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 QTAILQ_HEAD(, Slirp) slirp_instances =
  52     QTAILQ_HEAD_INITIALIZER(slirp_instances);
  53
  54 static struct in_addr dns_addr;
  55 #ifndef _WIN32
  56 static struct in6_addr dns6_addr;
  57 #endif
  58 static unsigned dns_addr_time;
  59 #ifndef _WIN32
  60 static unsigned dns6_addr_time;
  61 #endif
  62
  63 #define TIMEOUT_FAST 2  /* milliseconds */
  64 #define TIMEOUT_SLOW 499  /* milliseconds */
  65 /* for the aging of certain requests like DNS */
  66 #define TIMEOUT_DEFAULT 1000  /* milliseconds */
  67
  68 #ifdef _WIN32
  69
  70 int get_dns_addr(struct in_addr *pdns_addr)
  71 {
  72     FIXED_INFO *FixedInfo=NULL;
  73     ULONG    BufLen;
  74     DWORD    ret;
  75     IP_ADDR_STRING *pIPAddr;
  76     struct in_addr tmp_addr;
  77
  78     if (dns_addr.s_addr != 0 && (curtime - dns_addr_time) < TIMEOUT_DEFAULT) {
  79         *pdns_addr = dns_addr;
  80         return 0;
  81     }
  82
  83     FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
  84     BufLen = sizeof(FIXED_INFO);
  85
  86     if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen)) {
  87         if (FixedInfo) {
  88             GlobalFree(FixedInfo);
  89             FixedInfo = NULL;
  90         }
  91         FixedInfo = GlobalAlloc(GPTR, BufLen);
  92     }
  93
  94     if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS) {
  95         printf("GetNetworkParams failed. ret = %08x\n", (unsigned)ret );
  96         if (FixedInfo) {
  97             GlobalFree(FixedInfo);
  98             FixedInfo = NULL;
  99         }
 100         return -1;
 101     }
 102
 103     pIPAddr = &(FixedInfo->DnsServerList);
 104     inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
 105     *pdns_addr = tmp_addr;
 106     dns_addr = tmp_addr;
 107     dns_addr_time = curtime;
 108     if (FixedInfo) {
 109         GlobalFree(FixedInfo);
 110         FixedInfo = NULL;
 111     }
 112     return 0;
 113 }
 114
 115 int get_dns6_addr(struct in6_addr *pdns6_addr, uint32_t *scope_id)
 116 {
 117     return -1;
 118 }
 119
 120 static void winsock_cleanup(void)
 121 {
 122     WSACleanup();
 123 }
 124
 125 #else
 126
 127 static int get_dns_addr_cached(void *pdns_addr, void *cached_addr,
 128                                socklen_t addrlen,
 129                                struct stat *cached_stat, unsigned *cached_time)
 130 {
 131     struct stat old_stat;
 132     if (curtime - *cached_time < TIMEOUT_DEFAULT) {
 133         memcpy(pdns_addr, cached_addr, addrlen);
 134         return 0;
 135     }
 136     old_stat = *cached_stat;
 137     if (stat("/etc/resolv.conf", cached_stat) != 0) {
 138         return -1;
 139     }
 140     if (cached_stat->st_dev == old_stat.st_dev
 141         && cached_stat->st_ino == old_stat.st_ino
 142         && cached_stat->st_size == old_stat.st_size
 143         && cached_stat->st_mtime == old_stat.st_mtime) {
 144         memcpy(pdns_addr, cached_addr, addrlen);
 145         return 0;
 146     }
 147     return 1;
 148 }
 149
 150 static int get_dns_addr_resolv_conf(int af, void *pdns_addr, void *cached_addr,
 151                                     socklen_t addrlen, uint32_t *scope_id,
 152                                     unsigned *cached_time)
 153 {
 154     char buff[512];
 155     char buff2[257];
 156     FILE *f;
 157     int found = 0;
 158     void *tmp_addr = alloca(addrlen);
 159     unsigned if_index;
 160
 161     f = fopen("/etc/resolv.conf", "r");
 162     if (!f)
 163         return -1;
 164
 165     DEBUG_MISC("IP address of your DNS(s):");
 166     while (fgets(buff, 512, f) != NULL) {
 167         if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1) {
 168             char *c = strchr(buff2, '%');
 169             if (c) {
 170                 if_index = if_nametoindex(c + 1);
 171                 *c = '\0';
 172             } else {
 173                 if_index = 0;
 174             }
 175
 176             if (!inet_pton(af, buff2, tmp_addr)) {
 177                 continue;
 178             }
 179             /* If it's the first one, set it to dns_addr */
 180             if (!found) {
 181                 memcpy(pdns_addr, tmp_addr, addrlen);
 182                 memcpy(cached_addr, tmp_addr, addrlen);
 183                 if (scope_id) {
 184                     *scope_id = if_index;
 185                 }
 186                 *cached_time = curtime;
 187             }
 188
 189             if (++found > 3) {
 190                 DEBUG_MISC("  (more)");
 191                 break;
 192             } else if (slirp_debug & DBG_MISC) {
