sys/netpfil/ipfw/ip_fw2.c

   1 /*-
   2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
   3  *
   4  * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa
   5  *
   6  * Redistribution and use in source and binary forms, with or without
   7  * modification, are permitted provided that the following conditions
   8  * are met:
   9  * 1. Redistributions of source code must retain the above copyright
  10  *    notice, this list of conditions and the following disclaimer.
  11  * 2. Redistributions in binary form must reproduce the above copyright
  12  *    notice, this list of conditions and the following disclaimer in the
  13  *    documentation and/or other materials provided with the distribution.
  14  *
  15  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
  16  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  17  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  18  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
  19  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  20  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  21  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  22  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  23  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  24  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  25  * SUCH DAMAGE.
  26  */
  27
  28 #include <sys/cdefs.h>
  29 __FBSDID("$FreeBSD$");
  30
  31 /*
  32  * The FreeBSD IP packet firewall, main file
  33  */
  34
  35 #include "opt_ipfw.h"
  36 #include "opt_ipdivert.h"
  37 #include "opt_inet.h"
  38 #ifndef INET
  39 #error "IPFIREWALL requires INET"
  40 #endif /* INET */
  41 #include "opt_inet6.h"
  42
  43 #include <sys/param.h>
  44 #include <sys/systm.h>
  45 #include <sys/condvar.h>
  46 #include <sys/counter.h>
  47 #include <sys/eventhandler.h>
  48 #include <sys/malloc.h>
  49 #include <sys/mbuf.h>
  50 #include <sys/kernel.h>
  51 #include <sys/lock.h>
  52 #include <sys/jail.h>
  53 #include <sys/module.h>
  54 #include <sys/priv.h>
  55 #include <sys/proc.h>
  56 #include <sys/rwlock.h>
  57 #include <sys/rmlock.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/if_var.h>
  66 #include <net/route.h>
  67 #include <net/pfil.h>
  68 #include <net/vnet.h>
  69
  70 #include <netpfil/pf/pf_mtag.h>
  71
  72 #include <netinet/in.h>
  73 #include <netinet/in_var.h>
  74 #include <netinet/in_pcb.h>
  75 #include <netinet/ip.h>
  76 #include <netinet/ip_var.h>
  77 #include <netinet/ip_icmp.h>
  78 #include <netinet/ip_fw.h>
  79 #include <netinet/ip_carp.h>
  80 #include <netinet/pim.h>
  81 #include <netinet/tcp_var.h>
  82 #include <netinet/udp.h>
  83 #include <netinet/udp_var.h>
  84 #include <netinet/sctp.h>
  85 #include <netinet/sctp_crc32.h>
  86 #include <netinet/sctp_header.h>
  87
  88 #include <netinet/ip6.h>
  89 #include <netinet/icmp6.h>
  90 #include <netinet/in_fib.h>
  91 #ifdef INET6
  92 #include <netinet6/in6_fib.h>
  93 #include <netinet6/in6_pcb.h>
  94 #include <netinet6/scope6_var.h>
  95 #include <netinet6/ip6_var.h>
  96 #endif
  97
  98 #include <net/if_gre.h> /* for struct grehdr */
  99
 100 #include <netpfil/ipfw/ip_fw_private.h>
 101
 102 #include <machine/in_cksum.h>   /* XXX for in_cksum */
 103
 104 #ifdef MAC
 105 #include <security/mac/mac_framework.h>
 106 #endif
 107
 108 /*
 109  * static variables followed by global ones.
 110  * All ipfw global variables are here.
 111  */
 112
 113 VNET_DEFINE_STATIC(int, fw_deny_unknown_exthdrs);
 114 #define V_fw_deny_unknown_exthdrs       VNET(fw_deny_unknown_exthdrs)
 115
 116 VNET_DEFINE_STATIC(int, fw_permit_single_frag6) = 1;
 117 #define V_fw_permit_single_frag6        VNET(fw_permit_single_frag6)
 118
 119 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
 120 static int default_to_accept = 1;
 121 #else
 122 static int default_to_accept;
 123 #endif
 124
 125 VNET_DEFINE(int, autoinc_step);
 126 VNET_DEFINE(int, fw_one_pass) = 1;
 127
 128 VNET_DEFINE(unsigned int, fw_tables_max);
 129 VNET_DEFINE(unsigned int, fw_tables_sets) = 0;  /* Don't use set-aware tables */
 130 /* Use 128 tables by default */
 131 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT;
 132
 133 #ifndef LINEAR_SKIPTO
 134 static int jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
 135     int tablearg, int jump_backwards);
 136 #define JUMP(ch, f, num, targ, back)    jump_fast(ch, f, num, targ, back)
 137 #else
 138 static int jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
 139     int tablearg, int jump_backwards);
 140 #define JUMP(ch, f, num, targ, back)    jump_linear(ch, f, num, targ, back)
 141 #endif
 142
 143 /*
 144  * Each rule belongs to one of 32 different sets (0..31).
 145  * The variable set_disable contains one bit per set.
 146  * If the bit is set, all rules in the corresponding set
 147  * are disabled. Set RESVD_SET(31) is reserved for the default rule
 148  * and rules that are not deleted by the flush command,
 149  * and CANNOT be disabled.
 150  * Rules in set RESVD_SET can only be deleted individually.
 151  */
 152 VNET_DEFINE(u_int32_t, set_disable);
 153 #define V_set_disable                   VNET(set_disable)
 154
 155 VNET_DEFINE(int, fw_verbose);
 156 /* counter for ipfw_log(NULL...) */
 157 VNET_DEFINE(u_int64_t, norule_counter);
 158 VNET_DEFINE(int, verbose_limit);
 159
 160 /* layer3_chain contains the list of rules for layer 3 */
 161 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
 162
 163 /* ipfw_vnet_ready controls when we are open for business */
 164 VNET_DEFINE(int, ipfw_vnet_ready) = 0;
 165
 166 VNET_DEFINE(int, ipfw_nat_ready) = 0;
 167
 168 ipfw_nat_t *ipfw_nat_ptr = NULL;
 169 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
 170 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
 171 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
 172 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
 173 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
 174
 175 #ifdef SYSCTL_NODE
 176 uint32_t dummy_def = IPFW_DEFAULT_RULE;
 177 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS);
 178 static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS);
 179
 180 SYSBEGIN(f3)
 181
 182 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 183 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
 184     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
 185     "Only do a single pass through ipfw when using dummynet(4)");
 186 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
 187     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
 188     "Rule number auto-increment step");
 189 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose,
 190     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
 191     "Log matches to ipfw rules");
 192 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
 193     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
 194     "Set upper limit of matches of ipfw rules logged");
 195 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
 196     &dummy_def, 0,
 197     "The default/max possible rule number.");
 198 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
 199     CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU",
 200     "Maximum number of concurrently used tables");
 201 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets,
 202     CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW,
 203     0, 0, sysctl_ipfw_tables_sets, "IU",
 204     "Use per-set namespace for tables");
 205 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
 206     &default_to_accept, 0,
 207     "Make the default rule accept all packets.");
 208 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables);
 209 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count,
 210     CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
 211     "Number of static rules");
 212
 213 #ifdef INET6
 214 SYSCTL_DECL(_net_inet6_ip6);
 215 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 216 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
 217     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
 218     &VNET_NAME(fw_deny_unknown_exthdrs), 0,
 219     "Deny packets with unknown IPv6 Extension Headers");
 220 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
 221     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
 222     &VNET_NAME(fw_permit_single_frag6), 0,
 223     "Permit single packet IPv6 fragments");
 224 #endif /* INET6 */
 225
 226 SYSEND
 227
 228 #endif /* SYSCTL_NODE */
 229
 230
 231 /*
 232  * Some macros used in the various matching options.
 233  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
 234  * Other macros just cast void * into the appropriate type
 235  */
 236 #define L3HDR(T, ip)    ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
 237 #define TCP(p)          ((struct tcphdr *)(p))
 238 #define SCTP(p)         ((struct sctphdr *)(p))
 239 #define UDP(p)          ((struct udphdr *)(p))
 240 #define ICMP(p)         ((struct icmphdr *)(p))
 241 #define ICMP6(p)        ((struct icmp6_hdr *)(p))
 242
 243 static __inline int
 244 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
 245 {
 246         int type = icmp->icmp_type;
 247
 248         return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
 249 }
 250
 251 #define TT      ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
 252     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
 253
 254 static int
 255 is_icmp_query(struct icmphdr *icmp)
 256 {
 257         int type = icmp->icmp_type;
 258
 259         return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
 260 }
 261 #undef TT
 262
 263 /*
 264  * The following checks use two arrays of 8 or 16 bits to store the
 265  * bits that we want set or clear, respectively. They are in the
 266  * low and high half of cmd->arg1 or cmd->d[0].
 267  *
 268  * We scan options and store the bits we find set. We succeed if
 269  *
 270  *      (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
 271  *
 272  * The code is sometimes optimized not to store additional variables.
 273  */
 274
 275 static int
 276 flags_match(ipfw_insn *cmd, u_int8_t bits)
 277 {
 278         u_char want_clear;
 279         bits = ~bits;
 280
 281         if ( ((cmd->arg1 & 0xff) & bits) != 0)
 282                 return 0; /* some bits we want set were clear */
 283         want_clear = (cmd->arg1 >> 8) & 0xff;
 284         if ( (want_clear & bits) != want_clear)
 285                 return 0; /* some bits we want clear were set */
 286         return 1;
 287 }
 288
 289 static int
 290 ipopts_match(struct ip *ip, ipfw_insn *cmd)
 291 {
 292         int optlen, bits = 0;
 293         u_char *cp = (u_char *)(ip + 1);
 294         int x = (ip->ip_hl << 2) - sizeof (struct ip);
 295
 296         for (; x > 0; x -= optlen, cp += optlen) {
 297                 int opt = cp[IPOPT_OPTVAL];
 298
 299                 if (opt == IPOPT_EOL)
 300                         break;
 301                 if (opt == IPOPT_NOP)
 302                         optlen = 1;
 303                 else {
 304                         optlen = cp[IPOPT_OLEN];
 305                         if (optlen <= 0 || optlen > x)
 306                                 return 0; /* invalid or truncated */
 307                 }
 308                 switch (opt) {
 309
 310                 default:
 311                         break;
 312
 313                 case IPOPT_LSRR:
 314                         bits |= IP_FW_IPOPT_LSRR;
 315                         break;
 316
 317                 case IPOPT_SSRR:
 318                         bits |= IP_FW_IPOPT_SSRR;
 319                         break;
 320
 321                 case IPOPT_RR:
 322                         bits |= IP_FW_IPOPT_RR;
 323                         break;
 324
 325                 case IPOPT_TS:
 326                         bits |= IP_FW_IPOPT_TS;
 327                         break;
 328                 }
 329         }
 330         return (flags_match(cmd, bits));
 331 }
 332
 333 /*
 334  * Parse TCP options. The logic copied from tcp_dooptions().
 335  */
 336 static int
 337 tcpopts_parse(const struct tcphdr *tcp, uint16_t *mss)
 338 {
 339         const u_char *cp = (const u_char *)(tcp + 1);
 340         int optlen, bits = 0;
 341         int cnt = (tcp->th_off << 2) - sizeof(struct tcphdr);
 342
 343         for (; cnt > 0; cnt -= optlen, cp += optlen) {
 344                 int opt = cp[0];
 345                 if (opt == TCPOPT_EOL)
 346                         break;
 347                 if (opt == TCPOPT_NOP)
 348                         optlen = 1;
 349                 else {
 350                         if (cnt < 2)
 351                                 break;
 352                         optlen = cp[1];
 353                         if (optlen < 2 || optlen > cnt)
 354                                 break;
 355                 }
 356
 357                 switch (opt) {
 358                 default:
 359                         break;
 360
 361                 case TCPOPT_MAXSEG:
 362                         if (optlen != TCPOLEN_MAXSEG)
 363                                 break;
 364                         bits |= IP_FW_TCPOPT_MSS;
 365                         if (mss != NULL)
 366                                 *mss = be16dec(cp + 2);
 367                         break;
 368
 369                 case TCPOPT_WINDOW:
 370                         if (optlen == TCPOLEN_WINDOW)
 371                                 bits |= IP_FW_TCPOPT_WINDOW;
 372                         break;
 373
 374                 case TCPOPT_SACK_PERMITTED:
 375                         if (optlen == TCPOLEN_SACK_PERMITTED)
 376                                 bits |= IP_FW_TCPOPT_SACK;
 377                         break;
 378
 379                 case TCPOPT_SACK:
 380                         if (optlen > 2 && (optlen - 2) % TCPOLEN_SACK == 0)
 381                                 bits |= IP_FW_TCPOPT_SACK;
 382                         break;
 383
 384                 case TCPOPT_TIMESTAMP:
 385                         if (optlen == TCPOLEN_TIMESTAMP)
 386                                 bits |= IP_FW_TCPOPT_TS;
 387                         break;
 388                 }
 389         }
 390         return (bits);
 391 }
 392
 393 static int
 394 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
 395 {
 396
 397         return (flags_match(cmd, tcpopts_parse(tcp, NULL)));
 398 }
 399
 400 static int
 401 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain,
 402     uint32_t *tablearg)
 403 {
 404
 405         if (ifp == NULL)        /* no iface with this packet, match fails */
 406                 return (0);
 407
 408         /* Check by name or by IP address */
 409         if (cmd->name[0] != '\0') { /* match by name */
 410                 if (cmd->name[0] == '\1') /* use tablearg to match */
 411                         return ipfw_lookup_table(chain, cmd->p.kidx, 0,
 412                             &ifp->if_index, tablearg);
 413                 /* Check name */
 414                 if (cmd->p.glob) {
 415                         if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
 416                                 return(1);
 417                 } else {
 418                         if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
 419                                 return(1);
 420                 }
 421         } else {
 422 #if !defined(USERSPACE) && defined(__FreeBSD__) /* and OSX too ? */
 423                 struct ifaddr *ia;
 424
 425                 if_addr_rlock(ifp);
 426                 CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
 427                         if (ia->ifa_addr->sa_family != AF_INET)
 428                                 continue;
 429                         if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
 430                             (ia->ifa_addr))->sin_addr.s_addr) {
 431                                 if_addr_runlock(ifp);
 432                                 return(1);      /* match */
 433                         }
 434                 }
 435                 if_addr_runlock(ifp);
 436 #endif /* __FreeBSD__ */
 437         }
 438         return(0);      /* no match, fail ... */
 439 }
 440
 441 /*
 442  * The verify_path function checks if a route to the src exists and
 443  * if it is reachable via ifp (when provided).
 444  *
 445  * The 'verrevpath' option checks that the interface that an IP packet
 446  * arrives on is the same interface that traffic destined for the
 447  * packet's source address would be routed out of.
 448  * The 'versrcreach' option just checks that the source address is
 449  * reachable via any route (except default) in the routing table.
 450  * These two are a measure to block forged packets. This is also
 451  * commonly known as "anti-spoofing" or Unicast Reverse Path
 452  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
 453  * is purposely reminiscent of the Cisco IOS command,
 454  *
 455  *   ip verify unicast reverse-path
 456  *   ip verify unicast source reachable-via any
 457  *
 458  * which implements the same functionality. But note that the syntax
 459  * is misleading, and the check may be performed on all IP packets
 460  * whether unicast, multicast, or broadcast.
 461  */
 462 static int
 463 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
 464 {
 465 #if defined(USERSPACE) || !defined(__FreeBSD__)
 466         return 0;
 467 #else
 468         struct nhop4_basic nh4;
 469
 470         if (fib4_lookup_nh_basic(fib, src, NHR_IFAIF, 0, &nh4) != 0)
 471                 return (0);
 472
 473         /*
 474          * If ifp is provided, check for equality with rtentry.
 475          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 476          * in order to pass packets injected back by if_simloop():
 477          * routing entry (via lo0) for our own address
 478          * may exist, so we need to handle routing assymetry.
