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 #include "opt_ipfw.h"
  34 #if !defined(KLD_MODULE)
  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/in6_pcb.h>
  88 #include <netinet6/scope6_var.h>
  89 #include <netinet6/ip6_var.h>
  90 #endif
  91
  92 #include <machine/in_cksum.h>   /* XXX for in_cksum */
  93
  94 #ifdef MAC
  95 #include <security/mac/mac_framework.h>
  96 #endif
  97
  98 /*
  99  * static variables followed by global ones.
 100  * All ipfw global variables are here.
 101  */
 102
 103 /* ipfw_vnet_ready controls when we are open for business */
 104 static VNET_DEFINE(int, ipfw_vnet_ready) = 0;
 105 #define V_ipfw_vnet_ready       VNET(ipfw_vnet_ready)
 106
 107 static VNET_DEFINE(int, fw_deny_unknown_exthdrs);
 108 #define V_fw_deny_unknown_exthdrs       VNET(fw_deny_unknown_exthdrs)
 109
 110 static VNET_DEFINE(int, fw_permit_single_frag6) = 1;
 111 #define V_fw_permit_single_frag6        VNET(fw_permit_single_frag6)
 112
 113 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
 114 static int default_to_accept = 1;
 115 #else
 116 static int default_to_accept;
 117 #endif
 118
 119 VNET_DEFINE(int, autoinc_step);
 120 VNET_DEFINE(int, fw_one_pass) = 1;
 121
 122 /*
 123  * Each rule belongs to one of 32 different sets (0..31).
 124  * The variable set_disable contains one bit per set.
 125  * If the bit is set, all rules in the corresponding set
 126  * are disabled. Set RESVD_SET(31) is reserved for the default rule
 127  * and rules that are not deleted by the flush command,
 128  * and CANNOT be disabled.
 129  * Rules in set RESVD_SET can only be deleted individually.
 130  */
 131 VNET_DEFINE(u_int32_t, set_disable);
 132 #define V_set_disable                   VNET(set_disable)
 133
 134 VNET_DEFINE(int, fw_verbose);
 135 /* counter for ipfw_log(NULL...) */
 136 VNET_DEFINE(u_int64_t, norule_counter);
 137 VNET_DEFINE(int, verbose_limit);
 138
 139 /* layer3_chain contains the list of rules for layer 3 */
 140 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
 141
 142 ipfw_nat_t *ipfw_nat_ptr = NULL;
 143 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
 144 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
 145 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
 146 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
 147 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
 148
 149 #ifdef SYSCTL_NODE
 150 uint32_t dummy_def = IPFW_DEFAULT_RULE;
 151 uint32_t dummy_tables_max = IPFW_TABLES_MAX;
 152
 153 SYSBEGIN(f3)
 154
 155 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 156 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
 157     CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
 158     "Only do a single pass through ipfw when using dummynet(4)");
 159 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
 160     CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
 161     "Rule number auto-increment step");
 162 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose,
 163     CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
 164     "Log matches to ipfw rules");
 165 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
 166     CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
 167     "Set upper limit of matches of ipfw rules logged");
 168 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
 169     &dummy_def, 0,
 170     "The default/max possible rule number.");
 171 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_RD,
 172     &dummy_tables_max, 0,
 173     "The maximum number of tables.");
 174 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
 175     &default_to_accept, 0,
 176     "Make the default rule accept all packets.");
 177 TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept);
 178 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count,
 179     CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
 180     "Number of static rules");
 181
 182 #ifdef INET6
 183 SYSCTL_DECL(_net_inet6_ip6);
 184 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 185 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
 186     CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0,
 187     "Deny packets with unknown IPv6 Extension Headers");
 188 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
 189     CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_permit_single_frag6), 0,
 190     "Permit single packet IPv6 fragments");
 191 #endif /* INET6 */
 192
 193 SYSEND
 194
 195 #endif /* SYSCTL_NODE */
 196
 197
 198 /*
 199  * Some macros used in the various matching options.
 200  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
 201  * Other macros just cast void * into the appropriate type
 202  */
 203 #define L3HDR(T, ip)    ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
 204 #define TCP(p)          ((struct tcphdr *)(p))
 205 #define SCTP(p)         ((struct sctphdr *)(p))
 206 #define UDP(p)          ((struct udphdr *)(p))
 207 #define ICMP(p)         ((struct icmphdr *)(p))
 208 #define ICMP6(p)        ((struct icmp6_hdr *)(p))
 209
 210 static __inline int
 211 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
 212 {
 213         int type = icmp->icmp_type;
 214
 215         return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
 216 }
 217
 218 #define TT      ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
 219     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
 220
 221 static int
 222 is_icmp_query(struct icmphdr *icmp)
 223 {
 224         int type = icmp->icmp_type;
 225
 226         return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
 227 }
 228 #undef TT
 229
 230 /*
 231  * The following checks use two arrays of 8 or 16 bits to store the
 232  * bits that we want set or clear, respectively. They are in the
 233  * low and high half of cmd->arg1 or cmd->d[0].
 234  *
 235  * We scan options and store the bits we find set. We succeed if
 236  *
 237  *      (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
 238  *
 239  * The code is sometimes optimized not to store additional variables.
 240  */
 241
 242 static int
 243 flags_match(ipfw_insn *cmd, u_int8_t bits)
 244 {
 245         u_char want_clear;
 246         bits = ~bits;
 247
 248         if ( ((cmd->arg1 & 0xff) & bits) != 0)
 249                 return 0; /* some bits we want set were clear */
 250         want_clear = (cmd->arg1 >> 8) & 0xff;
 251         if ( (want_clear & bits) != want_clear)
 252                 return 0; /* some bits we want clear were set */
 253         return 1;
 254 }
 255
 256 static int
 257 ipopts_match(struct ip *ip, ipfw_insn *cmd)
 258 {
 259         int optlen, bits = 0;
 260         u_char *cp = (u_char *)(ip + 1);
 261         int x = (ip->ip_hl << 2) - sizeof (struct ip);
 262
 263         for (; x > 0; x -= optlen, cp += optlen) {
 264                 int opt = cp[IPOPT_OPTVAL];
 265
 266                 if (opt == IPOPT_EOL)
 267                         break;
 268                 if (opt == IPOPT_NOP)
 269                         optlen = 1;
 270                 else {
 271                         optlen = cp[IPOPT_OLEN];
 272                         if (optlen <= 0 || optlen > x)
 273                                 return 0; /* invalid or truncated */
 274                 }
 275                 switch (opt) {
 276
 277                 default:
 278                         break;
 279
 280                 case IPOPT_LSRR:
 281                         bits |= IP_FW_IPOPT_LSRR;
 282                         break;
 283
 284                 case IPOPT_SSRR:
 285                         bits |= IP_FW_IPOPT_SSRR;
 286                         break;
 287
 288                 case IPOPT_RR:
 289                         bits |= IP_FW_IPOPT_RR;
 290                         break;
 291
 292                 case IPOPT_TS:
 293                         bits |= IP_FW_IPOPT_TS;
 294                         break;
 295                 }
 296         }
 297         return (flags_match(cmd, bits));
 298 }
 299
 300 static int
 301 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
 302 {
 303         int optlen, bits = 0;
 304         u_char *cp = (u_char *)(tcp + 1);
 305         int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
 306
 307         for (; x > 0; x -= optlen, cp += optlen) {
 308                 int opt = cp[0];
 309                 if (opt == TCPOPT_EOL)
 310                         break;
 311                 if (opt == TCPOPT_NOP)
 312                         optlen = 1;
 313                 else {
 314                         optlen = cp[1];
 315                         if (optlen <= 0)
 316                                 break;
 317                 }
 318
 319                 switch (opt) {
 320
 321                 default:
 322                         break;
 323
 324                 case TCPOPT_MAXSEG:
 325                         bits |= IP_FW_TCPOPT_MSS;
 326                         break;
 327
 328                 case TCPOPT_WINDOW:
 329                         bits |= IP_FW_TCPOPT_WINDOW;
 330                         break;
 331
 332                 case TCPOPT_SACK_PERMITTED:
 333                 case TCPOPT_SACK:
 334                         bits |= IP_FW_TCPOPT_SACK;
 335                         break;
 336
 337                 case TCPOPT_TIMESTAMP:
 338                         bits |= IP_FW_TCPOPT_TS;
 339                         break;
 340
 341                 }
 342         }
 343         return (flags_match(cmd, bits));
 344 }
 345
 346 static int
 347 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd)
 348 {
 349         if (ifp == NULL)        /* no iface with this packet, match fails */
 350                 return 0;
 351         /* Check by name or by IP address */
 352         if (cmd->name[0] != '\0') { /* match by name */
 353                 /* Check name */
 354                 if (cmd->p.glob) {
 355                         if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
 356                                 return(1);
 357                 } else {
 358                         if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
 359                                 return(1);
 360                 }
 361         } else {
 362 #ifdef __FreeBSD__      /* and OSX too ? */
 363                 struct ifaddr *ia;
 364
 365                 if_addr_rlock(ifp);
 366                 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
 367                         if (ia->ifa_addr->sa_family != AF_INET)
 368                                 continue;
 369                         if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
 370                             (ia->ifa_addr))->sin_addr.s_addr) {
 371                                 if_addr_runlock(ifp);
 372                                 return(1);      /* match */
 373                         }
 374                 }
 375                 if_addr_runlock(ifp);
 376 #endif /* __FreeBSD__ */
 377         }
 378         return(0);      /* no match, fail ... */
 379 }
 380
 381 /*
 382  * The verify_path function checks if a route to the src exists and
 383  * if it is reachable via ifp (when provided).
 384  *
 385  * The 'verrevpath' option checks that the interface that an IP packet
 386  * arrives on is the same interface that traffic destined for the
 387  * packet's source address would be routed out of.
 388  * The 'versrcreach' option just checks that the source address is
 389  * reachable via any route (except default) in the routing table.
 390  * These two are a measure to block forged packets. This is also
 391  * commonly known as "anti-spoofing" or Unicast Reverse Path
 392  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
 393  * is purposely reminiscent of the Cisco IOS command,
 394  *
 395  *   ip verify unicast reverse-path
 396  *   ip verify unicast source reachable-via any
 397  *
 398  * which implements the same functionality. But note that the syntax
 399  * is misleading, and the check may be performed on all IP packets
 400  * whether unicast, multicast, or broadcast.
 401  */
 402 static int
 403 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
 404 {
 405 #ifndef __FreeBSD__
 406         return 0;
 407 #else
 408         struct route ro;
 409         struct sockaddr_in *dst;
 410
 411         bzero(&ro, sizeof(ro));
 412
 413         dst = (struct sockaddr_in *)&(ro.ro_dst);
 414         dst->sin_family = AF_INET;
 415         dst->sin_len = sizeof(*dst);
 416         dst->sin_addr = src;
 417         in_rtalloc_ign(&ro, 0, fib);
 418
 419         if (ro.ro_rt == NULL)
 420                 return 0;
 421
 422         /*
 423          * If ifp is provided, check for equality with rtentry.
 424          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 425          * in order to pass packets injected back by if_simloop():
 426          * if useloopback == 1 routing entry (via lo0) for our own address
 427          * may exist, so we need to handle routing assymetry.
 428          */
 429         if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
 430                 RTFREE(ro.ro_rt);
 431                 return 0;
 432         }
 433
 434         /* if no ifp provided, check if rtentry is not default route */
 435         if (ifp == NULL &&
 436              satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) {
 437                 RTFREE(ro.ro_rt);
 438                 return 0;
 439         }
 440
 441         /* or if this is a blackhole/reject route */
 442         if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
 443                 RTFREE(ro.ro_rt);
 444                 return 0;
 445         }
 446
 447         /* found valid route */
 448         RTFREE(ro.ro_rt);
 449         return 1;
 450 #endif /* __FreeBSD__ */
 451 }
 452
 453 #ifdef INET6
 454 /*
 455  * ipv6 specific rules here...
