sys/netinet/ip_fw2.c

   1 /*-
   2  * Copyright (c) 2002 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  * $FreeBSD$
  26  */
  27
  28 #define        DEB(x)
  29 #define        DDB(x) x
  30
  31 /*
  32  * Implement IP packet firewall (new version)
  33  */
  34
  35 #if !defined(KLD_MODULE)
  36 #include "opt_ipfw.h"
  37 #include "opt_ipdn.h"
  38 #include "opt_inet.h"
  39 #ifndef INET
  40 #error IPFIREWALL requires INET.
  41 #endif /* INET */
  42 #endif
  43 #include "opt_inet6.h"
  44 #include "opt_ipsec.h"
  45
  46 #include <sys/param.h>
  47 #include <sys/systm.h>
  48 #include <sys/condvar.h>
  49 #include <sys/malloc.h>
  50 #include <sys/mbuf.h>
  51 #include <sys/kernel.h>
  52 #include <sys/lock.h>
  53 #include <sys/jail.h>
  54 #include <sys/module.h>
  55 #include <sys/proc.h>
  56 #include <sys/rwlock.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/if.h>
  63 #include <net/radix.h>
  64 #include <net/route.h>
  65 #include <netinet/in.h>
  66 #include <netinet/in_systm.h>
  67 #include <netinet/in_var.h>
  68 #include <netinet/in_pcb.h>
  69 #include <netinet/ip.h>
  70 #include <netinet/ip_var.h>
  71 #include <netinet/ip_icmp.h>
  72 #include <netinet/ip_fw.h>
  73 #include <netinet/ip_divert.h>
  74 #include <netinet/ip_dummynet.h>
  75 #include <netinet/tcp.h>
  76 #include <netinet/tcp_timer.h>
  77 #include <netinet/tcp_var.h>
  78 #include <netinet/tcpip.h>
  79 #include <netinet/udp.h>
  80 #include <netinet/udp_var.h>
  81
  82 #include <netgraph/ng_ipfw.h>
  83
  84 #include <altq/if_altq.h>
  85
  86 #ifdef IPSEC
  87 #include <netinet6/ipsec.h>
  88 #endif
  89
  90 #include <netinet/ip6.h>
  91 #include <netinet/icmp6.h>
  92 #ifdef INET6
  93 #include <netinet6/scope6_var.h>
  94 #endif
  95
  96 #include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */
  97
  98 #include <machine/in_cksum.h>   /* XXX for in_cksum */
  99
 100 /*
 101  * set_disable contains one bit per set value (0..31).
 102  * If the bit is set, all rules with the corresponding set
 103  * are disabled. Set RESVD_SET(31) is reserved for the default rule
 104  * and rules that are not deleted by the flush command,
 105  * and CANNOT be disabled.
 106  * Rules in set RESVD_SET can only be deleted explicitly.
 107  */
 108 static u_int32_t set_disable;
 109
 110 static int fw_verbose;
 111 static int verbose_limit;
 112
 113 static struct callout ipfw_timeout;
 114 static uma_zone_t ipfw_dyn_rule_zone;
 115 #define IPFW_DEFAULT_RULE       65535
 116
 117 /*
 118  * Data structure to cache our ucred related
 119  * information. This structure only gets used if
 120  * the user specified UID/GID based constraints in
 121  * a firewall rule.
 122  */
 123 struct ip_fw_ugid {
 124         gid_t           fw_groups[NGROUPS];
 125         int             fw_ngroups;
 126         uid_t           fw_uid;
 127         int             fw_prid;
 128 };
 129
 130 #define IPFW_TABLES_MAX         128
 131 struct ip_fw_chain {
 132         struct ip_fw    *rules;         /* list of rules */
 133         struct ip_fw    *reap;          /* list of rules to reap */
 134         struct radix_node_head *tables[IPFW_TABLES_MAX];
 135         struct rwlock   rwmtx;
 136 };
 137 #define IPFW_LOCK_INIT(_chain) \
 138         rw_init(&(_chain)->rwmtx, "IPFW static rules")
 139 #define IPFW_LOCK_DESTROY(_chain)       rw_destroy(&(_chain)->rwmtx)
 140 #define IPFW_WLOCK_ASSERT(_chain)       do {                            \
 141         rw_assert(&(_chain)->rwmtx, RA_WLOCKED);                                        \
 142         NET_ASSERT_GIANT();                                             \
 143 } while (0)
 144
 145 #define IPFW_RLOCK(p) rw_rlock(&(p)->rwmtx)
 146 #define IPFW_RUNLOCK(p) rw_runlock(&(p)->rwmtx)
 147 #define IPFW_WLOCK(p) rw_wlock(&(p)->rwmtx)
 148 #define IPFW_WUNLOCK(p) rw_wunlock(&(p)->rwmtx)
 149
 150 /*
 151  * list of rules for layer 3
 152  */
 153 static struct ip_fw_chain layer3_chain;
 154
 155 MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's");
 156 MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables");
 157
 158 struct table_entry {
 159         struct radix_node       rn[2];
 160         struct sockaddr_in      addr, mask;
 161         u_int32_t               value;
 162 };
 163
 164 static int fw_debug = 1;
 165 static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */
 166
 167 extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS);
 168
 169 #ifdef SYSCTL_NODE
 170 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 171 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable,
 172     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &fw_enable, 0,
 173     ipfw_chg_hook, "I", "Enable ipfw");
 174 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW,
 175     &autoinc_step, 0, "Rule number autincrement step");
 176 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
 177     CTLFLAG_RW | CTLFLAG_SECURE3,
 178     &fw_one_pass, 0,
 179     "Only do a single pass through ipfw when using dummynet(4)");
 180 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW,
 181     &fw_debug, 0, "Enable printing of debug ip_fw statements");
 182 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose,
 183     CTLFLAG_RW | CTLFLAG_SECURE3,
 184     &fw_verbose, 0, "Log matches to ipfw rules");
 185 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW,
 186     &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged");
 187
 188 /*
 189  * Description of dynamic rules.
 190  *
 191  * Dynamic rules are stored in lists accessed through a hash table
 192  * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can
 193  * be modified through the sysctl variable dyn_buckets which is
 194  * updated when the table becomes empty.
 195  *
 196  * XXX currently there is only one list, ipfw_dyn.
 197  *
 198  * When a packet is received, its address fields are first masked
 199  * with the mask defined for the rule, then hashed, then matched
 200  * against the entries in the corresponding list.
 201  * Dynamic rules can be used for different purposes:
 202  *  + stateful rules;
 203  *  + enforcing limits on the number of sessions;
 204  *  + in-kernel NAT (not implemented yet)
 205  *
 206  * The lifetime of dynamic rules is regulated by dyn_*_lifetime,
 207  * measured in seconds and depending on the flags.
 208  *
 209  * The total number of dynamic rules is stored in dyn_count.
 210  * The max number of dynamic rules is dyn_max. When we reach
 211  * the maximum number of rules we do not create anymore. This is
 212  * done to avoid consuming too much memory, but also too much
 213  * time when searching on each packet (ideally, we should try instead
 214  * to put a limit on the length of the list on each bucket...).
 215  *
 216  * Each dynamic rule holds a pointer to the parent ipfw rule so
 217  * we know what action to perform. Dynamic rules are removed when
 218  * the parent rule is deleted. XXX we should make them survive.
 219  *
 220  * There are some limitations with dynamic rules -- we do not
 221  * obey the 'randomized match', and we do not do multiple
 222  * passes through the firewall. XXX check the latter!!!
 223  */
 224 static ipfw_dyn_rule **ipfw_dyn_v = NULL;
 225 static u_int32_t dyn_buckets = 256; /* must be power of 2 */
 226 static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */
 227
 228 static struct mtx ipfw_dyn_mtx;         /* mutex guarding dynamic rules */
 229 #define IPFW_DYN_LOCK_INIT() \
 230         mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF)
 231 #define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx)
 232 #define IPFW_DYN_LOCK()         mtx_lock(&ipfw_dyn_mtx)
 233 #define IPFW_DYN_UNLOCK()       mtx_unlock(&ipfw_dyn_mtx)
 234 #define IPFW_DYN_LOCK_ASSERT()  mtx_assert(&ipfw_dyn_mtx, MA_OWNED)
 235
 236 /*
 237  * Timeouts for various events in handing dynamic rules.
 238  */
 239 static u_int32_t dyn_ack_lifetime = 300;
 240 static u_int32_t dyn_syn_lifetime = 20;
 241 static u_int32_t dyn_fin_lifetime = 1;
 242 static u_int32_t dyn_rst_lifetime = 1;
 243 static u_int32_t dyn_udp_lifetime = 10;
 244 static u_int32_t dyn_short_lifetime = 5;
 245
 246 /*
 247  * Keepalives are sent if dyn_keepalive is set. They are sent every
 248  * dyn_keepalive_period seconds, in the last dyn_keepalive_interval
 249  * seconds of lifetime of a rule.
 250  * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower
 251  * than dyn_keepalive_period.
