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 | CTLFLAG_MPSAFE, 0,
 183     "Firewall");
 184 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
 185     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
 186     "Only do a single pass through ipfw when using dummynet(4)");
 187 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
 188     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
 189     "Rule number auto-increment step");
 190 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose,
 191     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
 192     "Log matches to ipfw rules");
 193 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
 194     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
 195     "Set upper limit of matches of ipfw rules logged");
 196 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
 197     &dummy_def, 0,
 198     "The default/max possible rule number.");
 199 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
 200     CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE,
 201     0, 0, sysctl_ipfw_table_num, "IU",
 202     "Maximum number of concurrently used tables");
 203 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets,
 204     CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE,
 205     0, 0, sysctl_ipfw_tables_sets, "IU",
 206     "Use per-set namespace for tables");
 207 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
 208     &default_to_accept, 0,
 209     "Make the default rule accept all packets.");
 210 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables);
 211 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count,
 212     CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
 213     "Number of static rules");
 214
 215 #ifdef INET6
 216 SYSCTL_DECL(_net_inet6_ip6);
 217 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
 218     "Firewall");
 219 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
 220     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
 221     &VNET_NAME(fw_deny_unknown_exthdrs), 0,
 222     "Deny packets with unknown IPv6 Extension Headers");
 223 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
 224     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
 225     &VNET_NAME(fw_permit_single_frag6), 0,
 226     "Permit single packet IPv6 fragments");
 227 #endif /* INET6 */
 228
 229 SYSEND
 230
 231 #endif /* SYSCTL_NODE */
 232
 233
 234 /*
 235  * Some macros used in the various matching options.
 236  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
 237  * Other macros just cast void * into the appropriate type
 238  */
 239 #define L3HDR(T, ip)    ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
 240 #define TCP(p)          ((struct tcphdr *)(p))
 241 #define SCTP(p)         ((struct sctphdr *)(p))
 242 #define UDP(p)          ((struct udphdr *)(p))
 243 #define ICMP(p)         ((struct icmphdr *)(p))
 244 #define ICMP6(p)        ((struct icmp6_hdr *)(p))
 245
 246 static __inline int
 247 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
 248 {
 249         int type = icmp->icmp_type;
 250
 251         return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
 252 }
 253
 254 #define TT      ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
 255     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
 256
 257 static int
 258 is_icmp_query(struct icmphdr *icmp)
 259 {
 260         int type = icmp->icmp_type;
 261
 262         return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
 263 }
 264 #undef TT
 265
 266 /*
 267  * The following checks use two arrays of 8 or 16 bits to store the
 268  * bits that we want set or clear, respectively. They are in the
 269  * low and high half of cmd->arg1 or cmd->d[0].
 270  *
 271  * We scan options and store the bits we find set. We succeed if
 272  *
 273  *      (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
 274  *
 275  * The code is sometimes optimized not to store additional variables.
 276  */
 277
 278 static int
 279 flags_match(ipfw_insn *cmd, u_int8_t bits)
 280 {
 281         u_char want_clear;
 282         bits = ~bits;
 283
 284         if ( ((cmd->arg1 & 0xff) & bits) != 0)
 285                 return 0; /* some bits we want set were clear */
 286         want_clear = (cmd->arg1 >> 8) & 0xff;
 287         if ( (want_clear & bits) != want_clear)
 288                 return 0; /* some bits we want clear were set */
 289         return 1;
 290 }
 291
 292 static int
 293 ipopts_match(struct ip *ip, ipfw_insn *cmd)
 294 {
 295         int optlen, bits = 0;
 296         u_char *cp = (u_char *)(ip + 1);
 297         int x = (ip->ip_hl << 2) - sizeof (struct ip);
 298
 299         for (; x > 0; x -= optlen, cp += optlen) {
 300                 int opt = cp[IPOPT_OPTVAL];
 301
 302                 if (opt == IPOPT_EOL)
 303                         break;
 304                 if (opt == IPOPT_NOP)
 305                         optlen = 1;
 306                 else {
 307                         optlen = cp[IPOPT_OLEN];
 308                         if (optlen <= 0 || optlen > x)
 309                                 return 0; /* invalid or truncated */
 310                 }
 311                 switch (opt) {
 312
 313                 default:
 314                         break;
 315
 316                 case IPOPT_LSRR:
 317                         bits |= IP_FW_IPOPT_LSRR;
 318                         break;
 319
 320                 case IPOPT_SSRR:
 321                         bits |= IP_FW_IPOPT_SSRR;
 322                         break;
 323
 324                 case IPOPT_RR:
 325                         bits |= IP_FW_IPOPT_RR;
 326                         break;
 327
 328                 case IPOPT_TS:
 329                         bits |= IP_FW_IPOPT_TS;
 330                         break;
 331                 }
 332         }
 333         return (flags_match(cmd, bits));
 334 }
 335
 336 /*
 337  * Parse TCP options. The logic copied from tcp_dooptions().
 338  */
 339 static int
 340 tcpopts_parse(const struct tcphdr *tcp, uint16_t *mss)
 341 {
 342         const u_char *cp = (const u_char *)(tcp + 1);
 343         int optlen, bits = 0;
 344         int cnt = (tcp->th_off << 2) - sizeof(struct tcphdr);
 345
 346         for (; cnt > 0; cnt -= optlen, cp += optlen) {
 347                 int opt = cp[0];
 348                 if (opt == TCPOPT_EOL)
 349                         break;
 350                 if (opt == TCPOPT_NOP)
 351                         optlen = 1;
 352                 else {
 353                         if (cnt < 2)
 354                                 break;
 355                         optlen = cp[1];
 356                         if (optlen < 2 || optlen > cnt)
 357                                 break;
 358                 }
 359
 360                 switch (opt) {
 361                 default:
 362                         break;
 363
 364                 case TCPOPT_MAXSEG:
 365                         if (optlen != TCPOLEN_MAXSEG)
 366                                 break;
 367                         bits |= IP_FW_TCPOPT_MSS;
 368                         if (mss != NULL)
 369                                 *mss = be16dec(cp + 2);
 370                         break;
 371
 372                 case TCPOPT_WINDOW:
 373                         if (optlen == TCPOLEN_WINDOW)
 374                                 bits |= IP_FW_TCPOPT_WINDOW;
 375                         break;
 376
 377                 case TCPOPT_SACK_PERMITTED:
 378                         if (optlen == TCPOLEN_SACK_PERMITTED)
 379                                 bits |= IP_FW_TCPOPT_SACK;
 380                         break;
 381
 382                 case TCPOPT_SACK:
 383                         if (optlen > 2 && (optlen - 2) % TCPOLEN_SACK == 0)
 384                                 bits |= IP_FW_TCPOPT_SACK;
 385                         break;
 386
 387                 case TCPOPT_TIMESTAMP:
 388                         if (optlen == TCPOLEN_TIMESTAMP)
 389                                 bits |= IP_FW_TCPOPT_TS;
 390                         break;
 391                 }
 392         }
 393         return (bits);
 394 }
 395
 396 static int
 397 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
 398 {
 399
 400         return (flags_match(cmd, tcpopts_parse(tcp, NULL)));
 401 }
 402
 403 static int
 404 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain,
 405     uint32_t *tablearg)
 406 {
 407
 408         if (ifp == NULL)        /* no iface with this packet, match fails */
 409                 return (0);
 410
 411         /* Check by name or by IP address */
 412         if (cmd->name[0] != '\0') { /* match by name */
 413                 if (cmd->name[0] == '\1') /* use tablearg to match */
 414                         return ipfw_lookup_table(chain, cmd->p.kidx, 0,
 415                             &ifp->if_index, tablearg);
 416                 /* Check name */
 417                 if (cmd->p.glob) {
 418                         if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
 419                                 return(1);
 420                 } else {
 421                         if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
 422                                 return(1);
 423                 }
 424         } else {
 425 #if !defined(USERSPACE) && defined(__FreeBSD__) /* and OSX too ? */
 426                 struct ifaddr *ia;
 427
 428                 NET_EPOCH_ASSERT();
 429
 430                 CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
 431                         if (ia->ifa_addr->sa_family != AF_INET)
 432                                 continue;
 433                         if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
 434                             (ia->ifa_addr))->sin_addr.s_addr)
 435                                 return (1);     /* match */
 436                 }
 437 #endif /* __FreeBSD__ */
 438         }
 439         return(0);      /* no match, fail ... */
 440 }
 441
 442 /*
 443  * The verify_path function checks if a route to the src exists and
 444  * if it is reachable via ifp (when provided).
 445  *
 446  * The 'verrevpath' option checks that the interface that an IP packet
 447  * arrives on is the same interface that traffic destined for the
 448  * packet's source address would be routed out of.
 449  * The 'versrcreach' option just checks that the source address is
 450  * reachable via any route (except default) in the routing table.
 451  * These two are a measure to block forged packets. This is also
 452  * commonly known as "anti-spoofing" or Unicast Reverse Path
 453  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
 454  * is purposely reminiscent of the Cisco IOS command,
 455  *
 456  *   ip verify unicast reverse-path
 457  *   ip verify unicast source reachable-via any
 458  *
 459  * which implements the same functionality. But note that the syntax
 460  * is misleading, and the check may be performed on all IP packets
 461  * whether unicast, multicast, or broadcast.
 462  */
 463 static int
 464 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
 465 {
 466 #if defined(USERSPACE) || !defined(__FreeBSD__)
 467         return 0;
 468 #else
 469         struct nhop4_basic nh4;
 470
 471         if (fib4_lookup_nh_basic(fib, src, NHR_IFAIF, 0, &nh4) != 0)
 472                 return (0);
 473
 474         /*
 475          * If ifp is provided, check for equality with rtentry.
 476          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 477          * in order to pass packets injected back by if_simloop():
 478          * routing entry (via lo0) for our own address
 479          * may exist, so we need to handle routing assymetry.
 480          */
 481         if (ifp != NULL && ifp != nh4.nh_ifp)
 482                 return (0);
 483
 484         /* if no ifp provided, check if rtentry is not default route */
 485         if (ifp == NULL && (nh4.nh_flags & NHF_DEFAULT) != 0)
 486                 return (0);
 487
 488         /* or if this is a blackhole/reject route */
 489         if (ifp == NULL && (nh4.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
 490                 return (0);
 491
 492         /* found valid route */
 493         return 1;
 494 #endif /* __FreeBSD__ */
 495 }
 496
 497 /*
 498  * Generate an SCTP packet containing an ABORT chunk. The verification tag
 499  * is given by vtag. The T-bit is set in the ABORT chunk if and only if
 500  * reflected is not 0.
 501  */
 502
 503 static struct mbuf *
 504 ipfw_send_abort(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t vtag,
 505     int reflected)
 506 {
 507         struct mbuf *m;
 508         struct ip *ip;
 509 #ifdef INET6
 510         struct ip6_hdr *ip6;
 511 #endif
 512         struct sctphdr *sctp;
 513         struct sctp_chunkhdr *chunk;
 514         u_int16_t hlen, plen, tlen;
 515
 516         MGETHDR(m, M_NOWAIT, MT_DATA);
 517         if (m == NULL)
 518                 return (NULL);
 519
 520         M_SETFIB(m, id->fib);
 521 #ifdef MAC
 522         if (replyto != NULL)
 523                 mac_netinet_firewall_reply(replyto, m);
 524         else
 525                 mac_netinet_firewall_send(m);
 526 #else
 527         (void)replyto;          /* don't warn about unused arg */
 528 #endif
 529
 530         switch (id->addr_type) {
 531         case 4:
 532                 hlen = sizeof(struct ip);
 533                 break;
 534 #ifdef INET6
 535         case 6:
 536                 hlen = sizeof(struct ip6_hdr);
 537                 break;
 538 #endif
 539         default:
 540                 /* XXX: log me?!? */
 541                 FREE_PKT(m);
 542                 return (NULL);
 543         }
 544         plen = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr);
 545         tlen = hlen + plen;
 546         m->m_data += max_linkhdr;
 547         m->m_flags |= M_SKIP_FIREWALL;
 548         m->m_pkthdr.len = m->m_len = tlen;
 549         m->m_pkthdr.rcvif = NULL;
 550         bzero(m->m_data, tlen);
 551
 552         switch (id->addr_type) {
 553         case 4:
 554                 ip = mtod(m, struct ip *);
 555
 556                 ip->ip_v = 4;
 557                 ip->ip_hl = sizeof(struct ip) >> 2;
 558                 ip->ip_tos = IPTOS_LOWDELAY;
 559                 ip->ip_len = htons(tlen);
 560                 ip->ip_id = htons(0);
 561                 ip->ip_off = htons(0);
 562                 ip->ip_ttl = V_ip_defttl;
 563                 ip->ip_p = IPPROTO_SCTP;
 564                 ip->ip_sum = 0;
 565                 ip->ip_src.s_addr = htonl(id->dst_ip);
 566                 ip->ip_dst.s_addr = htonl(id->src_ip);
 567
 568                 sctp = (struct sctphdr *)(ip + 1);
 569                 break;
 570 #ifdef INET6
 571         case 6:
 572                 ip6 = mtod(m, struct ip6_hdr *);
 573
 574                 ip6->ip6_vfc = IPV6_VERSION;
 575                 ip6->ip6_plen = htons(plen);
 576                 ip6->ip6_nxt = IPPROTO_SCTP;
 577                 ip6->ip6_hlim = IPV6_DEFHLIM;
 578                 ip6->ip6_src = id->dst_ip6;
 579                 ip6->ip6_dst = id->src_ip6;
 580
 581                 sctp = (struct sctphdr *)(ip6 + 1);
 582                 break;
 583 #endif
 584         }
 585
 586         sctp->src_port = htons(id->dst_port);
 587         sctp->dest_port = htons(id->src_port);
 588         sctp->v_tag = htonl(vtag);
 589         sctp->checksum = htonl(0);
 590
 591         chunk = (struct sctp_chunkhdr *)(sctp + 1);
 592         chunk->chunk_type = SCTP_ABORT_ASSOCIATION;
 593         chunk->chunk_flags = 0;
 594         if (reflected != 0) {
 595                 chunk->chunk_flags |= SCTP_HAD_NO_TCB;
 596         }
 597         chunk->chunk_length = htons(sizeof(struct sctp_chunkhdr));
 598
 599         sctp->checksum = sctp_calculate_cksum(m, hlen);
 600
 601         return (m);
 602 }
 603
 604 /*
 605  * Generate a TCP packet, containing either a RST or a keepalive.
 606  * When flags & TH_RST, we are sending a RST packet, because of a
 607  * "reset" action matched the packet.
 608  * Otherwise we are sending a keepalive, and flags & TH_
 609  * The 'replyto' mbuf is the mbuf being replied to, if any, and is required
 610  * so that MAC can label the reply appropriately.
