slirp/socket.c

   1 /*
   2  * Copyright (c) 1995 Danny Gasparovski.
   3  *
   4  * Please read the file COPYRIGHT for the
   5  * terms and conditions of the copyright.
   6  */
   7
   8 #include "qemu/osdep.h"
   9 #include "qemu-common.h"
  10 #include "slirp.h"
  11 #include "ip_icmp.h"
  12 #ifdef __sun__
  13 #include <sys/filio.h>
  14 #endif
  15
  16 static void sofcantrcvmore(struct socket *so);
  17 static void sofcantsendmore(struct socket *so);
  18
  19 struct socket *solookup(struct socket **last, struct socket *head,
  20         struct sockaddr_storage *lhost, struct sockaddr_storage *fhost)
  21 {
  22     struct socket *so = *last;
  23
  24     /* Optimisation */
  25     if (so != head && sockaddr_equal(&(so->lhost.ss), lhost)
  26             && (!fhost || sockaddr_equal(&so->fhost.ss, fhost))) {
  27         return so;
  28     }
  29
  30     for (so = head->so_next; so != head; so = so->so_next) {
  31         if (sockaddr_equal(&(so->lhost.ss), lhost)
  32                 && (!fhost || sockaddr_equal(&so->fhost.ss, fhost))) {
  33             *last = so;
  34             return so;
  35         }
  36     }
  37
  38     return (struct socket *)NULL;
  39 }
  40
  41 /*
  42  * Create a new socket, initialise the fields
  43  * It is the responsibility of the caller to
  44  * insque() it into the correct linked-list
  45  */
  46 struct socket *
  47 socreate(Slirp *slirp)
  48 {
  49   struct socket *so;
  50
  51   so = (struct socket *)malloc(sizeof(struct socket));
  52   if(so) {
  53     memset(so, 0, sizeof(struct socket));
  54     so->so_state = SS_NOFDREF;
  55     so->s = -1;
  56     so->slirp = slirp;
  57     so->pollfds_idx = -1;
  58   }
  59   return(so);
  60 }
  61
  62 /*
  63  * remque and free a socket, clobber cache
  64  */
  65 void
  66 sofree(struct socket *so)
  67 {
  68   Slirp *slirp = so->slirp;
  69   struct mbuf *ifm;
  70
  71   for (ifm = (struct mbuf *) slirp->if_fastq.qh_link;
  72        (struct quehead *) ifm != &slirp->if_fastq;
  73        ifm = ifm->ifq_next) {
  74     if (ifm->ifq_so == so) {
  75       ifm->ifq_so = NULL;
  76     }
  77   }
  78
  79   for (ifm = (struct mbuf *) slirp->if_batchq.qh_link;
  80        (struct quehead *) ifm != &slirp->if_batchq;
  81        ifm = ifm->ifq_next) {
  82     if (ifm->ifq_so == so) {
  83       ifm->ifq_so = NULL;
  84     }
  85   }
  86
  87   if (so->so_emu==EMU_RSH && so->extra) {
  88         sofree(so->extra);
  89         so->extra=NULL;
  90   }
  91   if (so == slirp->tcp_last_so) {
  92       slirp->tcp_last_so = &slirp->tcb;
  93   } else if (so == slirp->udp_last_so) {
  94       slirp->udp_last_so = &slirp->udb;
  95   } else if (so == slirp->icmp_last_so) {
  96       slirp->icmp_last_so = &slirp->icmp;
  97   }
  98   m_free(so->so_m);
  99
 100   if(so->so_next && so->so_prev)
 101     remque(so);  /* crashes if so is not in a queue */
 102
 103   if (so->so_tcpcb) {
 104       free(so->so_tcpcb);
 105   }
 106   free(so);
 107 }
 108
 109 size_t sopreprbuf(struct socket *so, struct iovec *iov, int *np)
 110 {
 111         int n, lss, total;
 112         struct sbuf *sb = &so->so_snd;
 113         int len = sb->sb_datalen - sb->sb_cc;
 114         int mss = so->so_tcpcb->t_maxseg;
 115
 116         DEBUG_CALL("sopreprbuf");
 117         DEBUG_ARG("so = %p", so);
 118
 119         if (len <= 0)
 120                 return 0;
 121
 122         iov[0].iov_base = sb->sb_wptr;
 123         iov[1].iov_base = NULL;
 124         iov[1].iov_len = 0;
 125         if (sb->sb_wptr < sb->sb_rptr) {
 126                 iov[0].iov_len = sb->sb_rptr - sb->sb_wptr;
 127                 /* Should never succeed, but... */
 128                 if (iov[0].iov_len > len)
 129                    iov[0].iov_len = len;
 130                 if (iov[0].iov_len > mss)
 131                    iov[0].iov_len -= iov[0].iov_len%mss;
 132                 n = 1;
 133         } else {
 134                 iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_wptr;
