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 static ssize_t calc_iov_length(const struct iovec *iov, int iovcnt)
 432 {
 433     size_t offset = 0;
 434     int i;
 435
 436     for (i = 0; i < iovcnt; i++)
 437         offset += iov[i].iov_len;
 438     return offset;
 439 }
 440
 441 ssize_t qemu_sendv_packet(VLANClientState *vc1, const struct iovec *iov,
 442                           int iovcnt)
 443 {
 444     VLANState *vlan = vc1->vlan;
 445     VLANClientState *vc;
 446     ssize_t max_len = 0;
 447
 448     if (vc1->link_down)
 449         return calc_iov_length(iov, iovcnt);
 450
 451     for (vc = vlan->first_client; vc != NULL; vc = vc->next) {
 452         ssize_t len = 0;
 453
 454         if (vc == vc1)
 455             continue;
 456
 457         if (vc->link_down)
 458             len = calc_iov_length(iov, iovcnt);
 459         if (vc->fd_readv)
 460             len = vc->fd_readv(vc->opaque, iov, iovcnt);
 461         else if (vc->fd_read)
 462             len = vc_sendv_compat(vc, iov, iovcnt);
 463
 464         max_len = MAX(max_len, len);
 465     }
 466
 467     return max_len;
 468 }
 469
 470 #if defined(CONFIG_SLIRP)
 471
 472 /* slirp network adapter */
 473
 474 static int slirp_inited;
 475 static int slirp_restrict;
 476 static char *slirp_ip;
 477 static VLANClientState *slirp_vc;
 478
 479 int slirp_can_output(void)
 480 {
 481     return !slirp_vc || qemu_can_send_packet(slirp_vc);
 482 }
 483
 484 void slirp_output(const uint8_t *pkt, int pkt_len)
 485 {
 486 #ifdef DEBUG_SLIRP
 487     printf("slirp output:\n");
 488     hex_dump(stdout, pkt, pkt_len);
 489 #endif
 490     if (!slirp_vc)
 491         return;
 492     qemu_send_packet(slirp_vc, pkt, pkt_len);
 493 }
 494
 495 int slirp_is_inited(void)
 496 {
 497     return slirp_inited;
 498 }
 499
 500 static void slirp_receive(void *opaque, const uint8_t *buf, int size)
 501 {
 502 #ifdef DEBUG_SLIRP
 503     printf("slirp input:\n");
 504     hex_dump(stdout, buf, size);
 505 #endif
 506     slirp_input(buf, size);
 507 }
 508
 509 static int net_slirp_init(VLANState *vlan, const char *model, const char *name)
 510 {
 511     if (!slirp_inited) {
 512         slirp_inited = 1;
 513         slirp_init(slirp_restrict, slirp_ip);
 514     }
 515     slirp_vc = qemu_new_vlan_client(vlan, model, name,
 516                                     slirp_receive, NULL, NULL);
 517     slirp_vc->info_str[0] = '\0';
 518     return 0;
 519 }
 520
 521 void net_slirp_redir(const char *redir_str)
 522 {
 523     int is_udp;
 524     char buf[256], *r;
 525     const char *p;
 526     struct in_addr guest_addr;
 527     int host_port, guest_port;
 528
 529     if (!slirp_inited) {
 530         slirp_inited = 1;
 531         slirp_init(slirp_restrict, slirp_ip);
 532     }
 533
 534     p = redir_str;
 535     if (get_str_sep(buf, sizeof(buf), &p, ':') < 0)
 536         goto fail;
 537     if (!strcmp(buf, "tcp")) {
 538         is_udp = 0;
 539     } else if (!strcmp(buf, "udp")) {
 540         is_udp = 1;
 541     } else {
 542         goto fail;
 543     }
 544
 545     if (get_str_sep(buf, sizeof(buf), &p, ':') < 0)
 546         goto fail;
 547     host_port = strtol(buf, &r, 0);
 548     if (r == buf)
 549         goto fail;
 550
 551     if (get_str_sep(buf, sizeof(buf), &p, ':') < 0)
 552         goto fail;
 553     if (buf[0] == '\0') {
 554         pstrcpy(buf, sizeof(buf), "10.0.2.15");
 555     }
 556     if (!inet_aton(buf, &guest_addr))
 557         goto fail;
 558
 559     guest_port = strtol(p, &r, 0);
 560     if (r == p)
 561         goto fail;
 562
 563     if (slirp_redir(is_udp, host_port, guest_addr, guest_port) < 0) {
 564         fprintf(stderr, "qemu: could not set up redirection\n");
 565         exit(1);
 566     }
 567     return;
 568  fail:
 569     fprintf(stderr, "qemu: syntax: -redir [tcp|udp]:host-port:[guest-host]:guest-port\n");
 570     exit(1);
 571 }
 572
 573 #ifndef _WIN32
 574
 575 static char smb_dir[1024];
 576
 577 static void erase_dir(char *dir_name)
 578 {
 579     DIR *d;
 580     struct dirent *de;
 581     char filename[1024];
 582
 583     /* erase all the files in the directory */
 584     if ((d = opendir(dir_name)) != 0) {
 585         for(;;) {
 586             de = readdir(d);
 587             if (!de)
 588                 break;
 589             if (strcmp(de->d_name, ".") != 0 &&
 590                 strcmp(de->d_name, "..") != 0) {
 591                 snprintf(filename, sizeof(filename), "%s/%s",
 592                          smb_dir, de->d_name);
 593                 if (unlink(filename) != 0)  /* is it a directory? */
 594                     erase_dir(filename);
 595             }
 596         }
 597         closedir(d);
 598         rmdir(dir_name);
 599     }
 600 }
 601
 602 /* automatic user mode samba server configuration */
 603 static void smb_exit(void)
 604 {
 605     erase_dir(smb_dir);
 606 }
 607
 608 /* automatic user mode samba server configuration */
 609 void net_slirp_smb(const char *exported_dir)
 610 {
 611     char smb_conf[1024];
 612     char smb_cmdline[1024];
 613     FILE *f;
 614
 615     if (!slirp_inited) {
 616         slirp_inited = 1;
 617         slirp_init(slirp_restrict, slirp_ip);
 618     }
 619
 620     /* XXX: better tmp dir construction */
 621     snprintf(smb_dir, sizeof(smb_dir), "/tmp/qemu-smb.%d", getpid());
 622     if (mkdir(smb_dir, 0700) < 0) {
 623         fprintf(stderr, "qemu: could not create samba server dir '%s'\n", smb_dir);
 624         exit(1);
 625     }
 626     snprintf(smb_conf, sizeof(smb_conf), "%s/%s", smb_dir, "smb.conf");
 627
 628     f = fopen(smb_conf, "w");
 629     if (!f) {
 630         fprintf(stderr, "qemu: could not create samba server configuration file '%s'\n", smb_conf);
 631         exit(1);
 632     }
 633     fprintf(f,
 634             "[global]\n"
 635             "private dir=%s\n"
 636             "smb ports=0\n"
 637             "socket address=127.0.0.1\n"
 638             "pid directory=%s\n"
 639             "lock directory=%s\n"
 640             "log file=%s/log.smbd\n"
 641             "smb passwd file=%s/smbpasswd\n"
 642             "security = share\n"
 643             "[qemu]\n"
 644             "path=%s\n"
 645             "read only=no\n"
