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 #include <netinet/in_fib.h>
  88 #ifdef INET6
  89 #include <netinet6/in6_fib.h>
  90 #include <netinet6/in6_pcb.h>
  91 #include <netinet6/scope6_var.h>
  92 #include <netinet6/ip6_var.h>
  93 #endif
  94
  95 #include <net/if_gre.h> /* for struct grehdr */
  96
  97 #include <netpfil/ipfw/ip_fw_private.h>
  98
  99 #include <machine/in_cksum.h>   /* XXX for in_cksum */
 100
 101 #ifdef MAC
 102 #include <security/mac/mac_framework.h>
 103 #endif
 104
 105 /*
 106  * static variables followed by global ones.
 107  * All ipfw global variables are here.
 108  */
 109
 110 static VNET_DEFINE(int, fw_deny_unknown_exthdrs);
 111 #define V_fw_deny_unknown_exthdrs       VNET(fw_deny_unknown_exthdrs)
 112
 113 static VNET_DEFINE(int, fw_permit_single_frag6) = 1;
 114 #define V_fw_permit_single_frag6        VNET(fw_permit_single_frag6)
 115
 116 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
 117 static int default_to_accept = 1;
 118 #else
 119 static int default_to_accept;
 120 #endif
 121
 122 VNET_DEFINE(int, autoinc_step);
 123 VNET_DEFINE(int, fw_one_pass) = 1;
 124
 125 VNET_DEFINE(unsigned int, fw_tables_max);
 126 VNET_DEFINE(unsigned int, fw_tables_sets) = 0;  /* Don't use set-aware tables */
 127 /* Use 128 tables by default */
 128 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT;
 129
 130 #ifndef LINEAR_SKIPTO
 131 static int jump_fast(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_fast(ch, f, num, targ, back)
 134 #else
 135 static int jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
 136     int tablearg, int jump_backwards);
 137 #define JUMP(ch, f, num, targ, back)    jump_linear(ch, f, num, targ, back)
 138 #endif
 139
 140 /*
 141  * Each rule belongs to one of 32 different sets (0..31).
 142  * The variable set_disable contains one bit per set.
 143  * If the bit is set, all rules in the corresponding set
 144  * are disabled. Set RESVD_SET(31) is reserved for the default rule
 145  * and rules that are not deleted by the flush command,
 146  * and CANNOT be disabled.
 147  * Rules in set RESVD_SET can only be deleted individually.
 148  */
 149 VNET_DEFINE(u_int32_t, set_disable);
 150 #define V_set_disable                   VNET(set_disable)
 151
 152 VNET_DEFINE(int, fw_verbose);
 153 /* counter for ipfw_log(NULL...) */
 154 VNET_DEFINE(u_int64_t, norule_counter);
 155 VNET_DEFINE(int, verbose_limit);
 156
 157 /* layer3_chain contains the list of rules for layer 3 */
 158 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
 159
 160 /* ipfw_vnet_ready controls when we are open for business */
 161 VNET_DEFINE(int, ipfw_vnet_ready) = 0;
 162
 163 VNET_DEFINE(int, ipfw_nat_ready) = 0;
 164
 165 ipfw_nat_t *ipfw_nat_ptr = NULL;
 166 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
 167 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
 168 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
 169 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
 170 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
 171
 172 #ifdef SYSCTL_NODE
 173 uint32_t dummy_def = IPFW_DEFAULT_RULE;
 174 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS);
 175 static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS);
 176
 177 SYSBEGIN(f3)
 178
 179 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 180 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
 181     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
 182     "Only do a single pass through ipfw when using dummynet(4)");
 183 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
 184     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
 185     "Rule number auto-increment step");
 186 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose,
 187     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
 188     "Log matches to ipfw rules");
 189 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
 190     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
 191     "Set upper limit of matches of ipfw rules logged");
 192 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
 193     &dummy_def, 0,
 194     "The default/max possible rule number.");
 195 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
 196     CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU",
 197     "Maximum number of concurrently used tables");
 198 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets,
 199     CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW,
 200     0, 0, sysctl_ipfw_tables_sets, "IU",
 201     "Use per-set namespace for tables");
 202 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
 203     &default_to_accept, 0,
 204     "Make the default rule accept all packets.");
 205 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables);
 206 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count,
 207     CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
 208     "Number of static rules");
 209
 210 #ifdef INET6
 211 SYSCTL_DECL(_net_inet6_ip6);
 212 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 213 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
 214     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
 215     &VNET_NAME(fw_deny_unknown_exthdrs), 0,
 216     "Deny packets with unknown IPv6 Extension Headers");
 217 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
 218     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
 219     &VNET_NAME(fw_permit_single_frag6), 0,
 220     "Permit single packet IPv6 fragments");
 221 #endif /* INET6 */
 222
 223 SYSEND
 224
 225 #endif /* SYSCTL_NODE */
 226
 227
 228 /*
 229  * Some macros used in the various matching options.
 230  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
 231  * Other macros just cast void * into the appropriate type
 232  */
 233 #define L3HDR(T, ip)    ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
 234 #define TCP(p)          ((struct tcphdr *)(p))
 235 #define SCTP(p)         ((struct sctphdr *)(p))
 236 #define UDP(p)          ((struct udphdr *)(p))
 237 #define ICMP(p)         ((struct icmphdr *)(p))
 238 #define ICMP6(p)        ((struct icmp6_hdr *)(p))
 239
 240 static __inline int
 241 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
 242 {
 243         int type = icmp->icmp_type;
 244
 245         return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
 246 }
 247
 248 #define TT      ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
 249     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
 250
 251 static int
 252 is_icmp_query(struct icmphdr *icmp)
 253 {
 254         int type = icmp->icmp_type;
 255
 256         return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
 257 }
 258 #undef TT
 259
 260 /*
 261  * The following checks use two arrays of 8 or 16 bits to store the
 262  * bits that we want set or clear, respectively. They are in the
 263  * low and high half of cmd->arg1 or cmd->d[0].
 264  *
 265  * We scan options and store the bits we find set. We succeed if
 266  *
 267  *      (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
 268  *
 269  * The code is sometimes optimized not to store additional variables.
 270  */
 271
 272 static int
 273 flags_match(ipfw_insn *cmd, u_int8_t bits)
 274 {
 275         u_char want_clear;
 276         bits = ~bits;
 277
 278         if ( ((cmd->arg1 & 0xff) & bits) != 0)
 279                 return 0; /* some bits we want set were clear */
 280         want_clear = (cmd->arg1 >> 8) & 0xff;
 281         if ( (want_clear & bits) != want_clear)
 282                 return 0; /* some bits we want clear were set */
 283         return 1;
 284 }
 285
 286 static int
 287 ipopts_match(struct ip *ip, ipfw_insn *cmd)
 288 {
 289         int optlen, bits = 0;
 290         u_char *cp = (u_char *)(ip + 1);
 291         int x = (ip->ip_hl << 2) - sizeof (struct ip);
 292
 293         for (; x > 0; x -= optlen, cp += optlen) {
 294                 int opt = cp[IPOPT_OPTVAL];
 295
 296                 if (opt == IPOPT_EOL)
 297                         break;
 298                 if (opt == IPOPT_NOP)
 299                         optlen = 1;
 300                 else {
 301                         optlen = cp[IPOPT_OLEN];
 302                         if (optlen <= 0 || optlen > x)
 303                                 return 0; /* invalid or truncated */
 304                 }
 305                 switch (opt) {
 306
 307                 default:
 308                         break;
 309
 310                 case IPOPT_LSRR:
 311                         bits |= IP_FW_IPOPT_LSRR;
 312                         break;
 313
 314                 case IPOPT_SSRR:
 315                         bits |= IP_FW_IPOPT_SSRR;
 316                         break;
 317
 318                 case IPOPT_RR:
 319                         bits |= IP_FW_IPOPT_RR;
 320                         break;
 321
 322                 case IPOPT_TS:
 323                         bits |= IP_FW_IPOPT_TS;
 324                         break;
 325                 }
 326         }
 327         return (flags_match(cmd, bits));
 328 }
 329
 330 static int
 331 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
 332 {
 333         int optlen, bits = 0;
 334         u_char *cp = (u_char *)(tcp + 1);
 335         int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
 336
 337         for (; x > 0; x -= optlen, cp += optlen) {
 338                 int opt = cp[0];
 339                 if (opt == TCPOPT_EOL)
 340                         break;
 341                 if (opt == TCPOPT_NOP)
 342                         optlen = 1;
 343                 else {
 344                         optlen = cp[1];
 345                         if (optlen <= 0)
 346                                 break;
 347                 }
 348
 349                 switch (opt) {
 350
 351                 default:
 352                         break;
 353
 354                 case TCPOPT_MAXSEG:
 355                         bits |= IP_FW_TCPOPT_MSS;
 356                         break;
 357
 358                 case TCPOPT_WINDOW:
 359                         bits |= IP_FW_TCPOPT_WINDOW;
 360                         break;
 361
 362                 case TCPOPT_SACK_PERMITTED:
 363                 case TCPOPT_SACK:
 364                         bits |= IP_FW_TCPOPT_SACK;
 365                         break;
 366
 367                 case TCPOPT_TIMESTAMP:
 368                         bits |= IP_FW_TCPOPT_TS;
 369                         break;
 370
 371                 }
 372         }
 373         return (flags_match(cmd, bits));
 374 }
 375
 376 static int
 377 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain,
 378     uint32_t *tablearg)
 379 {
 380
 381         if (ifp == NULL)        /* no iface with this packet, match fails */
 382                 return (0);
 383
 384         /* Check by name or by IP address */
 385         if (cmd->name[0] != '\0') { /* match by name */
 386                 if (cmd->name[0] == '\1') /* use tablearg to match */
 387                         return ipfw_lookup_table(chain, cmd->p.kidx, 0,
 388                             &ifp->if_index, tablearg);
 389                 /* Check name */
 390                 if (cmd->p.glob) {
 391                         if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
 392                                 return(1);
 393                 } else {
 394                         if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
 395                                 return(1);
 396                 }
 397         } else {
 398 #if !defined(USERSPACE) && defined(__FreeBSD__) /* and OSX too ? */
 399                 struct ifaddr *ia;
 400
 401                 if_addr_rlock(ifp);
 402                 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
 403                         if (ia->ifa_addr->sa_family != AF_INET)
 404                                 continue;
 405                         if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
 406                             (ia->ifa_addr))->sin_addr.s_addr) {
 407                                 if_addr_runlock(ifp);
 408                                 return(1);      /* match */
 409                         }
 410                 }
 411                 if_addr_runlock(ifp);
 412 #endif /* __FreeBSD__ */
 413         }
 414         return(0);      /* no match, fail ... */
 415 }
 416
 417 /*
 418  * The verify_path function checks if a route to the src exists and
 419  * if it is reachable via ifp (when provided).
 420  *
 421  * The 'verrevpath' option checks that the interface that an IP packet
 422  * arrives on is the same interface that traffic destined for the
 423  * packet's source address would be routed out of.
 424  * The 'versrcreach' option just checks that the source address is
 425  * reachable via any route (except default) in the routing table.
 426  * These two are a measure to block forged packets. This is also
 427  * commonly known as "anti-spoofing" or Unicast Reverse Path
 428  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
 429  * is purposely reminiscent of the Cisco IOS command,
 430  *
 431  *   ip verify unicast reverse-path
 432  *   ip verify unicast source reachable-via any
 433  *
 434  * which implements the same functionality. But note that the syntax
 435  * is misleading, and the check may be performed on all IP packets
 436  * whether unicast, multicast, or broadcast.
 437  */
 438 static int
 439 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
 440 {
 441 #if defined(USERSPACE) || !defined(__FreeBSD__)
 442         return 0;
 443 #else
 444         struct nhop4_basic nh4;
 445
 446         if (fib4_lookup_nh_basic(fib, src, NHR_IFAIF, 0, &nh4) != 0)
 447                 return (0);
 448
 449         /*
 450          * If ifp is provided, check for equality with rtentry.
 451          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 452          * in order to pass packets injected back by if_simloop():
 453          * routing entry (via lo0) for our own address
 454          * may exist, so we need to handle routing assymetry.
 455          */
 456         if (ifp != NULL && ifp != nh4.nh_ifp)
 457                 return (0);
 458
 459         /* if no ifp provided, check if rtentry is not default route */
 460         if (ifp == NULL && (nh4.nh_flags & NHF_DEFAULT) != 0)
 461                 return (0);
 462
 463         /* or if this is a blackhole/reject route */
 464         if (ifp == NULL && (nh4.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
 465                 return (0);
 466
 467         /* found valid route */
 468         return 1;
 469 #endif /* __FreeBSD__ */
 470 }
 471
 472 #ifdef INET6
 473 /*
 474  * ipv6 specific rules here...
