slirp/src/socket.c

   1 /*
   2  * Copyright (c) 1995 Danny Gasparovski.
   3  *
   4  * Please read the file COPYRIGHT for the
   5  * terms and conditions of the copyright.
   6  */
   7
   8 #include "slirp.h"
   9 #include "ip_icmp.h"
  10 #ifdef __sun__
  11 #include <sys/filio.h>
  12 #endif
  13
  14 static void sofcantrcvmore(struct socket *so);
  15 static void sofcantsendmore(struct socket *so);
  16
  17 struct socket *solookup(struct socket **last, struct socket *head,
  18         struct sockaddr_storage *lhost, struct sockaddr_storage *fhost)
  19 {
  20     struct socket *so = *last;
  21
  22     /* Optimisation */
  23     if (so != head && sockaddr_equal(&(so->lhost.ss), lhost)
  24             && (!fhost || sockaddr_equal(&so->fhost.ss, fhost))) {
  25         return so;
  26     }
  27
  28     for (so = head->so_next; so != head; so = so->so_next) {
  29         if (sockaddr_equal(&(so->lhost.ss), lhost)
  30                 && (!fhost || sockaddr_equal(&so->fhost.ss, fhost))) {
  31             *last = so;
  32             return so;
  33         }
  34     }
  35
  36     return (struct socket *)NULL;
  37 }
  38
  39 /*
  40  * Create a new socket, initialise the fields
  41  * It is the responsibility of the caller to
  42  * insque() it into the correct linked-list
  43  */
  44 struct socket *
  45 socreate(Slirp *slirp)
  46 {
  47     struct socket *so = g_new(struct socket, 1);
  48
  49     memset(so, 0, sizeof(struct socket));
  50     so->so_state = SS_NOFDREF;
  51     so->s = -1;
  52     so->slirp = slirp;
  53     so->pollfds_idx = -1;
  54
  55     return so;
  56 }
  57
  58 /*
  59  * Remove references to so from the given message queue.
  60  */
  61 static void
  62 soqfree(struct socket *so, struct quehead *qh)
  63 {
  64     struct mbuf *ifq;
  65
  66     for (ifq = (struct mbuf *) qh->qh_link;
  67              (struct quehead *) ifq != qh;
  68              ifq = ifq->ifq_next) {
  69         if (ifq->ifq_so == so) {
  70             struct mbuf *ifm;
  71             ifq->ifq_so = NULL;
  72             for (ifm = ifq->ifs_next; ifm != ifq; ifm = ifm->ifs_next) {
  73                 ifm->ifq_so = NULL;
  74             }
  75         }
  76     }
  77 }
  78
  79 /*
  80  * remque and free a socket, clobber cache
  81  */
  82 void
  83 sofree(struct socket *so)
  84 {
  85   Slirp *slirp = so->slirp;
  86
  87   soqfree(so, &slirp->if_fastq);
  88   soqfree(so, &slirp->if_batchq);
  89
  90   if (so == slirp->tcp_last_so) {
  91       slirp->tcp_last_so = &slirp->tcb;
  92   } else if (so == slirp->udp_last_so) {
  93       slirp->udp_last_so = &slirp->udb;
  94   } else if (so == slirp->icmp_last_so) {
  95       slirp->icmp_last_so = &slirp->icmp;
  96   }
  97   m_free(so->so_m);
  98
  99   if(so->so_next && so->so_prev)
 100     remque(so);  /* crashes if so is not in a queue */
 101
 102   if (so->so_tcpcb) {
 103       free(so->so_tcpcb);
 104   }
 105   g_free(so);
 106 }
 107
 108 size_t sopreprbuf(struct socket *so, struct iovec *iov, int *np)
 109 {
 110         int n, lss, total;
 111         struct sbuf *sb = &so->so_snd;
 112         int len = sb->sb_datalen - sb->sb_cc;
 113         int mss = so->so_tcpcb->t_maxseg;
 114
 115         DEBUG_CALL("sopreprbuf");
 116         DEBUG_ARG("so = %p", so);
 117
 118         if (len <= 0)
 119                 return 0;
 120
 121         iov[0].iov_base = sb->sb_wptr;
 122         iov[1].iov_base = NULL;
 123         iov[1].iov_len = 0;
 124         if (sb->sb_wptr < sb->sb_rptr) {
 125                 iov[0].iov_len = sb->sb_rptr - sb->sb_wptr;
 126                 /* Should never succeed, but... */
 127                 if (iov[0].iov_len > len)
 128                    iov[0].iov_len = len;
 129                 if (iov[0].iov_len > mss)
 130                    iov[0].iov_len -= iov[0].iov_len%mss;
 131                 n = 1;
 132         } else {
 133                 iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_wptr;
 134                 /* Should never succeed, but... */
 135                 if (iov[0].iov_len > len) iov[0].iov_len = len;
