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 int
 507 search_ip6_addr_net (struct in6_addr * ip6_addr)
 508 {
 509         struct ifnet *mdc;
 510         struct ifaddr *mdc2;
 511         struct in6_ifaddr *fdm;
 512         struct in6_addr copia;
 513
 514         TAILQ_FOREACH(mdc, &V_ifnet, if_link) {
 515                 if_addr_rlock(mdc);
 516                 TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) {
 517                         if (mdc2->ifa_addr->sa_family == AF_INET6) {
 518                                 fdm = (struct in6_ifaddr *)mdc2;
 519                                 copia = fdm->ia_addr.sin6_addr;
 520                                 /* need for leaving scope_id in the sock_addr */
 521                                 in6_clearscope(&copia);
 522                                 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) {
 523                                         if_addr_runlock(mdc);
 524                                         return 1;
 525                                 }
 526                         }
 527                 }
 528                 if_addr_runlock(mdc);
 529         }
 530         return 0;
 531 }
 532
 533 static int
 534 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
 535 {
 536         struct route_in6 ro;
 537         struct sockaddr_in6 *dst;
 538
 539         bzero(&ro, sizeof(ro));
 540
 541         dst = (struct sockaddr_in6 * )&(ro.ro_dst);
 542         dst->sin6_family = AF_INET6;
 543         dst->sin6_len = sizeof(*dst);
 544         dst->sin6_addr = *src;
 545
 546         in6_rtalloc_ign(&ro, 0, fib);
 547         if (ro.ro_rt == NULL)
 548                 return 0;
 549
 550         /*
 551          * if ifp is provided, check for equality with rtentry
 552          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 553          * to support the case of sending packets to an address of our own.
 554          * (where the former interface is the first argument of if_simloop()
 555          *  (=ifp), the latter is lo0)
 556          */
 557         if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
 558                 RTFREE(ro.ro_rt);
 559                 return 0;
 560         }
 561
 562         /* if no ifp provided, check if rtentry is not default route */
 563         if (ifp == NULL &&
 564             IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) {
 565                 RTFREE(ro.ro_rt);
 566                 return 0;
 567         }
 568
 569         /* or if this is a blackhole/reject route */
 570         if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
 571                 RTFREE(ro.ro_rt);
 572                 return 0;
 573         }
 574
 575         /* found valid route */
 576         RTFREE(ro.ro_rt);
 577         return 1;
 578
 579 }
 580
 581 static int
 582 is_icmp6_query(int icmp6_type)
 583 {
 584         if ((icmp6_type <= ICMP6_MAXTYPE) &&
 585             (icmp6_type == ICMP6_ECHO_REQUEST ||
 586             icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
 587             icmp6_type == ICMP6_WRUREQUEST ||
 588             icmp6_type == ICMP6_FQDN_QUERY ||
 589             icmp6_type == ICMP6_NI_QUERY))
 590                 return (1);
 591
 592         return (0);
 593 }
 594
 595 static void
 596 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
 597 {
 598         struct mbuf *m;
 599
 600         m = args->m;
 601         if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
 602                 struct tcphdr *tcp;
 603                 tcp = (struct tcphdr *)((char *)ip6 + hlen);
 604
 605                 if ((tcp->th_flags & TH_RST) == 0) {
 606                         struct mbuf *m0;
 607                         m0 = ipfw_send_pkt(args->m, &(args->f_id),
 608                             ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 609                             tcp->th_flags | TH_RST);
 610                         if (m0 != NULL)
 611                                 ip6_output(m0, NULL, NULL, 0, NULL, NULL,
 612                                     NULL);
 613                 }
 614                 FREE_PKT(m);
 615         } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
 616 #if 0
 617                 /*
 618                  * Unlike above, the mbufs need to line up with the ip6 hdr,
 619                  * as the contents are read. We need to m_adj() the
 620                  * needed amount.
 621                  * The mbuf will however be thrown away so we can adjust it.
 622                  * Remember we did an m_pullup on it already so we
 623                  * can make some assumptions about contiguousness.
 624                  */
 625                 if (args->L3offset)
 626                         m_adj(m, args->L3offset);
 627 #endif
 628                 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
 629         } else
 630                 FREE_PKT(m);
 631
 632         args->m = NULL;
 633 }
 634
 635 #endif /* INET6 */
 636
 637
 638 /*
 639  * sends a reject message, consuming the mbuf passed as an argument.
 640  */
 641 static void
 642 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
 643 {
 644
 645 #if 0
 646         /* XXX When ip is not guaranteed to be at mtod() we will
 647          * need to account for this */
 648          * The mbuf will however be thrown away so we can adjust it.
 649          * Remember we did an m_pullup on it already so we
 650          * can make some assumptions about contiguousness.
 651          */
 652         if (args->L3offset)
 653                 m_adj(m, args->L3offset);
 654 #endif
 655         if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
 656                 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
 657         } else if (args->f_id.proto == IPPROTO_TCP) {
 658                 struct tcphdr *const tcp =
 659                     L3HDR(struct tcphdr, mtod(args->m, struct ip *));
 660                 if ( (tcp->th_flags & TH_RST) == 0) {
 661                         struct mbuf *m;
 662                         m = ipfw_send_pkt(args->m, &(args->f_id),
 663                                 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 664                                 tcp->th_flags | TH_RST);
 665                         if (m != NULL)
 666                                 ip_output(m, NULL, NULL, 0, NULL, NULL);
 667                 }
 668                 FREE_PKT(args->m);
 669         } else
 670                 FREE_PKT(args->m);
 671         args->m = NULL;
 672 }
 673
 674 /*
 675  * Support for uid/gid/jail lookup. These tests are expensive
 676  * (because we may need to look into the list of active sockets)
 677  * so we cache the results. ugid_lookupp is 0 if we have not
 678  * yet done a lookup, 1 if we succeeded, and -1 if we tried
 679  * and failed. The function always returns the match value.
 680  * We could actually spare the variable and use *uc, setting
 681  * it to '(void *)check_uidgid if we have no info, NULL if
 682  * we tried and failed, or any other value if successful.
 683  */
 684 static int
 685 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
 686     struct ucred **uc)
 687 {
 688 #if defined(USERSPACE)
 689         return 0;       // not supported in userspace
 690 #else
 691 #ifndef __FreeBSD__
 692         /* XXX */
 693         return cred_check(insn, proto, oif,
 694             dst_ip, dst_port, src_ip, src_port,
 695             (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
 696 #else  /* FreeBSD */
 697         struct in_addr src_ip, dst_ip;
 698         struct inpcbinfo *pi;
 699         struct ipfw_flow_id *id;
 700         struct inpcb *pcb, *inp;
 701         struct ifnet *oif;
 702         int lookupflags;
 703         int match;
 704
 705         id = &args->f_id;
 706         inp = args->inp;
 707         oif = args->oif;
 708
 709         /*
 710          * Check to see if the UDP or TCP stack supplied us with
 711          * the PCB. If so, rather then holding a lock and looking
 712          * up the PCB, we can use the one that was supplied.
 713          */
 714         if (inp && *ugid_lookupp == 0) {
 715                 INP_LOCK_ASSERT(inp);
 716                 if (inp->inp_socket != NULL) {
 717                         *uc = crhold(inp->inp_cred);
 718                         *ugid_lookupp = 1;
 719                 } else
 720                         *ugid_lookupp = -1;
 721         }
 722         /*
 723          * If we have already been here and the packet has no
 724          * PCB entry associated with it, then we can safely
 725          * assume that this is a no match.
 726          */
 727         if (*ugid_lookupp == -1)
 728                 return (0);
 729         if (id->proto == IPPROTO_TCP) {
 730                 lookupflags = 0;
 731                 pi = &V_tcbinfo;
 732         } else if (id->proto == IPPROTO_UDP) {
 733                 lookupflags = INPLOOKUP_WILDCARD;
 734                 pi = &V_udbinfo;
 735         } else
 736                 return 0;
 737         lookupflags |= INPLOOKUP_RLOCKPCB;
 738         match = 0;
 739         if (*ugid_lookupp == 0) {
 740                 if (id->addr_type == 6) {
 741 #ifdef INET6
 742                         if (oif == NULL)
 743                                 pcb = in6_pcblookup_mbuf(pi,
 744                                     &id->src_ip6, htons(id->src_port),
 745                                     &id->dst_ip6, htons(id->dst_port),
 746                                     lookupflags, oif, args->m);
 747                         else
 748                                 pcb = in6_pcblookup_mbuf(pi,
 749                                     &id->dst_ip6, htons(id->dst_port),
 750                                     &id->src_ip6, htons(id->src_port),
 751                                     lookupflags, oif, args->m);
 752 #else
 753                         *ugid_lookupp = -1;
 754                         return (0);
 755 #endif
 756                 } else {
 757                         src_ip.s_addr = htonl(id->src_ip);
 758                         dst_ip.s_addr = htonl(id->dst_ip);
 759                         if (oif == NULL)
 760                                 pcb = in_pcblookup_mbuf(pi,
 761                                     src_ip, htons(id->src_port),
 762                                     dst_ip, htons(id->dst_port),
 763                                     lookupflags, oif, args->m);
 764                         else
 765                                 pcb = in_pcblookup_mbuf(pi,
 766                                     dst_ip, htons(id->dst_port),
 767                                     src_ip, htons(id->src_port),
 768                                     lookupflags, oif, args->m);
 769                 }
 770                 if (pcb != NULL) {
 771                         INP_RLOCK_ASSERT(pcb);
 772                         *uc = crhold(pcb->inp_cred);
 773                         *ugid_lookupp = 1;
 774                         INP_RUNLOCK(pcb);
 775                 }
 776                 if (*ugid_lookupp == 0) {
 777                         /*
 778                          * We tried and failed, set the variable to -1
 779                          * so we will not try again on this packet.
 780                          */
 781                         *ugid_lookupp = -1;
 782                         return (0);
 783                 }
 784         }
 785         if (insn->o.opcode == O_UID)
 786                 match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
 787         else if (insn->o.opcode == O_GID)
 788                 match = groupmember((gid_t)insn->d[0], *uc);
 789         else if (insn->o.opcode == O_JAIL)
 790                 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
 791         return (match);
 792 #endif /* __FreeBSD__ */
 793 #endif /* not supported in userspace */
 794 }
 795
 796 /*
 797  * Helper function to set args with info on the rule after the matching
 798  * one. slot is precise, whereas we guess rule_id as they are
 799  * assigned sequentially.
 800  */
 801 static inline void
 802 set_match(struct ip_fw_args *args, int slot,
 803         struct ip_fw_chain *chain)
 804 {
 805         args->rule.chain_id = chain->id;
 806         args->rule.slot = slot + 1; /* we use 0 as a marker */
 807         args->rule.rule_id = 1 + chain->map[slot]->id;
 808         args->rule.rulenum = chain->map[slot]->rulenum;
 809 }
 810
 811 #ifndef LINEAR_SKIPTO
 812 /*
 813  * Helper function to enable cached rule lookups using
 814  * cached_id and cached_pos fields in ipfw rule.
 815  */
 816 static int
 817 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
 818     int tablearg, int jump_backwards)
 819 {
 820         int f_pos;
 821
 822         /* If possible use cached f_pos (in f->cached_pos),
 823          * whose version is written in f->cached_id
 824          * (horrible hacks to avoid changing the ABI).
