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