slirp/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 #ifdef WITH_QEMU
  27 #include "state.h"
  28 #endif
  29
  30 #ifndef _WIN32
  31 #include <net/if.h>
  32 #endif
  33
  34 int slirp_debug;
  35
  36 /* Define to 1 if you want KEEPALIVE timers */
  37 bool slirp_do_keepalive;
  38
  39 /* host loopback address */
  40 struct in_addr loopback_addr;
  41 /* host loopback network mask */
  42 unsigned long loopback_mask;
  43
  44 /* emulated hosts use the MAC addr 52:55:IP:IP:IP:IP */
  45 static const uint8_t special_ethaddr[ETH_ALEN] = {
  46     0x52, 0x55, 0x00, 0x00, 0x00, 0x00
  47 };
  48
  49 unsigned curtime;
  50
  51 static struct in_addr dns_addr;
  52 #ifndef _WIN32
  53 static struct in6_addr dns6_addr;
  54 #endif
  55 static unsigned dns_addr_time;
  56 #ifndef _WIN32
  57 static unsigned dns6_addr_time;
  58 #endif
  59
  60 #define TIMEOUT_FAST 2  /* milliseconds */
  61 #define TIMEOUT_SLOW 499  /* milliseconds */
  62 /* for the aging of certain requests like DNS */
  63 #define TIMEOUT_DEFAULT 1000  /* milliseconds */
  64
  65 #ifdef _WIN32
  66
  67 int get_dns_addr(struct in_addr *pdns_addr)
  68 {
  69     FIXED_INFO *FixedInfo=NULL;
  70     ULONG    BufLen;
  71     DWORD    ret;
  72     IP_ADDR_STRING *pIPAddr;
  73     struct in_addr tmp_addr;
  74
  75     if (dns_addr.s_addr != 0 && (curtime - dns_addr_time) < TIMEOUT_DEFAULT) {
  76         *pdns_addr = dns_addr;
  77         return 0;
  78     }
  79
  80     FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
  81     BufLen = sizeof(FIXED_INFO);
  82
  83     if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen)) {
  84         if (FixedInfo) {
  85             GlobalFree(FixedInfo);
  86             FixedInfo = NULL;
  87         }
  88         FixedInfo = GlobalAlloc(GPTR, BufLen);
  89     }
  90
  91     if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS) {
  92         printf("GetNetworkParams failed. ret = %08x\n", (unsigned)ret );
  93         if (FixedInfo) {
  94             GlobalFree(FixedInfo);
  95             FixedInfo = NULL;
  96         }
  97         return -1;
  98     }
  99
 100     pIPAddr = &(FixedInfo->DnsServerList);
 101     inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
 102     *pdns_addr = tmp_addr;
 103     dns_addr = tmp_addr;
 104     dns_addr_time = curtime;
 105     if (FixedInfo) {
 106         GlobalFree(FixedInfo);
 107         FixedInfo = NULL;
 108     }
 109     return 0;
 110 }
 111
 112 int get_dns6_addr(struct in6_addr *pdns6_addr, uint32_t *scope_id)
 113 {
 114     return -1;
 115 }
 116
 117 static void winsock_cleanup(void)
 118 {
 119     WSACleanup();
 120 }
 121
 122 #else
 123
 124 static int get_dns_addr_cached(void *pdns_addr, void *cached_addr,
 125                                socklen_t addrlen,
 126                                struct stat *cached_stat, unsigned *cached_time)
 127 {
 128     struct stat old_stat;
 129     if (curtime - *cached_time < TIMEOUT_DEFAULT) {
 130         memcpy(pdns_addr, cached_addr, addrlen);
 131         return 0;
 132     }
 133     old_stat = *cached_stat;
 134     if (stat("/etc/resolv.conf", cached_stat) != 0) {
 135         return -1;
 136     }
 137     if (cached_stat->st_dev == old_stat.st_dev
 138         && cached_stat->st_ino == old_stat.st_ino
 139         && cached_stat->st_size == old_stat.st_size
 140         && cached_stat->st_mtime == old_stat.st_mtime) {
 141         memcpy(pdns_addr, cached_addr, addrlen);
 142         return 0;
 143     }
 144     return 1;
 145 }
 146
 147 static int get_dns_addr_resolv_conf(int af, void *pdns_addr, void *cached_addr,
 148                                     socklen_t addrlen, uint32_t *scope_id,
 149                                     unsigned *cached_time)
 150 {
 151     char buff[512];
 152     char buff2[257];
 153     FILE *f;
 154     int found = 0;
 155     void *tmp_addr = alloca(addrlen);
 156     unsigned if_index;
 157
 158     f = fopen("/etc/resolv.conf", "r");
 159     if (!f)
 160         return -1;
 161
 162     DEBUG_MISC("IP address of your DNS(s):");
 163     while (fgets(buff, 512, f) != NULL) {
 164         if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1) {
 165             char *c = strchr(buff2, '%');
 166             if (c) {
 167                 if_index = if_nametoindex(c + 1);
 168                 *c = '\0';
 169             } else {
 170                 if_index = 0;
 171             }
 172
 173             if (!inet_pton(af, buff2, tmp_addr)) {