 456  */
 457 static __inline int
 458 icmp6type_match (int type, ipfw_insn_u32 *cmd)
 459 {
 460         return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
 461 }
 462
 463 static int
 464 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
 465 {
 466         int i;
 467         for (i=0; i <= cmd->o.arg1; ++i )
 468                 if (curr_flow == cmd->d[i] )
 469                         return 1;
 470         return 0;
 471 }
 472
 473 /* support for IP6_*_ME opcodes */
 474 static int
 475 search_ip6_addr_net (struct in6_addr * ip6_addr)
 476 {
 477         struct ifnet *mdc;
 478         struct ifaddr *mdc2;
 479         struct in6_ifaddr *fdm;
 480         struct in6_addr copia;
 481
 482         TAILQ_FOREACH(mdc, &V_ifnet, if_link) {
 483                 if_addr_rlock(mdc);
 484                 TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) {
 485                         if (mdc2->ifa_addr->sa_family == AF_INET6) {
 486                                 fdm = (struct in6_ifaddr *)mdc2;
 487                                 copia = fdm->ia_addr.sin6_addr;
 488                                 /* need for leaving scope_id in the sock_addr */
 489                                 in6_clearscope(&copia);
 490                                 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) {
 491                                         if_addr_runlock(mdc);
 492                                         return 1;
 493                                 }
 494                         }
 495                 }
 496                 if_addr_runlock(mdc);
 497         }
 498         return 0;
 499 }
 500
 501 static int
 502 verify_path6(struct in6_addr *src, struct ifnet *ifp)
 503 {
 504         struct route_in6 ro;
 505         struct sockaddr_in6 *dst;
 506
 507         bzero(&ro, sizeof(ro));
 508
 509         dst = (struct sockaddr_in6 * )&(ro.ro_dst);
 510         dst->sin6_family = AF_INET6;
 511         dst->sin6_len = sizeof(*dst);
 512         dst->sin6_addr = *src;
 513         /* XXX MRT 0 for ipv6 at this time */
 514         rtalloc_ign((struct route *)&ro, 0);
 515
 516         if (ro.ro_rt == NULL)
 517                 return 0;
 518
 519         /*
 520          * if ifp is provided, check for equality with rtentry
 521          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 522          * to support the case of sending packets to an address of our own.
 523          * (where the former interface is the first argument of if_simloop()
 524          *  (=ifp), the latter is lo0)
 525          */
 526         if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
 527                 RTFREE(ro.ro_rt);
 528                 return 0;
 529         }
 530
 531         /* if no ifp provided, check if rtentry is not default route */
 532         if (ifp == NULL &&
 533             IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) {
 534                 RTFREE(ro.ro_rt);
 535                 return 0;
 536         }
 537
 538         /* or if this is a blackhole/reject route */
 539         if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
 540                 RTFREE(ro.ro_rt);
 541                 return 0;
 542         }
 543
 544         /* found valid route */
 545         RTFREE(ro.ro_rt);
 546         return 1;
 547
 548 }
 549
 550 static int
 551 is_icmp6_query(int icmp6_type)
 552 {
 553         if ((icmp6_type <= ICMP6_MAXTYPE) &&
 554             (icmp6_type == ICMP6_ECHO_REQUEST ||
 555             icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
 556             icmp6_type == ICMP6_WRUREQUEST ||
 557             icmp6_type == ICMP6_FQDN_QUERY ||
 558             icmp6_type == ICMP6_NI_QUERY))
 559                 return (1);
 560
 561         return (0);
 562 }
 563
 564 static void
 565 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
 566 {
 567         struct mbuf *m;
 568
 569         m = args->m;
 570         if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
 571                 struct tcphdr *tcp;
 572                 tcp = (struct tcphdr *)((char *)ip6 + hlen);
 573
 574                 if ((tcp->th_flags & TH_RST) == 0) {
 575                         struct mbuf *m0;
 576                         m0 = ipfw_send_pkt(args->m, &(args->f_id),
 577                             ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 578                             tcp->th_flags | TH_RST);
 579                         if (m0 != NULL)
 580                                 ip6_output(m0, NULL, NULL, 0, NULL, NULL,
 581                                     NULL);
 582                 }
 583                 FREE_PKT(m);
 584         } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
 585 #if 0
 586                 /*
 587                  * Unlike above, the mbufs need to line up with the ip6 hdr,
 588                  * as the contents are read. We need to m_adj() the
 589                  * needed amount.
 590                  * The mbuf will however be thrown away so we can adjust it.
 591                  * Remember we did an m_pullup on it already so we
 592                  * can make some assumptions about contiguousness.
 593                  */
 594                 if (args->L3offset)
 595                         m_adj(m, args->L3offset);
 596 #endif
 597                 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
 598         } else
 599                 FREE_PKT(m);
 600
 601         args->m = NULL;
 602 }
 603
 604 #endif /* INET6 */
 605
 606
 607 /*
 608  * sends a reject message, consuming the mbuf passed as an argument.
 609  */
 610 static void
 611 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
 612 {
 613
 614 #if 0
 615         /* XXX When ip is not guaranteed to be at mtod() we will
 616          * need to account for this */
 617          * The mbuf will however be thrown away so we can adjust it.
 618          * Remember we did an m_pullup on it already so we
 619          * can make some assumptions about contiguousness.
 620          */
 621         if (args->L3offset)
 622                 m_adj(m, args->L3offset);
 623 #endif
 624         if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
 625                 /* We need the IP header in host order for icmp_error(). */
 626                 SET_HOST_IPLEN(ip);
 627                 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
 628         } else if (args->f_id.proto == IPPROTO_TCP) {
 629                 struct tcphdr *const tcp =
 630                     L3HDR(struct tcphdr, mtod(args->m, struct ip *));
 631                 if ( (tcp->th_flags & TH_RST) == 0) {
 632                         struct mbuf *m;
 633                         m = ipfw_send_pkt(args->m, &(args->f_id),
 634                                 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 635                                 tcp->th_flags | TH_RST);
 636                         if (m != NULL)
 637                                 ip_output(m, NULL, NULL, 0, NULL, NULL);
 638                 }
 639                 FREE_PKT(args->m);
 640         } else
 641                 FREE_PKT(args->m);
 642         args->m = NULL;
 643 }
 644
 645 /*
 646  * Support for uid/gid/jail lookup. These tests are expensive
 647  * (because we may need to look into the list of active sockets)
 648  * so we cache the results. ugid_lookupp is 0 if we have not
 649  * yet done a lookup, 1 if we succeeded, and -1 if we tried
 650  * and failed. The function always returns the match value.
 651  * We could actually spare the variable and use *uc, setting
 652  * it to '(void *)check_uidgid if we have no info, NULL if
 653  * we tried and failed, or any other value if successful.
 654  */
 655 static int
 656 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
 657     struct ucred **uc)
 658 {
 659 #ifndef __FreeBSD__
 660         /* XXX */
 661         return cred_check(insn, proto, oif,
 662             dst_ip, dst_port, src_ip, src_port,
 663             (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
 664 #else  /* FreeBSD */
 665         struct in_addr src_ip, dst_ip;
 666         struct inpcbinfo *pi;
 667         struct ipfw_flow_id *id;
 668         struct inpcb *pcb, *inp;
 669         struct ifnet *oif;
 670         int lookupflags;
 671         int match;
 672
 673         id = &args->f_id;
 674         inp = args->inp;
 675         oif = args->oif;
 676
 677         /*
 678          * Check to see if the UDP or TCP stack supplied us with
 679          * the PCB. If so, rather then holding a lock and looking
 680          * up the PCB, we can use the one that was supplied.
 681          */
 682         if (inp && *ugid_lookupp == 0) {
 683                 INP_LOCK_ASSERT(inp);
 684                 if (inp->inp_socket != NULL) {
 685                         *uc = crhold(inp->inp_cred);
 686                         *ugid_lookupp = 1;
 687                 } else
 688                         *ugid_lookupp = -1;
 689         }
 690         /*
 691          * If we have already been here and the packet has no
 692          * PCB entry associated with it, then we can safely
 693          * assume that this is a no match.
 694          */
 695         if (*ugid_lookupp == -1)
 696                 return (0);
 697         if (id->proto == IPPROTO_TCP) {
 698                 lookupflags = 0;
 699                 pi = &V_tcbinfo;
 700         } else if (id->proto == IPPROTO_UDP) {
 701                 lookupflags = INPLOOKUP_WILDCARD;
 702                 pi = &V_udbinfo;
 703         } else
 704                 return 0;
 705         lookupflags |= INPLOOKUP_RLOCKPCB;
 706         match = 0;
 707         if (*ugid_lookupp == 0) {
 708                 if (id->addr_type == 6) {
 709 #ifdef INET6
 710                         if (oif == NULL)
 711                                 pcb = in6_pcblookup_mbuf(pi,
 712                                     &id->src_ip6, htons(id->src_port),
 713                                     &id->dst_ip6, htons(id->dst_port),
 714                                     lookupflags, oif, args->m);
 715                         else
 716                                 pcb = in6_pcblookup_mbuf(pi,
 717                                     &id->dst_ip6, htons(id->dst_port),
 718                                     &id->src_ip6, htons(id->src_port),
 719                                     lookupflags, oif, args->m);
 720 #else
 721                         *ugid_lookupp = -1;
 722                         return (0);
 723 #endif
 724                 } else {
 725                         src_ip.s_addr = htonl(id->src_ip);
 726                         dst_ip.s_addr = htonl(id->dst_ip);
 727                         if (oif == NULL)
 728                                 pcb = in_pcblookup_mbuf(pi,
 729                                     src_ip, htons(id->src_port),
 730                                     dst_ip, htons(id->dst_port),
 731                                     lookupflags, oif, args->m);
 732                         else
 733                                 pcb = in_pcblookup_mbuf(pi,
 734                                     dst_ip, htons(id->dst_port),
 735                                     src_ip, htons(id->src_port),
 736                                     lookupflags, oif, args->m);
 737                 }
 738                 if (pcb != NULL) {
 739                         INP_RLOCK_ASSERT(pcb);
 740                         *uc = crhold(pcb->inp_cred);
 741                         *ugid_lookupp = 1;
 742                         INP_RUNLOCK(pcb);
 743                 }
 744                 if (*ugid_lookupp == 0) {
 745                         /*
 746                          * We tried and failed, set the variable to -1
 747                          * so we will not try again on this packet.
 748                          */
 749                         *ugid_lookupp = -1;
 750                         return (0);
 751                 }
 752         }
 753         if (insn->o.opcode == O_UID)
 754                 match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
 755         else if (insn->o.opcode == O_GID)
 756                 match = groupmember((gid_t)insn->d[0], *uc);
 757         else if (insn->o.opcode == O_JAIL)
 758                 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
 759         return (match);
 760 #endif /* __FreeBSD__ */
 761 }
 762
 763 /*
 764  * Helper function to set args with info on the rule after the matching
 765  * one. slot is precise, whereas we guess rule_id as they are
 766  * assigned sequentially.
 767  */
 768 static inline void
 769 set_match(struct ip_fw_args *args, int slot,
 770         struct ip_fw_chain *chain)
 771 {
 772         args->rule.chain_id = chain->id;
 773         args->rule.slot = slot + 1; /* we use 0 as a marker */
 774         args->rule.rule_id = 1 + chain->map[slot]->id;
 775         args->rule.rulenum = chain->map[slot]->rulenum;
 776 }
 777
 778 /*
 779  * The main check routine for the firewall.
 780  *
 781  * All arguments are in args so we can modify them and return them
 782  * back to the caller.
 783  *
 784  * Parameters:
 785  *
 786  *      args->m (in/out) The packet; we set to NULL when/if we nuke it.
 787  *              Starts with the IP header.
 788  *      args->eh (in)   Mac header if present, NULL for layer3 packet.
 789  *      args->L3offset  Number of bytes bypassed if we came from L2.
 790  *                      e.g. often sizeof(eh)  ** NOTYET **
 791  *      args->oif       Outgoing interface, NULL if packet is incoming.
 792  *              The incoming interface is in the mbuf. (in)
 793  *      args->divert_rule (in/out)
 794  *              Skip up to the first rule past this rule number;
 795  *              upon return, non-zero port number for divert or tee.
