr[4] = (offset & 0xFF00) >> 8;
 170         macaddr[5] = offset & 0xFF;
 171         return 0;
 172     } else {
 173         for(i = 0; i < 6; i++) {
 174             macaddr[i] = strtol(p, (char **)&p, 16);
 175             if (i == 5) {
 176                 if (*p != '\0')
 177                     return -1;
 178             } else {
 179                 if (*p != ':' && *p != '-')
 180                     return -1;
 181                 p++;
 182             }
 183         }
 184         return 0;
 185     }
 186
 187     return -1;
 188 }
 189
 190 static int get_str_sep(char *buf, int buf_size, const char **pp, int sep)
 191 {
 192     const char *p, *p1;
 193     int len;
 194     p = *pp;
 195     p1 = strchr(p, sep);
 196     if (!p1)
 197         return -1;
 198     len = p1 - p;
 199     p1++;
 200     if (buf_size > 0) {
 201         if (len > buf_size - 1)
 202             len = buf_size - 1;
 203         memcpy(buf, p, len);
 204         buf[len] = '\0';
 205     }
 206     *pp = p1;
 207     return 0;
 208 }
 209
 210 int parse_host_src_port(struct sockaddr_in *haddr,
 211                         struct sockaddr_in *saddr,
 212                         const char *input_str)
 213 {
 214     char *str = strdup(input_str);
 215     char *host_str = str;
 216     char *src_str;
 217     const char *src_str2;
 218     char *ptr;
 219
 220     /*
 221      * Chop off any extra arguments at the end of the string which
 222      * would start with a comma, then fill in the src port information
 223      * if it was provided else use the "any address" and "any port".
 224      */
 225     if ((ptr = strchr(str,',')))
 226         *ptr = '\0';
 227
 228     if ((src_str = strchr(input_str,'@'))) {
 229         *src_str = '\0';
 230         src_str++;
 231     }
 232
 233     if (parse_host_port(haddr, host_str) < 0)
 234         goto fail;
 235
 236     src_str2 = src_str;
 237     if (!src_str || *src_str == '\0')
 238         src_str2 = ":0";
 239
 240     if (parse_host_port(saddr, src_str2) < 0)
 241         goto fail;
 242
 243     free(str);
 244     return(0);
 245
 246 fail:
 247     free(str);
 248     return -1;
 249 }
 250
 251 int parse_host_port(struct sockaddr_in *saddr, const char *str)
 252 {
 253     char buf[512];
 254     struct hostent *he;
 255     const char *p, *r;
 256     int port;
 257
 258     p = str;
 259     if (get_str_sep(buf, sizeof(buf), &p, ':') < 0)
 260         return -1;
 261     saddr->sin_family = AF_INET;
 262     if (buf[0] == '\0') {
 263         saddr->sin_addr.s_addr = 0;
 264     } else {
 265         if (qemu_isdigit(buf[0])) {
 266             if (!inet_aton(buf, &saddr->sin_addr))
 267                 return -1;
 268         } else {
 269             if ((he = gethostbyname(buf)) == NULL)
 270                 return - 1;
 271             saddr->sin_addr = *(struct in_addr *)he->h_addr;
 272         }
 273     }
 274     port = strtol(p, (char **)&r, 0);
 275     if (r == p)
 276         return -1;
 277     saddr->sin_port = htons(port);
 278     return 0;
 279 }
 280
 281 #if !defined(_WIN32) && 0
 282 static int parse_unix_path(struct sockaddr_un *uaddr, const char *str)
 283 {
 284     const char *p;
 285     int len;
 286
 287     len = MIN(108, strlen(str));
 288     p = strchr(str, ',');
 289     if (p)
 290         len = MIN(len, p - str);
 291
 292     memset(uaddr, 0, sizeof(*uaddr));
 293
 294     uaddr->sun_family = AF_UNIX;
 295     memcpy(uaddr->sun_path, str, len);
 296
 297     return 0;
 298 }
 299 #endif
 300
 301 void qemu_format_nic_info_str(VLANClientState *vc, uint8_t macaddr[6])
 302 {
 303     snprintf(vc->info_str, sizeof(vc->info_str),
 304              "model=%s,macaddr=%02x:%02x:%02x:%02x:%02x:%02x",
 305              vc->model,
 306              macaddr[0], macaddr[1], macaddr[2],
 307              macaddr[3], macaddr[4], macaddr[5]);
 308 }
 309
 310 static char *assign_name(VLANClientState *vc1, const char *model)
 311 {
 312     VLANState *vlan;
 313     char buf[256];
 314     int id = 0;
 315
 316     for (vlan = first_vlan; vlan; vlan = vlan->next) {
 317         VLANClientState *vc;
 318
 319         for (vc = vlan->first_client; vc; vc = vc->next)
 320             if (vc != vc1 && strcmp(vc->model, model) == 0)
 321                 id++;
 322     }
 323
 324     snprintf(buf, sizeof(buf), "%s.%d", model, id);
 325
 326     return strdup(buf);
 327 }
 328
 329 VLANClientState *qemu_new_vlan_client(VLANState *vlan,
 330                                       const char *model,
 331                                       const char *name,
 332                                       IOReadHandler *fd_read,
 333                                       IOCanRWHandler *fd_can_read,
 334                                       void *opaque)
 335 {
 336     VLANClientState *vc, **pvc;
 337     vc = qemu_mallocz(sizeof(VLANClientState));
 338     vc->model = strdup(model);
 339     if (name)
 340         vc->name = strdup(name);
 341     else
 342         vc->name = assign_name(vc, model);
 343     vc->fd_read = fd_read;
 344     vc->fd_can_read = fd_can_read;
 345     vc->opaque = opaque;
 346     vc->vlan = vlan;
 347
 348     vc->next = NULL;
 349     pvc = &vlan->first_client;
 350     while (*pvc != NULL)
 351         pvc = &(*pvc)->next;
 352     *pvc = vc;
 353     return vc;
 354 }
 355
 356 void qemu_del_vlan_client(VLANClientState *vc)
 357 {
 358     VLANClientState **pvc = &vc->vlan->first_client;
 359
 360     while (*pvc != NULL)
 361         if (*pvc == vc) {
 362             *pvc = vc->next;
 363             free(vc->name);
 364             free(vc->model);
 365             free(vc);
 366             break;
 367         } else
 368             pvc = &(*pvc)->next;
 369 }
 370
 371 int qemu_can_send_packet(VLANClientState *vc1)
 372 {
 373     VLANState *vlan = vc1->vlan;
 374     VLANClientState *vc;
 375
 376     for(vc = vlan->first_client; vc != NULL; vc = vc->next) {
 377         if (vc != vc1) {
 378             if (vc->fd_can_read && vc->fd_can_read(vc->opaque))
 379                 return 1;
 380         }
 381     }
 382     return 0;
 383 }
 384
 385 int qemu_send_packet(VLANClientState *vc1, const uint8_t *buf, int size)
 386 {
 387     VLANState *vlan = vc1->vlan;
 388     VLANClientState *vc;
 389     int ret = -EAGAIN;
 390
 391     if (vc1->link_down)
 392         return 0;
 393
 394 #ifdef DEBUG_NET
 395     printf("vlan %d send:\n", vlan->id);
 396     hex_dump(stdout, buf, size);
 397 #endif
 398     for(vc = vlan->first_client; vc != NULL; vc = vc->next) {
 399         if (vc != vc1 && !vc->link_down) {
 400             if (!vc->fd_can_read || vc->fd_can_read(vc->opaque)) {
 401                 vc->fd_read(vc->opaque, buf, size);
 402                 ret = 0;
 403             }
 404         }
 405     }
 406     return ret;
 407 }
 408
 409 static ssize_t vc_sendv_compat(VLANClientState *vc, const struct iovec *iov,
 410                                int iovcnt)
 411 {
 412     uint8_t buffer[4096];
 413     size_t offset = 0;
 414     int i;
 415
 416     for (i = 0; i < iovcnt; i++) {
 417         size_t len;
 418
