slirp/src/slirp.c

   1 /* SPDX-License-Identifier: MIT */
   2 /*
   3  * libslirp glue
   4  *
   5  * Copyright (c) 2004-2008 Fabrice Bellard
   6  *
   7  * Permission is hereby granted, free of charge, to any person obtaining a copy
   8  * of this software and associated documentation files (the "Software"), to deal
   9  * in the Software without restriction, including without limitation the rights
  10  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  11  * copies of the Software, and to permit persons to whom the Software is
  12  * furnished to do so, subject to the following conditions:
  13  *
  14  * The above copyright notice and this permission notice shall be included in
  15  * all copies or substantial portions of the Software.
  16  *
  17  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  18  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  19  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  20  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  21  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  22  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  23  * THE SOFTWARE.
  24  */
  25 #include "slirp.h"
  26
  27
  28 #ifndef _WIN32
  29 #include <net/if.h>
  30 #endif
  31
  32 int slirp_debug;
  33
  34 /* Define to 1 if you want KEEPALIVE timers */
  35 bool slirp_do_keepalive;
  36
  37 /* host loopback address */
  38 struct in_addr loopback_addr;
  39 /* host loopback network mask */
  40 unsigned long loopback_mask;
  41
  42 /* emulated hosts use the MAC addr 52:55:IP:IP:IP:IP */
  43 static const uint8_t special_ethaddr[ETH_ALEN] = {
  44     0x52, 0x55, 0x00, 0x00, 0x00, 0x00
  45 };
  46
  47 unsigned curtime;
  48
  49 static struct in_addr dns_addr;
  50 #ifndef _WIN32
  51 static struct in6_addr dns6_addr;
  52 #endif
  53 static unsigned dns_addr_time;
  54 #ifndef _WIN32
  55 static unsigned dns6_addr_time;
  56 #endif
  57
  58 #define TIMEOUT_FAST 2  /* milliseconds */
  59 #define TIMEOUT_SLOW 499  /* milliseconds */
  60 /* for the aging of certain requests like DNS */
  61 #define TIMEOUT_DEFAULT 1000  /* milliseconds */
  62
  63 #ifdef _WIN32
  64
  65 int get_dns_addr(struct in_addr *pdns_addr)
  66 {
  67     FIXED_INFO *FixedInfo=NULL;
  68     ULONG    BufLen;
  69     DWORD    ret;
  70     IP_ADDR_STRING *pIPAddr;
  71     struct in_addr tmp_addr;
  72
  73     if (dns_addr.s_addr != 0 && (curtime - dns_addr_time) < TIMEOUT_DEFAULT) {
  74         *pdns_addr = dns_addr;
  75         return 0;
  76     }
  77
  78     FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
  79     BufLen = sizeof(FIXED_INFO);
  80
  81     if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen)) {
  82         if (FixedInfo) {
  83             GlobalFree(FixedInfo);
  84             FixedInfo = NULL;
  85         }
  86         FixedInfo = GlobalAlloc(GPTR, BufLen);
  87     }
  88
  89     if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS) {
  90         printf("GetNetworkParams failed. ret = %08x\n", (unsigned)ret );
  91         if (FixedInfo) {
  92             GlobalFree(FixedInfo);
  93             FixedInfo = NULL;
  94         }
  95         return -1;
  96     }
  97
  98     pIPAddr = &(FixedInfo->DnsServerList);
  99     inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
 100     *pdns_addr = tmp_addr;
 101     dns_addr = tmp_addr;
 102     dns_addr_time = curtime;
 103     if (FixedInfo) {
 104         GlobalFree(FixedInfo);
 105         FixedInfo = NULL;
 106     }
 107     return 0;
 108 }
 109
 110 int get_dns6_addr(struct in6_addr *pdns6_addr, uint32_t *scope_id)
 111 {
 112     return -1;
 113 }
 114
 115 static void winsock_cleanup(void)
 116 {
 117     WSACleanup();
 118 }
 119
 120 #else
 121
 122 static int get_dns_addr_cached(void *pdns_addr, void *cached_addr,
 123                                socklen_t addrlen,
 124                                struct stat *cached_stat, unsigned *cached_time)
 125 {
 126     struct stat old_stat;
 127     if (curtime - *cached_time < TIMEOUT_DEFAULT) {
 128         memcpy(pdns_addr, cached_addr, addrlen);
 129         return 0;
 130     }
 131     old_stat = *cached_stat;
 132     if (stat("/etc/resolv.conf", cached_stat) != 0) {
 133         return -1;
 134     }
 135     if (cached_stat->st_dev == old_stat.st_dev
 136         && cached_stat->st_ino == old_stat.st_ino
 137         && cached_stat->st_size == old_stat.st_size
 138         && cached_stat->st_mtime == old_stat.st_mtime) {
 139         memcpy(pdns_addr, cached_addr, addrlen);
 140         return 0;
 141     }
 142     return 1;
 143 }
 144
 145 static int get_dns_addr_resolv_conf(int af, void *pdns_addr, void *cached_addr,
 146                                     socklen_t addrlen, uint32_t *scope_id,
 147                                     unsigned *cached_time)