 796  *
 797  *      args->rule      Pointer to the last matching rule (in/out)
 798  *      args->next_hop  Socket we are forwarding to (out).
 799  *      args->next_hop6 IPv6 next hop we are forwarding to (out).
 800  *      args->f_id      Addresses grabbed from the packet (out)
 801  *      args->rule.info a cookie depending on rule action
 802  *
 803  * Return value:
 804  *
 805  *      IP_FW_PASS      the packet must be accepted
 806  *      IP_FW_DENY      the packet must be dropped
 807  *      IP_FW_DIVERT    divert packet, port in m_tag
 808  *      IP_FW_TEE       tee packet, port in m_tag
 809  *      IP_FW_DUMMYNET  to dummynet, pipe in args->cookie
 810  *      IP_FW_NETGRAPH  into netgraph, cookie args->cookie
 811  *              args->rule contains the matching rule,
 812  *              args->rule.info has additional information.
 813  *
 814  */
 815 int
 816 ipfw_chk(struct ip_fw_args *args)
 817 {
 818
 819         /*
 820          * Local variables holding state while processing a packet:
 821          *
 822          * IMPORTANT NOTE: to speed up the processing of rules, there
 823          * are some assumption on the values of the variables, which
 824          * are documented here. Should you change them, please check
 825          * the implementation of the various instructions to make sure
 826          * that they still work.
 827          *
 828          * args->eh     The MAC header. It is non-null for a layer2
 829          *      packet, it is NULL for a layer-3 packet.
 830          * **notyet**
 831          * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
 832          *
 833          * m | args->m  Pointer to the mbuf, as received from the caller.
 834          *      It may change if ipfw_chk() does an m_pullup, or if it
 835          *      consumes the packet because it calls send_reject().
 836          *      XXX This has to change, so that ipfw_chk() never modifies
 837          *      or consumes the buffer.
 838          * ip   is the beginning of the ip(4 or 6) header.
 839          *      Calculated by adding the L3offset to the start of data.
 840          *      (Until we start using L3offset, the packet is
 841          *      supposed to start with the ip header).
 842          */
 843         struct mbuf *m = args->m;
 844         struct ip *ip = mtod(m, struct ip *);
 845
 846         /*
 847          * For rules which contain uid/gid or jail constraints, cache
 848          * a copy of the users credentials after the pcb lookup has been
 849          * executed. This will speed up the processing of rules with
 850          * these types of constraints, as well as decrease contention
 851          * on pcb related locks.
 852          */
 853 #ifndef __FreeBSD__
 854         struct bsd_ucred ucred_cache;
 855 #else
 856         struct ucred *ucred_cache = NULL;
 857 #endif
 858         int ucred_lookup = 0;
 859
 860         /*
 861          * oif | args->oif      If NULL, ipfw_chk has been called on the
 862          *      inbound path (ether_input, ip_input).
 863          *      If non-NULL, ipfw_chk has been called on the outbound path
 864          *      (ether_output, ip_output).
 865          */
 866         struct ifnet *oif = args->oif;
 867
 868         int f_pos = 0;          /* index of current rule in the array */
 869         int retval = 0;
 870
 871         /*
 872          * hlen The length of the IP header.
 873          */
 874         u_int hlen = 0;         /* hlen >0 means we have an IP pkt */
 875
 876         /*
 877          * offset       The offset of a fragment. offset != 0 means that
 878          *      we have a fragment at this offset of an IPv4 packet.
 879          *      offset == 0 means that (if this is an IPv4 packet)
 880          *      this is the first or only fragment.
 881          *      For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
 882          *      or there is a single packet fragement (fragement header added
 883          *      without needed).  We will treat a single packet fragment as if
 884          *      there was no fragment header (or log/block depending on the
 885          *      V_fw_permit_single_frag6 sysctl setting).
 886          */
 887         u_short offset = 0;
 888         u_short ip6f_mf = 0;
 889
 890         /*
 891          * Local copies of addresses. They are only valid if we have
 892          * an IP packet.
 893          *
 894          * proto        The protocol. Set to 0 for non-ip packets,
 895          *      or to the protocol read from the packet otherwise.
 896          *      proto != 0 means that we have an IPv4 packet.
 897          *
 898          * src_port, dst_port   port numbers, in HOST format. Only
 899          *      valid for TCP and UDP packets.
 900          *
 901          * src_ip, dst_ip       ip addresses, in NETWORK format.
 902          *      Only valid for IPv4 packets.
 903          */
 904         uint8_t proto;
 905         uint16_t src_port = 0, dst_port = 0;    /* NOTE: host format    */
 906         struct in_addr src_ip, dst_ip;          /* NOTE: network format */
 907         uint16_t iplen=0;
 908         int pktlen;
 909         uint16_t        etype = 0;      /* Host order stored ether type */
 910
 911         /*
 912          * dyn_dir = MATCH_UNKNOWN when rules unchecked,
 913          *      MATCH_NONE when checked and not matched (q = NULL),
 914          *      MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
 915          */
 916         int dyn_dir = MATCH_UNKNOWN;
 917         ipfw_dyn_rule *q = NULL;
 918         struct ip_fw_chain *chain = &V_layer3_chain;
 919
 920         /*
 921          * We store in ulp a pointer to the upper layer protocol header.
 922          * In the ipv4 case this is easy to determine from the header,
 923          * but for ipv6 we might have some additional headers in the middle.
 924          * ulp is NULL if not found.
 925          */
 926         void *ulp = NULL;               /* upper layer protocol pointer. */
 927
 928         /* XXX ipv6 variables */
 929         int is_ipv6 = 0;
 930         uint8_t icmp6_type = 0;
 931         uint16_t ext_hd = 0;    /* bits vector for extension header filtering */
 932         /* end of ipv6 variables */
 933
 934         int is_ipv4 = 0;
 935
 936         int done = 0;           /* flag to exit the outer loop */
 937
 938         if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
 939                 return (IP_FW_PASS);    /* accept */
 940
 941         dst_ip.s_addr = 0;              /* make sure it is initialized */
 942         src_ip.s_addr = 0;              /* make sure it is initialized */
 943         pktlen = m->m_pkthdr.len;
 944         args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
 945         proto = args->f_id.proto = 0;   /* mark f_id invalid */
 946                 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
 947
 948 /*
 949  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
 950  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
 951  * pointer might become stale after other pullups (but we never use it
 952  * this way).
 953  */
 954 #define PULLUP_TO(_len, p, T)   PULLUP_LEN(_len, p, sizeof(T))
 955 #define PULLUP_LEN(_len, p, T)                                  \
 956 do {                                                            \
 957         int x = (_len) + T;                                     \
 958         if ((m)->m_len < x) {                                   \
 959                 args->m = m = m_pullup(m, x);                   \
 960                 if (m == NULL)                                  \
 961                         goto pullup_failed;                     \
 962         }                                                       \
 963         p = (mtod(m, char *) + (_len));                         \
 964 } while (0)
 965
 966         /*
 967          * if we have an ether header,
 968          */
 969         if (args->eh)
 970                 etype = ntohs(args->eh->ether_type);
 971
 972         /* Identify IP packets and fill up variables. */
 973         if (pktlen >= sizeof(struct ip6_hdr) &&
 974             (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
 975                 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
 976                 is_ipv6 = 1;
 977                 args->f_id.addr_type = 6;
 978                 hlen = sizeof(struct ip6_hdr);
 979                 proto = ip6->ip6_nxt;
 980
 981                 /* Search extension headers to find upper layer protocols */
 982                 while (ulp == NULL && offset == 0) {
 983                         switch (proto) {
 984                         case IPPROTO_ICMPV6:
 985                                 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
 986                                 icmp6_type = ICMP6(ulp)->icmp6_type;
 987                                 break;
 988
 989                         case IPPROTO_TCP:
 990                                 PULLUP_TO(hlen, ulp, struct tcphdr);
 991                                 dst_port = TCP(ulp)->th_dport;
 992                                 src_port = TCP(ulp)->th_sport;
 993                                 /* save flags for dynamic rules */
 994                                 args->f_id._flags = TCP(ulp)->th_flags;
 995                                 break;
 996
 997                         case IPPROTO_SCTP:
 998                                 PULLUP_TO(hlen, ulp, struct sctphdr);
 999                                 src_port = SCTP(ulp)->src_port;
1000                                 dst_port = SCTP(ulp)->dest_port;
1001                                 break;
1002
1003                         case IPPROTO_UDP:
1004                                 PULLUP_TO(hlen, ulp, struct udphdr);
1005                                 dst_port = UDP(ulp)->uh_dport;
1006                                 src_port = UDP(ulp)->uh_sport;
1007                                 break;
1008
1009                         case IPPROTO_HOPOPTS:   /* RFC 2460 */
1010                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1011                                 ext_hd |= EXT_HOPOPTS;
1012                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1013                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1014                                 ulp = NULL;
1015                                 break;
1016
1017                         case IPPROTO_ROUTING:   /* RFC 2460 */
1018                                 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1019                                 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1020                                 case 0:
1021                                         ext_hd |= EXT_RTHDR0;
1022                                         break;
1023                                 case 2:
1024                                         ext_hd |= EXT_RTHDR2;
1025                                         break;
1026                                 default:
1027                                         if (V_fw_verbose)
1028                                                 printf("IPFW2: IPV6 - Unknown "
1029                                                     "Routing Header type(%d)\n",
1030                                                     ((struct ip6_rthdr *)
1031                                                     ulp)->ip6r_type);
1032                                         if (V_fw_deny_unknown_exthdrs)
1033                                             return (IP_FW_DENY);
1034                                         break;
1035                                 }
1036                                 ext_hd |= EXT_ROUTING;
1037                                 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1038                                 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1039                                 ulp = NULL;
1040                                 break;
1041
1042                         case IPPROTO_FRAGMENT:  /* RFC 2460 */
1043                                 PULLUP_TO(hlen, ulp, struct ip6_frag);
1044                                 ext_hd |= EXT_FRAGMENT;
1045                                 hlen += sizeof (struct ip6_frag);
1046                                 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1047                                 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1048                                         IP6F_OFF_MASK;
1049                                 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1050                                         IP6F_MORE_FRAG;
1051                                 if (V_fw_permit_single_frag6 == 0 &&
1052                                     offset == 0 && ip6f_mf == 0) {
1053                                         if (V_fw_verbose)
1054                                                 printf("IPFW2: IPV6 - Invalid "
1055                                                     "Fragment Header\n");
1056                                         if (V_fw_deny_unknown_exthdrs)
1057                                             return (IP_FW_DENY);
1058                                         break;
1059                                 }
1060                                 args->f_id.extra =
1061                                     ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1062                                 ulp = NULL;
1063                                 break;
1064
1065                         case IPPROTO_DSTOPTS:   /* RFC 2460 */
1066                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1067                                 ext_hd |= EXT_DSTOPTS;
1068                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1069                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1070                                 ulp = NULL;
1071                                 break;
1072
1073                         case IPPROTO_AH:        /* RFC 2402 */
1074                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1075                                 ext_hd |= EXT_AH;
1076                                 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1077                                 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1078                                 ulp = NULL;
1079                                 break;
1080
1081                         case IPPROTO_ESP:       /* RFC 2406 */
1082                                 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1083                                 /* Anything past Seq# is variable length and
1084                                  * data past this ext. header is encrypted. */
1085                                 ext_hd |= EXT_ESP;
1086                                 break;
1087
1088                         case IPPROTO_NONE:      /* RFC 2460 */
1089                                 /*
1090                                  * Packet ends here, and IPv6 header has
1091                                  * already been pulled up. If ip6e_len!=0
1092                                  * then octets must be ignored.