 193                 char s[INET6_ADDRSTRLEN];
 194                 const char *res = inet_ntop(af, tmp_addr, s, sizeof(s));
 195                 if (!res) {
 196                     res = "  (string conversion error)";
 197                 }
 198                 DEBUG_MISC("  %s", res);
 199             }
 200         }
 201     }
 202     fclose(f);
 203     if (!found)
 204         return -1;
 205     return 0;
 206 }
 207
 208 int get_dns_addr(struct in_addr *pdns_addr)
 209 {
 210     static struct stat dns_addr_stat;
 211
 212     if (dns_addr.s_addr != 0) {
 213         int ret;
 214         ret = get_dns_addr_cached(pdns_addr, &dns_addr, sizeof(dns_addr),
 215                                   &dns_addr_stat, &dns_addr_time);
 216         if (ret <= 0) {
 217             return ret;
 218         }
 219     }
 220     return get_dns_addr_resolv_conf(AF_INET, pdns_addr, &dns_addr,
 221                                     sizeof(dns_addr), NULL, &dns_addr_time);
 222 }
 223
 224 int get_dns6_addr(struct in6_addr *pdns6_addr, uint32_t *scope_id)
 225 {
 226     static struct stat dns6_addr_stat;
 227
 228     if (!in6_zero(&dns6_addr)) {
 229         int ret;
 230         ret = get_dns_addr_cached(pdns6_addr, &dns6_addr, sizeof(dns6_addr),
 231                                   &dns6_addr_stat, &dns6_addr_time);
 232         if (ret <= 0) {
 233             return ret;
 234         }
 235     }
 236     return get_dns_addr_resolv_conf(AF_INET6, pdns6_addr, &dns6_addr,
 237                                     sizeof(dns6_addr),
 238                                     scope_id, &dns6_addr_time);
 239 }
 240
 241 #endif
 242
 243 static void slirp_init_once(void)
 244 {
 245     static int initialized;
 246     const char *debug;
 247 #ifdef _WIN32
 248     WSADATA Data;
 249 #endif
 250
 251     if (initialized) {
 252         return;
 253     }
 254     initialized = 1;
 255
 256 #ifdef _WIN32
 257     WSAStartup(MAKEWORD(2,0), &Data);
 258     atexit(winsock_cleanup);
 259 #endif
 260
 261     loopback_addr.s_addr = htonl(INADDR_LOOPBACK);
 262     loopback_mask = htonl(IN_CLASSA_NET);
 263
 264     debug = g_getenv("SLIRP_DEBUG");
 265     if (debug) {
 266         const GDebugKey keys[] = {
 267             { "call", DBG_CALL },
 268             { "misc", DBG_MISC },
 269             { "error", DBG_ERROR },
 270             { "tftp", DBG_TFTP },
 271         };
 272         slirp_debug = g_parse_debug_string(debug, keys, G_N_ELEMENTS(keys));
 273     }
 274
 275
 276 }
 277
 278 Slirp *slirp_init(int restricted, bool in_enabled, struct in_addr vnetwork,
 279                   struct in_addr vnetmask, struct in_addr vhost,
 280                   bool in6_enabled,
 281                   struct in6_addr vprefix_addr6, uint8_t vprefix_len,
 282                   struct in6_addr vhost6, const char *vhostname,
 283                   const char *tftp_server_name,
 284                   const char *tftp_path, const char *bootfile,
 285                   struct in_addr vdhcp_start, struct in_addr vnameserver,
 286                   struct in6_addr vnameserver6, const char **vdnssearch,
 287                   const char *vdomainname,
 288                   const SlirpCb *callbacks,
 289                   void *opaque)
 290 {
 291     Slirp *slirp = g_malloc0(sizeof(Slirp));
 292
 293     slirp_init_once();
 294
 295     slirp->cb = callbacks;
 296     slirp->grand = g_rand_new();
 297     slirp->restricted = restricted;
 298
 299     slirp->in_enabled = in_enabled;
 300     slirp->in6_enabled = in6_enabled;
 301
 302     if_init(slirp);
 303     ip_init(slirp);
 304     ip6_init(slirp);
 305
 306     /* Initialise mbufs *after* setting the MTU */
 307     m_init(slirp);
 308
 309     slirp->vnetwork_addr = vnetwork;
 310     slirp->vnetwork_mask = vnetmask;
 311     slirp->vhost_addr = vhost;
 312     slirp->vprefix_addr6 = vprefix_addr6;
 313     slirp->vprefix_len = vprefix_len;
 314     slirp->vhost_addr6 = vhost6;
 315     if (vhostname) {
 316         slirp_pstrcpy(slirp->client_hostname, sizeof(slirp->client_hostname),
 317                       vhostname);
 318     }
 319     slirp->tftp_prefix = g_strdup(tftp_path);
 320     slirp->bootp_filename = g_strdup(bootfile);
 321     slirp->vdomainname = g_strdup(vdomainname);
 322     slirp->vdhcp_startaddr = vdhcp_start;
 323     slirp->vnameserver_addr = vnameserver;
 324     slirp->vnameserver_addr6 = vnameserver6;
 325     slirp->tftp_server_name = g_strdup(tftp_server_name);
 326
 327     if (vdnssearch) {
 328         translate_dnssearch(slirp, vdnssearch);
 329     }
 330
 331     slirp->opaque = opaque;
