sys/netpfil/ipfw/ip_fw2.c

   1 /*-
   2  * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa
   3  *
   4  * Redistribution and use in source and binary forms, with or without
   5  * modification, are permitted provided that the following conditions
   6  * are met:
   7  * 1. Redistributions of source code must retain the above copyright
   8  *    notice, this list of conditions and the following disclaimer.
   9  * 2. Redistributions in binary form must reproduce the above copyright
  10  *    notice, this list of conditions and the following disclaimer in the
  11  *    documentation and/or other materials provided with the distribution.
  12  *
  13  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
  14  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  15  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  16  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
  17  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  18  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  19  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  20  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  21  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  22  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  23  * SUCH DAMAGE.
  24  */
  25
  26 #include <sys/cdefs.h>
  27 __FBSDID("$FreeBSD$");
  28
  29 /*
  30  * The FreeBSD IP packet firewall, main file
  31  */
  32
  33 #include "opt_ipfw.h"
  34 #include "opt_ipdivert.h"
  35 #include "opt_inet.h"
  36 #ifndef INET
  37 #error "IPFIREWALL requires INET"
  38 #endif /* INET */
  39 #include "opt_inet6.h"
  40
  41 #include <sys/param.h>
  42 #include <sys/systm.h>
  43 #include <sys/condvar.h>
  44 #include <sys/counter.h>
  45 #include <sys/eventhandler.h>
  46 #include <sys/malloc.h>
  47 #include <sys/mbuf.h>
  48 #include <sys/kernel.h>
  49 #include <sys/lock.h>
  50 #include <sys/jail.h>
  51 #include <sys/module.h>
  52 #include <sys/priv.h>
  53 #include <sys/proc.h>
  54 #include <sys/rwlock.h>
  55 #include <sys/rmlock.h>
  56 #include <sys/socket.h>
  57 #include <sys/socketvar.h>
  58 #include <sys/sysctl.h>
  59 #include <sys/syslog.h>
  60 #include <sys/ucred.h>
  61 #include <net/ethernet.h> /* for ETHERTYPE_IP */
  62 #include <net/if.h>
  63 #include <net/if_var.h>
  64 #include <net/route.h>
  65 #include <net/pfil.h>
  66 #include <net/vnet.h>
  67
  68 #include <netpfil/pf/pf_mtag.h>
  69
  70 #include <netinet/in.h>
  71 #include <netinet/in_var.h>
  72 #include <netinet/in_pcb.h>
  73 #include <netinet/ip.h>
  74 #include <netinet/ip_var.h>
  75 #include <netinet/ip_icmp.h>
  76 #include <netinet/ip_fw.h>
  77 #include <netinet/ip_carp.h>
  78 #include <netinet/pim.h>
  79 #include <netinet/tcp_var.h>
  80 #include <netinet/udp.h>
  81 #include <netinet/udp_var.h>
  82 #include <netinet/sctp.h>
  83 #include <netinet/sctp_crc32.h>
  84 #include <netinet/sctp_header.h>
  85
  86 #include <netinet/ip6.h>
  87 #include <netinet/icmp6.h>
  88 #include <netinet/in_fib.h>
  89 #ifdef INET6
  90 #include <netinet6/in6_fib.h>
  91 #include <netinet6/in6_pcb.h>
  92 #include <netinet6/scope6_var.h>
  93 #include <netinet6/ip6_var.h>
  94 #endif
  95
  96 #include <net/if_gre.h> /* for struct grehdr */
  97
  98 #include <netpfil/ipfw/ip_fw_private.h>
  99
 100 #include <machine/in_cksum.h>   /* XXX for in_cksum */
 101
 102 #ifdef MAC
 103 #include <security/mac/mac_framework.h>
 104 #endif
 105
 106 /*
 107  * static variables followed by global ones.
 108  * All ipfw global variables are here.
 109  */
 110
 111 static VNET_DEFINE(int, fw_deny_unknown_exthdrs);
 112 #define V_fw_deny_unknown_exthdrs       VNET(fw_deny_unknown_exthdrs)
 113
 114 static VNET_DEFINE(int, fw_permit_single_frag6) = 1;
 115 #define V_fw_permit_single_frag6        VNET(fw_permit_single_frag6)
 116
 117 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
 118 static int default_to_accept = 1;
 119 #else
 120 static int default_to_accept;
 121 #endif
 122
 123 VNET_DEFINE(int, autoinc_step);
 124 VNET_DEFINE(int, fw_one_pass) = 1;
 125
 126 VNET_DEFINE(unsigned int, fw_tables_max);
 127 VNET_DEFINE(unsigned int, fw_tables_sets) = 0;  /* Don't use set-aware tables */
 128 /* Use 128 tables by default */
 129 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT;
 130
 131 #ifndef LINEAR_SKIPTO
 132 static int jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
 133     int tablearg, int jump_backwards);
 134 #define JUMP(ch, f, num, targ, back)    jump_fast(ch, f, num, targ, back)
 135 #else
 136 static int jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
 137     int tablearg, int jump_backwards);
 138 #define JUMP(ch, f, num, targ, back)    jump_linear(ch, f, num, targ, back)
 139 #endif
 140
 141 /*
 142  * Each rule belongs to one of 32 different sets (0..31).
 143  * The variable set_disable contains one bit per set.
 144  * If the bit is set, all rules in the corresponding set
 145  * are disabled. Set RESVD_SET(31) is reserved for the default rule
 146  * and rules that are not deleted by the flush command,
 147  * and CANNOT be disabled.
 148  * Rules in set RESVD_SET can only be deleted individually.
 149  */
 150 VNET_DEFINE(u_int32_t, set_disable);
 151 #define V_set_disable                   VNET(set_disable)
 152
 153 VNET_DEFINE(int, fw_verbose);
 154 /* counter for ipfw_log(NULL...) */
 155 VNET_DEFINE(u_int64_t, norule_counter);
 156 VNET_DEFINE(int, verbose_limit);
 157
 158 /* layer3_chain contains the list of rules for layer 3 */
 159 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
 160
 161 /* ipfw_vnet_ready controls when we are open for business */
 162 VNET_DEFINE(int, ipfw_vnet_ready) = 0;
 163
 164 VNET_DEFINE(int, ipfw_nat_ready) = 0;
 165
 166 ipfw_nat_t *ipfw_nat_ptr = NULL;
 167 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
 168 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
 169 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
 170 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
 171 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
 172
 173 #ifdef SYSCTL_NODE
 174 uint32_t dummy_def = IPFW_DEFAULT_RULE;
 175 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS);
 176 static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS);
 177
 178 SYSBEGIN(f3)
 179
 180 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 181 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
 182     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
 183     "Only do a single pass through ipfw when using dummynet(4)");
 184 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
 185     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
 186     "Rule number auto-increment step");
 187 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose,
 188     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
 189     "Log matches to ipfw rules");
 190 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
 191     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
 192     "Set upper limit of matches of ipfw rules logged");
 193 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
 194     &dummy_def, 0,
 195     "The default/max possible rule number.");
 196 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
 197     CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU",
 198     "Maximum number of concurrently used tables");
 199 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets,
 200     CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW,
 201     0, 0, sysctl_ipfw_tables_sets, "IU",
 202     "Use per-set namespace for tables");
 203 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
 204     &default_to_accept, 0,
 205     "Make the default rule accept all packets.");
 206 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables);
 207 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count,
 208     CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
 209     "Number of static rules");
 210
 211 #ifdef INET6
 212 SYSCTL_DECL(_net_inet6_ip6);
 213 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 214 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
 215     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
 216     &VNET_NAME(fw_deny_unknown_exthdrs), 0,
 217     "Deny packets with unknown IPv6 Extension Headers");
 218 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
 219     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
 220     &VNET_NAME(fw_permit_single_frag6), 0,
 221     "Permit single packet IPv6 fragments");
 222 #endif /* INET6 */
 223
 224 SYSEND
 225
 226 #endif /* SYSCTL_NODE */
 227
 228
 229 /*
 230  * Some macros used in the various matching options.
 231  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
 232  * Other macros just cast void * into the appropriate type
 233  */
 234 #define L3HDR(T, ip)    ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
 235 #define TCP(p)          ((struct tcphdr *)(p))
 236 #define SCTP(p)         ((struct sctphdr *)(p))
 237 #define UDP(p)          ((struct udphdr *)(p))
 238 #define ICMP(p)         ((struct icmphdr *)(p))
 239 #define ICMP6(p)        ((struct icmp6_hdr *)(p))
 240
 241 static __inline int
 242 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
 243 {
 244         int type = icmp->icmp_type;
 245
 246         return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
 247 }
 248
 249 #define TT      ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
 250     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
 251
 252 static int
 253 is_icmp_query(struct icmphdr *icmp)
 254 {
 255         int type = icmp->icmp_type;
 256
 257         return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
 258 }
 259 #undef TT
 260
 261 /*
 262  * The following checks use two arrays of 8 or 16 bits to store the
 263  * bits that we want set or clear, respectively. They are in the
 264  * low and high half of cmd->arg1 or cmd->d[0].
 265  *
 266  * We scan options and store the bits we find set. We succeed if
 267  *
 268  *      (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
 269  *
 270  * The code is sometimes optimized not to store additional variables.
 271  */
 272
 273 static int
 274 flags_match(ipfw_insn *cmd, u_int8_t bits)
 275 {
 276         u_char want_clear;
 277         bits = ~bits;
 278
 279         if ( ((cmd->arg1 & 0xff) & bits) != 0)
 280                 return 0; /* some bits we want set were clear */
 281         want_clear = (cmd->arg1 >> 8) & 0xff;
 282         if ( (want_clear & bits) != want_clear)
 283                 return 0; /* some bits we want clear were set */
 284         return 1;
 285 }
 286
 287 static int
 288 ipopts_match(struct ip *ip, ipfw_insn *cmd)
 289 {
 290         int optlen, bits = 0;
 291         u_char *cp = (u_char *)(ip + 1);
 292         int x = (ip->ip_hl << 2) - sizeof (struct ip);
 293
 294         for (; x > 0; x -= optlen, cp += optlen) {
 295                 int opt = cp[IPOPT_OPTVAL];
 296
 297                 if (opt == IPOPT_EOL)
 298                         break;
 299                 if (opt == IPOPT_NOP)
 300                         optlen = 1;
 301                 else {
 302                         optlen = cp[IPOPT_OLEN];
 303                         if (optlen <= 0 || optlen > x)
 304                                 return 0; /* invalid or truncated */
 305                 }
 306                 switch (opt) {
 307
 308                 default:
 309                         break;
 310
 311                 case IPOPT_LSRR:
 312                         bits |= IP_FW_IPOPT_LSRR;
 313                         break;
 314
 315                 case IPOPT_SSRR:
 316                         bits |= IP_FW_IPOPT_SSRR;
 317                         break;
 318
 319                 case IPOPT_RR:
 320                         bits |= IP_FW_IPOPT_RR;
 321                         break;
 322
 323                 case IPOPT_TS:
 324                         bits |= IP_FW_IPOPT_TS;
 325                         break;
 326                 }
 327         }
 328         return (flags_match(cmd, bits));
 329 }
 330
 331 static int
 332 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
 333 {
 334         int optlen, bits = 0;
 335         u_char *cp = (u_char *)(tcp + 1);
 336         int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
 337
 338         for (; x > 0; x -= optlen, cp += optlen) {
 339                 int opt = cp[0];
 340                 if (opt == TCPOPT_EOL)
 341                         break;
 342                 if (opt == TCPOPT_NOP)
 343                         optlen = 1;
 344                 else {
 345                         optlen = cp[1];
 346                         if (optlen <= 0)
 347                                 break;
 348                 }
 349
 350                 switch (opt) {
 351
 352                 default:
 353                         break;
 354
 355                 case TCPOPT_MAXSEG:
 356                         bits |= IP_FW_TCPOPT_MSS;
 357                         break;
 358
 359                 case TCPOPT_WINDOW:
 360                         bits |= IP_FW_TCPOPT_WINDOW;
 361                         break;
 362
 363                 case TCPOPT_SACK_PERMITTED:
 364                 case TCPOPT_SACK:
 365                         bits |= IP_FW_TCPOPT_SACK;
 366                         break;
 367
 368                 case TCPOPT_TIMESTAMP:
 369                         bits |= IP_FW_TCPOPT_TS;
 370                         break;
 371
 372                 }
 373         }
 374         return (flags_match(cmd, bits));
 375 }
 376
 377 static int
 378 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain,
 379     uint32_t *tablearg)
 380 {
 381
 382         if (ifp == NULL)        /* no iface with this packet, match fails */
 383                 return (0);
 384
 385         /* Check by name or by IP address */
 386         if (cmd->name[0] != '\0') { /* match by name */
 387                 if (cmd->name[0] == '\1') /* use tablearg to match */
 388                         return ipfw_lookup_table(chain, cmd->p.kidx, 0,
 389                             &ifp->if_index, tablearg);
 390                 /* Check name */
 391                 if (cmd->p.glob) {
 392                         if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
 393                                 return(1);
 394                 } else {
 395                         if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
 396                                 return(1);
 397                 }
 398         } else {
 399 #if !defined(USERSPACE) && defined(__FreeBSD__) /* and OSX too ? */
 400                 struct ifaddr *ia;
 401
 402                 if_addr_rlock(ifp);
 403                 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
 404                         if (ia->ifa_addr->sa_family != AF_INET)
 405                                 continue;
 406                         if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
 407                             (ia->ifa_addr))->sin_addr.s_addr) {
 408                                 if_addr_runlock(ifp);
 409                                 return(1);      /* match */
 410                         }
 411                 }
 412                 if_addr_runlock(ifp);
 413 #endif /* __FreeBSD__ */
 414         }
 415         return(0);      /* no match, fail ... */
 416 }
 417
 418 /*
 419  * The verify_path function checks if a route to the src exists and
 420  * if it is reachable via ifp (when provided).
 421  *
 422  * The 'verrevpath' option checks that the interface that an IP packet
 423  * arrives on is the same interface that traffic destined for the
 424  * packet's source address would be routed out of.
 425  * The 'versrcreach' option just checks that the source address is
 426  * reachable via any route (except default) in the routing table.
 427  * These two are a measure to block forged packets. This is also
 428  * commonly known as "anti-spoofing" or Unicast Reverse Path
 429  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
 430  * is purposely reminiscent of the Cisco IOS command,
 431  *
 432  *   ip verify unicast reverse-path
 433  *   ip verify unicast source reachable-via any
 434  *
 435  * which implements the same functionality. But note that the syntax
 436  * is misleading, and the check may be performed on all IP packets
 437  * whether unicast, multicast, or broadcast.
 438  */
 439 static int
 440 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
 441 {
 442 #if defined(USERSPACE) || !defined(__FreeBSD__)
 443         return 0;
 444 #else
 445         struct nhop4_basic nh4;
 446
 447         if (fib4_lookup_nh_basic(fib, src, NHR_IFAIF, 0, &nh4) != 0)
 448                 return (0);
 449
 450         /*
 451          * If ifp is provided, check for equality with rtentry.
 452          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 453          * in order to pass packets injected back by if_simloop():
 454          * routing entry (via lo0) for our own address
 455          * may exist, so we need to handle routing assymetry.
 456          */
 457         if (ifp != NULL && ifp != nh4.nh_ifp)
 458                 return (0);
 459
 460         /* if no ifp provided, check if rtentry is not default route */
 461         if (ifp == NULL && (nh4.nh_flags & NHF_DEFAULT) != 0)
 462                 return (0);
 463
 464         /* or if this is a blackhole/reject route */
 465         if (ifp == NULL && (nh4.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
 466                 return (0);
 467
 468         /* found valid route */
 469         return 1;
 470 #endif /* __FreeBSD__ */
 471 }
 472
 473 /*
 474  * Generate an SCTP packet containing an ABORT chunk. The verification tag
 475  * is given by vtag. The T-bit is set in the ABORT chunk if and only if
 476  * reflected is not 0.
