slirp/src/slirp.c

   1 /*
   2  * libslirp glue
   3  *
   4  * Copyright (c) 2004-2008 Fabrice Bellard
   5  *
   6  * Permission is hereby granted, free of charge, to any person obtaining a copy
   7  * of this software and associated documentation files (the "Software"), to deal
   8  * in the Software without restriction, including without limitation the rights
   9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  10  * copies of the Software, and to permit persons to whom the Software is
  11  * furnished to do so, subject to the following conditions:
  12  *
  13  * The above copyright notice and this permission notice shall be included in
  14  * all copies or substantial portions of the Software.
  15  *
  16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  22  * THE SOFTWARE.
  23  */
  24 #include "slirp.h"
  25
  26
  27 #ifndef _WIN32
  28 #include <net/if.h>
  29 #endif
  30
  31 int slirp_debug;
  32
  33 /* Define to 1 if you want KEEPALIVE timers */
  34 bool slirp_do_keepalive;
  35
  36 /* host loopback address */
  37 struct in_addr loopback_addr;
  38 /* host loopback network mask */
  39 unsigned long loopback_mask;
  40
  41 /* emulated hosts use the MAC addr 52:55:IP:IP:IP:IP */
  42 static const uint8_t special_ethaddr[ETH_ALEN] = {
  43     0x52, 0x55, 0x00, 0x00, 0x00, 0x00
  44 };
  45
  46 unsigned curtime;
  47
  48 static struct in_addr dns_addr;
  49 #ifndef _WIN32
  50 static struct in6_addr dns6_addr;
  51 #endif
  52 static unsigned dns_addr_time;
  53 #ifndef _WIN32
  54 static unsigned dns6_addr_time;
  55 #endif
  56
  57 #define TIMEOUT_FAST 2  /* milliseconds */
  58 #define TIMEOUT_SLOW 499  /* milliseconds */
  59 /* for the aging of certain requests like DNS */
  60 #define TIMEOUT_DEFAULT 1000  /* milliseconds */
  61
  62 #ifdef _WIN32
  63
  64 int get_dns_addr(struct in_addr *pdns_addr)
  65 {
  66     FIXED_INFO *FixedInfo=NULL;
  67     ULONG    BufLen;
  68     DWORD    ret;
  69     IP_ADDR_STRING *pIPAddr;
  70     struct in_addr tmp_addr;
  71
  72     if (dns_addr.s_addr != 0 && (curtime - dns_addr_time) < TIMEOUT_DEFAULT) {
  73         *pdns_addr = dns_addr;
  74         return 0;
  75     }
  76
  77     FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
  78     BufLen = sizeof(FIXED_INFO);
  79
  80     if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen)) {
  81         if (FixedInfo) {
  82             GlobalFree(FixedInfo);
  83             FixedInfo = NULL;
  84         }
  85         FixedInfo = GlobalAlloc(GPTR, BufLen);
  86     }
  87
  88     if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS) {
  89         printf("GetNetworkParams failed. ret = %08x\n", (unsigned)ret );
  90         if (FixedInfo) {
  91             GlobalFree(FixedInfo);
  92             FixedInfo = NULL;
  93         }
  94         return -1;
  95     }
  96
  97     pIPAddr = &(FixedInfo->DnsServerList);
  98     inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
  99     *pdns_addr = tmp_addr;
 100     dns_addr = tmp_addr;
 101     dns_addr_time = curtime;
 102     if (FixedInfo) {
 103         GlobalFree(FixedInfo);
 104         FixedInfo = NULL;
 105     }
 106     return 0;
 107 }
 108
 109 int get_dns6_addr(struct in6_addr *pdns6_addr, uint32_t *scope_id)
 110 {
 111     return -1;
 112 }
 113
 114 static void winsock_cleanup(void)
 115 {
 116     WSACleanup();
 117 }
 118
 119 #else
 120
 121 static int get_dns_addr_cached(void *pdns_addr, void *cached_addr,
 122                                socklen_t addrlen,
 123                                struct stat *cached_stat, unsigned *cached_time)
 124 {
 125     struct stat old_stat;
 126     if (curtime - *cached_time < TIMEOUT_DEFAULT) {
 127         memcpy(pdns_addr, cached_addr, addrlen);
 128         return 0;
 129     }
 130     old_stat = *cached_stat;
 131     if (stat("/etc/resolv.conf", cached_stat) != 0) {
 132         return -1;
 133     }
 134     if (cached_stat->st_dev == old_stat.st_dev
 135         && cached_stat->st_ino == old_stat.st_ino
 136         && cached_stat->st_size == old_stat.st_size
 137         && cached_stat->st_mtime == old_stat.st_mtime) {
 138         memcpy(pdns_addr, cached_addr, addrlen);
 139         return 0;
 140     }
 141     return 1;
 142 }
 143
 144 static int get_dns_addr_resolv_conf(int af, void *pdns_addr, void *cached_addr,
 145                                     socklen_t addrlen, uint32_t *scope_id,
 146                                     unsigned *cached_time)
 147 {
 148     char buff[512];
 149     char buff2[257];
 150     FILE *f;
 151     int found = 0;
 152     void *tmp_addr = alloca(addrlen);
 153     unsigned if_index;
 154
 155     f = fopen("/etc/resolv.conf", "r");