 174                 continue;
 175             }
 176             /* If it's the first one, set it to dns_addr */
 177             if (!found) {
 178                 memcpy(pdns_addr, tmp_addr, addrlen);
 179                 memcpy(cached_addr, tmp_addr, addrlen);
 180                 if (scope_id) {
 181                     *scope_id = if_index;
 182                 }
 183                 *cached_time = curtime;
 184             }
 185
 186             if (++found > 3) {
 187                 DEBUG_MISC("  (more)");
 188                 break;
 189             } else if (slirp_debug & DBG_MISC) {
 190                 char s[INET6_ADDRSTRLEN];
 191                 const char *res = inet_ntop(af, tmp_addr, s, sizeof(s));
 192                 if (!res) {
 193                     res = "  (string conversion error)";
 194                 }
 195                 DEBUG_MISC("  %s", res);
 196             }
 197         }
 198     }
 199     fclose(f);
 200     if (!found)
 201         return -1;
 202     return 0;
 203 }
 204
 205 int get_dns_addr(struct in_addr *pdns_addr)
 206 {
 207     static struct stat dns_addr_stat;
 208
 209     if (dns_addr.s_addr != 0) {
 210         int ret;
 211         ret = get_dns_addr_cached(pdns_addr, &dns_addr, sizeof(dns_addr),
 212                                   &dns_addr_stat, &dns_addr_time);
 213         if (ret <= 0) {
 214             return ret;
 215         }
 216     }
 217     return get_dns_addr_resolv_conf(AF_INET, pdns_addr, &dns_addr,
 218                                     sizeof(dns_addr), NULL, &dns_addr_time);
 219 }
 220
 221 int get_dns6_addr(struct in6_addr *pdns6_addr, uint32_t *scope_id)
 222 {
 223     static struct stat dns6_addr_stat;
 224
 225     if (!in6_zero(&dns6_addr)) {
 226         int ret;
 227         ret = get_dns_addr_cached(pdns6_addr, &dns6_addr, sizeof(dns6_addr),
 228                                   &dns6_addr_stat, &dns6_addr_time);
 229         if (ret <= 0) {
 230             return ret;
 231         }
 232     }
 233     return get_dns_addr_resolv_conf(AF_INET6, pdns6_addr, &dns6_addr,
 234                                     sizeof(dns6_addr),
 235                                     scope_id, &dns6_addr_time);
 236 }
 237
 238 #endif
 239
 240 static void slirp_init_once(void)
 241 {
 242     static int initialized;
 243     const char *debug;
 244 #ifdef _WIN32
 245     WSADATA Data;
 246 #endif
 247
 248     if (initialized) {
 249         return;
 250     }
 251     initialized = 1;
 252
 253 #ifdef _WIN32
 254     WSAStartup(MAKEWORD(2,0), &Data);
 255     atexit(winsock_cleanup);
 256 #endif
 257
 258     loopback_addr.s_addr = htonl(INADDR_LOOPBACK);
 259     loopback_mask = htonl(IN_CLASSA_NET);
 260
 261     debug = g_getenv("SLIRP_DEBUG");
 262     if (debug) {
 263         const GDebugKey keys[] = {
 264             { "call", DBG_CALL },
 265             { "misc", DBG_MISC },
 266             { "error", DBG_ERROR },
 267             { "tftp", DBG_TFTP },
 268         };
 269         slirp_debug = g_parse_debug_string(debug, keys, G_N_ELEMENTS(keys));
 270     }
 271
 272
 273 }
 274
 275 Slirp *slirp_init(int restricted, bool in_enabled, struct in_addr vnetwork,
 276                   struct in_addr vnetmask, struct in_addr vhost,
 277                   bool in6_enabled,
 278                   struct in6_addr vprefix_addr6, uint8_t vprefix_len,
 279                   struct in6_addr vhost6, const char *vhostname,
 280                   const char *tftp_server_name,
 281                   const char *tftp_path, const char *bootfile,
 282                   struct in_addr vdhcp_start, struct in_addr vnameserver,
 283                   struct in6_addr vnameserver6, const char **vdnssearch,
 284                   const char *vdomainname,
 285                   const SlirpCb *callbacks,
 286                   void *opaque)
 287 {
 288     Slirp *slirp = g_malloc0(sizeof(Slirp));
 289
 290     slirp_init_once();
 291
 292     slirp->opaque = opaque;
 293     slirp->cb = callbacks;
 294     slirp->grand = g_rand_new();
 295     slirp->restricted = restricted;
 296
 297     slirp->in_enabled = in_enabled;
 298     slirp->in6_enabled = in6_enabled;
 299
 300     if_init(slirp);
 301     ip_init(slirp);
 302     ip6_init(slirp);
 303
 304     /* Initialise mbufs *after* setting the MTU */
 305     m_init(slirp);
 306
 307     slirp->vnetwork_addr = vnetwork;
 308     slirp->vnetwork_mask = vnetmask;
 309     slirp->vhost_addr = vhost;
 310     slirp->vprefix_addr6 = vprefix_addr6;