 611  */
 612 struct mbuf *
 613 ipfw_send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq,
 614     u_int32_t ack, int flags)
 615 {
 616         struct mbuf *m = NULL;          /* stupid compiler */
 617         struct ip *h = NULL;            /* stupid compiler */
 618 #ifdef INET6
 619         struct ip6_hdr *h6 = NULL;
 620 #endif
 621         struct tcphdr *th = NULL;
 622         int len, dir;
 623
 624         MGETHDR(m, M_NOWAIT, MT_DATA);
 625         if (m == NULL)
 626                 return (NULL);
 627
 628         M_SETFIB(m, id->fib);
 629 #ifdef MAC
 630         if (replyto != NULL)
 631                 mac_netinet_firewall_reply(replyto, m);
 632         else
 633                 mac_netinet_firewall_send(m);
 634 #else
 635         (void)replyto;          /* don't warn about unused arg */
 636 #endif
 637
 638         switch (id->addr_type) {
 639         case 4:
 640                 len = sizeof(struct ip) + sizeof(struct tcphdr);
 641                 break;
 642 #ifdef INET6
 643         case 6:
 644                 len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
 645                 break;
 646 #endif
 647         default:
 648                 /* XXX: log me?!? */
 649                 FREE_PKT(m);
 650                 return (NULL);
 651         }
 652         dir = ((flags & (TH_SYN | TH_RST)) == TH_SYN);
 653
 654         m->m_data += max_linkhdr;
 655         m->m_flags |= M_SKIP_FIREWALL;
 656         m->m_pkthdr.len = m->m_len = len;
 657         m->m_pkthdr.rcvif = NULL;
 658         bzero(m->m_data, len);
 659
 660         switch (id->addr_type) {
 661         case 4:
 662                 h = mtod(m, struct ip *);
 663
 664                 /* prepare for checksum */
 665                 h->ip_p = IPPROTO_TCP;
 666                 h->ip_len = htons(sizeof(struct tcphdr));
 667                 if (dir) {
 668                         h->ip_src.s_addr = htonl(id->src_ip);
 669                         h->ip_dst.s_addr = htonl(id->dst_ip);
 670                 } else {
 671                         h->ip_src.s_addr = htonl(id->dst_ip);
 672                         h->ip_dst.s_addr = htonl(id->src_ip);
 673                 }
 674
 675                 th = (struct tcphdr *)(h + 1);
 676                 break;
 677 #ifdef INET6
 678         case 6:
 679                 h6 = mtod(m, struct ip6_hdr *);
 680
 681                 /* prepare for checksum */
 682                 h6->ip6_nxt = IPPROTO_TCP;
 683                 h6->ip6_plen = htons(sizeof(struct tcphdr));
 684                 if (dir) {
 685                         h6->ip6_src = id->src_ip6;
 686                         h6->ip6_dst = id->dst_ip6;
 687                 } else {
 688                         h6->ip6_src = id->dst_ip6;
 689                         h6->ip6_dst = id->src_ip6;
 690                 }
 691
 692                 th = (struct tcphdr *)(h6 + 1);
 693                 break;
 694 #endif
 695         }
 696
 697         if (dir) {
 698                 th->th_sport = htons(id->src_port);
 699                 th->th_dport = htons(id->dst_port);
 700         } else {
 701                 th->th_sport = htons(id->dst_port);
 702                 th->th_dport = htons(id->src_port);
 703         }
 704         th->th_off = sizeof(struct tcphdr) >> 2;
 705
 706         if (flags & TH_RST) {
 707                 if (flags & TH_ACK) {
 708                         th->th_seq = htonl(ack);
 709                         th->th_flags = TH_RST;
 710                 } else {
 711                         if (flags & TH_SYN)
 712                                 seq++;
 713                         th->th_ack = htonl(seq);
 714                         th->th_flags = TH_RST | TH_ACK;
 715                 }
 716         } else {
 717                 /*
 718                  * Keepalive - use caller provided sequence numbers
 719                  */
 720                 th->th_seq = htonl(seq);
 721                 th->th_ack = htonl(ack);
 722                 th->th_flags = TH_ACK;
 723         }
 724
 725         switch (id->addr_type) {
 726         case 4:
 727                 th->th_sum = in_cksum(m, len);
 728
 729                 /* finish the ip header */
 730                 h->ip_v = 4;
 731                 h->ip_hl = sizeof(*h) >> 2;
 732                 h->ip_tos = IPTOS_LOWDELAY;
 733                 h->ip_off = htons(0);
 734                 h->ip_len = htons(len);
 735                 h->ip_ttl = V_ip_defttl;
 736                 h->ip_sum = 0;
 737                 break;
 738 #ifdef INET6
 739         case 6:
 740                 th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(*h6),
 741                     sizeof(struct tcphdr));
 742
 743                 /* finish the ip6 header */
 744                 h6->ip6_vfc |= IPV6_VERSION;
 745                 h6->ip6_hlim = IPV6_DEFHLIM;
 746                 break;
 747 #endif
 748         }
 749
 750         return (m);
 751 }
 752
 753 #ifdef INET6
 754 /*
 755  * ipv6 specific rules here...
 756  */
 757 static __inline int
 758 icmp6type_match (int type, ipfw_insn_u32 *cmd)
 759 {
 760         return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
 761 }
 762
 763 static int
 764 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
 765 {
 766         int i;
 767         for (i=0; i <= cmd->o.arg1; ++i )
 768                 if (curr_flow == cmd->d[i] )
 769                         return 1;
 770         return 0;
 771 }
 772
 773 /* support for IP6_*_ME opcodes */
 774 static const struct in6_addr lla_mask = {{{
 775         0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff,
 776         0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
 777 }}};
 778
 779 static int
 780 ipfw_localip6(struct in6_addr *in6)
 781 {
 782         struct rm_priotracker in6_ifa_tracker;
 783         struct in6_ifaddr *ia;
 784
 785         if (IN6_IS_ADDR_MULTICAST(in6))
 786                 return (0);
 787
 788         if (!IN6_IS_ADDR_LINKLOCAL(in6))
 789                 return (in6_localip(in6));
 790
 791         IN6_IFADDR_RLOCK(&in6_ifa_tracker);
 792         CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
 793                 if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr))
 794                         continue;
 795                 if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
 796                     in6, &lla_mask)) {
 797                         IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
 798                         return (1);
 799                 }
 800         }
 801         IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
 802         return (0);
 803 }
 804
 805 static int
 806 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
 807 {
 808         struct nhop6_basic nh6;
 809
 810         if (IN6_IS_SCOPE_LINKLOCAL(src))
 811                 return (1);
 812
 813         if (fib6_lookup_nh_basic(fib, src, 0, NHR_IFAIF, 0, &nh6) != 0)
 814                 return (0);
 815
 816         /* If ifp is provided, check for equality with route table. */
 817         if (ifp != NULL && ifp != nh6.nh_ifp)
 818                 return (0);
 819
 820         /* if no ifp provided, check if rtentry is not default route */
 821         if (ifp == NULL && (nh6.nh_flags & NHF_DEFAULT) != 0)
 822                 return (0);
 823
 824         /* or if this is a blackhole/reject route */
 825         if (ifp == NULL && (nh6.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
 826                 return (0);
 827
 828         /* found valid route */
 829         return 1;
 830 }
 831
 832 static int
 833 is_icmp6_query(int icmp6_type)
 834 {
 835         if ((icmp6_type <= ICMP6_MAXTYPE) &&
 836             (icmp6_type == ICMP6_ECHO_REQUEST ||
 837             icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
 838             icmp6_type == ICMP6_WRUREQUEST ||
 839             icmp6_type == ICMP6_FQDN_QUERY ||
 840             icmp6_type == ICMP6_NI_QUERY))
 841                 return (1);
 842
 843         return (0);
 844 }
 845
 846 static int
 847 map_icmp_unreach(int code)
 848 {
 849
 850         /* RFC 7915 p4.2 */
 851         switch (code) {
 852         case ICMP_UNREACH_NET:
 853         case ICMP_UNREACH_HOST:
 854         case ICMP_UNREACH_SRCFAIL:
 855         case ICMP_UNREACH_NET_UNKNOWN:
 856         case ICMP_UNREACH_HOST_UNKNOWN:
 857         case ICMP_UNREACH_TOSNET:
 858         case ICMP_UNREACH_TOSHOST:
 859                 return (ICMP6_DST_UNREACH_NOROUTE);
 860         case ICMP_UNREACH_PORT:
 861                 return (ICMP6_DST_UNREACH_NOPORT);
 862         default:
 863                 /*
 864                  * Map the rest of codes into admit prohibited.
 865                  * XXX: unreach proto should be mapped into ICMPv6
 866                  * parameter problem, but we use only unreach type.
 867                  */
 868                 return (ICMP6_DST_UNREACH_ADMIN);
 869         }
 870 }
 871
 872 static void
 873 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
 874 {
 875         struct mbuf *m;
 876
 877         m = args->m;
 878         if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
 879                 struct tcphdr *tcp;
 880                 tcp = (struct tcphdr *)((char *)ip6 + hlen);
 881
 882                 if ((tcp->th_flags & TH_RST) == 0) {
 883                         struct mbuf *m0;
 884                         m0 = ipfw_send_pkt(args->m, &(args->f_id),
 885                             ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 886                             tcp->th_flags | TH_RST);
 887                         if (m0 != NULL)
 888                                 ip6_output(m0, NULL, NULL, 0, NULL, NULL,
 889                                     NULL);
 890                 }
 891                 FREE_PKT(m);
 892         } else if (code == ICMP6_UNREACH_ABORT &&
 893             args->f_id.proto == IPPROTO_SCTP) {
 894                 struct mbuf *m0;
 895                 struct sctphdr *sctp;
 896                 u_int32_t v_tag;
 897                 int reflected;
 898
 899                 sctp = (struct sctphdr *)((char *)ip6 + hlen);
 900                 reflected = 1;
 901                 v_tag = ntohl(sctp->v_tag);
 902                 /* Investigate the first chunk header if available */
 903                 if (m->m_len >= hlen + sizeof(struct sctphdr) +
 904                     sizeof(struct sctp_chunkhdr)) {
 905                         struct sctp_chunkhdr *chunk;
 906
 907                         chunk = (struct sctp_chunkhdr *)(sctp + 1);
 908                         switch (chunk->chunk_type) {
 909                         case SCTP_INITIATION:
 910                                 /*
 911                                  * Packets containing an INIT chunk MUST have
 912                                  * a zero v-tag.
 913                                  */
 914                                 if (v_tag != 0) {
 915                                         v_tag = 0;
 916                                         break;
 917                                 }
 918                                 /* INIT chunk MUST NOT be bundled */
 919                                 if (m->m_pkthdr.len >
 920                                     hlen + sizeof(struct sctphdr) +
 921                                     ntohs(chunk->chunk_length) + 3) {
 922                                         break;
 923                                 }
 924                                 /* Use the initiate tag if available */
 925                                 if ((m->m_len >= hlen + sizeof(struct sctphdr) +
 926                                     sizeof(struct sctp_chunkhdr) +
 927                                     offsetof(struct sctp_init, a_rwnd))) {
 928                                         struct sctp_init *init;
 929
 930                                         init = (struct sctp_init *)(chunk + 1);
 931                                         v_tag = ntohl(init->initiate_tag);
 932                                         reflected = 0;
 933                                 }
 934                                 break;
 935                         case SCTP_ABORT_ASSOCIATION:
 936                                 /*
 937                                  * If the packet contains an ABORT chunk, don't
 938                                  * reply.
 939                                  * XXX: We should search through all chunks,
 940                                  * but do not do that to avoid attacks.
 941                                  */
 942                                 v_tag = 0;
 943                                 break;
 944                         }
 945                 }
 946                 if (v_tag == 0) {
 947                         m0 = NULL;
 948                 } else {
 949                         m0 = ipfw_send_abort(args->m, &(args->f_id), v_tag,
 950                             reflected);
 951                 }
 952                 if (m0 != NULL)
 953                         ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
 954                 FREE_PKT(m);
 955         } else if (code != ICMP6_UNREACH_RST && code != ICMP6_UNREACH_ABORT) {
 956                 /* Send an ICMPv6 unreach. */
 957 #if 0
 958                 /*
 959                  * Unlike above, the mbufs need to line up with the ip6 hdr,
 960                  * as the contents are read. We need to m_adj() the
 961                  * needed amount.
 962                  * The mbuf will however be thrown away so we can adjust it.
 963                  * Remember we did an m_pullup on it already so we
 964                  * can make some assumptions about contiguousness.
 965                  */
 966                 if (args->L3offset)
 967                         m_adj(m, args->L3offset);
 968 #endif
 969                 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
 970         } else
 971                 FREE_PKT(m);
 972
 973         args->m = NULL;
 974 }
 975
 976 #endif /* INET6 */
 977
 978
 979 /*
 980  * sends a reject message, consuming the mbuf passed as an argument.
 981  */
 982 static void
 983 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
 984 {
 985
 986 #if 0
 987         /* XXX When ip is not guaranteed to be at mtod() we will
 988          * need to account for this */
 989          * The mbuf will however be thrown away so we can adjust it.
 990          * Remember we did an m_pullup on it already so we
 991          * can make some assumptions about contiguousness.
 992          */
 993         if (args->L3offset)
 994                 m_adj(m, args->L3offset);
 995 #endif
 996         if (code != ICMP_REJECT_RST && code != ICMP_REJECT_ABORT) {
 997                 /* Send an ICMP unreach */
 998                 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
 999         } else if (code == ICMP_REJECT_RST && args->f_id.proto == IPPROTO_TCP) {
1000                 struct tcphdr *const tcp =
1001                     L3HDR(struct tcphdr, mtod(args->m, struct ip *));
1002                 if ( (tcp->th_flags & TH_RST) == 0) {
1003                         struct mbuf *m;
1004                         m = ipfw_send_pkt(args->m, &(args->f_id),
1005                                 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
1006                                 tcp->th_flags | TH_RST);
1007                         if (m != NULL)
1008                                 ip_output(m, NULL, NULL, 0, NULL, NULL);
1009                 }
1010                 FREE_PKT(args->m);
1011         } else if (code == ICMP_REJECT_ABORT &&
1012             args->f_id.proto == IPPROTO_SCTP) {
1013                 struct mbuf *m;
1014                 struct sctphdr *sctp;
1015                 struct sctp_chunkhdr *chunk;
1016                 struct sctp_init *init;
1017                 u_int32_t v_tag;
1018                 int reflected;
1019
1020                 sctp = L3HDR(struct sctphdr, mtod(args->m, struct ip *));
1021                 reflected = 1;
1022                 v_tag = ntohl(sctp->v_tag);
1023                 if (iplen >= (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1024                     sizeof(struct sctp_chunkhdr)) {
1025                         /* Look at the first chunk header if available */
1026                         chunk = (struct sctp_chunkhdr *)(sctp + 1);
1027                         switch (chunk->chunk_type) {
1028                         case SCTP_INITIATION:
1029                                 /*
1030                                  * Packets containing an INIT chunk MUST have
1031                                  * a zero v-tag.
1032                                  */
1033                                 if (v_tag != 0) {
1034                                         v_tag = 0;
1035                                         break;
1036                                 }
1037                                 /* INIT chunk MUST NOT be bundled */
1038                                 if (iplen >
1039                                     (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1040                                     ntohs(chunk->chunk_length) + 3) {
1041                                         break;
1042                                 }
1043                                 /* Use the initiate tag if available */
1044                                 if ((iplen >= (ip->ip_hl << 2) +
1045                                     sizeof(struct sctphdr) +
1046                                     sizeof(struct sctp_chunkhdr) +
1047                                     offsetof(struct sctp_init, a_rwnd))) {
1048                                         init = (struct sctp_init *)(chunk + 1);
1049                                         v_tag = ntohl(init->initiate_tag);
1050                                         reflected = 0;
1051                                 }
1052                                 break;
1053                         case SCTP_ABORT_ASSOCIATION:
1054                                 /*
1055                                  * If the packet contains an ABORT chunk, don't
1056                                  * reply.
1057                                  * XXX: We should search through all chunks,
1058                                  * but do not do that to avoid attacks.