 419         len = MIN(sizeof(buffer) - offset, iov[i].iov_len);
 420         memcpy(buffer + offset, iov[i].iov_base, len);
 421         offset += len;
 422     }
 423
 424     vc->fd_read(vc->opaque, buffer, offset);
 425
 426     return offset;
 427 }
 428
 429 static ssize_t calc_iov_length(const struct iovec *iov, int iovcnt)
 430 {
 431     size_t offset = 0;
 432     int i;
 433
 434     for (i = 0; i < iovcnt; i++)
 435         offset += iov[i].iov_len;
 436     return offset;
 437 }
 438
 439 ssize_t qemu_sendv_packet(VLANClientState *vc1, const struct iovec *iov,
 440                           int iovcnt)
 441 {
 442     VLANState *vlan = vc1->vlan;
 443     VLANClientState *vc;
 444     ssize_t max_len = 0;
 445
 446     if (vc1->link_down)
 447         return calc_iov_length(iov, iovcnt);
 448
 449     for (vc = vlan->first_client; vc != NULL; vc = vc->next) {
 450         ssize_t len = 0;
 451
 452         if (vc == vc1)
 453             continue;
 454
 455         if (vc->link_down)
 456             len = calc_iov_length(iov, iovcnt);
 457         if (vc->fd_readv)
 458             len = vc->fd_readv(vc->opaque, iov, iovcnt);
 459         else if (vc->fd_read)
 460             len = vc_sendv_compat(vc, iov, iovcnt);
 461
 462         max_len = MAX(max_len, len);
 463     }
 464
 465     return max_len;
 466 }
 467
 468 #if defined(CONFIG_SLIRP)
 469
 470 /* slirp network adapter */
 471
 472 static int slirp_inited;
 473 static int slirp_restrict;
 474 static char *slirp_ip;
 475 static VLANClientState *slirp_vc;
 476
 477 int slirp_can_output(void)
 478 {
 479     return !slirp_vc || qemu_can_send_packet(slirp_vc);
 480 }
 481
 482 void slirp_output(const uint8_t *pkt, int pkt_len)
 483 {
 484 #ifdef DEBUG_SLIRP
 485     printf("slirp output:\n");
 486     hex_dump(stdout, pkt, pkt_len);
 487 #endif
 488     if (!slirp_vc)
 489         return;
 490     qemu_send_packet(slirp_vc, pkt, pkt_len);
 491 }
 492
 493 int slirp_is_inited(void)
 494 {
 495     return slirp_inited;
 496 }
 497
 498 static void slirp_receive(void *opaque, const uint8_t *buf, int size)
 499 {
 500 #ifdef DEBUG_SLIRP
 501     printf("slirp input:\n");
 502     hex_dump(stdout, buf, size);
 503 #endif
 504     slirp_input(buf, size);
 505 }
 506
 507 static int net_slirp_init(VLANState *vlan, const char *model, const char *name)
 508 {
 509     if (!slirp_inited) {
 510         slirp_inited = 1;
 511         slirp_init(slirp_restrict, slirp_ip);
 512     }
 513     slirp_vc = qemu_new_vlan_client(vlan, model, name,
 514                                     slirp_receive, NULL, NULL);
 515     slirp_vc->info_str[0] = '\0';
 516     return 0;
 517 }
 518
 519 void net_slirp_redir(const char *redir_str)
 520 {
 521     int is_udp;
 522     char buf[256], *r;
 523     const char *p;
 524     struct in_addr guest_addr;
 525     int host_port, guest_port;
 526
 527     if (!slirp_inited) {
 528         slirp_inited = 1;
 529         slirp_init(slirp_restrict, slirp_ip);
 530     }
 531
 532     p = redir_str;
 533     if (get_str_sep(buf, sizeof(buf), &p, ':') < 0)
 534         goto fail;
 535     if (!strcmp(buf, "tcp")) {
 536         is_udp = 0;
 537     } else if (!strcmp(buf, "udp")) {
 538         is_udp = 1;
 539     } else {
 540         goto fail;
 541     }
 542
 543     if (get_str_sep(buf, sizeof(buf), &p, ':') < 0)
 544         goto fail;
 545     host_port = strtol(buf, &r, 0);
 546     if (r == buf)
 547         goto fail;
 548
 549     if (get_str_sep(buf, sizeof(buf), &p, ':') < 0)
 550         goto fail;
 551     if (buf[0] == '\0') {
 552         pstrcpy(buf, sizeof(buf), "10.0.2.15");
 553     }
 554     if (!inet_aton(buf, &guest_addr))
 555         goto fail;
 556
 557     guest_port = strtol(p, &r, 0);
 558     if (r == p)
 559         goto fail;
 560
 561     if (slirp_redir(is_udp, host_port, guest_addr, guest_port) < 0) {
 562         fprintf(stderr, "qemu: could not set up redirection\n");
 563         exit(1);
 564     }
 565     return;
 566  fail:
 567     fprintf(stderr, "qemu: syntax: -redir [tcp|udp]:host-port:[guest-host]:guest-port\n");
 568     exit(1);
 569 }
 570
 571 #ifndef _WIN32
 572
 573 static char smb_dir[1024];
 574
 575 static void erase_dir(char *dir_name)
 576 {
 577     DIR *d;
 578     struct dirent *de;
 579     char filename[1024];
 580
 581     /* erase all the files in the directory */
 582     if ((d = opendir(dir_name)) != 0) {
 583         for(;;) {
 584             de = readdir(d);
 585             if (!de)
 586                 break;
 587             if (strcmp(de->d_name, ".") != 0 &&
 588                 strcmp(de->d_name, "..") != 0) {
 589                 snprintf(filename, sizeof(filename), "%s/%s",
 590                          smb_dir, de->d_name);
 591                 if (unlink(filename) != 0)  /* is it a directory? */
 592                     erase_dir(filename);
 593             }
 594         }
 595         closedir(d);
 596         rmdir(dir_name);
 597     }
 598 }
 599
 600 /* automatic user mode samba server configuration */
 601 static void smb_exit(void)
 602 {
 603     erase_dir(smb_dir);
 604 }
 605
 606 /* automatic user mode samba server configuration */
 607 void net_slirp_smb(const char *exported_dir)
 608 {
 609     char smb_conf[1024];
 610     char smb_cmdline[1024];
 611     FILE *f;
 612
 613     if (!slirp_inited) {
 614         slirp_inited = 1;
 615         slirp_init(slirp_restrict, slirp_ip);
 616     }
 617
 618     /* XXX: better tmp dir construction */
 619     snprintf(smb_dir, sizeof(smb_dir), "/tmp/qemu-smb.%d", getpid());
 620     if (mkdir(smb_dir, 0700) < 0) {
 621         fprintf(stderr, "qemu: could not create samba server dir '%s'\n", smb_dir);
 622         exit(1);
 623     }
 624     snprintf(smb_conf, sizeof(smb_conf), "%s/%s", smb_dir, "smb.conf");
 625
 626     f = fopen(smb_conf, "w");
 627     if (!f) {
 628         fprintf(stderr, "qemu: could not create samba server configuration file '%s'\n", smb_conf);
 629         exit(1);
 630     }
 631     fprintf(f,
 632             "[global]\n"
 633             "private dir=%s\n"
 634             "smb ports=0\n"
 635             "socket address=127.0.0.1\n"
 636             "pid directory=%s\n"
 637             "lock directory=%s\n"
 638             "log file=%s/log.smbd\n"
 639             "smb passwd file=%s/smbpasswd\n"
 640             "security = share\n"
 641             "[qemu]\n"
 642             "path=%s\n"
 643             "read only=no\n"
 644             "guest ok=yes\n",
 645             smb_dir,
 646             smb_dir,
 647             smb_dir,
 648             smb_dir,
 649             smb_dir,
 650             exported_dir
 651             );
 652     fclose(f);
 653     atexit(smb_exit);
 654
 655     snprintf(smb_cmdline, sizeof(smb_cmdline), "%s -s %s",
 656              SMBD_COMMAND, smb_conf);
 657
 658     slirp_add_exec(0, smb_cmdline, 4, 139);
 659 }
 660
 661 #endif /* !defined(_WIN32) */
 662 void do_info_slirp(void)
 663 {
 664     slirp_stats();
 665 }
 666
 667 #endif /* CONFIG_SLIRP */
 668
 669 #ifdef _WIN32
 670
 671 int tap_has_vnet_hdr(void *opaque)
 672 {
 673     return 0;