 825          */
 826         if (num != IP_FW_TARG && f->cached_id == chain->id)
 827                 f_pos = f->cached_pos;
 828         else {
 829                 int i = IP_FW_ARG_TABLEARG(chain, num, skipto);
 830                 /* make sure we do not jump backward */
 831                 if (jump_backwards == 0 && i <= f->rulenum)
 832                         i = f->rulenum + 1;
 833                 if (chain->idxmap != NULL)
 834                         f_pos = chain->idxmap[i];
 835                 else
 836                         f_pos = ipfw_find_rule(chain, i, 0);
 837                 /* update the cache */
 838                 if (num != IP_FW_TARG) {
 839                         f->cached_id = chain->id;
 840                         f->cached_pos = f_pos;
 841                 }
 842         }
 843
 844         return (f_pos);
 845 }
 846 #else
 847 /*
 848  * Helper function to enable real fast rule lookups.
 849  */
 850 static int
 851 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
 852     int tablearg, int jump_backwards)
 853 {
 854         int f_pos;
 855
 856         num = IP_FW_ARG_TABLEARG(chain, num, skipto);
 857         /* make sure we do not jump backward */
 858         if (jump_backwards == 0 && num <= f->rulenum)
 859                 num = f->rulenum + 1;
 860         f_pos = chain->idxmap[num];
 861
 862         return (f_pos);
 863 }
 864 #endif
 865
 866 #define TARG(k, f)      IP_FW_ARG_TABLEARG(chain, k, f)
 867 /*
 868  * The main check routine for the firewall.
 869  *
 870  * All arguments are in args so we can modify them and return them
 871  * back to the caller.
 872  *
 873  * Parameters:
 874  *
 875  *      args->m (in/out) The packet; we set to NULL when/if we nuke it.
 876  *              Starts with the IP header.
 877  *      args->eh (in)   Mac header if present, NULL for layer3 packet.
 878  *      args->L3offset  Number of bytes bypassed if we came from L2.
 879  *                      e.g. often sizeof(eh)  ** NOTYET **
 880  *      args->oif       Outgoing interface, NULL if packet is incoming.
 881  *              The incoming interface is in the mbuf. (in)
 882  *      args->divert_rule (in/out)
 883  *              Skip up to the first rule past this rule number;
 884  *              upon return, non-zero port number for divert or tee.
 885  *
 886  *      args->rule      Pointer to the last matching rule (in/out)
 887  *      args->next_hop  Socket we are forwarding to (out).
 888  *      args->next_hop6 IPv6 next hop we are forwarding to (out).
 889  *      args->f_id      Addresses grabbed from the packet (out)
 890  *      args->rule.info a cookie depending on rule action
 891  *
 892  * Return value:
 893  *
 894  *      IP_FW_PASS      the packet must be accepted
 895  *      IP_FW_DENY      the packet must be dropped
 896  *      IP_FW_DIVERT    divert packet, port in m_tag
 897  *      IP_FW_TEE       tee packet, port in m_tag
 898  *      IP_FW_DUMMYNET  to dummynet, pipe in args->cookie
 899  *      IP_FW_NETGRAPH  into netgraph, cookie args->cookie
 900  *              args->rule contains the matching rule,
 901  *              args->rule.info has additional information.
 902  *
 903  */
 904 int
 905 ipfw_chk(struct ip_fw_args *args)
 906 {
 907
 908         /*
 909          * Local variables holding state while processing a packet:
 910          *
 911          * IMPORTANT NOTE: to speed up the processing of rules, there
 912          * are some assumption on the values of the variables, which
 913          * are documented here. Should you change them, please check
 914          * the implementation of the various instructions to make sure
 915          * that they still work.
 916          *
 917          * args->eh     The MAC header. It is non-null for a layer2
 918          *      packet, it is NULL for a layer-3 packet.
 919          * **notyet**
 920          * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
 921          *
 922          * m | args->m  Pointer to the mbuf, as received from the caller.
 923          *      It may change if ipfw_chk() does an m_pullup, or if it
 924          *      consumes the packet because it calls send_reject().
 925          *      XXX This has to change, so that ipfw_chk() never modifies
 926          *      or consumes the buffer.
 927          * ip   is the beginning of the ip(4 or 6) header.
 928          *      Calculated by adding the L3offset to the start of data.
 929          *      (Until we start using L3offset, the packet is
 930          *      supposed to start with the ip header).
 931          */
 932         struct mbuf *m = args->m;
 933         struct ip *ip = mtod(m, struct ip *);
 934
 935         /*
 936          * For rules which contain uid/gid or jail constraints, cache
 937          * a copy of the users credentials after the pcb lookup has been
 938          * executed. This will speed up the processing of rules with
 939          * these types of constraints, as well as decrease contention
 940          * on pcb related locks.
 941          */
 942 #ifndef __FreeBSD__
 943         struct bsd_ucred ucred_cache;
 944 #else
 945         struct ucred *ucred_cache = NULL;
 946 #endif
 947         int ucred_lookup = 0;
 948
 949         /*
 950          * oif | args->oif      If NULL, ipfw_chk has been called on the
 951          *      inbound path (ether_input, ip_input).
 952          *      If non-NULL, ipfw_chk has been called on the outbound path
 953          *      (ether_output, ip_output).
 954          */
 955         struct ifnet *oif = args->oif;
 956
 957         int f_pos = 0;          /* index of current rule in the array */
 958         int retval = 0;
 959
 960         /*
 961          * hlen The length of the IP header.
 962          */
 963         u_int hlen = 0;         /* hlen >0 means we have an IP pkt */
 964
 965         /*
 966          * offset       The offset of a fragment. offset != 0 means that
 967          *      we have a fragment at this offset of an IPv4 packet.
 968          *      offset == 0 means that (if this is an IPv4 packet)
 969          *      this is the first or only fragment.
 970          *      For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
 971          *      or there is a single packet fragement (fragement header added
 972          *      without needed).  We will treat a single packet fragment as if
 973          *      there was no fragment header (or log/block depending on the
 974          *      V_fw_permit_single_frag6 sysctl setting).
 975          */
 976         u_short offset = 0;
 977         u_short ip6f_mf = 0;
 978
 979         /*
 980          * Local copies of addresses. They are only valid if we have
 981          * an IP packet.
 982          *
 983          * proto        The protocol. Set to 0 for non-ip packets,
 984          *      or to the protocol read from the packet otherwise.
 985          *      proto != 0 means that we have an IPv4 packet.
 986          *
 987          * src_port, dst_port   port numbers, in HOST format. Only
 988          *      valid for TCP and UDP packets.
 989          *
 990          * src_ip, dst_ip       ip addresses, in NETWORK format.
 991          *      Only valid for IPv4 packets.
 992          */
 993         uint8_t proto;
 994         uint16_t src_port = 0, dst_port = 0;    /* NOTE: host format    */
 995         struct in_addr src_ip, dst_ip;          /* NOTE: network format */
 996         uint16_t iplen=0;
 997         int pktlen;
 998         uint16_t        etype = 0;      /* Host order stored ether type */
 999
1000         /*
1001          * dyn_dir = MATCH_UNKNOWN when rules unchecked,
1002          *      MATCH_NONE when checked and not matched (q = NULL),
1003          *      MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
1004          */
1005         int dyn_dir = MATCH_UNKNOWN;
1006         ipfw_dyn_rule *q = NULL;
1007         struct ip_fw_chain *chain = &V_layer3_chain;
1008
1009         /*
1010          * We store in ulp a pointer to the upper layer protocol header.
1011          * In the ipv4 case this is easy to determine from the header,
1012          * but for ipv6 we might have some additional headers in the middle.
1013          * ulp is NULL if not found.
1014          */
1015         void *ulp = NULL;               /* upper layer protocol pointer. */
1016
1017         /* XXX ipv6 variables */
1018         int is_ipv6 = 0;
1019         uint8_t icmp6_type = 0;
1020         uint16_t ext_hd = 0;    /* bits vector for extension header filtering */
1021         /* end of ipv6 variables */
1022
1023         int is_ipv4 = 0;
1024
1025         int done = 0;           /* flag to exit the outer loop */
1026         IPFW_RLOCK_TRACKER;
1027
1028         if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
1029                 return (IP_FW_PASS);    /* accept */
1030
1031         dst_ip.s_addr = 0;              /* make sure it is initialized */
1032         src_ip.s_addr = 0;              /* make sure it is initialized */
1033         pktlen = m->m_pkthdr.len;
1034         args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
1035         proto = args->f_id.proto = 0;   /* mark f_id invalid */
1036                 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
1037
1038 /*
1039  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1040  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1041  * pointer might become stale after other pullups (but we never use it
1042  * this way).