1059                                  */
1060                                 v_tag = 0;
1061                                 break;
1062                         }
1063                 }
1064                 if (v_tag == 0) {
1065                         m = NULL;
1066                 } else {
1067                         m = ipfw_send_abort(args->m, &(args->f_id), v_tag,
1068                             reflected);
1069                 }
1070                 if (m != NULL)
1071                         ip_output(m, NULL, NULL, 0, NULL, NULL);
1072                 FREE_PKT(args->m);
1073         } else
1074                 FREE_PKT(args->m);
1075         args->m = NULL;
1076 }
1077
1078 /*
1079  * Support for uid/gid/jail lookup. These tests are expensive
1080  * (because we may need to look into the list of active sockets)
1081  * so we cache the results. ugid_lookupp is 0 if we have not
1082  * yet done a lookup, 1 if we succeeded, and -1 if we tried
1083  * and failed. The function always returns the match value.
1084  * We could actually spare the variable and use *uc, setting
1085  * it to '(void *)check_uidgid if we have no info, NULL if
1086  * we tried and failed, or any other value if successful.
1087  */
1088 static int
1089 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
1090     struct ucred **uc)
1091 {
1092 #if defined(USERSPACE)
1093         return 0;       // not supported in userspace
1094 #else
1095 #ifndef __FreeBSD__
1096         /* XXX */
1097         return cred_check(insn, proto, oif,
1098             dst_ip, dst_port, src_ip, src_port,
1099             (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
1100 #else  /* FreeBSD */
1101         struct in_addr src_ip, dst_ip;
1102         struct inpcbinfo *pi;
1103         struct ipfw_flow_id *id;
1104         struct inpcb *pcb, *inp;
1105         int lookupflags;
1106         int match;
1107
1108         id = &args->f_id;
1109         inp = args->inp;
1110
1111         /*
1112          * Check to see if the UDP or TCP stack supplied us with
1113          * the PCB. If so, rather then holding a lock and looking
1114          * up the PCB, we can use the one that was supplied.
1115          */
1116         if (inp && *ugid_lookupp == 0) {
1117                 INP_LOCK_ASSERT(inp);
1118                 if (inp->inp_socket != NULL) {
1119                         *uc = crhold(inp->inp_cred);
1120                         *ugid_lookupp = 1;
1121                 } else
1122                         *ugid_lookupp = -1;
1123         }
1124         /*
1125          * If we have already been here and the packet has no
1126          * PCB entry associated with it, then we can safely
1127          * assume that this is a no match.
1128          */
1129         if (*ugid_lookupp == -1)
1130                 return (0);
1131         if (id->proto == IPPROTO_TCP) {
1132                 lookupflags = 0;
1133                 pi = &V_tcbinfo;
1134         } else if (id->proto == IPPROTO_UDP) {
1135                 lookupflags = INPLOOKUP_WILDCARD;
1136                 pi = &V_udbinfo;
1137         } else if (id->proto == IPPROTO_UDPLITE) {
1138                 lookupflags = INPLOOKUP_WILDCARD;
1139                 pi = &V_ulitecbinfo;
1140         } else
1141                 return 0;
1142         lookupflags |= INPLOOKUP_RLOCKPCB;
1143         match = 0;
1144         if (*ugid_lookupp == 0) {
1145                 if (id->addr_type == 6) {
1146 #ifdef INET6
1147                         if (args->flags & IPFW_ARGS_IN)
1148                                 pcb = in6_pcblookup_mbuf(pi,
1149                                     &id->src_ip6, htons(id->src_port),
1150                                     &id->dst_ip6, htons(id->dst_port),
1151                                     lookupflags, NULL, args->m);
1152                         else
1153                                 pcb = in6_pcblookup_mbuf(pi,
1154                                     &id->dst_ip6, htons(id->dst_port),
1155                                     &id->src_ip6, htons(id->src_port),
1156                                     lookupflags, args->ifp, args->m);
1157 #else
1158                         *ugid_lookupp = -1;
1159                         return (0);
1160 #endif
1161                 } else {
1162                         src_ip.s_addr = htonl(id->src_ip);
1163                         dst_ip.s_addr = htonl(id->dst_ip);
1164                         if (args->flags & IPFW_ARGS_IN)
1165                                 pcb = in_pcblookup_mbuf(pi,
1166                                     src_ip, htons(id->src_port),
1167                                     dst_ip, htons(id->dst_port),
1168                                     lookupflags, NULL, args->m);
1169                         else
1170                                 pcb = in_pcblookup_mbuf(pi,
1171                                     dst_ip, htons(id->dst_port),
1172                                     src_ip, htons(id->src_port),
1173                                     lookupflags, args->ifp, args->m);
1174                 }
1175                 if (pcb != NULL) {
1176                         INP_RLOCK_ASSERT(pcb);
1177                         *uc = crhold(pcb->inp_cred);
1178                         *ugid_lookupp = 1;
1179                         INP_RUNLOCK(pcb);
1180                 }
1181                 if (*ugid_lookupp == 0) {
1182                         /*
1183                          * We tried and failed, set the variable to -1
1184                          * so we will not try again on this packet.
1185                          */
1186                         *ugid_lookupp = -1;
1187                         return (0);
1188                 }
1189         }
1190         if (insn->o.opcode == O_UID)
1191                 match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
1192         else if (insn->o.opcode == O_GID)
1193                 match = groupmember((gid_t)insn->d[0], *uc);
1194         else if (insn->o.opcode == O_JAIL)
1195                 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
1196         return (match);
1197 #endif /* __FreeBSD__ */
1198 #endif /* not supported in userspace */
1199 }
1200
1201 /*
1202  * Helper function to set args with info on the rule after the matching
1203  * one. slot is precise, whereas we guess rule_id as they are
1204  * assigned sequentially.
1205  */
1206 static inline void
1207 set_match(struct ip_fw_args *args, int slot,
1208         struct ip_fw_chain *chain)
1209 {
1210         args->rule.chain_id = chain->id;
1211         args->rule.slot = slot + 1; /* we use 0 as a marker */
1212         args->rule.rule_id = 1 + chain->map[slot]->id;
1213         args->rule.rulenum = chain->map[slot]->rulenum;
1214         args->flags |= IPFW_ARGS_REF;
1215 }
1216
1217 #ifndef LINEAR_SKIPTO
1218 /*
1219  * Helper function to enable cached rule lookups using
1220  * cached_id and cached_pos fields in ipfw rule.
1221  */
1222 static int
1223 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1224     int tablearg, int jump_backwards)
1225 {
1226         int f_pos;
1227
1228         /* If possible use cached f_pos (in f->cached_pos),
1229          * whose version is written in f->cached_id
1230          * (horrible hacks to avoid changing the ABI).
1231          */
1232         if (num != IP_FW_TARG && f->cached_id == chain->id)
1233                 f_pos = f->cached_pos;
1234         else {
1235                 int i = IP_FW_ARG_TABLEARG(chain, num, skipto);
1236                 /* make sure we do not jump backward */
1237                 if (jump_backwards == 0 && i <= f->rulenum)
1238                         i = f->rulenum + 1;
1239                 if (chain->idxmap != NULL)
1240                         f_pos = chain->idxmap[i];
1241                 else
1242                         f_pos = ipfw_find_rule(chain, i, 0);
1243                 /* update the cache */
1244                 if (num != IP_FW_TARG) {
1245                         f->cached_id = chain->id;
1246                         f->cached_pos = f_pos;
1247                 }
1248         }
1249
1250         return (f_pos);
1251 }
1252 #else
1253 /*
1254  * Helper function to enable real fast rule lookups.
1255  */
1256 static int
1257 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1258     int tablearg, int jump_backwards)
1259 {
1260         int f_pos;
1261
1262         num = IP_FW_ARG_TABLEARG(chain, num, skipto);
1263         /* make sure we do not jump backward */
1264         if (jump_backwards == 0 && num <= f->rulenum)
1265                 num = f->rulenum + 1;
1266         f_pos = chain->idxmap[num];
1267
1268         return (f_pos);
1269 }
1270 #endif
1271
1272 #define TARG(k, f)      IP_FW_ARG_TABLEARG(chain, k, f)
1273 /*
1274  * The main check routine for the firewall.
1275  *
1276  * All arguments are in args so we can modify them and return them
1277  * back to the caller.
1278  *
1279  * Parameters:
1280  *
1281  *      args->m (in/out) The packet; we set to NULL when/if we nuke it.
1282  *              Starts with the IP header.
1283  *      args->L3offset  Number of bytes bypassed if we came from L2.
1284  *                      e.g. often sizeof(eh)  ** NOTYET **
1285  *      args->ifp       Incoming or outgoing interface.
1286  *      args->divert_rule (in/out)
1287  *              Skip up to the first rule past this rule number;
1288  *              upon return, non-zero port number for divert or tee.
1289  *
1290  *      args->rule      Pointer to the last matching rule (in/out)
1291  *      args->next_hop  Socket we are forwarding to (out).
1292  *      args->next_hop6 IPv6 next hop we are forwarding to (out).
1293  *      args->f_id      Addresses grabbed from the packet (out)
1294  *      args->rule.info a cookie depending on rule action
1295  *
1296  * Return value:
1297  *
1298  *      IP_FW_PASS      the packet must be accepted
1299  *      IP_FW_DENY      the packet must be dropped
1300  *      IP_FW_DIVERT    divert packet, port in m_tag
1301  *      IP_FW_TEE       tee packet, port in m_tag
1302  *      IP_FW_DUMMYNET  to dummynet, pipe in args->cookie
1303  *      IP_FW_NETGRAPH  into netgraph, cookie args->cookie
1304  *              args->rule contains the matching rule,
1305  *              args->rule.info has additional information.
1306  *
1307  */
1308 int
1309 ipfw_chk(struct ip_fw_args *args)
1310 {
1311
1312         /*
1313          * Local variables holding state while processing a packet:
1314          *
1315          * IMPORTANT NOTE: to speed up the processing of rules, there
1316          * are some assumption on the values of the variables, which
1317          * are documented here. Should you change them, please check
1318          * the implementation of the various instructions to make sure
1319          * that they still work.
1320          *
1321          * m | args->m  Pointer to the mbuf, as received from the caller.
1322          *      It may change if ipfw_chk() does an m_pullup, or if it
1323          *      consumes the packet because it calls send_reject().
1324          *      XXX This has to change, so that ipfw_chk() never modifies
1325          *      or consumes the buffer.
1326          *      OR
1327          * args->mem    Pointer to contigous memory chunk.
1328          * ip   Is the beginning of the ip(4 or 6) header.
1329          * eh   Ethernet header in case if input is Layer2.
1330          */
1331         struct mbuf *m;
1332         struct ip *ip;
1333         struct ether_header *eh;
1334
1335         /*
1336          * For rules which contain uid/gid or jail constraints, cache
1337          * a copy of the users credentials after the pcb lookup has been
1338          * executed. This will speed up the processing of rules with
1339          * these types of constraints, as well as decrease contention
1340          * on pcb related locks.
1341          */
1342 #ifndef __FreeBSD__
1343         struct bsd_ucred ucred_cache;
1344 #else
1345         struct ucred *ucred_cache = NULL;
1346 #endif
1347         int ucred_lookup = 0;
1348         int f_pos = 0;          /* index of current rule in the array */
1349         int retval = 0;
1350         struct ifnet *oif, *iif;
1351
1352         /*
1353          * hlen The length of the IP header.
1354          */
1355         u_int hlen = 0;         /* hlen >0 means we have an IP pkt */
1356
1357         /*
1358          * offset       The offset of a fragment. offset != 0 means that
1359          *      we have a fragment at this offset of an IPv4 packet.
1360          *      offset == 0 means that (if this is an IPv4 packet)
1361          *      this is the first or only fragment.
1362          *      For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
1363          *      or there is a single packet fragment (fragment header added
1364          *      without needed).  We will treat a single packet fragment as if
1365          *      there was no fragment header (or log/block depending on the
1366          *      V_fw_permit_single_frag6 sysctl setting).
1367          */
1368         u_short offset = 0;
1369         u_short ip6f_mf = 0;
1370
1371         /*
1372          * Local copies of addresses. They are only valid if we have
1373          * an IP packet.
1374          *
1375          * proto        The protocol. Set to 0 for non-ip packets,
1376          *      or to the protocol read from the packet otherwise.
1377          *      proto != 0 means that we have an IPv4 packet.
1378          *
1379          * src_port, dst_port   port numbers, in HOST format. Only
1380          *      valid for TCP and UDP packets.
1381          *
1382          * src_ip, dst_ip       ip addresses, in NETWORK format.
1383          *      Only valid for IPv4 packets.
1384          */
1385         uint8_t proto;
1386         uint16_t src_port, dst_port;            /* NOTE: host format    */
1387         struct in_addr src_ip, dst_ip;          /* NOTE: network format */
1388         int iplen = 0;
1389         int pktlen;
1390
1391         struct ipfw_dyn_info dyn_info;
1392         struct ip_fw *q = NULL;
1393         struct ip_fw_chain *chain = &V_layer3_chain;
1394
1395         /*
1396          * We store in ulp a pointer to the upper layer protocol header.
1397          * In the ipv4 case this is easy to determine from the header,
1398          * but for ipv6 we might have some additional headers in the middle.
1399          * ulp is NULL if not found.
1400          */
1401         void *ulp = NULL;               /* upper layer protocol pointer. */
1402
1403         /* XXX ipv6 variables */
1404         int is_ipv6 = 0;
1405         uint8_t icmp6_type = 0;
1406         uint16_t ext_hd = 0;    /* bits vector for extension header filtering */
1407         /* end of ipv6 variables */
1408
1409         int is_ipv4 = 0;
1410
1411         int done = 0;           /* flag to exit the outer loop */
1412         IPFW_RLOCK_TRACKER;
1413         bool mem;
1414
1415         if ((mem = (args->flags & IPFW_ARGS_LENMASK))) {
1416                 if (args->flags & IPFW_ARGS_ETHER) {
1417                         eh = (struct ether_header *)args->mem;
1418                         if (eh->ether_type == htons(ETHERTYPE_VLAN))
1419                                 ip = (struct ip *)
1420                                     ((struct ether_vlan_header *)eh + 1);
1421                         else
1422                                 ip = (struct ip *)(eh + 1);
1423                 } else {
1424                         eh = NULL;
1425                         ip = (struct ip *)args->mem;
1426                 }
1427                 pktlen = IPFW_ARGS_LENGTH(args->flags);
1428                 args->f_id.fib = args->ifp->if_fib;     /* best guess */
1429         } else {
1430                 m = args->m;
1431                 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
1432                         return (IP_FW_PASS);    /* accept */
1433                 if (args->flags & IPFW_ARGS_ETHER) {
1434                         /* We need some amount of data to be contiguous. */
1435                         if (m->m_len < min(m->m_pkthdr.len, max_protohdr) &&
1436                             (args->m = m = m_pullup(m, min(m->m_pkthdr.len,
1437                             max_protohdr))) == NULL)
1438                                 goto pullup_failed;
1439                         eh = mtod(m, struct ether_header *);
1440                         ip = (struct ip *)(eh + 1);
1441                 } else {
1442                         eh = NULL;
1443                         ip = mtod(m, struct ip *);
1444                 }
1445                 pktlen = m->m_pkthdr.len;
1446                 args->f_id.fib = M_GETFIB(m); /* mbuf not altered */
1447         }
1448
1449         dst_ip.s_addr = 0;              /* make sure it is initialized */
1450         src_ip.s_addr = 0;              /* make sure it is initialized */
1451         src_port = dst_port = 0;
1452
1453         DYN_INFO_INIT(&dyn_info);
1454 /*
1455  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1456  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1457  * pointer might become stale after other pullups (but we never use it
1458  * this way).
