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 "qemu-common.h"
  25 #include "qemu/timer.h"
  26 #include "sysemu/char.h"
  27 #include "slirp.h"
  28 #include "hw/hw.h"
  29
  30 /* host loopback address */
  31 struct in_addr loopback_addr;
  32 /* host loopback network mask */
  33 unsigned long loopback_mask;
  34
  35 /* emulated hosts use the MAC addr 52:55:IP:IP:IP:IP */
  36 static const uint8_t special_ethaddr[ETH_ALEN] = {
  37     0x52, 0x55, 0x00, 0x00, 0x00, 0x00
  38 };
  39
  40 static const uint8_t zero_ethaddr[ETH_ALEN] = { 0, 0, 0, 0, 0, 0 };
  41
  42 u_int curtime;
  43
  44 static QTAILQ_HEAD(slirp_instances, Slirp) slirp_instances =
  45     QTAILQ_HEAD_INITIALIZER(slirp_instances);
  46
  47 static struct in_addr dns_addr;
  48 static u_int dns_addr_time;
  49
  50 #define TIMEOUT_FAST 2  /* milliseconds */
  51 #define TIMEOUT_SLOW 499  /* milliseconds */
  52 /* for the aging of certain requests like DNS */
  53 #define TIMEOUT_DEFAULT 1000  /* milliseconds */
  54
  55 #ifdef _WIN32
  56
  57 int get_dns_addr(struct in_addr *pdns_addr)
  58 {
  59     FIXED_INFO *FixedInfo=NULL;
  60     ULONG    BufLen;
  61     DWORD    ret;
  62     IP_ADDR_STRING *pIPAddr;
  63     struct in_addr tmp_addr;
  64
  65     if (dns_addr.s_addr != 0 && (curtime - dns_addr_time) < TIMEOUT_DEFAULT) {
  66         *pdns_addr = dns_addr;
  67         return 0;
  68     }
  69
  70     FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
  71     BufLen = sizeof(FIXED_INFO);
  72
  73     if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen)) {
  74         if (FixedInfo) {
  75             GlobalFree(FixedInfo);
  76             FixedInfo = NULL;
  77         }
  78         FixedInfo = GlobalAlloc(GPTR, BufLen);
  79     }
  80
  81     if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS) {
  82         printf("GetNetworkParams failed. ret = %08x\n", (u_int)ret );
  83         if (FixedInfo) {
  84             GlobalFree(FixedInfo);
  85             FixedInfo = NULL;
  86         }
  87         return -1;
  88     }
  89
  90     pIPAddr = &(FixedInfo->DnsServerList);
  91     inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
  92     *pdns_addr = tmp_addr;
  93     dns_addr = tmp_addr;
  94     dns_addr_time = curtime;
  95     if (FixedInfo) {
  96         GlobalFree(FixedInfo);
  97         FixedInfo = NULL;
  98     }
  99     return 0;
 100 }
 101
 102 static void winsock_cleanup(void)
 103 {
 104     WSACleanup();
 105 }
 106
 107 #else
 108
 109 static struct stat dns_addr_stat;
 110
 111 int get_dns_addr(struct in_addr *pdns_addr)
 112 {
 113     char buff[512];
 114     char buff2[257];
 115     FILE *f;
 116     int found = 0;
 117     struct in_addr tmp_addr;
 118
 119     if (dns_addr.s_addr != 0) {
 120         struct stat old_stat;
 121         if ((curtime - dns_addr_time) < TIMEOUT_DEFAULT) {
 122             *pdns_addr = dns_addr;
 123             return 0;
 124         }
 125         old_stat = dns_addr_stat;
 126         if (stat("/etc/resolv.conf", &dns_addr_stat) != 0)
 127             return -1;
 128         if ((dns_addr_stat.st_dev == old_stat.st_dev)
 129             && (dns_addr_stat.st_ino == old_stat.st_ino)
 130             && (dns_addr_stat.st_size == old_stat.st_size)
 131             && (dns_addr_stat.st_mtime == old_stat.st_mtime)) {
 132             *pdns_addr = dns_addr;
 133             return 0;
 134         }
 135     }
 136
 137     f = fopen("/etc/resolv.conf", "r");
 138     if (!f)
 139         return -1;
 140
 141 #ifdef DEBUG
 142     lprint("IP address of your DNS(s): ");
 143 #endif
 144     while (fgets(buff, 512, f) != NULL) {
 145         if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1) {
 146             if (!inet_aton(buff2, &tmp_addr))
 147                 continue;
 148             /* If it's the first one, set it to dns_addr */
 149             if (!found) {
 150                 *pdns_addr = tmp_addr;
 151                 dns_addr = tmp_addr;
 152                 dns_addr_time = curtime;
 153             }
 154 #ifdef DEBUG
 155             else
 156                 lprint(", ");
 157 #endif
 158             if (++found > 3) {
 159 #ifdef DEBUG
 160                 lprint("(more)");
 161 #endif
 162                 break;
 163             }
 164 #ifdef DEBUG
 165             else
 166                 lprint("%s", inet_ntoa(tmp_addr));
 167 #endif
 168         }
 169     }
 170     fclose(f);
 171     if (!found)
 172         return -1;
 173     return 0;
 174 }
 175
 176 #endif
 177
 178 static void slirp_init_once(void)
 179 {
 180     static int initialized;
 181 #ifdef _WIN32
 182     WSADATA Data;
 183 #endif
 184
 185     if (initialized) {
 186         return;
 187     }
 188     initialized = 1;
 189
 190 #ifdef _WIN32
 191     WSAStartup(MAKEWORD(2,0), &Data);
 192     atexit(winsock_cleanup);
 193 #endif
 194
 195     loopback_addr.s_addr = htonl(INADDR_LOOPBACK);
 196     loopback_mask = htonl(IN_CLASSA_NET);
 197 }
 198
 199 static void slirp_state_save(QEMUFile *f, void *opaque);
 200 static int slirp_state_load(QEMUFile *f, void *opaque, int version_id);
 201
 202 Slirp *slirp_init(int restricted, struct in_addr vnetwork,
 203                   struct in_addr vnetmask, struct in_addr vhost,
 204                   const char *vhostname, const char *tftp_path,