 156     if (!f)
 157         return -1;
 158
 159     DEBUG_MISC("IP address of your DNS(s):");
 160     while (fgets(buff, 512, f) != NULL) {
 161         if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1) {
 162             char *c = strchr(buff2, '%');
 163             if (c) {
 164                 if_index = if_nametoindex(c + 1);
 165                 *c = '\0';
 166             } else {
 167                 if_index = 0;
 168             }
 169
 170             if (!inet_pton(af, buff2, tmp_addr)) {
 171                 continue;
 172             }
 173             /* If it's the first one, set it to dns_addr */
 174             if (!found) {
 175                 memcpy(pdns_addr, tmp_addr, addrlen);
 176                 memcpy(cached_addr, tmp_addr, addrlen);
 177                 if (scope_id) {
 178                     *scope_id = if_index;
 179                 }
 180                 *cached_time = curtime;
 181             }
 182
 183             if (++found > 3) {
 184                 DEBUG_MISC("  (more)");
 185                 break;
 186             } else if (slirp_debug & DBG_MISC) {
 187                 char s[INET6_ADDRSTRLEN];
 188                 const char *res = inet_ntop(af, tmp_addr, s, sizeof(s));
 189                 if (!res) {
 190                     res = "  (string conversion error)";
 191                 }
 192                 DEBUG_MISC("  %s", res);
 193             }
 194         }
 195     }
 196     fclose(f);
 197     if (!found)
 198         return -1;
 199     return 0;
 200 }
 201
 202 int get_dns_addr(struct in_addr *pdns_addr)
 203 {
 204     static struct stat dns_addr_stat;
 205
 206     if (dns_addr.s_addr != 0) {
 207         int ret;
 208         ret = get_dns_addr_cached(pdns_addr, &dns_addr, sizeof(dns_addr),
 209                                   &dns_addr_stat, &dns_addr_time);
 210         if (ret <= 0) {
 211             return ret;
 212         }
 213     }
 214     return get_dns_addr_resolv_conf(AF_INET, pdns_addr, &dns_addr,
 215                                     sizeof(dns_addr), NULL, &dns_addr_time);
 216 }
 217
 218 int get_dns6_addr(struct in6_addr *pdns6_addr, uint32_t *scope_id)
 219 {
 220     static struct stat dns6_addr_stat;
 221
 222     if (!in6_zero(&dns6_addr)) {
 223         int ret;
 224         ret = get_dns_addr_cached(pdns6_addr, &dns6_addr, sizeof(dns6_addr),
 225                                   &dns6_addr_stat, &dns6_addr_time);
 226         if (ret <= 0) {
 227             return ret;
 228         }
 229     }
 230     return get_dns_addr_resolv_conf(AF_INET6, pdns6_addr, &dns6_addr,
 231                                     sizeof(dns6_addr),
 232                                     scope_id, &dns6_addr_time);
 233 }
 234
 235 #endif
 236
 237 static void slirp_init_once(void)
 238 {
 239     static int initialized;
 240     const char *debug;
 241 #ifdef _WIN32
 242     WSADATA Data;
 243 #endif
 244
 245     if (initialized) {
 246         return;
 247     }
 248     initialized = 1;
 249
 250 #ifdef _WIN32
 251     WSAStartup(MAKEWORD(2,0), &Data);
 252     atexit(winsock_cleanup);
 253 #endif
 254
 255     loopback_addr.s_addr = htonl(INADDR_LOOPBACK);
 256     loopback_mask = htonl(IN_CLASSA_NET);
 257
 258     debug = g_getenv("SLIRP_DEBUG");
 259     if (debug) {
 260         const GDebugKey keys[] = {
 261             { "call", DBG_CALL },
 262             { "misc", DBG_MISC },
 263             { "error", DBG_ERROR },
 264             { "tftp", DBG_TFTP },
 265         };
 266         slirp_debug = g_parse_debug_string(debug, keys, G_N_ELEMENTS(keys));
 267     }
 268
 269
 270 }
 271
 272 Slirp *slirp_init(int restricted, bool in_enabled, struct in_addr vnetwork,
 273                   struct in_addr vnetmask, struct in_addr vhost,
 274                   bool in6_enabled,
 275                   struct in6_addr vprefix_addr6, uint8_t vprefix_len,
 276                   struct in6_addr vhost6, const char *vhostname,
 277                   const char *tftp_server_name,
 278                   const char *tftp_path, const char *bootfile,
 279                   struct in_addr vdhcp_start, struct in_addr vnameserver,
 280                   struct in6_addr vnameserver6, const char **vdnssearch,
 281                   const char *vdomainname,
 282                   const SlirpCb *callbacks,
 283                   void *opaque)
 284 {
 285     Slirp *slirp = g_malloc0(sizeof(Slirp));
 286
 287     slirp_init_once();
 288
 289     slirp->opaque = opaque;
 290     slirp->cb = callbacks;
 291     slirp->grand = g_rand_new();
 292     slirp->restricted = restricted;
 293
 294     slirp->in_enabled = in_enabled;
 295     slirp->in6_enabled = in6_enabled;