 136                 len -= iov[0].iov_len;
 137                 if (len) {
 138                         iov[1].iov_base = sb->sb_data;
 139                         iov[1].iov_len = sb->sb_rptr - sb->sb_data;
 140                         if(iov[1].iov_len > len)
 141                            iov[1].iov_len = len;
 142                         total = iov[0].iov_len + iov[1].iov_len;
 143                         if (total > mss) {
 144                                 lss = total%mss;
 145                                 if (iov[1].iov_len > lss) {
 146                                         iov[1].iov_len -= lss;
 147                                         n = 2;
 148                                 } else {
 149                                         lss -= iov[1].iov_len;
 150                                         iov[0].iov_len -= lss;
 151                                         n = 1;
 152                                 }
 153                         } else
 154                                 n = 2;
 155                 } else {
 156                         if (iov[0].iov_len > mss)
 157                            iov[0].iov_len -= iov[0].iov_len%mss;
 158                         n = 1;
 159                 }
 160         }
 161         if (np)
 162                 *np = n;
 163
 164         return iov[0].iov_len + (n - 1) * iov[1].iov_len;
 165 }
 166
 167 /*
 168  * Read from so's socket into sb_snd, updating all relevant sbuf fields
 169  * NOTE: This will only be called if it is select()ed for reading, so
 170  * a read() of 0 (or less) means it's disconnected
 171  */
 172 int
 173 soread(struct socket *so)
 174 {
 175         int n, nn;
 176         struct sbuf *sb = &so->so_snd;
 177         struct iovec iov[2];
 178
 179         DEBUG_CALL("soread");
 180         DEBUG_ARG("so = %p", so);
 181
 182         /*
 183          * No need to check if there's enough room to read.
 184          * soread wouldn't have been called if there weren't
 185          */
 186         sopreprbuf(so, iov, &n);
 187
 188         nn = recv(so->s, iov[0].iov_base, iov[0].iov_len,0);
 189         if (nn <= 0) {
 190                 if (nn < 0 && (errno == EINTR || errno == EAGAIN))
 191                         return 0;
 192                 else {
 193                         int err;
 194                         socklen_t elen = sizeof err;
 195                         struct sockaddr_storage addr;
 196                         struct sockaddr *paddr = (struct sockaddr *) &addr;
 197                         socklen_t alen = sizeof addr;
 198
 199                         err = errno;
 200                         if (nn == 0) {
 201                                 if (getpeername(so->s, paddr, &alen) < 0) {
 202                                         err = errno;
 203                                 } else {
 204                                         getsockopt(so->s, SOL_SOCKET, SO_ERROR,
 205                                                 &err, &elen);
 206                                 }
 207                         }
 208
 209                         DEBUG_MISC(" --- soread() disconnected, nn = %d, errno = %d-%s",
 210                        nn, errno,strerror(errno));
 211                         sofcantrcvmore(so);
 212
 213                         if (err == ECONNRESET || err == ECONNREFUSED
 214                             || err == ENOTCONN || err == EPIPE) {
 215                                 tcp_drop(sototcpcb(so), err);
 216                         } else {
 217                                 tcp_sockclosed(sototcpcb(so));
 218                         }
 219                         return -1;
 220                 }
 221         }
 222
 223         /*
 224          * If there was no error, try and read the second time round
 225          * We read again if n = 2 (ie, there's another part of the buffer)
 226          * and we read as much as we could in the first read
 227          * We don't test for <= 0 this time, because there legitimately
 228          * might not be any more data (since the socket is non-blocking),
 229          * a close will be detected on next iteration.