 311     slirp->vprefix_len = vprefix_len;
 312     slirp->vhost_addr6 = vhost6;
 313     if (vhostname) {
 314         slirp_pstrcpy(slirp->client_hostname, sizeof(slirp->client_hostname),
 315                       vhostname);
 316     }
 317     slirp->tftp_prefix = g_strdup(tftp_path);
 318     slirp->bootp_filename = g_strdup(bootfile);
 319     slirp->vdomainname = g_strdup(vdomainname);
 320     slirp->vdhcp_startaddr = vdhcp_start;
 321     slirp->vnameserver_addr = vnameserver;
 322     slirp->vnameserver_addr6 = vnameserver6;
 323     slirp->tftp_server_name = g_strdup(tftp_server_name);
 324
 325     if (vdnssearch) {
 326         translate_dnssearch(slirp, vdnssearch);
 327     }
 328
 329 #ifdef WITH_QEMU
 330     slirp_state_register(slirp);
 331 #endif
 332     return slirp;
 333 }
 334
 335 void slirp_cleanup(Slirp *slirp)
 336 {
 337     struct gfwd_list *e, *next;
 338
 339     for (e = slirp->guestfwd_list; e; e = next) {
 340         next = e->ex_next;
 341         g_free(e->ex_exec);
 342         g_free(e);
 343     }
 344
 345 #ifdef WITH_QEMU
 346     slirp_state_unregister(slirp);
 347 #endif
 348     ip_cleanup(slirp);
 349     ip6_cleanup(slirp);
 350     m_cleanup(slirp);
 351
 352     g_rand_free(slirp->grand);
 353
 354     g_free(slirp->vdnssearch);
 355     g_free(slirp->tftp_prefix);
 356     g_free(slirp->bootp_filename);
 357     g_free(slirp->vdomainname);
 358     g_free(slirp);
 359 }
 360
 361 #define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
 362 #define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
 363
 364 static void slirp_update_timeout(Slirp *slirp, uint32_t *timeout)
 365 {
 366     uint32_t t;
 367
 368     if (*timeout <= TIMEOUT_FAST) {
 369         return;
 370     }
 371
 372     t = MIN(1000, *timeout);
 373
 374     /* If we have tcp timeout with slirp, then we will fill @timeout with
 375      * more precise value.
 376      */
 377     if (slirp->time_fasttimo) {
 378         *timeout = TIMEOUT_FAST;
 379         return;
 380     }
 381     if (slirp->do_slowtimo) {
 382         t = MIN(TIMEOUT_SLOW, t);
 383     }
 384     *timeout = t;
 385 }
 386
 387 void slirp_pollfds_fill(Slirp *slirp, uint32_t *timeout,
 388                         SlirpAddPollCb add_poll, void *opaque)
 389 {
 390     struct socket *so, *so_next;
 391
 392     /*
 393      * First, TCP sockets
 394      */
 395
 396     /*
 397      * *_slowtimo needs calling if there are IP fragments
 398      * in the fragment queue, or there are TCP connections active
 399      */
 400     slirp->do_slowtimo = ((slirp->tcb.so_next != &slirp->tcb) ||
 401                           (&slirp->ipq.ip_link != slirp->ipq.ip_link.next));
 402
 403     for (so = slirp->tcb.so_next; so != &slirp->tcb; so = so_next) {
 404         int events = 0;
 405
 406         so_next = so->so_next;
 407
 408         so->pollfds_idx = -1;
 409
 410         /*
 411          * See if we need a tcp_fasttimo
 412          */
 413         if (slirp->time_fasttimo == 0 &&
 414             so->so_tcpcb->t_flags & TF_DELACK) {
 415             slirp->time_fasttimo = curtime; /* Flag when want a fasttimo */
 416         }
 417
 418         /*
 419          * NOFDREF can include still connecting to local-host,
 420          * newly socreated() sockets etc. Don't want to select these.
 421          */
 422         if (so->so_state & SS_NOFDREF || so->s == -1) {
 423             continue;
 424         }
 425
 426         /*
 427          * Set for reading sockets which are accepting
 428          */
 429         if (so->so_state & SS_FACCEPTCONN) {
 430             so->pollfds_idx = add_poll(so->s,
 431                 SLIRP_POLL_IN | SLIRP_POLL_HUP | SLIRP_POLL_ERR, opaque);
 432             continue;
 433         }
 434
 435         /*
 436          * Set for writing sockets which are connecting
 437          */
 438         if (so->so_state & SS_ISFCONNECTING) {
 439             so->pollfds_idx = add_poll(so->s,
 440                 SLIRP_POLL_OUT | SLIRP_POLL_ERR, opaque);
 441             continue;
 442         }
 443
 444         /*
 445          * Set for writing if we are connected, can send more, and
 446          * we have something to send
 447          */
 448         if (CONN_CANFSEND(so) && so->so_rcv.sb_cc) {
 449             events |= SLIRP_POLL_OUT | SLIRP_POLL_ERR;
 450         }
 451
 452         /*
 453          * Set for reading (and urgent data) if we are connected, can
 454          * receive more, and we have room for it XXX /2 ?