1093                                  */
1094                                 ulp = ip; /* non-NULL to get out of loop. */
1095                                 break;
1096
1097                         case IPPROTO_OSPFIGP:
1098                                 /* XXX OSPF header check? */
1099                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1100                                 break;
1101
1102                         case IPPROTO_PIM:
1103                                 /* XXX PIM header check? */
1104                                 PULLUP_TO(hlen, ulp, struct pim);
1105                                 break;
1106
1107                         case IPPROTO_CARP:
1108                                 PULLUP_TO(hlen, ulp, struct carp_header);
1109                                 if (((struct carp_header *)ulp)->carp_version !=
1110                                     CARP_VERSION)
1111                                         return (IP_FW_DENY);
1112                                 if (((struct carp_header *)ulp)->carp_type !=
1113                                     CARP_ADVERTISEMENT)
1114                                         return (IP_FW_DENY);
1115                                 break;
1116
1117                         case IPPROTO_IPV6:      /* RFC 2893 */
1118                                 PULLUP_TO(hlen, ulp, struct ip6_hdr);
1119                                 break;
1120
1121                         case IPPROTO_IPV4:      /* RFC 2893 */
1122                                 PULLUP_TO(hlen, ulp, struct ip);
1123                                 break;
1124
1125                         default:
1126                                 if (V_fw_verbose)
1127                                         printf("IPFW2: IPV6 - Unknown "
1128                                             "Extension Header(%d), ext_hd=%x\n",
1129                                              proto, ext_hd);
1130                                 if (V_fw_deny_unknown_exthdrs)
1131                                     return (IP_FW_DENY);
1132                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1133                                 break;
1134                         } /*switch */
1135                 }
1136                 ip = mtod(m, struct ip *);
1137                 ip6 = (struct ip6_hdr *)ip;
1138                 args->f_id.src_ip6 = ip6->ip6_src;
1139                 args->f_id.dst_ip6 = ip6->ip6_dst;
1140                 args->f_id.src_ip = 0;
1141                 args->f_id.dst_ip = 0;
1142                 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1143         } else if (pktlen >= sizeof(struct ip) &&
1144             (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1145                 is_ipv4 = 1;
1146                 hlen = ip->ip_hl << 2;
1147                 args->f_id.addr_type = 4;
1148
1149                 /*
1150                  * Collect parameters into local variables for faster matching.
1151                  */
1152                 proto = ip->ip_p;
1153                 src_ip = ip->ip_src;
1154                 dst_ip = ip->ip_dst;
1155                 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1156                 iplen = ntohs(ip->ip_len);
1157                 pktlen = iplen < pktlen ? iplen : pktlen;
1158
1159                 if (offset == 0) {
1160                         switch (proto) {
1161                         case IPPROTO_TCP:
1162                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1163                                 dst_port = TCP(ulp)->th_dport;
1164                                 src_port = TCP(ulp)->th_sport;
1165                                 /* save flags for dynamic rules */
1166                                 args->f_id._flags = TCP(ulp)->th_flags;
1167                                 break;
1168
1169                         case IPPROTO_SCTP:
1170                                 PULLUP_TO(hlen, ulp, struct sctphdr);
1171                                 src_port = SCTP(ulp)->src_port;
1172                                 dst_port = SCTP(ulp)->dest_port;
1173                                 break;
1174
1175                         case IPPROTO_UDP:
1176                                 PULLUP_TO(hlen, ulp, struct udphdr);
1177                                 dst_port = UDP(ulp)->uh_dport;
1178                                 src_port = UDP(ulp)->uh_sport;
1179                                 break;
1180
1181                         case IPPROTO_ICMP:
1182                                 PULLUP_TO(hlen, ulp, struct icmphdr);
1183                                 //args->f_id.flags = ICMP(ulp)->icmp_type;
1184                                 break;
1185
1186                         default:
1187                                 break;
1188                         }
1189                 }
1190
1191                 ip = mtod(m, struct ip *);
1192                 args->f_id.src_ip = ntohl(src_ip.s_addr);
1193                 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1194         }
1195 #undef PULLUP_TO
1196         if (proto) { /* we may have port numbers, store them */
1197                 args->f_id.proto = proto;
1198                 args->f_id.src_port = src_port = ntohs(src_port);
1199                 args->f_id.dst_port = dst_port = ntohs(dst_port);
1200         }
1201
1202         IPFW_RLOCK(chain);
1203         if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1204                 IPFW_RUNLOCK(chain);
1205                 return (IP_FW_PASS);    /* accept */
1206         }
1207         if (args->rule.slot) {
1208                 /*
1209                  * Packet has already been tagged as a result of a previous
1210                  * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1211                  * REASS, NETGRAPH, DIVERT/TEE...)
1212                  * Validate the slot and continue from the next one
1213                  * if still present, otherwise do a lookup.
1214                  */
1215                 f_pos = (args->rule.chain_id == chain->id) ?
1216                     args->rule.slot :
1217                     ipfw_find_rule(chain, args->rule.rulenum,
1218                         args->rule.rule_id);
1219         } else {
1220                 f_pos = 0;
1221         }
1222
1223         /*
1224          * Now scan the rules, and parse microinstructions for each rule.
1225          * We have two nested loops and an inner switch. Sometimes we
1226          * need to break out of one or both loops, or re-enter one of
1227          * the loops with updated variables. Loop variables are:
1228          *
1229          *      f_pos (outer loop) points to the current rule.
1230          *              On output it points to the matching rule.
1231          *      done (outer loop) is used as a flag to break the loop.
1232          *      l (inner loop)  residual length of current rule.
1233          *              cmd points to the current microinstruction.
1234          *
1235          * We break the inner loop by setting l=0 and possibly
1236          * cmdlen=0 if we don't want to advance cmd.
1237          * We break the outer loop by setting done=1
1238          * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1239          * as needed.
1240          */
1241         for (; f_pos < chain->n_rules; f_pos++) {
1242                 ipfw_insn *cmd;
1243                 uint32_t tablearg = 0;
1244                 int l, cmdlen, skip_or; /* skip rest of OR block */
1245                 struct ip_fw *f;
1246
1247                 f = chain->map[f_pos];
1248                 if (V_set_disable & (1 << f->set) )
1249                         continue;
1250
1251                 skip_or = 0;
1252                 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1253                     l -= cmdlen, cmd += cmdlen) {
1254                         int match;
1255
1256                         /*
1257                          * check_body is a jump target used when we find a
1258                          * CHECK_STATE, and need to jump to the body of
1259                          * the target rule.
1260                          */
1261
1262 /* check_body: */
1263                         cmdlen = F_LEN(cmd);
1264                         /*
1265                          * An OR block (insn_1 || .. || insn_n) has the
1266                          * F_OR bit set in all but the last instruction.
1267                          * The first match will set "skip_or", and cause
1268                          * the following instructions to be skipped until
1269                          * past the one with the F_OR bit clear.
1270                          */
1271                         if (skip_or) {          /* skip this instruction */
1272                                 if ((cmd->len & F_OR) == 0)
1273                                         skip_or = 0;    /* next one is good */
1274                                 continue;
1275                         }
1276                         match = 0; /* set to 1 if we succeed */
1277
1278                         switch (cmd->opcode) {
1279                         /*
1280                          * The first set of opcodes compares the packet's
1281                          * fields with some pattern, setting 'match' if a
1282                          * match is found. At the end of the loop there is
1283                          * logic to deal with F_NOT and F_OR flags associated
1284                          * with the opcode.
1285                          */
1286                         case O_NOP:
1287                                 match = 1;
1288                                 break;
1289
1290                         case O_FORWARD_MAC:
1291                                 printf("ipfw: opcode %d unimplemented\n",
1292                                     cmd->opcode);
1293                                 break;
1294
1295                         case O_GID:
1296                         case O_UID:
1297                         case O_JAIL:
1298                                 /*
1299                                  * We only check offset == 0 && proto != 0,
1300                                  * as this ensures that we have a
1301                                  * packet with the ports info.
1302                                  */
1303                                 if (offset != 0)
1304                                         break;
1305                                 if (proto == IPPROTO_TCP ||
1306                                     proto == IPPROTO_UDP)
1307                                         match = check_uidgid(
1308                                                     (ipfw_insn_u32 *)cmd,
1309                                                     args, &ucred_lookup,
1310 #ifdef __FreeBSD__
1311                                                     &ucred_cache);
1312 #else
1313                                                     (void *)&ucred_cache);
1314 #endif
1315                                 break;
1316
1317                         case O_RECV:
1318                                 match = iface_match(m->m_pkthdr.rcvif,
1319                                     (ipfw_insn_if *)cmd);
1320                                 break;
1321
1322                         case O_XMIT:
1323                                 match = iface_match(oif, (ipfw_insn_if *)cmd);
1324                                 break;
1325
1326                         case O_VIA:
1327                                 match = iface_match(oif ? oif :
1328                                     m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd);
1329                                 break;
1330
1331                         case O_MACADDR2:
1332                                 if (args->eh != NULL) { /* have MAC header */
1333                                         u_int32_t *want = (u_int32_t *)
1334                                                 ((ipfw_insn_mac *)cmd)->addr;
1335                                         u_int32_t *mask = (u_int32_t *)
1336                                                 ((ipfw_insn_mac *)cmd)->mask;
1337                                         u_int32_t *hdr = (u_int32_t *)args->eh;
1338
1339                                         match =
1340                                             ( want[0] == (hdr[0] & mask[0]) &&
1341                                               want[1] == (hdr[1] & mask[1]) &&
1342                                               want[2] == (hdr[2] & mask[2]) );
1343                                 }
1344                                 break;
1345
1346                         case O_MAC_TYPE:
1347                                 if (args->eh != NULL) {
1348                                         u_int16_t *p =
1349                                             ((ipfw_insn_u16 *)cmd)->ports;
1350                                         int i;
1351
1352                                         for (i = cmdlen - 1; !match && i>0;
1353                                             i--, p += 2)
1354                                                 match = (etype >= p[0] &&
1355                                                     etype <= p[1]);
1356                                 }
1357                                 break;
1358
1359                         case O_FRAG:
1360                                 match = (offset != 0);
1361                                 break;
1362
1363                         case O_IN:      /* "out" is "not in" */
1364                                 match = (oif == NULL);
1365                                 break;
1366
1367                         case O_LAYER2:
1368                                 match = (args->eh != NULL);
1369                                 break;
1370
1371                         case O_DIVERTED:
1372                             {
1373                                 /* For diverted packets, args->rule.info
1374                                  * contains the divert port (in host format)
1375                                  * reason and direction.
1376                                  */
1377                                 uint32_t i = args->rule.info;
1378                                 match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1379                                     cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1380                             }
1381                                 break;
1382
1383                         case O_PROTO:
1384                                 /*
1385                                  * We do not allow an arg of 0 so the
1386                                  * check of "proto" only suffices.
1387                                  */
1388                                 match = (proto == cmd->arg1);
1389                                 break;
1390
1391                         case O_IP_SRC:
1392                                 match = is_ipv4 &&
1393                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1394                                     src_ip.s_addr);
1395                                 break;
1396
1397                         case O_IP_SRC_LOOKUP:
1398                         case O_IP_DST_LOOKUP:
1399                                 if (is_ipv4) {
1400                                     uint32_t key =
1401                                         (cmd->opcode == O_IP_DST_LOOKUP) ?
1402                                             dst_ip.s_addr : src_ip.s_addr;
1403                                     uint32_t v = 0;
1404
1405                                     if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1406                                         /* generic lookup. The key must be
1407                                          * in 32bit big-endian format.