 475  */
 476 static __inline int
 477 icmp6type_match (int type, ipfw_insn_u32 *cmd)
 478 {
 479         return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
 480 }
 481
 482 static int
 483 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
 484 {
 485         int i;
 486         for (i=0; i <= cmd->o.arg1; ++i )
 487                 if (curr_flow == cmd->d[i] )
 488                         return 1;
 489         return 0;
 490 }
 491
 492 /* support for IP6_*_ME opcodes */
 493 static const struct in6_addr lla_mask = {{{
 494         0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff,
 495         0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
 496 }}};
 497
 498 static int
 499 ipfw_localip6(struct in6_addr *in6)
 500 {
 501         struct rm_priotracker in6_ifa_tracker;
 502         struct in6_ifaddr *ia;
 503
 504         if (IN6_IS_ADDR_MULTICAST(in6))
 505                 return (0);
 506
 507         if (!IN6_IS_ADDR_LINKLOCAL(in6))
 508                 return (in6_localip(in6));
 509
 510         IN6_IFADDR_RLOCK(&in6_ifa_tracker);
 511         TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
 512                 if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr))
 513                         continue;
 514                 if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
 515                     in6, &lla_mask)) {
 516                         IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
 517                         return (1);
 518                 }
 519         }
 520         IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
 521         return (0);
 522 }
 523
 524 static int
 525 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
 526 {
 527         struct nhop6_basic nh6;
 528
 529         if (IN6_IS_SCOPE_LINKLOCAL(src))
 530                 return (1);
 531
 532         if (fib6_lookup_nh_basic(fib, src, 0, NHR_IFAIF, 0, &nh6) != 0)
 533                 return (0);
 534
 535         /* If ifp is provided, check for equality with route table. */
 536         if (ifp != NULL && ifp != nh6.nh_ifp)
 537                 return (0);
 538
 539         /* if no ifp provided, check if rtentry is not default route */
 540         if (ifp == NULL && (nh6.nh_flags & NHF_DEFAULT) != 0)
 541                 return (0);
 542
 543         /* or if this is a blackhole/reject route */
 544         if (ifp == NULL && (nh6.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
 545                 return (0);
 546
 547         /* found valid route */
 548         return 1;
 549 }
 550
 551 static int
 552 is_icmp6_query(int icmp6_type)
 553 {
 554         if ((icmp6_type <= ICMP6_MAXTYPE) &&
 555             (icmp6_type == ICMP6_ECHO_REQUEST ||
 556             icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
 557             icmp6_type == ICMP6_WRUREQUEST ||
 558             icmp6_type == ICMP6_FQDN_QUERY ||
 559             icmp6_type == ICMP6_NI_QUERY))
 560                 return (1);
 561
 562         return (0);
 563 }
 564
 565 static void
 566 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
 567 {
 568         struct mbuf *m;
 569
 570         m = args->m;
 571         if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
 572                 struct tcphdr *tcp;
 573                 tcp = (struct tcphdr *)((char *)ip6 + hlen);
 574
 575                 if ((tcp->th_flags & TH_RST) == 0) {
 576                         struct mbuf *m0;
 577                         m0 = ipfw_send_pkt(args->m, &(args->f_id),
 578                             ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 579                             tcp->th_flags | TH_RST);
 580                         if (m0 != NULL)
 581                                 ip6_output(m0, NULL, NULL, 0, NULL, NULL,
 582                                     NULL);
 583                 }
 584                 FREE_PKT(m);
 585         } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
 586 #if 0
 587                 /*
 588                  * Unlike above, the mbufs need to line up with the ip6 hdr,
 589                  * as the contents are read. We need to m_adj() the
 590                  * needed amount.
 591                  * The mbuf will however be thrown away so we can adjust it.
 592                  * Remember we did an m_pullup on it already so we
 593                  * can make some assumptions about contiguousness.
 594                  */
 595                 if (args->L3offset)
 596                         m_adj(m, args->L3offset);
 597 #endif
 598                 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
 599         } else
 600                 FREE_PKT(m);
 601
 602         args->m = NULL;
 603 }
 604
 605 #endif /* INET6 */
 606
 607
 608 /*
 609  * sends a reject message, consuming the mbuf passed as an argument.
 610  */
 611 static void
 612 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
 613 {
 614
 615 #if 0
 616         /* XXX When ip is not guaranteed to be at mtod() we will
 617          * need to account for this */
 618          * The mbuf will however be thrown away so we can adjust it.
 619          * Remember we did an m_pullup on it already so we
 620          * can make some assumptions about contiguousness.
 621          */
 622         if (args->L3offset)
 623                 m_adj(m, args->L3offset);
 624 #endif
 625         if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
 626                 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
 627         } else if (args->f_id.proto == IPPROTO_TCP) {
 628                 struct tcphdr *const tcp =
 629                     L3HDR(struct tcphdr, mtod(args->m, struct ip *));
 630                 if ( (tcp->th_flags & TH_RST) == 0) {
 631                         struct mbuf *m;
 632                         m = ipfw_send_pkt(args->m, &(args->f_id),
 633                                 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 634                                 tcp->th_flags | TH_RST);
 635                         if (m != NULL)
 636                                 ip_output(m, NULL, NULL, 0, NULL, NULL);
 637                 }
 638                 FREE_PKT(args->m);
 639         } else
 640                 FREE_PKT(args->m);
 641         args->m = NULL;
 642 }
 643
 644 /*
 645  * Support for uid/gid/jail lookup. These tests are expensive
 646  * (because we may need to look into the list of active sockets)
 647  * so we cache the results. ugid_lookupp is 0 if we have not
 648  * yet done a lookup, 1 if we succeeded, and -1 if we tried
 649  * and failed. The function always returns the match value.
 650  * We could actually spare the variable and use *uc, setting
 651  * it to '(void *)check_uidgid if we have no info, NULL if
 652  * we tried and failed, or any other value if successful.
 653  */
 654 static int
 655 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
 656     struct ucred **uc)
 657 {
 658 #if defined(USERSPACE)
 659         return 0;       // not supported in userspace
 660 #else
 661 #ifndef __FreeBSD__
 662         /* XXX */
 663         return cred_check(insn, proto, oif,
 664             dst_ip, dst_port, src_ip, src_port,
 665             (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
 666 #else  /* FreeBSD */
 667         struct in_addr src_ip, dst_ip;
 668         struct inpcbinfo *pi;
 669         struct ipfw_flow_id *id;
 670         struct inpcb *pcb, *inp;
 671         struct ifnet *oif;
 672         int lookupflags;
 673         int match;
 674
 675         id = &args->f_id;
 676         inp = args->inp;
 677         oif = args->oif;
 678
 679         /*
 680          * Check to see if the UDP or TCP stack supplied us with
 681          * the PCB. If so, rather then holding a lock and looking
 682          * up the PCB, we can use the one that was supplied.
 683          */
 684         if (inp && *ugid_lookupp == 0) {
 685                 INP_LOCK_ASSERT(inp);
 686                 if (inp->inp_socket != NULL) {
 687                         *uc = crhold(inp->inp_cred);
 688                         *ugid_lookupp = 1;
 689                 } else
 690                         *ugid_lookupp = -1;
 691         }
 692         /*
 693          * If we have already been here and the packet has no
 694          * PCB entry associated with it, then we can safely
 695          * assume that this is a no match.
 696          */
 697         if (*ugid_lookupp == -1)
 698                 return (0);
 699         if (id->proto == IPPROTO_TCP) {
 700                 lookupflags = 0;
 701                 pi = &V_tcbinfo;
 702         } else if (id->proto == IPPROTO_UDP) {
 703                 lookupflags = INPLOOKUP_WILDCARD;
 704                 pi = &V_udbinfo;
 705         } else
 706                 return 0;
 707         lookupflags |= INPLOOKUP_RLOCKPCB;
 708         match = 0;
 709         if (*ugid_lookupp == 0) {
 710                 if (id->addr_type == 6) {
 711 #ifdef INET6
 712                         if (oif == NULL)
 713                                 pcb = in6_pcblookup_mbuf(pi,
 714                                     &id->src_ip6, htons(id->src_port),
 715                                     &id->dst_ip6, htons(id->dst_port),
 716                                     lookupflags, oif, args->m);
 717                         else
 718                                 pcb = in6_pcblookup_mbuf(pi,
 719                                     &id->dst_ip6, htons(id->dst_port),
 720                                     &id->src_ip6, htons(id->src_port),
 721                                     lookupflags, oif, args->m);
 722 #else
 723                         *ugid_lookupp = -1;
 724                         return (0);
 725 #endif
 726                 } else {
 727                         src_ip.s_addr = htonl(id->src_ip);
 728                         dst_ip.s_addr = htonl(id->dst_ip);
 729                         if (oif == NULL)
 730                                 pcb = in_pcblookup_mbuf(pi,
 731                                     src_ip, htons(id->src_port),
 732                                     dst_ip, htons(id->dst_port),
 733                                     lookupflags, oif, args->m);
 734                         else
 735                                 pcb = in_pcblookup_mbuf(pi,
 736                                     dst_ip, htons(id->dst_port),
 737                                     src_ip, htons(id->src_port),
 738                                     lookupflags, oif, args->m);
 739                 }
 740                 if (pcb != NULL) {
 741                         INP_RLOCK_ASSERT(pcb);
 742                         *uc = crhold(pcb->inp_cred);
 743                         *ugid_lookupp = 1;
 744                         INP_RUNLOCK(pcb);
 745                 }
 746                 if (*ugid_lookupp == 0) {
 747                         /*
 748                          * We tried and failed, set the variable to -1
 749                          * so we will not try again on this packet.
 750                          */
 751                         *ugid_lookupp = -1;
 752                         return (0);
 753                 }
 754         }
 755         if (insn->o.opcode == O_UID)
 756                 match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
 757         else if (insn->o.opcode == O_GID)
 758                 match = groupmember((gid_t)insn->d[0], *uc);
 759         else if (insn->o.opcode == O_JAIL)
 760                 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
 761         return (match);
 762 #endif /* __FreeBSD__ */
 763 #endif /* not supported in userspace */
 764 }
 765
 766 /*
 767  * Helper function to set args with info on the rule after the matching
 768  * one. slot is precise, whereas we guess rule_id as they are
 769  * assigned sequentially.
 770  */
 771 static inline void
 772 set_match(struct ip_fw_args *args, int slot,
 773         struct ip_fw_chain *chain)
 774 {
 775         args->rule.chain_id = chain->id;
 776         args->rule.slot = slot + 1; /* we use 0 as a marker */
 777         args->rule.rule_id = 1 + chain->map[slot]->id;
 778         args->rule.rulenum = chain->map[slot]->rulenum;
 779 }
 780
 781 #ifndef LINEAR_SKIPTO
 782 /*
 783  * Helper function to enable cached rule lookups using
 784  * cached_id and cached_pos fields in ipfw rule.
 785  */
 786 static int
 787 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
 788     int tablearg, int jump_backwards)
 789 {
 790         int f_pos;
 791
 792         /* If possible use cached f_pos (in f->cached_pos),
 793          * whose version is written in f->cached_id
 794          * (horrible hacks to avoid changing the ABI).
 795          */
 796         if (num != IP_FW_TARG && f->cached_id == chain->id)
 797                 f_pos = f->cached_pos;
 798         else {
 799                 int i = IP_FW_ARG_TABLEARG(chain, num, skipto);
 800                 /* make sure we do not jump backward */
 801                 if (jump_backwards == 0 && i <= f->rulenum)
 802                         i = f->rulenum + 1;
 803                 if (chain->idxmap != NULL)
 804                         f_pos = chain->idxmap[i];
 805                 else
 806                         f_pos = ipfw_find_rule(chain, i, 0);
 807                 /* update the cache */
 808                 if (num != IP_FW_TARG) {
 809                         f->cached_id = chain->id;
 810                         f->cached_pos = f_pos;
 811                 }
 812         }
 813
 814         return (f_pos);
 815 }
 816 #else
 817 /*
 818  * Helper function to enable real fast rule lookups.
 819  */
 820 static int
 821 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
 822     int tablearg, int jump_backwards)
 823 {
 824         int f_pos;
 825
 826         num = IP_FW_ARG_TABLEARG(chain, num, skipto);
 827         /* make sure we do not jump backward */
 828         if (jump_backwards == 0 && num <= f->rulenum)
 829                 num = f->rulenum + 1;
 830         f_pos = chain->idxmap[num];
 831
 832         return (f_pos);
 833 }
 834 #endif
 835
 836 #define TARG(k, f)      IP_FW_ARG_TABLEARG(chain, k, f)
 837 /*
 838  * The main check routine for the firewall.
 839  *
 840  * All arguments are in args so we can modify them and return them
 841  * back to the caller.
 842  *
 843  * Parameters:
 844  *
 845  *      args->m (in/out) The packet; we set to NULL when/if we nuke it.
 846  *              Starts with the IP header.
 847  *      args->eh (in)   Mac header if present, NULL for layer3 packet.
 848  *      args->L3offset  Number of bytes bypassed if we came from L2.
 849  *                      e.g. often sizeof(eh)  ** NOTYET **
 850  *      args->oif       Outgoing interface, NULL if packet is incoming.
 851  *              The incoming interface is in the mbuf. (in)
 852  *      args->divert_rule (in/out)
 853  *              Skip up to the first rule past this rule number;
 854  *              upon return, non-zero port number for divert or tee.
 855  *
 856  *      args->rule      Pointer to the last matching rule (in/out)
 857  *      args->next_hop  Socket we are forwarding to (out).
 858  *      args->next_hop6 IPv6 next hop we are forwarding to (out).