 252  */
 253
 254 static u_int32_t dyn_keepalive_interval = 20;
 255 static u_int32_t dyn_keepalive_period = 5;
 256 static u_int32_t dyn_keepalive = 1;     /* do send keepalives */
 257
 258 static u_int32_t static_count;  /* # of static rules */
 259 static u_int32_t static_len;    /* size in bytes of static rules */
 260 static u_int32_t dyn_count;             /* # of dynamic rules */
 261 static u_int32_t dyn_max = 4096;        /* max # of dynamic rules */
 262
 263 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW,
 264     &dyn_buckets, 0, "Number of dyn. buckets");
 265 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD,
 266     &curr_dyn_buckets, 0, "Current Number of dyn. buckets");
 267 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD,
 268     &dyn_count, 0, "Number of dyn. rules");
 269 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW,
 270     &dyn_max, 0, "Max number of dyn. rules");
 271 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD,
 272     &static_count, 0, "Number of static rules");
 273 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW,
 274     &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks");
 275 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW,
 276     &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn");
 277 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW,
 278     &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin");
 279 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW,
 280     &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst");
 281 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW,
 282     &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP");
 283 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW,
 284     &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations");
 285 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW,
 286     &dyn_keepalive, 0, "Enable keepalives for dyn. rules");
 287
 288 #ifdef INET6
 289 /*
 290  * IPv6 specific variables
 291  */
 292 SYSCTL_DECL(_net_inet6_ip6);
 293
 294 static struct sysctl_ctx_list ip6_fw_sysctl_ctx;
 295 static struct sysctl_oid *ip6_fw_sysctl_tree;
 296 #endif /* INET6 */
 297 #endif /* SYSCTL_NODE */
 298
 299 static int fw_deny_unknown_exthdrs = 1;
 300
 301
 302 /*
 303  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
 304  * Other macros just cast void * into the appropriate type
 305  */
 306 #define L3HDR(T, ip)    ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
 307 #define TCP(p)          ((struct tcphdr *)(p))
 308 #define UDP(p)          ((struct udphdr *)(p))
 309 #define ICMP(p)         ((struct icmphdr *)(p))
 310 #define ICMP6(p)        ((struct icmp6_hdr *)(p))
 311
 312 static __inline int
 313 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
 314 {
 315         int type = icmp->icmp_type;
 316
 317         return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
 318 }
 319
 320 #define TT      ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
 321     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
 322
 323 static int
 324 is_icmp_query(struct icmphdr *icmp)
 325 {
 326         int type = icmp->icmp_type;
 327
 328         return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
 329 }
 330 #undef TT
 331
 332 /*
 333  * The following checks use two arrays of 8 or 16 bits to store the
 334  * bits that we want set or clear, respectively. They are in the
 335  * low and high half of cmd->arg1 or cmd->d[0].
 336  *
 337  * We scan options and store the bits we find set. We succeed if
 338  *
 339  *      (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
 340  *
 341  * The code is sometimes optimized not to store additional variables.
 342  */
 343
 344 static int
 345 flags_match(ipfw_insn *cmd, u_int8_t bits)
 346 {
 347         u_char want_clear;
 348         bits = ~bits;
 349
 350         if ( ((cmd->arg1 & 0xff) & bits) != 0)
 351                 return 0; /* some bits we want set were clear */
 352         want_clear = (cmd->arg1 >> 8) & 0xff;
 353         if ( (want_clear & bits) != want_clear)
 354                 return 0; /* some bits we want clear were set */
 355         return 1;
 356 }
 357
 358 static int
 359 ipopts_match(struct ip *ip, ipfw_insn *cmd)
 360 {
 361         int optlen, bits = 0;
 362         u_char *cp = (u_char *)(ip + 1);
 363         int x = (ip->ip_hl << 2) - sizeof (struct ip);
 364
 365         for (; x > 0; x -= optlen, cp += optlen) {
 366                 int opt = cp[IPOPT_OPTVAL];
 367
 368                 if (opt == IPOPT_EOL)
 369                         break;
 370                 if (opt == IPOPT_NOP)
 371                         optlen = 1;
 372                 else {
 373                         optlen = cp[IPOPT_OLEN];
 374                         if (optlen <= 0 || optlen > x)
 375                                 return 0; /* invalid or truncated */
 376                 }
 377                 switch (opt) {
 378
 379                 default:
 380                         break;
 381
 382                 case IPOPT_LSRR:
 383                         bits |= IP_FW_IPOPT_LSRR;
 384                         break;
 385
 386                 case IPOPT_SSRR:
 387                         bits |= IP_FW_IPOPT_SSRR;
 388                         break;
 389
 390                 case IPOPT_RR:
 391                         bits |= IP_FW_IPOPT_RR;
 392                         break;
 393
 394                 case IPOPT_TS:
 395                         bits |= IP_FW_IPOPT_TS;
 396                         break;
 397                 }
 398         }
 399         return (flags_match(cmd, bits));
 400 }
 401
 402 static int
 403 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
 404 {
 405         int optlen, bits = 0;
 406         u_char *cp = (u_char *)(tcp + 1);
 407         int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
 408
 409         for (; x > 0; x -= optlen, cp += optlen) {
 410                 int opt = cp[0];
 411                 if (opt == TCPOPT_EOL)
 412                         break;
 413                 if (opt == TCPOPT_NOP)
 414                         optlen = 1;
 415                 else {
 416                         optlen = cp[1];
 417                         if (optlen <= 0)
 418                                 break;
 419                 }
 420
 421                 switch (opt) {
 422
 423                 default:
 424                         break;
 425
 426                 case TCPOPT_MAXSEG:
 427                         bits |= IP_FW_TCPOPT_MSS;
 428                         break;
 429
 430                 case TCPOPT_WINDOW:
 431                         bits |= IP_FW_TCPOPT_WINDOW;
 432                         break;
 433
 434                 case TCPOPT_SACK_PERMITTED:
 435                 case TCPOPT_SACK:
 436                         bits |= IP_FW_TCPOPT_SACK;
 437                         break;
 438
 439                 case TCPOPT_TIMESTAMP:
 440                         bits |= IP_FW_TCPOPT_TS;
 441                         break;
 442
 443                 }
 444         }
 445         return (flags_match(cmd, bits));
 446 }
 447
 448 static int
 449 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd)
 450 {
 451         if (ifp == NULL)        /* no iface with this packet, match fails */
 452                 return 0;
 453         /* Check by name or by IP address */
 454         if (cmd->name[0] != '\0') { /* match by name */
 455                 /* Check name */
 456                 if (cmd->p.glob) {
 457                         if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
 458                                 return(1);
 459                 } else {
 460                         if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
 461                                 return(1);
 462                 }
 463         } else {
 464                 struct ifaddr *ia;
 465
 466                 /* XXX lock? */
 467                 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
 468                         if (ia->ifa_addr == NULL)
 469                                 continue;
 470                         if (ia->ifa_addr->sa_family != AF_INET)
 471                                 continue;
 472                         if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
 473                             (ia->ifa_addr))->sin_addr.s_addr)
 474                                 return(1);      /* match */
 475                 }
 476         }
 477         return(0);      /* no match, fail ... */
 478 }
 479
 480 /*
 481  * The verify_path function checks if a route to the src exists and
 482  * if it is reachable via ifp (when provided).
 483  *
 484  * The 'verrevpath' option checks that the interface that an IP packet
 485  * arrives on is the same interface that traffic destined for the
 486  * packet's source address would be routed out of.  The 'versrcreach'
 487  * option just checks that the source address is reachable via any route
 488  * (except default) in the routing table.  These two are a measure to block
 489  * forged packets.  This is also commonly known as "anti-spoofing" or Unicast
 490  * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
 491  * is purposely reminiscent of the Cisco IOS command,
 492  *
 493  *   ip verify unicast reverse-path
 494  *   ip verify unicast source reachable-via any
 495  *
 496  * which implements the same functionality. But note that syntax is
 497  * misleading. The check may be performed on all IP packets whether unicast,
 498  * multicast, or broadcast.
 499  */
 500 static int
 501 verify_path(struct in_addr src, struct ifnet *ifp)
 502 {
 503         struct route ro;
 504         struct sockaddr_in *dst;
 505
 506         bzero(&ro, sizeof(ro));
 507
 508         dst = (struct sockaddr_in *)&(ro.ro_dst);
 509         dst->sin_family = AF_INET;
 510         dst->sin_len = sizeof(*dst);
 511         dst->sin_addr = src;
 512         rtalloc_ign(&ro, RTF_CLONING);
 513
 514         if (ro.ro_rt == NULL)
 515                 return 0;
 516
 517         /*
 518          * If ifp is provided, check for equality with rtentry.
 519          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 520          * in order to pass packets injected back by if_simloop():
 521          * if useloopback == 1 routing entry (via lo0) for our own address
 522          * may exist, so we need to handle routing assymetry.
 523          */
 524         if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
 525                 RTFREE(ro.ro_rt);
 526                 return 0;
 527         }
 528
 529         /* if no ifp provided, check if rtentry is not default route */
 530         if (ifp == NULL &&
 531              satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) {
 532                 RTFREE(ro.ro_rt);
 533                 return 0;
 534         }
 535
 536         /* or if this is a blackhole/reject route */
 537         if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
 538                 RTFREE(ro.ro_rt);
 539                 return 0;
 540         }
 541
 542         /* found valid route */
 543         RTFREE(ro.ro_rt);
 544         return 1;
 545 }
 546
 547 #ifdef INET6
 548 /*
 549  * ipv6 specific rules here...
 550  */
 551 static __inline int
 552 icmp6type_match (int type, ipfw_insn_u32 *cmd)
 553 {
 554         return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
 555 }
 556
 557 static int
 558 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
 559 {
 560         int i;
 561         for (i=0; i <= cmd->o.arg1; ++i )
 562                 if (curr_flow == cmd->d[i] )
 563                         return 1;
 564         return 0;
 565 }
 566
 567 /* support for IP6_*_ME opcodes */
 568 static int
 569 search_ip6_addr_net (struct in6_addr * ip6_addr)
 570 {
 571         struct ifnet *mdc;
 572         struct ifaddr *mdc2;
 573         struct in6_ifaddr *fdm;
 574         struct in6_addr copia;
 575
 576         TAILQ_FOREACH(mdc, &ifnet, if_link)
 577                 for (mdc2 = mdc->if_addrlist.tqh_first; mdc2;
 578                     mdc2 = mdc2->ifa_list.tqe_next) {
 579                         if (!mdc2->ifa_addr)
 580                                 continue;
 581                         if (mdc2->ifa_addr->sa_family == AF_INET6) {
 582                                 fdm = (struct in6_ifaddr *)mdc2;
 583                                 copia = fdm->ia_addr.sin6_addr;
 584                                 /* need for leaving scope_id in the sock_addr */
 585                                 in6_clearscope(&copia);
 586                                 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia))
 587                                         return 1;
 588                         }
 589                 }
 590         return 0;
 591 }
 592
 593 static int
 594 verify_path6(struct in6_addr *src, struct ifnet *ifp)
 595 {
 596         struct route_in6 ro;
 597         struct sockaddr_in6 *dst;
 598
 599         bzero(&ro, sizeof(ro));
 600
 601         dst = (struct sockaddr_in6 * )&(ro.ro_dst);
 602         dst->sin6_family = AF_INET6;
 603         dst->sin6_len = sizeof(*dst);
 604         dst->sin6_addr = *src;
 605         rtalloc_ign((struct route *)&ro, RTF_CLONING);
 606
 607         if (ro.ro_rt == NULL)
 608                 return 0;
 609
 610         /*
 611          * if ifp is provided, check for equality with rtentry
 612          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 613          * to support the case of sending packets to an address of our own.
