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-char.h"
  26 #include "slirp.h"
  27 #include "hw/hw.h"
  28
  29 /* host address */
  30 struct in_addr our_addr;
  31 /* host dns address */
  32 struct in_addr dns_addr;
  33 /* host loopback address */
  34 struct in_addr loopback_addr;
  35
  36 /* address for slirp virtual addresses */
  37 struct in_addr special_addr;
  38 /* virtual address alias for host */
  39 struct in_addr alias_addr;
  40
  41 static const uint8_t special_ethaddr[6] = {
  42     0x52, 0x54, 0x00, 0x12, 0x35, 0x00
  43 };
  44
  45 /* ARP cache for the guest IP addresses (XXX: allow many entries) */
  46 uint8_t client_ethaddr[6];
  47 static struct in_addr client_ipaddr;
  48
  49 static const uint8_t zero_ethaddr[6] = { 0, 0, 0, 0, 0, 0 };
  50
  51 const char *slirp_special_ip = CTL_SPECIAL;
  52 int slirp_restrict;
  53 static int do_slowtimo;
  54 int link_up;
  55 struct timeval tt;
  56 FILE *lfd;
  57 struct ex_list *exec_list;
  58
  59 /* XXX: suppress those select globals */
  60 fd_set *global_readfds, *global_writefds, *global_xfds;
  61
  62 char slirp_hostname[33];
  63
  64 #ifdef _WIN32
  65
  66 static int get_dns_addr(struct in_addr *pdns_addr)
  67 {
  68     FIXED_INFO *FixedInfo=NULL;
  69     ULONG    BufLen;
  70     DWORD    ret;
  71     IP_ADDR_STRING *pIPAddr;
  72     struct in_addr tmp_addr;
  73
  74     FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
  75     BufLen = sizeof(FIXED_INFO);
  76
  77     if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen)) {
  78         if (FixedInfo) {
  79             GlobalFree(FixedInfo);
  80             FixedInfo = NULL;
  81         }
  82         FixedInfo = GlobalAlloc(GPTR, BufLen);
  83     }
  84
  85     if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS) {
  86         printf("GetNetworkParams failed. ret = %08x\n", (u_int)ret );
  87         if (FixedInfo) {
  88             GlobalFree(FixedInfo);
  89             FixedInfo = NULL;
  90         }
  91         return -1;
  92     }
  93
  94     pIPAddr = &(FixedInfo->DnsServerList);
  95     inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
  96     *pdns_addr = tmp_addr;
  97 #if 0
  98     printf( "DNS Servers:\n" );
  99     printf( "DNS Addr:%s\n", pIPAddr->IpAddress.String );
 100
 101     pIPAddr = FixedInfo -> DnsServerList.Next;
 102     while ( pIPAddr ) {
 103             printf( "DNS Addr:%s\n", pIPAddr ->IpAddress.String );
 104             pIPAddr = pIPAddr ->Next;
 105     }
 106 #endif
 107     if (FixedInfo) {
 108         GlobalFree(FixedInfo);
 109         FixedInfo = NULL;
 110     }
 111     return 0;
 112 }
 113
 114 #else
 115
 116 static int get_dns_addr(struct in_addr *pdns_addr)
 117 {
 118     char buff[512];
 119     char buff2[257];
 120     FILE *f;
 121     int found = 0;
 122     struct in_addr tmp_addr;
 123
 124     f = fopen("/etc/resolv.conf", "r");
 125     if (!f)
 126         return -1;
 127
 128 #ifdef DEBUG
 129     lprint("IP address of your DNS(s): ");
 130 #endif
 131     while (fgets(buff, 512, f) != NULL) {
 132         if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1) {
 133             if (!inet_aton(buff2, &tmp_addr))
 134                 continue;
 135             if (tmp_addr.s_addr == loopback_addr.s_addr)
 136                 tmp_addr = our_addr;
 137             /* If it's the first one, set it to dns_addr */
 138             if (!found)
 139                 *pdns_addr = tmp_addr;
 140 #ifdef DEBUG
 141             else
 142                 lprint(", ");
 143 #endif
 144             if (++found > 3) {
 145 #ifdef DEBUG
 146                 lprint("(more)");
 147 #endif
 148                 break;
 149             }
 150 #ifdef DEBUG
 151             else
 152                 lprint("%s", inet_ntoa(tmp_addr));
 153 #endif
 154         }
 155     }
 156     fclose(f);
 157     if (!found)
 158         return -1;
 159     return 0;
 160 }
 161
 162 #endif
 163
 164 #ifdef _WIN32
 165 static void slirp_cleanup(void)
 166 {
 167     WSACleanup();
 168 }
 169 #endif
 170
 171 static void slirp_state_save(QEMUFile *f, void *opaque);
 172 static int slirp_state_load(QEMUFile *f, void *opaque, int version_id);
 173
 174 void slirp_init(int restricted, char *special_ip)
 175 {
 176     //    debug_init("/tmp/slirp.log", DEBUG_DEFAULT);
 177
 178 #ifdef _WIN32
 179     {
 180         WSADATA Data;
 181         WSAStartup(MAKEWORD(2,0), &Data);
 182         atexit(slirp_cleanup);
 183     }
 184 #endif
 185
 186     link_up = 1;
 187     slirp_restrict = restricted;