 479          */
 480         if (ifp != NULL && ifp != nh4.nh_ifp)
 481                 return (0);
 482
 483         /* if no ifp provided, check if rtentry is not default route */
 484         if (ifp == NULL && (nh4.nh_flags & NHF_DEFAULT) != 0)
 485                 return (0);
 486
 487         /* or if this is a blackhole/reject route */
 488         if (ifp == NULL && (nh4.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
 489                 return (0);
 490
 491         /* found valid route */
 492         return 1;
 493 #endif /* __FreeBSD__ */
 494 }
 495
 496 /*
 497  * Generate an SCTP packet containing an ABORT chunk. The verification tag
 498  * is given by vtag. The T-bit is set in the ABORT chunk if and only if
 499  * reflected is not 0.
 500  */
 501
 502 static struct mbuf *
 503 ipfw_send_abort(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t vtag,
 504     int reflected)
 505 {
 506         struct mbuf *m;
 507         struct ip *ip;
 508 #ifdef INET6
 509         struct ip6_hdr *ip6;
 510 #endif
 511         struct sctphdr *sctp;
 512         struct sctp_chunkhdr *chunk;
 513         u_int16_t hlen, plen, tlen;
 514
 515         MGETHDR(m, M_NOWAIT, MT_DATA);
 516         if (m == NULL)
 517                 return (NULL);
 518
 519         M_SETFIB(m, id->fib);
 520 #ifdef MAC
 521         if (replyto != NULL)
 522                 mac_netinet_firewall_reply(replyto, m);
 523         else
 524                 mac_netinet_firewall_send(m);
 525 #else
 526         (void)replyto;          /* don't warn about unused arg */
 527 #endif
 528
 529         switch (id->addr_type) {
 530         case 4:
 531                 hlen = sizeof(struct ip);
 532                 break;
 533 #ifdef INET6
 534         case 6:
 535                 hlen = sizeof(struct ip6_hdr);
 536                 break;
 537 #endif
 538         default:
 539                 /* XXX: log me?!? */
 540                 FREE_PKT(m);
 541                 return (NULL);
 542         }
 543         plen = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr);
 544         tlen = hlen + plen;
 545         m->m_data += max_linkhdr;
 546         m->m_flags |= M_SKIP_FIREWALL;
 547         m->m_pkthdr.len = m->m_len = tlen;
 548         m->m_pkthdr.rcvif = NULL;
 549         bzero(m->m_data, tlen);
 550
 551         switch (id->addr_type) {
 552         case 4:
 553                 ip = mtod(m, struct ip *);
 554
 555                 ip->ip_v = 4;
 556                 ip->ip_hl = sizeof(struct ip) >> 2;
 557                 ip->ip_tos = IPTOS_LOWDELAY;
 558                 ip->ip_len = htons(tlen);
 559                 ip->ip_id = htons(0);
 560                 ip->ip_off = htons(0);
 561                 ip->ip_ttl = V_ip_defttl;
 562                 ip->ip_p = IPPROTO_SCTP;
 563                 ip->ip_sum = 0;
 564                 ip->ip_src.s_addr = htonl(id->dst_ip);
 565                 ip->ip_dst.s_addr = htonl(id->src_ip);
 566
 567                 sctp = (struct sctphdr *)(ip + 1);
 568                 break;
 569 #ifdef INET6
 570         case 6:
 571                 ip6 = mtod(m, struct ip6_hdr *);
 572
 573                 ip6->ip6_vfc = IPV6_VERSION;
 574                 ip6->ip6_plen = htons(plen);
 575                 ip6->ip6_nxt = IPPROTO_SCTP;
 576                 ip6->ip6_hlim = IPV6_DEFHLIM;
 577                 ip6->ip6_src = id->dst_ip6;
 578                 ip6->ip6_dst = id->src_ip6;
 579
 580                 sctp = (struct sctphdr *)(ip6 + 1);
 581                 break;
 582 #endif
 583         }
 584
 585         sctp->src_port = htons(id->dst_port);
 586         sctp->dest_port = htons(id->src_port);
 587         sctp->v_tag = htonl(vtag);
 588         sctp->checksum = htonl(0);
 589
 590         chunk = (struct sctp_chunkhdr *)(sctp + 1);
 591         chunk->chunk_type = SCTP_ABORT_ASSOCIATION;
 592         chunk->chunk_flags = 0;
 593         if (reflected != 0) {
 594                 chunk->chunk_flags |= SCTP_HAD_NO_TCB;
 595         }
 596         chunk->chunk_length = htons(sizeof(struct sctp_chunkhdr));
 597
 598         sctp->checksum = sctp_calculate_cksum(m, hlen);
 599
 600         return (m);
 601 }
 602
 603 /*
 604  * Generate a TCP packet, containing either a RST or a keepalive.
 605  * When flags & TH_RST, we are sending a RST packet, because of a
 606  * "reset" action matched the packet.
 607  * Otherwise we are sending a keepalive, and flags & TH_
 608  * The 'replyto' mbuf is the mbuf being replied to, if any, and is required
 609  * so that MAC can label the reply appropriately.
 610  */
 611 struct mbuf *
 612 ipfw_send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq,
 613     u_int32_t ack, int flags)
 614 {
 615         struct mbuf *m = NULL;          /* stupid compiler */
 616         struct ip *h = NULL;            /* stupid compiler */
 617 #ifdef INET6
 618         struct ip6_hdr *h6 = NULL;
 619 #endif
 620         struct tcphdr *th = NULL;
 621         int len, dir;
 622
 623         MGETHDR(m, M_NOWAIT, MT_DATA);
 624         if (m == NULL)
 625                 return (NULL);
 626
 627         M_SETFIB(m, id->fib);
 628 #ifdef MAC
 629         if (replyto != NULL)
 630                 mac_netinet_firewall_reply(replyto, m);
 631         else
 632                 mac_netinet_firewall_send(m);
 633 #else
 634         (void)replyto;          /* don't warn about unused arg */
 635 #endif
 636
 637         switch (id->addr_type) {
 638         case 4:
 639                 len = sizeof(struct ip) + sizeof(struct tcphdr);
 640                 break;
 641 #ifdef INET6
 642         case 6:
 643                 len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
 644                 break;
 645 #endif
 646         default:
 647                 /* XXX: log me?!? */
 648                 FREE_PKT(m);
 649                 return (NULL);
 650         }
 651         dir = ((flags & (TH_SYN | TH_RST)) == TH_SYN);
 652
 653         m->m_data += max_linkhdr;
 654         m->m_flags |= M_SKIP_FIREWALL;
 655         m->m_pkthdr.len = m->m_len = len;
 656         m->m_pkthdr.rcvif = NULL;
 657         bzero(m->m_data, len);
 658
 659         switch (id->addr_type) {
 660         case 4:
 661                 h = mtod(m, struct ip *);
 662
 663                 /* prepare for checksum */
 664                 h->ip_p = IPPROTO_TCP;
 665                 h->ip_len = htons(sizeof(struct tcphdr));
 666                 if (dir) {
 667                         h->ip_src.s_addr = htonl(id->src_ip);
 668                         h->ip_dst.s_addr = htonl(id->dst_ip);
 669                 } else {
 670                         h->ip_src.s_addr = htonl(id->dst_ip);
 671                         h->ip_dst.s_addr = htonl(id->src_ip);
 672                 }
 673
 674                 th = (struct tcphdr *)(h + 1);
 675                 break;
 676 #ifdef INET6
 677         case 6:
 678                 h6 = mtod(m, struct ip6_hdr *);
 679
 680                 /* prepare for checksum */
 681                 h6->ip6_nxt = IPPROTO_TCP;
 682                 h6->ip6_plen = htons(sizeof(struct tcphdr));
 683                 if (dir) {
 684                         h6->ip6_src = id->src_ip6;
 685                         h6->ip6_dst = id->dst_ip6;
 686                 } else {
 687                         h6->ip6_src = id->dst_ip6;
 688                         h6->ip6_dst = id->src_ip6;
 689                 }
 690
 691                 th = (struct tcphdr *)(h6 + 1);
 692                 break;
 693 #endif
 694         }
 695
 696         if (dir) {
 697                 th->th_sport = htons(id->src_port);
 698                 th->th_dport = htons(id->dst_port);
 699         } else {
 700                 th->th_sport = htons(id->dst_port);
 701                 th->th_dport = htons(id->src_port);
 702         }
 703         th->th_off = sizeof(struct tcphdr) >> 2;
 704
 705         if (flags & TH_RST) {
 706                 if (flags & TH_ACK) {
 707                         th->th_seq = htonl(ack);
 708                         th->th_flags = TH_RST;
 709                 } else {
 710                         if (flags & TH_SYN)
 711                                 seq++;
 712                         th->th_ack = htonl(seq);
 713                         th->th_flags = TH_RST | TH_ACK;
 714                 }
 715         } else {
 716                 /*
 717                  * Keepalive - use caller provided sequence numbers
 718                  */
 719                 th->th_seq = htonl(seq);
 720                 th->th_ack = htonl(ack);
 721                 th->th_flags = TH_ACK;
 722         }
 723
 724         switch (id->addr_type) {
 725         case 4:
 726                 th->th_sum = in_cksum(m, len);
 727
 728                 /* finish the ip header */
 729                 h->ip_v = 4;
 730                 h->ip_hl = sizeof(*h) >> 2;
 731                 h->ip_tos = IPTOS_LOWDELAY;
 732                 h->ip_off = htons(0);
 733                 h->ip_len = htons(len);
 734                 h->ip_ttl = V_ip_defttl;
 735                 h->ip_sum = 0;
 736                 break;
 737 #ifdef INET6
 738         case 6:
 739                 th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(*h6),
 740                     sizeof(struct tcphdr));
 741
 742                 /* finish the ip6 header */
 743                 h6->ip6_vfc |= IPV6_VERSION;
 744                 h6->ip6_hlim = IPV6_DEFHLIM;
 745                 break;
 746 #endif
 747         }
 748
 749         return (m);
 750 }
 751
 752 #ifdef INET6
 753 /*
 754  * ipv6 specific rules here...
 755  */
 756 static __inline int
 757 icmp6type_match (int type, ipfw_insn_u32 *cmd)
 758 {
 759         return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
 760 }
 761
 762 static int
 763 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
 764 {
 765         int i;
 766         for (i=0; i <= cmd->o.arg1; ++i )
 767                 if (curr_flow == cmd->d[i] )
 768                         return 1;
 769         return 0;
 770 }
 771
 772 /* support for IP6_*_ME opcodes */
 773 static const struct in6_addr lla_mask = {{{
 774         0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff,
 775         0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
 776 }}};
 777
 778 static int
 779 ipfw_localip6(struct in6_addr *in6)
 780 {
 781         struct rm_priotracker in6_ifa_tracker;
 782         struct in6_ifaddr *ia;
 783
 784         if (IN6_IS_ADDR_MULTICAST(in6))
 785                 return (0);
 786
 787         if (!IN6_IS_ADDR_LINKLOCAL(in6))
 788                 return (in6_localip(in6));
 789
 790         IN6_IFADDR_RLOCK(&in6_ifa_tracker);
 791         CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
 792                 if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr))
 793                         continue;
 794                 if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
 795                     in6, &lla_mask)) {
 796                         IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
 797                         return (1);
 798                 }
 799         }
 800         IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
 801         return (0);
 802 }
 803
 804 static int
 805 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
 806 {
 807         struct nhop6_basic nh6;
 808
 809         if (IN6_IS_SCOPE_LINKLOCAL(src))
 810                 return (1);
 811
 812         if (fib6_lookup_nh_basic(fib, src, 0, NHR_IFAIF, 0, &nh6) != 0)
 813                 return (0);
 814
 815         /* If ifp is provided, check for equality with route table. */
 816         if (ifp != NULL && ifp != nh6.nh_ifp)
 817                 return (0);
 818
 819         /* if no ifp provided, check if rtentry is not default route */
 820         if (ifp == NULL && (nh6.nh_flags & NHF_DEFAULT) != 0)
 821                 return (0);
 822
 823         /* or if this is a blackhole/reject route */
 824         if (ifp == NULL && (nh6.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
 825                 return (0);
 826
 827         /* found valid route */
 828         return 1;
 829 }
 830
 831 static int
 832 is_icmp6_query(int icmp6_type)
 833 {
 834         if ((icmp6_type <= ICMP6_MAXTYPE) &&
 835             (icmp6_type == ICMP6_ECHO_REQUEST ||
 836             icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
 837             icmp6_type == ICMP6_WRUREQUEST ||
 838             icmp6_type == ICMP6_FQDN_QUERY ||
 839             icmp6_type == ICMP6_NI_QUERY))
 840                 return (1);
 841
 842         return (0);
 843 }
 844
 845 static int
 846 map_icmp_unreach(int code)
 847 {
 848
 849         /* RFC 7915 p4.2 */
 850         switch (code) {
 851         case ICMP_UNREACH_NET:
 852         case ICMP_UNREACH_HOST:
 853         case ICMP_UNREACH_SRCFAIL:
 854         case ICMP_UNREACH_NET_UNKNOWN:
 855         case ICMP_UNREACH_HOST_UNKNOWN:
 856         case ICMP_UNREACH_TOSNET:
 857         case ICMP_UNREACH_TOSHOST:
 858                 return (ICMP6_DST_UNREACH_NOROUTE);
 859         case ICMP_UNREACH_PORT:
 860                 return (ICMP6_DST_UNREACH_NOPORT);
 861         default:
 862                 /*
 863                  * Map the rest of codes into admit prohibited.
 864                  * XXX: unreach proto should be mapped into ICMPv6
 865                  * parameter problem, but we use only unreach type.
 866                  */
 867                 return (ICMP6_DST_UNREACH_ADMIN);
 868         }
 869 }
 870
 871 static void
 872 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
 873 {
 874         struct mbuf *m;
 875
 876         m = args->m;
 877         if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
 878                 struct tcphdr *tcp;
 879                 tcp = (struct tcphdr *)((char *)ip6 + hlen);
 880
 881                 if ((tcp->th_flags & TH_RST) == 0) {
 882                         struct mbuf *m0;
 883                         m0 = ipfw_send_pkt(args->m, &(args->f_id),
 884                             ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 885                             tcp->th_flags | TH_RST);
 886                         if (m0 != NULL)
 887                                 ip6_output(m0, NULL, NULL, 0, NULL, NULL,
 888                                     NULL);
 889                 }
 890                 FREE_PKT(m);
 891         } else if (code == ICMP6_UNREACH_ABORT &&
 892             args->f_id.proto == IPPROTO_SCTP) {
 893                 struct mbuf *m0;
 894                 struct sctphdr *sctp;
 895                 u_int32_t v_tag;
 896                 int reflected;
 897
 898                 sctp = (struct sctphdr *)((char *)ip6 + hlen);
 899                 reflected = 1;
 900                 v_tag = ntohl(sctp->v_tag);
 901                 /* Investigate the first chunk header if available */
 902                 if (m->m_len >= hlen + sizeof(struct sctphdr) +
 903                     sizeof(struct sctp_chunkhdr)) {
 904                         struct sctp_chunkhdr *chunk;
 905
 906                         chunk = (struct sctp_chunkhdr *)(sctp + 1);
 907                         switch (chunk->chunk_type) {
 908                         case SCTP_INITIATION:
 909                                 /*
 910                                  * Packets containing an INIT chunk MUST have
 911                                  * a zero v-tag.
 912                                  */
 913                                 if (v_tag != 0) {
 914                                         v_tag = 0;
 915                                         break;
 916                                 }
 917                                 /* INIT chunk MUST NOT be bundled */
 918                                 if (m->m_pkthdr.len >
 919                                     hlen + sizeof(struct sctphdr) +
 920                                     ntohs(chunk->chunk_length) + 3) {
 921                                         break;
 922                                 }
 923                                 /* Use the initiate tag if available */
 924                                 if ((m->m_len >= hlen + sizeof(struct sctphdr) +
 925                                     sizeof(struct sctp_chunkhdr) +
 926                                     offsetof(struct sctp_init, a_rwnd))) {
 927                                         struct sctp_init *init;
 928
 929                                         init = (struct sctp_init *)(chunk + 1);
 930                                         v_tag = ntohl(init->initiate_tag);
 931                                         reflected = 0;
 932                                 }
 933                                 break;
 934                         case SCTP_ABORT_ASSOCIATION:
 935                                 /*
 936                                  * If the packet contains an ABORT chunk, don't
 937                                  * reply.