 859  *      args->f_id      Addresses grabbed from the packet (out)
 860  *      args->rule.info a cookie depending on rule action
 861  *
 862  * Return value:
 863  *
 864  *      IP_FW_PASS      the packet must be accepted
 865  *      IP_FW_DENY      the packet must be dropped
 866  *      IP_FW_DIVERT    divert packet, port in m_tag
 867  *      IP_FW_TEE       tee packet, port in m_tag
 868  *      IP_FW_DUMMYNET  to dummynet, pipe in args->cookie
 869  *      IP_FW_NETGRAPH  into netgraph, cookie args->cookie
 870  *              args->rule contains the matching rule,
 871  *              args->rule.info has additional information.
 872  *
 873  */
 874 int
 875 ipfw_chk(struct ip_fw_args *args)
 876 {
 877
 878         /*
 879          * Local variables holding state while processing a packet:
 880          *
 881          * IMPORTANT NOTE: to speed up the processing of rules, there
 882          * are some assumption on the values of the variables, which
 883          * are documented here. Should you change them, please check
 884          * the implementation of the various instructions to make sure
 885          * that they still work.
 886          *
 887          * args->eh     The MAC header. It is non-null for a layer2
 888          *      packet, it is NULL for a layer-3 packet.
 889          * **notyet**
 890          * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
 891          *
 892          * m | args->m  Pointer to the mbuf, as received from the caller.
 893          *      It may change if ipfw_chk() does an m_pullup, or if it
 894          *      consumes the packet because it calls send_reject().
 895          *      XXX This has to change, so that ipfw_chk() never modifies
 896          *      or consumes the buffer.
 897          * ip   is the beginning of the ip(4 or 6) header.
 898          *      Calculated by adding the L3offset to the start of data.
 899          *      (Until we start using L3offset, the packet is
 900          *      supposed to start with the ip header).
 901          */
 902         struct mbuf *m = args->m;
 903         struct ip *ip = mtod(m, struct ip *);
 904
 905         /*
 906          * For rules which contain uid/gid or jail constraints, cache
 907          * a copy of the users credentials after the pcb lookup has been
 908          * executed. This will speed up the processing of rules with
 909          * these types of constraints, as well as decrease contention
 910          * on pcb related locks.
 911          */
 912 #ifndef __FreeBSD__
 913         struct bsd_ucred ucred_cache;
 914 #else
 915         struct ucred *ucred_cache = NULL;
 916 #endif
 917         int ucred_lookup = 0;
 918
 919         /*
 920          * oif | args->oif      If NULL, ipfw_chk has been called on the
 921          *      inbound path (ether_input, ip_input).
 922          *      If non-NULL, ipfw_chk has been called on the outbound path
 923          *      (ether_output, ip_output).
 924          */
 925         struct ifnet *oif = args->oif;
 926
 927         int f_pos = 0;          /* index of current rule in the array */
 928         int retval = 0;
 929
 930         /*
 931          * hlen The length of the IP header.
 932          */
 933         u_int hlen = 0;         /* hlen >0 means we have an IP pkt */
 934
 935         /*
 936          * offset       The offset of a fragment. offset != 0 means that
 937          *      we have a fragment at this offset of an IPv4 packet.
 938          *      offset == 0 means that (if this is an IPv4 packet)
 939          *      this is the first or only fragment.
 940          *      For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
 941          *      or there is a single packet fragment (fragment header added
 942          *      without needed).  We will treat a single packet fragment as if
 943          *      there was no fragment header (or log/block depending on the
 944          *      V_fw_permit_single_frag6 sysctl setting).
 945          */
 946         u_short offset = 0;
 947         u_short ip6f_mf = 0;
 948
 949         /*
 950          * Local copies of addresses. They are only valid if we have
 951          * an IP packet.
 952          *
 953          * proto        The protocol. Set to 0 for non-ip packets,
 954          *      or to the protocol read from the packet otherwise.
 955          *      proto != 0 means that we have an IPv4 packet.
 956          *
 957          * src_port, dst_port   port numbers, in HOST format. Only
 958          *      valid for TCP and UDP packets.
 959          *
 960          * src_ip, dst_ip       ip addresses, in NETWORK format.
 961          *      Only valid for IPv4 packets.
 962          */
 963         uint8_t proto;
 964         uint16_t src_port = 0, dst_port = 0;    /* NOTE: host format    */
 965         struct in_addr src_ip, dst_ip;          /* NOTE: network format */
 966         uint16_t iplen=0;
 967         int pktlen;
 968         uint16_t        etype = 0;      /* Host order stored ether type */
 969
 970         /*
 971          * dyn_dir = MATCH_UNKNOWN when rules unchecked,
 972          *      MATCH_NONE when checked and not matched (q = NULL),
 973          *      MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
 974          */
 975         int dyn_dir = MATCH_UNKNOWN;
 976         uint16_t dyn_name = 0;
 977         ipfw_dyn_rule *q = NULL;
 978         struct ip_fw_chain *chain = &V_layer3_chain;
 979
 980         /*
 981          * We store in ulp a pointer to the upper layer protocol header.
 982          * In the ipv4 case this is easy to determine from the header,
 983          * but for ipv6 we might have some additional headers in the middle.
 984          * ulp is NULL if not found.
 985          */
 986         void *ulp = NULL;               /* upper layer protocol pointer. */
 987
 988         /* XXX ipv6 variables */
 989         int is_ipv6 = 0;
 990         uint8_t icmp6_type = 0;
 991         uint16_t ext_hd = 0;    /* bits vector for extension header filtering */
 992         /* end of ipv6 variables */
 993
 994         int is_ipv4 = 0;
 995
 996         int done = 0;           /* flag to exit the outer loop */
 997
 998         if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
 999                 return (IP_FW_PASS);    /* accept */
1000
1001         dst_ip.s_addr = 0;              /* make sure it is initialized */
1002         src_ip.s_addr = 0;              /* make sure it is initialized */
1003         pktlen = m->m_pkthdr.len;
1004         args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
1005         proto = args->f_id.proto = 0;   /* mark f_id invalid */
1006                 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
1007
1008 /*
1009  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1010  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1011  * pointer might become stale after other pullups (but we never use it
1012  * this way).
1013  */
1014 #define PULLUP_TO(_len, p, T)   PULLUP_LEN(_len, p, sizeof(T))
1015 #define PULLUP_LEN(_len, p, T)                                  \
1016 do {                                                            \
1017         int x = (_len) + T;                                     \
1018         if ((m)->m_len < x) {                                   \
1019                 args->m = m = m_pullup(m, x);                   \
1020                 if (m == NULL)                                  \
1021                         goto pullup_failed;                     \
1022         }                                                       \
1023         p = (mtod(m, char *) + (_len));                         \
1024 } while (0)
1025
1026         /*
1027          * if we have an ether header,
1028          */
1029         if (args->eh)
1030                 etype = ntohs(args->eh->ether_type);
1031
1032         /* Identify IP packets and fill up variables. */
1033         if (pktlen >= sizeof(struct ip6_hdr) &&
1034             (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
1035                 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1036                 is_ipv6 = 1;
1037                 args->f_id.addr_type = 6;
1038                 hlen = sizeof(struct ip6_hdr);
1039                 proto = ip6->ip6_nxt;
1040
1041                 /* Search extension headers to find upper layer protocols */
1042                 while (ulp == NULL && offset == 0) {
1043                         switch (proto) {
1044                         case IPPROTO_ICMPV6:
1045                                 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1046                                 icmp6_type = ICMP6(ulp)->icmp6_type;
1047                                 break;
1048
1049                         case IPPROTO_TCP:
1050                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1051                                 dst_port = TCP(ulp)->th_dport;
1052                                 src_port = TCP(ulp)->th_sport;
1053                                 /* save flags for dynamic rules */
1054                                 args->f_id._flags = TCP(ulp)->th_flags;
1055                                 break;
1056
1057                         case IPPROTO_SCTP:
1058                                 PULLUP_TO(hlen, ulp, struct sctphdr);
1059                                 src_port = SCTP(ulp)->src_port;
1060                                 dst_port = SCTP(ulp)->dest_port;
1061                                 break;
1062
1063                         case IPPROTO_UDP:
1064                                 PULLUP_TO(hlen, ulp, struct udphdr);
1065                                 dst_port = UDP(ulp)->uh_dport;
1066                                 src_port = UDP(ulp)->uh_sport;
1067                                 break;
1068
1069                         case IPPROTO_HOPOPTS:   /* RFC 2460 */
1070                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1071                                 ext_hd |= EXT_HOPOPTS;
1072                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1073                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1074                                 ulp = NULL;
1075                                 break;
1076
1077                         case IPPROTO_ROUTING:   /* RFC 2460 */
1078                                 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1079                                 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1080                                 case 0:
1081                                         ext_hd |= EXT_RTHDR0;
1082                                         break;
1083                                 case 2:
1084                                         ext_hd |= EXT_RTHDR2;
1085                                         break;
1086                                 default:
1087                                         if (V_fw_verbose)
1088                                                 printf("IPFW2: IPV6 - Unknown "
1089                                                     "Routing Header type(%d)\n",
1090                                                     ((struct ip6_rthdr *)
1091                                                     ulp)->ip6r_type);
1092                                         if (V_fw_deny_unknown_exthdrs)
1093                                             return (IP_FW_DENY);
1094                                         break;
1095                                 }
1096                                 ext_hd |= EXT_ROUTING;
1097                                 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1098                                 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1099                                 ulp = NULL;
1100                                 break;
1101
1102                         case IPPROTO_FRAGMENT:  /* RFC 2460 */
1103                                 PULLUP_TO(hlen, ulp, struct ip6_frag);
1104                                 ext_hd |= EXT_FRAGMENT;
1105                                 hlen += sizeof (struct ip6_frag);
1106                                 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1107                                 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1108                                         IP6F_OFF_MASK;
1109                                 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1110                                         IP6F_MORE_FRAG;
1111                                 if (V_fw_permit_single_frag6 == 0 &&
1112                                     offset == 0 && ip6f_mf == 0) {
1113                                         if (V_fw_verbose)
1114                                                 printf("IPFW2: IPV6 - Invalid "
1115                                                     "Fragment Header\n");
1116                                         if (V_fw_deny_unknown_exthdrs)
1117                                             return (IP_FW_DENY);
1118                                         break;
1119                                 }
1120                                 args->f_id.extra =
1121                                     ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1122                                 ulp = NULL;
1123                                 break;
1124
1125                         case IPPROTO_DSTOPTS:   /* RFC 2460 */
1126                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1127                                 ext_hd |= EXT_DSTOPTS;
1128                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1129                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1130                                 ulp = NULL;
1131                                 break;
1132
1133                         case IPPROTO_AH:        /* RFC 2402 */
1134                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1135                                 ext_hd |= EXT_AH;
1136                                 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1137                                 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1138                                 ulp = NULL;
1139                                 break;
1140
1141                         case IPPROTO_ESP:       /* RFC 2406 */
1142                                 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1143                                 /* Anything past Seq# is variable length and
1144                                  * data past this ext. header is encrypted. */
1145                                 ext_hd |= EXT_ESP;
1146                                 break;
1147
1148                         case IPPROTO_NONE:      /* RFC 2460 */
1149                                 /*
1150                                  * Packet ends here, and IPv6 header has
1151                                  * already been pulled up. If ip6e_len!=0
1152                                  * then octets must be ignored.
1153                                  */
1154                                 ulp = ip; /* non-NULL to get out of loop. */
1155                                 break;
1156
1157                         case IPPROTO_OSPFIGP:
1158                                 /* XXX OSPF header check? */
1159                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1160                                 break;
1161
1162                         case IPPROTO_PIM:
1163                                 /* XXX PIM header check? */
1164                                 PULLUP_TO(hlen, ulp, struct pim);
1165                                 break;
1166
1167                         case IPPROTO_GRE:       /* RFC 1701 */
1168                                 /* XXX GRE header check? */
1169                                 PULLUP_TO(hlen, ulp, struct grehdr);
1170                                 break;
1171
1172                         case IPPROTO_CARP:
1173                                 PULLUP_TO(hlen, ulp, struct carp_header);
1174                                 if (((struct carp_header *)ulp)->carp_version !=
1175                                     CARP_VERSION)
1176                                         return (IP_FW_DENY);
1177                                 if (((struct carp_header *)ulp)->carp_type !=
1178                                     CARP_ADVERTISEMENT)
1179                                         return (IP_FW_DENY);
1180                                 break;
1181
1182                         case IPPROTO_IPV6:      /* RFC 2893 */
1183                                 PULLUP_TO(hlen, ulp, struct ip6_hdr);
1184                                 break;
1185
1186                         case IPPROTO_IPV4:      /* RFC 2893 */
1187                                 PULLUP_TO(hlen, ulp, struct ip);
1188                                 break;
1189
1190                         default:
1191                                 if (V_fw_verbose)
1192                                         printf("IPFW2: IPV6 - Unknown "
1193                                             "Extension Header(%d), ext_hd=%x\n",
1194                                              proto, ext_hd);
1195                                 if (V_fw_deny_unknown_exthdrs)
1196                                     return (IP_FW_DENY);
1197                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1198                                 break;
1199                         } /*switch */
1200                 }
1201                 ip = mtod(m, struct ip *);
1202                 ip6 = (struct ip6_hdr *)ip;
1203                 args->f_id.src_ip6 = ip6->ip6_src;
1204                 args->f_id.dst_ip6 = ip6->ip6_dst;
1205                 args->f_id.src_ip = 0;
1206                 args->f_id.dst_ip = 0;
1207                 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1208         } else if (pktlen >= sizeof(struct ip) &&
1209             (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1210                 is_ipv4 = 1;
1211                 hlen = ip->ip_hl << 2;
1212                 args->f_id.addr_type = 4;
1213
1214                 /*
1215                  * Collect parameters into local variables for faster matching.
