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                         int err;
 180                         socklen_t slen = sizeof err;
 181
 182                         err = errno;
 183                         if (nn == 0) {
 184                                 getsockopt(so->s, SOL_SOCKET, SO_ERROR,
 185                                            &err, &slen);
 186                         }
 187
 188                         DEBUG_MISC((dfd, " --- soread() disconnected, nn = %d, errno = %d-%s\n", nn, errno,strerror(errno)));
 189                         sofcantrcvmore(so);
 190
 191                         if (err == ECONNRESET || err == ECONNREFUSED
 192                             || err == ENOTCONN || err == EPIPE) {
 193                                 tcp_drop(sototcpcb(so), err);
 194                         } else {
 195                                 tcp_sockclosed(sototcpcb(so));
 196                         }
 197                         return -1;
 198                 }
 199         }
 200
 201 #ifndef HAVE_READV
 202         /*
 203          * If there was no error, try and read the second time round
 204          * We read again if n = 2 (ie, there's another part of the buffer)
 205          * and we read as much as we could in the first read
 206          * We don't test for <= 0 this time, because there legitimately
 207          * might not be any more data (since the socket is non-blocking),
 208          * a close will be detected on next iteration.
 209          * A return of -1 won't (shouldn't) happen, since it didn't happen above
 210          */
 211         if (n == 2 && nn == iov[0].iov_len) {
 212             int ret;
 213             ret = qemu_recv(so->s, iov[1].iov_base, iov[1].iov_len,0);
 214             if (ret > 0)
 215                 nn += ret;
 216         }
 217
 218         DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
 219 #endif
 220
 221         /* Update fields */
 222         sb->sb_cc += nn;
 223         sb->sb_wptr += nn;
 224         if (sb->sb_wptr >= (sb->sb_data + sb->sb_datalen))
 225                 sb->sb_wptr -= sb->sb_datalen;
 226         return nn;
 227 }
 228
 229 int soreadbuf(struct socket *so, const char *buf, int size)
 230 {
 231     int n, nn, copy = size;
 232         struct sbuf *sb = &so->so_snd;
 233         struct iovec iov[2];
 234
 235         DEBUG_CALL("soreadbuf");
 236         DEBUG_ARG("so = %p", so);
 237
 238         /*
 239          * No need to check if there's enough room to read.
 240          * soread wouldn't have been called if there weren't
 241          */
 242         if (sopreprbuf(so, iov, &n) < size)
 243         goto err;
 244
 245     nn = MIN(iov[0].iov_len, copy);
 246     memcpy(iov[0].iov_base, buf, nn);
 247
 248     copy -= nn;
 249     buf += nn;
 250
 251     if (copy == 0)
 252         goto done;
 253
 254     memcpy(iov[1].iov_base, buf, copy);
 255
 256 done:
 257     /* Update fields */
 258         sb->sb_cc += size;
 259         sb->sb_wptr += size;
 260         if (sb->sb_wptr >= (sb->sb_data + sb->sb_datalen))
 261                 sb->sb_wptr -= sb->sb_datalen;
 262     return size;
 263 err:
 264
 265     sofcantrcvmore(so);
 266     tcp_sockclosed(sototcpcb(so));
 267     fprintf(stderr, "soreadbuf buffer to small");
 268     return -1;
 269 }
 270
 271 /*
 272  * Get urgent data
 273  *
 274  * When the socket is created, we set it SO_OOBINLINE,
 275  * so when OOB data arrives, we soread() it and everything
 276  * in the send buffer is sent as urgent data
 277  */
 278 int
 279 sorecvoob(struct socket *so)
 280 {
 281         struct tcpcb *tp = sototcpcb(so);
 282         int ret;
 283
 284         DEBUG_CALL("sorecvoob");
 285         DEBUG_ARG("so = %p", so);
 286
 287         /*
 288          * We take a guess at how much urgent data has arrived.