 614          * (where the former interface is the first argument of if_simloop()
 615          *  (=ifp), the latter is lo0)
 616          */
 617         if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
 618                 RTFREE(ro.ro_rt);
 619                 return 0;
 620         }
 621
 622         /* if no ifp provided, check if rtentry is not default route */
 623         if (ifp == NULL &&
 624             IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) {
 625                 RTFREE(ro.ro_rt);
 626                 return 0;
 627         }
 628
 629         /* or if this is a blackhole/reject route */
 630         if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
 631                 RTFREE(ro.ro_rt);
 632                 return 0;
 633         }
 634
 635         /* found valid route */
 636         RTFREE(ro.ro_rt);
 637         return 1;
 638
 639 }
 640 static __inline int
 641 hash_packet6(struct ipfw_flow_id *id)
 642 {
 643         u_int32_t i;
 644         i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^
 645             (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^
 646             (id->src_ip6.__u6_addr.__u6_addr32[2]) ^
 647             (id->src_ip6.__u6_addr.__u6_addr32[3]) ^
 648             (id->dst_port) ^ (id->src_port);
 649         return i;
 650 }
 651
 652 static int
 653 is_icmp6_query(int icmp6_type)
 654 {
 655         if ((icmp6_type <= ICMP6_MAXTYPE) &&
 656             (icmp6_type == ICMP6_ECHO_REQUEST ||
 657             icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
 658             icmp6_type == ICMP6_WRUREQUEST ||
 659             icmp6_type == ICMP6_FQDN_QUERY ||
 660             icmp6_type == ICMP6_NI_QUERY))
 661                 return (1);
 662
 663         return (0);
 664 }
 665
 666 static void
 667 send_reject6(struct ip_fw_args *args, int code, u_short offset, u_int hlen)
 668 {
 669         if (code == ICMP6_UNREACH_RST && offset == 0 &&
 670             args->f_id.proto == IPPROTO_TCP) {
 671                 struct ip6_hdr *ip6;
 672                 struct tcphdr *tcp;
 673                 tcp_seq ack, seq;
 674                 int flags;
 675                 struct {
 676                         struct ip6_hdr ip6;
 677                         struct tcphdr th;
 678                 } ti;
 679
 680                 if (args->m->m_len < (hlen+sizeof(struct tcphdr))) {
 681                         args->m = m_pullup(args->m, hlen+sizeof(struct tcphdr));
 682                         if (args->m == NULL)
 683                                 return;
 684                 }
 685
 686                 ip6 = mtod(args->m, struct ip6_hdr *);
 687                 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen);
 688
 689                 if ((tcp->th_flags & TH_RST) != 0) {
 690                         m_freem(args->m);
 691                         return;
 692                 }
 693
 694                 ti.ip6 = *ip6;
 695                 ti.th = *tcp;
 696                 ti.th.th_seq = ntohl(ti.th.th_seq);
 697                 ti.th.th_ack = ntohl(ti.th.th_ack);
 698                 ti.ip6.ip6_nxt = IPPROTO_TCP;
 699
 700                 if (ti.th.th_flags & TH_ACK) {
 701                         ack = 0;
 702                         seq = ti.th.th_ack;
 703                         flags = TH_RST;
 704                 } else {
 705                         ack = ti.th.th_seq;
 706                         if (((args->m)->m_flags & M_PKTHDR) != 0) {
 707                                 ack += (args->m)->m_pkthdr.len - hlen
 708                                         - (ti.th.th_off << 2);
 709                         } else if (ip6->ip6_plen) {
 710                                 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6)
 711                                         - hlen - (ti.th.th_off << 2);
 712                         } else {
 713                                 m_freem(args->m);
 714                                 return;
 715                         }
 716                         if (tcp->th_flags & TH_SYN)
 717                                 ack++;
 718                         seq = 0;
 719                         flags = TH_RST|TH_ACK;
 720                 }
 721                 bcopy(&ti, ip6, sizeof(ti));
 722                 tcp_respond(NULL, ip6, (struct tcphdr *)(ip6 + 1),
 723                         args->m, ack, seq, flags);
 724
 725         } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
 726                 icmp6_error(args->m, ICMP6_DST_UNREACH, code, 0);
 727
 728         } else
 729                 m_freem(args->m);
 730
 731         args->m = NULL;
 732 }
 733
 734 #endif /* INET6 */
 735
 736 static u_int64_t norule_counter;        /* counter for ipfw_log(NULL...) */
 737
 738 #define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0
 739 #define SNP(buf) buf, sizeof(buf)
 740
 741 /*
 742  * We enter here when we have a rule with O_LOG.
 743  * XXX this function alone takes about 2Kbytes of code!
 744  */
 745 static void
 746 ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args,
 747         struct mbuf *m, struct ifnet *oif, u_short offset)
 748 {
 749         struct ether_header *eh = args->eh;
 750         char *action;
 751         int limit_reached = 0;
 752         char action2[40], proto[128], fragment[32];
 753
 754         fragment[0] = '\0';
 755         proto[0] = '\0';
 756
 757         if (f == NULL) {        /* bogus pkt */
 758                 if (verbose_limit != 0 && norule_counter >= verbose_limit)
 759                         return;
 760                 norule_counter++;
 761                 if (norule_counter == verbose_limit)
 762                         limit_reached = verbose_limit;
 763                 action = "Refuse";
 764         } else {        /* O_LOG is the first action, find the real one */
 765                 ipfw_insn *cmd = ACTION_PTR(f);
 766                 ipfw_insn_log *l = (ipfw_insn_log *)cmd;
 767
 768                 if (l->max_log != 0 && l->log_left == 0)
 769                         return;
 770                 l->log_left--;
 771                 if (l->log_left == 0)
 772                         limit_reached = l->max_log;
 773                 cmd += F_LEN(cmd);      /* point to first action */
 774                 if (cmd->opcode == O_ALTQ) {
 775                         ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
 776
 777                         snprintf(SNPARGS(action2, 0), "Altq %d",
 778                                 altq->qid);
 779                         cmd += F_LEN(cmd);
 780                 }
 781                 if (cmd->opcode == O_PROB)
 782                         cmd += F_LEN(cmd);
 783
 784                 if (cmd->opcode == O_TAG)
 785                         cmd += F_LEN(cmd);
 786
 787                 action = action2;
 788                 switch (cmd->opcode) {
 789                 case O_DENY:
 790                         action = "Deny";
 791                         break;
 792
 793                 case O_REJECT:
 794                         if (cmd->arg1==ICMP_REJECT_RST)
 795                                 action = "Reset";
 796                         else if (cmd->arg1==ICMP_UNREACH_HOST)
 797                                 action = "Reject";
 798                         else
 799                                 snprintf(SNPARGS(action2, 0), "Unreach %d",
 800                                         cmd->arg1);
 801                         break;
 802
 803                 case O_UNREACH6:
 804                         if (cmd->arg1==ICMP6_UNREACH_RST)
 805                                 action = "Reset";
 806                         else
 807                                 snprintf(SNPARGS(action2, 0), "Unreach %d",
 808                                         cmd->arg1);
 809                         break;
 810
 811                 case O_ACCEPT:
 812                         action = "Accept";
 813                         break;
 814                 case O_COUNT:
 815                         action = "Count";
 816                         break;
 817                 case O_DIVERT:
 818                         snprintf(SNPARGS(action2, 0), "Divert %d",
 819                                 cmd->arg1);
 820                         break;
 821                 case O_TEE:
 822                         snprintf(SNPARGS(action2, 0), "Tee %d",
 823                                 cmd->arg1);
 824                         break;
 825                 case O_SKIPTO:
 826                         snprintf(SNPARGS(action2, 0), "SkipTo %d",
 827                                 cmd->arg1);
 828                         break;
 829                 case O_PIPE:
 830                         snprintf(SNPARGS(action2, 0), "Pipe %d",
 831                                 cmd->arg1);
 832                         break;
 833                 case O_QUEUE:
 834                         snprintf(SNPARGS(action2, 0), "Queue %d",
 835                                 cmd->arg1);
 836                         break;
 837                 case O_FORWARD_IP: {
 838                         ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd;
 839                         int len;
 840
 841                         len = snprintf(SNPARGS(action2, 0), "Forward to %s",
 842                                 inet_ntoa(sa->sa.sin_addr));
 843                         if (sa->sa.sin_port)
 844                                 snprintf(SNPARGS(action2, len), ":%d",
 845                                     sa->sa.sin_port);
 846                         }
 847                         break;
 848                 case O_NETGRAPH:
 849                         snprintf(SNPARGS(action2, 0), "Netgraph %d",
 850                                 cmd->arg1);
 851                         break;
 852                 case O_NGTEE:
 853                         snprintf(SNPARGS(action2, 0), "Ngtee %d",
 854                                 cmd->arg1);
 855                         break;
 856                 default:
 857                         action = "UNKNOWN";
 858                         break;
 859                 }
 860         }
 861
 862         if (hlen == 0) {        /* non-ip */
 863                 snprintf(SNPARGS(proto, 0), "MAC");
 864
 865         } else {
 866                 int len;
 867                 char src[48], dst[48];
 868                 struct icmphdr *icmp;
 869                 struct tcphdr *tcp;
 870                 struct udphdr *udp;
 871                 /* Initialize to make compiler happy. */
 872                 struct ip *ip = NULL;