1216                  */
1217                 proto = ip->ip_p;
1218                 src_ip = ip->ip_src;
1219                 dst_ip = ip->ip_dst;
1220                 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1221                 iplen = ntohs(ip->ip_len);
1222                 pktlen = iplen < pktlen ? iplen : pktlen;
1223
1224                 if (offset == 0) {
1225                         switch (proto) {
1226                         case IPPROTO_TCP:
1227                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1228                                 dst_port = TCP(ulp)->th_dport;
1229                                 src_port = TCP(ulp)->th_sport;
1230                                 /* save flags for dynamic rules */
1231                                 args->f_id._flags = TCP(ulp)->th_flags;
1232                                 break;
1233
1234                         case IPPROTO_SCTP:
1235                                 PULLUP_TO(hlen, ulp, struct sctphdr);
1236                                 src_port = SCTP(ulp)->src_port;
1237                                 dst_port = SCTP(ulp)->dest_port;
1238                                 break;
1239
1240                         case IPPROTO_UDP:
1241                                 PULLUP_TO(hlen, ulp, struct udphdr);
1242                                 dst_port = UDP(ulp)->uh_dport;
1243                                 src_port = UDP(ulp)->uh_sport;
1244                                 break;
1245
1246                         case IPPROTO_ICMP:
1247                                 PULLUP_TO(hlen, ulp, struct icmphdr);
1248                                 //args->f_id.flags = ICMP(ulp)->icmp_type;
1249                                 break;
1250
1251                         default:
1252                                 break;
1253                         }
1254                 }
1255
1256                 ip = mtod(m, struct ip *);
1257                 args->f_id.src_ip = ntohl(src_ip.s_addr);
1258                 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1259         }
1260 #undef PULLUP_TO
1261         if (proto) { /* we may have port numbers, store them */
1262                 args->f_id.proto = proto;
1263                 args->f_id.src_port = src_port = ntohs(src_port);
1264                 args->f_id.dst_port = dst_port = ntohs(dst_port);
1265         }
1266
1267         IPFW_PF_RLOCK(chain);
1268         if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1269                 IPFW_PF_RUNLOCK(chain);
1270                 return (IP_FW_PASS);    /* accept */
1271         }
1272         if (args->rule.slot) {
1273                 /*
1274                  * Packet has already been tagged as a result of a previous
1275                  * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1276                  * REASS, NETGRAPH, DIVERT/TEE...)
1277                  * Validate the slot and continue from the next one
1278                  * if still present, otherwise do a lookup.
1279                  */
1280                 f_pos = (args->rule.chain_id == chain->id) ?
1281                     args->rule.slot :
1282                     ipfw_find_rule(chain, args->rule.rulenum,
1283                         args->rule.rule_id);
1284         } else {
1285                 f_pos = 0;
1286         }
1287
1288         /*
1289          * Now scan the rules, and parse microinstructions for each rule.
1290          * We have two nested loops and an inner switch. Sometimes we
1291          * need to break out of one or both loops, or re-enter one of
1292          * the loops with updated variables. Loop variables are:
1293          *
1294          *      f_pos (outer loop) points to the current rule.
1295          *              On output it points to the matching rule.
1296          *      done (outer loop) is used as a flag to break the loop.
1297          *      l (inner loop)  residual length of current rule.
1298          *              cmd points to the current microinstruction.
1299          *
1300          * We break the inner loop by setting l=0 and possibly
1301          * cmdlen=0 if we don't want to advance cmd.
1302          * We break the outer loop by setting done=1
1303          * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1304          * as needed.
1305          */
1306         for (; f_pos < chain->n_rules; f_pos++) {
1307                 ipfw_insn *cmd;
1308                 uint32_t tablearg = 0;
1309                 int l, cmdlen, skip_or; /* skip rest of OR block */
1310                 struct ip_fw *f;
1311
1312                 f = chain->map[f_pos];
1313                 if (V_set_disable & (1 << f->set) )
1314                         continue;
1315
1316                 skip_or = 0;
1317                 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1318                     l -= cmdlen, cmd += cmdlen) {
1319                         int match;
1320
1321                         /*
1322                          * check_body is a jump target used when we find a
1323                          * CHECK_STATE, and need to jump to the body of
1324                          * the target rule.
1325                          */
1326
1327 /* check_body: */
1328                         cmdlen = F_LEN(cmd);
1329                         /*
1330                          * An OR block (insn_1 || .. || insn_n) has the
1331                          * F_OR bit set in all but the last instruction.
1332                          * The first match will set "skip_or", and cause
1333                          * the following instructions to be skipped until
1334                          * past the one with the F_OR bit clear.
1335                          */
1336                         if (skip_or) {          /* skip this instruction */
1337                                 if ((cmd->len & F_OR) == 0)
1338                                         skip_or = 0;    /* next one is good */
1339                                 continue;
1340                         }
1341                         match = 0; /* set to 1 if we succeed */
1342
1343                         switch (cmd->opcode) {
1344                         /*
1345                          * The first set of opcodes compares the packet's
1346                          * fields with some pattern, setting 'match' if a
1347                          * match is found. At the end of the loop there is
1348                          * logic to deal with F_NOT and F_OR flags associated
1349                          * with the opcode.
1350                          */
1351                         case O_NOP:
1352                                 match = 1;
1353                                 break;
1354
1355                         case O_FORWARD_MAC:
1356                                 printf("ipfw: opcode %d unimplemented\n",
1357                                     cmd->opcode);
1358                                 break;
1359
1360                         case O_GID:
1361                         case O_UID:
1362                         case O_JAIL:
1363                                 /*
1364                                  * We only check offset == 0 && proto != 0,
1365                                  * as this ensures that we have a
1366                                  * packet with the ports info.
1367                                  */
1368                                 if (offset != 0)
1369                                         break;
1370                                 if (proto == IPPROTO_TCP ||
1371                                     proto == IPPROTO_UDP)
1372                                         match = check_uidgid(
1373                                                     (ipfw_insn_u32 *)cmd,
1374                                                     args, &ucred_lookup,
1375 #ifdef __FreeBSD__
1376                                                     &ucred_cache);
1377 #else
1378                                                     (void *)&ucred_cache);
1379 #endif
1380                                 break;
1381
1382                         case O_RECV:
1383                                 match = iface_match(m->m_pkthdr.rcvif,
1384                                     (ipfw_insn_if *)cmd, chain, &tablearg);
1385                                 break;
1386
1387                         case O_XMIT:
1388                                 match = iface_match(oif, (ipfw_insn_if *)cmd,
1389                                     chain, &tablearg);
1390                                 break;
1391
1392                         case O_VIA:
1393                                 match = iface_match(oif ? oif :
1394                                     m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1395                                     chain, &tablearg);
1396                                 break;
1397
1398                         case O_MACADDR2:
1399                                 if (args->eh != NULL) { /* have MAC header */
1400                                         u_int32_t *want = (u_int32_t *)
1401                                                 ((ipfw_insn_mac *)cmd)->addr;
1402                                         u_int32_t *mask = (u_int32_t *)
1403                                                 ((ipfw_insn_mac *)cmd)->mask;
1404                                         u_int32_t *hdr = (u_int32_t *)args->eh;
1405
1406                                         match =
1407                                             ( want[0] == (hdr[0] & mask[0]) &&
1408                                               want[1] == (hdr[1] & mask[1]) &&
1409                                               want[2] == (hdr[2] & mask[2]) );
1410                                 }
1411                                 break;
1412
1413                         case O_MAC_TYPE:
1414                                 if (args->eh != NULL) {
1415                                         u_int16_t *p =
1416                                             ((ipfw_insn_u16 *)cmd)->ports;
1417                                         int i;
1418
1419                                         for (i = cmdlen - 1; !match && i>0;
1420                                             i--, p += 2)
1421                                                 match = (etype >= p[0] &&
1422                                                     etype <= p[1]);
1423                                 }
1424                                 break;
1425
1426                         case O_FRAG:
1427                                 match = (offset != 0);
1428                                 break;
1429
1430                         case O_IN:      /* "out" is "not in" */
1431                                 match = (oif == NULL);
1432                                 break;
1433
1434                         case O_LAYER2:
1435                                 match = (args->eh != NULL);
1436                                 break;
1437
1438                         case O_DIVERTED:
1439                             {
1440                                 /* For diverted packets, args->rule.info
1441                                  * contains the divert port (in host format)
1442                                  * reason and direction.
1443                                  */
1444                                 uint32_t i = args->rule.info;
1445                                 match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1446                                     cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1447                             }
1448                                 break;
1449
1450                         case O_PROTO:
1451                                 /*
1452                                  * We do not allow an arg of 0 so the
1453                                  * check of "proto" only suffices.
1454                                  */
1455                                 match = (proto == cmd->arg1);
1456                                 break;
1457
1458                         case O_IP_SRC:
1459                                 match = is_ipv4 &&
1460                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1461                                     src_ip.s_addr);
1462                                 break;
1463
1464                         case O_IP_DST_LOOKUP:
1465                         {
1466                                 void *pkey;
1467                                 uint32_t vidx, key;
1468                                 uint16_t keylen;
1469
1470                                 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1471                                         /* Determine lookup key type */
1472                                         vidx = ((ipfw_insn_u32 *)cmd)->d[1];
1473                                         if (vidx != 4 /* uid */ &&
1474                                             vidx != 5 /* jail */ &&
1475                                             is_ipv6 == 0 && is_ipv4 == 0)
1476                                                 break;
1477                                         /* Determine key length */
1478                                         if (vidx == 0 /* dst-ip */ ||
1479                                             vidx == 1 /* src-ip */)
1480                                                 keylen = is_ipv6 ?
1481                                                     sizeof(struct in6_addr):
1482                                                     sizeof(in_addr_t);
1483                                         else {
1484                                                 keylen = sizeof(key);
1485                                                 pkey = &key;
1486                                         }
1487                                         if (vidx == 0 /* dst-ip */)
1488                                                 pkey = is_ipv4 ? (void *)&dst_ip:
1489                                                     (void *)&args->f_id.dst_ip6;
1490                                         else if (vidx == 1 /* src-ip */)
1491                                                 pkey = is_ipv4 ? (void *)&src_ip:
1492                                                     (void *)&args->f_id.src_ip6;
1493                                         else if (vidx == 6 /* dscp */) {
1494                                                 if (is_ipv4)
1495                                                         key = ip->ip_tos >> 2;
1496                                                 else {
1497                                                         key = args->f_id.flow_id6;
1498                                                         key = (key & 0x0f) << 2 |
1499                                                             (key & 0xf000) >> 14;
1500                                                 }
1501                                                 key &= 0x3f;
1502                                         } else if (vidx == 2 /* dst-port */ ||
1503                                             vidx == 3 /* src-port */) {
1504                                                 /* Skip fragments */
1505                                                 if (offset != 0)
1506                                                         break;
1507                                                 /* Skip proto without ports */
1508                                                 if (proto != IPPROTO_TCP &&
1509                                                     proto != IPPROTO_UDP &&
1510                                                     proto != IPPROTO_SCTP)
1511                                                         break;
1512                                                 if (vidx == 2 /* dst-port */)
1513                                                         key = dst_port;
1514                                                 else
1515                                                         key = src_port;
1516                                         }
1517 #ifndef USERSPACE
1518                                         else if (vidx == 4 /* uid */ ||
1519                                             vidx == 5 /* jail */) {
1520                                                 check_uidgid(
1521                                                     (ipfw_insn_u32 *)cmd,
1522                                                     args, &ucred_lookup,
1523 #ifdef __FreeBSD__
1524                                                     &ucred_cache);
1525                                                 if (vidx == 4 /* uid */)
1526                                                         key = ucred_cache->cr_uid;
1527                                                 else if (vidx == 5 /* jail */)
1528                                                         key = ucred_cache->cr_prison->pr_id;
1529 #else /* !__FreeBSD__ */
1530                                                     (void *)&ucred_cache);
1531                                                 if (vidx == 4 /* uid */)
1532                                                         key = ucred_cache.uid;
1533                                                 else if (vidx == 5 /* jail */)
1534                                                         key = ucred_cache.xid;
1535 #endif /* !__FreeBSD__ */
1536                                         }
1537 #endif /* !USERSPACE */
1538                                         else
1539                                                 break;
1540                                         match = ipfw_lookup_table(chain,
1541                                             cmd->arg1, keylen, pkey, &vidx);
1542                                         if (!match)
1543                                                 break;
1544                                         tablearg = vidx;
1545                                         break;
1546                                 }
1547                                 /* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */
1548                                 /* FALLTHROUGH */
1549                         }
1550                         case O_IP_SRC_LOOKUP:
1551                         {
1552                                 void *pkey;
1553                                 uint32_t vidx;
1554                                 uint16_t keylen;
1555
1556                                 if (is_ipv4) {
1557                                         keylen = sizeof(in_addr_t);
1558                                         if (cmd->opcode == O_IP_DST_LOOKUP)
1559                                                 pkey = &dst_ip;
1560                                         else
1561                                                 pkey = &src_ip;
1562                                 } else if (is_ipv6) {
1563                                         keylen = sizeof(struct in6_addr);
1564                                         if (cmd->opcode == O_IP_DST_LOOKUP)
1565                                                 pkey = &args->f_id.dst_ip6;
1566                                         else
1567                                                 pkey = &args->f_id.src_ip6;
1568                                 } else
1569                                         break;
1570                                 match = ipfw_lookup_table(chain, cmd->arg1,
1571                                     keylen, pkey, &vidx);
1572                                 if (!match)
1573                                         break;
1574                                 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) {
1575                                         match = ((ipfw_insn_u32 *)cmd)->d[0] ==
1576                                             TARG_VAL(chain, vidx, tag);
1577                                         if (!match)
1578                                                 break;
1579                                 }
1580                                 tablearg = vidx;
1581                                 break;
1582                         }
1583
1584                         case O_IP_FLOW_LOOKUP:
1585                                 {
1586                                         uint32_t v = 0;
1587                                         match = ipfw_lookup_table(chain,
1588                                             cmd->arg1, 0, &args->f_id, &v);
1589                                         if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1590                                                 match = ((ipfw_insn_u32 *)cmd)->d[0] ==
1591                                                     TARG_VAL(chain, v, tag);
1592                                         if (match)
1593                                                 tablearg = v;
1594                                 }
1595                                 break;
1596                         case O_IP_SRC_MASK:
1597                         case O_IP_DST_MASK:
1598                                 if (is_ipv4) {
1599                                     uint32_t a =
1600                                         (cmd->opcode == O_IP_DST_MASK) ?