 332
 333 #ifdef WITH_QEMU
 334     slirp_state_register(slirp);
 335 #endif
 336     QTAILQ_INSERT_TAIL(&slirp_instances, slirp, entry);
 337
 338     return slirp;
 339 }
 340
 341 void slirp_cleanup(Slirp *slirp)
 342 {
 343     struct gfwd_list *e, *next;
 344
 345     for (e = slirp->guestfwd_list; e; e = next) {
 346         next = e->ex_next;
 347         g_free(e->ex_exec);
 348         g_free(e);
 349     }
 350
 351     QTAILQ_REMOVE(&slirp_instances, slirp, entry);
 352 #ifdef WITH_QEMU
 353     slirp_state_unregister(slirp);
 354 #endif
 355     ip_cleanup(slirp);
 356     ip6_cleanup(slirp);
 357     m_cleanup(slirp);
 358
 359     g_rand_free(slirp->grand);
 360
 361     g_free(slirp->vdnssearch);
 362     g_free(slirp->tftp_prefix);
 363     g_free(slirp->bootp_filename);
 364     g_free(slirp->vdomainname);
 365     g_free(slirp);
 366 }
 367
 368 #define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
 369 #define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
 370
 371 static void slirp_update_timeout(uint32_t *timeout)
 372 {
 373     Slirp *slirp;
 374     uint32_t t;
 375
 376     if (*timeout <= TIMEOUT_FAST) {
 377         return;
 378     }
 379
 380     t = MIN(1000, *timeout);
 381
 382     /* If we have tcp timeout with slirp, then we will fill @timeout with
 383      * more precise value.
 384      */
 385     QTAILQ_FOREACH(slirp, &slirp_instances, entry) {
 386         if (slirp->time_fasttimo) {
 387             *timeout = TIMEOUT_FAST;
 388             return;
 389         }
 390         if (slirp->do_slowtimo) {
 391             t = MIN(TIMEOUT_SLOW, t);
 392         }
 393     }
 394     *timeout = t;
 395 }
 396
 397 void slirp_pollfds_fill(GArray *pollfds, uint32_t *timeout)
 398 {
 399     Slirp *slirp;
 400     struct socket *so, *so_next;
 401
 402     if (QTAILQ_EMPTY(&slirp_instances)) {
 403         return;
 404     }
 405
 406     /*
 407      * First, TCP sockets
 408      */
 409
 410     QTAILQ_FOREACH(slirp, &slirp_instances, entry) {
 411         /*
 412          * *_slowtimo needs calling if there are IP fragments
 413          * in the fragment queue, or there are TCP connections active
 414          */
 415         slirp->do_slowtimo = ((slirp->tcb.so_next != &slirp->tcb) ||
 416                 (&slirp->ipq.ip_link != slirp->ipq.ip_link.next));
 417
 418         for (so = slirp->tcb.so_next; so != &slirp->tcb;
 419                 so = so_next) {
 420             int events = 0;
 421
 422             so_next = so->so_next;
 423
 424             so->pollfds_idx = -1;
 425
 426             /*
 427              * See if we need a tcp_fasttimo
 428              */
 429             if (slirp->time_fasttimo == 0 &&
 430                 so->so_tcpcb->t_flags & TF_DELACK) {
 431                 slirp->time_fasttimo = curtime; /* Flag when want a fasttimo */
 432             }
 433
 434             /*
 435              * NOFDREF can include still connecting to local-host,
 436              * newly socreated() sockets etc. Don't want to select these.
 437              */
 438             if (so->so_state & SS_NOFDREF || so->s == -1) {
 439                 continue;
 440             }
 441
 442             /*
 443              * Set for reading sockets which are accepting
 444              */
 445             if (so->so_state & SS_FACCEPTCONN) {
 446                 GPollFD pfd = {
 447                     .fd = so->s,
 448                     .events = G_IO_IN | G_IO_HUP | G_IO_ERR,
 449                 };
 450                 so->pollfds_idx = pollfds->len;
 451                 g_array_append_val(pollfds, pfd);
 452                 continue;
 453             }
 454
 455             /*
 456              * Set for writing sockets which are connecting
 457              */
 458             if (so->so_state & SS_ISFCONNECTING) {
 459                 GPollFD pfd = {
 460                     .fd = so->s,
 461                     .events = G_IO_OUT | G_IO_ERR,
 462                 };
 463                 so->pollfds_idx = pollfds->len;
 464                 g_array_append_val(pollfds, pfd);
 465                 continue;
 466             }
 467
 468             /*
 469              * Set for writing if we are connected, can send more, and
 470              * we have something to send
 471              */
 472             if (CONN_CANFSEND(so) && so->so_rcv.sb_cc) {
 473                 events |= G_IO_OUT | G_IO_ERR;
 474             }
 475
 476             /*
 477              * Set for reading (and urgent data) if we are connected, can
 478              * receive more, and we have room for it XXX /2 ?
