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/osdep.h"
  25 #include "qemu-common.h"
  26 #include "qemu/timer.h"
  27 #include "qemu/error-report.h"
  28 #include "chardev/char-fe.h"
  29 #include "migration/register.h"
  30 #include "slirp.h"
  31 #include "hw/hw.h"
  32 #include "qemu/cutils.h"
  33
  34 #ifndef _WIN32
  35 #include <net/if.h>
  36 #endif
  37
  38 /* host loopback address */
  39 struct in_addr loopback_addr;
  40 /* host loopback network mask */
  41 unsigned long loopback_mask;
  42
  43 /* emulated hosts use the MAC addr 52:55:IP:IP:IP:IP */
  44 static const uint8_t special_ethaddr[ETH_ALEN] = {
  45     0x52, 0x55, 0x00, 0x00, 0x00, 0x00
  46 };
  47
  48 u_int curtime;
  49
  50 static QTAILQ_HEAD(, Slirp) slirp_instances =
  51     QTAILQ_HEAD_INITIALIZER(slirp_instances);
  52
  53 static struct in_addr dns_addr;
  54 #ifndef _WIN32
  55 static struct in6_addr dns6_addr;
  56 #endif
  57 static u_int dns_addr_time;
  58 #ifndef _WIN32
  59 static u_int dns6_addr_time;
  60 #endif
  61
  62 #define TIMEOUT_FAST 2  /* milliseconds */
  63 #define TIMEOUT_SLOW 499  /* milliseconds */
  64 /* for the aging of certain requests like DNS */
  65 #define TIMEOUT_DEFAULT 1000  /* milliseconds */
  66
  67 #ifdef _WIN32
  68
  69 int get_dns_addr(struct in_addr *pdns_addr)
  70 {
  71     FIXED_INFO *FixedInfo=NULL;
  72     ULONG    BufLen;
  73     DWORD    ret;
  74     IP_ADDR_STRING *pIPAddr;
  75     struct in_addr tmp_addr;
  76
  77     if (dns_addr.s_addr != 0 && (curtime - dns_addr_time) < TIMEOUT_DEFAULT) {
  78         *pdns_addr = dns_addr;
  79         return 0;
  80     }
  81
  82     FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
  83     BufLen = sizeof(FIXED_INFO);
  84
  85     if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen)) {
  86         if (FixedInfo) {
  87             GlobalFree(FixedInfo);
  88             FixedInfo = NULL;
  89         }
  90         FixedInfo = GlobalAlloc(GPTR, BufLen);
  91     }
  92
  93     if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS) {
  94         printf("GetNetworkParams failed. ret = %08x\n", (u_int)ret );
  95         if (FixedInfo) {
  96             GlobalFree(FixedInfo);
  97             FixedInfo = NULL;
  98         }
  99         return -1;
 100     }
 101
 102     pIPAddr = &(FixedInfo->DnsServerList);
 103     inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
 104     *pdns_addr = tmp_addr;
 105     dns_addr = tmp_addr;
 106     dns_addr_time = curtime;
 107     if (FixedInfo) {
 108         GlobalFree(FixedInfo);
 109         FixedInfo = NULL;
 110     }
 111     return 0;
 112 }
 113
 114 int get_dns6_addr(struct in6_addr *pdns6_addr, uint32_t *scope_id)
 115 {
 116     return -1;
 117 }
 118
 119 static void winsock_cleanup(void)
 120 {
 121     WSACleanup();
 122 }
 123
 124 #else
 125
 126 static int get_dns_addr_cached(void *pdns_addr, void *cached_addr,
 127                                socklen_t addrlen,
 128                                struct stat *cached_stat, u_int *cached_time)
 129 {
 130     struct stat old_stat;
 131     if (curtime - *cached_time < TIMEOUT_DEFAULT) {
 132         memcpy(pdns_addr, cached_addr, addrlen);
 133         return 0;
 134     }
 135     old_stat = *cached_stat;
 136     if (stat("/etc/resolv.conf", cached_stat) != 0) {
 137         return -1;
 138     }
 139     if (cached_stat->st_dev == old_stat.st_dev
 140         && cached_stat->st_ino == old_stat.st_ino
 141         && cached_stat->st_size == old_stat.st_size
 142         && cached_stat->st_mtime == old_stat.st_mtime) {
 143         memcpy(pdns_addr, cached_addr, addrlen);
 144         return 0;
 145     }
 146     return 1;
 147 }
 148
 149 static int get_dns_addr_resolv_conf(int af, void *pdns_addr, void *cached_addr,
 150                                     socklen_t addrlen, uint32_t *scope_id,
 151                                     u_int *cached_time)
 152 {
 153     char buff[512];
 154     char buff2[257];
 155     FILE *f;
 156     int found = 0;
 157     void *tmp_addr = alloca(addrlen);
 158     unsigned if_index;
 159
 160     f = fopen("/etc/resolv.conf", "r");
 161     if (!f)
 162         return -1;
 163
 164 #ifdef DEBUG
 165     fprintf(stderr, "IP address of your DNS(s): ");
 166 #endif
 167     while (fgets(buff, 512, f) != NULL) {
 168         if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1) {
 169             char *c = strchr(buff2, '%');
 170             if (c) {
 171                 if_index = if_nametoindex(c + 1);
 172                 *c = '\0';
 173             } else {
 174                 if_index = 0;
 175             }
 176
 177             if (!inet_pton(af, buff2, tmp_addr)) {
 178                 continue;
 179             }
 180             /* If it's the first one, set it to dns_addr */
 181             if (!found) {
 182                 memcpy(pdns_addr, tmp_addr, addrlen);
 183                 memcpy(cached_addr, tmp_addr, addrlen);
 184                 if (scope_id) {
 185                     *scope_id = if_index;
 186                 }
 187                 *cached_time = curtime;
 188             }
 189 #ifdef DEBUG
 190             else
 191                 fprintf(stderr, ", ");
 192 #endif
 193             if (++found > 3) {
 194 #ifdef DEBUG
 195                 fprintf(stderr, "(more)");
 196 #endif
 197                 break;
 198             }
 199 #ifdef DEBUG
 200             else {
 201                 char s[INET6_ADDRSTRLEN];
 202                 const char *res = inet_ntop(af, tmp_addr, s, sizeof(s));
 203                 if (!res) {
 204                     res = "(string conversion error)";
 205                 }
 206                 fprintf(stderr, "%s", res);
 207             }
 208 #endif
 209         }
 210     }
 211     fclose(f);
 212     if (!found)
 213         return -1;
 214     return 0;
 215 }
 216
 217 int get_dns_addr(struct in_addr *pdns_addr)
 218 {
 219     static struct stat dns_addr_stat;
 220
 221     if (dns_addr.s_addr != 0) {
 222         int ret;
 223         ret = get_dns_addr_cached(pdns_addr, &dns_addr, sizeof(dns_addr),
 224                                   &dns_addr_stat, &dns_addr_time);
 225         if (ret <= 0) {
 226             return ret;
 227         }
 228     }
 229     return get_dns_addr_resolv_conf(AF_INET, pdns_addr, &dns_addr,
 230                                     sizeof(dns_addr), NULL, &dns_addr_time);
 231 }
 232
 233 int get_dns6_addr(struct in6_addr *pdns6_addr, uint32_t *scope_id)
 234 {
 235     static struct stat dns6_addr_stat;