 135                 /* Should never succeed, but... */
 136                 if (iov[0].iov_len > len) iov[0].iov_len = len;
 137                 len -= iov[0].iov_len;
 138                 if (len) {
 139                         iov[1].iov_base = sb->sb_data;
 140                         iov[1].iov_len = sb->sb_rptr - sb->sb_data;
 141                         if(iov[1].iov_len > len)
 142                            iov[1].iov_len = len;
 143                         total = iov[0].iov_len + iov[1].iov_len;
 144                         if (total > mss) {
 145                                 lss = total%mss;
 146                                 if (iov[1].iov_len > lss) {
 147                                         iov[1].iov_len -= lss;
 148                                         n = 2;
 149                                 } else {
 150                                         lss -= iov[1].iov_len;
 151                                         iov[0].iov_len -= lss;
 152                                         n = 1;
 153                                 }
 154                         } else
 155                                 n = 2;
 156                 } else {
 157                         if (iov[0].iov_len > mss)
 158                            iov[0].iov_len -= iov[0].iov_len%mss;
 159                         n = 1;
 160                 }
 161         }
 162         if (np)
 163                 *np = n;
 164
 165         return iov[0].iov_len + (n - 1) * iov[1].iov_len;
 166 }
 167
 168 /*
 169  * Read from so's socket into sb_snd, updating all relevant sbuf fields
 170  * NOTE: This will only be called if it is select()ed for reading, so
 171  * a read() of 0 (or less) means it's disconnected
 172  */
 173 int
 174 soread(struct socket *so)
 175 {
 176         int n, nn;
 177         struct sbuf *sb = &so->so_snd;
 178         struct iovec iov[2];
 179
 180         DEBUG_CALL("soread");
 181         DEBUG_ARG("so = %p", so);
 182
 183         /*
 184          * No need to check if there's enough room to read.
 185          * soread wouldn't have been called if there weren't
 186          */
 187         sopreprbuf(so, iov, &n);
 188
 189 #ifdef HAVE_READV
 190         nn = readv(so->s, (struct iovec *)iov, n);
 191         DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
 192 #else
 193         nn = qemu_recv(so->s, iov[0].iov_base, iov[0].iov_len,0);
 194 #endif
 195         if (nn <= 0) {
 196                 if (nn < 0 && (errno == EINTR || errno == EAGAIN))
 197                         return 0;
 198                 else {
 199                         int err;
 200                         socklen_t slen = sizeof err;
 201
 202                         err = errno;
 203                         if (nn == 0) {
 204                                 getsockopt(so->s, SOL_SOCKET, SO_ERROR,
 205                                            &err, &slen);
 206                         }
 207
 208                         DEBUG_MISC((dfd, " --- soread() disconnected, nn = %d, errno = %d-%s\n", nn, errno,strerror(errno)));
 209                         sofcantrcvmore(so);
 210
 211                         if (err == ECONNRESET || err == ECONNREFUSED
 212                             || err == ENOTCONN || err == EPIPE) {
 213                                 tcp_drop(sototcpcb(so), err);
 214                         } else {
 215                                 tcp_sockclosed(sototcpcb(so));
 216                         }
 217                         return -1;
 218                 }
 219         }
 220
 221 #ifndef HAVE_READV
 222         /*
 223          * If there was no error, try and read the second time round
 224          * We read again if n = 2 (ie, there's another part of the buffer)
 225          * and we read as much as we could in the first read
 226          * We don't test for <= 0 this time, because there legitimately
 227          * might not be any more data (since the socket is non-blocking),
 228          * a close will be detected on next iteration.