 230          * A return of -1 won't (shouldn't) happen, since it didn't happen above
 231          */
 232         if (n == 2 && nn == iov[0].iov_len) {
 233             int ret;
 234             ret = recv(so->s, iov[1].iov_base, iov[1].iov_len,0);
 235             if (ret > 0)
 236                 nn += ret;
 237         }
 238
 239         DEBUG_MISC(" ... read nn = %d bytes", nn);
 240
 241         /* Update fields */
 242         sb->sb_cc += nn;
 243         sb->sb_wptr += nn;
 244         if (sb->sb_wptr >= (sb->sb_data + sb->sb_datalen))
 245                 sb->sb_wptr -= sb->sb_datalen;
 246         return nn;
 247 }
 248
 249 int soreadbuf(struct socket *so, const char *buf, int size)
 250 {
 251     int n, nn, copy = size;
 252         struct sbuf *sb = &so->so_snd;
 253         struct iovec iov[2];
 254
 255         DEBUG_CALL("soreadbuf");
 256         DEBUG_ARG("so = %p", so);
 257
 258         /*
 259          * No need to check if there's enough room to read.
 260          * soread wouldn't have been called if there weren't
 261          */
 262         if (sopreprbuf(so, iov, &n) < size)
 263         goto err;
 264
 265     nn = MIN(iov[0].iov_len, copy);
 266     memcpy(iov[0].iov_base, buf, nn);
 267
 268     copy -= nn;
 269     buf += nn;
 270
 271     if (copy == 0)
 272         goto done;
 273
 274     memcpy(iov[1].iov_base, buf, copy);
 275
 276 done:
 277     /* Update fields */
 278         sb->sb_cc += size;
 279         sb->sb_wptr += size;
 280         if (sb->sb_wptr >= (sb->sb_data + sb->sb_datalen))
 281                 sb->sb_wptr -= sb->sb_datalen;
 282     return size;
 283 err:
 284
 285     sofcantrcvmore(so);
 286     tcp_sockclosed(sototcpcb(so));
 287     g_critical("soreadbuf buffer too small");
 288     return -1;
 289 }
 290
 291 /*
 292  * Get urgent data
 293  *
 294  * When the socket is created, we set it SO_OOBINLINE,
 295  * so when OOB data arrives, we soread() it and everything
 296  * in the send buffer is sent as urgent data
 297  */
 298 int
 299 sorecvoob(struct socket *so)
 300 {
 301         struct tcpcb *tp = sototcpcb(so);
 302         int ret;
 303
 304         DEBUG_CALL("sorecvoob");
 305         DEBUG_ARG("so = %p", so);
 306
 307         /*
 308          * We take a guess at how much urgent data has arrived.
 309          * In most situations, when urgent data arrives, the next
 310          * read() should get all the urgent data.  This guess will
 311          * be wrong however if more data arrives just after the
 312          * urgent data, or the read() doesn't return all the
 313          * urgent data.
 314          */
 315         ret = soread(so);
 316         if (ret > 0) {
 317             tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
 318             tp->t_force = 1;
 319             tcp_output(tp);
 320             tp->t_force = 0;
 321         }
 322
 323         return ret;
 324 }
 325
 326 /*
 327  * Send urgent data
 328  * There's a lot duplicated code here, but...
 329  */
 330 int
 331 sosendoob(struct socket *so)
 332 {
 333         struct sbuf *sb = &so->so_rcv;
 334         char buff[2048]; /* XXX Shouldn't be sending more oob data than this */
 335
 336         int n;
 337
 338         DEBUG_CALL("sosendoob");
 339         DEBUG_ARG("so = %p", so);
 340         DEBUG_ARG("sb->sb_cc = %d", sb->sb_cc);
 341
 342         if (so->so_urgc > 2048)
 343            so->so_urgc = 2048; /* XXXX */
 344
 345         if (sb->sb_rptr < sb->sb_wptr) {
 346                 /* We can send it directly */
 347                 n = slirp_send(so, sb->sb_rptr, so->so_urgc, (MSG_OOB)); /* |MSG_DONTWAIT)); */
 348         } else {
 349                 /*
 350                  * Since there's no sendv or sendtov like writev,
 351                  * we must copy all data to a linear buffer then
 352                  * send it all
 353                  */
 354                 uint32_t urgc = so->so_urgc;
 355                 int len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
 356                 if (len > urgc) {
 357                         len = urgc;
 358                 }
 359                 memcpy(buff, sb->sb_rptr, len);
 360                 urgc -= len;
 361                 if (urgc) {
 362                         n = sb->sb_wptr - sb->sb_data;
 363                         if (n > urgc) {
 364                                 n = urgc;
 365                         }
 366                         memcpy((buff + len), sb->sb_data, n);
 367                         len += n;