 205                   const char *bootfile, struct in_addr vdhcp_start,
 206                   struct in_addr vnameserver, const char **vdnssearch,
 207                   void *opaque)
 208 {
 209     Slirp *slirp = g_malloc0(sizeof(Slirp));
 210
 211     slirp_init_once();
 212
 213     slirp->restricted = restricted;
 214
 215     if_init(slirp);
 216     ip_init(slirp);
 217
 218     /* Initialise mbufs *after* setting the MTU */
 219     m_init(slirp);
 220
 221     slirp->vnetwork_addr = vnetwork;
 222     slirp->vnetwork_mask = vnetmask;
 223     slirp->vhost_addr = vhost;
 224     if (vhostname) {
 225         pstrcpy(slirp->client_hostname, sizeof(slirp->client_hostname),
 226                 vhostname);
 227     }
 228     slirp->tftp_prefix = g_strdup(tftp_path);
 229     slirp->bootp_filename = g_strdup(bootfile);
 230     slirp->vdhcp_startaddr = vdhcp_start;
 231     slirp->vnameserver_addr = vnameserver;
 232
 233     if (vdnssearch) {
 234         translate_dnssearch(slirp, vdnssearch);
 235     }
 236
 237     slirp->opaque = opaque;
 238
 239     register_savevm(NULL, "slirp", 0, 3,
 240                     slirp_state_save, slirp_state_load, slirp);
 241
 242     QTAILQ_INSERT_TAIL(&slirp_instances, slirp, entry);
 243
 244     return slirp;
 245 }
 246
 247 void slirp_cleanup(Slirp *slirp)
 248 {
 249     QTAILQ_REMOVE(&slirp_instances, slirp, entry);
 250
 251     unregister_savevm(NULL, "slirp", slirp);
 252
 253     ip_cleanup(slirp);
 254     m_cleanup(slirp);
 255
 256     g_free(slirp->vdnssearch);
 257     g_free(slirp->tftp_prefix);
 258     g_free(slirp->bootp_filename);
 259     g_free(slirp);
 260 }
 261
 262 #define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
 263 #define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
 264
 265 static void slirp_update_timeout(uint32_t *timeout)
 266 {
 267     Slirp *slirp;
 268     uint32_t t;
 269
 270     if (*timeout <= TIMEOUT_FAST) {
 271         return;
 272     }
 273
 274     t = MIN(1000, *timeout);
 275
 276     /* If we have tcp timeout with slirp, then we will fill @timeout with
 277      * more precise value.
 278      */
 279     QTAILQ_FOREACH(slirp, &slirp_instances, entry) {
 280         if (slirp->time_fasttimo) {
 281             *timeout = TIMEOUT_FAST;
 282             return;
 283         }
 284         if (slirp->do_slowtimo) {
 285             t = MIN(TIMEOUT_SLOW, t);
 286         }
 287     }
 288     *timeout = t;
 289 }
 290
 291 void slirp_pollfds_fill(GArray *pollfds, uint32_t *timeout)
 292 {
 293     Slirp *slirp;
 294     struct socket *so, *so_next;
 295
 296     if (QTAILQ_EMPTY(&slirp_instances)) {
 297         return;
 298     }
 299
 300     /*
 301      * First, TCP sockets
 302      */
 303
 304     QTAILQ_FOREACH(slirp, &slirp_instances, entry) {
 305         /*
 306          * *_slowtimo needs calling if there are IP fragments
 307          * in the fragment queue, or there are TCP connections active
 308          */
 309         slirp->do_slowtimo = ((slirp->tcb.so_next != &slirp->tcb) ||
 310                 (&slirp->ipq.ip_link != slirp->ipq.ip_link.next));
 311
 312         for (so = slirp->tcb.so_next; so != &slirp->tcb;
 313                 so = so_next) {
 314             int events = 0;
 315
 316             so_next = so->so_next;
 317
 318             so->pollfds_idx = -1;
 319
 320             /*
 321              * See if we need a tcp_fasttimo
 322              */
 323             if (slirp->time_fasttimo == 0 &&
 324                 so->so_tcpcb->t_flags & TF_DELACK) {
 325                 slirp->time_fasttimo = curtime; /* Flag when want a fasttimo */
 326             }
 327
 328             /*
 329              * NOFDREF can include still connecting to local-host,
 330              * newly socreated() sockets etc. Don't want to select these.
 331              */
 332             if (so->so_state & SS_NOFDREF || so->s == -1) {
 333                 continue;
 334             }
 335
 336             /*
 337              * Set for reading sockets which are accepting
 338              */
 339             if (so->so_state & SS_FACCEPTCONN) {
 340                 GPollFD pfd = {
 341                     .fd = so->s,
 342                     .events = G_IO_IN | G_IO_HUP | G_IO_ERR,
 343                 };
 344                 so->pollfds_idx = pollfds->len;
 345                 g_array_append_val(pollfds, pfd);
 346                 continue;
 347             }
 348
 349             /*
 350              * Set for writing sockets which are connecting
 351              */
 352             if (so->so_state & SS_ISFCONNECTING) {
 353                 GPollFD pfd = {
 354                     .fd = so->s,
 355                     .events = G_IO_OUT | G_IO_ERR,
 356                 };
 357                 so->pollfds_idx = pollfds->len;
 358                 g_array_append_val(pollfds, pfd);
 359                 continue;
 360             }
 361
 362             /*
 363              * Set for writing if we are connected, can send more, and
 364              * we have something to send
 365              */
 366             if (CONN_CANFSEND(so) && so->so_rcv.sb_cc) {
 367                 events |= G_IO_OUT | G_IO_ERR;
 368             }
 369
 370             /*
 371              * Set for reading (and urgent data) if we are connected, can
 372              * receive more, and we have room for it XXX /2 ?