 477  */
 478
 479 static struct mbuf *
 480 ipfw_send_abort(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t vtag,
 481     int reflected)
 482 {
 483         struct mbuf *m;
 484         struct ip *ip;
 485 #ifdef INET6
 486         struct ip6_hdr *ip6;
 487 #endif
 488         struct sctphdr *sctp;
 489         struct sctp_chunkhdr *chunk;
 490         u_int16_t hlen, plen, tlen;
 491
 492         MGETHDR(m, M_NOWAIT, MT_DATA);
 493         if (m == NULL)
 494                 return (NULL);
 495
 496         M_SETFIB(m, id->fib);
 497 #ifdef MAC
 498         if (replyto != NULL)
 499                 mac_netinet_firewall_reply(replyto, m);
 500         else
 501                 mac_netinet_firewall_send(m);
 502 #else
 503         (void)replyto;          /* don't warn about unused arg */
 504 #endif
 505
 506         switch (id->addr_type) {
 507         case 4:
 508                 hlen = sizeof(struct ip);
 509                 break;
 510 #ifdef INET6
 511         case 6:
 512                 hlen = sizeof(struct ip6_hdr);
 513                 break;
 514 #endif
 515         default:
 516                 /* XXX: log me?!? */
 517                 FREE_PKT(m);
 518                 return (NULL);
 519         }
 520         plen = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr);
 521         tlen = hlen + plen;
 522         m->m_data += max_linkhdr;
 523         m->m_flags |= M_SKIP_FIREWALL;
 524         m->m_pkthdr.len = m->m_len = tlen;
 525         m->m_pkthdr.rcvif = NULL;
 526         bzero(m->m_data, tlen);
 527
 528         switch (id->addr_type) {
 529         case 4:
 530                 ip = mtod(m, struct ip *);
 531
 532                 ip->ip_v = 4;
 533                 ip->ip_hl = sizeof(struct ip) >> 2;
 534                 ip->ip_tos = IPTOS_LOWDELAY;
 535                 ip->ip_len = htons(tlen);
 536                 ip->ip_id = htons(0);
 537                 ip->ip_off = htons(0);
 538                 ip->ip_ttl = V_ip_defttl;
 539                 ip->ip_p = IPPROTO_SCTP;
 540                 ip->ip_sum = 0;
 541                 ip->ip_src.s_addr = htonl(id->dst_ip);
 542                 ip->ip_dst.s_addr = htonl(id->src_ip);
 543
 544                 sctp = (struct sctphdr *)(ip + 1);
 545                 break;
 546 #ifdef INET6
 547         case 6:
 548                 ip6 = mtod(m, struct ip6_hdr *);
 549
 550                 ip6->ip6_vfc = IPV6_VERSION;
 551                 ip6->ip6_plen = htons(plen);
 552                 ip6->ip6_nxt = IPPROTO_SCTP;
 553                 ip6->ip6_hlim = IPV6_DEFHLIM;
 554                 ip6->ip6_src = id->dst_ip6;
 555                 ip6->ip6_dst = id->src_ip6;
 556
 557                 sctp = (struct sctphdr *)(ip6 + 1);
 558                 break;
 559 #endif
 560         }
 561
 562         sctp->src_port = htons(id->dst_port);
 563         sctp->dest_port = htons(id->src_port);
 564         sctp->v_tag = htonl(vtag);
 565         sctp->checksum = htonl(0);
 566
 567         chunk = (struct sctp_chunkhdr *)(sctp + 1);
 568         chunk->chunk_type = SCTP_ABORT_ASSOCIATION;
 569         chunk->chunk_flags = 0;
 570         if (reflected != 0) {
 571                 chunk->chunk_flags |= SCTP_HAD_NO_TCB;
 572         }
 573         chunk->chunk_length = htons(sizeof(struct sctp_chunkhdr));
 574
 575         sctp->checksum = sctp_calculate_cksum(m, hlen);
 576
 577         return (m);
 578 }
 579
 580 /*
 581  * Generate a TCP packet, containing either a RST or a keepalive.
 582  * When flags & TH_RST, we are sending a RST packet, because of a
 583  * "reset" action matched the packet.
 584  * Otherwise we are sending a keepalive, and flags & TH_
 585  * The 'replyto' mbuf is the mbuf being replied to, if any, and is required
 586  * so that MAC can label the reply appropriately.
 587  */
 588 struct mbuf *
 589 ipfw_send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq,
 590     u_int32_t ack, int flags)
 591 {
 592         struct mbuf *m = NULL;          /* stupid compiler */
 593         struct ip *h = NULL;            /* stupid compiler */
 594 #ifdef INET6
 595         struct ip6_hdr *h6 = NULL;
 596 #endif
 597         struct tcphdr *th = NULL;
 598         int len, dir;
 599
 600         MGETHDR(m, M_NOWAIT, MT_DATA);
 601         if (m == NULL)
 602                 return (NULL);
 603
 604         M_SETFIB(m, id->fib);
 605 #ifdef MAC
 606         if (replyto != NULL)
 607                 mac_netinet_firewall_reply(replyto, m);
 608         else
 609                 mac_netinet_firewall_send(m);
 610 #else
 611         (void)replyto;          /* don't warn about unused arg */
 612 #endif
 613
 614         switch (id->addr_type) {
 615         case 4:
 616                 len = sizeof(struct ip) + sizeof(struct tcphdr);
 617                 break;
 618 #ifdef INET6
 619         case 6:
 620                 len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
 621                 break;
 622 #endif
 623         default:
 624                 /* XXX: log me?!? */
 625                 FREE_PKT(m);
 626                 return (NULL);
 627         }
 628         dir = ((flags & (TH_SYN | TH_RST)) == TH_SYN);
 629
 630         m->m_data += max_linkhdr;
 631         m->m_flags |= M_SKIP_FIREWALL;
 632         m->m_pkthdr.len = m->m_len = len;
 633         m->m_pkthdr.rcvif = NULL;
 634         bzero(m->m_data, len);
 635
 636         switch (id->addr_type) {
 637         case 4:
 638                 h = mtod(m, struct ip *);
 639
 640                 /* prepare for checksum */
 641                 h->ip_p = IPPROTO_TCP;
 642                 h->ip_len = htons(sizeof(struct tcphdr));
 643                 if (dir) {
 644                         h->ip_src.s_addr = htonl(id->src_ip);
 645                         h->ip_dst.s_addr = htonl(id->dst_ip);
 646                 } else {
 647                         h->ip_src.s_addr = htonl(id->dst_ip);
 648                         h->ip_dst.s_addr = htonl(id->src_ip);
 649                 }
 650
 651                 th = (struct tcphdr *)(h + 1);
 652                 break;
 653 #ifdef INET6
 654         case 6:
 655                 h6 = mtod(m, struct ip6_hdr *);
 656
 657                 /* prepare for checksum */
 658                 h6->ip6_nxt = IPPROTO_TCP;
 659                 h6->ip6_plen = htons(sizeof(struct tcphdr));
 660                 if (dir) {
 661                         h6->ip6_src = id->src_ip6;
 662                         h6->ip6_dst = id->dst_ip6;
 663                 } else {
 664                         h6->ip6_src = id->dst_ip6;
 665                         h6->ip6_dst = id->src_ip6;
 666                 }
 667
 668                 th = (struct tcphdr *)(h6 + 1);
 669                 break;
 670 #endif
 671         }
 672
 673         if (dir) {
 674                 th->th_sport = htons(id->src_port);
 675                 th->th_dport = htons(id->dst_port);
 676         } else {
 677                 th->th_sport = htons(id->dst_port);
 678                 th->th_dport = htons(id->src_port);
 679         }
 680         th->th_off = sizeof(struct tcphdr) >> 2;
 681
 682         if (flags & TH_RST) {
 683                 if (flags & TH_ACK) {
 684                         th->th_seq = htonl(ack);
 685                         th->th_flags = TH_RST;
 686                 } else {
 687                         if (flags & TH_SYN)
 688                                 seq++;
 689                         th->th_ack = htonl(seq);
 690                         th->th_flags = TH_RST | TH_ACK;
 691                 }
 692         } else {
 693                 /*
 694                  * Keepalive - use caller provided sequence numbers
 695                  */
 696                 th->th_seq = htonl(seq);
 697                 th->th_ack = htonl(ack);
 698                 th->th_flags = TH_ACK;
 699         }
 700
 701         switch (id->addr_type) {
 702         case 4:
 703                 th->th_sum = in_cksum(m, len);
 704
 705                 /* finish the ip header */
 706                 h->ip_v = 4;
 707                 h->ip_hl = sizeof(*h) >> 2;
 708                 h->ip_tos = IPTOS_LOWDELAY;
 709                 h->ip_off = htons(0);
 710                 h->ip_len = htons(len);
 711                 h->ip_ttl = V_ip_defttl;
 712                 h->ip_sum = 0;
 713                 break;
 714 #ifdef INET6
 715         case 6:
 716                 th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(*h6),
 717                     sizeof(struct tcphdr));
 718
 719                 /* finish the ip6 header */
 720                 h6->ip6_vfc |= IPV6_VERSION;
 721                 h6->ip6_hlim = IPV6_DEFHLIM;
 722                 break;
 723 #endif
 724         }
 725
 726         return (m);
 727 }
 728
 729 #ifdef INET6
 730 /*
 731  * ipv6 specific rules here...
 732  */
 733 static __inline int
 734 icmp6type_match (int type, ipfw_insn_u32 *cmd)
 735 {
 736         return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
 737 }
 738
 739 static int
 740 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
 741 {
 742         int i;
 743         for (i=0; i <= cmd->o.arg1; ++i )
 744                 if (curr_flow == cmd->d[i] )
 745                         return 1;
 746         return 0;
 747 }
 748
 749 /* support for IP6_*_ME opcodes */
 750 static const struct in6_addr lla_mask = {{{
 751         0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff,
 752         0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
 753 }}};
 754
 755 static int
 756 ipfw_localip6(struct in6_addr *in6)
 757 {
 758         struct rm_priotracker in6_ifa_tracker;
 759         struct in6_ifaddr *ia;
 760
 761         if (IN6_IS_ADDR_MULTICAST(in6))
 762                 return (0);
 763
 764         if (!IN6_IS_ADDR_LINKLOCAL(in6))
 765                 return (in6_localip(in6));
 766
 767         IN6_IFADDR_RLOCK(&in6_ifa_tracker);
 768         TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
 769                 if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr))
 770                         continue;
 771                 if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
 772                     in6, &lla_mask)) {
 773                         IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
 774                         return (1);
 775                 }
 776         }
 777         IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
 778         return (0);
 779 }
 780
 781 static int
 782 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
 783 {
 784         struct nhop6_basic nh6;
 785
 786         if (IN6_IS_SCOPE_LINKLOCAL(src))
 787                 return (1);
 788
 789         if (fib6_lookup_nh_basic(fib, src, 0, NHR_IFAIF, 0, &nh6) != 0)
 790                 return (0);
 791
 792         /* If ifp is provided, check for equality with route table. */
 793         if (ifp != NULL && ifp != nh6.nh_ifp)
 794                 return (0);
 795
 796         /* if no ifp provided, check if rtentry is not default route */
 797         if (ifp == NULL && (nh6.nh_flags & NHF_DEFAULT) != 0)
 798                 return (0);
 799
 800         /* or if this is a blackhole/reject route */
 801         if (ifp == NULL && (nh6.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
 802                 return (0);
 803
 804         /* found valid route */
 805         return 1;
 806 }
 807
 808 static int
 809 is_icmp6_query(int icmp6_type)
 810 {
 811         if ((icmp6_type <= ICMP6_MAXTYPE) &&
 812             (icmp6_type == ICMP6_ECHO_REQUEST ||
 813             icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
 814             icmp6_type == ICMP6_WRUREQUEST ||
 815             icmp6_type == ICMP6_FQDN_QUERY ||
 816             icmp6_type == ICMP6_NI_QUERY))
 817                 return (1);
 818
 819         return (0);
 820 }
 821
 822 static int
 823 map_icmp_unreach(int code)
 824 {
 825
 826         /* RFC 7915 p4.2 */
 827         switch (code) {
 828         case ICMP_UNREACH_NET:
 829         case ICMP_UNREACH_HOST:
 830         case ICMP_UNREACH_SRCFAIL:
 831         case ICMP_UNREACH_NET_UNKNOWN:
 832         case ICMP_UNREACH_HOST_UNKNOWN:
 833         case ICMP_UNREACH_TOSNET:
 834         case ICMP_UNREACH_TOSHOST:
 835                 return (ICMP6_DST_UNREACH_NOROUTE);
 836         case ICMP_UNREACH_PORT:
 837                 return (ICMP6_DST_UNREACH_NOPORT);
 838         default:
 839                 /*
 840                  * Map the rest of codes into admit prohibited.
 841                  * XXX: unreach proto should be mapped into ICMPv6
 842                  * parameter problem, but we use only unreach type.
 843                  */
 844                 return (ICMP6_DST_UNREACH_ADMIN);
 845         }
 846 }
 847
 848 static void
 849 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
 850 {
 851         struct mbuf *m;
 852
 853         m = args->m;
 854         if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
 855                 struct tcphdr *tcp;
 856                 tcp = (struct tcphdr *)((char *)ip6 + hlen);
 857
 858                 if ((tcp->th_flags & TH_RST) == 0) {
 859                         struct mbuf *m0;
 860                         m0 = ipfw_send_pkt(args->m, &(args->f_id),
 861                             ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 862                             tcp->th_flags | TH_RST);
 863                         if (m0 != NULL)
 864                                 ip6_output(m0, NULL, NULL, 0, NULL, NULL,
 865                                     NULL);
 866                 }
 867                 FREE_PKT(m);
 868         } else if (code == ICMP6_UNREACH_ABORT &&
 869             args->f_id.proto == IPPROTO_SCTP) {
 870                 struct mbuf *m0;
 871                 struct sctphdr *sctp;
 872                 u_int32_t v_tag;
 873                 int reflected;
 874
 875                 sctp = (struct sctphdr *)((char *)ip6 + hlen);
 876                 reflected = 1;
 877                 v_tag = ntohl(sctp->v_tag);
 878                 /* Investigate the first chunk header if available */
 879                 if (m->m_len >= hlen + sizeof(struct sctphdr) +
 880                     sizeof(struct sctp_chunkhdr)) {
 881                         struct sctp_chunkhdr *chunk;
 882
 883                         chunk = (struct sctp_chunkhdr *)(sctp + 1);
 884                         switch (chunk->chunk_type) {
 885                         case SCTP_INITIATION:
 886                                 /*
 887                                  * Packets containing an INIT chunk MUST have
 888                                  * a zero v-tag.
 889                                  */
 890                                 if (v_tag != 0) {
 891                                         v_tag = 0;
 892                                         break;
 893                                 }
 894                                 /* INIT chunk MUST NOT be bundled */
 895                                 if (m->m_pkthdr.len >
 896                                     hlen + sizeof(struct sctphdr) +
 897                                     ntohs(chunk->chunk_length) + 3) {
 898                                         break;
 899                                 }
 900                                 /* Use the initiate tag if available */
 901                                 if ((m->m_len >= hlen + sizeof(struct sctphdr) +
 902                                     sizeof(struct sctp_chunkhdr) +
 903                                     offsetof(struct sctp_init, a_rwnd))) {
 904                                         struct sctp_init *init;
 905
 906                                         init = (struct sctp_init *)(chunk + 1);
 907                                         v_tag = ntohl(init->initiate_tag);
 908                                         reflected = 0;
 909                                 }
 910                                 break;
 911                         case SCTP_ABORT_ASSOCIATION:
 912                                 /*
 913                                  * If the packet contains an ABORT chunk, don't
 914                                  * reply.