 148 {
 149     char buff[512];
 150     char buff2[257];
 151     FILE *f;
 152     int found = 0;
 153     void *tmp_addr = alloca(addrlen);
 154     unsigned if_index;
 155
 156     f = fopen("/etc/resolv.conf", "r");
 157     if (!f)
 158         return -1;
 159
 160     DEBUG_MISC("IP address of your DNS(s):");
 161     while (fgets(buff, 512, f) != NULL) {
 162         if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1) {
 163             char *c = strchr(buff2, '%');
 164             if (c) {
 165                 if_index = if_nametoindex(c + 1);
 166                 *c = '\0';
 167             } else {
 168                 if_index = 0;
 169             }
 170
 171             if (!inet_pton(af, buff2, tmp_addr)) {
 172                 continue;
 173             }
 174             /* If it's the first one, set it to dns_addr */
 175             if (!found) {
 176                 memcpy(pdns_addr, tmp_addr, addrlen);
 177                 memcpy(cached_addr, tmp_addr, addrlen);
 178                 if (scope_id) {
 179                     *scope_id = if_index;
 180                 }
 181                 *cached_time = curtime;
 182             }
 183
 184             if (++found > 3) {
 185                 DEBUG_MISC("  (more)");
 186                 break;
 187             } else if (slirp_debug & DBG_MISC) {
 188                 char s[INET6_ADDRSTRLEN];
 189                 const char *res = inet_ntop(af, tmp_addr, s, sizeof(s));
 190                 if (!res) {
 191                     res = "  (string conversion error)";
 192                 }
 193                 DEBUG_MISC("  %s", res);
 194             }
 195         }
 196     }
 197     fclose(f);
 198     if (!found)
 199         return -1;
 200     return 0;
 201 }
 202
 203 int get_dns_addr(struct in_addr *pdns_addr)
 204 {
 205     static struct stat dns_addr_stat;
 206
 207     if (dns_addr.s_addr != 0) {
 208         int ret;
 209         ret = get_dns_addr_cached(pdns_addr, &dns_addr, sizeof(dns_addr),
 210                                   &dns_addr_stat, &dns_addr_time);
 211         if (ret <= 0) {
 212             return ret;
 213         }
 214     }
 215     return get_dns_addr_resolv_conf(AF_INET, pdns_addr, &dns_addr,
 216                                     sizeof(dns_addr), NULL, &dns_addr_time);
 217 }
 218
 219 int get_dns6_addr(struct in6_addr *pdns6_addr, uint32_t *scope_id)
 220 {
 221     static struct stat dns6_addr_stat;
 222
 223     if (!in6_zero(&dns6_addr)) {
 224         int ret;
 225         ret = get_dns_addr_cached(pdns6_addr, &dns6_addr, sizeof(dns6_addr),
 226                                   &dns6_addr_stat, &dns6_addr_time);
 227         if (ret <= 0) {
 228             return ret;
 229         }
 230     }
 231     return get_dns_addr_resolv_conf(AF_INET6, pdns6_addr, &dns6_addr,
 232                                     sizeof(dns6_addr),
 233                                     scope_id, &dns6_addr_time);
 234 }
 235
 236 #endif
 237
 238 static void slirp_init_once(void)
 239 {
 240     static int initialized;
 241     const char *debug;
 242 #ifdef _WIN32
 243     WSADATA Data;
 244 #endif
 245
 246     if (initialized) {
 247         return;
 248     }
 249     initialized = 1;
 250
 251 #ifdef _WIN32
 252     WSAStartup(MAKEWORD(2,0), &Data);
 253     atexit(winsock_cleanup);
 254 #endif
 255
 256     loopback_addr.s_addr = htonl(INADDR_LOOPBACK);
 257     loopback_mask = htonl(IN_CLASSA_NET);
 258
 259     debug = g_getenv("SLIRP_DEBUG");
 260     if (debug) {
 261         const GDebugKey keys[] = {
 262             { "call", DBG_CALL },
 263             { "misc", DBG_MISC },
 264             { "error", DBG_ERROR },
 265             { "tftp", DBG_TFTP },
 266         };
 267         slirp_debug = g_parse_debug_string(debug, keys, G_N_ELEMENTS(keys));
 268     }
 269
 270
 271 }
 272
 273 Slirp *slirp_init(int restricted, bool in_enabled, struct in_addr vnetwork,
 274                   struct in_addr vnetmask, struct in_addr vhost,
 275                   bool in6_enabled,
 276                   struct in6_addr vprefix_addr6, uint8_t vprefix_len,
 277                   struct in6_addr vhost6, const char *vhostname,
 278                   const char *tftp_server_name,
 279                   const char *tftp_path, const char *bootfile,
 280                   struct in_addr vdhcp_start, struct in_addr vnameserver,
 281                   struct in6_addr vnameserver6, const char **vdnssearch,
 282                   const char *vdomainname,
 283                   const SlirpCb *callbacks,
 284                   void *opaque)
 285 {
 286     Slirp *slirp = g_malloc0(sizeof(Slirp));
 287
 288     slirp_init_once();
 289
 290     slirp->opaque = opaque;
 291     slirp->cb = callbacks;
 292     slirp->grand = g_rand_new();
