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