 296
 297     if_init(slirp);
 298     ip_init(slirp);
 299     ip6_init(slirp);
 300
 301     /* Initialise mbufs *after* setting the MTU */
 302     m_init(slirp);
 303
 304     slirp->vnetwork_addr = vnetwork;
 305     slirp->vnetwork_mask = vnetmask;
 306     slirp->vhost_addr = vhost;
 307     slirp->vprefix_addr6 = vprefix_addr6;
 308     slirp->vprefix_len = vprefix_len;
 309     slirp->vhost_addr6 = vhost6;
 310     if (vhostname) {
 311         slirp_pstrcpy(slirp->client_hostname, sizeof(slirp->client_hostname),
 312                       vhostname);
 313     }
 314     slirp->tftp_prefix = g_strdup(tftp_path);
 315     slirp->bootp_filename = g_strdup(bootfile);
 316     slirp->vdomainname = g_strdup(vdomainname);
 317     slirp->vdhcp_startaddr = vdhcp_start;
 318     slirp->vnameserver_addr = vnameserver;
 319     slirp->vnameserver_addr6 = vnameserver6;
 320     slirp->tftp_server_name = g_strdup(tftp_server_name);
 321
 322     if (vdnssearch) {
 323         translate_dnssearch(slirp, vdnssearch);
 324     }
 325
 326     return slirp;
 327 }
 328
 329 void slirp_cleanup(Slirp *slirp)
 330 {
 331     struct gfwd_list *e, *next;
 332
 333     for (e = slirp->guestfwd_list; e; e = next) {
 334         next = e->ex_next;
 335         g_free(e->ex_exec);
 336         g_free(e);
 337     }
 338
 339     ip_cleanup(slirp);
 340     ip6_cleanup(slirp);
 341     m_cleanup(slirp);
 342
 343     g_rand_free(slirp->grand);
 344
 345     g_free(slirp->vdnssearch);
 346     g_free(slirp->tftp_prefix);
 347     g_free(slirp->bootp_filename);
 348     g_free(slirp->vdomainname);
 349     g_free(slirp);
 350 }
 351
 352 #define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
 353 #define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
 354
 355 static void slirp_update_timeout(Slirp *slirp, uint32_t *timeout)
 356 {
 357     uint32_t t;
 358
 359     if (*timeout <= TIMEOUT_FAST) {
 360         return;
 361     }
 362
 363     t = MIN(1000, *timeout);
 364
 365     /* If we have tcp timeout with slirp, then we will fill @timeout with
 366      * more precise value.
 367      */
 368     if (slirp->time_fasttimo) {
 369         *timeout = TIMEOUT_FAST;
 370         return;
 371     }
 372     if (slirp->do_slowtimo) {
 373         t = MIN(TIMEOUT_SLOW, t);
 374     }
 375     *timeout = t;
 376 }
 377
 378 void slirp_pollfds_fill(Slirp *slirp, uint32_t *timeout,
 379                         SlirpAddPollCb add_poll, void *opaque)
 380 {
 381     struct socket *so, *so_next;
 382
 383     /*
 384      * First, TCP sockets
 385      */
 386
 387     /*
 388      * *_slowtimo needs calling if there are IP fragments
 389      * in the fragment queue, or there are TCP connections active
 390      */
 391     slirp->do_slowtimo = ((slirp->tcb.so_next != &slirp->tcb) ||
 392                           (&slirp->ipq.ip_link != slirp->ipq.ip_link.next));
 393
 394     for (so = slirp->tcb.so_next; so != &slirp->tcb; so = so_next) {
 395         int events = 0;
 396
 397         so_next = so->so_next;
 398
 399         so->pollfds_idx = -1;
 400
 401         /*
 402          * See if we need a tcp_fasttimo
 403          */
 404         if (slirp->time_fasttimo == 0 &&
 405             so->so_tcpcb->t_flags & TF_DELACK) {
 406             slirp->time_fasttimo = curtime; /* Flag when want a fasttimo */
 407         }
 408
 409         /*
 410          * NOFDREF can include still connecting to local-host,
 411          * newly socreated() sockets etc. Don't want to select these.
 412          */
 413         if (so->so_state & SS_NOFDREF || so->s == -1) {
 414             continue;
 415         }
 416
 417         /*
 418          * Set for reading sockets which are accepting
 419          */
 420         if (so->so_state & SS_FACCEPTCONN) {
 421             so->pollfds_idx = add_poll(so->s,
 422                 SLIRP_POLL_IN | SLIRP_POLL_HUP | SLIRP_POLL_ERR, opaque);
 423             continue;
 424         }
 425
 426         /*
 427          * Set for writing sockets which are connecting
 428          */
 429         if (so->so_state & SS_ISFCONNECTING) {
 430             so->pollfds_idx = add_poll(so->s,
 431                 SLIRP_POLL_OUT | SLIRP_POLL_ERR, opaque);
 432             continue;
 433         }
 434
 435         /*
 436          * Set for writing if we are connected, can send more, and
 437          * we have something to send
 438          */
 439         if (CONN_CANFSEND(so) && so->so_rcv.sb_cc) {
 440             events |= SLIRP_POLL_OUT | SLIRP_POLL_ERR;
 441         }
 442
 443         /*
 444          * Set for reading (and urgent data) if we are connected, can
 445          * receive more, and we have room for it XXX /2 ?