 293     slirp->restricted = restricted;
 294
 295     slirp->in_enabled = in_enabled;
 296     slirp->in6_enabled = in6_enabled;
 297
 298     if_init(slirp);
 299     ip_init(slirp);
 300     ip6_init(slirp);
 301
 302     /* Initialise mbufs *after* setting the MTU */
 303     m_init(slirp);
 304
 305     slirp->vnetwork_addr = vnetwork;
 306     slirp->vnetwork_mask = vnetmask;
 307     slirp->vhost_addr = vhost;
 308     slirp->vprefix_addr6 = vprefix_addr6;
 309     slirp->vprefix_len = vprefix_len;
 310     slirp->vhost_addr6 = vhost6;
 311     if (vhostname) {
 312         slirp_pstrcpy(slirp->client_hostname, sizeof(slirp->client_hostname),
 313                       vhostname);
 314     }
 315     slirp->tftp_prefix = g_strdup(tftp_path);
 316     slirp->bootp_filename = g_strdup(bootfile);
 317     slirp->vdomainname = g_strdup(vdomainname);
 318     slirp->vdhcp_startaddr = vdhcp_start;
 319     slirp->vnameserver_addr = vnameserver;
 320     slirp->vnameserver_addr6 = vnameserver6;
 321     slirp->tftp_server_name = g_strdup(tftp_server_name);
 322
 323     if (vdnssearch) {
 324         translate_dnssearch(slirp, vdnssearch);
 325     }
 326
 327     return slirp;
 328 }
 329
 330 void slirp_cleanup(Slirp *slirp)
 331 {
 332     struct gfwd_list *e, *next;
 333
 334     for (e = slirp->guestfwd_list; e; e = next) {
 335         next = e->ex_next;
 336         g_free(e->ex_exec);
 337         g_free(e);
 338     }
 339
 340     ip_cleanup(slirp);
 341     ip6_cleanup(slirp);
 342     m_cleanup(slirp);
 343
 344     g_rand_free(slirp->grand);
 345
 346     g_free(slirp->vdnssearch);
 347     g_free(slirp->tftp_prefix);
 348     g_free(slirp->bootp_filename);
 349     g_free(slirp->vdomainname);
 350     g_free(slirp);
 351 }
 352
 353 #define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
 354 #define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
 355
 356 static void slirp_update_timeout(Slirp *slirp, uint32_t *timeout)
 357 {
 358     uint32_t t;
 359
 360     if (*timeout <= TIMEOUT_FAST) {
 361         return;
 362     }
 363
 364     t = MIN(1000, *timeout);
 365
 366     /* If we have tcp timeout with slirp, then we will fill @timeout with
 367      * more precise value.
 368      */
 369     if (slirp->time_fasttimo) {
 370         *timeout = TIMEOUT_FAST;
 371         return;
 372     }
 373     if (slirp->do_slowtimo) {
 374         t = MIN(TIMEOUT_SLOW, t);
 375     }
 376     *timeout = t;
 377 }
 378
 379 void slirp_pollfds_fill(Slirp *slirp, uint32_t *timeout,
 380                         SlirpAddPollCb add_poll, void *opaque)
 381 {
 382     struct socket *so, *so_next;
 383
 384     /*
 385      * First, TCP sockets
 386      */
 387
 388     /*
 389      * *_slowtimo needs calling if there are IP fragments
 390      * in the fragment queue, or there are TCP connections active
 391      */
 392     slirp->do_slowtimo = ((slirp->tcb.so_next != &slirp->tcb) ||
 393                           (&slirp->ipq.ip_link != slirp->ipq.ip_link.next));
 394
 395     for (so = slirp->tcb.so_next; so != &slirp->tcb; so = so_next) {
 396         int events = 0;
 397
 398         so_next = so->so_next;
 399
 400         so->pollfds_idx = -1;
 401
 402         /*
 403          * See if we need a tcp_fasttimo
 404          */
 405         if (slirp->time_fasttimo == 0 &&
 406             so->so_tcpcb->t_flags & TF_DELACK) {
 407             slirp->time_fasttimo = curtime; /* Flag when want a fasttimo */
 408         }
 409
 410         /*
 411          * NOFDREF can include still connecting to local-host,
 412          * newly socreated() sockets etc. Don't want to select these.
 413          */
 414         if (so->so_state & SS_NOFDREF || so->s == -1) {
 415             continue;
 416         }
 417
 418         /*
 419          * Set for reading sockets which are accepting
 420          */
 421         if (so->so_state & SS_FACCEPTCONN) {
 422             so->pollfds_idx = add_poll(so->s,
 423                 SLIRP_POLL_IN | SLIRP_POLL_HUP | SLIRP_POLL_ERR, opaque);
 424             continue;
 425         }
 426
 427         /*
 428          * Set for writing sockets which are connecting
 429          */
 430         if (so->so_state & SS_ISFCONNECTING) {
 431             so->pollfds_idx = add_poll(so->s,
 432                 SLIRP_POLL_OUT | SLIRP_POLL_ERR, opaque);
 433             continue;
 434         }
 435
 436         /*
 437          * Set for writing if we are connected, can send more, and
 438          * we have something to send
 439          */
 440         if (CONN_CANFSEND(so) && so->so_rcv.sb_cc) {
 441             events |= SLIRP_POLL_OUT | SLIRP_POLL_ERR;
 442         }
 443
 444         /*
 445          * Set for reading (and urgent data) if we are connected, can
 446          * receive more, and we have room for it XXX /2 ?