 289          * In most situations, when urgent data arrives, the next
 290          * read() should get all the urgent data.  This guess will
 291          * be wrong however if more data arrives just after the
 292          * urgent data, or the read() doesn't return all the
 293          * urgent data.
 294          */
 295         ret = soread(so);
 296         if (ret > 0) {
 297             tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
 298             tp->t_force = 1;
 299             tcp_output(tp);
 300             tp->t_force = 0;
 301         }
 302
 303         return ret;
 304 }
 305
 306 /*
 307  * Send urgent data
 308  * There's a lot duplicated code here, but...
 309  */
 310 int
 311 sosendoob(struct socket *so)
 312 {
 313         struct sbuf *sb = &so->so_rcv;
 314         char buff[2048]; /* XXX Shouldn't be sending more oob data than this */
 315
 316         int n, len;
 317
 318         DEBUG_CALL("sosendoob");
 319         DEBUG_ARG("so = %p", so);
 320         DEBUG_ARG("sb->sb_cc = %d", sb->sb_cc);
 321
 322         if (so->so_urgc > 2048)
 323            so->so_urgc = 2048; /* XXXX */
 324
 325         if (sb->sb_rptr < sb->sb_wptr) {
 326                 /* We can send it directly */
 327                 n = slirp_send(so, sb->sb_rptr, so->so_urgc, (MSG_OOB)); /* |MSG_DONTWAIT)); */
 328                 so->so_urgc -= n;
 329
 330                 DEBUG_MISC((dfd, " --- sent %d bytes urgent data, %d urgent bytes left\n", n, so->so_urgc));
 331         } else {
 332                 /*
 333                  * Since there's no sendv or sendtov like writev,
 334                  * we must copy all data to a linear buffer then
 335                  * send it all
 336                  */
 337                 len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
 338                 if (len > so->so_urgc) len = so->so_urgc;
 339                 memcpy(buff, sb->sb_rptr, len);
 340                 so->so_urgc -= len;
 341                 if (so->so_urgc) {
 342                         n = sb->sb_wptr - sb->sb_data;
 343                         if (n > so->so_urgc) n = so->so_urgc;
 344                         memcpy((buff + len), sb->sb_data, n);
 345                         so->so_urgc -= n;
 346                         len += n;
 347                 }
 348                 n = slirp_send(so, buff, len, (MSG_OOB)); /* |MSG_DONTWAIT)); */
 349 #ifdef DEBUG
 350                 if (n != len)
 351                    DEBUG_ERROR((dfd, "Didn't send all data urgently XXXXX\n"));
 352 #endif
 353                 DEBUG_MISC((dfd, " ---2 sent %d bytes urgent data, %d urgent bytes left\n", n, so->so_urgc));
 354         }
 355
 356         sb->sb_cc -= n;
 357         sb->sb_rptr += n;
 358         if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
 359                 sb->sb_rptr -= sb->sb_datalen;
 360
 361         return n;
 362 }
 363
 364 /*
 365  * Write data from so_rcv to so's socket,
 366  * updating all sbuf field as necessary
 367  */
 368 int
 369 sowrite(struct socket *so)
 370 {
 371         int  n,nn;
 372         struct sbuf *sb = &so->so_rcv;
 373         int len = sb->sb_cc;
 374         struct iovec iov[2];
 375
 376         DEBUG_CALL("sowrite");
 377         DEBUG_ARG("so = %p", so);
 378
 379         if (so->so_urgc) {
 380                 sosendoob(so);
 381                 if (sb->sb_cc == 0)
 382                         return 0;
 383         }
 384
 385         /*
 386          * No need to check if there's something to write,
 387          * sowrite wouldn't have been called otherwise