 674 }
 675
 676 void tap_using_vnet_hdr(void *opaque, int using_vnet_hdr)
 677 {
 678 }
 679
 680 #else /* !defined(_WIN32) */
 681
 682 /* Maximum GSO packet size (64k) plus plenty of room for
 683  * the ethernet and virtio_net headers
 684  */
 685 #define TAP_BUFSIZE (4096 + 65536)
 686
 687 #ifdef IFF_VNET_HDR
 688 #include <linux/virtio_net.h>
 689 #endif
 690
 691 typedef struct TAPState {
 692     VLANClientState *vc;
 693     int fd;
 694     char down_script[1024];
 695     char down_script_arg[128];
 696     char buf[TAP_BUFSIZE];
 697     int size;
 698     unsigned int has_vnet_hdr : 1;
 699     unsigned int using_vnet_hdr : 1;
 700 } TAPState;
 701
 702 #ifdef HAVE_IOVEC
 703 static ssize_t tap_receive_iov(void *opaque, const struct iovec *iov,
 704                                int iovcnt)
 705 {
 706     TAPState *s = opaque;
 707     ssize_t len;
 708
 709     do {
 710         len = writev(s->fd, iov, iovcnt);
 711     } while (len == -1 && (errno == EINTR || errno == EAGAIN));
 712
 713     return len;
 714 }
 715 #endif
 716
 717 static void tap_receive(void *opaque, const uint8_t *buf, int size)
 718 {
 719     struct iovec iov[2];
 720     int i = 0;
 721
 722 #ifdef IFF_VNET_HDR
 723     TAPState *s = opaque;
 724     struct virtio_net_hdr hdr = { 0, };
 725
 726     if (s->has_vnet_hdr && !s->using_vnet_hdr) {
 727         iov[i].iov_base = &hdr;
 728         iov[i].iov_len  = sizeof(hdr);
 729         i++;
 730     }
 731 #endif
 732
 733     iov[i].iov_base = (char *) buf;
 734     iov[i].iov_len  = size;
 735     i++;
 736
 737     tap_receive_iov(opaque, iov, i);
 738 }
 739
 740 static int tap_can_send(void *opaque)
 741 {
 742     TAPState *s = opaque;
 743     VLANClientState *vc;
 744     int can_receive = 0;
 745
 746     /* Check to see if any of our clients can receive a packet */
 747     for (vc = s->vc->vlan->first_client; vc; vc = vc->next) {
 748         /* Skip ourselves */
 749         if (vc == s->vc)
 750             continue;
 751
 752         if (!vc->fd_can_read) {
 753             /* no fd_can_read handler, they always can receive */
 754             can_receive = 1;
 755         } else
 756             can_receive = vc->fd_can_read(vc->opaque);
 757
 758         /* Once someone can receive, we try to send a packet */
 759         if (can_receive)
 760             break;
 761     }
 762
 763     return can_receive;
 764 }
 765
 766 static int tap_send_packet(TAPState *s)
 767 {
 768     uint8_t *buf = (uint8_t *)s->buf;
 769     int size = s->size;
 770
 771 #ifdef IFF_VNET_HDR
 772     if (s->has_vnet_hdr && !s->using_vnet_hdr) {
 773         buf += sizeof(struct virtio_net_hdr);
 774         size -= sizeof(struct virtio_net_hdr);
 775     }
 776 #endif
 777
 778     return qemu_send_packet(s->vc, buf, size);
 779 }
 780
 781 static void tap_send(void *opaque)
 782 {
 783     TAPState *s = opaque;
 784
 785     /* First try to send any buffered packet */
 786     if (s->size > 0) {
 787         int err;
 788
 789         /* If noone can receive the packet, buffer it */
 790         err = tap_send_packet(s);
 791         if (err == -EAGAIN)
 792             return;
 793     }
 794
 795     /* Read packets until we hit EAGAIN */
 796     do {
 797 #ifdef __sun__
 798         struct strbuf sbuf;
 799         int f = 0;
 800         sbuf.maxlen = sizeof(s->buf);
 801         sbuf.buf = s->buf;
 802         s->size = getmsg(s->fd, NULL, &sbuf, &f) >=0 ? sbuf.len : -1;
 803 #else
 804         // FIXME: kvm_sleep_begin();
 805         s->size = read(s->fd, s->buf, sizeof(s->buf));
 806         // FIXME: kvm_sleep_end();
 807 #endif
 808
 809         if (s->size == -1 && errno == EINTR)
 810             continue;
 811
 812         if (s->size > 0) {
 813             int err;
 814
 815             /* If noone can receive the packet, buffer it */
 816             err = tap_send_packet(s);
 817             if (err == -EAGAIN)
 818                 break;
 819         }
 820     } while (s->size > 0);
 821 }
 822
 823 int tap_has_vnet_hdr(void *opaque)
 824 {
 825     VLANClientState *vc = opaque;
 826     TAPState *s = vc->opaque;
 827
 828     return s ? s->has_vnet_hdr : 0;
 829 }
 830
 831 void tap_using_vnet_hdr(void *opaque, int using_vnet_hdr)
 832 {
 833     VLANClientState *vc = opaque;
 834     TAPState *s = vc->opaque;
 835
 836     if (!s || !s->has_vnet_hdr)
 837         return;
 838
 839     s->using_vnet_hdr = using_vnet_hdr != 0;
 840 }
 841
 842 static int tap_probe_vnet_hdr(int fd)
 843 {
 844 #if defined(TUNGETIFF) && defined(IFF_VNET_HDR)
 845     struct ifreq ifr;
 846
 847     if (ioctl(fd, TUNGETIFF, &ifr) != 0) {
 848         fprintf(stderr, "TUNGETIFF ioctl() failed: %s\n", strerror(errno));
 849         return 0;
 850     }
 851
 852     return ifr.ifr_flags & IFF_VNET_HDR;
 853 #else
 854     return 0;
 855 #endif
 856 }
 857
 858 #ifdef TUNSETOFFLOAD
 859 static void tap_set_offload(VLANClientState *vc, int csum, int tso4, int tso6,
 860                             int ecn)
 861 {
 862     TAPState *s = vc->opaque;
 863     unsigned int offload = 0;
 864
 865     if (csum) {
 866         offload |= TUN_F_CSUM;
 867         if (tso4)
 868             offload |= TUN_F_TSO4;
 869         if (tso6)
 870             offload |= TUN_F_TSO6;
 871         if ((tso4 || tso6) && ecn)
 872             offload |= TUN_F_TSO_ECN;
 873     }
 874
 875     if (ioctl(s->fd, TUNSETOFFLOAD, offload) != 0)
 876         fprintf(stderr, "TUNSETOFFLOAD ioctl() failed: %s\n",
 877                 strerror(errno));
 878 }
 879 #endif /* TUNSETOFFLOAD */
 880
 881 /* fd support */
 882
 883 static TAPState *net_tap_fd_init(VLANState *vlan,
 884                                  const char *model,
 885                                  const char *name,
 886                                  int fd,
 887                                  int vnet_hdr)
 888 {
 889     TAPState *s;
 890
 891     s = qemu_mallocz(sizeof(TAPState));
 892     s->fd = fd;
 893     s->has_vnet_hdr = vnet_hdr != 0;
 894     s->vc = qemu_new_vlan_client(vlan, model, name, tap_receive, NULL, s);
 895 #ifdef HAVE_IOVEC
 896     s->vc->fd_readv = tap_receive_iov;
 897 #endif
 898 #ifdef TUNSETOFFLOAD
 899     s->vc->set_offload = tap_set_offload;
 900 #endif
 901     qemu_set_fd_handler2(s->fd, tap_can_send, tap_send, NULL, s);
 902     snprintf(s->vc->info_str, sizeof(s->vc->info_str), "fd=%d", fd);
 903     return s;
 904 }
 905
 906 #if defined (_BSD) || defined (__FreeBSD_kernel__)
 907 static int tap_open(char *ifname, int ifname_size)
 908 {
 909     int fd;
 910     char *dev;
 911     struct stat s;
 912
 913     TFR(fd = open("/dev/tap", O_RDWR));
 914     if (fd < 0) {
 915         fprintf(stderr, "warning: could not open /dev/tap: no virtual network emulation\n");
 916         return -1;
 917     }
 918
 919     fstat(fd, &s);
 920     dev = devname(s.st_rdev, S_IFCHR);
 921     pstrcpy(ifname, ifname_size, dev);
 922
 923     fcntl(fd, F_SETFL, O_NONBLOCK);
 924     return fd;
 925 }
 926 #elif defined(__sun__)
 927 #define TUNNEWPPA       (('T'<<16) | 0x0001)
 928 /*
 929  * Allocate TAP device, returns opened fd.