 447          */
 448         if (CONN_CANFRCV(so) &&
 449             (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2))) {
 450             events |= SLIRP_POLL_IN | SLIRP_POLL_HUP |
 451                       SLIRP_POLL_ERR | SLIRP_POLL_PRI;
 452         }
 453
 454         if (events) {
 455             so->pollfds_idx = add_poll(so->s, events, opaque);
 456         }
 457     }
 458
 459     /*
 460      * UDP sockets
 461      */
 462     for (so = slirp->udb.so_next; so != &slirp->udb; so = so_next) {
 463         so_next = so->so_next;
 464
 465         so->pollfds_idx = -1;
 466
 467         /*
 468          * See if it's timed out
 469          */
 470         if (so->so_expire) {
 471             if (so->so_expire <= curtime) {
 472                 udp_detach(so);
 473                 continue;
 474             } else {
 475                 slirp->do_slowtimo = true; /* Let socket expire */
 476             }
 477         }
 478
 479         /*
 480          * When UDP packets are received from over the
 481          * link, they're sendto()'d straight away, so
 482          * no need for setting for writing
 483          * Limit the number of packets queued by this session
 484          * to 4.  Note that even though we try and limit this
 485          * to 4 packets, the session could have more queued
 486          * if the packets needed to be fragmented
 487          * (XXX <= 4 ?)
 488          */
 489         if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4) {
 490             so->pollfds_idx = add_poll(so->s,
 491                 SLIRP_POLL_IN | SLIRP_POLL_HUP | SLIRP_POLL_ERR, opaque);
 492         }
 493     }
 494
 495     /*
 496      * ICMP sockets
 497      */
 498     for (so = slirp->icmp.so_next; so != &slirp->icmp; so = so_next) {
 499         so_next = so->so_next;
 500
 501         so->pollfds_idx = -1;
 502
 503         /*
 504          * See if it's timed out
 505          */
 506         if (so->so_expire) {
 507             if (so->so_expire <= curtime) {
 508                 icmp_detach(so);
 509                 continue;
 510             } else {
 511                 slirp->do_slowtimo = true; /* Let socket expire */
 512             }
 513         }
 514
 515         if (so->so_state & SS_ISFCONNECTED) {
 516             so->pollfds_idx = add_poll(so->s,
 517                 SLIRP_POLL_IN | SLIRP_POLL_HUP | SLIRP_POLL_ERR, opaque);
 518         }
 519     }
 520
 521     slirp_update_timeout(slirp, timeout);
 522 }
 523
 524 void slirp_pollfds_poll(Slirp *slirp, int select_error,
 525                         SlirpGetREventsCb get_revents, void *opaque)
 526 {
 527     struct socket *so, *so_next;
 528     int ret;
 529
 530     curtime = slirp->cb->clock_get_ns(slirp->opaque) / SCALE_MS;
 531
 532     /*
 533      * See if anything has timed out
 534      */
 535     if (slirp->time_fasttimo &&
 536         ((curtime - slirp->time_fasttimo) >= TIMEOUT_FAST)) {
 537         tcp_fasttimo(slirp);
 538         slirp->time_fasttimo = 0;
 539     }
 540     if (slirp->do_slowtimo &&
 541         ((curtime - slirp->last_slowtimo) >= TIMEOUT_SLOW)) {
 542         ip_slowtimo(slirp);
 543         tcp_slowtimo(slirp);
 544         slirp->last_slowtimo = curtime;
 545     }
 546
 547     /*
 548      * Check sockets
 549      */
 550     if (!select_error) {
 551         /*
 552          * Check TCP sockets
 553          */
 554         for (so = slirp->tcb.so_next; so != &slirp->tcb;
 555              so = so_next) {
 556             int revents;
 557
 558             so_next = so->so_next;
 559
 560             revents = 0;
 561             if (so->pollfds_idx != -1) {
 562                 revents = get_revents(so->pollfds_idx, opaque);
 563             }
 564
 565             if (so->so_state & SS_NOFDREF || so->s == -1) {
 566                 continue;
 567             }
 568
 569             /*
 570              * Check for URG data
 571              * This will soread as well, so no need to
 572              * test for SLIRP_POLL_IN below if this succeeds
 573              */
 574             if (revents & SLIRP_POLL_PRI) {
 575                 ret = sorecvoob(so);
 576                 if (ret < 0) {
 577                     /* Socket error might have resulted in the socket being
 578                      * removed, do not try to do anything more with it. */
 579                     continue;
 580                 }
 581             }
 582             /*
 583              * Check sockets for reading
 584              */
 585             else if (revents &
 586                      (SLIRP_POLL_IN | SLIRP_POLL_HUP | SLIRP_POLL_ERR)) {
 587                 /*
 588                  * Check for incoming connections
 589                  */
 590                 if (so->so_state & SS_FACCEPTCONN) {
 591                     tcp_connect(so);
 592                     continue;
 593                 } /* else */
 594                 ret = soread(so);
 595
 596                 /* Output it if we read something */
 597                 if (ret > 0) {
 598                     tcp_output(sototcpcb(so));
 599                 }
 600                 if (ret < 0) {