 446          */
 447         if (CONN_CANFRCV(so) &&
 448             (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2))) {
 449             events |= SLIRP_POLL_IN | SLIRP_POLL_HUP |
 450                       SLIRP_POLL_ERR | SLIRP_POLL_PRI;
 451         }
 452
 453         if (events) {
 454             so->pollfds_idx = add_poll(so->s, events, opaque);
 455         }
 456     }
 457
 458     /*
 459      * UDP sockets
 460      */
 461     for (so = slirp->udb.so_next; so != &slirp->udb; so = so_next) {
 462         so_next = so->so_next;
 463
 464         so->pollfds_idx = -1;
 465
 466         /*
 467          * See if it's timed out
 468          */
 469         if (so->so_expire) {
 470             if (so->so_expire <= curtime) {
 471                 udp_detach(so);
 472                 continue;
 473             } else {
 474                 slirp->do_slowtimo = true; /* Let socket expire */
 475             }
 476         }
 477
 478         /*
 479          * When UDP packets are received from over the
 480          * link, they're sendto()'d straight away, so
 481          * no need for setting for writing
 482          * Limit the number of packets queued by this session
 483          * to 4.  Note that even though we try and limit this
 484          * to 4 packets, the session could have more queued
 485          * if the packets needed to be fragmented
 486          * (XXX <= 4 ?)
 487          */
 488         if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4) {
 489             so->pollfds_idx = add_poll(so->s,
 490                 SLIRP_POLL_IN | SLIRP_POLL_HUP | SLIRP_POLL_ERR, opaque);
 491         }
 492     }
 493
 494     /*
 495      * ICMP sockets
 496      */
 497     for (so = slirp->icmp.so_next; so != &slirp->icmp; so = so_next) {
 498         so_next = so->so_next;
 499
 500         so->pollfds_idx = -1;
 501
 502         /*
 503          * See if it's timed out
 504          */
 505         if (so->so_expire) {
 506             if (so->so_expire <= curtime) {
 507                 icmp_detach(so);
 508                 continue;
 509             } else {
 510                 slirp->do_slowtimo = true; /* Let socket expire */
 511             }
 512         }
 513
 514         if (so->so_state & SS_ISFCONNECTED) {
 515             so->pollfds_idx = add_poll(so->s,
 516                 SLIRP_POLL_IN | SLIRP_POLL_HUP | SLIRP_POLL_ERR, opaque);
 517         }
 518     }
 519
 520     slirp_update_timeout(slirp, timeout);
 521 }
 522
 523 void slirp_pollfds_poll(Slirp *slirp, int select_error,
 524                         SlirpGetREventsCb get_revents, void *opaque)
 525 {
 526     struct socket *so, *so_next;
 527     int ret;
 528
 529     curtime = slirp->cb->clock_get_ns(slirp->opaque) / SCALE_MS;
 530
 531     /*
 532      * See if anything has timed out
 533      */
 534     if (slirp->time_fasttimo &&
 535         ((curtime - slirp->time_fasttimo) >= TIMEOUT_FAST)) {
 536         tcp_fasttimo(slirp);
 537         slirp->time_fasttimo = 0;
 538     }
 539     if (slirp->do_slowtimo &&
 540         ((curtime - slirp->last_slowtimo) >= TIMEOUT_SLOW)) {
 541         ip_slowtimo(slirp);
 542         tcp_slowtimo(slirp);
 543         slirp->last_slowtimo = curtime;
 544     }
 545
 546     /*
 547      * Check sockets
 548      */
 549     if (!select_error) {
 550         /*
 551          * Check TCP sockets
 552          */
 553         for (so = slirp->tcb.so_next; so != &slirp->tcb;
 554              so = so_next) {
 555             int revents;
 556
 557             so_next = so->so_next;
 558
 559             revents = 0;
 560             if (so->pollfds_idx != -1) {
 561                 revents = get_revents(so->pollfds_idx, opaque);
 562             }
 563
 564             if (so->so_state & SS_NOFDREF || so->s == -1) {
 565                 continue;
 566             }
 567
 568             /*
 569              * Check for URG data
 570              * This will soread as well, so no need to
 571              * test for SLIRP_POLL_IN below if this succeeds
 572              */
 573             if (revents & SLIRP_POLL_PRI) {
 574                 ret = sorecvoob(so);
 575                 if (ret < 0) {
 576                     /* Socket error might have resulted in the socket being
 577                      * removed, do not try to do anything more with it. */
 578                     continue;
 579                 }
 580             }
 581             /*
 582              * Check sockets for reading
 583              */
 584             else if (revents &
 585                      (SLIRP_POLL_IN | SLIRP_POLL_HUP | SLIRP_POLL_ERR)) {
 586                 /*
 587                  * Check for incoming connections
 588                  */
 589                 if (so->so_state & SS_FACCEPTCONN) {
 590                     tcp_connect(so);
 591                     continue;
 592                 } /* else */
 593                 ret = soread(so);
 594
 595                 /* Output it if we read something */
 596                 if (ret > 0) {
 597                     tcp_output(sototcpcb(so));
 598                 }
 599                 if (ret < 0) {
