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