sys/netinet/ipfw/ip_fw2.c

   1 /*-
   2  * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa
   3  *
   4  * Redistribution and use in source and binary forms, with or without
   5  * modification, are permitted provided that the following conditions
   6  * are met:
   7  * 1. Redistributions of source code must retain the above copyright
   8  *    notice, this list of conditions and the following disclaimer.
   9  * 2. Redistributions in binary form must reproduce the above copyright
  10  *    notice, this list of conditions and the following disclaimer in the
  11  *    documentation and/or other materials provided with the distribution.
  12  *
  13  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
  14  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  15  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  16  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
  17  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  18  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  19  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  20  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  21  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  22  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  23  * SUCH DAMAGE.
  24  */
  25
  26 #include <sys/cdefs.h>
  27 __FBSDID("$FreeBSD$");
  28
  29 /*
  30  * The FreeBSD IP packet firewall, main file
  31  */
  32
  33 #if !defined(KLD_MODULE)
  34 #include "opt_ipfw.h"
  35 #include "opt_ipdivert.h"
  36 #include "opt_ipdn.h"
  37 #include "opt_inet.h"
  38 #ifndef INET
  39 #error IPFIREWALL requires INET.
  40 #endif /* INET */
  41 #endif
  42 #include "opt_inet6.h"
  43 #include "opt_ipsec.h"
  44
  45 #include <sys/param.h>
  46 #include <sys/systm.h>
  47 #include <sys/condvar.h>
  48 #include <sys/eventhandler.h>
  49 #include <sys/malloc.h>
  50 #include <sys/mbuf.h>
  51 #include <sys/kernel.h>
  52 #include <sys/lock.h>
  53 #include <sys/jail.h>
  54 #include <sys/module.h>
  55 #include <sys/priv.h>
  56 #include <sys/proc.h>
  57 #include <sys/rwlock.h>
  58 #include <sys/socket.h>
  59 #include <sys/socketvar.h>
  60 #include <sys/sysctl.h>
  61 #include <sys/syslog.h>
  62 #include <sys/ucred.h>
  63 #include <net/ethernet.h> /* for ETHERTYPE_IP */
  64 #include <net/if.h>
  65 #include <net/route.h>
  66 #include <net/pf_mtag.h>
  67 #include <net/vnet.h>
  68
  69 #include <netinet/in.h>
  70 #include <netinet/in_var.h>
  71 #include <netinet/in_pcb.h>
  72 #include <netinet/ip.h>
  73 #include <netinet/ip_var.h>
  74 #include <netinet/ip_icmp.h>
  75 #include <netinet/ip_fw.h>
  76 #include <netinet/ipfw/ip_fw_private.h>
  77 #include <netinet/ip_carp.h>
  78 #include <netinet/pim.h>
  79 #include <netinet/tcp_var.h>
  80 #include <netinet/udp.h>
  81 #include <netinet/udp_var.h>
  82 #include <netinet/sctp.h>
  83
  84 #include <netinet/ip6.h>
  85 #include <netinet/icmp6.h>
  86 #ifdef INET6
  87 #include <netinet6/scope6_var.h>
  88 #include <netinet6/ip6_var.h>
  89 #endif
  90
  91 #include <machine/in_cksum.h>   /* XXX for in_cksum */
  92
  93 #ifdef MAC
  94 #include <security/mac/mac_framework.h>
  95 #endif
  96
  97 /*
  98  * static variables followed by global ones.
  99  * All ipfw global variables are here.
 100  */
 101
 102 /* ipfw_vnet_ready controls when we are open for business */
 103 static VNET_DEFINE(int, ipfw_vnet_ready) = 0;
 104 #define V_ipfw_vnet_ready       VNET(ipfw_vnet_ready)
 105
 106 static VNET_DEFINE(int, fw_deny_unknown_exthdrs);
 107 #define V_fw_deny_unknown_exthdrs       VNET(fw_deny_unknown_exthdrs)
 108
 109 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
 110 static int default_to_accept = 1;
 111 #else
 112 static int default_to_accept;
 113 #endif
 114
 115 VNET_DEFINE(int, autoinc_step);
 116
 117 /*
 118  * Each rule belongs to one of 32 different sets (0..31).
 119  * The variable set_disable contains one bit per set.
 120  * If the bit is set, all rules in the corresponding set
 121  * are disabled. Set RESVD_SET(31) is reserved for the default rule
 122  * and rules that are not deleted by the flush command,
 123  * and CANNOT be disabled.
 124  * Rules in set RESVD_SET can only be deleted individually.
 125  */
 126 VNET_DEFINE(u_int32_t, set_disable);
 127 #define V_set_disable                   VNET(set_disable)
 128
 129 VNET_DEFINE(int, fw_verbose);
 130 /* counter for ipfw_log(NULL...) */
 131 VNET_DEFINE(u_int64_t, norule_counter);
 132 VNET_DEFINE(int, verbose_limit);
 133
 134 /* layer3_chain contains the list of rules for layer 3 */
 135 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
 136
 137 ipfw_nat_t *ipfw_nat_ptr = NULL;
 138 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
 139 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
 140 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
 141 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
 142 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
 143
 144 #ifdef SYSCTL_NODE
 145 uint32_t dummy_def = IPFW_DEFAULT_RULE;
 146 uint32_t dummy_tables_max = IPFW_TABLES_MAX;
 147
 148 SYSBEGIN(f3)
 149
 150 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 151 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
 152     CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
 153     "Only do a single pass through ipfw when using dummynet(4)");
 154 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
 155     CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
 156     "Rule number auto-increment step");
 157 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose,
 158     CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
 159     "Log matches to ipfw rules");
 160 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
 161     CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
 162     "Set upper limit of matches of ipfw rules logged");
 163 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
 164     &dummy_def, 0,
 165     "The default/max possible rule number.");
 166 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_RD,
 167     &dummy_tables_max, 0,
 168     "The maximum number of tables.");
 169 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
 170     &default_to_accept, 0,
 171     "Make the default rule accept all packets.");
 172 TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept);
 173 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count,
 174     CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
 175     "Number of static rules");
 176
 177 #ifdef INET6
 178 SYSCTL_DECL(_net_inet6_ip6);
 179 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 180 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
 181     CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0,
 182     "Deny packets with unknown IPv6 Extension Headers");
 183 #endif /* INET6 */
 184
 185 SYSEND
 186
 187 #endif /* SYSCTL_NODE */
 188
 189
 190 /*
 191  * Some macros used in the various matching options.
 192  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
 193  * Other macros just cast void * into the appropriate type
 194  */
 195 #define L3HDR(T, ip)    ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
 196 #define TCP(p)          ((struct tcphdr *)(p))
 197 #define SCTP(p)         ((struct sctphdr *)(p))
 198 #define UDP(p)          ((struct udphdr *)(p))
 199 #define ICMP(p)         ((struct icmphdr *)(p))
 200 #define ICMP6(p)        ((struct icmp6_hdr *)(p))
 201
 202 static __inline int
 203 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
 204 {
 205         int type = icmp->icmp_type;
 206
 207         return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
 208 }
 209
 210 #define TT      ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
 211     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
 212
 213 static int
 214 is_icmp_query(struct icmphdr *icmp)
 215 {
 216         int type = icmp->icmp_type;
 217
 218         return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
 219 }
 220 #undef TT
 221
 222 /*
 223  * The following checks use two arrays of 8 or 16 bits to store the
 224  * bits that we want set or clear, respectively. They are in the
 225  * low and high half of cmd->arg1 or cmd->d[0].
 226  *
 227  * We scan options and store the bits we find set. We succeed if
 228  *
 229  *      (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
 230  *
 231  * The code is sometimes optimized not to store additional variables.
 232  */
 233
 234 static int
 235 flags_match(ipfw_insn *cmd, u_int8_t bits)
 236 {
 237         u_char want_clear;
 238         bits = ~bits;
 239
 240         if ( ((cmd->arg1 & 0xff) & bits) != 0)
 241                 return 0; /* some bits we want set were clear */
 242         want_clear = (cmd->arg1 >> 8) & 0xff;
 243         if ( (want_clear & bits) != want_clear)
 244                 return 0; /* some bits we want clear were set */
 245         return 1;
 246 }
 247
 248 static int
 249 ipopts_match(struct ip *ip, ipfw_insn *cmd)
 250 {
 251         int optlen, bits = 0;
 252         u_char *cp = (u_char *)(ip + 1);
 253         int x = (ip->ip_hl << 2) - sizeof (struct ip);
 254
 255         for (; x > 0; x -= optlen, cp += optlen) {
 256                 int opt = cp[IPOPT_OPTVAL];
 257
 258                 if (opt == IPOPT_EOL)
 259                         break;
 260                 if (opt == IPOPT_NOP)
 261                         optlen = 1;
 262                 else {
 263                         optlen = cp[IPOPT_OLEN];
 264                         if (optlen <= 0 || optlen > x)
 265                                 return 0; /* invalid or truncated */
 266                 }
 267                 switch (opt) {
 268
 269                 default:
 270                         break;
 271
 272                 case IPOPT_LSRR:
 273                         bits |= IP_FW_IPOPT_LSRR;
 274                         break;
 275
 276                 case IPOPT_SSRR:
 277                         bits |= IP_FW_IPOPT_SSRR;
 278                         break;
 279
 280                 case IPOPT_RR:
 281                         bits |= IP_FW_IPOPT_RR;
 282                         break;
 283
 284                 case IPOPT_TS:
 285                         bits |= IP_FW_IPOPT_TS;
 286                         break;
 287                 }
 288         }
 289         return (flags_match(cmd, bits));
 290 }
 291
 292 static int
 293 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
 294 {
 295         int optlen, bits = 0;
 296         u_char *cp = (u_char *)(tcp + 1);
 297         int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
 298
 299         for (; x > 0; x -= optlen, cp += optlen) {
 300                 int opt = cp[0];
 301                 if (opt == TCPOPT_EOL)
 302                         break;
 303                 if (opt == TCPOPT_NOP)
 304                         optlen = 1;
 305                 else {
 306                         optlen = cp[1];
 307                         if (optlen <= 0)
 308                                 break;
 309                 }
 310
 311                 switch (opt) {
 312
 313                 default:
 314                         break;
 315
 316                 case TCPOPT_MAXSEG:
 317                         bits |= IP_FW_TCPOPT_MSS;
 318                         break;
 319
 320                 case TCPOPT_WINDOW:
 321                         bits |= IP_FW_TCPOPT_WINDOW;
 322                         break;
 323
 324                 case TCPOPT_SACK_PERMITTED:
 325                 case TCPOPT_SACK:
 326                         bits |= IP_FW_TCPOPT_SACK;
 327                         break;
 328
 329                 case TCPOPT_TIMESTAMP:
 330                         bits |= IP_FW_TCPOPT_TS;
 331                         break;
 332
 333                 }
 334         }
 335         return (flags_match(cmd, bits));
 336 }
 337
 338 static int
 339 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd)
 340 {
 341         if (ifp == NULL)        /* no iface with this packet, match fails */
 342                 return 0;
 343         /* Check by name or by IP address */
 344         if (cmd->name[0] != '\0') { /* match by name */
 345                 /* Check name */
 346                 if (cmd->p.glob) {
 347                         if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
 348                                 return(1);
 349                 } else {
 350                         if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
 351                                 return(1);
 352                 }
 353         } else {
 354 #ifdef  __FreeBSD__     /* and OSX too ? */
 355                 struct ifaddr *ia;
 356
 357                 if_addr_rlock(ifp);
 358                 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
 359                         if (ia->ifa_addr->sa_family != AF_INET)
 360                                 continue;
 361                         if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
 362                             (ia->ifa_addr))->sin_addr.s_addr) {
 363                                 if_addr_runlock(ifp);
 364                                 return(1);      /* match */
 365                         }
 366                 }
 367                 if_addr_runlock(ifp);
 368 #endif /* __FreeBSD__ */
 369         }
 370         return(0);      /* no match, fail ... */
 371 }
 372
 373 /*
 374  * The verify_path function checks if a route to the src exists and
 375  * if it is reachable via ifp (when provided).