 188
 189     if_init();
 190     ip_init();
 191
 192     /* Initialise mbufs *after* setting the MTU */
 193     m_init();
 194
 195     /* set default addresses */
 196     inet_aton("127.0.0.1", &loopback_addr);
 197
 198     if (get_dns_addr(&dns_addr) < 0) {
 199         dns_addr = loopback_addr;
 200         fprintf (stderr, "Warning: No DNS servers found\n");
 201     }
 202
 203     if (special_ip)
 204         slirp_special_ip = special_ip;
 205
 206     inet_aton(slirp_special_ip, &special_addr);
 207     alias_addr.s_addr = special_addr.s_addr | htonl(CTL_ALIAS);
 208     getouraddr();
 209     register_savevm("slirp", 0, 1, slirp_state_save, slirp_state_load, NULL);
 210 }
 211
 212 #define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
 213 #define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
 214 #define UPD_NFDS(x) if (nfds < (x)) nfds = (x)
 215
 216 /*
 217  * curtime kept to an accuracy of 1ms
 218  */
 219 #ifdef _WIN32
 220 static void updtime(void)
 221 {
 222     struct _timeb tb;
 223
 224     _ftime(&tb);
 225     curtime = (u_int)tb.time * (u_int)1000;
 226     curtime += (u_int)tb.millitm;
 227 }
 228 #else
 229 static void updtime(void)
 230 {
 231         gettimeofday(&tt, NULL);
 232
 233         curtime = (u_int)tt.tv_sec * (u_int)1000;
 234         curtime += (u_int)tt.tv_usec / (u_int)1000;
 235
 236         if ((tt.tv_usec % 1000) >= 500)
 237            curtime++;
 238 }
 239 #endif
 240
 241 void slirp_select_fill(int *pnfds,
 242                        fd_set *readfds, fd_set *writefds, fd_set *xfds)
 243 {
 244     struct socket *so, *so_next;
 245     struct timeval timeout;
 246     int nfds;
 247     int tmp_time;
 248
 249     /* fail safe */
 250     global_readfds = NULL;
 251     global_writefds = NULL;
 252     global_xfds = NULL;
 253
 254     nfds = *pnfds;
 255         /*
 256          * First, TCP sockets
 257          */
 258         do_slowtimo = 0;
 259         if (link_up) {
 260                 /*
 261                  * *_slowtimo needs calling if there are IP fragments
 262                  * in the fragment queue, or there are TCP connections active
 263                  */
 264                 do_slowtimo = ((tcb.so_next != &tcb) ||
 265                 (&ipq.ip_link != ipq.ip_link.next));
 266
 267                 for (so = tcb.so_next; so != &tcb; so = so_next) {
 268                         so_next = so->so_next;
 269
 270                         /*
 271                          * See if we need a tcp_fasttimo
 272                          */
 273                         if (time_fasttimo == 0 && so->so_tcpcb->t_flags & TF_DELACK)
 274                            time_fasttimo = curtime; /* Flag when we want a fasttimo */
 275
 276                         /*
 277                          * NOFDREF can include still connecting to local-host,
 278                          * newly socreated() sockets etc. Don't want to select these.
 279                          */
 280                         if (so->so_state & SS_NOFDREF || so->s == -1)
 281                            continue;
 282
 283                         /*
 284                          * Set for reading sockets which are accepting
 285                          */
 286                         if (so->so_state & SS_FACCEPTCONN) {
 287                                 FD_SET(so->s, readfds);
 288                                 UPD_NFDS(so->s);
 289                                 continue;
 290                         }
 291
 292                         /*
 293                          * Set for writing sockets which are connecting
 294                          */
 295                         if (so->so_state & SS_ISFCONNECTING) {
 296                                 FD_SET(so->s, writefds);
 297                                 UPD_NFDS(so->s);
 298                                 continue;
 299                         }
 300
 301                         /*
 302                          * Set for writing if we are connected, can send more, and
 303                          * we have something to send
 304                          */
 305                         if (CONN_CANFSEND(so) && so->so_rcv.sb_cc) {
 306                                 FD_SET(so->s, writefds);
 307                                 UPD_NFDS(so->s);
 308                         }
 309
 310                         /*
 311                          * Set for reading (and urgent data) if we are connected, can
 312                          * receive more, and we have room for it XXX /2 ?
