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 #include "opt_ipsec.h"
  41
  42 #include <sys/param.h>
  43 #include <sys/systm.h>
  44 #include <sys/condvar.h>
  45 #include <sys/counter.h>
  46 #include <sys/eventhandler.h>
  47 #include <sys/malloc.h>
  48 #include <sys/mbuf.h>
  49 #include <sys/kernel.h>
  50 #include <sys/lock.h>
  51 #include <sys/jail.h>
  52 #include <sys/module.h>
  53 #include <sys/priv.h>
  54 #include <sys/proc.h>
  55 #include <sys/rwlock.h>
  56 #include <sys/rmlock.h>
  57 #include <sys/socket.h>
  58 #include <sys/socketvar.h>
  59 #include <sys/sysctl.h>
  60 #include <sys/syslog.h>
  61 #include <sys/ucred.h>
  62 #include <net/ethernet.h> /* for ETHERTYPE_IP */
  63 #include <net/if.h>
  64 #include <net/if_var.h>
  65 #include <net/route.h>
  66 #include <net/pfil.h>
  67 #include <net/vnet.h>
  68
  69 #include <netpfil/pf/pf_mtag.h>
  70
  71 #include <netinet/in.h>
  72 #include <netinet/in_var.h>
  73 #include <netinet/in_pcb.h>
  74 #include <netinet/ip.h>
  75 #include <netinet/ip_var.h>
  76 #include <netinet/ip_icmp.h>
  77 #include <netinet/ip_fw.h>
  78 #include <netinet/ip_carp.h>
  79 #include <netinet/pim.h>
  80 #include <netinet/tcp_var.h>
  81 #include <netinet/udp.h>
  82 #include <netinet/udp_var.h>
  83 #include <netinet/sctp.h>
  84
  85 #include <netinet/ip6.h>
  86 #include <netinet/icmp6.h>
  87 #ifdef INET6
  88 #include <netinet6/in6_pcb.h>
  89 #include <netinet6/scope6_var.h>
  90 #include <netinet6/ip6_var.h>
  91 #endif
  92
  93 #include <netpfil/ipfw/ip_fw_private.h>
  94
  95 #include <machine/in_cksum.h>   /* XXX for in_cksum */
  96
  97 #ifdef MAC
  98 #include <security/mac/mac_framework.h>
  99 #endif
 100
 101 /*
 102  * static variables followed by global ones.
 103  * All ipfw global variables are here.
 104  */
 105
 106 static VNET_DEFINE(int, fw_deny_unknown_exthdrs);
 107 #define V_fw_deny_unknown_exthdrs       VNET(fw_deny_unknown_exthdrs)
 108
 109 static VNET_DEFINE(int, fw_permit_single_frag6) = 1;
 110 #define V_fw_permit_single_frag6        VNET(fw_permit_single_frag6)
 111
 112 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
 113 static int default_to_accept = 1;
 114 #else
 115 static int default_to_accept;
 116 #endif
 117
 118 VNET_DEFINE(int, autoinc_step);
 119 VNET_DEFINE(int, fw_one_pass) = 1;
 120
 121 VNET_DEFINE(unsigned int, fw_tables_max);
 122 VNET_DEFINE(unsigned int, fw_tables_sets) = 0;  /* Don't use set-aware tables */
 123 /* Use 128 tables by default */
 124 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT;
 125
 126 #ifndef LINEAR_SKIPTO
 127 static int jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
 128     int tablearg, int jump_backwards);
 129 #define JUMP(ch, f, num, targ, back)    jump_fast(ch, f, num, targ, back)
 130 #else
 131 static int jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
 132     int tablearg, int jump_backwards);
 133 #define JUMP(ch, f, num, targ, back)    jump_linear(ch, f, num, targ, back)
 134 #endif
 135
 136 /*
 137  * Each rule belongs to one of 32 different sets (0..31).
 138  * The variable set_disable contains one bit per set.
 139  * If the bit is set, all rules in the corresponding set
 140  * are disabled. Set RESVD_SET(31) is reserved for the default rule
 141  * and rules that are not deleted by the flush command,
 142  * and CANNOT be disabled.
 143  * Rules in set RESVD_SET can only be deleted individually.
 144  */
 145 VNET_DEFINE(u_int32_t, set_disable);
 146 #define V_set_disable                   VNET(set_disable)
 147
 148 VNET_DEFINE(int, fw_verbose);
 149 /* counter for ipfw_log(NULL...) */
 150 VNET_DEFINE(u_int64_t, norule_counter);
 151 VNET_DEFINE(int, verbose_limit);
 152
 153 /* layer3_chain contains the list of rules for layer 3 */
 154 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
 155
 156 /* ipfw_vnet_ready controls when we are open for business */
 157 VNET_DEFINE(int, ipfw_vnet_ready) = 0;
 158
 159 VNET_DEFINE(int, ipfw_nat_ready) = 0;
 160
 161 ipfw_nat_t *ipfw_nat_ptr = NULL;
 162 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
 163 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
 164 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
 165 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
 166 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
 167
 168 #ifdef SYSCTL_NODE
 169 uint32_t dummy_def = IPFW_DEFAULT_RULE;
 170 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS);
 171 static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS);
 172
 173 SYSBEGIN(f3)
 174
 175 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 176 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
 177     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
 178     "Only do a single pass through ipfw when using dummynet(4)");
 179 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
 180     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
 181     "Rule number auto-increment step");
 182 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose,
 183     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
 184     "Log matches to ipfw rules");
 185 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
 186     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
 187     "Set upper limit of matches of ipfw rules logged");
 188 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
 189     &dummy_def, 0,
 190     "The default/max possible rule number.");
 191 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
 192     CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU",
 193     "Maximum number of concurrently used tables");
 194 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets,
 195     CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW,
 196     0, 0, sysctl_ipfw_tables_sets, "IU",
 197     "Use per-set namespace for tables");
 198 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
 199     &default_to_accept, 0,
 200     "Make the default rule accept all packets.");
 201 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables);
 202 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count,
 203     CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
 204     "Number of static rules");
 205
 206 #ifdef INET6
 207 SYSCTL_DECL(_net_inet6_ip6);
 208 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 209 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
 210     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
 211     &VNET_NAME(fw_deny_unknown_exthdrs), 0,
 212     "Deny packets with unknown IPv6 Extension Headers");
 213 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
 214     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
 215     &VNET_NAME(fw_permit_single_frag6), 0,
 216     "Permit single packet IPv6 fragments");
 217 #endif /* INET6 */
 218
 219 SYSEND
 220
 221 #endif /* SYSCTL_NODE */
 222
 223
 224 /*
 225  * Some macros used in the various matching options.
 226  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
 227  * Other macros just cast void * into the appropriate type
 228  */
 229 #define L3HDR(T, ip)    ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
 230 #define TCP(p)          ((struct tcphdr *)(p))
 231 #define SCTP(p)         ((struct sctphdr *)(p))
 232 #define UDP(p)          ((struct udphdr *)(p))
 233 #define ICMP(p)         ((struct icmphdr *)(p))
 234 #define ICMP6(p)        ((struct icmp6_hdr *)(p))
 235
 236 static __inline int
 237 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
 238 {
 239         int type = icmp->icmp_type;
 240
 241         return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
 242 }
 243
 244 #define TT      ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
 245     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
 246
 247 static int
 248 is_icmp_query(struct icmphdr *icmp)
 249 {
 250         int type = icmp->icmp_type;
 251
 252         return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
 253 }
 254 #undef TT
 255
 256 /*
 257  * The following checks use two arrays of 8 or 16 bits to store the
 258  * bits that we want set or clear, respectively. They are in the
 259  * low and high half of cmd->arg1 or cmd->d[0].
 260  *
 261  * We scan options and store the bits we find set. We succeed if
 262  *
 263  *      (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
 264  *
 265  * The code is sometimes optimized not to store additional variables.
 266  */
 267
 268 static int
 269 flags_match(ipfw_insn *cmd, u_int8_t bits)
 270 {
 271         u_char want_clear;
 272         bits = ~bits;
 273
 274         if ( ((cmd->arg1 & 0xff) & bits) != 0)
 275                 return 0; /* some bits we want set were clear */
 276         want_clear = (cmd->arg1 >> 8) & 0xff;
 277         if ( (want_clear & bits) != want_clear)
 278                 return 0; /* some bits we want clear were set */
 279         return 1;
 280 }
 281
 282 static int
 283 ipopts_match(struct ip *ip, ipfw_insn *cmd)
 284 {
 285         int optlen, bits = 0;
 286         u_char *cp = (u_char *)(ip + 1);
 287         int x = (ip->ip_hl << 2) - sizeof (struct ip);
 288
 289         for (; x > 0; x -= optlen, cp += optlen) {
 290                 int opt = cp[IPOPT_OPTVAL];
 291
 292                 if (opt == IPOPT_EOL)
 293                         break;
 294                 if (opt == IPOPT_NOP)
 295                         optlen = 1;
 296                 else {
 297                         optlen = cp[IPOPT_OLEN];
 298                         if (optlen <= 0 || optlen > x)
 299                                 return 0; /* invalid or truncated */
 300                 }
 301                 switch (opt) {
 302
 303                 default:
 304                         break;
 305
 306                 case IPOPT_LSRR:
 307                         bits |= IP_FW_IPOPT_LSRR;
 308                         break;
 309
 310                 case IPOPT_SSRR:
 311                         bits |= IP_FW_IPOPT_SSRR;
 312                         break;
 313
 314                 case IPOPT_RR:
 315                         bits |= IP_FW_IPOPT_RR;
 316                         break;
 317
 318                 case IPOPT_TS:
 319                         bits |= IP_FW_IPOPT_TS;
 320                         break;
 321                 }
 322         }
 323         return (flags_match(cmd, bits));
 324 }
 325
 326 static int
 327 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
 328 {
 329         int optlen, bits = 0;
 330         u_char *cp = (u_char *)(tcp + 1);
 331         int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
 332
 333         for (; x > 0; x -= optlen, cp += optlen) {
 334                 int opt = cp[0];
 335                 if (opt == TCPOPT_EOL)
 336                         break;
 337                 if (opt == TCPOPT_NOP)
 338                         optlen = 1;
 339                 else {
 340                         optlen = cp[1];
 341                         if (optlen <= 0)
 342                                 break;
 343                 }
 344
 345                 switch (opt) {
 346
 347                 default:
 348                         break;
 349
 350                 case TCPOPT_MAXSEG:
 351                         bits |= IP_FW_TCPOPT_MSS;
 352                         break;
 353
 354                 case TCPOPT_WINDOW:
 355                         bits |= IP_FW_TCPOPT_WINDOW;
 356                         break;
 357
 358                 case TCPOPT_SACK_PERMITTED:
 359                 case TCPOPT_SACK:
 360                         bits |= IP_FW_TCPOPT_SACK;
 361                         break;
 362
 363                 case TCPOPT_TIMESTAMP:
 364                         bits |= IP_FW_TCPOPT_TS;
 365                         break;
 366
 367                 }
 368         }
 369         return (flags_match(cmd, bits));
 370 }
 371
 372 static int
 373 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain,
 374     uint32_t *tablearg)
 375 {
 376
 377         if (ifp == NULL)        /* no iface with this packet, match fails */
 378                 return (0);
 379
 380         /* Check by name or by IP address */
 381         if (cmd->name[0] != '\0') { /* match by name */
 382                 if (cmd->name[0] == '\1') /* use tablearg to match */
 383                         return ipfw_lookup_table_extended(chain, cmd->p.kidx, 0,
 384                                 &ifp->if_index, tablearg);
 385                 /* Check name */
 386                 if (cmd->p.glob) {
 387                         if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
 388                                 return(1);
 389                 } else {
 390                         if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
 391                                 return(1);
 392                 }
 393         } else {
 394 #if !defined(USERSPACE) && defined(__FreeBSD__) /* and OSX too ? */
 395                 struct ifaddr *ia;
 396
 397                 if_addr_rlock(ifp);
 398                 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
 399                         if (ia->ifa_addr->sa_family != AF_INET)
 400                                 continue;
 401                         if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
 402                             (ia->ifa_addr))->sin_addr.s_addr) {
 403                                 if_addr_runlock(ifp);
 404                                 return(1);      /* match */
 405                         }
 406                 }
 407                 if_addr_runlock(ifp);
 408 #endif /* __FreeBSD__ */
 409         }
 410         return(0);      /* no match, fail ... */
 411 }
 412
 413 /*
 414  * The verify_path function checks if a route to the src exists and
 415  * if it is reachable via ifp (when provided).
 416  *
 417  * The 'verrevpath' option checks that the interface that an IP packet
 418  * arrives on is the same interface that traffic destined for the
 419  * packet's source address would be routed out of.
 420  * The 'versrcreach' option just checks that the source address is
 421  * reachable via any route (except default) in the routing table.
 422  * These two are a measure to block forged packets. This is also
 423  * commonly known as "anti-spoofing" or Unicast Reverse Path
 424  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
 425  * is purposely reminiscent of the Cisco IOS command,
 426  *
 427  *   ip verify unicast reverse-path
 428  *   ip verify unicast source reachable-via any
 429  *
 430  * which implements the same functionality. But note that the syntax
 431  * is misleading, and the check may be performed on all IP packets
 432  * whether unicast, multicast, or broadcast.