 373              */
 374             if (CONN_CANFRCV(so) &&
 375                 (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2))) {
 376                 events |= G_IO_IN | G_IO_HUP | G_IO_ERR | G_IO_PRI;
 377             }
 378
 379             if (events) {
 380                 GPollFD pfd = {
 381                     .fd = so->s,
 382                     .events = events,
 383                 };
 384                 so->pollfds_idx = pollfds->len;
 385                 g_array_append_val(pollfds, pfd);
 386             }
 387         }
 388
 389         /*
 390          * UDP sockets
 391          */
 392         for (so = slirp->udb.so_next; so != &slirp->udb;
 393                 so = so_next) {
 394             so_next = so->so_next;
 395
 396             so->pollfds_idx = -1;
 397
 398             /*
 399              * See if it's timed out
 400              */
 401             if (so->so_expire) {
 402                 if (so->so_expire <= curtime) {
 403                     udp_detach(so);
 404                     continue;
 405                 } else {
 406                     slirp->do_slowtimo = true; /* Let socket expire */
 407                 }
 408             }
 409
 410             /*
 411              * When UDP packets are received from over the
 412              * link, they're sendto()'d straight away, so
 413              * no need for setting for writing
 414              * Limit the number of packets queued by this session
 415              * to 4.  Note that even though we try and limit this
 416              * to 4 packets, the session could have more queued
 417              * if the packets needed to be fragmented
 418              * (XXX <= 4 ?)
 419              */
 420             if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4) {
 421                 GPollFD pfd = {
 422                     .fd = so->s,
 423                     .events = G_IO_IN | G_IO_HUP | G_IO_ERR,
 424                 };
 425                 so->pollfds_idx = pollfds->len;
 426                 g_array_append_val(pollfds, pfd);
 427             }
 428         }
 429
 430         /*
 431          * ICMP sockets
 432          */
 433         for (so = slirp->icmp.so_next; so != &slirp->icmp;
 434                 so = so_next) {
 435             so_next = so->so_next;
 436
 437             so->pollfds_idx = -1;
 438
 439             /*
 440              * See if it's timed out
 441              */
 442             if (so->so_expire) {
 443                 if (so->so_expire <= curtime) {
 444                     icmp_detach(so);
 445                     continue;
 446                 } else {
 447                     slirp->do_slowtimo = true; /* Let socket expire */
 448                 }
 449             }
 450
 451             if (so->so_state & SS_ISFCONNECTED) {
 452                 GPollFD pfd = {
 453                     .fd = so->s,
 454                     .events = G_IO_IN | G_IO_HUP | G_IO_ERR,
 455                 };
 456                 so->pollfds_idx = pollfds->len;
 457                 g_array_append_val(pollfds, pfd);
 458             }
 459         }
 460     }
 461     slirp_update_timeout(timeout);
 462 }
 463
 464 void slirp_pollfds_poll(GArray *pollfds, int select_error)
 465 {
 466     Slirp *slirp;
 467     struct socket *so, *so_next;
 468     int ret;
 469
 470     if (QTAILQ_EMPTY(&slirp_instances)) {
 471         return;
 472     }
 473
 474     curtime = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
 475
 476     QTAILQ_FOREACH(slirp, &slirp_instances, entry) {
 477         /*
 478          * See if anything has timed out
 479          */
 480         if (slirp->time_fasttimo &&
 481             ((curtime - slirp->time_fasttimo) >= TIMEOUT_FAST)) {
 482             tcp_fasttimo(slirp);
 483             slirp->time_fasttimo = 0;
 484         }
 485         if (slirp->do_slowtimo &&
 486             ((curtime - slirp->last_slowtimo) >= TIMEOUT_SLOW)) {
 487             ip_slowtimo(slirp);
 488             tcp_slowtimo(slirp);
 489             slirp->last_slowtimo = curtime;
 490         }
 491
 492         /*
 493          * Check sockets
 494          */
 495         if (!select_error) {
 496             /*
 497              * Check TCP sockets
 498              */
 499             for (so = slirp->tcb.so_next; so != &slirp->tcb;
 500                     so = so_next) {
 501                 int revents;
 502
 503                 so_next = so->so_next;
 504
 505                 revents = 0;
 506                 if (so->pollfds_idx != -1) {
 507                     revents = g_array_index(pollfds, GPollFD,
 508                                             so->pollfds_idx).revents;
 509                 }
 510
 511                 if (so->so_state & SS_NOFDREF || so->s == -1) {
 512                     continue;
 513                 }
 514
 515                 /*
 516                  * Check for URG data
 517                  * This will soread as well, so no need to
 518                  * test for G_IO_IN below if this succeeds
 519                  */
 520                 if (revents & G_IO_PRI) {
 521                     sorecvoob(so);
 522                 }
 523                 /*
 524                  * Check sockets for reading
 525                  */
 526                 else if (revents & (G_IO_IN | G_IO_HUP | G_IO_ERR)) {