 600                     /* Socket error might have resulted in the socket being
 601                      * removed, do not try to do anything more with it. */
 602                     continue;
 603                 }
 604             }
 605
 606             /*
 607              * Check sockets for writing
 608              */
 609             if (!(so->so_state & SS_NOFDREF) &&
 610                 (revents & (SLIRP_POLL_OUT | SLIRP_POLL_ERR))) {
 611                 /*
 612                  * Check for non-blocking, still-connecting sockets
 613                  */
 614                 if (so->so_state & SS_ISFCONNECTING) {
 615                     /* Connected */
 616                     so->so_state &= ~SS_ISFCONNECTING;
 617
 618                     ret = send(so->s, (const void *) &ret, 0, 0);
 619                     if (ret < 0) {
 620                         /* XXXXX Must fix, zero bytes is a NOP */
 621                         if (errno == EAGAIN || errno == EWOULDBLOCK ||
 622                             errno == EINPROGRESS || errno == ENOTCONN) {
 623                             continue;
 624                         }
 625
 626                         /* else failed */
 627                         so->so_state &= SS_PERSISTENT_MASK;
 628                         so->so_state |= SS_NOFDREF;
 629                     }
 630                     /* else so->so_state &= ~SS_ISFCONNECTING; */
 631
 632                     /*
 633                      * Continue tcp_input
 634                      */
 635                     tcp_input((struct mbuf *)NULL, sizeof(struct ip), so,
 636                               so->so_ffamily);
 637                     /* continue; */
 638                 } else {
 639                     ret = sowrite(so);
 640                     if (ret > 0) {
 641                         /* Call tcp_output in case we need to send a window
 642                          * update to the guest, otherwise it will be stuck
 643                          * until it sends a window probe. */
 644                         tcp_output(sototcpcb(so));
 645                     }
 646                 }
 647             }
 648         }
 649
 650         /*
 651          * Now UDP sockets.
 652          * Incoming packets are sent straight away, they're not buffered.
 653          * Incoming UDP data isn't buffered either.
 654          */
 655         for (so = slirp->udb.so_next; so != &slirp->udb;
 656              so = so_next) {
 657             int revents;
 658
 659             so_next = so->so_next;
 660
 661             revents = 0;
 662             if (so->pollfds_idx != -1) {
 663                 revents = get_revents(so->pollfds_idx, opaque);
 664             }
 665
 666             if (so->s != -1 &&
 667                 (revents & (SLIRP_POLL_IN | SLIRP_POLL_HUP | SLIRP_POLL_ERR))) {
 668                 sorecvfrom(so);
 669             }
 670         }
 671
 672         /*
 673          * Check incoming ICMP relies.
 674          */
 675         for (so = slirp->icmp.so_next; so != &slirp->icmp;
 676              so = so_next) {
 677             int revents;
 678
 679             so_next = so->so_next;
 680
 681             revents = 0;
 682             if (so->pollfds_idx != -1) {
 683                 revents = get_revents(so->pollfds_idx, opaque);
 684             }
 685
 686             if (so->s != -1 &&
 687                 (revents & (SLIRP_POLL_IN | SLIRP_POLL_HUP | SLIRP_POLL_ERR))) {
 688                 icmp_receive(so);
 689             }
 690         }
 691     }
 692
 693     if_start(slirp);
 694 }
 695
 696 static void arp_input(Slirp *slirp, const uint8_t *pkt, int pkt_len)
 697 {
 698     struct slirp_arphdr *ah = (struct slirp_arphdr *)(pkt + ETH_HLEN);
 699     uint8_t arp_reply[MAX(ETH_HLEN + sizeof(struct slirp_arphdr), 64)];
 700     struct ethhdr *reh = (struct ethhdr *)arp_reply;
 701     struct slirp_arphdr *rah = (struct slirp_arphdr *)(arp_reply + ETH_HLEN);
 702     int ar_op;
 703     struct gfwd_list *ex_ptr;
 704
 705     if (!slirp->in_enabled) {
 706         return;
 707     }
 708
 709     ar_op = ntohs(ah->ar_op);
 710     switch(ar_op) {
 711     case ARPOP_REQUEST:
 712         if (ah->ar_tip == ah->ar_sip) {
 713             /* Gratuitous ARP */
 714             arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
 715             return;
 716         }
 717
 718         if ((ah->ar_tip & slirp->vnetwork_mask.s_addr) ==
 719             slirp->vnetwork_addr.s_addr) {
 720             if (ah->ar_tip == slirp->vnameserver_addr.s_addr ||
 721                 ah->ar_tip == slirp->vhost_addr.s_addr)
 722                 goto arp_ok;
 723             /* TODO: IPv6 */
 724             for (ex_ptr = slirp->guestfwd_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
 725                 if (ex_ptr->ex_addr.s_addr == ah->ar_tip)
 726                     goto arp_ok;
 727             }
 728             return;
 729         arp_ok:
 730             memset(arp_reply, 0, sizeof(arp_reply));
 731
 732             arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
 733
 734             /* ARP request for alias/dns mac address */