 433  */
 434 static int
 435 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
 436 {
 437 #if defined(USERSPACE) || !defined(__FreeBSD__)
 438         return 0;
 439 #else
 440         struct route ro;
 441         struct sockaddr_in *dst;
 442
 443         bzero(&ro, sizeof(ro));
 444
 445         dst = (struct sockaddr_in *)&(ro.ro_dst);
 446         dst->sin_family = AF_INET;
 447         dst->sin_len = sizeof(*dst);
 448         dst->sin_addr = src;
 449         in_rtalloc_ign(&ro, 0, fib);
 450
 451         if (ro.ro_rt == NULL)
 452                 return 0;
 453
 454         /*
 455          * If ifp is provided, check for equality with rtentry.
 456          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 457          * in order to pass packets injected back by if_simloop():
 458          * routing entry (via lo0) for our own address
 459          * may exist, so we need to handle routing assymetry.
 460          */
 461         if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
 462                 RTFREE(ro.ro_rt);
 463                 return 0;
 464         }
 465
 466         /* if no ifp provided, check if rtentry is not default route */
 467         if (ifp == NULL &&
 468              satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) {
 469                 RTFREE(ro.ro_rt);
 470                 return 0;
 471         }
 472
 473         /* or if this is a blackhole/reject route */
 474         if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
 475                 RTFREE(ro.ro_rt);
 476                 return 0;
 477         }
 478
 479         /* found valid route */
 480         RTFREE(ro.ro_rt);
 481         return 1;
 482 #endif /* __FreeBSD__ */
 483 }
 484
 485 #ifdef INET6
 486 /*
 487  * ipv6 specific rules here...
 488  */
 489 static __inline int
 490 icmp6type_match (int type, ipfw_insn_u32 *cmd)
 491 {
 492         return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
 493 }
 494
 495 static int
 496 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
 497 {
 498         int i;
 499         for (i=0; i <= cmd->o.arg1; ++i )
 500                 if (curr_flow == cmd->d[i] )
 501                         return 1;
 502         return 0;
 503 }
 504
 505 /* support for IP6_*_ME opcodes */
 506 static const struct in6_addr lla_mask = {{{
 507         0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff,
 508         0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
 509 }}};
 510
 511 static int
 512 ipfw_localip6(struct in6_addr *in6)
 513 {
 514         struct rm_priotracker in6_ifa_tracker;
 515         struct in6_ifaddr *ia;
 516
 517         if (IN6_IS_ADDR_MULTICAST(in6))
 518                 return (0);
 519
 520         if (!IN6_IS_ADDR_LINKLOCAL(in6))
 521                 return (in6_localip(in6));
 522
 523         IN6_IFADDR_RLOCK(&in6_ifa_tracker);
 524         TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
 525                 if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr))
 526                         continue;
 527                 if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
 528                     in6, &lla_mask)) {
 529                         IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
 530                         return (1);
 531                 }
 532         }
 533         IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
 534         return (0);
 535 }
 536
 537 static int
 538 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
 539 {
 540         struct route_in6 ro;
 541         struct sockaddr_in6 *dst;
 542
 543         bzero(&ro, sizeof(ro));
 544
 545         dst = (struct sockaddr_in6 * )&(ro.ro_dst);
 546         dst->sin6_family = AF_INET6;
 547         dst->sin6_len = sizeof(*dst);
 548         dst->sin6_addr = *src;
 549
 550         in6_rtalloc_ign(&ro, 0, fib);
 551         if (ro.ro_rt == NULL)
 552                 return 0;
 553
 554         /*
 555          * if ifp is provided, check for equality with rtentry
 556          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 557          * to support the case of sending packets to an address of our own.
 558          * (where the former interface is the first argument of if_simloop()
 559          *  (=ifp), the latter is lo0)
 560          */
 561         if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
 562                 RTFREE(ro.ro_rt);
 563                 return 0;
 564         }
 565
 566         /* if no ifp provided, check if rtentry is not default route */
 567         if (ifp == NULL &&
 568             IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) {
 569                 RTFREE(ro.ro_rt);
 570                 return 0;
 571         }
 572
 573         /* or if this is a blackhole/reject route */
 574         if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
 575                 RTFREE(ro.ro_rt);
 576                 return 0;
 577         }
 578
 579         /* found valid route */
 580         RTFREE(ro.ro_rt);
 581         return 1;
 582
 583 }
 584
 585 static int
 586 is_icmp6_query(int icmp6_type)
 587 {
 588         if ((icmp6_type <= ICMP6_MAXTYPE) &&
 589             (icmp6_type == ICMP6_ECHO_REQUEST ||
 590             icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
 591             icmp6_type == ICMP6_WRUREQUEST ||
 592             icmp6_type == ICMP6_FQDN_QUERY ||
 593             icmp6_type == ICMP6_NI_QUERY))
 594                 return (1);
 595
 596         return (0);
 597 }
 598
 599 static void
 600 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
 601 {
 602         struct mbuf *m;
 603
 604         m = args->m;
 605         if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
 606                 struct tcphdr *tcp;
 607                 tcp = (struct tcphdr *)((char *)ip6 + hlen);
 608
 609                 if ((tcp->th_flags & TH_RST) == 0) {
 610                         struct mbuf *m0;
 611                         m0 = ipfw_send_pkt(args->m, &(args->f_id),
 612                             ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 613                             tcp->th_flags | TH_RST);
 614                         if (m0 != NULL)
 615                                 ip6_output(m0, NULL, NULL, 0, NULL, NULL,
 616                                     NULL);
 617                 }
 618                 FREE_PKT(m);
 619         } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
 620 #if 0
 621                 /*
 622                  * Unlike above, the mbufs need to line up with the ip6 hdr,
 623                  * as the contents are read. We need to m_adj() the
 624                  * needed amount.
 625                  * The mbuf will however be thrown away so we can adjust it.
 626                  * Remember we did an m_pullup on it already so we
 627                  * can make some assumptions about contiguousness.
 628                  */
 629                 if (args->L3offset)
 630                         m_adj(m, args->L3offset);
 631 #endif
 632                 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
 633         } else
 634                 FREE_PKT(m);
 635
 636         args->m = NULL;
 637 }
 638
 639 #endif /* INET6 */
 640
 641
 642 /*
 643  * sends a reject message, consuming the mbuf passed as an argument.
 644  */
 645 static void
 646 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
 647 {
 648
 649 #if 0
 650         /* XXX When ip is not guaranteed to be at mtod() we will
 651          * need to account for this */
 652          * The mbuf will however be thrown away so we can adjust it.
 653          * Remember we did an m_pullup on it already so we
 654          * can make some assumptions about contiguousness.
 655          */
 656         if (args->L3offset)
 657                 m_adj(m, args->L3offset);
 658 #endif
 659         if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
 660                 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
 661         } else if (args->f_id.proto == IPPROTO_TCP) {
 662                 struct tcphdr *const tcp =
 663                     L3HDR(struct tcphdr, mtod(args->m, struct ip *));
 664                 if ( (tcp->th_flags & TH_RST) == 0) {
 665                         struct mbuf *m;
 666                         m = ipfw_send_pkt(args->m, &(args->f_id),
 667                                 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 668                                 tcp->th_flags | TH_RST);
 669                         if (m != NULL)
 670                                 ip_output(m, NULL, NULL, 0, NULL, NULL);
 671                 }
 672                 FREE_PKT(args->m);
 673         } else
 674                 FREE_PKT(args->m);
 675         args->m = NULL;
 676 }
 677
 678 /*
 679  * Support for uid/gid/jail lookup. These tests are expensive
 680  * (because we may need to look into the list of active sockets)
 681  * so we cache the results. ugid_lookupp is 0 if we have not
 682  * yet done a lookup, 1 if we succeeded, and -1 if we tried
 683  * and failed. The function always returns the match value.
 684  * We could actually spare the variable and use *uc, setting
 685  * it to '(void *)check_uidgid if we have no info, NULL if
 686  * we tried and failed, or any other value if successful.
 687  */
 688 static int
 689 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
 690     struct ucred **uc)
 691 {
 692 #if defined(USERSPACE)
 693         return 0;       // not supported in userspace
 694 #else
 695 #ifndef __FreeBSD__
 696         /* XXX */
 697         return cred_check(insn, proto, oif,
 698             dst_ip, dst_port, src_ip, src_port,
 699             (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
 700 #else  /* FreeBSD */
 701         struct in_addr src_ip, dst_ip;
 702         struct inpcbinfo *pi;
 703         struct ipfw_flow_id *id;
 704         struct inpcb *pcb, *inp;
 705         struct ifnet *oif;
 706         int lookupflags;
 707         int match;
 708
 709         id = &args->f_id;
 710         inp = args->inp;
 711         oif = args->oif;
 712
 713         /*
 714          * Check to see if the UDP or TCP stack supplied us with
 715          * the PCB. If so, rather then holding a lock and looking
 716          * up the PCB, we can use the one that was supplied.
 717          */
 718         if (inp && *ugid_lookupp == 0) {
 719                 INP_LOCK_ASSERT(inp);
 720                 if (inp->inp_socket != NULL) {
 721                         *uc = crhold(inp->inp_cred);
 722                         *ugid_lookupp = 1;
 723                 } else
 724                         *ugid_lookupp = -1;
 725         }
 726         /*
 727          * If we have already been here and the packet has no
 728          * PCB entry associated with it, then we can safely
 729          * assume that this is a no match.
 730          */
 731         if (*ugid_lookupp == -1)
 732                 return (0);
 733         if (id->proto == IPPROTO_TCP) {
 734                 lookupflags = 0;
 735                 pi = &V_tcbinfo;
 736         } else if (id->proto == IPPROTO_UDP) {
 737                 lookupflags = INPLOOKUP_WILDCARD;
 738                 pi = &V_udbinfo;
 739         } else
 740                 return 0;
 741         lookupflags |= INPLOOKUP_RLOCKPCB;
 742         match = 0;
 743         if (*ugid_lookupp == 0) {
 744                 if (id->addr_type == 6) {
 745 #ifdef INET6
 746                         if (oif == NULL)
 747                                 pcb = in6_pcblookup_mbuf(pi,
 748                                     &id->src_ip6, htons(id->src_port),
 749                                     &id->dst_ip6, htons(id->dst_port),
 750                                     lookupflags, oif, args->m);
 751                         else
 752                                 pcb = in6_pcblookup_mbuf(pi,
 753                                     &id->dst_ip6, htons(id->dst_port),
 754                                     &id->src_ip6, htons(id->src_port),
 755                                     lookupflags, oif, args->m);
 756 #else
 757                         *ugid_lookupp = -1;
 758                         return (0);
 759 #endif
 760                 } else {
 761                         src_ip.s_addr = htonl(id->src_ip);
 762                         dst_ip.s_addr = htonl(id->dst_ip);
 763                         if (oif == NULL)
 764                                 pcb = in_pcblookup_mbuf(pi,
 765                                     src_ip, htons(id->src_port),
 766                                     dst_ip, htons(id->dst_port),
 767                                     lookupflags, oif, args->m);
 768                         else
 769                                 pcb = in_pcblookup_mbuf(pi,
 770                                     dst_ip, htons(id->dst_port),
 771                                     src_ip, htons(id->src_port),
 772                                     lookupflags, oif, args->m);
 773                 }
 774                 if (pcb != NULL) {
 775                         INP_RLOCK_ASSERT(pcb);
 776                         *uc = crhold(pcb->inp_cred);
 777                         *ugid_lookupp = 1;
 778                         INP_RUNLOCK(pcb);
 779                 }
 780                 if (*ugid_lookupp == 0) {
 781                         /*
 782                          * We tried and failed, set the variable to -1
 783                          * so we will not try again on this packet.
 784                          */
 785                         *ugid_lookupp = -1;
 786                         return (0);
 787                 }
 788         }
 789         if (insn->o.opcode == O_UID)
 790                 match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
 791         else if (insn->o.opcode == O_GID)
 792                 match = groupmember((gid_t)insn->d[0], *uc);
 793         else if (insn->o.opcode == O_JAIL)
 794                 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
 795         return (match);
 796 #endif /* __FreeBSD__ */
 797 #endif /* not supported in userspace */
 798 }
 799
 800 /*
 801  * Helper function to set args with info on the rule after the matching
 802  * one. slot is precise, whereas we guess rule_id as they are
 803  * assigned sequentially.
 804  */
 805 static inline void
 806 set_match(struct ip_fw_args *args, int slot,
 807         struct ip_fw_chain *chain)
 808 {
 809         args->rule.chain_id = chain->id;
 810         args->rule.slot = slot + 1; /* we use 0 as a marker */
 811         args->rule.rule_id = 1 + chain->map[slot]->id;
 812         args->rule.rulenum = chain->map[slot]->rulenum;
 813 }
 814
 815 #ifndef LINEAR_SKIPTO
 816 /*
 817  * Helper function to enable cached rule lookups using
 818  * cached_id and cached_pos fields in ipfw rule.