 388          */
 389
 390         iov[0].iov_base = sb->sb_rptr;
 391         iov[1].iov_base = NULL;
 392         iov[1].iov_len = 0;
 393         if (sb->sb_rptr < sb->sb_wptr) {
 394                 iov[0].iov_len = sb->sb_wptr - sb->sb_rptr;
 395                 /* Should never succeed, but... */
 396                 if (iov[0].iov_len > len) iov[0].iov_len = len;
 397                 n = 1;
 398         } else {
 399                 iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
 400                 if (iov[0].iov_len > len) iov[0].iov_len = len;
 401                 len -= iov[0].iov_len;
 402                 if (len) {
 403                         iov[1].iov_base = sb->sb_data;
 404                         iov[1].iov_len = sb->sb_wptr - sb->sb_data;
 405                         if (iov[1].iov_len > len) iov[1].iov_len = len;
 406                         n = 2;
 407                 } else
 408                         n = 1;
 409         }
 410         /* Check if there's urgent data to send, and if so, send it */
 411
 412 #ifdef HAVE_READV
 413         nn = writev(so->s, (const struct iovec *)iov, n);
 414
 415         DEBUG_MISC((dfd, "  ... wrote nn = %d bytes\n", nn));
 416 #else
 417         nn = slirp_send(so, iov[0].iov_base, iov[0].iov_len,0);
 418 #endif
 419         /* This should never happen, but people tell me it does *shrug* */
 420         if (nn < 0 && (errno == EAGAIN || errno == EINTR))
 421                 return 0;
 422
 423         if (nn <= 0) {
 424                 DEBUG_MISC((dfd, " --- sowrite disconnected, so->so_state = %x, errno = %d\n",
 425                         so->so_state, errno));
 426                 sofcantsendmore(so);
 427                 tcp_sockclosed(sototcpcb(so));
 428                 return -1;
 429         }
 430
 431 #ifndef HAVE_READV
 432         if (n == 2 && nn == iov[0].iov_len) {
 433             int ret;
 434             ret = slirp_send(so, iov[1].iov_base, iov[1].iov_len,0);
 435             if (ret > 0)
 436                 nn += ret;
 437         }
 438         DEBUG_MISC((dfd, "  ... wrote nn = %d bytes\n", nn));
 439 #endif
 440
 441         /* Update sbuf */
 442         sb->sb_cc -= nn;
 443         sb->sb_rptr += nn;
 444         if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
 445                 sb->sb_rptr -= sb->sb_datalen;
 446
 447         /*
 448          * If in DRAIN mode, and there's no more data, set
 449          * it CANTSENDMORE
 450          */
 451         if ((so->so_state & SS_FWDRAIN) && sb->sb_cc == 0)
 452                 sofcantsendmore(so);
 453
 454         return nn;
 455 }
 456
 457 /*
 458  * recvfrom() a UDP socket
 459  */
 460 void
 461 sorecvfrom(struct socket *so)
 462 {
 463         struct sockaddr_storage addr;
 464         struct sockaddr_storage saddr, daddr;
 465         socklen_t addrlen = sizeof(struct sockaddr_storage);
 466
 467         DEBUG_CALL("sorecvfrom");
 468         DEBUG_ARG("so = %p", so);
 469
 470         if (so->so_type == IPPROTO_ICMP) {   /* This is a "ping" reply */
 471           char buff[256];
 472           int len;
 473
 474           len = recvfrom(so->s, buff, 256, 0,
 475                          (struct sockaddr *)&addr, &addrlen);
 476           /* XXX Check if reply is "correct"? */
 477
 478           if(len == -1 || len == 0) {
 479             u_char code=ICMP_UNREACH_PORT;
 480
 481             if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
 482             else if(errno == ENETUNREACH) code=ICMP_UNREACH_NET;
 483
 484             DEBUG_MISC((dfd," udp icmp rx errno = %d-%s\n",