 368                 }
 369                 n = slirp_send(so, buff, len, (MSG_OOB)); /* |MSG_DONTWAIT)); */
 370 #ifdef DEBUG
 371                 if (n != len) {
 372                         DEBUG_ERROR("Didn't send all data urgently XXXXX");
 373                 }
 374 #endif
 375         }
 376
 377         if (n < 0) {
 378                 return n;
 379         }
 380         so->so_urgc -= n;
 381         DEBUG_MISC(" ---2 sent %d bytes urgent data, %d urgent bytes left", n, so->so_urgc);
 382
 383         sb->sb_cc -= n;
 384         sb->sb_rptr += n;
 385         if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
 386                 sb->sb_rptr -= sb->sb_datalen;
 387
 388         return n;
 389 }
 390
 391 /*
 392  * Write data from so_rcv to so's socket,
 393  * updating all sbuf field as necessary
 394  */
 395 int
 396 sowrite(struct socket *so)
 397 {
 398         int  n,nn;
 399         struct sbuf *sb = &so->so_rcv;
 400         int len = sb->sb_cc;
 401         struct iovec iov[2];
 402
 403         DEBUG_CALL("sowrite");
 404         DEBUG_ARG("so = %p", so);
 405
 406         if (so->so_urgc) {
 407                 uint32_t expected = so->so_urgc;
 408                 if (sosendoob(so) < expected) {
 409                         /* Treat a short write as a fatal error too,
 410                          * rather than continuing on and sending the urgent
 411                          * data as if it were non-urgent and leaving the
 412                          * so_urgc count wrong.
 413                          */
 414                         goto err_disconnected;
 415                 }
 416                 if (sb->sb_cc == 0)
 417                         return 0;
 418         }
 419
 420         /*
 421          * No need to check if there's something to write,
 422          * sowrite wouldn't have been called otherwise
 423          */
 424
 425         iov[0].iov_base = sb->sb_rptr;
 426         iov[1].iov_base = NULL;
 427         iov[1].iov_len = 0;
 428         if (sb->sb_rptr < sb->sb_wptr) {
 429                 iov[0].iov_len = sb->sb_wptr - sb->sb_rptr;
 430                 /* Should never succeed, but... */
 431                 if (iov[0].iov_len > len) iov[0].iov_len = len;
 432                 n = 1;
 433         } else {
 434                 iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
 435                 if (iov[0].iov_len > len) iov[0].iov_len = len;
 436                 len -= iov[0].iov_len;
 437                 if (len) {
 438                         iov[1].iov_base = sb->sb_data;
 439                         iov[1].iov_len = sb->sb_wptr - sb->sb_data;
 440                         if (iov[1].iov_len > len) iov[1].iov_len = len;
 441                         n = 2;
 442                 } else
 443                         n = 1;
 444         }
 445         /* Check if there's urgent data to send, and if so, send it */
 446
 447         nn = slirp_send(so, iov[0].iov_base, iov[0].iov_len,0);
 448         /* This should never happen, but people tell me it does *shrug* */
 449         if (nn < 0 && (errno == EAGAIN || errno == EINTR))
 450                 return 0;
 451
 452         if (nn <= 0) {
 453                 goto err_disconnected;
 454         }
 455
 456         if (n == 2 && nn == iov[0].iov_len) {
 457             int ret;
 458             ret = slirp_send(so, iov[1].iov_base, iov[1].iov_len,0);
 459             if (ret > 0)
 460                 nn += ret;
 461         }
 462         DEBUG_MISC("  ... wrote nn = %d bytes", nn);
 463
 464         /* Update sbuf */
 465         sb->sb_cc -= nn;
 466         sb->sb_rptr += nn;
 467         if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
 468                 sb->sb_rptr -= sb->sb_datalen;
 469
 470         /*
 471          * If in DRAIN mode, and there's no more data, set
 472          * it CANTSENDMORE
 473          */
 474         if ((so->so_state & SS_FWDRAIN) && sb->sb_cc == 0)
 475                 sofcantsendmore(so);
 476
 477         return nn;
 478
 479 err_disconnected:
 480         DEBUG_MISC(" --- sowrite disconnected, so->so_state = %x, errno = %d",
 481                so->so_state, errno);
 482         sofcantsendmore(so);
 483         tcp_sockclosed(sototcpcb(so));
 484         return -1;
 485 }
 486
 487 /*
 488  * recvfrom() a UDP socket
 489  */
 490 void
 491 sorecvfrom(struct socket *so)
 492 {
 493         struct sockaddr_storage addr;
 494         struct sockaddr_storage saddr, daddr;
 495         socklen_t addrlen = sizeof(struct sockaddr_storage);