1408                                          */
1409                                         v = ((ipfw_insn_u32 *)cmd)->d[1];
1410                                         if (v == 0)
1411                                             key = dst_ip.s_addr;
1412                                         else if (v == 1)
1413                                             key = src_ip.s_addr;
1414                                         else if (v == 6) /* dscp */
1415                                             key = (ip->ip_tos >> 2) & 0x3f;
1416                                         else if (offset != 0)
1417                                             break;
1418                                         else if (proto != IPPROTO_TCP &&
1419                                                 proto != IPPROTO_UDP)
1420                                             break;
1421                                         else if (v == 2)
1422                                             key = htonl(dst_port);
1423                                         else if (v == 3)
1424                                             key = htonl(src_port);
1425                                         else if (v == 4 || v == 5) {
1426                                             check_uidgid(
1427                                                 (ipfw_insn_u32 *)cmd,
1428                                                 args, &ucred_lookup,
1429 #ifdef __FreeBSD__
1430                                                 &ucred_cache);
1431                                             if (v == 4 /* O_UID */)
1432                                                 key = ucred_cache->cr_uid;
1433                                             else if (v == 5 /* O_JAIL */)
1434                                                 key = ucred_cache->cr_prison->pr_id;
1435 #else /* !__FreeBSD__ */
1436                                                 (void *)&ucred_cache);
1437                                             if (v ==4 /* O_UID */)
1438                                                 key = ucred_cache.uid;
1439                                             else if (v == 5 /* O_JAIL */)
1440                                                 key = ucred_cache.xid;
1441 #endif /* !__FreeBSD__ */
1442                                             key = htonl(key);
1443                                         } else
1444                                             break;
1445                                     }
1446                                     match = ipfw_lookup_table(chain,
1447                                         cmd->arg1, key, &v);
1448                                     if (!match)
1449                                         break;
1450                                     if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1451                                         match =
1452                                             ((ipfw_insn_u32 *)cmd)->d[0] == v;
1453                                     else
1454                                         tablearg = v;
1455                                 }
1456                                 break;
1457
1458                         case O_IP_SRC_MASK:
1459                         case O_IP_DST_MASK:
1460                                 if (is_ipv4) {
1461                                     uint32_t a =
1462                                         (cmd->opcode == O_IP_DST_MASK) ?
1463                                             dst_ip.s_addr : src_ip.s_addr;
1464                                     uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
1465                                     int i = cmdlen-1;
1466
1467                                     for (; !match && i>0; i-= 2, p+= 2)
1468                                         match = (p[0] == (a & p[1]));
1469                                 }
1470                                 break;
1471
1472                         case O_IP_SRC_ME:
1473                                 if (is_ipv4) {
1474                                         struct ifnet *tif;
1475
1476                                         INADDR_TO_IFP(src_ip, tif);
1477                                         match = (tif != NULL);
1478                                         break;
1479                                 }
1480 #ifdef INET6
1481                                 /* FALLTHROUGH */
1482                         case O_IP6_SRC_ME:
1483                                 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6);
1484 #endif
1485                                 break;
1486
1487                         case O_IP_DST_SET:
1488                         case O_IP_SRC_SET:
1489                                 if (is_ipv4) {
1490                                         u_int32_t *d = (u_int32_t *)(cmd+1);
1491                                         u_int32_t addr =
1492                                             cmd->opcode == O_IP_DST_SET ?
1493                                                 args->f_id.dst_ip :
1494                                                 args->f_id.src_ip;
1495
1496                                             if (addr < d[0])
1497                                                     break;
1498                                             addr -= d[0]; /* subtract base */
1499                                             match = (addr < cmd->arg1) &&
1500                                                 ( d[ 1 + (addr>>5)] &
1501                                                   (1<<(addr & 0x1f)) );
1502                                 }
1503                                 break;
1504
1505                         case O_IP_DST:
1506                                 match = is_ipv4 &&
1507                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1508                                     dst_ip.s_addr);
1509                                 break;
1510
1511                         case O_IP_DST_ME:
1512                                 if (is_ipv4) {
1513                                         struct ifnet *tif;
1514
1515                                         INADDR_TO_IFP(dst_ip, tif);
1516                                         match = (tif != NULL);
1517                                         break;
1518                                 }
1519 #ifdef INET6
1520                                 /* FALLTHROUGH */
1521                         case O_IP6_DST_ME:
1522                                 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6);
1523 #endif
1524                                 break;
1525
1526
1527                         case O_IP_SRCPORT:
1528                         case O_IP_DSTPORT:
1529                                 /*
1530                                  * offset == 0 && proto != 0 is enough
1531                                  * to guarantee that we have a
1532                                  * packet with port info.
1533                                  */
1534                                 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1535                                     && offset == 0) {
1536                                         u_int16_t x =
1537                                             (cmd->opcode == O_IP_SRCPORT) ?
1538                                                 src_port : dst_port ;
1539                                         u_int16_t *p =
1540                                             ((ipfw_insn_u16 *)cmd)->ports;
1541                                         int i;
1542
1543                                         for (i = cmdlen - 1; !match && i>0;
1544                                             i--, p += 2)
1545                                                 match = (x>=p[0] && x<=p[1]);
1546                                 }
1547                                 break;
1548
1549                         case O_ICMPTYPE:
1550                                 match = (offset == 0 && proto==IPPROTO_ICMP &&
1551                                     icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
1552                                 break;
1553
1554 #ifdef INET6
1555                         case O_ICMP6TYPE:
1556                                 match = is_ipv6 && offset == 0 &&
1557                                     proto==IPPROTO_ICMPV6 &&
1558                                     icmp6type_match(
1559                                         ICMP6(ulp)->icmp6_type,
1560                                         (ipfw_insn_u32 *)cmd);
1561                                 break;
1562 #endif /* INET6 */
1563
1564                         case O_IPOPT:
1565                                 match = (is_ipv4 &&
1566                                     ipopts_match(ip, cmd) );
1567                                 break;
1568
1569                         case O_IPVER:
1570                                 match = (is_ipv4 &&
1571                                     cmd->arg1 == ip->ip_v);
1572                                 break;
1573
1574                         case O_IPID:
1575                         case O_IPLEN:
1576                         case O_IPTTL:
1577                                 if (is_ipv4) {  /* only for IP packets */
1578                                     uint16_t x;
1579                                     uint16_t *p;
1580                                     int i;
1581
1582                                     if (cmd->opcode == O_IPLEN)
1583                                         x = iplen;
1584                                     else if (cmd->opcode == O_IPTTL)
1585                                         x = ip->ip_ttl;
1586                                     else /* must be IPID */
1587                                         x = ntohs(ip->ip_id);
1588                                     if (cmdlen == 1) {
1589                                         match = (cmd->arg1 == x);
1590                                         break;
1591                                     }
1592                                     /* otherwise we have ranges */
1593                                     p = ((ipfw_insn_u16 *)cmd)->ports;
1594                                     i = cmdlen - 1;
1595                                     for (; !match && i>0; i--, p += 2)
1596                                         match = (x >= p[0] && x <= p[1]);
1597                                 }
1598                                 break;
1599
1600                         case O_IPPRECEDENCE:
1601                                 match = (is_ipv4 &&
1602                                     (cmd->arg1 == (ip->ip_tos & 0xe0)) );
1603                                 break;
1604
1605                         case O_IPTOS:
1606                                 match = (is_ipv4 &&
1607                                     flags_match(cmd, ip->ip_tos));
1608                                 break;
1609
1610                         case O_TCPDATALEN:
1611                                 if (proto == IPPROTO_TCP && offset == 0) {
1612                                     struct tcphdr *tcp;
1613                                     uint16_t x;
1614                                     uint16_t *p;
1615                                     int i;
1616
1617                                     tcp = TCP(ulp);
1618                                     x = iplen -
1619                                         ((ip->ip_hl + tcp->th_off) << 2);
1620                                     if (cmdlen == 1) {
1621                                         match = (cmd->arg1 == x);
1622                                         break;
1623                                     }
1624                                     /* otherwise we have ranges */
1625                                     p = ((ipfw_insn_u16 *)cmd)->ports;
1626                                     i = cmdlen - 1;
1627                                     for (; !match && i>0; i--, p += 2)
1628                                         match = (x >= p[0] && x <= p[1]);
1629                                 }
1630                                 break;
1631
1632                         case O_TCPFLAGS:
1633                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1634                                     flags_match(cmd, TCP(ulp)->th_flags));
1635                                 break;
1636
1637                         case O_TCPOPTS:
1638                                 PULLUP_LEN(hlen, ulp, (TCP(ulp)->th_off << 2));
1639                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1640                                     tcpopts_match(TCP(ulp), cmd));
1641                                 break;
1642
1643                         case O_TCPSEQ:
1644                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1645                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
1646                                         TCP(ulp)->th_seq);
1647                                 break;
1648
1649                         case O_TCPACK:
1650                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1651                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
1652                                         TCP(ulp)->th_ack);
1653                                 break;
1654
1655                         case O_TCPWIN:
1656                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1657                                     cmd->arg1 == TCP(ulp)->th_win);
1658                                 break;
1659
1660                         case O_ESTAB:
1661                                 /* reject packets which have SYN only */
1662                                 /* XXX should i also check for TH_ACK ? */
1663                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1664                                     (TCP(ulp)->th_flags &
1665                                      (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1666                                 break;
1667
1668                         case O_ALTQ: {
1669                                 struct pf_mtag *at;
1670                                 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
1671
1672                                 match = 1;
1673                                 at = pf_find_mtag(m);
1674                                 if (at != NULL && at->qid != 0)
1675                                         break;
1676                                 at = pf_get_mtag(m);
1677                                 if (at == NULL) {
1678                                         /*
1679                                          * Let the packet fall back to the
1680                                          * default ALTQ.
1681                                          */
1682                                         break;
1683                                 }
1684                                 at->qid = altq->qid;
1685                                 at->hdr = ip;
1686                                 break;
1687                         }
1688
1689                         case O_LOG:
1690                                 ipfw_log(f, hlen, args, m,
1691                                     oif, offset | ip6f_mf, tablearg, ip);
1692                                 match = 1;
1693                                 break;
1694
1695                         case O_PROB:
1696                                 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
1697                                 break;
1698
1699                         case O_VERREVPATH:
1700                                 /* Outgoing packets automatically pass/match */
1701                                 match = ((oif != NULL) ||
1702                                     (m->m_pkthdr.rcvif == NULL) ||
1703                                     (
1704 #ifdef INET6
1705                                     is_ipv6 ?
1706                                         verify_path6(&(args->f_id.src_ip6),
1707                                             m->m_pkthdr.rcvif) :
1708 #endif
1709                                     verify_path(src_ip, m->m_pkthdr.rcvif,
1710                                         args->f_id.fib)));
1711                                 break;
1712
1713                         case O_VERSRCREACH:
1714                                 /* Outgoing packets automatically pass/match */
1715                                 match = (hlen > 0 && ((oif != NULL) ||
1716 #ifdef INET6
1717                                     is_ipv6 ?
1718                                         verify_path6(&(args->f_id.src_ip6),
1719                                             NULL) :
1720 #endif
1721                                     verify_path(src_ip, NULL, args->f_id.fib)));
1722                                 break;
1723
1724                         case O_ANTISPOOF:
1725                                 /* Outgoing packets automatically pass/match */
1726                                 if (oif == NULL && hlen > 0 &&
1727                                     (  (is_ipv4 && in_localaddr(src_ip))
1728 #ifdef INET6
1729                                     || (is_ipv6 &&
1730                                         in6_localaddr(&(args->f_id.src_ip6)))
1731 #endif
1732                                     ))
1733                                         match =
1734 #ifdef INET6
1735                                             is_ipv6 ? verify_path6(
1736                                                 &(args->f_id.src_ip6),
1737                                                 m->m_pkthdr.rcvif) :
1738 #endif
1739                                             verify_path(src_ip,
1740                                                 m->m_pkthdr.rcvif,
1741                                                 args->f_id.fib);
1742                                 else
1743                                         match = 1;
1744                                 break;
1745
1746                         case O_IPSEC:
1747 #ifdef IPSEC
1748                                 match = (m_tag_find(m,
1749                                     PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
1750 #endif
1751                                 /* otherwise no match */
1752                                 break;
1753
1754 #ifdef INET6
1755                         case O_IP6_SRC:
1756                                 match = is_ipv6 &&
1757                                     IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
1758                                     &((ipfw_insn_ip6 *)cmd)->addr6);
1759                                 break;
1760
1761                         case O_IP6_DST:
1762                                 match = is_ipv6 &&
1763                                 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
1764                                     &((ipfw_insn_ip6 *)cmd)->addr6);
1765                                 break;
1766                         case O_IP6_SRC_MASK:
1767                         case O_IP6_DST_MASK:
1768                                 if (is_ipv6) {
1769                                         int i = cmdlen - 1;
1770                                         struct in6_addr p;
1771                                         struct in6_addr *d =
1772                                             &((ipfw_insn_ip6 *)cmd)->addr6;
1773
1774                                         for (; !match && i > 0; d += 2,
1775                                             i -= F_INSN_SIZE(struct in6_addr)
1776                                             * 2) {
1777                                                 p = (cmd->opcode ==
1778                                                     O_IP6_SRC_MASK) ?
1779                                                     args->f_id.src_ip6:
1780                                                     args->f_id.dst_ip6;
1781                                                 APPLY_MASK(&p, &d[1]);
1782                                                 match =
1783                                                     IN6_ARE_ADDR_EQUAL(&d[0],
1784                                                     &p);
1785                                         }
1786                                 }
1787                                 break;
1788
1789                         case O_FLOW6ID:
1790                                 match = is_ipv6 &&
1791                                     flow6id_match(args->f_id.flow_id6,
1792                                     (ipfw_insn_u32 *) cmd);
1793                                 break;
1794
1795                         case O_EXT_HDR:
1796                                 match = is_ipv6 &&
1797                                     (ext_hd & ((ipfw_insn *) cmd)->arg1);
1798                                 break;
1799
1800                         case O_IP6:
1801                                 match = is_ipv6;
1802                                 break;
1803 #endif
1804
1805                         case O_IP4:
1806                                 match = is_ipv4;
1807                                 break;
1808
1809                         case O_TAG: {
1810                                 struct m_tag *mtag;
1811                                 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
1812                                     tablearg : cmd->arg1;
1813
1814                                 /* Packet is already tagged with this tag? */
1815                                 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
1816
1817                                 /* We have `untag' action when F_NOT flag is
1818                                  * present. And we must remove this mtag from
1819                                  * mbuf and reset `match' to zero (`match' will
1820                                  * be inversed later).