 313                          */
 314                         if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2))) {
 315                                 FD_SET(so->s, readfds);
 316                                 FD_SET(so->s, xfds);
 317                                 UPD_NFDS(so->s);
 318                         }
 319                 }
 320
 321                 /*
 322                  * UDP sockets
 323                  */
 324                 for (so = udb.so_next; so != &udb; so = so_next) {
 325                         so_next = so->so_next;
 326
 327                         /*
 328                          * See if it's timed out
 329                          */
 330                         if (so->so_expire) {
 331                                 if (so->so_expire <= curtime) {
 332                                         udp_detach(so);
 333                                         continue;
 334                                 } else
 335                                         do_slowtimo = 1; /* Let socket expire */
 336                         }
 337
 338                         /*
 339                          * When UDP packets are received from over the
 340                          * link, they're sendto()'d straight away, so
 341                          * no need for setting for writing
 342                          * Limit the number of packets queued by this session
 343                          * to 4.  Note that even though we try and limit this
 344                          * to 4 packets, the session could have more queued
 345                          * if the packets needed to be fragmented
 346                          * (XXX <= 4 ?)
 347                          */
 348                         if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4) {
 349                                 FD_SET(so->s, readfds);
 350                                 UPD_NFDS(so->s);
 351                         }
 352                 }
 353         }
 354
 355         /*
 356          * Setup timeout to use minimum CPU usage, especially when idle
 357          */
 358
 359         /*
 360          * First, see the timeout needed by *timo
 361          */
 362         timeout.tv_sec = 0;
 363         timeout.tv_usec = -1;
 364         /*
 365          * If a slowtimo is needed, set timeout to 500ms from the last
 366          * slow timeout. If a fast timeout is needed, set timeout within
 367          * 200ms of when it was requested.
 368          */
 369         if (do_slowtimo) {
 370                 /* XXX + 10000 because some select()'s aren't that accurate */
 371                 timeout.tv_usec = ((500 - (curtime - last_slowtimo)) * 1000) + 10000;
 372                 if (timeout.tv_usec < 0)
 373                    timeout.tv_usec = 0;
 374                 else if (timeout.tv_usec > 510000)
 375                    timeout.tv_usec = 510000;
 376
 377                 /* Can only fasttimo if we also slowtimo */
 378                 if (time_fasttimo) {
 379                         tmp_time = (200 - (curtime - time_fasttimo)) * 1000;
 380                         if (tmp_time < 0)
 381                            tmp_time = 0;
 382
 383                         /* Choose the smallest of the 2 */
 384                         if (tmp_time < timeout.tv_usec)
 385                            timeout.tv_usec = (u_int)tmp_time;
 386                 }
 387         }
 388         *pnfds = nfds;
 389 }
 390
 391 void slirp_select_poll(fd_set *readfds, fd_set *writefds, fd_set *xfds)
 392 {
 393     struct socket *so, *so_next;
 394     int ret;
 395
 396     global_readfds = readfds;
 397     global_writefds = writefds;
 398     global_xfds = xfds;
 399
 400         /* Update time */
 401         updtime();
 402
 403         /*
 404          * See if anything has timed out
 405          */
 406         if (link_up) {
 407                 if (time_fasttimo && ((curtime - time_fasttimo) >= 2)) {
 408                         tcp_fasttimo();
 409                         time_fasttimo = 0;
 410                 }
 411                 if (do_slowtimo && ((curtime - last_slowtimo) >= 499)) {
 412                         ip_slowtimo();
 413                         tcp_slowtimo();
 414                         last_slowtimo = curtime;
 415                 }
 416         }
 417
 418         /*
 419          * Check sockets
 420          */
 421         if (link_up) {
 422                 /*
 423                  * Check TCP sockets
 424                  */
 425                 for (so = tcb.so_next; so != &tcb; so = so_next) {
 426                         so_next = so->so_next;
 427
 428                         /*
 429                          * FD_ISSET is meaningless on these sockets
 430                          * (and they can crash the program)
 431                          */
 432                         if (so->so_state & SS_NOFDREF || so->s == -1)
 433                            continue;
 434
 435                         /*
 436                          * Check for URG data
 437                          * This will soread as well, so no need to
 438                          * test for readfds below if this succeeds
 439                          */
 440                         if (FD_ISSET(so->s, xfds))
 441                            sorecvoob(so);
 442                         /*
 443                          * Check sockets for reading
 444                          */
 445                         else if (FD_ISSET(so->s, readfds)) {
 446                                 /*
 447                                  * Check for incoming connections
 448                                  */
 449                                 if (so->so_state & SS_FACCEPTCONN) {
 450                                         tcp_connect(so);
 451                                         continue;
 452                                 } /* else */
 453                                 ret = soread(so);
 454
 455                                 /* Output it if we read something */
 456                                 if (ret > 0)
 457                                    tcp_output(sototcpcb(so));
 458                         }
 459
 460                         /*
 461                          * Check sockets for writing
 462                          */
 463                         if (FD_ISSET(so->s, writefds)) {
 464                           /*
 465                            * Check for non-blocking, still-connecting sockets
 466                            */
 467                           if (so->so_state & SS_ISFCONNECTING) {
 468                             /* Connected */
 469                             so->so_state &= ~SS_ISFCONNECTING;
 470
 471                             ret = send(so->s, &ret, 0, 0);
 472                             if (ret < 0) {
 473                               /* XXXXX Must fix, zero bytes is a NOP */
 474                               if (errno == EAGAIN || errno == EWOULDBLOCK ||
 475                                   errno == EINPROGRESS || errno == ENOTCONN)
 476                                 continue;
 477
 478                               /* else failed */
 479                               so->so_state = SS_NOFDREF;
 480                             }
 481                             /* else so->so_state &= ~SS_ISFCONNECTING; */
 482
 483                             /*
 484                              * Continue tcp_input
 485                              */
 486                             tcp_input((struct mbuf *)NULL, sizeof(struct ip), so);
 487                             /* continue; */
 488                           } else
 489                             ret = sowrite(so);
 490                           /*
 491                            * XXXXX If we wrote something (a lot), there
 492                            * could be a need for a window update.