1043  */
1044 #define PULLUP_TO(_len, p, T)   PULLUP_LEN(_len, p, sizeof(T))
1045 #define PULLUP_LEN(_len, p, T)                                  \
1046 do {                                                            \
1047         int x = (_len) + T;                                     \
1048         if ((m)->m_len < x) {                                   \
1049                 args->m = m = m_pullup(m, x);                   \
1050                 if (m == NULL)                                  \
1051                         goto pullup_failed;                     \
1052         }                                                       \
1053         p = (mtod(m, char *) + (_len));                         \
1054 } while (0)
1055
1056         /*
1057          * if we have an ether header,
1058          */
1059         if (args->eh)
1060                 etype = ntohs(args->eh->ether_type);
1061
1062         /* Identify IP packets and fill up variables. */
1063         if (pktlen >= sizeof(struct ip6_hdr) &&
1064             (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
1065                 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1066                 is_ipv6 = 1;
1067                 args->f_id.addr_type = 6;
1068                 hlen = sizeof(struct ip6_hdr);
1069                 proto = ip6->ip6_nxt;
1070
1071                 /* Search extension headers to find upper layer protocols */
1072                 while (ulp == NULL && offset == 0) {
1073                         switch (proto) {
1074                         case IPPROTO_ICMPV6:
1075                                 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1076                                 icmp6_type = ICMP6(ulp)->icmp6_type;
1077                                 break;
1078
1079                         case IPPROTO_TCP:
1080                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1081                                 dst_port = TCP(ulp)->th_dport;
1082                                 src_port = TCP(ulp)->th_sport;
1083                                 /* save flags for dynamic rules */
1084                                 args->f_id._flags = TCP(ulp)->th_flags;
1085                                 break;
1086
1087                         case IPPROTO_SCTP:
1088                                 PULLUP_TO(hlen, ulp, struct sctphdr);
1089                                 src_port = SCTP(ulp)->src_port;
1090                                 dst_port = SCTP(ulp)->dest_port;
1091                                 break;
1092
1093                         case IPPROTO_UDP:
1094                                 PULLUP_TO(hlen, ulp, struct udphdr);
1095                                 dst_port = UDP(ulp)->uh_dport;
1096                                 src_port = UDP(ulp)->uh_sport;
1097                                 break;
1098
1099                         case IPPROTO_HOPOPTS:   /* RFC 2460 */
1100                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1101                                 ext_hd |= EXT_HOPOPTS;
1102                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1103                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1104                                 ulp = NULL;
1105                                 break;
1106
1107                         case IPPROTO_ROUTING:   /* RFC 2460 */
1108                                 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1109                                 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1110                                 case 0:
1111                                         ext_hd |= EXT_RTHDR0;
1112                                         break;
1113                                 case 2:
1114                                         ext_hd |= EXT_RTHDR2;
1115                                         break;
1116                                 default:
1117                                         if (V_fw_verbose)
1118                                                 printf("IPFW2: IPV6 - Unknown "
1119                                                     "Routing Header type(%d)\n",
1120                                                     ((struct ip6_rthdr *)
1121                                                     ulp)->ip6r_type);
1122                                         if (V_fw_deny_unknown_exthdrs)
1123                                             return (IP_FW_DENY);
1124                                         break;
1125                                 }
1126                                 ext_hd |= EXT_ROUTING;
1127                                 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1128                                 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1129                                 ulp = NULL;
1130                                 break;
1131
1132                         case IPPROTO_FRAGMENT:  /* RFC 2460 */
1133                                 PULLUP_TO(hlen, ulp, struct ip6_frag);
1134                                 ext_hd |= EXT_FRAGMENT;
1135                                 hlen += sizeof (struct ip6_frag);
1136                                 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1137                                 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1138                                         IP6F_OFF_MASK;
1139                                 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1140                                         IP6F_MORE_FRAG;
1141                                 if (V_fw_permit_single_frag6 == 0 &&
1142                                     offset == 0 && ip6f_mf == 0) {
1143                                         if (V_fw_verbose)
1144                                                 printf("IPFW2: IPV6 - Invalid "
1145                                                     "Fragment Header\n");
1146                                         if (V_fw_deny_unknown_exthdrs)
1147                                             return (IP_FW_DENY);
1148                                         break;
1149                                 }
1150                                 args->f_id.extra =
1151                                     ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1152                                 ulp = NULL;
1153                                 break;
1154
1155                         case IPPROTO_DSTOPTS:   /* RFC 2460 */
1156                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1157                                 ext_hd |= EXT_DSTOPTS;
1158                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1159                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1160                                 ulp = NULL;
1161                                 break;
1162
1163                         case IPPROTO_AH:        /* RFC 2402 */
1164                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1165                                 ext_hd |= EXT_AH;
1166                                 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1167                                 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1168                                 ulp = NULL;
1169                                 break;
1170
1171                         case IPPROTO_ESP:       /* RFC 2406 */
1172                                 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1173                                 /* Anything past Seq# is variable length and
1174                                  * data past this ext. header is encrypted. */
1175                                 ext_hd |= EXT_ESP;
1176                                 break;
1177
1178                         case IPPROTO_NONE:      /* RFC 2460 */
1179                                 /*
1180                                  * Packet ends here, and IPv6 header has
1181                                  * already been pulled up. If ip6e_len!=0
1182                                  * then octets must be ignored.
1183                                  */
1184                                 ulp = ip; /* non-NULL to get out of loop. */
1185                                 break;
1186
1187                         case IPPROTO_OSPFIGP:
1188                                 /* XXX OSPF header check? */
1189                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1190                                 break;
1191
1192                         case IPPROTO_PIM:
1193                                 /* XXX PIM header check? */
1194                                 PULLUP_TO(hlen, ulp, struct pim);
1195                                 break;
1196
1197                         case IPPROTO_CARP:
1198                                 PULLUP_TO(hlen, ulp, struct carp_header);
1199                                 if (((struct carp_header *)ulp)->carp_version !=
1200                                     CARP_VERSION)
1201                                         return (IP_FW_DENY);
1202                                 if (((struct carp_header *)ulp)->carp_type !=
1203                                     CARP_ADVERTISEMENT)
1204                                         return (IP_FW_DENY);
1205                                 break;
1206
1207                         case IPPROTO_IPV6:      /* RFC 2893 */
1208                                 PULLUP_TO(hlen, ulp, struct ip6_hdr);
1209                                 break;
1210
1211                         case IPPROTO_IPV4:      /* RFC 2893 */
1212                                 PULLUP_TO(hlen, ulp, struct ip);
1213                                 break;
1214
1215                         default:
1216                                 if (V_fw_verbose)
1217                                         printf("IPFW2: IPV6 - Unknown "
1218                                             "Extension Header(%d), ext_hd=%x\n",
1219                                              proto, ext_hd);
1220                                 if (V_fw_deny_unknown_exthdrs)
1221                                     return (IP_FW_DENY);
1222                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1223                                 break;
1224                         } /*switch */
1225                 }
1226                 ip = mtod(m, struct ip *);
1227                 ip6 = (struct ip6_hdr *)ip;
1228                 args->f_id.src_ip6 = ip6->ip6_src;
1229                 args->f_id.dst_ip6 = ip6->ip6_dst;
1230                 args->f_id.src_ip = 0;
1231                 args->f_id.dst_ip = 0;
1232                 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1233         } else if (pktlen >= sizeof(struct ip) &&
1234             (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1235                 is_ipv4 = 1;
1236                 hlen = ip->ip_hl << 2;
1237                 args->f_id.addr_type = 4;
1238
1239                 /*
1240                  * Collect parameters into local variables for faster matching.
1241                  */
1242                 proto = ip->ip_p;
1243                 src_ip = ip->ip_src;
1244                 dst_ip = ip->ip_dst;
1245                 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1246                 iplen = ntohs(ip->ip_len);
1247                 pktlen = iplen < pktlen ? iplen : pktlen;
1248
1249                 if (offset == 0) {
1250                         switch (proto) {
1251                         case IPPROTO_TCP:
1252                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1253                                 dst_port = TCP(ulp)->th_dport;
1254                                 src_port = TCP(ulp)->th_sport;
1255                                 /* save flags for dynamic rules */
1256                                 args->f_id._flags = TCP(ulp)->th_flags;
1257                                 break;
1258
1259                         case IPPROTO_SCTP:
1260                                 PULLUP_TO(hlen, ulp, struct sctphdr);
1261                                 src_port = SCTP(ulp)->src_port;
1262                                 dst_port = SCTP(ulp)->dest_port;
1263                                 break;
1264
1265                         case IPPROTO_UDP:
1266                                 PULLUP_TO(hlen, ulp, struct udphdr);
1267                                 dst_port = UDP(ulp)->uh_dport;
1268                                 src_port = UDP(ulp)->uh_sport;
1269                                 break;
1270
1271                         case IPPROTO_ICMP:
1272                                 PULLUP_TO(hlen, ulp, struct icmphdr);
1273                                 //args->f_id.flags = ICMP(ulp)->icmp_type;
1274                                 break;
1275
1276                         default:
1277                                 break;
1278                         }
1279                 }
1280
1281                 ip = mtod(m, struct ip *);
1282                 args->f_id.src_ip = ntohl(src_ip.s_addr);
1283                 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1284         }
1285 #undef PULLUP_TO
1286         if (proto) { /* we may have port numbers, store them */
1287                 args->f_id.proto = proto;
1288                 args->f_id.src_port = src_port = ntohs(src_port);
1289                 args->f_id.dst_port = dst_port = ntohs(dst_port);
1290         }
1291
1292         IPFW_PF_RLOCK(chain);
1293         if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1294                 IPFW_PF_RUNLOCK(chain);
1295                 return (IP_FW_PASS);    /* accept */
1296         }
1297         if (args->rule.slot) {
1298                 /*
1299                  * Packet has already been tagged as a result of a previous
1300                  * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1301                  * REASS, NETGRAPH, DIVERT/TEE...)
1302                  * Validate the slot and continue from the next one
1303                  * if still present, otherwise do a lookup.
1304                  */
1305                 f_pos = (args->rule.chain_id == chain->id) ?
1306                     args->rule.slot :
1307                     ipfw_find_rule(chain, args->rule.rulenum,
1308                         args->rule.rule_id);
1309         } else {
1310                 f_pos = 0;
1311         }
1312
1313         /*
1314          * Now scan the rules, and parse microinstructions for each rule.
1315          * We have two nested loops and an inner switch. Sometimes we
1316          * need to break out of one or both loops, or re-enter one of
1317          * the loops with updated variables. Loop variables are:
1318          *
1319          *      f_pos (outer loop) points to the current rule.
1320          *              On output it points to the matching rule.
1321          *      done (outer loop) is used as a flag to break the loop.
1322          *      l (inner loop)  residual length of current rule.
1323          *              cmd points to the current microinstruction.
1324          *
1325          * We break the inner loop by setting l=0 and possibly
1326          * cmdlen=0 if we don't want to advance cmd.
1327          * We break the outer loop by setting done=1
1328          * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1329          * as needed.
1330          */
1331         for (; f_pos < chain->n_rules; f_pos++) {
1332                 ipfw_insn *cmd;
1333                 uint32_t tablearg = 0;
1334                 int l, cmdlen, skip_or; /* skip rest of OR block */
1335                 struct ip_fw *f;
1336
1337                 f = chain->map[f_pos];
1338                 if (V_set_disable & (1 << f->set) )
1339                         continue;
1340
1341                 skip_or = 0;
1342                 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1343                     l -= cmdlen, cmd += cmdlen) {
1344                         int match;
1345
1346                         /*
1347                          * check_body is a jump target used when we find a
1348                          * CHECK_STATE, and need to jump to the body of
1349                          * the target rule.
1350                          */
1351
1352 /* check_body: */
1353                         cmdlen = F_LEN(cmd);
1354                         /*
1355                          * An OR block (insn_1 || .. || insn_n) has the
1356                          * F_OR bit set in all but the last instruction.
1357                          * The first match will set "skip_or", and cause
1358                          * the following instructions to be skipped until
1359                          * past the one with the F_OR bit clear.
1360                          */
1361                         if (skip_or) {          /* skip this instruction */
1362                                 if ((cmd->len & F_OR) == 0)
1363                                         skip_or = 0;    /* next one is good */
1364                                 continue;
1365                         }
1366                         match = 0; /* set to 1 if we succeed */
1367
1368                         switch (cmd->opcode) {
1369                         /*
1370                          * The first set of opcodes compares the packet's
1371                          * fields with some pattern, setting 'match' if a
1372                          * match is found. At the end of the loop there is
1373                          * logic to deal with F_NOT and F_OR flags associated
1374                          * with the opcode.
1375                          */
1376                         case O_NOP:
1377                                 match = 1;
1378                                 break;
1379
1380                         case O_FORWARD_MAC:
1381                                 printf("ipfw: opcode %d unimplemented\n",
1382                                     cmd->opcode);
1383                                 break;
1384
1385                         case O_GID:
1386                         case O_UID:
1387                         case O_JAIL:
1388                                 /*
1389                                  * We only check offset == 0 && proto != 0,
1390                                  * as this ensures that we have a
1391                                  * packet with the ports info.