1459  */
1460 #define PULLUP_TO(_len, p, T)   PULLUP_LEN(_len, p, sizeof(T))
1461 #define EHLEN   (eh != NULL ? ((char *)ip - (char *)eh) : 0)
1462 #define _PULLUP_LOCKED(_len, p, T, unlock)                      \
1463 do {                                                            \
1464         int x = (_len) + T + EHLEN;                             \
1465         if (mem) {                                              \
1466                 if (__predict_false(pktlen < x)) {              \
1467                         unlock;                                 \
1468                         goto pullup_failed;                     \
1469                 }                                               \
1470                 p = (char *)args->mem + (_len) + EHLEN;         \
1471         } else {                                                \
1472                 if (__predict_false((m)->m_len < x)) {          \
1473                         args->m = m = m_pullup(m, x);           \
1474                         if (m == NULL) {                        \
1475                                 unlock;                         \
1476                                 goto pullup_failed;             \
1477                         }                                       \
1478                 }                                               \
1479                 p = mtod(m, char *) + (_len) + EHLEN;           \
1480         }                                                       \
1481 } while (0)
1482
1483 #define PULLUP_LEN(_len, p, T)  _PULLUP_LOCKED(_len, p, T, )
1484 #define PULLUP_LEN_LOCKED(_len, p, T)   \
1485     _PULLUP_LOCKED(_len, p, T, IPFW_PF_RUNLOCK(chain)); \
1486     UPDATE_POINTERS()
1487 /*
1488  * In case pointers got stale after pullups, update them.
1489  */
1490 #define UPDATE_POINTERS()                                       \
1491 do {                                                            \
1492         if (!mem) {                                             \
1493                 if (eh != NULL) {                               \
1494                         eh = mtod(m, struct ether_header *);    \
1495                         ip = (struct ip *)(eh + 1);             \
1496                 } else                                          \
1497                         ip = mtod(m, struct ip *);              \
1498                 args->m = m;                                    \
1499         }                                                       \
1500 } while (0)
1501
1502         /* Identify IP packets and fill up variables. */
1503         if (pktlen >= sizeof(struct ip6_hdr) &&
1504             (eh == NULL || eh->ether_type == htons(ETHERTYPE_IPV6)) &&
1505             ip->ip_v == 6) {
1506                 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1507
1508                 is_ipv6 = 1;
1509                 args->flags |= IPFW_ARGS_IP6;
1510                 hlen = sizeof(struct ip6_hdr);
1511                 proto = ip6->ip6_nxt;
1512                 /* Search extension headers to find upper layer protocols */
1513                 while (ulp == NULL && offset == 0) {
1514                         switch (proto) {
1515                         case IPPROTO_ICMPV6:
1516                                 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1517                                 icmp6_type = ICMP6(ulp)->icmp6_type;
1518                                 break;
1519
1520                         case IPPROTO_TCP:
1521                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1522                                 dst_port = TCP(ulp)->th_dport;
1523                                 src_port = TCP(ulp)->th_sport;
1524                                 /* save flags for dynamic rules */
1525                                 args->f_id._flags = TCP(ulp)->th_flags;
1526                                 break;
1527
1528                         case IPPROTO_SCTP:
1529                                 if (pktlen >= hlen + sizeof(struct sctphdr) +
1530                                     sizeof(struct sctp_chunkhdr) +
1531                                     offsetof(struct sctp_init, a_rwnd))
1532                                         PULLUP_LEN(hlen, ulp,
1533                                             sizeof(struct sctphdr) +
1534                                             sizeof(struct sctp_chunkhdr) +
1535                                             offsetof(struct sctp_init, a_rwnd));
1536                                 else if (pktlen >= hlen + sizeof(struct sctphdr))
1537                                         PULLUP_LEN(hlen, ulp, pktlen - hlen);
1538                                 else
1539                                         PULLUP_LEN(hlen, ulp,
1540                                             sizeof(struct sctphdr));
1541                                 src_port = SCTP(ulp)->src_port;
1542                                 dst_port = SCTP(ulp)->dest_port;
1543                                 break;
1544
1545                         case IPPROTO_UDP:
1546                         case IPPROTO_UDPLITE:
1547                                 PULLUP_TO(hlen, ulp, struct udphdr);
1548                                 dst_port = UDP(ulp)->uh_dport;
1549                                 src_port = UDP(ulp)->uh_sport;
1550                                 break;
1551
1552                         case IPPROTO_HOPOPTS:   /* RFC 2460 */
1553                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1554                                 ext_hd |= EXT_HOPOPTS;
1555                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1556                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1557                                 ulp = NULL;
1558                                 break;
1559
1560                         case IPPROTO_ROUTING:   /* RFC 2460 */
1561                                 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1562                                 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1563                                 case 0:
1564                                         ext_hd |= EXT_RTHDR0;
1565                                         break;
1566                                 case 2:
1567                                         ext_hd |= EXT_RTHDR2;
1568                                         break;
1569                                 default:
1570                                         if (V_fw_verbose)
1571                                                 printf("IPFW2: IPV6 - Unknown "
1572                                                     "Routing Header type(%d)\n",
1573                                                     ((struct ip6_rthdr *)
1574                                                     ulp)->ip6r_type);
1575                                         if (V_fw_deny_unknown_exthdrs)
1576                                             return (IP_FW_DENY);
1577                                         break;
1578                                 }
1579                                 ext_hd |= EXT_ROUTING;
1580                                 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1581                                 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1582                                 ulp = NULL;
1583                                 break;
1584
1585                         case IPPROTO_FRAGMENT:  /* RFC 2460 */
1586                                 PULLUP_TO(hlen, ulp, struct ip6_frag);
1587                                 ext_hd |= EXT_FRAGMENT;
1588                                 hlen += sizeof (struct ip6_frag);
1589                                 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1590                                 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1591                                         IP6F_OFF_MASK;
1592                                 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1593                                         IP6F_MORE_FRAG;
1594                                 if (V_fw_permit_single_frag6 == 0 &&
1595                                     offset == 0 && ip6f_mf == 0) {
1596                                         if (V_fw_verbose)
1597                                                 printf("IPFW2: IPV6 - Invalid "
1598                                                     "Fragment Header\n");
1599                                         if (V_fw_deny_unknown_exthdrs)
1600                                             return (IP_FW_DENY);
1601                                         break;
1602                                 }
1603                                 args->f_id.extra =
1604                                     ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1605                                 ulp = NULL;
1606                                 break;
1607
1608                         case IPPROTO_DSTOPTS:   /* RFC 2460 */
1609                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1610                                 ext_hd |= EXT_DSTOPTS;
1611                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1612                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1613                                 ulp = NULL;
1614                                 break;
1615
1616                         case IPPROTO_AH:        /* RFC 2402 */
1617                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1618                                 ext_hd |= EXT_AH;
1619                                 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1620                                 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1621                                 ulp = NULL;
1622                                 break;
1623
1624                         case IPPROTO_ESP:       /* RFC 2406 */
1625                                 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1626                                 /* Anything past Seq# is variable length and
1627                                  * data past this ext. header is encrypted. */
1628                                 ext_hd |= EXT_ESP;
1629                                 break;
1630
1631                         case IPPROTO_NONE:      /* RFC 2460 */
1632                                 /*
1633                                  * Packet ends here, and IPv6 header has
1634                                  * already been pulled up. If ip6e_len!=0
1635                                  * then octets must be ignored.
1636                                  */
1637                                 ulp = ip; /* non-NULL to get out of loop. */
1638                                 break;
1639
1640                         case IPPROTO_OSPFIGP:
1641                                 /* XXX OSPF header check? */
1642                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1643                                 break;
1644
1645                         case IPPROTO_PIM:
1646                                 /* XXX PIM header check? */
1647                                 PULLUP_TO(hlen, ulp, struct pim);
1648                                 break;
1649
1650                         case IPPROTO_GRE:       /* RFC 1701 */
1651                                 /* XXX GRE header check? */
1652                                 PULLUP_TO(hlen, ulp, struct grehdr);
1653                                 break;
1654
1655                         case IPPROTO_CARP:
1656                                 PULLUP_TO(hlen, ulp, offsetof(
1657                                     struct carp_header, carp_counter));
1658                                 if (CARP_ADVERTISEMENT !=
1659                                     ((struct carp_header *)ulp)->carp_type)
1660                                         return (IP_FW_DENY);
1661                                 break;
1662
1663                         case IPPROTO_IPV6:      /* RFC 2893 */
1664                                 PULLUP_TO(hlen, ulp, struct ip6_hdr);
1665                                 break;
1666
1667                         case IPPROTO_IPV4:      /* RFC 2893 */
1668                                 PULLUP_TO(hlen, ulp, struct ip);
1669                                 break;
1670
1671                         default:
1672                                 if (V_fw_verbose)
1673                                         printf("IPFW2: IPV6 - Unknown "
1674                                             "Extension Header(%d), ext_hd=%x\n",
1675                                              proto, ext_hd);
1676                                 if (V_fw_deny_unknown_exthdrs)
1677                                     return (IP_FW_DENY);
1678                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1679                                 break;
1680                         } /*switch */
1681                 }
1682                 UPDATE_POINTERS();
1683                 ip6 = (struct ip6_hdr *)ip;
1684                 args->f_id.addr_type = 6;
1685                 args->f_id.src_ip6 = ip6->ip6_src;
1686                 args->f_id.dst_ip6 = ip6->ip6_dst;
1687                 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1688                 iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6);
1689         } else if (pktlen >= sizeof(struct ip) &&
1690             (eh == NULL || eh->ether_type == htons(ETHERTYPE_IP)) &&
1691             ip->ip_v == 4) {
1692                 is_ipv4 = 1;
1693                 args->flags |= IPFW_ARGS_IP4;
1694                 hlen = ip->ip_hl << 2;
1695                 /*
1696                  * Collect parameters into local variables for faster
1697                  * matching.
1698                  */
1699                 proto = ip->ip_p;
1700                 src_ip = ip->ip_src;
1701                 dst_ip = ip->ip_dst;
1702                 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1703                 iplen = ntohs(ip->ip_len);
1704
1705                 if (offset == 0) {
1706                         switch (proto) {
1707                         case IPPROTO_TCP:
1708                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1709                                 dst_port = TCP(ulp)->th_dport;
1710                                 src_port = TCP(ulp)->th_sport;
1711                                 /* save flags for dynamic rules */
1712                                 args->f_id._flags = TCP(ulp)->th_flags;
1713                                 break;
1714
1715                         case IPPROTO_SCTP:
1716                                 if (pktlen >= hlen + sizeof(struct sctphdr) +
1717                                     sizeof(struct sctp_chunkhdr) +
1718                                     offsetof(struct sctp_init, a_rwnd))
1719                                         PULLUP_LEN(hlen, ulp,
1720                                             sizeof(struct sctphdr) +
1721                                             sizeof(struct sctp_chunkhdr) +
1722                                             offsetof(struct sctp_init, a_rwnd));
1723                                 else if (pktlen >= hlen + sizeof(struct sctphdr))
1724                                         PULLUP_LEN(hlen, ulp, pktlen - hlen);
1725                                 else
1726                                         PULLUP_LEN(hlen, ulp,
1727                                             sizeof(struct sctphdr));
1728                                 src_port = SCTP(ulp)->src_port;
1729                                 dst_port = SCTP(ulp)->dest_port;
1730                                 break;
1731
1732                         case IPPROTO_UDP:
1733                         case IPPROTO_UDPLITE:
1734                                 PULLUP_TO(hlen, ulp, struct udphdr);
1735                                 dst_port = UDP(ulp)->uh_dport;
1736                                 src_port = UDP(ulp)->uh_sport;
1737                                 break;
1738
1739                         case IPPROTO_ICMP:
1740                                 PULLUP_TO(hlen, ulp, struct icmphdr);
1741                                 //args->f_id.flags = ICMP(ulp)->icmp_type;
1742                                 break;
1743
1744                         default:
1745                                 break;
1746                         }
1747                 } else {
1748                         if (offset == 1 && proto == IPPROTO_TCP) {
1749                                 /* RFC 3128 */
1750                                 goto pullup_failed;
1751                         }
1752                 }
1753
1754                 UPDATE_POINTERS();
1755                 args->f_id.addr_type = 4;
1756                 args->f_id.src_ip = ntohl(src_ip.s_addr);
1757                 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1758         } else {
1759                 proto = 0;
1760                 dst_ip.s_addr = src_ip.s_addr = 0;
1761
1762                 args->f_id.addr_type = 1; /* XXX */
1763         }
1764 #undef PULLUP_TO
1765         pktlen = iplen < pktlen ? iplen: pktlen;
1766
1767         /* Properly initialize the rest of f_id */
1768         args->f_id.proto = proto;
1769         args->f_id.src_port = src_port = ntohs(src_port);
1770         args->f_id.dst_port = dst_port = ntohs(dst_port);
1771
1772         IPFW_PF_RLOCK(chain);
1773         if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1774                 IPFW_PF_RUNLOCK(chain);
1775                 return (IP_FW_PASS);    /* accept */
1776         }
1777         if (args->flags & IPFW_ARGS_REF) {
1778                 /*
1779                  * Packet has already been tagged as a result of a previous
1780                  * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1781                  * REASS, NETGRAPH, DIVERT/TEE...)
1782                  * Validate the slot and continue from the next one
1783                  * if still present, otherwise do a lookup.
1784                  */
1785                 f_pos = (args->rule.chain_id == chain->id) ?
1786                     args->rule.slot :
1787                     ipfw_find_rule(chain, args->rule.rulenum,
1788                         args->rule.rule_id);
1789         } else {
1790                 f_pos = 0;
1791         }
1792
1793         if (args->flags & IPFW_ARGS_IN) {
1794                 iif = args->ifp;
1795                 oif = NULL;
1796         } else {
1797                 MPASS(args->flags & IPFW_ARGS_OUT);
1798                 iif = mem ? NULL : m_rcvif(m);
1799                 oif = args->ifp;
1800         }
1801
1802         /*
1803          * Now scan the rules, and parse microinstructions for each rule.
1804          * We have two nested loops and an inner switch. Sometimes we
1805          * need to break out of one or both loops, or re-enter one of
1806          * the loops with updated variables. Loop variables are:
1807          *
1808          *      f_pos (outer loop) points to the current rule.
1809          *              On output it points to the matching rule.
1810          *      done (outer loop) is used as a flag to break the loop.
1811          *      l (inner loop)  residual length of current rule.
1812          *              cmd points to the current microinstruction.
1813          *
1814          * We break the inner loop by setting l=0 and possibly
1815          * cmdlen=0 if we don't want to advance cmd.
1816          * We break the outer loop by setting done=1
1817          * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1818          * as needed.
1819          */
1820         for (; f_pos < chain->n_rules; f_pos++) {
1821                 ipfw_insn *cmd;
1822                 uint32_t tablearg = 0;
1823                 int l, cmdlen, skip_or; /* skip rest of OR block */
1824                 struct ip_fw *f;
1825
1826                 f = chain->map[f_pos];
1827                 if (V_set_disable & (1 << f->set) )
1828                         continue;
1829
1830                 skip_or = 0;
1831                 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1832                     l -= cmdlen, cmd += cmdlen) {
1833                         int match;
1834
1835                         /*
1836                          * check_body is a jump target used when we find a
1837                          * CHECK_STATE, and need to jump to the body of
1838                          * the target rule.
1839                          */
1840
1841 /* check_body: */
1842                         cmdlen = F_LEN(cmd);
1843                         /*
1844                          * An OR block (insn_1 || .. || insn_n) has the
1845                          * F_OR bit set in all but the last instruction.
1846                          * The first match will set "skip_or", and cause
1847                          * the following instructions to be skipped until
1848                          * past the one with the F_OR bit clear.
1849                          */
1850                         if (skip_or) {          /* skip this instruction */
1851                                 if ((cmd->len & F_OR) == 0)
1852                                         skip_or = 0;    /* next one is good */
1853                                 continue;
1854                         }
1855                         match = 0; /* set to 1 if we succeed */
1856
1857                         switch (cmd->opcode) {
1858                         /*
1859                          * The first set of opcodes compares the packet's
1860                          * fields with some pattern, setting 'match' if a
1861                          * match is found. At the end of the loop there is
1862                          * logic to deal with F_NOT and F_OR flags associated
1863                          * with the opcode.