 455          */
 456         if (CONN_CANFRCV(so) &&
 457             (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2))) {
 458             events |= SLIRP_POLL_IN | SLIRP_POLL_HUP |
 459                       SLIRP_POLL_ERR | SLIRP_POLL_PRI;
 460         }
 461
 462         if (events) {
 463             so->pollfds_idx = add_poll(so->s, events, opaque);
 464         }
 465     }
 466
 467     /*
 468      * UDP sockets
 469      */
 470     for (so = slirp->udb.so_next; so != &slirp->udb; so = so_next) {
 471         so_next = so->so_next;
 472
 473         so->pollfds_idx = -1;
 474
 475         /*
 476          * See if it's timed out
 477          */
 478         if (so->so_expire) {
 479             if (so->so_expire <= curtime) {
 480                 udp_detach(so);
 481                 continue;
 482             } else {
 483                 slirp->do_slowtimo = true; /* Let socket expire */
 484             }
 485         }
 486
 487         /*
 488          * When UDP packets are received from over the
 489          * link, they're sendto()'d straight away, so
 490          * no need for setting for writing
 491          * Limit the number of packets queued by this session
 492          * to 4.  Note that even though we try and limit this
 493          * to 4 packets, the session could have more queued
 494          * if the packets needed to be fragmented
 495          * (XXX <= 4 ?)
 496          */
 497         if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4) {
 498             so->pollfds_idx = add_poll(so->s,
 499                 SLIRP_POLL_IN | SLIRP_POLL_HUP | SLIRP_POLL_ERR, opaque);
 500         }
 501     }
 502
 503     /*
 504      * ICMP sockets
 505      */
 506     for (so = slirp->icmp.so_next; so != &slirp->icmp; so = so_next) {
 507         so_next = so->so_next;
 508
 509         so->pollfds_idx = -1;
 510
 511         /*
 512          * See if it's timed out
 513          */
 514         if (so->so_expire) {
 515             if (so->so_expire <= curtime) {
 516                 icmp_detach(so);
 517                 continue;
 518             } else {
 519                 slirp->do_slowtimo = true; /* Let socket expire */
 520             }
 521         }
 522
 523         if (so->so_state & SS_ISFCONNECTED) {
 524             so->pollfds_idx = add_poll(so->s,
 525                 SLIRP_POLL_IN | SLIRP_POLL_HUP | SLIRP_POLL_ERR, opaque);
 526         }
 527     }
 528
 529     slirp_update_timeout(slirp, timeout);
 530 }
 531
 532 void slirp_pollfds_poll(Slirp *slirp, int select_error,
 533                         SlirpGetREventsCb get_revents, void *opaque)
 534 {
 535     struct socket *so, *so_next;
 536     int ret;
 537
 538     curtime = slirp->cb->clock_get_ns(slirp->opaque) / SCALE_MS;
 539
 540     /*
 541      * See if anything has timed out
 542      */
 543     if (slirp->time_fasttimo &&
 544         ((curtime - slirp->time_fasttimo) >= TIMEOUT_FAST)) {
 545         tcp_fasttimo(slirp);
 546         slirp->time_fasttimo = 0;
 547     }
 548     if (slirp->do_slowtimo &&
 549         ((curtime - slirp->last_slowtimo) >= TIMEOUT_SLOW)) {
 550         ip_slowtimo(slirp);
 551         tcp_slowtimo(slirp);
 552         slirp->last_slowtimo = curtime;
 553     }
 554
 555     /*
 556      * Check sockets
 557      */
 558     if (!select_error) {
 559         /*
 560          * Check TCP sockets
 561          */
 562         for (so = slirp->tcb.so_next; so != &slirp->tcb;
 563              so = so_next) {
 564             int revents;
 565
 566             so_next = so->so_next;
 567
 568             revents = 0;
 569             if (so->pollfds_idx != -1) {
 570                 revents = get_revents(so->pollfds_idx, opaque);
 571             }
 572
 573             if (so->so_state & SS_NOFDREF || so->s == -1) {
 574                 continue;
 575             }
 576
 577             /*
 578              * Check for URG data
 579              * This will soread as well, so no need to
 580              * test for SLIRP_POLL_IN below if this succeeds
 581              */
 582             if (revents & SLIRP_POLL_PRI) {
 583                 ret = sorecvoob(so);
 584                 if (ret < 0) {
 585                     /* Socket error might have resulted in the socket being
 586                      * removed, do not try to do anything more with it. */
 587                     continue;
 588                 }
 589             }
 590             /*
 591              * Check sockets for reading
 592              */
 593             else if (revents &
 594                      (SLIRP_POLL_IN | SLIRP_POLL_HUP | SLIRP_POLL_ERR)) {
 595                 /*
 596                  * Check for incoming connections
 597                  */
 598                 if (so->so_state & SS_FACCEPTCONN) {
 599                     tcp_connect(so);
 600                     continue;
 601                 } /* else */
 602                 ret = soread(so);
 603
 604                 /* Output it if we read something */
 605                 if (ret > 0) {
 606                     tcp_output(sototcpcb(so));
 607                 }
 608                 if (ret < 0) {