 873 #ifdef INET6
 874                 struct ip6_hdr *ip6 = NULL;
 875                 struct icmp6_hdr *icmp6;
 876 #endif
 877                 src[0] = '\0';
 878                 dst[0] = '\0';
 879 #ifdef INET6
 880                 if (args->f_id.addr_type == 6) {
 881                         snprintf(src, sizeof(src), "[%s]",
 882                             ip6_sprintf(&args->f_id.src_ip6));
 883                         snprintf(dst, sizeof(dst), "[%s]",
 884                             ip6_sprintf(&args->f_id.dst_ip6));
 885
 886                         ip6 = (struct ip6_hdr *)mtod(m, struct ip6_hdr *);
 887                         tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen);
 888                         udp = (struct udphdr *)(mtod(args->m, char *) + hlen);
 889                 } else
 890 #endif
 891                 {
 892                         ip = mtod(m, struct ip *);
 893                         tcp = L3HDR(struct tcphdr, ip);
 894                         udp = L3HDR(struct udphdr, ip);
 895
 896                         inet_ntoa_r(ip->ip_src, src);
 897                         inet_ntoa_r(ip->ip_dst, dst);
 898                 }
 899
 900                 switch (args->f_id.proto) {
 901                 case IPPROTO_TCP:
 902                         len = snprintf(SNPARGS(proto, 0), "TCP %s", src);
 903                         if (offset == 0)
 904                                 snprintf(SNPARGS(proto, len), ":%d %s:%d",
 905                                     ntohs(tcp->th_sport),
 906                                     dst,
 907                                     ntohs(tcp->th_dport));
 908                         else
 909                                 snprintf(SNPARGS(proto, len), " %s", dst);
 910                         break;
 911
 912                 case IPPROTO_UDP:
 913                         len = snprintf(SNPARGS(proto, 0), "UDP %s", src);
 914                         if (offset == 0)
 915                                 snprintf(SNPARGS(proto, len), ":%d %s:%d",
 916                                     ntohs(udp->uh_sport),
 917                                     dst,
 918                                     ntohs(udp->uh_dport));
 919                         else
 920                                 snprintf(SNPARGS(proto, len), " %s", dst);
 921                         break;
 922
 923                 case IPPROTO_ICMP:
 924                         icmp = L3HDR(struct icmphdr, ip);
 925                         if (offset == 0)
 926                                 len = snprintf(SNPARGS(proto, 0),
 927                                     "ICMP:%u.%u ",
 928                                     icmp->icmp_type, icmp->icmp_code);
 929                         else
 930                                 len = snprintf(SNPARGS(proto, 0), "ICMP ");
 931                         len += snprintf(SNPARGS(proto, len), "%s", src);
 932                         snprintf(SNPARGS(proto, len), " %s", dst);
 933                         break;
 934 #ifdef INET6
 935                 case IPPROTO_ICMPV6:
 936                         icmp6 = (struct icmp6_hdr *)(mtod(args->m, char *) + hlen);
 937                         if (offset == 0)
 938                                 len = snprintf(SNPARGS(proto, 0),
 939                                     "ICMPv6:%u.%u ",
 940                                     icmp6->icmp6_type, icmp6->icmp6_code);
 941                         else
 942                                 len = snprintf(SNPARGS(proto, 0), "ICMPv6 ");
 943                         len += snprintf(SNPARGS(proto, len), "%s", src);
 944                         snprintf(SNPARGS(proto, len), " %s", dst);
 945                         break;
 946 #endif
 947                 default:
 948                         len = snprintf(SNPARGS(proto, 0), "P:%d %s",
 949                             args->f_id.proto, src);
 950                         snprintf(SNPARGS(proto, len), " %s", dst);
 951                         break;
 952                 }
 953
 954 #ifdef INET6
 955                 if (args->f_id.addr_type == 6) {
 956                         if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG))
 957                                 snprintf(SNPARGS(fragment, 0),
 958                                     " (frag %08x:%d@%d%s)",
 959                                     args->f_id.frag_id6,
 960                                     ntohs(ip6->ip6_plen) - hlen,
 961                                     ntohs(offset & IP6F_OFF_MASK) << 3,
 962                                     (offset & IP6F_MORE_FRAG) ? "+" : "");
 963                 } else
 964 #endif
 965                 {
 966                         int ip_off, ip_len;
 967                         if (eh != NULL) { /* layer 2 packets are as on the wire */
 968                                 ip_off = ntohs(ip->ip_off);
 969                                 ip_len = ntohs(ip->ip_len);
 970                         } else {
 971                                 ip_off = ip->ip_off;
 972                                 ip_len = ip->ip_len;
 973                         }
 974                         if (ip_off & (IP_MF | IP_OFFMASK))
 975                                 snprintf(SNPARGS(fragment, 0),
 976                                     " (frag %d:%d@%d%s)",
 977                                     ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2),
 978                                     offset << 3,
 979                                     (ip_off & IP_MF) ? "+" : "");
 980                 }
 981         }
 982         if (oif || m->m_pkthdr.rcvif)
 983                 log(LOG_SECURITY | LOG_INFO,
 984                     "ipfw: %d %s %s %s via %s%s\n",
 985                     f ? f->rulenum : -1,
 986                     action, proto, oif ? "out" : "in",
 987                     oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname,
 988                     fragment);
 989         else
 990                 log(LOG_SECURITY | LOG_INFO,
 991                     "ipfw: %d %s %s [no if info]%s\n",
 992                     f ? f->rulenum : -1,
 993                     action, proto, fragment);
 994         if (limit_reached)
 995                 log(LOG_SECURITY | LOG_NOTICE,
 996                     "ipfw: limit %d reached on entry %d\n",
 997                     limit_reached, f ? f->rulenum : -1);
 998 }
 999
1000 /*
1001  * IMPORTANT: the hash function for dynamic rules must be commutative
1002  * in source and destination (ip,port), because rules are bidirectional
1003  * and we want to find both in the same bucket.
1004  */
1005 static __inline int
1006 hash_packet(struct ipfw_flow_id *id)
1007 {
1008         u_int32_t i;
1009
1010 #ifdef INET6
1011         if (IS_IP6_FLOW_ID(id))
1012                 i = hash_packet6(id);
1013         else
1014 #endif /* INET6 */
1015         i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port);
1016         i &= (curr_dyn_buckets - 1);
1017         return i;
1018 }
1019
1020 /**
1021  * unlink a dynamic rule from a chain. prev is a pointer to
1022  * the previous one, q is a pointer to the rule to delete,
1023  * head is a pointer to the head of the queue.
1024  * Modifies q and potentially also head.
1025  */
1026 #define UNLINK_DYN_RULE(prev, head, q) {                                \
1027         ipfw_dyn_rule *old_q = q;                                       \
1028                                                                         \
1029         /* remove a refcount to the parent */                           \
1030         if (q->dyn_type == O_LIMIT)                                     \
1031                 q->parent->count--;                                     \
1032         DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\
1033                 (q->id.src_ip), (q->id.src_port),                       \
1034                 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); )      \
1035         if (prev != NULL)                                               \
1036                 prev->next = q = q->next;                               \
1037         else                                                            \
1038                 head = q = q->next;                                     \
1039         dyn_count--;                                                    \
1040         uma_zfree(ipfw_dyn_rule_zone, old_q); }
1041
1042 #define TIME_LEQ(a,b)       ((int)((a)-(b)) <= 0)
1043
1044 /**
1045  * Remove dynamic rules pointing to "rule", or all of them if rule == NULL.
1046  *
1047  * If keep_me == NULL, rules are deleted even if not expired,
1048  * otherwise only expired rules are removed.
1049  *
1050  * The value of the second parameter is also used to point to identify
1051  * a rule we absolutely do not want to remove (e.g. because we are
1052  * holding a reference to it -- this is the case with O_LIMIT_PARENT
1053  * rules). The pointer is only used for comparison, so any non-null
1054  * value will do.
1055  */
1056 static void
1057 remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me)
1058 {
1059         static u_int32_t last_remove = 0;
1060
1061 #define FORCE (keep_me == NULL)
1062
1063         ipfw_dyn_rule *prev, *q;
1064         int i, pass = 0, max_pass = 0;
1065
1066         IPFW_DYN_LOCK_ASSERT();
1067
1068         if (ipfw_dyn_v == NULL || dyn_count == 0)
1069                 return;
1070         /* do not expire more than once per second, it is useless */
1071         if (!FORCE && last_remove == time_uptime)
1072                 return;
1073         last_remove = time_uptime;
1074
1075         /*
1076          * because O_LIMIT refer to parent rules, during the first pass only
1077          * remove child and mark any pending LIMIT_PARENT, and remove
1078          * them in a second pass.
1079          */
1080 next_pass:
1081         for (i = 0 ; i < curr_dyn_buckets ; i++) {
1082                 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) {
1083                         /*
1084                          * Logic can become complex here, so we split tests.
1085                          */
1086                         if (q == keep_me)
1087                                 goto next;
1088                         if (rule != NULL && rule != q->rule)
1089                                 goto next; /* not the one we are looking for */
1090                         if (q->dyn_type == O_LIMIT_PARENT) {
1091                                 /*
1092                                  * handle parent in the second pass,
1093                                  * record we need one.