 236
 237     if (!in6_zero(&dns6_addr)) {
 238         int ret;
 239         ret = get_dns_addr_cached(pdns6_addr, &dns6_addr, sizeof(dns6_addr),
 240                                   &dns6_addr_stat, &dns6_addr_time);
 241         if (ret <= 0) {
 242             return ret;
 243         }
 244     }
 245     return get_dns_addr_resolv_conf(AF_INET6, pdns6_addr, &dns6_addr,
 246                                     sizeof(dns6_addr),
 247                                     scope_id, &dns6_addr_time);
 248 }
 249
 250 #endif
 251
 252 static void slirp_init_once(void)
 253 {
 254     static int initialized;
 255 #ifdef _WIN32
 256     WSADATA Data;
 257 #endif
 258
 259     if (initialized) {
 260         return;
 261     }
 262     initialized = 1;
 263
 264 #ifdef _WIN32
 265     WSAStartup(MAKEWORD(2,0), &Data);
 266     atexit(winsock_cleanup);
 267 #endif
 268
 269     loopback_addr.s_addr = htonl(INADDR_LOOPBACK);
 270     loopback_mask = htonl(IN_CLASSA_NET);
 271 }
 272
 273 static void slirp_state_save(QEMUFile *f, void *opaque);
 274 static int slirp_state_load(QEMUFile *f, void *opaque, int version_id);
 275
 276 static SaveVMHandlers savevm_slirp_state = {
 277     .save_state = slirp_state_save,
 278     .load_state = slirp_state_load,
 279 };
 280
 281 Slirp *slirp_init(int restricted, bool in_enabled, struct in_addr vnetwork,
 282                   struct in_addr vnetmask, struct in_addr vhost,
 283                   bool in6_enabled,
 284                   struct in6_addr vprefix_addr6, uint8_t vprefix_len,
 285                   struct in6_addr vhost6, const char *vhostname,
 286                   const char *tftp_server_name,
 287                   const char *tftp_path, const char *bootfile,
 288                   struct in_addr vdhcp_start, struct in_addr vnameserver,
 289                   struct in6_addr vnameserver6, const char **vdnssearch,
 290                   const char *vdomainname, void *opaque)
 291 {
 292     Slirp *slirp = g_malloc0(sizeof(Slirp));
 293
 294     slirp_init_once();
 295
 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         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     register_savevm_live(NULL, "slirp", 0, 4, &savevm_slirp_state, slirp);
 334
 335     QTAILQ_INSERT_TAIL(&slirp_instances, slirp, entry);
 336
 337     return slirp;
 338 }
 339
 340 void slirp_cleanup(Slirp *slirp)
 341 {
 342     QTAILQ_REMOVE(&slirp_instances, slirp, entry);
 343
 344     unregister_savevm(NULL, "slirp", slirp);
 345
 346     ip_cleanup(slirp);
 347     ip6_cleanup(slirp);
 348     m_cleanup(slirp);
 349
 350     g_rand_free(slirp->grand);
 351
 352     g_free(slirp->vdnssearch);
 353     g_free(slirp->tftp_prefix);
 354     g_free(slirp->bootp_filename);
 355     g_free(slirp->vdomainname);
 356     g_free(slirp);
 357 }
 358
 359 #define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
 360 #define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
 361
 362 static void slirp_update_timeout(uint32_t *timeout)
 363 {
 364     Slirp *slirp;
 365     uint32_t t;
 366
 367     if (*timeout <= TIMEOUT_FAST) {
 368         return;
 369     }
 370
 371     t = MIN(1000, *timeout);
 372
 373     /* If we have tcp timeout with slirp, then we will fill @timeout with
 374      * more precise value.
 375      */
 376     QTAILQ_FOREACH(slirp, &slirp_instances, entry) {
 377         if (slirp->time_fasttimo) {
 378             *timeout = TIMEOUT_FAST;
 379             return;
 380         }
 381         if (slirp->do_slowtimo) {
 382             t = MIN(TIMEOUT_SLOW, t);
 383         }
 384     }
 385     *timeout = t;
 386 }
 387
 388 void slirp_pollfds_fill(GArray *pollfds, uint32_t *timeout)
 389 {
 390     Slirp *slirp;
 391     struct socket *so, *so_next;
 392
 393     if (QTAILQ_EMPTY(&slirp_instances)) {
 394         return;
 395     }
 396
 397     /*
 398      * First, TCP sockets
 399      */
 400
 401     QTAILQ_FOREACH(slirp, &slirp_instances, entry) {
 402         /*
 403          * *_slowtimo needs calling if there are IP fragments
 404          * in the fragment queue, or there are TCP connections active
 405          */
 406         slirp->do_slowtimo = ((slirp->tcb.so_next != &slirp->tcb) ||
 407                 (&slirp->ipq.ip_link != slirp->ipq.ip_link.next));
 408
 409         for (so = slirp->tcb.so_next; so != &slirp->tcb;
 410                 so = so_next) {
 411             int events = 0;
 412
 413             so_next = so->so_next;
 414
 415             so->pollfds_idx = -1;
 416
 417             /*
 418              * See if we need a tcp_fasttimo
 419              */
 420             if (slirp->time_fasttimo == 0 &&
 421                 so->so_tcpcb->t_flags & TF_DELACK) {
 422                 slirp->time_fasttimo = curtime; /* Flag when want a fasttimo */
 423             }
 424
 425             /*
 426              * NOFDREF can include still connecting to local-host,
 427              * newly socreated() sockets etc. Don't want to select these.
 428              */
 429             if (so->so_state & SS_NOFDREF || so->s == -1) {
 430                 continue;
 431             }
 432
 433             /*
 434              * Set for reading sockets which are accepting
 435              */
 436             if (so->so_state & SS_FACCEPTCONN) {
 437                 GPollFD pfd = {
 438                     .fd = so->s,
 439                     .events = G_IO_IN | G_IO_HUP | G_IO_ERR,
 440                 };
 441                 so->pollfds_idx = pollfds->len;
 442                 g_array_append_val(pollfds, pfd);
 443                 continue;
 444             }
 445
 446             /*
 447              * Set for writing sockets which are connecting
 448              */
 449             if (so->so_state & SS_ISFCONNECTING) {
 450                 GPollFD pfd = {
 451                     .fd = so->s,
 452                     .events = G_IO_OUT | G_IO_ERR,
 453                 };
 454                 so->pollfds_idx = pollfds->len;
 455                 g_array_append_val(pollfds, pfd);
 456                 continue;
 457             }
 458
 459             /*
 460              * Set for writing if we are connected, can send more, and
 461              * we have something to send
 462              */
 463             if (CONN_CANFSEND(so) && so->so_rcv.sb_cc) {
 464                 events |= G_IO_OUT | G_IO_ERR;
 465             }
 466
 467             /*
 468              * Set for reading (and urgent data) if we are connected, can
 469              * receive more, and we have room for it XXX /2 ?