 601                     /* Socket error might have resulted in the socket being
 602                      * removed, do not try to do anything more with it. */
 603                     continue;
 604                 }
 605             }
 606
 607             /*
 608              * Check sockets for writing
 609              */
 610             if (!(so->so_state & SS_NOFDREF) &&
 611                 (revents & (SLIRP_POLL_OUT | SLIRP_POLL_ERR))) {
 612                 /*
 613                  * Check for non-blocking, still-connecting sockets
 614                  */
 615                 if (so->so_state & SS_ISFCONNECTING) {
 616                     /* Connected */
 617                     so->so_state &= ~SS_ISFCONNECTING;
 618
 619                     ret = send(so->s, (const void *) &ret, 0, 0);
 620                     if (ret < 0) {
 621                         /* XXXXX Must fix, zero bytes is a NOP */
 622                         if (errno == EAGAIN || errno == EWOULDBLOCK ||
 623                             errno == EINPROGRESS || errno == ENOTCONN) {
 624                             continue;
 625                         }
 626
 627                         /* else failed */
 628                         so->so_state &= SS_PERSISTENT_MASK;
 629                         so->so_state |= SS_NOFDREF;
 630                     }
 631                     /* else so->so_state &= ~SS_ISFCONNECTING; */
 632
 633                     /*
 634                      * Continue tcp_input
 635                      */
 636                     tcp_input((struct mbuf *)NULL, sizeof(struct ip), so,
 637                               so->so_ffamily);
 638                     /* continue; */
 639                 } else {
 640                     ret = sowrite(so);
 641                     if (ret > 0) {
 642                         /* Call tcp_output in case we need to send a window
 643                          * update to the guest, otherwise it will be stuck
 644                          * until it sends a window probe. */
 645                         tcp_output(sototcpcb(so));
 646                     }
 647                 }
 648             }
 649         }
 650
 651         /*
 652          * Now UDP sockets.
 653          * Incoming packets are sent straight away, they're not buffered.
 654          * Incoming UDP data isn't buffered either.
 655          */
 656         for (so = slirp->udb.so_next; so != &slirp->udb;
 657              so = so_next) {
 658             int revents;
 659
 660             so_next = so->so_next;
 661
 662             revents = 0;
 663             if (so->pollfds_idx != -1) {
 664                 revents = get_revents(so->pollfds_idx, opaque);
 665             }
 666
 667             if (so->s != -1 &&
 668                 (revents & (SLIRP_POLL_IN | SLIRP_POLL_HUP | SLIRP_POLL_ERR))) {
 669                 sorecvfrom(so);
 670             }
 671         }
 672
 673         /*
 674          * Check incoming ICMP relies.
 675          */
 676         for (so = slirp->icmp.so_next; so != &slirp->icmp;
 677              so = so_next) {
 678             int revents;
 679
 680             so_next = so->so_next;
 681
 682             revents = 0;
 683             if (so->pollfds_idx != -1) {
 684                 revents = get_revents(so->pollfds_idx, opaque);
 685             }
 686
 687             if (so->s != -1 &&
 688                 (revents & (SLIRP_POLL_IN | SLIRP_POLL_HUP | SLIRP_POLL_ERR))) {
 689                 icmp_receive(so);
 690             }
 691         }
 692     }
 693
 694     if_start(slirp);
 695 }
 696
 697 static void arp_input(Slirp *slirp, const uint8_t *pkt, int pkt_len)
 698 {
 699     struct slirp_arphdr *ah = (struct slirp_arphdr *)(pkt + ETH_HLEN);
 700     uint8_t arp_reply[MAX(ETH_HLEN + sizeof(struct slirp_arphdr), 64)];
 701     struct ethhdr *reh = (struct ethhdr *)arp_reply;
 702     struct slirp_arphdr *rah = (struct slirp_arphdr *)(arp_reply + ETH_HLEN);
 703     int ar_op;
 704     struct gfwd_list *ex_ptr;
 705
 706     if (!slirp->in_enabled) {
 707         return;
 708     }
 709
 710     ar_op = ntohs(ah->ar_op);
 711     switch(ar_op) {
 712     case ARPOP_REQUEST:
 713         if (ah->ar_tip == ah->ar_sip) {
 714             /* Gratuitous ARP */
 715             arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
 716             return;
 717         }
 718
 719         if ((ah->ar_tip & slirp->vnetwork_mask.s_addr) ==
 720             slirp->vnetwork_addr.s_addr) {
 721             if (ah->ar_tip == slirp->vnameserver_addr.s_addr ||
 722                 ah->ar_tip == slirp->vhost_addr.s_addr)
 723                 goto arp_ok;
 724             /* TODO: IPv6 */
 725             for (ex_ptr = slirp->guestfwd_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
 726                 if (ex_ptr->ex_addr.s_addr == ah->ar_tip)
 727                     goto arp_ok;
 728             }
 729             return;
 730         arp_ok:
 731             memset(arp_reply, 0, sizeof(arp_reply));
 732
 733             arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
 734
 735             /* ARP request for alias/dns mac address */