 915                                  * XXX: We should search through all chunks,
 916                                  *      but don't do to avoid attacks.
 917                                  */
 918                                 v_tag = 0;
 919                                 break;
 920                         }
 921                 }
 922                 if (v_tag == 0) {
 923                         m0 = NULL;
 924                 } else {
 925                         m0 = ipfw_send_abort(args->m, &(args->f_id), v_tag,
 926                             reflected);
 927                 }
 928                 if (m0 != NULL)
 929                         ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
 930                 FREE_PKT(m);
 931         } else if (code != ICMP6_UNREACH_RST && code != ICMP6_UNREACH_ABORT) {
 932                 /* Send an ICMPv6 unreach. */
 933 #if 0
 934                 /*
 935                  * Unlike above, the mbufs need to line up with the ip6 hdr,
 936                  * as the contents are read. We need to m_adj() the
 937                  * needed amount.
 938                  * The mbuf will however be thrown away so we can adjust it.
 939                  * Remember we did an m_pullup on it already so we
 940                  * can make some assumptions about contiguousness.
 941                  */
 942                 if (args->L3offset)
 943                         m_adj(m, args->L3offset);
 944 #endif
 945                 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
 946         } else
 947                 FREE_PKT(m);
 948
 949         args->m = NULL;
 950 }
 951
 952 #endif /* INET6 */
 953
 954
 955 /*
 956  * sends a reject message, consuming the mbuf passed as an argument.
 957  */
 958 static void
 959 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
 960 {
 961
 962 #if 0
 963         /* XXX When ip is not guaranteed to be at mtod() we will
 964          * need to account for this */
 965          * The mbuf will however be thrown away so we can adjust it.
 966          * Remember we did an m_pullup on it already so we
 967          * can make some assumptions about contiguousness.
 968          */
 969         if (args->L3offset)
 970                 m_adj(m, args->L3offset);
 971 #endif
 972         if (code != ICMP_REJECT_RST && code != ICMP_REJECT_ABORT) {
 973                 /* Send an ICMP unreach */
 974                 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
 975         } else if (code == ICMP_REJECT_RST && args->f_id.proto == IPPROTO_TCP) {
 976                 struct tcphdr *const tcp =
 977                     L3HDR(struct tcphdr, mtod(args->m, struct ip *));
 978                 if ( (tcp->th_flags & TH_RST) == 0) {
 979                         struct mbuf *m;
 980                         m = ipfw_send_pkt(args->m, &(args->f_id),
 981                                 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 982                                 tcp->th_flags | TH_RST);
 983                         if (m != NULL)
 984                                 ip_output(m, NULL, NULL, 0, NULL, NULL);
 985                 }
 986                 FREE_PKT(args->m);
 987         } else if (code == ICMP_REJECT_ABORT &&
 988             args->f_id.proto == IPPROTO_SCTP) {
 989                 struct mbuf *m;
 990                 struct sctphdr *sctp;
 991                 struct sctp_chunkhdr *chunk;
 992                 struct sctp_init *init;
 993                 u_int32_t v_tag;
 994                 int reflected;
 995
 996                 sctp = L3HDR(struct sctphdr, mtod(args->m, struct ip *));
 997                 reflected = 1;
 998                 v_tag = ntohl(sctp->v_tag);
 999                 if (iplen >= (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1000                     sizeof(struct sctp_chunkhdr)) {
1001                         /* Look at the first chunk header if available */
1002                         chunk = (struct sctp_chunkhdr *)(sctp + 1);
1003                         switch (chunk->chunk_type) {
1004                         case SCTP_INITIATION:
1005                                 /*
1006                                  * Packets containing an INIT chunk MUST have
1007                                  * a zero v-tag.
1008                                  */
1009                                 if (v_tag != 0) {
1010                                         v_tag = 0;
1011                                         break;
1012                                 }
1013                                 /* INIT chunk MUST NOT be bundled */
1014                                 if (iplen >
1015                                     (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1016                                     ntohs(chunk->chunk_length) + 3) {
1017                                         break;
1018                                 }
1019                                 /* Use the initiate tag if available */
1020                                 if ((iplen >= (ip->ip_hl << 2) +
1021                                     sizeof(struct sctphdr) +
1022                                     sizeof(struct sctp_chunkhdr) +
1023                                     offsetof(struct sctp_init, a_rwnd))) {
1024                                         init = (struct sctp_init *)(chunk + 1);
1025                                         v_tag = ntohl(init->initiate_tag);
1026                                         reflected = 0;
1027                                 }
1028                                 break;
1029                         case SCTP_ABORT_ASSOCIATION:
1030                                 /*
1031                                  * If the packet contains an ABORT chunk, don't
1032                                  * reply.
1033                                  * XXX: We should search through all chunks,
1034                                  * but don't do to avoid attacks.
1035                                  */
1036                                 v_tag = 0;
1037                                 break;
1038                         }
1039                 }
1040                 if (v_tag == 0) {
1041                         m = NULL;
1042                 } else {
1043                         m = ipfw_send_abort(args->m, &(args->f_id), v_tag,
1044                             reflected);
1045                 }
1046                 if (m != NULL)
1047                         ip_output(m, NULL, NULL, 0, NULL, NULL);
1048                 FREE_PKT(args->m);
1049         } else
1050                 FREE_PKT(args->m);
1051         args->m = NULL;
1052 }
1053
1054 /*
1055  * Support for uid/gid/jail lookup. These tests are expensive
1056  * (because we may need to look into the list of active sockets)
1057  * so we cache the results. ugid_lookupp is 0 if we have not
1058  * yet done a lookup, 1 if we succeeded, and -1 if we tried
1059  * and failed. The function always returns the match value.
1060  * We could actually spare the variable and use *uc, setting
1061  * it to '(void *)check_uidgid if we have no info, NULL if
1062  * we tried and failed, or any other value if successful.
1063  */
1064 static int
1065 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
1066     struct ucred **uc)
1067 {
1068 #if defined(USERSPACE)
1069         return 0;       // not supported in userspace
1070 #else
1071 #ifndef __FreeBSD__
1072         /* XXX */
1073         return cred_check(insn, proto, oif,
1074             dst_ip, dst_port, src_ip, src_port,
1075             (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
1076 #else  /* FreeBSD */
1077         struct in_addr src_ip, dst_ip;
1078         struct inpcbinfo *pi;
1079         struct ipfw_flow_id *id;
1080         struct inpcb *pcb, *inp;
1081         struct ifnet *oif;
1082         int lookupflags;
1083         int match;
1084
1085         id = &args->f_id;
1086         inp = args->inp;
1087         oif = args->oif;
1088
1089         /*
1090          * Check to see if the UDP or TCP stack supplied us with
1091          * the PCB. If so, rather then holding a lock and looking
1092          * up the PCB, we can use the one that was supplied.
1093          */
1094         if (inp && *ugid_lookupp == 0) {
1095                 INP_LOCK_ASSERT(inp);
1096                 if (inp->inp_socket != NULL) {
1097                         *uc = crhold(inp->inp_cred);
1098                         *ugid_lookupp = 1;
1099                 } else
1100                         *ugid_lookupp = -1;
1101         }
1102         /*
1103          * If we have already been here and the packet has no
1104          * PCB entry associated with it, then we can safely
1105          * assume that this is a no match.
1106          */
1107         if (*ugid_lookupp == -1)
1108                 return (0);
1109         if (id->proto == IPPROTO_TCP) {
1110                 lookupflags = 0;
1111                 pi = &V_tcbinfo;
1112         } else if (id->proto == IPPROTO_UDP) {
1113                 lookupflags = INPLOOKUP_WILDCARD;
1114                 pi = &V_udbinfo;
1115         } else if (id->proto == IPPROTO_UDPLITE) {
1116                 lookupflags = INPLOOKUP_WILDCARD;
1117                 pi = &V_ulitecbinfo;
1118         } else
1119                 return 0;
1120         lookupflags |= INPLOOKUP_RLOCKPCB;
1121         match = 0;
1122         if (*ugid_lookupp == 0) {
1123                 if (id->addr_type == 6) {
1124 #ifdef INET6
1125                         if (oif == NULL)
1126                                 pcb = in6_pcblookup_mbuf(pi,
1127                                     &id->src_ip6, htons(id->src_port),
1128                                     &id->dst_ip6, htons(id->dst_port),
1129                                     lookupflags, oif, args->m);
1130                         else
1131                                 pcb = in6_pcblookup_mbuf(pi,
1132                                     &id->dst_ip6, htons(id->dst_port),
1133                                     &id->src_ip6, htons(id->src_port),
1134                                     lookupflags, oif, args->m);
1135 #else
1136                         *ugid_lookupp = -1;
1137                         return (0);
1138 #endif
1139                 } else {
1140                         src_ip.s_addr = htonl(id->src_ip);
1141                         dst_ip.s_addr = htonl(id->dst_ip);
1142                         if (oif == NULL)
1143                                 pcb = in_pcblookup_mbuf(pi,
1144                                     src_ip, htons(id->src_port),
1145                                     dst_ip, htons(id->dst_port),
1146                                     lookupflags, oif, args->m);
1147                         else
1148                                 pcb = in_pcblookup_mbuf(pi,
1149                                     dst_ip, htons(id->dst_port),
1150                                     src_ip, htons(id->src_port),
1151                                     lookupflags, oif, args->m);
1152                 }
1153                 if (pcb != NULL) {
1154                         INP_RLOCK_ASSERT(pcb);
1155                         *uc = crhold(pcb->inp_cred);
1156                         *ugid_lookupp = 1;
1157                         INP_RUNLOCK(pcb);
1158                 }
1159                 if (*ugid_lookupp == 0) {
1160                         /*
1161                          * We tried and failed, set the variable to -1
1162                          * so we will not try again on this packet.
1163                          */
1164                         *ugid_lookupp = -1;
1165                         return (0);
1166                 }
1167         }
1168         if (insn->o.opcode == O_UID)
1169                 match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
1170         else if (insn->o.opcode == O_GID)
1171                 match = groupmember((gid_t)insn->d[0], *uc);
1172         else if (insn->o.opcode == O_JAIL)
1173                 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
1174         return (match);
1175 #endif /* __FreeBSD__ */
1176 #endif /* not supported in userspace */
1177 }
1178
1179 /*
1180  * Helper function to set args with info on the rule after the matching
1181  * one. slot is precise, whereas we guess rule_id as they are
1182  * assigned sequentially.
1183  */
1184 static inline void
1185 set_match(struct ip_fw_args *args, int slot,
1186         struct ip_fw_chain *chain)
1187 {
1188         args->rule.chain_id = chain->id;
1189         args->rule.slot = slot + 1; /* we use 0 as a marker */
1190         args->rule.rule_id = 1 + chain->map[slot]->id;
1191         args->rule.rulenum = chain->map[slot]->rulenum;
1192 }
1193
1194 #ifndef LINEAR_SKIPTO
1195 /*
1196  * Helper function to enable cached rule lookups using
1197  * cached_id and cached_pos fields in ipfw rule.
1198  */
1199 static int
1200 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1201     int tablearg, int jump_backwards)
1202 {
1203         int f_pos;
1204
1205         /* If possible use cached f_pos (in f->cached_pos),
1206          * whose version is written in f->cached_id
1207          * (horrible hacks to avoid changing the ABI).
1208          */
1209         if (num != IP_FW_TARG && f->cached_id == chain->id)
1210                 f_pos = f->cached_pos;
1211         else {
1212                 int i = IP_FW_ARG_TABLEARG(chain, num, skipto);
1213                 /* make sure we do not jump backward */
1214                 if (jump_backwards == 0 && i <= f->rulenum)
1215                         i = f->rulenum + 1;
1216                 if (chain->idxmap != NULL)
1217                         f_pos = chain->idxmap[i];
1218                 else
1219                         f_pos = ipfw_find_rule(chain, i, 0);
1220                 /* update the cache */
1221                 if (num != IP_FW_TARG) {
1222                         f->cached_id = chain->id;
1223                         f->cached_pos = f_pos;
1224                 }
1225         }
1226
1227         return (f_pos);
1228 }
1229 #else
1230 /*
1231  * Helper function to enable real fast rule lookups.
1232  */
1233 static int
1234 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1235     int tablearg, int jump_backwards)
1236 {
1237         int f_pos;
1238
1239         num = IP_FW_ARG_TABLEARG(chain, num, skipto);
1240         /* make sure we do not jump backward */
1241         if (jump_backwards == 0 && num <= f->rulenum)
1242                 num = f->rulenum + 1;
1243         f_pos = chain->idxmap[num];
1244
1245         return (f_pos);
1246 }
1247 #endif
1248
1249 #define TARG(k, f)      IP_FW_ARG_TABLEARG(chain, k, f)
1250 /*
1251  * The main check routine for the firewall.
1252  *
1253  * All arguments are in args so we can modify them and return them
1254  * back to the caller.
1255  *
1256  * Parameters:
1257  *
1258  *      args->m (in/out) The packet; we set to NULL when/if we nuke it.
1259  *              Starts with the IP header.
1260  *      args->eh (in)   Mac header if present, NULL for layer3 packet.
1261  *      args->L3offset  Number of bytes bypassed if we came from L2.
1262  *                      e.g. often sizeof(eh)  ** NOTYET **
1263  *      args->oif       Outgoing interface, NULL if packet is incoming.
1264  *              The incoming interface is in the mbuf. (in)
1265  *      args->divert_rule (in/out)
1266  *              Skip up to the first rule past this rule number;
1267  *              upon return, non-zero port number for divert or tee.
1268  *
1269  *      args->rule      Pointer to the last matching rule (in/out)
1270  *      args->next_hop  Socket we are forwarding to (out).
1271  *      args->next_hop6 IPv6 next hop we are forwarding to (out).
1272  *      args->f_id      Addresses grabbed from the packet (out)
1273  *      args->rule.info a cookie depending on rule action
1274  *
1275  * Return value:
1276  *
1277  *      IP_FW_PASS      the packet must be accepted
1278  *      IP_FW_DENY      the packet must be dropped
1279  *      IP_FW_DIVERT    divert packet, port in m_tag
1280  *      IP_FW_TEE       tee packet, port in m_tag
1281  *      IP_FW_DUMMYNET  to dummynet, pipe in args->cookie
1282  *      IP_FW_NETGRAPH  into netgraph, cookie args->cookie
1283  *              args->rule contains the matching rule,
1284  *              args->rule.info has additional information.
1285  *
1286  */
1287 int
1288 ipfw_chk(struct ip_fw_args *args)
1289 {
1290
1291         /*
1292          * Local variables holding state while processing a packet:
1293          *
1294          * IMPORTANT NOTE: to speed up the processing of rules, there
1295          * are some assumption on the values of the variables, which
1296          * are documented here. Should you change them, please check
1297          * the implementation of the various instructions to make sure
1298          * that they still work.
1299          *
1300          * args->eh     The MAC header. It is non-null for a layer2
1301          *      packet, it is NULL for a layer-3 packet.
1302          * **notyet**
1303          * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
1304          *
1305          * m | args->m  Pointer to the mbuf, as received from the caller.
1306          *      It may change if ipfw_chk() does an m_pullup, or if it
1307          *      consumes the packet because it calls send_reject().
1308          *      XXX This has to change, so that ipfw_chk() never modifies
1309          *      or consumes the buffer.
1310          * ip   is the beginning of the ip(4 or 6) header.
1311          *      Calculated by adding the L3offset to the start of data.
1312          *      (Until we start using L3offset, the packet is
1313          *      supposed to start with the ip header).
1314          */
1315         struct mbuf *m = args->m;
1316         struct ip *ip = mtod(m, struct ip *);
1317
1318         /*
1319          * For rules which contain uid/gid or jail constraints, cache
1320          * a copy of the users credentials after the pcb lookup has been
1321          * executed. This will speed up the processing of rules with
1322          * these types of constraints, as well as decrease contention
1323          * on pcb related locks.