 485                         errno,strerror(errno)));
 486             icmp_send_error(so->so_m, ICMP_UNREACH, code, 0, strerror(errno));
 487           } else {
 488             icmp_reflect(so->so_m);
 489             so->so_m = NULL; /* Don't m_free() it again! */
 490           }
 491           /* No need for this socket anymore, udp_detach it */
 492           udp_detach(so);
 493         } else {                                /* A "normal" UDP packet */
 494           struct mbuf *m;
 495           int len;
 496 #ifdef _WIN32
 497           unsigned long n;
 498 #else
 499           int n;
 500 #endif
 501
 502           m = m_get(so->slirp);
 503           if (!m) {
 504               return;
 505           }
 506           switch (so->so_ffamily) {
 507           case AF_INET:
 508               m->m_data += IF_MAXLINKHDR + sizeof(struct udpiphdr);
 509               break;
 510           case AF_INET6:
 511               m->m_data += IF_MAXLINKHDR + sizeof(struct ip6)
 512                                          + sizeof(struct udphdr);
 513               break;
 514           default:
 515               g_assert_not_reached();
 516               break;
 517           }
 518
 519           /*
 520            * XXX Shouldn't FIONREAD packets destined for port 53,
 521            * but I don't know the max packet size for DNS lookups
 522            */
 523           len = M_FREEROOM(m);
 524           /* if (so->so_fport != htons(53)) { */
 525           ioctlsocket(so->s, FIONREAD, &n);
 526
 527           if (n > len) {
 528             n = (m->m_data - m->m_dat) + m->m_len + n + 1;
 529             m_inc(m, n);
 530             len = M_FREEROOM(m);
 531           }
 532           /* } */
 533
 534           m->m_len = recvfrom(so->s, m->m_data, len, 0,
 535                               (struct sockaddr *)&addr, &addrlen);
 536           DEBUG_MISC((dfd, " did recvfrom %d, errno = %d-%s\n",
 537                       m->m_len, errno,strerror(errno)));
 538           if(m->m_len<0) {
 539             /* Report error as ICMP */
 540             switch (so->so_lfamily) {
 541             uint8_t code;
 542             case AF_INET:
 543               code = ICMP_UNREACH_PORT;
 544
 545               if (errno == EHOSTUNREACH) {
 546                 code = ICMP_UNREACH_HOST;
 547               } else if (errno == ENETUNREACH) {
 548                 code = ICMP_UNREACH_NET;
 549               }
 550
 551               DEBUG_MISC((dfd, " rx error, tx icmp ICMP_UNREACH:%i\n", code));
 552               icmp_send_error(so->so_m, ICMP_UNREACH, code, 0, strerror(errno));
 553               break;
 554             case AF_INET6:
 555               code = ICMP6_UNREACH_PORT;
 556
 557               if (errno == EHOSTUNREACH) {
 558                 code = ICMP6_UNREACH_ADDRESS;
 559               } else if (errno == ENETUNREACH) {
 560                 code = ICMP6_UNREACH_NO_ROUTE;
 561               }
 562
 563               DEBUG_MISC((dfd, " rx error, tx icmp6 ICMP_UNREACH:%i\n", code));
 564               icmp6_send_error(so->so_m, ICMP6_UNREACH, code);
 565               break;
 566             default:
 567               g_assert_not_reached();
 568               break;
 569             }
 570             m_free(m);
 571           } else {
 572           /*
 573            * Hack: domain name lookup will be used the most for UDP,
 574            * and since they'll only be used once there's no need
 575            * for the 4 minute (or whatever) timeout... So we time them
 576            * out much quicker (10 seconds  for now...)