1392                                  */
1393                                 if (offset != 0)
1394                                         break;
1395                                 if (proto == IPPROTO_TCP ||
1396                                     proto == IPPROTO_UDP)
1397                                         match = check_uidgid(
1398                                                     (ipfw_insn_u32 *)cmd,
1399                                                     args, &ucred_lookup,
1400 #ifdef __FreeBSD__
1401                                                     &ucred_cache);
1402 #else
1403                                                     (void *)&ucred_cache);
1404 #endif
1405                                 break;
1406
1407                         case O_RECV:
1408                                 match = iface_match(m->m_pkthdr.rcvif,
1409                                     (ipfw_insn_if *)cmd, chain, &tablearg);
1410                                 break;
1411
1412                         case O_XMIT:
1413                                 match = iface_match(oif, (ipfw_insn_if *)cmd,
1414                                     chain, &tablearg);
1415                                 break;
1416
1417                         case O_VIA:
1418                                 match = iface_match(oif ? oif :
1419                                     m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1420                                     chain, &tablearg);
1421                                 break;
1422
1423                         case O_MACADDR2:
1424                                 if (args->eh != NULL) { /* have MAC header */
1425                                         u_int32_t *want = (u_int32_t *)
1426                                                 ((ipfw_insn_mac *)cmd)->addr;
1427                                         u_int32_t *mask = (u_int32_t *)
1428                                                 ((ipfw_insn_mac *)cmd)->mask;
1429                                         u_int32_t *hdr = (u_int32_t *)args->eh;
1430
1431                                         match =
1432                                             ( want[0] == (hdr[0] & mask[0]) &&
1433                                               want[1] == (hdr[1] & mask[1]) &&
1434                                               want[2] == (hdr[2] & mask[2]) );
1435                                 }
1436                                 break;
1437
1438                         case O_MAC_TYPE:
1439                                 if (args->eh != NULL) {
1440                                         u_int16_t *p =
1441                                             ((ipfw_insn_u16 *)cmd)->ports;
1442                                         int i;
1443
1444                                         for (i = cmdlen - 1; !match && i>0;
1445                                             i--, p += 2)
1446                                                 match = (etype >= p[0] &&
1447                                                     etype <= p[1]);
1448                                 }
1449                                 break;
1450
1451                         case O_FRAG:
1452                                 match = (offset != 0);
1453                                 break;
1454
1455                         case O_IN:      /* "out" is "not in" */
1456                                 match = (oif == NULL);
1457                                 break;
1458
1459                         case O_LAYER2:
1460                                 match = (args->eh != NULL);
1461                                 break;
1462
1463                         case O_DIVERTED:
1464                             {
1465                                 /* For diverted packets, args->rule.info
1466                                  * contains the divert port (in host format)
1467                                  * reason and direction.
1468                                  */
1469                                 uint32_t i = args->rule.info;
1470                                 match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1471                                     cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1472                             }
1473                                 break;
1474
1475                         case O_PROTO:
1476                                 /*
1477                                  * We do not allow an arg of 0 so the
1478                                  * check of "proto" only suffices.
1479                                  */
1480                                 match = (proto == cmd->arg1);
1481                                 break;
1482
1483                         case O_IP_SRC:
1484                                 match = is_ipv4 &&
1485                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1486                                     src_ip.s_addr);
1487                                 break;
1488
1489                         case O_IP_SRC_LOOKUP:
1490                         case O_IP_DST_LOOKUP:
1491                                 if (is_ipv4) {
1492                                     uint32_t key =
1493                                         (cmd->opcode == O_IP_DST_LOOKUP) ?
1494                                             dst_ip.s_addr : src_ip.s_addr;
1495                                     uint32_t v = 0;
1496
1497                                     if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1498                                         /* generic lookup. The key must be
1499                                          * in 32bit big-endian format.
1500                                          */
1501                                         v = ((ipfw_insn_u32 *)cmd)->d[1];
1502                                         if (v == 0)
1503                                             key = dst_ip.s_addr;
1504                                         else if (v == 1)
1505                                             key = src_ip.s_addr;
1506                                         else if (v == 6) /* dscp */
1507                                             key = (ip->ip_tos >> 2) & 0x3f;
1508                                         else if (offset != 0)
1509                                             break;
1510                                         else if (proto != IPPROTO_TCP &&
1511                                                 proto != IPPROTO_UDP)
1512                                             break;
1513                                         else if (v == 2)
1514                                             key = dst_port;
1515                                         else if (v == 3)
1516                                             key = src_port;
1517 #ifndef USERSPACE
1518                                         else if (v == 4 || v == 5) {
1519                                             check_uidgid(
1520                                                 (ipfw_insn_u32 *)cmd,
1521                                                 args, &ucred_lookup,
1522 #ifdef __FreeBSD__
1523                                                 &ucred_cache);
1524                                             if (v == 4 /* O_UID */)
1525                                                 key = ucred_cache->cr_uid;
1526                                             else if (v == 5 /* O_JAIL */)
1527                                                 key = ucred_cache->cr_prison->pr_id;
1528 #else /* !__FreeBSD__ */
1529                                                 (void *)&ucred_cache);
1530                                             if (v ==4 /* O_UID */)
1531                                                 key = ucred_cache.uid;
1532                                             else if (v == 5 /* O_JAIL */)
1533                                                 key = ucred_cache.xid;
1534 #endif /* !__FreeBSD__ */
1535                                         } else
1536 #endif /* !USERSPACE */
1537                                             break;
1538                                     }
1539                                     match = ipfw_lookup_table(chain,
1540                                         cmd->arg1, key, &v);
1541                                     if (!match)
1542                                         break;
1543                                     if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1544                                         match =
1545                                             ((ipfw_insn_u32 *)cmd)->d[0] == v;
1546                                     else
1547                                         tablearg = v;
1548                                 } else if (is_ipv6) {
1549                                         uint32_t v = 0;
1550                                         void *pkey = (cmd->opcode == O_IP_DST_LOOKUP) ?
1551                                                 &args->f_id.dst_ip6: &args->f_id.src_ip6;
1552                                         match = ipfw_lookup_table_extended(chain,
1553                                                         cmd->arg1,
1554                                                         sizeof(struct in6_addr),
1555                                                         pkey, &v);
1556                                         if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1557                                                 match = ((ipfw_insn_u32 *)cmd)->d[0] == v;
1558                                         if (match)
1559                                                 tablearg = v;
1560                                 }
1561                                 break;
1562
1563                         case O_IP_FLOW_LOOKUP:
1564                                 {
1565                                         uint32_t v = 0;
1566                                         match = ipfw_lookup_table_extended(chain,
1567                                             cmd->arg1, 0, &args->f_id, &v);
1568                                         if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1569                                                 match = ((ipfw_insn_u32 *)cmd)->d[0] == v;
1570                                         if (match)
1571                                                 tablearg = v;
1572                                 }
1573                                 break;
1574                         case O_IP_SRC_MASK:
1575                         case O_IP_DST_MASK:
1576                                 if (is_ipv4) {
1577                                     uint32_t a =
1578                                         (cmd->opcode == O_IP_DST_MASK) ?
1579                                             dst_ip.s_addr : src_ip.s_addr;
1580                                     uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
1581                                     int i = cmdlen-1;
1582
1583                                     for (; !match && i>0; i-= 2, p+= 2)
1584                                         match = (p[0] == (a & p[1]));
1585                                 }
1586                                 break;
1587
1588                         case O_IP_SRC_ME:
1589                                 if (is_ipv4) {
1590                                         struct ifnet *tif;
1591
1592                                         INADDR_TO_IFP(src_ip, tif);
1593                                         match = (tif != NULL);
1594                                         break;
1595                                 }
1596 #ifdef INET6
1597                                 /* FALLTHROUGH */
1598                         case O_IP6_SRC_ME:
1599                                 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6);
1600 #endif
1601                                 break;
1602
1603                         case O_IP_DST_SET:
1604                         case O_IP_SRC_SET:
1605                                 if (is_ipv4) {
1606                                         u_int32_t *d = (u_int32_t *)(cmd+1);
1607                                         u_int32_t addr =
1608                                             cmd->opcode == O_IP_DST_SET ?
1609                                                 args->f_id.dst_ip :
1610                                                 args->f_id.src_ip;
1611
1612                                             if (addr < d[0])
1613                                                     break;
1614                                             addr -= d[0]; /* subtract base */
1615                                             match = (addr < cmd->arg1) &&
1616                                                 ( d[ 1 + (addr>>5)] &
1617                                                   (1<<(addr & 0x1f)) );
1618                                 }
1619                                 break;
1620
1621                         case O_IP_DST:
1622                                 match = is_ipv4 &&
1623                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1624                                     dst_ip.s_addr);
1625                                 break;
1626
1627                         case O_IP_DST_ME:
1628                                 if (is_ipv4) {
1629                                         struct ifnet *tif;
1630
1631                                         INADDR_TO_IFP(dst_ip, tif);
1632                                         match = (tif != NULL);
1633                                         break;
1634                                 }
1635 #ifdef INET6
1636                                 /* FALLTHROUGH */
1637                         case O_IP6_DST_ME:
1638                                 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6);
1639 #endif
1640                                 break;
1641
1642
1643                         case O_IP_SRCPORT:
1644                         case O_IP_DSTPORT:
1645                                 /*
1646                                  * offset == 0 && proto != 0 is enough
1647                                  * to guarantee that we have a
1648                                  * packet with port info.
1649                                  */
1650                                 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1651                                     && offset == 0) {
1652                                         u_int16_t x =
1653                                             (cmd->opcode == O_IP_SRCPORT) ?