 493                            * In the worst case, the remote will send
 494                            * a window probe to get things going again
 495                            */
 496                         }
 497
 498                         /*
 499                          * Probe a still-connecting, non-blocking socket
 500                          * to check if it's still alive
 501                          */
 502 #ifdef PROBE_CONN
 503                         if (so->so_state & SS_ISFCONNECTING) {
 504                           ret = recv(so->s, (char *)&ret, 0,0);
 505
 506                           if (ret < 0) {
 507                             /* XXX */
 508                             if (errno == EAGAIN || errno == EWOULDBLOCK ||
 509                                 errno == EINPROGRESS || errno == ENOTCONN)
 510                               continue; /* Still connecting, continue */
 511
 512                             /* else failed */
 513                             so->so_state = SS_NOFDREF;
 514
 515                             /* tcp_input will take care of it */
 516                           } else {
 517                             ret = send(so->s, &ret, 0,0);
 518                             if (ret < 0) {
 519                               /* XXX */
 520                               if (errno == EAGAIN || errno == EWOULDBLOCK ||
 521                                   errno == EINPROGRESS || errno == ENOTCONN)
 522                                 continue;
 523                               /* else failed */
 524                               so->so_state = SS_NOFDREF;
 525                             } else
 526                               so->so_state &= ~SS_ISFCONNECTING;
 527
 528                           }
 529                           tcp_input((struct mbuf *)NULL, sizeof(struct ip),so);
 530                         } /* SS_ISFCONNECTING */
 531 #endif
 532                 }
 533
 534                 /*
 535                  * Now UDP sockets.
 536                  * Incoming packets are sent straight away, they're not buffered.
 537                  * Incoming UDP data isn't buffered either.
 538                  */
 539                 for (so = udb.so_next; so != &udb; so = so_next) {
 540                         so_next = so->so_next;
 541
 542                         if (so->s != -1 && FD_ISSET(so->s, readfds)) {
 543                             sorecvfrom(so);
 544                         }
 545                 }
 546         }
 547
 548         /*
 549          * See if we can start outputting
 550          */
 551         if (if_queued && link_up)
 552            if_start();
 553
 554         /* clear global file descriptor sets.
 555          * these reside on the stack in vl.c
 556          * so they're unusable if we're not in
 557          * slirp_select_fill or slirp_select_poll.
 558          */
 559          global_readfds = NULL;
 560          global_writefds = NULL;
 561          global_xfds = NULL;
 562 }
 563
 564 #define ETH_ALEN 6
 565 #define ETH_HLEN 14
 566
 567 #define ETH_P_IP        0x0800          /* Internet Protocol packet     */
 568 #define ETH_P_ARP       0x0806          /* Address Resolution packet    */
 569
 570 #define ARPOP_REQUEST   1               /* ARP request                  */
 571 #define ARPOP_REPLY     2               /* ARP reply                    */
 572
 573 struct ethhdr
 574 {
 575         unsigned char   h_dest[ETH_ALEN];       /* destination eth addr */
 576         unsigned char   h_source[ETH_ALEN];     /* source ether addr    */
 577         unsigned short  h_proto;                /* packet type ID field */
 578 };
 579
 580 struct arphdr
 581 {
 582         unsigned short  ar_hrd;         /* format of hardware address   */
 583         unsigned short  ar_pro;         /* format of protocol address   */
 584         unsigned char   ar_hln;         /* length of hardware address   */
 585         unsigned char   ar_pln;         /* length of protocol address   */
 586         unsigned short  ar_op;          /* ARP opcode (command)         */
 587
 588          /*
 589           *      Ethernet looks like this : This bit is variable sized however...