 376  *
 377  * The 'verrevpath' option checks that the interface that an IP packet
 378  * arrives on is the same interface that traffic destined for the
 379  * packet's source address would be routed out of.
 380  * The 'versrcreach' option just checks that the source address is
 381  * reachable via any route (except default) in the routing table.
 382  * These two are a measure to block forged packets. This is also
 383  * commonly known as "anti-spoofing" or Unicast Reverse Path
 384  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
 385  * is purposely reminiscent of the Cisco IOS command,
 386  *
 387  *   ip verify unicast reverse-path
 388  *   ip verify unicast source reachable-via any
 389  *
 390  * which implements the same functionality. But note that the syntax
 391  * is misleading, and the check may be performed on all IP packets
 392  * whether unicast, multicast, or broadcast.
 393  */
 394 static int
 395 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
 396 {
 397 #ifndef __FreeBSD__
 398         return 0;
 399 #else
 400         struct route ro;
 401         struct sockaddr_in *dst;
 402
 403         bzero(&ro, sizeof(ro));
 404
 405         dst = (struct sockaddr_in *)&(ro.ro_dst);
 406         dst->sin_family = AF_INET;
 407         dst->sin_len = sizeof(*dst);
 408         dst->sin_addr = src;
 409         in_rtalloc_ign(&ro, 0, fib);
 410
 411         if (ro.ro_rt == NULL)
 412                 return 0;
 413
 414         /*
 415          * If ifp is provided, check for equality with rtentry.
 416          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 417          * in order to pass packets injected back by if_simloop():
 418          * if useloopback == 1 routing entry (via lo0) for our own address
 419          * may exist, so we need to handle routing assymetry.
 420          */
 421         if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
 422                 RTFREE(ro.ro_rt);
 423                 return 0;
 424         }
 425
 426         /* if no ifp provided, check if rtentry is not default route */
 427         if (ifp == NULL &&
 428              satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) {
 429                 RTFREE(ro.ro_rt);
 430                 return 0;
 431         }
 432
 433         /* or if this is a blackhole/reject route */
 434         if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
 435                 RTFREE(ro.ro_rt);
 436                 return 0;
 437         }
 438
 439         /* found valid route */
 440         RTFREE(ro.ro_rt);
 441         return 1;
 442 #endif /* __FreeBSD__ */
 443 }
 444
 445 #ifdef INET6
 446 /*
 447  * ipv6 specific rules here...
 448  */
 449 static __inline int
 450 icmp6type_match (int type, ipfw_insn_u32 *cmd)
 451 {
 452         return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
 453 }
 454
 455 static int
 456 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
 457 {
 458         int i;
 459         for (i=0; i <= cmd->o.arg1; ++i )
 460                 if (curr_flow == cmd->d[i] )
 461                         return 1;
 462         return 0;
 463 }
 464
 465 /* support for IP6_*_ME opcodes */
 466 static int
 467 search_ip6_addr_net (struct in6_addr * ip6_addr)
 468 {
 469         struct ifnet *mdc;
 470         struct ifaddr *mdc2;
 471         struct in6_ifaddr *fdm;
 472         struct in6_addr copia;
 473
 474         TAILQ_FOREACH(mdc, &V_ifnet, if_link) {
 475                 if_addr_rlock(mdc);
 476                 TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) {
 477                         if (mdc2->ifa_addr->sa_family == AF_INET6) {
 478                                 fdm = (struct in6_ifaddr *)mdc2;
 479                                 copia = fdm->ia_addr.sin6_addr;
 480                                 /* need for leaving scope_id in the sock_addr */
 481                                 in6_clearscope(&copia);
 482                                 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) {
 483                                         if_addr_runlock(mdc);
 484                                         return 1;
 485                                 }
 486                         }
 487                 }
 488                 if_addr_runlock(mdc);
 489         }
 490         return 0;
 491 }
 492
 493 static int
 494 verify_path6(struct in6_addr *src, struct ifnet *ifp)
 495 {
 496         struct route_in6 ro;
 497         struct sockaddr_in6 *dst;
 498
 499         bzero(&ro, sizeof(ro));
 500
 501         dst = (struct sockaddr_in6 * )&(ro.ro_dst);
 502         dst->sin6_family = AF_INET6;
 503         dst->sin6_len = sizeof(*dst);
 504         dst->sin6_addr = *src;
 505         /* XXX MRT 0 for ipv6 at this time */
 506         rtalloc_ign((struct route *)&ro, 0);
 507
 508         if (ro.ro_rt == NULL)
 509                 return 0;
 510
 511         /*
 512          * if ifp is provided, check for equality with rtentry
 513          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 514          * to support the case of sending packets to an address of our own.
 515          * (where the former interface is the first argument of if_simloop()
 516          *  (=ifp), the latter is lo0)
 517          */
 518         if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
 519                 RTFREE(ro.ro_rt);
 520                 return 0;
 521         }
 522
 523         /* if no ifp provided, check if rtentry is not default route */
 524         if (ifp == NULL &&
 525             IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) {
 526                 RTFREE(ro.ro_rt);
 527                 return 0;
 528         }
 529
 530         /* or if this is a blackhole/reject route */
 531         if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
 532                 RTFREE(ro.ro_rt);
 533                 return 0;
 534         }
 535
 536         /* found valid route */
 537         RTFREE(ro.ro_rt);
 538         return 1;
 539
 540 }
 541
 542 static int
 543 is_icmp6_query(int icmp6_type)
 544 {
 545         if ((icmp6_type <= ICMP6_MAXTYPE) &&
 546             (icmp6_type == ICMP6_ECHO_REQUEST ||
 547             icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
 548             icmp6_type == ICMP6_WRUREQUEST ||
 549             icmp6_type == ICMP6_FQDN_QUERY ||
 550             icmp6_type == ICMP6_NI_QUERY))
 551                 return (1);
 552
 553         return (0);
 554 }
 555
 556 static void
 557 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
 558 {
 559         struct mbuf *m;
 560
 561         m = args->m;
 562         if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
 563                 struct tcphdr *tcp;
 564                 tcp = (struct tcphdr *)((char *)ip6 + hlen);
 565
 566                 if ((tcp->th_flags & TH_RST) == 0) {
 567                         struct mbuf *m0;
 568                         m0 = ipfw_send_pkt(args->m, &(args->f_id),
 569                             ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 570                             tcp->th_flags | TH_RST);
 571                         if (m0 != NULL)
 572                                 ip6_output(m0, NULL, NULL, 0, NULL, NULL,
 573                                     NULL);
 574                 }
 575                 FREE_PKT(m);
 576         } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
 577 #if 0
 578                 /*
 579                  * Unlike above, the mbufs need to line up with the ip6 hdr,
 580                  * as the contents are read. We need to m_adj() the
 581                  * needed amount.
 582                  * The mbuf will however be thrown away so we can adjust it.
 583                  * Remember we did an m_pullup on it already so we
 584                  * can make some assumptions about contiguousness.
 585                  */
 586                 if (args->L3offset)
 587                         m_adj(m, args->L3offset);
 588 #endif
 589                 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
 590         } else
 591                 FREE_PKT(m);
 592
 593         args->m = NULL;
 594 }
 595
 596 #endif /* INET6 */
 597
 598
 599 /*
 600  * sends a reject message, consuming the mbuf passed as an argument.
 601  */
 602 static void
 603 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
 604 {
 605
 606 #if 0
 607         /* XXX When ip is not guaranteed to be at mtod() we will
 608          * need to account for this */
 609          * The mbuf will however be thrown away so we can adjust it.
 610          * Remember we did an m_pullup on it already so we
 611          * can make some assumptions about contiguousness.
 612          */
 613         if (args->L3offset)
 614                 m_adj(m, args->L3offset);
 615 #endif
 616         if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
 617                 /* We need the IP header in host order for icmp_error(). */
 618                 SET_HOST_IPLEN(ip);
 619                 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
 620         } else if (args->f_id.proto == IPPROTO_TCP) {
 621                 struct tcphdr *const tcp =
 622                     L3HDR(struct tcphdr, mtod(args->m, struct ip *));
 623                 if ( (tcp->th_flags & TH_RST) == 0) {
 624                         struct mbuf *m;
 625                         m = ipfw_send_pkt(args->m, &(args->f_id),
 626                                 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 627                                 tcp->th_flags | TH_RST);
 628                         if (m != NULL)
 629                                 ip_output(m, NULL, NULL, 0, NULL, NULL);
 630                 }
 631                 FREE_PKT(args->m);
 632         } else
 633                 FREE_PKT(args->m);
 634         args->m = NULL;
 635 }
 636
 637 /*
 638  * Support for uid/gid/jail lookup. These tests are expensive
 639  * (because we may need to look into the list of active sockets)
 640  * so we cache the results. ugid_lookupp is 0 if we have not
 641  * yet done a lookup, 1 if we succeeded, and -1 if we tried
 642  * and failed. The function always returns the match value.
 643  * We could actually spare the variable and use *uc, setting
 644  * it to '(void *)check_uidgid if we have no info, NULL if
 645  * we tried and failed, or any other value if successful.
 646  */
 647 static int
 648 check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif,
 649     struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip,
 650     u_int16_t src_port, int *ugid_lookupp,
 651     struct ucred **uc, struct inpcb *inp)
 652 {
 653 #ifndef __FreeBSD__
 654         return cred_check(insn, proto, oif,
 655             dst_ip, dst_port, src_ip, src_port,
 656             (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
 657 #else  /* FreeBSD */
 658         struct inpcbinfo *pi;
 659         int wildcard;
 660         struct inpcb *pcb;
 661         int match;
 662
 663         /*
 664          * Check to see if the UDP or TCP stack supplied us with
 665          * the PCB. If so, rather then holding a lock and looking
 666          * up the PCB, we can use the one that was supplied.
 667          */
 668         if (inp && *ugid_lookupp == 0) {
 669                 INP_LOCK_ASSERT(inp);
 670                 if (inp->inp_socket != NULL) {
 671                         *uc = crhold(inp->inp_cred);
 672                         *ugid_lookupp = 1;
 673                 } else
 674                         *ugid_lookupp = -1;
 675         }
 676         /*
 677          * If we have already been here and the packet has no
 678          * PCB entry associated with it, then we can safely
 679          * assume that this is a no match.
 680          */
 681         if (*ugid_lookupp == -1)
 682                 return (0);
 683         if (proto == IPPROTO_TCP) {
 684                 wildcard = 0;
 685                 pi = &V_tcbinfo;
 686         } else if (proto == IPPROTO_UDP) {
 687                 wildcard = INPLOOKUP_WILDCARD;
 688                 pi = &V_udbinfo;
 689         } else
 690                 return 0;
 691         match = 0;
 692         if (*ugid_lookupp == 0) {
 693                 INP_INFO_RLOCK(pi);
 694                 pcb =  (oif) ?