 577            */
 578             if (so->so_expire) {
 579               if (so->so_fport == htons(53))
 580                 so->so_expire = curtime + SO_EXPIREFAST;
 581               else
 582                 so->so_expire = curtime + SO_EXPIRE;
 583             }
 584
 585             /*
 586              * If this packet was destined for CTL_ADDR,
 587              * make it look like that's where it came from
 588              */
 589             saddr = addr;
 590             sotranslate_in(so, &saddr);
 591             daddr = so->lhost.ss;
 592
 593             switch (so->so_ffamily) {
 594             case AF_INET:
 595                 udp_output(so, m, (struct sockaddr_in *) &saddr,
 596                            (struct sockaddr_in *) &daddr,
 597                            so->so_iptos);
 598                 break;
 599             case AF_INET6:
 600                 udp6_output(so, m, (struct sockaddr_in6 *) &saddr,
 601                             (struct sockaddr_in6 *) &daddr);
 602                 break;
 603             default:
 604                 g_assert_not_reached();
 605                 break;
 606             }
 607           } /* rx error */
 608         } /* if ping packet */
 609 }
 610
 611 /*
 612  * sendto() a socket
 613  */
 614 int
 615 sosendto(struct socket *so, struct mbuf *m)
 616 {
 617         int ret;
 618         struct sockaddr_storage addr;
 619
 620         DEBUG_CALL("sosendto");
 621         DEBUG_ARG("so = %p", so);
 622         DEBUG_ARG("m = %p", m);
 623
 624         addr = so->fhost.ss;
 625         DEBUG_CALL(" sendto()ing)");
 626         sotranslate_out(so, &addr);
 627
 628         /* Don't care what port we get */
 629         ret = sendto(so->s, m->m_data, m->m_len, 0,
 630                      (struct sockaddr *)&addr, sockaddr_size(&addr));
 631         if (ret < 0)
 632                 return -1;
 633
 634         /*
 635          * Kill the socket if there's no reply in 4 minutes,
 636          * but only if it's an expirable socket
 637          */
 638         if (so->so_expire)
 639                 so->so_expire = curtime + SO_EXPIRE;
 640         so->so_state &= SS_PERSISTENT_MASK;
 641         so->so_state |= SS_ISFCONNECTED; /* So that it gets select()ed */
 642         return 0;
 643 }
 644
 645 /*
 646  * Listen for incoming TCP connections
 647  */
 648 struct socket *
 649 tcp_listen(Slirp *slirp, uint32_t haddr, u_int hport, uint32_t laddr,
 650            u_int lport, int flags)
 651 {
 652         struct sockaddr_in addr;
 653         struct socket *so;
 654         int s, opt = 1;
 655         socklen_t addrlen = sizeof(addr);
 656         memset(&addr, 0, addrlen);
 657
 658         DEBUG_CALL("tcp_listen");
 659         DEBUG_ARG("haddr = %x", haddr);
 660         DEBUG_ARG("hport = %d", hport);
 661         DEBUG_ARG("laddr = %x", laddr);
 662         DEBUG_ARG("lport = %d", lport);
 663         DEBUG_ARG("flags = %x", flags);
 664
 665         so = socreate(slirp);
 666         if (!so) {
 667           return NULL;
 668         }
 669
 670         /* Don't tcp_attach... we don't need so_snd nor so_rcv */
 671         if ((so->so_tcpcb = tcp_newtcpcb(so)) == NULL) {
 672                 free(so);
 673                 return NULL;
 674         }
 675         insque(so, &slirp->tcb);
 676
 677         /*
 678          * SS_FACCEPTONCE sockets must time out.