 609                     /* Socket error might have resulted in the socket being
 610                      * removed, do not try to do anything more with it. */
 611                     continue;
 612                 }
 613             }
 614
 615             /*
 616              * Check sockets for writing
 617              */
 618             if (!(so->so_state & SS_NOFDREF) &&
 619                 (revents & (SLIRP_POLL_OUT | SLIRP_POLL_ERR))) {
 620                 /*
 621                  * Check for non-blocking, still-connecting sockets
 622                  */
 623                 if (so->so_state & SS_ISFCONNECTING) {
 624                     /* Connected */
 625                     so->so_state &= ~SS_ISFCONNECTING;
 626
 627                     ret = send(so->s, (const void *) &ret, 0, 0);
 628                     if (ret < 0) {
 629                         /* XXXXX Must fix, zero bytes is a NOP */
 630                         if (errno == EAGAIN || errno == EWOULDBLOCK ||
 631                             errno == EINPROGRESS || errno == ENOTCONN) {
 632                             continue;
 633                         }
 634
 635                         /* else failed */
 636                         so->so_state &= SS_PERSISTENT_MASK;
 637                         so->so_state |= SS_NOFDREF;
 638                     }
 639                     /* else so->so_state &= ~SS_ISFCONNECTING; */
 640
 641                     /*
 642                      * Continue tcp_input
 643                      */
 644                     tcp_input((struct mbuf *)NULL, sizeof(struct ip), so,
 645                               so->so_ffamily);
 646                     /* continue; */
 647                 } else {
 648                     ret = sowrite(so);
 649                     if (ret > 0) {
 650                         /* Call tcp_output in case we need to send a window
 651                          * update to the guest, otherwise it will be stuck
 652                          * until it sends a window probe. */
 653                         tcp_output(sototcpcb(so));
 654                     }
 655                 }
 656             }
 657         }
 658
 659         /*
 660          * Now UDP sockets.
 661          * Incoming packets are sent straight away, they're not buffered.
 662          * Incoming UDP data isn't buffered either.
 663          */
 664         for (so = slirp->udb.so_next; so != &slirp->udb;
 665              so = so_next) {
 666             int revents;
 667
 668             so_next = so->so_next;
 669
 670             revents = 0;
 671             if (so->pollfds_idx != -1) {
 672                 revents = get_revents(so->pollfds_idx, opaque);
 673             }
 674
 675             if (so->s != -1 &&
 676                 (revents & (SLIRP_POLL_IN | SLIRP_POLL_HUP | SLIRP_POLL_ERR))) {
 677                 sorecvfrom(so);
 678             }
 679         }
 680
 681         /*
 682          * Check incoming ICMP relies.
 683          */
 684         for (so = slirp->icmp.so_next; so != &slirp->icmp;
 685              so = so_next) {
 686             int revents;
 687
 688             so_next = so->so_next;
 689
 690             revents = 0;
 691             if (so->pollfds_idx != -1) {
 692                 revents = get_revents(so->pollfds_idx, opaque);
 693             }
 694
 695             if (so->s != -1 &&
 696                 (revents & (SLIRP_POLL_IN | SLIRP_POLL_HUP | SLIRP_POLL_ERR))) {
 697                 icmp_receive(so);
 698             }
 699         }
 700     }
 701
 702     if_start(slirp);
 703 }
 704
 705 static void arp_input(Slirp *slirp, const uint8_t *pkt, int pkt_len)
 706 {
 707     struct slirp_arphdr *ah = (struct slirp_arphdr *)(pkt + ETH_HLEN);
 708     uint8_t arp_reply[MAX(ETH_HLEN + sizeof(struct slirp_arphdr), 64)];
 709     struct ethhdr *reh = (struct ethhdr *)arp_reply;
 710     struct slirp_arphdr *rah = (struct slirp_arphdr *)(arp_reply + ETH_HLEN);
 711     int ar_op;
 712     struct gfwd_list *ex_ptr;
 713
 714     if (!slirp->in_enabled) {
 715         return;
 716     }
 717
 718     ar_op = ntohs(ah->ar_op);
 719     switch(ar_op) {
 720     case ARPOP_REQUEST:
 721         if (ah->ar_tip == ah->ar_sip) {
 722             /* Gratuitous ARP */
 723             arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
 724             return;
 725         }
 726
 727         if ((ah->ar_tip & slirp->vnetwork_mask.s_addr) ==
 728             slirp->vnetwork_addr.s_addr) {
 729             if (ah->ar_tip == slirp->vnameserver_addr.s_addr ||
 730                 ah->ar_tip == slirp->vhost_addr.s_addr)
 731                 goto arp_ok;
 732             for (ex_ptr = slirp->guestfwd_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
 733                 if (ex_ptr->ex_addr.s_addr == ah->ar_tip)
 734                     goto arp_ok;
 735             }
 736             return;
 737         arp_ok:
 738             memset(arp_reply, 0, sizeof(arp_reply));
 739
 740             arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
 741
 742             /* ARP request for alias/dns mac address */
