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