1821                                  * Otherwise we should allocate new mtag and
1822                                  * push it into mbuf.
1823                                  */
1824                                 if (cmd->len & F_NOT) { /* `untag' action */
1825                                         if (mtag != NULL)
1826                                                 m_tag_delete(m, mtag);
1827                                         match = 0;
1828                                 } else {
1829                                         if (mtag == NULL) {
1830                                                 mtag = m_tag_alloc( MTAG_IPFW,
1831                                                     tag, 0, M_NOWAIT);
1832                                                 if (mtag != NULL)
1833                                                         m_tag_prepend(m, mtag);
1834                                         }
1835                                         match = 1;
1836                                 }
1837                                 break;
1838                         }
1839
1840                         case O_FIB: /* try match the specified fib */
1841                                 if (args->f_id.fib == cmd->arg1)
1842                                         match = 1;
1843                                 break;
1844
1845                         case O_SOCKARG: {
1846                                 struct inpcb *inp = args->inp;
1847                                 struct inpcbinfo *pi;
1848
1849                                 if (is_ipv6) /* XXX can we remove this ? */
1850                                         break;
1851
1852                                 if (proto == IPPROTO_TCP)
1853                                         pi = &V_tcbinfo;
1854                                 else if (proto == IPPROTO_UDP)
1855                                         pi = &V_udbinfo;
1856                                 else
1857                                         break;
1858
1859                                 /*
1860                                  * XXXRW: so_user_cookie should almost
1861                                  * certainly be inp_user_cookie?
1862                                  */
1863
1864                                 /* For incomming packet, lookup up the
1865                                 inpcb using the src/dest ip/port tuple */
1866                                 if (inp == NULL) {
1867                                         inp = in_pcblookup(pi,
1868                                                 src_ip, htons(src_port),
1869                                                 dst_ip, htons(dst_port),
1870                                                 INPLOOKUP_RLOCKPCB, NULL);
1871                                         if (inp != NULL) {
1872                                                 tablearg =
1873                                                     inp->inp_socket->so_user_cookie;
1874                                                 if (tablearg)
1875                                                         match = 1;
1876                                                 INP_RUNLOCK(inp);
1877                                         }
1878                                 } else {
1879                                         if (inp->inp_socket) {
1880                                                 tablearg =
1881                                                     inp->inp_socket->so_user_cookie;
1882                                                 if (tablearg)
1883                                                         match = 1;
1884                                         }
1885                                 }
1886                                 break;
1887                         }
1888
1889                         case O_TAGGED: {
1890                                 struct m_tag *mtag;
1891                                 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
1892                                     tablearg : cmd->arg1;
1893
1894                                 if (cmdlen == 1) {
1895                                         match = m_tag_locate(m, MTAG_IPFW,
1896                                             tag, NULL) != NULL;
1897                                         break;
1898                                 }
1899
1900                                 /* we have ranges */
1901                                 for (mtag = m_tag_first(m);
1902                                     mtag != NULL && !match;
1903                                     mtag = m_tag_next(m, mtag)) {
1904                                         uint16_t *p;
1905                                         int i;
1906
1907                                         if (mtag->m_tag_cookie != MTAG_IPFW)
1908                                                 continue;
1909
1910                                         p = ((ipfw_insn_u16 *)cmd)->ports;
1911                                         i = cmdlen - 1;
1912                                         for(; !match && i > 0; i--, p += 2)
1913                                                 match =
1914                                                     mtag->m_tag_id >= p[0] &&
1915                                                     mtag->m_tag_id <= p[1];
1916                                 }
1917                                 break;
1918                         }
1919
1920                         /*
1921                          * The second set of opcodes represents 'actions',
1922                          * i.e. the terminal part of a rule once the packet
1923                          * matches all previous patterns.
1924                          * Typically there is only one action for each rule,
1925                          * and the opcode is stored at the end of the rule
1926                          * (but there are exceptions -- see below).
1927                          *
1928                          * In general, here we set retval and terminate the
1929                          * outer loop (would be a 'break 3' in some language,
1930                          * but we need to set l=0, done=1)
1931                          *
1932                          * Exceptions:
1933                          * O_COUNT and O_SKIPTO actions:
1934                          *   instead of terminating, we jump to the next rule
1935                          *   (setting l=0), or to the SKIPTO target (setting
1936                          *   f/f_len, cmd and l as needed), respectively.
1937                          *
1938                          * O_TAG, O_LOG and O_ALTQ action parameters:
1939                          *   perform some action and set match = 1;
1940                          *
1941                          * O_LIMIT and O_KEEP_STATE: these opcodes are
1942                          *   not real 'actions', and are stored right
1943                          *   before the 'action' part of the rule.
1944                          *   These opcodes try to install an entry in the
1945                          *   state tables; if successful, we continue with
1946                          *   the next opcode (match=1; break;), otherwise
1947                          *   the packet must be dropped (set retval,
1948                          *   break loops with l=0, done=1)
1949                          *
1950                          * O_PROBE_STATE and O_CHECK_STATE: these opcodes
1951                          *   cause a lookup of the state table, and a jump
1952                          *   to the 'action' part of the parent rule
1953                          *   if an entry is found, or
1954                          *   (CHECK_STATE only) a jump to the next rule if
1955                          *   the entry is not found.
1956                          *   The result of the lookup is cached so that
1957                          *   further instances of these opcodes become NOPs.
1958                          *   The jump to the next rule is done by setting
1959                          *   l=0, cmdlen=0.
1960                          */
1961                         case O_LIMIT:
1962                         case O_KEEP_STATE:
1963                                 if (ipfw_install_state(f,
1964                                     (ipfw_insn_limit *)cmd, args, tablearg)) {
1965                                         /* error or limit violation */
1966                                         retval = IP_FW_DENY;
1967                                         l = 0;  /* exit inner loop */
1968                                         done = 1; /* exit outer loop */
1969                                 }
1970                                 match = 1;
1971                                 break;
1972
1973                         case O_PROBE_STATE:
1974                         case O_CHECK_STATE:
1975                                 /*
1976                                  * dynamic rules are checked at the first
1977                                  * keep-state or check-state occurrence,
1978                                  * with the result being stored in dyn_dir.
1979                                  * The compiler introduces a PROBE_STATE
1980                                  * instruction for us when we have a
1981                                  * KEEP_STATE (because PROBE_STATE needs
1982                                  * to be run first).
1983                                  */
1984                                 if (dyn_dir == MATCH_UNKNOWN &&
1985                                     (q = ipfw_lookup_dyn_rule(&args->f_id,
1986                                      &dyn_dir, proto == IPPROTO_TCP ?
1987                                         TCP(ulp) : NULL))
1988                                         != NULL) {
1989                                         /*
1990                                          * Found dynamic entry, update stats
1991                                          * and jump to the 'action' part of
1992                                          * the parent rule by setting
1993                                          * f, cmd, l and clearing cmdlen.
1994                                          */
1995                                         q->pcnt++;
1996                                         q->bcnt += pktlen;
1997                                         /* XXX we would like to have f_pos
1998                                          * readily accessible in the dynamic
1999                                          * rule, instead of having to
2000                                          * lookup q->rule.
2001                                          */
2002                                         f = q->rule;
2003                                         f_pos = ipfw_find_rule(chain,
2004                                                 f->rulenum, f->id);
2005                                         cmd = ACTION_PTR(f);
2006                                         l = f->cmd_len - f->act_ofs;
2007                                         ipfw_dyn_unlock();
2008                                         cmdlen = 0;
2009                                         match = 1;
2010                                         break;
2011                                 }
2012                                 /*
2013                                  * Dynamic entry not found. If CHECK_STATE,
2014                                  * skip to next rule, if PROBE_STATE just
2015                                  * ignore and continue with next opcode.
2016                                  */
2017                                 if (cmd->opcode == O_CHECK_STATE)
2018                                         l = 0;  /* exit inner loop */
2019                                 match = 1;
2020                                 break;
2021
2022                         case O_ACCEPT:
2023                                 retval = 0;     /* accept */
2024                                 l = 0;          /* exit inner loop */
2025                                 done = 1;       /* exit outer loop */
2026                                 break;
2027
2028                         case O_PIPE:
2029                         case O_QUEUE:
2030                                 set_match(args, f_pos, chain);
2031                                 args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ?
2032                                         tablearg : cmd->arg1;
2033                                 if (cmd->opcode == O_PIPE)
2034                                         args->rule.info |= IPFW_IS_PIPE;
2035                                 if (V_fw_one_pass)
2036                                         args->rule.info |= IPFW_ONEPASS;
2037                                 retval = IP_FW_DUMMYNET;
2038                                 l = 0;          /* exit inner loop */
2039                                 done = 1;       /* exit outer loop */
2040                                 break;
2041
2042                         case O_DIVERT:
2043                         case O_TEE:
2044                                 if (args->eh) /* not on layer 2 */
2045                                     break;
2046                                 /* otherwise this is terminal */
2047                                 l = 0;          /* exit inner loop */
2048                                 done = 1;       /* exit outer loop */
2049                                 retval = (cmd->opcode == O_DIVERT) ?
2050                                         IP_FW_DIVERT : IP_FW_TEE;
2051                                 set_match(args, f_pos, chain);
2052                                 args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ?
2053                                     tablearg : cmd->arg1;
2054                                 break;
2055
2056                         case O_COUNT:
2057                                 f->pcnt++;      /* update stats */
2058                                 f->bcnt += pktlen;
2059                                 f->timestamp = time_uptime;
2060                                 l = 0;          /* exit inner loop */
2061                                 break;
2062
2063                         case O_SKIPTO:
2064                             f->pcnt++;  /* update stats */
2065                             f->bcnt += pktlen;
2066                             f->timestamp = time_uptime;
2067                             /* If possible use cached f_pos (in f->next_rule),
2068                              * whose version is written in f->next_rule
2069                              * (horrible hacks to avoid changing the ABI).
2070                              */
2071                             if (cmd->arg1 != IP_FW_TABLEARG &&
2072                                     (uintptr_t)f->x_next == chain->id) {
2073                                 f_pos = (uintptr_t)f->next_rule;
2074                             } else {
2075                                 int i = (cmd->arg1 == IP_FW_TABLEARG) ?