 695                         in_pcblookup_hash(pi,
 696                                 dst_ip, htons(dst_port),
 697                                 src_ip, htons(src_port),
 698                                 wildcard, oif) :
 699                         in_pcblookup_hash(pi,
 700                                 src_ip, htons(src_port),
 701                                 dst_ip, htons(dst_port),
 702                                 wildcard, NULL);
 703                 if (pcb != NULL) {
 704                         *uc = crhold(pcb->inp_cred);
 705                         *ugid_lookupp = 1;
 706                 }
 707                 INP_INFO_RUNLOCK(pi);
 708                 if (*ugid_lookupp == 0) {
 709                         /*
 710                          * We tried and failed, set the variable to -1
 711                          * so we will not try again on this packet.
 712                          */
 713                         *ugid_lookupp = -1;
 714                         return (0);
 715                 }
 716         }
 717         if (insn->o.opcode == O_UID)
 718                 match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
 719         else if (insn->o.opcode == O_GID)
 720                 match = groupmember((gid_t)insn->d[0], *uc);
 721         else if (insn->o.opcode == O_JAIL)
 722                 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
 723         return match;
 724 #endif /* __FreeBSD__ */
 725 }
 726
 727 /*
 728  * Helper function to set args with info on the rule after the matching
 729  * one. slot is precise, whereas we guess rule_id as they are
 730  * assigned sequentially.
 731  */
 732 static inline void
 733 set_match(struct ip_fw_args *args, int slot,
 734         struct ip_fw_chain *chain)
 735 {
 736         args->rule.chain_id = chain->id;
 737         args->rule.slot = slot + 1; /* we use 0 as a marker */
 738         args->rule.rule_id = 1 + chain->map[slot]->id;
 739         args->rule.rulenum = chain->map[slot]->rulenum;
 740 }
 741
 742 /*
 743  * The main check routine for the firewall.
 744  *
 745  * All arguments are in args so we can modify them and return them
 746  * back to the caller.
 747  *
 748  * Parameters:
 749  *
 750  *      args->m (in/out) The packet; we set to NULL when/if we nuke it.
 751  *              Starts with the IP header.
 752  *      args->eh (in)   Mac header if present, NULL for layer3 packet.
 753  *      args->L3offset  Number of bytes bypassed if we came from L2.
 754  *                      e.g. often sizeof(eh)  ** NOTYET **
 755  *      args->oif       Outgoing interface, NULL if packet is incoming.
 756  *              The incoming interface is in the mbuf. (in)
 757  *      args->divert_rule (in/out)
 758  *              Skip up to the first rule past this rule number;
 759  *              upon return, non-zero port number for divert or tee.
 760  *
 761  *      args->rule      Pointer to the last matching rule (in/out)
 762  *      args->next_hop  Socket we are forwarding to (out).
 763  *      args->f_id      Addresses grabbed from the packet (out)
 764  *      args->rule.info a cookie depending on rule action
 765  *
 766  * Return value:
 767  *
 768  *      IP_FW_PASS      the packet must be accepted
 769  *      IP_FW_DENY      the packet must be dropped
 770  *      IP_FW_DIVERT    divert packet, port in m_tag
 771  *      IP_FW_TEE       tee packet, port in m_tag
 772  *      IP_FW_DUMMYNET  to dummynet, pipe in args->cookie
 773  *      IP_FW_NETGRAPH  into netgraph, cookie args->cookie
 774  *              args->rule contains the matching rule,
 775  *              args->rule.info has additional information.
 776  *
 777  */
 778 int
 779 ipfw_chk(struct ip_fw_args *args)
 780 {
 781
 782         /*
 783          * Local variables holding state while processing a packet:
 784          *
 785          * IMPORTANT NOTE: to speed up the processing of rules, there
 786          * are some assumption on the values of the variables, which
 787          * are documented here. Should you change them, please check
 788          * the implementation of the various instructions to make sure
 789          * that they still work.
 790          *
 791          * args->eh     The MAC header. It is non-null for a layer2
 792          *      packet, it is NULL for a layer-3 packet.
 793          * **notyet**
 794          * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
 795          *
 796          * m | args->m  Pointer to the mbuf, as received from the caller.
 797          *      It may change if ipfw_chk() does an m_pullup, or if it
 798          *      consumes the packet because it calls send_reject().
 799          *      XXX This has to change, so that ipfw_chk() never modifies
 800          *      or consumes the buffer.
 801          * ip   is the beginning of the ip(4 or 6) header.
 802          *      Calculated by adding the L3offset to the start of data.
 803          *      (Until we start using L3offset, the packet is
 804          *      supposed to start with the ip header).
 805          */
 806         struct mbuf *m = args->m;
 807         struct ip *ip = mtod(m, struct ip *);
 808
 809         /*
 810          * For rules which contain uid/gid or jail constraints, cache
 811          * a copy of the users credentials after the pcb lookup has been
 812          * executed. This will speed up the processing of rules with
 813          * these types of constraints, as well as decrease contention
 814          * on pcb related locks.
 815          */
 816 #ifndef __FreeBSD__
 817         struct bsd_ucred ucred_cache;
 818 #else
 819         struct ucred *ucred_cache = NULL;
 820 #endif
 821         int ucred_lookup = 0;
 822
 823         /*
 824          * oif | args->oif      If NULL, ipfw_chk has been called on the
 825          *      inbound path (ether_input, ip_input).
 826          *      If non-NULL, ipfw_chk has been called on the outbound path
 827          *      (ether_output, ip_output).
 828          */
 829         struct ifnet *oif = args->oif;
 830
 831         int f_pos = 0;          /* index of current rule in the array */
 832         int retval = 0;
 833
 834         /*
 835          * hlen The length of the IP header.
 836          */
 837         u_int hlen = 0;         /* hlen >0 means we have an IP pkt */
 838
 839         /*
 840          * offset       The offset of a fragment. offset != 0 means that
 841          *      we have a fragment at this offset of an IPv4 packet.
 842          *      offset == 0 means that (if this is an IPv4 packet)
 843          *      this is the first or only fragment.
 844          *      For IPv6 offset == 0 means there is no Fragment Header.
 845          *      If offset != 0 for IPv6 always use correct mask to
 846          *      get the correct offset because we add IP6F_MORE_FRAG
 847          *      to be able to dectect the first fragment which would
 848          *      otherwise have offset = 0.
 849          */
 850         u_short offset = 0;
 851
 852         /*
 853          * Local copies of addresses. They are only valid if we have
 854          * an IP packet.
 855          *
 856          * proto        The protocol. Set to 0 for non-ip packets,
 857          *      or to the protocol read from the packet otherwise.
 858          *      proto != 0 means that we have an IPv4 packet.
 859          *
 860          * src_port, dst_port   port numbers, in HOST format. Only
 861          *      valid for TCP and UDP packets.
 862          *
 863          * src_ip, dst_ip       ip addresses, in NETWORK format.
 864          *      Only valid for IPv4 packets.
 865          */
 866         uint8_t proto;
 867         uint16_t src_port = 0, dst_port = 0;    /* NOTE: host format    */
 868         struct in_addr src_ip, dst_ip;          /* NOTE: network format */
 869         uint16_t iplen=0;
 870         int pktlen;
 871         uint16_t        etype = 0;      /* Host order stored ether type */
 872
 873         /*
 874          * dyn_dir = MATCH_UNKNOWN when rules unchecked,
 875          *      MATCH_NONE when checked and not matched (q = NULL),
 876          *      MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
 877          */
 878         int dyn_dir = MATCH_UNKNOWN;
 879         ipfw_dyn_rule *q = NULL;
 880         struct ip_fw_chain *chain = &V_layer3_chain;
 881
 882         /*
 883          * We store in ulp a pointer to the upper layer protocol header.
 884          * In the ipv4 case this is easy to determine from the header,
 885          * but for ipv6 we might have some additional headers in the middle.
 886          * ulp is NULL if not found.
 887          */
 888         void *ulp = NULL;               /* upper layer protocol pointer. */
 889
 890         /* XXX ipv6 variables */
 891         int is_ipv6 = 0;
 892         uint8_t icmp6_type = 0;
 893         uint16_t ext_hd = 0;    /* bits vector for extension header filtering */
 894         /* end of ipv6 variables */
 895
 896         int is_ipv4 = 0;
 897
 898         int done = 0;           /* flag to exit the outer loop */
 899
 900         if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
 901                 return (IP_FW_PASS);    /* accept */
 902
 903         dst_ip.s_addr = 0;              /* make sure it is initialized */
 904         src_ip.s_addr = 0;              /* make sure it is initialized */
 905         pktlen = m->m_pkthdr.len;
 906         args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
 907         proto = args->f_id.proto = 0;   /* mark f_id invalid */
 908                 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
 909
 910 /*
 911  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
 912  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
 913  * pointer might become stale after other pullups (but we never use it
 914  * this way).