 819  */
 820 static int
 821 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
 822     int tablearg, int jump_backwards)
 823 {
 824         int f_pos;
 825
 826         /* If possible use cached f_pos (in f->cached_pos),
 827          * whose version is written in f->cached_id
 828          * (horrible hacks to avoid changing the ABI).
 829          */
 830         if (num != IP_FW_TARG && f->cached_id == chain->id)
 831                 f_pos = f->cached_pos;
 832         else {
 833                 int i = IP_FW_ARG_TABLEARG(chain, num, skipto);
 834                 /* make sure we do not jump backward */
 835                 if (jump_backwards == 0 && i <= f->rulenum)
 836                         i = f->rulenum + 1;
 837                 if (chain->idxmap != NULL)
 838                         f_pos = chain->idxmap[i];
 839                 else
 840                         f_pos = ipfw_find_rule(chain, i, 0);
 841                 /* update the cache */
 842                 if (num != IP_FW_TARG) {
 843                         f->cached_id = chain->id;
 844                         f->cached_pos = f_pos;
 845                 }
 846         }
 847
 848         return (f_pos);
 849 }
 850 #else
 851 /*
 852  * Helper function to enable real fast rule lookups.
 853  */
 854 static int
 855 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
 856     int tablearg, int jump_backwards)
 857 {
 858         int f_pos;
 859
 860         num = IP_FW_ARG_TABLEARG(chain, num, skipto);
 861         /* make sure we do not jump backward */
 862         if (jump_backwards == 0 && num <= f->rulenum)
 863                 num = f->rulenum + 1;
 864         f_pos = chain->idxmap[num];
 865
 866         return (f_pos);
 867 }
 868 #endif
 869
 870 #define TARG(k, f)      IP_FW_ARG_TABLEARG(chain, k, f)
 871 /*
 872  * The main check routine for the firewall.
 873  *
 874  * All arguments are in args so we can modify them and return them
 875  * back to the caller.
 876  *
 877  * Parameters:
 878  *
 879  *      args->m (in/out) The packet; we set to NULL when/if we nuke it.
 880  *              Starts with the IP header.
 881  *      args->eh (in)   Mac header if present, NULL for layer3 packet.
 882  *      args->L3offset  Number of bytes bypassed if we came from L2.
 883  *                      e.g. often sizeof(eh)  ** NOTYET **
 884  *      args->oif       Outgoing interface, NULL if packet is incoming.
 885  *              The incoming interface is in the mbuf. (in)
 886  *      args->divert_rule (in/out)
 887  *              Skip up to the first rule past this rule number;
 888  *              upon return, non-zero port number for divert or tee.
 889  *
 890  *      args->rule      Pointer to the last matching rule (in/out)
 891  *      args->next_hop  Socket we are forwarding to (out).
 892  *      args->next_hop6 IPv6 next hop we are forwarding to (out).
 893  *      args->f_id      Addresses grabbed from the packet (out)
 894  *      args->rule.info a cookie depending on rule action
 895  *
 896  * Return value:
 897  *
 898  *      IP_FW_PASS      the packet must be accepted
 899  *      IP_FW_DENY      the packet must be dropped
 900  *      IP_FW_DIVERT    divert packet, port in m_tag
 901  *      IP_FW_TEE       tee packet, port in m_tag
 902  *      IP_FW_DUMMYNET  to dummynet, pipe in args->cookie
 903  *      IP_FW_NETGRAPH  into netgraph, cookie args->cookie
 904  *              args->rule contains the matching rule,
 905  *              args->rule.info has additional information.
 906  *
 907  */
 908 int
 909 ipfw_chk(struct ip_fw_args *args)
 910 {
 911
 912         /*
 913          * Local variables holding state while processing a packet:
 914          *
 915          * IMPORTANT NOTE: to speed up the processing of rules, there
 916          * are some assumption on the values of the variables, which
 917          * are documented here. Should you change them, please check
 918          * the implementation of the various instructions to make sure
 919          * that they still work.
 920          *
 921          * args->eh     The MAC header. It is non-null for a layer2
 922          *      packet, it is NULL for a layer-3 packet.
 923          * **notyet**
 924          * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
 925          *
 926          * m | args->m  Pointer to the mbuf, as received from the caller.
 927          *      It may change if ipfw_chk() does an m_pullup, or if it
 928          *      consumes the packet because it calls send_reject().
 929          *      XXX This has to change, so that ipfw_chk() never modifies
 930          *      or consumes the buffer.
 931          * ip   is the beginning of the ip(4 or 6) header.
 932          *      Calculated by adding the L3offset to the start of data.
 933          *      (Until we start using L3offset, the packet is
 934          *      supposed to start with the ip header).
 935          */
 936         struct mbuf *m = args->m;
 937         struct ip *ip = mtod(m, struct ip *);
 938
 939         /*
 940          * For rules which contain uid/gid or jail constraints, cache
 941          * a copy of the users credentials after the pcb lookup has been
 942          * executed. This will speed up the processing of rules with
 943          * these types of constraints, as well as decrease contention
 944          * on pcb related locks.
 945          */
 946 #ifndef __FreeBSD__
 947         struct bsd_ucred ucred_cache;
 948 #else
 949         struct ucred *ucred_cache = NULL;
 950 #endif
 951         int ucred_lookup = 0;
 952
 953         /*
 954          * oif | args->oif      If NULL, ipfw_chk has been called on the
 955          *      inbound path (ether_input, ip_input).
 956          *      If non-NULL, ipfw_chk has been called on the outbound path
 957          *      (ether_output, ip_output).
 958          */
 959         struct ifnet *oif = args->oif;
 960
 961         int f_pos = 0;          /* index of current rule in the array */
 962         int retval = 0;
 963
 964         /*
 965          * hlen The length of the IP header.
 966          */
 967         u_int hlen = 0;         /* hlen >0 means we have an IP pkt */
 968
 969         /*
 970          * offset       The offset of a fragment. offset != 0 means that
 971          *      we have a fragment at this offset of an IPv4 packet.
 972          *      offset == 0 means that (if this is an IPv4 packet)
 973          *      this is the first or only fragment.
 974          *      For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
 975          *      or there is a single packet fragement (fragement header added
 976          *      without needed).  We will treat a single packet fragment as if
 977          *      there was no fragment header (or log/block depending on the
 978          *      V_fw_permit_single_frag6 sysctl setting).
 979          */
 980         u_short offset = 0;
 981         u_short ip6f_mf = 0;
 982
 983         /*
 984          * Local copies of addresses. They are only valid if we have
 985          * an IP packet.
 986          *
 987          * proto        The protocol. Set to 0 for non-ip packets,
 988          *      or to the protocol read from the packet otherwise.
 989          *      proto != 0 means that we have an IPv4 packet.
 990          *
 991          * src_port, dst_port   port numbers, in HOST format. Only
 992          *      valid for TCP and UDP packets.
 993          *
 994          * src_ip, dst_ip       ip addresses, in NETWORK format.
 995          *      Only valid for IPv4 packets.
 996          */
 997         uint8_t proto;
 998         uint16_t src_port = 0, dst_port = 0;    /* NOTE: host format    */
 999         struct in_addr src_ip, dst_ip;          /* NOTE: network format */
1000         uint16_t iplen=0;
1001         int pktlen;
1002         uint16_t        etype = 0;      /* Host order stored ether type */
1003
1004         /*
1005          * dyn_dir = MATCH_UNKNOWN when rules unchecked,
1006          *      MATCH_NONE when checked and not matched (q = NULL),
1007          *      MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
1008          */
1009         int dyn_dir = MATCH_UNKNOWN;
1010         ipfw_dyn_rule *q = NULL;
1011         struct ip_fw_chain *chain = &V_layer3_chain;
1012
1013         /*
1014          * We store in ulp a pointer to the upper layer protocol header.
1015          * In the ipv4 case this is easy to determine from the header,
1016          * but for ipv6 we might have some additional headers in the middle.
1017          * ulp is NULL if not found.
1018          */
1019         void *ulp = NULL;               /* upper layer protocol pointer. */
1020
1021         /* XXX ipv6 variables */
1022         int is_ipv6 = 0;
1023         uint8_t icmp6_type = 0;
1024         uint16_t ext_hd = 0;    /* bits vector for extension header filtering */
1025         /* end of ipv6 variables */
1026
1027         int is_ipv4 = 0;
1028
1029         int done = 0;           /* flag to exit the outer loop */
1030         IPFW_RLOCK_TRACKER;
1031
1032         if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
1033                 return (IP_FW_PASS);    /* accept */
1034
1035         dst_ip.s_addr = 0;              /* make sure it is initialized */
1036         src_ip.s_addr = 0;              /* make sure it is initialized */
1037         pktlen = m->m_pkthdr.len;
1038         args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
1039         proto = args->f_id.proto = 0;   /* mark f_id invalid */
1040                 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
1041
1042 /*
1043  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1044  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1045  * pointer might become stale after other pullups (but we never use it
1046  * this way).
1047  */
1048 #define PULLUP_TO(_len, p, T)   PULLUP_LEN(_len, p, sizeof(T))
1049 #define PULLUP_LEN(_len, p, T)                                  \
1050 do {                                                            \
1051         int x = (_len) + T;                                     \
1052         if ((m)->m_len < x) {                                   \
1053                 args->m = m = m_pullup(m, x);                   \
1054                 if (m == NULL)                                  \
1055                         goto pullup_failed;                     \
1056         }                                                       \
1057         p = (mtod(m, char *) + (_len));                         \
1058 } while (0)
1059
1060         /*
1061          * if we have an ether header,
1062          */
1063         if (args->eh)
1064                 etype = ntohs(args->eh->ether_type);
1065
1066         /* Identify IP packets and fill up variables. */
1067         if (pktlen >= sizeof(struct ip6_hdr) &&
1068             (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
1069                 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1070                 is_ipv6 = 1;
1071                 args->f_id.addr_type = 6;
1072                 hlen = sizeof(struct ip6_hdr);
1073                 proto = ip6->ip6_nxt;
1074
1075                 /* Search extension headers to find upper layer protocols */
1076                 while (ulp == NULL && offset == 0) {
1077                         switch (proto) {
1078                         case IPPROTO_ICMPV6:
1079                                 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1080                                 icmp6_type = ICMP6(ulp)->icmp6_type;
1081                                 break;
1082
1083                         case IPPROTO_TCP:
1084                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1085                                 dst_port = TCP(ulp)->th_dport;
1086                                 src_port = TCP(ulp)->th_sport;
1087                                 /* save flags for dynamic rules */
1088                                 args->f_id._flags = TCP(ulp)->th_flags;
1089                                 break;
1090
1091                         case IPPROTO_SCTP:
1092                                 PULLUP_TO(hlen, ulp, struct sctphdr);
1093                                 src_port = SCTP(ulp)->src_port;
1094                                 dst_port = SCTP(ulp)->dest_port;
1095                                 break;
1096
1097                         case IPPROTO_UDP:
1098                                 PULLUP_TO(hlen, ulp, struct udphdr);
1099                                 dst_port = UDP(ulp)->uh_dport;
1100                                 src_port = UDP(ulp)->uh_sport;
1101                                 break;
1102
1103                         case IPPROTO_HOPOPTS:   /* RFC 2460 */
1104                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1105                                 ext_hd |= EXT_HOPOPTS;
1106                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1107                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1108                                 ulp = NULL;
1109                                 break;
1110
1111                         case IPPROTO_ROUTING:   /* RFC 2460 */
1112                                 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1113                                 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1114                                 case 0:
1115                                         ext_hd |= EXT_RTHDR0;
1116                                         break;
1117                                 case 2:
1118                                         ext_hd |= EXT_RTHDR2;
1119                                         break;
1120                                 default:
1121                                         if (V_fw_verbose)
1122                                                 printf("IPFW2: IPV6 - Unknown "
1123                                                     "Routing Header type(%d)\n",
1124                                                     ((struct ip6_rthdr *)
1125                                                     ulp)->ip6r_type);
1126                                         if (V_fw_deny_unknown_exthdrs)
1127                                             return (IP_FW_DENY);
1128                                         break;
1129                                 }
1130                                 ext_hd |= EXT_ROUTING;
1131                                 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1132                                 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1133                                 ulp = NULL;
1134                                 break;
1135
1136                         case IPPROTO_FRAGMENT:  /* RFC 2460 */
1137                                 PULLUP_TO(hlen, ulp, struct ip6_frag);
1138                                 ext_hd |= EXT_FRAGMENT;
1139                                 hlen += sizeof (struct ip6_frag);
1140                                 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1141                                 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1142                                         IP6F_OFF_MASK;
1143                                 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1144                                         IP6F_MORE_FRAG;
1145                                 if (V_fw_permit_single_frag6 == 0 &&
1146                                     offset == 0 && ip6f_mf == 0) {
1147                                         if (V_fw_verbose)
1148                                                 printf("IPFW2: IPV6 - Invalid "
1149                                                     "Fragment Header\n");
1150                                         if (V_fw_deny_unknown_exthdrs)
1151                                             return (IP_FW_DENY);
1152                                         break;
1153                                 }
1154                                 args->f_id.extra =
1155                                     ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1156                                 ulp = NULL;
1157                                 break;
1158
1159                         case IPPROTO_DSTOPTS:   /* RFC 2460 */
1160                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1161                                 ext_hd |= EXT_DSTOPTS;
1162                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1163                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1164                                 ulp = NULL;
1165                                 break;
1166
1167                         case IPPROTO_AH:        /* RFC 2402 */
1168                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1169                                 ext_hd |= EXT_AH;
1170                                 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1171                                 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1172                                 ulp = NULL;
1173                                 break;
1174
1175                         case IPPROTO_ESP:       /* RFC 2406 */
1176                                 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1177                                 /* Anything past Seq# is variable length and
1178                                  * data past this ext. header is encrypted. */
1179                                 ext_hd |= EXT_ESP;
1180                                 break;
1181
1182                         case IPPROTO_NONE:      /* RFC 2460 */
1183                                 /*
1184                                  * Packet ends here, and IPv6 header has
1185                                  * already been pulled up. If ip6e_len!=0
1186                                  * then octets must be ignored.