1601                                             dst_ip.s_addr : src_ip.s_addr;
1602                                     uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
1603                                     int i = cmdlen-1;
1604
1605                                     for (; !match && i>0; i-= 2, p+= 2)
1606                                         match = (p[0] == (a & p[1]));
1607                                 }
1608                                 break;
1609
1610                         case O_IP_SRC_ME:
1611                                 if (is_ipv4) {
1612                                         struct ifnet *tif;
1613
1614                                         INADDR_TO_IFP(src_ip, tif);
1615                                         match = (tif != NULL);
1616                                         break;
1617                                 }
1618 #ifdef INET6
1619                                 /* FALLTHROUGH */
1620                         case O_IP6_SRC_ME:
1621                                 match= is_ipv6 && ipfw_localip6(&args->f_id.src_ip6);
1622 #endif
1623                                 break;
1624
1625                         case O_IP_DST_SET:
1626                         case O_IP_SRC_SET:
1627                                 if (is_ipv4) {
1628                                         u_int32_t *d = (u_int32_t *)(cmd+1);
1629                                         u_int32_t addr =
1630                                             cmd->opcode == O_IP_DST_SET ?
1631                                                 args->f_id.dst_ip :
1632                                                 args->f_id.src_ip;
1633
1634                                             if (addr < d[0])
1635                                                     break;
1636                                             addr -= d[0]; /* subtract base */
1637                                             match = (addr < cmd->arg1) &&
1638                                                 ( d[ 1 + (addr>>5)] &
1639                                                   (1<<(addr & 0x1f)) );
1640                                 }
1641                                 break;
1642
1643                         case O_IP_DST:
1644                                 match = is_ipv4 &&
1645                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1646                                     dst_ip.s_addr);
1647                                 break;
1648
1649                         case O_IP_DST_ME:
1650                                 if (is_ipv4) {
1651                                         struct ifnet *tif;
1652
1653                                         INADDR_TO_IFP(dst_ip, tif);
1654                                         match = (tif != NULL);
1655                                         break;
1656                                 }
1657 #ifdef INET6
1658                                 /* FALLTHROUGH */
1659                         case O_IP6_DST_ME:
1660                                 match= is_ipv6 && ipfw_localip6(&args->f_id.dst_ip6);
1661 #endif
1662                                 break;
1663
1664
1665                         case O_IP_SRCPORT:
1666                         case O_IP_DSTPORT:
1667                                 /*
1668                                  * offset == 0 && proto != 0 is enough
1669                                  * to guarantee that we have a
1670                                  * packet with port info.
1671                                  */
1672                                 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1673                                     && offset == 0) {
1674                                         u_int16_t x =
1675                                             (cmd->opcode == O_IP_SRCPORT) ?
1676                                                 src_port : dst_port ;
1677                                         u_int16_t *p =
1678                                             ((ipfw_insn_u16 *)cmd)->ports;
1679                                         int i;
1680
1681                                         for (i = cmdlen - 1; !match && i>0;
1682                                             i--, p += 2)
1683                                                 match = (x>=p[0] && x<=p[1]);
1684                                 }
1685                                 break;
1686
1687                         case O_ICMPTYPE:
1688                                 match = (offset == 0 && proto==IPPROTO_ICMP &&
1689                                     icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
1690                                 break;
1691
1692 #ifdef INET6
1693                         case O_ICMP6TYPE:
1694                                 match = is_ipv6 && offset == 0 &&
1695                                     proto==IPPROTO_ICMPV6 &&
1696                                     icmp6type_match(
1697                                         ICMP6(ulp)->icmp6_type,
1698                                         (ipfw_insn_u32 *)cmd);
1699                                 break;
1700 #endif /* INET6 */
1701
1702                         case O_IPOPT:
1703                                 match = (is_ipv4 &&
1704                                     ipopts_match(ip, cmd) );
1705                                 break;
1706
1707                         case O_IPVER:
1708                                 match = (is_ipv4 &&
1709                                     cmd->arg1 == ip->ip_v);
1710                                 break;
1711
1712                         case O_IPID:
1713                         case O_IPLEN:
1714                         case O_IPTTL:
1715                                 if (is_ipv4) {  /* only for IP packets */
1716                                     uint16_t x;
1717                                     uint16_t *p;
1718                                     int i;
1719
1720                                     if (cmd->opcode == O_IPLEN)
1721                                         x = iplen;
1722                                     else if (cmd->opcode == O_IPTTL)
1723                                         x = ip->ip_ttl;
1724                                     else /* must be IPID */
1725                                         x = ntohs(ip->ip_id);
1726                                     if (cmdlen == 1) {
1727                                         match = (cmd->arg1 == x);
1728                                         break;
1729                                     }
1730                                     /* otherwise we have ranges */
1731                                     p = ((ipfw_insn_u16 *)cmd)->ports;
1732                                     i = cmdlen - 1;
1733                                     for (; !match && i>0; i--, p += 2)
1734                                         match = (x >= p[0] && x <= p[1]);
1735                                 }
1736                                 break;
1737
1738                         case O_IPPRECEDENCE:
1739                                 match = (is_ipv4 &&
1740                                     (cmd->arg1 == (ip->ip_tos & 0xe0)) );
1741                                 break;
1742
1743                         case O_IPTOS:
1744                                 match = (is_ipv4 &&
1745                                     flags_match(cmd, ip->ip_tos));
1746                                 break;
1747
1748                         case O_DSCP:
1749                             {
1750                                 uint32_t *p;
1751                                 uint16_t x;
1752
1753                                 p = ((ipfw_insn_u32 *)cmd)->d;
1754
1755                                 if (is_ipv4)
1756                                         x = ip->ip_tos >> 2;
1757                                 else if (is_ipv6) {
1758                                         uint8_t *v;
1759                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
1760                                         x = (*v & 0x0F) << 2;
1761                                         v++;
1762                                         x |= *v >> 6;
1763                                 } else
1764                                         break;
1765
1766                                 /* DSCP bitmask is stored as low_u32 high_u32 */
1767                                 if (x >= 32)
1768                                         match = *(p + 1) & (1 << (x - 32));
1769                                 else
1770                                         match = *p & (1 << x);
1771                             }
1772                                 break;
1773
1774                         case O_TCPDATALEN:
1775                                 if (proto == IPPROTO_TCP && offset == 0) {
1776                                     struct tcphdr *tcp;
1777                                     uint16_t x;
1778                                     uint16_t *p;
1779                                     int i;
1780
1781                                     tcp = TCP(ulp);
1782                                     x = iplen -
1783                                         ((ip->ip_hl + tcp->th_off) << 2);
1784                                     if (cmdlen == 1) {
1785                                         match = (cmd->arg1 == x);
1786                                         break;
1787                                     }
1788                                     /* otherwise we have ranges */
1789                                     p = ((ipfw_insn_u16 *)cmd)->ports;
1790                                     i = cmdlen - 1;
1791                                     for (; !match && i>0; i--, p += 2)
1792                                         match = (x >= p[0] && x <= p[1]);
1793                                 }
1794                                 break;
1795
1796                         case O_TCPFLAGS:
1797                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1798                                     flags_match(cmd, TCP(ulp)->th_flags));
1799                                 break;
1800
1801                         case O_TCPOPTS:
1802                                 if (proto == IPPROTO_TCP && offset == 0 && ulp){
1803                                         PULLUP_LEN(hlen, ulp,
1804                                             (TCP(ulp)->th_off << 2));
1805                                         match = tcpopts_match(TCP(ulp), cmd);
1806                                 }
1807                                 break;
1808
1809                         case O_TCPSEQ:
1810                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1811                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
1812                                         TCP(ulp)->th_seq);
1813                                 break;
1814
1815                         case O_TCPACK:
1816                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1817                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
1818                                         TCP(ulp)->th_ack);
1819                                 break;
1820
1821                         case O_TCPWIN:
1822                                 if (proto == IPPROTO_TCP && offset == 0) {
1823                                     uint16_t x;
1824                                     uint16_t *p;
1825                                     int i;
1826
1827                                     x = ntohs(TCP(ulp)->th_win);
1828                                     if (cmdlen == 1) {
1829                                         match = (cmd->arg1 == x);
1830                                         break;
1831                                     }
1832                                     /* Otherwise we have ranges. */
1833                                     p = ((ipfw_insn_u16 *)cmd)->ports;
1834                                     i = cmdlen - 1;
1835                                     for (; !match && i > 0; i--, p += 2)
1836                                         match = (x >= p[0] && x <= p[1]);
1837                                 }
1838                                 break;
1839
1840                         case O_ESTAB:
1841                                 /* reject packets which have SYN only */
1842                                 /* XXX should i also check for TH_ACK ? */
1843                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1844                                     (TCP(ulp)->th_flags &
1845                                      (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1846                                 break;
1847
1848                         case O_ALTQ: {
1849                                 struct pf_mtag *at;
1850                                 struct m_tag *mtag;
1851                                 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
1852
1853                                 /*
1854                                  * ALTQ uses mbuf tags from another
1855                                  * packet filtering system - pf(4).
1856                                  * We allocate a tag in its format
1857                                  * and fill it in, pretending to be pf(4).
1858                                  */
1859                                 match = 1;
1860                                 at = pf_find_mtag(m);
1861                                 if (at != NULL && at->qid != 0)
1862                                         break;
1863                                 mtag = m_tag_get(PACKET_TAG_PF,
1864                                     sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
1865                                 if (mtag == NULL) {
1866                                         /*
1867                                          * Let the packet fall back to the
1868                                          * default ALTQ.
1869                                          */
1870                                         break;
1871                                 }
1872                                 m_tag_prepend(m, mtag);
1873                                 at = (struct pf_mtag *)(mtag + 1);
1874                                 at->qid = altq->qid;
1875                                 at->hdr = ip;
1876                                 break;
1877                         }
1878
1879                         case O_LOG:
1880                                 ipfw_log(chain, f, hlen, args, m,
1881                                     oif, offset | ip6f_mf, tablearg, ip);
1882                                 match = 1;
1883                                 break;
1884
1885                         case O_PROB:
1886                                 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
1887                                 break;
1888
1889                         case O_VERREVPATH:
1890                                 /* Outgoing packets automatically pass/match */
1891                                 match = ((oif != NULL) ||
1892                                     (m->m_pkthdr.rcvif == NULL) ||
1893                                     (
1894 #ifdef INET6
1895                                     is_ipv6 ?
1896                                         verify_path6(&(args->f_id.src_ip6),
1897                                             m->m_pkthdr.rcvif, args->f_id.fib) :
1898 #endif
1899                                     verify_path(src_ip, m->m_pkthdr.rcvif,
1900                                         args->f_id.fib)));
1901                                 break;
1902
1903                         case O_VERSRCREACH:
1904                                 /* Outgoing packets automatically pass/match */
1905                                 match = (hlen > 0 && ((oif != NULL) ||
1906 #ifdef INET6
1907                                     is_ipv6 ?