 646             "guest ok=yes\n",
 647             smb_dir,
 648             smb_dir,
 649             smb_dir,
 650             smb_dir,
 651             smb_dir,
 652             exported_dir
 653             );
 654     fclose(f);
 655     atexit(smb_exit);
 656
 657     snprintf(smb_cmdline, sizeof(smb_cmdline), "%s -s %s",
 658              SMBD_COMMAND, smb_conf);
 659
 660     slirp_add_exec(0, smb_cmdline, 4, 139);
 661 }
 662
 663 #endif /* !defined(_WIN32) */
 664 void do_info_slirp(void)
 665 {
 666     slirp_stats();
 667 }
 668
 669 #endif /* CONFIG_SLIRP */
 670
 671 #ifdef _WIN32
 672
 673 int tap_has_vnet_hdr(void *opaque)
 674 {
 675     return 0;
 676 }
 677
 678 void tap_using_vnet_hdr(void *opaque, int using_vnet_hdr)
 679 {
 680 }
 681
 682 #else /* !defined(_WIN32) */
 683
 684 /* Maximum GSO packet size (64k) plus plenty of room for
 685  * the ethernet and virtio_net headers
 686  */
 687 #define TAP_BUFSIZE (4096 + 65536)
 688
 689 #ifdef IFF_VNET_HDR
 690 #include <linux/virtio_net.h>
 691 #endif
 692
 693 typedef struct TAPState {
 694     VLANClientState *vc;
 695     int fd;
 696     char down_script[1024];
 697     char down_script_arg[128];
 698     char buf[TAP_BUFSIZE];
 699     int size;
 700     unsigned int has_vnet_hdr : 1;
 701     unsigned int using_vnet_hdr : 1;
 702 } TAPState;
 703
 704 #ifdef HAVE_IOVEC
 705 static ssize_t tap_receive_iov(void *opaque, const struct iovec *iov,
 706                                int iovcnt)
 707 {
 708     TAPState *s = opaque;
 709     ssize_t len;
 710
 711     do {
 712         len = writev(s->fd, iov, iovcnt);
 713     } while (len == -1 && (errno == EINTR || errno == EAGAIN));
 714
 715     return len;
 716 }
 717 #endif
 718
 719 static void tap_receive(void *opaque, const uint8_t *buf, int size)
 720 {
 721     struct iovec iov[2];
 722     int i = 0;
 723
 724 #ifdef IFF_VNET_HDR
 725     TAPState *s = opaque;
 726     struct virtio_net_hdr hdr = { 0, };
 727
 728     if (s->has_vnet_hdr && !s->using_vnet_hdr) {
 729         iov[i].iov_base = &hdr;
 730         iov[i].iov_len  = sizeof(hdr);
 731         i++;
 732     }
 733 #endif
 734
 735     iov[i].iov_base = (char *) buf;
 736     iov[i].iov_len  = size;
 737     i++;
 738
 739     tap_receive_iov(opaque, iov, i);
 740 }
 741
 742 static int tap_can_send(void *opaque)
 743 {
 744     TAPState *s = opaque;
 745     VLANClientState *vc;
 746     int can_receive = 0;
 747
 748     /* Check to see if any of our clients can receive a packet */
 749     for (vc = s->vc->vlan->first_client; vc; vc = vc->next) {
 750         /* Skip ourselves */
 751         if (vc == s->vc)
 752             continue;
 753
 754         if (!vc->fd_can_read) {
 755             /* no fd_can_read handler, they always can receive */
 756             can_receive = 1;
 757         } else
 758             can_receive = vc->fd_can_read(vc->opaque);
 759
 760         /* Once someone can receive, we try to send a packet */
 761         if (can_receive)
 762             break;
 763     }
 764
 765     return can_receive;
 766 }
 767
 768 static int tap_send_packet(TAPState *s)
 769 {
 770     uint8_t *buf = s->buf;
 771     int size = s->size;
 772
 773 #ifdef IFF_VNET_HDR
 774     if (s->has_vnet_hdr && !s->using_vnet_hdr) {
 775         buf += sizeof(struct virtio_net_hdr);
 776         size -= sizeof(struct virtio_net_hdr);
 777     }
 778 #endif
 779
 780     return qemu_send_packet(s->vc, buf, size);
 781 }
 782
 783 static void tap_send(void *opaque)
 784 {
 785     TAPState *s = opaque;
 786
 787     /* First try to send any buffered packet */
 788     if (s->size > 0) {
 789         int err;
 790
 791         /* If noone can receive the packet, buffer it */
 792         err = tap_send_packet(s);
 793         if (err == -EAGAIN)
 794             return;
 795     }
 796
 797     /* Read packets until we hit EAGAIN */
 798     do {
 799 #ifdef __sun__
 800         struct strbuf sbuf;
 801         int f = 0;
 802         sbuf.maxlen = sizeof(s->buf);
 803         sbuf.buf = s->buf;
 804         s->size = getmsg(s->fd, NULL, &sbuf, &f) >=0 ? sbuf.len : -1;
 805 #else
 806         // FIXME: kvm_sleep_begin();
 807         s->size = read(s->fd, s->buf, sizeof(s->buf));
 808         // FIXME: kvm_sleep_end();
 809 #endif
 810
 811         if (s->size == -1 && errno == EINTR)
 812             continue;
 813
 814         if (s->size > 0) {
 815             int err;
 816
 817             /* If noone can receive the packet, buffer it */
 818             err = tap_send_packet(s);
 819             if (err == -EAGAIN)
 820                 break;
 821         }
 822     } while (s->size > 0);
 823 }
 824
 825 int tap_has_vnet_hdr(void *opaque)
 826 {
 827     VLANClientState *vc = opaque;
 828     TAPState *s = vc->opaque;
 829
 830     return s ? s->has_vnet_hdr : 0;
 831 }
 832
 833 void tap_using_vnet_hdr(void *opaque, int using_vnet_hdr)
 834 {
 835     VLANClientState *vc = opaque;
 836     TAPState *s = vc->opaque;
 837
 838     if (!s || !s->has_vnet_hdr)
 839         return;
 840
 841     s->using_vnet_hdr = using_vnet_hdr != 0;
 842 }
 843
 844 static int tap_probe_vnet_hdr(int fd)
 845 {
 846 #if defined(TUNGETIFF) && defined(IFF_VNET_HDR)
 847     struct ifreq ifr;
 848
 849     if (ioctl(fd, TUNGETIFF, &ifr) != 0) {
 850         fprintf(stderr, "TUNGETIFF ioctl() failed: %s\n", strerror(errno));
 851         return 0;
 852     }
 853
 854     return ifr.ifr_flags & IFF_VNET_HDR;
 855 #else
 856     return 0;
 857 #endif
 858 }
 859
 860 #ifdef TUNSETOFFLOAD
 861 static void tap_set_offload(VLANClientState *vc, int csum, int tso4, int tso6,
 862                             int ecn)
 863 {
 864     TAPState *s = vc->opaque;
 865     unsigned int offload = 0;
 866
 867     if (csum) {
 868         offload |= TUN_F_CSUM;
 869         if (tso4)
 870             offload |= TUN_F_TSO4;
 871         if (tso6)
 872             offload |= TUN_F_TSO6;
 873         if ((tso4 || tso6) && ecn)
 874             offload |= TUN_F_TSO_ECN;
 875     }
 876
 877     if (ioctl(s->fd, TUNSETOFFLOAD, offload) != 0)
 878         fprintf(stderr, "TUNSETOFFLOAD ioctl() failed: %s\n",
 879                 strerror(errno));
 880 }
 881 #endif /* TUNSETOFFLOAD */
 882
 883 /* fd support */