1094                                  */
1095                                 max_pass = 1;
1096                                 if (pass == 0)
1097                                         goto next;
1098                                 if (FORCE && q->count != 0 ) {
1099                                         /* XXX should not happen! */
1100                                         printf("ipfw: OUCH! cannot remove rule,"
1101                                              " count %d\n", q->count);
1102                                 }
1103                         } else {
1104                                 if (!FORCE &&
1105                                     !TIME_LEQ( q->expire, time_uptime ))
1106                                         goto next;
1107                         }
1108              if (q->dyn_type != O_LIMIT_PARENT || !q->count) {
1109                      UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
1110                      continue;
1111              }
1112 next:
1113                         prev=q;
1114                         q=q->next;
1115                 }
1116         }
1117         if (pass++ < max_pass)
1118                 goto next_pass;
1119 }
1120
1121
1122 /**
1123  * lookup a dynamic rule.
1124  */
1125 static ipfw_dyn_rule *
1126 lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction,
1127         struct tcphdr *tcp)
1128 {
1129         /*
1130          * stateful ipfw extensions.
1131          * Lookup into dynamic session queue
1132          */
1133 #define MATCH_REVERSE   0
1134 #define MATCH_FORWARD   1
1135 #define MATCH_NONE      2
1136 #define MATCH_UNKNOWN   3
1137         int i, dir = MATCH_NONE;
1138         ipfw_dyn_rule *prev, *q=NULL;
1139
1140         IPFW_DYN_LOCK_ASSERT();
1141
1142         if (ipfw_dyn_v == NULL)
1143                 goto done;      /* not found */
1144         i = hash_packet( pkt );
1145         for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) {
1146                 if (q->dyn_type == O_LIMIT_PARENT && q->count)
1147                         goto next;
1148                 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */
1149                         UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
1150                         continue;
1151                 }
1152                 if (pkt->proto == q->id.proto &&
1153                     q->dyn_type != O_LIMIT_PARENT) {
1154                         if (IS_IP6_FLOW_ID(pkt)) {
1155                             if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6),
1156                                 &(q->id.src_ip6)) &&
1157                             IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6),
1158                                 &(q->id.dst_ip6)) &&
1159                             pkt->src_port == q->id.src_port &&
1160                             pkt->dst_port == q->id.dst_port ) {
1161                                 dir = MATCH_FORWARD;
1162                                 break;
1163                             }
1164                             if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6),
1165                                     &(q->id.dst_ip6)) &&
1166                                 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6),
1167                                     &(q->id.src_ip6)) &&
1168                                 pkt->src_port == q->id.dst_port &&
1169                                 pkt->dst_port == q->id.src_port ) {
1170                                     dir = MATCH_REVERSE;
1171                                     break;
1172                             }
1173                         } else {
1174                             if (pkt->src_ip == q->id.src_ip &&
1175                                 pkt->dst_ip == q->id.dst_ip &&
1176                                 pkt->src_port == q->id.src_port &&
1177                                 pkt->dst_port == q->id.dst_port ) {
1178                                     dir = MATCH_FORWARD;
1179                                     break;
1180                             }
1181                             if (pkt->src_ip == q->id.dst_ip &&
1182                                 pkt->dst_ip == q->id.src_ip &&
1183                                 pkt->src_port == q->id.dst_port &&
1184                                 pkt->dst_port == q->id.src_port ) {
1185                                     dir = MATCH_REVERSE;
1186                                     break;
1187                             }
1188                         }
1189                 }
1190 next:
1191                 prev = q;
1192                 q = q->next;
1193         }
1194         if (q == NULL)
1195                 goto done; /* q = NULL, not found */
1196
1197         if ( prev != NULL) { /* found and not in front */
1198                 prev->next = q->next;
1199                 q->next = ipfw_dyn_v[i];
1200                 ipfw_dyn_v[i] = q;
1201         }
1202         if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */
1203                 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST);
1204
1205 #define BOTH_SYN        (TH_SYN | (TH_SYN << 8))
1206 #define BOTH_FIN        (TH_FIN | (TH_FIN << 8))
1207                 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8);
1208                 switch (q->state) {
1209                 case TH_SYN:                            /* opening */
1210                         q->expire = time_uptime + dyn_syn_lifetime;
1211                         break;
1212
1213                 case BOTH_SYN:                  /* move to established */
1214                 case BOTH_SYN | TH_FIN :        /* one side tries to close */
1215                 case BOTH_SYN | (TH_FIN << 8) :
1216                         if (tcp) {
1217 #define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0)
1218                             u_int32_t ack = ntohl(tcp->th_ack);
1219                             if (dir == MATCH_FORWARD) {
1220                                 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd))
1221                                     q->ack_fwd = ack;
1222                                 else { /* ignore out-of-sequence */
1223                                     break;
1224                                 }
1225                             } else {
1226                                 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev))
1227                                     q->ack_rev = ack;
1228                                 else { /* ignore out-of-sequence */
1229                                     break;
1230                                 }
1231                             }
1232                         }
1233                         q->expire = time_uptime + dyn_ack_lifetime;
1234                         break;
1235
1236                 case BOTH_SYN | BOTH_FIN:       /* both sides closed */
1237                         if (dyn_fin_lifetime >= dyn_keepalive_period)
1238                                 dyn_fin_lifetime = dyn_keepalive_period - 1;
1239                         q->expire = time_uptime + dyn_fin_lifetime;
1240                         break;
1241
1242                 default:
1243 #if 0
1244                         /*
1245                          * reset or some invalid combination, but can also
1246                          * occur if we use keep-state the wrong way.
1247                          */
1248                         if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0)
1249                                 printf("invalid state: 0x%x\n", q->state);
1250 #endif
1251                         if (dyn_rst_lifetime >= dyn_keepalive_period)
1252                                 dyn_rst_lifetime = dyn_keepalive_period - 1;
1253                         q->expire = time_uptime + dyn_rst_lifetime;
1254                         break;
1255                 }
1256         } else if (pkt->proto == IPPROTO_UDP) {
1257                 q->expire = time_uptime + dyn_udp_lifetime;
1258         } else {
1259                 /* other protocols */
1260                 q->expire = time_uptime + dyn_short_lifetime;
1261         }
1262 done:
1263         if (match_direction)
1264                 *match_direction = dir;
1265         return q;
1266 }
1267
1268 static ipfw_dyn_rule *
1269 lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction,
1270         struct tcphdr *tcp)
1271 {
1272         ipfw_dyn_rule *q;
1273
1274         IPFW_DYN_LOCK();
1275         q = lookup_dyn_rule_locked(pkt, match_direction, tcp);
1276         if (q == NULL)
1277                 IPFW_DYN_UNLOCK();
1278         /* NB: return table locked when q is not NULL */
1279         return q;
1280 }
1281
1282 static void
1283 realloc_dynamic_table(void)
1284 {
1285         IPFW_DYN_LOCK_ASSERT();
1286
1287         /*
1288          * Try reallocation, make sure we have a power of 2 and do
1289          * not allow more than 64k entries. In case of overflow,
1290          * default to 1024.
1291          */
1292
1293         if (dyn_buckets > 65536)
1294                 dyn_buckets = 1024;
1295         if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */
1296                 dyn_buckets = curr_dyn_buckets; /* reset */
1297                 return;
1298         }
1299         curr_dyn_buckets = dyn_buckets;
1300         if (ipfw_dyn_v != NULL)
1301                 free(ipfw_dyn_v, M_IPFW);
1302         for (;;) {
1303                 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *),
1304                        M_IPFW, M_NOWAIT | M_ZERO);
1305                 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2)
1306                         break;
1307                 curr_dyn_buckets /= 2;
1308         }
1309 }
1310
1311 /**
1312  * Install state of type 'type' for a dynamic session.
1313  * The hash table contains two type of rules:
1314  * - regular rules (O_KEEP_STATE)
1315  * - rules for sessions with limited number of sess per user
1316  *   (O_LIMIT). When they are created, the parent is
1317  *   increased by 1, and decreased on delete. In this case,
1318  *   the third parameter is the parent rule and not the chain.
1319  * - "parent" rules for the above (O_LIMIT_PARENT).
1320  */
1321 static ipfw_dyn_rule *
1322 add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule)
1323 {
1324         ipfw_dyn_rule *r;
1325         int i;
1326
1327         IPFW_DYN_LOCK_ASSERT();
1328
1329         if (ipfw_dyn_v == NULL ||
1330             (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) {
1331                 realloc_dynamic_table();
1332                 if (ipfw_dyn_v == NULL)
1333                         return NULL; /* failed ! */
1334         }
1335         i = hash_packet(id);
1336
1337         r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO);
1338         if (r == NULL) {
1339                 printf ("ipfw: sorry cannot allocate state\n");
1340                 return NULL;
1341         }
1342
1343         /* increase refcount on parent, and set pointer */
1344         if (dyn_type == O_LIMIT) {
1345                 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule;
1346                 if ( parent->dyn_type != O_LIMIT_PARENT)
1347                         panic("invalid parent");
1348                 parent->count++;
1349                 r->parent = parent;
1350                 rule = parent->rule;
1351         }
1352
1353         r->id = *id;
1354         r->expire = time_uptime + dyn_syn_lifetime;
1355         r->rule = rule;
1356         r->dyn_type = dyn_type;
1357         r->pcnt = r->bcnt = 0;
1358         r->count = 0;
1359
1360         r->bucket = i;
1361         r->next = ipfw_dyn_v[i];
1362         ipfw_dyn_v[i] = r;
1363         dyn_count++;
1364         DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n",
1365            dyn_type,
1366            (r->id.src_ip), (r->id.src_port),
1367            (r->id.dst_ip), (r->id.dst_port),
1368            dyn_count ); )
1369         return r;
1370 }
1371
1372 /**
1373  * lookup dynamic parent rule using pkt and rule as search keys.
1374  * If the lookup fails, then install one.