 930  * Stores dev name in the first arg(must be large enough).
 931  */
 932 int tap_alloc(char *dev, size_t dev_size)
 933 {
 934     int tap_fd, if_fd, ppa = -1;
 935     static int ip_fd = 0;
 936     char *ptr;
 937
 938     static int arp_fd = 0;
 939     int ip_muxid, arp_muxid;
 940     struct strioctl  strioc_if, strioc_ppa;
 941     int link_type = I_PLINK;;
 942     struct lifreq ifr;
 943     char actual_name[32] = "";
 944
 945     memset(&ifr, 0x0, sizeof(ifr));
 946
 947     if( *dev ){
 948        ptr = dev;
 949        while( *ptr && !qemu_isdigit((int)*ptr) ) ptr++;
 950        ppa = atoi(ptr);
 951     }
 952
 953     /* Check if IP device was opened */
 954     if( ip_fd )
 955        close(ip_fd);
 956
 957     TFR(ip_fd = open("/dev/udp", O_RDWR, 0));
 958     if (ip_fd < 0) {
 959        syslog(LOG_ERR, "Can't open /dev/ip (actually /dev/udp)");
 960        return -1;
 961     }
 962
 963     TFR(tap_fd = open("/dev/tap", O_RDWR, 0));
 964     if (tap_fd < 0) {
 965        syslog(LOG_ERR, "Can't open /dev/tap");
 966        return -1;
 967     }
 968
 969     /* Assign a new PPA and get its unit number. */
 970     strioc_ppa.ic_cmd = TUNNEWPPA;
 971     strioc_ppa.ic_timout = 0;
 972     strioc_ppa.ic_len = sizeof(ppa);
 973     strioc_ppa.ic_dp = (char *)&ppa;
 974     if ((ppa = ioctl (tap_fd, I_STR, &strioc_ppa)) < 0)
 975        syslog (LOG_ERR, "Can't assign new interface");
 976
 977     TFR(if_fd = open("/dev/tap", O_RDWR, 0));
 978     if (if_fd < 0) {
 979        syslog(LOG_ERR, "Can't open /dev/tap (2)");
 980        return -1;
 981     }
 982     if(ioctl(if_fd, I_PUSH, "ip") < 0){
 983        syslog(LOG_ERR, "Can't push IP module");
 984        return -1;
 985     }
 986
 987     if (ioctl(if_fd, SIOCGLIFFLAGS, &ifr) < 0)
 988         syslog(LOG_ERR, "Can't get flags\n");
 989
 990     snprintf (actual_name, 32, "tap%d", ppa);
 991     pstrcpy(ifr.lifr_name, sizeof(ifr.lifr_name), actual_name);
 992
 993     ifr.lifr_ppa = ppa;
 994     /* Assign ppa according to the unit number returned by tun device */
 995
 996     if (ioctl (if_fd, SIOCSLIFNAME, &ifr) < 0)
 997         syslog (LOG_ERR, "Can't set PPA %d", ppa);
 998     if (ioctl(if_fd, SIOCGLIFFLAGS, &ifr) <0)
 999         syslog (LOG_ERR, "Can't get flags\n");
1000     /* Push arp module to if_fd */
1001     if (ioctl (if_fd, I_PUSH, "arp") < 0)
1002         syslog (LOG_ERR, "Can't push ARP module (2)");
1003
1004     /* Push arp module to ip_fd */
1005     if (ioctl (ip_fd, I_POP, NULL) < 0)
1006         syslog (LOG_ERR, "I_POP failed\n");
1007     if (ioctl (ip_fd, I_PUSH, "arp") < 0)
1008         syslog (LOG_ERR, "Can't push ARP module (3)\n");
1009     /* Open arp_fd */
1010     TFR(arp_fd = open ("/dev/tap", O_RDWR, 0));
1011     if (arp_fd < 0)
1012        syslog (LOG_ERR, "Can't open %s\n", "/dev/tap");
1013
1014     /* Set ifname to arp */
1015     strioc_if.ic_cmd = SIOCSLIFNAME;
1016     strioc_if.ic_timout = 0;
1017     strioc_if.ic_len = sizeof(ifr);
1018     strioc_if.ic_dp = (char *)&ifr;
1019     if (ioctl(arp_fd, I_STR, &strioc_if) < 0){
1020         syslog (LOG_ERR, "Can't set ifname to arp\n");
1021     }
1022
1023     if((ip_muxid = ioctl(ip_fd, I_LINK, if_fd)) < 0){
1024        syslog(LOG_ERR, "Can't link TAP device to IP");
1025        return -1;
1026     }
1027
1028     if ((arp_muxid = ioctl (ip_fd, link_type, arp_fd)) < 0)
1029         syslog (LOG_ERR, "Can't link TAP device to ARP");
1030
1031     close (if_fd);
1032
1033     memset(&ifr, 0x0, sizeof(ifr));
1034     pstrcpy(ifr.lifr_name, sizeof(ifr.lifr_name), actual_name);
1035     ifr.lifr_ip_muxid  = ip_muxid;
1036     ifr.lifr_arp_muxid = arp_muxid;
1037
1038     if (ioctl (ip_fd, SIOCSLIFMUXID, &ifr) < 0)
1039     {
1040       ioctl (ip_fd, I_PUNLINK , arp_muxid);
1041       ioctl (ip_fd, I_PUNLINK, ip_muxid);
1042       syslog (LOG_ERR, "Can't set multiplexor id");
1043     }
1044
1045     snprintf(dev, dev_size, "tap%d", ppa);
1046     return tap_fd;
1047 }
1048
1049 static int tap_open(char *ifname, int ifname_size, int *vnet_hdr)
1050 {
1051     char  dev[10]="";
1052     int fd;
1053     if( (fd = tap_alloc(dev, sizeof(dev))) < 0 ){
1054        fprintf(stderr, "Cannot allocate TAP device\n");
1055        return -1;
1056     }
1057     pstrcpy(ifname, ifname_size, dev);
1058     fcntl(fd, F_SETFL, O_NONBLOCK);
1059     return fd;
1060 }
1061 #elif defined (_AIX)
1062 static int tap_open(char *ifname, int ifname_size)
1063 {
1064     fprintf (stderr, "no tap on AIX\n");
1065     return -1;
1066 }
1067 #else
1068 static int tap_open(char *ifname, int ifname_size, int *vnet_hdr)
1069 {
1070     struct ifreq ifr;
1071     int fd, ret;
1072
1073     TFR(fd = open("/dev/net/tun", O_RDWR));
1074     if (fd < 0) {
1075         fprintf(stderr, "warning: could not open /dev/net/tun: no virtual network emulation\n");
1076         return -1;
1077     }
1078     memset(&ifr, 0, sizeof(ifr));
1079     ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
1080
1081 #if defined(TUNGETFEATURES) && defined(IFF_VNET_HDR)
1082     {
1083         unsigned int features;
1084
1085         if (ioctl(fd, TUNGETFEATURES, &features) == 0 &&
1086             features & IFF_VNET_HDR) {
1087             *vnet_hdr = 1;
1088             ifr.ifr_flags |= IFF_VNET_HDR;
1089         }
1090     }
1091 #endif
1092
1093     if (ifname[0] != '\0')
1094         pstrcpy(ifr.ifr_name, IFNAMSIZ, ifname);
1095     else
1096         pstrcpy(ifr.ifr_name, IFNAMSIZ, "tap%d");
1097     ret = ioctl(fd, TUNSETIFF, (void *) &ifr);
1098     if (ret != 0) {
1099         fprintf(stderr, "warning: could not configure /dev/net/tun: no virtual network emulation\n");
1100         close(fd);
1101         return -1;
1102     }
1103     pstrcpy(ifname, ifname_size, ifr.ifr_name);
1104     fcntl(fd, F_SETFL, O_NONBLOCK);
1105     return fd;
1106 }
1107 #endif
1108
1109 static int launch_script(const char *setup_script, const char *ifname, int fd)
1110 {
1111     int pid, status;
1112     char *args[3];
1113     char **parg;
1114
1115         /* try to launch network script */
1116         pid = fork();
1117         if (pid >= 0) {
1118             if (pid == 0) {
1119                 int open_max = sysconf (_SC_OPEN_MAX), i;
1120                 for (i = 0; i < open_max; i++)
1121                     if (i != STDIN_FILENO &&
1122                         i != STDOUT_FILENO &&
1123                         i != STDERR_FILENO &&
1124                         i != fd)
1125                         close(i);
1126
1127                 parg = args;
1128                 *parg++ = (char *)setup_script;
1129                 *parg++ = (char *)ifname;
1130                 *parg++ = NULL;
1131                 execv(setup_script, args);
1132                 _exit(1);
1133             }
1134             while (waitpid(pid, &status, 0) != pid);
1135             if (!WIFEXITED(status) ||
1136                 WEXITSTATUS(status) != 0) {
1137                 fprintf(stderr, "%s: could not launch network script\n",
1138                         setup_script);
1139                 return -1;
1140             }
1141         }
1142     return 0;
1143 }
1144