1324          */
1325 #ifndef __FreeBSD__
1326         struct bsd_ucred ucred_cache;
1327 #else
1328         struct ucred *ucred_cache = NULL;
1329 #endif
1330         int ucred_lookup = 0;
1331
1332         /*
1333          * oif | args->oif      If NULL, ipfw_chk has been called on the
1334          *      inbound path (ether_input, ip_input).
1335          *      If non-NULL, ipfw_chk has been called on the outbound path
1336          *      (ether_output, ip_output).
1337          */
1338         struct ifnet *oif = args->oif;
1339
1340         int f_pos = 0;          /* index of current rule in the array */
1341         int retval = 0;
1342
1343         /*
1344          * hlen The length of the IP header.
1345          */
1346         u_int hlen = 0;         /* hlen >0 means we have an IP pkt */
1347
1348         /*
1349          * offset       The offset of a fragment. offset != 0 means that
1350          *      we have a fragment at this offset of an IPv4 packet.
1351          *      offset == 0 means that (if this is an IPv4 packet)
1352          *      this is the first or only fragment.
1353          *      For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
1354          *      or there is a single packet fragment (fragment header added
1355          *      without needed).  We will treat a single packet fragment as if
1356          *      there was no fragment header (or log/block depending on the
1357          *      V_fw_permit_single_frag6 sysctl setting).
1358          */
1359         u_short offset = 0;
1360         u_short ip6f_mf = 0;
1361
1362         /*
1363          * Local copies of addresses. They are only valid if we have
1364          * an IP packet.
1365          *
1366          * proto        The protocol. Set to 0 for non-ip packets,
1367          *      or to the protocol read from the packet otherwise.
1368          *      proto != 0 means that we have an IPv4 packet.
1369          *
1370          * src_port, dst_port   port numbers, in HOST format. Only
1371          *      valid for TCP and UDP packets.
1372          *
1373          * src_ip, dst_ip       ip addresses, in NETWORK format.
1374          *      Only valid for IPv4 packets.
1375          */
1376         uint8_t proto;
1377         uint16_t src_port = 0, dst_port = 0;    /* NOTE: host format    */
1378         struct in_addr src_ip, dst_ip;          /* NOTE: network format */
1379         int iplen = 0;
1380         int pktlen;
1381         uint16_t        etype = 0;      /* Host order stored ether type */
1382
1383         /*
1384          * dyn_dir = MATCH_UNKNOWN when rules unchecked,
1385          *      MATCH_NONE when checked and not matched (q = NULL),
1386          *      MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
1387          */
1388         struct ipfw_dyn_info dyn_info;
1389         struct ip_fw *q = NULL;
1390         struct ip_fw_chain *chain = &V_layer3_chain;
1391
1392         /*
1393          * We store in ulp a pointer to the upper layer protocol header.
1394          * In the ipv4 case this is easy to determine from the header,
1395          * but for ipv6 we might have some additional headers in the middle.
1396          * ulp is NULL if not found.
1397          */
1398         void *ulp = NULL;               /* upper layer protocol pointer. */
1399
1400         /* XXX ipv6 variables */
1401         int is_ipv6 = 0;
1402         uint8_t icmp6_type = 0;
1403         uint16_t ext_hd = 0;    /* bits vector for extension header filtering */
1404         /* end of ipv6 variables */
1405
1406         int is_ipv4 = 0;
1407
1408         int done = 0;           /* flag to exit the outer loop */
1409
1410         if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
1411                 return (IP_FW_PASS);    /* accept */
1412
1413         dst_ip.s_addr = 0;              /* make sure it is initialized */
1414         src_ip.s_addr = 0;              /* make sure it is initialized */
1415         pktlen = m->m_pkthdr.len;
1416         args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
1417         proto = args->f_id.proto = 0;   /* mark f_id invalid */
1418                 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
1419
1420         DYN_INFO_INIT(&dyn_info);
1421 /*
1422  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1423  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1424  * pointer might become stale after other pullups (but we never use it
1425  * this way).
1426  */
1427 #define PULLUP_TO(_len, p, T)   PULLUP_LEN(_len, p, sizeof(T))
1428 #define PULLUP_LEN(_len, p, T)                                  \
1429 do {                                                            \
1430         int x = (_len) + T;                                     \
1431         if ((m)->m_len < x) {                                   \
1432                 args->m = m = m_pullup(m, x);                   \
1433                 if (m == NULL)                                  \
1434                         goto pullup_failed;                     \
1435         }                                                       \
1436         p = (mtod(m, char *) + (_len));                         \
1437 } while (0)
1438
1439         /*
1440          * if we have an ether header,
1441          */
1442         if (args->eh)
1443                 etype = ntohs(args->eh->ether_type);
1444
1445         /* Identify IP packets and fill up variables. */
1446         if (pktlen >= sizeof(struct ip6_hdr) &&
1447             (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
1448                 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1449                 is_ipv6 = 1;
1450                 args->f_id.addr_type = 6;
1451                 hlen = sizeof(struct ip6_hdr);
1452                 proto = ip6->ip6_nxt;
1453
1454                 /* Search extension headers to find upper layer protocols */
1455                 while (ulp == NULL && offset == 0) {
1456                         switch (proto) {
1457                         case IPPROTO_ICMPV6:
1458                                 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1459                                 icmp6_type = ICMP6(ulp)->icmp6_type;
1460                                 break;
1461
1462                         case IPPROTO_TCP:
1463                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1464                                 dst_port = TCP(ulp)->th_dport;
1465                                 src_port = TCP(ulp)->th_sport;
1466                                 /* save flags for dynamic rules */
1467                                 args->f_id._flags = TCP(ulp)->th_flags;
1468                                 break;
1469
1470                         case IPPROTO_SCTP:
1471                                 if (pktlen >= hlen + sizeof(struct sctphdr) +
1472                                     sizeof(struct sctp_chunkhdr) +
1473                                     offsetof(struct sctp_init, a_rwnd))
1474                                         PULLUP_LEN(hlen, ulp,
1475                                             sizeof(struct sctphdr) +
1476                                             sizeof(struct sctp_chunkhdr) +
1477                                             offsetof(struct sctp_init, a_rwnd));
1478                                 else if (pktlen >= hlen + sizeof(struct sctphdr))
1479                                         PULLUP_LEN(hlen, ulp, pktlen - hlen);
1480                                 else
1481                                         PULLUP_LEN(hlen, ulp,
1482                                             sizeof(struct sctphdr));
1483                                 src_port = SCTP(ulp)->src_port;
1484                                 dst_port = SCTP(ulp)->dest_port;
1485                                 break;
1486
1487                         case IPPROTO_UDP:
1488                         case IPPROTO_UDPLITE:
1489                                 PULLUP_TO(hlen, ulp, struct udphdr);
1490                                 dst_port = UDP(ulp)->uh_dport;
1491                                 src_port = UDP(ulp)->uh_sport;
1492                                 break;
1493
1494                         case IPPROTO_HOPOPTS:   /* RFC 2460 */
1495                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1496                                 ext_hd |= EXT_HOPOPTS;
1497                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1498                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1499                                 ulp = NULL;
1500                                 break;
1501
1502                         case IPPROTO_ROUTING:   /* RFC 2460 */
1503                                 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1504                                 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1505                                 case 0:
1506                                         ext_hd |= EXT_RTHDR0;
1507                                         break;
1508                                 case 2:
1509                                         ext_hd |= EXT_RTHDR2;
1510                                         break;
1511                                 default:
1512                                         if (V_fw_verbose)
1513                                                 printf("IPFW2: IPV6 - Unknown "
1514                                                     "Routing Header type(%d)\n",
1515                                                     ((struct ip6_rthdr *)
1516                                                     ulp)->ip6r_type);
1517                                         if (V_fw_deny_unknown_exthdrs)
1518                                             return (IP_FW_DENY);
1519                                         break;
1520                                 }
1521                                 ext_hd |= EXT_ROUTING;
1522                                 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1523                                 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1524                                 ulp = NULL;
1525                                 break;
1526
1527                         case IPPROTO_FRAGMENT:  /* RFC 2460 */
1528                                 PULLUP_TO(hlen, ulp, struct ip6_frag);
1529                                 ext_hd |= EXT_FRAGMENT;
1530                                 hlen += sizeof (struct ip6_frag);
1531                                 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1532                                 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1533                                         IP6F_OFF_MASK;
1534                                 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1535                                         IP6F_MORE_FRAG;
1536                                 if (V_fw_permit_single_frag6 == 0 &&
1537                                     offset == 0 && ip6f_mf == 0) {
1538                                         if (V_fw_verbose)
1539                                                 printf("IPFW2: IPV6 - Invalid "
1540                                                     "Fragment Header\n");
1541                                         if (V_fw_deny_unknown_exthdrs)
1542                                             return (IP_FW_DENY);
1543                                         break;
1544                                 }
1545                                 args->f_id.extra =
1546                                     ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1547                                 ulp = NULL;
1548                                 break;
1549
1550                         case IPPROTO_DSTOPTS:   /* RFC 2460 */
1551                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1552                                 ext_hd |= EXT_DSTOPTS;
1553                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1554                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1555                                 ulp = NULL;
1556                                 break;
1557
1558                         case IPPROTO_AH:        /* RFC 2402 */
1559                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1560                                 ext_hd |= EXT_AH;
1561                                 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1562                                 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1563                                 ulp = NULL;
1564                                 break;
1565
1566                         case IPPROTO_ESP:       /* RFC 2406 */
1567                                 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1568                                 /* Anything past Seq# is variable length and
1569                                  * data past this ext. header is encrypted. */
1570                                 ext_hd |= EXT_ESP;
1571                                 break;
1572
1573                         case IPPROTO_NONE:      /* RFC 2460 */
1574                                 /*
1575                                  * Packet ends here, and IPv6 header has
1576                                  * already been pulled up. If ip6e_len!=0
1577                                  * then octets must be ignored.
1578                                  */
1579                                 ulp = ip; /* non-NULL to get out of loop. */
1580                                 break;
1581
1582                         case IPPROTO_OSPFIGP:
1583                                 /* XXX OSPF header check? */
1584                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1585                                 break;
1586
1587                         case IPPROTO_PIM:
1588                                 /* XXX PIM header check? */
1589                                 PULLUP_TO(hlen, ulp, struct pim);
1590                                 break;
1591
1592                         case IPPROTO_GRE:       /* RFC 1701 */
1593                                 /* XXX GRE header check? */
1594                                 PULLUP_TO(hlen, ulp, struct grehdr);
1595                                 break;
1596
1597                         case IPPROTO_CARP:
1598                                 PULLUP_TO(hlen, ulp, struct carp_header);
1599                                 if (((struct carp_header *)ulp)->carp_version !=
1600                                     CARP_VERSION)
1601                                         return (IP_FW_DENY);
1602                                 if (((struct carp_header *)ulp)->carp_type !=
1603                                     CARP_ADVERTISEMENT)
1604                                         return (IP_FW_DENY);
1605                                 break;
1606
1607                         case IPPROTO_IPV6:      /* RFC 2893 */
1608                                 PULLUP_TO(hlen, ulp, struct ip6_hdr);
1609                                 break;
1610
1611                         case IPPROTO_IPV4:      /* RFC 2893 */
1612                                 PULLUP_TO(hlen, ulp, struct ip);
1613                                 break;
1614
1615                         default:
1616                                 if (V_fw_verbose)
1617                                         printf("IPFW2: IPV6 - Unknown "
1618                                             "Extension Header(%d), ext_hd=%x\n",
1619                                              proto, ext_hd);
1620                                 if (V_fw_deny_unknown_exthdrs)
1621                                     return (IP_FW_DENY);
1622                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1623                                 break;
1624                         } /*switch */
1625                 }
1626                 ip = mtod(m, struct ip *);
1627                 ip6 = (struct ip6_hdr *)ip;
1628                 args->f_id.src_ip6 = ip6->ip6_src;
1629                 args->f_id.dst_ip6 = ip6->ip6_dst;
1630                 args->f_id.src_ip = 0;
1631                 args->f_id.dst_ip = 0;
1632                 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1633                 iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6);
1634         } else if (pktlen >= sizeof(struct ip) &&
1635             (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1636                 is_ipv4 = 1;
1637                 hlen = ip->ip_hl << 2;
1638                 args->f_id.addr_type = 4;
1639
1640                 /*
1641                  * Collect parameters into local variables for faster matching.
1642                  */
1643                 proto = ip->ip_p;
1644                 src_ip = ip->ip_src;
1645                 dst_ip = ip->ip_dst;
1646                 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1647                 iplen = ntohs(ip->ip_len);
1648
1649                 if (offset == 0) {
1650                         switch (proto) {
1651                         case IPPROTO_TCP:
1652                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1653                                 dst_port = TCP(ulp)->th_dport;
1654                                 src_port = TCP(ulp)->th_sport;
1655                                 /* save flags for dynamic rules */
1656                                 args->f_id._flags = TCP(ulp)->th_flags;
1657                                 break;
1658
1659                         case IPPROTO_SCTP:
1660                                 if (pktlen >= hlen + sizeof(struct sctphdr) +
1661                                     sizeof(struct sctp_chunkhdr) +
1662                                     offsetof(struct sctp_init, a_rwnd))
1663                                         PULLUP_LEN(hlen, ulp,
1664                                             sizeof(struct sctphdr) +
1665                                             sizeof(struct sctp_chunkhdr) +
1666                                             offsetof(struct sctp_init, a_rwnd));
1667                                 else if (pktlen >= hlen + sizeof(struct sctphdr))
1668                                         PULLUP_LEN(hlen, ulp, pktlen - hlen);
1669                                 else
1670                                         PULLUP_LEN(hlen, ulp,
1671                                             sizeof(struct sctphdr));
1672                                 src_port = SCTP(ulp)->src_port;
1673                                 dst_port = SCTP(ulp)->dest_port;
1674                                 break;
1675
1676                         case IPPROTO_UDP:
1677                         case IPPROTO_UDPLITE:
1678                                 PULLUP_TO(hlen, ulp, struct udphdr);
1679                                 dst_port = UDP(ulp)->uh_dport;
1680                                 src_port = UDP(ulp)->uh_sport;
1681                                 break;
1682
1683                         case IPPROTO_ICMP:
1684                                 PULLUP_TO(hlen, ulp, struct icmphdr);
1685                                 //args->f_id.flags = ICMP(ulp)->icmp_type;
1686                                 break;
1687
1688                         default:
1689                                 break;
1690                         }
1691                 }
1692
1693                 ip = mtod(m, struct ip *);
1694                 args->f_id.src_ip = ntohl(src_ip.s_addr);
1695                 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1696         }
1697 #undef PULLUP_TO
1698         pktlen = iplen < pktlen ? iplen: pktlen;
1699         if (proto) { /* we may have port numbers, store them */
1700                 args->f_id.proto = proto;
1701                 args->f_id.src_port = src_port = ntohs(src_port);
1702                 args->f_id.dst_port = dst_port = ntohs(dst_port);
1703         }
1704
1705         IPFW_PF_RLOCK(chain);
1706         if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1707                 IPFW_PF_RUNLOCK(chain);
1708                 return (IP_FW_PASS);    /* accept */
1709         }
1710         if (args->rule.slot) {
1711                 /*
1712                  * Packet has already been tagged as a result of a previous
1713                  * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1714                  * REASS, NETGRAPH, DIVERT/TEE...)
1715                  * Validate the slot and continue from the next one
1716                  * if still present, otherwise do a lookup.
1717                  */
1718                 f_pos = (args->rule.chain_id == chain->id) ?
1719                     args->rule.slot :
1720                     ipfw_find_rule(chain, args->rule.rulenum,
1721                         args->rule.rule_id);
1722         } else {
1723                 f_pos = 0;
1724         }
1725
1726         /*
1727          * Now scan the rules, and parse microinstructions for each rule.