 229          * A return of -1 won't (shouldn't) happen, since it didn't happen above
 230          */
 231         if (n == 2 && nn == iov[0].iov_len) {
 232             int ret;
 233             ret = qemu_recv(so->s, iov[1].iov_base, iov[1].iov_len,0);
 234             if (ret > 0)
 235                 nn += ret;
 236         }
 237
 238         DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
 239 #endif
 240
 241         /* Update fields */
 242         sb->sb_cc += nn;
 243         sb->sb_wptr += nn;
 244         if (sb->sb_wptr >= (sb->sb_data + sb->sb_datalen))
 245                 sb->sb_wptr -= sb->sb_datalen;
 246         return nn;
 247 }
 248
 249 int soreadbuf(struct socket *so, const char *buf, int size)
 250 {
 251     int n, nn, copy = size;
 252         struct sbuf *sb = &so->so_snd;
 253         struct iovec iov[2];
 254
 255         DEBUG_CALL("soreadbuf");
 256         DEBUG_ARG("so = %p", so);
 257
 258         /*
 259          * No need to check if there's enough room to read.
 260          * soread wouldn't have been called if there weren't
 261          */
 262         if (sopreprbuf(so, iov, &n) < size)
 263         goto err;
 264
 265     nn = MIN(iov[0].iov_len, copy);
 266     memcpy(iov[0].iov_base, buf, nn);
 267
 268     copy -= nn;
 269     buf += nn;
 270
 271     if (copy == 0)
 272         goto done;
 273
 274     memcpy(iov[1].iov_base, buf, copy);
 275
 276 done:
 277     /* Update fields */
 278         sb->sb_cc += size;
 279         sb->sb_wptr += size;
 280         if (sb->sb_wptr >= (sb->sb_data + sb->sb_datalen))
 281                 sb->sb_wptr -= sb->sb_datalen;
 282     return size;
 283 err:
 284
 285     sofcantrcvmore(so);
 286     tcp_sockclosed(sototcpcb(so));
 287     fprintf(stderr, "soreadbuf buffer to small");
 288     return -1;
 289 }
 290
 291 /*
 292  * Get urgent data
 293  *
 294  * When the socket is created, we set it SO_OOBINLINE,
 295  * so when OOB data arrives, we soread() it and everything
 296  * in the send buffer is sent as urgent data
 297  */
 298 int
 299 sorecvoob(struct socket *so)
 300 {
 301         struct tcpcb *tp = sototcpcb(so);
 302         int ret;
 303
 304         DEBUG_CALL("sorecvoob");
 305         DEBUG_ARG("so = %p", so);
 306
 307         /*
 308          * We take a guess at how much urgent data has arrived.
 309          * In most situations, when urgent data arrives, the next
 310          * read() should get all the urgent data.  This guess will
 311          * be wrong however if more data arrives just after the
 312          * urgent data, or the read() doesn't return all the
 313          * urgent data.
 314          */
 315         ret = soread(so);
 316         if (ret > 0) {
 317             tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
 318             tp->t_force = 1;
 319             tcp_output(tp);
 320             tp->t_force = 0;
 321         }
 322
 323         return ret;
 324 }
 325
 326 /*
 327  * Send urgent data
 328  * There's a lot duplicated code here, but...
 329  */
 330 int
 331 sosendoob(struct socket *so)
 332 {
 333         struct sbuf *sb = &so->so_rcv;
 334         char buff[2048]; /* XXX Shouldn't be sending more oob data than this */
 335
 336         int n, len;
 337
 338         DEBUG_CALL("sosendoob");
 339         DEBUG_ARG("so = %p", so);
 340         DEBUG_ARG("sb->sb_cc = %d", sb->sb_cc);
 341
 342         if (so->so_urgc > 2048)
 343            so->so_urgc = 2048; /* XXXX */
 344
 345         if (sb->sb_rptr < sb->sb_wptr) {
 346                 /* We can send it directly */
 347                 n = slirp_send(so, sb->sb_rptr, so->so_urgc, (MSG_OOB)); /* |MSG_DONTWAIT)); */
 348         } else {
 349                 /*
 350                  * Since there's no sendv or sendtov like writev,
 351                  * we must copy all data to a linear buffer then
 352                  * send it all
 353                  */
 354                 uint32_t urgc = so->so_urgc;
 355                 len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
 356                 if (len > urgc) {
 357                         len = urgc;
 358                 }
 359                 memcpy(buff, sb->sb_rptr, len);
 360                 urgc -= len;
 361                 if (urgc) {
 362                         n = sb->sb_wptr - sb->sb_data;
 363                         if (n > urgc) {
 364                                 n = urgc;
 365                         }
 366                         memcpy((buff + len), sb->sb_data, n);
 367                         len += n;