1187                                  */
1188                                 ulp = ip; /* non-NULL to get out of loop. */
1189                                 break;
1190
1191                         case IPPROTO_OSPFIGP:
1192                                 /* XXX OSPF header check? */
1193                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1194                                 break;
1195
1196                         case IPPROTO_PIM:
1197                                 /* XXX PIM header check? */
1198                                 PULLUP_TO(hlen, ulp, struct pim);
1199                                 break;
1200
1201                         case IPPROTO_CARP:
1202                                 PULLUP_TO(hlen, ulp, struct carp_header);
1203                                 if (((struct carp_header *)ulp)->carp_version !=
1204                                     CARP_VERSION)
1205                                         return (IP_FW_DENY);
1206                                 if (((struct carp_header *)ulp)->carp_type !=
1207                                     CARP_ADVERTISEMENT)
1208                                         return (IP_FW_DENY);
1209                                 break;
1210
1211                         case IPPROTO_IPV6:      /* RFC 2893 */
1212                                 PULLUP_TO(hlen, ulp, struct ip6_hdr);
1213                                 break;
1214
1215                         case IPPROTO_IPV4:      /* RFC 2893 */
1216                                 PULLUP_TO(hlen, ulp, struct ip);
1217                                 break;
1218
1219                         default:
1220                                 if (V_fw_verbose)
1221                                         printf("IPFW2: IPV6 - Unknown "
1222                                             "Extension Header(%d), ext_hd=%x\n",
1223                                              proto, ext_hd);
1224                                 if (V_fw_deny_unknown_exthdrs)
1225                                     return (IP_FW_DENY);
1226                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1227                                 break;
1228                         } /*switch */
1229                 }
1230                 ip = mtod(m, struct ip *);
1231                 ip6 = (struct ip6_hdr *)ip;
1232                 args->f_id.src_ip6 = ip6->ip6_src;
1233                 args->f_id.dst_ip6 = ip6->ip6_dst;
1234                 args->f_id.src_ip = 0;
1235                 args->f_id.dst_ip = 0;
1236                 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1237         } else if (pktlen >= sizeof(struct ip) &&
1238             (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1239                 is_ipv4 = 1;
1240                 hlen = ip->ip_hl << 2;
1241                 args->f_id.addr_type = 4;
1242
1243                 /*
1244                  * Collect parameters into local variables for faster matching.
1245                  */
1246                 proto = ip->ip_p;
1247                 src_ip = ip->ip_src;
1248                 dst_ip = ip->ip_dst;
1249                 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1250                 iplen = ntohs(ip->ip_len);
1251                 pktlen = iplen < pktlen ? iplen : pktlen;
1252
1253                 if (offset == 0) {
1254                         switch (proto) {
1255                         case IPPROTO_TCP:
1256                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1257                                 dst_port = TCP(ulp)->th_dport;
1258                                 src_port = TCP(ulp)->th_sport;
1259                                 /* save flags for dynamic rules */
1260                                 args->f_id._flags = TCP(ulp)->th_flags;
1261                                 break;
1262
1263                         case IPPROTO_SCTP:
1264                                 PULLUP_TO(hlen, ulp, struct sctphdr);
1265                                 src_port = SCTP(ulp)->src_port;
1266                                 dst_port = SCTP(ulp)->dest_port;
1267                                 break;
1268
1269                         case IPPROTO_UDP:
1270                                 PULLUP_TO(hlen, ulp, struct udphdr);
1271                                 dst_port = UDP(ulp)->uh_dport;
1272                                 src_port = UDP(ulp)->uh_sport;
1273                                 break;
1274
1275                         case IPPROTO_ICMP:
1276                                 PULLUP_TO(hlen, ulp, struct icmphdr);
1277                                 //args->f_id.flags = ICMP(ulp)->icmp_type;
1278                                 break;
1279
1280                         default:
1281                                 break;
1282                         }
1283                 }
1284
1285                 ip = mtod(m, struct ip *);
1286                 args->f_id.src_ip = ntohl(src_ip.s_addr);
1287                 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1288         }
1289 #undef PULLUP_TO
1290         if (proto) { /* we may have port numbers, store them */
1291                 args->f_id.proto = proto;
1292                 args->f_id.src_port = src_port = ntohs(src_port);
1293                 args->f_id.dst_port = dst_port = ntohs(dst_port);
1294         }
1295
1296         IPFW_PF_RLOCK(chain);
1297         if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1298                 IPFW_PF_RUNLOCK(chain);
1299                 return (IP_FW_PASS);    /* accept */
1300         }
1301         if (args->rule.slot) {
1302                 /*
1303                  * Packet has already been tagged as a result of a previous
1304                  * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1305                  * REASS, NETGRAPH, DIVERT/TEE...)
1306                  * Validate the slot and continue from the next one
1307                  * if still present, otherwise do a lookup.
1308                  */
1309                 f_pos = (args->rule.chain_id == chain->id) ?
1310                     args->rule.slot :
1311                     ipfw_find_rule(chain, args->rule.rulenum,
1312                         args->rule.rule_id);
1313         } else {
1314                 f_pos = 0;
1315         }
1316
1317         /*
1318          * Now scan the rules, and parse microinstructions for each rule.
1319          * We have two nested loops and an inner switch. Sometimes we
1320          * need to break out of one or both loops, or re-enter one of
1321          * the loops with updated variables. Loop variables are:
1322          *
1323          *      f_pos (outer loop) points to the current rule.
1324          *              On output it points to the matching rule.
1325          *      done (outer loop) is used as a flag to break the loop.
1326          *      l (inner loop)  residual length of current rule.
1327          *              cmd points to the current microinstruction.
1328          *
1329          * We break the inner loop by setting l=0 and possibly
1330          * cmdlen=0 if we don't want to advance cmd.
1331          * We break the outer loop by setting done=1
1332          * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1333          * as needed.
1334          */
1335         for (; f_pos < chain->n_rules; f_pos++) {
1336                 ipfw_insn *cmd;
1337                 uint32_t tablearg = 0;
1338                 int l, cmdlen, skip_or; /* skip rest of OR block */
1339                 struct ip_fw *f;
1340
1341                 f = chain->map[f_pos];
1342                 if (V_set_disable & (1 << f->set) )
1343                         continue;
1344
1345                 skip_or = 0;
1346                 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1347                     l -= cmdlen, cmd += cmdlen) {
1348                         int match;
1349
1350                         /*
1351                          * check_body is a jump target used when we find a
1352                          * CHECK_STATE, and need to jump to the body of
1353                          * the target rule.
1354                          */
1355
1356 /* check_body: */
1357                         cmdlen = F_LEN(cmd);
1358                         /*
1359                          * An OR block (insn_1 || .. || insn_n) has the
1360                          * F_OR bit set in all but the last instruction.
1361                          * The first match will set "skip_or", and cause
1362                          * the following instructions to be skipped until
1363                          * past the one with the F_OR bit clear.
1364                          */
1365                         if (skip_or) {          /* skip this instruction */
1366                                 if ((cmd->len & F_OR) == 0)
1367                                         skip_or = 0;    /* next one is good */
1368                                 continue;
1369                         }
1370                         match = 0; /* set to 1 if we succeed */
1371
1372                         switch (cmd->opcode) {
1373                         /*
1374                          * The first set of opcodes compares the packet's
1375                          * fields with some pattern, setting 'match' if a
1376                          * match is found. At the end of the loop there is
1377                          * logic to deal with F_NOT and F_OR flags associated
1378                          * with the opcode.
1379                          */
1380                         case O_NOP:
1381                                 match = 1;
1382                                 break;
1383
1384                         case O_FORWARD_MAC:
1385                                 printf("ipfw: opcode %d unimplemented\n",
1386                                     cmd->opcode);
1387                                 break;
1388
1389                         case O_GID:
1390                         case O_UID:
1391                         case O_JAIL:
1392                                 /*
1393                                  * We only check offset == 0 && proto != 0,
1394                                  * as this ensures that we have a
1395                                  * packet with the ports info.
1396                                  */
1397                                 if (offset != 0)
1398                                         break;
1399                                 if (proto == IPPROTO_TCP ||
1400                                     proto == IPPROTO_UDP)
1401                                         match = check_uidgid(
1402                                                     (ipfw_insn_u32 *)cmd,
1403                                                     args, &ucred_lookup,
1404 #ifdef __FreeBSD__
1405                                                     &ucred_cache);
1406 #else
1407                                                     (void *)&ucred_cache);
1408 #endif
1409                                 break;
1410
1411                         case O_RECV:
1412                                 match = iface_match(m->m_pkthdr.rcvif,
1413                                     (ipfw_insn_if *)cmd, chain, &tablearg);
1414                                 break;
1415
1416                         case O_XMIT:
1417                                 match = iface_match(oif, (ipfw_insn_if *)cmd,
1418                                     chain, &tablearg);
1419                                 break;
1420
1421                         case O_VIA:
1422                                 match = iface_match(oif ? oif :
1423                                     m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1424                                     chain, &tablearg);
1425                                 break;
1426
1427                         case O_MACADDR2:
1428                                 if (args->eh != NULL) { /* have MAC header */
1429                                         u_int32_t *want = (u_int32_t *)
1430                                                 ((ipfw_insn_mac *)cmd)->addr;
1431                                         u_int32_t *mask = (u_int32_t *)
1432                                                 ((ipfw_insn_mac *)cmd)->mask;
1433                                         u_int32_t *hdr = (u_int32_t *)args->eh;
1434
1435                                         match =
1436                                             ( want[0] == (hdr[0] & mask[0]) &&
1437                                               want[1] == (hdr[1] & mask[1]) &&
1438                                               want[2] == (hdr[2] & mask[2]) );
1439                                 }
1440                                 break;
1441
1442                         case O_MAC_TYPE:
1443                                 if (args->eh != NULL) {
1444                                         u_int16_t *p =
1445                                             ((ipfw_insn_u16 *)cmd)->ports;
1446                                         int i;
1447
1448                                         for (i = cmdlen - 1; !match && i>0;
1449                                             i--, p += 2)
1450                                                 match = (etype >= p[0] &&
1451                                                     etype <= p[1]);
1452                                 }
1453                                 break;
1454
1455                         case O_FRAG:
1456                                 match = (offset != 0);
1457                                 break;
1458
1459                         case O_IN:      /* "out" is "not in" */
1460                                 match = (oif == NULL);
1461                                 break;
1462
1463                         case O_LAYER2:
1464                                 match = (args->eh != NULL);
1465                                 break;
1466
1467                         case O_DIVERTED:
1468                             {
1469                                 /* For diverted packets, args->rule.info
1470                                  * contains the divert port (in host format)
1471                                  * reason and direction.
1472                                  */
1473                                 uint32_t i = args->rule.info;
1474                                 match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1475                                     cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1476                             }
1477                                 break;
1478
1479                         case O_PROTO:
1480                                 /*
1481                                  * We do not allow an arg of 0 so the
1482                                  * check of "proto" only suffices.
1483                                  */
1484                                 match = (proto == cmd->arg1);
1485                                 break;
1486
1487                         case O_IP_SRC:
1488                                 match = is_ipv4 &&
1489                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1490                                     src_ip.s_addr);
1491                                 break;
1492
1493                         case O_IP_SRC_LOOKUP:
1494                         case O_IP_DST_LOOKUP:
1495                                 if (is_ipv4) {
1496                                     uint32_t key =
1497                                         (cmd->opcode == O_IP_DST_LOOKUP) ?
1498                                             dst_ip.s_addr : src_ip.s_addr;
1499                                     uint32_t v = 0;
1500
1501                                     if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1502                                         /* generic lookup. The key must be
1503                                          * in 32bit big-endian format.