 479              */
 480             if (CONN_CANFRCV(so) &&
 481                 (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2))) {
 482                 events |= G_IO_IN | G_IO_HUP | G_IO_ERR | G_IO_PRI;
 483             }
 484
 485             if (events) {
 486                 GPollFD pfd = {
 487                     .fd = so->s,
 488                     .events = events,
 489                 };
 490                 so->pollfds_idx = pollfds->len;
 491                 g_array_append_val(pollfds, pfd);
 492             }
 493         }
 494
 495         /*
 496          * UDP sockets
 497          */
 498         for (so = slirp->udb.so_next; so != &slirp->udb;
 499                 so = so_next) {
 500             so_next = so->so_next;
 501
 502             so->pollfds_idx = -1;
 503
 504             /*
 505              * See if it's timed out
 506              */
 507             if (so->so_expire) {
 508                 if (so->so_expire <= curtime) {
 509                     udp_detach(so);
 510                     continue;
 511                 } else {
 512                     slirp->do_slowtimo = true; /* Let socket expire */
 513                 }
 514             }
 515
 516             /*
 517              * When UDP packets are received from over the
 518              * link, they're sendto()'d straight away, so
 519              * no need for setting for writing
 520              * Limit the number of packets queued by this session
 521              * to 4.  Note that even though we try and limit this
 522              * to 4 packets, the session could have more queued
 523              * if the packets needed to be fragmented
 524              * (XXX <= 4 ?)
 525              */
 526             if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4) {
 527                 GPollFD pfd = {
 528                     .fd = so->s,
 529                     .events = G_IO_IN | G_IO_HUP | G_IO_ERR,
 530                 };
 531                 so->pollfds_idx = pollfds->len;
 532                 g_array_append_val(pollfds, pfd);
 533             }
 534         }
 535
 536         /*
 537          * ICMP sockets
 538          */
 539         for (so = slirp->icmp.so_next; so != &slirp->icmp;
 540                 so = so_next) {
 541             so_next = so->so_next;
 542
 543             so->pollfds_idx = -1;
 544
 545             /*
 546              * See if it's timed out
 547              */
 548             if (so->so_expire) {
 549                 if (so->so_expire <= curtime) {
 550                     icmp_detach(so);
 551                     continue;
 552                 } else {
 553                     slirp->do_slowtimo = true; /* Let socket expire */
 554                 }
 555             }
 556
 557             if (so->so_state & SS_ISFCONNECTED) {
 558                 GPollFD pfd = {
 559                     .fd = so->s,
 560                     .events = G_IO_IN | G_IO_HUP | G_IO_ERR,
 561                 };
 562                 so->pollfds_idx = pollfds->len;
 563                 g_array_append_val(pollfds, pfd);
 564             }
 565         }
 566     }
 567     slirp_update_timeout(timeout);
 568 }
 569
 570 void slirp_pollfds_poll(GArray *pollfds, int select_error)
 571 {
 572     Slirp *slirp = QTAILQ_FIRST(&slirp_instances);
 573     struct socket *so, *so_next;
 574     int ret;
 575
 576     if (!slirp) {
 577         return;
 578     }
 579
 580     curtime = slirp->cb->clock_get_ns() / SCALE_MS;
 581
 582     QTAILQ_FOREACH(slirp, &slirp_instances, entry) {
 583         /*
 584          * See if anything has timed out
 585          */
 586         if (slirp->time_fasttimo &&
 587             ((curtime - slirp->time_fasttimo) >= TIMEOUT_FAST)) {
 588             tcp_fasttimo(slirp);
 589             slirp->time_fasttimo = 0;
 590         }
 591         if (slirp->do_slowtimo &&
 592             ((curtime - slirp->last_slowtimo) >= TIMEOUT_SLOW)) {
 593             ip_slowtimo(slirp);
 594             tcp_slowtimo(slirp);
 595             slirp->last_slowtimo = curtime;
 596         }
 597
 598         /*
 599          * Check sockets
 600          */
 601         if (!select_error) {
 602             /*
 603              * Check TCP sockets
 604              */
 605             for (so = slirp->tcb.so_next; so != &slirp->tcb;
 606                     so = so_next) {
 607                 int revents;
 608
 609                 so_next = so->so_next;
 610
 611                 revents = 0;
 612                 if (so->pollfds_idx != -1) {
 613                     revents = g_array_index(pollfds, GPollFD,
 614                                             so->pollfds_idx).revents;
 615                 }
 616
 617                 if (so->so_state & SS_NOFDREF || so->s == -1) {
 618                     continue;
 619                 }
 620
 621                 /*
 622                  * Check for URG data
 623                  * This will soread as well, so no need to
 624                  * test for G_IO_IN below if this succeeds
 625                  */
 626                 if (revents & G_IO_PRI) {
 627                     ret = sorecvoob(so);
 628                     if (ret < 0) {
 629                         /* Socket error might have resulted in the socket being
 630                          * removed, do not try to do anything more with it. */
 631                         continue;
 632                     }
 633                 }
 634                 /*
 635                  * Check sockets for reading
 636                  */
 637                 else if (revents & (G_IO_IN | G_IO_HUP | G_IO_ERR)) {