1864                          */
1865                         case O_NOP:
1866                                 match = 1;
1867                                 break;
1868
1869                         case O_FORWARD_MAC:
1870                                 printf("ipfw: opcode %d unimplemented\n",
1871                                     cmd->opcode);
1872                                 break;
1873
1874                         case O_GID:
1875                         case O_UID:
1876                         case O_JAIL:
1877                                 /*
1878                                  * We only check offset == 0 && proto != 0,
1879                                  * as this ensures that we have a
1880                                  * packet with the ports info.
1881                                  */
1882                                 if (offset != 0)
1883                                         break;
1884                                 if (proto == IPPROTO_TCP ||
1885                                     proto == IPPROTO_UDP ||
1886                                     proto == IPPROTO_UDPLITE)
1887                                         match = check_uidgid(
1888                                                     (ipfw_insn_u32 *)cmd,
1889                                                     args, &ucred_lookup,
1890 #ifdef __FreeBSD__
1891                                                     &ucred_cache);
1892 #else
1893                                                     (void *)&ucred_cache);
1894 #endif
1895                                 break;
1896
1897                         case O_RECV:
1898                                 match = iface_match(iif, (ipfw_insn_if *)cmd,
1899                                     chain, &tablearg);
1900                                 break;
1901
1902                         case O_XMIT:
1903                                 match = iface_match(oif, (ipfw_insn_if *)cmd,
1904                                     chain, &tablearg);
1905                                 break;
1906
1907                         case O_VIA:
1908                                 match = iface_match(args->ifp,
1909                                     (ipfw_insn_if *)cmd, chain, &tablearg);
1910                                 break;
1911
1912                         case O_MACADDR2:
1913                                 if (args->flags & IPFW_ARGS_ETHER) {
1914                                         u_int32_t *want = (u_int32_t *)
1915                                                 ((ipfw_insn_mac *)cmd)->addr;
1916                                         u_int32_t *mask = (u_int32_t *)
1917                                                 ((ipfw_insn_mac *)cmd)->mask;
1918                                         u_int32_t *hdr = (u_int32_t *)eh;
1919
1920                                         match =
1921                                             ( want[0] == (hdr[0] & mask[0]) &&
1922                                               want[1] == (hdr[1] & mask[1]) &&
1923                                               want[2] == (hdr[2] & mask[2]) );
1924                                 }
1925                                 break;
1926
1927                         case O_MAC_TYPE:
1928                                 if (args->flags & IPFW_ARGS_ETHER) {
1929                                         u_int16_t *p =
1930                                             ((ipfw_insn_u16 *)cmd)->ports;
1931                                         int i;
1932
1933                                         for (i = cmdlen - 1; !match && i>0;
1934                                             i--, p += 2)
1935                                                 match =
1936                                                     (ntohs(eh->ether_type) >=
1937                                                     p[0] &&
1938                                                     ntohs(eh->ether_type) <=
1939                                                     p[1]);
1940                                 }
1941                                 break;
1942
1943                         case O_FRAG:
1944                                 match = (offset != 0);
1945                                 break;
1946
1947                         case O_IN:      /* "out" is "not in" */
1948                                 match = (oif == NULL);
1949                                 break;
1950
1951                         case O_LAYER2:
1952                                 match = (args->flags & IPFW_ARGS_ETHER);
1953                                 break;
1954
1955                         case O_DIVERTED:
1956                                 if ((args->flags & IPFW_ARGS_REF) == 0)
1957                                         break;
1958                                 /*
1959                                  * For diverted packets, args->rule.info
1960                                  * contains the divert port (in host format)
1961                                  * reason and direction.
1962                                  */
1963                                 match = ((args->rule.info & IPFW_IS_MASK) ==
1964                                     IPFW_IS_DIVERT) && (
1965                                     ((args->rule.info & IPFW_INFO_IN) ?
1966                                         1: 2) & cmd->arg1);
1967                                 break;
1968
1969                         case O_PROTO:
1970                                 /*
1971                                  * We do not allow an arg of 0 so the
1972                                  * check of "proto" only suffices.
1973                                  */
1974                                 match = (proto == cmd->arg1);
1975                                 break;
1976
1977                         case O_IP_SRC:
1978                                 match = is_ipv4 &&
1979                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1980                                     src_ip.s_addr);
1981                                 break;
1982
1983                         case O_IP_DST_LOOKUP:
1984                         {
1985                                 void *pkey;
1986                                 uint32_t vidx, key;
1987                                 uint16_t keylen;
1988
1989                                 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1990                                         /* Determine lookup key type */
1991                                         vidx = ((ipfw_insn_u32 *)cmd)->d[1];
1992                                         if (vidx != 4 /* uid */ &&
1993                                             vidx != 5 /* jail */ &&
1994                                             is_ipv6 == 0 && is_ipv4 == 0)
1995                                                 break;
1996                                         /* Determine key length */
1997                                         if (vidx == 0 /* dst-ip */ ||
1998                                             vidx == 1 /* src-ip */)
1999                                                 keylen = is_ipv6 ?
2000                                                     sizeof(struct in6_addr):
2001                                                     sizeof(in_addr_t);
2002                                         else {
2003                                                 keylen = sizeof(key);
2004                                                 pkey = &key;
2005                                         }
2006                                         if (vidx == 0 /* dst-ip */)
2007                                                 pkey = is_ipv4 ? (void *)&dst_ip:
2008                                                     (void *)&args->f_id.dst_ip6;
2009                                         else if (vidx == 1 /* src-ip */)
2010                                                 pkey = is_ipv4 ? (void *)&src_ip:
2011                                                     (void *)&args->f_id.src_ip6;
2012                                         else if (vidx == 6 /* dscp */) {
2013                                                 if (is_ipv4)
2014                                                         key = ip->ip_tos >> 2;
2015                                                 else {
2016                                                         key = args->f_id.flow_id6;
2017                                                         key = (key & 0x0f) << 2 |
2018                                                             (key & 0xf000) >> 14;
2019                                                 }
2020                                                 key &= 0x3f;
2021                                         } else if (vidx == 2 /* dst-port */ ||
2022                                             vidx == 3 /* src-port */) {
2023                                                 /* Skip fragments */
2024                                                 if (offset != 0)
2025                                                         break;
2026                                                 /* Skip proto without ports */
2027                                                 if (proto != IPPROTO_TCP &&
2028                                                     proto != IPPROTO_UDP &&
2029                                                     proto != IPPROTO_UDPLITE &&
2030                                                     proto != IPPROTO_SCTP)
2031                                                         break;
2032                                                 if (vidx == 2 /* dst-port */)
2033                                                         key = dst_port;
2034                                                 else
2035                                                         key = src_port;
2036                                         }
2037 #ifndef USERSPACE
2038                                         else if (vidx == 4 /* uid */ ||
2039                                             vidx == 5 /* jail */) {
2040                                                 check_uidgid(
2041                                                     (ipfw_insn_u32 *)cmd,
2042                                                     args, &ucred_lookup,
2043 #ifdef __FreeBSD__
2044                                                     &ucred_cache);
2045                                                 if (vidx == 4 /* uid */)
2046                                                         key = ucred_cache->cr_uid;
2047                                                 else if (vidx == 5 /* jail */)
2048                                                         key = ucred_cache->cr_prison->pr_id;
2049 #else /* !__FreeBSD__ */
2050                                                     (void *)&ucred_cache);
2051                                                 if (vidx == 4 /* uid */)
2052                                                         key = ucred_cache.uid;
2053                                                 else if (vidx == 5 /* jail */)
2054                                                         key = ucred_cache.xid;
2055 #endif /* !__FreeBSD__ */
2056                                         }
2057 #endif /* !USERSPACE */
2058                                         else
2059                                                 break;
2060                                         match = ipfw_lookup_table(chain,
2061                                             cmd->arg1, keylen, pkey, &vidx);
2062                                         if (!match)
2063                                                 break;
2064                                         tablearg = vidx;
2065                                         break;
2066                                 }
2067                                 /* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */
2068                                 /* FALLTHROUGH */
2069                         }
2070                         case O_IP_SRC_LOOKUP:
2071                         {
2072                                 void *pkey;
2073                                 uint32_t vidx;
2074                                 uint16_t keylen;
2075
2076                                 if (is_ipv4) {
2077                                         keylen = sizeof(in_addr_t);
2078                                         if (cmd->opcode == O_IP_DST_LOOKUP)
2079                                                 pkey = &dst_ip;
2080                                         else
2081                                                 pkey = &src_ip;
2082                                 } else if (is_ipv6) {
2083                                         keylen = sizeof(struct in6_addr);
2084                                         if (cmd->opcode == O_IP_DST_LOOKUP)
2085                                                 pkey = &args->f_id.dst_ip6;
2086                                         else
2087                                                 pkey = &args->f_id.src_ip6;
2088                                 } else
2089                                         break;
2090                                 match = ipfw_lookup_table(chain, cmd->arg1,
2091                                     keylen, pkey, &vidx);
2092                                 if (!match)
2093                                         break;
2094                                 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) {
2095                                         match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2096                                             TARG_VAL(chain, vidx, tag);
2097                                         if (!match)
2098                                                 break;
2099                                 }
2100                                 tablearg = vidx;
2101                                 break;
2102                         }
2103
2104                         case O_IP_FLOW_LOOKUP:
2105                                 {
2106                                         uint32_t v = 0;
2107                                         match = ipfw_lookup_table(chain,
2108                                             cmd->arg1, 0, &args->f_id, &v);
2109                                         if (!match)
2110                                                 break;
2111                                         if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
2112                                                 match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2113                                                     TARG_VAL(chain, v, tag);
2114                                         if (match)
2115                                                 tablearg = v;
2116                                 }
2117                                 break;
2118                         case O_IP_SRC_MASK:
2119                         case O_IP_DST_MASK:
2120                                 if (is_ipv4) {
2121                                     uint32_t a =
2122                                         (cmd->opcode == O_IP_DST_MASK) ?
2123                                             dst_ip.s_addr : src_ip.s_addr;
2124                                     uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2125                                     int i = cmdlen-1;
2126
2127                                     for (; !match && i>0; i-= 2, p+= 2)
2128                                         match = (p[0] == (a & p[1]));
2129                                 }
2130                                 break;
2131
2132                         case O_IP_SRC_ME:
2133                                 if (is_ipv4) {
2134                                         match = in_localip(src_ip);
2135                                         break;
2136                                 }
2137 #ifdef INET6
2138                                 /* FALLTHROUGH */
2139                         case O_IP6_SRC_ME:
2140                                 match = is_ipv6 &&
2141                                     ipfw_localip6(&args->f_id.src_ip6);
2142 #endif
2143                                 break;
2144
2145                         case O_IP_DST_SET:
2146                         case O_IP_SRC_SET:
2147                                 if (is_ipv4) {
2148                                         u_int32_t *d = (u_int32_t *)(cmd+1);
2149                                         u_int32_t addr =
2150                                             cmd->opcode == O_IP_DST_SET ?
2151                                                 args->f_id.dst_ip :
2152                                                 args->f_id.src_ip;
2153
2154                                             if (addr < d[0])
2155                                                     break;
2156                                             addr -= d[0]; /* subtract base */
2157                                             match = (addr < cmd->arg1) &&
2158                                                 ( d[ 1 + (addr>>5)] &
2159                                                   (1<<(addr & 0x1f)) );
2160                                 }
2161                                 break;
2162
2163                         case O_IP_DST:
2164                                 match = is_ipv4 &&
2165                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2166                                     dst_ip.s_addr);
2167                                 break;
2168
2169                         case O_IP_DST_ME:
2170                                 if (is_ipv4) {
2171                                         match = in_localip(dst_ip);
2172                                         break;
2173                                 }
2174 #ifdef INET6
2175                                 /* FALLTHROUGH */
2176                         case O_IP6_DST_ME:
2177                                 match = is_ipv6 &&
2178                                     ipfw_localip6(&args->f_id.dst_ip6);
2179 #endif
2180                                 break;
2181
2182
2183                         case O_IP_SRCPORT:
2184                         case O_IP_DSTPORT:
2185                                 /*
2186                                  * offset == 0 && proto != 0 is enough
2187                                  * to guarantee that we have a
2188                                  * packet with port info.
2189                                  */
2190                                 if ((proto == IPPROTO_UDP ||
2191                                     proto == IPPROTO_UDPLITE ||
2192                                     proto == IPPROTO_TCP ||
2193                                     proto == IPPROTO_SCTP) && offset == 0) {
2194                                         u_int16_t x =
2195                                             (cmd->opcode == O_IP_SRCPORT) ?
2196                                                 src_port : dst_port ;
2197                                         u_int16_t *p =
2198                                             ((ipfw_insn_u16 *)cmd)->ports;
2199                                         int i;
2200
2201                                         for (i = cmdlen - 1; !match && i>0;
2202                                             i--, p += 2)
2203                                                 match = (x>=p[0] && x<=p[1]);
2204                                 }
2205                                 break;
2206
2207                         case O_ICMPTYPE:
2208                                 match = (offset == 0 && proto==IPPROTO_ICMP &&
2209                                     icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
2210                                 break;
2211
2212 #ifdef INET6
2213                         case O_ICMP6TYPE:
2214                                 match = is_ipv6 && offset == 0 &&
2215                                     proto==IPPROTO_ICMPV6 &&
2216                                     icmp6type_match(
2217                                         ICMP6(ulp)->icmp6_type,
2218                                         (ipfw_insn_u32 *)cmd);
2219                                 break;
2220 #endif /* INET6 */
2221
2222                         case O_IPOPT:
2223                                 match = (is_ipv4 &&
2224                                     ipopts_match(ip, cmd) );
2225                                 break;
2226
2227                         case O_IPVER:
2228                                 match = (is_ipv4 &&
2229                                     cmd->arg1 == ip->ip_v);
2230                                 break;
2231
2232                         case O_IPID:
2233                         case O_IPTTL:
2234                                 if (!is_ipv4)
2235                                         break;
2236                         case O_IPLEN:
2237                                 {       /* only for IP packets */
2238                                     uint16_t x;
2239                                     uint16_t *p;
2240                                     int i;
2241
2242                                     if (cmd->opcode == O_IPLEN)
2243                                         x = iplen;
2244                                     else if (cmd->opcode == O_IPTTL)
2245                                         x = ip->ip_ttl;
2246                                     else /* must be IPID */
2247                                         x = ntohs(ip->ip_id);
2248                                     if (cmdlen == 1) {
2249                                         match = (cmd->arg1 == x);
2250                                         break;
2251                                     }
2252                                     /* otherwise we have ranges */
2253                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2254                                     i = cmdlen - 1;
2255                                     for (; !match && i>0; i--, p += 2)
2256                                         match = (x >= p[0] && x <= p[1]);
2257                                 }
2258                                 break;
2259
2260                         case O_IPPRECEDENCE:
2261                                 match = (is_ipv4 &&
2262                                     (cmd->arg1 == (ip->ip_tos & 0xe0)) );
2263                                 break;
2264
2265                         case O_IPTOS:
2266                                 match = (is_ipv4 &&
2267                                     flags_match(cmd, ip->ip_tos));
2268                                 break;
2269
2270                         case O_DSCP:
2271                             {
2272                                 uint32_t *p;
2273                                 uint16_t x;
2274
2275                                 p = ((ipfw_insn_u32 *)cmd)->d;
2276
2277                                 if (is_ipv4)
2278                                         x = ip->ip_tos >> 2;
2279                                 else if (is_ipv6) {
2280                                         uint8_t *v;
2281                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
2282                                         x = (*v & 0x0F) << 2;
2283                                         v++;
2284                                         x |= *v >> 6;
2285                                 } else
2286                                         break;
2287
2288                                 /* DSCP bitmask is stored as low_u32 high_u32 */
2289                                 if (x >= 32)
2290                                         match = *(p + 1) & (1 << (x - 32));
2291                                 else
2292                                         match = *p & (1 << x);
2293                             }
2294                                 break;
2295
2296                         case O_TCPDATALEN:
2297                                 if (proto == IPPROTO_TCP && offset == 0) {
2298                                     struct tcphdr *tcp;
2299                                     uint16_t x;
2300                                     uint16_t *p;
2301                                     int i;
2302 #ifdef INET6
2303                                     if (is_ipv6) {
2304                                             struct ip6_hdr *ip6;
2305
2306                                             ip6 = (struct ip6_hdr *)ip;
2307                                             if (ip6->ip6_plen == 0) {
2308                                                     /*
2309                                                      * Jumbo payload is not
2310                                                      * supported by this
2311                                                      * opcode.