1504                                          */
1505                                         v = ((ipfw_insn_u32 *)cmd)->d[1];
1506                                         if (v == 0)
1507                                             key = dst_ip.s_addr;
1508                                         else if (v == 1)
1509                                             key = src_ip.s_addr;
1510                                         else if (v == 6) /* dscp */
1511                                             key = (ip->ip_tos >> 2) & 0x3f;
1512                                         else if (offset != 0)
1513                                             break;
1514                                         else if (proto != IPPROTO_TCP &&
1515                                                 proto != IPPROTO_UDP)
1516                                             break;
1517                                         else if (v == 2)
1518                                             key = dst_port;
1519                                         else if (v == 3)
1520                                             key = src_port;
1521 #ifndef USERSPACE
1522                                         else if (v == 4 || v == 5) {
1523                                             check_uidgid(
1524                                                 (ipfw_insn_u32 *)cmd,
1525                                                 args, &ucred_lookup,
1526 #ifdef __FreeBSD__
1527                                                 &ucred_cache);
1528                                             if (v == 4 /* O_UID */)
1529                                                 key = ucred_cache->cr_uid;
1530                                             else if (v == 5 /* O_JAIL */)
1531                                                 key = ucred_cache->cr_prison->pr_id;
1532 #else /* !__FreeBSD__ */
1533                                                 (void *)&ucred_cache);
1534                                             if (v ==4 /* O_UID */)
1535                                                 key = ucred_cache.uid;
1536                                             else if (v == 5 /* O_JAIL */)
1537                                                 key = ucred_cache.xid;
1538 #endif /* !__FreeBSD__ */
1539                                         }
1540 #endif /* !USERSPACE */
1541                                         else
1542                                             break;
1543                                     }
1544                                     match = ipfw_lookup_table(chain,
1545                                         cmd->arg1, key, &v);
1546                                     if (!match)
1547                                         break;
1548                                     if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1549                                         match =
1550                                             ((ipfw_insn_u32 *)cmd)->d[0] == v;
1551                                     else
1552                                         tablearg = v;
1553                                 } else if (is_ipv6) {
1554                                         uint32_t v = 0;
1555                                         void *pkey = (cmd->opcode == O_IP_DST_LOOKUP) ?
1556                                                 &args->f_id.dst_ip6: &args->f_id.src_ip6;
1557                                         match = ipfw_lookup_table_extended(chain,
1558                                                         cmd->arg1,
1559                                                         sizeof(struct in6_addr),
1560                                                         pkey, &v);
1561                                         if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1562                                                 match = ((ipfw_insn_u32 *)cmd)->d[0] == v;
1563                                         if (match)
1564                                                 tablearg = v;
1565                                 }
1566                                 break;
1567
1568                         case O_IP_FLOW_LOOKUP:
1569                                 {
1570                                         uint32_t v = 0;
1571                                         match = ipfw_lookup_table_extended(chain,
1572                                             cmd->arg1, 0, &args->f_id, &v);
1573                                         if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1574                                                 match = ((ipfw_insn_u32 *)cmd)->d[0] == v;
1575                                         if (match)
1576                                                 tablearg = v;
1577                                 }
1578                                 break;
1579                         case O_IP_SRC_MASK:
1580                         case O_IP_DST_MASK:
1581                                 if (is_ipv4) {
1582                                     uint32_t a =
1583                                         (cmd->opcode == O_IP_DST_MASK) ?
1584                                             dst_ip.s_addr : src_ip.s_addr;
1585                                     uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
1586                                     int i = cmdlen-1;
1587
1588                                     for (; !match && i>0; i-= 2, p+= 2)
1589                                         match = (p[0] == (a & p[1]));
1590                                 }
1591                                 break;
1592
1593                         case O_IP_SRC_ME:
1594                                 if (is_ipv4) {
1595                                         struct ifnet *tif;
1596
1597                                         INADDR_TO_IFP(src_ip, tif);
1598                                         match = (tif != NULL);
1599                                         break;
1600                                 }
1601 #ifdef INET6
1602                                 /* FALLTHROUGH */
1603                         case O_IP6_SRC_ME:
1604                                 match= is_ipv6 && ipfw_localip6(&args->f_id.src_ip6);
1605 #endif
1606                                 break;
1607
1608                         case O_IP_DST_SET:
1609                         case O_IP_SRC_SET:
1610                                 if (is_ipv4) {
1611                                         u_int32_t *d = (u_int32_t *)(cmd+1);
1612                                         u_int32_t addr =
1613                                             cmd->opcode == O_IP_DST_SET ?
1614                                                 args->f_id.dst_ip :
1615                                                 args->f_id.src_ip;
1616
1617                                             if (addr < d[0])
1618                                                     break;
1619                                             addr -= d[0]; /* subtract base */
1620                                             match = (addr < cmd->arg1) &&
1621                                                 ( d[ 1 + (addr>>5)] &
1622                                                   (1<<(addr & 0x1f)) );
1623                                 }
1624                                 break;
1625
1626                         case O_IP_DST:
1627                                 match = is_ipv4 &&
1628                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1629                                     dst_ip.s_addr);
1630                                 break;
1631
1632                         case O_IP_DST_ME:
1633                                 if (is_ipv4) {
1634                                         struct ifnet *tif;
1635
1636                                         INADDR_TO_IFP(dst_ip, tif);
1637                                         match = (tif != NULL);
1638                                         break;
1639                                 }
1640 #ifdef INET6
1641                                 /* FALLTHROUGH */
1642                         case O_IP6_DST_ME:
1643                                 match= is_ipv6 && ipfw_localip6(&args->f_id.dst_ip6);
1644 #endif
1645                                 break;
1646
1647
1648                         case O_IP_SRCPORT:
1649                         case O_IP_DSTPORT:
1650                                 /*
1651                                  * offset == 0 && proto != 0 is enough
1652                                  * to guarantee that we have a
1653                                  * packet with port info.
1654                                  */
1655                                 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1656                                     && offset == 0) {
1657                                         u_int16_t x =
1658                                             (cmd->opcode == O_IP_SRCPORT) ?
1659                                                 src_port : dst_port ;
1660                                         u_int16_t *p =
1661                                             ((ipfw_insn_u16 *)cmd)->ports;
1662                                         int i;
1663
1664                                         for (i = cmdlen - 1; !match && i>0;
1665                                             i--, p += 2)
1666                                                 match = (x>=p[0] && x<=p[1]);
1667                                 }
1668                                 break;
1669
1670                         case O_ICMPTYPE:
1671                                 match = (offset == 0 && proto==IPPROTO_ICMP &&
1672                                     icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
1673                                 break;
1674
1675 #ifdef INET6
1676                         case O_ICMP6TYPE:
1677                                 match = is_ipv6 && offset == 0 &&
1678                                     proto==IPPROTO_ICMPV6 &&
1679                                     icmp6type_match(
1680                                         ICMP6(ulp)->icmp6_type,
1681                                         (ipfw_insn_u32 *)cmd);
1682                                 break;
1683 #endif /* INET6 */
1684
1685                         case O_IPOPT:
1686                                 match = (is_ipv4 &&
1687                                     ipopts_match(ip, cmd) );
1688                                 break;
1689
1690                         case O_IPVER:
1691                                 match = (is_ipv4 &&
1692                                     cmd->arg1 == ip->ip_v);
1693                                 break;
1694
1695                         case O_IPID:
1696                         case O_IPLEN:
1697                         case O_IPTTL:
1698                                 if (is_ipv4) {  /* only for IP packets */
1699                                     uint16_t x;
1700                                     uint16_t *p;
1701                                     int i;
1702
1703                                     if (cmd->opcode == O_IPLEN)
1704                                         x = iplen;
1705                                     else if (cmd->opcode == O_IPTTL)
1706                                         x = ip->ip_ttl;
1707                                     else /* must be IPID */
1708                                         x = ntohs(ip->ip_id);
1709                                     if (cmdlen == 1) {
1710                                         match = (cmd->arg1 == x);
1711                                         break;
1712                                     }
1713                                     /* otherwise we have ranges */
1714                                     p = ((ipfw_insn_u16 *)cmd)->ports;
1715                                     i = cmdlen - 1;
1716                                     for (; !match && i>0; i--, p += 2)
1717                                         match = (x >= p[0] && x <= p[1]);
1718                                 }
1719                                 break;
1720
1721                         case O_IPPRECEDENCE:
1722                                 match = (is_ipv4 &&
1723                                     (cmd->arg1 == (ip->ip_tos & 0xe0)) );
1724                                 break;
1725
1726                         case O_IPTOS:
1727                                 match = (is_ipv4 &&
1728                                     flags_match(cmd, ip->ip_tos));
1729                                 break;
1730
1731                         case O_DSCP:
1732                             {
1733                                 uint32_t *p;
1734                                 uint16_t x;
1735
1736                                 p = ((ipfw_insn_u32 *)cmd)->d;
1737
1738                                 if (is_ipv4)
1739                                         x = ip->ip_tos >> 2;
1740                                 else if (is_ipv6) {
1741                                         uint8_t *v;
1742                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
1743                                         x = (*v & 0x0F) << 2;
1744                                         v++;
1745                                         x |= *v >> 6;
1746                                 } else
1747                                         break;
1748
1749                                 /* DSCP bitmask is stored as low_u32 high_u32 */
1750                                 if (x > 32)
1751                                         match = *(p + 1) & (1 << (x - 32));
1752                                 else
1753                                         match = *p & (1 << x);
1754                             }
1755                                 break;
1756
1757                         case O_TCPDATALEN:
1758                                 if (proto == IPPROTO_TCP && offset == 0) {
1759                                     struct tcphdr *tcp;
1760                                     uint16_t x;
1761                                     uint16_t *p;
1762                                     int i;
1763
1764                                     tcp = TCP(ulp);
1765                                     x = iplen -
1766                                         ((ip->ip_hl + tcp->th_off) << 2);
1767                                     if (cmdlen == 1) {
1768                                         match = (cmd->arg1 == x);
1769                                         break;
1770                                     }
1771                                     /* otherwise we have ranges */
1772                                     p = ((ipfw_insn_u16 *)cmd)->ports;
1773                                     i = cmdlen - 1;
1774                                     for (; !match && i>0; i--, p += 2)
1775                                         match = (x >= p[0] && x <= p[1]);
1776                                 }
1777                                 break;
1778
1779                         case O_TCPFLAGS:
1780                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1781                                     flags_match(cmd, TCP(ulp)->th_flags));
1782                                 break;
1783
1784                         case O_TCPOPTS:
1785                                 if (proto == IPPROTO_TCP && offset == 0 && ulp){
1786                                         PULLUP_LEN(hlen, ulp,
1787                                             (TCP(ulp)->th_off << 2));
1788                                         match = tcpopts_match(TCP(ulp), cmd);
1789                                 }
1790                                 break;
1791
1792                         case O_TCPSEQ:
1793                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1794                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
1795                                         TCP(ulp)->th_seq);
1796                                 break;
1797
1798                         case O_TCPACK:
1799                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1800                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
1801                                         TCP(ulp)->th_ack);
1802                                 break;
1803
1804                         case O_TCPWIN:
1805                                 if (proto == IPPROTO_TCP && offset == 0) {
1806                                     uint16_t x;
1807                                     uint16_t *p;
1808                                     int i;
1809
1810                                     x = ntohs(TCP(ulp)->th_win);
1811                                     if (cmdlen == 1) {
1812                                         match = (cmd->arg1 == x);
1813                                         break;
1814                                     }
1815                                     /* Otherwise we have ranges. */
1816                                     p = ((ipfw_insn_u16 *)cmd)->ports;
1817                                     i = cmdlen - 1;
1818                                     for (; !match && i > 0; i--, p += 2)
1819                                         match = (x >= p[0] && x <= p[1]);
1820                                 }
1821                                 break;
1822
1823                         case O_ESTAB:
1824                                 /* reject packets which have SYN only */
1825                                 /* XXX should i also check for TH_ACK ? */
1826                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1827                                     (TCP(ulp)->th_flags &
1828                                      (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1829                                 break;
1830
1831                         case O_ALTQ: {
1832                                 struct pf_mtag *at;
1833                                 struct m_tag *mtag;
1834                                 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
1835
1836                                 /*
1837                                  * ALTQ uses mbuf tags from another
1838                                  * packet filtering system - pf(4).
1839                                  * We allocate a tag in its format
1840                                  * and fill it in, pretending to be pf(4).
1841                                  */
1842                                 match = 1;
1843                                 at = pf_find_mtag(m);
1844                                 if (at != NULL && at->qid != 0)
1845                                         break;
1846                                 mtag = m_tag_get(PACKET_TAG_PF,
1847                                     sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
1848                                 if (mtag == NULL) {
1849                                         /*
1850                                          * Let the packet fall back to the
1851                                          * default ALTQ.
1852                                          */
1853                                         break;
1854                                 }
1855                                 m_tag_prepend(m, mtag);
1856                                 at = (struct pf_mtag *)(mtag + 1);
1857                                 at->qid = altq->qid;
1858                                 at->hdr = ip;
1859                                 break;
1860                         }
1861
1862                         case O_LOG:
1863                                 ipfw_log(chain, f, hlen, args, m,
1864                                     oif, offset | ip6f_mf, tablearg, ip);
1865                                 match = 1;
1866                                 break;
1867
1868                         case O_PROB:
1869                                 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
1870                                 break;
1871
1872                         case O_VERREVPATH:
1873                                 /* Outgoing packets automatically pass/match */
1874                                 match = ((oif != NULL) ||
1875                                     (m->m_pkthdr.rcvif == NULL) ||
1876                                     (
1877 #ifdef INET6
1878                                     is_ipv6 ?