2076                                         tablearg : cmd->arg1;
2077                                 /* make sure we do not jump backward */
2078                                 if (i <= f->rulenum)
2079                                     i = f->rulenum + 1;
2080                                 f_pos = ipfw_find_rule(chain, i, 0);
2081                                 /* update the cache */
2082                                 if (cmd->arg1 != IP_FW_TABLEARG) {
2083                                     f->next_rule =
2084                                         (void *)(uintptr_t)f_pos;
2085                                     f->x_next =
2086                                         (void *)(uintptr_t)chain->id;
2087                                 }
2088                             }
2089                             /*
2090                              * Skip disabled rules, and re-enter
2091                              * the inner loop with the correct
2092                              * f_pos, f, l and cmd.
2093                              * Also clear cmdlen and skip_or
2094                              */
2095                             for (; f_pos < chain->n_rules - 1 &&
2096                                     (V_set_disable &
2097                                      (1 << chain->map[f_pos]->set));
2098                                     f_pos++)
2099                                 ;
2100                             /* Re-enter the inner loop at the skipto rule. */
2101                             f = chain->map[f_pos];
2102                             l = f->cmd_len;
2103                             cmd = f->cmd;
2104                             match = 1;
2105                             cmdlen = 0;
2106                             skip_or = 0;
2107                             continue;
2108                             break;      /* not reached */
2109
2110                         case O_CALLRETURN: {
2111                                 /*
2112                                  * Implementation of `subroutine' call/return,
2113                                  * in the stack carried in an mbuf tag. This
2114                                  * is different from `skipto' in that any call
2115                                  * address is possible (`skipto' must prevent
2116                                  * backward jumps to avoid endless loops).
2117                                  * We have `return' action when F_NOT flag is
2118                                  * present. The `m_tag_id' field is used as
2119                                  * stack pointer.
2120                                  */
2121                                 struct m_tag *mtag;
2122                                 uint16_t jmpto, *stack;
2123
2124 #define IS_CALL         ((cmd->len & F_NOT) == 0)
2125 #define IS_RETURN       ((cmd->len & F_NOT) != 0)
2126                                 /*
2127                                  * Hand-rolled version of m_tag_locate() with
2128                                  * wildcard `type'.
2129                                  * If not already tagged, allocate new tag.
2130                                  */
2131                                 mtag = m_tag_first(m);
2132                                 while (mtag != NULL) {
2133                                         if (mtag->m_tag_cookie ==
2134                                             MTAG_IPFW_CALL)
2135                                                 break;
2136                                         mtag = m_tag_next(m, mtag);
2137                                 }
2138                                 if (mtag == NULL && IS_CALL) {
2139                                         mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2140                                             IPFW_CALLSTACK_SIZE *
2141                                             sizeof(uint16_t), M_NOWAIT);
2142                                         if (mtag != NULL)
2143                                                 m_tag_prepend(m, mtag);
2144                                 }
2145
2146                                 /*
2147                                  * On error both `call' and `return' just
2148                                  * continue with next rule.
2149                                  */
2150                                 if (IS_RETURN && (mtag == NULL ||
2151                                     mtag->m_tag_id == 0)) {
2152                                         l = 0;          /* exit inner loop */
2153                                         break;
2154                                 }
2155                                 if (IS_CALL && (mtag == NULL ||
2156                                     mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2157                                         printf("ipfw: call stack error, "
2158                                             "go to next rule\n");
2159                                         l = 0;          /* exit inner loop */
2160                                         break;
2161                                 }
2162
2163                                 f->pcnt++;      /* update stats */
2164                                 f->bcnt += pktlen;
2165                                 f->timestamp = time_uptime;
2166                                 stack = (uint16_t *)(mtag + 1);
2167
2168                                 /*
2169                                  * The `call' action may use cached f_pos
2170                                  * (in f->next_rule), whose version is written
2171                                  * in f->next_rule.
2172                                  * The `return' action, however, doesn't have
2173                                  * fixed jump address in cmd->arg1 and can't use
2174                                  * cache.
2175                                  */
2176                                 if (IS_CALL) {
2177                                         stack[mtag->m_tag_id] = f->rulenum;
2178                                         mtag->m_tag_id++;
2179                                         if (cmd->arg1 != IP_FW_TABLEARG &&
2180                                             (uintptr_t)f->x_next == chain->id) {
2181                                                 f_pos = (uintptr_t)f->next_rule;
2182                                         } else {
2183                                                 jmpto = (cmd->arg1 ==
2184                                                     IP_FW_TABLEARG) ? tablearg:
2185                                                     cmd->arg1;
2186                                                 f_pos = ipfw_find_rule(chain,
2187                                                     jmpto, 0);
2188                                                 /* update the cache */
2189                                                 if (cmd->arg1 !=
2190                                                     IP_FW_TABLEARG) {
2191                                                         f->next_rule =
2192                                                             (void *)(uintptr_t)
2193                                                             f_pos;
2194                                                         f->x_next =
2195                                                             (void *)(uintptr_t)
2196                                                             chain->id;
2197                                                 }
2198                                         }
2199                                 } else {        /* `return' action */
2200                                         mtag->m_tag_id--;
2201                                         jmpto = stack[mtag->m_tag_id] + 1;
2202                                         f_pos = ipfw_find_rule(chain, jmpto, 0);
2203                                 }
2204
2205                                 /*
2206                                  * Skip disabled rules, and re-enter
2207                                  * the inner loop with the correct
2208                                  * f_pos, f, l and cmd.
2209                                  * Also clear cmdlen and skip_or
2210                                  */
2211                                 for (; f_pos < chain->n_rules - 1 &&
2212                                     (V_set_disable &
2213                                     (1 << chain->map[f_pos]->set)); f_pos++)
2214                                         ;
2215                                 /* Re-enter the inner loop at the dest rule. */
2216                                 f = chain->map[f_pos];
2217                                 l = f->cmd_len;
2218                                 cmd = f->cmd;
2219                                 cmdlen = 0;
2220                                 skip_or = 0;
2221                                 continue;
2222                                 break;  /* NOTREACHED */
2223                         }
2224 #undef IS_CALL
2225 #undef IS_RETURN
2226
2227                         case O_REJECT:
2228                                 /*
2229                                  * Drop the packet and send a reject notice
2230                                  * if the packet is not ICMP (or is an ICMP
2231                                  * query), and it is not multicast/broadcast.
2232                                  */
2233                                 if (hlen > 0 && is_ipv4 && offset == 0 &&
2234                                     (proto != IPPROTO_ICMP ||
2235                                      is_icmp_query(ICMP(ulp))) &&
2236                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2237                                     !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2238                                         send_reject(args, cmd->arg1, iplen, ip);
2239                                         m = args->m;
2240                                 }
2241                                 /* FALLTHROUGH */
2242 #ifdef INET6
2243                         case O_UNREACH6:
2244                                 if (hlen > 0 && is_ipv6 &&
2245                                     ((offset & IP6F_OFF_MASK) == 0) &&
2246                                     (proto != IPPROTO_ICMPV6 ||
2247                                      (is_icmp6_query(icmp6_type) == 1)) &&
2248                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2249                                     !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2250                                         send_reject6(
2251                                             args, cmd->arg1, hlen,
2252                                             (struct ip6_hdr *)ip);
2253                                         m = args->m;
2254                                 }
2255                                 /* FALLTHROUGH */
2256 #endif
2257                         case O_DENY:
2258                                 retval = IP_FW_DENY;
2259                                 l = 0;          /* exit inner loop */
2260                                 done = 1;       /* exit outer loop */
2261                                 break;
2262
2263                         case O_FORWARD_IP:
2264                                 if (args->eh)   /* not valid on layer2 pkts */
2265                                         break;
2266                                 if (q == NULL || q->rule != f ||
2267                                     dyn_dir == MATCH_FORWARD) {
2268                                     struct sockaddr_in *sa;
2269                                     sa = &(((ipfw_insn_sa *)cmd)->sa);
2270                                     if (sa->sin_addr.s_addr == INADDR_ANY) {
2271                                         bcopy(sa, &args->hopstore,
2272                                                         sizeof(*sa));
2273                                         args->hopstore.sin_addr.s_addr =
2274                                                     htonl(tablearg);
2275                                         args->next_hop = &args->hopstore;
2276                                     } else {
2277                                         args->next_hop = sa;
2278                                     }
2279                                 }
2280                                 retval = IP_FW_PASS;
2281                                 l = 0;          /* exit inner loop */
2282                                 done = 1;       /* exit outer loop */
2283                                 break;
2284
2285 #ifdef INET6
2286                         case O_FORWARD_IP6:
2287                                 if (args->eh)   /* not valid on layer2 pkts */
2288                                         break;
2289                                 if (q == NULL || q->rule != f ||
2290                                     dyn_dir == MATCH_FORWARD) {
2291                                         struct sockaddr_in6 *sin6;
2292
2293                                         sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2294                                         args->next_hop6 = sin6;
2295                                 }
2296                                 retval = IP_FW_PASS;
2297                                 l = 0;          /* exit inner loop */
2298                                 done = 1;       /* exit outer loop */
2299                                 break;
2300 #endif
2301
2302                         case O_NETGRAPH:
2303                         case O_NGTEE:
2304                                 set_match(args, f_pos, chain);
2305                                 args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ?
2306                                         tablearg : cmd->arg1;
2307                                 if (V_fw_one_pass)
2308                                         args->rule.info |= IPFW_ONEPASS;
2309                                 retval = (cmd->opcode == O_NETGRAPH) ?
2310                                     IP_FW_NETGRAPH : IP_FW_NGTEE;
2311                                 l = 0;          /* exit inner loop */
2312                                 done = 1;       /* exit outer loop */
2313                                 break;
2314
2315                         case O_SETFIB: {
2316                                 uint32_t fib;
2317
2318                                 f->pcnt++;      /* update stats */
2319                                 f->bcnt += pktlen;
2320                                 f->timestamp = time_uptime;
2321                                 fib = (cmd->arg1 == IP_FW_TABLEARG) ? tablearg:
2322                                     cmd->arg1;
2323                                 if (fib >= rt_numfibs)
2324                                         fib = 0;
2325                                 M_SETFIB(m, fib);
2326                                 args->f_id.fib = fib;
2327                                 l = 0;          /* exit inner loop */
2328                                 break;
2329                         }
2330
2331                         case O_NAT:
2332                                 if (!IPFW_NAT_LOADED) {
2333                                     retval = IP_FW_DENY;
2334                                 } else {
2335                                     struct cfg_nat *t;
2336                                     int nat_id;
2337
2338                                     set_match(args, f_pos, chain);
2339                                     /* Check if this is 'global' nat rule */
2340                                     if (cmd->arg1 == 0) {
2341                                             retval = ipfw_nat_ptr(args, NULL, m);
2342                                             l = 0;
2343                                             done = 1;
2344                                             break;
2345                                     }
2346                                     t = ((ipfw_insn_nat *)cmd)->nat;
2347                                     if (t == NULL) {
2348                                         nat_id = (cmd->arg1 == IP_FW_TABLEARG) ?
2349                                                 tablearg : cmd->arg1;
2350                                         t = (*lookup_nat_ptr)(&chain->nat, nat_id);
2351
2352                                         if (t == NULL) {
2353                                             retval = IP_FW_DENY;
2354                                             l = 0;      /* exit inner loop */
2355                                             done = 1;   /* exit outer loop */
2356                                             break;
2357                                         }
2358                                         if (cmd->arg1 != IP_FW_TABLEARG)
2359                                             ((ipfw_insn_nat *)cmd)->nat = t;
2360                                     }
2361                                     retval = ipfw_nat_ptr(args, t, m);
2362                                 }
2363                                 l = 0;          /* exit inner loop */
2364                                 done = 1;       /* exit outer loop */
2365                                 break;
2366
2367                         case O_REASS: {
2368                                 int ip_off;
2369
2370                                 f->pcnt++;
2371                                 f->bcnt += pktlen;
2372                                 l = 0;  /* in any case exit inner loop */
2373                                 ip_off = ntohs(ip->ip_off);
2374
2375                                 /* if not fragmented, go to next rule */
2376                                 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
2377                                     break;
2378                                 /*
2379                                  * ip_reass() expects len & off in host
2380                                  * byte order.