 938                                  * XXX: We should search through all chunks,
 939                                  *      but don't do to avoid attacks.
 940                                  */
 941                                 v_tag = 0;
 942                                 break;
 943                         }
 944                 }
 945                 if (v_tag == 0) {
 946                         m0 = NULL;
 947                 } else {
 948                         m0 = ipfw_send_abort(args->m, &(args->f_id), v_tag,
 949                             reflected);
 950                 }
 951                 if (m0 != NULL)
 952                         ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
 953                 FREE_PKT(m);
 954         } else if (code != ICMP6_UNREACH_RST && code != ICMP6_UNREACH_ABORT) {
 955                 /* Send an ICMPv6 unreach. */
 956 #if 0
 957                 /*
 958                  * Unlike above, the mbufs need to line up with the ip6 hdr,
 959                  * as the contents are read. We need to m_adj() the
 960                  * needed amount.
 961                  * The mbuf will however be thrown away so we can adjust it.
 962                  * Remember we did an m_pullup on it already so we
 963                  * can make some assumptions about contiguousness.
 964                  */
 965                 if (args->L3offset)
 966                         m_adj(m, args->L3offset);
 967 #endif
 968                 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
 969         } else
 970                 FREE_PKT(m);
 971
 972         args->m = NULL;
 973 }
 974
 975 #endif /* INET6 */
 976
 977
 978 /*
 979  * sends a reject message, consuming the mbuf passed as an argument.
 980  */
 981 static void
 982 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
 983 {
 984
 985 #if 0
 986         /* XXX When ip is not guaranteed to be at mtod() we will
 987          * need to account for this */
 988          * The mbuf will however be thrown away so we can adjust it.
 989          * Remember we did an m_pullup on it already so we
 990          * can make some assumptions about contiguousness.
 991          */
 992         if (args->L3offset)
 993                 m_adj(m, args->L3offset);
 994 #endif
 995         if (code != ICMP_REJECT_RST && code != ICMP_REJECT_ABORT) {
 996                 /* Send an ICMP unreach */
 997                 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
 998         } else if (code == ICMP_REJECT_RST && args->f_id.proto == IPPROTO_TCP) {
 999                 struct tcphdr *const tcp =
1000                     L3HDR(struct tcphdr, mtod(args->m, struct ip *));
1001                 if ( (tcp->th_flags & TH_RST) == 0) {
1002                         struct mbuf *m;
1003                         m = ipfw_send_pkt(args->m, &(args->f_id),
1004                                 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
1005                                 tcp->th_flags | TH_RST);
1006                         if (m != NULL)
1007                                 ip_output(m, NULL, NULL, 0, NULL, NULL);
1008                 }
1009                 FREE_PKT(args->m);
1010         } else if (code == ICMP_REJECT_ABORT &&
1011             args->f_id.proto == IPPROTO_SCTP) {
1012                 struct mbuf *m;
1013                 struct sctphdr *sctp;
1014                 struct sctp_chunkhdr *chunk;
1015                 struct sctp_init *init;
1016                 u_int32_t v_tag;
1017                 int reflected;
1018
1019                 sctp = L3HDR(struct sctphdr, mtod(args->m, struct ip *));
1020                 reflected = 1;
1021                 v_tag = ntohl(sctp->v_tag);
1022                 if (iplen >= (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1023                     sizeof(struct sctp_chunkhdr)) {
1024                         /* Look at the first chunk header if available */
1025                         chunk = (struct sctp_chunkhdr *)(sctp + 1);
1026                         switch (chunk->chunk_type) {
1027                         case SCTP_INITIATION:
1028                                 /*
1029                                  * Packets containing an INIT chunk MUST have
1030                                  * a zero v-tag.
1031                                  */
1032                                 if (v_tag != 0) {
1033                                         v_tag = 0;
1034                                         break;
1035                                 }
1036                                 /* INIT chunk MUST NOT be bundled */
1037                                 if (iplen >
1038                                     (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1039                                     ntohs(chunk->chunk_length) + 3) {
1040                                         break;
1041                                 }
1042                                 /* Use the initiate tag if available */
1043                                 if ((iplen >= (ip->ip_hl << 2) +
1044                                     sizeof(struct sctphdr) +
1045                                     sizeof(struct sctp_chunkhdr) +
1046                                     offsetof(struct sctp_init, a_rwnd))) {
1047                                         init = (struct sctp_init *)(chunk + 1);
1048                                         v_tag = ntohl(init->initiate_tag);
1049                                         reflected = 0;
1050                                 }
1051                                 break;
1052                         case SCTP_ABORT_ASSOCIATION:
1053                                 /*
1054                                  * If the packet contains an ABORT chunk, don't
1055                                  * reply.
1056                                  * XXX: We should search through all chunks,
1057                                  * but don't do to avoid attacks.
1058                                  */
1059                                 v_tag = 0;
1060                                 break;
1061                         }
1062                 }
1063                 if (v_tag == 0) {
1064                         m = NULL;
1065                 } else {
1066                         m = ipfw_send_abort(args->m, &(args->f_id), v_tag,
1067                             reflected);
1068                 }
1069                 if (m != NULL)
1070                         ip_output(m, NULL, NULL, 0, NULL, NULL);
1071                 FREE_PKT(args->m);
1072         } else
1073                 FREE_PKT(args->m);
1074         args->m = NULL;
1075 }
1076
1077 /*
1078  * Support for uid/gid/jail lookup. These tests are expensive
1079  * (because we may need to look into the list of active sockets)
1080  * so we cache the results. ugid_lookupp is 0 if we have not
1081  * yet done a lookup, 1 if we succeeded, and -1 if we tried
1082  * and failed. The function always returns the match value.
1083  * We could actually spare the variable and use *uc, setting
1084  * it to '(void *)check_uidgid if we have no info, NULL if
1085  * we tried and failed, or any other value if successful.
1086  */
1087 static int
1088 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
1089     struct ucred **uc)
1090 {
1091 #if defined(USERSPACE)
1092         return 0;       // not supported in userspace
1093 #else
1094 #ifndef __FreeBSD__
1095         /* XXX */
1096         return cred_check(insn, proto, oif,
1097             dst_ip, dst_port, src_ip, src_port,
1098             (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
1099 #else  /* FreeBSD */
1100         struct in_addr src_ip, dst_ip;
1101         struct inpcbinfo *pi;
1102         struct ipfw_flow_id *id;
1103         struct inpcb *pcb, *inp;
1104         struct ifnet *oif;
1105         int lookupflags;
1106         int match;
1107
1108         id = &args->f_id;
1109         inp = args->inp;
1110         oif = args->oif;
1111
1112         /*
1113          * Check to see if the UDP or TCP stack supplied us with
1114          * the PCB. If so, rather then holding a lock and looking
1115          * up the PCB, we can use the one that was supplied.
1116          */
1117         if (inp && *ugid_lookupp == 0) {
1118                 INP_LOCK_ASSERT(inp);
1119                 if (inp->inp_socket != NULL) {
1120                         *uc = crhold(inp->inp_cred);
1121                         *ugid_lookupp = 1;
1122                 } else
1123                         *ugid_lookupp = -1;
1124         }
1125         /*
1126          * If we have already been here and the packet has no
1127          * PCB entry associated with it, then we can safely
1128          * assume that this is a no match.
1129          */
1130         if (*ugid_lookupp == -1)
1131                 return (0);
1132         if (id->proto == IPPROTO_TCP) {
1133                 lookupflags = 0;
1134                 pi = &V_tcbinfo;
1135         } else if (id->proto == IPPROTO_UDP) {
1136                 lookupflags = INPLOOKUP_WILDCARD;
1137                 pi = &V_udbinfo;
1138         } else if (id->proto == IPPROTO_UDPLITE) {
1139                 lookupflags = INPLOOKUP_WILDCARD;
1140                 pi = &V_ulitecbinfo;
1141         } else
1142                 return 0;
1143         lookupflags |= INPLOOKUP_RLOCKPCB;
1144         match = 0;
1145         if (*ugid_lookupp == 0) {
1146                 if (id->addr_type == 6) {
1147 #ifdef INET6
1148                         if (oif == NULL)
1149                                 pcb = in6_pcblookup_mbuf(pi,
1150                                     &id->src_ip6, htons(id->src_port),
1151                                     &id->dst_ip6, htons(id->dst_port),
1152                                     lookupflags, oif, args->m);
1153                         else
1154                                 pcb = in6_pcblookup_mbuf(pi,
1155                                     &id->dst_ip6, htons(id->dst_port),
1156                                     &id->src_ip6, htons(id->src_port),
1157                                     lookupflags, oif, args->m);
1158 #else
1159                         *ugid_lookupp = -1;
1160                         return (0);
1161 #endif
1162                 } else {
1163                         src_ip.s_addr = htonl(id->src_ip);
1164                         dst_ip.s_addr = htonl(id->dst_ip);
1165                         if (oif == NULL)
1166                                 pcb = in_pcblookup_mbuf(pi,
1167                                     src_ip, htons(id->src_port),
1168                                     dst_ip, htons(id->dst_port),
1169                                     lookupflags, oif, args->m);
1170                         else
1171                                 pcb = in_pcblookup_mbuf(pi,
1172                                     dst_ip, htons(id->dst_port),
1173                                     src_ip, htons(id->src_port),
1174                                     lookupflags, oif, args->m);
1175                 }
1176                 if (pcb != NULL) {
1177                         INP_RLOCK_ASSERT(pcb);
1178                         *uc = crhold(pcb->inp_cred);
1179                         *ugid_lookupp = 1;
1180                         INP_RUNLOCK(pcb);
1181                 }
1182                 if (*ugid_lookupp == 0) {
1183                         /*
1184                          * We tried and failed, set the variable to -1
1185                          * so we will not try again on this packet.
1186                          */
1187                         *ugid_lookupp = -1;
1188                         return (0);
1189                 }
1190         }
1191         if (insn->o.opcode == O_UID)
1192                 match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
1193         else if (insn->o.opcode == O_GID)
1194                 match = groupmember((gid_t)insn->d[0], *uc);
1195         else if (insn->o.opcode == O_JAIL)
1196                 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
1197         return (match);
1198 #endif /* __FreeBSD__ */
1199 #endif /* not supported in userspace */
1200 }
1201
1202 /*
1203  * Helper function to set args with info on the rule after the matching
1204  * one. slot is precise, whereas we guess rule_id as they are
1205  * assigned sequentially.
1206  */
1207 static inline void
1208 set_match(struct ip_fw_args *args, int slot,
1209         struct ip_fw_chain *chain)
1210 {
1211         args->rule.chain_id = chain->id;
1212         args->rule.slot = slot + 1; /* we use 0 as a marker */
1213         args->rule.rule_id = 1 + chain->map[slot]->id;
1214         args->rule.rulenum = chain->map[slot]->rulenum;
1215         args->flags |= IPFW_ARGS_REF;
1216 }
1217
1218 #ifndef LINEAR_SKIPTO
1219 /*
1220  * Helper function to enable cached rule lookups using
1221  * cached_id and cached_pos fields in ipfw rule.
1222  */
1223 static int
1224 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1225     int tablearg, int jump_backwards)
1226 {
1227         int f_pos;
1228
1229         /* If possible use cached f_pos (in f->cached_pos),
1230          * whose version is written in f->cached_id
1231          * (horrible hacks to avoid changing the ABI).
1232          */
1233         if (num != IP_FW_TARG && f->cached_id == chain->id)
1234                 f_pos = f->cached_pos;
1235         else {
1236                 int i = IP_FW_ARG_TABLEARG(chain, num, skipto);
1237                 /* make sure we do not jump backward */
1238                 if (jump_backwards == 0 && i <= f->rulenum)
1239                         i = f->rulenum + 1;
1240                 if (chain->idxmap != NULL)
1241                         f_pos = chain->idxmap[i];
1242                 else
1243                         f_pos = ipfw_find_rule(chain, i, 0);
1244                 /* update the cache */
1245                 if (num != IP_FW_TARG) {
1246                         f->cached_id = chain->id;
1247                         f->cached_pos = f_pos;
1248                 }
1249         }
1250
1251         return (f_pos);
1252 }
1253 #else
1254 /*
1255  * Helper function to enable real fast rule lookups.
1256  */
1257 static int
1258 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1259     int tablearg, int jump_backwards)
1260 {
1261         int f_pos;
1262
1263         num = IP_FW_ARG_TABLEARG(chain, num, skipto);
1264         /* make sure we do not jump backward */
1265         if (jump_backwards == 0 && num <= f->rulenum)
1266                 num = f->rulenum + 1;
1267         f_pos = chain->idxmap[num];
1268
1269         return (f_pos);
1270 }
1271 #endif
1272
1273 #define TARG(k, f)      IP_FW_ARG_TABLEARG(chain, k, f)
1274 /*
1275  * The main check routine for the firewall.
1276  *
1277  * All arguments are in args so we can modify them and return them
1278  * back to the caller.
1279  *
1280  * Parameters:
1281  *
1282  *      args->m (in/out) The packet; we set to NULL when/if we nuke it.
1283  *              Starts with the IP header.
1284  *      args->eh (in)   Mac header if present, NULL for layer3 packet.
1285  *      args->L3offset  Number of bytes bypassed if we came from L2.
1286  *                      e.g. often sizeof(eh)  ** NOTYET **
1287  *      args->oif       Outgoing interface, NULL if packet is incoming.
1288  *              The incoming interface is in the mbuf. (in)
1289  *      args->divert_rule (in/out)
1290  *              Skip up to the first rule past this rule number;
1291  *              upon return, non-zero port number for divert or tee.
1292  *
1293  *      args->rule      Pointer to the last matching rule (in/out)
1294  *      args->next_hop  Socket we are forwarding to (out).
1295  *      args->next_hop6 IPv6 next hop we are forwarding to (out).
1296  *      args->f_id      Addresses grabbed from the packet (out)
1297  *      args->rule.info a cookie depending on rule action
1298  *
1299  * Return value:
1300  *
1301  *      IP_FW_PASS      the packet must be accepted
1302  *      IP_FW_DENY      the packet must be dropped
1303  *      IP_FW_DIVERT    divert packet, port in m_tag
1304  *      IP_FW_TEE       tee packet, port in m_tag
1305  *      IP_FW_DUMMYNET  to dummynet, pipe in args->cookie
1306  *      IP_FW_NETGRAPH  into netgraph, cookie args->cookie
1307  *              args->rule contains the matching rule,
1308  *              args->rule.info has additional information.
1309  *
1310  */
1311 int
1312 ipfw_chk(struct ip_fw_args *args)
1313 {
1314
1315         /*
1316          * Local variables holding state while processing a packet:
1317          *
1318          * IMPORTANT NOTE: to speed up the processing of rules, there
1319          * are some assumption on the values of the variables, which
1320          * are documented here. Should you change them, please check
1321          * the implementation of the various instructions to make sure
1322          * that they still work.
1323          *
1324          * args->eh     The MAC header. It is non-null for a layer2
1325          *      packet, it is NULL for a layer-3 packet.
1326          * **notyet**
1327          * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
1328          *
1329          * m | args->m  Pointer to the mbuf, as received from the caller.
1330          *      It may change if ipfw_chk() does an m_pullup, or if it
1331          *      consumes the packet because it calls send_reject().
1332          *      XXX This has to change, so that ipfw_chk() never modifies
1333          *      or consumes the buffer.
1334          * ip   is the beginning of the ip(4 or 6) header.
1335          *      Calculated by adding the L3offset to the start of data.
1336          *      (Until we start using L3offset, the packet is
1337          *      supposed to start with the ip header).
1338          */
1339         struct mbuf *m = args->m;
1340         struct ip *ip = mtod(m, struct ip *);
1341
1342         /*
1343          * For rules which contain uid/gid or jail constraints, cache
1344          * a copy of the users credentials after the pcb lookup has been
1345          * executed. This will speed up the processing of rules with
1346          * these types of constraints, as well as decrease contention
1347          * on pcb related locks.
1348          */
1349 #ifndef __FreeBSD__
1350         struct bsd_ucred ucred_cache;
1351 #else
1352         struct ucred *ucred_cache = NULL;
1353 #endif
1354         int ucred_lookup = 0;
1355
1356         /*
1357          * oif | args->oif      If NULL, ipfw_chk has been called on the
1358          *      inbound path (ether_input, ip_input).