1654                                                 src_port : dst_port ;
1655                                         u_int16_t *p =
1656                                             ((ipfw_insn_u16 *)cmd)->ports;
1657                                         int i;
1658
1659                                         for (i = cmdlen - 1; !match && i>0;
1660                                             i--, p += 2)
1661                                                 match = (x>=p[0] && x<=p[1]);
1662                                 }
1663                                 break;
1664
1665                         case O_ICMPTYPE:
1666                                 match = (offset == 0 && proto==IPPROTO_ICMP &&
1667                                     icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
1668                                 break;
1669
1670 #ifdef INET6
1671                         case O_ICMP6TYPE:
1672                                 match = is_ipv6 && offset == 0 &&
1673                                     proto==IPPROTO_ICMPV6 &&
1674                                     icmp6type_match(
1675                                         ICMP6(ulp)->icmp6_type,
1676                                         (ipfw_insn_u32 *)cmd);
1677                                 break;
1678 #endif /* INET6 */
1679
1680                         case O_IPOPT:
1681                                 match = (is_ipv4 &&
1682                                     ipopts_match(ip, cmd) );
1683                                 break;
1684
1685                         case O_IPVER:
1686                                 match = (is_ipv4 &&
1687                                     cmd->arg1 == ip->ip_v);
1688                                 break;
1689
1690                         case O_IPID:
1691                         case O_IPLEN:
1692                         case O_IPTTL:
1693                                 if (is_ipv4) {  /* only for IP packets */
1694                                     uint16_t x;
1695                                     uint16_t *p;
1696                                     int i;
1697
1698                                     if (cmd->opcode == O_IPLEN)
1699                                         x = iplen;
1700                                     else if (cmd->opcode == O_IPTTL)
1701                                         x = ip->ip_ttl;
1702                                     else /* must be IPID */
1703                                         x = ntohs(ip->ip_id);
1704                                     if (cmdlen == 1) {
1705                                         match = (cmd->arg1 == x);
1706                                         break;
1707                                     }
1708                                     /* otherwise we have ranges */
1709                                     p = ((ipfw_insn_u16 *)cmd)->ports;
1710                                     i = cmdlen - 1;
1711                                     for (; !match && i>0; i--, p += 2)
1712                                         match = (x >= p[0] && x <= p[1]);
1713                                 }
1714                                 break;
1715
1716                         case O_IPPRECEDENCE:
1717                                 match = (is_ipv4 &&
1718                                     (cmd->arg1 == (ip->ip_tos & 0xe0)) );
1719                                 break;
1720
1721                         case O_IPTOS:
1722                                 match = (is_ipv4 &&
1723                                     flags_match(cmd, ip->ip_tos));
1724                                 break;
1725
1726                         case O_DSCP:
1727                             {
1728                                 uint32_t *p;
1729                                 uint16_t x;
1730
1731                                 p = ((ipfw_insn_u32 *)cmd)->d;
1732
1733                                 if (is_ipv4)
1734                                         x = ip->ip_tos >> 2;
1735                                 else if (is_ipv6) {
1736                                         uint8_t *v;
1737                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
1738                                         x = (*v & 0x0F) << 2;
1739                                         v++;
1740                                         x |= *v >> 6;
1741                                 } else
1742                                         break;
1743
1744                                 /* DSCP bitmask is stored as low_u32 high_u32 */
1745                                 if (x > 32)
1746                                         match = *(p + 1) & (1 << (x - 32));
1747                                 else
1748                                         match = *p & (1 << x);
1749                             }
1750                                 break;
1751
1752                         case O_TCPDATALEN:
1753                                 if (proto == IPPROTO_TCP && offset == 0) {
1754                                     struct tcphdr *tcp;
1755                                     uint16_t x;
1756                                     uint16_t *p;
1757                                     int i;
1758
1759                                     tcp = TCP(ulp);
1760                                     x = iplen -
1761                                         ((ip->ip_hl + tcp->th_off) << 2);
1762                                     if (cmdlen == 1) {
1763                                         match = (cmd->arg1 == x);
1764                                         break;
1765                                     }
1766                                     /* otherwise we have ranges */
1767                                     p = ((ipfw_insn_u16 *)cmd)->ports;
1768                                     i = cmdlen - 1;
1769                                     for (; !match && i>0; i--, p += 2)
1770                                         match = (x >= p[0] && x <= p[1]);
1771                                 }
1772                                 break;
1773
1774                         case O_TCPFLAGS:
1775                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1776                                     flags_match(cmd, TCP(ulp)->th_flags));
1777                                 break;
1778
1779                         case O_TCPOPTS:
1780                                 if (proto == IPPROTO_TCP && offset == 0 && ulp){
1781                                         PULLUP_LEN(hlen, ulp,
1782                                             (TCP(ulp)->th_off << 2));
1783                                         match = tcpopts_match(TCP(ulp), cmd);
1784                                 }
1785                                 break;
1786
1787                         case O_TCPSEQ:
1788                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1789                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
1790                                         TCP(ulp)->th_seq);
1791                                 break;
1792
1793                         case O_TCPACK:
1794                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1795                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
1796                                         TCP(ulp)->th_ack);
1797                                 break;
1798
1799                         case O_TCPWIN:
1800                                 if (proto == IPPROTO_TCP && offset == 0) {
1801                                     uint16_t x;
1802                                     uint16_t *p;
1803                                     int i;
1804
1805                                     x = ntohs(TCP(ulp)->th_win);
1806                                     if (cmdlen == 1) {
1807                                         match = (cmd->arg1 == x);
1808                                         break;
1809                                     }
1810                                     /* Otherwise we have ranges. */
1811                                     p = ((ipfw_insn_u16 *)cmd)->ports;
1812                                     i = cmdlen - 1;
1813                                     for (; !match && i > 0; i--, p += 2)
1814                                         match = (x >= p[0] && x <= p[1]);
1815                                 }
1816                                 break;
1817
1818                         case O_ESTAB:
1819                                 /* reject packets which have SYN only */
1820                                 /* XXX should i also check for TH_ACK ? */
1821                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1822                                     (TCP(ulp)->th_flags &
1823                                      (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1824                                 break;
1825
1826                         case O_ALTQ: {
1827                                 struct pf_mtag *at;
1828                                 struct m_tag *mtag;
1829                                 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
1830
1831                                 /*
1832                                  * ALTQ uses mbuf tags from another
1833                                  * packet filtering system - pf(4).
1834                                  * We allocate a tag in its format
1835                                  * and fill it in, pretending to be pf(4).
1836                                  */
1837                                 match = 1;
1838                                 at = pf_find_mtag(m);
1839                                 if (at != NULL && at->qid != 0)
1840                                         break;
1841                                 mtag = m_tag_get(PACKET_TAG_PF,
1842                                     sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
1843                                 if (mtag == NULL) {
1844                                         /*
1845                                          * Let the packet fall back to the
1846                                          * default ALTQ.
1847                                          */
1848                                         break;
1849                                 }
1850                                 m_tag_prepend(m, mtag);
1851                                 at = (struct pf_mtag *)(mtag + 1);
1852                                 at->qid = altq->qid;
1853                                 at->hdr = ip;
1854                                 break;
1855                         }
1856
1857                         case O_LOG:
1858                                 ipfw_log(chain, f, hlen, args, m,
1859                                     oif, offset | ip6f_mf, tablearg, ip);
1860                                 match = 1;
1861                                 break;
1862
1863                         case O_PROB:
1864                                 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
1865                                 break;
1866
1867                         case O_VERREVPATH:
1868                                 /* Outgoing packets automatically pass/match */
1869                                 match = ((oif != NULL) ||
1870                                     (m->m_pkthdr.rcvif == NULL) ||
1871                                     (
1872 #ifdef INET6
1873                                     is_ipv6 ?
1874                                         verify_path6(&(args->f_id.src_ip6),
1875                                             m->m_pkthdr.rcvif, args->f_id.fib) :
1876 #endif
1877                                     verify_path(src_ip, m->m_pkthdr.rcvif,
1878                                         args->f_id.fib)));
1879                                 break;
1880
1881                         case O_VERSRCREACH:
1882                                 /* Outgoing packets automatically pass/match */
1883                                 match = (hlen > 0 && ((oif != NULL) ||
1884 #ifdef INET6
1885                                     is_ipv6 ?
1886                                         verify_path6(&(args->f_id.src_ip6),
1887                                             NULL, args->f_id.fib) :
1888 #endif
1889                                     verify_path(src_ip, NULL, args->f_id.fib)));
1890                                 break;
1891
1892                         case O_ANTISPOOF:
1893                                 /* Outgoing packets automatically pass/match */
1894                                 if (oif == NULL && hlen > 0 &&
1895                                     (  (is_ipv4 && in_localaddr(src_ip))
1896 #ifdef INET6
1897                                     || (is_ipv6 &&
1898                                         in6_localaddr(&(args->f_id.src_ip6)))
1899 #endif
1900                                     ))
1901                                         match =
1902 #ifdef INET6
1903                                             is_ipv6 ? verify_path6(
1904                                                 &(args->f_id.src_ip6),
1905                                                 m->m_pkthdr.rcvif,
1906                                                 args->f_id.fib) :
1907 #endif
1908                                             verify_path(src_ip,
1909                                                 m->m_pkthdr.rcvif,
1910                                                 args->f_id.fib);
1911                                 else
1912                                         match = 1;
1913                                 break;
1914
1915                         case O_IPSEC:
1916 #ifdef IPSEC
1917                                 match = (m_tag_find(m,
1918                                     PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
1919 #endif
1920                                 /* otherwise no match */
1921                                 break;
1922
1923 #ifdef INET6
1924                         case O_IP6_SRC:
1925                                 match = is_ipv6 &&
1926                                     IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
1927                                     &((ipfw_insn_ip6 *)cmd)->addr6);
1928                                 break;
1929
1930                         case O_IP6_DST:
1931                                 match = is_ipv6 &&
1932                                 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
1933                                     &((ipfw_insn_ip6 *)cmd)->addr6);
1934                                 break;
1935                         case O_IP6_SRC_MASK:
1936                         case O_IP6_DST_MASK:
1937                                 if (is_ipv6) {
1938                                         int i = cmdlen - 1;
1939                                         struct in6_addr p;
1940                                         struct in6_addr *d =
1941                                             &((ipfw_insn_ip6 *)cmd)->addr6;
1942
1943                                         for (; !match && i > 0; d += 2,
1944                                             i -= F_INSN_SIZE(struct in6_addr)
1945                                             * 2) {
1946                                                 p = (cmd->opcode ==
1947                                                     O_IP6_SRC_MASK) ?
1948                                                     args->f_id.src_ip6:
1949                                                     args->f_id.dst_ip6;
1950                                                 APPLY_MASK(&p, &d[1]);
1951                                                 match =
1952                                                     IN6_ARE_ADDR_EQUAL(&d[0],
1953                                                     &p);
1954                                         }
1955                                 }
1956                                 break;
1957
1958                         case O_FLOW6ID:
1959                                 match = is_ipv6 &&
1960                                     flow6id_match(args->f_id.flow_id6,
1961                                     (ipfw_insn_u32 *) cmd);
1962                                 break;
1963
1964                         case O_EXT_HDR:
1965                                 match = is_ipv6 &&
1966                                     (ext_hd & ((ipfw_insn *) cmd)->arg1);
1967                                 break;
1968
1969                         case O_IP6:
1970                                 match = is_ipv6;
1971                                 break;
1972 #endif
1973
1974                         case O_IP4:
1975                                 match = is_ipv4;
1976                                 break;
1977
1978                         case O_TAG: {
1979                                 struct m_tag *mtag;
1980                                 uint32_t tag = TARG(cmd->arg1, tag);
1981
1982                                 /* Packet is already tagged with this tag? */
1983                                 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
1984
1985                                 /* We have `untag' action when F_NOT flag is
1986                                  * present. And we must remove this mtag from
1987                                  * mbuf and reset `match' to zero (`match' will
1988                                  * be inversed later).
1989                                  * Otherwise we should allocate new mtag and
1990                                  * push it into mbuf.
1991                                  */
1992                                 if (cmd->len & F_NOT) { /* `untag' action */
1993                                         if (mtag != NULL)
1994                                                 m_tag_delete(m, mtag);
1995                                         match = 0;
1996                                 } else {
1997                                         if (mtag == NULL) {
1998                                                 mtag = m_tag_alloc( MTAG_IPFW,
1999                                                     tag, 0, M_NOWAIT);
2000                                                 if (mtag != NULL)
2001                                                         m_tag_prepend(m, mtag);
2002                                         }
2003                                         match = 1;
2004                                 }
2005                                 break;
2006                         }
2007
2008                         case O_FIB: /* try match the specified fib */
2009                                 if (args->f_id.fib == cmd->arg1)
2010                                         match = 1;
2011                                 break;
2012
2013                         case O_SOCKARG: {
2014 #ifndef USERSPACE       /* not supported in userspace */
2015                                 struct inpcb *inp = args->inp;
2016                                 struct inpcbinfo *pi;
2017
2018                                 if (is_ipv6) /* XXX can we remove this ? */
2019                                         break;
2020
2021                                 if (proto == IPPROTO_TCP)
2022                                         pi = &V_tcbinfo;
2023                                 else if (proto == IPPROTO_UDP)
2024                                         pi = &V_udbinfo;
2025                                 else
2026                                         break;
2027
2028                                 /*
2029                                  * XXXRW: so_user_cookie should almost
2030                                  * certainly be inp_user_cookie?