1145 static int net_tap_init(VLANState *vlan, const char *model,
1146                         const char *name, const char *ifname1,
1147                         const char *setup_script, const char *down_script)
1148 {
1149     TAPState *s;
1150     int fd;
1151     int vnet_hdr;
1152     char ifname[128];
1153
1154     if (ifname1 != NULL)
1155         pstrcpy(ifname, sizeof(ifname), ifname1);
1156     else
1157         ifname[0] = '\0';
1158     vnet_hdr = 0;
1159     TFR(fd = tap_open(ifname, sizeof(ifname), &vnet_hdr));
1160     if (fd < 0)
1161         return -1;
1162
1163     if (!setup_script || !strcmp(setup_script, "no"))
1164         setup_script = "";
1165     if (setup_script[0] != '\0') {
1166         if (launch_script(setup_script, ifname, fd))
1167             return -1;
1168     }
1169     s = net_tap_fd_init(vlan, model, name, fd, vnet_hdr);
1170     if (!s)
1171         return -1;
1172
1173     snprintf(s->vc->info_str, sizeof(s->vc->info_str),
1174              "ifname=%s,script=%s,downscript=%s",
1175              ifname, setup_script, down_script);
1176     if (down_script && strcmp(down_script, "no")) {
1177         snprintf(s->down_script, sizeof(s->down_script), "%s", down_script);
1178         snprintf(s->down_script_arg, sizeof(s->down_script_arg), "%s", ifname);
1179     }
1180     return 0;
1181 }
1182
1183 #endif /* !_WIN32 */
1184
1185 #if defined(CONFIG_VDE)
1186 typedef struct VDEState {
1187     VLANClientState *vc;
1188     VDECONN *vde;
1189 } VDEState;
1190
1191 static void vde_to_qemu(void *opaque)
1192 {
1193     VDEState *s = opaque;
1194     uint8_t buf[4096];
1195     int size;
1196
1197     size = vde_recv(s->vde, buf, sizeof(buf), 0);
1198     if (size > 0) {
1199         qemu_send_packet(s->vc, buf, size);
1200     }
1201 }
1202
1203 static void vde_from_qemu(void *opaque, const uint8_t *buf, int size)
1204 {
1205     VDEState *s = opaque;
1206     int ret;
1207     for(;;) {
1208         ret = vde_send(s->vde, buf, size, 0);
1209         if (ret < 0 && errno == EINTR) {
1210         } else {
1211             break;
1212         }
1213     }
1214 }
1215
1216 static int net_vde_init(VLANState *vlan, const char *model,
1217                         const char *name, const char *sock,
1218                         int port, const char *group, int mode)
1219 {
1220     VDEState *s;
1221     char *init_group = strlen(group) ? (char *)group : NULL;
1222     char *init_sock = strlen(sock) ? (char *)sock : NULL;
1223
1224     struct vde_open_args args = {
1225         .port = port,
1226         .group = init_group,
1227         .mode = mode,
1228     };
1229
1230     s = qemu_mallocz(sizeof(VDEState));
1231     s->vde = vde_open(init_sock, "QEMU", &args);
1232     if (!s->vde){
1233         free(s);
1234         return -1;
1235     }
1236     s->vc = qemu_new_vlan_client(vlan, model, name, vde_from_qemu, NULL, s);
1237     qemu_set_fd_handler(vde_datafd(s->vde), vde_to_qemu, NULL, s);
1238     snprintf(s->vc->info_str, sizeof(s->vc->info_str), "sock=%s,fd=%d",
1239              sock, vde_datafd(s->vde));
1240     return 0;
1241 }
1242 #endif
1243
1244 /* network connection */
1245 typedef struct NetSocketState {
1246     VLANClientState *vc;
1247     int fd;
1248     int state; /* 0 = getting length, 1 = getting data */
1249     int index;
1250     int packet_len;
1251     uint8_t buf[4096];
1252     struct sockaddr_in dgram_dst; /* contains inet host and port destination iff connectionless (SOCK_DGRAM) */
1253 } NetSocketState;
1254
1255 typedef struct NetSocketListenState {
1256     VLANState *vlan;
1257     char *model;
1258     char *name;
1259     int fd;
1260 } NetSocketListenState;
1261
1262 /* XXX: we consider we can send the whole packet without blocking */
1263 static void net_socket_receive(void *opaque, const uint8_t *buf, int size)
1264 {
1265     NetSocketState *s = opaque;
1266     uint32_t len;
1267     len = htonl(size);
1268
1269     send_all(s->fd, (const uint8_t *)&len, sizeof(len));
1270     send_all(s->fd, buf, size);
1271 }
1272
1273 static void net_socket_receive_dgram(void *opaque, const uint8_t *buf, int size)
1274 {
1275     NetSocketState *s = opaque;
1276     sendto(s->fd, buf, size, 0,
1277            (struct sockaddr *)&s->dgram_dst, sizeof(s->dgram_dst));
1278 }
1279
1280 static void net_socket_send(void *opaque)
1281 {
1282     NetSocketState *s = opaque;
1283     int l, size, err;
1284     uint8_t buf1[4096];
1285     const uint8_t *buf;
1286
1287     size = recv(s->fd, buf1, sizeof(buf1), 0);
1288     if (size < 0) {
1289         err = socket_error();
1290         if (err != EWOULDBLOCK)
1291             goto eoc;
1292     } else if (size == 0) {
1293         /* end of connection */
1294     eoc:
1295         qemu_set_fd_handler(s->fd, NULL, NULL, NULL);
1296         closesocket(s->fd);
1297         return;
1298     }
1299     buf = buf1;
1300     while (size > 0) {
1301         /* reassemble a packet from the network */
1302         switch(s->state) {
1303         case 0:
1304             l = 4 - s->index;
1305             if (l > size)
1306                 l = size;
1307             memcpy(s->buf + s->index, buf, l);
1308             buf += l;
1309             size -= l;
1310             s->index += l;
1311             if (s->index == 4) {
1312                 /* got length */
1313                 s->packet_len = ntohl(*(uint32_t *)s->buf);
1314                 s->index = 0;
1315                 s->state = 1;
1316             }
1317             break;
1318         case 1:
1319             l = s->packet_len - s->index;
1320             if (l > size)
1321                 l = size;
1322             memcpy(s->buf + s->index, buf, l);
1323             s->index += l;
1324             buf += l;
1325             size -= l;
1326             if (s->index >= s->packet_len) {
1327                 qemu_send_packet(s->vc, s->buf, s->packet_len);
1328                 s->index = 0;
1329                 s->state = 0;
1330             }
1331             break;
1332         }
1333     }
1334 }
1335
1336 static void net_socket_send_dgram(void *opaque)
1337 {
1338     NetSocketState *s = opaque;
1339     int size;
1340
1341     size = recv(s->fd, s->buf, sizeof(s->buf), 0);
1342     if (size < 0)
1343         return;
1344     if (size == 0) {
1345         /* end of connection */
1346         qemu_set_fd_handler(s->fd, NULL, NULL, NULL);
1347         return;
1348     }
1349     qemu_send_packet(s->vc, s->buf, size);
1350 }
1351
1352 static int net_socket_mcast_create(struct sockaddr_in *mcastaddr)
1353 {
1354     struct ip_mreq imr;
1355     int fd;
1356     int val, ret;
1357     if (!IN_MULTICAST(ntohl(mcastaddr->sin_addr.s_addr))) {
1358         fprintf(stderr, "qemu: error: specified mcastaddr \"%s\" (0x%08x) does not contain a multicast address\n",
1359                 inet_ntoa(mcastaddr->sin_addr),
1360                 (int)ntohl(mcastaddr->sin_addr.s_addr));
1361         return -1;
1362
1363     }
1364     fd = socket(PF_INET, SOCK_DGRAM, 0);
1365     if (fd < 0) {
1366         perror("socket(PF_INET, SOCK_DGRAM)");
1367         return -1;
1368     }
1369
1370     val = 1;