 590           */
 591         unsigned char           ar_sha[ETH_ALEN];       /* sender hardware address      */
 592         unsigned char           ar_sip[4];              /* sender IP address            */
 593         unsigned char           ar_tha[ETH_ALEN];       /* target hardware address      */
 594         unsigned char           ar_tip[4];              /* target IP address            */
 595 };
 596
 597 static void arp_input(const uint8_t *pkt, int pkt_len)
 598 {
 599     struct ethhdr *eh = (struct ethhdr *)pkt;
 600     struct arphdr *ah = (struct arphdr *)(pkt + ETH_HLEN);
 601     uint8_t arp_reply[ETH_HLEN + sizeof(struct arphdr)];
 602     struct ethhdr *reh = (struct ethhdr *)arp_reply;
 603     struct arphdr *rah = (struct arphdr *)(arp_reply + ETH_HLEN);
 604     int ar_op;
 605     struct ex_list *ex_ptr;
 606
 607     ar_op = ntohs(ah->ar_op);
 608     switch(ar_op) {
 609     case ARPOP_REQUEST:
 610         if (!memcmp(ah->ar_tip, &special_addr, 3)) {
 611             if (ah->ar_tip[3] == CTL_DNS || ah->ar_tip[3] == CTL_ALIAS)
 612                 goto arp_ok;
 613             for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
 614                 if (ex_ptr->ex_addr == ah->ar_tip[3])
 615                     goto arp_ok;
 616             }
 617             return;
 618         arp_ok:
 619             /* XXX: make an ARP request to have the client address */
 620             memcpy(client_ethaddr, eh->h_source, ETH_ALEN);
 621
 622             /* ARP request for alias/dns mac address */
 623             memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
 624             memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
 625             reh->h_source[5] = ah->ar_tip[3];
 626             reh->h_proto = htons(ETH_P_ARP);
 627
 628             rah->ar_hrd = htons(1);
 629             rah->ar_pro = htons(ETH_P_IP);
 630             rah->ar_hln = ETH_ALEN;
 631             rah->ar_pln = 4;
 632             rah->ar_op = htons(ARPOP_REPLY);
 633             memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
 634             memcpy(rah->ar_sip, ah->ar_tip, 4);
 635             memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
 636             memcpy(rah->ar_tip, ah->ar_sip, 4);
 637             slirp_output(arp_reply, sizeof(arp_reply));
 638         }
 639         break;
 640     case ARPOP_REPLY:
 641         /* reply to request of client mac address ? */
 642         if (!memcmp(client_ethaddr, zero_ethaddr, ETH_ALEN) &&
 643             !memcmp(ah->ar_sip, &client_ipaddr.s_addr, 4)) {
 644             memcpy(client_ethaddr, ah->ar_sha, ETH_ALEN);
 645         }
 646         break;
 647     default:
 648         break;
 649     }
 650 }
 651
 652 void slirp_input(const uint8_t *pkt, int pkt_len)
 653 {
 654     struct mbuf *m;
 655     int proto;
 656
 657     if (pkt_len < ETH_HLEN)
 658         return;
 659
 660     proto = ntohs(*(uint16_t *)(pkt + 12));
 661     switch(proto) {
 662     case ETH_P_ARP:
 663         arp_input(pkt, pkt_len);
 664         break;
 665     case ETH_P_IP:
 666         m = m_get();
 667         if (!m)
 668             return;
 669         /* Note: we add to align the IP header */
 670         if (M_FREEROOM(m) < pkt_len + 2) {
 671             m_inc(m, pkt_len + 2);
 672         }
 673         m->m_len = pkt_len + 2;
 674         memcpy(m->m_data + 2, pkt, pkt_len);
 675
 676         m->m_data += 2 + ETH_HLEN;
 677         m->m_len -= 2 + ETH_HLEN;
 678
 679         ip_input(m);
 680         break;
 681     default:
 682         break;
 683     }
 684 }
 685
 686 /* output the IP packet to the ethernet device */
 687 void if_encap(const uint8_t *ip_data, int ip_data_len)
 688 {
 689     uint8_t buf[1600];
 690     struct ethhdr *eh = (struct ethhdr *)buf;
 691
 692     if (ip_data_len + ETH_HLEN > sizeof(buf))
 693         return;
 694
 695     if (!memcmp(client_ethaddr, zero_ethaddr, ETH_ALEN)) {
 696         uint8_t arp_req[ETH_HLEN + sizeof(struct arphdr)];