1879                                         verify_path6(&(args->f_id.src_ip6),
1880                                             m->m_pkthdr.rcvif, args->f_id.fib) :
1881 #endif
1882                                     verify_path(src_ip, m->m_pkthdr.rcvif,
1883                                         args->f_id.fib)));
1884                                 break;
1885
1886                         case O_VERSRCREACH:
1887                                 /* Outgoing packets automatically pass/match */
1888                                 match = (hlen > 0 && ((oif != NULL) ||
1889 #ifdef INET6
1890                                     is_ipv6 ?
1891                                         verify_path6(&(args->f_id.src_ip6),
1892                                             NULL, args->f_id.fib) :
1893 #endif
1894                                     verify_path(src_ip, NULL, args->f_id.fib)));
1895                                 break;
1896
1897                         case O_ANTISPOOF:
1898                                 /* Outgoing packets automatically pass/match */
1899                                 if (oif == NULL && hlen > 0 &&
1900                                     (  (is_ipv4 && in_localaddr(src_ip))
1901 #ifdef INET6
1902                                     || (is_ipv6 &&
1903                                         in6_localaddr(&(args->f_id.src_ip6)))
1904 #endif
1905                                     ))
1906                                         match =
1907 #ifdef INET6
1908                                             is_ipv6 ? verify_path6(
1909                                                 &(args->f_id.src_ip6),
1910                                                 m->m_pkthdr.rcvif,
1911                                                 args->f_id.fib) :
1912 #endif
1913                                             verify_path(src_ip,
1914                                                 m->m_pkthdr.rcvif,
1915                                                 args->f_id.fib);
1916                                 else
1917                                         match = 1;
1918                                 break;
1919
1920                         case O_IPSEC:
1921 #ifdef IPSEC
1922                                 match = (m_tag_find(m,
1923                                     PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
1924 #endif
1925                                 /* otherwise no match */
1926                                 break;
1927
1928 #ifdef INET6
1929                         case O_IP6_SRC:
1930                                 match = is_ipv6 &&
1931                                     IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
1932                                     &((ipfw_insn_ip6 *)cmd)->addr6);
1933                                 break;
1934
1935                         case O_IP6_DST:
1936                                 match = is_ipv6 &&
1937                                 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
1938                                     &((ipfw_insn_ip6 *)cmd)->addr6);
1939                                 break;
1940                         case O_IP6_SRC_MASK:
1941                         case O_IP6_DST_MASK:
1942                                 if (is_ipv6) {
1943                                         int i = cmdlen - 1;
1944                                         struct in6_addr p;
1945                                         struct in6_addr *d =
1946                                             &((ipfw_insn_ip6 *)cmd)->addr6;
1947
1948                                         for (; !match && i > 0; d += 2,
1949                                             i -= F_INSN_SIZE(struct in6_addr)
1950                                             * 2) {
1951                                                 p = (cmd->opcode ==
1952                                                     O_IP6_SRC_MASK) ?
1953                                                     args->f_id.src_ip6:
1954                                                     args->f_id.dst_ip6;
1955                                                 APPLY_MASK(&p, &d[1]);
1956                                                 match =
1957                                                     IN6_ARE_ADDR_EQUAL(&d[0],
1958                                                     &p);
1959                                         }
1960                                 }
1961                                 break;
1962
1963                         case O_FLOW6ID:
1964                                 match = is_ipv6 &&
1965                                     flow6id_match(args->f_id.flow_id6,
1966                                     (ipfw_insn_u32 *) cmd);
1967                                 break;
1968
1969                         case O_EXT_HDR:
1970                                 match = is_ipv6 &&
1971                                     (ext_hd & ((ipfw_insn *) cmd)->arg1);
1972                                 break;
1973
1974                         case O_IP6:
1975                                 match = is_ipv6;
1976                                 break;
1977 #endif
1978
1979                         case O_IP4:
1980                                 match = is_ipv4;
1981                                 break;
1982
1983                         case O_TAG: {
1984                                 struct m_tag *mtag;
1985                                 uint32_t tag = TARG(cmd->arg1, tag);
1986
1987                                 /* Packet is already tagged with this tag? */
1988                                 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
1989
1990                                 /* We have `untag' action when F_NOT flag is
1991                                  * present. And we must remove this mtag from
1992                                  * mbuf and reset `match' to zero (`match' will
1993                                  * be inversed later).
1994                                  * Otherwise we should allocate new mtag and
1995                                  * push it into mbuf.
1996                                  */
1997                                 if (cmd->len & F_NOT) { /* `untag' action */
1998                                         if (mtag != NULL)
1999                                                 m_tag_delete(m, mtag);
2000                                         match = 0;
2001                                 } else {
2002                                         if (mtag == NULL) {
2003                                                 mtag = m_tag_alloc( MTAG_IPFW,
2004                                                     tag, 0, M_NOWAIT);
2005                                                 if (mtag != NULL)
2006                                                         m_tag_prepend(m, mtag);
2007                                         }
2008                                         match = 1;
2009                                 }
2010                                 break;
2011                         }
2012
2013                         case O_FIB: /* try match the specified fib */
2014                                 if (args->f_id.fib == cmd->arg1)
2015                                         match = 1;
2016                                 break;
2017
2018                         case O_SOCKARG: {
2019 #ifndef USERSPACE       /* not supported in userspace */
2020                                 struct inpcb *inp = args->inp;
2021                                 struct inpcbinfo *pi;
2022
2023                                 if (is_ipv6) /* XXX can we remove this ? */
2024                                         break;
2025
2026                                 if (proto == IPPROTO_TCP)
2027                                         pi = &V_tcbinfo;
2028                                 else if (proto == IPPROTO_UDP)
2029                                         pi = &V_udbinfo;
2030                                 else
2031                                         break;
2032
2033                                 /*
2034                                  * XXXRW: so_user_cookie should almost
2035                                  * certainly be inp_user_cookie?
2036                                  */
2037
2038                                 /* For incomming packet, lookup up the
2039                                 inpcb using the src/dest ip/port tuple */
2040                                 if (inp == NULL) {
2041                                         inp = in_pcblookup(pi,
2042                                                 src_ip, htons(src_port),
2043                                                 dst_ip, htons(dst_port),
2044                                                 INPLOOKUP_RLOCKPCB, NULL);
2045                                         if (inp != NULL) {
2046                                                 tablearg =
2047                                                     inp->inp_socket->so_user_cookie;
2048                                                 if (tablearg)
2049                                                         match = 1;
2050                                                 INP_RUNLOCK(inp);
2051                                         }
2052                                 } else {
2053                                         if (inp->inp_socket) {
2054                                                 tablearg =
2055                                                     inp->inp_socket->so_user_cookie;
2056                                                 if (tablearg)
2057                                                         match = 1;
2058                                         }
2059                                 }
2060 #endif /* !USERSPACE */
2061                                 break;
2062                         }
2063
2064                         case O_TAGGED: {
2065                                 struct m_tag *mtag;
2066                                 uint32_t tag = TARG(cmd->arg1, tag);
2067
2068                                 if (cmdlen == 1) {
2069                                         match = m_tag_locate(m, MTAG_IPFW,
2070                                             tag, NULL) != NULL;
2071                                         break;
2072                                 }
2073
2074                                 /* we have ranges */
2075                                 for (mtag = m_tag_first(m);
2076                                     mtag != NULL && !match;
2077                                     mtag = m_tag_next(m, mtag)) {
2078                                         uint16_t *p;
2079                                         int i;
2080
2081                                         if (mtag->m_tag_cookie != MTAG_IPFW)
2082                                                 continue;
2083
2084                                         p = ((ipfw_insn_u16 *)cmd)->ports;
2085                                         i = cmdlen - 1;
2086                                         for(; !match && i > 0; i--, p += 2)
2087                                                 match =
2088                                                     mtag->m_tag_id >= p[0] &&
2089                                                     mtag->m_tag_id <= p[1];
2090                                 }
2091                                 break;
2092                         }
2093
2094                         /*
2095                          * The second set of opcodes represents 'actions',
2096                          * i.e. the terminal part of a rule once the packet
2097                          * matches all previous patterns.
2098                          * Typically there is only one action for each rule,
2099                          * and the opcode is stored at the end of the rule
2100                          * (but there are exceptions -- see below).
2101                          *
2102                          * In general, here we set retval and terminate the
2103                          * outer loop (would be a 'break 3' in some language,
2104                          * but we need to set l=0, done=1)
2105                          *
2106                          * Exceptions:
2107                          * O_COUNT and O_SKIPTO actions:
2108                          *   instead of terminating, we jump to the next rule
2109                          *   (setting l=0), or to the SKIPTO target (setting
2110                          *   f/f_len, cmd and l as needed), respectively.
2111                          *
2112                          * O_TAG, O_LOG and O_ALTQ action parameters:
2113                          *   perform some action and set match = 1;
2114                          *
2115                          * O_LIMIT and O_KEEP_STATE: these opcodes are
2116                          *   not real 'actions', and are stored right
2117                          *   before the 'action' part of the rule.
2118                          *   These opcodes try to install an entry in the
2119                          *   state tables; if successful, we continue with
2120                          *   the next opcode (match=1; break;), otherwise
2121                          *   the packet must be dropped (set retval,
2122                          *   break loops with l=0, done=1)
2123                          *
2124                          * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2125                          *   cause a lookup of the state table, and a jump
2126                          *   to the 'action' part of the parent rule
2127                          *   if an entry is found, or
2128                          *   (CHECK_STATE only) a jump to the next rule if
2129                          *   the entry is not found.
2130                          *   The result of the lookup is cached so that
2131                          *   further instances of these opcodes become NOPs.
2132                          *   The jump to the next rule is done by setting
2133                          *   l=0, cmdlen=0.
2134                          */
2135                         case O_LIMIT:
2136                         case O_KEEP_STATE:
2137                                 if (ipfw_install_state(chain, f,
2138                                     (ipfw_insn_limit *)cmd, args, tablearg)) {
2139                                         /* error or limit violation */
2140                                         retval = IP_FW_DENY;
2141                                         l = 0;  /* exit inner loop */
2142                                         done = 1; /* exit outer loop */
2143                                 }
2144                                 match = 1;
2145                                 break;
2146
2147                         case O_PROBE_STATE:
2148                         case O_CHECK_STATE:
2149                                 /*
2150                                  * dynamic rules are checked at the first
2151                                  * keep-state or check-state occurrence,
2152                                  * with the result being stored in dyn_dir.
2153                                  * The compiler introduces a PROBE_STATE
2154                                  * instruction for us when we have a
2155                                  * KEEP_STATE (because PROBE_STATE needs
2156                                  * to be run first).
2157                                  */
2158                                 if (dyn_dir == MATCH_UNKNOWN &&
2159                                     (q = ipfw_lookup_dyn_rule(&args->f_id,
2160                                      &dyn_dir, proto == IPPROTO_TCP ?
2161                                         TCP(ulp) : NULL))
2162                                         != NULL) {
2163                                         /*
2164                                          * Found dynamic entry, update stats
2165                                          * and jump to the 'action' part of
2166                                          * the parent rule by setting
2167                                          * f, cmd, l and clearing cmdlen.
2168                                          */
2169                                         IPFW_INC_DYN_COUNTER(q, pktlen);
2170                                         /* XXX we would like to have f_pos
2171                                          * readily accessible in the dynamic
2172                                          * rule, instead of having to
2173                                          * lookup q->rule.
2174                                          */
2175                                         f = q->rule;
2176                                         f_pos = ipfw_find_rule(chain,
2177                                                 f->rulenum, f->id);
2178                                         cmd = ACTION_PTR(f);
2179                                         l = f->cmd_len - f->act_ofs;
2180                                         ipfw_dyn_unlock(q);
2181                                         cmdlen = 0;
2182                                         match = 1;
2183                                         break;
2184                                 }
2185                                 /*
2186                                  * Dynamic entry not found. If CHECK_STATE,
2187                                  * skip to next rule, if PROBE_STATE just
2188                                  * ignore and continue with next opcode.
2189                                  */
2190                                 if (cmd->opcode == O_CHECK_STATE)
2191                                         l = 0;  /* exit inner loop */
2192                                 match = 1;
2193                                 break;
2194
2195                         case O_ACCEPT:
2196                                 retval = 0;     /* accept */
2197                                 l = 0;          /* exit inner loop */
2198                                 done = 1;       /* exit outer loop */
2199                                 break;
2200
2201                         case O_PIPE:
2202                         case O_QUEUE:
2203                                 set_match(args, f_pos, chain);
2204                                 args->rule.info = TARG(cmd->arg1, pipe);
2205                                 if (cmd->opcode == O_PIPE)
2206                                         args->rule.info |= IPFW_IS_PIPE;
2207                                 if (V_fw_one_pass)
2208                                         args->rule.info |= IPFW_ONEPASS;
2209                                 retval = IP_FW_DUMMYNET;
2210                                 l = 0;          /* exit inner loop */
2211                                 done = 1;       /* exit outer loop */
2212                                 break;
2213
2214                         case O_DIVERT:
2215                         case O_TEE:
2216                                 if (args->eh) /* not on layer 2 */
2217                                     break;
2218                                 /* otherwise this is terminal */
2219                                 l = 0;          /* exit inner loop */
2220                                 done = 1;       /* exit outer loop */
2221                                 retval = (cmd->opcode == O_DIVERT) ?