 884
 885 static TAPState *net_tap_fd_init(VLANState *vlan,
 886                                  const char *model,
 887                                  const char *name,
 888                                  int fd,
 889                                  int vnet_hdr)
 890 {
 891     TAPState *s;
 892
 893     s = qemu_mallocz(sizeof(TAPState));
 894     if (!s)
 895         return NULL;
 896     s->fd = fd;
 897     s->has_vnet_hdr = vnet_hdr != 0;
 898     s->vc = qemu_new_vlan_client(vlan, model, name, tap_receive, NULL, s);
 899 #ifdef HAVE_IOVEC
 900     s->vc->fd_readv = tap_receive_iov;
 901 #endif
 902 #ifdef TUNSETOFFLOAD
 903     s->vc->set_offload = tap_set_offload;
 904 #endif
 905     qemu_set_fd_handler2(s->fd, tap_can_send, tap_send, NULL, s);
 906     snprintf(s->vc->info_str, sizeof(s->vc->info_str), "fd=%d", fd);
 907     return s;
 908 }
 909
 910 #if defined (_BSD) || defined (__FreeBSD_kernel__)
 911 static int tap_open(char *ifname, int ifname_size)
 912 {
 913     int fd;
 914     char *dev;
 915     struct stat s;
 916
 917     TFR(fd = open("/dev/tap", O_RDWR));
 918     if (fd < 0) {
 919         fprintf(stderr, "warning: could not open /dev/tap: no virtual network emulation\n");
 920         return -1;
 921     }
 922
 923     fstat(fd, &s);
 924     dev = devname(s.st_rdev, S_IFCHR);
 925     pstrcpy(ifname, ifname_size, dev);
 926
 927     fcntl(fd, F_SETFL, O_NONBLOCK);
 928     return fd;
 929 }
 930 #elif defined(__sun__)
 931 #define TUNNEWPPA       (('T'<<16) | 0x0001)
 932 /*
 933  * Allocate TAP device, returns opened fd.
 934  * Stores dev name in the first arg(must be large enough).
 935  */
 936 int tap_alloc(char *dev, size_t dev_size)
 937 {
 938     int tap_fd, if_fd, ppa = -1;
 939     static int ip_fd = 0;
 940     char *ptr;
 941
 942     static int arp_fd = 0;
 943     int ip_muxid, arp_muxid;
 944     struct strioctl  strioc_if, strioc_ppa;
 945     int link_type = I_PLINK;;
 946     struct lifreq ifr;
 947     char actual_name[32] = "";
 948
 949     memset(&ifr, 0x0, sizeof(ifr));
 950
 951     if( *dev ){
 952        ptr = dev;
 953        while( *ptr && !qemu_isdigit((int)*ptr) ) ptr++;
 954        ppa = atoi(ptr);
 955     }
 956
 957     /* Check if IP device was opened */
 958     if( ip_fd )
 959        close(ip_fd);
 960
 961     TFR(ip_fd = open("/dev/udp", O_RDWR, 0));
 962     if (ip_fd < 0) {
 963        syslog(LOG_ERR, "Can't open /dev/ip (actually /dev/udp)");
 964        return -1;
 965     }
 966
 967     TFR(tap_fd = open("/dev/tap", O_RDWR, 0));
 968     if (tap_fd < 0) {
 969        syslog(LOG_ERR, "Can't open /dev/tap");
 970        return -1;
 971     }
 972
 973     /* Assign a new PPA and get its unit number. */
 974     strioc_ppa.ic_cmd = TUNNEWPPA;
 975     strioc_ppa.ic_timout = 0;
 976     strioc_ppa.ic_len = sizeof(ppa);
 977     strioc_ppa.ic_dp = (char *)&ppa;
 978     if ((ppa = ioctl (tap_fd, I_STR, &strioc_ppa)) < 0)
 979        syslog (LOG_ERR, "Can't assign new interface");
 980
 981     TFR(if_fd = open("/dev/tap", O_RDWR, 0));
 982     if (if_fd < 0) {
 983        syslog(LOG_ERR, "Can't open /dev/tap (2)");
 984        return -1;
 985     }
 986     if(ioctl(if_fd, I_PUSH, "ip") < 0){
 987        syslog(LOG_ERR, "Can't push IP module");
 988        return -1;
 989     }
 990
 991     if (ioctl(if_fd, SIOCGLIFFLAGS, &ifr) < 0)
 992         syslog(LOG_ERR, "Can't get flags\n");
 993
 994     snprintf (actual_name, 32, "tap%d", ppa);
 995     pstrcpy(ifr.lifr_name, sizeof(ifr.lifr_name), actual_name);
 996
 997     ifr.lifr_ppa = ppa;
 998     /* Assign ppa according to the unit number returned by tun device */
 999
1000     if (ioctl (if_fd, SIOCSLIFNAME, &ifr) < 0)
1001         syslog (LOG_ERR, "Can't set PPA %d", ppa);
1002     if (ioctl(if_fd, SIOCGLIFFLAGS, &ifr) <0)
1003         syslog (LOG_ERR, "Can't get flags\n");
1004     /* Push arp module to if_fd */
1005     if (ioctl (if_fd, I_PUSH, "arp") < 0)
1006         syslog (LOG_ERR, "Can't push ARP module (2)");
1007
1008     /* Push arp module to ip_fd */
1009     if (ioctl (ip_fd, I_POP, NULL) < 0)
1010         syslog (LOG_ERR, "I_POP failed\n");
1011     if (ioctl (ip_fd, I_PUSH, "arp") < 0)
1012         syslog (LOG_ERR, "Can't push ARP module (3)\n");
1013     /* Open arp_fd */
1014     TFR(arp_fd = open ("/dev/tap", O_RDWR, 0));
1015     if (arp_fd < 0)
1016        syslog (LOG_ERR, "Can't open %s\n", "/dev/tap");
1017
1018     /* Set ifname to arp */
1019     strioc_if.ic_cmd = SIOCSLIFNAME;
1020     strioc_if.ic_timout = 0;
1021     strioc_if.ic_len = sizeof(ifr);
1022     strioc_if.ic_dp = (char *)&ifr;
1023     if (ioctl(arp_fd, I_STR, &strioc_if) < 0){
1024         syslog (LOG_ERR, "Can't set ifname to arp\n");
1025     }
1026
1027     if((ip_muxid = ioctl(ip_fd, I_LINK, if_fd)) < 0){
1028        syslog(LOG_ERR, "Can't link TAP device to IP");
1029        return -1;
1030     }
1031
1032     if ((arp_muxid = ioctl (ip_fd, link_type, arp_fd)) < 0)
1033         syslog (LOG_ERR, "Can't link TAP device to ARP");
1034
1035     close (if_fd);
1036
1037     memset(&ifr, 0x0, sizeof(ifr));
1038     pstrcpy(ifr.lifr_name, sizeof(ifr.lifr_name), actual_name);
1039     ifr.lifr_ip_muxid  = ip_muxid;
1040     ifr.lifr_arp_muxid = arp_muxid;
1041
1042     if (ioctl (ip_fd, SIOCSLIFMUXID, &ifr) < 0)
1043     {
1044       ioctl (ip_fd, I_PUNLINK , arp_muxid);
1045       ioctl (ip_fd, I_PUNLINK, ip_muxid);
1046       syslog (LOG_ERR, "Can't set multiplexor id");
1047     }
1048
1049     snprintf(dev, dev_size, "tap%d", ppa);
1050     return tap_fd;
1051 }
1052
1053 static int tap_open(char *ifname, int ifname_size, int *vnet_hdr)
1054 {
1055     char  dev[10]="";
1056     int fd;
1057     if( (fd = tap_alloc(dev, sizeof(dev))) < 0 ){
1058        fprintf(stderr, "Cannot allocate TAP device\n");
1059        return -1;
1060     }
1061     pstrcpy(ifname, ifname_size, dev);
1062     fcntl(fd, F_SETFL, O_NONBLOCK);
1063     return fd;
1064 }
1065 #elif defined (_AIX)
1066 static int tap_open(char *ifname, int ifname_size)
1067 {
1068     fprintf (stderr, "no tap on AIX\n");
1069     return -1;
1070 }
1071 #else
1072 static int tap_open(char *ifname, int ifname_size, int *vnet_hdr)
1073 {
1074     struct ifreq ifr;
1075     int fd, ret;
1076