 638                     /*
 639                      * Check for incoming connections
 640                      */
 641                     if (so->so_state & SS_FACCEPTCONN) {
 642                         tcp_connect(so);
 643                         continue;
 644                     } /* else */
 645                     ret = soread(so);
 646
 647                     /* Output it if we read something */
 648                     if (ret > 0) {
 649                         tcp_output(sototcpcb(so));
 650                     }
 651                     if (ret < 0) {
 652                         /* Socket error might have resulted in the socket being
 653                          * removed, do not try to do anything more with it. */
 654                         continue;
 655                     }
 656                 }
 657
 658                 /*
 659                  * Check sockets for writing
 660                  */
 661                 if (!(so->so_state & SS_NOFDREF) &&
 662                         (revents & (G_IO_OUT | G_IO_ERR))) {
 663                     /*
 664                      * Check for non-blocking, still-connecting sockets
 665                      */
 666                     if (so->so_state & SS_ISFCONNECTING) {
 667                         /* Connected */
 668                         so->so_state &= ~SS_ISFCONNECTING;
 669
 670                         ret = send(so->s, (const void *) &ret, 0, 0);
 671                         if (ret < 0) {
 672                             /* XXXXX Must fix, zero bytes is a NOP */
 673                             if (errno == EAGAIN || errno == EWOULDBLOCK ||
 674                                 errno == EINPROGRESS || errno == ENOTCONN) {
 675                                 continue;
 676                             }
 677
 678                             /* else failed */
 679                             so->so_state &= SS_PERSISTENT_MASK;
 680                             so->so_state |= SS_NOFDREF;
 681                         }
 682                         /* else so->so_state &= ~SS_ISFCONNECTING; */
 683
 684                         /*
 685                          * Continue tcp_input
 686                          */
 687                         tcp_input((struct mbuf *)NULL, sizeof(struct ip), so,
 688                                   so->so_ffamily);
 689                         /* continue; */
 690                     } else {
 691                         ret = sowrite(so);
 692                         if (ret > 0) {
 693                             /* Call tcp_output in case we need to send a window
 694                              * update to the guest, otherwise it will be stuck
 695                              * until it sends a window probe. */
 696                             tcp_output(sototcpcb(so));
 697                         }
 698                     }
 699                 }
 700             }
 701
 702             /*
 703              * Now UDP sockets.
 704              * Incoming packets are sent straight away, they're not buffered.
 705              * Incoming UDP data isn't buffered either.
 706              */
 707             for (so = slirp->udb.so_next; so != &slirp->udb;
 708                     so = so_next) {
 709                 int revents;
 710
 711                 so_next = so->so_next;
 712
 713                 revents = 0;
 714                 if (so->pollfds_idx != -1) {
 715                     revents = g_array_index(pollfds, GPollFD,
 716                             so->pollfds_idx).revents;
 717                 }
 718
 719                 if (so->s != -1 &&
 720                     (revents & (G_IO_IN | G_IO_HUP | G_IO_ERR))) {
 721                     sorecvfrom(so);
 722                 }
 723             }
 724
 725             /*
 726              * Check incoming ICMP relies.
 727              */
 728             for (so = slirp->icmp.so_next; so != &slirp->icmp;
 729                     so = so_next) {
 730                     int revents;
 731
 732                     so_next = so->so_next;
 733
 734                     revents = 0;
 735                     if (so->pollfds_idx != -1) {
 736                         revents = g_array_index(pollfds, GPollFD,
 737                                                 so->pollfds_idx).revents;
 738                     }
 739
 740                     if (so->s != -1 &&
 741                         (revents & (G_IO_IN | G_IO_HUP | G_IO_ERR))) {
 742                     icmp_receive(so);
 743                 }
 744             }
 745         }
 746
 747         if_start(slirp);
 748     }
 749 }
 750
 751 static void arp_input(Slirp *slirp, const uint8_t *pkt, int pkt_len)
 752 {
 753     struct slirp_arphdr *ah = (struct slirp_arphdr *)(pkt + ETH_HLEN);
 754     uint8_t arp_reply[MAX(ETH_HLEN + sizeof(struct slirp_arphdr), 64)];
 755     struct ethhdr *reh = (struct ethhdr *)arp_reply;
 756     struct slirp_arphdr *rah = (struct slirp_arphdr *)(arp_reply + ETH_HLEN);
 757     int ar_op;
 758     struct gfwd_list *ex_ptr;
 759
 760     if (!slirp->in_enabled) {
 761         return;
 762     }
 763
 764     ar_op = ntohs(ah->ar_op);
 765     switch(ar_op) {
 766     case ARPOP_REQUEST:
 767         if (ah->ar_tip == ah->ar_sip) {
 768             /* Gratuitous ARP */
 769             arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
 770             return;
 771         }
 772
 773         if ((ah->ar_tip & slirp->vnetwork_mask.s_addr) ==
 774             slirp->vnetwork_addr.s_addr) {
 775             if (ah->ar_tip == slirp->vnameserver_addr.s_addr ||
 776                 ah->ar_tip == slirp->vhost_addr.s_addr)