 470              */
 471             if (CONN_CANFRCV(so) &&
 472                 (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2))) {
 473                 events |= G_IO_IN | G_IO_HUP | G_IO_ERR | G_IO_PRI;
 474             }
 475
 476             if (events) {
 477                 GPollFD pfd = {
 478                     .fd = so->s,
 479                     .events = events,
 480                 };
 481                 so->pollfds_idx = pollfds->len;
 482                 g_array_append_val(pollfds, pfd);
 483             }
 484         }
 485
 486         /*
 487          * UDP sockets
 488          */
 489         for (so = slirp->udb.so_next; so != &slirp->udb;
 490                 so = so_next) {
 491             so_next = so->so_next;
 492
 493             so->pollfds_idx = -1;
 494
 495             /*
 496              * See if it's timed out
 497              */
 498             if (so->so_expire) {
 499                 if (so->so_expire <= curtime) {
 500                     udp_detach(so);
 501                     continue;
 502                 } else {
 503                     slirp->do_slowtimo = true; /* Let socket expire */
 504                 }
 505             }
 506
 507             /*
 508              * When UDP packets are received from over the
 509              * link, they're sendto()'d straight away, so
 510              * no need for setting for writing
 511              * Limit the number of packets queued by this session
 512              * to 4.  Note that even though we try and limit this
 513              * to 4 packets, the session could have more queued
 514              * if the packets needed to be fragmented
 515              * (XXX <= 4 ?)
 516              */
 517             if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4) {
 518                 GPollFD pfd = {
 519                     .fd = so->s,
 520                     .events = G_IO_IN | G_IO_HUP | G_IO_ERR,
 521                 };
 522                 so->pollfds_idx = pollfds->len;
 523                 g_array_append_val(pollfds, pfd);
 524             }
 525         }
 526
 527         /*
 528          * ICMP sockets
 529          */
 530         for (so = slirp->icmp.so_next; so != &slirp->icmp;
 531                 so = so_next) {
 532             so_next = so->so_next;
 533
 534             so->pollfds_idx = -1;
 535
 536             /*
 537              * See if it's timed out
 538              */
 539             if (so->so_expire) {
 540                 if (so->so_expire <= curtime) {
 541                     icmp_detach(so);
 542                     continue;
 543                 } else {
 544                     slirp->do_slowtimo = true; /* Let socket expire */
 545                 }
 546             }
 547
 548             if (so->so_state & SS_ISFCONNECTED) {
 549                 GPollFD pfd = {
 550                     .fd = so->s,
 551                     .events = G_IO_IN | G_IO_HUP | G_IO_ERR,
 552                 };
 553                 so->pollfds_idx = pollfds->len;
 554                 g_array_append_val(pollfds, pfd);
 555             }
 556         }
 557     }
 558     slirp_update_timeout(timeout);
 559 }
 560
 561 void slirp_pollfds_poll(GArray *pollfds, int select_error)
 562 {
 563     Slirp *slirp;
 564     struct socket *so, *so_next;
 565     int ret;
 566
 567     if (QTAILQ_EMPTY(&slirp_instances)) {
 568         return;
 569     }
 570
 571     curtime = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
 572
 573     QTAILQ_FOREACH(slirp, &slirp_instances, entry) {
 574         /*
 575          * See if anything has timed out
 576          */
 577         if (slirp->time_fasttimo &&
 578             ((curtime - slirp->time_fasttimo) >= TIMEOUT_FAST)) {
 579             tcp_fasttimo(slirp);
 580             slirp->time_fasttimo = 0;
 581         }
 582         if (slirp->do_slowtimo &&
 583             ((curtime - slirp->last_slowtimo) >= TIMEOUT_SLOW)) {
 584             ip_slowtimo(slirp);
 585             tcp_slowtimo(slirp);
 586             slirp->last_slowtimo = curtime;
 587         }
 588
 589         /*
 590          * Check sockets
 591          */
 592         if (!select_error) {
 593             /*
 594              * Check TCP sockets
 595              */
 596             for (so = slirp->tcb.so_next; so != &slirp->tcb;
 597                     so = so_next) {
 598                 int revents;
 599
 600                 so_next = so->so_next;
 601
 602                 revents = 0;
 603                 if (so->pollfds_idx != -1) {
 604                     revents = g_array_index(pollfds, GPollFD,
 605                                             so->pollfds_idx).revents;
 606                 }
 607
 608                 if (so->so_state & SS_NOFDREF || so->s == -1) {
 609                     continue;
 610                 }
 611
 612                 /*
 613                  * Check for URG data
 614                  * This will soread as well, so no need to
 615                  * test for G_IO_IN below if this succeeds
 616                  */
 617                 if (revents & G_IO_PRI) {
 618                     ret = sorecvoob(so);
 619                     if (ret < 0) {
 620                         /* Socket error might have resulted in the socket being
 621                          * removed, do not try to do anything more with it. */
 622                         continue;
 623                     }
 624                 }
 625                 /*
 626                  * Check sockets for reading
 627                  */
 628                 else if (revents & (G_IO_IN | G_IO_HUP | G_IO_ERR)) {
 629                     /*
 630                      * Check for incoming connections
 631                      */
 632                     if (so->so_state & SS_FACCEPTCONN) {
 633                         tcp_connect(so);
 634                         continue;
 635                     } /* else */
 636                     ret = soread(so);
 637
 638                     /* Output it if we read something */
 639                     if (ret > 0) {
 640                         tcp_output(sototcpcb(so));
 641                     }
 642                     if (ret < 0) {
 643                         /* Socket error might have resulted in the socket being
 644                          * removed, do not try to do anything more with it. */
 645                         continue;
 646                     }
 647                 }
 648
 649                 /*
 650                  * Check sockets for writing
 651                  */
 652                 if (!(so->so_state & SS_NOFDREF) &&
 653                         (revents & (G_IO_OUT | G_IO_ERR))) {
 654                     /*
 655                      * Check for non-blocking, still-connecting sockets
 656                      */
 657                     if (so->so_state & SS_ISFCONNECTING) {
 658                         /* Connected */
 659                         so->so_state &= ~SS_ISFCONNECTING;
 660
 661                         ret = send(so->s, (const void *) &ret, 0, 0);
 662                         if (ret < 0) {
 663                             /* XXXXX Must fix, zero bytes is a NOP */
 664                             if (errno == EAGAIN || errno == EWOULDBLOCK ||
 665                                 errno == EINPROGRESS || errno == ENOTCONN) {
 666                                 continue;
 667                             }
 668
 669                             /* else failed */
 670                             so->so_state &= SS_PERSISTENT_MASK;
 671                             so->so_state |= SS_NOFDREF;
 672                         }
 673                         /* else so->so_state &= ~SS_ISFCONNECTING; */
 674
 675                         /*
 676                          * Continue tcp_input
 677                          */
 678                         tcp_input((struct mbuf *)NULL, sizeof(struct ip), so,
 679                                   so->so_ffamily);
 680                         /* continue; */
 681                     } else {