1728          * We have two nested loops and an inner switch. Sometimes we
1729          * need to break out of one or both loops, or re-enter one of
1730          * the loops with updated variables. Loop variables are:
1731          *
1732          *      f_pos (outer loop) points to the current rule.
1733          *              On output it points to the matching rule.
1734          *      done (outer loop) is used as a flag to break the loop.
1735          *      l (inner loop)  residual length of current rule.
1736          *              cmd points to the current microinstruction.
1737          *
1738          * We break the inner loop by setting l=0 and possibly
1739          * cmdlen=0 if we don't want to advance cmd.
1740          * We break the outer loop by setting done=1
1741          * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1742          * as needed.
1743          */
1744         for (; f_pos < chain->n_rules; f_pos++) {
1745                 ipfw_insn *cmd;
1746                 uint32_t tablearg = 0;
1747                 int l, cmdlen, skip_or; /* skip rest of OR block */
1748                 struct ip_fw *f;
1749
1750                 f = chain->map[f_pos];
1751                 if (V_set_disable & (1 << f->set) )
1752                         continue;
1753
1754                 skip_or = 0;
1755                 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1756                     l -= cmdlen, cmd += cmdlen) {
1757                         int match;
1758
1759                         /*
1760                          * check_body is a jump target used when we find a
1761                          * CHECK_STATE, and need to jump to the body of
1762                          * the target rule.
1763                          */
1764
1765 /* check_body: */
1766                         cmdlen = F_LEN(cmd);
1767                         /*
1768                          * An OR block (insn_1 || .. || insn_n) has the
1769                          * F_OR bit set in all but the last instruction.
1770                          * The first match will set "skip_or", and cause
1771                          * the following instructions to be skipped until
1772                          * past the one with the F_OR bit clear.
1773                          */
1774                         if (skip_or) {          /* skip this instruction */
1775                                 if ((cmd->len & F_OR) == 0)
1776                                         skip_or = 0;    /* next one is good */
1777                                 continue;
1778                         }
1779                         match = 0; /* set to 1 if we succeed */
1780
1781                         switch (cmd->opcode) {
1782                         /*
1783                          * The first set of opcodes compares the packet's
1784                          * fields with some pattern, setting 'match' if a
1785                          * match is found. At the end of the loop there is
1786                          * logic to deal with F_NOT and F_OR flags associated
1787                          * with the opcode.
1788                          */
1789                         case O_NOP:
1790                                 match = 1;
1791                                 break;
1792
1793                         case O_FORWARD_MAC:
1794                                 printf("ipfw: opcode %d unimplemented\n",
1795                                     cmd->opcode);
1796                                 break;
1797
1798                         case O_GID:
1799                         case O_UID:
1800                         case O_JAIL:
1801                                 /*
1802                                  * We only check offset == 0 && proto != 0,
1803                                  * as this ensures that we have a
1804                                  * packet with the ports info.
1805                                  */
1806                                 if (offset != 0)
1807                                         break;
1808                                 if (proto == IPPROTO_TCP ||
1809                                     proto == IPPROTO_UDP ||
1810                                     proto == IPPROTO_UDPLITE)
1811                                         match = check_uidgid(
1812                                                     (ipfw_insn_u32 *)cmd,
1813                                                     args, &ucred_lookup,
1814 #ifdef __FreeBSD__
1815                                                     &ucred_cache);
1816 #else
1817                                                     (void *)&ucred_cache);
1818 #endif
1819                                 break;
1820
1821                         case O_RECV:
1822                                 match = iface_match(m->m_pkthdr.rcvif,
1823                                     (ipfw_insn_if *)cmd, chain, &tablearg);
1824                                 break;
1825
1826                         case O_XMIT:
1827                                 match = iface_match(oif, (ipfw_insn_if *)cmd,
1828                                     chain, &tablearg);
1829                                 break;
1830
1831                         case O_VIA:
1832                                 match = iface_match(oif ? oif :
1833                                     m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1834                                     chain, &tablearg);
1835                                 break;
1836
1837                         case O_MACADDR2:
1838                                 if (args->eh != NULL) { /* have MAC header */
1839                                         u_int32_t *want = (u_int32_t *)
1840                                                 ((ipfw_insn_mac *)cmd)->addr;
1841                                         u_int32_t *mask = (u_int32_t *)
1842                                                 ((ipfw_insn_mac *)cmd)->mask;
1843                                         u_int32_t *hdr = (u_int32_t *)args->eh;
1844
1845                                         match =
1846                                             ( want[0] == (hdr[0] & mask[0]) &&
1847                                               want[1] == (hdr[1] & mask[1]) &&
1848                                               want[2] == (hdr[2] & mask[2]) );
1849                                 }
1850                                 break;
1851
1852                         case O_MAC_TYPE:
1853                                 if (args->eh != NULL) {
1854                                         u_int16_t *p =
1855                                             ((ipfw_insn_u16 *)cmd)->ports;
1856                                         int i;
1857
1858                                         for (i = cmdlen - 1; !match && i>0;
1859                                             i--, p += 2)
1860                                                 match = (etype >= p[0] &&
1861                                                     etype <= p[1]);
1862                                 }
1863                                 break;
1864
1865                         case O_FRAG:
1866                                 match = (offset != 0);
1867                                 break;
1868
1869                         case O_IN:      /* "out" is "not in" */
1870                                 match = (oif == NULL);
1871                                 break;
1872
1873                         case O_LAYER2:
1874                                 match = (args->eh != NULL);
1875                                 break;
1876
1877                         case O_DIVERTED:
1878                             {
1879                                 /* For diverted packets, args->rule.info
1880                                  * contains the divert port (in host format)
1881                                  * reason and direction.
1882                                  */
1883                                 uint32_t i = args->rule.info;
1884                                 match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1885                                     cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1886                             }
1887                                 break;
1888
1889                         case O_PROTO:
1890                                 /*
1891                                  * We do not allow an arg of 0 so the
1892                                  * check of "proto" only suffices.
1893                                  */
1894                                 match = (proto == cmd->arg1);
1895                                 break;
1896
1897                         case O_IP_SRC:
1898                                 match = is_ipv4 &&
1899                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1900                                     src_ip.s_addr);
1901                                 break;
1902
1903                         case O_IP_DST_LOOKUP:
1904                         {
1905                                 void *pkey;
1906                                 uint32_t vidx, key;
1907                                 uint16_t keylen;
1908
1909                                 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1910                                         /* Determine lookup key type */
1911                                         vidx = ((ipfw_insn_u32 *)cmd)->d[1];
1912                                         if (vidx != 4 /* uid */ &&
1913                                             vidx != 5 /* jail */ &&
1914                                             is_ipv6 == 0 && is_ipv4 == 0)
1915                                                 break;
1916                                         /* Determine key length */
1917                                         if (vidx == 0 /* dst-ip */ ||
1918                                             vidx == 1 /* src-ip */)
1919                                                 keylen = is_ipv6 ?
1920                                                     sizeof(struct in6_addr):
1921                                                     sizeof(in_addr_t);
1922                                         else {
1923                                                 keylen = sizeof(key);
1924                                                 pkey = &key;
1925                                         }
1926                                         if (vidx == 0 /* dst-ip */)
1927                                                 pkey = is_ipv4 ? (void *)&dst_ip:
1928                                                     (void *)&args->f_id.dst_ip6;
1929                                         else if (vidx == 1 /* src-ip */)
1930                                                 pkey = is_ipv4 ? (void *)&src_ip:
1931                                                     (void *)&args->f_id.src_ip6;
1932                                         else if (vidx == 6 /* dscp */) {
1933                                                 if (is_ipv4)
1934                                                         key = ip->ip_tos >> 2;
1935                                                 else {
1936                                                         key = args->f_id.flow_id6;
1937                                                         key = (key & 0x0f) << 2 |
1938                                                             (key & 0xf000) >> 14;
1939                                                 }
1940                                                 key &= 0x3f;
1941                                         } else if (vidx == 2 /* dst-port */ ||
1942                                             vidx == 3 /* src-port */) {
1943                                                 /* Skip fragments */
1944                                                 if (offset != 0)
1945                                                         break;
1946                                                 /* Skip proto without ports */
1947                                                 if (proto != IPPROTO_TCP &&
1948                                                     proto != IPPROTO_UDP &&
1949                                                     proto != IPPROTO_UDPLITE &&
1950                                                     proto != IPPROTO_SCTP)
1951                                                         break;
1952                                                 if (vidx == 2 /* dst-port */)
1953                                                         key = dst_port;
1954                                                 else
1955                                                         key = src_port;
1956                                         }
1957 #ifndef USERSPACE
1958                                         else if (vidx == 4 /* uid */ ||
1959                                             vidx == 5 /* jail */) {
1960                                                 check_uidgid(
1961                                                     (ipfw_insn_u32 *)cmd,
1962                                                     args, &ucred_lookup,
1963 #ifdef __FreeBSD__
1964                                                     &ucred_cache);
1965                                                 if (vidx == 4 /* uid */)
1966                                                         key = ucred_cache->cr_uid;
1967                                                 else if (vidx == 5 /* jail */)
1968                                                         key = ucred_cache->cr_prison->pr_id;
1969 #else /* !__FreeBSD__ */
1970                                                     (void *)&ucred_cache);
1971                                                 if (vidx == 4 /* uid */)
1972                                                         key = ucred_cache.uid;
1973                                                 else if (vidx == 5 /* jail */)
1974                                                         key = ucred_cache.xid;
1975 #endif /* !__FreeBSD__ */
1976                                         }
1977 #endif /* !USERSPACE */
1978                                         else
1979                                                 break;
1980                                         match = ipfw_lookup_table(chain,
1981                                             cmd->arg1, keylen, pkey, &vidx);
1982                                         if (!match)
1983                                                 break;
1984                                         tablearg = vidx;
1985                                         break;
1986                                 }
1987                                 /* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */
1988                                 /* FALLTHROUGH */
1989                         }
1990                         case O_IP_SRC_LOOKUP:
1991                         {
1992                                 void *pkey;
1993                                 uint32_t vidx;
1994                                 uint16_t keylen;
1995
1996                                 if (is_ipv4) {
1997                                         keylen = sizeof(in_addr_t);
1998                                         if (cmd->opcode == O_IP_DST_LOOKUP)
1999                                                 pkey = &dst_ip;
2000                                         else
2001                                                 pkey = &src_ip;
2002                                 } else if (is_ipv6) {
2003                                         keylen = sizeof(struct in6_addr);
2004                                         if (cmd->opcode == O_IP_DST_LOOKUP)
2005                                                 pkey = &args->f_id.dst_ip6;
2006                                         else
2007                                                 pkey = &args->f_id.src_ip6;
2008                                 } else
2009                                         break;
2010                                 match = ipfw_lookup_table(chain, cmd->arg1,
2011                                     keylen, pkey, &vidx);
2012                                 if (!match)
2013                                         break;
2014                                 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) {
2015                                         match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2016                                             TARG_VAL(chain, vidx, tag);
2017                                         if (!match)
2018                                                 break;
2019                                 }
2020                                 tablearg = vidx;
2021                                 break;
2022                         }
2023
2024                         case O_IP_FLOW_LOOKUP:
2025                                 {
2026                                         uint32_t v = 0;
2027                                         match = ipfw_lookup_table(chain,
2028                                             cmd->arg1, 0, &args->f_id, &v);
2029                                         if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
2030                                                 match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2031                                                     TARG_VAL(chain, v, tag);
2032                                         if (match)
2033                                                 tablearg = v;
2034                                 }
2035                                 break;
2036                         case O_IP_SRC_MASK:
2037                         case O_IP_DST_MASK:
2038                                 if (is_ipv4) {
2039                                     uint32_t a =
2040                                         (cmd->opcode == O_IP_DST_MASK) ?
2041                                             dst_ip.s_addr : src_ip.s_addr;
2042                                     uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2043                                     int i = cmdlen-1;
2044
2045                                     for (; !match && i>0; i-= 2, p+= 2)
2046                                         match = (p[0] == (a & p[1]));
2047                                 }
2048                                 break;
2049
2050                         case O_IP_SRC_ME:
2051                                 if (is_ipv4) {
2052                                         match = in_localip(src_ip);
2053                                         break;
2054                                 }
2055 #ifdef INET6
2056                                 /* FALLTHROUGH */
2057                         case O_IP6_SRC_ME:
2058                                 match= is_ipv6 && ipfw_localip6(&args->f_id.src_ip6);
2059 #endif
2060                                 break;
2061
2062                         case O_IP_DST_SET:
2063                         case O_IP_SRC_SET:
2064                                 if (is_ipv4) {
2065                                         u_int32_t *d = (u_int32_t *)(cmd+1);
2066                                         u_int32_t addr =
2067                                             cmd->opcode == O_IP_DST_SET ?
2068                                                 args->f_id.dst_ip :
2069                                                 args->f_id.src_ip;
2070
2071                                             if (addr < d[0])
2072                                                     break;
2073                                             addr -= d[0]; /* subtract base */
2074                                             match = (addr < cmd->arg1) &&
2075                                                 ( d[ 1 + (addr>>5)] &
2076                                                   (1<<(addr & 0x1f)) );
2077                                 }
2078                                 break;
2079
2080                         case O_IP_DST:
2081                                 match = is_ipv4 &&
2082                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2083                                     dst_ip.s_addr);
2084                                 break;
2085
2086                         case O_IP_DST_ME:
2087                                 if (is_ipv4) {
2088                                         match = in_localip(dst_ip);
2089                                         break;
2090                                 }
2091 #ifdef INET6
2092                                 /* FALLTHROUGH */
2093                         case O_IP6_DST_ME:
2094                                 match= is_ipv6 && ipfw_localip6(&args->f_id.dst_ip6);
2095 #endif
2096                                 break;
2097
2098
2099                         case O_IP_SRCPORT:
2100                         case O_IP_DSTPORT:
2101                                 /*
2102                                  * offset == 0 && proto != 0 is enough
2103                                  * to guarantee that we have a
2104                                  * packet with port info.
2105                                  */
2106                                 if ((proto == IPPROTO_UDP ||
2107                                     proto == IPPROTO_UDPLITE ||
2108                                     proto == IPPROTO_TCP ||
2109                                     proto==IPPROTO_SCTP) && offset == 0) {
2110                                         u_int16_t x =
2111                                             (cmd->opcode == O_IP_SRCPORT) ?