 697         struct ethhdr *reh = (struct ethhdr *)arp_req;
 698         struct arphdr *rah = (struct arphdr *)(arp_req + ETH_HLEN);
 699         const struct ip *iph = (const struct ip *)ip_data;
 700
 701         /* If the client addr is not known, there is no point in
 702            sending the packet to it. Normally the sender should have
 703            done an ARP request to get its MAC address. Here we do it
 704            in place of sending the packet and we hope that the sender
 705            will retry sending its packet. */
 706         memset(reh->h_dest, 0xff, ETH_ALEN);
 707         memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 1);
 708         reh->h_source[5] = CTL_ALIAS;
 709         reh->h_proto = htons(ETH_P_ARP);
 710         rah->ar_hrd = htons(1);
 711         rah->ar_pro = htons(ETH_P_IP);
 712         rah->ar_hln = ETH_ALEN;
 713         rah->ar_pln = 4;
 714         rah->ar_op = htons(ARPOP_REQUEST);
 715         /* source hw addr */
 716         memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN - 1);
 717         rah->ar_sha[5] = CTL_ALIAS;
 718         /* source IP */
 719         memcpy(rah->ar_sip, &alias_addr, 4);
 720         /* target hw addr (none) */
 721         memset(rah->ar_tha, 0, ETH_ALEN);
 722         /* target IP */
 723         memcpy(rah->ar_tip, &iph->ip_dst, 4);
 724         client_ipaddr = iph->ip_dst;
 725         slirp_output(arp_req, sizeof(arp_req));
 726     } else {
 727         memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
 728         memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
 729         /* XXX: not correct */
 730         eh->h_source[5] = CTL_ALIAS;
 731         eh->h_proto = htons(ETH_P_IP);
 732         memcpy(buf + sizeof(struct ethhdr), ip_data, ip_data_len);
 733         slirp_output(buf, ip_data_len + ETH_HLEN);
 734     }
 735 }
 736
 737 int slirp_redir(int is_udp, int host_port,
 738                 struct in_addr guest_addr, int guest_port)
 739 {
 740     if (is_udp) {
 741         if (!udp_listen(htons(host_port), guest_addr.s_addr,
 742                         htons(guest_port), 0))
 743             return -1;
 744     } else {
 745         if (!solisten(htons(host_port), guest_addr.s_addr,
 746                       htons(guest_port), 0))
 747             return -1;
 748     }
 749     return 0;
 750 }
 751
 752 int slirp_add_exec(int do_pty, const void *args, int addr_low_byte,
 753                   int guest_port)
 754 {
 755     return add_exec(&exec_list, do_pty, (char *)args,
 756                     addr_low_byte, htons(guest_port));
 757 }
 758
 759 ssize_t slirp_send(struct socket *so, const void *buf, size_t len, int flags)
 760 {
 761         if (so->s == -1 && so->extra) {
 762                 qemu_chr_write(so->extra, buf, len);
 763                 return len;
 764         }
 765
 766         return send(so->s, buf, len, flags);
 767 }
 768
 769 static struct socket *slirp_find_ctl_socket(int addr_low_byte, int guest_port)
 770 {
 771         struct socket *so;
 772
 773         for (so = tcb.so_next; so != &tcb; so = so->so_next) {
 774                 if ((so->so_faddr.s_addr & htonl(0xffffff00)) ==
 775                                 special_addr.s_addr
 776                                 && (ntohl(so->so_faddr.s_addr) & 0xff) ==
 777                                 addr_low_byte
 778                                 && htons(so->so_fport) == guest_port)
 779                         return so;
 780         }
 781
 782         return NULL;
 783 }
 784
 785 size_t slirp_socket_can_recv(int addr_low_byte, int guest_port)
 786 {
 787         struct iovec iov[2];
 788         struct socket *so;
 789
 790     if (!link_up)
 791         return 0;
 792
 793         so = slirp_find_ctl_socket(addr_low_byte, guest_port);
 794
 795         if (!so || so->so_state & SS_NOFDREF)
 796                 return 0;
 797
 798         if (!CONN_CANFRCV(so) || so->so_snd.sb_cc >= (so->so_snd.sb_datalen/2))
 799                 return 0;
 800
 801         return sopreprbuf(so, iov, NULL);