 368                 }
 369                 n = slirp_send(so, buff, len, (MSG_OOB)); /* |MSG_DONTWAIT)); */
 370         }
 371
 372 #ifdef DEBUG
 373         if (n != len) {
 374                 DEBUG_ERROR((dfd, "Didn't send all data urgently XXXXX\n"));
 375         }
 376 #endif
 377         if (n < 0) {
 378                 return n;
 379         }
 380         so->so_urgc -= n;
 381         DEBUG_MISC((dfd, " ---2 sent %d bytes urgent data, %d urgent bytes left\n", n, so->so_urgc));
 382
 383         sb->sb_cc -= n;
 384         sb->sb_rptr += n;
 385         if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
 386                 sb->sb_rptr -= sb->sb_datalen;
 387
 388         return n;
 389 }
 390
 391 /*
 392  * Write data from so_rcv to so's socket,
 393  * updating all sbuf field as necessary
 394  */
 395 int
 396 sowrite(struct socket *so)
 397 {
 398         int  n,nn;
 399         struct sbuf *sb = &so->so_rcv;
 400         int len = sb->sb_cc;
 401         struct iovec iov[2];
 402
 403         DEBUG_CALL("sowrite");
 404         DEBUG_ARG("so = %p", so);
 405
 406         if (so->so_urgc) {
 407                 uint32_t expected = so->so_urgc;
 408                 if (sosendoob(so) < expected) {
 409                         /* Treat a short write as a fatal error too,
 410                          * rather than continuing on and sending the urgent
 411                          * data as if it were non-urgent and leaving the
 412                          * so_urgc count wrong.
 413                          */
 414                         goto err_disconnected;
 415                 }
 416                 if (sb->sb_cc == 0)
 417                         return 0;
 418         }
 419
 420         /*
 421          * No need to check if there's something to write,
 422          * sowrite wouldn't have been called otherwise
 423          */
 424
 425         iov[0].iov_base = sb->sb_rptr;
 426         iov[1].iov_base = NULL;
 427         iov[1].iov_len = 0;
 428         if (sb->sb_rptr < sb->sb_wptr) {
 429                 iov[0].iov_len = sb->sb_wptr - sb->sb_rptr;
 430                 /* Should never succeed, but... */
 431                 if (iov[0].iov_len > len) iov[0].iov_len = len;
 432                 n = 1;
 433         } else {
 434                 iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
 435                 if (iov[0].iov_len > len) iov[0].iov_len = len;
 436                 len -= iov[0].iov_len;
 437                 if (len) {
 438                         iov[1].iov_base = sb->sb_data;
 439                         iov[1].iov_len = sb->sb_wptr - sb->sb_data;
 440                         if (iov[1].iov_len > len) iov[1].iov_len = len;
 441                         n = 2;
 442                 } else
 443                         n = 1;
 444         }
 445         /* Check if there's urgent data to send, and if so, send it */
 446
 447 #ifdef HAVE_READV
 448         nn = writev(so->s, (const struct iovec *)iov, n);
 449
 450         DEBUG_MISC((dfd, "  ... wrote nn = %d bytes\n", nn));
 451 #else
 452         nn = slirp_send(so, iov[0].iov_base, iov[0].iov_len,0);
 453 #endif
 454         /* This should never happen, but people tell me it does *shrug* */
 455         if (nn < 0 && (errno == EAGAIN || errno == EINTR))
 456                 return 0;
 457
 458         if (nn <= 0) {
 459                 goto err_disconnected;
 460         }
 461
 462 #ifndef HAVE_READV
 463         if (n == 2 && nn == iov[0].iov_len) {
 464             int ret;
 465             ret = slirp_send(so, iov[1].iov_base, iov[1].iov_len,0);
 466             if (ret > 0)
 467                 nn += ret;
 468         }
 469         DEBUG_MISC((dfd, "  ... wrote nn = %d bytes\n", nn));
 470 #endif
 471
 472         /* Update sbuf */
 473         sb->sb_cc -= nn;
 474         sb->sb_rptr += nn;
 475         if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
 476                 sb->sb_rptr -= sb->sb_datalen;
 477
 478         /*
 479          * If in DRAIN mode, and there's no more data, set
 480          * it CANTSENDMORE
 481          */
 482         if ((so->so_state & SS_FWDRAIN) && sb->sb_cc == 0)
 483                 sofcantsendmore(so);
 484
 485         return nn;
 486
 487 err_disconnected:
 488         DEBUG_MISC((dfd, " --- sowrite disconnected, so->so_state = %x, errno = %d\n",
 489                     so->so_state, errno));
 490         sofcantsendmore(so);
 491         tcp_sockclosed(sototcpcb(so));
 492         return -1;
 493 }
 494
 495 /*