1908                                         verify_path6(&(args->f_id.src_ip6),
1909                                             NULL, args->f_id.fib) :
1910 #endif
1911                                     verify_path(src_ip, NULL, args->f_id.fib)));
1912                                 break;
1913
1914                         case O_ANTISPOOF:
1915                                 /* Outgoing packets automatically pass/match */
1916                                 if (oif == NULL && hlen > 0 &&
1917                                     (  (is_ipv4 && in_localaddr(src_ip))
1918 #ifdef INET6
1919                                     || (is_ipv6 &&
1920                                         in6_localaddr(&(args->f_id.src_ip6)))
1921 #endif
1922                                     ))
1923                                         match =
1924 #ifdef INET6
1925                                             is_ipv6 ? verify_path6(
1926                                                 &(args->f_id.src_ip6),
1927                                                 m->m_pkthdr.rcvif,
1928                                                 args->f_id.fib) :
1929 #endif
1930                                             verify_path(src_ip,
1931                                                 m->m_pkthdr.rcvif,
1932                                                 args->f_id.fib);
1933                                 else
1934                                         match = 1;
1935                                 break;
1936
1937                         case O_IPSEC:
1938 #ifdef IPSEC
1939                                 match = (m_tag_find(m,
1940                                     PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
1941 #endif
1942                                 /* otherwise no match */
1943                                 break;
1944
1945 #ifdef INET6
1946                         case O_IP6_SRC:
1947                                 match = is_ipv6 &&
1948                                     IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
1949                                     &((ipfw_insn_ip6 *)cmd)->addr6);
1950                                 break;
1951
1952                         case O_IP6_DST:
1953                                 match = is_ipv6 &&
1954                                 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
1955                                     &((ipfw_insn_ip6 *)cmd)->addr6);
1956                                 break;
1957                         case O_IP6_SRC_MASK:
1958                         case O_IP6_DST_MASK:
1959                                 if (is_ipv6) {
1960                                         int i = cmdlen - 1;
1961                                         struct in6_addr p;
1962                                         struct in6_addr *d =
1963                                             &((ipfw_insn_ip6 *)cmd)->addr6;
1964
1965                                         for (; !match && i > 0; d += 2,
1966                                             i -= F_INSN_SIZE(struct in6_addr)
1967                                             * 2) {
1968                                                 p = (cmd->opcode ==
1969                                                     O_IP6_SRC_MASK) ?
1970                                                     args->f_id.src_ip6:
1971                                                     args->f_id.dst_ip6;
1972                                                 APPLY_MASK(&p, &d[1]);
1973                                                 match =
1974                                                     IN6_ARE_ADDR_EQUAL(&d[0],
1975                                                     &p);
1976                                         }
1977                                 }
1978                                 break;
1979
1980                         case O_FLOW6ID:
1981                                 match = is_ipv6 &&
1982                                     flow6id_match(args->f_id.flow_id6,
1983                                     (ipfw_insn_u32 *) cmd);
1984                                 break;
1985
1986                         case O_EXT_HDR:
1987                                 match = is_ipv6 &&
1988                                     (ext_hd & ((ipfw_insn *) cmd)->arg1);
1989                                 break;
1990
1991                         case O_IP6:
1992                                 match = is_ipv6;
1993                                 break;
1994 #endif
1995
1996                         case O_IP4:
1997                                 match = is_ipv4;
1998                                 break;
1999
2000                         case O_TAG: {
2001                                 struct m_tag *mtag;
2002                                 uint32_t tag = TARG(cmd->arg1, tag);
2003
2004                                 /* Packet is already tagged with this tag? */
2005                                 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
2006
2007                                 /* We have `untag' action when F_NOT flag is
2008                                  * present. And we must remove this mtag from
2009                                  * mbuf and reset `match' to zero (`match' will
2010                                  * be inversed later).
2011                                  * Otherwise we should allocate new mtag and
2012                                  * push it into mbuf.
2013                                  */
2014                                 if (cmd->len & F_NOT) { /* `untag' action */
2015                                         if (mtag != NULL)
2016                                                 m_tag_delete(m, mtag);
2017                                         match = 0;
2018                                 } else {
2019                                         if (mtag == NULL) {
2020                                                 mtag = m_tag_alloc( MTAG_IPFW,
2021                                                     tag, 0, M_NOWAIT);
2022                                                 if (mtag != NULL)
2023                                                         m_tag_prepend(m, mtag);
2024                                         }
2025                                         match = 1;
2026                                 }
2027                                 break;
2028                         }
2029
2030                         case O_FIB: /* try match the specified fib */
2031                                 if (args->f_id.fib == cmd->arg1)
2032                                         match = 1;
2033                                 break;
2034
2035                         case O_SOCKARG: {
2036 #ifndef USERSPACE       /* not supported in userspace */
2037                                 struct inpcb *inp = args->inp;
2038                                 struct inpcbinfo *pi;
2039
2040                                 if (is_ipv6) /* XXX can we remove this ? */
2041                                         break;
2042
2043                                 if (proto == IPPROTO_TCP)
2044                                         pi = &V_tcbinfo;
2045                                 else if (proto == IPPROTO_UDP)
2046                                         pi = &V_udbinfo;
2047                                 else
2048                                         break;
2049
2050                                 /*
2051                                  * XXXRW: so_user_cookie should almost
2052                                  * certainly be inp_user_cookie?
2053                                  */
2054
2055                                 /* For incoming packet, lookup up the
2056                                 inpcb using the src/dest ip/port tuple */
2057                                 if (inp == NULL) {
2058                                         inp = in_pcblookup(pi,
2059                                                 src_ip, htons(src_port),
2060                                                 dst_ip, htons(dst_port),
2061                                                 INPLOOKUP_RLOCKPCB, NULL);
2062                                         if (inp != NULL) {
2063                                                 tablearg =
2064                                                     inp->inp_socket->so_user_cookie;
2065                                                 if (tablearg)
2066                                                         match = 1;
2067                                                 INP_RUNLOCK(inp);
2068                                         }
2069                                 } else {
2070                                         if (inp->inp_socket) {
2071                                                 tablearg =
2072                                                     inp->inp_socket->so_user_cookie;
2073                                                 if (tablearg)
2074                                                         match = 1;
2075                                         }
2076                                 }
2077 #endif /* !USERSPACE */
2078                                 break;
2079                         }
2080
2081                         case O_TAGGED: {
2082                                 struct m_tag *mtag;
2083                                 uint32_t tag = TARG(cmd->arg1, tag);
2084
2085                                 if (cmdlen == 1) {
2086                                         match = m_tag_locate(m, MTAG_IPFW,
2087                                             tag, NULL) != NULL;
2088                                         break;
2089                                 }
2090
2091                                 /* we have ranges */
2092                                 for (mtag = m_tag_first(m);
2093                                     mtag != NULL && !match;
2094                                     mtag = m_tag_next(m, mtag)) {
2095                                         uint16_t *p;
2096                                         int i;
2097
2098                                         if (mtag->m_tag_cookie != MTAG_IPFW)
2099                                                 continue;
2100
2101                                         p = ((ipfw_insn_u16 *)cmd)->ports;
2102                                         i = cmdlen - 1;
2103                                         for(; !match && i > 0; i--, p += 2)
2104                                                 match =
2105                                                     mtag->m_tag_id >= p[0] &&
2106                                                     mtag->m_tag_id <= p[1];
2107                                 }
2108                                 break;
2109                         }
2110
2111                         /*
2112                          * The second set of opcodes represents 'actions',
2113                          * i.e. the terminal part of a rule once the packet
2114                          * matches all previous patterns.
2115                          * Typically there is only one action for each rule,
2116                          * and the opcode is stored at the end of the rule
2117                          * (but there are exceptions -- see below).
2118                          *
2119                          * In general, here we set retval and terminate the
2120                          * outer loop (would be a 'break 3' in some language,
2121                          * but we need to set l=0, done=1)
2122                          *
2123                          * Exceptions:
2124                          * O_COUNT and O_SKIPTO actions:
2125                          *   instead of terminating, we jump to the next rule
2126                          *   (setting l=0), or to the SKIPTO target (setting
2127                          *   f/f_len, cmd and l as needed), respectively.
2128                          *
2129                          * O_TAG, O_LOG and O_ALTQ action parameters:
2130                          *   perform some action and set match = 1;
2131                          *
2132                          * O_LIMIT and O_KEEP_STATE: these opcodes are
2133                          *   not real 'actions', and are stored right
2134                          *   before the 'action' part of the rule.
2135                          *   These opcodes try to install an entry in the
2136                          *   state tables; if successful, we continue with
2137                          *   the next opcode (match=1; break;), otherwise
2138                          *   the packet must be dropped (set retval,
2139                          *   break loops with l=0, done=1)
2140                          *
2141                          * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2142                          *   cause a lookup of the state table, and a jump
2143                          *   to the 'action' part of the parent rule
2144                          *   if an entry is found, or
2145                          *   (CHECK_STATE only) a jump to the next rule if
2146                          *   the entry is not found.
2147                          *   The result of the lookup is cached so that
2148                          *   further instances of these opcodes become NOPs.
2149                          *   The jump to the next rule is done by setting
2150                          *   l=0, cmdlen=0.
2151                          */
2152                         case O_LIMIT:
2153                         case O_KEEP_STATE:
2154                                 if (ipfw_install_state(chain, f,
2155                                     (ipfw_insn_limit *)cmd, args, tablearg)) {
2156                                         /* error or limit violation */
2157                                         retval = IP_FW_DENY;
2158                                         l = 0;  /* exit inner loop */
2159                                         done = 1; /* exit outer loop */
2160                                 }
2161                                 match = 1;
2162                                 break;
2163
2164                         case O_PROBE_STATE:
2165                         case O_CHECK_STATE:
2166                                 /*
2167                                  * dynamic rules are checked at the first
2168                                  * keep-state or check-state occurrence,
2169                                  * with the result being stored in dyn_dir
2170                                  * and dyn_name.
2171                                  * The compiler introduces a PROBE_STATE
2172                                  * instruction for us when we have a
2173                                  * KEEP_STATE (because PROBE_STATE needs
2174                                  * to be run first).
2175                                  *
2176                                  * (dyn_dir == MATCH_UNKNOWN) means this is
2177                                  * first lookup for such f_id. Do lookup.
2178                                  *
2179                                  * (dyn_dir != MATCH_UNKNOWN &&
2180                                  *  dyn_name != 0 && dyn_name != cmd->arg1)
2181                                  * means previous lookup didn't find dynamic
2182                                  * rule for specific state name and current
2183                                  * lookup will search rule with another state
2184                                  * name. Redo lookup.
2185                                  *
2186                                  * (dyn_dir != MATCH_UNKNOWN && dyn_name == 0)
2187                                  * means previous lookup was for `any' name
2188                                  * and it didn't find rule. No need to do
2189                                  * lookup again.
2190                                  */
2191                                 if ((dyn_dir == MATCH_UNKNOWN ||
2192                                     (dyn_name != 0 &&
2193                                     dyn_name != cmd->arg1)) &&
2194                                     (q = ipfw_lookup_dyn_rule(&args->f_id,
2195                                      &dyn_dir, proto == IPPROTO_TCP ?
2196                                      TCP(ulp): NULL,
2197                                      (dyn_name = cmd->arg1))) != NULL) {
2198                                         /*
2199                                          * Found dynamic entry, update stats
2200                                          * and jump to the 'action' part of
2201                                          * the parent rule by setting
2202                                          * f, cmd, l and clearing cmdlen.
2203                                          */
2204                                         IPFW_INC_DYN_COUNTER(q, pktlen);
2205                                         /* XXX we would like to have f_pos
2206                                          * readily accessible in the dynamic
2207                                          * rule, instead of having to
2208                                          * lookup q->rule.
2209                                          */
2210                                         f = q->rule;
2211                                         f_pos = ipfw_find_rule(chain,
2212                                                 f->rulenum, f->id);
2213                                         cmd = ACTION_PTR(f);
2214                                         l = f->cmd_len - f->act_ofs;
2215                                         ipfw_dyn_unlock(q);
2216                                         cmdlen = 0;
2217                                         match = 1;
2218                                         break;
2219                                 }
2220                                 /*
2221                                  * Dynamic entry not found. If CHECK_STATE,
2222                                  * skip to next rule, if PROBE_STATE just
2223                                  * ignore and continue with next opcode.
2224                                  */
2225                                 if (cmd->opcode == O_CHECK_STATE)
2226                                         l = 0;  /* exit inner loop */
2227                                 match = 1;
2228                                 break;
2229
2230                         case O_ACCEPT:
2231                                 retval = 0;     /* accept */
2232                                 l = 0;          /* exit inner loop */
2233                                 done = 1;       /* exit outer loop */
2234                                 break;
2235
2236                         case O_PIPE:
2237                         case O_QUEUE:
2238                                 set_match(args, f_pos, chain);
2239                                 args->rule.info = TARG(cmd->arg1, pipe);
2240                                 if (cmd->opcode == O_PIPE)
2241                                         args->rule.info |= IPFW_IS_PIPE;
2242                                 if (V_fw_one_pass)
2243                                         args->rule.info |= IPFW_ONEPASS;
2244                                 retval = IP_FW_DUMMYNET;
2245                                 l = 0;          /* exit inner loop */
2246                                 done = 1;       /* exit outer loop */
2247                                 break;
2248
2249                         case O_DIVERT:
2250                         case O_TEE:
2251                                 if (args->eh) /* not on layer 2 */
2252                                     break;
2253                                 /* otherwise this is terminal */
2254                                 l = 0;          /* exit inner loop */
2255                                 done = 1;       /* exit outer loop */
2256                                 retval = (cmd->opcode == O_DIVERT) ?
2257                                         IP_FW_DIVERT : IP_FW_TEE;
2258                                 set_match(args, f_pos, chain);
2259                                 args->rule.info = TARG(cmd->arg1, divert);
2260                                 break;
2261
2262                         case O_COUNT:
2263                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2264                                 l = 0;          /* exit inner loop */
2265                                 break;
2266
2267                         case O_SKIPTO:
2268                             IPFW_INC_RULE_COUNTER(f, pktlen);
2269                             f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2270                             /*
2271                              * Skip disabled rules, and re-enter
2272                              * the inner loop with the correct
2273                              * f_pos, f, l and cmd.