 682                         ret = sowrite(so);
 683                         if (ret > 0) {
 684                             /* Call tcp_output in case we need to send a window
 685                              * update to the guest, otherwise it will be stuck
 686                              * until it sends a window probe. */
 687                             tcp_output(sototcpcb(so));
 688                         }
 689                     }
 690                 }
 691
 692                 /*
 693                  * Probe a still-connecting, non-blocking socket
 694                  * to check if it's still alive
 695                  */
 696 #ifdef PROBE_CONN
 697                 if (so->so_state & SS_ISFCONNECTING) {
 698                     ret = qemu_recv(so->s, &ret, 0, 0);
 699
 700                     if (ret < 0) {
 701                         /* XXX */
 702                         if (errno == EAGAIN || errno == EWOULDBLOCK ||
 703                             errno == EINPROGRESS || errno == ENOTCONN) {
 704                             continue; /* Still connecting, continue */
 705                         }
 706
 707                         /* else failed */
 708                         so->so_state &= SS_PERSISTENT_MASK;
 709                         so->so_state |= SS_NOFDREF;
 710
 711                         /* tcp_input will take care of it */
 712                     } else {
 713                         ret = send(so->s, &ret, 0, 0);
 714                         if (ret < 0) {
 715                             /* XXX */
 716                             if (errno == EAGAIN || errno == EWOULDBLOCK ||
 717                                 errno == EINPROGRESS || errno == ENOTCONN) {
 718                                 continue;
 719                             }
 720                             /* else failed */
 721                             so->so_state &= SS_PERSISTENT_MASK;
 722                             so->so_state |= SS_NOFDREF;
 723                         } else {
 724                             so->so_state &= ~SS_ISFCONNECTING;
 725                         }
 726
 727                     }
 728                     tcp_input((struct mbuf *)NULL, sizeof(struct ip), so,
 729                               so->so_ffamily);
 730                 } /* SS_ISFCONNECTING */
 731 #endif
 732             }
 733
 734             /*
 735              * Now UDP sockets.
 736              * Incoming packets are sent straight away, they're not buffered.
 737              * Incoming UDP data isn't buffered either.
 738              */
 739             for (so = slirp->udb.so_next; so != &slirp->udb;
 740                     so = so_next) {
 741                 int revents;
 742
 743                 so_next = so->so_next;
 744
 745                 revents = 0;
 746                 if (so->pollfds_idx != -1) {
 747                     revents = g_array_index(pollfds, GPollFD,
 748                             so->pollfds_idx).revents;
 749                 }
 750
 751                 if (so->s != -1 &&
 752                     (revents & (G_IO_IN | G_IO_HUP | G_IO_ERR))) {
 753                     sorecvfrom(so);
 754                 }
 755             }
 756
 757             /*
 758              * Check incoming ICMP relies.
 759              */
 760             for (so = slirp->icmp.so_next; so != &slirp->icmp;
 761                     so = so_next) {
 762                     int revents;
 763
 764                     so_next = so->so_next;
 765
 766                     revents = 0;
 767                     if (so->pollfds_idx != -1) {
 768                         revents = g_array_index(pollfds, GPollFD,
 769                                                 so->pollfds_idx).revents;
 770                     }
 771
 772                     if (so->s != -1 &&
 773                         (revents & (G_IO_IN | G_IO_HUP | G_IO_ERR))) {
 774                     icmp_receive(so);
 775                 }
 776             }
 777         }
 778
 779         if_start(slirp);
 780     }
 781 }
 782
 783 static void arp_input(Slirp *slirp, const uint8_t *pkt, int pkt_len)
 784 {
 785     struct slirp_arphdr *ah = (struct slirp_arphdr *)(pkt + ETH_HLEN);
 786     uint8_t arp_reply[MAX(ETH_HLEN + sizeof(struct slirp_arphdr), 64)];
 787     struct ethhdr *reh = (struct ethhdr *)arp_reply;
 788     struct slirp_arphdr *rah = (struct slirp_arphdr *)(arp_reply + ETH_HLEN);
 789     int ar_op;
 790     struct ex_list *ex_ptr;
 791
 792     if (!slirp->in_enabled) {
 793         return;
 794     }
 795
 796     ar_op = ntohs(ah->ar_op);
 797     switch(ar_op) {
 798     case ARPOP_REQUEST:
 799         if (ah->ar_tip == ah->ar_sip) {
 800             /* Gratuitous ARP */
 801             arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
 802             return;
 803         }
 804
 805         if ((ah->ar_tip & slirp->vnetwork_mask.s_addr) ==
 806             slirp->vnetwork_addr.s_addr) {
 807             if (ah->ar_tip == slirp->vnameserver_addr.s_addr ||
 808                 ah->ar_tip == slirp->vhost_addr.s_addr)
 809                 goto arp_ok;
 810             for (ex_ptr = slirp->exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
 811                 if (ex_ptr->ex_addr.s_addr == ah->ar_tip)
 812                     goto arp_ok;
 813             }
 814             return;
 815         arp_ok:
 816             memset(arp_reply, 0, sizeof(arp_reply));
 817
 818             arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
 819
 820             /* ARP request for alias/dns mac address */
 821             memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
 822             memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 4);
 823             memcpy(&reh->h_source[2], &ah->ar_tip, 4);
 824             reh->h_proto = htons(ETH_P_ARP);
 825
 826             rah->ar_hrd = htons(1);
 827             rah->ar_pro = htons(ETH_P_IP);
 828             rah->ar_hln = ETH_ALEN;
 829             rah->ar_pln = 4;
 830             rah->ar_op = htons(ARPOP_REPLY);
 831             memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
 832             rah->ar_sip = ah->ar_tip;
 833             memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
 834             rah->ar_tip = ah->ar_sip;
 835             slirp_output(slirp->opaque, arp_reply, sizeof(arp_reply));
 836         }
 837         break;
 838     case ARPOP_REPLY:
 839         arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
 840         break;
 841     default:
 842         break;
 843     }
 844 }
 845
 846 void slirp_input(Slirp *slirp, const uint8_t *pkt, int pkt_len)
 847 {
 848     struct mbuf *m;
 849     int proto;
 850
 851     if (pkt_len < ETH_HLEN)
 852         return;
 853
 854     proto = ntohs(*(uint16_t *)(pkt + 12));
 855     switch(proto) {
 856     case ETH_P_ARP:
 857         arp_input(slirp, pkt, pkt_len);
 858         break;
 859     case ETH_P_IP:
 860     case ETH_P_IPV6:
 861         m = m_get(slirp);
 862         if (!m)
 863             return;
 864         /* Note: we add 2 to align the IP header on 4 bytes,
 865          * and add the margin for the tcpiphdr overhead  */
 866         if (M_FREEROOM(m) < pkt_len + TCPIPHDR_DELTA + 2) {
 867             m_inc(m, pkt_len + TCPIPHDR_DELTA + 2);
 868         }
 869         m->m_len = pkt_len + TCPIPHDR_DELTA + 2;
 870         memcpy(m->m_data + TCPIPHDR_DELTA + 2, pkt, pkt_len);
 871
 872         m->m_data += TCPIPHDR_DELTA + 2 + ETH_HLEN;
 873         m->m_len -= TCPIPHDR_DELTA + 2 + ETH_HLEN;
 874
 875         if (proto == ETH_P_IP) {
 876             ip_input(m);
 877         } else if (proto == ETH_P_IPV6) {
 878             ip6_input(m);
 879         }
 880         break;
 881
 882     case ETH_P_NCSI:
 883         ncsi_input(slirp, pkt, pkt_len);
 884         break;
 885
 886     default:
 887         break;
 888     }
 889 }
 890
 891 /* Prepare the IPv4 packet to be sent to the ethernet device. Returns 1 if no
 892  * packet should be sent, 0 if the packet must be re-queued, 2 if the packet
 893  * is ready to go.