2112                                                 src_port : dst_port ;
2113                                         u_int16_t *p =
2114                                             ((ipfw_insn_u16 *)cmd)->ports;
2115                                         int i;
2116
2117                                         for (i = cmdlen - 1; !match && i>0;
2118                                             i--, p += 2)
2119                                                 match = (x>=p[0] && x<=p[1]);
2120                                 }
2121                                 break;
2122
2123                         case O_ICMPTYPE:
2124                                 match = (offset == 0 && proto==IPPROTO_ICMP &&
2125                                     icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
2126                                 break;
2127
2128 #ifdef INET6
2129                         case O_ICMP6TYPE:
2130                                 match = is_ipv6 && offset == 0 &&
2131                                     proto==IPPROTO_ICMPV6 &&
2132                                     icmp6type_match(
2133                                         ICMP6(ulp)->icmp6_type,
2134                                         (ipfw_insn_u32 *)cmd);
2135                                 break;
2136 #endif /* INET6 */
2137
2138                         case O_IPOPT:
2139                                 match = (is_ipv4 &&
2140                                     ipopts_match(ip, cmd) );
2141                                 break;
2142
2143                         case O_IPVER:
2144                                 match = (is_ipv4 &&
2145                                     cmd->arg1 == ip->ip_v);
2146                                 break;
2147
2148                         case O_IPID:
2149                         case O_IPLEN:
2150                         case O_IPTTL:
2151                                 if (is_ipv4) {  /* only for IP packets */
2152                                     uint16_t x;
2153                                     uint16_t *p;
2154                                     int i;
2155
2156                                     if (cmd->opcode == O_IPLEN)
2157                                         x = iplen;
2158                                     else if (cmd->opcode == O_IPTTL)
2159                                         x = ip->ip_ttl;
2160                                     else /* must be IPID */
2161                                         x = ntohs(ip->ip_id);
2162                                     if (cmdlen == 1) {
2163                                         match = (cmd->arg1 == x);
2164                                         break;
2165                                     }
2166                                     /* otherwise we have ranges */
2167                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2168                                     i = cmdlen - 1;
2169                                     for (; !match && i>0; i--, p += 2)
2170                                         match = (x >= p[0] && x <= p[1]);
2171                                 }
2172                                 break;
2173
2174                         case O_IPPRECEDENCE:
2175                                 match = (is_ipv4 &&
2176                                     (cmd->arg1 == (ip->ip_tos & 0xe0)) );
2177                                 break;
2178
2179                         case O_IPTOS:
2180                                 match = (is_ipv4 &&
2181                                     flags_match(cmd, ip->ip_tos));
2182                                 break;
2183
2184                         case O_DSCP:
2185                             {
2186                                 uint32_t *p;
2187                                 uint16_t x;
2188
2189                                 p = ((ipfw_insn_u32 *)cmd)->d;
2190
2191                                 if (is_ipv4)
2192                                         x = ip->ip_tos >> 2;
2193                                 else if (is_ipv6) {
2194                                         uint8_t *v;
2195                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
2196                                         x = (*v & 0x0F) << 2;
2197                                         v++;
2198                                         x |= *v >> 6;
2199                                 } else
2200                                         break;
2201
2202                                 /* DSCP bitmask is stored as low_u32 high_u32 */
2203                                 if (x >= 32)
2204                                         match = *(p + 1) & (1 << (x - 32));
2205                                 else
2206                                         match = *p & (1 << x);
2207                             }
2208                                 break;
2209
2210                         case O_TCPDATALEN:
2211                                 if (proto == IPPROTO_TCP && offset == 0) {
2212                                     struct tcphdr *tcp;
2213                                     uint16_t x;
2214                                     uint16_t *p;
2215                                     int i;
2216 #ifdef INET6
2217                                     if (is_ipv6) {
2218                                             struct ip6_hdr *ip6;
2219
2220                                             ip6 = (struct ip6_hdr *)ip;
2221                                             if (ip6->ip6_plen == 0) {
2222                                                     /*
2223                                                      * Jumbo payload is not
2224                                                      * supported by this
2225                                                      * opcode.
2226                                                      */
2227                                                     break;
2228                                             }
2229                                             x = iplen - hlen;
2230                                     } else
2231 #endif /* INET6 */
2232                                             x = iplen - (ip->ip_hl << 2);
2233                                     tcp = TCP(ulp);
2234                                     x -= tcp->th_off << 2;
2235                                     if (cmdlen == 1) {
2236                                         match = (cmd->arg1 == x);
2237                                         break;
2238                                     }
2239                                     /* otherwise we have ranges */
2240                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2241                                     i = cmdlen - 1;
2242                                     for (; !match && i>0; i--, p += 2)
2243                                         match = (x >= p[0] && x <= p[1]);
2244                                 }
2245                                 break;
2246
2247                         case O_TCPFLAGS:
2248                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2249                                     flags_match(cmd, TCP(ulp)->th_flags));
2250                                 break;
2251
2252                         case O_TCPOPTS:
2253                                 if (proto == IPPROTO_TCP && offset == 0 && ulp){
2254                                         PULLUP_LEN(hlen, ulp,
2255                                             (TCP(ulp)->th_off << 2));
2256                                         match = tcpopts_match(TCP(ulp), cmd);
2257                                 }
2258                                 break;
2259
2260                         case O_TCPSEQ:
2261                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2262                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
2263                                         TCP(ulp)->th_seq);
2264                                 break;
2265
2266                         case O_TCPACK:
2267                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2268                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
2269                                         TCP(ulp)->th_ack);
2270                                 break;
2271
2272                         case O_TCPWIN:
2273                                 if (proto == IPPROTO_TCP && offset == 0) {
2274                                     uint16_t x;
2275                                     uint16_t *p;
2276                                     int i;
2277
2278                                     x = ntohs(TCP(ulp)->th_win);
2279                                     if (cmdlen == 1) {
2280                                         match = (cmd->arg1 == x);
2281                                         break;
2282                                     }
2283                                     /* Otherwise we have ranges. */
2284                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2285                                     i = cmdlen - 1;
2286                                     for (; !match && i > 0; i--, p += 2)
2287                                         match = (x >= p[0] && x <= p[1]);
2288                                 }
2289                                 break;
2290
2291                         case O_ESTAB:
2292                                 /* reject packets which have SYN only */
2293                                 /* XXX should i also check for TH_ACK ? */
2294                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2295                                     (TCP(ulp)->th_flags &
2296                                      (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
2297                                 break;
2298
2299                         case O_ALTQ: {
2300                                 struct pf_mtag *at;
2301                                 struct m_tag *mtag;
2302                                 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
2303
2304                                 /*
2305                                  * ALTQ uses mbuf tags from another
2306                                  * packet filtering system - pf(4).
2307                                  * We allocate a tag in its format
2308                                  * and fill it in, pretending to be pf(4).
2309                                  */
2310                                 match = 1;
2311                                 at = pf_find_mtag(m);
2312                                 if (at != NULL && at->qid != 0)
2313                                         break;
2314                                 mtag = m_tag_get(PACKET_TAG_PF,
2315                                     sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
2316                                 if (mtag == NULL) {
2317                                         /*
2318                                          * Let the packet fall back to the
2319                                          * default ALTQ.
2320                                          */
2321                                         break;
2322                                 }
2323                                 m_tag_prepend(m, mtag);
2324                                 at = (struct pf_mtag *)(mtag + 1);
2325                                 at->qid = altq->qid;
2326                                 at->hdr = ip;
2327                                 break;
2328                         }
2329
2330                         case O_LOG:
2331                                 ipfw_log(chain, f, hlen, args, m,
2332                                     oif, offset | ip6f_mf, tablearg, ip);
2333                                 match = 1;
2334                                 break;
2335
2336                         case O_PROB:
2337                                 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
2338                                 break;
2339
2340                         case O_VERREVPATH:
2341                                 /* Outgoing packets automatically pass/match */
2342                                 match = ((oif != NULL) ||
2343                                     (m->m_pkthdr.rcvif == NULL) ||
2344                                     (
2345 #ifdef INET6
2346                                     is_ipv6 ?
2347                                         verify_path6(&(args->f_id.src_ip6),
2348                                             m->m_pkthdr.rcvif, args->f_id.fib) :
2349 #endif
2350                                     verify_path(src_ip, m->m_pkthdr.rcvif,
2351                                         args->f_id.fib)));
2352                                 break;
2353
2354                         case O_VERSRCREACH:
2355                                 /* Outgoing packets automatically pass/match */
2356                                 match = (hlen > 0 && ((oif != NULL) ||
2357 #ifdef INET6
2358                                     is_ipv6 ?
2359                                         verify_path6(&(args->f_id.src_ip6),
2360                                             NULL, args->f_id.fib) :
2361 #endif
2362                                     verify_path(src_ip, NULL, args->f_id.fib)));
2363                                 break;
2364
2365                         case O_ANTISPOOF:
2366                                 /* Outgoing packets automatically pass/match */
2367                                 if (oif == NULL && hlen > 0 &&
2368                                     (  (is_ipv4 && in_localaddr(src_ip))
2369 #ifdef INET6
2370                                     || (is_ipv6 &&
2371                                         in6_localaddr(&(args->f_id.src_ip6)))
2372 #endif
2373                                     ))
2374                                         match =
2375 #ifdef INET6
2376                                             is_ipv6 ? verify_path6(
2377                                                 &(args->f_id.src_ip6),
2378                                                 m->m_pkthdr.rcvif,
2379                                                 args->f_id.fib) :
2380 #endif
2381                                             verify_path(src_ip,
2382                                                 m->m_pkthdr.rcvif,
2383                                                 args->f_id.fib);
2384                                 else
2385                                         match = 1;
2386                                 break;
2387
2388                         case O_IPSEC:
2389                                 match = (m_tag_find(m,
2390                                     PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
2391                                 /* otherwise no match */
2392                                 break;
2393
2394 #ifdef INET6
2395                         case O_IP6_SRC:
2396                                 match = is_ipv6 &&
2397                                     IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
2398                                     &((ipfw_insn_ip6 *)cmd)->addr6);
2399                                 break;
2400
2401                         case O_IP6_DST:
2402                                 match = is_ipv6 &&
2403                                 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
2404                                     &((ipfw_insn_ip6 *)cmd)->addr6);
2405                                 break;
2406                         case O_IP6_SRC_MASK:
2407                         case O_IP6_DST_MASK:
2408                                 if (is_ipv6) {
2409                                         int i = cmdlen - 1;
2410                                         struct in6_addr p;
2411                                         struct in6_addr *d =
2412                                             &((ipfw_insn_ip6 *)cmd)->addr6;
2413
2414                                         for (; !match && i > 0; d += 2,
2415                                             i -= F_INSN_SIZE(struct in6_addr)
2416                                             * 2) {
2417                                                 p = (cmd->opcode ==
2418                                                     O_IP6_SRC_MASK) ?
2419                                                     args->f_id.src_ip6:
2420                                                     args->f_id.dst_ip6;
2421                                                 APPLY_MASK(&p, &d[1]);
2422                                                 match =
2423                                                     IN6_ARE_ADDR_EQUAL(&d[0],
2424                                                     &p);
2425                                         }
2426                                 }
2427                                 break;
2428
2429                         case O_FLOW6ID:
2430                                 match = is_ipv6 &&
2431                                     flow6id_match(args->f_id.flow_id6,
2432                                     (ipfw_insn_u32 *) cmd);
2433                                 break;
2434
2435                         case O_EXT_HDR:
2436                                 match = is_ipv6 &&
2437                                     (ext_hd & ((ipfw_insn *) cmd)->arg1);
2438                                 break;
2439
2440                         case O_IP6:
2441                                 match = is_ipv6;
2442                                 break;
2443 #endif
2444
2445                         case O_IP4:
2446                                 match = is_ipv4;
2447                                 break;
2448
2449                         case O_TAG: {
2450                                 struct m_tag *mtag;
2451                                 uint32_t tag = TARG(cmd->arg1, tag);
2452
2453                                 /* Packet is already tagged with this tag? */
2454                                 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
2455
2456                                 /* We have `untag' action when F_NOT flag is
2457                                  * present. And we must remove this mtag from
2458                                  * mbuf and reset `match' to zero (`match' will
2459                                  * be inversed later).
2460                                  * Otherwise we should allocate new mtag and
2461                                  * push it into mbuf.
2462                                  */
2463                                 if (cmd->len & F_NOT) { /* `untag' action */
2464                                         if (mtag != NULL)
2465                                                 m_tag_delete(m, mtag);
2466                                         match = 0;
2467                                 } else {
2468                                         if (mtag == NULL) {
2469                                                 mtag = m_tag_alloc( MTAG_IPFW,
2470                                                     tag, 0, M_NOWAIT);
2471                                                 if (mtag != NULL)
2472                                                         m_tag_prepend(m, mtag);
2473                                         }
2474                                         match = 1;
2475                                 }
2476                                 break;
2477                         }
2478
2479                         case O_FIB: /* try match the specified fib */
2480                                 if (args->f_id.fib == cmd->arg1)
2481                                         match = 1;
2482                                 break;
2483
2484                         case O_SOCKARG: {
2485 #ifndef USERSPACE       /* not supported in userspace */
2486                                 struct inpcb *inp = args->inp;
2487                                 struct inpcbinfo *pi;
2488
2489                                 if (is_ipv6) /* XXX can we remove this ? */
2490                                         break;
2491
2492                                 if (proto == IPPROTO_TCP)
2493                                         pi = &V_tcbinfo;
2494                                 else if (proto == IPPROTO_UDP)
2495                                         pi = &V_udbinfo;
2496                                 else if (proto == IPPROTO_UDPLITE)
2497                                         pi = &V_ulitecbinfo;
2498                                 else
2499                                         break;
2500
2501                                 /*
2502                                  * XXXRW: so_user_cookie should almost
2503                                  * certainly be inp_user_cookie?
2504                                  */
2505
2506                                 /* For incoming packet, lookup up the
2507                                 inpcb using the src/dest ip/port tuple */
2508                                 if (inp == NULL) {
2509                                         inp = in_pcblookup(pi,
2510                                                 src_ip, htons(src_port),
2511                                                 dst_ip, htons(dst_port),
2512                                                 INPLOOKUP_RLOCKPCB, NULL);
2513                                         if (inp != NULL) {
2514                                                 tablearg =
2515                                                     inp->inp_socket->so_user_cookie;
2516                                                 if (tablearg)
2517                                                         match = 1;
2518                                                 INP_RUNLOCK(inp);
2519                                         }
2520                                 } else {
2521                                         if (inp->inp_socket) {
2522                                                 tablearg =
2523                                                     inp->inp_socket->so_user_cookie;
2524                                                 if (tablearg)
2525                                                         match = 1;
2526                                         }
2527                                 }
2528 #endif /* !USERSPACE */
2529                                 break;
2530                         }
2531
2532                         case O_TAGGED: {
2533                                 struct m_tag *mtag;
2534                                 uint32_t tag = TARG(cmd->arg1, tag);
2535
2536                                 if (cmdlen == 1) {
2537                                         match = m_tag_locate(m, MTAG_IPFW,
2538                                             tag, NULL) != NULL;
2539                                         break;
2540                                 }
2541
2542                                 /* we have ranges */
2543                                 for (mtag = m_tag_first(m);
2544                                     mtag != NULL && !match;
2545                                     mtag = m_tag_next(m, mtag)) {
2546                                         uint16_t *p;
2547                                         int i;
2548
2549                                         if (mtag->m_tag_cookie != MTAG_IPFW)
2550                                                 continue;
2551
2552                                         p = ((ipfw_insn_u16 *)cmd)->ports;
2553                                         i = cmdlen - 1;
2554                                         for(; !match && i > 0; i--, p += 2)
2555                                                 match =
2556                                                     mtag->m_tag_id >= p[0] &&
2557                                                     mtag->m_tag_id <= p[1];
2558                                 }
2559                                 break;
2560                         }
2561
2562                         /*
2563                          * The second set of opcodes represents 'actions',
2564                          * i.e. the terminal part of a rule once the packet
2565                          * matches all previous patterns.
2566                          * Typically there is only one action for each rule,
2567                          * and the opcode is stored at the end of the rule
2568                          * (but there are exceptions -- see below).
2569                          *
2570                          * In general, here we set retval and terminate the
2571                          * outer loop (would be a 'break 3' in some language,
2572                          * but we need to set l=0, done=1)
2573                          *
2574                          * Exceptions:
2575                          * O_COUNT and O_SKIPTO actions:
2576                          *   instead of terminating, we jump to the next rule
2577                          *   (setting l=0), or to the SKIPTO target (setting
2578                          *   f/f_len, cmd and l as needed), respectively.
2579                          *
2580                          * O_TAG, O_LOG and O_ALTQ action parameters:
2581                          *   perform some action and set match = 1;
2582                          *
2583                          * O_LIMIT and O_KEEP_STATE: these opcodes are
2584                          *   not real 'actions', and are stored right
2585                          *   before the 'action' part of the rule (one
2586                          *   exception is O_SKIP_ACTION which could be
2587                          *   between these opcodes and 'action' one).