 915  */
 916 #define PULLUP_TO(_len, p, T)                                   \
 917 do {                                                            \
 918         int x = (_len) + sizeof(T);                             \
 919         if ((m)->m_len < x) {                                   \
 920                 args->m = m = m_pullup(m, x);                   \
 921                 if (m == NULL)                                  \
 922                         goto pullup_failed;                     \
 923         }                                                       \
 924         p = (mtod(m, char *) + (_len));                         \
 925 } while (0)
 926
 927         /*
 928          * if we have an ether header,
 929          */
 930         if (args->eh)
 931                 etype = ntohs(args->eh->ether_type);
 932
 933         /* Identify IP packets and fill up variables. */
 934         if (pktlen >= sizeof(struct ip6_hdr) &&
 935             (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
 936                 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
 937                 is_ipv6 = 1;
 938                 args->f_id.addr_type = 6;
 939                 hlen = sizeof(struct ip6_hdr);
 940                 proto = ip6->ip6_nxt;
 941
 942                 /* Search extension headers to find upper layer protocols */
 943                 while (ulp == NULL) {
 944                         switch (proto) {
 945                         case IPPROTO_ICMPV6:
 946                                 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
 947                                 icmp6_type = ICMP6(ulp)->icmp6_type;
 948                                 break;
 949
 950                         case IPPROTO_TCP:
 951                                 PULLUP_TO(hlen, ulp, struct tcphdr);
 952                                 dst_port = TCP(ulp)->th_dport;
 953                                 src_port = TCP(ulp)->th_sport;
 954                                 /* save flags for dynamic rules */
 955                                 args->f_id._flags = TCP(ulp)->th_flags;
 956                                 break;
 957
 958                         case IPPROTO_SCTP:
 959                                 PULLUP_TO(hlen, ulp, struct sctphdr);
 960                                 src_port = SCTP(ulp)->src_port;
 961                                 dst_port = SCTP(ulp)->dest_port;
 962                                 break;
 963
 964                         case IPPROTO_UDP:
 965                                 PULLUP_TO(hlen, ulp, struct udphdr);
 966                                 dst_port = UDP(ulp)->uh_dport;
 967                                 src_port = UDP(ulp)->uh_sport;
 968                                 break;
 969
 970                         case IPPROTO_HOPOPTS:   /* RFC 2460 */
 971                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
 972                                 ext_hd |= EXT_HOPOPTS;
 973                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
 974                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
 975                                 ulp = NULL;
 976                                 break;
 977
 978                         case IPPROTO_ROUTING:   /* RFC 2460 */
 979                                 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
 980                                 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
 981                                 case 0:
 982                                         ext_hd |= EXT_RTHDR0;
 983                                         break;
 984                                 case 2:
 985                                         ext_hd |= EXT_RTHDR2;
 986                                         break;
 987                                 default:
 988                                         printf("IPFW2: IPV6 - Unknown Routing "
 989                                             "Header type(%d)\n",
 990                                             ((struct ip6_rthdr *)ulp)->ip6r_type);
 991                                         if (V_fw_deny_unknown_exthdrs)
 992                                             return (IP_FW_DENY);
 993                                         break;
 994                                 }
 995                                 ext_hd |= EXT_ROUTING;
 996                                 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
 997                                 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
 998                                 ulp = NULL;
 999                                 break;
1000
1001                         case IPPROTO_FRAGMENT:  /* RFC 2460 */
1002                                 PULLUP_TO(hlen, ulp, struct ip6_frag);
1003                                 ext_hd |= EXT_FRAGMENT;
1004                                 hlen += sizeof (struct ip6_frag);
1005                                 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1006                                 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1007                                         IP6F_OFF_MASK;
1008                                 /* Add IP6F_MORE_FRAG for offset of first
1009                                  * fragment to be != 0. */
1010                                 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg &
1011                                         IP6F_MORE_FRAG;
1012                                 if (offset == 0) {
1013                                         printf("IPFW2: IPV6 - Invalid Fragment "
1014                                             "Header\n");
1015                                         if (V_fw_deny_unknown_exthdrs)
1016                                             return (IP_FW_DENY);
1017                                         break;
1018                                 }
1019                                 args->f_id.extra =
1020                                     ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1021                                 ulp = NULL;
1022                                 break;
1023
1024                         case IPPROTO_DSTOPTS:   /* RFC 2460 */
1025                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1026                                 ext_hd |= EXT_DSTOPTS;
1027                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1028                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1029                                 ulp = NULL;
1030                                 break;
1031
1032                         case IPPROTO_AH:        /* RFC 2402 */
1033                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1034                                 ext_hd |= EXT_AH;
1035                                 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1036                                 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1037                                 ulp = NULL;
1038                                 break;
1039
1040                         case IPPROTO_ESP:       /* RFC 2406 */
1041                                 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1042                                 /* Anything past Seq# is variable length and
1043                                  * data past this ext. header is encrypted. */
1044                                 ext_hd |= EXT_ESP;
1045                                 break;
1046
1047                         case IPPROTO_NONE:      /* RFC 2460 */
1048                                 /*
1049                                  * Packet ends here, and IPv6 header has
1050                                  * already been pulled up. If ip6e_len!=0
1051                                  * then octets must be ignored.
1052                                  */
1053                                 ulp = ip; /* non-NULL to get out of loop. */
1054                                 break;
1055
1056                         case IPPROTO_OSPFIGP:
1057                                 /* XXX OSPF header check? */
1058                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1059                                 break;
1060
1061                         case IPPROTO_PIM:
1062                                 /* XXX PIM header check? */
1063                                 PULLUP_TO(hlen, ulp, struct pim);
1064                                 break;
1065
1066                         case IPPROTO_CARP:
1067                                 PULLUP_TO(hlen, ulp, struct carp_header);
1068                                 if (((struct carp_header *)ulp)->carp_version !=
1069                                     CARP_VERSION)
1070                                         return (IP_FW_DENY);
1071                                 if (((struct carp_header *)ulp)->carp_type !=
1072                                     CARP_ADVERTISEMENT)
1073                                         return (IP_FW_DENY);
1074                                 break;
1075
1076                         case IPPROTO_IPV6:      /* RFC 2893 */
1077                                 PULLUP_TO(hlen, ulp, struct ip6_hdr);
1078                                 break;
1079
1080                         case IPPROTO_IPV4:      /* RFC 2893 */
1081                                 PULLUP_TO(hlen, ulp, struct ip);
1082                                 break;
1083
1084                         default:
1085                                 printf("IPFW2: IPV6 - Unknown Extension "
1086                                     "Header(%d), ext_hd=%x\n", proto, ext_hd);
1087                                 if (V_fw_deny_unknown_exthdrs)
1088                                     return (IP_FW_DENY);
1089                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1090                                 break;
1091                         } /*switch */
1092                 }
1093                 ip = mtod(m, struct ip *);
1094                 ip6 = (struct ip6_hdr *)ip;
1095                 args->f_id.src_ip6 = ip6->ip6_src;
1096                 args->f_id.dst_ip6 = ip6->ip6_dst;
1097                 args->f_id.src_ip = 0;
1098                 args->f_id.dst_ip = 0;
1099                 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1100         } else if (pktlen >= sizeof(struct ip) &&
1101             (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1102                 is_ipv4 = 1;
1103                 hlen = ip->ip_hl << 2;
1104                 args->f_id.addr_type = 4;
1105
1106                 /*
1107                  * Collect parameters into local variables for faster matching.
1108                  */
1109                 proto = ip->ip_p;
1110                 src_ip = ip->ip_src;
1111                 dst_ip = ip->ip_dst;
1112                 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1113                 iplen = ntohs(ip->ip_len);
1114                 pktlen = iplen < pktlen ? iplen : pktlen;
1115
1116                 if (offset == 0) {
1117                         switch (proto) {
1118                         case IPPROTO_TCP:
1119                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1120                                 dst_port = TCP(ulp)->th_dport;
1121                                 src_port = TCP(ulp)->th_sport;
1122                                 /* save flags for dynamic rules */
1123                                 args->f_id._flags = TCP(ulp)->th_flags;
1124                                 break;
1125
1126                         case IPPROTO_SCTP:
1127                                 PULLUP_TO(hlen, ulp, struct sctphdr);
1128                                 src_port = SCTP(ulp)->src_port;
1129                                 dst_port = SCTP(ulp)->dest_port;
1130                                 break;
1131
1132                         case IPPROTO_UDP:
1133                                 PULLUP_TO(hlen, ulp, struct udphdr);
1134                                 dst_port = UDP(ulp)->uh_dport;
1135                                 src_port = UDP(ulp)->uh_sport;
1136                                 break;
1137
1138                         case IPPROTO_ICMP:
1139                                 PULLUP_TO(hlen, ulp, struct icmphdr);
1140                                 //args->f_id.flags = ICMP(ulp)->icmp_type;
1141                                 break;
1142
1143                         default:
1144                                 break;
1145                         }
1146                 }
1147
1148                 ip = mtod(m, struct ip *);
1149                 args->f_id.src_ip = ntohl(src_ip.s_addr);
1150                 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1151         }
1152 #undef PULLUP_TO
1153         if (proto) { /* we may have port numbers, store them */
1154                 args->f_id.proto = proto;
1155                 args->f_id.src_port = src_port = ntohs(src_port);
1156                 args->f_id.dst_port = dst_port = ntohs(dst_port);
1157         }
1158
1159         IPFW_RLOCK(chain);
1160         if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1161                 IPFW_RUNLOCK(chain);
1162                 return (IP_FW_PASS);    /* accept */
1163         }
1164         if (args->rule.slot) {
1165                 /*
1166                  * Packet has already been tagged as a result of a previous
1167                  * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1168                  * REASS, NETGRAPH, DIVERT/TEE...)
1169                  * Validate the slot and continue from the next one
1170                  * if still present, otherwise do a lookup.
1171                  */
1172                 f_pos = (args->rule.chain_id == chain->id) ?
1173                     args->rule.slot :
1174                     ipfw_find_rule(chain, args->rule.rulenum,
1175                         args->rule.rule_id);
1176         } else {
1177                 f_pos = 0;
1178         }
1179
1180         /*
1181          * Now scan the rules, and parse microinstructions for each rule.
1182          * We have two nested loops and an inner switch. Sometimes we
1183          * need to break out of one or both loops, or re-enter one of
1184          * the loops with updated variables. Loop variables are:
1185          *
1186          *      f_pos (outer loop) points to the current rule.
1187          *              On output it points to the matching rule.
1188          *      done (outer loop) is used as a flag to break the loop.
1189          *      l (inner loop)  residual length of current rule.
1190          *              cmd points to the current microinstruction.
1191          *
1192          * We break the inner loop by setting l=0 and possibly
1193          * cmdlen=0 if we don't want to advance cmd.
1194          * We break the outer loop by setting done=1
1195          * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1196          * as needed.
1197          */
1198         for (; f_pos < chain->n_rules; f_pos++) {
1199                 ipfw_insn *cmd;
1200                 uint32_t tablearg = 0;
1201                 int l, cmdlen, skip_or; /* skip rest of OR block */
1202                 struct ip_fw *f;
1203
1204                 f = chain->map[f_pos];
1205                 if (V_set_disable & (1 << f->set) )
1206                         continue;
1207
1208                 skip_or = 0;
1209                 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1210                     l -= cmdlen, cmd += cmdlen) {
1211                         int match;
1212
1213                         /*
1214                          * check_body is a jump target used when we find a
1215                          * CHECK_STATE, and need to jump to the body of
1216                          * the target rule.
1217                          */
1218
1219 /* check_body: */
1220                         cmdlen = F_LEN(cmd);
1221                         /*
1222                          * An OR block (insn_1 || .. || insn_n) has the
1223                          * F_OR bit set in all but the last instruction.
1224                          * The first match will set "skip_or", and cause
1225                          * the following instructions to be skipped until
1226                          * past the one with the F_OR bit clear.
1227                          */
1228                         if (skip_or) {          /* skip this instruction */
1229                                 if ((cmd->len & F_OR) == 0)
1230                                         skip_or = 0;    /* next one is good */
1231                                 continue;
1232                         }
1233                         match = 0; /* set to 1 if we succeed */
1234
1235                         switch (cmd->opcode) {
1236                         /*
1237                          * The first set of opcodes compares the packet's
1238                          * fields with some pattern, setting 'match' if a
1239                          * match is found. At the end of the loop there is
1240                          * logic to deal with F_NOT and F_OR flags associated
1241                          * with the opcode.
1242                          */
1243                         case O_NOP:
1244                                 match = 1;
1245                                 break;
1246
1247                         case O_FORWARD_MAC:
1248                                 printf("ipfw: opcode %d unimplemented\n",
1249                                     cmd->opcode);
1250                                 break;
1251
1252                         case O_GID:
1253                         case O_UID:
1254                         case O_JAIL:
1255                                 /*
1256                                  * We only check offset == 0 && proto != 0,
1257                                  * as this ensures that we have a
1258                                  * packet with the ports info.