 743             memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
 744             memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 4);
 745             memcpy(&reh->h_source[2], &ah->ar_tip, 4);
 746             reh->h_proto = htons(ETH_P_ARP);
 747
 748             rah->ar_hrd = htons(1);
 749             rah->ar_pro = htons(ETH_P_IP);
 750             rah->ar_hln = ETH_ALEN;
 751             rah->ar_pln = 4;
 752             rah->ar_op = htons(ARPOP_REPLY);
 753             memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
 754             rah->ar_sip = ah->ar_tip;
 755             memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
 756             rah->ar_tip = ah->ar_sip;
 757             slirp_send_packet_all(slirp, arp_reply, sizeof(arp_reply));
 758         }
 759         break;
 760     case ARPOP_REPLY:
 761         arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
 762         break;
 763     default:
 764         break;
 765     }
 766 }
 767
 768 void slirp_input(Slirp *slirp, const uint8_t *pkt, int pkt_len)
 769 {
 770     struct mbuf *m;
 771     int proto;
 772
 773     if (pkt_len < ETH_HLEN)
 774         return;
 775
 776     proto = (((uint16_t) pkt[12]) << 8) + pkt[13];
 777     switch(proto) {
 778     case ETH_P_ARP:
 779         arp_input(slirp, pkt, pkt_len);
 780         break;
 781     case ETH_P_IP:
 782     case ETH_P_IPV6:
 783         m = m_get(slirp);
 784         if (!m)
 785             return;
 786         /* Note: we add 2 to align the IP header on 4 bytes,
 787          * and add the margin for the tcpiphdr overhead  */
 788         if (M_FREEROOM(m) < pkt_len + TCPIPHDR_DELTA + 2) {
 789             m_inc(m, pkt_len + TCPIPHDR_DELTA + 2);
 790         }
 791         m->m_len = pkt_len + TCPIPHDR_DELTA + 2;
 792         memcpy(m->m_data + TCPIPHDR_DELTA + 2, pkt, pkt_len);
 793
 794         m->m_data += TCPIPHDR_DELTA + 2 + ETH_HLEN;
 795         m->m_len -= TCPIPHDR_DELTA + 2 + ETH_HLEN;
 796
 797         if (proto == ETH_P_IP) {
 798             ip_input(m);
 799         } else if (proto == ETH_P_IPV6) {
 800             ip6_input(m);
 801         }
 802         break;
 803
 804     case ETH_P_NCSI:
 805         ncsi_input(slirp, pkt, pkt_len);
 806         break;
 807
 808     default:
 809         break;
 810     }
 811 }
 812
 813 /* Prepare the IPv4 packet to be sent to the ethernet device. Returns 1 if no
 814  * packet should be sent, 0 if the packet must be re-queued, 2 if the packet
 815  * is ready to go.
 816  */
 817 static int if_encap4(Slirp *slirp, struct mbuf *ifm, struct ethhdr *eh,
 818         uint8_t ethaddr[ETH_ALEN])
 819 {
 820     const struct ip *iph = (const struct ip *)ifm->m_data;
 821
 822     if (iph->ip_dst.s_addr == 0) {
 823         /* 0.0.0.0 can not be a destination address, something went wrong,
 824          * avoid making it worse */
 825         return 1;
 826     }
 827     if (!arp_table_search(slirp, iph->ip_dst.s_addr, ethaddr)) {
 828         uint8_t arp_req[ETH_HLEN + sizeof(struct slirp_arphdr)];
 829         struct ethhdr *reh = (struct ethhdr *)arp_req;
 830         struct slirp_arphdr *rah = (struct slirp_arphdr *)(arp_req + ETH_HLEN);
 831
 832         if (!ifm->resolution_requested) {
 833             /* If the client addr is not known, send an ARP request */
 834             memset(reh->h_dest, 0xff, ETH_ALEN);
 835             memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 4);
 836             memcpy(&reh->h_source[2], &slirp->vhost_addr, 4);
 837             reh->h_proto = htons(ETH_P_ARP);
 838             rah->ar_hrd = htons(1);
 839             rah->ar_pro = htons(ETH_P_IP);
 840             rah->ar_hln = ETH_ALEN;
 841             rah->ar_pln = 4;
 842             rah->ar_op = htons(ARPOP_REQUEST);
 843
 844             /* source hw addr */
 845             memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN - 4);
 846             memcpy(&rah->ar_sha[2], &slirp->vhost_addr, 4);
 847
 848             /* source IP */
 849             rah->ar_sip = slirp->vhost_addr.s_addr;
 850
 851             /* target hw addr (none) */
 852             memset(rah->ar_tha, 0, ETH_ALEN);
 853
 854             /* target IP */
 855             rah->ar_tip = iph->ip_dst.s_addr;
 856             slirp->client_ipaddr = iph->ip_dst;
 857             slirp_send_packet_all(slirp, arp_req, sizeof(arp_req));
 858             ifm->resolution_requested = true;
 859
 860             /* Expire request and drop outgoing packet after 1 second */
 861             ifm->expiration_date =
 862                 slirp->cb->clock_get_ns(slirp->opaque) + 1000000000ULL;
 863         }
 864         return 0;
 865     } else {
 866         memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 4);
 867         /* XXX: not correct */
 868         memcpy(&eh->h_source[2], &slirp->vhost_addr, 4);
 869         eh->h_proto = htons(ETH_P_IP);
 870
 871         /* Send this */
 872         return 2;
 873     }
 874 }
 875
 876 /* Prepare the IPv6 packet to be sent to the ethernet device. Returns 1 if no
 877  * packet should be sent, 0 if the packet must be re-queued, 2 if the packet
 878  * is ready to go.