2312                                                      */
2313                                                     break;
2314                                             }
2315                                             x = iplen - hlen;
2316                                     } else
2317 #endif /* INET6 */
2318                                             x = iplen - (ip->ip_hl << 2);
2319                                     tcp = TCP(ulp);
2320                                     x -= tcp->th_off << 2;
2321                                     if (cmdlen == 1) {
2322                                         match = (cmd->arg1 == x);
2323                                         break;
2324                                     }
2325                                     /* otherwise we have ranges */
2326                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2327                                     i = cmdlen - 1;
2328                                     for (; !match && i>0; i--, p += 2)
2329                                         match = (x >= p[0] && x <= p[1]);
2330                                 }
2331                                 break;
2332
2333                         case O_TCPFLAGS:
2334                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2335                                     flags_match(cmd, TCP(ulp)->th_flags));
2336                                 break;
2337
2338                         case O_TCPOPTS:
2339                                 if (proto == IPPROTO_TCP && offset == 0 && ulp){
2340                                         PULLUP_LEN_LOCKED(hlen, ulp,
2341                                             (TCP(ulp)->th_off << 2));
2342                                         match = tcpopts_match(TCP(ulp), cmd);
2343                                 }
2344                                 break;
2345
2346                         case O_TCPSEQ:
2347                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2348                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
2349                                         TCP(ulp)->th_seq);
2350                                 break;
2351
2352                         case O_TCPACK:
2353                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2354                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
2355                                         TCP(ulp)->th_ack);
2356                                 break;
2357
2358                         case O_TCPMSS:
2359                                 if (proto == IPPROTO_TCP &&
2360                                     (args->f_id._flags & TH_SYN) != 0 &&
2361                                     ulp != NULL) {
2362                                         uint16_t mss, *p;
2363                                         int i;
2364
2365                                         PULLUP_LEN_LOCKED(hlen, ulp,
2366                                             (TCP(ulp)->th_off << 2));
2367                                         if ((tcpopts_parse(TCP(ulp), &mss) &
2368                                             IP_FW_TCPOPT_MSS) == 0)
2369                                                 break;
2370                                         if (cmdlen == 1) {
2371                                                 match = (cmd->arg1 == mss);
2372                                                 break;
2373                                         }
2374                                         /* Otherwise we have ranges. */
2375                                         p = ((ipfw_insn_u16 *)cmd)->ports;
2376                                         i = cmdlen - 1;
2377                                         for (; !match && i > 0; i--, p += 2)
2378                                                 match = (mss >= p[0] &&
2379                                                     mss <= p[1]);
2380                                 }
2381                                 break;
2382
2383                         case O_TCPWIN:
2384                                 if (proto == IPPROTO_TCP && offset == 0) {
2385                                     uint16_t x;
2386                                     uint16_t *p;
2387                                     int i;
2388
2389                                     x = ntohs(TCP(ulp)->th_win);
2390                                     if (cmdlen == 1) {
2391                                         match = (cmd->arg1 == x);
2392                                         break;
2393                                     }
2394                                     /* Otherwise we have ranges. */
2395                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2396                                     i = cmdlen - 1;
2397                                     for (; !match && i > 0; i--, p += 2)
2398                                         match = (x >= p[0] && x <= p[1]);
2399                                 }
2400                                 break;
2401
2402                         case O_ESTAB:
2403                                 /* reject packets which have SYN only */
2404                                 /* XXX should i also check for TH_ACK ? */
2405                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2406                                     (TCP(ulp)->th_flags &
2407                                      (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
2408                                 break;
2409
2410                         case O_ALTQ: {
2411                                 struct pf_mtag *at;
2412                                 struct m_tag *mtag;
2413                                 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
2414
2415                                 /*
2416                                  * ALTQ uses mbuf tags from another
2417                                  * packet filtering system - pf(4).
2418                                  * We allocate a tag in its format
2419                                  * and fill it in, pretending to be pf(4).
2420                                  */
2421                                 match = 1;
2422                                 at = pf_find_mtag(m);
2423                                 if (at != NULL && at->qid != 0)
2424                                         break;
2425                                 mtag = m_tag_get(PACKET_TAG_PF,
2426                                     sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
2427                                 if (mtag == NULL) {
2428                                         /*
2429                                          * Let the packet fall back to the
2430                                          * default ALTQ.
2431                                          */
2432                                         break;
2433                                 }
2434                                 m_tag_prepend(m, mtag);
2435                                 at = (struct pf_mtag *)(mtag + 1);
2436                                 at->qid = altq->qid;
2437                                 at->hdr = ip;
2438                                 break;
2439                         }
2440
2441                         case O_LOG:
2442                                 ipfw_log(chain, f, hlen, args,
2443                                     offset | ip6f_mf, tablearg, ip);
2444                                 match = 1;
2445                                 break;
2446
2447                         case O_PROB:
2448                                 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
2449                                 break;
2450
2451                         case O_VERREVPATH:
2452                                 /* Outgoing packets automatically pass/match */
2453                                 match = (args->flags & IPFW_ARGS_OUT ||
2454                                     (
2455 #ifdef INET6
2456                                     is_ipv6 ?
2457                                         verify_path6(&(args->f_id.src_ip6),
2458                                             iif, args->f_id.fib) :
2459 #endif
2460                                     verify_path(src_ip, iif, args->f_id.fib)));
2461                                 break;
2462
2463                         case O_VERSRCREACH:
2464                                 /* Outgoing packets automatically pass/match */
2465                                 match = (hlen > 0 && ((oif != NULL) || (
2466 #ifdef INET6
2467                                     is_ipv6 ?
2468                                         verify_path6(&(args->f_id.src_ip6),
2469                                             NULL, args->f_id.fib) :
2470 #endif
2471                                     verify_path(src_ip, NULL, args->f_id.fib))));
2472                                 break;
2473
2474                         case O_ANTISPOOF:
2475                                 /* Outgoing packets automatically pass/match */
2476                                 if (oif == NULL && hlen > 0 &&
2477                                     (  (is_ipv4 && in_localaddr(src_ip))
2478 #ifdef INET6
2479                                     || (is_ipv6 &&
2480                                         in6_localaddr(&(args->f_id.src_ip6)))
2481 #endif
2482                                     ))
2483                                         match =
2484 #ifdef INET6
2485                                             is_ipv6 ? verify_path6(
2486                                                 &(args->f_id.src_ip6), iif,
2487                                                 args->f_id.fib) :
2488 #endif
2489                                             verify_path(src_ip, iif,
2490                                                 args->f_id.fib);
2491                                 else
2492                                         match = 1;
2493                                 break;
2494
2495                         case O_IPSEC:
2496                                 match = (m_tag_find(m,
2497                                     PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
2498                                 /* otherwise no match */
2499                                 break;
2500
2501 #ifdef INET6
2502                         case O_IP6_SRC:
2503                                 match = is_ipv6 &&
2504                                     IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
2505                                     &((ipfw_insn_ip6 *)cmd)->addr6);
2506                                 break;
2507
2508                         case O_IP6_DST:
2509                                 match = is_ipv6 &&
2510                                 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
2511                                     &((ipfw_insn_ip6 *)cmd)->addr6);
2512                                 break;
2513                         case O_IP6_SRC_MASK:
2514                         case O_IP6_DST_MASK:
2515                                 if (is_ipv6) {
2516                                         int i = cmdlen - 1;
2517                                         struct in6_addr p;
2518                                         struct in6_addr *d =
2519                                             &((ipfw_insn_ip6 *)cmd)->addr6;
2520
2521                                         for (; !match && i > 0; d += 2,
2522                                             i -= F_INSN_SIZE(struct in6_addr)
2523                                             * 2) {
2524                                                 p = (cmd->opcode ==
2525                                                     O_IP6_SRC_MASK) ?
2526                                                     args->f_id.src_ip6:
2527                                                     args->f_id.dst_ip6;
2528                                                 APPLY_MASK(&p, &d[1]);
2529                                                 match =
2530                                                     IN6_ARE_ADDR_EQUAL(&d[0],
2531                                                     &p);
2532                                         }
2533                                 }
2534                                 break;
2535
2536                         case O_FLOW6ID:
2537                                 match = is_ipv6 &&
2538                                     flow6id_match(args->f_id.flow_id6,
2539                                     (ipfw_insn_u32 *) cmd);
2540                                 break;
2541
2542                         case O_EXT_HDR:
2543                                 match = is_ipv6 &&
2544                                     (ext_hd & ((ipfw_insn *) cmd)->arg1);
2545                                 break;
2546
2547                         case O_IP6:
2548                                 match = is_ipv6;
2549                                 break;
2550 #endif
2551
2552                         case O_IP4:
2553                                 match = is_ipv4;
2554                                 break;
2555
2556                         case O_TAG: {
2557                                 struct m_tag *mtag;
2558                                 uint32_t tag = TARG(cmd->arg1, tag);
2559
2560                                 /* Packet is already tagged with this tag? */
2561                                 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
2562
2563                                 /* We have `untag' action when F_NOT flag is
2564                                  * present. And we must remove this mtag from
2565                                  * mbuf and reset `match' to zero (`match' will
2566                                  * be inversed later).
2567                                  * Otherwise we should allocate new mtag and
2568                                  * push it into mbuf.
2569                                  */
2570                                 if (cmd->len & F_NOT) { /* `untag' action */
2571                                         if (mtag != NULL)
2572                                                 m_tag_delete(m, mtag);
2573                                         match = 0;
2574                                 } else {
2575                                         if (mtag == NULL) {
2576                                                 mtag = m_tag_alloc( MTAG_IPFW,
2577                                                     tag, 0, M_NOWAIT);
2578                                                 if (mtag != NULL)
2579                                                         m_tag_prepend(m, mtag);
2580                                         }
2581                                         match = 1;
2582                                 }
2583                                 break;
2584                         }
2585
2586                         case O_FIB: /* try match the specified fib */
2587                                 if (args->f_id.fib == cmd->arg1)
2588                                         match = 1;
2589                                 break;
2590
2591                         case O_SOCKARG: {
2592 #ifndef USERSPACE       /* not supported in userspace */
2593                                 struct inpcb *inp = args->inp;
2594                                 struct inpcbinfo *pi;
2595
2596                                 if (is_ipv6) /* XXX can we remove this ? */
2597                                         break;
2598
2599                                 if (proto == IPPROTO_TCP)
2600                                         pi = &V_tcbinfo;
2601                                 else if (proto == IPPROTO_UDP)
2602                                         pi = &V_udbinfo;
2603                                 else if (proto == IPPROTO_UDPLITE)
2604                                         pi = &V_ulitecbinfo;
2605                                 else
2606                                         break;
2607
2608                                 /*
2609                                  * XXXRW: so_user_cookie should almost
2610                                  * certainly be inp_user_cookie?
2611                                  */
2612
2613                                 /* For incoming packet, lookup up the
2614                                 inpcb using the src/dest ip/port tuple */
2615                                 if (inp == NULL) {
2616                                         inp = in_pcblookup(pi,
2617                                                 src_ip, htons(src_port),
2618                                                 dst_ip, htons(dst_port),
2619                                                 INPLOOKUP_RLOCKPCB, NULL);
2620                                         if (inp != NULL) {
2621                                                 tablearg =
2622                                                     inp->inp_socket->so_user_cookie;
2623                                                 if (tablearg)
2624                                                         match = 1;
2625                                                 INP_RUNLOCK(inp);
2626                                         }
2627                                 } else {
2628                                         if (inp->inp_socket) {
2629                                                 tablearg =
2630                                                     inp->inp_socket->so_user_cookie;
2631                                                 if (tablearg)
2632                                                         match = 1;
2633                                         }
2634                                 }
2635 #endif /* !USERSPACE */
2636                                 break;
2637                         }
2638
2639                         case O_TAGGED: {
2640                                 struct m_tag *mtag;
2641                                 uint32_t tag = TARG(cmd->arg1, tag);
2642
2643                                 if (cmdlen == 1) {
2644                                         match = m_tag_locate(m, MTAG_IPFW,
2645                                             tag, NULL) != NULL;
2646                                         break;
2647                                 }
2648
2649                                 /* we have ranges */
2650                                 for (mtag = m_tag_first(m);
2651                                     mtag != NULL && !match;
2652                                     mtag = m_tag_next(m, mtag)) {
2653                                         uint16_t *p;
2654                                         int i;
2655
2656                                         if (mtag->m_tag_cookie != MTAG_IPFW)
2657                                                 continue;
2658
2659                                         p = ((ipfw_insn_u16 *)cmd)->ports;
2660                                         i = cmdlen - 1;
2661                                         for(; !match && i > 0; i--, p += 2)
2662                                                 match =
2663                                                     mtag->m_tag_id >= p[0] &&
2664                                                     mtag->m_tag_id <= p[1];
2665                                 }
2666                                 break;
2667                         }
2668
2669                         /*
2670                          * The second set of opcodes represents 'actions',
2671                          * i.e. the terminal part of a rule once the packet
2672                          * matches all previous patterns.
2673                          * Typically there is only one action for each rule,
2674                          * and the opcode is stored at the end of the rule
2675                          * (but there are exceptions -- see below).
2676                          *
2677                          * In general, here we set retval and terminate the
2678                          * outer loop (would be a 'break 3' in some language,
2679                          * but we need to set l=0, done=1)
2680                          *
2681                          * Exceptions:
2682                          * O_COUNT and O_SKIPTO actions:
2683                          *   instead of terminating, we jump to the next rule
2684                          *   (setting l=0), or to the SKIPTO target (setting
2685                          *   f/f_len, cmd and l as needed), respectively.
2686                          *
2687                          * O_TAG, O_LOG and O_ALTQ action parameters:
2688                          *   perform some action and set match = 1;
2689                          *
2690                          * O_LIMIT and O_KEEP_STATE: these opcodes are
2691                          *   not real 'actions', and are stored right
2692                          *   before the 'action' part of the rule (one
2693                          *   exception is O_SKIP_ACTION which could be
2694                          *   between these opcodes and 'action' one).
2695                          *   These opcodes try to install an entry in the
2696                          *   state tables; if successful, we continue with
2697                          *   the next opcode (match=1; break;), otherwise
2698                          *   the packet must be dropped (set retval,
2699                          *   break loops with l=0, done=1)
2700                          *
2701                          * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2702                          *   cause a lookup of the state table, and a jump
2703                          *   to the 'action' part of the parent rule
2704                          *   if an entry is found, or
2705                          *   (CHECK_STATE only) a jump to the next rule if
2706                          *   the entry is not found.