 527                     /*
 528                      * Check for incoming connections
 529                      */
 530                     if (so->so_state & SS_FACCEPTCONN) {
 531                         tcp_connect(so);
 532                         continue;
 533                     } /* else */
 534                     ret = soread(so);
 535
 536                     /* Output it if we read something */
 537                     if (ret > 0) {
 538                         tcp_output(sototcpcb(so));
 539                     }
 540                 }
 541
 542                 /*
 543                  * Check sockets for writing
 544                  */
 545                 if (!(so->so_state & SS_NOFDREF) &&
 546                         (revents & (G_IO_OUT | G_IO_ERR))) {
 547                     /*
 548                      * Check for non-blocking, still-connecting sockets
 549                      */
 550                     if (so->so_state & SS_ISFCONNECTING) {
 551                         /* Connected */
 552                         so->so_state &= ~SS_ISFCONNECTING;
 553
 554                         ret = send(so->s, (const void *) &ret, 0, 0);
 555                         if (ret < 0) {
 556                             /* XXXXX Must fix, zero bytes is a NOP */
 557                             if (errno == EAGAIN || errno == EWOULDBLOCK ||
 558                                 errno == EINPROGRESS || errno == ENOTCONN) {
 559                                 continue;
 560                             }
 561
 562                             /* else failed */
 563                             so->so_state &= SS_PERSISTENT_MASK;
 564                             so->so_state |= SS_NOFDREF;
 565                         }
 566                         /* else so->so_state &= ~SS_ISFCONNECTING; */
 567
 568                         /*
 569                          * Continue tcp_input
 570                          */
 571                         tcp_input((struct mbuf *)NULL, sizeof(struct ip), so);
 572                         /* continue; */
 573                     } else {
 574                         ret = sowrite(so);
 575                     }
 576                     /*
 577                      * XXXXX If we wrote something (a lot), there
 578                      * could be a need for a window update.
 579                      * In the worst case, the remote will send
 580                      * a window probe to get things going again
 581                      */
 582                 }
 583
 584                 /*
 585                  * Probe a still-connecting, non-blocking socket
 586                  * to check if it's still alive
 587                  */
 588 #ifdef PROBE_CONN
 589                 if (so->so_state & SS_ISFCONNECTING) {
 590                     ret = qemu_recv(so->s, &ret, 0, 0);
 591
 592                     if (ret < 0) {
 593                         /* XXX */
 594                         if (errno == EAGAIN || errno == EWOULDBLOCK ||
 595                             errno == EINPROGRESS || errno == ENOTCONN) {
 596                             continue; /* Still connecting, continue */
 597                         }
 598
 599                         /* else failed */
 600                         so->so_state &= SS_PERSISTENT_MASK;
 601                         so->so_state |= SS_NOFDREF;
 602
 603                         /* tcp_input will take care of it */
 604                     } else {
 605                         ret = send(so->s, &ret, 0, 0);
 606                         if (ret < 0) {
 607                             /* XXX */
 608                             if (errno == EAGAIN || errno == EWOULDBLOCK ||
 609                                 errno == EINPROGRESS || errno == ENOTCONN) {
 610                                 continue;
 611                             }
 612                             /* else failed */
 613                             so->so_state &= SS_PERSISTENT_MASK;
 614                             so->so_state |= SS_NOFDREF;
 615                         } else {
 616                             so->so_state &= ~SS_ISFCONNECTING;
 617                         }
 618
 619                     }
 620                     tcp_input((struct mbuf *)NULL, sizeof(struct ip), so);
 621                 } /* SS_ISFCONNECTING */
 622 #endif
 623             }
 624
 625             /*
 626              * Now UDP sockets.
 627              * Incoming packets are sent straight away, they're not buffered.
 628              * Incoming UDP data isn't buffered either.
 629              */
 630             for (so = slirp->udb.so_next; so != &slirp->udb;
 631                     so = so_next) {
 632                 int revents;
 633
 634                 so_next = so->so_next;
 635
 636                 revents = 0;
 637                 if (so->pollfds_idx != -1) {
 638                     revents = g_array_index(pollfds, GPollFD,
 639                             so->pollfds_idx).revents;
 640                 }
 641
 642                 if (so->s != -1 &&
 643                     (revents & (G_IO_IN | G_IO_HUP | G_IO_ERR))) {
 644                     sorecvfrom(so);
 645                 }
 646             }
 647
 648             /*
 649              * Check incoming ICMP relies.