 735             memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
 736             memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 4);
 737             memcpy(&reh->h_source[2], &ah->ar_tip, 4);
 738             reh->h_proto = htons(ETH_P_ARP);
 739
 740             rah->ar_hrd = htons(1);
 741             rah->ar_pro = htons(ETH_P_IP);
 742             rah->ar_hln = ETH_ALEN;
 743             rah->ar_pln = 4;
 744             rah->ar_op = htons(ARPOP_REPLY);
 745             memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
 746             rah->ar_sip = ah->ar_tip;
 747             memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
 748             rah->ar_tip = ah->ar_sip;
 749             slirp_send_packet_all(slirp, arp_reply, sizeof(arp_reply));
 750         }
 751         break;
 752     case ARPOP_REPLY:
 753         arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
 754         break;
 755     default:
 756         break;
 757     }
 758 }
 759
 760 void slirp_input(Slirp *slirp, const uint8_t *pkt, int pkt_len)
 761 {
 762     struct mbuf *m;
 763     int proto;
 764
 765     if (pkt_len < ETH_HLEN)
 766         return;
 767
 768     proto = (((uint16_t) pkt[12]) << 8) + pkt[13];
 769     switch(proto) {
 770     case ETH_P_ARP:
 771         arp_input(slirp, pkt, pkt_len);
 772         break;
 773     case ETH_P_IP:
 774     case ETH_P_IPV6:
 775         m = m_get(slirp);
 776         if (!m)
 777             return;
 778         /* Note: we add 2 to align the IP header on 4 bytes,
 779          * and add the margin for the tcpiphdr overhead  */
 780         if (M_FREEROOM(m) < pkt_len + TCPIPHDR_DELTA + 2) {
 781             m_inc(m, pkt_len + TCPIPHDR_DELTA + 2);
 782         }
 783         m->m_len = pkt_len + TCPIPHDR_DELTA + 2;
 784         memcpy(m->m_data + TCPIPHDR_DELTA + 2, pkt, pkt_len);
 785
 786         m->m_data += TCPIPHDR_DELTA + 2 + ETH_HLEN;
 787         m->m_len -= TCPIPHDR_DELTA + 2 + ETH_HLEN;
 788
 789         if (proto == ETH_P_IP) {
 790             ip_input(m);
 791         } else if (proto == ETH_P_IPV6) {
 792             ip6_input(m);
 793         }
 794         break;
 795
 796     case ETH_P_NCSI:
 797         ncsi_input(slirp, pkt, pkt_len);
 798         break;
 799
 800     default:
 801         break;
 802     }
 803 }
 804
 805 /* Prepare the IPv4 packet to be sent to the ethernet device. Returns 1 if no
 806  * packet should be sent, 0 if the packet must be re-queued, 2 if the packet
 807  * is ready to go.
 808  */
 809 static int if_encap4(Slirp *slirp, struct mbuf *ifm, struct ethhdr *eh,
 810         uint8_t ethaddr[ETH_ALEN])
 811 {
 812     const struct ip *iph = (const struct ip *)ifm->m_data;
 813
 814     if (iph->ip_dst.s_addr == 0) {
 815         /* 0.0.0.0 can not be a destination address, something went wrong,
 816          * avoid making it worse */
 817         return 1;
 818     }
 819     if (!arp_table_search(slirp, iph->ip_dst.s_addr, ethaddr)) {
 820         uint8_t arp_req[ETH_HLEN + sizeof(struct slirp_arphdr)];
 821         struct ethhdr *reh = (struct ethhdr *)arp_req;
 822         struct slirp_arphdr *rah = (struct slirp_arphdr *)(arp_req + ETH_HLEN);
 823
 824         if (!ifm->resolution_requested) {
 825             /* If the client addr is not known, send an ARP request */
 826             memset(reh->h_dest, 0xff, ETH_ALEN);
 827             memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 4);
 828             memcpy(&reh->h_source[2], &slirp->vhost_addr, 4);
 829             reh->h_proto = htons(ETH_P_ARP);
 830             rah->ar_hrd = htons(1);
 831             rah->ar_pro = htons(ETH_P_IP);
 832             rah->ar_hln = ETH_ALEN;
 833             rah->ar_pln = 4;
 834             rah->ar_op = htons(ARPOP_REQUEST);
 835
 836             /* source hw addr */
 837             memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN - 4);
 838             memcpy(&rah->ar_sha[2], &slirp->vhost_addr, 4);
 839
 840             /* source IP */
 841             rah->ar_sip = slirp->vhost_addr.s_addr;
 842
 843             /* target hw addr (none) */
 844             memset(rah->ar_tha, 0, ETH_ALEN);
 845
 846             /* target IP */
 847             rah->ar_tip = iph->ip_dst.s_addr;
 848             slirp->client_ipaddr = iph->ip_dst;
 849             slirp_send_packet_all(slirp, arp_req, sizeof(arp_req));
 850             ifm->resolution_requested = true;
 851
 852             /* Expire request and drop outgoing packet after 1 second */
 853             ifm->expiration_date =
 854                 slirp->cb->clock_get_ns(slirp->opaque) + 1000000000ULL;
 855         }
 856         return 0;
 857     } else {
 858         memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 4);
 859         /* XXX: not correct */
 860         memcpy(&eh->h_source[2], &slirp->vhost_addr, 4);
 861         eh->h_proto = htons(ETH_P_IP);
 862
 863         /* Send this */
 864         return 2;
 865     }
 866 }
 867
 868 /* Prepare the IPv6 packet to be sent to the ethernet device. Returns 1 if no
 869  * packet should be sent, 0 if the packet must be re-queued, 2 if the packet
 870  * is ready to go.