1359          *      If non-NULL, ipfw_chk has been called on the outbound path
1360          *      (ether_output, ip_output).
1361          */
1362         struct ifnet *oif = args->oif;
1363
1364         int f_pos = 0;          /* index of current rule in the array */
1365         int retval = 0;
1366
1367         /*
1368          * hlen The length of the IP header.
1369          */
1370         u_int hlen = 0;         /* hlen >0 means we have an IP pkt */
1371
1372         /*
1373          * offset       The offset of a fragment. offset != 0 means that
1374          *      we have a fragment at this offset of an IPv4 packet.
1375          *      offset == 0 means that (if this is an IPv4 packet)
1376          *      this is the first or only fragment.
1377          *      For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
1378          *      or there is a single packet fragment (fragment header added
1379          *      without needed).  We will treat a single packet fragment as if
1380          *      there was no fragment header (or log/block depending on the
1381          *      V_fw_permit_single_frag6 sysctl setting).
1382          */
1383         u_short offset = 0;
1384         u_short ip6f_mf = 0;
1385
1386         /*
1387          * Local copies of addresses. They are only valid if we have
1388          * an IP packet.
1389          *
1390          * proto        The protocol. Set to 0 for non-ip packets,
1391          *      or to the protocol read from the packet otherwise.
1392          *      proto != 0 means that we have an IPv4 packet.
1393          *
1394          * src_port, dst_port   port numbers, in HOST format. Only
1395          *      valid for TCP and UDP packets.
1396          *
1397          * src_ip, dst_ip       ip addresses, in NETWORK format.
1398          *      Only valid for IPv4 packets.
1399          */
1400         uint8_t proto;
1401         uint16_t src_port, dst_port;            /* NOTE: host format    */
1402         struct in_addr src_ip, dst_ip;          /* NOTE: network format */
1403         int iplen = 0;
1404         int pktlen;
1405         uint16_t etype;                 /* Host order stored ether type */
1406
1407         struct ipfw_dyn_info dyn_info;
1408         struct ip_fw *q = NULL;
1409         struct ip_fw_chain *chain = &V_layer3_chain;
1410
1411         /*
1412          * We store in ulp a pointer to the upper layer protocol header.
1413          * In the ipv4 case this is easy to determine from the header,
1414          * but for ipv6 we might have some additional headers in the middle.
1415          * ulp is NULL if not found.
1416          */
1417         void *ulp = NULL;               /* upper layer protocol pointer. */
1418
1419         /* XXX ipv6 variables */
1420         int is_ipv6 = 0;
1421         uint8_t icmp6_type = 0;
1422         uint16_t ext_hd = 0;    /* bits vector for extension header filtering */
1423         /* end of ipv6 variables */
1424
1425         int is_ipv4 = 0;
1426
1427         int done = 0;           /* flag to exit the outer loop */
1428
1429         if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
1430                 return (IP_FW_PASS);    /* accept */
1431
1432         dst_ip.s_addr = 0;              /* make sure it is initialized */
1433         src_ip.s_addr = 0;              /* make sure it is initialized */
1434         src_port = dst_port = 0;
1435         pktlen = m->m_pkthdr.len;
1436
1437         DYN_INFO_INIT(&dyn_info);
1438 /*
1439  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1440  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1441  * pointer might become stale after other pullups (but we never use it
1442  * this way).
1443  */
1444 #define PULLUP_TO(_len, p, T)   PULLUP_LEN(_len, p, sizeof(T))
1445 #define _PULLUP_LOCKED(_len, p, T, unlock)                      \
1446 do {                                                            \
1447         int x = (_len) + T;                                     \
1448         if ((m)->m_len < x) {                                   \
1449                 args->m = m = m_pullup(m, x);                   \
1450                 if (m == NULL) {                                \
1451                         unlock;                                 \
1452                         goto pullup_failed;                     \
1453                 }                                               \
1454         }                                                       \
1455         p = (mtod(m, char *) + (_len));                         \
1456 } while (0)
1457
1458 #define PULLUP_LEN(_len, p, T)  _PULLUP_LOCKED(_len, p, T, )
1459 #define PULLUP_LEN_LOCKED(_len, p, T)   \
1460     _PULLUP_LOCKED(_len, p, T, IPFW_PF_RUNLOCK(chain)); \
1461     UPDATE_POINTERS()
1462 /*
1463  * In case pointers got stale after pullups, update them.
1464  */
1465 #define UPDATE_POINTERS()                       \
1466 do {                                            \
1467         ip = mtod(m, struct ip *);              \
1468 } while (0)
1469
1470         /*
1471          * if we have an ether header,
1472          */
1473         if (args->flags & IPFW_ARGS_ETHER)
1474                 etype = ntohs(args->eh->ether_type);
1475         else
1476                 etype = 0;
1477
1478         /* Identify IP packets and fill up variables. */
1479         if (pktlen >= sizeof(struct ip6_hdr) &&
1480             (etype == 0 || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
1481                 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1482
1483                 is_ipv6 = 1;
1484                 hlen = sizeof(struct ip6_hdr);
1485                 proto = ip6->ip6_nxt;
1486                 /* Search extension headers to find upper layer protocols */
1487                 while (ulp == NULL && offset == 0) {
1488                         switch (proto) {
1489                         case IPPROTO_ICMPV6:
1490                                 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1491                                 icmp6_type = ICMP6(ulp)->icmp6_type;
1492                                 break;
1493
1494                         case IPPROTO_TCP:
1495                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1496                                 dst_port = TCP(ulp)->th_dport;
1497                                 src_port = TCP(ulp)->th_sport;
1498                                 /* save flags for dynamic rules */
1499                                 args->f_id._flags = TCP(ulp)->th_flags;
1500                                 break;
1501
1502                         case IPPROTO_SCTP:
1503                                 if (pktlen >= hlen + sizeof(struct sctphdr) +
1504                                     sizeof(struct sctp_chunkhdr) +
1505                                     offsetof(struct sctp_init, a_rwnd))
1506                                         PULLUP_LEN(hlen, ulp,
1507                                             sizeof(struct sctphdr) +
1508                                             sizeof(struct sctp_chunkhdr) +
1509                                             offsetof(struct sctp_init, a_rwnd));
1510                                 else if (pktlen >= hlen + sizeof(struct sctphdr))
1511                                         PULLUP_LEN(hlen, ulp, pktlen - hlen);
1512                                 else
1513                                         PULLUP_LEN(hlen, ulp,
1514                                             sizeof(struct sctphdr));
1515                                 src_port = SCTP(ulp)->src_port;
1516                                 dst_port = SCTP(ulp)->dest_port;
1517                                 break;
1518
1519                         case IPPROTO_UDP:
1520                         case IPPROTO_UDPLITE:
1521                                 PULLUP_TO(hlen, ulp, struct udphdr);
1522                                 dst_port = UDP(ulp)->uh_dport;
1523                                 src_port = UDP(ulp)->uh_sport;
1524                                 break;
1525
1526                         case IPPROTO_HOPOPTS:   /* RFC 2460 */
1527                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1528                                 ext_hd |= EXT_HOPOPTS;
1529                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1530                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1531                                 ulp = NULL;
1532                                 break;
1533
1534                         case IPPROTO_ROUTING:   /* RFC 2460 */
1535                                 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1536                                 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1537                                 case 0:
1538                                         ext_hd |= EXT_RTHDR0;
1539                                         break;
1540                                 case 2:
1541                                         ext_hd |= EXT_RTHDR2;
1542                                         break;
1543                                 default:
1544                                         if (V_fw_verbose)
1545                                                 printf("IPFW2: IPV6 - Unknown "
1546                                                     "Routing Header type(%d)\n",
1547                                                     ((struct ip6_rthdr *)
1548                                                     ulp)->ip6r_type);
1549                                         if (V_fw_deny_unknown_exthdrs)
1550                                             return (IP_FW_DENY);
1551                                         break;
1552                                 }
1553                                 ext_hd |= EXT_ROUTING;
1554                                 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1555                                 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1556                                 ulp = NULL;
1557                                 break;
1558
1559                         case IPPROTO_FRAGMENT:  /* RFC 2460 */
1560                                 PULLUP_TO(hlen, ulp, struct ip6_frag);
1561                                 ext_hd |= EXT_FRAGMENT;
1562                                 hlen += sizeof (struct ip6_frag);
1563                                 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1564                                 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1565                                         IP6F_OFF_MASK;
1566                                 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1567                                         IP6F_MORE_FRAG;
1568                                 if (V_fw_permit_single_frag6 == 0 &&
1569                                     offset == 0 && ip6f_mf == 0) {
1570                                         if (V_fw_verbose)
1571                                                 printf("IPFW2: IPV6 - Invalid "
1572                                                     "Fragment Header\n");
1573                                         if (V_fw_deny_unknown_exthdrs)
1574                                             return (IP_FW_DENY);
1575                                         break;
1576                                 }
1577                                 args->f_id.extra =
1578                                     ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1579                                 ulp = NULL;
1580                                 break;
1581
1582                         case IPPROTO_DSTOPTS:   /* RFC 2460 */
1583                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1584                                 ext_hd |= EXT_DSTOPTS;
1585                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1586                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1587                                 ulp = NULL;
1588                                 break;
1589
1590                         case IPPROTO_AH:        /* RFC 2402 */
1591                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1592                                 ext_hd |= EXT_AH;
1593                                 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1594                                 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1595                                 ulp = NULL;
1596                                 break;
1597
1598                         case IPPROTO_ESP:       /* RFC 2406 */
1599                                 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1600                                 /* Anything past Seq# is variable length and
1601                                  * data past this ext. header is encrypted. */
1602                                 ext_hd |= EXT_ESP;
1603                                 break;
1604
1605                         case IPPROTO_NONE:      /* RFC 2460 */
1606                                 /*
1607                                  * Packet ends here, and IPv6 header has
1608                                  * already been pulled up. If ip6e_len!=0
1609                                  * then octets must be ignored.
1610                                  */
1611                                 ulp = ip; /* non-NULL to get out of loop. */
1612                                 break;
1613
1614                         case IPPROTO_OSPFIGP:
1615                                 /* XXX OSPF header check? */
1616                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1617                                 break;
1618
1619                         case IPPROTO_PIM:
1620                                 /* XXX PIM header check? */
1621                                 PULLUP_TO(hlen, ulp, struct pim);
1622                                 break;
1623
1624                         case IPPROTO_GRE:       /* RFC 1701 */
1625                                 /* XXX GRE header check? */
1626                                 PULLUP_TO(hlen, ulp, struct grehdr);
1627                                 break;
1628
1629                         case IPPROTO_CARP:
1630                                 PULLUP_TO(hlen, ulp, offsetof(
1631                                     struct carp_header, carp_counter));
1632                                 if (CARP_ADVERTISEMENT !=
1633                                     ((struct carp_header *)ulp)->carp_type)
1634                                         return (IP_FW_DENY);
1635                                 break;
1636
1637                         case IPPROTO_IPV6:      /* RFC 2893 */
1638                                 PULLUP_TO(hlen, ulp, struct ip6_hdr);
1639                                 break;
1640
1641                         case IPPROTO_IPV4:      /* RFC 2893 */
1642                                 PULLUP_TO(hlen, ulp, struct ip);
1643                                 break;
1644
1645                         default:
1646                                 if (V_fw_verbose)
1647                                         printf("IPFW2: IPV6 - Unknown "
1648                                             "Extension Header(%d), ext_hd=%x\n",
1649                                              proto, ext_hd);
1650                                 if (V_fw_deny_unknown_exthdrs)
1651                                     return (IP_FW_DENY);
1652                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1653                                 break;
1654                         } /*switch */
1655                 }
1656                 ip = mtod(m, struct ip *);
1657                 ip6 = (struct ip6_hdr *)ip;
1658                 args->f_id.addr_type = 6;
1659                 args->f_id.src_ip6 = ip6->ip6_src;
1660                 args->f_id.dst_ip6 = ip6->ip6_dst;
1661                 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1662                 iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6);
1663         } else if (pktlen >= sizeof(struct ip) &&
1664             (etype == 0 || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1665                 is_ipv4 = 1;
1666                 hlen = ip->ip_hl << 2;
1667                 /*
1668                  * Collect parameters into local variables for faster
1669                  * matching.
1670                  */
1671                 proto = ip->ip_p;
1672                 src_ip = ip->ip_src;
1673                 dst_ip = ip->ip_dst;
1674                 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1675                 iplen = ntohs(ip->ip_len);
1676
1677                 if (offset == 0) {
1678                         switch (proto) {
1679                         case IPPROTO_TCP:
1680                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1681                                 dst_port = TCP(ulp)->th_dport;
1682                                 src_port = TCP(ulp)->th_sport;
1683                                 /* save flags for dynamic rules */
1684                                 args->f_id._flags = TCP(ulp)->th_flags;
1685                                 break;
1686
1687                         case IPPROTO_SCTP:
1688                                 if (pktlen >= hlen + sizeof(struct sctphdr) +
1689                                     sizeof(struct sctp_chunkhdr) +
1690                                     offsetof(struct sctp_init, a_rwnd))
1691                                         PULLUP_LEN(hlen, ulp,
1692                                             sizeof(struct sctphdr) +
1693                                             sizeof(struct sctp_chunkhdr) +
1694                                             offsetof(struct sctp_init, a_rwnd));
1695                                 else if (pktlen >= hlen + sizeof(struct sctphdr))
1696                                         PULLUP_LEN(hlen, ulp, pktlen - hlen);
1697                                 else
1698                                         PULLUP_LEN(hlen, ulp,
1699                                             sizeof(struct sctphdr));
1700                                 src_port = SCTP(ulp)->src_port;
1701                                 dst_port = SCTP(ulp)->dest_port;
1702                                 break;
1703
1704                         case IPPROTO_UDP:
1705                         case IPPROTO_UDPLITE:
1706                                 PULLUP_TO(hlen, ulp, struct udphdr);
1707                                 dst_port = UDP(ulp)->uh_dport;
1708                                 src_port = UDP(ulp)->uh_sport;
1709                                 break;
1710
1711                         case IPPROTO_ICMP:
1712                                 PULLUP_TO(hlen, ulp, struct icmphdr);
1713                                 //args->f_id.flags = ICMP(ulp)->icmp_type;
1714                                 break;
1715
1716                         default:
1717                                 break;
1718                         }
1719                 } else {
1720                         if (offset == 1 && proto == IPPROTO_TCP) {
1721                                 /* RFC 3128 */
1722                                 goto pullup_failed;
1723                         }
1724                 }
1725
1726                 ip = mtod(m, struct ip *);
1727                 args->f_id.addr_type = 4;
1728                 args->f_id.src_ip = ntohl(src_ip.s_addr);
1729                 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1730         } else {
1731                 proto = 0;
1732                 dst_ip.s_addr = src_ip.s_addr = 0;
1733
1734                 args->f_id.addr_type = 1; /* XXX */
1735         }
1736 #undef PULLUP_TO
1737         pktlen = iplen < pktlen ? iplen: pktlen;
1738
1739         /* Properly initialize the rest of f_id */
1740         args->f_id.proto = proto;
1741         args->f_id.src_port = src_port = ntohs(src_port);
1742         args->f_id.dst_port = dst_port = ntohs(dst_port);
1743         args->f_id.fib = M_GETFIB(m);
1744
1745         IPFW_PF_RLOCK(chain);
1746         if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1747                 IPFW_PF_RUNLOCK(chain);
1748                 return (IP_FW_PASS);    /* accept */
1749         }
1750         if (args->flags & IPFW_ARGS_REF) {
1751                 /*
1752                  * Packet has already been tagged as a result of a previous
1753                  * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1754                  * REASS, NETGRAPH, DIVERT/TEE...)
1755                  * Validate the slot and continue from the next one
1756                  * if still present, otherwise do a lookup.
1757                  */
1758                 f_pos = (args->rule.chain_id == chain->id) ?
1759                     args->rule.slot :
1760                     ipfw_find_rule(chain, args->rule.rulenum,
1761                         args->rule.rule_id);
1762         } else {
1763                 f_pos = 0;
1764         }
1765
1766         /*
1767          * Now scan the rules, and parse microinstructions for each rule.