1375  */
1376 static ipfw_dyn_rule *
1377 lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule)
1378 {
1379         ipfw_dyn_rule *q;
1380         int i;
1381
1382         IPFW_DYN_LOCK_ASSERT();
1383
1384         if (ipfw_dyn_v) {
1385                 int is_v6 = IS_IP6_FLOW_ID(pkt);
1386                 i = hash_packet( pkt );
1387                 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next)
1388                         if (q->dyn_type == O_LIMIT_PARENT &&
1389                             rule== q->rule &&
1390                             pkt->proto == q->id.proto &&
1391                             pkt->src_port == q->id.src_port &&
1392                             pkt->dst_port == q->id.dst_port &&
1393                             (
1394                                 (is_v6 &&
1395                                  IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6),
1396                                         &(q->id.src_ip6)) &&
1397                                  IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6),
1398                                         &(q->id.dst_ip6))) ||
1399                                 (!is_v6 &&
1400                                  pkt->src_ip == q->id.src_ip &&
1401                                  pkt->dst_ip == q->id.dst_ip)
1402                             )
1403                         ) {
1404                                 q->expire = time_uptime + dyn_short_lifetime;
1405                                 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);)
1406                                 return q;
1407                         }
1408         }
1409         return add_dyn_rule(pkt, O_LIMIT_PARENT, rule);
1410 }
1411
1412 /**
1413  * Install dynamic state for rule type cmd->o.opcode
1414  *
1415  * Returns 1 (failure) if state is not installed because of errors or because
1416  * session limitations are enforced.
1417  */
1418 static int
1419 install_state(struct ip_fw *rule, ipfw_insn_limit *cmd,
1420     struct ip_fw_args *args, uint32_t tablearg)
1421 {
1422         static int last_log;
1423
1424         ipfw_dyn_rule *q;
1425
1426         DEB(
1427         printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n",
1428             __func__, cmd->o.opcode,
1429             (args->f_id.src_ip), (args->f_id.src_port),
1430             (args->f_id.dst_ip), (args->f_id.dst_port));
1431         )
1432
1433         IPFW_DYN_LOCK();
1434
1435         q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL);
1436
1437         if (q != NULL) {        /* should never occur */
1438                 if (last_log != time_uptime) {
1439                         last_log = time_uptime;
1440                         printf("ipfw: %s: entry already present, done\n",
1441                             __func__);
1442                 }
1443                 IPFW_DYN_UNLOCK();
1444                 return (0);
1445         }
1446
1447         if (dyn_count >= dyn_max)
1448                 /* Run out of slots, try to remove any expired rule. */
1449                 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1);
1450
1451         if (dyn_count >= dyn_max) {
1452                 if (last_log != time_uptime) {
1453                         last_log = time_uptime;
1454                         printf("ipfw: %s: Too many dynamic rules\n", __func__);
1455                 }
1456                 IPFW_DYN_UNLOCK();
1457                 return (1);     /* cannot install, notify caller */
1458         }
1459
1460         switch (cmd->o.opcode) {
1461         case O_KEEP_STATE:      /* bidir rule */
1462                 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule);
1463                 break;
1464
1465         case O_LIMIT: {         /* limit number of sessions */
1466                 struct ipfw_flow_id id;
1467                 ipfw_dyn_rule *parent;
1468                 uint32_t conn_limit;
1469                 uint16_t limit_mask = cmd->limit_mask;
1470
1471                 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ?
1472                     tablearg : cmd->conn_limit;
1473
1474                 DEB(
1475                 if (cmd->conn_limit == IP_FW_TABLEARG)
1476                         printf("ipfw: %s: O_LIMIT rule, conn_limit: %u "
1477                             "(tablearg)\n", __func__, conn_limit);
1478                 else
1479                         printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n",
1480                             __func__, conn_limit);
1481                 )
1482
1483                 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0;
1484                 id.proto = args->f_id.proto;
1485                 id.addr_type = args->f_id.addr_type;
1486
1487                 if (IS_IP6_FLOW_ID (&(args->f_id))) {
1488                         if (limit_mask & DYN_SRC_ADDR)
1489                                 id.src_ip6 = args->f_id.src_ip6;
1490                         if (limit_mask & DYN_DST_ADDR)
1491                                 id.dst_ip6 = args->f_id.dst_ip6;
1492                 } else {
1493                         if (limit_mask & DYN_SRC_ADDR)
1494                                 id.src_ip = args->f_id.src_ip;
1495                         if (limit_mask & DYN_DST_ADDR)
1496                                 id.dst_ip = args->f_id.dst_ip;
1497                 }
1498                 if (limit_mask & DYN_SRC_PORT)
1499                         id.src_port = args->f_id.src_port;
1500                 if (limit_mask & DYN_DST_PORT)
1501                         id.dst_port = args->f_id.dst_port;
1502                 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) {
1503                         printf("ipfw: %s: add parent failed\n", __func__);
1504                         IPFW_DYN_UNLOCK();
1505                         return (1);
1506                 }
1507
1508                 if (parent->count >= conn_limit) {
1509                         /* See if we can remove some expired rule. */
1510                         remove_dyn_rule(rule, parent);
1511                         if (parent->count >= conn_limit) {
1512                                 if (fw_verbose && last_log != time_uptime) {
1513                                         last_log = time_uptime;
1514                                         log(LOG_SECURITY | LOG_DEBUG,
1515                                             "drop session, too many entries\n");
1516                                 }
1517                                 IPFW_DYN_UNLOCK();
1518                                 return (1);
1519                         }
1520                 }
1521                 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent);
1522                 break;
1523         }
1524         default:
1525                 printf("ipfw: %s: unknown dynamic rule type %u\n",
1526                     __func__, cmd->o.opcode);
1527                 IPFW_DYN_UNLOCK();
1528                 return (1);
1529         }
1530
1531         /* XXX just set lifetime */
1532         lookup_dyn_rule_locked(&args->f_id, NULL, NULL);
1533
1534         IPFW_DYN_UNLOCK();
1535         return (0);
1536 }
1537
1538 /*
1539  * Generate a TCP packet, containing either a RST or a keepalive.
1540  * When flags & TH_RST, we are sending a RST packet, because of a
1541  * "reset" action matched the packet.
1542  * Otherwise we are sending a keepalive, and flags & TH_
1543  */
1544 static struct mbuf *
1545 send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags)
1546 {
1547         struct mbuf *m;
1548         struct ip *ip;
1549         struct tcphdr *tcp;
1550
1551         MGETHDR(m, M_DONTWAIT, MT_DATA);
1552         if (m == 0)
1553                 return (NULL);
1554         m->m_pkthdr.rcvif = (struct ifnet *)0;
1555         m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr);
1556         m->m_data += max_linkhdr;
1557
1558         ip = mtod(m, struct ip *);
1559         bzero(ip, m->m_len);
1560         tcp = (struct tcphdr *)(ip + 1); /* no IP options */
1561         ip->ip_p = IPPROTO_TCP;
1562         tcp->th_off = 5;
1563         /*
1564          * Assume we are sending a RST (or a keepalive in the reverse
1565          * direction), swap src and destination addresses and ports.
1566          */
1567         ip->ip_src.s_addr = htonl(id->dst_ip);
1568         ip->ip_dst.s_addr = htonl(id->src_ip);
1569         tcp->th_sport = htons(id->dst_port);
1570         tcp->th_dport = htons(id->src_port);
1571         if (flags & TH_RST) {   /* we are sending a RST */
1572                 if (flags & TH_ACK) {
1573                         tcp->th_seq = htonl(ack);
1574                         tcp->th_ack = htonl(0);
1575                         tcp->th_flags = TH_RST;
1576                 } else {
1577                         if (flags & TH_SYN)
1578                                 seq++;
1579                         tcp->th_seq = htonl(0);
1580                         tcp->th_ack = htonl(seq);
1581                         tcp->th_flags = TH_RST | TH_ACK;
1582                 }
1583         } else {
1584                 /*
1585                  * We are sending a keepalive. flags & TH_SYN determines
1586                  * the direction, forward if set, reverse if clear.
1587                  * NOTE: seq and ack are always assumed to be correct
1588                  * as set by the caller. This may be confusing...
1589                  */
1590                 if (flags & TH_SYN) {
1591                         /*
1592                          * we have to rewrite the correct addresses!
1593                          */
1594                         ip->ip_dst.s_addr = htonl(id->dst_ip);
1595                         ip->ip_src.s_addr = htonl(id->src_ip);
1596                         tcp->th_dport = htons(id->dst_port);
1597                         tcp->th_sport = htons(id->src_port);
1598                 }
1599                 tcp->th_seq = htonl(seq);
1600                 tcp->th_ack = htonl(ack);
1601                 tcp->th_flags = TH_ACK;
1602         }
1603         /*
1604          * set ip_len to the payload size so we can compute
1605          * the tcp checksum on the pseudoheader
1606          * XXX check this, could save a couple of words ?
1607          */
1608         ip->ip_len = htons(sizeof(struct tcphdr));
1609         tcp->th_sum = in_cksum(m, m->m_pkthdr.len);
1610         /*
1611          * now fill fields left out earlier
1612          */
1613         ip->ip_ttl = ip_defttl;
1614         ip->ip_len = m->m_pkthdr.len;
1615         m->m_flags |= M_SKIP_FIREWALL;
1616         return (m);
1617 }
1618
1619 /*
1620  * sends a reject message, consuming the mbuf passed as an argument.
1621  */
1622 static void
1623 send_reject(struct ip_fw_args *args, int code, u_short offset, int ip_len)
1624 {
1625
1626         if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
1627                 /* We need the IP header in host order for icmp_error(). */
1628                 if (args->eh != NULL) {
1629                         struct ip *ip = mtod(args->m, struct ip *);
1630                         ip->ip_len = ntohs(ip->ip_len);
1631                         ip->ip_off = ntohs(ip->ip_off);
1632                 }
1633                 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
1634         } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) {
1635                 struct tcphdr *const tcp =
1636                     L3HDR(struct tcphdr, mtod(args->m, struct ip *));
1637                 if ( (tcp->th_flags & TH_RST) == 0) {
1638                         struct mbuf *m;
1639                         m = send_pkt(&(args->f_id), ntohl(tcp->th_seq),
1640                                 ntohl(tcp->th_ack),
1641                                 tcp->th_flags | TH_RST);
1642                         if (m != NULL)
1643                                 ip_output(m, NULL, NULL, 0, NULL, NULL);
1644                 }
1645                 m_freem(args->m);
1646         } else
1647                 m_freem(args->m);
1648         args->m = NULL;
1649 }
1650
1651 /**
1652  *
1653  * Given an ip_fw *, lookup_next_rule will return a pointer
1654  * to the next rule, which can be either the jump
1655  * target (for skipto instructions) or the next one in the list (in
1656  * all other cases including a missing jump target).