 650              */
 651             for (so = slirp->icmp.so_next; so != &slirp->icmp;
 652                     so = so_next) {
 653                     int revents;
 654
 655                     so_next = so->so_next;
 656
 657                     revents = 0;
 658                     if (so->pollfds_idx != -1) {
 659                         revents = g_array_index(pollfds, GPollFD,
 660                                                 so->pollfds_idx).revents;
 661                     }
 662
 663                     if (so->s != -1 &&
 664                         (revents & (G_IO_IN | G_IO_HUP | G_IO_ERR))) {
 665                     icmp_receive(so);
 666                 }
 667             }
 668         }
 669
 670         if_start(slirp);
 671     }
 672 }
 673
 674 static void arp_input(Slirp *slirp, const uint8_t *pkt, int pkt_len)
 675 {
 676     struct arphdr *ah = (struct arphdr *)(pkt + ETH_HLEN);
 677     uint8_t arp_reply[max(ETH_HLEN + sizeof(struct arphdr), 64)];
 678     struct ethhdr *reh = (struct ethhdr *)arp_reply;
 679     struct arphdr *rah = (struct arphdr *)(arp_reply + ETH_HLEN);
 680     int ar_op;
 681     struct ex_list *ex_ptr;
 682
 683     ar_op = ntohs(ah->ar_op);
 684     switch(ar_op) {
 685     case ARPOP_REQUEST:
 686         if (ah->ar_tip == ah->ar_sip) {
 687             /* Gratuitous ARP */
 688             arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
 689             return;
 690         }
 691
 692         if ((ah->ar_tip & slirp->vnetwork_mask.s_addr) ==
 693             slirp->vnetwork_addr.s_addr) {
 694             if (ah->ar_tip == slirp->vnameserver_addr.s_addr ||
 695                 ah->ar_tip == slirp->vhost_addr.s_addr)
 696                 goto arp_ok;
 697             for (ex_ptr = slirp->exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
 698                 if (ex_ptr->ex_addr.s_addr == ah->ar_tip)
 699                     goto arp_ok;
 700             }
 701             return;
 702         arp_ok:
 703             memset(arp_reply, 0, sizeof(arp_reply));
 704
 705             arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
 706
 707             /* ARP request for alias/dns mac address */
 708             memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
 709             memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 4);
 710             memcpy(&reh->h_source[2], &ah->ar_tip, 4);
 711             reh->h_proto = htons(ETH_P_ARP);
 712
 713             rah->ar_hrd = htons(1);
 714             rah->ar_pro = htons(ETH_P_IP);
 715             rah->ar_hln = ETH_ALEN;
 716             rah->ar_pln = 4;
 717             rah->ar_op = htons(ARPOP_REPLY);
 718             memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
 719             rah->ar_sip = ah->ar_tip;
 720             memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
 721             rah->ar_tip = ah->ar_sip;
 722             slirp_output(slirp->opaque, arp_reply, sizeof(arp_reply));
 723         }
 724         break;
 725     case ARPOP_REPLY:
 726         arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
 727         break;
 728     default:
 729         break;
 730     }
 731 }
 732
 733 void slirp_input(Slirp *slirp, const uint8_t *pkt, int pkt_len)
 734 {
 735     struct mbuf *m;
 736     int proto;
 737
 738     if (pkt_len < ETH_HLEN)
 739         return;
 740
 741     proto = ntohs(*(uint16_t *)(pkt + 12));
 742     switch(proto) {
 743     case ETH_P_ARP:
 744         arp_input(slirp, pkt, pkt_len);
 745         break;
 746     case ETH_P_IP:
 747         m = m_get(slirp);
 748         if (!m)
 749             return;
 750         /* Note: we add to align the IP header */
 751         if (M_FREEROOM(m) < pkt_len + 2) {
 752             m_inc(m, pkt_len + 2);
 753         }
 754         m->m_len = pkt_len + 2;
 755         memcpy(m->m_data + 2, pkt, pkt_len);
 756
 757         m->m_data += 2 + ETH_HLEN;
 758         m->m_len -= 2 + ETH_HLEN;
 759
 760         ip_input(m);
 761         break;
 762     default:
 763         break;
 764     }
 765 }
 766
 767 /* Output the IP packet to the ethernet device. Returns 0 if the packet must be
 768  * re-queued.
 769  */
 770 int if_encap(Slirp *slirp, struct mbuf *ifm)
 771 {
 772     uint8_t buf[1600];
 773     struct ethhdr *eh = (struct ethhdr *)buf;
 774     uint8_t ethaddr[ETH_ALEN];
 775     const struct ip *iph = (const struct ip *)ifm->m_data;
 776
 777     if (ifm->m_len + ETH_HLEN > sizeof(buf)) {
 778         return 1;
 779     }
 780
 781     if (!arp_table_search(slirp, iph->ip_dst.s_addr, ethaddr)) {
 782         uint8_t arp_req[ETH_HLEN + sizeof(struct arphdr)];
 783         struct ethhdr *reh = (struct ethhdr *)arp_req;
 784         struct arphdr *rah = (struct arphdr *)(arp_req + ETH_HLEN);
 785
 786         if (!ifm->arp_requested) {