2381                                  */
2382                                 SET_HOST_IPLEN(ip);
2383
2384                                 args->m = m = ip_reass(m);
2385
2386                                 /*
2387                                  * do IP header checksum fixup.
2388                                  */
2389                                 if (m == NULL) { /* fragment got swallowed */
2390                                     retval = IP_FW_DENY;
2391                                 } else { /* good, packet complete */
2392                                     int hlen;
2393
2394                                     ip = mtod(m, struct ip *);
2395                                     hlen = ip->ip_hl << 2;
2396                                     SET_NET_IPLEN(ip);
2397                                     ip->ip_sum = 0;
2398                                     if (hlen == sizeof(struct ip))
2399                                         ip->ip_sum = in_cksum_hdr(ip);
2400                                     else
2401                                         ip->ip_sum = in_cksum(m, hlen);
2402                                     retval = IP_FW_REASS;
2403                                     set_match(args, f_pos, chain);
2404                                 }
2405                                 done = 1;       /* exit outer loop */
2406                                 break;
2407                         }
2408
2409                         default:
2410                                 panic("-- unknown opcode %d\n", cmd->opcode);
2411                         } /* end of switch() on opcodes */
2412                         /*
2413                          * if we get here with l=0, then match is irrelevant.
2414                          */
2415
2416                         if (cmd->len & F_NOT)
2417                                 match = !match;
2418
2419                         if (match) {
2420                                 if (cmd->len & F_OR)
2421                                         skip_or = 1;
2422                         } else {
2423                                 if (!(cmd->len & F_OR)) /* not an OR block, */
2424                                         break;          /* try next rule    */
2425                         }
2426
2427                 }       /* end of inner loop, scan opcodes */
2428 #undef PULLUP_LEN
2429
2430                 if (done)
2431                         break;
2432
2433 /* next_rule:; */       /* try next rule                */
2434
2435         }               /* end of outer for, scan rules */
2436
2437         if (done) {
2438                 struct ip_fw *rule = chain->map[f_pos];
2439                 /* Update statistics */
2440                 rule->pcnt++;
2441                 rule->bcnt += pktlen;
2442                 rule->timestamp = time_uptime;
2443         } else {
2444                 retval = IP_FW_DENY;
2445                 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2446         }
2447         IPFW_RUNLOCK(chain);
2448 #ifdef __FreeBSD__
2449         if (ucred_cache != NULL)
2450                 crfree(ucred_cache);
2451 #endif
2452         return (retval);
2453
2454 pullup_failed:
2455         if (V_fw_verbose)
2456                 printf("ipfw: pullup failed\n");
2457         return (IP_FW_DENY);
2458 }
2459
2460 /*
2461  * Module and VNET glue
2462  */
2463
2464 /*
2465  * Stuff that must be initialised only on boot or module load
2466  */
2467 static int
2468 ipfw_init(void)
2469 {
2470         int error = 0;
2471
2472         ipfw_dyn_attach();
2473         /*
2474          * Only print out this stuff the first time around,
2475          * when called from the sysinit code.
2476          */
2477         printf("ipfw2 "
2478 #ifdef INET6
2479                 "(+ipv6) "
2480 #endif
2481                 "initialized, divert %s, nat %s, "
2482                 "rule-based forwarding "
2483 #ifdef IPFIREWALL_FORWARD
2484                 "enabled, "
2485 #else
2486                 "disabled, "
2487 #endif
2488                 "default to %s, logging ",
2489 #ifdef IPDIVERT
2490                 "enabled",
2491 #else
2492                 "loadable",
2493 #endif
2494 #ifdef IPFIREWALL_NAT
2495                 "enabled",
2496 #else
2497                 "loadable",
2498 #endif
2499                 default_to_accept ? "accept" : "deny");
2500
2501         /*
2502          * Note: V_xxx variables can be accessed here but the vnet specific
2503          * initializer may not have been called yet for the VIMAGE case.
2504          * Tuneables will have been processed. We will print out values for
2505          * the default vnet.
2506          * XXX This should all be rationalized AFTER 8.0
2507          */
2508         if (V_fw_verbose == 0)
2509                 printf("disabled\n");
2510         else if (V_verbose_limit == 0)
2511                 printf("unlimited\n");
2512         else
2513                 printf("limited to %d packets/entry by default\n",
2514                     V_verbose_limit);
2515
2516         ipfw_log_bpf(1); /* init */
2517         return (error);
2518 }
2519
2520 /*
2521  * Called for the removal of the last instance only on module unload.
2522  */
2523 static void
2524 ipfw_destroy(void)
2525 {
2526
2527         ipfw_log_bpf(0); /* uninit */
2528         ipfw_dyn_detach();
2529         printf("IP firewall unloaded\n");
2530 }
2531
2532 /*
2533  * Stuff that must be initialized for every instance
2534  * (including the first of course).
2535  */
2536 static int
2537 vnet_ipfw_init(const void *unused)
2538 {
2539         int error;
2540         struct ip_fw *rule = NULL;
2541         struct ip_fw_chain *chain;
2542
2543         chain = &V_layer3_chain;
2544
2545         /* First set up some values that are compile time options */
2546         V_autoinc_step = 100;   /* bounded to 1..1000 in add_rule() */
2547         V_fw_deny_unknown_exthdrs = 1;
2548 #ifdef IPFIREWALL_VERBOSE
2549         V_fw_verbose = 1;
2550 #endif
2551 #ifdef IPFIREWALL_VERBOSE_LIMIT
2552         V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2553 #endif
2554 #ifdef IPFIREWALL_NAT
2555         LIST_INIT(&chain->nat);
2556 #endif
2557
2558         /* insert the default rule and create the initial map */
2559         chain->n_rules = 1;
2560         chain->static_len = sizeof(struct ip_fw);
2561         chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_NOWAIT | M_ZERO);
2562         if (chain->map)
2563                 rule = malloc(chain->static_len, M_IPFW, M_NOWAIT | M_ZERO);
2564         if (rule == NULL) {
2565                 if (chain->map)
2566                         free(chain->map, M_IPFW);
2567                 printf("ipfw2: ENOSPC initializing default rule "
2568                         "(support disabled)\n");
2569                 return (ENOSPC);
2570         }
2571         error = ipfw_init_tables(chain);
2572         if (error) {
2573                 panic("init_tables"); /* XXX Marko fix this ! */
2574         }
2575
2576         /* fill and insert the default rule */
2577         rule->act_ofs = 0;
2578         rule->rulenum = IPFW_DEFAULT_RULE;
2579         rule->cmd_len = 1;
2580         rule->set = RESVD_SET;
2581         rule->cmd[0].len = 1;
2582         rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
2583         chain->rules = chain->default_rule = chain->map[0] = rule;
2584         chain->id = rule->id = 1;
2585
2586         IPFW_LOCK_INIT(chain);
2587         ipfw_dyn_init();
2588
2589         /* First set up some values that are compile time options */
2590         V_ipfw_vnet_ready = 1;          /* Open for business */
2591
2592         /*
2593          * Hook the sockopt handler, and the layer2 (V_ip_fw_chk_ptr)
2594          * and pfil hooks for ipv4 and ipv6. Even if the latter two fail
2595          * we still keep the module alive because the sockopt and
2596          * layer2 paths are still useful.
2597          * ipfw[6]_hook return 0 on success, ENOENT on failure,
2598          * so we can ignore the exact return value and just set a flag.
2599          *
2600          * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
2601          * changes in the underlying (per-vnet) variables trigger
2602          * immediate hook()/unhook() calls.
2603          * In layer2 we have the same behaviour, except that V_ether_ipfw
2604          * is checked on each packet because there are no pfil hooks.
2605          */
2606         V_ip_fw_ctl_ptr = ipfw_ctl;
2607         V_ip_fw_chk_ptr = ipfw_chk;
2608         error = ipfw_attach_hooks(1);
2609         return (error);
2610 }
2611
2612 /*
2613  * Called for the removal of each instance.
2614  */
2615 static int
2616 vnet_ipfw_uninit(const void *unused)
2617 {
2618         struct ip_fw *reap, *rule;
2619         struct ip_fw_chain *chain = &V_layer3_chain;
2620         int i;
2621
2622         V_ipfw_vnet_ready = 0; /* tell new callers to go away */
2623         /*
2624          * disconnect from ipv4, ipv6, layer2 and sockopt.
2625          * Then grab, release and grab again the WLOCK so we make
2626          * sure the update is propagated and nobody will be in.
2627          */
2628         (void)ipfw_attach_hooks(0 /* detach */);
2629         V_ip_fw_chk_ptr = NULL;
2630         V_ip_fw_ctl_ptr = NULL;
2631         IPFW_UH_WLOCK(chain);
2632         IPFW_UH_WUNLOCK(chain);
2633         IPFW_UH_WLOCK(chain);
2634
2635         IPFW_WLOCK(chain);
2636         IPFW_WUNLOCK(chain);
2637         IPFW_WLOCK(chain);
2638
2639         ipfw_dyn_uninit(0);     /* run the callout_drain */
2640         ipfw_destroy_tables(chain);
2641         reap = NULL;
2642         for (i = 0; i < chain->n_rules; i++) {
2643                 rule = chain->map[i];
2644                 rule->x_next = reap;
2645                 reap = rule;
2646         }
2647         if (chain->map)
2648                 free(chain->map, M_IPFW);
2649         IPFW_WUNLOCK(chain);
2650         IPFW_UH_WUNLOCK(chain);
2651         if (reap != NULL)
2652                 ipfw_reap_rules(reap);
2653         IPFW_LOCK_DESTROY(chain);
2654         ipfw_dyn_uninit(1);     /* free the remaining parts */
2655         return 0;
2656 }
2657
2658 /*
2659  * Module event handler.
2660  * In general we have the choice of handling most of these events by the
2661  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
2662  * use the SYSINIT handlers as they are more capable of expressing the
2663  * flow of control during module and vnet operations, so this is just
2664  * a skeleton. Note there is no SYSINIT equivalent of the module
2665  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
2666  */
2667 static int
2668 ipfw_modevent(module_t mod, int type, void *unused)
2669 {
2670         int err = 0;
2671
2672         switch (type) {
2673         case MOD_LOAD:
2674                 /* Called once at module load or
2675                  * system boot if compiled in. */
2676                 break;
2677         case MOD_QUIESCE:
2678                 /* Called before unload. May veto unloading. */
2679                 break;
2680         case MOD_UNLOAD:
2681                 /* Called during unload. */
2682                 break;
2683         case MOD_SHUTDOWN:
2684                 /* Called during system shutdown. */
2685                 break;
2686         default:
2687                 err = EOPNOTSUPP;
2688                 break;
2689         }
2690         return err;
2691 }
2692
2693 static moduledata_t ipfwmod = {
2694         "ipfw",
2695         ipfw_modevent,
2696         0
2697 };
2698
2699 /* Define startup order. */
2700 #define IPFW_SI_SUB_FIREWALL    SI_SUB_PROTO_IFATTACHDOMAIN
2701 #define IPFW_MODEVENT_ORDER     (SI_ORDER_ANY - 255) /* On boot slot in here. */
2702 #define IPFW_MODULE_ORDER       (IPFW_MODEVENT_ORDER + 1) /* A little later. */
2703 #define IPFW_VNET_ORDER         (IPFW_MODEVENT_ORDER + 2) /* Later still. */
2704
2705 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
2706 MODULE_VERSION(ipfw, 2);
2707 /* should declare some dependencies here */
2708
2709 /*
2710  * Starting up. Done in order after ipfwmod() has been called.
2711  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
2712  */
2713 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2714             ipfw_init, NULL);
2715 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2716             vnet_ipfw_init, NULL);
2717
2718 /*
2719  * Closing up shop. These are done in REVERSE ORDER, but still
2720  * after ipfwmod() has been called. Not called on reboot.
2721  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
2722  * or when the module is unloaded.
2723  */
2724 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2725             ipfw_destroy, NULL);
2726 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2727             vnet_ipfw_uninit, NULL);
2728 /* end of file */