 871  */
 872 static int if_encap6(Slirp *slirp, struct mbuf *ifm, struct ethhdr *eh,
 873         uint8_t ethaddr[ETH_ALEN])
 874 {
 875     const struct ip6 *ip6h = mtod(ifm, const struct ip6 *);
 876     if (!ndp_table_search(slirp, ip6h->ip_dst, ethaddr)) {
 877         if (!ifm->resolution_requested) {
 878             ndp_send_ns(slirp, ip6h->ip_dst);
 879             ifm->resolution_requested = true;
 880             ifm->expiration_date = slirp->cb->clock_get_ns(slirp->opaque) + 1000000000ULL;
 881         }
 882         return 0;
 883     } else {
 884         eh->h_proto = htons(ETH_P_IPV6);
 885         in6_compute_ethaddr(ip6h->ip_src, eh->h_source);
 886
 887         /* Send this */
 888         return 2;
 889     }
 890 }
 891
 892 /* Output the IP packet to the ethernet device. Returns 0 if the packet must be
 893  * re-queued.
 894  */
 895 int if_encap(Slirp *slirp, struct mbuf *ifm)
 896 {
 897     uint8_t buf[1600];
 898     struct ethhdr *eh = (struct ethhdr *)buf;
 899     uint8_t ethaddr[ETH_ALEN];
 900     const struct ip *iph = (const struct ip *)ifm->m_data;
 901     int ret;
 902
 903     if (ifm->m_len + ETH_HLEN > sizeof(buf)) {
 904         return 1;
 905     }
 906
 907     switch (iph->ip_v) {
 908     case IPVERSION:
 909         ret = if_encap4(slirp, ifm, eh, ethaddr);
 910         if (ret < 2) {
 911             return ret;
 912         }
 913         break;
 914
 915     case IP6VERSION:
 916         ret = if_encap6(slirp, ifm, eh, ethaddr);
 917         if (ret < 2) {
 918             return ret;
 919         }
 920         break;
 921
 922     default:
 923         g_assert_not_reached();
 924         break;
 925     }
 926
 927     memcpy(eh->h_dest, ethaddr, ETH_ALEN);
 928     DEBUG_ARG("src = %02x:%02x:%02x:%02x:%02x:%02x",
 929               eh->h_source[0], eh->h_source[1], eh->h_source[2],
 930               eh->h_source[3], eh->h_source[4], eh->h_source[5]);
 931     DEBUG_ARG("dst = %02x:%02x:%02x:%02x:%02x:%02x",
 932               eh->h_dest[0], eh->h_dest[1], eh->h_dest[2],
 933               eh->h_dest[3], eh->h_dest[4], eh->h_dest[5]);
 934     memcpy(buf + sizeof(struct ethhdr), ifm->m_data, ifm->m_len);
 935     slirp_send_packet_all(slirp, buf, ifm->m_len + ETH_HLEN);
 936     return 1;
 937 }
 938
 939 /* Drop host forwarding rule, return 0 if found. */
 940 /* TODO: IPv6 */
 941 int slirp_remove_hostfwd(Slirp *slirp, int is_udp, struct in_addr host_addr,
 942                          int host_port)
 943 {
 944     struct socket *so;
 945     struct socket *head = (is_udp ? &slirp->udb : &slirp->tcb);
 946     struct sockaddr_in addr;
 947     int port = htons(host_port);
 948     socklen_t addr_len;
 949
 950     for (so = head->so_next; so != head; so = so->so_next) {
 951         addr_len = sizeof(addr);
 952         if ((so->so_state & SS_HOSTFWD) &&
 953             getsockname(so->s, (struct sockaddr *)&addr, &addr_len) == 0 &&
 954             addr.sin_addr.s_addr == host_addr.s_addr &&
 955             addr.sin_port == port) {
 956             so->slirp->cb->unregister_poll_fd(so->s, so->slirp->opaque);
 957             closesocket(so->s);
 958             sofree(so);
 959             return 0;
 960         }
 961     }
 962
 963     return -1;
 964 }
 965
 966 /* TODO: IPv6 */
 967 int slirp_add_hostfwd(Slirp *slirp, int is_udp, struct in_addr host_addr,
 968                       int host_port, struct in_addr guest_addr, int guest_port)
 969 {
 970     if (!guest_addr.s_addr) {
 971         guest_addr = slirp->vdhcp_startaddr;
 972     }
 973     if (is_udp) {
 974         if (!udp_listen(slirp, host_addr.s_addr, htons(host_port),
 975                         guest_addr.s_addr, htons(guest_port), SS_HOSTFWD))
 976             return -1;
 977     } else {
 978         if (!tcp_listen(slirp, host_addr.s_addr, htons(host_port),
 979                         guest_addr.s_addr, htons(guest_port), SS_HOSTFWD))
 980             return -1;
 981     }
 982     return 0;
 983 }
 984
 985 /* TODO: IPv6 */
 986 static bool
 987 check_guestfwd(Slirp *slirp, struct in_addr *guest_addr, int guest_port)
 988 {
 989     struct gfwd_list *tmp_ptr;
 990
 991     if (!guest_addr->s_addr) {