1259                                  */
1260                                 if (offset!=0)
1261                                         break;
1262                                 if (is_ipv6) /* XXX to be fixed later */
1263                                         break;
1264                                 if (proto == IPPROTO_TCP ||
1265                                     proto == IPPROTO_UDP)
1266                                         match = check_uidgid(
1267                                                     (ipfw_insn_u32 *)cmd,
1268                                                     proto, oif,
1269                                                     dst_ip, dst_port,
1270                                                     src_ip, src_port, &ucred_lookup,
1271 #ifdef __FreeBSD__
1272                                                     &ucred_cache, args->inp);
1273 #else
1274                                                     (void *)&ucred_cache,
1275                                                     (struct inpcb *)args->m);
1276 #endif
1277                                 break;
1278
1279                         case O_RECV:
1280                                 match = iface_match(m->m_pkthdr.rcvif,
1281                                     (ipfw_insn_if *)cmd);
1282                                 break;
1283
1284                         case O_XMIT:
1285                                 match = iface_match(oif, (ipfw_insn_if *)cmd);
1286                                 break;
1287
1288                         case O_VIA:
1289                                 match = iface_match(oif ? oif :
1290                                     m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd);
1291                                 break;
1292
1293                         case O_MACADDR2:
1294                                 if (args->eh != NULL) { /* have MAC header */
1295                                         u_int32_t *want = (u_int32_t *)
1296                                                 ((ipfw_insn_mac *)cmd)->addr;
1297                                         u_int32_t *mask = (u_int32_t *)
1298                                                 ((ipfw_insn_mac *)cmd)->mask;
1299                                         u_int32_t *hdr = (u_int32_t *)args->eh;
1300
1301                                         match =
1302                                             ( want[0] == (hdr[0] & mask[0]) &&
1303                                               want[1] == (hdr[1] & mask[1]) &&
1304                                               want[2] == (hdr[2] & mask[2]) );
1305                                 }
1306                                 break;
1307
1308                         case O_MAC_TYPE:
1309                                 if (args->eh != NULL) {
1310                                         u_int16_t *p =
1311                                             ((ipfw_insn_u16 *)cmd)->ports;
1312                                         int i;
1313
1314                                         for (i = cmdlen - 1; !match && i>0;
1315                                             i--, p += 2)
1316                                                 match = (etype >= p[0] &&
1317                                                     etype <= p[1]);
1318                                 }
1319                                 break;
1320
1321                         case O_FRAG:
1322                                 match = (offset != 0);
1323                                 break;
1324
1325                         case O_IN:      /* "out" is "not in" */
1326                                 match = (oif == NULL);
1327                                 break;
1328
1329                         case O_LAYER2:
1330                                 match = (args->eh != NULL);
1331                                 break;
1332
1333                         case O_DIVERTED:
1334                             {
1335                                 /* For diverted packets, args->rule.info
1336                                  * contains the divert port (in host format)
1337                                  * reason and direction.
1338                                  */
1339                                 uint32_t i = args->rule.info;
1340                                 match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1341                                     cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1342                             }
1343                                 break;
1344
1345                         case O_PROTO:
1346                                 /*
1347                                  * We do not allow an arg of 0 so the
1348                                  * check of "proto" only suffices.
1349                                  */
1350                                 match = (proto == cmd->arg1);
1351                                 break;
1352
1353                         case O_IP_SRC:
1354                                 match = is_ipv4 &&
1355                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1356                                     src_ip.s_addr);
1357                                 break;
1358
1359                         case O_IP_SRC_LOOKUP:
1360                         case O_IP_DST_LOOKUP:
1361                                 if (is_ipv4) {
1362                                     uint32_t key =
1363                                         (cmd->opcode == O_IP_DST_LOOKUP) ?
1364                                             dst_ip.s_addr : src_ip.s_addr;
1365                                     uint32_t v = 0;
1366
1367                                     if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1368                                         /* generic lookup. The key must be
1369                                          * in 32bit big-endian format.
1370                                          */
1371                                         v = ((ipfw_insn_u32 *)cmd)->d[1];
1372                                         if (v == 0)
1373                                             key = dst_ip.s_addr;
1374                                         else if (v == 1)
1375                                             key = src_ip.s_addr;
1376                                         else if (v == 6) /* dscp */
1377                                             key = (ip->ip_tos >> 2) & 0x3f;
1378                                         else if (offset != 0)
1379                                             break;
1380                                         else if (proto != IPPROTO_TCP &&
1381                                                 proto != IPPROTO_UDP)
1382                                             break;
1383                                         else if (v == 2)
1384                                             key = htonl(dst_port);
1385                                         else if (v == 3)
1386                                             key = htonl(src_port);
1387                                         else if (v == 4 || v == 5) {
1388                                             check_uidgid(
1389                                                 (ipfw_insn_u32 *)cmd,
1390                                                 proto, oif,
1391                                                 dst_ip, dst_port,
1392                                                 src_ip, src_port, &ucred_lookup,
1393 #ifdef __FreeBSD__
1394                                                 &ucred_cache, args->inp);
1395                                             if (v == 4 /* O_UID */)
1396                                                 key = ucred_cache->cr_uid;
1397                                             else if (v == 5 /* O_JAIL */)
1398                                                 key = ucred_cache->cr_prison->pr_id;
1399 #else /* !__FreeBSD__ */
1400                                                 (void *)&ucred_cache,
1401                                                 (struct inpcb *)args->m);
1402                                             if (v ==4 /* O_UID */)
1403                                                 key = ucred_cache.uid;
1404                                             else if (v == 5 /* O_JAIL */)
1405                                                 key = ucred_cache.xid;
1406 #endif /* !__FreeBSD__ */
1407                                             key = htonl(key);
1408                                         } else
1409                                             break;
1410                                     }
1411                                     match = ipfw_lookup_table(chain,
1412                                         cmd->arg1, key, &v);
1413                                     if (!match)
1414                                         break;
1415                                     if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1416                                         match =
1417                                             ((ipfw_insn_u32 *)cmd)->d[0] == v;
1418                                     else
1419                                         tablearg = v;
1420                                 }
1421                                 break;
1422
1423                         case O_IP_SRC_MASK:
1424                         case O_IP_DST_MASK:
1425                                 if (is_ipv4) {
1426                                     uint32_t a =
1427                                         (cmd->opcode == O_IP_DST_MASK) ?
1428                                             dst_ip.s_addr : src_ip.s_addr;
1429                                     uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
1430                                     int i = cmdlen-1;
1431
1432                                     for (; !match && i>0; i-= 2, p+= 2)
1433                                         match = (p[0] == (a & p[1]));
1434                                 }
1435                                 break;
1436
1437                         case O_IP_SRC_ME:
1438                                 if (is_ipv4) {
1439                                         struct ifnet *tif;
1440
1441                                         INADDR_TO_IFP(src_ip, tif);
1442                                         match = (tif != NULL);
1443                                         break;
1444                                 }
1445 #ifdef INET6
1446                                 /* FALLTHROUGH */
1447                         case O_IP6_SRC_ME:
1448                                 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6);
1449 #endif
1450                                 break;
1451
1452                         case O_IP_DST_SET:
1453                         case O_IP_SRC_SET:
1454                                 if (is_ipv4) {
1455                                         u_int32_t *d = (u_int32_t *)(cmd+1);
1456                                         u_int32_t addr =
1457                                             cmd->opcode == O_IP_DST_SET ?
1458                                                 args->f_id.dst_ip :
1459                                                 args->f_id.src_ip;
1460
1461                                             if (addr < d[0])
1462                                                     break;
1463                                             addr -= d[0]; /* subtract base */
1464                                             match = (addr < cmd->arg1) &&
1465                                                 ( d[ 1 + (addr>>5)] &
1466                                                   (1<<(addr & 0x1f)) );
1467                                 }
1468                                 break;
1469
1470                         case O_IP_DST:
1471                                 match = is_ipv4 &&
1472                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1473                                     dst_ip.s_addr);
1474                                 break;
1475
1476                         case O_IP_DST_ME:
1477                                 if (is_ipv4) {
1478                                         struct ifnet *tif;
1479
1480                                         INADDR_TO_IFP(dst_ip, tif);
1481                                         match = (tif != NULL);
1482                                         break;
1483                                 }
1484 #ifdef INET6
1485                                 /* FALLTHROUGH */
1486                         case O_IP6_DST_ME:
1487                                 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6);
1488 #endif
1489                                 break;
1490
1491
1492                         case O_IP_SRCPORT:
1493                         case O_IP_DSTPORT:
1494                                 /*
1495                                  * offset == 0 && proto != 0 is enough
1496                                  * to guarantee that we have a
1497                                  * packet with port info.
1498                                  */
1499                                 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1500                                     && offset == 0) {
1501                                         u_int16_t x =
1502                                             (cmd->opcode == O_IP_SRCPORT) ?
1503                                                 src_port : dst_port ;
1504                                         u_int16_t *p =
1505                                             ((ipfw_insn_u16 *)cmd)->ports;
1506                                         int i;
1507
1508                                         for (i = cmdlen - 1; !match && i>0;
1509                                             i--, p += 2)
1510                                                 match = (x>=p[0] && x<=p[1]);
1511                                 }
1512                                 break;
1513
1514                         case O_ICMPTYPE:
1515                                 match = (offset == 0 && proto==IPPROTO_ICMP &&
1516                                     icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
1517                                 break;
1518
1519 #ifdef INET6
1520                         case O_ICMP6TYPE:
1521                                 match = is_ipv6 && offset == 0 &&
1522                                     proto==IPPROTO_ICMPV6 &&
1523                                     icmp6type_match(
1524                                         ICMP6(ulp)->icmp6_type,
1525                                         (ipfw_insn_u32 *)cmd);
1526                                 break;
1527 #endif /* INET6 */
1528
1529                         case O_IPOPT:
1530                                 match = (is_ipv4 &&
1531                                     ipopts_match(ip, cmd) );
1532                                 break;
1533
1534                         case O_IPVER:
1535                                 match = (is_ipv4 &&
1536                                     cmd->arg1 == ip->ip_v);
1537                                 break;
1538
1539                         case O_IPID:
1540                         case O_IPLEN:
1541                         case O_IPTTL:
1542                                 if (is_ipv4) {  /* only for IP packets */
1543                                     uint16_t x;
1544                                     uint16_t *p;
1545                                     int i;
1546
1547                                     if (cmd->opcode == O_IPLEN)
1548                                         x = iplen;
1549                                     else if (cmd->opcode == O_IPTTL)
1550                                         x = ip->ip_ttl;
1551                                     else /* must be IPID */
1552                                         x = ntohs(ip->ip_id);
1553                                     if (cmdlen == 1) {
1554                                         match = (cmd->arg1 == x);
1555                                         break;
1556                                     }
1557                                     /* otherwise we have ranges */
1558                                     p = ((ipfw_insn_u16 *)cmd)->ports;
1559                                     i = cmdlen - 1;
1560                                     for (; !match && i>0; i--, p += 2)
1561                                         match = (x >= p[0] && x <= p[1]);
1562                                 }
1563                                 break;
1564
1565                         case O_IPPRECEDENCE:
1566                                 match = (is_ipv4 &&
1567                                     (cmd->arg1 == (ip->ip_tos & 0xe0)) );
1568                                 break;
1569
1570                         case O_IPTOS:
1571                                 match = (is_ipv4 &&
1572                                     flags_match(cmd, ip->ip_tos));
1573                                 break;
1574
1575                         case O_TCPDATALEN:
1576                                 if (proto == IPPROTO_TCP && offset == 0) {
1577                                     struct tcphdr *tcp;
1578                                     uint16_t x;
1579                                     uint16_t *p;
1580                                     int i;
1581
1582                                     tcp = TCP(ulp);
1583                                     x = iplen -
1584                                         ((ip->ip_hl + tcp->th_off) << 2);
1585                                     if (cmdlen == 1) {
1586                                         match = (cmd->arg1 == x);
1587                                         break;
1588                                     }
1589                                     /* otherwise we have ranges */
1590                                     p = ((ipfw_insn_u16 *)cmd)->ports;
1591                                     i = cmdlen - 1;
1592                                     for (; !match && i>0; i--, p += 2)
1593                                         match = (x >= p[0] && x <= p[1]);
1594                                 }
1595                                 break;
1596
1597                         case O_TCPFLAGS:
1598                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1599                                     flags_match(cmd, TCP(ulp)->th_flags));
1600                                 break;
1601
1602                         case O_TCPOPTS:
1603                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1604                                     tcpopts_match(TCP(ulp), cmd));
1605                                 break;
1606
1607                         case O_TCPSEQ:
1608                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1609                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
1610                                         TCP(ulp)->th_seq);
1611                                 break;
1612
1613                         case O_TCPACK:
1614                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1615                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
1616                                         TCP(ulp)->th_ack);
1617                                 break;
1618
1619                         case O_TCPWIN:
1620                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1621                                     cmd->arg1 == TCP(ulp)->th_win);
1622                                 break;
1623
1624                         case O_ESTAB:
1625                                 /* reject packets which have SYN only */
1626                                 /* XXX should i also check for TH_ACK ? */
1627                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1628                                     (TCP(ulp)->th_flags &
1629                                      (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1630                                 break;
1631
1632                         case O_ALTQ: {
1633                                 struct pf_mtag *at;
1634                                 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
1635
1636                                 match = 1;
1637                                 at = pf_find_mtag(m);
1638                                 if (at != NULL && at->qid != 0)
1639                                         break;
1640                                 at = pf_get_mtag(m);
1641                                 if (at == NULL) {
1642                                         /*
1643                                          * Let the packet fall back to the
1644                                          * default ALTQ.