2274                              * Also clear cmdlen and skip_or
2275                              */
2276                             for (; f_pos < chain->n_rules - 1 &&
2277                                     (V_set_disable &
2278                                      (1 << chain->map[f_pos]->set));
2279                                     f_pos++)
2280                                 ;
2281                             /* Re-enter the inner loop at the skipto rule. */
2282                             f = chain->map[f_pos];
2283                             l = f->cmd_len;
2284                             cmd = f->cmd;
2285                             match = 1;
2286                             cmdlen = 0;
2287                             skip_or = 0;
2288                             continue;
2289                             break;      /* not reached */
2290
2291                         case O_CALLRETURN: {
2292                                 /*
2293                                  * Implementation of `subroutine' call/return,
2294                                  * in the stack carried in an mbuf tag. This
2295                                  * is different from `skipto' in that any call
2296                                  * address is possible (`skipto' must prevent
2297                                  * backward jumps to avoid endless loops).
2298                                  * We have `return' action when F_NOT flag is
2299                                  * present. The `m_tag_id' field is used as
2300                                  * stack pointer.
2301                                  */
2302                                 struct m_tag *mtag;
2303                                 uint16_t jmpto, *stack;
2304
2305 #define IS_CALL         ((cmd->len & F_NOT) == 0)
2306 #define IS_RETURN       ((cmd->len & F_NOT) != 0)
2307                                 /*
2308                                  * Hand-rolled version of m_tag_locate() with
2309                                  * wildcard `type'.
2310                                  * If not already tagged, allocate new tag.
2311                                  */
2312                                 mtag = m_tag_first(m);
2313                                 while (mtag != NULL) {
2314                                         if (mtag->m_tag_cookie ==
2315                                             MTAG_IPFW_CALL)
2316                                                 break;
2317                                         mtag = m_tag_next(m, mtag);
2318                                 }
2319                                 if (mtag == NULL && IS_CALL) {
2320                                         mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2321                                             IPFW_CALLSTACK_SIZE *
2322                                             sizeof(uint16_t), M_NOWAIT);
2323                                         if (mtag != NULL)
2324                                                 m_tag_prepend(m, mtag);
2325                                 }
2326
2327                                 /*
2328                                  * On error both `call' and `return' just
2329                                  * continue with next rule.
2330                                  */
2331                                 if (IS_RETURN && (mtag == NULL ||
2332                                     mtag->m_tag_id == 0)) {
2333                                         l = 0;          /* exit inner loop */
2334                                         break;
2335                                 }
2336                                 if (IS_CALL && (mtag == NULL ||
2337                                     mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2338                                         printf("ipfw: call stack error, "
2339                                             "go to next rule\n");
2340                                         l = 0;          /* exit inner loop */
2341                                         break;
2342                                 }
2343
2344                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2345                                 stack = (uint16_t *)(mtag + 1);
2346
2347                                 /*
2348                                  * The `call' action may use cached f_pos
2349                                  * (in f->next_rule), whose version is written
2350                                  * in f->next_rule.
2351                                  * The `return' action, however, doesn't have
2352                                  * fixed jump address in cmd->arg1 and can't use
2353                                  * cache.
2354                                  */
2355                                 if (IS_CALL) {
2356                                         stack[mtag->m_tag_id] = f->rulenum;
2357                                         mtag->m_tag_id++;
2358                                         f_pos = JUMP(chain, f, cmd->arg1,
2359                                             tablearg, 1);
2360                                 } else {        /* `return' action */
2361                                         mtag->m_tag_id--;
2362                                         jmpto = stack[mtag->m_tag_id] + 1;
2363                                         f_pos = ipfw_find_rule(chain, jmpto, 0);
2364                                 }
2365
2366                                 /*
2367                                  * Skip disabled rules, and re-enter
2368                                  * the inner loop with the correct
2369                                  * f_pos, f, l and cmd.
2370                                  * Also clear cmdlen and skip_or
2371                                  */
2372                                 for (; f_pos < chain->n_rules - 1 &&
2373                                     (V_set_disable &
2374                                     (1 << chain->map[f_pos]->set)); f_pos++)
2375                                         ;
2376                                 /* Re-enter the inner loop at the dest rule. */
2377                                 f = chain->map[f_pos];
2378                                 l = f->cmd_len;
2379                                 cmd = f->cmd;
2380                                 cmdlen = 0;
2381                                 skip_or = 0;
2382                                 continue;
2383                                 break;  /* NOTREACHED */
2384                         }
2385 #undef IS_CALL
2386 #undef IS_RETURN
2387
2388                         case O_REJECT:
2389                                 /*
2390                                  * Drop the packet and send a reject notice
2391                                  * if the packet is not ICMP (or is an ICMP
2392                                  * query), and it is not multicast/broadcast.
2393                                  */
2394                                 if (hlen > 0 && is_ipv4 && offset == 0 &&
2395                                     (proto != IPPROTO_ICMP ||
2396                                      is_icmp_query(ICMP(ulp))) &&
2397                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2398                                     !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2399                                         send_reject(args, cmd->arg1, iplen, ip);
2400                                         m = args->m;
2401                                 }
2402                                 /* FALLTHROUGH */
2403 #ifdef INET6
2404                         case O_UNREACH6:
2405                                 if (hlen > 0 && is_ipv6 &&
2406                                     ((offset & IP6F_OFF_MASK) == 0) &&
2407                                     (proto != IPPROTO_ICMPV6 ||
2408                                      (is_icmp6_query(icmp6_type) == 1)) &&
2409                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2410                                     !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2411                                         send_reject6(
2412                                             args, cmd->arg1, hlen,
2413                                             (struct ip6_hdr *)ip);
2414                                         m = args->m;
2415                                 }
2416                                 /* FALLTHROUGH */
2417 #endif
2418                         case O_DENY:
2419                                 retval = IP_FW_DENY;
2420                                 l = 0;          /* exit inner loop */
2421                                 done = 1;       /* exit outer loop */
2422                                 break;
2423
2424                         case O_FORWARD_IP:
2425                                 if (args->eh)   /* not valid on layer2 pkts */
2426                                         break;
2427                                 if (q == NULL || q->rule != f ||
2428                                     dyn_dir == MATCH_FORWARD) {
2429                                     struct sockaddr_in *sa;
2430
2431                                     sa = &(((ipfw_insn_sa *)cmd)->sa);
2432                                     if (sa->sin_addr.s_addr == INADDR_ANY) {
2433 #ifdef INET6
2434                                         /*
2435                                          * We use O_FORWARD_IP opcode for
2436                                          * fwd rule with tablearg, but tables
2437                                          * now support IPv6 addresses. And
2438                                          * when we are inspecting IPv6 packet,
2439                                          * we can use nh6 field from
2440                                          * table_value as next_hop6 address.
2441                                          */
2442                                         if (is_ipv6) {
2443                                                 struct sockaddr_in6 *sa6;
2444
2445                                                 sa6 = args->next_hop6 =
2446                                                     &args->hopstore6;
2447                                                 sa6->sin6_family = AF_INET6;
2448                                                 sa6->sin6_len = sizeof(*sa6);
2449                                                 sa6->sin6_addr = TARG_VAL(
2450                                                     chain, tablearg, nh6);
2451                                                 /*
2452                                                  * Set sin6_scope_id only for
2453                                                  * link-local unicast addresses.
2454                                                  */
2455                                                 if (IN6_IS_ADDR_LINKLOCAL(
2456                                                     &sa6->sin6_addr))
2457                                                         sa6->sin6_scope_id =
2458                                                             TARG_VAL(chain,
2459                                                                 tablearg,
2460                                                                 zoneid);
2461                                         } else
2462 #endif
2463                                         {
2464                                                 sa = args->next_hop =
2465                                                     &args->hopstore;
2466                                                 sa->sin_family = AF_INET;
2467                                                 sa->sin_len = sizeof(*sa);
2468                                                 sa->sin_addr.s_addr = htonl(
2469                                                     TARG_VAL(chain, tablearg,
2470                                                     nh4));
2471                                         }
2472                                     } else {
2473                                         args->next_hop = sa;
2474                                     }
2475                                 }
2476                                 retval = IP_FW_PASS;
2477                                 l = 0;          /* exit inner loop */
2478                                 done = 1;       /* exit outer loop */
2479                                 break;
2480
2481 #ifdef INET6
2482                         case O_FORWARD_IP6:
2483                                 if (args->eh)   /* not valid on layer2 pkts */
2484                                         break;
2485                                 if (q == NULL || q->rule != f ||
2486                                     dyn_dir == MATCH_FORWARD) {
2487                                         struct sockaddr_in6 *sin6;
2488
2489                                         sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2490                                         args->next_hop6 = sin6;
2491                                 }
2492                                 retval = IP_FW_PASS;
2493                                 l = 0;          /* exit inner loop */
2494                                 done = 1;       /* exit outer loop */
2495                                 break;
2496 #endif
2497
2498                         case O_NETGRAPH:
2499                         case O_NGTEE:
2500                                 set_match(args, f_pos, chain);
2501                                 args->rule.info = TARG(cmd->arg1, netgraph);
2502                                 if (V_fw_one_pass)
2503                                         args->rule.info |= IPFW_ONEPASS;
2504                                 retval = (cmd->opcode == O_NETGRAPH) ?
2505                                     IP_FW_NETGRAPH : IP_FW_NGTEE;
2506                                 l = 0;          /* exit inner loop */
2507                                 done = 1;       /* exit outer loop */
2508                                 break;
2509
2510                         case O_SETFIB: {
2511                                 uint32_t fib;
2512
2513                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2514                                 fib = TARG(cmd->arg1, fib) & 0x7FFF;
2515                                 if (fib >= rt_numfibs)
2516                                         fib = 0;
2517                                 M_SETFIB(m, fib);
2518                                 args->f_id.fib = fib;
2519                                 l = 0;          /* exit inner loop */
2520                                 break;
2521                         }
2522
2523                         case O_SETDSCP: {
2524                                 uint16_t code;
2525
2526                                 code = TARG(cmd->arg1, dscp) & 0x3F;
2527                                 l = 0;          /* exit inner loop */
2528                                 if (is_ipv4) {
2529                                         uint16_t old;
2530
2531                                         old = *(uint16_t *)ip;
2532                                         ip->ip_tos = (code << 2) |
2533                                             (ip->ip_tos & 0x03);
2534                                         ip->ip_sum = cksum_adjust(ip->ip_sum,
2535                                             old, *(uint16_t *)ip);
2536                                 } else if (is_ipv6) {
2537                                         uint8_t *v;
2538
2539                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
2540                                         *v = (*v & 0xF0) | (code >> 2);
2541                                         v++;
2542                                         *v = (*v & 0x3F) | ((code & 0x03) << 6);
2543                                 } else
2544                                         break;
2545
2546                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2547                                 break;
2548                         }
2549
2550                         case O_NAT:
2551                                 l = 0;          /* exit inner loop */
2552                                 done = 1;       /* exit outer loop */
2553                                 if (!IPFW_NAT_LOADED) {
2554                                     retval = IP_FW_DENY;
2555                                     break;
2556                                 }
2557
2558                                 struct cfg_nat *t;
2559                                 int nat_id;
2560
2561                                 set_match(args, f_pos, chain);
2562                                 /* Check if this is 'global' nat rule */
2563                                 if (cmd->arg1 == IP_FW_NAT44_GLOBAL) {
2564                                         retval = ipfw_nat_ptr(args, NULL, m);
2565                                         break;
2566                                 }
2567                                 t = ((ipfw_insn_nat *)cmd)->nat;
2568                                 if (t == NULL) {
2569                                         nat_id = TARG(cmd->arg1, nat);
2570                                         t = (*lookup_nat_ptr)(&chain->nat, nat_id);
2571
2572                                         if (t == NULL) {
2573                                             retval = IP_FW_DENY;
2574                                             break;
2575                                         }
2576                                         if (cmd->arg1 != IP_FW_TARG)
2577                                             ((ipfw_insn_nat *)cmd)->nat = t;
2578                                 }
2579                                 retval = ipfw_nat_ptr(args, t, m);
2580                                 break;
2581
2582                         case O_REASS: {
2583                                 int ip_off;
2584
2585                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2586                                 l = 0;  /* in any case exit inner loop */
2587                                 ip_off = ntohs(ip->ip_off);
2588
2589                                 /* if not fragmented, go to next rule */
2590                                 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
2591                                     break;
2592
2593                                 args->m = m = ip_reass(m);
2594
2595                                 /*
2596                                  * do IP header checksum fixup.
2597                                  */
2598                                 if (m == NULL) { /* fragment got swallowed */
2599                                     retval = IP_FW_DENY;
2600                                 } else { /* good, packet complete */
2601                                     int hlen;
2602
2603                                     ip = mtod(m, struct ip *);
2604                                     hlen = ip->ip_hl << 2;
2605                                     ip->ip_sum = 0;
2606                                     if (hlen == sizeof(struct ip))
2607                                         ip->ip_sum = in_cksum_hdr(ip);
2608                                     else
2609                                         ip->ip_sum = in_cksum(m, hlen);
2610                                     retval = IP_FW_REASS;
2611                                     set_match(args, f_pos, chain);
2612                                 }
2613                                 done = 1;       /* exit outer loop */
2614                                 break;
2615                         }
2616                         case O_EXTERNAL_ACTION:
2617                                 l = 0; /* in any case exit inner loop */
2618                                 retval = ipfw_run_eaction(chain, args,
2619                                     cmd, &done);
2620                                 /*
2621                                  * If both @retval and @done are zero,
2622                                  * consider this as rule matching and
2623                                  * update counters.