 496  * recvfrom() a UDP socket
 497  */
 498 void
 499 sorecvfrom(struct socket *so)
 500 {
 501         struct sockaddr_storage addr;
 502         struct sockaddr_storage saddr, daddr;
 503         socklen_t addrlen = sizeof(struct sockaddr_storage);
 504
 505         DEBUG_CALL("sorecvfrom");
 506         DEBUG_ARG("so = %p", so);
 507
 508         if (so->so_type == IPPROTO_ICMP) {   /* This is a "ping" reply */
 509           char buff[256];
 510           int len;
 511
 512           len = recvfrom(so->s, buff, 256, 0,
 513                          (struct sockaddr *)&addr, &addrlen);
 514           /* XXX Check if reply is "correct"? */
 515
 516           if(len == -1 || len == 0) {
 517             u_char code=ICMP_UNREACH_PORT;
 518
 519             if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
 520             else if(errno == ENETUNREACH) code=ICMP_UNREACH_NET;
 521
 522             DEBUG_MISC((dfd," udp icmp rx errno = %d-%s\n",
 523                         errno,strerror(errno)));
 524             icmp_send_error(so->so_m, ICMP_UNREACH, code, 0, strerror(errno));
 525           } else {
 526             icmp_reflect(so->so_m);
 527             so->so_m = NULL; /* Don't m_free() it again! */
 528           }
 529           /* No need for this socket anymore, udp_detach it */
 530           udp_detach(so);
 531         } else {                                /* A "normal" UDP packet */
 532           struct mbuf *m;
 533           int len;
 534 #ifdef _WIN32
 535           unsigned long n;
 536 #else
 537           int n;
 538 #endif
 539
 540           m = m_get(so->slirp);
 541           if (!m) {
 542               return;
 543           }
 544           switch (so->so_ffamily) {
 545           case AF_INET:
 546               m->m_data += IF_MAXLINKHDR + sizeof(struct udpiphdr);
 547               break;
 548           case AF_INET6:
 549               m->m_data += IF_MAXLINKHDR + sizeof(struct ip6)
 550                                          + sizeof(struct udphdr);
 551               break;
 552           default:
 553               g_assert_not_reached();
 554               break;
 555           }
 556
 557           /*
 558            * XXX Shouldn't FIONREAD packets destined for port 53,
 559            * but I don't know the max packet size for DNS lookups
 560            */
 561           len = M_FREEROOM(m);
 562           /* if (so->so_fport != htons(53)) { */
 563           ioctlsocket(so->s, FIONREAD, &n);
 564
 565           if (n > len) {
 566             n = (m->m_data - m->m_dat) + m->m_len + n + 1;
 567             m_inc(m, n);
 568             len = M_FREEROOM(m);
 569           }
 570           /* } */
 571
 572           m->m_len = recvfrom(so->s, m->m_data, len, 0,
 573                               (struct sockaddr *)&addr, &addrlen);
 574           DEBUG_MISC((dfd, " did recvfrom %d, errno = %d-%s\n",
 575                       m->m_len, errno,strerror(errno)));
 576           if(m->m_len<0) {
 577             /* Report error as ICMP */
 578             switch (so->so_lfamily) {
 579             uint8_t code;
 580             case AF_INET:
 581               code = ICMP_UNREACH_PORT;
 582
 583               if (errno == EHOSTUNREACH) {
 584                 code = ICMP_UNREACH_HOST;
 585               } else if (errno == ENETUNREACH) {
 586                 code = ICMP_UNREACH_NET;
 587               }
 588
 589               DEBUG_MISC((dfd, " rx error, tx icmp ICMP_UNREACH:%i\n", code));
 590               icmp_send_error(so->so_m, ICMP_UNREACH, code, 0, strerror(errno));
 591               break;
 592             case AF_INET6:
 593               code = ICMP6_UNREACH_PORT;
 594
 595               if (errno == EHOSTUNREACH) {
 596                 code = ICMP6_UNREACH_ADDRESS;
 597               } else if (errno == ENETUNREACH) {
 598                 code = ICMP6_UNREACH_NO_ROUTE;
 599               }
 600
 601               DEBUG_MISC((dfd, " rx error, tx icmp6 ICMP_UNREACH:%i\n", code));
 602               icmp6_send_error(so->so_m, ICMP6_UNREACH, code);
 603               break;
 604             default:
 605               g_assert_not_reached();
 606               break;
 607             }
 608             m_free(m);
 609           } else {
 610           /*
 611            * Hack: domain name lookup will be used the most for UDP,
 612            * and since they'll only be used once there's no need
 613            * for the 4 minute (or whatever) timeout... So we time them
 614            * out much quicker (10 seconds  for now...)