 894  */
 895 static int if_encap4(Slirp *slirp, struct mbuf *ifm, struct ethhdr *eh,
 896         uint8_t ethaddr[ETH_ALEN])
 897 {
 898     const struct ip *iph = (const struct ip *)ifm->m_data;
 899
 900     if (iph->ip_dst.s_addr == 0) {
 901         /* 0.0.0.0 can not be a destination address, something went wrong,
 902          * avoid making it worse */
 903         return 1;
 904     }
 905     if (!arp_table_search(slirp, iph->ip_dst.s_addr, ethaddr)) {
 906         uint8_t arp_req[ETH_HLEN + sizeof(struct slirp_arphdr)];
 907         struct ethhdr *reh = (struct ethhdr *)arp_req;
 908         struct slirp_arphdr *rah = (struct slirp_arphdr *)(arp_req + ETH_HLEN);
 909
 910         if (!ifm->resolution_requested) {
 911             /* If the client addr is not known, send an ARP request */
 912             memset(reh->h_dest, 0xff, ETH_ALEN);
 913             memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 4);
 914             memcpy(&reh->h_source[2], &slirp->vhost_addr, 4);
 915             reh->h_proto = htons(ETH_P_ARP);
 916             rah->ar_hrd = htons(1);
 917             rah->ar_pro = htons(ETH_P_IP);
 918             rah->ar_hln = ETH_ALEN;
 919             rah->ar_pln = 4;
 920             rah->ar_op = htons(ARPOP_REQUEST);
 921
 922             /* source hw addr */
 923             memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN - 4);
 924             memcpy(&rah->ar_sha[2], &slirp->vhost_addr, 4);
 925
 926             /* source IP */
 927             rah->ar_sip = slirp->vhost_addr.s_addr;
 928
 929             /* target hw addr (none) */
 930             memset(rah->ar_tha, 0, ETH_ALEN);
 931
 932             /* target IP */
 933             rah->ar_tip = iph->ip_dst.s_addr;
 934             slirp->client_ipaddr = iph->ip_dst;
 935             slirp_output(slirp->opaque, arp_req, sizeof(arp_req));
 936             ifm->resolution_requested = true;
 937
 938             /* Expire request and drop outgoing packet after 1 second */
 939             ifm->expiration_date = qemu_clock_get_ns(QEMU_CLOCK_REALTIME) + 1000000000ULL;
 940         }
 941         return 0;
 942     } else {
 943         memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 4);
 944         /* XXX: not correct */
 945         memcpy(&eh->h_source[2], &slirp->vhost_addr, 4);
 946         eh->h_proto = htons(ETH_P_IP);
 947
 948         /* Send this */
 949         return 2;
 950     }
 951 }
 952
 953 /* Prepare the IPv6 packet to be sent to the ethernet device. Returns 1 if no
 954  * packet should be sent, 0 if the packet must be re-queued, 2 if the packet
 955  * is ready to go.
 956  */
 957 static int if_encap6(Slirp *slirp, struct mbuf *ifm, struct ethhdr *eh,
 958         uint8_t ethaddr[ETH_ALEN])
 959 {
 960     const struct ip6 *ip6h = mtod(ifm, const struct ip6 *);
 961     if (!ndp_table_search(slirp, ip6h->ip_dst, ethaddr)) {
 962         if (!ifm->resolution_requested) {
 963             ndp_send_ns(slirp, ip6h->ip_dst);
 964             ifm->resolution_requested = true;
 965             ifm->expiration_date =
 966                 qemu_clock_get_ns(QEMU_CLOCK_REALTIME) + 1000000000ULL;
 967         }
 968         return 0;
 969     } else {
 970         eh->h_proto = htons(ETH_P_IPV6);
 971         in6_compute_ethaddr(ip6h->ip_src, eh->h_source);
 972
 973         /* Send this */
 974         return 2;
 975     }
 976 }
 977
 978 /* Output the IP packet to the ethernet device. Returns 0 if the packet must be
 979  * re-queued.