2624                                  */
2625                                 if (retval == 0 && done == 0) {
2626                                         IPFW_INC_RULE_COUNTER(f, pktlen);
2627                                         /*
2628                                          * Reset the result of the last
2629                                          * dynamic state lookup.
2630                                          * External action can change
2631                                          * @args content, and it may be
2632                                          * used for new state lookup later.
2633                                          */
2634                                         dyn_dir = MATCH_UNKNOWN;
2635                                 }
2636                                 break;
2637
2638                         default:
2639                                 panic("-- unknown opcode %d\n", cmd->opcode);
2640                         } /* end of switch() on opcodes */
2641                         /*
2642                          * if we get here with l=0, then match is irrelevant.
2643                          */
2644
2645                         if (cmd->len & F_NOT)
2646                                 match = !match;
2647
2648                         if (match) {
2649                                 if (cmd->len & F_OR)
2650                                         skip_or = 1;
2651                         } else {
2652                                 if (!(cmd->len & F_OR)) /* not an OR block, */
2653                                         break;          /* try next rule    */
2654                         }
2655
2656                 }       /* end of inner loop, scan opcodes */
2657 #undef PULLUP_LEN
2658
2659                 if (done)
2660                         break;
2661
2662 /* next_rule:; */       /* try next rule                */
2663
2664         }               /* end of outer for, scan rules */
2665
2666         if (done) {
2667                 struct ip_fw *rule = chain->map[f_pos];
2668                 /* Update statistics */
2669                 IPFW_INC_RULE_COUNTER(rule, pktlen);
2670         } else {
2671                 retval = IP_FW_DENY;
2672                 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2673         }
2674         IPFW_PF_RUNLOCK(chain);
2675 #ifdef __FreeBSD__
2676         if (ucred_cache != NULL)
2677                 crfree(ucred_cache);
2678 #endif
2679         return (retval);
2680
2681 pullup_failed:
2682         if (V_fw_verbose)
2683                 printf("ipfw: pullup failed\n");
2684         return (IP_FW_DENY);
2685 }
2686
2687 /*
2688  * Set maximum number of tables that can be used in given VNET ipfw instance.
2689  */
2690 #ifdef SYSCTL_NODE
2691 static int
2692 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
2693 {
2694         int error;
2695         unsigned int ntables;
2696
2697         ntables = V_fw_tables_max;
2698
2699         error = sysctl_handle_int(oidp, &ntables, 0, req);
2700         /* Read operation or some error */
2701         if ((error != 0) || (req->newptr == NULL))
2702                 return (error);
2703
2704         return (ipfw_resize_tables(&V_layer3_chain, ntables));
2705 }
2706
2707 /*
2708  * Switches table namespace between global and per-set.
2709  */
2710 static int
2711 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
2712 {
2713         int error;
2714         unsigned int sets;
2715
2716         sets = V_fw_tables_sets;
2717
2718         error = sysctl_handle_int(oidp, &sets, 0, req);
2719         /* Read operation or some error */
2720         if ((error != 0) || (req->newptr == NULL))
2721                 return (error);
2722
2723         return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
2724 }
2725 #endif
2726
2727 /*
2728  * Module and VNET glue
2729  */
2730
2731 /*
2732  * Stuff that must be initialised only on boot or module load
2733  */
2734 static int
2735 ipfw_init(void)
2736 {
2737         int error = 0;
2738
2739         /*
2740          * Only print out this stuff the first time around,
2741          * when called from the sysinit code.
2742          */
2743         printf("ipfw2 "
2744 #ifdef INET6
2745                 "(+ipv6) "
2746 #endif
2747                 "initialized, divert %s, nat %s, "
2748                 "default to %s, logging ",
2749 #ifdef IPDIVERT
2750                 "enabled",
2751 #else
2752                 "loadable",
2753 #endif
2754 #ifdef IPFIREWALL_NAT
2755                 "enabled",
2756 #else
2757                 "loadable",
2758 #endif
2759                 default_to_accept ? "accept" : "deny");
2760
2761         /*
2762          * Note: V_xxx variables can be accessed here but the vnet specific
2763          * initializer may not have been called yet for the VIMAGE case.
2764          * Tuneables will have been processed. We will print out values for
2765          * the default vnet.
2766          * XXX This should all be rationalized AFTER 8.0
2767          */
2768         if (V_fw_verbose == 0)
2769                 printf("disabled\n");
2770         else if (V_verbose_limit == 0)
2771                 printf("unlimited\n");
2772         else
2773                 printf("limited to %d packets/entry by default\n",
2774                     V_verbose_limit);
2775
2776         /* Check user-supplied table count for validness */
2777         if (default_fw_tables > IPFW_TABLES_MAX)
2778           default_fw_tables = IPFW_TABLES_MAX;
2779
2780         ipfw_init_sopt_handler();
2781         ipfw_init_obj_rewriter();
2782         ipfw_iface_init();
2783         return (error);
2784 }
2785
2786 /*
2787  * Called for the removal of the last instance only on module unload.
2788  */
2789 static void
2790 ipfw_destroy(void)
2791 {
2792
2793         ipfw_iface_destroy();
2794         ipfw_destroy_sopt_handler();
2795         ipfw_destroy_obj_rewriter();
2796         printf("IP firewall unloaded\n");
2797 }
2798
2799 /*
2800  * Stuff that must be initialized for every instance
2801  * (including the first of course).
2802  */
2803 static int
2804 vnet_ipfw_init(const void *unused)
2805 {
2806         int error, first;
2807         struct ip_fw *rule = NULL;
2808         struct ip_fw_chain *chain;
2809
2810         chain = &V_layer3_chain;
2811
2812         first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
2813
2814         /* First set up some values that are compile time options */
2815         V_autoinc_step = 100;   /* bounded to 1..1000 in add_rule() */
2816         V_fw_deny_unknown_exthdrs = 1;
2817 #ifdef IPFIREWALL_VERBOSE
2818         V_fw_verbose = 1;
2819 #endif
2820 #ifdef IPFIREWALL_VERBOSE_LIMIT
2821         V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2822 #endif
2823 #ifdef IPFIREWALL_NAT
2824         LIST_INIT(&chain->nat);
2825 #endif
2826
2827         /* Init shared services hash table */
2828         ipfw_init_srv(chain);
2829
2830         ipfw_init_counters();
2831         /* insert the default rule and create the initial map */
2832         chain->n_rules = 1;
2833         chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_WAITOK | M_ZERO);
2834         rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
2835
2836         /* Set initial number of tables */
2837         V_fw_tables_max = default_fw_tables;
2838         error = ipfw_init_tables(chain, first);
2839         if (error) {
2840                 printf("ipfw2: setting up tables failed\n");
2841                 free(chain->map, M_IPFW);
2842                 free(rule, M_IPFW);
2843                 return (ENOSPC);
2844         }
2845
2846         /* fill and insert the default rule */
2847         rule->act_ofs = 0;
2848         rule->rulenum = IPFW_DEFAULT_RULE;
2849         rule->cmd_len = 1;
2850         rule->set = RESVD_SET;
2851         rule->cmd[0].len = 1;
2852         rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
2853         chain->default_rule = chain->map[0] = rule;
2854         chain->id = rule->id = 1;
2855         /* Pre-calculate rules length for legacy dump format */
2856         chain->static_len = sizeof(struct ip_fw_rule0);
2857
2858         IPFW_LOCK_INIT(chain);
2859         ipfw_dyn_init(chain);
2860         ipfw_eaction_init(chain, first);
2861 #ifdef LINEAR_SKIPTO
2862         ipfw_init_skipto_cache(chain);
2863 #endif
2864         ipfw_bpf_init(first);
2865
2866         /* First set up some values that are compile time options */
2867         V_ipfw_vnet_ready = 1;          /* Open for business */
2868
2869         /*
2870          * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
2871          * Even if the latter two fail we still keep the module alive
2872          * because the sockopt and layer2 paths are still useful.
2873          * ipfw[6]_hook return 0 on success, ENOENT on failure,
2874          * so we can ignore the exact return value and just set a flag.
2875          *
2876          * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
2877          * changes in the underlying (per-vnet) variables trigger
2878          * immediate hook()/unhook() calls.
2879          * In layer2 we have the same behaviour, except that V_ether_ipfw
2880          * is checked on each packet because there are no pfil hooks.
2881          */
2882         V_ip_fw_ctl_ptr = ipfw_ctl3;
2883         error = ipfw_attach_hooks(1);
2884         return (error);
2885 }
2886
2887 /*
2888  * Called for the removal of each instance.
2889  */
2890 static int
2891 vnet_ipfw_uninit(const void *unused)
2892 {
2893         struct ip_fw *reap;
2894         struct ip_fw_chain *chain = &V_layer3_chain;
2895         int i, last;
2896
2897         V_ipfw_vnet_ready = 0; /* tell new callers to go away */
2898         /*
2899          * disconnect from ipv4, ipv6, layer2 and sockopt.
2900          * Then grab, release and grab again the WLOCK so we make
2901          * sure the update is propagated and nobody will be in.
2902          */
2903         (void)ipfw_attach_hooks(0 /* detach */);
2904         V_ip_fw_ctl_ptr = NULL;
2905
2906         last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
2907
2908         IPFW_UH_WLOCK(chain);
2909         IPFW_UH_WUNLOCK(chain);
2910
2911         ipfw_dyn_uninit(0);     /* run the callout_drain */
2912
2913         IPFW_UH_WLOCK(chain);
2914
2915         reap = NULL;
2916         IPFW_WLOCK(chain);
2917         for (i = 0; i < chain->n_rules; i++)
2918                 ipfw_reap_add(chain, &reap, chain->map[i]);
2919         free(chain->map, M_IPFW);
2920 #ifdef LINEAR_SKIPTO
2921         ipfw_destroy_skipto_cache(chain);
2922 #endif
2923         IPFW_WUNLOCK(chain);
2924         IPFW_UH_WUNLOCK(chain);
2925         ipfw_destroy_tables(chain, last);
2926         ipfw_eaction_uninit(chain, last);
2927         if (reap != NULL)
2928                 ipfw_reap_rules(reap);
2929         vnet_ipfw_iface_destroy(chain);
2930         ipfw_destroy_srv(chain);
2931         IPFW_LOCK_DESTROY(chain);
2932         ipfw_dyn_uninit(1);     /* free the remaining parts */
2933         ipfw_destroy_counters();
2934         ipfw_bpf_uninit(last);
2935         return (0);
2936 }
2937
2938 /*
2939  * Module event handler.
2940  * In general we have the choice of handling most of these events by the
2941  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
2942  * use the SYSINIT handlers as they are more capable of expressing the
2943  * flow of control during module and vnet operations, so this is just
2944  * a skeleton. Note there is no SYSINIT equivalent of the module
2945  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
2946  */
2947 static int
2948 ipfw_modevent(module_t mod, int type, void *unused)
2949 {
2950         int err = 0;
2951
2952         switch (type) {
2953         case MOD_LOAD:
2954                 /* Called once at module load or
2955                  * system boot if compiled in. */
2956                 break;
2957         case MOD_QUIESCE:
2958                 /* Called before unload. May veto unloading. */
2959                 break;
2960         case MOD_UNLOAD:
2961                 /* Called during unload. */
2962                 break;
2963         case MOD_SHUTDOWN:
2964                 /* Called during system shutdown. */
2965                 break;
2966         default:
2967                 err = EOPNOTSUPP;
2968                 break;
2969         }
2970         return err;
2971 }
2972
2973 static moduledata_t ipfwmod = {
2974         "ipfw",
2975         ipfw_modevent,
2976         0
2977 };
2978
2979 /* Define startup order. */
2980 #define IPFW_SI_SUB_FIREWALL    SI_SUB_PROTO_FIREWALL
2981 #define IPFW_MODEVENT_ORDER     (SI_ORDER_ANY - 255) /* On boot slot in here. */
2982 #define IPFW_MODULE_ORDER       (IPFW_MODEVENT_ORDER + 1) /* A little later. */
2983 #define IPFW_VNET_ORDER         (IPFW_MODEVENT_ORDER + 2) /* Later still. */
2984
2985 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
2986 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
2987 MODULE_VERSION(ipfw, 3);
2988 /* should declare some dependencies here */
2989
2990 /*
2991  * Starting up. Done in order after ipfwmod() has been called.
2992  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
2993  */
2994 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2995             ipfw_init, NULL);
2996 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2997             vnet_ipfw_init, NULL);
2998
2999 /*
3000  * Closing up shop. These are done in REVERSE ORDER, but still
3001  * after ipfwmod() has been called. Not called on reboot.
3002  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
3003  * or when the module is unloaded.
3004  */
3005 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3006             ipfw_destroy, NULL);
3007 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3008             vnet_ipfw_uninit, NULL);
3009 /* end of file */