1371     ret=setsockopt(fd, SOL_SOCKET, SO_REUSEADDR,
1372                    (const char *)&val, sizeof(val));
1373     if (ret < 0) {
1374         perror("setsockopt(SOL_SOCKET, SO_REUSEADDR)");
1375         goto fail;
1376     }
1377
1378     ret = bind(fd, (struct sockaddr *)mcastaddr, sizeof(*mcastaddr));
1379     if (ret < 0) {
1380         perror("bind");
1381         goto fail;
1382     }
1383
1384     /* Add host to multicast group */
1385     imr.imr_multiaddr = mcastaddr->sin_addr;
1386     imr.imr_interface.s_addr = htonl(INADDR_ANY);
1387
1388     ret = setsockopt(fd, IPPROTO_IP, IP_ADD_MEMBERSHIP,
1389                      (const char *)&imr, sizeof(struct ip_mreq));
1390     if (ret < 0) {
1391         perror("setsockopt(IP_ADD_MEMBERSHIP)");
1392         goto fail;
1393     }
1394
1395     /* Force mcast msgs to loopback (eg. several QEMUs in same host */
1396     val = 1;
1397     ret=setsockopt(fd, IPPROTO_IP, IP_MULTICAST_LOOP,
1398                    (const char *)&val, sizeof(val));
1399     if (ret < 0) {
1400         perror("setsockopt(SOL_IP, IP_MULTICAST_LOOP)");
1401         goto fail;
1402     }
1403
1404     socket_set_nonblock(fd);
1405     return fd;
1406 fail:
1407     if (fd >= 0)
1408         closesocket(fd);
1409     return -1;
1410 }
1411
1412 static NetSocketState *net_socket_fd_init_dgram(VLANState *vlan,
1413                                                 const char *model,
1414                                                 const char *name,
1415                                                 int fd, int is_connected)
1416 {
1417     struct sockaddr_in saddr;
1418     int newfd;
1419     socklen_t saddr_len;
1420     NetSocketState *s;
1421
1422     /* fd passed: multicast: "learn" dgram_dst address from bound address and save it
1423      * Because this may be "shared" socket from a "master" process, datagrams would be recv()
1424      * by ONLY ONE process: we must "clone" this dgram socket --jjo
1425      */
1426
1427     if (is_connected) {
1428         if (getsockname(fd, (struct sockaddr *) &saddr, &saddr_len) == 0) {
1429             /* must be bound */
1430             if (saddr.sin_addr.s_addr==0) {
1431                 fprintf(stderr, "qemu: error: init_dgram: fd=%d unbound, cannot setup multicast dst addr\n",
1432                         fd);
1433                 return NULL;
1434             }
1435             /* clone dgram socket */
1436             newfd = net_socket_mcast_create(&saddr);
1437             if (newfd < 0) {
1438                 /* error already reported by net_socket_mcast_create() */
1439                 close(fd);
1440                 return NULL;
1441             }
1442             /* clone newfd to fd, close newfd */
1443             dup2(newfd, fd);
1444             close(newfd);
1445
1446         } else {
1447             fprintf(stderr, "qemu: error: init_dgram: fd=%d failed getsockname(): %s\n",
1448                     fd, strerror(errno));
1449             return NULL;
1450         }
1451     }
1452
1453     s = qemu_mallocz(sizeof(NetSocketState));
1454     s->fd = fd;
1455
1456     s->vc = qemu_new_vlan_client(vlan, model, name, net_socket_receive_dgram, NULL, s);
1457     qemu_set_fd_handler(s->fd, net_socket_send_dgram, NULL, s);
1458
1459     /* mcast: save bound address as dst */
1460     if (is_connected) s->dgram_dst=saddr;
1461
1462     snprintf(s->vc->info_str, sizeof(s->vc->info_str),
1463             "socket: fd=%d (%s mcast=%s:%d)",
1464             fd, is_connected? "cloned" : "",
1465             inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
1466     return s;
1467 }
1468
1469 static void net_socket_connect(void *opaque)
1470 {
1471     NetSocketState *s = opaque;
1472     qemu_set_fd_handler(s->fd, net_socket_send, NULL, s);
1473 }
1474
1475 static NetSocketState *net_socket_fd_init_stream(VLANState *vlan,
1476                                                  const char *model,
1477                                                  const char *name,
1478                                                  int fd, int is_connected)
1479 {
1480     NetSocketState *s;
1481     s = qemu_mallocz(sizeof(NetSocketState));
1482     s->fd = fd;
1483     s->vc = qemu_new_vlan_client(vlan, model, name,
1484                                  net_socket_receive, NULL, s);
1485     snprintf(s->vc->info_str, sizeof(s->vc->info_str),
1486              "socket: fd=%d", fd);
1487     if (is_connected) {
1488         net_socket_connect(s);
1489     } else {
1490         qemu_set_fd_handler(s->fd, NULL, net_socket_connect, s);
1491     }
1492     return s;
1493 }
1494
1495 static NetSocketState *net_socket_fd_init(VLANState *vlan,
1496                                           const char *model, const char *name,
1497                                           int fd, int is_connected)
1498 {
1499     int so_type=-1, optlen=sizeof(so_type);
1500
1501     if(getsockopt(fd, SOL_SOCKET, SO_TYPE, (char *)&so_type,
1502         (socklen_t *)&optlen)< 0) {
1503         fprintf(stderr, "qemu: error: getsockopt(SO_TYPE) for fd=%d failed\n", fd);
1504         return NULL;
1505     }
1506     switch(so_type) {
1507     case SOCK_DGRAM:
1508         return net_socket_fd_init_dgram(vlan, model, name, fd, is_connected);
1509     case SOCK_STREAM:
1510         return net_socket_fd_init_stream(vlan, model, name, fd, is_connected);
1511     default:
1512         /* who knows ... this could be a eg. a pty, do warn and continue as stream */
1513         fprintf(stderr, "qemu: warning: socket type=%d for fd=%d is not SOCK_DGRAM or SOCK_STREAM\n", so_type, fd);
1514         return net_socket_fd_init_stream(vlan, model, name, fd, is_connected);
1515     }
1516     return NULL;
1517 }
1518
1519 static void net_socket_accept(void *opaque)
1520 {
1521     NetSocketListenState *s = opaque;
1522     NetSocketState *s1;
1523     struct sockaddr_in saddr;
1524     socklen_t len;
1525     int fd;
1526
1527     for(;;) {
1528         len = sizeof(saddr);
1529         fd = accept(s->fd, (struct sockaddr *)&saddr, &len);
1530         if (fd < 0 && errno != EINTR) {
1531             return;
1532         } else if (fd >= 0) {
1533             break;
1534         }
1535     }
1536     s1 = net_socket_fd_init(s->vlan, s->model, s->name, fd, 1);
1537     if (!s1) {
1538         closesocket(fd);
1539     } else {
1540         snprintf(s1->vc->info_str, sizeof(s1->vc->info_str),
1541                  "socket: connection from %s:%d",
1542                  inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
1543     }
1544 }
1545
1546 static int net_socket_listen_init(VLANState *vlan,
1547                                   const char *model,
1548                                   const char *name,
1549                                   const char *host_str)
1550 {
1551     NetSocketListenState *s;
1552     int fd, val, ret;
1553     struct sockaddr_in saddr;
1554
1555     if (parse_host_port(&saddr, host_str) < 0)
1556         return -1;
1557
1558     s = qemu_mallocz(sizeof(NetSocketListenState));
1559
1560     fd = socket(PF_INET, SOCK_STREAM, 0);
1561     if (fd < 0) {
1562         perror("socket");
1563         return -1;