 736             memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
 737             memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 4);
 738             memcpy(&reh->h_source[2], &ah->ar_tip, 4);
 739             reh->h_proto = htons(ETH_P_ARP);
 740
 741             rah->ar_hrd = htons(1);
 742             rah->ar_pro = htons(ETH_P_IP);
 743             rah->ar_hln = ETH_ALEN;
 744             rah->ar_pln = 4;
 745             rah->ar_op = htons(ARPOP_REPLY);
 746             memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
 747             rah->ar_sip = ah->ar_tip;
 748             memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
 749             rah->ar_tip = ah->ar_sip;
 750             slirp_send_packet_all(slirp, arp_reply, sizeof(arp_reply));
 751         }
 752         break;
 753     case ARPOP_REPLY:
 754         arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
 755         break;
 756     default:
 757         break;
 758     }
 759 }
 760
 761 void slirp_input(Slirp *slirp, const uint8_t *pkt, int pkt_len)
 762 {
 763     struct mbuf *m;
 764     int proto;
 765
 766     if (pkt_len < ETH_HLEN)
 767         return;
 768
 769     proto = (((uint16_t) pkt[12]) << 8) + pkt[13];
 770     switch(proto) {
 771     case ETH_P_ARP:
 772         arp_input(slirp, pkt, pkt_len);
 773         break;
 774     case ETH_P_IP:
 775     case ETH_P_IPV6:
 776         m = m_get(slirp);
 777         if (!m)
 778             return;
 779         /* Note: we add 2 to align the IP header on 4 bytes,
 780          * and add the margin for the tcpiphdr overhead  */
 781         if (M_FREEROOM(m) < pkt_len + TCPIPHDR_DELTA + 2) {
 782             m_inc(m, pkt_len + TCPIPHDR_DELTA + 2);
 783         }
 784         m->m_len = pkt_len + TCPIPHDR_DELTA + 2;
 785         memcpy(m->m_data + TCPIPHDR_DELTA + 2, pkt, pkt_len);
 786
 787         m->m_data += TCPIPHDR_DELTA + 2 + ETH_HLEN;
 788         m->m_len -= TCPIPHDR_DELTA + 2 + ETH_HLEN;
 789
 790         if (proto == ETH_P_IP) {
 791             ip_input(m);
 792         } else if (proto == ETH_P_IPV6) {
 793             ip6_input(m);
 794         }
 795         break;
 796
 797     case ETH_P_NCSI:
 798         ncsi_input(slirp, pkt, pkt_len);
 799         break;
 800
 801     default:
 802         break;
 803     }
 804 }
 805
 806 /* Prepare the IPv4 packet to be sent to the ethernet device. Returns 1 if no
 807  * packet should be sent, 0 if the packet must be re-queued, 2 if the packet
 808  * is ready to go.
 809  */
 810 static int if_encap4(Slirp *slirp, struct mbuf *ifm, struct ethhdr *eh,
 811         uint8_t ethaddr[ETH_ALEN])
 812 {
 813     const struct ip *iph = (const struct ip *)ifm->m_data;
 814
 815     if (iph->ip_dst.s_addr == 0) {
 816         /* 0.0.0.0 can not be a destination address, something went wrong,
 817          * avoid making it worse */
 818         return 1;
 819     }
 820     if (!arp_table_search(slirp, iph->ip_dst.s_addr, ethaddr)) {
 821         uint8_t arp_req[ETH_HLEN + sizeof(struct slirp_arphdr)];
 822         struct ethhdr *reh = (struct ethhdr *)arp_req;
 823         struct slirp_arphdr *rah = (struct slirp_arphdr *)(arp_req + ETH_HLEN);
 824
 825         if (!ifm->resolution_requested) {
 826             /* If the client addr is not known, send an ARP request */
 827             memset(reh->h_dest, 0xff, ETH_ALEN);
 828             memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 4);
 829             memcpy(&reh->h_source[2], &slirp->vhost_addr, 4);
 830             reh->h_proto = htons(ETH_P_ARP);
 831             rah->ar_hrd = htons(1);
 832             rah->ar_pro = htons(ETH_P_IP);
 833             rah->ar_hln = ETH_ALEN;
 834             rah->ar_pln = 4;
 835             rah->ar_op = htons(ARPOP_REQUEST);
 836
 837             /* source hw addr */
 838             memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN - 4);
 839             memcpy(&rah->ar_sha[2], &slirp->vhost_addr, 4);
 840
 841             /* source IP */
 842             rah->ar_sip = slirp->vhost_addr.s_addr;
 843
 844             /* target hw addr (none) */
 845             memset(rah->ar_tha, 0, ETH_ALEN);
 846
 847             /* target IP */
 848             rah->ar_tip = iph->ip_dst.s_addr;
 849             slirp->client_ipaddr = iph->ip_dst;
 850             slirp_send_packet_all(slirp, arp_req, sizeof(arp_req));
 851             ifm->resolution_requested = true;
 852
 853             /* Expire request and drop outgoing packet after 1 second */
 854             ifm->expiration_date =
 855                 slirp->cb->clock_get_ns(slirp->opaque) + 1000000000ULL;
 856         }
 857         return 0;
 858     } else {
 859         memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 4);
 860         /* XXX: not correct */
 861         memcpy(&eh->h_source[2], &slirp->vhost_addr, 4);
 862         eh->h_proto = htons(ETH_P_IP);
 863
 864         /* Send this */
 865         return 2;
 866     }
 867 }
 868
 869 /* Prepare the IPv6 packet to be sent to the ethernet device. Returns 1 if no
 870  * packet should be sent, 0 if the packet must be re-queued, 2 if the packet
 871  * is ready to go.