 787             /* If the client addr is not known, send an ARP request */
 788             memset(reh->h_dest, 0xff, ETH_ALEN);
 789             memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 4);
 790             memcpy(&reh->h_source[2], &slirp->vhost_addr, 4);
 791             reh->h_proto = htons(ETH_P_ARP);
 792             rah->ar_hrd = htons(1);
 793             rah->ar_pro = htons(ETH_P_IP);
 794             rah->ar_hln = ETH_ALEN;
 795             rah->ar_pln = 4;
 796             rah->ar_op = htons(ARPOP_REQUEST);
 797
 798             /* source hw addr */
 799             memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN - 4);
 800             memcpy(&rah->ar_sha[2], &slirp->vhost_addr, 4);
 801
 802             /* source IP */
 803             rah->ar_sip = slirp->vhost_addr.s_addr;
 804
 805             /* target hw addr (none) */
 806             memset(rah->ar_tha, 0, ETH_ALEN);
 807
 808             /* target IP */
 809             rah->ar_tip = iph->ip_dst.s_addr;
 810             slirp->client_ipaddr = iph->ip_dst;
 811             slirp_output(slirp->opaque, arp_req, sizeof(arp_req));
 812             ifm->arp_requested = true;
 813
 814             /* Expire request and drop outgoing packet after 1 second */
 815             ifm->expiration_date = qemu_clock_get_ns(QEMU_CLOCK_REALTIME) + 1000000000ULL;
 816         }
 817         return 0;
 818     } else {
 819         memcpy(eh->h_dest, ethaddr, ETH_ALEN);
 820         memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 4);
 821         /* XXX: not correct */
 822         memcpy(&eh->h_source[2], &slirp->vhost_addr, 4);
 823         eh->h_proto = htons(ETH_P_IP);
 824         memcpy(buf + sizeof(struct ethhdr), ifm->m_data, ifm->m_len);
 825         slirp_output(slirp->opaque, buf, ifm->m_len + ETH_HLEN);
 826         return 1;
 827     }
 828 }
 829
 830 /* Drop host forwarding rule, return 0 if found. */
 831 int slirp_remove_hostfwd(Slirp *slirp, int is_udp, struct in_addr host_addr,
 832                          int host_port)
 833 {
 834     struct socket *so;
 835     struct socket *head = (is_udp ? &slirp->udb : &slirp->tcb);
 836     struct sockaddr_in addr;
 837     int port = htons(host_port);
 838     socklen_t addr_len;
 839
 840     for (so = head->so_next; so != head; so = so->so_next) {
 841         addr_len = sizeof(addr);
 842         if ((so->so_state & SS_HOSTFWD) &&
 843             getsockname(so->s, (struct sockaddr *)&addr, &addr_len) == 0 &&
 844             addr.sin_addr.s_addr == host_addr.s_addr &&
 845             addr.sin_port == port) {
 846             close(so->s);
 847             sofree(so);
 848             return 0;
 849         }
 850     }
 851
 852     return -1;
 853 }
 854
 855 int slirp_add_hostfwd(Slirp *slirp, int is_udp, struct in_addr host_addr,
 856                       int host_port, struct in_addr guest_addr, int guest_port)
 857 {
 858     if (!guest_addr.s_addr) {
 859         guest_addr = slirp->vdhcp_startaddr;
 860     }
 861     if (is_udp) {
 862         if (!udp_listen(slirp, host_addr.s_addr, htons(host_port),
 863                         guest_addr.s_addr, htons(guest_port), SS_HOSTFWD))
 864             return -1;
 865     } else {
 866         if (!tcp_listen(slirp, host_addr.s_addr, htons(host_port),
 867                         guest_addr.s_addr, htons(guest_port), SS_HOSTFWD))
 868             return -1;
 869     }
 870     return 0;
 871 }
 872
 873 int slirp_add_exec(Slirp *slirp, int do_pty, const void *args,
 874                    struct in_addr *guest_addr, int guest_port)
 875 {
 876     if (!guest_addr->s_addr) {
 877         guest_addr->s_addr = slirp->vnetwork_addr.s_addr |
 878             (htonl(0x0204) & ~slirp->vnetwork_mask.s_addr);
 879     }
 880     if ((guest_addr->s_addr & slirp->vnetwork_mask.s_addr) !=
 881         slirp->vnetwork_addr.s_addr ||
 882         guest_addr->s_addr == slirp->vhost_addr.s_addr ||
 883         guest_addr->s_addr == slirp->vnameserver_addr.s_addr) {
 884         return -1;
 885     }
 886     return add_exec(&slirp->exec_list, do_pty, (char *)args, *guest_addr,
 887                     htons(guest_port));
 888 }
 889
 890 ssize_t slirp_send(struct socket *so, const void *buf, size_t len, int flags)
 891 {
 892     if (so->s == -1 && so->extra) {
 893         qemu_chr_fe_write(so->extra, buf, len);
 894         return len;
 895     }
 896
 897     return send(so->s, buf, len, flags);
 898 }
 899
 900 static struct socket *
 901 slirp_find_ctl_socket(Slirp *slirp, struct in_addr guest_addr, int guest_port)
 902 {
 903     struct socket *so;
 904
 905     for (so = slirp->tcb.so_next; so != &slirp->tcb; so = so->so_next) {
 906         if (so->so_faddr.s_addr == guest_addr.s_addr &&
 907             htons(so->so_fport) == guest_port) {
 908             return so;
 909         }