1077     TFR(fd = open("/dev/net/tun", O_RDWR));
1078     if (fd < 0) {
1079         fprintf(stderr, "warning: could not open /dev/net/tun: no virtual network emulation\n");
1080         return -1;
1081     }
1082     memset(&ifr, 0, sizeof(ifr));
1083     ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
1084
1085 #if defined(TUNGETFEATURES) && defined(IFF_VNET_HDR)
1086     {
1087         unsigned int features;
1088
1089         if (ioctl(fd, TUNGETFEATURES, &features) == 0 &&
1090             features & IFF_VNET_HDR) {
1091             *vnet_hdr = 1;
1092             ifr.ifr_flags |= IFF_VNET_HDR;
1093         }
1094     }
1095 #endif
1096
1097     if (ifname[0] != '\0')
1098         pstrcpy(ifr.ifr_name, IFNAMSIZ, ifname);
1099     else
1100         pstrcpy(ifr.ifr_name, IFNAMSIZ, "tap%d");
1101     ret = ioctl(fd, TUNSETIFF, (void *) &ifr);
1102     if (ret != 0) {
1103         fprintf(stderr, "warning: could not configure /dev/net/tun: no virtual network emulation\n");
1104         close(fd);
1105         return -1;
1106     }
1107     pstrcpy(ifname, ifname_size, ifr.ifr_name);
1108     fcntl(fd, F_SETFL, O_NONBLOCK);
1109     return fd;
1110 }
1111 #endif
1112
1113 static int launch_script(const char *setup_script, const char *ifname, int fd)
1114 {
1115     int pid, status;
1116     char *args[3];
1117     char **parg;
1118
1119         /* try to launch network script */
1120         pid = fork();
1121         if (pid >= 0) {
1122             if (pid == 0) {
1123                 int open_max = sysconf (_SC_OPEN_MAX), i;
1124                 for (i = 0; i < open_max; i++)
1125                     if (i != STDIN_FILENO &&
1126                         i != STDOUT_FILENO &&
1127                         i != STDERR_FILENO &&
1128                         i != fd)
1129                         close(i);
1130
1131                 parg = args;
1132                 *parg++ = (char *)setup_script;
1133                 *parg++ = (char *)ifname;
1134                 *parg++ = NULL;
1135                 execv(setup_script, args);
1136                 _exit(1);
1137             }
1138             while (waitpid(pid, &status, 0) != pid);
1139             if (!WIFEXITED(status) ||
1140                 WEXITSTATUS(status) != 0) {
1141                 fprintf(stderr, "%s: could not launch network script\n",
1142                         setup_script);
1143                 return -1;
1144             }
1145         }
1146     return 0;
1147 }
1148
1149 static int net_tap_init(VLANState *vlan, const char *model,
1150                         const char *name, const char *ifname1,
1151                         const char *setup_script, const char *down_script)
1152 {
1153     TAPState *s;
1154     int fd;
1155     int vnet_hdr;
1156     char ifname[128];
1157
1158     if (ifname1 != NULL)
1159         pstrcpy(ifname, sizeof(ifname), ifname1);
1160     else
1161         ifname[0] = '\0';
1162     vnet_hdr = 0;
1163     TFR(fd = tap_open(ifname, sizeof(ifname), &vnet_hdr));
1164     if (fd < 0)
1165         return -1;
1166
1167     if (!setup_script || !strcmp(setup_script, "no"))
1168         setup_script = "";
1169     if (setup_script[0] != '\0') {
1170         if (launch_script(setup_script, ifname, fd))
1171             return -1;
1172     }
1173     s = net_tap_fd_init(vlan, model, name, fd, vnet_hdr);
1174     if (!s)
1175         return -1;
1176
1177     snprintf(s->vc->info_str, sizeof(s->vc->info_str),
1178              "ifname=%s,script=%s,downscript=%s",
1179              ifname, setup_script, down_script);
1180     if (down_script && strcmp(down_script, "no")) {
1181         snprintf(s->down_script, sizeof(s->down_script), "%s", down_script);
1182         snprintf(s->down_script_arg, sizeof(s->down_script_arg), "%s", ifname);
1183     }
1184     return 0;
1185 }
1186
1187 #endif /* !_WIN32 */
1188
1189 #if defined(CONFIG_VDE)
1190 typedef struct VDEState {
1191     VLANClientState *vc;
1192     VDECONN *vde;
1193 } VDEState;
1194
1195 static void vde_to_qemu(void *opaque)
1196 {
1197     VDEState *s = opaque;
1198     uint8_t buf[4096];
1199     int size;
1200
1201     size = vde_recv(s->vde, buf, sizeof(buf), 0);
1202     if (size > 0) {
1203         qemu_send_packet(s->vc, buf, size);
1204     }
1205 }
1206
1207 static void vde_from_qemu(void *opaque, const uint8_t *buf, int size)
1208 {
1209     VDEState *s = opaque;
1210     int ret;
1211     for(;;) {
1212         ret = vde_send(s->vde, buf, size, 0);
1213         if (ret < 0 && errno == EINTR) {
1214         } else {
1215             break;
1216         }
1217     }
1218 }
1219
1220 static int net_vde_init(VLANState *vlan, const char *model,
1221                         const char *name, const char *sock,
1222                         int port, const char *group, int mode)
1223 {
1224     VDEState *s;
1225     char *init_group = strlen(group) ? (char *)group : NULL;
1226     char *init_sock = strlen(sock) ? (char *)sock : NULL;
1227
1228     struct vde_open_args args = {
1229         .port = port,
1230         .group = init_group,
1231         .mode = mode,
1232     };
1233
1234     s = qemu_mallocz(sizeof(VDEState));
1235     if (!s)
1236         return -1;
1237     s->vde = vde_open(init_sock, "QEMU", &args);
1238     if (!s->vde){
1239         free(s);
1240         return -1;
1241     }
1242     s->vc = qemu_new_vlan_client(vlan, model, name, vde_from_qemu, NULL, s);
1243     qemu_set_fd_handler(vde_datafd(s->vde), vde_to_qemu, NULL, s);
1244     snprintf(s->vc->info_str, sizeof(s->vc->info_str), "sock=%s,fd=%d",
1245              sock, vde_datafd(s->vde));
1246     return 0;
1247 }
1248 #endif
1249
1250 /* network connection */
1251 typedef struct NetSocketState {
1252     VLANClientState *vc;
1253     int fd;
1254     int state; /* 0 = getting length, 1 = getting data */
1255     int index;
1256     int packet_len;
1257     uint8_t buf[4096];
1258     struct sockaddr_in dgram_dst; /* contains inet host and port destination iff connectionless (SOCK_DGRAM) */
1259 } NetSocketState;
1260
1261 typedef struct NetSocketListenState {
1262     VLANState *vlan;
1263     char *model;
1264     char *name;
1265     int fd;
1266 } NetSocketListenState;
1267
1268 /* XXX: we consider we can send the whole packet without blocking */