1657  * The result is also written in the "next_rule" field of the rule.
1658  * Backward jumps are not allowed, so start looking from the next
1659  * rule...
1660  *
1661  * This never returns NULL -- in case we do not have an exact match,
1662  * the next rule is returned. When the ruleset is changed,
1663  * pointers are flushed so we are always correct.
1664  */
1665
1666 static struct ip_fw *
1667 lookup_next_rule(struct ip_fw *me)
1668 {
1669         struct ip_fw *rule = NULL;
1670         ipfw_insn *cmd;
1671
1672         /* look for action, in case it is a skipto */
1673         cmd = ACTION_PTR(me);
1674         if (cmd->opcode == O_LOG)
1675                 cmd += F_LEN(cmd);
1676         if (cmd->opcode == O_ALTQ)
1677                 cmd += F_LEN(cmd);
1678         if (cmd->opcode == O_TAG)
1679                 cmd += F_LEN(cmd);
1680         if ( cmd->opcode == O_SKIPTO )
1681                 for (rule = me->next; rule ; rule = rule->next)
1682                         if (rule->rulenum >= cmd->arg1)
1683                                 break;
1684         if (rule == NULL)                       /* failure or not a skipto */
1685                 rule = me->next;
1686         me->next_rule = rule;
1687         return rule;
1688 }
1689
1690 static int
1691 add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr,
1692         uint8_t mlen, uint32_t value)
1693 {
1694         struct radix_node_head *rnh;
1695         struct table_entry *ent;
1696
1697         if (tbl >= IPFW_TABLES_MAX)
1698                 return (EINVAL);
1699         rnh = ch->tables[tbl];
1700         ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO);
1701         if (ent == NULL)
1702                 return (ENOMEM);
1703         ent->value = value;
1704         ent->addr.sin_len = ent->mask.sin_len = 8;
1705         ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0);
1706         ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr;
1707         IPFW_WLOCK(&layer3_chain);
1708         if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) ==
1709             NULL) {
1710                 IPFW_WUNLOCK(&layer3_chain);
1711                 free(ent, M_IPFW_TBL);
1712                 return (EEXIST);
1713         }
1714         IPFW_WUNLOCK(&layer3_chain);
1715         return (0);
1716 }
1717
1718 static int
1719 del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr,
1720         uint8_t mlen)
1721 {
1722         struct radix_node_head *rnh;
1723         struct table_entry *ent;
1724         struct sockaddr_in sa, mask;
1725
1726         if (tbl >= IPFW_TABLES_MAX)
1727                 return (EINVAL);
1728         rnh = ch->tables[tbl];
1729         sa.sin_len = mask.sin_len = 8;
1730         mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0);
1731         sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr;
1732         IPFW_WLOCK(ch);
1733         ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh);
1734         if (ent == NULL) {
1735                 IPFW_WUNLOCK(ch);
1736                 return (ESRCH);
1737         }
1738         IPFW_WUNLOCK(ch);
1739         free(ent, M_IPFW_TBL);
1740         return (0);
1741 }
1742
1743 static int
1744 flush_table_entry(struct radix_node *rn, void *arg)
1745 {
1746         struct radix_node_head * const rnh = arg;
1747         struct table_entry *ent;
1748
1749         ent = (struct table_entry *)
1750             rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh);
1751         if (ent != NULL)
1752                 free(ent, M_IPFW_TBL);
1753         return (0);
1754 }
1755
1756 static int
1757 flush_table(struct ip_fw_chain *ch, uint16_t tbl)
1758 {
1759         struct radix_node_head *rnh;
1760
1761         IPFW_WLOCK_ASSERT(ch);
1762
1763         if (tbl >= IPFW_TABLES_MAX)
1764                 return (EINVAL);
1765         rnh = ch->tables[tbl];
1766         KASSERT(rnh != NULL, ("NULL IPFW table"));
1767         rnh->rnh_walktree(rnh, flush_table_entry, rnh);
1768         return (0);
1769 }
1770
1771 static void
1772 flush_tables(struct ip_fw_chain *ch)
1773 {
1774         uint16_t tbl;
1775
1776         IPFW_WLOCK_ASSERT(ch);
1777
1778         for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++)
1779                 flush_table(ch, tbl);
1780 }
1781
1782 static int
1783 init_tables(struct ip_fw_chain *ch)
1784 {
1785         int i;
1786         uint16_t j;
1787
1788         for (i = 0; i < IPFW_TABLES_MAX; i++) {
1789                 if (!rn_inithead((void **)&ch->tables[i], 32)) {
1790                         for (j = 0; j < i; j++) {
1791                                 (void) flush_table(ch, j);
1792                         }
1793                         return (ENOMEM);
1794                 }
1795         }
1796         return (0);
1797 }
1798
1799 static int
1800 lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr,
1801         uint32_t *val)
1802 {
1803         struct radix_node_head *rnh;
1804         struct table_entry *ent;
1805         struct sockaddr_in sa;
1806
1807         if (tbl >= IPFW_TABLES_MAX)
1808                 return (0);
1809         rnh = ch->tables[tbl];
1810         sa.sin_len = 8;
1811         sa.sin_addr.s_addr = addr;
1812         ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh));
1813         if (ent != NULL) {
1814                 *val = ent->value;
1815                 return (1);
1816         }
1817         return (0);
1818 }
1819
1820 static int
1821 count_table_entry(struct radix_node *rn, void *arg)
1822 {
1823         u_int32_t * const cnt = arg;
1824
1825         (*cnt)++;
1826         return (0);
1827 }
1828
1829 static int
1830 count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt)
1831 {
1832         struct radix_node_head *rnh;
1833
1834         if (tbl >= IPFW_TABLES_MAX)
1835                 return (EINVAL);
1836         rnh = ch->tables[tbl];
1837         *cnt = 0;
1838         rnh->rnh_walktree(rnh, count_table_entry, cnt);
1839         return (0);
1840 }
1841
1842 static int
1843 dump_table_entry(struct radix_node *rn, void *arg)
1844 {
1845         struct table_entry * const n = (struct table_entry *)rn;
1846         ipfw_table * const tbl = arg;
1847         ipfw_table_entry *ent;
1848
1849         if (tbl->cnt == tbl->size)
1850                 return (1);
1851         ent = &tbl->ent[tbl->cnt];
1852         ent->tbl = tbl->tbl;
1853         if (in_nullhost(n->mask.sin_addr))
1854                 ent->masklen = 0;
1855         else
1856                 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr));
1857         ent->addr = n->addr.sin_addr.s_addr;
1858         ent->value = n->value;
1859         tbl->cnt++;
1860         return (0);
1861 }
1862
1863 static int
1864 dump_table(struct ip_fw_chain *ch, ipfw_table *tbl)
1865 {
1866         struct radix_node_head *rnh;
1867
1868         if (tbl->tbl >= IPFW_TABLES_MAX)
1869                 return (EINVAL);
1870         rnh = ch->tables[tbl->tbl];
1871         tbl->cnt = 0;
1872         rnh->rnh_walktree(rnh, dump_table_entry, tbl);
1873         return (0);
1874 }
1875
1876 static void
1877 fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp)
1878 {
1879         struct ucred *cr;
1880
1881         if (inp->inp_socket != NULL) {
1882                 cr = inp->inp_socket->so_cred;
1883                 ugp->fw_prid = jailed(cr) ?
1884                     cr->cr_prison->pr_id : -1;
1885                 ugp->fw_uid = cr->cr_uid;
1886                 ugp->fw_ngroups = cr->cr_ngroups;
1887                 bcopy(cr->cr_groups, ugp->fw_groups,
1888                     sizeof(ugp->fw_groups));
1889         }
1890 }
1891
1892 static int
1893 check_uidgid(ipfw_insn_u32 *insn,
1894         int proto, struct ifnet *oif,
1895         struct in_addr dst_ip, u_int16_t dst_port,
1896         struct in_addr src_ip, u_int16_t src_port,
1897         struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp)
1898 {
1899         struct inpcbinfo *pi;
1900         int wildcard;
1901         struct inpcb *pcb;
1902         int match;
1903         gid_t *gp;
1904
1905         /*
1906          * Check to see if the UDP or TCP stack supplied us with
1907          * the PCB. If so, rather then holding a lock and looking
1908          * up the PCB, we can use the one that was supplied.
1909          */
1910         if (inp && *lookup == 0) {
1911                 INP_LOCK_ASSERT(inp);
1912                 if (inp->inp_socket != NULL) {
1913                         fill_ugid_cache(inp, ugp);
1914                         *lookup = 1;
1915                 }
1916         }
1917         /*
1918          * If we have already been here and the packet has no
1919          * PCB entry associated with it, then we can safely
1920          * assume that this is a no match.
1921          */
1922         if (*lookup == -1)
1923                 return (0);
1924         if (proto == IPPROTO_TCP) {
1925                 wildcard = 0;
1926                 pi = &tcbinfo;
1927         } else if (proto == IPPROTO_UDP) {
1928                 wildcard = 1;
1929                 pi = &udbinfo;
1930         } else
1931                 return 0;
1932         match = 0;
1933         if (*lookup == 0) {
1934                 INP_INFO_RLOCK(pi);
1935                 pcb =  (oif) ?