 910     }
 911     return NULL;
 912 }
 913
 914 size_t slirp_socket_can_recv(Slirp *slirp, struct in_addr guest_addr,
 915                              int guest_port)
 916 {
 917     struct iovec iov[2];
 918     struct socket *so;
 919
 920     so = slirp_find_ctl_socket(slirp, guest_addr, guest_port);
 921
 922     if (!so || so->so_state & SS_NOFDREF) {
 923         return 0;
 924     }
 925
 926     if (!CONN_CANFRCV(so) || so->so_snd.sb_cc >= (so->so_snd.sb_datalen/2)) {
 927         return 0;
 928     }
 929
 930     return sopreprbuf(so, iov, NULL);
 931 }
 932
 933 void slirp_socket_recv(Slirp *slirp, struct in_addr guest_addr, int guest_port,
 934                        const uint8_t *buf, int size)
 935 {
 936     int ret;
 937     struct socket *so = slirp_find_ctl_socket(slirp, guest_addr, guest_port);
 938
 939     if (!so)
 940         return;
 941
 942     ret = soreadbuf(so, (const char *)buf, size);
 943
 944     if (ret > 0)
 945         tcp_output(sototcpcb(so));
 946 }
 947
 948 static void slirp_tcp_save(QEMUFile *f, struct tcpcb *tp)
 949 {
 950     int i;
 951
 952     qemu_put_sbe16(f, tp->t_state);
 953     for (i = 0; i < TCPT_NTIMERS; i++)
 954         qemu_put_sbe16(f, tp->t_timer[i]);
 955     qemu_put_sbe16(f, tp->t_rxtshift);
 956     qemu_put_sbe16(f, tp->t_rxtcur);
 957     qemu_put_sbe16(f, tp->t_dupacks);
 958     qemu_put_be16(f, tp->t_maxseg);
 959     qemu_put_sbyte(f, tp->t_force);
 960     qemu_put_be16(f, tp->t_flags);
 961     qemu_put_be32(f, tp->snd_una);
 962     qemu_put_be32(f, tp->snd_nxt);
 963     qemu_put_be32(f, tp->snd_up);
 964     qemu_put_be32(f, tp->snd_wl1);
 965     qemu_put_be32(f, tp->snd_wl2);
 966     qemu_put_be32(f, tp->iss);
 967     qemu_put_be32(f, tp->snd_wnd);
 968     qemu_put_be32(f, tp->rcv_wnd);
 969     qemu_put_be32(f, tp->rcv_nxt);
 970     qemu_put_be32(f, tp->rcv_up);
 971     qemu_put_be32(f, tp->irs);
 972     qemu_put_be32(f, tp->rcv_adv);
 973     qemu_put_be32(f, tp->snd_max);
 974     qemu_put_be32(f, tp->snd_cwnd);
 975     qemu_put_be32(f, tp->snd_ssthresh);
 976     qemu_put_sbe16(f, tp->t_idle);
 977     qemu_put_sbe16(f, tp->t_rtt);
 978     qemu_put_be32(f, tp->t_rtseq);
 979     qemu_put_sbe16(f, tp->t_srtt);
 980     qemu_put_sbe16(f, tp->t_rttvar);
 981     qemu_put_be16(f, tp->t_rttmin);
 982     qemu_put_be32(f, tp->max_sndwnd);
 983     qemu_put_byte(f, tp->t_oobflags);
 984     qemu_put_byte(f, tp->t_iobc);
 985     qemu_put_sbe16(f, tp->t_softerror);
 986     qemu_put_byte(f, tp->snd_scale);
 987     qemu_put_byte(f, tp->rcv_scale);
 988     qemu_put_byte(f, tp->request_r_scale);
 989     qemu_put_byte(f, tp->requested_s_scale);
 990     qemu_put_be32(f, tp->ts_recent);
 991     qemu_put_be32(f, tp->ts_recent_age);
 992     qemu_put_be32(f, tp->last_ack_sent);
 993 }
 994
 995 static void slirp_sbuf_save(QEMUFile *f, struct sbuf *sbuf)
 996 {
 997     uint32_t off;
 998
 999     qemu_put_be32(f, sbuf->sb_cc);
1000     qemu_put_be32(f, sbuf->sb_datalen);
1001     off = (uint32_t)(sbuf->sb_wptr - sbuf->sb_data);
1002     qemu_put_sbe32(f, off);
1003     off = (uint32_t)(sbuf->sb_rptr - sbuf->sb_data);
1004     qemu_put_sbe32(f, off);
1005     qemu_put_buffer(f, (unsigned char*)sbuf->sb_data, sbuf->sb_datalen);
1006 }
1007
1008 static void slirp_socket_save(QEMUFile *f, struct socket *so)
1009 {
1010     qemu_put_be32(f, so->so_urgc);
1011     qemu_put_be32(f, so->so_faddr.s_addr);
1012     qemu_put_be32(f, so->so_laddr.s_addr);
1013     qemu_put_be16(f, so->so_fport);
1014     qemu_put_be16(f, so->so_lport);
1015     qemu_put_byte(f, so->so_iptos);
1016     qemu_put_byte(f, so->so_emu);
1017     qemu_put_byte(f, so->so_type);
1018     qemu_put_be32(f, so->so_state);
1019     slirp_sbuf_save(f, &so->so_rcv);
1020     slirp_sbuf_save(f, &so->so_snd);
1021     slirp_tcp_save(f, so->so_tcpcb);
1022 }
1023
1024 static void slirp_bootp_save(QEMUFile *f, Slirp *slirp)
1025 {
1026     int i;
1027
1028     for (i = 0; i < NB_BOOTP_CLIENTS; i++) {
1029         qemu_put_be16(f, slirp->bootp_clients[i].allocated);
1030         qemu_put_buffer(f, slirp->bootp_clients[i].macaddr, 6);
1031     }
1032 }
1033
1034 static void slirp_state_save(QEMUFile *f, void *opaque)
1035 {
1036     Slirp *slirp = opaque;
1037     struct ex_list *ex_ptr;
1038
1039     for (ex_ptr = slirp->exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1040         if (ex_ptr->ex_pty == 3) {
1041             struct socket *so;
1042             so = slirp_find_ctl_socket(slirp, ex_ptr->ex_addr,
1043                                        ntohs(ex_ptr->ex_fport));
1044             if (!so)
1045                 continue;
1046
1047             qemu_put_byte(f, 42);
1048             slirp_socket_save(f, so);
1049         }
1050     qemu_put_byte(f, 0);
1051
1052     qemu_put_be16(f, slirp->ip_id);
1053