1768          * We have two nested loops and an inner switch. Sometimes we
1769          * need to break out of one or both loops, or re-enter one of
1770          * the loops with updated variables. Loop variables are:
1771          *
1772          *      f_pos (outer loop) points to the current rule.
1773          *              On output it points to the matching rule.
1774          *      done (outer loop) is used as a flag to break the loop.
1775          *      l (inner loop)  residual length of current rule.
1776          *              cmd points to the current microinstruction.
1777          *
1778          * We break the inner loop by setting l=0 and possibly
1779          * cmdlen=0 if we don't want to advance cmd.
1780          * We break the outer loop by setting done=1
1781          * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1782          * as needed.
1783          */
1784         for (; f_pos < chain->n_rules; f_pos++) {
1785                 ipfw_insn *cmd;
1786                 uint32_t tablearg = 0;
1787                 int l, cmdlen, skip_or; /* skip rest of OR block */
1788                 struct ip_fw *f;
1789
1790                 f = chain->map[f_pos];
1791                 if (V_set_disable & (1 << f->set) )
1792                         continue;
1793
1794                 skip_or = 0;
1795                 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1796                     l -= cmdlen, cmd += cmdlen) {
1797                         int match;
1798
1799                         /*
1800                          * check_body is a jump target used when we find a
1801                          * CHECK_STATE, and need to jump to the body of
1802                          * the target rule.
1803                          */
1804
1805 /* check_body: */
1806                         cmdlen = F_LEN(cmd);
1807                         /*
1808                          * An OR block (insn_1 || .. || insn_n) has the
1809                          * F_OR bit set in all but the last instruction.
1810                          * The first match will set "skip_or", and cause
1811                          * the following instructions to be skipped until
1812                          * past the one with the F_OR bit clear.
1813                          */
1814                         if (skip_or) {          /* skip this instruction */
1815                                 if ((cmd->len & F_OR) == 0)
1816                                         skip_or = 0;    /* next one is good */
1817                                 continue;
1818                         }
1819                         match = 0; /* set to 1 if we succeed */
1820
1821                         switch (cmd->opcode) {
1822                         /*
1823                          * The first set of opcodes compares the packet's
1824                          * fields with some pattern, setting 'match' if a
1825                          * match is found. At the end of the loop there is
1826                          * logic to deal with F_NOT and F_OR flags associated
1827                          * with the opcode.
1828                          */
1829                         case O_NOP:
1830                                 match = 1;
1831                                 break;
1832
1833                         case O_FORWARD_MAC:
1834                                 printf("ipfw: opcode %d unimplemented\n",
1835                                     cmd->opcode);
1836                                 break;
1837
1838                         case O_GID:
1839                         case O_UID:
1840                         case O_JAIL:
1841                                 /*
1842                                  * We only check offset == 0 && proto != 0,
1843                                  * as this ensures that we have a
1844                                  * packet with the ports info.
1845                                  */
1846                                 if (offset != 0)
1847                                         break;
1848                                 if (proto == IPPROTO_TCP ||
1849                                     proto == IPPROTO_UDP ||
1850                                     proto == IPPROTO_UDPLITE)
1851                                         match = check_uidgid(
1852                                                     (ipfw_insn_u32 *)cmd,
1853                                                     args, &ucred_lookup,
1854 #ifdef __FreeBSD__
1855                                                     &ucred_cache);
1856 #else
1857                                                     (void *)&ucred_cache);
1858 #endif
1859                                 break;
1860
1861                         case O_RECV:
1862                                 match = iface_match(m->m_pkthdr.rcvif,
1863                                     (ipfw_insn_if *)cmd, chain, &tablearg);
1864                                 break;
1865
1866                         case O_XMIT:
1867                                 match = iface_match(oif, (ipfw_insn_if *)cmd,
1868                                     chain, &tablearg);
1869                                 break;
1870
1871                         case O_VIA:
1872                                 match = iface_match(oif ? oif :
1873                                     m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1874                                     chain, &tablearg);
1875                                 break;
1876
1877                         case O_MACADDR2:
1878                                 if (args->flags & IPFW_ARGS_ETHER) {
1879                                         u_int32_t *want = (u_int32_t *)
1880                                                 ((ipfw_insn_mac *)cmd)->addr;
1881                                         u_int32_t *mask = (u_int32_t *)
1882                                                 ((ipfw_insn_mac *)cmd)->mask;
1883                                         u_int32_t *hdr = (u_int32_t *)args->eh;
1884
1885                                         match =
1886                                             ( want[0] == (hdr[0] & mask[0]) &&
1887                                               want[1] == (hdr[1] & mask[1]) &&
1888                                               want[2] == (hdr[2] & mask[2]) );
1889                                 }
1890                                 break;
1891
1892                         case O_MAC_TYPE:
1893                                 if (args->flags & IPFW_ARGS_ETHER) {
1894                                         u_int16_t *p =
1895                                             ((ipfw_insn_u16 *)cmd)->ports;
1896                                         int i;
1897
1898                                         for (i = cmdlen - 1; !match && i>0;
1899                                             i--, p += 2)
1900                                                 match = (etype >= p[0] &&
1901                                                     etype <= p[1]);
1902                                 }
1903                                 break;
1904
1905                         case O_FRAG:
1906                                 match = (offset != 0);
1907                                 break;
1908
1909                         case O_IN:      /* "out" is "not in" */
1910                                 match = (oif == NULL);
1911                                 break;
1912
1913                         case O_LAYER2:
1914                                 match = (args->flags & IPFW_ARGS_ETHER);
1915                                 break;
1916
1917                         case O_DIVERTED:
1918                                 if ((args->flags & IPFW_ARGS_REF) == 0)
1919                                         break;
1920                                 /*
1921                                  * For diverted packets, args->rule.info
1922                                  * contains the divert port (in host format)
1923                                  * reason and direction.
1924                                  */
1925                                 match = ((args->rule.info & IPFW_IS_MASK) ==
1926                                     IPFW_IS_DIVERT) && (
1927                                     ((args->rule.info & IPFW_INFO_IN) ?
1928                                         1: 2) & cmd->arg1);
1929                                 break;
1930
1931                         case O_PROTO:
1932                                 /*
1933                                  * We do not allow an arg of 0 so the
1934                                  * check of "proto" only suffices.
1935                                  */
1936                                 match = (proto == cmd->arg1);
1937                                 break;
1938
1939                         case O_IP_SRC:
1940                                 match = is_ipv4 &&
1941                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1942                                     src_ip.s_addr);
1943                                 break;
1944
1945                         case O_IP_DST_LOOKUP:
1946                         {
1947                                 void *pkey;
1948                                 uint32_t vidx, key;
1949                                 uint16_t keylen;
1950
1951                                 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1952                                         /* Determine lookup key type */
1953                                         vidx = ((ipfw_insn_u32 *)cmd)->d[1];
1954                                         if (vidx != 4 /* uid */ &&
1955                                             vidx != 5 /* jail */ &&
1956                                             is_ipv6 == 0 && is_ipv4 == 0)
1957                                                 break;
1958                                         /* Determine key length */
1959                                         if (vidx == 0 /* dst-ip */ ||
1960                                             vidx == 1 /* src-ip */)
1961                                                 keylen = is_ipv6 ?
1962                                                     sizeof(struct in6_addr):
1963                                                     sizeof(in_addr_t);
1964                                         else {
1965                                                 keylen = sizeof(key);
1966                                                 pkey = &key;
1967                                         }
1968                                         if (vidx == 0 /* dst-ip */)
1969                                                 pkey = is_ipv4 ? (void *)&dst_ip:
1970                                                     (void *)&args->f_id.dst_ip6;
1971                                         else if (vidx == 1 /* src-ip */)
1972                                                 pkey = is_ipv4 ? (void *)&src_ip:
1973                                                     (void *)&args->f_id.src_ip6;
1974                                         else if (vidx == 6 /* dscp */) {
1975                                                 if (is_ipv4)
1976                                                         key = ip->ip_tos >> 2;
1977                                                 else {
1978                                                         key = args->f_id.flow_id6;
1979                                                         key = (key & 0x0f) << 2 |
1980                                                             (key & 0xf000) >> 14;
1981                                                 }
1982                                                 key &= 0x3f;
1983                                         } else if (vidx == 2 /* dst-port */ ||
1984                                             vidx == 3 /* src-port */) {
1985                                                 /* Skip fragments */
1986                                                 if (offset != 0)
1987                                                         break;
1988                                                 /* Skip proto without ports */
1989                                                 if (proto != IPPROTO_TCP &&
1990                                                     proto != IPPROTO_UDP &&
1991                                                     proto != IPPROTO_UDPLITE &&
1992                                                     proto != IPPROTO_SCTP)
1993                                                         break;
1994                                                 if (vidx == 2 /* dst-port */)
1995                                                         key = dst_port;
1996                                                 else
1997                                                         key = src_port;
1998                                         }
1999 #ifndef USERSPACE
2000                                         else if (vidx == 4 /* uid */ ||
2001                                             vidx == 5 /* jail */) {
2002                                                 check_uidgid(
2003                                                     (ipfw_insn_u32 *)cmd,
2004                                                     args, &ucred_lookup,
2005 #ifdef __FreeBSD__
2006                                                     &ucred_cache);
2007                                                 if (vidx == 4 /* uid */)
2008                                                         key = ucred_cache->cr_uid;
2009                                                 else if (vidx == 5 /* jail */)
2010                                                         key = ucred_cache->cr_prison->pr_id;
2011 #else /* !__FreeBSD__ */
2012                                                     (void *)&ucred_cache);
2013                                                 if (vidx == 4 /* uid */)
2014                                                         key = ucred_cache.uid;
2015                                                 else if (vidx == 5 /* jail */)
2016                                                         key = ucred_cache.xid;
2017 #endif /* !__FreeBSD__ */
2018                                         }
2019 #endif /* !USERSPACE */
2020                                         else
2021                                                 break;
2022                                         match = ipfw_lookup_table(chain,
2023                                             cmd->arg1, keylen, pkey, &vidx);
2024                                         if (!match)
2025                                                 break;
2026                                         tablearg = vidx;
2027                                         break;
2028                                 }
2029                                 /* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */
2030                                 /* FALLTHROUGH */
2031                         }
2032                         case O_IP_SRC_LOOKUP:
2033                         {
2034                                 void *pkey;
2035                                 uint32_t vidx;
2036                                 uint16_t keylen;
2037
2038                                 if (is_ipv4) {
2039                                         keylen = sizeof(in_addr_t);
2040                                         if (cmd->opcode == O_IP_DST_LOOKUP)
2041                                                 pkey = &dst_ip;
2042                                         else
2043                                                 pkey = &src_ip;
2044                                 } else if (is_ipv6) {
2045                                         keylen = sizeof(struct in6_addr);
2046                                         if (cmd->opcode == O_IP_DST_LOOKUP)
2047                                                 pkey = &args->f_id.dst_ip6;
2048                                         else
2049                                                 pkey = &args->f_id.src_ip6;
2050                                 } else
2051                                         break;
2052                                 match = ipfw_lookup_table(chain, cmd->arg1,
2053                                     keylen, pkey, &vidx);
2054                                 if (!match)
2055                                         break;
2056                                 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) {
2057                                         match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2058                                             TARG_VAL(chain, vidx, tag);
2059                                         if (!match)
2060                                                 break;
2061                                 }
2062                                 tablearg = vidx;
2063                                 break;
2064                         }
2065
2066                         case O_IP_FLOW_LOOKUP:
2067                                 {
2068                                         uint32_t v = 0;
2069                                         match = ipfw_lookup_table(chain,
2070                                             cmd->arg1, 0, &args->f_id, &v);
2071                                         if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
2072                                                 match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2073                                                     TARG_VAL(chain, v, tag);
2074                                         if (match)
2075                                                 tablearg = v;
2076                                 }
2077                                 break;
2078                         case O_IP_SRC_MASK:
2079                         case O_IP_DST_MASK:
2080                                 if (is_ipv4) {
2081                                     uint32_t a =
2082                                         (cmd->opcode == O_IP_DST_MASK) ?
2083                                             dst_ip.s_addr : src_ip.s_addr;
2084                                     uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2085                                     int i = cmdlen-1;
2086
2087                                     for (; !match && i>0; i-= 2, p+= 2)
2088                                         match = (p[0] == (a & p[1]));
2089                                 }
2090                                 break;
2091
2092                         case O_IP_SRC_ME:
2093                                 if (is_ipv4) {
2094                                         match = in_localip(src_ip);
2095                                         break;
2096                                 }
2097 #ifdef INET6
2098                                 /* FALLTHROUGH */
2099                         case O_IP6_SRC_ME:
2100                                 match = is_ipv6 &&
2101                                     ipfw_localip6(&args->f_id.src_ip6);
2102 #endif
2103                                 break;
2104
2105                         case O_IP_DST_SET:
2106                         case O_IP_SRC_SET:
2107                                 if (is_ipv4) {
2108                                         u_int32_t *d = (u_int32_t *)(cmd+1);
2109                                         u_int32_t addr =
2110                                             cmd->opcode == O_IP_DST_SET ?
2111                                                 args->f_id.dst_ip :
2112                                                 args->f_id.src_ip;
2113
2114                                             if (addr < d[0])
2115                                                     break;
2116                                             addr -= d[0]; /* subtract base */
2117                                             match = (addr < cmd->arg1) &&
2118                                                 ( d[ 1 + (addr>>5)] &
2119                                                   (1<<(addr & 0x1f)) );
2120                                 }
2121                                 break;
2122
2123                         case O_IP_DST:
2124                                 match = is_ipv4 &&
2125                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2126                                     dst_ip.s_addr);
2127                                 break;
2128
2129                         case O_IP_DST_ME:
2130                                 if (is_ipv4) {
2131                                         match = in_localip(dst_ip);
2132                                         break;
2133                                 }
2134 #ifdef INET6
2135                                 /* FALLTHROUGH */
2136                         case O_IP6_DST_ME:
2137                                 match = is_ipv6 &&
2138                                     ipfw_localip6(&args->f_id.dst_ip6);
2139 #endif
2140                                 break;
2141
2142
2143                         case O_IP_SRCPORT:
2144                         case O_IP_DSTPORT:
2145                                 /*
2146                                  * offset == 0 && proto != 0 is enough
2147                                  * to guarantee that we have a
2148                                  * packet with port info.
2149                                  */
2150                                 if ((proto == IPPROTO_UDP ||
2151                                     proto == IPPROTO_UDPLITE ||
2152                                     proto == IPPROTO_TCP ||
2153                                     proto == IPPROTO_SCTP) && offset == 0) {
2154                                         u_int16_t x =
2155                                             (cmd->opcode == O_IP_SRCPORT) ?