 777                 goto arp_ok;
 778             for (ex_ptr = slirp->guestfwd_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
 779                 if (ex_ptr->ex_addr.s_addr == ah->ar_tip)
 780                     goto arp_ok;
 781             }
 782             return;
 783         arp_ok:
 784             memset(arp_reply, 0, sizeof(arp_reply));
 785
 786             arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
 787
 788             /* ARP request for alias/dns mac address */
 789             memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
 790             memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 4);
 791             memcpy(&reh->h_source[2], &ah->ar_tip, 4);
 792             reh->h_proto = htons(ETH_P_ARP);
 793
 794             rah->ar_hrd = htons(1);
 795             rah->ar_pro = htons(ETH_P_IP);
 796             rah->ar_hln = ETH_ALEN;
 797             rah->ar_pln = 4;
 798             rah->ar_op = htons(ARPOP_REPLY);
 799             memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
 800             rah->ar_sip = ah->ar_tip;
 801             memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
 802             rah->ar_tip = ah->ar_sip;
 803             slirp_send_packet_all(slirp, arp_reply, sizeof(arp_reply));
 804         }
 805         break;
 806     case ARPOP_REPLY:
 807         arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
 808         break;
 809     default:
 810         break;
 811     }
 812 }
 813
 814 void slirp_input(Slirp *slirp, const uint8_t *pkt, int pkt_len)
 815 {
 816     struct mbuf *m;
 817     int proto;
 818
 819     if (pkt_len < ETH_HLEN)
 820         return;
 821
 822     proto = (((uint16_t) pkt[12]) << 8) + pkt[13];
 823     switch(proto) {
 824     case ETH_P_ARP:
 825         arp_input(slirp, pkt, pkt_len);
 826         break;
 827     case ETH_P_IP:
 828     case ETH_P_IPV6:
 829         m = m_get(slirp);
 830         if (!m)
 831             return;
 832         /* Note: we add 2 to align the IP header on 4 bytes,
 833          * and add the margin for the tcpiphdr overhead  */
 834         if (M_FREEROOM(m) < pkt_len + TCPIPHDR_DELTA + 2) {
 835             m_inc(m, pkt_len + TCPIPHDR_DELTA + 2);
 836         }
 837         m->m_len = pkt_len + TCPIPHDR_DELTA + 2;
 838         memcpy(m->m_data + TCPIPHDR_DELTA + 2, pkt, pkt_len);
 839
 840         m->m_data += TCPIPHDR_DELTA + 2 + ETH_HLEN;
 841         m->m_len -= TCPIPHDR_DELTA + 2 + ETH_HLEN;
 842
 843         if (proto == ETH_P_IP) {
 844             ip_input(m);
 845         } else if (proto == ETH_P_IPV6) {
 846             ip6_input(m);
 847         }
 848         break;
 849
 850     case ETH_P_NCSI:
 851         ncsi_input(slirp, pkt, pkt_len);
 852         break;
 853
 854     default:
 855         break;
 856     }
 857 }
 858
 859 /* Prepare the IPv4 packet to be sent to the ethernet device. Returns 1 if no
 860  * packet should be sent, 0 if the packet must be re-queued, 2 if the packet
 861  * is ready to go.
 862  */
 863 static int if_encap4(Slirp *slirp, struct mbuf *ifm, struct ethhdr *eh,
 864         uint8_t ethaddr[ETH_ALEN])
 865 {
 866     const struct ip *iph = (const struct ip *)ifm->m_data;
 867
 868     if (iph->ip_dst.s_addr == 0) {
 869         /* 0.0.0.0 can not be a destination address, something went wrong,
 870          * avoid making it worse */
 871         return 1;
 872     }
 873     if (!arp_table_search(slirp, iph->ip_dst.s_addr, ethaddr)) {
 874         uint8_t arp_req[ETH_HLEN + sizeof(struct slirp_arphdr)];
 875         struct ethhdr *reh = (struct ethhdr *)arp_req;
 876         struct slirp_arphdr *rah = (struct slirp_arphdr *)(arp_req + ETH_HLEN);
 877
 878         if (!ifm->resolution_requested) {
 879             /* If the client addr is not known, send an ARP request */
 880             memset(reh->h_dest, 0xff, ETH_ALEN);
 881             memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 4);
 882             memcpy(&reh->h_source[2], &slirp->vhost_addr, 4);
 883             reh->h_proto = htons(ETH_P_ARP);
 884             rah->ar_hrd = htons(1);
 885             rah->ar_pro = htons(ETH_P_IP);
 886             rah->ar_hln = ETH_ALEN;
 887             rah->ar_pln = 4;
 888             rah->ar_op = htons(ARPOP_REQUEST);
 889
 890             /* source hw addr */
 891             memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN - 4);
 892             memcpy(&rah->ar_sha[2], &slirp->vhost_addr, 4);
 893
 894             /* source IP */
 895             rah->ar_sip = slirp->vhost_addr.s_addr;
 896
 897             /* target hw addr (none) */
 898             memset(rah->ar_tha, 0, ETH_ALEN);
 899
 900             /* target IP */
 901             rah->ar_tip = iph->ip_dst.s_addr;
 902             slirp->client_ipaddr = iph->ip_dst;
 903             slirp_send_packet_all(slirp, arp_req, sizeof(arp_req));
 904             ifm->resolution_requested = true;
 905
 906             /* Expire request and drop outgoing packet after 1 second */
 907             ifm->expiration_date = slirp->cb->clock_get_ns() + 1000000000ULL;
 908         }
 909         return 0;
 910     } else {
 911         memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 4);
 912         /* XXX: not correct */
 913         memcpy(&eh->h_source[2], &slirp->vhost_addr, 4);
 914         eh->h_proto = htons(ETH_P_IP);
 915
 916         /* Send this */
 917         return 2;
 918     }
 919 }
 920
 921 /* Prepare the IPv6 packet to be sent to the ethernet device. Returns 1 if no
 922  * packet should be sent, 0 if the packet must be re-queued, 2 if the packet
 923  * is ready to go.
