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