 879  */
 880 static int if_encap6(Slirp *slirp, struct mbuf *ifm, struct ethhdr *eh,
 881         uint8_t ethaddr[ETH_ALEN])
 882 {
 883     const struct ip6 *ip6h = mtod(ifm, const struct ip6 *);
 884     if (!ndp_table_search(slirp, ip6h->ip_dst, ethaddr)) {
 885         if (!ifm->resolution_requested) {
 886             ndp_send_ns(slirp, ip6h->ip_dst);
 887             ifm->resolution_requested = true;
 888             ifm->expiration_date = slirp->cb->clock_get_ns(slirp->opaque) + 1000000000ULL;
 889         }
 890         return 0;
 891     } else {
 892         eh->h_proto = htons(ETH_P_IPV6);
 893         in6_compute_ethaddr(ip6h->ip_src, eh->h_source);
 894
 895         /* Send this */
 896         return 2;
 897     }
 898 }
 899
 900 /* Output the IP packet to the ethernet device. Returns 0 if the packet must be
 901  * re-queued.
 902  */
 903 int if_encap(Slirp *slirp, struct mbuf *ifm)
 904 {
 905     uint8_t buf[1600];
 906     struct ethhdr *eh = (struct ethhdr *)buf;
 907     uint8_t ethaddr[ETH_ALEN];
 908     const struct ip *iph = (const struct ip *)ifm->m_data;
 909     int ret;
 910
 911     if (ifm->m_len + ETH_HLEN > sizeof(buf)) {
 912         return 1;
 913     }
 914
 915     switch (iph->ip_v) {
 916     case IPVERSION:
 917         ret = if_encap4(slirp, ifm, eh, ethaddr);
 918         if (ret < 2) {
 919             return ret;
 920         }
 921         break;
 922
 923     case IP6VERSION:
 924         ret = if_encap6(slirp, ifm, eh, ethaddr);
 925         if (ret < 2) {
 926             return ret;
 927         }
 928         break;
 929
 930     default:
 931         g_assert_not_reached();
 932         break;
 933     }
 934
 935     memcpy(eh->h_dest, ethaddr, ETH_ALEN);
 936     DEBUG_ARG("src = %02x:%02x:%02x:%02x:%02x:%02x",
 937               eh->h_source[0], eh->h_source[1], eh->h_source[2],
 938               eh->h_source[3], eh->h_source[4], eh->h_source[5]);
 939     DEBUG_ARG("dst = %02x:%02x:%02x:%02x:%02x:%02x",
 940               eh->h_dest[0], eh->h_dest[1], eh->h_dest[2],
 941               eh->h_dest[3], eh->h_dest[4], eh->h_dest[5]);
 942     memcpy(buf + sizeof(struct ethhdr), ifm->m_data, ifm->m_len);
 943     slirp_send_packet_all(slirp, buf, ifm->m_len + ETH_HLEN);
 944     return 1;
 945 }
 946
 947 /* Drop host forwarding rule, return 0 if found. */
 948 int slirp_remove_hostfwd(Slirp *slirp, int is_udp, struct in_addr host_addr,
 949                          int host_port)
 950 {
 951     struct socket *so;
 952     struct socket *head = (is_udp ? &slirp->udb : &slirp->tcb);
 953     struct sockaddr_in addr;
 954     int port = htons(host_port);
 955     socklen_t addr_len;
 956
 957     for (so = head->so_next; so != head; so = so->so_next) {
 958         addr_len = sizeof(addr);
 959         if ((so->so_state & SS_HOSTFWD) &&
 960             getsockname(so->s, (struct sockaddr *)&addr, &addr_len) == 0 &&
 961             addr.sin_addr.s_addr == host_addr.s_addr &&
 962             addr.sin_port == port) {
 963             so->slirp->cb->unregister_poll_fd(so->s, so->slirp->opaque);
 964             closesocket(so->s);
 965             sofree(so);
 966             return 0;
 967         }
 968     }
 969
 970     return -1;
 971 }
 972
 973 int slirp_add_hostfwd(Slirp *slirp, int is_udp, struct in_addr host_addr,
 974                       int host_port, struct in_addr guest_addr, int guest_port)
 975 {
 976     if (!guest_addr.s_addr) {
 977         guest_addr = slirp->vdhcp_startaddr;
 978     }
 979     if (is_udp) {
 980         if (!udp_listen(slirp, host_addr.s_addr, htons(host_port),
 981                         guest_addr.s_addr, htons(guest_port), SS_HOSTFWD))
 982             return -1;
 983     } else {
 984         if (!tcp_listen(slirp, host_addr.s_addr, htons(host_port),
 985                         guest_addr.s_addr, htons(guest_port), SS_HOSTFWD))
 986             return -1;
 987     }
 988     return 0;
 989 }
 990
 991 static bool
 992 check_guestfwd(Slirp *slirp, struct in_addr *guest_addr, int guest_port)
 993 {
 994     struct gfwd_list *tmp_ptr;
 995
 996     if (!guest_addr->s_addr) {