 802 }
 803
 804 void slirp_socket_recv(int addr_low_byte, int guest_port, const uint8_t *buf,
 805         int size)
 806 {
 807     int ret;
 808     struct socket *so = slirp_find_ctl_socket(addr_low_byte, guest_port);
 809
 810     if (!so)
 811         return;
 812
 813     ret = soreadbuf(so, (const char *)buf, size);
 814
 815     if (ret > 0)
 816         tcp_output(sototcpcb(so));
 817 }
 818
 819 static void slirp_tcp_save(QEMUFile *f, struct tcpcb *tp)
 820 {
 821     int i;
 822
 823     qemu_put_sbe16(f, tp->t_state);
 824     for (i = 0; i < TCPT_NTIMERS; i++)
 825         qemu_put_sbe16(f, tp->t_timer[i]);
 826     qemu_put_sbe16(f, tp->t_rxtshift);
 827     qemu_put_sbe16(f, tp->t_rxtcur);
 828     qemu_put_sbe16(f, tp->t_dupacks);
 829     qemu_put_be16(f, tp->t_maxseg);
 830     qemu_put_sbyte(f, tp->t_force);
 831     qemu_put_be16(f, tp->t_flags);
 832     qemu_put_be32(f, tp->snd_una);
 833     qemu_put_be32(f, tp->snd_nxt);
 834     qemu_put_be32(f, tp->snd_up);
 835     qemu_put_be32(f, tp->snd_wl1);
 836     qemu_put_be32(f, tp->snd_wl2);
 837     qemu_put_be32(f, tp->iss);
 838     qemu_put_be32(f, tp->snd_wnd);
 839     qemu_put_be32(f, tp->rcv_wnd);
 840     qemu_put_be32(f, tp->rcv_nxt);
 841     qemu_put_be32(f, tp->rcv_up);
 842     qemu_put_be32(f, tp->irs);
 843     qemu_put_be32(f, tp->rcv_adv);
 844     qemu_put_be32(f, tp->snd_max);
 845     qemu_put_be32(f, tp->snd_cwnd);
 846     qemu_put_be32(f, tp->snd_ssthresh);
 847     qemu_put_sbe16(f, tp->t_idle);
 848     qemu_put_sbe16(f, tp->t_rtt);
 849     qemu_put_be32(f, tp->t_rtseq);
 850     qemu_put_sbe16(f, tp->t_srtt);
 851     qemu_put_sbe16(f, tp->t_rttvar);
 852     qemu_put_be16(f, tp->t_rttmin);
 853     qemu_put_be32(f, tp->max_sndwnd);
 854     qemu_put_byte(f, tp->t_oobflags);
 855     qemu_put_byte(f, tp->t_iobc);
 856     qemu_put_sbe16(f, tp->t_softerror);
 857     qemu_put_byte(f, tp->snd_scale);
 858     qemu_put_byte(f, tp->rcv_scale);
 859     qemu_put_byte(f, tp->request_r_scale);
 860     qemu_put_byte(f, tp->requested_s_scale);
 861     qemu_put_be32(f, tp->ts_recent);
 862     qemu_put_be32(f, tp->ts_recent_age);
 863     qemu_put_be32(f, tp->last_ack_sent);
 864 }
 865
 866 static void slirp_sbuf_save(QEMUFile *f, struct sbuf *sbuf)
 867 {
 868     uint32_t off;
 869
 870     qemu_put_be32(f, sbuf->sb_cc);
 871     qemu_put_be32(f, sbuf->sb_datalen);
 872     off = (uint32_t)(sbuf->sb_wptr - sbuf->sb_data);
 873     qemu_put_sbe32(f, off);
 874     off = (uint32_t)(sbuf->sb_rptr - sbuf->sb_data);
 875     qemu_put_sbe32(f, off);
 876     qemu_put_buffer(f, (unsigned char*)sbuf->sb_data, sbuf->sb_datalen);
 877 }
 878
 879 static void slirp_socket_save(QEMUFile *f, struct socket *so)
 880 {
 881     qemu_put_be32(f, so->so_urgc);
 882     qemu_put_be32(f, so->so_faddr.s_addr);
 883     qemu_put_be32(f, so->so_laddr.s_addr);
 884     qemu_put_be16(f, so->so_fport);
 885     qemu_put_be16(f, so->so_lport);
 886     qemu_put_byte(f, so->so_iptos);
 887     qemu_put_byte(f, so->so_emu);
 888     qemu_put_byte(f, so->so_type);
 889     qemu_put_be32(f, so->so_state);
 890     slirp_sbuf_save(f, &so->so_rcv);
 891     slirp_sbuf_save(f, &so->so_snd);
 892     slirp_tcp_save(f, so->so_tcpcb);
 893 }
 894
 895 static void slirp_state_save(QEMUFile *f, void *opaque)
 896 {
 897     struct ex_list *ex_ptr;
 898
 899     for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
 900         if (ex_ptr->ex_pty == 3) {
 901             struct socket *so;
 902             so = slirp_find_ctl_socket(ex_ptr->ex_addr, ntohs(ex_ptr->ex_fport));
 903             if (!so)
 904                 continue;
 905
 906             qemu_put_byte(f, 42);
 907             slirp_socket_save(f, so);
 908         }
 909     qemu_put_byte(f, 0);
 910 }
 911
 912 static void slirp_tcp_load(QEMUFile *f, struct tcpcb *tp)
 913 {
 914     int i;
 915
 916     tp->t_state = qemu_get_sbe16(f);
 917     for (i = 0; i < TCPT_NTIMERS; i++)
 918         tp->t_timer[i] = qemu_get_sbe16(f);