 980  */
 981 int if_encap(Slirp *slirp, struct mbuf *ifm)
 982 {
 983     uint8_t buf[1600];
 984     struct ethhdr *eh = (struct ethhdr *)buf;
 985     uint8_t ethaddr[ETH_ALEN];
 986     const struct ip *iph = (const struct ip *)ifm->m_data;
 987     int ret;
 988
 989     if (ifm->m_len + ETH_HLEN > sizeof(buf)) {
 990         return 1;
 991     }
 992
 993     switch (iph->ip_v) {
 994     case IPVERSION:
 995         ret = if_encap4(slirp, ifm, eh, ethaddr);
 996         if (ret < 2) {
 997             return ret;
 998         }
 999         break;
1000
1001     case IP6VERSION:
1002         ret = if_encap6(slirp, ifm, eh, ethaddr);
1003         if (ret < 2) {
1004             return ret;
1005         }
1006         break;
1007
1008     default:
1009         g_assert_not_reached();
1010         break;
1011     }
1012
1013     memcpy(eh->h_dest, ethaddr, ETH_ALEN);
1014     DEBUG_ARGS((dfd, " src = %02x:%02x:%02x:%02x:%02x:%02x\n",
1015                 eh->h_source[0], eh->h_source[1], eh->h_source[2],
1016                 eh->h_source[3], eh->h_source[4], eh->h_source[5]));
1017     DEBUG_ARGS((dfd, " dst = %02x:%02x:%02x:%02x:%02x:%02x\n",
1018                 eh->h_dest[0], eh->h_dest[1], eh->h_dest[2],
1019                 eh->h_dest[3], eh->h_dest[4], eh->h_dest[5]));
1020     memcpy(buf + sizeof(struct ethhdr), ifm->m_data, ifm->m_len);
1021     slirp_output(slirp->opaque, buf, ifm->m_len + ETH_HLEN);
1022     return 1;
1023 }
1024
1025 /* Drop host forwarding rule, return 0 if found. */
1026 int slirp_remove_hostfwd(Slirp *slirp, int is_udp, struct in_addr host_addr,
1027                          int host_port)
1028 {
1029     struct socket *so;
1030     struct socket *head = (is_udp ? &slirp->udb : &slirp->tcb);
1031     struct sockaddr_in addr;
1032     int port = htons(host_port);
1033     socklen_t addr_len;
1034
1035     for (so = head->so_next; so != head; so = so->so_next) {
1036         addr_len = sizeof(addr);
1037         if ((so->so_state & SS_HOSTFWD) &&
1038             getsockname(so->s, (struct sockaddr *)&addr, &addr_len) == 0 &&
1039             addr.sin_addr.s_addr == host_addr.s_addr &&
1040             addr.sin_port == port) {
1041             close(so->s);
1042             sofree(so);
1043             return 0;
1044         }
1045     }
1046
1047     return -1;
1048 }
1049
1050 int slirp_add_hostfwd(Slirp *slirp, int is_udp, struct in_addr host_addr,
1051                       int host_port, struct in_addr guest_addr, int guest_port)
1052 {
1053     if (!guest_addr.s_addr) {
1054         guest_addr = slirp->vdhcp_startaddr;
1055     }
1056     if (is_udp) {
1057         if (!udp_listen(slirp, host_addr.s_addr, htons(host_port),
1058                         guest_addr.s_addr, htons(guest_port), SS_HOSTFWD))
1059             return -1;
1060     } else {
1061         if (!tcp_listen(slirp, host_addr.s_addr, htons(host_port),
1062                         guest_addr.s_addr, htons(guest_port), SS_HOSTFWD))
1063             return -1;
1064     }
1065     return 0;
1066 }
1067
1068 int slirp_add_exec(Slirp *slirp, int do_pty, const void *args,
1069                    struct in_addr *guest_addr, int guest_port)
1070 {
1071     if (!guest_addr->s_addr) {
1072         guest_addr->s_addr = slirp->vnetwork_addr.s_addr |
1073             (htonl(0x0204) & ~slirp->vnetwork_mask.s_addr);
1074     }
1075     if ((guest_addr->s_addr & slirp->vnetwork_mask.s_addr) !=
1076         slirp->vnetwork_addr.s_addr ||
1077         guest_addr->s_addr == slirp->vhost_addr.s_addr ||
1078         guest_addr->s_addr == slirp->vnameserver_addr.s_addr) {
1079         return -1;
1080     }
1081     return add_exec(&slirp->exec_list, do_pty, (char *)args, *guest_addr,
1082                     htons(guest_port));
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->extra) {
1088         /* XXX this blocks entire thread. Rewrite to use
1089          * qemu_chr_fe_write and background I/O callbacks */
1090         qemu_chr_fe_write_all(so->extra, buf, len);
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 static 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 static int slirp_tcp_post_load(void *opaque, int version)
1157 {
1158     tcp_template((struct tcpcb *)opaque);
1159
1160     return 0;
1161 }
1162
1163 static const VMStateDescription vmstate_slirp_tcp = {
1164     .name = "slirp-tcp",
1165     .version_id = 0,
1166     .post_load = slirp_tcp_post_load,
1167     .fields = (VMStateField[]) {
1168         VMSTATE_INT16(t_state, struct tcpcb),
1169         VMSTATE_INT16_ARRAY(t_timer, struct tcpcb, TCPT_NTIMERS),
1170         VMSTATE_INT16(t_rxtshift, struct tcpcb),
1171         VMSTATE_INT16(t_rxtcur, struct tcpcb),
1172         VMSTATE_INT16(t_dupacks, struct tcpcb),
1173         VMSTATE_UINT16(t_maxseg, struct tcpcb),
1174         VMSTATE_UINT8(t_force, struct tcpcb),
1175         VMSTATE_UINT16(t_flags, struct tcpcb),
1176         VMSTATE_UINT32(snd_una, struct tcpcb),
1177         VMSTATE_UINT32(snd_nxt, struct tcpcb),
1178         VMSTATE_UINT32(snd_up, struct tcpcb),
1179         VMSTATE_UINT32(snd_wl1, struct tcpcb),
1180         VMSTATE_UINT32(snd_wl2, struct tcpcb),
1181         VMSTATE_UINT32(iss, struct tcpcb),
1182         VMSTATE_UINT32(snd_wnd, struct tcpcb),
1183         VMSTATE_UINT32(rcv_wnd, struct tcpcb),
1184         VMSTATE_UINT32(rcv_nxt, struct tcpcb),
1185         VMSTATE_UINT32(rcv_up, struct tcpcb),
1186         VMSTATE_UINT32(irs, struct tcpcb),
1187         VMSTATE_UINT32(rcv_adv, struct tcpcb),
1188         VMSTATE_UINT32(snd_max, struct tcpcb),
1189         VMSTATE_UINT32(snd_cwnd, struct tcpcb),
1190         VMSTATE_UINT32(snd_ssthresh, struct tcpcb),
1191         VMSTATE_INT16(t_idle, struct tcpcb),
1192         VMSTATE_INT16(t_rtt, struct tcpcb),
1193         VMSTATE_UINT32(t_rtseq, struct tcpcb),
1194         VMSTATE_INT16(t_srtt, struct tcpcb),
1195         VMSTATE_INT16(t_rttvar, struct tcpcb),
1196         VMSTATE_UINT16(t_rttmin, struct tcpcb),
1197         VMSTATE_UINT32(max_sndwnd, struct tcpcb),
1198         VMSTATE_UINT8(t_oobflags, struct tcpcb),
1199         VMSTATE_UINT8(t_iobc, struct tcpcb),
1200         VMSTATE_INT16(t_softerror, struct tcpcb),
1201         VMSTATE_UINT8(snd_scale, struct tcpcb),
1202         VMSTATE_UINT8(rcv_scale, struct tcpcb),
1203         VMSTATE_UINT8(request_r_scale, struct tcpcb),
1204         VMSTATE_UINT8(requested_s_scale, struct tcpcb),
1205         VMSTATE_UINT32(ts_recent, struct tcpcb),
1206         VMSTATE_UINT32(ts_recent_age, struct tcpcb),
1207         VMSTATE_UINT32(last_ack_sent, struct tcpcb),
1208         VMSTATE_END_OF_LIST()
1209     }
1210 };
1211
1212 /* The sbuf has a pair of pointers that are migrated as offsets;
1213  * we calculate the offsets and restore the pointers using
1214  * pre_save/post_load on a tmp structure.