2156                                                 src_port : dst_port ;
2157                                         u_int16_t *p =
2158                                             ((ipfw_insn_u16 *)cmd)->ports;
2159                                         int i;
2160
2161                                         for (i = cmdlen - 1; !match && i>0;
2162                                             i--, p += 2)
2163                                                 match = (x>=p[0] && x<=p[1]);
2164                                 }
2165                                 break;
2166
2167                         case O_ICMPTYPE:
2168                                 match = (offset == 0 && proto==IPPROTO_ICMP &&
2169                                     icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
2170                                 break;
2171
2172 #ifdef INET6
2173                         case O_ICMP6TYPE:
2174                                 match = is_ipv6 && offset == 0 &&
2175                                     proto==IPPROTO_ICMPV6 &&
2176                                     icmp6type_match(
2177                                         ICMP6(ulp)->icmp6_type,
2178                                         (ipfw_insn_u32 *)cmd);
2179                                 break;
2180 #endif /* INET6 */
2181
2182                         case O_IPOPT:
2183                                 match = (is_ipv4 &&
2184                                     ipopts_match(ip, cmd) );
2185                                 break;
2186
2187                         case O_IPVER:
2188                                 match = (is_ipv4 &&
2189                                     cmd->arg1 == ip->ip_v);
2190                                 break;
2191
2192                         case O_IPID:
2193                         case O_IPTTL:
2194                                 if (!is_ipv4)
2195                                         break;
2196                         case O_IPLEN:
2197                                 {       /* only for IP packets */
2198                                     uint16_t x;
2199                                     uint16_t *p;
2200                                     int i;
2201
2202                                     if (cmd->opcode == O_IPLEN)
2203                                         x = iplen;
2204                                     else if (cmd->opcode == O_IPTTL)
2205                                         x = ip->ip_ttl;
2206                                     else /* must be IPID */
2207                                         x = ntohs(ip->ip_id);
2208                                     if (cmdlen == 1) {
2209                                         match = (cmd->arg1 == x);
2210                                         break;
2211                                     }
2212                                     /* otherwise we have ranges */
2213                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2214                                     i = cmdlen - 1;
2215                                     for (; !match && i>0; i--, p += 2)
2216                                         match = (x >= p[0] && x <= p[1]);
2217                                 }
2218                                 break;
2219
2220                         case O_IPPRECEDENCE:
2221                                 match = (is_ipv4 &&
2222                                     (cmd->arg1 == (ip->ip_tos & 0xe0)) );
2223                                 break;
2224
2225                         case O_IPTOS:
2226                                 match = (is_ipv4 &&
2227                                     flags_match(cmd, ip->ip_tos));
2228                                 break;
2229
2230                         case O_DSCP:
2231                             {
2232                                 uint32_t *p;
2233                                 uint16_t x;
2234
2235                                 p = ((ipfw_insn_u32 *)cmd)->d;
2236
2237                                 if (is_ipv4)
2238                                         x = ip->ip_tos >> 2;
2239                                 else if (is_ipv6) {
2240                                         uint8_t *v;
2241                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
2242                                         x = (*v & 0x0F) << 2;
2243                                         v++;
2244                                         x |= *v >> 6;
2245                                 } else
2246                                         break;
2247
2248                                 /* DSCP bitmask is stored as low_u32 high_u32 */
2249                                 if (x >= 32)
2250                                         match = *(p + 1) & (1 << (x - 32));
2251                                 else
2252                                         match = *p & (1 << x);
2253                             }
2254                                 break;
2255
2256                         case O_TCPDATALEN:
2257                                 if (proto == IPPROTO_TCP && offset == 0) {
2258                                     struct tcphdr *tcp;
2259                                     uint16_t x;
2260                                     uint16_t *p;
2261                                     int i;
2262 #ifdef INET6
2263                                     if (is_ipv6) {
2264                                             struct ip6_hdr *ip6;
2265
2266                                             ip6 = (struct ip6_hdr *)ip;
2267                                             if (ip6->ip6_plen == 0) {
2268                                                     /*
2269                                                      * Jumbo payload is not
2270                                                      * supported by this
2271                                                      * opcode.
2272                                                      */
2273                                                     break;
2274                                             }
2275                                             x = iplen - hlen;
2276                                     } else
2277 #endif /* INET6 */
2278                                             x = iplen - (ip->ip_hl << 2);
2279                                     tcp = TCP(ulp);
2280                                     x -= tcp->th_off << 2;
2281                                     if (cmdlen == 1) {
2282                                         match = (cmd->arg1 == x);
2283                                         break;
2284                                     }
2285                                     /* otherwise we have ranges */
2286                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2287                                     i = cmdlen - 1;
2288                                     for (; !match && i>0; i--, p += 2)
2289                                         match = (x >= p[0] && x <= p[1]);
2290                                 }
2291                                 break;
2292
2293                         case O_TCPFLAGS:
2294                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2295                                     flags_match(cmd, TCP(ulp)->th_flags));
2296                                 break;
2297
2298                         case O_TCPOPTS:
2299                                 if (proto == IPPROTO_TCP && offset == 0 && ulp){
2300                                         PULLUP_LEN_LOCKED(hlen, ulp,
2301                                             (TCP(ulp)->th_off << 2));
2302                                         match = tcpopts_match(TCP(ulp), cmd);
2303                                 }
2304                                 break;
2305
2306                         case O_TCPSEQ:
2307                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2308                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
2309                                         TCP(ulp)->th_seq);
2310                                 break;
2311
2312                         case O_TCPACK:
2313                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2314                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
2315                                         TCP(ulp)->th_ack);
2316                                 break;
2317
2318                         case O_TCPMSS:
2319                                 if (proto == IPPROTO_TCP &&
2320                                     (args->f_id._flags & TH_SYN) != 0 &&
2321                                     ulp != NULL) {
2322                                         uint16_t mss, *p;
2323                                         int i;
2324
2325                                         PULLUP_LEN_LOCKED(hlen, ulp,
2326                                             (TCP(ulp)->th_off << 2));
2327                                         if ((tcpopts_parse(TCP(ulp), &mss) &
2328                                             IP_FW_TCPOPT_MSS) == 0)
2329                                                 break;
2330                                         if (cmdlen == 1) {
2331                                                 match = (cmd->arg1 == mss);
2332                                                 break;
2333                                         }
2334                                         /* Otherwise we have ranges. */
2335                                         p = ((ipfw_insn_u16 *)cmd)->ports;
2336                                         i = cmdlen - 1;
2337                                         for (; !match && i > 0; i--, p += 2)
2338                                                 match = (mss >= p[0] &&
2339                                                     mss <= p[1]);
2340                                 }
2341                                 break;
2342
2343                         case O_TCPWIN:
2344                                 if (proto == IPPROTO_TCP && offset == 0) {
2345                                     uint16_t x;
2346                                     uint16_t *p;
2347                                     int i;
2348
2349                                     x = ntohs(TCP(ulp)->th_win);
2350                                     if (cmdlen == 1) {
2351                                         match = (cmd->arg1 == x);
2352                                         break;
2353                                     }
2354                                     /* Otherwise we have ranges. */
2355                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2356                                     i = cmdlen - 1;
2357                                     for (; !match && i > 0; i--, p += 2)
2358                                         match = (x >= p[0] && x <= p[1]);
2359                                 }
2360                                 break;
2361
2362                         case O_ESTAB:
2363                                 /* reject packets which have SYN only */
2364                                 /* XXX should i also check for TH_ACK ? */
2365                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2366                                     (TCP(ulp)->th_flags &
2367                                      (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
2368                                 break;
2369
2370                         case O_ALTQ: {
2371                                 struct pf_mtag *at;
2372                                 struct m_tag *mtag;
2373                                 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
2374
2375                                 /*
2376                                  * ALTQ uses mbuf tags from another
2377                                  * packet filtering system - pf(4).
2378                                  * We allocate a tag in its format
2379                                  * and fill it in, pretending to be pf(4).
2380                                  */
2381                                 match = 1;
2382                                 at = pf_find_mtag(m);
2383                                 if (at != NULL && at->qid != 0)
2384                                         break;
2385                                 mtag = m_tag_get(PACKET_TAG_PF,
2386                                     sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
2387                                 if (mtag == NULL) {
2388                                         /*
2389                                          * Let the packet fall back to the
2390                                          * default ALTQ.
2391                                          */
2392                                         break;
2393                                 }
2394                                 m_tag_prepend(m, mtag);
2395                                 at = (struct pf_mtag *)(mtag + 1);
2396                                 at->qid = altq->qid;
2397                                 at->hdr = ip;
2398                                 break;
2399                         }
2400
2401                         case O_LOG:
2402                                 ipfw_log(chain, f, hlen, args, m,
2403                                     oif, offset | ip6f_mf, tablearg, ip);
2404                                 match = 1;
2405                                 break;
2406
2407                         case O_PROB:
2408                                 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
2409                                 break;
2410
2411                         case O_VERREVPATH:
2412                                 /* Outgoing packets automatically pass/match */
2413                                 match = ((oif != NULL) ||
2414                                     (m->m_pkthdr.rcvif == NULL) ||
2415                                     (
2416 #ifdef INET6
2417                                     is_ipv6 ?
2418                                         verify_path6(&(args->f_id.src_ip6),
2419                                             m->m_pkthdr.rcvif, args->f_id.fib) :
2420 #endif
2421                                     verify_path(src_ip, m->m_pkthdr.rcvif,
2422                                         args->f_id.fib)));
2423                                 break;
2424
2425                         case O_VERSRCREACH:
2426                                 /* Outgoing packets automatically pass/match */
2427                                 match = (hlen > 0 && ((oif != NULL) || (
2428 #ifdef INET6
2429                                     is_ipv6 ?
2430                                         verify_path6(&(args->f_id.src_ip6),
2431                                             NULL, args->f_id.fib) :
2432 #endif
2433                                     verify_path(src_ip, NULL, args->f_id.fib))));
2434                                 break;
2435
2436                         case O_ANTISPOOF:
2437                                 /* Outgoing packets automatically pass/match */
2438                                 if (oif == NULL && hlen > 0 &&
2439                                     (  (is_ipv4 && in_localaddr(src_ip))
2440 #ifdef INET6
2441                                     || (is_ipv6 &&
2442                                         in6_localaddr(&(args->f_id.src_ip6)))
2443 #endif
2444                                     ))
2445                                         match =
2446 #ifdef INET6
2447                                             is_ipv6 ? verify_path6(
2448                                                 &(args->f_id.src_ip6),
2449                                                 m->m_pkthdr.rcvif,
2450                                                 args->f_id.fib) :
2451 #endif
2452                                             verify_path(src_ip,
2453                                                 m->m_pkthdr.rcvif,
2454                                                 args->f_id.fib);
2455                                 else
2456                                         match = 1;
2457                                 break;
2458
2459                         case O_IPSEC:
2460                                 match = (m_tag_find(m,
2461                                     PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
2462                                 /* otherwise no match */
2463                                 break;
2464
2465 #ifdef INET6
2466                         case O_IP6_SRC:
2467                                 match = is_ipv6 &&
2468                                     IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
2469                                     &((ipfw_insn_ip6 *)cmd)->addr6);
2470                                 break;
2471
2472                         case O_IP6_DST:
2473                                 match = is_ipv6 &&
2474                                 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
2475                                     &((ipfw_insn_ip6 *)cmd)->addr6);
2476                                 break;
2477                         case O_IP6_SRC_MASK:
2478                         case O_IP6_DST_MASK:
2479                                 if (is_ipv6) {
2480                                         int i = cmdlen - 1;
2481                                         struct in6_addr p;
2482                                         struct in6_addr *d =
2483                                             &((ipfw_insn_ip6 *)cmd)->addr6;
2484
2485                                         for (; !match && i > 0; d += 2,
2486                                             i -= F_INSN_SIZE(struct in6_addr)
2487                                             * 2) {
2488                                                 p = (cmd->opcode ==
2489                                                     O_IP6_SRC_MASK) ?
2490                                                     args->f_id.src_ip6:
2491                                                     args->f_id.dst_ip6;
2492                                                 APPLY_MASK(&p, &d[1]);
2493                                                 match =
2494                                                     IN6_ARE_ADDR_EQUAL(&d[0],
2495                                                     &p);
2496                                         }
2497                                 }
2498                                 break;
2499
2500                         case O_FLOW6ID:
2501                                 match = is_ipv6 &&
2502                                     flow6id_match(args->f_id.flow_id6,
2503                                     (ipfw_insn_u32 *) cmd);
2504                                 break;
2505
2506                         case O_EXT_HDR:
2507                                 match = is_ipv6 &&
2508                                     (ext_hd & ((ipfw_insn *) cmd)->arg1);
2509                                 break;
2510
2511                         case O_IP6:
2512                                 match = is_ipv6;
2513                                 break;
2514 #endif
2515
2516                         case O_IP4:
2517                                 match = is_ipv4;
2518                                 break;
2519
2520                         case O_TAG: {
2521                                 struct m_tag *mtag;
2522                                 uint32_t tag = TARG(cmd->arg1, tag);
2523
2524                                 /* Packet is already tagged with this tag? */
2525                                 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
2526
2527                                 /* We have `untag' action when F_NOT flag is
2528                                  * present. And we must remove this mtag from
2529                                  * mbuf and reset `match' to zero (`match' will
2530                                  * be inversed later).
2531                                  * Otherwise we should allocate new mtag and
2532                                  * push it into mbuf.
2533                                  */
2534                                 if (cmd->len & F_NOT) { /* `untag' action */
2535                                         if (mtag != NULL)
2536                                                 m_tag_delete(m, mtag);
2537                                         match = 0;
2538                                 } else {
2539                                         if (mtag == NULL) {
2540                                                 mtag = m_tag_alloc( MTAG_IPFW,
2541                                                     tag, 0, M_NOWAIT);
2542                                                 if (mtag != NULL)
2543                                                         m_tag_prepend(m, mtag);
2544                                         }
2545                                         match = 1;
2546                                 }
2547                                 break;
2548                         }
2549
2550                         case O_FIB: /* try match the specified fib */
2551                                 if (args->f_id.fib == cmd->arg1)
2552                                         match = 1;
2553                                 break;
2554
2555                         case O_SOCKARG: {
2556 #ifndef USERSPACE       /* not supported in userspace */
2557                                 struct inpcb *inp = args->inp;
2558                                 struct inpcbinfo *pi;
2559
2560                                 if (is_ipv6) /* XXX can we remove this ? */
2561                                         break;
2562
2563                                 if (proto == IPPROTO_TCP)
2564                                         pi = &V_tcbinfo;
2565                                 else if (proto == IPPROTO_UDP)
2566                                         pi = &V_udbinfo;
2567                                 else if (proto == IPPROTO_UDPLITE)
2568                                         pi = &V_ulitecbinfo;
2569                                 else
2570                                         break;
2571
2572                                 /*
2573                                  * XXXRW: so_user_cookie should almost
2574                                  * certainly be inp_user_cookie?
2575                                  */
2576
2577                                 /* For incoming packet, lookup up the
2578                                 inpcb using the src/dest ip/port tuple */
2579                                 if (inp == NULL) {
2580                                         inp = in_pcblookup(pi,
2581                                                 src_ip, htons(src_port),
2582                                                 dst_ip, htons(dst_port),
2583                                                 INPLOOKUP_RLOCKPCB, NULL);
2584                                         if (inp != NULL) {
2585                                                 tablearg =
2586                                                     inp->inp_socket->so_user_cookie;
2587                                                 if (tablearg)
2588                                                         match = 1;
2589                                                 INP_RUNLOCK(inp);
2590                                         }
2591                                 } else {
2592                                         if (inp->inp_socket) {
2593                                                 tablearg =
2594                                                     inp->inp_socket->so_user_cookie;
2595                                                 if (tablearg)
2596                                                         match = 1;
2597                                         }
2598                                 }
2599 #endif /* !USERSPACE */
2600                                 break;
2601                         }
2602
2603                         case O_TAGGED: {
2604                                 struct m_tag *mtag;
2605                                 uint32_t tag = TARG(cmd->arg1, tag);
2606
2607                                 if (cmdlen == 1) {
2608                                         match = m_tag_locate(m, MTAG_IPFW,
2609                                             tag, NULL) != NULL;
2610                                         break;
2611                                 }
2612
2613                                 /* we have ranges */
2614                                 for (mtag = m_tag_first(m);
2615                                     mtag != NULL && !match;
2616                                     mtag = m_tag_next(m, mtag)) {
2617                                         uint16_t *p;
2618                                         int i;
2619
2620                                         if (mtag->m_tag_cookie != MTAG_IPFW)
2621                                                 continue;
2622
2623                                         p = ((ipfw_insn_u16 *)cmd)->ports;
2624                                         i = cmdlen - 1;
2625                                         for(; !match && i > 0; i--, p += 2)
2626                                                 match =
2627                                                     mtag->m_tag_id >= p[0] &&
2628                                                     mtag->m_tag_id <= p[1];
2629                                 }
2630                                 break;
2631                         }
2632
2633                         /*
2634                          * The second set of opcodes represents 'actions',
2635                          * i.e. the terminal part of a rule once the packet
2636                          * matches all previous patterns.
2637                          * Typically there is only one action for each rule,
2638                          * and the opcode is stored at the end of the rule
2639                          * (but there are exceptions -- see below).
2640                          *
2641                          * In general, here we set retval and terminate the
2642                          * outer loop (would be a 'break 3' in some language,
2643                          * but we need to set l=0, done=1)
2644                          *
2645                          * Exceptions:
2646                          * O_COUNT and O_SKIPTO actions:
2647                          *   instead of terminating, we jump to the next rule
2648                          *   (setting l=0), or to the SKIPTO target (setting
2649                          *   f/f_len, cmd and l as needed), respectively.