 919     tp->t_rxtshift = qemu_get_sbe16(f);
 920     tp->t_rxtcur = qemu_get_sbe16(f);
 921     tp->t_dupacks = qemu_get_sbe16(f);
 922     tp->t_maxseg = qemu_get_be16(f);
 923     tp->t_force = qemu_get_sbyte(f);
 924     tp->t_flags = qemu_get_be16(f);
 925     tp->snd_una = qemu_get_be32(f);
 926     tp->snd_nxt = qemu_get_be32(f);
 927     tp->snd_up = qemu_get_be32(f);
 928     tp->snd_wl1 = qemu_get_be32(f);
 929     tp->snd_wl2 = qemu_get_be32(f);
 930     tp->iss = qemu_get_be32(f);
 931     tp->snd_wnd = qemu_get_be32(f);
 932     tp->rcv_wnd = qemu_get_be32(f);
 933     tp->rcv_nxt = qemu_get_be32(f);
 934     tp->rcv_up = qemu_get_be32(f);
 935     tp->irs = qemu_get_be32(f);
 936     tp->rcv_adv = qemu_get_be32(f);
 937     tp->snd_max = qemu_get_be32(f);
 938     tp->snd_cwnd = qemu_get_be32(f);
 939     tp->snd_ssthresh = qemu_get_be32(f);
 940     tp->t_idle = qemu_get_sbe16(f);
 941     tp->t_rtt = qemu_get_sbe16(f);
 942     tp->t_rtseq = qemu_get_be32(f);
 943     tp->t_srtt = qemu_get_sbe16(f);
 944     tp->t_rttvar = qemu_get_sbe16(f);
 945     tp->t_rttmin = qemu_get_be16(f);
 946     tp->max_sndwnd = qemu_get_be32(f);
 947     tp->t_oobflags = qemu_get_byte(f);
 948     tp->t_iobc = qemu_get_byte(f);
 949     tp->t_softerror = qemu_get_sbe16(f);
 950     tp->snd_scale = qemu_get_byte(f);
 951     tp->rcv_scale = qemu_get_byte(f);
 952     tp->request_r_scale = qemu_get_byte(f);
 953     tp->requested_s_scale = qemu_get_byte(f);
 954     tp->ts_recent = qemu_get_be32(f);
 955     tp->ts_recent_age = qemu_get_be32(f);
 956     tp->last_ack_sent = qemu_get_be32(f);
 957     tcp_template(tp);
 958 }
 959
 960 static int slirp_sbuf_load(QEMUFile *f, struct sbuf *sbuf)
 961 {
 962     uint32_t off, sb_cc, sb_datalen;
 963
 964     sb_cc = qemu_get_be32(f);
 965     sb_datalen = qemu_get_be32(f);
 966
 967     sbreserve(sbuf, sb_datalen);
 968
 969     if (sbuf->sb_datalen != sb_datalen)
 970         return -ENOMEM;
 971
 972     sbuf->sb_cc = sb_cc;
 973
 974     off = qemu_get_sbe32(f);
 975     sbuf->sb_wptr = sbuf->sb_data + off;
 976     off = qemu_get_sbe32(f);
 977     sbuf->sb_rptr = sbuf->sb_data + off;
 978     qemu_get_buffer(f, (unsigned char*)sbuf->sb_data, sbuf->sb_datalen);
 979
 980     return 0;
 981 }
 982
 983 static int slirp_socket_load(QEMUFile *f, struct socket *so)
 984 {
 985     if (tcp_attach(so) < 0)
 986         return -ENOMEM;
 987
 988     so->so_urgc = qemu_get_be32(f);
 989     so->so_faddr.s_addr = qemu_get_be32(f);
 990     so->so_laddr.s_addr = qemu_get_be32(f);
 991     so->so_fport = qemu_get_be16(f);
 992     so->so_lport = qemu_get_be16(f);
 993     so->so_iptos = qemu_get_byte(f);
 994     so->so_emu = qemu_get_byte(f);
 995     so->so_type = qemu_get_byte(f);
 996     so->so_state = qemu_get_be32(f);
 997     if (slirp_sbuf_load(f, &so->so_rcv) < 0)
 998         return -ENOMEM;
 999     if (slirp_sbuf_load(f, &so->so_snd) < 0)
1000         return -ENOMEM;
1001     slirp_tcp_load(f, so->so_tcpcb);
1002
1003     return 0;
1004 }
1005
1006 static int slirp_state_load(QEMUFile *f, void *opaque, int version_id)
1007 {
1008     struct ex_list *ex_ptr;
1009     int r;
1010
1011     while ((r = qemu_get_byte(f))) {
1012         int ret;
1013         struct socket *so = socreate();
1014
1015         if (!so)
1016             return -ENOMEM;
1017
1018         ret = slirp_socket_load(f, so);
1019
1020         if (ret < 0)
1021             return ret;
1022
1023         if ((so->so_faddr.s_addr & htonl(0xffffff00)) != special_addr.s_addr)
1024             return -EINVAL;
1025
1026         for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1027             if (ex_ptr->ex_pty == 3 &&
1028                     (ntohl(so->so_faddr.s_addr) & 0xff) == ex_ptr->ex_addr &&
1029                     so->so_fport == ex_ptr->ex_fport)
1030                 break;
1031
1032         if (!ex_ptr)
1033             return -EINVAL;
1034
1035         so->extra = (void *)ex_ptr->ex_exec;
1036     }
1037
1038     return 0;
1039 }