 496
 497         DEBUG_CALL("sorecvfrom");
 498         DEBUG_ARG("so = %p", so);
 499
 500         if (so->so_type == IPPROTO_ICMP) {   /* This is a "ping" reply */
 501           char buff[256];
 502           int len;
 503
 504           len = recvfrom(so->s, buff, 256, 0,
 505                          (struct sockaddr *)&addr, &addrlen);
 506           /* XXX Check if reply is "correct"? */
 507
 508           if(len == -1 || len == 0) {
 509             uint8_t code=ICMP_UNREACH_PORT;
 510
 511             if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
 512             else if(errno == ENETUNREACH) code=ICMP_UNREACH_NET;
 513
 514             DEBUG_MISC(" udp icmp rx errno = %d-%s",
 515                    errno,strerror(errno));
 516             icmp_send_error(so->so_m, ICMP_UNREACH, code, 0, strerror(errno));
 517           } else {
 518             icmp_reflect(so->so_m);
 519             so->so_m = NULL; /* Don't m_free() it again! */
 520           }
 521           /* No need for this socket anymore, udp_detach it */
 522           udp_detach(so);
 523         } else {                                /* A "normal" UDP packet */
 524           struct mbuf *m;
 525           int len;
 526 #ifdef _WIN32
 527           unsigned long n;
 528 #else
 529           int n;
 530 #endif
 531
 532           if (ioctlsocket(so->s, FIONREAD, &n) != 0) {
 533               DEBUG_MISC(" ioctlsocket errno = %d-%s\n",
 534                          errno,strerror(errno));
 535               return;
 536           }
 537           if (n == 0) {
 538               return;
 539           }
 540
 541           m = m_get(so->slirp);
 542           if (!m) {
 543               return;
 544           }
 545           switch (so->so_ffamily) {
 546           case AF_INET:
 547               m->m_data += IF_MAXLINKHDR + sizeof(struct udpiphdr);
 548               break;
 549           case AF_INET6:
 550               m->m_data += IF_MAXLINKHDR + sizeof(struct ip6)
 551                                          + sizeof(struct udphdr);
 552               break;
 553           default:
 554               g_assert_not_reached();
 555               break;
 556           }
 557
 558           /*
 559            * XXX Shouldn't FIONREAD packets destined for port 53,
 560            * but I don't know the max packet size for DNS lookups
 561            */
 562           len = M_FREEROOM(m);
 563           /* if (so->so_fport != htons(53)) { */
 564
 565           if (n > len) {
 566             n = (m->m_data - m->m_dat) + m->m_len + n + 1;
 567             m_inc(m, n);
 568             len = M_FREEROOM(m);
 569           }
 570           /* } */
 571
 572           m->m_len = recvfrom(so->s, m->m_data, len, 0,
 573                               (struct sockaddr *)&addr, &addrlen);
 574           DEBUG_MISC(" did recvfrom %d, errno = %d-%s",
 575                  m->m_len, errno,strerror(errno));
 576           if(m->m_len<0) {
 577             /* Report error as ICMP */
 578             switch (so->so_lfamily) {
 579             uint8_t code;
 580             case AF_INET:
 581               code = ICMP_UNREACH_PORT;
 582
 583               if (errno == EHOSTUNREACH) {
 584                 code = ICMP_UNREACH_HOST;
 585               } else if (errno == ENETUNREACH) {
 586                 code = ICMP_UNREACH_NET;
 587               }
 588
 589               DEBUG_MISC(" rx error, tx icmp ICMP_UNREACH:%i", code);
 590               icmp_send_error(so->so_m, ICMP_UNREACH, code, 0, strerror(errno));
 591               break;
 592             case AF_INET6:
 593               code = ICMP6_UNREACH_PORT;
 594
 595               if (errno == EHOSTUNREACH) {
 596                 code = ICMP6_UNREACH_ADDRESS;
 597               } else if (errno == ENETUNREACH) {
 598                 code = ICMP6_UNREACH_NO_ROUTE;
 599               }
 600
 601               DEBUG_MISC(" rx error, tx icmp6 ICMP_UNREACH:%i", code);
 602               icmp6_send_error(so->so_m, ICMP6_UNREACH, code);
 603               break;
 604             default:
 605               g_assert_not_reached();
 606               break;
 607             }
 608             m_free(m);
 609           } else {
 610           /*
 611            * Hack: domain name lookup will be used the most for UDP,
 612            * and since they'll only be used once there's no need
 613            * for the 4 minute (or whatever) timeout... So we time them
 614            * out much quicker (10 seconds  for now...)