2650                          *
2651                          * O_TAG, O_LOG and O_ALTQ action parameters:
2652                          *   perform some action and set match = 1;
2653                          *
2654                          * O_LIMIT and O_KEEP_STATE: these opcodes are
2655                          *   not real 'actions', and are stored right
2656                          *   before the 'action' part of the rule (one
2657                          *   exception is O_SKIP_ACTION which could be
2658                          *   between these opcodes and 'action' one).
2659                          *   These opcodes try to install an entry in the
2660                          *   state tables; if successful, we continue with
2661                          *   the next opcode (match=1; break;), otherwise
2662                          *   the packet must be dropped (set retval,
2663                          *   break loops with l=0, done=1)
2664                          *
2665                          * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2666                          *   cause a lookup of the state table, and a jump
2667                          *   to the 'action' part of the parent rule
2668                          *   if an entry is found, or
2669                          *   (CHECK_STATE only) a jump to the next rule if
2670                          *   the entry is not found.
2671                          *   The result of the lookup is cached so that
2672                          *   further instances of these opcodes become NOPs.
2673                          *   The jump to the next rule is done by setting
2674                          *   l=0, cmdlen=0.
2675                          *
2676                          * O_SKIP_ACTION: this opcode is not a real 'action'
2677                          *  either, and is stored right before the 'action'
2678                          *  part of the rule, right after the O_KEEP_STATE
2679                          *  opcode. It causes match failure so the real
2680                          *  'action' could be executed only if the rule
2681                          *  is checked via dynamic rule from the state
2682                          *  table, as in such case execution starts
2683                          *  from the true 'action' opcode directly.
2684                          *
2685                          */
2686                         case O_LIMIT:
2687                         case O_KEEP_STATE:
2688                                 if (ipfw_dyn_install_state(chain, f,
2689                                     (ipfw_insn_limit *)cmd, args, ulp,
2690                                     pktlen, &dyn_info, tablearg)) {
2691                                         /* error or limit violation */
2692                                         retval = IP_FW_DENY;
2693                                         l = 0;  /* exit inner loop */
2694                                         done = 1; /* exit outer loop */
2695                                 }
2696                                 match = 1;
2697                                 break;
2698
2699                         case O_PROBE_STATE:
2700                         case O_CHECK_STATE:
2701                                 /*
2702                                  * dynamic rules are checked at the first
2703                                  * keep-state or check-state occurrence,
2704                                  * with the result being stored in dyn_info.
2705                                  * The compiler introduces a PROBE_STATE
2706                                  * instruction for us when we have a
2707                                  * KEEP_STATE (because PROBE_STATE needs
2708                                  * to be run first).
2709                                  */
2710                                 if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) &&
2711                                     (q = ipfw_dyn_lookup_state(args, ulp,
2712                                     pktlen, cmd, &dyn_info)) != NULL) {
2713                                         /*
2714                                          * Found dynamic entry, jump to the
2715                                          * 'action' part of the parent rule
2716                                          * by setting f, cmd, l and clearing
2717                                          * cmdlen.
2718                                          */
2719                                         f = q;
2720                                         f_pos = dyn_info.f_pos;
2721                                         cmd = ACTION_PTR(f);
2722                                         l = f->cmd_len - f->act_ofs;
2723                                         cmdlen = 0;
2724                                         match = 1;
2725                                         break;
2726                                 }
2727                                 /*
2728                                  * Dynamic entry not found. If CHECK_STATE,
2729                                  * skip to next rule, if PROBE_STATE just
2730                                  * ignore and continue with next opcode.
2731                                  */
2732                                 if (cmd->opcode == O_CHECK_STATE)
2733                                         l = 0;  /* exit inner loop */
2734                                 match = 1;
2735                                 break;
2736
2737                         case O_SKIP_ACTION:
2738                                 match = 0;      /* skip to the next rule */
2739                                 l = 0;          /* exit inner loop */
2740                                 break;
2741
2742                         case O_ACCEPT:
2743                                 retval = 0;     /* accept */
2744                                 l = 0;          /* exit inner loop */
2745                                 done = 1;       /* exit outer loop */
2746                                 break;
2747
2748                         case O_PIPE:
2749                         case O_QUEUE:
2750                                 set_match(args, f_pos, chain);
2751                                 args->rule.info = TARG(cmd->arg1, pipe);
2752                                 if (cmd->opcode == O_PIPE)
2753                                         args->rule.info |= IPFW_IS_PIPE;
2754                                 if (V_fw_one_pass)
2755                                         args->rule.info |= IPFW_ONEPASS;
2756                                 retval = IP_FW_DUMMYNET;
2757                                 l = 0;          /* exit inner loop */
2758                                 done = 1;       /* exit outer loop */
2759                                 break;
2760
2761                         case O_DIVERT:
2762                         case O_TEE:
2763                                 if (args->flags & IPFW_ARGS_ETHER)
2764                                         break;  /* not on layer 2 */
2765                                 /* otherwise this is terminal */
2766                                 l = 0;          /* exit inner loop */
2767                                 done = 1;       /* exit outer loop */
2768                                 retval = (cmd->opcode == O_DIVERT) ?
2769                                         IP_FW_DIVERT : IP_FW_TEE;
2770                                 set_match(args, f_pos, chain);
2771                                 args->rule.info = TARG(cmd->arg1, divert);
2772                                 break;
2773
2774                         case O_COUNT:
2775                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2776                                 l = 0;          /* exit inner loop */
2777                                 break;
2778
2779                         case O_SKIPTO:
2780                             IPFW_INC_RULE_COUNTER(f, pktlen);
2781                             f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2782                             /*
2783                              * Skip disabled rules, and re-enter
2784                              * the inner loop with the correct
2785                              * f_pos, f, l and cmd.
2786                              * Also clear cmdlen and skip_or
2787                              */
2788                             for (; f_pos < chain->n_rules - 1 &&
2789                                     (V_set_disable &
2790                                      (1 << chain->map[f_pos]->set));
2791                                     f_pos++)
2792                                 ;
2793                             /* Re-enter the inner loop at the skipto rule. */
2794                             f = chain->map[f_pos];
2795                             l = f->cmd_len;
2796                             cmd = f->cmd;
2797                             match = 1;
2798                             cmdlen = 0;
2799                             skip_or = 0;
2800                             continue;
2801                             break;      /* not reached */
2802
2803                         case O_CALLRETURN: {
2804                                 /*
2805                                  * Implementation of `subroutine' call/return,
2806                                  * in the stack carried in an mbuf tag. This
2807                                  * is different from `skipto' in that any call
2808                                  * address is possible (`skipto' must prevent
2809                                  * backward jumps to avoid endless loops).
2810                                  * We have `return' action when F_NOT flag is
2811                                  * present. The `m_tag_id' field is used as
2812                                  * stack pointer.
2813                                  */
2814                                 struct m_tag *mtag;
2815                                 uint16_t jmpto, *stack;
2816
2817 #define IS_CALL         ((cmd->len & F_NOT) == 0)
2818 #define IS_RETURN       ((cmd->len & F_NOT) != 0)
2819                                 /*
2820                                  * Hand-rolled version of m_tag_locate() with
2821                                  * wildcard `type'.
2822                                  * If not already tagged, allocate new tag.
2823                                  */
2824                                 mtag = m_tag_first(m);
2825                                 while (mtag != NULL) {
2826                                         if (mtag->m_tag_cookie ==
2827                                             MTAG_IPFW_CALL)
2828                                                 break;
2829                                         mtag = m_tag_next(m, mtag);
2830                                 }
2831                                 if (mtag == NULL && IS_CALL) {
2832                                         mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2833                                             IPFW_CALLSTACK_SIZE *
2834                                             sizeof(uint16_t), M_NOWAIT);
2835                                         if (mtag != NULL)
2836                                                 m_tag_prepend(m, mtag);
2837                                 }
2838
2839                                 /*
2840                                  * On error both `call' and `return' just
2841                                  * continue with next rule.
2842                                  */
2843                                 if (IS_RETURN && (mtag == NULL ||
2844                                     mtag->m_tag_id == 0)) {
2845                                         l = 0;          /* exit inner loop */
2846                                         break;
2847                                 }
2848                                 if (IS_CALL && (mtag == NULL ||
2849                                     mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2850                                         printf("ipfw: call stack error, "
2851                                             "go to next rule\n");
2852                                         l = 0;          /* exit inner loop */
2853                                         break;
2854                                 }
2855
2856                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2857                                 stack = (uint16_t *)(mtag + 1);
2858
2859                                 /*
2860                                  * The `call' action may use cached f_pos
2861                                  * (in f->next_rule), whose version is written
2862                                  * in f->next_rule.
2863                                  * The `return' action, however, doesn't have
2864                                  * fixed jump address in cmd->arg1 and can't use
2865                                  * cache.
2866                                  */
2867                                 if (IS_CALL) {
2868                                         stack[mtag->m_tag_id] = f->rulenum;
2869                                         mtag->m_tag_id++;
2870                                         f_pos = JUMP(chain, f, cmd->arg1,
2871                                             tablearg, 1);
2872                                 } else {        /* `return' action */
2873                                         mtag->m_tag_id--;
2874                                         jmpto = stack[mtag->m_tag_id] + 1;
2875                                         f_pos = ipfw_find_rule(chain, jmpto, 0);
2876                                 }
2877
2878                                 /*
2879                                  * Skip disabled rules, and re-enter
2880                                  * the inner loop with the correct
2881                                  * f_pos, f, l and cmd.
2882                                  * Also clear cmdlen and skip_or
2883                                  */
2884                                 for (; f_pos < chain->n_rules - 1 &&
2885                                     (V_set_disable &
2886                                     (1 << chain->map[f_pos]->set)); f_pos++)
2887                                         ;
2888                                 /* Re-enter the inner loop at the dest rule. */
2889                                 f = chain->map[f_pos];
2890                                 l = f->cmd_len;
2891                                 cmd = f->cmd;
2892                                 cmdlen = 0;
2893                                 skip_or = 0;
2894                                 continue;
2895                                 break;  /* NOTREACHED */
2896                         }
2897 #undef IS_CALL
2898 #undef IS_RETURN
2899
2900                         case O_REJECT:
2901                                 /*
2902                                  * Drop the packet and send a reject notice
2903                                  * if the packet is not ICMP (or is an ICMP
2904                                  * query), and it is not multicast/broadcast.
2905                                  */
2906                                 if (hlen > 0 && is_ipv4 && offset == 0 &&
2907                                     (proto != IPPROTO_ICMP ||
2908                                      is_icmp_query(ICMP(ulp))) &&
2909                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2910                                     !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2911                                         send_reject(args, cmd->arg1, iplen, ip);
2912                                         m = args->m;
2913                                 }
2914                                 /* FALLTHROUGH */
2915 #ifdef INET6
2916                         case O_UNREACH6:
2917                                 if (hlen > 0 && is_ipv6 &&
2918                                     ((offset & IP6F_OFF_MASK) == 0) &&
2919                                     (proto != IPPROTO_ICMPV6 ||
2920                                      (is_icmp6_query(icmp6_type) == 1)) &&
2921                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2922                                     !IN6_IS_ADDR_MULTICAST(
2923                                         &args->f_id.dst_ip6)) {
2924                                         send_reject6(args,
2925                                             cmd->opcode == O_REJECT ?
2926                                             map_icmp_unreach(cmd->arg1):
2927                                             cmd->arg1, hlen,
2928                                             (struct ip6_hdr *)ip);
2929                                         m = args->m;
2930                                 }
2931                                 /* FALLTHROUGH */
2932 #endif
2933                         case O_DENY:
2934                                 retval = IP_FW_DENY;
2935                                 l = 0;          /* exit inner loop */
2936                                 done = 1;       /* exit outer loop */
2937                                 break;
2938
2939                         case O_FORWARD_IP:
2940                                 if (args->flags & IPFW_ARGS_ETHER)
2941                                         break;  /* not valid on layer2 pkts */
2942                                 if (q != f ||
2943                                     dyn_info.direction == MATCH_FORWARD) {
2944                                     struct sockaddr_in *sa;
2945
2946                                     sa = &(((ipfw_insn_sa *)cmd)->sa);
2947                                     if (sa->sin_addr.s_addr == INADDR_ANY) {
2948 #ifdef INET6
2949                                         /*
2950                                          * We use O_FORWARD_IP opcode for
2951                                          * fwd rule with tablearg, but tables
2952                                          * now support IPv6 addresses. And
2953                                          * when we are inspecting IPv6 packet,
2954                                          * we can use nh6 field from
2955                                          * table_value as next_hop6 address.
2956                                          */
2957                                         if (is_ipv6) {
2958                                                 struct ip_fw_nh6 *nh6;
2959
2960                                                 args->flags |= IPFW_ARGS_NH6;
2961                                                 nh6 = &args->hopstore6;
2962                                                 nh6->sin6_addr = TARG_VAL(
2963                                                     chain, tablearg, nh6);
2964                                                 nh6->sin6_port = sa->sin_port;
2965                                                 nh6->sin6_scope_id = TARG_VAL(
2966                                                     chain, tablearg, zoneid);
2967                                         } else
2968 #endif
2969                                         {
2970                                                 args->flags |= IPFW_ARGS_NH4;
2971                                                 args->hopstore.sin_port =
2972                                                     sa->sin_port;
2973                                                 sa = &args->hopstore;
2974                                                 sa->sin_family = AF_INET;
2975                                                 sa->sin_len = sizeof(*sa);
2976                                                 sa->sin_addr.s_addr = htonl(
2977                                                     TARG_VAL(chain, tablearg,
2978                                                     nh4));
2979                                         }
2980                                     } else {
2981                                             args->flags |= IPFW_ARGS_NH4PTR;
2982                                             args->next_hop = sa;
2983                                     }
2984                                 }
2985                                 retval = IP_FW_PASS;
2986                                 l = 0;          /* exit inner loop */
2987                                 done = 1;       /* exit outer loop */
2988                                 break;
2989
2990 #ifdef INET6
2991                         case O_FORWARD_IP6:
2992                                 if (args->flags & IPFW_ARGS_ETHER)
2993                                         break;  /* not valid on layer2 pkts */
2994                                 if (q != f ||
2995                                     dyn_info.direction == MATCH_FORWARD) {
2996                                         struct sockaddr_in6 *sin6;
2997
2998                                         sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2999                                         args->flags |= IPFW_ARGS_NH6PTR;
3000                                         args->next_hop6 = sin6;
3001                                 }
3002                                 retval = IP_FW_PASS;
3003                                 l = 0;          /* exit inner loop */
3004                                 done = 1;       /* exit outer loop */
3005                                 break;
3006 #endif
3007
3008                         case O_NETGRAPH:
3009                         case O_NGTEE:
3010                                 set_match(args, f_pos, chain);
3011                                 args->rule.info = TARG(cmd->arg1, netgraph);
3012                                 if (V_fw_one_pass)
3013                                         args->rule.info |= IPFW_ONEPASS;
3014                                 retval = (cmd->opcode == O_NETGRAPH) ?
3015                                     IP_FW_NETGRAPH : IP_FW_NGTEE;
3016                                 l = 0;          /* exit inner loop */
3017                                 done = 1;       /* exit outer loop */
3018                                 break;
3019
3020                         case O_SETFIB: {
3021                                 uint32_t fib;
3022
3023                                 IPFW_INC_RULE_COUNTER(f, pktlen);
3024                                 fib = TARG(cmd->arg1, fib) & 0x7FFF;
3025                                 if (fib >= rt_numfibs)
3026                                         fib = 0;
3027                                 M_SETFIB(m, fib);
3028                                 args->f_id.fib = fib; /* XXX */
3029                                 l = 0;          /* exit inner loop */
3030                                 break;
3031                         }
3032
3033                         case O_SETDSCP: {
3034                                 uint16_t code;
3035
3036                                 code = TARG(cmd->arg1, dscp) & 0x3F;
3037                                 l = 0;          /* exit inner loop */
3038                                 if (is_ipv4) {
3039                                         uint16_t old;
3040
3041                                         old = *(uint16_t *)ip;
3042                                         ip->ip_tos = (code << 2) |
3043                                             (ip->ip_tos & 0x03);
3044                                         ip->ip_sum = cksum_adjust(ip->ip_sum,
3045                                             old, *(uint16_t *)ip);
3046                                 } else if (is_ipv6) {
3047                                         uint8_t *v;
3048
3049                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
3050                                         *v = (*v & 0xF0) | (code >> 2);
3051                                         v++;
3052                                         *v = (*v & 0x3F) | ((code & 0x03) << 6);
3053                                 } else
3054                                         break;
3055
3056                                 IPFW_INC_RULE_COUNTER(f, pktlen);
3057                                 break;
3058                         }
3059
3060                         case O_NAT:
3061                                 l = 0;          /* exit inner loop */
3062                                 done = 1;       /* exit outer loop */
3063                                 /*
3064                                  * Ensure that we do not invoke NAT handler for
3065                                  * non IPv4 packets. Libalias expects only IPv4.