1269 static void net_socket_receive(void *opaque, const uint8_t *buf, int size)
1270 {
1271     NetSocketState *s = opaque;
1272     uint32_t len;
1273     len = htonl(size);
1274
1275     send_all(s->fd, (const uint8_t *)&len, sizeof(len));
1276     send_all(s->fd, buf, size);
1277 }
1278
1279 static void net_socket_receive_dgram(void *opaque, const uint8_t *buf, int size)
1280 {
1281     NetSocketState *s = opaque;
1282     sendto(s->fd, buf, size, 0,
1283            (struct sockaddr *)&s->dgram_dst, sizeof(s->dgram_dst));
1284 }
1285
1286 static void net_socket_send(void *opaque)
1287 {
1288     NetSocketState *s = opaque;
1289     int l, size, err;
1290     uint8_t buf1[4096];
1291     const uint8_t *buf;
1292
1293     size = recv(s->fd, buf1, sizeof(buf1), 0);
1294     if (size < 0) {
1295         err = socket_error();
1296         if (err != EWOULDBLOCK)
1297             goto eoc;
1298     } else if (size == 0) {
1299         /* end of connection */
1300     eoc:
1301         qemu_set_fd_handler(s->fd, NULL, NULL, NULL);
1302         closesocket(s->fd);
1303         return;
1304     }
1305     buf = buf1;
1306     while (size > 0) {
1307         /* reassemble a packet from the network */
1308         switch(s->state) {
1309         case 0:
1310             l = 4 - s->index;
1311             if (l > size)
1312                 l = size;
1313             memcpy(s->buf + s->index, buf, l);
1314             buf += l;
1315             size -= l;
1316             s->index += l;
1317             if (s->index == 4) {
1318                 /* got length */
1319                 s->packet_len = ntohl(*(uint32_t *)s->buf);
1320                 s->index = 0;
1321                 s->state = 1;
1322             }
1323             break;
1324         case 1:
1325             l = s->packet_len - s->index;
1326             if (l > size)
1327                 l = size;
1328             memcpy(s->buf + s->index, buf, l);
1329             s->index += l;
1330             buf += l;
1331             size -= l;
1332             if (s->index >= s->packet_len) {
1333                 qemu_send_packet(s->vc, s->buf, s->packet_len);
1334                 s->index = 0;
1335                 s->state = 0;
1336             }
1337             break;
1338         }
1339     }
1340 }
1341
1342 static void net_socket_send_dgram(void *opaque)
1343 {
1344     NetSocketState *s = opaque;
1345     int size;
1346
1347     size = recv(s->fd, s->buf, sizeof(s->buf), 0);
1348     if (size < 0)
1349         return;
1350     if (size == 0) {
1351         /* end of connection */
1352         qemu_set_fd_handler(s->fd, NULL, NULL, NULL);
1353         return;
1354     }
1355     qemu_send_packet(s->vc, s->buf, size);
1356 }
1357
1358 static int net_socket_mcast_create(struct sockaddr_in *mcastaddr)
1359 {
1360     struct ip_mreq imr;
1361     int fd;
1362     int val, ret;
1363     if (!IN_MULTICAST(ntohl(mcastaddr->sin_addr.s_addr))) {
1364         fprintf(stderr, "qemu: error: specified mcastaddr \"%s\" (0x%08x) does not contain a multicast address\n",
1365                 inet_ntoa(mcastaddr->sin_addr),
1366                 (int)ntohl(mcastaddr->sin_addr.s_addr));
1367         return -1;
1368
1369     }
1370     fd = socket(PF_INET, SOCK_DGRAM, 0);
1371     if (fd < 0) {
1372         perror("socket(PF_INET, SOCK_DGRAM)");
1373         return -1;
1374     }
1375
1376     val = 1;
1377     ret=setsockopt(fd, SOL_SOCKET, SO_REUSEADDR,
1378                    (const char *)&val, sizeof(val));
1379     if (ret < 0) {
1380         perror("setsockopt(SOL_SOCKET, SO_REUSEADDR)");
1381         goto fail;
1382     }
1383
1384     ret = bind(fd, (struct sockaddr *)mcastaddr, sizeof(*mcastaddr));
1385     if (ret < 0) {
1386         perror("bind");
1387         goto fail;
1388     }
1389
1390     /* Add host to multicast group */
1391     imr.imr_multiaddr = mcastaddr->sin_addr;
1392     imr.imr_interface.s_addr = htonl(INADDR_ANY);
1393
1394     ret = setsockopt(fd, IPPROTO_IP, IP_ADD_MEMBERSHIP,
1395                      (const char *)&imr, sizeof(struct ip_mreq));
1396     if (ret < 0) {
1397         perror("setsockopt(IP_ADD_MEMBERSHIP)");
1398         goto fail;
1399     }
1400
1401     /* Force mcast msgs to loopback (eg. several QEMUs in same host */
1402     val = 1;
1403     ret=setsockopt(fd, IPPROTO_IP, IP_MULTICAST_LOOP,
1404                    (const char *)&val, sizeof(val));
1405     if (ret < 0) {
1406         perror("setsockopt(SOL_IP, IP_MULTICAST_LOOP)");
1407         goto fail;
1408     }
1409
1410     socket_set_nonblock(fd);
1411     return fd;
1412 fail:
1413     if (fd >= 0)
1414         closesocket(fd);
1415     return -1;
1416 }
1417
1418 static NetSocketState *net_socket_fd_init_dgram(VLANState *vlan,
1419                                                 const char *model,
1420                                                 const char *name,
1421                                                 int fd, int is_connected)
1422 {
1423     struct sockaddr_in saddr;
1424     int newfd;
1425     socklen_t saddr_len;
1426     NetSocketState *s;
1427
1428     /* fd passed: multicast: "learn" dgram_dst address from bound address and save it
1429      * Because this may be "shared" socket from a "master" process, datagrams would be recv()
1430      * by ONLY ONE process: we must "clone" this dgram socket --jjo
1431      */
1432
1433     if (is_connected) {
1434         if (getsockname(fd, (struct sockaddr *) &saddr, &saddr_len) == 0) {
1435             /* must be bound */
1436             if (saddr.sin_addr.s_addr==0) {
1437                 fprintf(stderr, "qemu: error: init_dgram: fd=%d unbound, cannot setup multicast dst addr\n",
1438                         fd);
1439                 return NULL;
1440             }
1441             /* clone dgram socket */
1442             newfd = net_socket_mcast_create(&saddr);
1443             if (newfd < 0) {
1444                 /* error already reported by net_socket_mcast_create() */
1445                 close(fd);
1446                 return NULL;
1447             }
1448             /* clone newfd to fd, close newfd */
1449             dup2(newfd, fd);
1450             close(newfd);
1451
1452         } else {
1453             fprintf(stderr, "qemu: error: init_dgram: fd=%d failed getsockname(): %s\n",