1215  */
1216 struct sbuf_tmp {
1217     struct sbuf *parent;
1218     uint32_t roff, woff;
1219 };
1220
1221 static int sbuf_tmp_pre_save(void *opaque)
1222 {
1223     struct sbuf_tmp *tmp = opaque;
1224     tmp->woff = tmp->parent->sb_wptr - tmp->parent->sb_data;
1225     tmp->roff = tmp->parent->sb_rptr - tmp->parent->sb_data;
1226
1227     return 0;
1228 }
1229
1230 static int sbuf_tmp_post_load(void *opaque, int version)
1231 {
1232     struct sbuf_tmp *tmp = opaque;
1233     uint32_t requested_len = tmp->parent->sb_datalen;
1234
1235     /* Allocate the buffer space used by the field after the tmp */
1236     sbreserve(tmp->parent, tmp->parent->sb_datalen);
1237
1238     if (tmp->parent->sb_datalen != requested_len) {
1239         return -ENOMEM;
1240     }
1241     if (tmp->woff >= requested_len ||
1242         tmp->roff >= requested_len) {
1243         error_report("invalid sbuf offsets r/w=%u/%u len=%u",
1244                      tmp->roff, tmp->woff, requested_len);
1245         return -EINVAL;
1246     }
1247
1248     tmp->parent->sb_wptr = tmp->parent->sb_data + tmp->woff;
1249     tmp->parent->sb_rptr = tmp->parent->sb_data + tmp->roff;
1250
1251     return 0;
1252 }
1253
1254
1255 static const VMStateDescription vmstate_slirp_sbuf_tmp = {
1256     .name = "slirp-sbuf-tmp",
1257     .post_load = sbuf_tmp_post_load,
1258     .pre_save  = sbuf_tmp_pre_save,
1259     .version_id = 0,
1260     .fields = (VMStateField[]) {
1261         VMSTATE_UINT32(woff, struct sbuf_tmp),
1262         VMSTATE_UINT32(roff, struct sbuf_tmp),
1263         VMSTATE_END_OF_LIST()
1264     }
1265 };
1266
1267 static const VMStateDescription vmstate_slirp_sbuf = {
1268     .name = "slirp-sbuf",
1269     .version_id = 0,
1270     .fields = (VMStateField[]) {
1271         VMSTATE_UINT32(sb_cc, struct sbuf),
1272         VMSTATE_UINT32(sb_datalen, struct sbuf),
1273         VMSTATE_WITH_TMP(struct sbuf, struct sbuf_tmp, vmstate_slirp_sbuf_tmp),
1274         VMSTATE_VBUFFER_UINT32(sb_data, struct sbuf, 0, NULL, sb_datalen),
1275         VMSTATE_END_OF_LIST()
1276     }
1277 };
1278
1279 static bool slirp_older_than_v4(void *opaque, int version_id)
1280 {
1281     return version_id < 4;
1282 }
1283
1284 static bool slirp_family_inet(void *opaque, int version_id)
1285 {
1286     union slirp_sockaddr *ssa = (union slirp_sockaddr *)opaque;
1287     return ssa->ss.ss_family == AF_INET;
1288 }
1289
1290 static int slirp_socket_pre_load(void *opaque)
1291 {
1292     struct socket *so = opaque;
1293     if (tcp_attach(so) < 0) {
1294         return -ENOMEM;
1295     }
1296     /* Older versions don't load these fields */
1297     so->so_ffamily = AF_INET;
1298     so->so_lfamily = AF_INET;
1299     return 0;
1300 }
1301
1302 #ifndef _WIN32
1303 #define VMSTATE_SIN4_ADDR(f, s, t) VMSTATE_UINT32_TEST(f, s, t)
1304 #else
1305 /* Win uses u_long rather than uint32_t - but it's still 32bits long */
1306 #define VMSTATE_SIN4_ADDR(f, s, t) VMSTATE_SINGLE_TEST(f, s, t, 0, \
1307                                        vmstate_info_uint32, u_long)
1308 #endif
1309
1310 /* The OS provided ss_family field isn't that portable; it's size
1311  * and type varies (16/8 bit, signed, unsigned)
1312  * and the values it contains aren't fully portable.