 924  */
 925 static int if_encap6(Slirp *slirp, struct mbuf *ifm, struct ethhdr *eh,
 926         uint8_t ethaddr[ETH_ALEN])
 927 {
 928     const struct ip6 *ip6h = mtod(ifm, const struct ip6 *);
 929     if (!ndp_table_search(slirp, ip6h->ip_dst, ethaddr)) {
 930         if (!ifm->resolution_requested) {
 931             ndp_send_ns(slirp, ip6h->ip_dst);
 932             ifm->resolution_requested = true;
 933             ifm->expiration_date = slirp->cb->clock_get_ns() + 1000000000ULL;
 934         }
 935         return 0;
 936     } else {
 937         eh->h_proto = htons(ETH_P_IPV6);
 938         in6_compute_ethaddr(ip6h->ip_src, eh->h_source);
 939
 940         /* Send this */
 941         return 2;
 942     }
 943 }
 944
 945 /* Output the IP packet to the ethernet device. Returns 0 if the packet must be
 946  * re-queued.
 947  */
 948 int if_encap(Slirp *slirp, struct mbuf *ifm)
 949 {
 950     uint8_t buf[1600];
 951     struct ethhdr *eh = (struct ethhdr *)buf;
 952     uint8_t ethaddr[ETH_ALEN];
 953     const struct ip *iph = (const struct ip *)ifm->m_data;
 954     int ret;
 955
 956     if (ifm->m_len + ETH_HLEN > sizeof(buf)) {
 957         return 1;
 958     }
 959
 960     switch (iph->ip_v) {
 961     case IPVERSION:
 962         ret = if_encap4(slirp, ifm, eh, ethaddr);
 963         if (ret < 2) {
 964             return ret;
 965         }
 966         break;
 967
 968     case IP6VERSION:
 969         ret = if_encap6(slirp, ifm, eh, ethaddr);
 970         if (ret < 2) {
 971             return ret;
 972         }
 973         break;
 974
 975     default:
 976         g_assert_not_reached();
 977         break;
 978     }
 979
 980     memcpy(eh->h_dest, ethaddr, ETH_ALEN);
 981     DEBUG_ARG("src = %02x:%02x:%02x:%02x:%02x:%02x",
 982               eh->h_source[0], eh->h_source[1], eh->h_source[2],
 983               eh->h_source[3], eh->h_source[4], eh->h_source[5]);
 984     DEBUG_ARG("dst = %02x:%02x:%02x:%02x:%02x:%02x",
 985               eh->h_dest[0], eh->h_dest[1], eh->h_dest[2],
 986               eh->h_dest[3], eh->h_dest[4], eh->h_dest[5]);
 987     memcpy(buf + sizeof(struct ethhdr), ifm->m_data, ifm->m_len);
 988     slirp_send_packet_all(slirp, buf, ifm->m_len + ETH_HLEN);
 989     return 1;
 990 }
 991
 992 /* Drop host forwarding rule, return 0 if found. */
 993 int slirp_remove_hostfwd(Slirp *slirp, int is_udp, struct in_addr host_addr,
 994                          int host_port)
 995 {
 996     struct socket *so;
 997     struct socket *head = (is_udp ? &slirp->udb : &slirp->tcb);
 998     struct sockaddr_in addr;
 999     int port = htons(host_port);
1000     socklen_t addr_len;
1001
1002     for (so = head->so_next; so != head; so = so->so_next) {
1003         addr_len = sizeof(addr);
1004         if ((so->so_state & SS_HOSTFWD) &&
1005             getsockname(so->s, (struct sockaddr *)&addr, &addr_len) == 0 &&
1006             addr.sin_addr.s_addr == host_addr.s_addr &&
1007             addr.sin_port == port) {
1008             so->slirp->cb->unregister_poll_fd(so->s);
1009             slirp_closesocket(so->s);
1010             sofree(so);
1011             return 0;
1012         }
1013     }
1014
1015     return -1;
1016 }
1017
1018 int slirp_add_hostfwd(Slirp *slirp, int is_udp, struct in_addr host_addr,
1019                       int host_port, struct in_addr guest_addr, int guest_port)
1020 {
1021     if (!guest_addr.s_addr) {
1022         guest_addr = slirp->vdhcp_startaddr;
1023     }
1024     if (is_udp) {
1025         if (!udp_listen(slirp, host_addr.s_addr, htons(host_port),
1026                         guest_addr.s_addr, htons(guest_port), SS_HOSTFWD))
1027             return -1;
1028     } else {
1029         if (!tcp_listen(slirp, host_addr.s_addr, htons(host_port),
1030                         guest_addr.s_addr, htons(guest_port), SS_HOSTFWD))
1031             return -1;
1032     }
1033     return 0;
1034 }
1035
1036 static bool
1037 check_guestfwd(Slirp *slirp, struct in_addr *guest_addr, int guest_port)
1038 {
1039     struct gfwd_list *tmp_ptr;
1040
1041     if (!guest_addr->s_addr) {