 872  */
 873 static int if_encap6(Slirp *slirp, struct mbuf *ifm, struct ethhdr *eh,
 874         uint8_t ethaddr[ETH_ALEN])
 875 {
 876     const struct ip6 *ip6h = mtod(ifm, const struct ip6 *);
 877     if (!ndp_table_search(slirp, ip6h->ip_dst, ethaddr)) {
 878         if (!ifm->resolution_requested) {
 879             ndp_send_ns(slirp, ip6h->ip_dst);
 880             ifm->resolution_requested = true;
 881             ifm->expiration_date = slirp->cb->clock_get_ns(slirp->opaque) + 1000000000ULL;
 882         }
 883         return 0;
 884     } else {
 885         eh->h_proto = htons(ETH_P_IPV6);
 886         in6_compute_ethaddr(ip6h->ip_src, eh->h_source);
 887
 888         /* Send this */
 889         return 2;
 890     }
 891 }
 892
 893 /* Output the IP packet to the ethernet device. Returns 0 if the packet must be
 894  * re-queued.
 895  */
 896 int if_encap(Slirp *slirp, struct mbuf *ifm)
 897 {
 898     uint8_t buf[1600];
 899     struct ethhdr *eh = (struct ethhdr *)buf;
 900     uint8_t ethaddr[ETH_ALEN];
 901     const struct ip *iph = (const struct ip *)ifm->m_data;
 902     int ret;
 903
 904     if (ifm->m_len + ETH_HLEN > sizeof(buf)) {
 905         return 1;
 906     }
 907
 908     switch (iph->ip_v) {
 909     case IPVERSION:
 910         ret = if_encap4(slirp, ifm, eh, ethaddr);
 911         if (ret < 2) {
 912             return ret;
 913         }
 914         break;
 915
 916     case IP6VERSION:
 917         ret = if_encap6(slirp, ifm, eh, ethaddr);
 918         if (ret < 2) {
 919             return ret;
 920         }
 921         break;
 922
 923     default:
 924         g_assert_not_reached();
 925         break;
 926     }
 927
 928     memcpy(eh->h_dest, ethaddr, ETH_ALEN);
 929     DEBUG_ARG("src = %02x:%02x:%02x:%02x:%02x:%02x",
 930               eh->h_source[0], eh->h_source[1], eh->h_source[2],
 931               eh->h_source[3], eh->h_source[4], eh->h_source[5]);
 932     DEBUG_ARG("dst = %02x:%02x:%02x:%02x:%02x:%02x",
 933               eh->h_dest[0], eh->h_dest[1], eh->h_dest[2],
 934               eh->h_dest[3], eh->h_dest[4], eh->h_dest[5]);
 935     memcpy(buf + sizeof(struct ethhdr), ifm->m_data, ifm->m_len);
 936     slirp_send_packet_all(slirp, buf, ifm->m_len + ETH_HLEN);
 937     return 1;
 938 }
 939
 940 /* Drop host forwarding rule, return 0 if found. */
 941 /* TODO: IPv6 */
 942 int slirp_remove_hostfwd(Slirp *slirp, int is_udp, struct in_addr host_addr,
 943                          int host_port)
 944 {
 945     struct socket *so;
 946     struct socket *head = (is_udp ? &slirp->udb : &slirp->tcb);
 947     struct sockaddr_in addr;
 948     int port = htons(host_port);
 949     socklen_t addr_len;
 950
 951     for (so = head->so_next; so != head; so = so->so_next) {
 952         addr_len = sizeof(addr);
 953         if ((so->so_state & SS_HOSTFWD) &&
 954             getsockname(so->s, (struct sockaddr *)&addr, &addr_len) == 0 &&
 955             addr.sin_addr.s_addr == host_addr.s_addr &&
 956             addr.sin_port == port) {
 957             so->slirp->cb->unregister_poll_fd(so->s, so->slirp->opaque);
 958             closesocket(so->s);
 959             sofree(so);
 960             return 0;
 961         }
 962     }
 963
 964     return -1;
 965 }
 966
 967 /* TODO: IPv6 */
 968 int slirp_add_hostfwd(Slirp *slirp, int is_udp, struct in_addr host_addr,
 969                       int host_port, struct in_addr guest_addr, int guest_port)
 970 {
 971     if (!guest_addr.s_addr) {
 972         guest_addr = slirp->vdhcp_startaddr;
 973     }
 974     if (is_udp) {
 975         if (!udp_listen(slirp, host_addr.s_addr, htons(host_port),
 976                         guest_addr.s_addr, htons(guest_port), SS_HOSTFWD))
 977             return -1;
 978     } else {
 979         if (!tcp_listen(slirp, host_addr.s_addr, htons(host_port),
 980                         guest_addr.s_addr, htons(guest_port), SS_HOSTFWD))
 981             return -1;
 982     }
 983     return 0;
 984 }
 985
 986 /* TODO: IPv6 */
 987 static bool
 988 check_guestfwd(Slirp *slirp, struct in_addr *guest_addr, int guest_port)
 989 {
 990     struct gfwd_list *tmp_ptr;
 991
 992     if (!guest_addr->s_addr) {