2222                                         IP_FW_DIVERT : IP_FW_TEE;
2223                                 set_match(args, f_pos, chain);
2224                                 args->rule.info = TARG(cmd->arg1, divert);
2225                                 break;
2226
2227                         case O_COUNT:
2228                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2229                                 l = 0;          /* exit inner loop */
2230                                 break;
2231
2232                         case O_SKIPTO:
2233                             IPFW_INC_RULE_COUNTER(f, pktlen);
2234                             f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2235                             /*
2236                              * Skip disabled rules, and re-enter
2237                              * the inner loop with the correct
2238                              * f_pos, f, l and cmd.
2239                              * Also clear cmdlen and skip_or
2240                              */
2241                             for (; f_pos < chain->n_rules - 1 &&
2242                                     (V_set_disable &
2243                                      (1 << chain->map[f_pos]->set));
2244                                     f_pos++)
2245                                 ;
2246                             /* Re-enter the inner loop at the skipto rule. */
2247                             f = chain->map[f_pos];
2248                             l = f->cmd_len;
2249                             cmd = f->cmd;
2250                             match = 1;
2251                             cmdlen = 0;
2252                             skip_or = 0;
2253                             continue;
2254                             break;      /* not reached */
2255
2256                         case O_CALLRETURN: {
2257                                 /*
2258                                  * Implementation of `subroutine' call/return,
2259                                  * in the stack carried in an mbuf tag. This
2260                                  * is different from `skipto' in that any call
2261                                  * address is possible (`skipto' must prevent
2262                                  * backward jumps to avoid endless loops).
2263                                  * We have `return' action when F_NOT flag is
2264                                  * present. The `m_tag_id' field is used as
2265                                  * stack pointer.
2266                                  */
2267                                 struct m_tag *mtag;
2268                                 uint16_t jmpto, *stack;
2269
2270 #define IS_CALL         ((cmd->len & F_NOT) == 0)
2271 #define IS_RETURN       ((cmd->len & F_NOT) != 0)
2272                                 /*
2273                                  * Hand-rolled version of m_tag_locate() with
2274                                  * wildcard `type'.
2275                                  * If not already tagged, allocate new tag.
2276                                  */
2277                                 mtag = m_tag_first(m);
2278                                 while (mtag != NULL) {
2279                                         if (mtag->m_tag_cookie ==
2280                                             MTAG_IPFW_CALL)
2281                                                 break;
2282                                         mtag = m_tag_next(m, mtag);
2283                                 }
2284                                 if (mtag == NULL && IS_CALL) {
2285                                         mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2286                                             IPFW_CALLSTACK_SIZE *
2287                                             sizeof(uint16_t), M_NOWAIT);
2288                                         if (mtag != NULL)
2289                                                 m_tag_prepend(m, mtag);
2290                                 }
2291
2292                                 /*
2293                                  * On error both `call' and `return' just
2294                                  * continue with next rule.
2295                                  */
2296                                 if (IS_RETURN && (mtag == NULL ||
2297                                     mtag->m_tag_id == 0)) {
2298                                         l = 0;          /* exit inner loop */
2299                                         break;
2300                                 }
2301                                 if (IS_CALL && (mtag == NULL ||
2302                                     mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2303                                         printf("ipfw: call stack error, "
2304                                             "go to next rule\n");
2305                                         l = 0;          /* exit inner loop */
2306                                         break;
2307                                 }
2308
2309                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2310                                 stack = (uint16_t *)(mtag + 1);
2311
2312                                 /*
2313                                  * The `call' action may use cached f_pos
2314                                  * (in f->next_rule), whose version is written
2315                                  * in f->next_rule.
2316                                  * The `return' action, however, doesn't have
2317                                  * fixed jump address in cmd->arg1 and can't use
2318                                  * cache.
2319                                  */
2320                                 if (IS_CALL) {
2321                                         stack[mtag->m_tag_id] = f->rulenum;
2322                                         mtag->m_tag_id++;
2323                                         f_pos = JUMP(chain, f, cmd->arg1,
2324                                             tablearg, 1);
2325                                 } else {        /* `return' action */
2326                                         mtag->m_tag_id--;
2327                                         jmpto = stack[mtag->m_tag_id] + 1;
2328                                         f_pos = ipfw_find_rule(chain, jmpto, 0);
2329                                 }
2330
2331                                 /*
2332                                  * Skip disabled rules, and re-enter
2333                                  * the inner loop with the correct
2334                                  * f_pos, f, l and cmd.
2335                                  * Also clear cmdlen and skip_or
2336                                  */
2337                                 for (; f_pos < chain->n_rules - 1 &&
2338                                     (V_set_disable &
2339                                     (1 << chain->map[f_pos]->set)); f_pos++)
2340                                         ;
2341                                 /* Re-enter the inner loop at the dest rule. */
2342                                 f = chain->map[f_pos];
2343                                 l = f->cmd_len;
2344                                 cmd = f->cmd;
2345                                 cmdlen = 0;
2346                                 skip_or = 0;
2347                                 continue;
2348                                 break;  /* NOTREACHED */
2349                         }
2350 #undef IS_CALL
2351 #undef IS_RETURN
2352
2353                         case O_REJECT:
2354                                 /*
2355                                  * Drop the packet and send a reject notice
2356                                  * if the packet is not ICMP (or is an ICMP
2357                                  * query), and it is not multicast/broadcast.
2358                                  */
2359                                 if (hlen > 0 && is_ipv4 && offset == 0 &&
2360                                     (proto != IPPROTO_ICMP ||
2361                                      is_icmp_query(ICMP(ulp))) &&
2362                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2363                                     !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2364                                         send_reject(args, cmd->arg1, iplen, ip);
2365                                         m = args->m;
2366                                 }
2367                                 /* FALLTHROUGH */
2368 #ifdef INET6
2369                         case O_UNREACH6:
2370                                 if (hlen > 0 && is_ipv6 &&
2371                                     ((offset & IP6F_OFF_MASK) == 0) &&
2372                                     (proto != IPPROTO_ICMPV6 ||
2373                                      (is_icmp6_query(icmp6_type) == 1)) &&
2374                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2375                                     !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2376                                         send_reject6(
2377                                             args, cmd->arg1, hlen,
2378                                             (struct ip6_hdr *)ip);
2379                                         m = args->m;
2380                                 }
2381                                 /* FALLTHROUGH */
2382 #endif
2383                         case O_DENY:
2384                                 retval = IP_FW_DENY;
2385                                 l = 0;          /* exit inner loop */
2386                                 done = 1;       /* exit outer loop */
2387                                 break;
2388
2389                         case O_FORWARD_IP:
2390                                 if (args->eh)   /* not valid on layer2 pkts */
2391                                         break;
2392                                 if (q == NULL || q->rule != f ||
2393                                     dyn_dir == MATCH_FORWARD) {
2394                                     struct sockaddr_in *sa;
2395
2396                                     sa = &(((ipfw_insn_sa *)cmd)->sa);
2397                                     if (sa->sin_addr.s_addr == INADDR_ANY) {
2398 #ifdef INET6
2399                                         /*
2400                                          * We use O_FORWARD_IP opcode for
2401                                          * fwd rule with tablearg, but tables
2402                                          * now support IPv6 addresses. And
2403                                          * when we are inspecting IPv6 packet,
2404                                          * we can use nh6 field from
2405                                          * table_value as next_hop6 address.
2406                                          */
2407                                         if (is_ipv6) {
2408                                                 struct sockaddr_in6 *sa6;
2409
2410                                                 sa6 = args->next_hop6 =
2411                                                     &args->hopstore6;
2412                                                 sa6->sin6_family = AF_INET6;
2413                                                 sa6->sin6_len = sizeof(*sa6);
2414                                                 sa6->sin6_addr = TARG_VAL(
2415                                                     chain, tablearg, nh6);
2416                                                 /*
2417                                                  * Set sin6_scope_id only for
2418                                                  * link-local unicast addresses.
2419                                                  */
2420                                                 if (IN6_IS_ADDR_LINKLOCAL(
2421                                                     &sa6->sin6_addr))
2422                                                         sa6->sin6_scope_id =
2423                                                             TARG_VAL(chain,
2424                                                                 tablearg,
2425                                                                 zoneid);
2426                                         } else
2427 #endif
2428                                         {
2429                                                 sa = args->next_hop =
2430                                                     &args->hopstore;
2431                                                 sa->sin_family = AF_INET;
2432                                                 sa->sin_len = sizeof(*sa);
2433                                                 sa->sin_addr.s_addr = htonl(
2434                                                     TARG_VAL(chain, tablearg,
2435                                                     nh4));
2436                                         }
2437                                     } else {
2438                                         args->next_hop = sa;
2439                                     }
2440                                 }
2441                                 retval = IP_FW_PASS;
2442                                 l = 0;          /* exit inner loop */
2443                                 done = 1;       /* exit outer loop */
2444                                 break;
2445
2446 #ifdef INET6
2447                         case O_FORWARD_IP6:
2448                                 if (args->eh)   /* not valid on layer2 pkts */
2449                                         break;
2450                                 if (q == NULL || q->rule != f ||
2451                                     dyn_dir == MATCH_FORWARD) {
2452                                         struct sockaddr_in6 *sin6;
2453
2454                                         sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2455                                         args->next_hop6 = sin6;
2456                                 }
2457                                 retval = IP_FW_PASS;
2458                                 l = 0;          /* exit inner loop */
2459                                 done = 1;       /* exit outer loop */
2460                                 break;
2461 #endif
2462
2463                         case O_NETGRAPH:
2464                         case O_NGTEE:
2465                                 set_match(args, f_pos, chain);
2466                                 args->rule.info = TARG(cmd->arg1, netgraph);
2467                                 if (V_fw_one_pass)
2468                                         args->rule.info |= IPFW_ONEPASS;
2469                                 retval = (cmd->opcode == O_NETGRAPH) ?
2470                                     IP_FW_NETGRAPH : IP_FW_NGTEE;
2471                                 l = 0;          /* exit inner loop */
2472                                 done = 1;       /* exit outer loop */
2473                                 break;
2474
2475                         case O_SETFIB: {
2476                                 uint32_t fib;
2477
2478                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2479                                 fib = TARG(cmd->arg1, fib) & 0x7FFFF;
2480                                 if (fib >= rt_numfibs)
2481                                         fib = 0;
2482                                 M_SETFIB(m, fib);
2483                                 args->f_id.fib = fib;
2484                                 l = 0;          /* exit inner loop */
2485                                 break;
2486                         }
2487
2488                         case O_SETDSCP: {
2489                                 uint16_t code;
2490
2491                                 code = TARG(cmd->arg1, dscp) & 0x3F;
2492                                 l = 0;          /* exit inner loop */
2493                                 if (is_ipv4) {
2494                                         uint16_t old;
2495
2496                                         old = *(uint16_t *)ip;
2497                                         ip->ip_tos = (code << 2) |
2498                                             (ip->ip_tos & 0x03);
2499                                         ip->ip_sum = cksum_adjust(ip->ip_sum,
2500                                             old, *(uint16_t *)ip);
2501                                 } else if (is_ipv6) {
2502                                         uint8_t *v;
2503
2504                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
2505                                         *v = (*v & 0xF0) | (code >> 2);
2506                                         v++;
2507                                         *v = (*v & 0x3F) | ((code & 0x03) << 6);
2508                                 } else
2509                                         break;
2510
2511                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2512                                 break;
2513                         }
2514
2515                         case O_NAT:
2516                                 l = 0;          /* exit inner loop */
2517                                 done = 1;       /* exit outer loop */
2518                                 if (!IPFW_NAT_LOADED) {
2519                                     retval = IP_FW_DENY;
2520                                     break;
2521                                 }
2522
2523                                 struct cfg_nat *t;
2524                                 int nat_id;
2525
2526                                 set_match(args, f_pos, chain);
2527                                 /* Check if this is 'global' nat rule */
2528                                 if (cmd->arg1 == 0) {
2529                                         retval = ipfw_nat_ptr(args, NULL, m);
2530                                         break;
2531                                 }
2532                                 t = ((ipfw_insn_nat *)cmd)->nat;
2533                                 if (t == NULL) {
2534                                         nat_id = TARG(cmd->arg1, nat);
2535                                         t = (*lookup_nat_ptr)(&chain->nat, nat_id);
2536
2537                                         if (t == NULL) {
2538                                             retval = IP_FW_DENY;
2539                                             break;
2540                                         }
2541                                         if (cmd->arg1 != IP_FW_TARG)
2542                                             ((ipfw_insn_nat *)cmd)->nat = t;
2543                                 }
2544                                 retval = ipfw_nat_ptr(args, t, m);
2545                                 break;
2546
2547                         case O_REASS: {
2548                                 int ip_off;
2549
2550                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2551                                 l = 0;  /* in any case exit inner loop */
2552                                 ip_off = ntohs(ip->ip_off);
2553
2554                                 /* if not fragmented, go to next rule */
2555                                 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
2556                                     break;
2557
2558                                 args->m = m = ip_reass(m);
2559
2560                                 /*
2561                                  * do IP header checksum fixup.