1054     slirp_bootp_save(f, slirp);
1055 }
1056
1057 static void slirp_tcp_load(QEMUFile *f, struct tcpcb *tp)
1058 {
1059     int i;
1060
1061     tp->t_state = qemu_get_sbe16(f);
1062     for (i = 0; i < TCPT_NTIMERS; i++)
1063         tp->t_timer[i] = qemu_get_sbe16(f);
1064     tp->t_rxtshift = qemu_get_sbe16(f);
1065     tp->t_rxtcur = qemu_get_sbe16(f);
1066     tp->t_dupacks = qemu_get_sbe16(f);
1067     tp->t_maxseg = qemu_get_be16(f);
1068     tp->t_force = qemu_get_sbyte(f);
1069     tp->t_flags = qemu_get_be16(f);
1070     tp->snd_una = qemu_get_be32(f);
1071     tp->snd_nxt = qemu_get_be32(f);
1072     tp->snd_up = qemu_get_be32(f);
1073     tp->snd_wl1 = qemu_get_be32(f);
1074     tp->snd_wl2 = qemu_get_be32(f);
1075     tp->iss = qemu_get_be32(f);
1076     tp->snd_wnd = qemu_get_be32(f);
1077     tp->rcv_wnd = qemu_get_be32(f);
1078     tp->rcv_nxt = qemu_get_be32(f);
1079     tp->rcv_up = qemu_get_be32(f);
1080     tp->irs = qemu_get_be32(f);
1081     tp->rcv_adv = qemu_get_be32(f);
1082     tp->snd_max = qemu_get_be32(f);
1083     tp->snd_cwnd = qemu_get_be32(f);
1084     tp->snd_ssthresh = qemu_get_be32(f);
1085     tp->t_idle = qemu_get_sbe16(f);
1086     tp->t_rtt = qemu_get_sbe16(f);
1087     tp->t_rtseq = qemu_get_be32(f);
1088     tp->t_srtt = qemu_get_sbe16(f);
1089     tp->t_rttvar = qemu_get_sbe16(f);
1090     tp->t_rttmin = qemu_get_be16(f);
1091     tp->max_sndwnd = qemu_get_be32(f);
1092     tp->t_oobflags = qemu_get_byte(f);
1093     tp->t_iobc = qemu_get_byte(f);
1094     tp->t_softerror = qemu_get_sbe16(f);
1095     tp->snd_scale = qemu_get_byte(f);
1096     tp->rcv_scale = qemu_get_byte(f);
1097     tp->request_r_scale = qemu_get_byte(f);
1098     tp->requested_s_scale = qemu_get_byte(f);
1099     tp->ts_recent = qemu_get_be32(f);
1100     tp->ts_recent_age = qemu_get_be32(f);
1101     tp->last_ack_sent = qemu_get_be32(f);
1102     tcp_template(tp);
1103 }
1104
1105 static int slirp_sbuf_load(QEMUFile *f, struct sbuf *sbuf)
1106 {
1107     uint32_t off, sb_cc, sb_datalen;
1108
1109     sb_cc = qemu_get_be32(f);
1110     sb_datalen = qemu_get_be32(f);
1111
1112     sbreserve(sbuf, sb_datalen);
1113
1114     if (sbuf->sb_datalen != sb_datalen)
1115         return -ENOMEM;
1116
1117     sbuf->sb_cc = sb_cc;
1118
1119     off = qemu_get_sbe32(f);
1120     sbuf->sb_wptr = sbuf->sb_data + off;
1121     off = qemu_get_sbe32(f);
1122     sbuf->sb_rptr = sbuf->sb_data + off;
1123     qemu_get_buffer(f, (unsigned char*)sbuf->sb_data, sbuf->sb_datalen);
1124
1125     return 0;
1126 }
1127
1128 static int slirp_socket_load(QEMUFile *f, struct socket *so)
1129 {
1130     if (tcp_attach(so) < 0)
1131         return -ENOMEM;
1132
1133     so->so_urgc = qemu_get_be32(f);
1134     so->so_faddr.s_addr = qemu_get_be32(f);
1135     so->so_laddr.s_addr = qemu_get_be32(f);
1136     so->so_fport = qemu_get_be16(f);
1137     so->so_lport = qemu_get_be16(f);
1138     so->so_iptos = qemu_get_byte(f);
1139     so->so_emu = qemu_get_byte(f);
1140     so->so_type = qemu_get_byte(f);
1141     so->so_state = qemu_get_be32(f);
1142     if (slirp_sbuf_load(f, &so->so_rcv) < 0)
1143         return -ENOMEM;
1144     if (slirp_sbuf_load(f, &so->so_snd) < 0)
1145         return -ENOMEM;
1146     slirp_tcp_load(f, so->so_tcpcb);
1147
1148     return 0;
1149 }
1150
1151 static void slirp_bootp_load(QEMUFile *f, Slirp *slirp)
1152 {
1153     int i;
1154
1155     for (i = 0; i < NB_BOOTP_CLIENTS; i++) {
1156         slirp->bootp_clients[i].allocated = qemu_get_be16(f);
1157         qemu_get_buffer(f, slirp->bootp_clients[i].macaddr, 6);
1158     }
1159 }
1160
1161 static int slirp_state_load(QEMUFile *f, void *opaque, int version_id)
1162 {
1163     Slirp *slirp = opaque;
1164     struct ex_list *ex_ptr;
1165
1166     while (qemu_get_byte(f)) {
1167         int ret;
1168         struct socket *so = socreate(slirp);
1169
1170         if (!so)
1171             return -ENOMEM;
1172
1173         ret = slirp_socket_load(f, so);
1174
1175         if (ret < 0)
1176             return ret;
1177
1178         if ((so->so_faddr.s_addr & slirp->vnetwork_mask.s_addr) !=
1179             slirp->vnetwork_addr.s_addr) {
1180             return -EINVAL;
1181         }
1182         for (ex_ptr = slirp->exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
1183             if (ex_ptr->ex_pty == 3 &&
1184                 so->so_faddr.s_addr == ex_ptr->ex_addr.s_addr &&
1185                 so->so_fport == ex_ptr->ex_fport) {
1186                 break;
1187             }
1188         }
1189         if (!ex_ptr)
1190             return -EINVAL;
1191
1192         so->extra = (void *)ex_ptr->ex_exec;
1193     }
1194
1195     if (version_id >= 2) {
1196         slirp->ip_id = qemu_get_be16(f);
1197     }
1198
1199     if (version_id >= 3) {
1200         slirp_bootp_load(f, slirp);
1201     }
1202
1203     return 0;
1204 }