 993         guest_addr->s_addr = slirp->vnetwork_addr.s_addr |
 994             (htonl(0x0204) & ~slirp->vnetwork_mask.s_addr);
 995     }
 996     if ((guest_addr->s_addr & slirp->vnetwork_mask.s_addr) !=
 997         slirp->vnetwork_addr.s_addr ||
 998         guest_addr->s_addr == slirp->vhost_addr.s_addr ||
 999         guest_addr->s_addr == slirp->vnameserver_addr.s_addr) {
1000         return false;
1001     }
1002
1003     /* check if the port is "bound" */
1004     for (tmp_ptr = slirp->guestfwd_list; tmp_ptr; tmp_ptr = tmp_ptr->ex_next) {
1005         if (guest_port == tmp_ptr->ex_fport &&
1006             guest_addr->s_addr == tmp_ptr->ex_addr.s_addr)
1007             return false;
1008     }
1009
1010     return true;
1011 }
1012
1013 int slirp_add_exec(Slirp *slirp, const char *cmdline,
1014                    struct in_addr *guest_addr, int guest_port)
1015 {
1016     if (!check_guestfwd(slirp, guest_addr, guest_port)) {
1017         return -1;
1018     }
1019
1020     add_exec(&slirp->guestfwd_list, cmdline, *guest_addr, htons(guest_port));
1021     return 0;
1022 }
1023
1024 int slirp_add_guestfwd(Slirp *slirp, SlirpWriteCb write_cb, void *opaque,
1025                        struct in_addr *guest_addr, int guest_port)
1026 {
1027     if (!check_guestfwd(slirp, guest_addr, guest_port)) {
1028         return -1;
1029     }
1030
1031     add_guestfwd(&slirp->guestfwd_list, write_cb, opaque,
1032                  *guest_addr, htons(guest_port));
1033     return 0;
1034 }
1035
1036 ssize_t slirp_send(struct socket *so, const void *buf, size_t len, int flags)
1037 {
1038     if (so->s == -1 && so->guestfwd) {
1039         /* XXX this blocks entire thread. Rewrite to use
1040          * qemu_chr_fe_write and background I/O callbacks */
1041         so->guestfwd->write_cb(buf, len, so->guestfwd->opaque);
1042         return len;
1043     }
1044
1045     if (so->s == -1) {
1046         /*
1047          * This should in theory not happen but it is hard to be
1048          * sure because some code paths will end up with so->s == -1
1049          * on a failure but don't dispose of the struct socket.
1050          * Check specifically, so we don't pass -1 to send().
1051          */
1052         errno = EBADF;
1053         return -1;
1054     }
1055
1056     return send(so->s, buf, len, flags);
1057 }
1058
1059 struct socket *
1060 slirp_find_ctl_socket(Slirp *slirp, struct in_addr guest_addr, int guest_port)
1061 {
1062     struct socket *so;
1063
1064     /* TODO: IPv6 */
1065     for (so = slirp->tcb.so_next; so != &slirp->tcb; so = so->so_next) {
1066         if (so->so_faddr.s_addr == guest_addr.s_addr &&
1067             htons(so->so_fport) == guest_port) {
1068             return so;
1069         }
1070     }
1071     return NULL;
1072 }
1073
1074 size_t slirp_socket_can_recv(Slirp *slirp, struct in_addr guest_addr,
1075                              int guest_port)
1076 {
1077     struct iovec iov[2];
1078     struct socket *so;
1079
1080     so = slirp_find_ctl_socket(slirp, guest_addr, guest_port);
1081
1082     if (!so || so->so_state & SS_NOFDREF) {
1083         return 0;
1084     }
1085
1086     if (!CONN_CANFRCV(so) || so->so_snd.sb_cc >= (so->so_snd.sb_datalen/2)) {
1087         return 0;
1088     }
1089
1090     return sopreprbuf(so, iov, NULL);
1091 }
1092
1093 void slirp_socket_recv(Slirp *slirp, struct in_addr guest_addr, int guest_port,
1094                        const uint8_t *buf, int size)
1095 {
1096     int ret;
1097     struct socket *so = slirp_find_ctl_socket(slirp, guest_addr, guest_port);
1098
1099     if (!so)
1100         return;
1101
1102     ret = soreadbuf(so, (const char *)buf, size);
1103
1104     if (ret > 0)
1105         tcp_output(sototcpcb(so));
1106 }
1107
1108 void slirp_send_packet_all(Slirp *slirp, const void *buf, size_t len)
1109 {
1110     ssize_t ret = slirp->cb->send_packet(buf, len, slirp->opaque);
1111
1112     if (ret < 0) {
1113         g_critical("Failed to send packet, ret: %ld", (long) ret);
1114     } else if (ret < len) {
1115         DEBUG_ERROR("send_packet() didn't send all data: %ld < %lu",
1116                 (long) ret, (unsigned long) len);
1117     }
1118 }