 615            */
 616             if (so->so_expire) {
 617               if (so->so_fport == htons(53))
 618                 so->so_expire = curtime + SO_EXPIREFAST;
 619               else
 620                 so->so_expire = curtime + SO_EXPIRE;
 621             }
 622
 623             /*
 624              * If this packet was destined for CTL_ADDR,
 625              * make it look like that's where it came from
 626              */
 627             saddr = addr;
 628             sotranslate_in(so, &saddr);
 629             daddr = so->lhost.ss;
 630
 631             switch (so->so_ffamily) {
 632             case AF_INET:
 633                 udp_output(so, m, (struct sockaddr_in *) &saddr,
 634                            (struct sockaddr_in *) &daddr,
 635                            so->so_iptos);
 636                 break;
 637             case AF_INET6:
 638                 udp6_output(so, m, (struct sockaddr_in6 *) &saddr,
 639                             (struct sockaddr_in6 *) &daddr);
 640                 break;
 641             default:
 642                 g_assert_not_reached();
 643                 break;
 644             }
 645           } /* rx error */
 646         } /* if ping packet */
 647 }
 648
 649 /*
 650  * sendto() a socket
 651  */
 652 int
 653 sosendto(struct socket *so, struct mbuf *m)
 654 {
 655         int ret;
 656         struct sockaddr_storage addr;
 657
 658         DEBUG_CALL("sosendto");
 659         DEBUG_ARG("so = %p", so);
 660         DEBUG_ARG("m = %p", m);
 661
 662         addr = so->fhost.ss;
 663         DEBUG_CALL(" sendto()ing)");
 664         sotranslate_out(so, &addr);
 665
 666         /* Don't care what port we get */
 667         ret = sendto(so->s, m->m_data, m->m_len, 0,
 668                      (struct sockaddr *)&addr, sockaddr_size(&addr));
 669         if (ret < 0)
 670                 return -1;
 671
 672         /*
 673          * Kill the socket if there's no reply in 4 minutes,
 674          * but only if it's an expirable socket
 675          */
 676         if (so->so_expire)
 677                 so->so_expire = curtime + SO_EXPIRE;
 678         so->so_state &= SS_PERSISTENT_MASK;
 679         so->so_state |= SS_ISFCONNECTED; /* So that it gets select()ed */
 680         return 0;
 681 }
 682
 683 /*
 684  * Listen for incoming TCP connections
 685  */
 686 struct socket *
 687 tcp_listen(Slirp *slirp, uint32_t haddr, u_int hport, uint32_t laddr,
 688            u_int lport, int flags)
 689 {
 690         struct sockaddr_in addr;
 691         struct socket *so;
 692         int s, opt = 1;
 693         socklen_t addrlen = sizeof(addr);
 694         memset(&addr, 0, addrlen);
 695
 696         DEBUG_CALL("tcp_listen");
 697         DEBUG_ARG("haddr = %x", haddr);
 698         DEBUG_ARG("hport = %d", hport);
 699         DEBUG_ARG("laddr = %x", laddr);
 700         DEBUG_ARG("lport = %d", lport);
 701         DEBUG_ARG("flags = %x", flags);
 702
 703         so = socreate(slirp);
 704         if (!so) {
 705           return NULL;
 706         }
 707
 708         /* Don't tcp_attach... we don't need so_snd nor so_rcv */
 709         if ((so->so_tcpcb = tcp_newtcpcb(so)) == NULL) {
 710                 free(so);
 711                 return NULL;
 712         }
 713         insque(so, &slirp->tcb);
 714
 715         /*
 716          * SS_FACCEPTONCE sockets must time out.