2562                                  */
2563                                 if (m == NULL) { /* fragment got swallowed */
2564                                     retval = IP_FW_DENY;
2565                                 } else { /* good, packet complete */
2566                                     int hlen;
2567
2568                                     ip = mtod(m, struct ip *);
2569                                     hlen = ip->ip_hl << 2;
2570                                     ip->ip_sum = 0;
2571                                     if (hlen == sizeof(struct ip))
2572                                         ip->ip_sum = in_cksum_hdr(ip);
2573                                     else
2574                                         ip->ip_sum = in_cksum(m, hlen);
2575                                     retval = IP_FW_REASS;
2576                                     set_match(args, f_pos, chain);
2577                                 }
2578                                 done = 1;       /* exit outer loop */
2579                                 break;
2580                         }
2581
2582                         default:
2583                                 panic("-- unknown opcode %d\n", cmd->opcode);
2584                         } /* end of switch() on opcodes */
2585                         /*
2586                          * if we get here with l=0, then match is irrelevant.
2587                          */
2588
2589                         if (cmd->len & F_NOT)
2590                                 match = !match;
2591
2592                         if (match) {
2593                                 if (cmd->len & F_OR)
2594                                         skip_or = 1;
2595                         } else {
2596                                 if (!(cmd->len & F_OR)) /* not an OR block, */
2597                                         break;          /* try next rule    */
2598                         }
2599
2600                 }       /* end of inner loop, scan opcodes */
2601 #undef PULLUP_LEN
2602
2603                 if (done)
2604                         break;
2605
2606 /* next_rule:; */       /* try next rule                */
2607
2608         }               /* end of outer for, scan rules */
2609
2610         if (done) {
2611                 struct ip_fw *rule = chain->map[f_pos];
2612                 /* Update statistics */
2613                 IPFW_INC_RULE_COUNTER(rule, pktlen);
2614         } else {
2615                 retval = IP_FW_DENY;
2616                 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2617         }
2618         IPFW_PF_RUNLOCK(chain);
2619 #ifdef __FreeBSD__
2620         if (ucred_cache != NULL)
2621                 crfree(ucred_cache);
2622 #endif
2623         return (retval);
2624
2625 pullup_failed:
2626         if (V_fw_verbose)
2627                 printf("ipfw: pullup failed\n");
2628         return (IP_FW_DENY);
2629 }
2630
2631 /*
2632  * Set maximum number of tables that can be used in given VNET ipfw instance.
2633  */
2634 #ifdef SYSCTL_NODE
2635 static int
2636 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
2637 {
2638         int error;
2639         unsigned int ntables;
2640
2641         ntables = V_fw_tables_max;
2642
2643         error = sysctl_handle_int(oidp, &ntables, 0, req);
2644         /* Read operation or some error */
2645         if ((error != 0) || (req->newptr == NULL))
2646                 return (error);
2647
2648         return (ipfw_resize_tables(&V_layer3_chain, ntables));
2649 }
2650
2651 /*
2652  * Switches table namespace between global and per-set.
2653  */
2654 static int
2655 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
2656 {
2657         int error;
2658         unsigned int sets;
2659
2660         sets = V_fw_tables_sets;
2661
2662         error = sysctl_handle_int(oidp, &sets, 0, req);
2663         /* Read operation or some error */
2664         if ((error != 0) || (req->newptr == NULL))
2665                 return (error);
2666
2667         return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
2668 }
2669 #endif
2670
2671 /*
2672  * Module and VNET glue
2673  */
2674
2675 /*
2676  * Stuff that must be initialised only on boot or module load
2677  */
2678 static int
2679 ipfw_init(void)
2680 {
2681         int error = 0;
2682
2683         /*
2684          * Only print out this stuff the first time around,
2685          * when called from the sysinit code.
2686          */
2687         printf("ipfw2 "
2688 #ifdef INET6
2689                 "(+ipv6) "
2690 #endif
2691                 "initialized, divert %s, nat %s, "
2692                 "default to %s, logging ",
2693 #ifdef IPDIVERT
2694                 "enabled",
2695 #else
2696                 "loadable",
2697 #endif
2698 #ifdef IPFIREWALL_NAT
2699                 "enabled",
2700 #else
2701                 "loadable",
2702 #endif
2703                 default_to_accept ? "accept" : "deny");
2704
2705         /*
2706          * Note: V_xxx variables can be accessed here but the vnet specific
2707          * initializer may not have been called yet for the VIMAGE case.
2708          * Tuneables will have been processed. We will print out values for
2709          * the default vnet.
2710          * XXX This should all be rationalized AFTER 8.0
2711          */
2712         if (V_fw_verbose == 0)
2713                 printf("disabled\n");
2714         else if (V_verbose_limit == 0)
2715                 printf("unlimited\n");
2716         else
2717                 printf("limited to %d packets/entry by default\n",
2718                     V_verbose_limit);
2719
2720         /* Check user-supplied table count for validness */
2721         if (default_fw_tables > IPFW_TABLES_MAX)
2722           default_fw_tables = IPFW_TABLES_MAX;
2723
2724         ipfw_init_sopt_handler();
2725         ipfw_log_bpf(1); /* init */
2726         ipfw_iface_init();
2727         return (error);
2728 }
2729
2730 /*
2731  * Called for the removal of the last instance only on module unload.
2732  */
2733 static void
2734 ipfw_destroy(void)
2735 {
2736
2737         ipfw_iface_destroy();
2738         ipfw_log_bpf(0); /* uninit */
2739         ipfw_destroy_sopt_handler();
2740         printf("IP firewall unloaded\n");
2741 }
2742
2743 /*
2744  * Stuff that must be initialized for every instance
2745  * (including the first of course).
2746  */
2747 static int
2748 vnet_ipfw_init(const void *unused)
2749 {
2750         int error, first;
2751         struct ip_fw *rule = NULL;
2752         struct ip_fw_chain *chain;
2753
2754         chain = &V_layer3_chain;
2755
2756         first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
2757
2758         /* First set up some values that are compile time options */
2759         V_autoinc_step = 100;   /* bounded to 1..1000 in add_rule() */
2760         V_fw_deny_unknown_exthdrs = 1;
2761 #ifdef IPFIREWALL_VERBOSE
2762         V_fw_verbose = 1;
2763 #endif
2764 #ifdef IPFIREWALL_VERBOSE_LIMIT
2765         V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2766 #endif
2767 #ifdef IPFIREWALL_NAT
2768         LIST_INIT(&chain->nat);
2769 #endif
2770
2771         /* Init shared services hash table */
2772         ipfw_init_srv(chain);
2773
2774         ipfw_init_obj_rewriter();
2775         ipfw_init_counters();
2776         /* insert the default rule and create the initial map */
2777         chain->n_rules = 1;
2778         chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_WAITOK | M_ZERO);
2779         rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
2780
2781         /* Set initial number of tables */
2782         V_fw_tables_max = default_fw_tables;
2783         error = ipfw_init_tables(chain, first);
2784         if (error) {
2785                 printf("ipfw2: setting up tables failed\n");
2786                 free(chain->map, M_IPFW);
2787                 free(rule, M_IPFW);
2788                 return (ENOSPC);
2789         }
2790
2791         /* fill and insert the default rule */
2792         rule->act_ofs = 0;
2793         rule->rulenum = IPFW_DEFAULT_RULE;
2794         rule->cmd_len = 1;
2795         rule->set = RESVD_SET;
2796         rule->cmd[0].len = 1;
2797         rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
2798         chain->default_rule = chain->map[0] = rule;
2799         chain->id = rule->id = 1;
2800         /* Pre-calculate rules length for legacy dump format */
2801         chain->static_len = sizeof(struct ip_fw_rule0);
2802
2803         IPFW_LOCK_INIT(chain);
2804         ipfw_dyn_init(chain);
2805 #ifdef LINEAR_SKIPTO
2806         ipfw_init_skipto_cache(chain);
2807 #endif
2808
2809         /* First set up some values that are compile time options */
2810         V_ipfw_vnet_ready = 1;          /* Open for business */
2811
2812         /*
2813          * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
2814          * Even if the latter two fail we still keep the module alive
2815          * because the sockopt and layer2 paths are still useful.
2816          * ipfw[6]_hook return 0 on success, ENOENT on failure,
2817          * so we can ignore the exact return value and just set a flag.
2818          *
2819          * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
2820          * changes in the underlying (per-vnet) variables trigger
2821          * immediate hook()/unhook() calls.
2822          * In layer2 we have the same behaviour, except that V_ether_ipfw
2823          * is checked on each packet because there are no pfil hooks.
2824          */
2825         V_ip_fw_ctl_ptr = ipfw_ctl3;
2826         error = ipfw_attach_hooks(1);
2827         return (error);
2828 }
2829
2830 /*
2831  * Called for the removal of each instance.
2832  */
2833 static int
2834 vnet_ipfw_uninit(const void *unused)
2835 {
2836         struct ip_fw *reap;
2837         struct ip_fw_chain *chain = &V_layer3_chain;
2838         int i, last;
2839
2840         V_ipfw_vnet_ready = 0; /* tell new callers to go away */
2841         /*
2842          * disconnect from ipv4, ipv6, layer2 and sockopt.
2843          * Then grab, release and grab again the WLOCK so we make
2844          * sure the update is propagated and nobody will be in.
2845          */
2846         (void)ipfw_attach_hooks(0 /* detach */);
2847         V_ip_fw_ctl_ptr = NULL;
2848
2849         last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
2850
2851         IPFW_UH_WLOCK(chain);
2852         IPFW_UH_WUNLOCK(chain);
2853         IPFW_UH_WLOCK(chain);
2854
2855         IPFW_WLOCK(chain);
2856         ipfw_dyn_uninit(0);     /* run the callout_drain */
2857         IPFW_WUNLOCK(chain);
2858
2859         reap = NULL;
2860         IPFW_WLOCK(chain);
2861         for (i = 0; i < chain->n_rules; i++)
2862                 ipfw_reap_add(chain, &reap, chain->map[i]);
2863         free(chain->map, M_IPFW);
2864 #ifdef LINEAR_SKIPTO
2865         ipfw_destroy_skipto_cache(chain);
2866 #endif
2867         IPFW_WUNLOCK(chain);
2868         IPFW_UH_WUNLOCK(chain);
2869         ipfw_destroy_tables(chain, last);
2870         if (reap != NULL)
2871                 ipfw_reap_rules(reap);
2872         vnet_ipfw_iface_destroy(chain);
2873         ipfw_destroy_srv(chain);
2874         IPFW_LOCK_DESTROY(chain);
2875         ipfw_dyn_uninit(1);     /* free the remaining parts */
2876         ipfw_destroy_counters();
2877         ipfw_destroy_obj_rewriter();
2878         return (0);
2879 }
2880
2881 /*
2882  * Module event handler.
2883  * In general we have the choice of handling most of these events by the
2884  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
2885  * use the SYSINIT handlers as they are more capable of expressing the
2886  * flow of control during module and vnet operations, so this is just
2887  * a skeleton. Note there is no SYSINIT equivalent of the module
2888  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
2889  */
2890 static int
2891 ipfw_modevent(module_t mod, int type, void *unused)
2892 {
2893         int err = 0;
2894
2895         switch (type) {
2896         case MOD_LOAD:
2897                 /* Called once at module load or
2898                  * system boot if compiled in. */
2899                 break;
2900         case MOD_QUIESCE:
2901                 /* Called before unload. May veto unloading. */
2902                 break;
2903         case MOD_UNLOAD:
2904                 /* Called during unload. */
2905                 break;
2906         case MOD_SHUTDOWN:
2907                 /* Called during system shutdown. */
2908                 break;
2909         default:
2910                 err = EOPNOTSUPP;
2911                 break;
2912         }
2913         return err;
2914 }
2915
2916 static moduledata_t ipfwmod = {
2917         "ipfw",
2918         ipfw_modevent,
2919         0
2920 };
2921
2922 /* Define startup order. */
2923 #define IPFW_SI_SUB_FIREWALL    SI_SUB_PROTO_IFATTACHDOMAIN
2924 #define IPFW_MODEVENT_ORDER     (SI_ORDER_ANY - 255) /* On boot slot in here. */
2925 #define IPFW_MODULE_ORDER       (IPFW_MODEVENT_ORDER + 1) /* A little later. */
2926 #define IPFW_VNET_ORDER         (IPFW_MODEVENT_ORDER + 2) /* Later still. */
2927
2928 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
2929 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
2930 MODULE_VERSION(ipfw, 3);
2931 /* should declare some dependencies here */
2932
2933 /*
2934  * Starting up. Done in order after ipfwmod() has been called.
2935  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
2936  */
2937 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2938             ipfw_init, NULL);
2939 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2940             vnet_ipfw_init, NULL);
2941
2942 /*
2943  * Closing up shop. These are done in REVERSE ORDER, but still
2944  * after ipfwmod() has been called. Not called on reboot.
2945  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
2946  * or when the module is unloaded.
2947  */
2948 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2949             ipfw_destroy, NULL);
2950 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2951             vnet_ipfw_uninit, NULL);
2952 /* end of file */