2707                          *   The result of the lookup is cached so that
2708                          *   further instances of these opcodes become NOPs.
2709                          *   The jump to the next rule is done by setting
2710                          *   l=0, cmdlen=0.
2711                          *
2712                          * O_SKIP_ACTION: this opcode is not a real 'action'
2713                          *  either, and is stored right before the 'action'
2714                          *  part of the rule, right after the O_KEEP_STATE
2715                          *  opcode. It causes match failure so the real
2716                          *  'action' could be executed only if the rule
2717                          *  is checked via dynamic rule from the state
2718                          *  table, as in such case execution starts
2719                          *  from the true 'action' opcode directly.
2720                          *
2721                          */
2722                         case O_LIMIT:
2723                         case O_KEEP_STATE:
2724                                 if (ipfw_dyn_install_state(chain, f,
2725                                     (ipfw_insn_limit *)cmd, args, ulp,
2726                                     pktlen, &dyn_info, tablearg)) {
2727                                         /* error or limit violation */
2728                                         retval = IP_FW_DENY;
2729                                         l = 0;  /* exit inner loop */
2730                                         done = 1; /* exit outer loop */
2731                                 }
2732                                 match = 1;
2733                                 break;
2734
2735                         case O_PROBE_STATE:
2736                         case O_CHECK_STATE:
2737                                 /*
2738                                  * dynamic rules are checked at the first
2739                                  * keep-state or check-state occurrence,
2740                                  * with the result being stored in dyn_info.
2741                                  * The compiler introduces a PROBE_STATE
2742                                  * instruction for us when we have a
2743                                  * KEEP_STATE (because PROBE_STATE needs
2744                                  * to be run first).
2745                                  */
2746                                 if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) &&
2747                                     (q = ipfw_dyn_lookup_state(args, ulp,
2748                                     pktlen, cmd, &dyn_info)) != NULL) {
2749                                         /*
2750                                          * Found dynamic entry, jump to the
2751                                          * 'action' part of the parent rule
2752                                          * by setting f, cmd, l and clearing
2753                                          * cmdlen.
2754                                          */
2755                                         f = q;
2756                                         f_pos = dyn_info.f_pos;
2757                                         cmd = ACTION_PTR(f);
2758                                         l = f->cmd_len - f->act_ofs;
2759                                         cmdlen = 0;
2760                                         match = 1;
2761                                         break;
2762                                 }
2763                                 /*
2764                                  * Dynamic entry not found. If CHECK_STATE,
2765                                  * skip to next rule, if PROBE_STATE just
2766                                  * ignore and continue with next opcode.
2767                                  */
2768                                 if (cmd->opcode == O_CHECK_STATE)
2769                                         l = 0;  /* exit inner loop */
2770                                 match = 1;
2771                                 break;
2772
2773                         case O_SKIP_ACTION:
2774                                 match = 0;      /* skip to the next rule */
2775                                 l = 0;          /* exit inner loop */
2776                                 break;
2777
2778                         case O_ACCEPT:
2779                                 retval = 0;     /* accept */
2780                                 l = 0;          /* exit inner loop */
2781                                 done = 1;       /* exit outer loop */
2782                                 break;
2783
2784                         case O_PIPE:
2785                         case O_QUEUE:
2786                                 set_match(args, f_pos, chain);
2787                                 args->rule.info = TARG(cmd->arg1, pipe);
2788                                 if (cmd->opcode == O_PIPE)
2789                                         args->rule.info |= IPFW_IS_PIPE;
2790                                 if (V_fw_one_pass)
2791                                         args->rule.info |= IPFW_ONEPASS;
2792                                 retval = IP_FW_DUMMYNET;
2793                                 l = 0;          /* exit inner loop */
2794                                 done = 1;       /* exit outer loop */
2795                                 break;
2796
2797                         case O_DIVERT:
2798                         case O_TEE:
2799                                 if (args->flags & IPFW_ARGS_ETHER)
2800                                         break;  /* not on layer 2 */
2801                                 /* otherwise this is terminal */
2802                                 l = 0;          /* exit inner loop */
2803                                 done = 1;       /* exit outer loop */
2804                                 retval = (cmd->opcode == O_DIVERT) ?
2805                                         IP_FW_DIVERT : IP_FW_TEE;
2806                                 set_match(args, f_pos, chain);
2807                                 args->rule.info = TARG(cmd->arg1, divert);
2808                                 break;
2809
2810                         case O_COUNT:
2811                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2812                                 l = 0;          /* exit inner loop */
2813                                 break;
2814
2815                         case O_SKIPTO:
2816                             IPFW_INC_RULE_COUNTER(f, pktlen);
2817                             f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2818                             /*
2819                              * Skip disabled rules, and re-enter
2820                              * the inner loop with the correct
2821                              * f_pos, f, l and cmd.
2822                              * Also clear cmdlen and skip_or
2823                              */
2824                             for (; f_pos < chain->n_rules - 1 &&
2825                                     (V_set_disable &
2826                                      (1 << chain->map[f_pos]->set));
2827                                     f_pos++)
2828                                 ;
2829                             /* Re-enter the inner loop at the skipto rule. */
2830                             f = chain->map[f_pos];
2831                             l = f->cmd_len;
2832                             cmd = f->cmd;
2833                             match = 1;
2834                             cmdlen = 0;
2835                             skip_or = 0;
2836                             continue;
2837                             break;      /* not reached */
2838
2839                         case O_CALLRETURN: {
2840                                 /*
2841                                  * Implementation of `subroutine' call/return,
2842                                  * in the stack carried in an mbuf tag. This
2843                                  * is different from `skipto' in that any call
2844                                  * address is possible (`skipto' must prevent
2845                                  * backward jumps to avoid endless loops).
2846                                  * We have `return' action when F_NOT flag is
2847                                  * present. The `m_tag_id' field is used as
2848                                  * stack pointer.
2849                                  */
2850                                 struct m_tag *mtag;
2851                                 uint16_t jmpto, *stack;
2852
2853 #define IS_CALL         ((cmd->len & F_NOT) == 0)
2854 #define IS_RETURN       ((cmd->len & F_NOT) != 0)
2855                                 /*
2856                                  * Hand-rolled version of m_tag_locate() with
2857                                  * wildcard `type'.
2858                                  * If not already tagged, allocate new tag.
2859                                  */
2860                                 mtag = m_tag_first(m);
2861                                 while (mtag != NULL) {
2862                                         if (mtag->m_tag_cookie ==
2863                                             MTAG_IPFW_CALL)
2864                                                 break;
2865                                         mtag = m_tag_next(m, mtag);
2866                                 }
2867                                 if (mtag == NULL && IS_CALL) {
2868                                         mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2869                                             IPFW_CALLSTACK_SIZE *
2870                                             sizeof(uint16_t), M_NOWAIT);
2871                                         if (mtag != NULL)
2872                                                 m_tag_prepend(m, mtag);
2873                                 }
2874
2875                                 /*
2876                                  * On error both `call' and `return' just
2877                                  * continue with next rule.
2878                                  */
2879                                 if (IS_RETURN && (mtag == NULL ||
2880                                     mtag->m_tag_id == 0)) {
2881                                         l = 0;          /* exit inner loop */
2882                                         break;
2883                                 }
2884                                 if (IS_CALL && (mtag == NULL ||
2885                                     mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2886                                         printf("ipfw: call stack error, "
2887                                             "go to next rule\n");
2888                                         l = 0;          /* exit inner loop */
2889                                         break;
2890                                 }
2891
2892                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2893                                 stack = (uint16_t *)(mtag + 1);
2894
2895                                 /*
2896                                  * The `call' action may use cached f_pos
2897                                  * (in f->next_rule), whose version is written
2898                                  * in f->next_rule.
2899                                  * The `return' action, however, doesn't have
2900                                  * fixed jump address in cmd->arg1 and can't use
2901                                  * cache.
2902                                  */
2903                                 if (IS_CALL) {
2904                                         stack[mtag->m_tag_id] = f->rulenum;
2905                                         mtag->m_tag_id++;
2906                                         f_pos = JUMP(chain, f, cmd->arg1,
2907                                             tablearg, 1);
2908                                 } else {        /* `return' action */
2909                                         mtag->m_tag_id--;
2910                                         jmpto = stack[mtag->m_tag_id] + 1;
2911                                         f_pos = ipfw_find_rule(chain, jmpto, 0);
2912                                 }
2913
2914                                 /*
2915                                  * Skip disabled rules, and re-enter
2916                                  * the inner loop with the correct
2917                                  * f_pos, f, l and cmd.
2918                                  * Also clear cmdlen and skip_or
2919                                  */
2920                                 for (; f_pos < chain->n_rules - 1 &&
2921                                     (V_set_disable &
2922                                     (1 << chain->map[f_pos]->set)); f_pos++)
2923                                         ;
2924                                 /* Re-enter the inner loop at the dest rule. */
2925                                 f = chain->map[f_pos];
2926                                 l = f->cmd_len;
2927                                 cmd = f->cmd;
2928                                 cmdlen = 0;
2929                                 skip_or = 0;
2930                                 continue;
2931                                 break;  /* NOTREACHED */
2932                         }
2933 #undef IS_CALL
2934 #undef IS_RETURN
2935
2936                         case O_REJECT:
2937                                 /*
2938                                  * Drop the packet and send a reject notice
2939                                  * if the packet is not ICMP (or is an ICMP
2940                                  * query), and it is not multicast/broadcast.
2941                                  */
2942                                 if (hlen > 0 && is_ipv4 && offset == 0 &&
2943                                     (proto != IPPROTO_ICMP ||
2944                                      is_icmp_query(ICMP(ulp))) &&
2945                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2946                                     !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2947                                         send_reject(args, cmd->arg1, iplen, ip);
2948                                         m = args->m;
2949                                 }
2950                                 /* FALLTHROUGH */
2951 #ifdef INET6
2952                         case O_UNREACH6:
2953                                 if (hlen > 0 && is_ipv6 &&
2954                                     ((offset & IP6F_OFF_MASK) == 0) &&
2955                                     (proto != IPPROTO_ICMPV6 ||
2956                                      (is_icmp6_query(icmp6_type) == 1)) &&
2957                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2958                                     !IN6_IS_ADDR_MULTICAST(
2959                                         &args->f_id.dst_ip6)) {
2960                                         send_reject6(args,
2961                                             cmd->opcode == O_REJECT ?
2962                                             map_icmp_unreach(cmd->arg1):
2963                                             cmd->arg1, hlen,
2964                                             (struct ip6_hdr *)ip);
2965                                         m = args->m;
2966                                 }
2967                                 /* FALLTHROUGH */
2968 #endif
2969                         case O_DENY:
2970                                 retval = IP_FW_DENY;
2971                                 l = 0;          /* exit inner loop */
2972                                 done = 1;       /* exit outer loop */
2973                                 break;
2974
2975                         case O_FORWARD_IP:
2976                                 if (args->flags & IPFW_ARGS_ETHER)
2977                                         break;  /* not valid on layer2 pkts */
2978                                 if (q != f ||
2979                                     dyn_info.direction == MATCH_FORWARD) {
2980                                     struct sockaddr_in *sa;
2981
2982                                     sa = &(((ipfw_insn_sa *)cmd)->sa);
2983                                     if (sa->sin_addr.s_addr == INADDR_ANY) {
2984 #ifdef INET6
2985                                         /*
2986                                          * We use O_FORWARD_IP opcode for
2987                                          * fwd rule with tablearg, but tables
2988                                          * now support IPv6 addresses. And
2989                                          * when we are inspecting IPv6 packet,
2990                                          * we can use nh6 field from
2991                                          * table_value as next_hop6 address.
2992                                          */
2993                                         if (is_ipv6) {
2994                                                 struct ip_fw_nh6 *nh6;
2995
2996                                                 args->flags |= IPFW_ARGS_NH6;
2997                                                 nh6 = &args->hopstore6;
2998                                                 nh6->sin6_addr = TARG_VAL(
2999                                                     chain, tablearg, nh6);
3000                                                 nh6->sin6_port = sa->sin_port;
3001                                                 nh6->sin6_scope_id = TARG_VAL(
3002                                                     chain, tablearg, zoneid);
3003                                         } else
3004 #endif
3005                                         {
3006                                                 args->flags |= IPFW_ARGS_NH4;
3007                                                 args->hopstore.sin_port =
3008                                                     sa->sin_port;
3009                                                 sa = &args->hopstore;
3010                                                 sa->sin_family = AF_INET;
3011                                                 sa->sin_len = sizeof(*sa);
3012                                                 sa->sin_addr.s_addr = htonl(
3013                                                     TARG_VAL(chain, tablearg,
3014                                                     nh4));
3015                                         }
3016                                     } else {
3017                                             args->flags |= IPFW_ARGS_NH4PTR;
3018                                             args->next_hop = sa;
3019                                     }
3020                                 }
3021                                 retval = IP_FW_PASS;
3022                                 l = 0;          /* exit inner loop */
3023                                 done = 1;       /* exit outer loop */
3024                                 break;
3025
3026 #ifdef INET6
3027                         case O_FORWARD_IP6:
3028                                 if (args->flags & IPFW_ARGS_ETHER)
3029                                         break;  /* not valid on layer2 pkts */
3030                                 if (q != f ||
3031                                     dyn_info.direction == MATCH_FORWARD) {
3032                                         struct sockaddr_in6 *sin6;
3033
3034                                         sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
3035                                         args->flags |= IPFW_ARGS_NH6PTR;
3036                                         args->next_hop6 = sin6;
3037                                 }
3038                                 retval = IP_FW_PASS;
3039                                 l = 0;          /* exit inner loop */
3040                                 done = 1;       /* exit outer loop */
3041                                 break;
3042 #endif
3043
3044                         case O_NETGRAPH:
3045                         case O_NGTEE:
3046                                 set_match(args, f_pos, chain);
3047                                 args->rule.info = TARG(cmd->arg1, netgraph);
3048                                 if (V_fw_one_pass)
3049                                         args->rule.info |= IPFW_ONEPASS;
3050                                 retval = (cmd->opcode == O_NETGRAPH) ?
3051                                     IP_FW_NETGRAPH : IP_FW_NGTEE;
3052                                 l = 0;          /* exit inner loop */
3053                                 done = 1;       /* exit outer loop */
3054                                 break;
3055
3056                         case O_SETFIB: {
3057                                 uint32_t fib;
3058
3059                                 IPFW_INC_RULE_COUNTER(f, pktlen);
3060                                 fib = TARG(cmd->arg1, fib) & 0x7FFF;
3061                                 if (fib >= rt_numfibs)
3062                                         fib = 0;
3063                                 M_SETFIB(m, fib);
3064                                 args->f_id.fib = fib; /* XXX */
3065                                 l = 0;          /* exit inner loop */
3066                                 break;
3067                         }
3068
3069                         case O_SETDSCP: {
3070                                 uint16_t code;
3071
3072                                 code = TARG(cmd->arg1, dscp) & 0x3F;
3073                                 l = 0;          /* exit inner loop */
3074                                 if (is_ipv4) {
3075                                         uint16_t old;
3076
3077                                         old = *(uint16_t *)ip;
3078                                         ip->ip_tos = (code << 2) |
3079                                             (ip->ip_tos & 0x03);
3080                                         ip->ip_sum = cksum_adjust(ip->ip_sum,
3081                                             old, *(uint16_t *)ip);
3082                                 } else if (is_ipv6) {
3083                                         uint8_t *v;
3084
3085                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
3086                                         *v = (*v & 0xF0) | (code >> 2);
3087                                         v++;
3088                                         *v = (*v & 0x3F) | ((code & 0x03) << 6);
3089                                 } else
3090                                         break;
3091
3092                                 IPFW_INC_RULE_COUNTER(f, pktlen);
3093                                 break;
3094                         }
3095
3096                         case O_NAT:
3097                                 l = 0;          /* exit inner loop */
3098                                 done = 1;       /* exit outer loop */
3099                                 /*
3100                                  * Ensure that we do not invoke NAT handler for
3101                                  * non IPv4 packets. Libalias expects only IPv4.