2588                          *   These opcodes try to install an entry in the
2589                          *   state tables; if successful, we continue with
2590                          *   the next opcode (match=1; break;), otherwise
2591                          *   the packet must be dropped (set retval,
2592                          *   break loops with l=0, done=1)
2593                          *
2594                          * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2595                          *   cause a lookup of the state table, and a jump
2596                          *   to the 'action' part of the parent rule
2597                          *   if an entry is found, or
2598                          *   (CHECK_STATE only) a jump to the next rule if
2599                          *   the entry is not found.
2600                          *   The result of the lookup is cached so that
2601                          *   further instances of these opcodes become NOPs.
2602                          *   The jump to the next rule is done by setting
2603                          *   l=0, cmdlen=0.
2604                          *
2605                          * O_SKIP_ACTION: this opcode is not a real 'action'
2606                          *  either, and is stored right before the 'action'
2607                          *  part of the rule, right after the O_KEEP_STATE
2608                          *  opcode. It causes match failure so the real
2609                          *  'action' could be executed only if the rule
2610                          *  is checked via dynamic rule from the state
2611                          *  table, as in such case execution starts
2612                          *  from the true 'action' opcode directly.
2613                          *
2614                          */
2615                         case O_LIMIT:
2616                         case O_KEEP_STATE:
2617                                 if (ipfw_dyn_install_state(chain, f,
2618                                     (ipfw_insn_limit *)cmd, args, ulp,
2619                                     pktlen, &dyn_info, tablearg)) {
2620                                         /* error or limit violation */
2621                                         retval = IP_FW_DENY;
2622                                         l = 0;  /* exit inner loop */
2623                                         done = 1; /* exit outer loop */
2624                                 }
2625                                 match = 1;
2626                                 break;
2627
2628                         case O_PROBE_STATE:
2629                         case O_CHECK_STATE:
2630                                 /*
2631                                  * dynamic rules are checked at the first
2632                                  * keep-state or check-state occurrence,
2633                                  * with the result being stored in dyn_info.
2634                                  * The compiler introduces a PROBE_STATE
2635                                  * instruction for us when we have a
2636                                  * KEEP_STATE (because PROBE_STATE needs
2637                                  * to be run first).
2638                                  */
2639                                 if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) &&
2640                                     (q = ipfw_dyn_lookup_state(args, ulp,
2641                                     pktlen, cmd, &dyn_info)) != NULL) {
2642                                         /*
2643                                          * Found dynamic entry, jump to the
2644                                          * 'action' part of the parent rule
2645                                          * by setting f, cmd, l and clearing
2646                                          * cmdlen.
2647                                          */
2648                                         f = q;
2649                                         f_pos = dyn_info.f_pos;
2650                                         cmd = ACTION_PTR(f);
2651                                         l = f->cmd_len - f->act_ofs;
2652                                         cmdlen = 0;
2653                                         match = 1;
2654                                         break;
2655                                 }
2656                                 /*
2657                                  * Dynamic entry not found. If CHECK_STATE,
2658                                  * skip to next rule, if PROBE_STATE just
2659                                  * ignore and continue with next opcode.
2660                                  */
2661                                 if (cmd->opcode == O_CHECK_STATE)
2662                                         l = 0;  /* exit inner loop */
2663                                 match = 1;
2664                                 break;
2665
2666                         case O_SKIP_ACTION:
2667                                 match = 0;      /* skip to the next rule */
2668                                 l = 0;          /* exit inner loop */
2669                                 break;
2670
2671                         case O_ACCEPT:
2672                                 retval = 0;     /* accept */
2673                                 l = 0;          /* exit inner loop */
2674                                 done = 1;       /* exit outer loop */
2675                                 break;
2676
2677                         case O_PIPE:
2678                         case O_QUEUE:
2679                                 set_match(args, f_pos, chain);
2680                                 args->rule.info = TARG(cmd->arg1, pipe);
2681                                 if (cmd->opcode == O_PIPE)
2682                                         args->rule.info |= IPFW_IS_PIPE;
2683                                 if (V_fw_one_pass)
2684                                         args->rule.info |= IPFW_ONEPASS;
2685                                 retval = IP_FW_DUMMYNET;
2686                                 l = 0;          /* exit inner loop */
2687                                 done = 1;       /* exit outer loop */
2688                                 break;
2689
2690                         case O_DIVERT:
2691                         case O_TEE:
2692                                 if (args->eh) /* not on layer 2 */
2693                                     break;
2694                                 /* otherwise this is terminal */
2695                                 l = 0;          /* exit inner loop */
2696                                 done = 1;       /* exit outer loop */
2697                                 retval = (cmd->opcode == O_DIVERT) ?
2698                                         IP_FW_DIVERT : IP_FW_TEE;
2699                                 set_match(args, f_pos, chain);
2700                                 args->rule.info = TARG(cmd->arg1, divert);
2701                                 break;
2702
2703                         case O_COUNT:
2704                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2705                                 l = 0;          /* exit inner loop */
2706                                 break;
2707
2708                         case O_SKIPTO:
2709                             IPFW_INC_RULE_COUNTER(f, pktlen);
2710                             f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2711                             /*
2712                              * Skip disabled rules, and re-enter
2713                              * the inner loop with the correct
2714                              * f_pos, f, l and cmd.
2715                              * Also clear cmdlen and skip_or
2716                              */
2717                             for (; f_pos < chain->n_rules - 1 &&
2718                                     (V_set_disable &
2719                                      (1 << chain->map[f_pos]->set));
2720                                     f_pos++)
2721                                 ;
2722                             /* Re-enter the inner loop at the skipto rule. */
2723                             f = chain->map[f_pos];
2724                             l = f->cmd_len;
2725                             cmd = f->cmd;
2726                             match = 1;
2727                             cmdlen = 0;
2728                             skip_or = 0;
2729                             continue;
2730                             break;      /* not reached */
2731
2732                         case O_CALLRETURN: {
2733                                 /*
2734                                  * Implementation of `subroutine' call/return,
2735                                  * in the stack carried in an mbuf tag. This
2736                                  * is different from `skipto' in that any call
2737                                  * address is possible (`skipto' must prevent
2738                                  * backward jumps to avoid endless loops).
2739                                  * We have `return' action when F_NOT flag is
2740                                  * present. The `m_tag_id' field is used as
2741                                  * stack pointer.
2742                                  */
2743                                 struct m_tag *mtag;
2744                                 uint16_t jmpto, *stack;
2745
2746 #define IS_CALL         ((cmd->len & F_NOT) == 0)
2747 #define IS_RETURN       ((cmd->len & F_NOT) != 0)
2748                                 /*
2749                                  * Hand-rolled version of m_tag_locate() with
2750                                  * wildcard `type'.
2751                                  * If not already tagged, allocate new tag.
2752                                  */
2753                                 mtag = m_tag_first(m);
2754                                 while (mtag != NULL) {
2755                                         if (mtag->m_tag_cookie ==
2756                                             MTAG_IPFW_CALL)
2757                                                 break;
2758                                         mtag = m_tag_next(m, mtag);
2759                                 }
2760                                 if (mtag == NULL && IS_CALL) {
2761                                         mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2762                                             IPFW_CALLSTACK_SIZE *
2763                                             sizeof(uint16_t), M_NOWAIT);
2764                                         if (mtag != NULL)
2765                                                 m_tag_prepend(m, mtag);
2766                                 }
2767
2768                                 /*
2769                                  * On error both `call' and `return' just
2770                                  * continue with next rule.
2771                                  */
2772                                 if (IS_RETURN && (mtag == NULL ||
2773                                     mtag->m_tag_id == 0)) {
2774                                         l = 0;          /* exit inner loop */
2775                                         break;
2776                                 }
2777                                 if (IS_CALL && (mtag == NULL ||
2778                                     mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2779                                         printf("ipfw: call stack error, "
2780                                             "go to next rule\n");
2781                                         l = 0;          /* exit inner loop */
2782                                         break;
2783                                 }
2784
2785                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2786                                 stack = (uint16_t *)(mtag + 1);
2787
2788                                 /*
2789                                  * The `call' action may use cached f_pos
2790                                  * (in f->next_rule), whose version is written
2791                                  * in f->next_rule.
2792                                  * The `return' action, however, doesn't have
2793                                  * fixed jump address in cmd->arg1 and can't use
2794                                  * cache.
2795                                  */
2796                                 if (IS_CALL) {
2797                                         stack[mtag->m_tag_id] = f->rulenum;
2798                                         mtag->m_tag_id++;
2799                                         f_pos = JUMP(chain, f, cmd->arg1,
2800                                             tablearg, 1);
2801                                 } else {        /* `return' action */
2802                                         mtag->m_tag_id--;
2803                                         jmpto = stack[mtag->m_tag_id] + 1;
2804                                         f_pos = ipfw_find_rule(chain, jmpto, 0);
2805                                 }
2806
2807                                 /*
2808                                  * Skip disabled rules, and re-enter
2809                                  * the inner loop with the correct
2810                                  * f_pos, f, l and cmd.
2811                                  * Also clear cmdlen and skip_or
2812                                  */
2813                                 for (; f_pos < chain->n_rules - 1 &&
2814                                     (V_set_disable &
2815                                     (1 << chain->map[f_pos]->set)); f_pos++)
2816                                         ;
2817                                 /* Re-enter the inner loop at the dest rule. */
2818                                 f = chain->map[f_pos];
2819                                 l = f->cmd_len;
2820                                 cmd = f->cmd;
2821                                 cmdlen = 0;
2822                                 skip_or = 0;
2823                                 continue;
2824                                 break;  /* NOTREACHED */
2825                         }
2826 #undef IS_CALL
2827 #undef IS_RETURN
2828
2829                         case O_REJECT:
2830                                 /*
2831                                  * Drop the packet and send a reject notice
2832                                  * if the packet is not ICMP (or is an ICMP
2833                                  * query), and it is not multicast/broadcast.
2834                                  */
2835                                 if (hlen > 0 && is_ipv4 && offset == 0 &&
2836                                     (proto != IPPROTO_ICMP ||
2837                                      is_icmp_query(ICMP(ulp))) &&
2838                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2839                                     !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2840                                         send_reject(args, cmd->arg1, iplen, ip);
2841                                         m = args->m;
2842                                 }
2843                                 /* FALLTHROUGH */
2844 #ifdef INET6
2845                         case O_UNREACH6:
2846                                 if (hlen > 0 && is_ipv6 &&
2847                                     ((offset & IP6F_OFF_MASK) == 0) &&
2848                                     (proto != IPPROTO_ICMPV6 ||
2849                                      (is_icmp6_query(icmp6_type) == 1)) &&
2850                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2851                                     !IN6_IS_ADDR_MULTICAST(
2852                                         &args->f_id.dst_ip6)) {
2853                                         send_reject6(args,
2854                                             cmd->opcode == O_REJECT ?
2855                                             map_icmp_unreach(cmd->arg1):
2856                                             cmd->arg1, hlen,
2857                                             (struct ip6_hdr *)ip);
2858                                         m = args->m;
2859                                 }
2860                                 /* FALLTHROUGH */
2861 #endif
2862                         case O_DENY:
2863                                 retval = IP_FW_DENY;
2864                                 l = 0;          /* exit inner loop */
2865                                 done = 1;       /* exit outer loop */
2866                                 break;
2867
2868                         case O_FORWARD_IP:
2869                                 if (args->eh)   /* not valid on layer2 pkts */
2870                                         break;
2871                                 if (q != f ||
2872                                     dyn_info.direction == MATCH_FORWARD) {
2873                                     struct sockaddr_in *sa;
2874
2875                                     sa = &(((ipfw_insn_sa *)cmd)->sa);
2876                                     if (sa->sin_addr.s_addr == INADDR_ANY) {
2877 #ifdef INET6
2878                                         /*
2879                                          * We use O_FORWARD_IP opcode for
2880                                          * fwd rule with tablearg, but tables
2881                                          * now support IPv6 addresses. And
2882                                          * when we are inspecting IPv6 packet,
2883                                          * we can use nh6 field from
2884                                          * table_value as next_hop6 address.
2885                                          */
2886                                         if (is_ipv6) {
2887                                                 struct sockaddr_in6 *sa6;
2888
2889                                                 sa6 = args->next_hop6 =
2890                                                     &args->hopstore6;
2891                                                 sa6->sin6_family = AF_INET6;
2892                                                 sa6->sin6_len = sizeof(*sa6);
2893                                                 sa6->sin6_addr = TARG_VAL(
2894                                                     chain, tablearg, nh6);
2895                                                 sa6->sin6_port = sa->sin_port;
2896                                                 /*
2897                                                  * Set sin6_scope_id only for
2898                                                  * link-local unicast addresses.
2899                                                  */
2900                                                 if (IN6_IS_ADDR_LINKLOCAL(
2901                                                     &sa6->sin6_addr))
2902                                                         sa6->sin6_scope_id =
2903                                                             TARG_VAL(chain,
2904                                                                 tablearg,
2905                                                                 zoneid);
2906                                         } else
2907 #endif
2908                                         {
2909                                                 args->hopstore.sin_port =
2910                                                     sa->sin_port;
2911                                                 sa = args->next_hop =
2912                                                     &args->hopstore;
2913                                                 sa->sin_family = AF_INET;
2914                                                 sa->sin_len = sizeof(*sa);
2915                                                 sa->sin_addr.s_addr = htonl(
2916                                                     TARG_VAL(chain, tablearg,
2917                                                     nh4));
2918                                         }
2919                                     } else {
2920                                         args->next_hop = sa;
2921                                     }
2922                                 }
2923                                 retval = IP_FW_PASS;
2924                                 l = 0;          /* exit inner loop */
2925                                 done = 1;       /* exit outer loop */
2926                                 break;
2927
2928 #ifdef INET6
2929                         case O_FORWARD_IP6:
2930                                 if (args->eh)   /* not valid on layer2 pkts */
2931                                         break;
2932                                 if (q != f ||
2933                                     dyn_info.direction == MATCH_FORWARD) {
2934                                         struct sockaddr_in6 *sin6;
2935
2936                                         sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2937                                         args->next_hop6 = sin6;
2938                                 }
2939                                 retval = IP_FW_PASS;
2940                                 l = 0;          /* exit inner loop */
2941                                 done = 1;       /* exit outer loop */
2942                                 break;
2943 #endif
2944
2945                         case O_NETGRAPH:
2946                         case O_NGTEE:
2947                                 set_match(args, f_pos, chain);
2948                                 args->rule.info = TARG(cmd->arg1, netgraph);
2949                                 if (V_fw_one_pass)
2950                                         args->rule.info |= IPFW_ONEPASS;
2951                                 retval = (cmd->opcode == O_NETGRAPH) ?
2952                                     IP_FW_NETGRAPH : IP_FW_NGTEE;
2953                                 l = 0;          /* exit inner loop */
2954                                 done = 1;       /* exit outer loop */
2955                                 break;
2956
2957                         case O_SETFIB: {
2958                                 uint32_t fib;
2959
2960                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2961                                 fib = TARG(cmd->arg1, fib) & 0x7FFF;
2962                                 if (fib >= rt_numfibs)
2963                                         fib = 0;
2964                                 M_SETFIB(m, fib);
2965                                 args->f_id.fib = fib;
2966                                 l = 0;          /* exit inner loop */
2967                                 break;
2968                         }
2969
2970                         case O_SETDSCP: {
2971                                 uint16_t code;
2972
2973                                 code = TARG(cmd->arg1, dscp) & 0x3F;
2974                                 l = 0;          /* exit inner loop */
2975                                 if (is_ipv4) {
2976                                         uint16_t old;
2977
2978                                         old = *(uint16_t *)ip;
2979                                         ip->ip_tos = (code << 2) |
2980                                             (ip->ip_tos & 0x03);
2981                                         ip->ip_sum = cksum_adjust(ip->ip_sum,
2982                                             old, *(uint16_t *)ip);
2983                                 } else if (is_ipv6) {
2984                                         uint8_t *v;
2985
2986                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
2987                                         *v = (*v & 0xF0) | (code >> 2);
2988                                         v++;
2989                                         *v = (*v & 0x3F) | ((code & 0x03) << 6);
2990                                 } else
2991                                         break;
2992
2993                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2994                                 break;
2995                         }
2996
2997                         case O_NAT:
2998                                 l = 0;          /* exit inner loop */
2999                                 done = 1;       /* exit outer loop */
3000                                 /*
3001                                  * Ensure that we do not invoke NAT handler for
3002                                  * non IPv4 packets. Libalias expects only IPv4.