1645                                          */
1646                                         break;
1647                                 }
1648                                 at->qid = altq->qid;
1649                                 if (is_ipv4)
1650                                         at->af = AF_INET;
1651                                 else
1652                                         at->af = AF_LINK;
1653                                 at->hdr = ip;
1654                                 break;
1655                         }
1656
1657                         case O_LOG:
1658                                 ipfw_log(f, hlen, args, m,
1659                                             oif, offset, tablearg, ip);
1660                                 match = 1;
1661                                 break;
1662
1663                         case O_PROB:
1664                                 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
1665                                 break;
1666
1667                         case O_VERREVPATH:
1668                                 /* Outgoing packets automatically pass/match */
1669                                 match = ((oif != NULL) ||
1670                                     (m->m_pkthdr.rcvif == NULL) ||
1671                                     (
1672 #ifdef INET6
1673                                     is_ipv6 ?
1674                                         verify_path6(&(args->f_id.src_ip6),
1675                                             m->m_pkthdr.rcvif) :
1676 #endif
1677                                     verify_path(src_ip, m->m_pkthdr.rcvif,
1678                                         args->f_id.fib)));
1679                                 break;
1680
1681                         case O_VERSRCREACH:
1682                                 /* Outgoing packets automatically pass/match */
1683                                 match = (hlen > 0 && ((oif != NULL) ||
1684 #ifdef INET6
1685                                     is_ipv6 ?
1686                                         verify_path6(&(args->f_id.src_ip6),
1687                                             NULL) :
1688 #endif
1689                                     verify_path(src_ip, NULL, args->f_id.fib)));
1690                                 break;
1691
1692                         case O_ANTISPOOF:
1693                                 /* Outgoing packets automatically pass/match */
1694                                 if (oif == NULL && hlen > 0 &&
1695                                     (  (is_ipv4 && in_localaddr(src_ip))
1696 #ifdef INET6
1697                                     || (is_ipv6 &&
1698                                         in6_localaddr(&(args->f_id.src_ip6)))
1699 #endif
1700                                     ))
1701                                         match =
1702 #ifdef INET6
1703                                             is_ipv6 ? verify_path6(
1704                                                 &(args->f_id.src_ip6),
1705                                                 m->m_pkthdr.rcvif) :
1706 #endif
1707                                             verify_path(src_ip,
1708                                                 m->m_pkthdr.rcvif,
1709                                                 args->f_id.fib);
1710                                 else
1711                                         match = 1;
1712                                 break;
1713
1714                         case O_IPSEC:
1715 #ifdef IPSEC
1716                                 match = (m_tag_find(m,
1717                                     PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
1718 #endif
1719                                 /* otherwise no match */
1720                                 break;
1721
1722 #ifdef INET6
1723                         case O_IP6_SRC:
1724                                 match = is_ipv6 &&
1725                                     IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
1726                                     &((ipfw_insn_ip6 *)cmd)->addr6);
1727                                 break;
1728
1729                         case O_IP6_DST:
1730                                 match = is_ipv6 &&
1731                                 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
1732                                     &((ipfw_insn_ip6 *)cmd)->addr6);
1733                                 break;
1734                         case O_IP6_SRC_MASK:
1735                         case O_IP6_DST_MASK:
1736                                 if (is_ipv6) {
1737                                         int i = cmdlen - 1;
1738                                         struct in6_addr p;
1739                                         struct in6_addr *d =
1740                                             &((ipfw_insn_ip6 *)cmd)->addr6;
1741
1742                                         for (; !match && i > 0; d += 2,
1743                                             i -= F_INSN_SIZE(struct in6_addr)
1744                                             * 2) {
1745                                                 p = (cmd->opcode ==
1746                                                     O_IP6_SRC_MASK) ?
1747                                                     args->f_id.src_ip6:
1748                                                     args->f_id.dst_ip6;
1749                                                 APPLY_MASK(&p, &d[1]);
1750                                                 match =
1751                                                     IN6_ARE_ADDR_EQUAL(&d[0],
1752                                                     &p);
1753                                         }
1754                                 }
1755                                 break;
1756
1757                         case O_FLOW6ID:
1758                                 match = is_ipv6 &&
1759                                     flow6id_match(args->f_id.flow_id6,
1760                                     (ipfw_insn_u32 *) cmd);
1761                                 break;
1762
1763                         case O_EXT_HDR:
1764                                 match = is_ipv6 &&
1765                                     (ext_hd & ((ipfw_insn *) cmd)->arg1);
1766                                 break;
1767
1768                         case O_IP6:
1769                                 match = is_ipv6;
1770                                 break;
1771 #endif
1772
1773                         case O_IP4:
1774                                 match = is_ipv4;
1775                                 break;
1776
1777                         case O_TAG: {
1778                                 struct m_tag *mtag;
1779                                 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
1780                                     tablearg : cmd->arg1;
1781
1782                                 /* Packet is already tagged with this tag? */
1783                                 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
1784
1785                                 /* We have `untag' action when F_NOT flag is
1786                                  * present. And we must remove this mtag from
1787                                  * mbuf and reset `match' to zero (`match' will
1788                                  * be inversed later).
1789                                  * Otherwise we should allocate new mtag and
1790                                  * push it into mbuf.
1791                                  */
1792                                 if (cmd->len & F_NOT) { /* `untag' action */
1793                                         if (mtag != NULL)
1794                                                 m_tag_delete(m, mtag);
1795                                         match = 0;
1796                                 } else {
1797                                         if (mtag == NULL) {
1798                                                 mtag = m_tag_alloc( MTAG_IPFW,
1799                                                     tag, 0, M_NOWAIT);
1800                                                 if (mtag != NULL)
1801                                                         m_tag_prepend(m, mtag);
1802                                         }
1803                                         match = 1;
1804                                 }
1805                                 break;
1806                         }
1807
1808                         case O_FIB: /* try match the specified fib */
1809                                 if (args->f_id.fib == cmd->arg1)
1810                                         match = 1;
1811                                 break;
1812
1813                         case O_TAGGED: {
1814                                 struct m_tag *mtag;
1815                                 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
1816                                     tablearg : cmd->arg1;
1817
1818                                 if (cmdlen == 1) {
1819                                         match = m_tag_locate(m, MTAG_IPFW,
1820                                             tag, NULL) != NULL;
1821                                         break;
1822                                 }
1823
1824                                 /* we have ranges */
1825                                 for (mtag = m_tag_first(m);
1826                                     mtag != NULL && !match;
1827                                     mtag = m_tag_next(m, mtag)) {
1828                                         uint16_t *p;
1829                                         int i;
1830
1831                                         if (mtag->m_tag_cookie != MTAG_IPFW)
1832                                                 continue;
1833
1834                                         p = ((ipfw_insn_u16 *)cmd)->ports;
1835                                         i = cmdlen - 1;
1836                                         for(; !match && i > 0; i--, p += 2)
1837                                                 match =
1838                                                     mtag->m_tag_id >= p[0] &&
1839                                                     mtag->m_tag_id <= p[1];
1840                                 }
1841                                 break;
1842                         }
1843
1844                         /*
1845                          * The second set of opcodes represents 'actions',
1846                          * i.e. the terminal part of a rule once the packet
1847                          * matches all previous patterns.
1848                          * Typically there is only one action for each rule,
1849                          * and the opcode is stored at the end of the rule
1850                          * (but there are exceptions -- see below).
1851                          *
1852                          * In general, here we set retval and terminate the
1853                          * outer loop (would be a 'break 3' in some language,
1854                          * but we need to set l=0, done=1)
1855                          *
1856                          * Exceptions:
1857                          * O_COUNT and O_SKIPTO actions:
1858                          *   instead of terminating, we jump to the next rule
1859                          *   (setting l=0), or to the SKIPTO target (setting
1860                          *   f/f_len, cmd and l as needed), respectively.
1861                          *
1862                          * O_TAG, O_LOG and O_ALTQ action parameters:
1863                          *   perform some action and set match = 1;
1864                          *
1865                          * O_LIMIT and O_KEEP_STATE: these opcodes are
1866                          *   not real 'actions', and are stored right
1867                          *   before the 'action' part of the rule.
1868                          *   These opcodes try to install an entry in the
1869                          *   state tables; if successful, we continue with
1870                          *   the next opcode (match=1; break;), otherwise
1871                          *   the packet must be dropped (set retval,
1872                          *   break loops with l=0, done=1)
1873                          *
1874                          * O_PROBE_STATE and O_CHECK_STATE: these opcodes
1875                          *   cause a lookup of the state table, and a jump
1876                          *   to the 'action' part of the parent rule
1877                          *   if an entry is found, or
1878                          *   (CHECK_STATE only) a jump to the next rule if
1879                          *   the entry is not found.
1880                          *   The result of the lookup is cached so that
1881                          *   further instances of these opcodes become NOPs.
1882                          *   The jump to the next rule is done by setting
1883                          *   l=0, cmdlen=0.
1884                          */
1885                         case O_LIMIT:
1886                         case O_KEEP_STATE:
1887                                 if (ipfw_install_state(f,
1888                                     (ipfw_insn_limit *)cmd, args, tablearg)) {
1889                                         /* error or limit violation */
1890                                         retval = IP_FW_DENY;
1891                                         l = 0;  /* exit inner loop */
1892                                         done = 1; /* exit outer loop */
1893                                 }
1894                                 match = 1;
1895                                 break;
1896
1897                         case O_PROBE_STATE:
1898                         case O_CHECK_STATE:
1899                                 /*
1900                                  * dynamic rules are checked at the first
1901                                  * keep-state or check-state occurrence,
1902                                  * with the result being stored in dyn_dir.