 717          */
 718         if (flags & SS_FACCEPTONCE)
 719            so->so_tcpcb->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT*2;
 720
 721         so->so_state &= SS_PERSISTENT_MASK;
 722         so->so_state |= (SS_FACCEPTCONN | flags);
 723         so->so_lfamily = AF_INET;
 724         so->so_lport = lport; /* Kept in network format */
 725         so->so_laddr.s_addr = laddr; /* Ditto */
 726
 727         addr.sin_family = AF_INET;
 728         addr.sin_addr.s_addr = haddr;
 729         addr.sin_port = hport;
 730
 731         if (((s = qemu_socket(AF_INET,SOCK_STREAM,0)) < 0) ||
 732             (socket_set_fast_reuse(s) < 0) ||
 733             (bind(s,(struct sockaddr *)&addr, sizeof(addr)) < 0) ||
 734             (listen(s,1) < 0)) {
 735                 int tmperrno = errno; /* Don't clobber the real reason we failed */
 736
 737                 if (s >= 0) {
 738                     closesocket(s);
 739                 }
 740                 sofree(so);
 741                 /* Restore the real errno */
 742 #ifdef _WIN32
 743                 WSASetLastError(tmperrno);
 744 #else
 745                 errno = tmperrno;
 746 #endif
 747                 return NULL;
 748         }
 749         qemu_setsockopt(s, SOL_SOCKET, SO_OOBINLINE, &opt, sizeof(int));
 750
 751         getsockname(s,(struct sockaddr *)&addr,&addrlen);
 752         so->so_ffamily = AF_INET;
 753         so->so_fport = addr.sin_port;
 754         if (addr.sin_addr.s_addr == 0 || addr.sin_addr.s_addr == loopback_addr.s_addr)
 755            so->so_faddr = slirp->vhost_addr;
 756         else
 757            so->so_faddr = addr.sin_addr;
 758
 759         so->s = s;
 760         return so;
 761 }
 762
 763 /*
 764  * Various session state calls
 765  * XXX Should be #define's
 766  * The socket state stuff needs work, these often get call 2 or 3
 767  * times each when only 1 was needed
 768  */
 769 void
 770 soisfconnecting(struct socket *so)
 771 {
 772         so->so_state &= ~(SS_NOFDREF|SS_ISFCONNECTED|SS_FCANTRCVMORE|
 773                           SS_FCANTSENDMORE|SS_FWDRAIN);
 774         so->so_state |= SS_ISFCONNECTING; /* Clobber other states */
 775 }
 776
 777 void
 778 soisfconnected(struct socket *so)
 779 {
 780         so->so_state &= ~(SS_ISFCONNECTING|SS_FWDRAIN|SS_NOFDREF);
 781         so->so_state |= SS_ISFCONNECTED; /* Clobber other states */
 782 }
 783
 784 static void
 785 sofcantrcvmore(struct socket *so)
 786 {
 787         if ((so->so_state & SS_NOFDREF) == 0) {
 788                 shutdown(so->s,0);
 789         }
 790         so->so_state &= ~(SS_ISFCONNECTING);
 791         if (so->so_state & SS_FCANTSENDMORE) {
 792            so->so_state &= SS_PERSISTENT_MASK;
 793            so->so_state |= SS_NOFDREF; /* Don't select it */
 794         } else {
 795            so->so_state |= SS_FCANTRCVMORE;
 796         }
 797 }
 798
 799 static void
 800 sofcantsendmore(struct socket *so)
 801 {
 802         if ((so->so_state & SS_NOFDREF) == 0) {
 803             shutdown(so->s,1);           /* send FIN to fhost */
 804         }
 805         so->so_state &= ~(SS_ISFCONNECTING);
 806         if (so->so_state & SS_FCANTRCVMORE) {
 807            so->so_state &= SS_PERSISTENT_MASK;
 808            so->so_state |= SS_NOFDREF; /* as above */
 809         } else {
 810            so->so_state |= SS_FCANTSENDMORE;
 811         }
 812 }
 813
 814 /*
 815  * Set write drain mode
 816  * Set CANTSENDMORE once all data has been write()n
 817  */
 818 void
 819 sofwdrain(struct socket *so)
 820 {
 821         if (so->so_rcv.sb_cc)
 822                 so->so_state |= SS_FWDRAIN;
 823         else
 824                 sofcantsendmore(so);
 825 }
 826
 827 /*
 828  * Translate addr in host addr when it is a virtual address
 829  */