 997         guest_addr->s_addr = slirp->vnetwork_addr.s_addr |
 998             (htonl(0x0204) & ~slirp->vnetwork_mask.s_addr);
 999     }
1000     if ((guest_addr->s_addr & slirp->vnetwork_mask.s_addr) !=
1001         slirp->vnetwork_addr.s_addr ||
1002         guest_addr->s_addr == slirp->vhost_addr.s_addr ||
1003         guest_addr->s_addr == slirp->vnameserver_addr.s_addr) {
1004         return false;
1005     }
1006
1007     /* check if the port is "bound" */
1008     for (tmp_ptr = slirp->guestfwd_list; tmp_ptr; tmp_ptr = tmp_ptr->ex_next) {
1009         if (guest_port == tmp_ptr->ex_fport &&
1010             guest_addr->s_addr == tmp_ptr->ex_addr.s_addr)
1011             return false;
1012     }
1013
1014     return true;
1015 }
1016
1017 int slirp_add_exec(Slirp *slirp, const char *cmdline,
1018                    struct in_addr *guest_addr, int guest_port)
1019 {
1020     if (!check_guestfwd(slirp, guest_addr, guest_port)) {
1021         return -1;
1022     }
1023
1024     add_exec(&slirp->guestfwd_list, cmdline, *guest_addr, htons(guest_port));
1025     return 0;
1026 }
1027
1028 int slirp_add_guestfwd(Slirp *slirp, SlirpWriteCb write_cb, void *opaque,
1029                        struct in_addr *guest_addr, int guest_port)
1030 {
1031     if (!check_guestfwd(slirp, guest_addr, guest_port)) {
1032         return -1;
1033     }
1034
1035     add_guestfwd(&slirp->guestfwd_list, write_cb, opaque,
1036                  *guest_addr, htons(guest_port));
1037     return 0;
1038 }
1039
1040 ssize_t slirp_send(struct socket *so, const void *buf, size_t len, int flags)
1041 {
1042     if (so->s == -1 && so->guestfwd) {
1043         /* XXX this blocks entire thread. Rewrite to use
1044          * qemu_chr_fe_write and background I/O callbacks */
1045         so->guestfwd->write_cb(buf, len, so->guestfwd->opaque);
1046         return len;
1047     }
1048
1049     if (so->s == -1) {
1050         /*
1051          * This should in theory not happen but it is hard to be
1052          * sure because some code paths will end up with so->s == -1
1053          * on a failure but don't dispose of the struct socket.
1054          * Check specifically, so we don't pass -1 to send().
1055          */
1056         errno = EBADF;
1057         return -1;
1058     }
1059
1060     return send(so->s, buf, len, flags);
1061 }
1062
1063 struct socket *
1064 slirp_find_ctl_socket(Slirp *slirp, struct in_addr guest_addr, int guest_port)
1065 {
1066     struct socket *so;
1067
1068     for (so = slirp->tcb.so_next; so != &slirp->tcb; so = so->so_next) {
1069         if (so->so_faddr.s_addr == guest_addr.s_addr &&
1070             htons(so->so_fport) == guest_port) {
1071             return so;
1072         }
1073     }
1074     return NULL;
1075 }
1076
1077 size_t slirp_socket_can_recv(Slirp *slirp, struct in_addr guest_addr,
1078                              int guest_port)
1079 {
1080     struct iovec iov[2];
1081     struct socket *so;
1082
1083     so = slirp_find_ctl_socket(slirp, guest_addr, guest_port);
1084
1085     if (!so || so->so_state & SS_NOFDREF) {
1086         return 0;
1087     }
1088
1089     if (!CONN_CANFRCV(so) || so->so_snd.sb_cc >= (so->so_snd.sb_datalen/2)) {
1090         return 0;
1091     }
1092
1093     return sopreprbuf(so, iov, NULL);
1094 }
1095
1096 void slirp_socket_recv(Slirp *slirp, struct in_addr guest_addr, int guest_port,
1097                        const uint8_t *buf, int size)
1098 {
1099     int ret;
1100     struct socket *so = slirp_find_ctl_socket(slirp, guest_addr, guest_port);
1101
1102     if (!so)
1103         return;
1104
1105     ret = soreadbuf(so, (const char *)buf, size);
1106
1107     if (ret > 0)
1108         tcp_output(sototcpcb(so));
1109 }
1110
1111 void slirp_send_packet_all(Slirp *slirp, const void *buf, size_t len)
1112 {
1113     ssize_t ret = slirp->cb->send_packet(buf, len, slirp->opaque);
1114
1115     if (ret < 0) {
1116         g_critical("Failed to send packet, ret: %ld", (long) ret);
1117     } else if (ret < len) {
1118         DEBUG_ERROR("send_packet() didn't send all data: %ld < %lu",
1119                 (long) ret, (unsigned long) len);
1120     }
1121 }