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