1042         guest_addr->s_addr = slirp->vnetwork_addr.s_addr |
1043             (htonl(0x0204) & ~slirp->vnetwork_mask.s_addr);
1044     }
1045     if ((guest_addr->s_addr & slirp->vnetwork_mask.s_addr) !=
1046         slirp->vnetwork_addr.s_addr ||
1047         guest_addr->s_addr == slirp->vhost_addr.s_addr ||
1048         guest_addr->s_addr == slirp->vnameserver_addr.s_addr) {
1049         return false;
1050     }
1051
1052     /* check if the port is "bound" */
1053     for (tmp_ptr = slirp->guestfwd_list; tmp_ptr; tmp_ptr = tmp_ptr->ex_next) {
1054         if (guest_port == tmp_ptr->ex_fport &&
1055             guest_addr->s_addr == tmp_ptr->ex_addr.s_addr)
1056             return false;
1057     }
1058
1059     return true;
1060 }
1061
1062 int slirp_add_exec(Slirp *slirp, const char *cmdline,
1063                    struct in_addr *guest_addr, int guest_port)
1064 {
1065     if (!check_guestfwd(slirp, guest_addr, guest_port)) {
1066         return -1;
1067     }
1068
1069     add_exec(&slirp->guestfwd_list, cmdline, *guest_addr, htons(guest_port));
1070     return 0;
1071 }
1072
1073 int slirp_add_guestfwd(Slirp *slirp, SlirpWriteCb write_cb, void *opaque,
1074                        struct in_addr *guest_addr, int guest_port)
1075 {
1076     if (!check_guestfwd(slirp, guest_addr, guest_port)) {
1077         return -1;
1078     }
1079
1080     add_guestfwd(&slirp->guestfwd_list, write_cb, opaque,
1081                  *guest_addr, htons(guest_port));
1082     return 0;
1083 }
1084
1085 ssize_t slirp_send(struct socket *so, const void *buf, size_t len, int flags)
1086 {
1087     if (so->s == -1 && so->guestfwd) {
1088         /* XXX this blocks entire thread. Rewrite to use
1089          * qemu_chr_fe_write and background I/O callbacks */
1090         so->guestfwd->write_cb(buf, len, so->guestfwd->opaque);
1091         return len;
1092     }
1093
1094     if (so->s == -1) {
1095         /*
1096          * This should in theory not happen but it is hard to be
1097          * sure because some code paths will end up with so->s == -1
1098          * on a failure but don't dispose of the struct socket.
1099          * Check specifically, so we don't pass -1 to send().
1100          */
1101         errno = EBADF;
1102         return -1;
1103     }
1104
1105     return send(so->s, buf, len, flags);
1106 }
1107
1108 struct socket *
1109 slirp_find_ctl_socket(Slirp *slirp, struct in_addr guest_addr, int guest_port)
1110 {
1111     struct socket *so;
1112
1113     for (so = slirp->tcb.so_next; so != &slirp->tcb; so = so->so_next) {
1114         if (so->so_faddr.s_addr == guest_addr.s_addr &&
1115             htons(so->so_fport) == guest_port) {
1116             return so;
1117         }
1118     }
1119     return NULL;
1120 }
1121
1122 size_t slirp_socket_can_recv(Slirp *slirp, struct in_addr guest_addr,
1123                              int guest_port)
1124 {
1125     struct iovec iov[2];
1126     struct socket *so;
1127
1128     so = slirp_find_ctl_socket(slirp, guest_addr, guest_port);
1129
1130     if (!so || so->so_state & SS_NOFDREF) {
1131         return 0;
1132     }
1133
1134     if (!CONN_CANFRCV(so) || so->so_snd.sb_cc >= (so->so_snd.sb_datalen/2)) {
1135         return 0;
1136     }
1137
1138     return sopreprbuf(so, iov, NULL);
1139 }
1140
1141 void slirp_socket_recv(Slirp *slirp, struct in_addr guest_addr, int guest_port,
1142                        const uint8_t *buf, int size)
1143 {
1144     int ret;
1145     struct socket *so = slirp_find_ctl_socket(slirp, guest_addr, guest_port);
1146
1147     if (!so)
1148         return;
1149
1150     ret = soreadbuf(so, (const char *)buf, size);
1151
1152     if (ret > 0)
1153         tcp_output(sototcpcb(so));
1154 }
1155
1156 void slirp_send_packet_all(Slirp *slirp, const void *buf, size_t len)
1157 {
1158     ssize_t ret = slirp->cb->send_packet(buf, len, slirp->opaque);
1159
1160     if (ret < 0) {
1161         g_critical("Failed to send packet, ret: %ld", (long) ret);
1162     } else if (ret < len) {
1163         DEBUG_ERROR("send_packet() didn't send all data: %ld < %lu",
1164                 (long) ret, (unsigned long) len);
1165     }
1166 }