 679          */
 680         if (flags & SS_FACCEPTONCE)
 681            so->so_tcpcb->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT*2;
 682
 683         so->so_state &= SS_PERSISTENT_MASK;
 684         so->so_state |= (SS_FACCEPTCONN | flags);
 685         so->so_lfamily = AF_INET;
 686         so->so_lport = lport; /* Kept in network format */
 687         so->so_laddr.s_addr = laddr; /* Ditto */
 688
 689         addr.sin_family = AF_INET;
 690         addr.sin_addr.s_addr = haddr;
 691         addr.sin_port = hport;
 692
 693         if (((s = qemu_socket(AF_INET,SOCK_STREAM,0)) < 0) ||
 694             (socket_set_fast_reuse(s) < 0) ||
 695             (bind(s,(struct sockaddr *)&addr, sizeof(addr)) < 0) ||
 696             (listen(s,1) < 0)) {
 697                 int tmperrno = errno; /* Don't clobber the real reason we failed */
 698
 699                 close(s);
 700                 sofree(so);
 701                 /* Restore the real errno */
 702 #ifdef _WIN32
 703                 WSASetLastError(tmperrno);
 704 #else
 705                 errno = tmperrno;
 706 #endif
 707                 return NULL;
 708         }
 709         qemu_setsockopt(s, SOL_SOCKET, SO_OOBINLINE, &opt, sizeof(int));
 710
 711         getsockname(s,(struct sockaddr *)&addr,&addrlen);
 712         so->so_ffamily = AF_INET;
 713         so->so_fport = addr.sin_port;
 714         if (addr.sin_addr.s_addr == 0 || addr.sin_addr.s_addr == loopback_addr.s_addr)
 715            so->so_faddr = slirp->vhost_addr;
 716         else
 717            so->so_faddr = addr.sin_addr;
 718
 719         so->s = s;
 720         return so;
 721 }
 722
 723 /*
 724  * Various session state calls
 725  * XXX Should be #define's
 726  * The socket state stuff needs work, these often get call 2 or 3
 727  * times each when only 1 was needed
 728  */
 729 void
 730 soisfconnecting(struct socket *so)
 731 {
 732         so->so_state &= ~(SS_NOFDREF|SS_ISFCONNECTED|SS_FCANTRCVMORE|
 733                           SS_FCANTSENDMORE|SS_FWDRAIN);
 734         so->so_state |= SS_ISFCONNECTING; /* Clobber other states */
 735 }
 736
 737 void
 738 soisfconnected(struct socket *so)
 739 {
 740         so->so_state &= ~(SS_ISFCONNECTING|SS_FWDRAIN|SS_NOFDREF);
 741         so->so_state |= SS_ISFCONNECTED; /* Clobber other states */
 742 }
 743
 744 static void
 745 sofcantrcvmore(struct socket *so)
 746 {
 747         if ((so->so_state & SS_NOFDREF) == 0) {
 748                 shutdown(so->s,0);
 749         }
 750         so->so_state &= ~(SS_ISFCONNECTING);
 751         if (so->so_state & SS_FCANTSENDMORE) {
 752            so->so_state &= SS_PERSISTENT_MASK;
 753            so->so_state |= SS_NOFDREF; /* Don't select it */
 754         } else {
 755            so->so_state |= SS_FCANTRCVMORE;
 756         }
 757 }
 758
 759 static void
 760 sofcantsendmore(struct socket *so)
 761 {
 762         if ((so->so_state & SS_NOFDREF) == 0) {
 763             shutdown(so->s,1);           /* send FIN to fhost */
 764         }
 765         so->so_state &= ~(SS_ISFCONNECTING);
 766         if (so->so_state & SS_FCANTRCVMORE) {
 767            so->so_state &= SS_PERSISTENT_MASK;
 768            so->so_state |= SS_NOFDREF; /* as above */
 769         } else {
 770            so->so_state |= SS_FCANTSENDMORE;
 771         }
 772 }
 773
 774 /*
 775  * Set write drain mode
 776  * Set CANTSENDMORE once all data has been write()n
 777  */
 778 void
 779 sofwdrain(struct socket *so)
 780 {
 781         if (so->so_rcv.sb_cc)
 782                 so->so_state |= SS_FWDRAIN;
 783         else
 784                 sofcantsendmore(so);
 785 }
 786
 787 /*
 788  * Translate addr in host addr when it is a virtual address
 789  */
 790 void sotranslate_out(struct socket *so, struct sockaddr_storage *addr)