 615            */
 616             if (so->so_expire) {
 617               if (so->so_fport == htons(53))
 618                 so->so_expire = curtime + SO_EXPIREFAST;
 619               else
 620                 so->so_expire = curtime + SO_EXPIRE;
 621             }
 622
 623             /*
 624              * If this packet was destined for CTL_ADDR,
 625              * make it look like that's where it came from
 626              */
 627             saddr = addr;
 628             sotranslate_in(so, &saddr);
 629             daddr = so->lhost.ss;
 630
 631             switch (so->so_ffamily) {
 632             case AF_INET:
 633                 udp_output(so, m, (struct sockaddr_in *) &saddr,
 634                            (struct sockaddr_in *) &daddr,
 635                            so->so_iptos);
 636                 break;
 637             case AF_INET6:
 638                 udp6_output(so, m, (struct sockaddr_in6 *) &saddr,
 639                             (struct sockaddr_in6 *) &daddr);
 640                 break;
 641             default:
 642                 g_assert_not_reached();
 643                 break;
 644             }
 645           } /* rx error */
 646         } /* if ping packet */
 647 }
 648
 649 /*
 650  * sendto() a socket
 651  */
 652 int
 653 sosendto(struct socket *so, struct mbuf *m)
 654 {
 655         int ret;
 656         struct sockaddr_storage addr;
 657
 658         DEBUG_CALL("sosendto");
 659         DEBUG_ARG("so = %p", so);
 660         DEBUG_ARG("m = %p", m);
 661
 662         addr = so->fhost.ss;
 663         DEBUG_CALL(" sendto()ing)");
 664         sotranslate_out(so, &addr);
 665
 666         /* Don't care what port we get */
 667         ret = sendto(so->s, m->m_data, m->m_len, 0,
 668                      (struct sockaddr *)&addr, sockaddr_size(&addr));
 669         if (ret < 0)
 670                 return -1;
 671
 672         /*
 673          * Kill the socket if there's no reply in 4 minutes,
 674          * but only if it's an expirable socket
 675          */
 676         if (so->so_expire)
 677                 so->so_expire = curtime + SO_EXPIRE;
 678         so->so_state &= SS_PERSISTENT_MASK;
 679         so->so_state |= SS_ISFCONNECTED; /* So that it gets select()ed */
 680         return 0;
 681 }
 682
 683 /*
 684  * Listen for incoming TCP connections
 685  */
 686 struct socket *
 687 tcp_listen(Slirp *slirp, uint32_t haddr, unsigned hport, uint32_t laddr,
 688            unsigned lport, int flags)
 689 {
 690         /* TODO: IPv6 */
 691         struct sockaddr_in addr;
 692         struct socket *so;
 693         int s, opt = 1;
 694         socklen_t addrlen = sizeof(addr);
 695         memset(&addr, 0, addrlen);
 696
 697         DEBUG_CALL("tcp_listen");
 698         DEBUG_ARG("haddr = %s", inet_ntoa((struct in_addr){.s_addr = haddr}));
 699         DEBUG_ARG("hport = %d", ntohs(hport));
 700         DEBUG_ARG("laddr = %s", inet_ntoa((struct in_addr){.s_addr = laddr}));
 701         DEBUG_ARG("lport = %d", ntohs(lport));
 702         DEBUG_ARG("flags = %x", flags);
 703
 704         so = socreate(slirp);
 705
 706         /* Don't tcp_attach... we don't need so_snd nor so_rcv */
 707         if ((so->so_tcpcb = tcp_newtcpcb(so)) == NULL) {
 708             g_free(so);
 709             return NULL;
 710         }
 711         insque(so, &slirp->tcb);
 712
 713         /*
 714          * SS_FACCEPTONCE sockets must time out.