2031                                  */
2032
2033                                 /* For incomming packet, lookup up the
2034                                 inpcb using the src/dest ip/port tuple */
2035                                 if (inp == NULL) {
2036                                         inp = in_pcblookup(pi,
2037                                                 src_ip, htons(src_port),
2038                                                 dst_ip, htons(dst_port),
2039                                                 INPLOOKUP_RLOCKPCB, NULL);
2040                                         if (inp != NULL) {
2041                                                 tablearg =
2042                                                     inp->inp_socket->so_user_cookie;
2043                                                 if (tablearg)
2044                                                         match = 1;
2045                                                 INP_RUNLOCK(inp);
2046                                         }
2047                                 } else {
2048                                         if (inp->inp_socket) {
2049                                                 tablearg =
2050                                                     inp->inp_socket->so_user_cookie;
2051                                                 if (tablearg)
2052                                                         match = 1;
2053                                         }
2054                                 }
2055 #endif /* !USERSPACE */
2056                                 break;
2057                         }
2058
2059                         case O_TAGGED: {
2060                                 struct m_tag *mtag;
2061                                 uint32_t tag = TARG(cmd->arg1, tag);
2062
2063                                 if (cmdlen == 1) {
2064                                         match = m_tag_locate(m, MTAG_IPFW,
2065                                             tag, NULL) != NULL;
2066                                         break;
2067                                 }
2068
2069                                 /* we have ranges */
2070                                 for (mtag = m_tag_first(m);
2071                                     mtag != NULL && !match;
2072                                     mtag = m_tag_next(m, mtag)) {
2073                                         uint16_t *p;
2074                                         int i;
2075
2076                                         if (mtag->m_tag_cookie != MTAG_IPFW)
2077                                                 continue;
2078
2079                                         p = ((ipfw_insn_u16 *)cmd)->ports;
2080                                         i = cmdlen - 1;
2081                                         for(; !match && i > 0; i--, p += 2)
2082                                                 match =
2083                                                     mtag->m_tag_id >= p[0] &&
2084                                                     mtag->m_tag_id <= p[1];
2085                                 }
2086                                 break;
2087                         }
2088
2089                         /*
2090                          * The second set of opcodes represents 'actions',
2091                          * i.e. the terminal part of a rule once the packet
2092                          * matches all previous patterns.
2093                          * Typically there is only one action for each rule,
2094                          * and the opcode is stored at the end of the rule
2095                          * (but there are exceptions -- see below).
2096                          *
2097                          * In general, here we set retval and terminate the
2098                          * outer loop (would be a 'break 3' in some language,
2099                          * but we need to set l=0, done=1)
2100                          *
2101                          * Exceptions:
2102                          * O_COUNT and O_SKIPTO actions:
2103                          *   instead of terminating, we jump to the next rule
2104                          *   (setting l=0), or to the SKIPTO target (setting
2105                          *   f/f_len, cmd and l as needed), respectively.
2106                          *
2107                          * O_TAG, O_LOG and O_ALTQ action parameters:
2108                          *   perform some action and set match = 1;
2109                          *
2110                          * O_LIMIT and O_KEEP_STATE: these opcodes are
2111                          *   not real 'actions', and are stored right
2112                          *   before the 'action' part of the rule.
2113                          *   These opcodes try to install an entry in the
2114                          *   state tables; if successful, we continue with
2115                          *   the next opcode (match=1; break;), otherwise
2116                          *   the packet must be dropped (set retval,
2117                          *   break loops with l=0, done=1)
2118                          *
2119                          * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2120                          *   cause a lookup of the state table, and a jump
2121                          *   to the 'action' part of the parent rule
2122                          *   if an entry is found, or
2123                          *   (CHECK_STATE only) a jump to the next rule if
2124                          *   the entry is not found.
2125                          *   The result of the lookup is cached so that
2126                          *   further instances of these opcodes become NOPs.
2127                          *   The jump to the next rule is done by setting
2128                          *   l=0, cmdlen=0.
2129                          */
2130                         case O_LIMIT:
2131                         case O_KEEP_STATE:
2132                                 if (ipfw_install_state(chain, f,
2133                                     (ipfw_insn_limit *)cmd, args, tablearg)) {
2134                                         /* error or limit violation */
2135                                         retval = IP_FW_DENY;
2136                                         l = 0;  /* exit inner loop */
2137                                         done = 1; /* exit outer loop */
2138                                 }
2139                                 match = 1;
2140                                 break;
2141
2142                         case O_PROBE_STATE:
2143                         case O_CHECK_STATE:
2144                                 /*
2145                                  * dynamic rules are checked at the first
2146                                  * keep-state or check-state occurrence,
2147                                  * with the result being stored in dyn_dir.
2148                                  * The compiler introduces a PROBE_STATE
2149                                  * instruction for us when we have a
2150                                  * KEEP_STATE (because PROBE_STATE needs
2151                                  * to be run first).
2152                                  */
2153                                 if (dyn_dir == MATCH_UNKNOWN &&
2154                                     (q = ipfw_lookup_dyn_rule(&args->f_id,
2155                                      &dyn_dir, proto == IPPROTO_TCP ?
2156                                         TCP(ulp) : NULL))
2157                                         != NULL) {
2158                                         /*
2159                                          * Found dynamic entry, update stats
2160                                          * and jump to the 'action' part of
2161                                          * the parent rule by setting
2162                                          * f, cmd, l and clearing cmdlen.
2163                                          */
2164                                         IPFW_INC_DYN_COUNTER(q, pktlen);
2165                                         /* XXX we would like to have f_pos
2166                                          * readily accessible in the dynamic
2167                                          * rule, instead of having to
2168                                          * lookup q->rule.
2169                                          */
2170                                         f = q->rule;
2171                                         f_pos = ipfw_find_rule(chain,
2172                                                 f->rulenum, f->id);
2173                                         cmd = ACTION_PTR(f);
2174                                         l = f->cmd_len - f->act_ofs;
2175                                         ipfw_dyn_unlock(q);
2176                                         cmdlen = 0;
2177                                         match = 1;
2178                                         break;
2179                                 }
2180                                 /*
2181                                  * Dynamic entry not found. If CHECK_STATE,
2182                                  * skip to next rule, if PROBE_STATE just
2183                                  * ignore and continue with next opcode.
2184                                  */
2185                                 if (cmd->opcode == O_CHECK_STATE)
2186                                         l = 0;  /* exit inner loop */
2187                                 match = 1;
2188                                 break;
2189
2190                         case O_ACCEPT:
2191                                 retval = 0;     /* accept */
2192                                 l = 0;          /* exit inner loop */
2193                                 done = 1;       /* exit outer loop */
2194                                 break;
2195
2196                         case O_PIPE:
2197                         case O_QUEUE:
2198                                 set_match(args, f_pos, chain);
2199                                 args->rule.info = TARG(cmd->arg1, pipe);
2200                                 if (cmd->opcode == O_PIPE)
2201                                         args->rule.info |= IPFW_IS_PIPE;
2202                                 if (V_fw_one_pass)
2203                                         args->rule.info |= IPFW_ONEPASS;
2204                                 retval = IP_FW_DUMMYNET;
2205                                 l = 0;          /* exit inner loop */
2206                                 done = 1;       /* exit outer loop */
2207                                 break;
2208
2209                         case O_DIVERT:
2210                         case O_TEE:
2211                                 if (args->eh) /* not on layer 2 */
2212                                     break;
2213                                 /* otherwise this is terminal */
2214                                 l = 0;          /* exit inner loop */
2215                                 done = 1;       /* exit outer loop */
2216                                 retval = (cmd->opcode == O_DIVERT) ?
2217                                         IP_FW_DIVERT : IP_FW_TEE;
2218                                 set_match(args, f_pos, chain);
2219                                 args->rule.info = TARG(cmd->arg1, divert);
2220                                 break;
2221
2222                         case O_COUNT:
2223                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2224                                 l = 0;          /* exit inner loop */
2225                                 break;
2226
2227                         case O_SKIPTO:
2228                             IPFW_INC_RULE_COUNTER(f, pktlen);
2229                             f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2230                             /*
2231                              * Skip disabled rules, and re-enter
2232                              * the inner loop with the correct
2233                              * f_pos, f, l and cmd.
2234                              * Also clear cmdlen and skip_or
2235                              */
2236                             for (; f_pos < chain->n_rules - 1 &&
2237                                     (V_set_disable &
2238                                      (1 << chain->map[f_pos]->set));
2239                                     f_pos++)
2240                                 ;
2241                             /* Re-enter the inner loop at the skipto rule. */
2242                             f = chain->map[f_pos];
2243                             l = f->cmd_len;
2244                             cmd = f->cmd;
2245                             match = 1;
2246                             cmdlen = 0;
2247                             skip_or = 0;
2248                             continue;
2249                             break;      /* not reached */
2250
2251                         case O_CALLRETURN: {
2252                                 /*
2253                                  * Implementation of `subroutine' call/return,
2254                                  * in the stack carried in an mbuf tag. This
2255                                  * is different from `skipto' in that any call
2256                                  * address is possible (`skipto' must prevent
2257                                  * backward jumps to avoid endless loops).
2258                                  * We have `return' action when F_NOT flag is
2259                                  * present. The `m_tag_id' field is used as
2260                                  * stack pointer.
2261                                  */
2262                                 struct m_tag *mtag;
2263                                 uint16_t jmpto, *stack;
2264
2265 #define IS_CALL         ((cmd->len & F_NOT) == 0)
2266 #define IS_RETURN       ((cmd->len & F_NOT) != 0)
2267                                 /*
2268                                  * Hand-rolled version of m_tag_locate() with
2269                                  * wildcard `type'.
2270                                  * If not already tagged, allocate new tag.
2271                                  */
2272                                 mtag = m_tag_first(m);
2273                                 while (mtag != NULL) {
2274                                         if (mtag->m_tag_cookie ==
2275                                             MTAG_IPFW_CALL)
2276                                                 break;
2277                                         mtag = m_tag_next(m, mtag);
2278                                 }
2279                                 if (mtag == NULL && IS_CALL) {
2280                                         mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2281                                             IPFW_CALLSTACK_SIZE *
2282                                             sizeof(uint16_t), M_NOWAIT);
2283                                         if (mtag != NULL)
2284                                                 m_tag_prepend(m, mtag);
2285                                 }
2286
2287                                 /*
2288                                  * On error both `call' and `return' just
2289                                  * continue with next rule.
2290                                  */
2291                                 if (IS_RETURN && (mtag == NULL ||
2292                                     mtag->m_tag_id == 0)) {
2293                                         l = 0;          /* exit inner loop */
2294                                         break;
2295                                 }
2296                                 if (IS_CALL && (mtag == NULL ||
2297                                     mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2298                                         printf("ipfw: call stack error, "
2299                                             "go to next rule\n");
2300                                         l = 0;          /* exit inner loop */
2301                                         break;
2302                                 }
2303
2304                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2305                                 stack = (uint16_t *)(mtag + 1);
2306
2307                                 /*
2308                                  * The `call' action may use cached f_pos
2309                                  * (in f->next_rule), whose version is written
2310                                  * in f->next_rule.