3066                                  */
3067                                 if (!is_ipv4 || !IPFW_NAT_LOADED) {
3068                                     retval = IP_FW_DENY;
3069                                     break;
3070                                 }
3071
3072                                 struct cfg_nat *t;
3073                                 int nat_id;
3074
3075                                 args->rule.info = 0;
3076                                 set_match(args, f_pos, chain);
3077                                 /* Check if this is 'global' nat rule */
3078                                 if (cmd->arg1 == IP_FW_NAT44_GLOBAL) {
3079                                         retval = ipfw_nat_ptr(args, NULL, m);
3080                                         break;
3081                                 }
3082                                 t = ((ipfw_insn_nat *)cmd)->nat;
3083                                 if (t == NULL) {
3084                                         nat_id = TARG(cmd->arg1, nat);
3085                                         t = (*lookup_nat_ptr)(&chain->nat, nat_id);
3086
3087                                         if (t == NULL) {
3088                                             retval = IP_FW_DENY;
3089                                             break;
3090                                         }
3091                                         if (cmd->arg1 != IP_FW_TARG)
3092                                             ((ipfw_insn_nat *)cmd)->nat = t;
3093                                 }
3094                                 retval = ipfw_nat_ptr(args, t, m);
3095                                 break;
3096
3097                         case O_REASS: {
3098                                 int ip_off;
3099
3100                                 l = 0;  /* in any case exit inner loop */
3101                                 if (is_ipv6) /* IPv6 is not supported yet */
3102                                         break;
3103                                 IPFW_INC_RULE_COUNTER(f, pktlen);
3104                                 ip_off = ntohs(ip->ip_off);
3105
3106                                 /* if not fragmented, go to next rule */
3107                                 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
3108                                     break;
3109
3110                                 args->m = m = ip_reass(m);
3111
3112                                 /*
3113                                  * do IP header checksum fixup.
3114                                  */
3115                                 if (m == NULL) { /* fragment got swallowed */
3116                                     retval = IP_FW_DENY;
3117                                 } else { /* good, packet complete */
3118                                     int hlen;
3119
3120                                     ip = mtod(m, struct ip *);
3121                                     hlen = ip->ip_hl << 2;
3122                                     ip->ip_sum = 0;
3123                                     if (hlen == sizeof(struct ip))
3124                                         ip->ip_sum = in_cksum_hdr(ip);
3125                                     else
3126                                         ip->ip_sum = in_cksum(m, hlen);
3127                                     retval = IP_FW_REASS;
3128                                     args->rule.info = 0;
3129                                     set_match(args, f_pos, chain);
3130                                 }
3131                                 done = 1;       /* exit outer loop */
3132                                 break;
3133                         }
3134                         case O_EXTERNAL_ACTION:
3135                                 l = 0; /* in any case exit inner loop */
3136                                 retval = ipfw_run_eaction(chain, args,
3137                                     cmd, &done);
3138                                 /*
3139                                  * If both @retval and @done are zero,
3140                                  * consider this as rule matching and
3141                                  * update counters.
3142                                  */
3143                                 if (retval == 0 && done == 0) {
3144                                         IPFW_INC_RULE_COUNTER(f, pktlen);
3145                                         /*
3146                                          * Reset the result of the last
3147                                          * dynamic state lookup.
3148                                          * External action can change
3149                                          * @args content, and it may be
3150                                          * used for new state lookup later.
3151                                          */
3152                                         DYN_INFO_INIT(&dyn_info);
3153                                 }
3154                                 break;
3155
3156                         default:
3157                                 panic("-- unknown opcode %d\n", cmd->opcode);
3158                         } /* end of switch() on opcodes */
3159                         /*
3160                          * if we get here with l=0, then match is irrelevant.
3161                          */
3162
3163                         if (cmd->len & F_NOT)
3164                                 match = !match;
3165
3166                         if (match) {
3167                                 if (cmd->len & F_OR)
3168                                         skip_or = 1;
3169                         } else {
3170                                 if (!(cmd->len & F_OR)) /* not an OR block, */
3171                                         break;          /* try next rule    */
3172                         }
3173
3174                 }       /* end of inner loop, scan opcodes */
3175 #undef PULLUP_LEN
3176 #undef PULLUP_LEN_LOCKED
3177
3178                 if (done)
3179                         break;
3180
3181 /* next_rule:; */       /* try next rule                */
3182
3183         }               /* end of outer for, scan rules */
3184
3185         if (done) {
3186                 struct ip_fw *rule = chain->map[f_pos];
3187                 /* Update statistics */
3188                 IPFW_INC_RULE_COUNTER(rule, pktlen);
3189         } else {
3190                 retval = IP_FW_DENY;
3191                 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3192         }
3193         IPFW_PF_RUNLOCK(chain);
3194 #ifdef __FreeBSD__
3195         if (ucred_cache != NULL)
3196                 crfree(ucred_cache);
3197 #endif
3198         return (retval);
3199
3200 pullup_failed:
3201         if (V_fw_verbose)
3202                 printf("ipfw: pullup failed\n");
3203         return (IP_FW_DENY);
3204 }
3205
3206 /*
3207  * Set maximum number of tables that can be used in given VNET ipfw instance.
3208  */
3209 #ifdef SYSCTL_NODE
3210 static int
3211 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
3212 {
3213         int error;
3214         unsigned int ntables;
3215
3216         ntables = V_fw_tables_max;
3217
3218         error = sysctl_handle_int(oidp, &ntables, 0, req);
3219         /* Read operation or some error */
3220         if ((error != 0) || (req->newptr == NULL))
3221                 return (error);
3222
3223         return (ipfw_resize_tables(&V_layer3_chain, ntables));
3224 }
3225
3226 /*
3227  * Switches table namespace between global and per-set.
3228  */
3229 static int
3230 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
3231 {
3232         int error;
3233         unsigned int sets;
3234
3235         sets = V_fw_tables_sets;
3236
3237         error = sysctl_handle_int(oidp, &sets, 0, req);
3238         /* Read operation or some error */
3239         if ((error != 0) || (req->newptr == NULL))
3240                 return (error);
3241
3242         return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
3243 }
3244 #endif
3245
3246 /*
3247  * Module and VNET glue
3248  */
3249
3250 /*
3251  * Stuff that must be initialised only on boot or module load
3252  */
3253 static int
3254 ipfw_init(void)
3255 {
3256         int error = 0;
3257
3258         /*
3259          * Only print out this stuff the first time around,
3260          * when called from the sysinit code.
3261          */
3262         printf("ipfw2 "
3263 #ifdef INET6
3264                 "(+ipv6) "
3265 #endif
3266                 "initialized, divert %s, nat %s, "
3267                 "default to %s, logging ",
3268 #ifdef IPDIVERT
3269                 "enabled",
3270 #else
3271                 "loadable",
3272 #endif
3273 #ifdef IPFIREWALL_NAT
3274                 "enabled",
3275 #else
3276                 "loadable",
3277 #endif
3278                 default_to_accept ? "accept" : "deny");
3279
3280         /*
3281          * Note: V_xxx variables can be accessed here but the vnet specific
3282          * initializer may not have been called yet for the VIMAGE case.
3283          * Tuneables will have been processed. We will print out values for
3284          * the default vnet.
3285          * XXX This should all be rationalized AFTER 8.0
3286          */
3287         if (V_fw_verbose == 0)
3288                 printf("disabled\n");
3289         else if (V_verbose_limit == 0)
3290                 printf("unlimited\n");
3291         else
3292                 printf("limited to %d packets/entry by default\n",
3293                     V_verbose_limit);
3294
3295         /* Check user-supplied table count for validness */
3296         if (default_fw_tables > IPFW_TABLES_MAX)
3297           default_fw_tables = IPFW_TABLES_MAX;
3298
3299         ipfw_init_sopt_handler();
3300         ipfw_init_obj_rewriter();
3301         ipfw_iface_init();
3302         return (error);
3303 }
3304
3305 /*
3306  * Called for the removal of the last instance only on module unload.
3307  */
3308 static void
3309 ipfw_destroy(void)
3310 {
3311
3312         ipfw_iface_destroy();
3313         ipfw_destroy_sopt_handler();
3314         ipfw_destroy_obj_rewriter();
3315         printf("IP firewall unloaded\n");
3316 }
3317
3318 /*
3319  * Stuff that must be initialized for every instance
3320  * (including the first of course).
3321  */
3322 static int
3323 vnet_ipfw_init(const void *unused)
3324 {
3325         int error, first;
3326         struct ip_fw *rule = NULL;
3327         struct ip_fw_chain *chain;
3328
3329         chain = &V_layer3_chain;
3330
3331         first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3332
3333         /* First set up some values that are compile time options */
3334         V_autoinc_step = 100;   /* bounded to 1..1000 in add_rule() */
3335         V_fw_deny_unknown_exthdrs = 1;
3336 #ifdef IPFIREWALL_VERBOSE
3337         V_fw_verbose = 1;
3338 #endif
3339 #ifdef IPFIREWALL_VERBOSE_LIMIT
3340         V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
3341 #endif
3342 #ifdef IPFIREWALL_NAT
3343         LIST_INIT(&chain->nat);
3344 #endif
3345
3346         /* Init shared services hash table */
3347         ipfw_init_srv(chain);
3348
3349         ipfw_init_counters();
3350         /* Set initial number of tables */
3351         V_fw_tables_max = default_fw_tables;
3352         error = ipfw_init_tables(chain, first);
3353         if (error) {
3354                 printf("ipfw2: setting up tables failed\n");
3355                 free(chain->map, M_IPFW);
3356                 free(rule, M_IPFW);
3357                 return (ENOSPC);
3358         }
3359
3360         IPFW_LOCK_INIT(chain);
3361
3362         /* fill and insert the default rule */
3363         rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
3364         rule->flags |= IPFW_RULE_NOOPT;
3365         rule->cmd_len = 1;
3366         rule->cmd[0].len = 1;
3367         rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
3368         chain->default_rule = rule;
3369         ipfw_add_protected_rule(chain, rule, 0);
3370
3371         ipfw_dyn_init(chain);
3372         ipfw_eaction_init(chain, first);
3373 #ifdef LINEAR_SKIPTO
3374         ipfw_init_skipto_cache(chain);
3375 #endif
3376         ipfw_bpf_init(first);
3377
3378         /* First set up some values that are compile time options */
3379         V_ipfw_vnet_ready = 1;          /* Open for business */
3380
3381         /*
3382          * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
3383          * Even if the latter two fail we still keep the module alive
3384          * because the sockopt and layer2 paths are still useful.
3385          * ipfw[6]_hook return 0 on success, ENOENT on failure,
3386          * so we can ignore the exact return value and just set a flag.
3387          *
3388          * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
3389          * changes in the underlying (per-vnet) variables trigger
3390          * immediate hook()/unhook() calls.
3391          * In layer2 we have the same behaviour, except that V_ether_ipfw
3392          * is checked on each packet because there are no pfil hooks.
3393          */
3394         V_ip_fw_ctl_ptr = ipfw_ctl3;
3395         error = ipfw_attach_hooks(1);
3396         return (error);
3397 }
3398
3399 /*
3400  * Called for the removal of each instance.
3401  */
3402 static int
3403 vnet_ipfw_uninit(const void *unused)
3404 {
3405         struct ip_fw *reap;
3406         struct ip_fw_chain *chain = &V_layer3_chain;
3407         int i, last;
3408
3409         V_ipfw_vnet_ready = 0; /* tell new callers to go away */
3410         /*
3411          * disconnect from ipv4, ipv6, layer2 and sockopt.
3412          * Then grab, release and grab again the WLOCK so we make
3413          * sure the update is propagated and nobody will be in.
3414          */
3415         (void)ipfw_attach_hooks(0 /* detach */);
3416         V_ip_fw_ctl_ptr = NULL;
3417
3418         last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3419
3420         IPFW_UH_WLOCK(chain);
3421         IPFW_UH_WUNLOCK(chain);
3422
3423         ipfw_dyn_uninit(0);     /* run the callout_drain */
3424
3425         IPFW_UH_WLOCK(chain);
3426
3427         reap = NULL;
3428         IPFW_WLOCK(chain);
3429         for (i = 0; i < chain->n_rules; i++)
3430                 ipfw_reap_add(chain, &reap, chain->map[i]);
3431         free(chain->map, M_IPFW);
3432 #ifdef LINEAR_SKIPTO
3433         ipfw_destroy_skipto_cache(chain);
3434 #endif
3435         IPFW_WUNLOCK(chain);
3436         IPFW_UH_WUNLOCK(chain);
3437         ipfw_destroy_tables(chain, last);
3438         ipfw_eaction_uninit(chain, last);
3439         if (reap != NULL)
3440                 ipfw_reap_rules(reap);
3441         vnet_ipfw_iface_destroy(chain);
3442         ipfw_destroy_srv(chain);
3443         IPFW_LOCK_DESTROY(chain);
3444         ipfw_dyn_uninit(1);     /* free the remaining parts */
3445         ipfw_destroy_counters();
3446         ipfw_bpf_uninit(last);
3447         return (0);
3448 }
3449
3450 /*
3451  * Module event handler.
3452  * In general we have the choice of handling most of these events by the
3453  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
3454  * use the SYSINIT handlers as they are more capable of expressing the
3455  * flow of control during module and vnet operations, so this is just
3456  * a skeleton. Note there is no SYSINIT equivalent of the module
3457  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
3458  */
3459 static int
3460 ipfw_modevent(module_t mod, int type, void *unused)
3461 {
3462         int err = 0;
3463
3464         switch (type) {
3465         case MOD_LOAD:
3466                 /* Called once at module load or
3467                  * system boot if compiled in. */
3468                 break;
3469         case MOD_QUIESCE:
3470                 /* Called before unload. May veto unloading. */
3471                 break;
3472         case MOD_UNLOAD:
3473                 /* Called during unload. */
3474                 break;
3475         case MOD_SHUTDOWN:
3476                 /* Called during system shutdown. */
3477                 break;
3478         default:
3479                 err = EOPNOTSUPP;
3480                 break;
3481         }
3482         return err;
3483 }
3484
3485 static moduledata_t ipfwmod = {
3486         "ipfw",
3487         ipfw_modevent,
3488         0
3489 };
3490
3491 /* Define startup order. */
3492 #define IPFW_SI_SUB_FIREWALL    SI_SUB_PROTO_FIREWALL
3493 #define IPFW_MODEVENT_ORDER     (SI_ORDER_ANY - 255) /* On boot slot in here. */
3494 #define IPFW_MODULE_ORDER       (IPFW_MODEVENT_ORDER + 1) /* A little later. */
3495 #define IPFW_VNET_ORDER         (IPFW_MODEVENT_ORDER + 2) /* Later still. */
3496
3497 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
3498 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
3499 MODULE_VERSION(ipfw, 3);
3500 /* should declare some dependencies here */
3501
3502 /*
3503  * Starting up. Done in order after ipfwmod() has been called.
3504  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
3505  */
3506 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3507             ipfw_init, NULL);
3508 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3509             vnet_ipfw_init, NULL);
3510
3511 /*
3512  * Closing up shop. These are done in REVERSE ORDER, but still
3513  * after ipfwmod() has been called. Not called on reboot.
3514  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
3515  * or when the module is unloaded.
3516  */
3517 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3518             ipfw_destroy, NULL);
3519 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3520             vnet_ipfw_uninit, NULL);
3521 /* end of file */