 715          */
 716         if (flags & SS_FACCEPTONCE)
 717            so->so_tcpcb->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT*2;
 718
 719         so->so_state &= SS_PERSISTENT_MASK;
 720         so->so_state |= (SS_FACCEPTCONN | flags);
 721         so->so_lfamily = AF_INET;
 722         so->so_lport = lport; /* Kept in network format */
 723         so->so_laddr.s_addr = laddr; /* Ditto */
 724
 725         addr.sin_family = AF_INET;
 726         addr.sin_addr.s_addr = haddr;
 727         addr.sin_port = hport;
 728
 729         if (((s = slirp_socket(AF_INET,SOCK_STREAM,0)) < 0) ||
 730             (slirp_socket_set_fast_reuse(s) < 0) ||
 731             (bind(s,(struct sockaddr *)&addr, sizeof(addr)) < 0) ||
 732             (listen(s,1) < 0)) {
 733                 int tmperrno = errno; /* Don't clobber the real reason we failed */
 734
 735                 if (s >= 0) {
 736                     closesocket(s);
 737                 }
 738                 sofree(so);
 739                 /* Restore the real errno */
 740 #ifdef _WIN32
 741                 WSASetLastError(tmperrno);
 742 #else
 743                 errno = tmperrno;
 744 #endif
 745                 return NULL;
 746         }
 747         setsockopt(s, SOL_SOCKET, SO_OOBINLINE, &opt, sizeof(int));
 748         opt = 1;
 749         setsockopt(s, IPPROTO_TCP, TCP_NODELAY, &opt, sizeof(int));
 750
 751         getsockname(s,(struct sockaddr *)&addr,&addrlen);
 752         so->so_ffamily = AF_INET;
 753         so->so_fport = addr.sin_port;
 754         if (addr.sin_addr.s_addr == 0 || addr.sin_addr.s_addr == loopback_addr.s_addr)
 755            so->so_faddr = slirp->vhost_addr;
 756         else
 757            so->so_faddr = addr.sin_addr;
 758
 759         so->s = s;
 760         return so;
 761 }
 762
 763 /*
 764  * Various session state calls
 765  * XXX Should be #define's
 766  * The socket state stuff needs work, these often get call 2 or 3
 767  * times each when only 1 was needed
 768  */
 769 void
 770 soisfconnecting(struct socket *so)
 771 {
 772         so->so_state &= ~(SS_NOFDREF|SS_ISFCONNECTED|SS_FCANTRCVMORE|
 773                           SS_FCANTSENDMORE|SS_FWDRAIN);
 774         so->so_state |= SS_ISFCONNECTING; /* Clobber other states */
 775 }
 776
 777 void
 778 soisfconnected(struct socket *so)
 779 {
 780         so->so_state &= ~(SS_ISFCONNECTING|SS_FWDRAIN|SS_NOFDREF);
 781         so->so_state |= SS_ISFCONNECTED; /* Clobber other states */
 782 }
 783
 784 static void
 785 sofcantrcvmore(struct socket *so)
 786 {
 787         if ((so->so_state & SS_NOFDREF) == 0) {
 788                 shutdown(so->s,0);
 789         }
 790         so->so_state &= ~(SS_ISFCONNECTING);
 791         if (so->so_state & SS_FCANTSENDMORE) {
 792            so->so_state &= SS_PERSISTENT_MASK;
 793            so->so_state |= SS_NOFDREF; /* Don't select it */
 794         } else {
 795            so->so_state |= SS_FCANTRCVMORE;
 796         }
 797 }
 798
 799 static void
 800 sofcantsendmore(struct socket *so)
 801 {
 802         if ((so->so_state & SS_NOFDREF) == 0) {
 803             shutdown(so->s,1);           /* send FIN to fhost */
 804         }
 805         so->so_state &= ~(SS_ISFCONNECTING);
 806         if (so->so_state & SS_FCANTRCVMORE) {
 807            so->so_state &= SS_PERSISTENT_MASK;
 808            so->so_state |= SS_NOFDREF; /* as above */
 809         } else {
 810            so->so_state |= SS_FCANTSENDMORE;
 811         }
 812 }
 813
 814 /*
 815  * Set write drain mode
 816  * Set CANTSENDMORE once all data has been write()n
 817  */
 818 void
 819 sofwdrain(struct socket *so)
 820 {
 821         if (so->so_rcv.sb_cc)
 822                 so->so_state |= SS_FWDRAIN;
 823         else
 824                 sofcantsendmore(so);
 825 }
 826
 827 /*
 828  * Translate addr in host addr when it is a virtual address
 829  */
 830 void sotranslate_out(struct socket *so, struct sockaddr_storage *addr)