1454                     fd, strerror(errno));
1455             return NULL;
1456         }
1457     }
1458
1459     s = qemu_mallocz(sizeof(NetSocketState));
1460     if (!s)
1461         return NULL;
1462     s->fd = fd;
1463
1464     s->vc = qemu_new_vlan_client(vlan, model, name, net_socket_receive_dgram, NULL, s);
1465     qemu_set_fd_handler(s->fd, net_socket_send_dgram, NULL, s);
1466
1467     /* mcast: save bound address as dst */
1468     if (is_connected) s->dgram_dst=saddr;
1469
1470     snprintf(s->vc->info_str, sizeof(s->vc->info_str),
1471             "socket: fd=%d (%s mcast=%s:%d)",
1472             fd, is_connected? "cloned" : "",
1473             inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
1474     return s;
1475 }
1476
1477 static void net_socket_connect(void *opaque)
1478 {
1479     NetSocketState *s = opaque;
1480     qemu_set_fd_handler(s->fd, net_socket_send, NULL, s);
1481 }
1482
1483 static NetSocketState *net_socket_fd_init_stream(VLANState *vlan,
1484                                                  const char *model,
1485                                                  const char *name,
1486                                                  int fd, int is_connected)
1487 {
1488     NetSocketState *s;
1489     s = qemu_mallocz(sizeof(NetSocketState));
1490     if (!s)
1491         return NULL;
1492     s->fd = fd;
1493     s->vc = qemu_new_vlan_client(vlan, model, name,
1494                                  net_socket_receive, NULL, s);
1495     snprintf(s->vc->info_str, sizeof(s->vc->info_str),
1496              "socket: fd=%d", fd);
1497     if (is_connected) {
1498         net_socket_connect(s);
1499     } else {
1500         qemu_set_fd_handler(s->fd, NULL, net_socket_connect, s);
1501     }
1502     return s;
1503 }
1504
1505 static NetSocketState *net_socket_fd_init(VLANState *vlan,
1506                                           const char *model, const char *name,
1507                                           int fd, int is_connected)
1508 {
1509     int so_type=-1, optlen=sizeof(so_type);
1510
1511     if(getsockopt(fd, SOL_SOCKET, SO_TYPE, (char *)&so_type,
1512         (socklen_t *)&optlen)< 0) {
1513         fprintf(stderr, "qemu: error: getsockopt(SO_TYPE) for fd=%d failed\n", fd);
1514         return NULL;
1515     }
1516     switch(so_type) {
1517     case SOCK_DGRAM:
1518         return net_socket_fd_init_dgram(vlan, model, name, fd, is_connected);
1519     case SOCK_STREAM:
1520         return net_socket_fd_init_stream(vlan, model, name, fd, is_connected);
1521     default:
1522         /* who knows ... this could be a eg. a pty, do warn and continue as stream */
1523         fprintf(stderr, "qemu: warning: socket type=%d for fd=%d is not SOCK_DGRAM or SOCK_STREAM\n", so_type, fd);
1524         return net_socket_fd_init_stream(vlan, model, name, fd, is_connected);
1525     }
1526     return NULL;
1527 }
1528
1529 static void net_socket_accept(void *opaque)
1530 {
1531     NetSocketListenState *s = opaque;
1532     NetSocketState *s1;
1533     struct sockaddr_in saddr;
1534     socklen_t len;
1535     int fd;
1536
1537     for(;;) {
1538         len = sizeof(saddr);
1539         fd = accept(s->fd, (struct sockaddr *)&saddr, &len);
1540         if (fd < 0 && errno != EINTR) {
1541             return;
1542         } else if (fd >= 0) {
1543             break;
1544         }
1545     }
1546     s1 = net_socket_fd_init(s->vlan, s->model, s->name, fd, 1);
1547     if (!s1) {
1548         closesocket(fd);
1549     } else {
1550         snprintf(s1->vc->info_str, sizeof(s1->vc->info_str),
1551                  "socket: connection from %s:%d",
1552                  inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
1553     }
1554 }
1555
1556 static int net_socket_listen_init(VLANState *vlan,
1557                                   const char *model,
1558                                   const char *name,
1559                                   const char *host_str)
1560 {
1561     NetSocketListenState *s;
1562     int fd, val, ret;
1563     struct sockaddr_in saddr;
1564
1565     if (parse_host_port(&saddr, host_str) < 0)
1566         return -1;
1567
1568     s = qemu_mallocz(sizeof(NetSocketListenState));
1569     if (!s)
1570         return -1;
1571
1572     fd = socket(PF_INET, SOCK_STREAM, 0);
1573     if (fd < 0) {
1574         perror("socket");
1575         return -1;
1576     }
1577     socket_set_nonblock(fd);
1578
1579     /* allow fast reuse */
1580     val = 1;
1581     setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, (const char *)&val, sizeof(val));
1582
1583     ret = bind(fd, (struct sockaddr *)&saddr, sizeof(saddr));
1584     if (ret < 0) {
1585         perror("bind");
1586         return -1;
1587     }
1588     ret = listen(fd, 0);
1589     if (ret < 0) {
1590         perror("listen");
1591         return -1;
1592     }
1593     s->vlan = vlan;
1594     s->model = strdup(model);
1595     s->name = strdup(name);
1596     s->fd = fd;
1597     qemu_set_fd_handler(fd, net_socket_accept, NULL, s);
1598     return 0;
1599 }
1600
1601 static int net_socket_connect_init(VLANState *vlan,
1602                                    const char *model,
1603                                    const char *name,
1604                                    const char *host_str)
1605 {
1606     NetSocketState *s;
1607     int fd, connected, ret, err;
1608     struct sockaddr_in saddr;
1609
1610     if (parse_host_port(&saddr, host_str) < 0)
1611         return -1;
1612
1613     fd = socket(PF_INET, SOCK_STREAM, 0);
1614     if (fd < 0) {
1615         perror("socket");
1616         return -1;
1617     }
1618     socket_set_nonblock(fd);
1619
1620     connected = 0;
1621     for(;;) {
1622         ret = connect(fd, (struct sockaddr *)&saddr, sizeof(saddr));
1623         if (ret < 0) {
1624             err = socket_error();
1625             if (err == EINTR || err == EWOULDBLOCK) {
1626             } else if (err == EINPROGRESS) {
1627                 break;
1628 #ifdef _WIN32
1629             } else if (err == WSAEALREADY) {
1630                 break;
1631 #endif
1632             } else {
1633                 perror("connect");
1634                 closesocket(fd);