2311                                  * The `return' action, however, doesn't have
2312                                  * fixed jump address in cmd->arg1 and can't use
2313                                  * cache.
2314                                  */
2315                                 if (IS_CALL) {
2316                                         stack[mtag->m_tag_id] = f->rulenum;
2317                                         mtag->m_tag_id++;
2318                                         f_pos = JUMP(chain, f, cmd->arg1,
2319                                             tablearg, 1);
2320                                 } else {        /* `return' action */
2321                                         mtag->m_tag_id--;
2322                                         jmpto = stack[mtag->m_tag_id] + 1;
2323                                         f_pos = ipfw_find_rule(chain, jmpto, 0);
2324                                 }
2325
2326                                 /*
2327                                  * Skip disabled rules, and re-enter
2328                                  * the inner loop with the correct
2329                                  * f_pos, f, l and cmd.
2330                                  * Also clear cmdlen and skip_or
2331                                  */
2332                                 for (; f_pos < chain->n_rules - 1 &&
2333                                     (V_set_disable &
2334                                     (1 << chain->map[f_pos]->set)); f_pos++)
2335                                         ;
2336                                 /* Re-enter the inner loop at the dest rule. */
2337                                 f = chain->map[f_pos];
2338                                 l = f->cmd_len;
2339                                 cmd = f->cmd;
2340                                 cmdlen = 0;
2341                                 skip_or = 0;
2342                                 continue;
2343                                 break;  /* NOTREACHED */
2344                         }
2345 #undef IS_CALL
2346 #undef IS_RETURN
2347
2348                         case O_REJECT:
2349                                 /*
2350                                  * Drop the packet and send a reject notice
2351                                  * if the packet is not ICMP (or is an ICMP
2352                                  * query), and it is not multicast/broadcast.
2353                                  */
2354                                 if (hlen > 0 && is_ipv4 && offset == 0 &&
2355                                     (proto != IPPROTO_ICMP ||
2356                                      is_icmp_query(ICMP(ulp))) &&
2357                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2358                                     !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2359                                         send_reject(args, cmd->arg1, iplen, ip);
2360                                         m = args->m;
2361                                 }
2362                                 /* FALLTHROUGH */
2363 #ifdef INET6
2364                         case O_UNREACH6:
2365                                 if (hlen > 0 && is_ipv6 &&
2366                                     ((offset & IP6F_OFF_MASK) == 0) &&
2367                                     (proto != IPPROTO_ICMPV6 ||
2368                                      (is_icmp6_query(icmp6_type) == 1)) &&
2369                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2370                                     !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2371                                         send_reject6(
2372                                             args, cmd->arg1, hlen,
2373                                             (struct ip6_hdr *)ip);
2374                                         m = args->m;
2375                                 }
2376                                 /* FALLTHROUGH */
2377 #endif
2378                         case O_DENY:
2379                                 retval = IP_FW_DENY;
2380                                 l = 0;          /* exit inner loop */
2381                                 done = 1;       /* exit outer loop */
2382                                 break;
2383
2384                         case O_FORWARD_IP:
2385                                 if (args->eh)   /* not valid on layer2 pkts */
2386                                         break;
2387                                 if (q == NULL || q->rule != f ||
2388                                     dyn_dir == MATCH_FORWARD) {
2389                                     struct sockaddr_in *sa;
2390
2391                                     sa = &(((ipfw_insn_sa *)cmd)->sa);
2392                                     if (sa->sin_addr.s_addr == INADDR_ANY) {
2393 #ifdef INET6
2394                                         /*
2395                                          * We use O_FORWARD_IP opcode for
2396                                          * fwd rule with tablearg, but tables
2397                                          * now support IPv6 addresses. And
2398                                          * when we are inspecting IPv6 packet,
2399                                          * we can use nh6 field from
2400                                          * table_value as next_hop6 address.
2401                                          */
2402                                         if (is_ipv6) {
2403                                                 struct sockaddr_in6 *sa6;
2404
2405                                                 sa6 = args->next_hop6 =
2406                                                     &args->hopstore6;
2407                                                 sa6->sin6_family = AF_INET6;
2408                                                 sa6->sin6_len = sizeof(*sa6);
2409                                                 sa6->sin6_addr = TARG_VAL(
2410                                                     chain, tablearg, nh6);
2411                                                 /*
2412                                                  * Set sin6_scope_id only for
2413                                                  * link-local unicast addresses.
2414                                                  */
2415                                                 if (IN6_IS_ADDR_LINKLOCAL(
2416                                                     &sa6->sin6_addr))
2417                                                         sa6->sin6_scope_id =
2418                                                             TARG_VAL(chain,
2419                                                                 tablearg,
2420                                                                 zoneid);
2421                                         } else
2422 #endif
2423                                         {
2424                                                 sa = args->next_hop =
2425                                                     &args->hopstore;
2426                                                 sa->sin_family = AF_INET;
2427                                                 sa->sin_len = sizeof(*sa);
2428                                                 sa->sin_addr.s_addr = htonl(
2429                                                     TARG_VAL(chain, tablearg,
2430                                                     nh4));
2431                                         }
2432                                     } else {
2433                                         args->next_hop = sa;
2434                                     }
2435                                 }
2436                                 retval = IP_FW_PASS;
2437                                 l = 0;          /* exit inner loop */
2438                                 done = 1;       /* exit outer loop */
2439                                 break;
2440
2441 #ifdef INET6
2442                         case O_FORWARD_IP6:
2443                                 if (args->eh)   /* not valid on layer2 pkts */
2444                                         break;
2445                                 if (q == NULL || q->rule != f ||
2446                                     dyn_dir == MATCH_FORWARD) {
2447                                         struct sockaddr_in6 *sin6;
2448
2449                                         sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2450                                         args->next_hop6 = sin6;
2451                                 }
2452                                 retval = IP_FW_PASS;
2453                                 l = 0;          /* exit inner loop */
2454                                 done = 1;       /* exit outer loop */
2455                                 break;
2456 #endif
2457
2458                         case O_NETGRAPH:
2459                         case O_NGTEE:
2460                                 set_match(args, f_pos, chain);
2461                                 args->rule.info = TARG(cmd->arg1, netgraph);
2462                                 if (V_fw_one_pass)
2463                                         args->rule.info |= IPFW_ONEPASS;
2464                                 retval = (cmd->opcode == O_NETGRAPH) ?
2465                                     IP_FW_NETGRAPH : IP_FW_NGTEE;
2466                                 l = 0;          /* exit inner loop */
2467                                 done = 1;       /* exit outer loop */
2468                                 break;
2469
2470                         case O_SETFIB: {
2471                                 uint32_t fib;
2472
2473                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2474                                 fib = TARG(cmd->arg1, fib) & 0x7FFFF;
2475                                 if (fib >= rt_numfibs)
2476                                         fib = 0;
2477                                 M_SETFIB(m, fib);
2478                                 args->f_id.fib = fib;
2479                                 l = 0;          /* exit inner loop */
2480                                 break;
2481                         }
2482
2483                         case O_SETDSCP: {
2484                                 uint16_t code;
2485
2486                                 code = TARG(cmd->arg1, dscp) & 0x3F;
2487                                 l = 0;          /* exit inner loop */
2488                                 if (is_ipv4) {
2489                                         uint16_t a;
2490
2491                                         a = ip->ip_tos;
2492                                         ip->ip_tos = (code << 2) | (ip->ip_tos & 0x03);
2493                                         a += ntohs(ip->ip_sum) - ip->ip_tos;
2494                                         ip->ip_sum = htons(a);
2495                                 } else if (is_ipv6) {
2496                                         uint8_t *v;
2497
2498                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
2499                                         *v = (*v & 0xF0) | (code >> 2);
2500                                         v++;
2501                                         *v = (*v & 0x3F) | ((code & 0x03) << 6);
2502                                 } else
2503                                         break;
2504
2505                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2506                                 break;
2507                         }
2508
2509                         case O_NAT:
2510                                 l = 0;          /* exit inner loop */
2511                                 done = 1;       /* exit outer loop */
2512                                 if (!IPFW_NAT_LOADED) {
2513                                     retval = IP_FW_DENY;
2514                                     break;
2515                                 }
2516
2517                                 struct cfg_nat *t;
2518                                 int nat_id;
2519
2520                                 set_match(args, f_pos, chain);
2521                                 /* Check if this is 'global' nat rule */
2522                                 if (cmd->arg1 == 0) {
2523                                         retval = ipfw_nat_ptr(args, NULL, m);
2524                                         break;
2525                                 }
2526                                 t = ((ipfw_insn_nat *)cmd)->nat;
2527                                 if (t == NULL) {
2528                                         nat_id = TARG(cmd->arg1, nat);
2529                                         t = (*lookup_nat_ptr)(&chain->nat, nat_id);
2530
2531                                         if (t == NULL) {
2532                                             retval = IP_FW_DENY;
2533                                             break;
2534                                         }
2535                                         if (cmd->arg1 != IP_FW_TARG)
2536                                             ((ipfw_insn_nat *)cmd)->nat = t;
2537                                 }
2538                                 retval = ipfw_nat_ptr(args, t, m);
2539                                 break;
2540
2541                         case O_REASS: {
2542                                 int ip_off;
2543
2544                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2545                                 l = 0;  /* in any case exit inner loop */
2546                                 ip_off = ntohs(ip->ip_off);
2547
2548                                 /* if not fragmented, go to next rule */
2549                                 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
2550                                     break;
2551
2552                                 args->m = m = ip_reass(m);
2553
2554                                 /*
2555                                  * do IP header checksum fixup.
2556                                  */
2557                                 if (m == NULL) { /* fragment got swallowed */
2558                                     retval = IP_FW_DENY;
2559                                 } else { /* good, packet complete */
2560                                     int hlen;
2561
2562                                     ip = mtod(m, struct ip *);
2563                                     hlen = ip->ip_hl << 2;
2564                                     ip->ip_sum = 0;
2565                                     if (hlen == sizeof(struct ip))
2566                                         ip->ip_sum = in_cksum_hdr(ip);
2567                                     else
2568                                         ip->ip_sum = in_cksum(m, hlen);
2569                                     retval = IP_FW_REASS;
2570                                     set_match(args, f_pos, chain);
2571                                 }
2572                                 done = 1;       /* exit outer loop */
2573                                 break;
2574                         }
2575
2576                         default:
2577                                 panic("-- unknown opcode %d\n", cmd->opcode);
2578                         } /* end of switch() on opcodes */
2579                         /*
2580                          * if we get here with l=0, then match is irrelevant.