 830 void sotranslate_out(struct socket *so, struct sockaddr_storage *addr)
 831 {
 832     Slirp *slirp = so->slirp;
 833     struct sockaddr_in *sin = (struct sockaddr_in *)addr;
 834     struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)addr;
 835
 836     switch (addr->ss_family) {
 837     case AF_INET:
 838         if ((so->so_faddr.s_addr & slirp->vnetwork_mask.s_addr) ==
 839                 slirp->vnetwork_addr.s_addr) {
 840             /* It's an alias */
 841             if (so->so_faddr.s_addr == slirp->vnameserver_addr.s_addr) {
 842                 if (get_dns_addr(&sin->sin_addr) < 0) {
 843                     sin->sin_addr = loopback_addr;
 844                 }
 845             } else {
 846                 sin->sin_addr = loopback_addr;
 847             }
 848         }
 849
 850         DEBUG_MISC((dfd, " addr.sin_port=%d, "
 851             "addr.sin_addr.s_addr=%.16s\n",
 852             ntohs(sin->sin_port), inet_ntoa(sin->sin_addr)));
 853         break;
 854
 855     case AF_INET6:
 856         if (in6_equal_net(&so->so_faddr6, &slirp->vprefix_addr6,
 857                     slirp->vprefix_len)) {
 858             if (in6_equal(&so->so_faddr6, &slirp->vnameserver_addr6)) {
 859                 uint32_t scope_id;
 860                 if (get_dns6_addr(&sin6->sin6_addr, &scope_id) >= 0) {
 861                     sin6->sin6_scope_id = scope_id;
 862                 } else {
 863                     sin6->sin6_addr = in6addr_loopback;
 864                 }
 865             } else {
 866                 sin6->sin6_addr = in6addr_loopback;
 867             }
 868         }
 869         break;
 870
 871     default:
 872         break;
 873     }
 874 }
 875
 876 void sotranslate_in(struct socket *so, struct sockaddr_storage *addr)
 877 {
 878     Slirp *slirp = so->slirp;
 879     struct sockaddr_in *sin = (struct sockaddr_in *)addr;
 880     struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)addr;
 881
 882     switch (addr->ss_family) {
 883     case AF_INET:
 884         if ((so->so_faddr.s_addr & slirp->vnetwork_mask.s_addr) ==
 885             slirp->vnetwork_addr.s_addr) {
 886             uint32_t inv_mask = ~slirp->vnetwork_mask.s_addr;
 887
 888             if ((so->so_faddr.s_addr & inv_mask) == inv_mask) {
 889                 sin->sin_addr = slirp->vhost_addr;
 890             } else if (sin->sin_addr.s_addr == loopback_addr.s_addr ||
 891                        so->so_faddr.s_addr != slirp->vhost_addr.s_addr) {
 892                 sin->sin_addr = so->so_faddr;
 893             }
 894         }
 895         break;
 896
 897     case AF_INET6:
 898         if (in6_equal_net(&so->so_faddr6, &slirp->vprefix_addr6,
 899                     slirp->vprefix_len)) {
 900             if (in6_equal(&sin6->sin6_addr, &in6addr_loopback)
 901                     || !in6_equal(&so->so_faddr6, &slirp->vhost_addr6)) {
 902                 sin6->sin6_addr = so->so_faddr6;
 903             }
 904         }
 905         break;
 906
 907     default:
 908         break;
 909     }
 910 }
 911
 912 /*
 913  * Translate connections from localhost to the real hostname
 914  */
 915 void sotranslate_accept(struct socket *so)
 916 {
 917     Slirp *slirp = so->slirp;
 918
 919     switch (so->so_ffamily) {
 920     case AF_INET:
 921         if (so->so_faddr.s_addr == INADDR_ANY ||
 922             (so->so_faddr.s_addr & loopback_mask) ==
 923             (loopback_addr.s_addr & loopback_mask)) {
 924            so->so_faddr = slirp->vhost_addr;
 925         }
 926         break;
 927
 928    case AF_INET6:
 929         if (in6_equal(&so->so_faddr6, &in6addr_any) ||
 930                 in6_equal(&so->so_faddr6, &in6addr_loopback)) {
 931            so->so_faddr6 = slirp->vhost_addr6;
 932         }
 933         break;
 934
 935     default:
 936         break;
 937     }
 938 }