1635                 return -1;
1636             }
1637         } else {
1638             connected = 1;
1639             break;
1640         }
1641     }
1642     s = net_socket_fd_init(vlan, model, name, fd, connected);
1643     if (!s)
1644         return -1;
1645     snprintf(s->vc->info_str, sizeof(s->vc->info_str),
1646              "socket: connect to %s:%d",
1647              inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
1648     return 0;
1649 }
1650
1651 static int net_socket_mcast_init(VLANState *vlan,
1652                                  const char *model,
1653                                  const char *name,
1654                                  const char *host_str)
1655 {
1656     NetSocketState *s;
1657     int fd;
1658     struct sockaddr_in saddr;
1659
1660     if (parse_host_port(&saddr, host_str) < 0)
1661         return -1;
1662
1663
1664     fd = net_socket_mcast_create(&saddr);
1665     if (fd < 0)
1666         return -1;
1667
1668     s = net_socket_fd_init(vlan, model, name, fd, 0);
1669     if (!s)
1670         return -1;
1671
1672     s->dgram_dst = saddr;
1673
1674     snprintf(s->vc->info_str, sizeof(s->vc->info_str),
1675              "socket: mcast=%s:%d",
1676              inet_ntoa(saddr.sin_addr), ntohs(saddr.sin_port));
1677     return 0;
1678
1679 }
1680
1681 /* find or alloc a new VLAN */
1682 VLANState *qemu_find_vlan(int id)
1683 {
1684     VLANState **pvlan, *vlan;
1685     for(vlan = first_vlan; vlan != NULL; vlan = vlan->next) {
1686         if (vlan->id == id)
1687             return vlan;
1688     }
1689     vlan = qemu_mallocz(sizeof(VLANState));
1690     if (!vlan)
1691         return NULL;
1692     vlan->id = id;
1693     vlan->next = NULL;
1694     pvlan = &first_vlan;
1695     while (*pvlan != NULL)
1696         pvlan = &(*pvlan)->next;
1697     *pvlan = vlan;
1698     return vlan;
1699 }
1700
1701 void qemu_check_nic_model(NICInfo *nd, const char *model)
1702 {
1703     const char *models[2];
1704
1705     models[0] = model;
1706     models[1] = NULL;
1707
1708     qemu_check_nic_model_list(nd, models, model);
1709 }
1710
1711 void qemu_check_nic_model_list(NICInfo *nd, const char * const *models,
1712                                const char *default_model)
1713 {
1714     int i, exit_status = 0;
1715
1716     if (!nd->model)
1717         nd->model = strdup(default_model);
1718
1719     if (strcmp(nd->model, "?") != 0) {
1720         for (i = 0 ; models[i]; i++)
1721             if (strcmp(nd->model, models[i]) == 0)
1722                 return;
1723
1724         fprintf(stderr, "qemu: Unsupported NIC model: %s\n", nd->model);
1725         exit_status = 1;
1726     }
1727
1728     fprintf(stderr, "qemu: Supported NIC models: ");
1729     for (i = 0 ; models[i]; i++)
1730         fprintf(stderr, "%s%c", models[i], models[i+1] ? ',' : '\n');
1731
1732     exit(exit_status);
1733 }
1734
1735 int net_client_init(const char *device, const char *p)
1736 {
1737     char buf[1024];
1738     int vlan_id, ret;
1739     VLANState *vlan;
1740     char *name = NULL;
1741
1742     vlan_id = 0;
1743     if (get_param_value(buf, sizeof(buf), "vlan", p)) {
1744         vlan_id = strtol(buf, NULL, 0);
1745     }
1746     vlan = qemu_find_vlan(vlan_id);
1747     if (!vlan) {
1748         fprintf(stderr, "Could not create vlan %d\n", vlan_id);
1749         return -1;
1750     }
1751     if (get_param_value(buf, sizeof(buf), "name", p)) {
1752         name = strdup(buf);
1753     }
1754     if (!strcmp(device, "nic")) {
1755         NICInfo *nd;
1756         uint8_t *macaddr;
1757
1758         if (nb_nics >= MAX_NICS) {
1759             fprintf(stderr, "Too Many NICs\n");
1760             return -1;
1761         }
1762         nd = &nd_table[nb_nics];
1763         macaddr = nd->macaddr;
1764         macaddr[0] = 0x52;
1765         macaddr[1] = 0x54;
1766         macaddr[2] = 0x00;
1767         macaddr[3] = 0x12;
1768         macaddr[4] = 0x34;
1769         macaddr[5] = 0x56 + nb_nics;
1770
1771         if (get_param_value(buf, sizeof(buf), "macaddr", p)) {
1772             if (parse_macaddr(macaddr, buf) < 0) {
1773                 fprintf(stderr, "invalid syntax for ethernet address\n");
1774                 return -1;
1775             }
1776         }
1777         if (get_param_value(buf, sizeof(buf), "model", p)) {
1778             nd->model = strdup(buf);
1779         }
1780         nd->vlan = vlan;
1781         nd->name = name;
1782         name = NULL;
1783         nb_nics++;
1784         vlan->nb_guest_devs++;
1785         ret = 0;
1786     } else
1787     if (!strcmp(device, "none")) {
1788         /* does nothing. It is needed to signal that no network cards
1789            are wanted */
1790         ret = 0;
1791     } else
1792 #ifdef CONFIG_SLIRP
1793     if (!strcmp(device, "user")) {
1794         if (get_param_value(buf, sizeof(buf), "hostname", p)) {
1795             pstrcpy(slirp_hostname, sizeof(slirp_hostname), buf);
1796         }
1797         if (get_param_value(buf, sizeof(buf), "restrict", p)) {
1798             slirp_restrict = (buf[0] == 'y') ? 1 : 0;
1799         }
1800         if (get_param_value(buf, sizeof(buf), "ip", p)) {
1801             slirp_ip = strdup(buf);
1802         }
1803         vlan->nb_host_devs++;
1804         ret = net_slirp_init(vlan, device, name);
1805     } else
1806 #endif
1807 #ifdef _WIN32
1808     if (!strcmp(device, "tap")) {
1809         char ifname[64];
1810         if (get_param_value(ifname, sizeof(ifname), "ifname", p) <= 0) {
1811             fprintf(stderr, "tap: no interface name\n");
1812             return -1;
1813         }
1814         vlan->nb_host_devs++;
1815         ret = tap_win32_init(vlan, device, name, ifname);
1816     } else
1817 #elif defined (_AIX)
1818 #else
1819     if (!strcmp(device, "tap")) {
1820         char ifname[64];
1821         char setup_script[1024], down_script[1024];
1822         int fd;
1823         vlan->nb_host_devs++;
1824         if (get_param_value(buf, sizeof(buf), "fd", p) > 0) {
1825             fd = strtol(buf, NULL, 0);
1826             fcntl(fd, F_SETFL, O_NONBLOCK);
1827             ret = -1;
1828                 if (net_tap_fd_init(vlan, device, name, fd,
1829                                     tap_probe_vnet_hdr(fd)))