1903                                  * The compiler introduces a PROBE_STATE
1904                                  * instruction for us when we have a
1905                                  * KEEP_STATE (because PROBE_STATE needs
1906                                  * to be run first).
1907                                  */
1908                                 if (dyn_dir == MATCH_UNKNOWN &&
1909                                     (q = ipfw_lookup_dyn_rule(&args->f_id,
1910                                      &dyn_dir, proto == IPPROTO_TCP ?
1911                                         TCP(ulp) : NULL))
1912                                         != NULL) {
1913                                         /*
1914                                          * Found dynamic entry, update stats
1915                                          * and jump to the 'action' part of
1916                                          * the parent rule by setting
1917                                          * f, cmd, l and clearing cmdlen.
1918                                          */
1919                                         q->pcnt++;
1920                                         q->bcnt += pktlen;
1921                                         /* XXX we would like to have f_pos
1922                                          * readily accessible in the dynamic
1923                                          * rule, instead of having to
1924                                          * lookup q->rule.
1925                                          */
1926                                         f = q->rule;
1927                                         f_pos = ipfw_find_rule(chain,
1928                                                 f->rulenum, f->id);
1929                                         cmd = ACTION_PTR(f);
1930                                         l = f->cmd_len - f->act_ofs;
1931                                         ipfw_dyn_unlock();
1932                                         cmdlen = 0;
1933                                         match = 1;
1934                                         break;
1935                                 }
1936                                 /*
1937                                  * Dynamic entry not found. If CHECK_STATE,
1938                                  * skip to next rule, if PROBE_STATE just
1939                                  * ignore and continue with next opcode.
1940                                  */
1941                                 if (cmd->opcode == O_CHECK_STATE)
1942                                         l = 0;  /* exit inner loop */
1943                                 match = 1;
1944                                 break;
1945
1946                         case O_ACCEPT:
1947                                 retval = 0;     /* accept */
1948                                 l = 0;          /* exit inner loop */
1949                                 done = 1;       /* exit outer loop */
1950                                 break;
1951
1952                         case O_PIPE:
1953                         case O_QUEUE:
1954                                 set_match(args, f_pos, chain);
1955                                 args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ?
1956                                         tablearg : cmd->arg1;
1957                                 if (cmd->opcode == O_PIPE)
1958                                         args->rule.info |= IPFW_IS_PIPE;
1959                                 if (V_fw_one_pass)
1960                                         args->rule.info |= IPFW_ONEPASS;
1961                                 retval = IP_FW_DUMMYNET;
1962                                 l = 0;          /* exit inner loop */
1963                                 done = 1;       /* exit outer loop */
1964                                 break;
1965
1966                         case O_DIVERT:
1967                         case O_TEE:
1968                                 if (args->eh) /* not on layer 2 */
1969                                     break;
1970                                 /* otherwise this is terminal */
1971                                 l = 0;          /* exit inner loop */
1972                                 done = 1;       /* exit outer loop */
1973                                 retval = (cmd->opcode == O_DIVERT) ?
1974                                         IP_FW_DIVERT : IP_FW_TEE;
1975                                 set_match(args, f_pos, chain);
1976                                 args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ?
1977                                     tablearg : cmd->arg1;
1978                                 break;
1979
1980                         case O_COUNT:
1981                                 f->pcnt++;      /* update stats */
1982                                 f->bcnt += pktlen;
1983                                 f->timestamp = time_uptime;
1984                                 l = 0;          /* exit inner loop */
1985                                 break;
1986
1987                         case O_SKIPTO:
1988                             f->pcnt++;  /* update stats */
1989                             f->bcnt += pktlen;
1990                             f->timestamp = time_uptime;
1991                             /* If possible use cached f_pos (in f->next_rule),
1992                              * whose version is written in f->next_rule
1993                              * (horrible hacks to avoid changing the ABI).
1994                              */
1995                             if (cmd->arg1 != IP_FW_TABLEARG &&
1996                                     (uintptr_t)f->x_next == chain->id) {
1997                                 f_pos = (uintptr_t)f->next_rule;
1998                             } else {
1999                                 int i = (cmd->arg1 == IP_FW_TABLEARG) ?
2000                                         tablearg : cmd->arg1;
2001                                 /* make sure we do not jump backward */
2002                                 if (i <= f->rulenum)
2003                                     i = f->rulenum + 1;
2004                                 f_pos = ipfw_find_rule(chain, i, 0);
2005                                 /* update the cache */
2006                                 if (cmd->arg1 != IP_FW_TABLEARG) {
2007                                     f->next_rule =
2008                                         (void *)(uintptr_t)f_pos;
2009                                     f->x_next =
2010                                         (void *)(uintptr_t)chain->id;
2011                                 }
2012                             }
2013                             /*
2014                              * Skip disabled rules, and re-enter
2015                              * the inner loop with the correct
2016                              * f_pos, f, l and cmd.
2017                              * Also clear cmdlen and skip_or
2018                              */
2019                             for (; f_pos < chain->n_rules - 1 &&
2020                                     (V_set_disable &
2021                                      (1 << chain->map[f_pos]->set));
2022                                     f_pos++)
2023                                 ;
2024                             /* Re-enter the inner loop at the skipto rule. */
2025                             f = chain->map[f_pos];
2026                             l = f->cmd_len;
2027                             cmd = f->cmd;
2028                             match = 1;
2029                             cmdlen = 0;
2030                             skip_or = 0;
2031                             continue;
2032                             break;      /* not reached */
2033
2034                         case O_REJECT:
2035                                 /*
2036                                  * Drop the packet and send a reject notice
2037                                  * if the packet is not ICMP (or is an ICMP
2038                                  * query), and it is not multicast/broadcast.
2039                                  */
2040                                 if (hlen > 0 && is_ipv4 && offset == 0 &&
2041                                     (proto != IPPROTO_ICMP ||
2042                                      is_icmp_query(ICMP(ulp))) &&
2043                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2044                                     !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2045                                         send_reject(args, cmd->arg1, iplen, ip);
2046                                         m = args->m;
2047                                 }
2048                                 /* FALLTHROUGH */
2049 #ifdef INET6
2050                         case O_UNREACH6:
2051                                 if (hlen > 0 && is_ipv6 &&
2052                                     ((offset & IP6F_OFF_MASK) == 0) &&
2053                                     (proto != IPPROTO_ICMPV6 ||
2054                                      (is_icmp6_query(icmp6_type) == 1)) &&
2055                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2056                                     !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2057                                         send_reject6(
2058                                             args, cmd->arg1, hlen,
2059                                             (struct ip6_hdr *)ip);
2060                                         m = args->m;
2061                                 }
2062                                 /* FALLTHROUGH */
2063 #endif
2064                         case O_DENY:
2065                                 retval = IP_FW_DENY;
2066                                 l = 0;          /* exit inner loop */
2067                                 done = 1;       /* exit outer loop */
2068                                 break;
2069
2070                         case O_FORWARD_IP:
2071                                 if (args->eh)   /* not valid on layer2 pkts */
2072                                         break;
2073                                 if (!q || dyn_dir == MATCH_FORWARD) {
2074                                     struct sockaddr_in *sa;
2075                                     sa = &(((ipfw_insn_sa *)cmd)->sa);
2076                                     if (sa->sin_addr.s_addr == INADDR_ANY) {
2077                                         bcopy(sa, &args->hopstore,
2078                                                         sizeof(*sa));
2079                                         args->hopstore.sin_addr.s_addr =
2080                                                     htonl(tablearg);
2081                                         args->next_hop = &args->hopstore;
2082                                     } else {
2083                                         args->next_hop = sa;
2084                                     }
2085                                 }
2086                                 retval = IP_FW_PASS;
2087                                 l = 0;          /* exit inner loop */
2088                                 done = 1;       /* exit outer loop */
2089                                 break;
2090
2091                         case O_NETGRAPH:
2092                         case O_NGTEE:
2093                                 set_match(args, f_pos, chain);
2094                                 args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ?
2095                                         tablearg : cmd->arg1;
2096                                 if (V_fw_one_pass)
2097                                         args->rule.info |= IPFW_ONEPASS;
2098                                 retval = (cmd->opcode == O_NETGRAPH) ?
2099                                     IP_FW_NETGRAPH : IP_FW_NGTEE;
2100                                 l = 0;          /* exit inner loop */
2101                                 done = 1;       /* exit outer loop */
2102                                 break;
2103
2104                         case O_SETFIB: {
2105                                 uint32_t fib;
2106
2107                                 f->pcnt++;      /* update stats */
2108                                 f->bcnt += pktlen;
2109                                 f->timestamp = time_uptime;
2110                                 fib = (cmd->arg1 == IP_FW_TABLEARG) ? tablearg:
2111                                     cmd->arg1;
2112                                 if (fib >= rt_numfibs)
2113                                         fib = 0;
2114                                 M_SETFIB(m, fib);
2115                                 args->f_id.fib = fib;
2116                                 l = 0;          /* exit inner loop */
2117                                 break;
2118                         }
2119
2120                         case O_NAT:
2121                                 if (!IPFW_NAT_LOADED) {
2122                                     retval = IP_FW_DENY;
2123                                 } else {
2124                                     struct cfg_nat *t;
2125                                     int nat_id;
2126
2127                                     set_match(args, f_pos, chain);
2128                                     t = ((ipfw_insn_nat *)cmd)->nat;
2129                                     if (t == NULL) {
2130                                         nat_id = (cmd->arg1 == IP_FW_TABLEARG) ?
2131                                                 tablearg : cmd->arg1;
2132                                         t = (*lookup_nat_ptr)(&chain->nat, nat_id);
2133
2134                                         if (t == NULL) {
2135                                             retval = IP_FW_DENY;
2136                                             l = 0;      /* exit inner loop */
2137                                             done = 1;   /* exit outer loop */
2138                                             break;
2139                                         }
2140                                         if (cmd->arg1 != IP_FW_TABLEARG)
2141                                             ((ipfw_insn_nat *)cmd)->nat = t;
2142                                     }
2143                                     retval = ipfw_nat_ptr(args, t, m);
2144                                 }
2145                                 l = 0;          /* exit inner loop */
2146                                 done = 1;       /* exit outer loop */
2147                                 break;
2148
2149                         case O_REASS: {
2150                                 int ip_off;
2151
2152                                 f->pcnt++;
2153                                 f->bcnt += pktlen;
2154                                 l = 0;  /* in any case exit inner loop */
2155                                 ip_off = ntohs(ip->ip_off);
2156
2157                                 /* if not fragmented, go to next rule */
2158                                 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
2159                                     break;
2160                                 /*
2161                                  * ip_reass() expects len & off in host
2162                                  * byte order.
2163                                  */
2164                                 SET_HOST_IPLEN(ip);
2165
2166                                 args->m = m = ip_reass(m);
2167
2168                                 /*
2169                                  * do IP header checksum fixup.