1564     }
1565     socket_set_nonblock(fd);
1566
1567     /* allow fast reuse */
1568     val = 1;
1569     setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, (const char *)&val, sizeof(val));
1570
1571     ret = bind(fd, (struct sockaddr *)&saddr, sizeof(saddr));
1572     if (ret < 0) {
1573         perror("bind");
1574         return -1;
1575     }
1576     ret = listen(fd, 0);
1577     if (ret < 0) {
1578         perror("listen");
1579         return -1;
1580     }
1581     s->vlan = vlan;
1582     s->model = strdup(model);
1583     s->name = strdup(name);
1584     s->fd = fd;
1585     qemu_set_fd_handler(fd, net_socket_accept, NULL, s);
1586     return 0;
1587 }
1588
1589 static int net_socket_connect_init(VLANState *vlan,
1590                                    const char *model,
1591                                    const char *name,
1592                                    const char *host_str)
1593 {
1594     NetSocketState *s;
1595     int fd, connected, ret, err;
1596     struct sockaddr_in saddr;
1597
1598     if (parse_host_port(&saddr, host_str) < 0)
1599         return -1;
1600
1601     fd = socket(PF_INET, SOCK_STREAM, 0);
1602     if (fd < 0) {
1603         perror("socket");
1604         return -1;
1605     }
1606     socket_set_nonblock(fd);
1607
1608     connected = 0;
1609     for(;;) {
1610         ret = connect(fd, (struct sockaddr *)&saddr, sizeof(saddr));
1611         if (ret < 0) {
1612             err = socket_error();
1613             if (err == EINTR || err == EWOULDBLOCK) {
1614             } else if (err == EINPROGRESS) {
1615                 break;
1616 #ifdef _WIN32
1617             } else if (err == WSAEALREADY) {
1618                 break;
1619 #endif
1620             } else {
1621                 perror("connect");
1622                 closesocket(fd);
1623                 return -1;
1624             }
1625         } else {
1626             connected = 1;
1627             break;
1628         }
1629     }
1630     s = net_socket_fd_init(vlan, model, name, fd, connected);
1631     if (!s)
1632         return -1;
1633     snprintf(s->vc->info_str, sizeof(s->vc->info_str),
1634              "socket: connect to %s:%d",
1635              inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
1636     return 0;
1637 }
1638
1639 static int net_socket_mcast_init(VLANState *vlan,
1640                                  const char *model,
1641                                  const char *name,
1642                                  const char *host_str)
1643 {
1644     NetSocketState *s;
1645     int fd;
1646     struct sockaddr_in saddr;
1647
1648     if (parse_host_port(&saddr, host_str) < 0)
1649         return -1;
1650
1651
1652     fd = net_socket_mcast_create(&saddr);
1653     if (fd < 0)
1654         return -1;
1655
1656     s = net_socket_fd_init(vlan, model, name, fd, 0);
1657     if (!s)
1658         return -1;
1659
1660     s->dgram_dst = saddr;
1661
1662     snprintf(s->vc->info_str, sizeof(s->vc->info_str),
1663              "socket: mcast=%s:%d",
1664              inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
1665     return 0;
1666
1667 }
1668
1669 /* find or alloc a new VLAN */
1670 VLANState *qemu_find_vlan(int id)
1671 {
1672     VLANState **pvlan, *vlan;
1673     for(vlan = first_vlan; vlan != NULL; vlan = vlan->next) {
1674         if (vlan->id == id)
1675             return vlan;
1676     }
1677     vlan = qemu_mallocz(sizeof(VLANState));
1678     vlan->id = id;
1679     vlan->next = NULL;
1680     pvlan = &first_vlan;
1681     while (*pvlan != NULL)
1682         pvlan = &(*pvlan)->next;
1683     *pvlan = vlan;
1684     return vlan;
1685 }
1686
1687 void qemu_check_nic_model(NICInfo *nd, const char *model)
1688 {
1689     const char *models[2];
1690
1691     models[0] = model;
1692     models[1] = NULL;
1693
1694     qemu_check_nic_model_list(nd, models, model);
1695 }
1696
1697 void qemu_check_nic_model_list(NICInfo *nd, const char * const *models,
1698                                const char *default_model)
1699 {
1700     int i, exit_status = 0;
1701
1702     if (!nd->model)
1703         nd->model = strdup(default_model);
1704
1705     if (strcmp(nd->model, "?") != 0) {
1706         for (i = 0 ; models[i]; i++)
1707             if (strcmp(nd->model, models[i]) == 0)
1708                 return;
1709
1710         fprintf(stderr, "qemu: Unsupported NIC model: %s\n", nd->model);
1711         exit_status = 1;
1712     }
1713
1714     fprintf(stderr, "qemu: Supported NIC models: ");
1715     for (i = 0 ; models[i]; i++)
1716         fprintf(stderr, "%s%c", models[i], models[i+1] ? ',' : '\n');
1717
1718     exit(exit_status);
1719 }
1720
1721 int net_client_init(const char *device, const char *p)
1722 {
1723     char buf[1024];
1724     int vlan_id, ret;
1725     VLANState *vlan;
1726     char *name = NULL;
1727
1728     vlan_id = 0;
1729     if (get_param_value(buf, sizeof(buf), "vlan", p)) {
1730         vlan_id = strtol(buf, NULL, 0);
1731     }
1732     vlan = qemu_find_vlan(vlan_id);
1733     if (!vlan) {
1734         fprintf(stderr, "Could not create vlan %d\n", vlan_id);
1735         return -1;
1736     }
1737     if (get_param_value(buf, sizeof(buf), "name", p)) {
1738         name = strdup(buf);
1739     }
1740     if (!strcmp(device, "nic")) {
1741         NICInfo *nd;
1742         uint8_t *macaddr;
1743
1744         if (nb_nics >= MAX_NICS) {
1745             fprintf(stderr, "Too Many NICs\n");
1746             return -1;
1747         }
1748         nd = &nd_table[nb_nics];
1749         macaddr = nd->macaddr;
1750         macaddr[0] = 0x52;
1751         macaddr[1] = 0x54;
1752         macaddr[2] = 0x00;
1753         macaddr[3] = 0x12;
1754         macaddr[4] = 0x34;
1755         macaddr[5] = 0x56 + nb_nics;
1756
1757         if (get_param_value(buf, sizeof(buf), "macaddr", p)) {
1758             if (parse_macaddr(macaddr, buf) < 0) {
1759                 fprintf(stderr, "invalid syntax for ethernet address\n");
1760                 return -1;
1761             }
1762         }
1763         if (get_param_value(buf, sizeof(buf), "model", p)) {
1764             nd->model = strdup(buf);
1765         }
1766         nd->vlan = vlan;
1767         nd->name = name;
1768         name = NULL;
1769         nb_nics++;
1770         vlan->nb_guest_devs++;
1771         ret = 0;
1772     } else
1773     if (!strcmp(device, "none")) {
1774         /* does nothing. It is needed to signal that no network cards
1775            are wanted */
1776         ret = 0;
1777     } else
1778 #ifdef CONFIG_SLIRP
1779     if (!strcmp(device, "user")) {
1780         if (get_param_value(buf, sizeof(buf), "hostname", p)) {
1781             pstrcpy(slirp_hostname, sizeof(slirp_hostname), buf);
1782         }
1783         if (get_param_value(buf, sizeof(buf), "restrict", p)) {
1784             slirp_restrict = (buf[0] == 'y') ? 1 : 0;
1785         }
1786         if (get_param_value(buf, sizeof(buf), "ip", p)) {
1787             slirp_ip = strdup(buf);
1788         }
1789         vlan->nb_host_devs++;
1790         ret = net_slirp_init(vlan, device, name);
1791     } else
1792 #endif
1793 #ifdef _WIN32
1794     if (!strcmp(device, "tap")) {