1830                 ret = 0;
1831         } else {
1832             if (get_param_value(ifname, sizeof(ifname), "ifname", p) <= 0) {
1833                 ifname[0] = '\0';
1834             }
1835             if (get_param_value(setup_script, sizeof(setup_script), "script", p) == 0) {
1836                 pstrcpy(setup_script, sizeof(setup_script), DEFAULT_NETWORK_SCRIPT);
1837             }
1838             if (get_param_value(down_script, sizeof(down_script), "downscript", p) == 0) {
1839                 pstrcpy(down_script, sizeof(down_script), DEFAULT_NETWORK_DOWN_SCRIPT);
1840             }
1841             ret = net_tap_init(vlan, device, name, ifname, setup_script, down_script);
1842         }
1843     } else
1844 #endif
1845     if (!strcmp(device, "socket")) {
1846         if (get_param_value(buf, sizeof(buf), "fd", p) > 0) {
1847             int fd;
1848             fd = strtol(buf, NULL, 0);
1849             ret = -1;
1850             if (net_socket_fd_init(vlan, device, name, fd, 1))
1851                 ret = 0;
1852         } else if (get_param_value(buf, sizeof(buf), "listen", p) > 0) {
1853             ret = net_socket_listen_init(vlan, device, name, buf);
1854         } else if (get_param_value(buf, sizeof(buf), "connect", p) > 0) {
1855             ret = net_socket_connect_init(vlan, device, name, buf);
1856         } else if (get_param_value(buf, sizeof(buf), "mcast", p) > 0) {
1857             ret = net_socket_mcast_init(vlan, device, name, buf);
1858         } else {
1859             fprintf(stderr, "Unknown socket options: %s\n", p);
1860             return -1;
1861         }
1862         vlan->nb_host_devs++;
1863     } else
1864 #ifdef CONFIG_VDE
1865     if (!strcmp(device, "vde")) {
1866         char vde_sock[1024], vde_group[512];
1867         int vde_port, vde_mode;
1868         vlan->nb_host_devs++;
1869         if (get_param_value(vde_sock, sizeof(vde_sock), "sock", p) <= 0) {
1870             vde_sock[0] = '\0';
1871         }
1872         if (get_param_value(buf, sizeof(buf), "port", p) > 0) {
1873             vde_port = strtol(buf, NULL, 10);
1874         } else {
1875             vde_port = 0;
1876         }
1877         if (get_param_value(vde_group, sizeof(vde_group), "group", p) <= 0) {
1878             vde_group[0] = '\0';
1879         }
1880         if (get_param_value(buf, sizeof(buf), "mode", p) > 0) {
1881             vde_mode = strtol(buf, NULL, 8);
1882         } else {
1883             vde_mode = 0700;
1884         }
1885         ret = net_vde_init(vlan, device, name, vde_sock, vde_port, vde_group, vde_mode);
1886     } else
1887 #endif
1888     {
1889         fprintf(stderr, "Unknown network device: %s\n", device);
1890         if (name)
1891             free(name);
1892         return -1;
1893     }
1894     if (ret < 0) {
1895         fprintf(stderr, "Could not initialize device '%s'\n", device);
1896     }
1897     if (name)
1898         free(name);
1899     return ret;
1900 }
1901
1902 void net_client_uninit(NICInfo *nd)
1903 {
1904     nd->vlan->nb_guest_devs--; /* XXX: free vlan on last reference */
1905     nb_nics--;
1906     nd->used = 0;
1907     free((void *)nd->model);
1908 }
1909
1910 int net_client_parse(const char *str)
1911 {
1912     const char *p;
1913     char *q;
1914     char device[64];
1915
1916     p = str;
1917     q = device;
1918     while (*p != '\0' && *p != ',') {
1919         if ((q - device) < sizeof(device) - 1)
1920             *q++ = *p;
1921         p++;
1922     }
1923     *q = '\0';
1924     if (*p == ',')
1925         p++;
1926
1927     return net_client_init(device, p);
1928 }
1929
1930 void do_info_network(void)
1931 {
1932     VLANState *vlan;
1933     VLANClientState *vc;
1934
1935     for(vlan = first_vlan; vlan != NULL; vlan = vlan->next) {
1936         term_printf("VLAN %d devices:\n", vlan->id);
1937         for(vc = vlan->first_client; vc != NULL; vc = vc->next)
1938             term_printf("  %s: %s\n", vc->name, vc->info_str);
1939     }
1940 }
1941
1942 int do_set_link(const char *name, const char *up_or_down)
1943 {
1944     VLANState *vlan;
1945     VLANClientState *vc = NULL;
1946
1947     for (vlan = first_vlan; vlan != NULL; vlan = vlan->next)
1948         for (vc = vlan->first_client; vc != NULL; vc = vc->next)
1949             if (strcmp(vc->name, name) == 0)
1950                 goto done;
1951 done:
1952
1953     if (!vc) {
1954         term_printf("could not find network device '%s'", name);
1955         return 0;
1956     }
1957
1958     if (strcmp(up_or_down, "up") == 0)
1959         vc->link_down = 0;
1960     else if (strcmp(up_or_down, "down") == 0)
1961         vc->link_down = 1;
1962     else
1963         term_printf("invalid link status '%s'; only 'up' or 'down' valid\n",
1964                     up_or_down);
1965
1966     if (vc->link_status_changed)
1967         vc->link_status_changed(vc);
1968
1969     return 1;
1970 }
1971
1972 void net_cleanup(void)
1973 {
1974     VLANState *vlan;
1975
1976 #if !defined(_WIN32)
1977     /* close network clients */
1978     for(vlan = first_vlan; vlan != NULL; vlan = vlan->next) {
1979         VLANClientState *vc;
1980
1981         for(vc = vlan->first_client; vc != NULL; vc = vc->next) {
1982             if (vc->fd_read == tap_receive) {
1983                 TAPState *s = vc->opaque;
1984
1985                 if (s->down_script[0])
1986                     launch_script(s->down_script, s->down_script_arg, s->fd);
1987             }
1988 #if defined(CONFIG_VDE)
1989             if (vc->fd_read == vde_from_qemu) {
1990                 VDEState *s = vc->opaque;
1991                 vde_close(s->vde);
1992             }
1993 #endif
1994         }
1995     }
1996 #endif
1997 }
1998
1999 void net_client_check(void)
2000 {
2001     VLANState *vlan;
2002
2003     for(vlan = first_vlan; vlan != NULL; vlan = vlan->next) {
2004         if (vlan->nb_guest_devs == 0 && vlan->nb_host_devs == 0)
2005             continue;
2006         if (vlan->nb_guest_devs == 0)
2007             fprintf(stderr, "Warning: vlan %d with no nics\n", vlan->id);
2008         if (vlan->nb_host_devs == 0)
2009             fprintf(stderr,
2010                     "Warning: vlan %d is not connected to host network\n",
2011                     vlan->id);
2012     }
2013 }