2170                                  */
2171                                 if (m == NULL) { /* fragment got swallowed */
2172                                     retval = IP_FW_DENY;
2173                                 } else { /* good, packet complete */
2174                                     int hlen;
2175
2176                                     ip = mtod(m, struct ip *);
2177                                     hlen = ip->ip_hl << 2;
2178                                     SET_NET_IPLEN(ip);
2179                                     ip->ip_sum = 0;
2180                                     if (hlen == sizeof(struct ip))
2181                                         ip->ip_sum = in_cksum_hdr(ip);
2182                                     else
2183                                         ip->ip_sum = in_cksum(m, hlen);
2184                                     retval = IP_FW_REASS;
2185                                     set_match(args, f_pos, chain);
2186                                 }
2187                                 done = 1;       /* exit outer loop */
2188                                 break;
2189                         }
2190
2191                         default:
2192                                 panic("-- unknown opcode %d\n", cmd->opcode);
2193                         } /* end of switch() on opcodes */
2194                         /*
2195                          * if we get here with l=0, then match is irrelevant.
2196                          */
2197
2198                         if (cmd->len & F_NOT)
2199                                 match = !match;
2200
2201                         if (match) {
2202                                 if (cmd->len & F_OR)
2203                                         skip_or = 1;
2204                         } else {
2205                                 if (!(cmd->len & F_OR)) /* not an OR block, */
2206                                         break;          /* try next rule    */
2207                         }
2208
2209                 }       /* end of inner loop, scan opcodes */
2210
2211                 if (done)
2212                         break;
2213
2214 /* next_rule:; */       /* try next rule                */
2215
2216         }               /* end of outer for, scan rules */
2217
2218         if (done) {
2219                 struct ip_fw *rule = chain->map[f_pos];
2220                 /* Update statistics */
2221                 rule->pcnt++;
2222                 rule->bcnt += pktlen;
2223                 rule->timestamp = time_uptime;
2224         } else {
2225                 retval = IP_FW_DENY;
2226                 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2227         }
2228         IPFW_RUNLOCK(chain);
2229 #ifdef __FreeBSD__
2230         if (ucred_cache != NULL)
2231                 crfree(ucred_cache);
2232 #endif
2233         return (retval);
2234
2235 pullup_failed:
2236         if (V_fw_verbose)
2237                 printf("ipfw: pullup failed\n");
2238         return (IP_FW_DENY);
2239 }
2240
2241 /*
2242  * Module and VNET glue
2243  */
2244
2245 /*
2246  * Stuff that must be initialised only on boot or module load
2247  */
2248 static int
2249 ipfw_init(void)
2250 {
2251         int error = 0;
2252
2253         ipfw_dyn_attach();
2254         /*
2255          * Only print out this stuff the first time around,
2256          * when called from the sysinit code.
2257          */
2258         printf("ipfw2 "
2259 #ifdef INET6
2260                 "(+ipv6) "
2261 #endif
2262                 "initialized, divert %s, nat %s, "
2263                 "rule-based forwarding "
2264 #ifdef IPFIREWALL_FORWARD
2265                 "enabled, "
2266 #else
2267                 "disabled, "
2268 #endif
2269                 "default to %s, logging ",
2270 #ifdef IPDIVERT
2271                 "enabled",
2272 #else
2273                 "loadable",
2274 #endif
2275 #ifdef IPFIREWALL_NAT
2276                 "enabled",
2277 #else
2278                 "loadable",
2279 #endif
2280                 default_to_accept ? "accept" : "deny");
2281
2282         /*
2283          * Note: V_xxx variables can be accessed here but the vnet specific
2284          * initializer may not have been called yet for the VIMAGE case.
2285          * Tuneables will have been processed. We will print out values for
2286          * the default vnet.
2287          * XXX This should all be rationalized AFTER 8.0
2288          */
2289         if (V_fw_verbose == 0)
2290                 printf("disabled\n");
2291         else if (V_verbose_limit == 0)
2292                 printf("unlimited\n");
2293         else
2294                 printf("limited to %d packets/entry by default\n",
2295                     V_verbose_limit);
2296
2297         ipfw_log_bpf(1); /* init */
2298         return (error);
2299 }
2300
2301 /*
2302  * Called for the removal of the last instance only on module unload.
2303  */
2304 static void
2305 ipfw_destroy(void)
2306 {
2307
2308         ipfw_log_bpf(0); /* uninit */
2309         ipfw_dyn_detach();
2310         printf("IP firewall unloaded\n");
2311 }
2312
2313 /*
2314  * Stuff that must be initialized for every instance
2315  * (including the first of course).
2316  */
2317 static int
2318 vnet_ipfw_init(const void *unused)
2319 {
2320         int error;
2321         struct ip_fw *rule = NULL;
2322         struct ip_fw_chain *chain;
2323
2324         chain = &V_layer3_chain;
2325
2326         /* First set up some values that are compile time options */
2327         V_autoinc_step = 100;   /* bounded to 1..1000 in add_rule() */
2328         V_fw_deny_unknown_exthdrs = 1;
2329 #ifdef IPFIREWALL_VERBOSE
2330         V_fw_verbose = 1;
2331 #endif
2332 #ifdef IPFIREWALL_VERBOSE_LIMIT
2333         V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2334 #endif
2335 #ifdef IPFIREWALL_NAT
2336         LIST_INIT(&chain->nat);
2337 #endif
2338
2339         /* insert the default rule and create the initial map */
2340         chain->n_rules = 1;
2341         chain->static_len = sizeof(struct ip_fw);
2342         chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_NOWAIT | M_ZERO);
2343         if (chain->map)
2344                 rule = malloc(chain->static_len, M_IPFW, M_NOWAIT | M_ZERO);
2345         if (rule == NULL) {
2346                 if (chain->map)
2347                         free(chain->map, M_IPFW);
2348                 printf("ipfw2: ENOSPC initializing default rule "
2349                         "(support disabled)\n");
2350                 return (ENOSPC);
2351         }
2352         error = ipfw_init_tables(chain);
2353         if (error) {
2354                 panic("init_tables"); /* XXX Marko fix this ! */
2355         }
2356
2357         /* fill and insert the default rule */
2358         rule->act_ofs = 0;
2359         rule->rulenum = IPFW_DEFAULT_RULE;
2360         rule->cmd_len = 1;
2361         rule->set = RESVD_SET;
2362         rule->cmd[0].len = 1;
2363         rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
2364         chain->rules = chain->default_rule = chain->map[0] = rule;
2365         chain->id = rule->id = 1;
2366
2367         IPFW_LOCK_INIT(chain);
2368         ipfw_dyn_init();
2369
2370         /* First set up some values that are compile time options */
2371         V_ipfw_vnet_ready = 1;          /* Open for business */
2372
2373         /*
2374          * Hook the sockopt handler, and the layer2 (V_ip_fw_chk_ptr)
2375          * and pfil hooks for ipv4 and ipv6. Even if the latter two fail
2376          * we still keep the module alive because the sockopt and
2377          * layer2 paths are still useful.
2378          * ipfw[6]_hook return 0 on success, ENOENT on failure,
2379          * so we can ignore the exact return value and just set a flag.
2380          *
2381          * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
2382          * changes in the underlying (per-vnet) variables trigger
2383          * immediate hook()/unhook() calls.
2384          * In layer2 we have the same behaviour, except that V_ether_ipfw
2385          * is checked on each packet because there are no pfil hooks.
2386          */
2387         V_ip_fw_ctl_ptr = ipfw_ctl;
2388         V_ip_fw_chk_ptr = ipfw_chk;
2389         error = ipfw_attach_hooks(1);
2390         return (error);
2391 }
2392
2393 /*
2394  * Called for the removal of each instance.
2395  */
2396 static int
2397 vnet_ipfw_uninit(const void *unused)
2398 {
2399         struct ip_fw *reap, *rule;
2400         struct ip_fw_chain *chain = &V_layer3_chain;
2401         int i;
2402
2403         V_ipfw_vnet_ready = 0; /* tell new callers to go away */
2404         /*
2405          * disconnect from ipv4, ipv6, layer2 and sockopt.
2406          * Then grab, release and grab again the WLOCK so we make
2407          * sure the update is propagated and nobody will be in.
2408          */
2409         (void)ipfw_attach_hooks(0 /* detach */);
2410         V_ip_fw_chk_ptr = NULL;
2411         V_ip_fw_ctl_ptr = NULL;
2412         IPFW_UH_WLOCK(chain);
2413         IPFW_UH_WUNLOCK(chain);
2414         IPFW_UH_WLOCK(chain);
2415
2416         IPFW_WLOCK(chain);
2417         IPFW_WUNLOCK(chain);
2418         IPFW_WLOCK(chain);
2419
2420         ipfw_dyn_uninit(0);     /* run the callout_drain */
2421         ipfw_destroy_tables(chain);
2422         reap = NULL;
2423         for (i = 0; i < chain->n_rules; i++) {
2424                 rule = chain->map[i];
2425                 rule->x_next = reap;
2426                 reap = rule;
2427         }
2428         if (chain->map)
2429                 free(chain->map, M_IPFW);
2430         IPFW_WUNLOCK(chain);
2431         IPFW_UH_WUNLOCK(chain);
2432         if (reap != NULL)
2433                 ipfw_reap_rules(reap);
2434         IPFW_LOCK_DESTROY(chain);
2435         ipfw_dyn_uninit(1);     /* free the remaining parts */
2436         return 0;
2437 }
2438
2439 /*
2440  * Module event handler.
2441  * In general we have the choice of handling most of these events by the
2442  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
2443  * use the SYSINIT handlers as they are more capable of expressing the
2444  * flow of control during module and vnet operations, so this is just
2445  * a skeleton. Note there is no SYSINIT equivalent of the module
2446  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
2447  */
2448 static int
2449 ipfw_modevent(module_t mod, int type, void *unused)
2450 {
2451         int err = 0;
2452
2453         switch (type) {
2454         case MOD_LOAD:
2455                 /* Called once at module load or
2456                  * system boot if compiled in. */
2457                 break;
2458         case MOD_QUIESCE:
2459                 /* Called before unload. May veto unloading. */
2460                 break;
2461         case MOD_UNLOAD:
2462                 /* Called during unload. */
2463                 break;
2464         case MOD_SHUTDOWN:
2465                 /* Called during system shutdown. */
2466                 break;
2467         default:
2468                 err = EOPNOTSUPP;
2469                 break;
2470         }
2471         return err;
2472 }
2473
2474 static moduledata_t ipfwmod = {
2475         "ipfw",
2476         ipfw_modevent,
2477         0
2478 };
2479
2480 /* Define startup order. */
2481 #define IPFW_SI_SUB_FIREWALL    SI_SUB_PROTO_IFATTACHDOMAIN
2482 #define IPFW_MODEVENT_ORDER     (SI_ORDER_ANY - 255) /* On boot slot in here. */
2483 #define IPFW_MODULE_ORDER       (IPFW_MODEVENT_ORDER + 1) /* A little later. */
2484 #define IPFW_VNET_ORDER         (IPFW_MODEVENT_ORDER + 2) /* Later still. */
2485
2486 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
2487 MODULE_VERSION(ipfw, 2);
2488 /* should declare some dependencies here */
2489
2490 /*
2491  * Starting up. Done in order after ipfwmod() has been called.
2492  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
2493  */
2494 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2495             ipfw_init, NULL);
2496 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2497             vnet_ipfw_init, NULL);
2498
2499 /*
2500  * Closing up shop. These are done in REVERSE ORDER, but still
2501  * after ipfwmod() has been called. Not called on reboot.
2502  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
2503  * or when the module is unloaded.
2504  */
2505 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2506             ipfw_destroy, NULL);
2507 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2508             vnet_ipfw_uninit, NULL);
2509 /* end of file */