3003                                  */
3004                                 if (!is_ipv4 || !IPFW_NAT_LOADED) {
3005                                     retval = IP_FW_DENY;
3006                                     break;
3007                                 }
3008
3009                                 struct cfg_nat *t;
3010                                 int nat_id;
3011
3012                                 set_match(args, f_pos, chain);
3013                                 /* Check if this is 'global' nat rule */
3014                                 if (cmd->arg1 == IP_FW_NAT44_GLOBAL) {
3015                                         retval = ipfw_nat_ptr(args, NULL, m);
3016                                         break;
3017                                 }
3018                                 t = ((ipfw_insn_nat *)cmd)->nat;
3019                                 if (t == NULL) {
3020                                         nat_id = TARG(cmd->arg1, nat);
3021                                         t = (*lookup_nat_ptr)(&chain->nat, nat_id);
3022
3023                                         if (t == NULL) {
3024                                             retval = IP_FW_DENY;
3025                                             break;
3026                                         }
3027                                         if (cmd->arg1 != IP_FW_TARG)
3028                                             ((ipfw_insn_nat *)cmd)->nat = t;
3029                                 }
3030                                 retval = ipfw_nat_ptr(args, t, m);
3031                                 break;
3032
3033                         case O_REASS: {
3034                                 int ip_off;
3035
3036                                 l = 0;  /* in any case exit inner loop */
3037                                 if (is_ipv6) /* IPv6 is not supported yet */
3038                                         break;
3039                                 IPFW_INC_RULE_COUNTER(f, pktlen);
3040                                 ip_off = ntohs(ip->ip_off);
3041
3042                                 /* if not fragmented, go to next rule */
3043                                 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
3044                                     break;
3045
3046                                 args->m = m = ip_reass(m);
3047
3048                                 /*
3049                                  * do IP header checksum fixup.
3050                                  */
3051                                 if (m == NULL) { /* fragment got swallowed */
3052                                     retval = IP_FW_DENY;
3053                                 } else { /* good, packet complete */
3054                                     int hlen;
3055
3056                                     ip = mtod(m, struct ip *);
3057                                     hlen = ip->ip_hl << 2;
3058                                     ip->ip_sum = 0;
3059                                     if (hlen == sizeof(struct ip))
3060                                         ip->ip_sum = in_cksum_hdr(ip);
3061                                     else
3062                                         ip->ip_sum = in_cksum(m, hlen);
3063                                     retval = IP_FW_REASS;
3064                                     set_match(args, f_pos, chain);
3065                                 }
3066                                 done = 1;       /* exit outer loop */
3067                                 break;
3068                         }
3069                         case O_EXTERNAL_ACTION:
3070                                 l = 0; /* in any case exit inner loop */
3071                                 retval = ipfw_run_eaction(chain, args,
3072                                     cmd, &done);
3073                                 /*
3074                                  * If both @retval and @done are zero,
3075                                  * consider this as rule matching and
3076                                  * update counters.
3077                                  */
3078                                 if (retval == 0 && done == 0) {
3079                                         IPFW_INC_RULE_COUNTER(f, pktlen);
3080                                         /*
3081                                          * Reset the result of the last
3082                                          * dynamic state lookup.
3083                                          * External action can change
3084                                          * @args content, and it may be
3085                                          * used for new state lookup later.
3086                                          */
3087                                         DYN_INFO_INIT(&dyn_info);
3088                                 }
3089                                 break;
3090
3091                         default:
3092                                 panic("-- unknown opcode %d\n", cmd->opcode);
3093                         } /* end of switch() on opcodes */
3094                         /*
3095                          * if we get here with l=0, then match is irrelevant.
3096                          */
3097
3098                         if (cmd->len & F_NOT)
3099                                 match = !match;
3100
3101                         if (match) {
3102                                 if (cmd->len & F_OR)
3103                                         skip_or = 1;
3104                         } else {
3105                                 if (!(cmd->len & F_OR)) /* not an OR block, */
3106                                         break;          /* try next rule    */
3107                         }
3108
3109                 }       /* end of inner loop, scan opcodes */
3110 #undef PULLUP_LEN
3111
3112                 if (done)
3113                         break;
3114
3115 /* next_rule:; */       /* try next rule                */
3116
3117         }               /* end of outer for, scan rules */
3118
3119         if (done) {
3120                 struct ip_fw *rule = chain->map[f_pos];
3121                 /* Update statistics */
3122                 IPFW_INC_RULE_COUNTER(rule, pktlen);
3123         } else {
3124                 retval = IP_FW_DENY;
3125                 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3126         }
3127         IPFW_PF_RUNLOCK(chain);
3128 #ifdef __FreeBSD__
3129         if (ucred_cache != NULL)
3130                 crfree(ucred_cache);
3131 #endif
3132         return (retval);
3133
3134 pullup_failed:
3135         if (V_fw_verbose)
3136                 printf("ipfw: pullup failed\n");
3137         return (IP_FW_DENY);
3138 }
3139
3140 /*
3141  * Set maximum number of tables that can be used in given VNET ipfw instance.
3142  */
3143 #ifdef SYSCTL_NODE
3144 static int
3145 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
3146 {
3147         int error;
3148         unsigned int ntables;
3149
3150         ntables = V_fw_tables_max;
3151
3152         error = sysctl_handle_int(oidp, &ntables, 0, req);
3153         /* Read operation or some error */
3154         if ((error != 0) || (req->newptr == NULL))
3155                 return (error);
3156
3157         return (ipfw_resize_tables(&V_layer3_chain, ntables));
3158 }
3159
3160 /*
3161  * Switches table namespace between global and per-set.
3162  */
3163 static int
3164 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
3165 {
3166         int error;
3167         unsigned int sets;
3168
3169         sets = V_fw_tables_sets;
3170
3171         error = sysctl_handle_int(oidp, &sets, 0, req);
3172         /* Read operation or some error */
3173         if ((error != 0) || (req->newptr == NULL))
3174                 return (error);
3175
3176         return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
3177 }
3178 #endif
3179
3180 /*
3181  * Module and VNET glue
3182  */
3183
3184 /*
3185  * Stuff that must be initialised only on boot or module load
3186  */
3187 static int
3188 ipfw_init(void)
3189 {
3190         int error = 0;
3191
3192         /*
3193          * Only print out this stuff the first time around,
3194          * when called from the sysinit code.
3195          */
3196         printf("ipfw2 "
3197 #ifdef INET6
3198                 "(+ipv6) "
3199 #endif
3200                 "initialized, divert %s, nat %s, "
3201                 "default to %s, logging ",
3202 #ifdef IPDIVERT
3203                 "enabled",
3204 #else
3205                 "loadable",
3206 #endif
3207 #ifdef IPFIREWALL_NAT
3208                 "enabled",
3209 #else
3210                 "loadable",
3211 #endif
3212                 default_to_accept ? "accept" : "deny");
3213
3214         /*
3215          * Note: V_xxx variables can be accessed here but the vnet specific
3216          * initializer may not have been called yet for the VIMAGE case.
3217          * Tuneables will have been processed. We will print out values for
3218          * the default vnet.
3219          * XXX This should all be rationalized AFTER 8.0
3220          */
3221         if (V_fw_verbose == 0)
3222                 printf("disabled\n");
3223         else if (V_verbose_limit == 0)
3224                 printf("unlimited\n");
3225         else
3226                 printf("limited to %d packets/entry by default\n",
3227                     V_verbose_limit);
3228
3229         /* Check user-supplied table count for validness */
3230         if (default_fw_tables > IPFW_TABLES_MAX)
3231           default_fw_tables = IPFW_TABLES_MAX;
3232
3233         ipfw_init_sopt_handler();
3234         ipfw_init_obj_rewriter();
3235         ipfw_iface_init();
3236         return (error);
3237 }
3238
3239 /*
3240  * Called for the removal of the last instance only on module unload.
3241  */
3242 static void
3243 ipfw_destroy(void)
3244 {
3245
3246         ipfw_iface_destroy();
3247         ipfw_destroy_sopt_handler();
3248         ipfw_destroy_obj_rewriter();
3249         printf("IP firewall unloaded\n");
3250 }
3251
3252 /*
3253  * Stuff that must be initialized for every instance
3254  * (including the first of course).
3255  */
3256 static int
3257 vnet_ipfw_init(const void *unused)
3258 {
3259         int error, first;
3260         struct ip_fw *rule = NULL;
3261         struct ip_fw_chain *chain;
3262
3263         chain = &V_layer3_chain;
3264
3265         first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3266
3267         /* First set up some values that are compile time options */
3268         V_autoinc_step = 100;   /* bounded to 1..1000 in add_rule() */
3269         V_fw_deny_unknown_exthdrs = 1;
3270 #ifdef IPFIREWALL_VERBOSE
3271         V_fw_verbose = 1;
3272 #endif
3273 #ifdef IPFIREWALL_VERBOSE_LIMIT
3274         V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
3275 #endif
3276 #ifdef IPFIREWALL_NAT
3277         LIST_INIT(&chain->nat);
3278 #endif
3279
3280         /* Init shared services hash table */
3281         ipfw_init_srv(chain);
3282
3283         ipfw_init_counters();
3284         /* Set initial number of tables */
3285         V_fw_tables_max = default_fw_tables;
3286         error = ipfw_init_tables(chain, first);
3287         if (error) {
3288                 printf("ipfw2: setting up tables failed\n");
3289                 free(chain->map, M_IPFW);
3290                 free(rule, M_IPFW);
3291                 return (ENOSPC);
3292         }
3293
3294         IPFW_LOCK_INIT(chain);
3295
3296         /* fill and insert the default rule */
3297         rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
3298         rule->cmd_len = 1;
3299         rule->cmd[0].len = 1;
3300         rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
3301         chain->default_rule = rule;
3302         ipfw_add_protected_rule(chain, rule, 0);
3303
3304         ipfw_dyn_init(chain);
3305         ipfw_eaction_init(chain, first);
3306 #ifdef LINEAR_SKIPTO
3307         ipfw_init_skipto_cache(chain);
3308 #endif
3309         ipfw_bpf_init(first);
3310
3311         /* First set up some values that are compile time options */
3312         V_ipfw_vnet_ready = 1;          /* Open for business */
3313
3314         /*
3315          * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
3316          * Even if the latter two fail we still keep the module alive
3317          * because the sockopt and layer2 paths are still useful.
3318          * ipfw[6]_hook return 0 on success, ENOENT on failure,
3319          * so we can ignore the exact return value and just set a flag.
3320          *
3321          * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
3322          * changes in the underlying (per-vnet) variables trigger
3323          * immediate hook()/unhook() calls.
3324          * In layer2 we have the same behaviour, except that V_ether_ipfw
3325          * is checked on each packet because there are no pfil hooks.
3326          */
3327         V_ip_fw_ctl_ptr = ipfw_ctl3;
3328         error = ipfw_attach_hooks(1);
3329         return (error);
3330 }
3331
3332 /*
3333  * Called for the removal of each instance.
3334  */
3335 static int
3336 vnet_ipfw_uninit(const void *unused)
3337 {
3338         struct ip_fw *reap;
3339         struct ip_fw_chain *chain = &V_layer3_chain;
3340         int i, last;
3341
3342         V_ipfw_vnet_ready = 0; /* tell new callers to go away */
3343         /*
3344          * disconnect from ipv4, ipv6, layer2 and sockopt.
3345          * Then grab, release and grab again the WLOCK so we make
3346          * sure the update is propagated and nobody will be in.
3347          */
3348         (void)ipfw_attach_hooks(0 /* detach */);
3349         V_ip_fw_ctl_ptr = NULL;
3350
3351         last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3352
3353         IPFW_UH_WLOCK(chain);
3354         IPFW_UH_WUNLOCK(chain);
3355
3356         ipfw_dyn_uninit(0);     /* run the callout_drain */
3357
3358         IPFW_UH_WLOCK(chain);
3359
3360         reap = NULL;
3361         IPFW_WLOCK(chain);
3362         for (i = 0; i < chain->n_rules; i++)
3363                 ipfw_reap_add(chain, &reap, chain->map[i]);
3364         free(chain->map, M_IPFW);
3365 #ifdef LINEAR_SKIPTO
3366         ipfw_destroy_skipto_cache(chain);
3367 #endif
3368         IPFW_WUNLOCK(chain);
3369         IPFW_UH_WUNLOCK(chain);
3370         ipfw_destroy_tables(chain, last);
3371         ipfw_eaction_uninit(chain, last);
3372         if (reap != NULL)
3373                 ipfw_reap_rules(reap);
3374         vnet_ipfw_iface_destroy(chain);
3375         ipfw_destroy_srv(chain);
3376         IPFW_LOCK_DESTROY(chain);
3377         ipfw_dyn_uninit(1);     /* free the remaining parts */
3378         ipfw_destroy_counters();
3379         ipfw_bpf_uninit(last);
3380         return (0);
3381 }
3382
3383 /*
3384  * Module event handler.
3385  * In general we have the choice of handling most of these events by the
3386  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
3387  * use the SYSINIT handlers as they are more capable of expressing the
3388  * flow of control during module and vnet operations, so this is just
3389  * a skeleton. Note there is no SYSINIT equivalent of the module
3390  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
3391  */
3392 static int
3393 ipfw_modevent(module_t mod, int type, void *unused)
3394 {
3395         int err = 0;
3396
3397         switch (type) {
3398         case MOD_LOAD:
3399                 /* Called once at module load or
3400                  * system boot if compiled in. */
3401                 break;
3402         case MOD_QUIESCE:
3403                 /* Called before unload. May veto unloading. */
3404                 break;
3405         case MOD_UNLOAD:
3406                 /* Called during unload. */
3407                 break;
3408         case MOD_SHUTDOWN:
3409                 /* Called during system shutdown. */
3410                 break;
3411         default:
3412                 err = EOPNOTSUPP;
3413                 break;
3414         }
3415         return err;
3416 }
3417
3418 static moduledata_t ipfwmod = {
3419         "ipfw",
3420         ipfw_modevent,
3421         0
3422 };
3423
3424 /* Define startup order. */
3425 #define IPFW_SI_SUB_FIREWALL    SI_SUB_PROTO_FIREWALL
3426 #define IPFW_MODEVENT_ORDER     (SI_ORDER_ANY - 255) /* On boot slot in here. */
3427 #define IPFW_MODULE_ORDER       (IPFW_MODEVENT_ORDER + 1) /* A little later. */
3428 #define IPFW_VNET_ORDER         (IPFW_MODEVENT_ORDER + 2) /* Later still. */
3429
3430 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
3431 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
3432 MODULE_VERSION(ipfw, 3);
3433 /* should declare some dependencies here */
3434
3435 /*
3436  * Starting up. Done in order after ipfwmod() has been called.
3437  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
3438  */
3439 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3440             ipfw_init, NULL);
3441 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3442             vnet_ipfw_init, NULL);
3443
3444 /*
3445  * Closing up shop. These are done in REVERSE ORDER, but still
3446  * after ipfwmod() has been called. Not called on reboot.
3447  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
3448  * or when the module is unloaded.
3449  */
3450 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3451             ipfw_destroy, NULL);
3452 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3453             vnet_ipfw_uninit, NULL);
3454 /* end of file */