2581                          */
2582
2583                         if (cmd->len & F_NOT)
2584                                 match = !match;
2585
2586                         if (match) {
2587                                 if (cmd->len & F_OR)
2588                                         skip_or = 1;
2589                         } else {
2590                                 if (!(cmd->len & F_OR)) /* not an OR block, */
2591                                         break;          /* try next rule    */
2592                         }
2593
2594                 }       /* end of inner loop, scan opcodes */
2595 #undef PULLUP_LEN
2596
2597                 if (done)
2598                         break;
2599
2600 /* next_rule:; */       /* try next rule                */
2601
2602         }               /* end of outer for, scan rules */
2603
2604         if (done) {
2605                 struct ip_fw *rule = chain->map[f_pos];
2606                 /* Update statistics */
2607                 IPFW_INC_RULE_COUNTER(rule, pktlen);
2608         } else {
2609                 retval = IP_FW_DENY;
2610                 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2611         }
2612         IPFW_PF_RUNLOCK(chain);
2613 #ifdef __FreeBSD__
2614         if (ucred_cache != NULL)
2615                 crfree(ucred_cache);
2616 #endif
2617         return (retval);
2618
2619 pullup_failed:
2620         if (V_fw_verbose)
2621                 printf("ipfw: pullup failed\n");
2622         return (IP_FW_DENY);
2623 }
2624
2625 /*
2626  * Set maximum number of tables that can be used in given VNET ipfw instance.
2627  */
2628 #ifdef SYSCTL_NODE
2629 static int
2630 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
2631 {
2632         int error;
2633         unsigned int ntables;
2634
2635         ntables = V_fw_tables_max;
2636
2637         error = sysctl_handle_int(oidp, &ntables, 0, req);
2638         /* Read operation or some error */
2639         if ((error != 0) || (req->newptr == NULL))
2640                 return (error);
2641
2642         return (ipfw_resize_tables(&V_layer3_chain, ntables));
2643 }
2644
2645 /*
2646  * Switches table namespace between global and per-set.
2647  */
2648 static int
2649 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
2650 {
2651         int error;
2652         unsigned int sets;
2653
2654         sets = V_fw_tables_sets;
2655
2656         error = sysctl_handle_int(oidp, &sets, 0, req);
2657         /* Read operation or some error */
2658         if ((error != 0) || (req->newptr == NULL))
2659                 return (error);
2660
2661         return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
2662 }
2663 #endif
2664
2665 /*
2666  * Module and VNET glue
2667  */
2668
2669 /*
2670  * Stuff that must be initialised only on boot or module load
2671  */
2672 static int
2673 ipfw_init(void)
2674 {
2675         int error = 0;
2676
2677         /*
2678          * Only print out this stuff the first time around,
2679          * when called from the sysinit code.
2680          */
2681         printf("ipfw2 "
2682 #ifdef INET6
2683                 "(+ipv6) "
2684 #endif
2685                 "initialized, divert %s, nat %s, "
2686                 "default to %s, logging ",
2687 #ifdef IPDIVERT
2688                 "enabled",
2689 #else
2690                 "loadable",
2691 #endif
2692 #ifdef IPFIREWALL_NAT
2693                 "enabled",
2694 #else
2695                 "loadable",
2696 #endif
2697                 default_to_accept ? "accept" : "deny");
2698
2699         /*
2700          * Note: V_xxx variables can be accessed here but the vnet specific
2701          * initializer may not have been called yet for the VIMAGE case.
2702          * Tuneables will have been processed. We will print out values for
2703          * the default vnet.
2704          * XXX This should all be rationalized AFTER 8.0
2705          */
2706         if (V_fw_verbose == 0)
2707                 printf("disabled\n");
2708         else if (V_verbose_limit == 0)
2709                 printf("unlimited\n");
2710         else
2711                 printf("limited to %d packets/entry by default\n",
2712                     V_verbose_limit);
2713
2714         /* Check user-supplied table count for validness */
2715         if (default_fw_tables > IPFW_TABLES_MAX)
2716           default_fw_tables = IPFW_TABLES_MAX;
2717
2718         ipfw_init_sopt_handler();
2719         ipfw_log_bpf(1); /* init */
2720         ipfw_iface_init();
2721         return (error);
2722 }
2723
2724 /*
2725  * Called for the removal of the last instance only on module unload.
2726  */
2727 static void
2728 ipfw_destroy(void)
2729 {
2730
2731         ipfw_iface_destroy();
2732         ipfw_log_bpf(0); /* uninit */
2733         ipfw_destroy_sopt_handler();
2734         printf("IP firewall unloaded\n");
2735 }
2736
2737 /*
2738  * Stuff that must be initialized for every instance
2739  * (including the first of course).
2740  */
2741 static int
2742 vnet_ipfw_init(const void *unused)
2743 {
2744         int error, first;
2745         struct ip_fw *rule = NULL;
2746         struct ip_fw_chain *chain;
2747
2748         chain = &V_layer3_chain;
2749
2750         first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
2751
2752         /* First set up some values that are compile time options */
2753         V_autoinc_step = 100;   /* bounded to 1..1000 in add_rule() */
2754         V_fw_deny_unknown_exthdrs = 1;
2755 #ifdef IPFIREWALL_VERBOSE
2756         V_fw_verbose = 1;
2757 #endif
2758 #ifdef IPFIREWALL_VERBOSE_LIMIT
2759         V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2760 #endif
2761 #ifdef IPFIREWALL_NAT
2762         LIST_INIT(&chain->nat);
2763 #endif
2764
2765         ipfw_init_counters();
2766         /* insert the default rule and create the initial map */
2767         chain->n_rules = 1;
2768         chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_WAITOK | M_ZERO);
2769         rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
2770
2771         /* Set initial number of tables */
2772         V_fw_tables_max = default_fw_tables;
2773         error = ipfw_init_tables(chain, first);
2774         if (error) {
2775                 printf("ipfw2: setting up tables failed\n");
2776                 free(chain->map, M_IPFW);
2777                 free(rule, M_IPFW);
2778                 return (ENOSPC);
2779         }
2780
2781         /* fill and insert the default rule */
2782         rule->act_ofs = 0;
2783         rule->rulenum = IPFW_DEFAULT_RULE;
2784         rule->cmd_len = 1;
2785         rule->set = RESVD_SET;
2786         rule->cmd[0].len = 1;
2787         rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
2788         chain->default_rule = chain->map[0] = rule;
2789         chain->id = rule->id = 1;
2790         /* Pre-calculate rules length for legacy dump format */
2791         chain->static_len = sizeof(struct ip_fw_rule0);
2792
2793         IPFW_LOCK_INIT(chain);
2794         ipfw_dyn_init(chain);
2795 #ifdef LINEAR_SKIPTO
2796         ipfw_init_skipto_cache(chain);
2797 #endif
2798
2799         /* First set up some values that are compile time options */
2800         V_ipfw_vnet_ready = 1;          /* Open for business */
2801
2802         /*
2803          * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
2804          * Even if the latter two fail we still keep the module alive
2805          * because the sockopt and layer2 paths are still useful.
2806          * ipfw[6]_hook return 0 on success, ENOENT on failure,
2807          * so we can ignore the exact return value and just set a flag.
2808          *
2809          * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
2810          * changes in the underlying (per-vnet) variables trigger
2811          * immediate hook()/unhook() calls.
2812          * In layer2 we have the same behaviour, except that V_ether_ipfw
2813          * is checked on each packet because there are no pfil hooks.
2814          */
2815         V_ip_fw_ctl_ptr = ipfw_ctl3;
2816         error = ipfw_attach_hooks(1);
2817         return (error);
2818 }
2819
2820 /*
2821  * Called for the removal of each instance.
2822  */
2823 static int
2824 vnet_ipfw_uninit(const void *unused)
2825 {
2826         struct ip_fw *reap;
2827         struct ip_fw_chain *chain = &V_layer3_chain;
2828         int i, last;
2829
2830         V_ipfw_vnet_ready = 0; /* tell new callers to go away */
2831         /*
2832          * disconnect from ipv4, ipv6, layer2 and sockopt.
2833          * Then grab, release and grab again the WLOCK so we make
2834          * sure the update is propagated and nobody will be in.
2835          */
2836         (void)ipfw_attach_hooks(0 /* detach */);
2837         V_ip_fw_ctl_ptr = NULL;
2838
2839         last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
2840
2841         IPFW_UH_WLOCK(chain);
2842         IPFW_UH_WUNLOCK(chain);
2843         IPFW_UH_WLOCK(chain);
2844
2845         IPFW_WLOCK(chain);
2846         ipfw_dyn_uninit(0);     /* run the callout_drain */
2847         IPFW_WUNLOCK(chain);
2848
2849         reap = NULL;
2850         IPFW_WLOCK(chain);
2851         for (i = 0; i < chain->n_rules; i++)
2852                 ipfw_reap_add(chain, &reap, chain->map[i]);
2853         free(chain->map, M_IPFW);
2854 #ifdef LINEAR_SKIPTO
2855         ipfw_destroy_skipto_cache(chain);
2856 #endif
2857         IPFW_WUNLOCK(chain);
2858         IPFW_UH_WUNLOCK(chain);
2859         ipfw_destroy_tables(chain, last);
2860         if (reap != NULL)
2861                 ipfw_reap_rules(reap);
2862         vnet_ipfw_iface_destroy(chain);
2863         IPFW_LOCK_DESTROY(chain);
2864         ipfw_dyn_uninit(1);     /* free the remaining parts */
2865         ipfw_destroy_counters();
2866         return (0);
2867 }
2868
2869 /*
2870  * Module event handler.
2871  * In general we have the choice of handling most of these events by the
2872  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
2873  * use the SYSINIT handlers as they are more capable of expressing the
2874  * flow of control during module and vnet operations, so this is just
2875  * a skeleton. Note there is no SYSINIT equivalent of the module
2876  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
2877  */
2878 static int
2879 ipfw_modevent(module_t mod, int type, void *unused)
2880 {
2881         int err = 0;
2882
2883         switch (type) {
2884         case MOD_LOAD:
2885                 /* Called once at module load or
2886                  * system boot if compiled in. */
2887                 break;
2888         case MOD_QUIESCE:
2889                 /* Called before unload. May veto unloading. */
2890                 break;
2891         case MOD_UNLOAD:
2892                 /* Called during unload. */
2893                 break;
2894         case MOD_SHUTDOWN:
2895                 /* Called during system shutdown. */
2896                 break;
2897         default:
2898                 err = EOPNOTSUPP;
2899                 break;
2900         }
2901         return err;
2902 }
2903
2904 static moduledata_t ipfwmod = {
2905         "ipfw",
2906         ipfw_modevent,
2907         0
2908 };
2909
2910 /* Define startup order. */
2911 #define IPFW_SI_SUB_FIREWALL    SI_SUB_PROTO_IFATTACHDOMAIN
2912 #define IPFW_MODEVENT_ORDER     (SI_ORDER_ANY - 255) /* On boot slot in here. */
2913 #define IPFW_MODULE_ORDER       (IPFW_MODEVENT_ORDER + 1) /* A little later. */
2914 #define IPFW_VNET_ORDER         (IPFW_MODEVENT_ORDER + 2) /* Later still. */
2915
2916 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
2917 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
2918 MODULE_VERSION(ipfw, 3);
2919 /* should declare some dependencies here */
2920
2921 /*
2922  * Starting up. Done in order after ipfwmod() has been called.
2923  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
2924  */
2925 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2926             ipfw_init, NULL);
2927 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2928             vnet_ipfw_init, NULL);
2929
2930 /*
2931  * Closing up shop. These are done in REVERSE ORDER, but still
2932  * after ipfwmod() has been called. Not called on reboot.
2933  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
2934  * or when the module is unloaded.
2935  */
2936 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2937             ipfw_destroy, NULL);
2938 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2939             vnet_ipfw_uninit, NULL);
2940 /* end of file */