3102                                  */
3103                                 if (!is_ipv4 || !IPFW_NAT_LOADED) {
3104                                     retval = IP_FW_DENY;
3105                                     break;
3106                                 }
3107
3108                                 struct cfg_nat *t;
3109                                 int nat_id;
3110
3111                                 args->rule.info = 0;
3112                                 set_match(args, f_pos, chain);
3113                                 /* Check if this is 'global' nat rule */
3114                                 if (cmd->arg1 == IP_FW_NAT44_GLOBAL) {
3115                                         retval = ipfw_nat_ptr(args, NULL, m);
3116                                         break;
3117                                 }
3118                                 t = ((ipfw_insn_nat *)cmd)->nat;
3119                                 if (t == NULL) {
3120                                         nat_id = TARG(cmd->arg1, nat);
3121                                         t = (*lookup_nat_ptr)(&chain->nat, nat_id);
3122
3123                                         if (t == NULL) {
3124                                             retval = IP_FW_DENY;
3125                                             break;
3126                                         }
3127                                         if (cmd->arg1 != IP_FW_TARG)
3128                                             ((ipfw_insn_nat *)cmd)->nat = t;
3129                                 }
3130                                 retval = ipfw_nat_ptr(args, t, m);
3131                                 break;
3132
3133                         case O_REASS: {
3134                                 int ip_off;
3135
3136                                 l = 0;  /* in any case exit inner loop */
3137                                 if (is_ipv6) /* IPv6 is not supported yet */
3138                                         break;
3139                                 IPFW_INC_RULE_COUNTER(f, pktlen);
3140                                 ip_off = ntohs(ip->ip_off);
3141
3142                                 /* if not fragmented, go to next rule */
3143                                 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
3144                                     break;
3145
3146                                 args->m = m = ip_reass(m);
3147
3148                                 /*
3149                                  * do IP header checksum fixup.
3150                                  */
3151                                 if (m == NULL) { /* fragment got swallowed */
3152                                     retval = IP_FW_DENY;
3153                                 } else { /* good, packet complete */
3154                                     int hlen;
3155
3156                                     ip = mtod(m, struct ip *);
3157                                     hlen = ip->ip_hl << 2;
3158                                     ip->ip_sum = 0;
3159                                     if (hlen == sizeof(struct ip))
3160                                         ip->ip_sum = in_cksum_hdr(ip);
3161                                     else
3162                                         ip->ip_sum = in_cksum(m, hlen);
3163                                     retval = IP_FW_REASS;
3164                                     args->rule.info = 0;
3165                                     set_match(args, f_pos, chain);
3166                                 }
3167                                 done = 1;       /* exit outer loop */
3168                                 break;
3169                         }
3170                         case O_EXTERNAL_ACTION:
3171                                 l = 0; /* in any case exit inner loop */
3172                                 retval = ipfw_run_eaction(chain, args,
3173                                     cmd, &done);
3174                                 /*
3175                                  * If both @retval and @done are zero,
3176                                  * consider this as rule matching and
3177                                  * update counters.
3178                                  */
3179                                 if (retval == 0 && done == 0) {
3180                                         IPFW_INC_RULE_COUNTER(f, pktlen);
3181                                         /*
3182                                          * Reset the result of the last
3183                                          * dynamic state lookup.
3184                                          * External action can change
3185                                          * @args content, and it may be
3186                                          * used for new state lookup later.
3187                                          */
3188                                         DYN_INFO_INIT(&dyn_info);
3189                                 }
3190                                 break;
3191
3192                         default:
3193                                 panic("-- unknown opcode %d\n", cmd->opcode);
3194                         } /* end of switch() on opcodes */
3195                         /*
3196                          * if we get here with l=0, then match is irrelevant.
3197                          */
3198
3199                         if (cmd->len & F_NOT)
3200                                 match = !match;
3201
3202                         if (match) {
3203                                 if (cmd->len & F_OR)
3204                                         skip_or = 1;
3205                         } else {
3206                                 if (!(cmd->len & F_OR)) /* not an OR block, */
3207                                         break;          /* try next rule    */
3208                         }
3209
3210                 }       /* end of inner loop, scan opcodes */
3211 #undef PULLUP_LEN
3212 #undef PULLUP_LEN_LOCKED
3213
3214                 if (done)
3215                         break;
3216
3217 /* next_rule:; */       /* try next rule                */
3218
3219         }               /* end of outer for, scan rules */
3220
3221         if (done) {
3222                 struct ip_fw *rule = chain->map[f_pos];
3223                 /* Update statistics */
3224                 IPFW_INC_RULE_COUNTER(rule, pktlen);
3225         } else {
3226                 retval = IP_FW_DENY;
3227                 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3228         }
3229         IPFW_PF_RUNLOCK(chain);
3230 #ifdef __FreeBSD__
3231         if (ucred_cache != NULL)
3232                 crfree(ucred_cache);
3233 #endif
3234         return (retval);
3235
3236 pullup_failed:
3237         if (V_fw_verbose)
3238                 printf("ipfw: pullup failed\n");
3239         return (IP_FW_DENY);
3240 }
3241
3242 /*
3243  * Set maximum number of tables that can be used in given VNET ipfw instance.
3244  */
3245 #ifdef SYSCTL_NODE
3246 static int
3247 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
3248 {
3249         int error;
3250         unsigned int ntables;
3251
3252         ntables = V_fw_tables_max;
3253
3254         error = sysctl_handle_int(oidp, &ntables, 0, req);
3255         /* Read operation or some error */
3256         if ((error != 0) || (req->newptr == NULL))
3257                 return (error);
3258
3259         return (ipfw_resize_tables(&V_layer3_chain, ntables));
3260 }
3261
3262 /*
3263  * Switches table namespace between global and per-set.
3264  */
3265 static int
3266 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
3267 {
3268         int error;
3269         unsigned int sets;
3270
3271         sets = V_fw_tables_sets;
3272
3273         error = sysctl_handle_int(oidp, &sets, 0, req);
3274         /* Read operation or some error */
3275         if ((error != 0) || (req->newptr == NULL))
3276                 return (error);
3277
3278         return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
3279 }
3280 #endif
3281
3282 /*
3283  * Module and VNET glue
3284  */
3285
3286 /*
3287  * Stuff that must be initialised only on boot or module load
3288  */
3289 static int
3290 ipfw_init(void)
3291 {
3292         int error = 0;
3293
3294         /*
3295          * Only print out this stuff the first time around,
3296          * when called from the sysinit code.
3297          */
3298         printf("ipfw2 "
3299 #ifdef INET6
3300                 "(+ipv6) "
3301 #endif
3302                 "initialized, divert %s, nat %s, "
3303                 "default to %s, logging ",
3304 #ifdef IPDIVERT
3305                 "enabled",
3306 #else
3307                 "loadable",
3308 #endif
3309 #ifdef IPFIREWALL_NAT
3310                 "enabled",
3311 #else
3312                 "loadable",
3313 #endif
3314                 default_to_accept ? "accept" : "deny");
3315
3316         /*
3317          * Note: V_xxx variables can be accessed here but the vnet specific
3318          * initializer may not have been called yet for the VIMAGE case.
3319          * Tuneables will have been processed. We will print out values for
3320          * the default vnet.
3321          * XXX This should all be rationalized AFTER 8.0
3322          */
3323         if (V_fw_verbose == 0)
3324                 printf("disabled\n");
3325         else if (V_verbose_limit == 0)
3326                 printf("unlimited\n");
3327         else
3328                 printf("limited to %d packets/entry by default\n",
3329                     V_verbose_limit);
3330
3331         /* Check user-supplied table count for validness */
3332         if (default_fw_tables > IPFW_TABLES_MAX)
3333           default_fw_tables = IPFW_TABLES_MAX;
3334
3335         ipfw_init_sopt_handler();
3336         ipfw_init_obj_rewriter();
3337         ipfw_iface_init();
3338         return (error);
3339 }
3340
3341 /*
3342  * Called for the removal of the last instance only on module unload.
3343  */
3344 static void
3345 ipfw_destroy(void)
3346 {
3347
3348         ipfw_iface_destroy();
3349         ipfw_destroy_sopt_handler();
3350         ipfw_destroy_obj_rewriter();
3351         printf("IP firewall unloaded\n");
3352 }
3353
3354 /*
3355  * Stuff that must be initialized for every instance
3356  * (including the first of course).
3357  */
3358 static int
3359 vnet_ipfw_init(const void *unused)
3360 {
3361         int error, first;
3362         struct ip_fw *rule = NULL;
3363         struct ip_fw_chain *chain;
3364
3365         chain = &V_layer3_chain;
3366
3367         first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3368
3369         /* First set up some values that are compile time options */
3370         V_autoinc_step = 100;   /* bounded to 1..1000 in add_rule() */
3371         V_fw_deny_unknown_exthdrs = 1;
3372 #ifdef IPFIREWALL_VERBOSE
3373         V_fw_verbose = 1;
3374 #endif
3375 #ifdef IPFIREWALL_VERBOSE_LIMIT
3376         V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
3377 #endif
3378 #ifdef IPFIREWALL_NAT
3379         LIST_INIT(&chain->nat);
3380 #endif
3381
3382         /* Init shared services hash table */
3383         ipfw_init_srv(chain);
3384
3385         ipfw_init_counters();
3386         /* Set initial number of tables */
3387         V_fw_tables_max = default_fw_tables;
3388         error = ipfw_init_tables(chain, first);
3389         if (error) {
3390                 printf("ipfw2: setting up tables failed\n");
3391                 free(chain->map, M_IPFW);
3392                 free(rule, M_IPFW);
3393                 return (ENOSPC);
3394         }
3395
3396         IPFW_LOCK_INIT(chain);
3397
3398         /* fill and insert the default rule */
3399         rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
3400         rule->flags |= IPFW_RULE_NOOPT;
3401         rule->cmd_len = 1;
3402         rule->cmd[0].len = 1;
3403         rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
3404         chain->default_rule = rule;
3405         ipfw_add_protected_rule(chain, rule, 0);
3406
3407         ipfw_dyn_init(chain);
3408         ipfw_eaction_init(chain, first);
3409 #ifdef LINEAR_SKIPTO
3410         ipfw_init_skipto_cache(chain);
3411 #endif
3412         ipfw_bpf_init(first);
3413
3414         /* First set up some values that are compile time options */
3415         V_ipfw_vnet_ready = 1;          /* Open for business */
3416
3417         /*
3418          * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
3419          * Even if the latter two fail we still keep the module alive
3420          * because the sockopt and layer2 paths are still useful.
3421          * ipfw[6]_hook return 0 on success, ENOENT on failure,
3422          * so we can ignore the exact return value and just set a flag.
3423          *
3424          * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
3425          * changes in the underlying (per-vnet) variables trigger
3426          * immediate hook()/unhook() calls.
3427          * In layer2 we have the same behaviour, except that V_ether_ipfw
3428          * is checked on each packet because there are no pfil hooks.
3429          */
3430         V_ip_fw_ctl_ptr = ipfw_ctl3;
3431         error = ipfw_attach_hooks();
3432         return (error);
3433 }
3434
3435 /*
3436  * Called for the removal of each instance.
3437  */
3438 static int
3439 vnet_ipfw_uninit(const void *unused)
3440 {
3441         struct ip_fw *reap;
3442         struct ip_fw_chain *chain = &V_layer3_chain;
3443         int i, last;
3444
3445         V_ipfw_vnet_ready = 0; /* tell new callers to go away */
3446         /*
3447          * disconnect from ipv4, ipv6, layer2 and sockopt.
3448          * Then grab, release and grab again the WLOCK so we make
3449          * sure the update is propagated and nobody will be in.
3450          */
3451         ipfw_detach_hooks();
3452         V_ip_fw_ctl_ptr = NULL;
3453
3454         last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3455
3456         IPFW_UH_WLOCK(chain);
3457         IPFW_UH_WUNLOCK(chain);
3458
3459         ipfw_dyn_uninit(0);     /* run the callout_drain */
3460
3461         IPFW_UH_WLOCK(chain);
3462
3463         reap = NULL;
3464         IPFW_WLOCK(chain);
3465         for (i = 0; i < chain->n_rules; i++)
3466                 ipfw_reap_add(chain, &reap, chain->map[i]);
3467         free(chain->map, M_IPFW);
3468 #ifdef LINEAR_SKIPTO
3469         ipfw_destroy_skipto_cache(chain);
3470 #endif
3471         IPFW_WUNLOCK(chain);
3472         IPFW_UH_WUNLOCK(chain);
3473         ipfw_destroy_tables(chain, last);
3474         ipfw_eaction_uninit(chain, last);
3475         if (reap != NULL)
3476                 ipfw_reap_rules(reap);
3477         vnet_ipfw_iface_destroy(chain);
3478         ipfw_destroy_srv(chain);
3479         IPFW_LOCK_DESTROY(chain);
3480         ipfw_dyn_uninit(1);     /* free the remaining parts */
3481         ipfw_destroy_counters();
3482         ipfw_bpf_uninit(last);
3483         return (0);
3484 }
3485
3486 /*
3487  * Module event handler.
3488  * In general we have the choice of handling most of these events by the
3489  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
3490  * use the SYSINIT handlers as they are more capable of expressing the
3491  * flow of control during module and vnet operations, so this is just
3492  * a skeleton. Note there is no SYSINIT equivalent of the module
3493  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
3494  */
3495 static int
3496 ipfw_modevent(module_t mod, int type, void *unused)
3497 {
3498         int err = 0;
3499
3500         switch (type) {
3501         case MOD_LOAD:
3502                 /* Called once at module load or
3503                  * system boot if compiled in. */
3504                 break;
3505         case MOD_QUIESCE:
3506                 /* Called before unload. May veto unloading. */
3507                 break;
3508         case MOD_UNLOAD:
3509                 /* Called during unload. */
3510                 break;
3511         case MOD_SHUTDOWN:
3512                 /* Called during system shutdown. */
3513                 break;
3514         default:
3515                 err = EOPNOTSUPP;
3516                 break;
3517         }
3518         return err;
3519 }
3520
3521 static moduledata_t ipfwmod = {
3522         "ipfw",
3523         ipfw_modevent,
3524         0
3525 };
3526
3527 /* Define startup order. */
3528 #define IPFW_SI_SUB_FIREWALL    SI_SUB_PROTO_FIREWALL
3529 #define IPFW_MODEVENT_ORDER     (SI_ORDER_ANY - 255) /* On boot slot in here. */
3530 #define IPFW_MODULE_ORDER       (IPFW_MODEVENT_ORDER + 1) /* A little later. */
3531 #define IPFW_VNET_ORDER         (IPFW_MODEVENT_ORDER + 2) /* Later still. */
3532
3533 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
3534 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
3535 MODULE_VERSION(ipfw, 3);
3536 /* should declare some dependencies here */
3537
3538 /*
3539  * Starting up. Done in order after ipfwmod() has been called.
3540  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
3541  */
3542 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3543             ipfw_init, NULL);
3544 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3545             vnet_ipfw_init, NULL);
3546
3547 /*
3548  * Closing up shop. These are done in REVERSE ORDER, but still
3549  * after ipfwmod() has been called. Not called on reboot.
3550  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
3551  * or when the module is unloaded.
3552  */
3553 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3554             ipfw_destroy, NULL);
3555 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3556             vnet_ipfw_uninit, NULL);
3557 /* end of file */