 831 {
 832     Slirp *slirp = so->slirp;
 833     struct sockaddr_in *sin = (struct sockaddr_in *)addr;
 834     struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)addr;
 835
 836     switch (addr->ss_family) {
 837     case AF_INET:
 838         if ((so->so_faddr.s_addr & slirp->vnetwork_mask.s_addr) ==
 839                 slirp->vnetwork_addr.s_addr) {
 840             /* It's an alias */
 841             if (so->so_faddr.s_addr == slirp->vnameserver_addr.s_addr) {
 842                 if (get_dns_addr(&sin->sin_addr) < 0) {
 843                     sin->sin_addr = loopback_addr;
 844                 }
 845             } else {
 846                 sin->sin_addr = loopback_addr;
 847             }
 848         }
 849
 850         DEBUG_MISC(" addr.sin_port=%d, addr.sin_addr.s_addr=%.16s",
 851                    ntohs(sin->sin_port), inet_ntoa(sin->sin_addr));
 852         break;
 853
 854     case AF_INET6:
 855         if (in6_equal_net(&so->so_faddr6, &slirp->vprefix_addr6,
 856                     slirp->vprefix_len)) {
 857             if (in6_equal(&so->so_faddr6, &slirp->vnameserver_addr6)) {
 858                 uint32_t scope_id;
 859                 if (get_dns6_addr(&sin6->sin6_addr, &scope_id) >= 0) {
 860                     sin6->sin6_scope_id = scope_id;
 861                 } else {
 862                     sin6->sin6_addr = in6addr_loopback;
 863                 }
 864             } else {
 865                 sin6->sin6_addr = in6addr_loopback;
 866             }
 867         }
 868         break;
 869
 870     default:
 871         break;
 872     }
 873 }
 874
 875 void sotranslate_in(struct socket *so, struct sockaddr_storage *addr)
 876 {
 877     Slirp *slirp = so->slirp;
 878     struct sockaddr_in *sin = (struct sockaddr_in *)addr;
 879     struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)addr;
 880
 881     switch (addr->ss_family) {
 882     case AF_INET:
 883         if ((so->so_faddr.s_addr & slirp->vnetwork_mask.s_addr) ==
 884             slirp->vnetwork_addr.s_addr) {
 885             uint32_t inv_mask = ~slirp->vnetwork_mask.s_addr;
 886
 887             if ((so->so_faddr.s_addr & inv_mask) == inv_mask) {
 888                 sin->sin_addr = slirp->vhost_addr;
 889             } else if (sin->sin_addr.s_addr == loopback_addr.s_addr ||
 890                        so->so_faddr.s_addr != slirp->vhost_addr.s_addr) {
 891                 sin->sin_addr = so->so_faddr;
 892             }
 893         }
 894         break;
 895
 896     case AF_INET6:
 897         if (in6_equal_net(&so->so_faddr6, &slirp->vprefix_addr6,
 898                     slirp->vprefix_len)) {
 899             if (in6_equal(&sin6->sin6_addr, &in6addr_loopback)
 900                     || !in6_equal(&so->so_faddr6, &slirp->vhost_addr6)) {
 901                 sin6->sin6_addr = so->so_faddr6;
 902             }
 903         }
 904         break;
 905
 906     default:
 907         break;
 908     }
 909 }
 910
 911 /*
 912  * Translate connections from localhost to the real hostname
 913  */
 914 void sotranslate_accept(struct socket *so)
 915 {
 916     Slirp *slirp = so->slirp;
 917
 918     switch (so->so_ffamily) {
 919     case AF_INET:
 920         if (so->so_faddr.s_addr == INADDR_ANY ||
 921             (so->so_faddr.s_addr & loopback_mask) ==
 922             (loopback_addr.s_addr & loopback_mask)) {
 923            so->so_faddr = slirp->vhost_addr;
 924         }
 925         break;
 926
 927    case AF_INET6:
 928         if (in6_equal(&so->so_faddr6, &in6addr_any) ||
 929                 in6_equal(&so->so_faddr6, &in6addr_loopback)) {
 930            so->so_faddr6 = slirp->vhost_addr6;
 931         }
 932         break;
 933
 934     default:
 935         break;
 936     }
 937 }
 938
 939 void sodrop(struct socket *s, int num)
 940 {
 941     if (sbdrop(&s->so_snd, num)) {
 942         s->slirp->cb->notify(s->slirp->opaque);
 943     }
 944 }