 791 {
 792     Slirp *slirp = so->slirp;
 793     struct sockaddr_in *sin = (struct sockaddr_in *)addr;
 794     struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)addr;
 795
 796     switch (addr->ss_family) {
 797     case AF_INET:
 798         if ((so->so_faddr.s_addr & slirp->vnetwork_mask.s_addr) ==
 799                 slirp->vnetwork_addr.s_addr) {
 800             /* It's an alias */
 801             if (so->so_faddr.s_addr == slirp->vnameserver_addr.s_addr) {
 802                 if (get_dns_addr(&sin->sin_addr) < 0) {
 803                     sin->sin_addr = loopback_addr;
 804                 }
 805             } else {
 806                 sin->sin_addr = loopback_addr;
 807             }
 808         }
 809
 810         DEBUG_MISC((dfd, " addr.sin_port=%d, "
 811             "addr.sin_addr.s_addr=%.16s\n",
 812             ntohs(sin->sin_port), inet_ntoa(sin->sin_addr)));
 813         break;
 814
 815     case AF_INET6:
 816         if (in6_equal_net(&so->so_faddr6, &slirp->vprefix_addr6,
 817                     slirp->vprefix_len)) {
 818             if (in6_equal(&so->so_faddr6, &slirp->vnameserver_addr6)) {
 819                 uint32_t scope_id;
 820                 if (get_dns6_addr(&sin6->sin6_addr, &scope_id) >= 0) {
 821                     sin6->sin6_scope_id = scope_id;
 822                 } else {
 823                     sin6->sin6_addr = in6addr_loopback;
 824                 }
 825             } else {
 826                 sin6->sin6_addr = in6addr_loopback;
 827             }
 828         }
 829         break;
 830
 831     default:
 832         break;
 833     }
 834 }
 835
 836 void sotranslate_in(struct socket *so, struct sockaddr_storage *addr)
 837 {
 838     Slirp *slirp = so->slirp;
 839     struct sockaddr_in *sin = (struct sockaddr_in *)addr;
 840     struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)addr;
 841
 842     switch (addr->ss_family) {
 843     case AF_INET:
 844         if ((so->so_faddr.s_addr & slirp->vnetwork_mask.s_addr) ==
 845             slirp->vnetwork_addr.s_addr) {
 846             uint32_t inv_mask = ~slirp->vnetwork_mask.s_addr;
 847
 848             if ((so->so_faddr.s_addr & inv_mask) == inv_mask) {
 849                 sin->sin_addr = slirp->vhost_addr;
 850             } else if (sin->sin_addr.s_addr == loopback_addr.s_addr ||
 851                        so->so_faddr.s_addr != slirp->vhost_addr.s_addr) {
 852                 sin->sin_addr = so->so_faddr;
 853             }
 854         }
 855         break;
 856
 857     case AF_INET6:
 858         if (in6_equal_net(&so->so_faddr6, &slirp->vprefix_addr6,
 859                     slirp->vprefix_len)) {
 860             if (in6_equal(&sin6->sin6_addr, &in6addr_loopback)
 861                     || !in6_equal(&so->so_faddr6, &slirp->vhost_addr6)) {
 862                 sin6->sin6_addr = so->so_faddr6;
 863             }
 864         }
 865         break;
 866
 867     default:
 868         break;
 869     }
 870 }
 871
 872 /*
 873  * Translate connections from localhost to the real hostname
 874  */
 875 void sotranslate_accept(struct socket *so)
 876 {
 877     Slirp *slirp = so->slirp;
 878
 879     switch (so->so_ffamily) {
 880     case AF_INET:
 881         if (so->so_faddr.s_addr == INADDR_ANY ||
 882             (so->so_faddr.s_addr & loopback_mask) ==
 883             (loopback_addr.s_addr & loopback_mask)) {
 884            so->so_faddr = slirp->vhost_addr;
 885         }
 886         break;
 887
 888    case AF_INET6:
 889         if (in6_equal(&so->so_faddr6, &in6addr_any) ||
 890                 in6_equal(&so->so_faddr6, &in6addr_loopback)) {
 891            so->so_faddr6 = slirp->vhost_addr6;
 892         }
 893         break;
 894
 895     default:
 896         break;
 897     }
 898 }