1795         char ifname[64];
1796         if (get_param_value(ifname, sizeof(ifname), "ifname", p) <= 0) {
1797             fprintf(stderr, "tap: no interface name\n");
1798             return -1;
1799         }
1800         vlan->nb_host_devs++;
1801         ret = tap_win32_init(vlan, device, name, ifname);
1802     } else
1803 #elif defined (_AIX)
1804 #else
1805     if (!strcmp(device, "tap")) {
1806         char ifname[64];
1807         char setup_script[1024], down_script[1024];
1808         int fd;
1809         vlan->nb_host_devs++;
1810         if (get_param_value(buf, sizeof(buf), "fd", p) > 0) {
1811             fd = strtol(buf, NULL, 0);
1812             fcntl(fd, F_SETFL, O_NONBLOCK);
1813             ret = -1;
1814                 if (net_tap_fd_init(vlan, device, name, fd,
1815                                     tap_probe_vnet_hdr(fd)))
1816                 ret = 0;
1817         } else {
1818             if (get_param_value(ifname, sizeof(ifname), "ifname", p) <= 0) {
1819                 ifname[0] = '\0';
1820             }
1821             if (get_param_value(setup_script, sizeof(setup_script), "script", p) == 0) {
1822                 pstrcpy(setup_script, sizeof(setup_script), DEFAULT_NETWORK_SCRIPT);
1823             }
1824             if (get_param_value(down_script, sizeof(down_script), "downscript", p) == 0) {
1825                 pstrcpy(down_script, sizeof(down_script), DEFAULT_NETWORK_DOWN_SCRIPT);
1826             }
1827             ret = net_tap_init(vlan, device, name, ifname, setup_script, down_script);
1828         }
1829     } else
1830 #endif
1831     if (!strcmp(device, "socket")) {
1832         if (get_param_value(buf, sizeof(buf), "fd", p) > 0) {
1833             int fd;
1834             fd = strtol(buf, NULL, 0);
1835             ret = -1;
1836             if (net_socket_fd_init(vlan, device, name, fd, 1))
1837                 ret = 0;
1838         } else if (get_param_value(buf, sizeof(buf), "listen", p) > 0) {
1839             ret = net_socket_listen_init(vlan, device, name, buf);
1840         } else if (get_param_value(buf, sizeof(buf), "connect", p) > 0) {
1841             ret = net_socket_connect_init(vlan, device, name, buf);
1842         } else if (get_param_value(buf, sizeof(buf), "mcast", p) > 0) {
1843             ret = net_socket_mcast_init(vlan, device, name, buf);
1844         } else {
1845             fprintf(stderr, "Unknown socket options: %s\n", p);
1846             return -1;
1847         }
1848         vlan->nb_host_devs++;
1849     } else
1850 #ifdef CONFIG_VDE
1851     if (!strcmp(device, "vde")) {
1852         char vde_sock[1024], vde_group[512];
1853         int vde_port, vde_mode;
1854         vlan->nb_host_devs++;
1855         if (get_param_value(vde_sock, sizeof(vde_sock), "sock", p) <= 0) {
1856             vde_sock[0] = '\0';
1857         }
1858         if (get_param_value(buf, sizeof(buf), "port", p) > 0) {
1859             vde_port = strtol(buf, NULL, 10);
1860         } else {
1861             vde_port = 0;
1862         }
1863         if (get_param_value(vde_group, sizeof(vde_group), "group", p) <= 0) {
1864             vde_group[0] = '\0';
1865         }
1866         if (get_param_value(buf, sizeof(buf), "mode", p) > 0) {
1867             vde_mode = strtol(buf, NULL, 8);
1868         } else {
1869             vde_mode = 0700;
1870         }
1871         ret = net_vde_init(vlan, device, name, vde_sock, vde_port, vde_group, vde_mode);
1872     } else
1873 #endif
1874     {
1875         fprintf(stderr, "Unknown network device: %s\n", device);
1876         if (name)
1877             free(name);
1878         return -1;
1879     }
1880     if (ret < 0) {
1881         fprintf(stderr, "Could not initialize device '%s'\n", device);
1882     }
1883     if (name)
1884         free(name);
1885     return ret;
1886 }
1887
1888 void net_client_uninit(NICInfo *nd)
1889 {
1890     nd->vlan->nb_guest_devs--; /* XXX: free vlan on last reference */
1891     nb_nics--;
1892     nd->used = 0;
1893     free((void *)nd->model);
1894 }
1895
1896 int net_client_parse(const char *str)
1897 {
1898     const char *p;
1899     char *q;
1900     char device[64];
1901
1902     p = str;
1903     q = device;
1904     while (*p != '\0' && *p != ',') {
1905         if ((q - device) < sizeof(device) - 1)
1906             *q++ = *p;
1907         p++;
1908     }
1909     *q = '\0';
1910     if (*p == ',')
1911         p++;
1912
1913     return net_client_init(device, p);
1914 }
1915
1916 void do_info_network(void)
1917 {
1918     VLANState *vlan;
1919     VLANClientState *vc;
1920
1921     for(vlan = first_vlan; vlan != NULL; vlan = vlan->next) {
1922         term_printf("VLAN %d devices:\n", vlan->id);
1923         for(vc = vlan->first_client; vc != NULL; vc = vc->next)
1924             term_printf("  %s: %s\n", vc->name, vc->info_str);
1925     }
1926 }
1927
1928 int do_set_link(const char *name, const char *up_or_down)
1929 {
1930     VLANState *vlan;
1931     VLANClientState *vc = NULL;
1932
1933     for (vlan = first_vlan; vlan != NULL; vlan = vlan->next)
1934         for (vc = vlan->first_client; vc != NULL; vc = vc->next)
1935             if (strcmp(vc->name, name) == 0)
1936                 goto done;
1937 done:
1938
1939     if (!vc) {
1940         term_printf("could not find network device '%s'", name);
1941         return 0;
1942     }
1943
1944     if (strcmp(up_or_down, "up") == 0)
1945         vc->link_down = 0;
1946     else if (strcmp(up_or_down, "down") == 0)
1947         vc->link_down = 1;
1948     else
1949         term_printf("invalid link status '%s'; only 'up' or 'down' valid\n",
1950                     up_or_down);
1951
1952     if (vc->link_status_changed)
1953         vc->link_status_changed(vc);
1954
1955     return 1;
1956 }
1957
1958 void net_cleanup(void)
1959 {
1960     VLANState *vlan;
1961
1962 #if !defined(_WIN32)
1963     /* close network clients */
1964     for(vlan = first_vlan; vlan != NULL; vlan = vlan->next) {
1965         VLANClientState *vc;
1966
1967         for(vc = vlan->first_client; vc != NULL; vc = vc->next) {
1968             if (vc->fd_read == tap_receive) {
1969                 TAPState *s = vc->opaque;
1970
1971                 if (s->down_script[0])
1972                     launch_script(s->down_script, s->down_script_arg, s->fd);
1973             }
1974 #if defined(CONFIG_VDE)
1975             if (vc->fd_read == vde_from_qemu) {
1976                 VDEState *s = vc->opaque;
1977                 vde_close(s->vde);
1978             }
1979 #endif
1980         }
1981     }
1982 #endif
1983 }
1984
1985 void net_client_check(void)
1986 {
1987     VLANState *vlan;
1988
1989     for(vlan = first_vlan; vlan != NULL; vlan = vlan->next) {
1990         if (vlan->nb_guest_devs == 0 && vlan->nb_host_devs == 0)
1991             continue;
1992         if (vlan->nb_guest_devs == 0)
1993             fprintf(stderr, "Warning: vlan %d with no nics\n", vlan->id);
1994         if (vlan->nb_host_devs == 0)
1995             fprintf(stderr,
1996                     "Warning: vlan %d is not connected to host network\n",
1997                     vlan->id);
1998     }
1999 }