 992         guest_addr->s_addr = slirp->vnetwork_addr.s_addr |
 993             (htonl(0x0204) & ~slirp->vnetwork_mask.s_addr);
 994     }
 995     if ((guest_addr->s_addr & slirp->vnetwork_mask.s_addr) !=
 996         slirp->vnetwork_addr.s_addr ||
 997         guest_addr->s_addr == slirp->vhost_addr.s_addr ||
 998         guest_addr->s_addr == slirp->vnameserver_addr.s_addr) {
 999         return false;
1000     }
1001
1002     /* check if the port is "bound" */
1003     for (tmp_ptr = slirp->guestfwd_list; tmp_ptr; tmp_ptr = tmp_ptr->ex_next) {
1004         if (guest_port == tmp_ptr->ex_fport &&
1005             guest_addr->s_addr == tmp_ptr->ex_addr.s_addr)
1006             return false;
1007     }
1008
1009     return true;
1010 }
1011
1012 int slirp_add_exec(Slirp *slirp, const char *cmdline,
1013                    struct in_addr *guest_addr, int guest_port)
1014 {
1015     if (!check_guestfwd(slirp, guest_addr, guest_port)) {
1016         return -1;
1017     }
1018
1019     add_exec(&slirp->guestfwd_list, cmdline, *guest_addr, htons(guest_port));
1020     return 0;
1021 }
1022
1023 int slirp_add_guestfwd(Slirp *slirp, SlirpWriteCb write_cb, void *opaque,
1024                        struct in_addr *guest_addr, int guest_port)
1025 {
1026     if (!check_guestfwd(slirp, guest_addr, guest_port)) {
1027         return -1;
1028     }
1029
1030     add_guestfwd(&slirp->guestfwd_list, write_cb, opaque,
1031                  *guest_addr, htons(guest_port));
1032     return 0;
1033 }
1034
1035 ssize_t slirp_send(struct socket *so, const void *buf, size_t len, int flags)
1036 {
1037     if (so->s == -1 && so->guestfwd) {
1038         /* XXX this blocks entire thread. Rewrite to use
1039          * qemu_chr_fe_write and background I/O callbacks */
1040         so->guestfwd->write_cb(buf, len, so->guestfwd->opaque);
1041         return len;
1042     }
1043
1044     if (so->s == -1) {
1045         /*
1046          * This should in theory not happen but it is hard to be
1047          * sure because some code paths will end up with so->s == -1
1048          * on a failure but don't dispose of the struct socket.
1049          * Check specifically, so we don't pass -1 to send().
1050          */
1051         errno = EBADF;
1052         return -1;
1053     }
1054
1055     return send(so->s, buf, len, flags);
1056 }
1057
1058 struct socket *
1059 slirp_find_ctl_socket(Slirp *slirp, struct in_addr guest_addr, int guest_port)
1060 {
1061     struct socket *so;
1062
1063     /* TODO: IPv6 */
1064     for (so = slirp->tcb.so_next; so != &slirp->tcb; so = so->so_next) {
1065         if (so->so_faddr.s_addr == guest_addr.s_addr &&
1066             htons(so->so_fport) == guest_port) {
1067             return so;
1068         }
1069     }
1070     return NULL;
1071 }
1072
1073 size_t slirp_socket_can_recv(Slirp *slirp, struct in_addr guest_addr,
1074                              int guest_port)
1075 {
1076     struct iovec iov[2];
1077     struct socket *so;
1078
1079     so = slirp_find_ctl_socket(slirp, guest_addr, guest_port);
1080
1081     if (!so || so->so_state & SS_NOFDREF) {
1082         return 0;
1083     }
1084
1085     if (!CONN_CANFRCV(so) || so->so_snd.sb_cc >= (so->so_snd.sb_datalen/2)) {
1086         return 0;
1087     }
1088
1089     return sopreprbuf(so, iov, NULL);
1090 }
1091
1092 void slirp_socket_recv(Slirp *slirp, struct in_addr guest_addr, int guest_port,
1093                        const uint8_t *buf, int size)
1094 {
1095     int ret;
1096     struct socket *so = slirp_find_ctl_socket(slirp, guest_addr, guest_port);
1097
1098     if (!so)
1099         return;
1100
1101     ret = soreadbuf(so, (const char *)buf, size);
1102
1103     if (ret > 0)
1104         tcp_output(sototcpcb(so));
1105 }
1106
1107 void slirp_send_packet_all(Slirp *slirp, const void *buf, size_t len)
1108 {
1109     ssize_t ret = slirp->cb->send_packet(buf, len, slirp->opaque);
1110
1111     if (ret < 0) {
1112         g_critical("Failed to send packet, ret: %ld", (long) ret);
1113     } else if (ret < len) {
1114         DEBUG_ERROR("send_packet() didn't send all data: %ld < %lu",
1115                 (long) ret, (unsigned long) len);
1116     }
1117 }