1313  */
1314 typedef struct SS_FamilyTmpStruct {
1315     union slirp_sockaddr    *parent;
1316     uint16_t                 portable_family;
1317 } SS_FamilyTmpStruct;
1318
1319 #define SS_FAMILY_MIG_IPV4   2  /* Linux, BSD, Win... */
1320 #define SS_FAMILY_MIG_IPV6  10  /* Linux */
1321 #define SS_FAMILY_MIG_OTHER 0xffff
1322
1323 static int ss_family_pre_save(void *opaque)
1324 {
1325     SS_FamilyTmpStruct *tss = opaque;
1326
1327     tss->portable_family = SS_FAMILY_MIG_OTHER;
1328
1329     if (tss->parent->ss.ss_family == AF_INET) {
1330         tss->portable_family = SS_FAMILY_MIG_IPV4;
1331     } else if (tss->parent->ss.ss_family == AF_INET6) {
1332         tss->portable_family = SS_FAMILY_MIG_IPV6;
1333     }
1334
1335     return 0;
1336 }
1337
1338 static int ss_family_post_load(void *opaque, int version_id)
1339 {
1340     SS_FamilyTmpStruct *tss = opaque;
1341
1342     switch (tss->portable_family) {
1343     case SS_FAMILY_MIG_IPV4:
1344         tss->parent->ss.ss_family = AF_INET;
1345         break;
1346     case SS_FAMILY_MIG_IPV6:
1347     case 23: /* compatibility: AF_INET6 from mingw */
1348     case 28: /* compatibility: AF_INET6 from FreeBSD sys/socket.h */
1349         tss->parent->ss.ss_family = AF_INET6;
1350         break;
1351     default:
1352         error_report("invalid ss_family type %x", tss->portable_family);
1353         return -EINVAL;
1354     }
1355
1356     return 0;
1357 }
1358
1359 static const VMStateDescription vmstate_slirp_ss_family = {
1360     .name = "slirp-socket-addr/ss_family",
1361     .pre_save  = ss_family_pre_save,
1362     .post_load = ss_family_post_load,
1363     .fields = (VMStateField[]) {
1364         VMSTATE_UINT16(portable_family, SS_FamilyTmpStruct),
1365         VMSTATE_END_OF_LIST()
1366     }
1367 };
1368
1369 static const VMStateDescription vmstate_slirp_socket_addr = {
1370     .name = "slirp-socket-addr",
1371     .version_id = 4,
1372     .fields = (VMStateField[]) {
1373         VMSTATE_WITH_TMP(union slirp_sockaddr, SS_FamilyTmpStruct,
1374                             vmstate_slirp_ss_family),
1375         VMSTATE_SIN4_ADDR(sin.sin_addr.s_addr, union slirp_sockaddr,
1376                             slirp_family_inet),
1377         VMSTATE_UINT16_TEST(sin.sin_port, union slirp_sockaddr,
1378                             slirp_family_inet),
1379
1380 #if 0
1381         /* Untested: Needs checking by someone with IPv6 test */
1382         VMSTATE_BUFFER_TEST(sin6.sin6_addr, union slirp_sockaddr,
1383                             slirp_family_inet6),
1384         VMSTATE_UINT16_TEST(sin6.sin6_port, union slirp_sockaddr,
1385                             slirp_family_inet6),
1386         VMSTATE_UINT32_TEST(sin6.sin6_flowinfo, union slirp_sockaddr,
1387                             slirp_family_inet6),
1388         VMSTATE_UINT32_TEST(sin6.sin6_scope_id, union slirp_sockaddr,
1389                             slirp_family_inet6),
1390 #endif
1391
1392         VMSTATE_END_OF_LIST()
1393     }
1394 };
1395
1396 static const VMStateDescription vmstate_slirp_socket = {
1397     .name = "slirp-socket",
1398     .version_id = 4,
1399     .pre_load = slirp_socket_pre_load,
1400     .fields = (VMStateField[]) {
1401         VMSTATE_UINT32(so_urgc, struct socket),
1402         /* Pre-v4 versions */
1403         VMSTATE_SIN4_ADDR(so_faddr.s_addr, struct socket,
1404                             slirp_older_than_v4),
1405         VMSTATE_SIN4_ADDR(so_laddr.s_addr, struct socket,
1406                             slirp_older_than_v4),
1407         VMSTATE_UINT16_TEST(so_fport, struct socket, slirp_older_than_v4),
1408         VMSTATE_UINT16_TEST(so_lport, struct socket, slirp_older_than_v4),
1409         /* v4 and newer */
1410         VMSTATE_STRUCT(fhost, struct socket, 4, vmstate_slirp_socket_addr,
1411                        union slirp_sockaddr),
1412         VMSTATE_STRUCT(lhost, struct socket, 4, vmstate_slirp_socket_addr,
1413                        union slirp_sockaddr),
1414
1415         VMSTATE_UINT8(so_iptos, struct socket),
1416         VMSTATE_UINT8(so_emu, struct socket),
1417         VMSTATE_UINT8(so_type, struct socket),
1418         VMSTATE_INT32(so_state, struct socket),
1419         VMSTATE_STRUCT(so_rcv, struct socket, 0, vmstate_slirp_sbuf,
1420                        struct sbuf),
1421         VMSTATE_STRUCT(so_snd, struct socket, 0, vmstate_slirp_sbuf,
1422                        struct sbuf),
1423         VMSTATE_STRUCT_POINTER(so_tcpcb, struct socket, vmstate_slirp_tcp,
1424                        struct tcpcb),
1425         VMSTATE_END_OF_LIST()
1426     }
1427 };
1428
1429 static const VMStateDescription vmstate_slirp_bootp_client = {
1430     .name = "slirp_bootpclient",
1431     .fields = (VMStateField[]) {
1432         VMSTATE_UINT16(allocated, BOOTPClient),
1433         VMSTATE_BUFFER(macaddr, BOOTPClient),
1434         VMSTATE_END_OF_LIST()
1435     }
1436 };
1437
1438 static const VMStateDescription vmstate_slirp = {
1439     .name = "slirp",
1440     .version_id = 4,
1441     .fields = (VMStateField[]) {
1442         VMSTATE_UINT16_V(ip_id, Slirp, 2),
1443         VMSTATE_STRUCT_ARRAY(bootp_clients, Slirp, NB_BOOTP_CLIENTS, 3,
1444                              vmstate_slirp_bootp_client, BOOTPClient),
1445         VMSTATE_END_OF_LIST()
1446     }
1447 };
1448
1449 static void slirp_state_save(QEMUFile *f, void *opaque)
1450 {
1451     Slirp *slirp = opaque;
1452     struct ex_list *ex_ptr;
1453
1454     for (ex_ptr = slirp->exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1455         if (ex_ptr->ex_pty == 3) {
1456             struct socket *so;
1457             so = slirp_find_ctl_socket(slirp, ex_ptr->ex_addr,
1458                                        ntohs(ex_ptr->ex_fport));
1459             if (!so)
1460                 continue;
1461
1462             qemu_put_byte(f, 42);
1463             vmstate_save_state(f, &vmstate_slirp_socket, so, NULL);
1464         }
1465     qemu_put_byte(f, 0);
1466
1467     vmstate_save_state(f, &vmstate_slirp, slirp, NULL);
1468 }
1469
1470
1471 static int slirp_state_load(QEMUFile *f, void *opaque, int version_id)
1472 {
1473     Slirp *slirp = opaque;
1474     struct ex_list *ex_ptr;
1475
1476     while (qemu_get_byte(f)) {
1477         int ret;
1478         struct socket *so = socreate(slirp);
1479
1480         ret = vmstate_load_state(f, &vmstate_slirp_socket, so, version_id);
1481
1482         if (ret < 0)
1483             return ret;
1484
1485         if ((so->so_faddr.s_addr & slirp->vnetwork_mask.s_addr) !=
1486             slirp->vnetwork_addr.s_addr) {
1487             return -EINVAL;
1488         }
1489         for (ex_ptr = slirp->exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
1490             if (ex_ptr->ex_pty == 3 &&
1491                 so->so_faddr.s_addr == ex_ptr->ex_addr.s_addr &&
1492                 so->so_fport == ex_ptr->ex_fport) {
1493                 break;
1494             }
1495         }
1496         if (!ex_ptr)
1497             return -EINVAL;
1498
1499         so->extra = (void *)ex_ptr->ex_exec;
1500     }
1501
1502     return vmstate_load_state(f, &vmstate_slirp, slirp, version_id);
1503 }