1936                         in_pcblookup_hash(pi,
1937                                 dst_ip, htons(dst_port),
1938                                 src_ip, htons(src_port),
1939                                 wildcard, oif) :
1940                         in_pcblookup_hash(pi,
1941                                 src_ip, htons(src_port),
1942                                 dst_ip, htons(dst_port),
1943                                 wildcard, NULL);
1944                 if (pcb != NULL) {
1945                         INP_LOCK(pcb);
1946                         if (pcb->inp_socket != NULL) {
1947                                 fill_ugid_cache(pcb, ugp);
1948                                 *lookup = 1;
1949                         }
1950                         INP_UNLOCK(pcb);
1951                 }
1952                 INP_INFO_RUNLOCK(pi);
1953                 if (*lookup == 0) {
1954                         /*
1955                          * If the lookup did not yield any results, there
1956                          * is no sense in coming back and trying again. So
1957                          * we can set lookup to -1 and ensure that we wont
1958                          * bother the pcb system again.
1959                          */
1960                         *lookup = -1;
1961                         return (0);
1962                 }
1963         }
1964         if (insn->o.opcode == O_UID)
1965                 match = (ugp->fw_uid == (uid_t)insn->d[0]);
1966         else if (insn->o.opcode == O_GID) {
1967                 for (gp = ugp->fw_groups;
1968                         gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++)
1969                         if (*gp == (gid_t)insn->d[0]) {
1970                                 match = 1;
1971                                 break;
1972                         }
1973         } else if (insn->o.opcode == O_JAIL)
1974                 match = (ugp->fw_prid == (int)insn->d[0]);
1975         return match;
1976 }
1977
1978 /*
1979  * The main check routine for the firewall.
1980  *
1981  * All arguments are in args so we can modify them and return them
1982  * back to the caller.
1983  *
1984  * Parameters:
1985  *
1986  *      args->m (in/out) The packet; we set to NULL when/if we nuke it.
1987  *              Starts with the IP header.
1988  *      args->eh (in)   Mac header if present, or NULL for layer3 packet.
1989  *      args->oif       Outgoing interface, or NULL if packet is incoming.
1990  *              The incoming interface is in the mbuf. (in)
1991  *      args->divert_rule (in/out)
1992  *              Skip up to the first rule past this rule number;
1993  *              upon return, non-zero port number for divert or tee.
1994  *
1995  *      args->rule      Pointer to the last matching rule (in/out)
1996  *      args->next_hop  Socket we are forwarding to (out).
1997  *      args->f_id      Addresses grabbed from the packet (out)
1998  *      args->cookie    a cookie depending on rule action
1999  *
2000  * Return value:
2001  *
2002  *      IP_FW_PASS      the packet must be accepted
2003  *      IP_FW_DENY      the packet must be dropped
2004  *      IP_FW_DIVERT    divert packet, port in m_tag
2005  *      IP_FW_TEE       tee packet, port in m_tag
2006  *      IP_FW_DUMMYNET  to dummynet, pipe in args->cookie
2007  *      IP_FW_NETGRAPH  into netgraph, cookie args->cookie
2008  *
2009  */
2010
2011 int
2012 ipfw_chk(struct ip_fw_args *args)
2013 {
2014         /*
2015          * Local variables hold state during the processing of a packet.
2016          *
2017          * IMPORTANT NOTE: to speed up the processing of rules, there
2018          * are some assumption on the values of the variables, which
2019          * are documented here. Should you change them, please check
2020          * the implementation of the various instructions to make sure
2021          * that they still work.
2022          *
2023          * args->eh     The MAC header. It is non-null for a layer2
2024          *      packet, it is NULL for a layer-3 packet.
2025          *
2026          * m | args->m  Pointer to the mbuf, as received from the caller.
2027          *      It may change if ipfw_chk() does an m_pullup, or if it
2028          *      consumes the packet because it calls send_reject().
2029          *      XXX This has to change, so that ipfw_chk() never modifies
2030          *      or consumes the buffer.
2031          * ip   is simply an alias of the value of m, and it is kept
2032          *      in sync with it (the packet is  supposed to start with
2033          *      the ip header).
2034          */
2035         struct mbuf *m = args->m;
2036         struct ip *ip = mtod(m, struct ip *);
2037
2038         /*
2039          * For rules which contain uid/gid or jail constraints, cache
2040          * a copy of the users credentials after the pcb lookup has been
2041          * executed. This will speed up the processing of rules with
2042          * these types of constraints, as well as decrease contention
2043          * on pcb related locks.
2044          */
2045         struct ip_fw_ugid fw_ugid_cache;
2046         int ugid_lookup = 0;
2047
2048         /*
2049          * divinput_flags       If non-zero, set to the IP_FW_DIVERT_*_FLAG
2050          *      associated with a packet input on a divert socket.  This
2051          *      will allow to distinguish traffic and its direction when
2052          *      it originates from a divert socket.
2053          */
2054         u_int divinput_flags = 0;
2055
2056         /*
2057          * oif | args->oif      If NULL, ipfw_chk has been called on the
2058          *      inbound path (ether_input, ip_input).
2059          *      If non-NULL, ipfw_chk has been called on the outbound path
2060          *      (ether_output, ip_output).
2061          */
2062         struct ifnet *oif = args->oif;
2063
2064         struct ip_fw *f = NULL;         /* matching rule */
2065         int retval = 0;
2066
2067         /*
2068          * hlen The length of the IP header.
2069          */
2070         u_int hlen = 0;         /* hlen >0 means we have an IP pkt */
2071
2072         /*
2073          * offset       The offset of a fragment. offset != 0 means that
2074          *      we have a fragment at this offset of an IPv4 packet.
2075          *      offset == 0 means that (if this is an IPv4 packet)
2076          *      this is the first or only fragment.
2077          *      For IPv6 offset == 0 means there is no Fragment Header.
2078          *      If offset != 0 for IPv6 always use correct mask to
2079          *      get the correct offset because we add IP6F_MORE_FRAG
2080          *      to be able to dectect the first fragment which would
2081          *      otherwise have offset = 0.
2082          */
2083         u_short offset = 0;
2084
2085         /*
2086          * Local copies of addresses. They are only valid if we have
2087          * an IP packet.
2088          *
2089          * proto        The protocol. Set to 0 for non-ip packets,
2090          *      or to the protocol read from the packet otherwise.
2091          *      proto != 0 means that we have an IPv4 packet.
2092          *
2093          * src_port, dst_port   port numbers, in HOST format. Only
2094          *      valid for TCP and UDP packets.
2095          *
2096          * src_ip, dst_ip       ip addresses, in NETWORK format.
2097          *      Only valid for IPv4 packets.
2098          */
2099         u_int8_t proto;
2100         u_int16_t src_port = 0, dst_port = 0;   /* NOTE: host format    */
2101         struct in_addr src_ip, dst_ip;          /* NOTE: network format */
2102         u_int16_t ip_len=0;
2103         int pktlen;
2104
2105         /*
2106          * dyn_dir = MATCH_UNKNOWN when rules unchecked,
2107          *      MATCH_NONE when checked and not matched (q = NULL),
2108          *      MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
2109          */
2110         int dyn_dir = MATCH_UNKNOWN;
2111         ipfw_dyn_rule *q = NULL;
2112         struct ip_fw_chain *chain = &layer3_chain;
2113         struct m_tag *mtag;
2114
2115         /*
2116          * We store in ulp a pointer to the upper layer protocol header.
2117          * In the ipv4 case this is easy to determine from the header,
2118          * but for ipv6 we might have some additional headers in the middle.
2119          * ulp is NULL if not found.
2120          */
2121         void *ulp = NULL;               /* upper layer protocol pointer. */
2122         /* XXX ipv6 variables */
2123         int is_ipv6 = 0;
2124         u_int16_t ext_hd = 0;   /* bits vector for extension header filtering */
2125         /* end of ipv6 variables */
2126         int is_ipv4 = 0;
2127
2128         if (m->m_flags & M_SKIP_FIREWALL)
2129                 return (IP_FW_PASS);    /* accept */
2130
2131         pktlen = m->m_pkthdr.len;
2132         proto = args->f_id.proto = 0;   /* mark f_id invalid */
2133                 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
2134
2135 /*
2136  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
2137  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
2138  * pointer might become stale after other pullups (but we never use it
2139  * this way).
2140  */
2141 #define PULLUP_TO(len, p, T)                                            \
2142 do {                                                                    \
2143         int x = (len) + sizeof(T);                                      \
2144         if ((m)->m_len < x) {                                           \
2145                 args->m = m = m_pullup(m, x);                           \
2146                 if (m == NULL)                                          \
2147                         goto pullup_failed;                             \
2148         }                                                               \
2149         p = (mtod(m, char *) + (len));                                  \
2150 } while (0)
2151
2152         /* Identify IP packets and fill up variables. */
2153         if (pktlen >= sizeof(struct ip6_hdr) &&
2154             (args->eh == NULL || ntohs(args->eh->ether_type)==ETHERTYPE_IPV6) &&
2155             mtod(m, struct ip *)->ip_v == 6) {
2156                 is_ipv6 = 1;
2157                 args->f_id.addr_type = 6;
2158                 hlen = sizeof(struct ip6_hdr);
2159                 proto = mtod(m, struct ip6_hdr *)->ip6_nxt;
2160
2161                 /* Search extension headers to find upper layer protocols */
2162                 while (ulp == NULL) {
2163                         switch (proto) {
2164                         case IPPROTO_ICMPV6:
2165                                 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
2166                                 args->f_id.flags = ICMP6(ulp)->icmp6_type;
2167                                 break;
2168
2169                         case IPPROTO_TCP:
2170                                 PULLUP_TO(hlen, ulp, struct tcphdr);
2171                                 dst_port = TCP(ulp)->th_dport;
2172                                 src_port = TCP(ulp)->th_sport;
2173                                 args->f_id.flags = TCP(ulp)->th_flags;
2174                                 break;
2175
2176                         case IPPROTO_UDP:
2177                                 PULLUP_TO(hlen, ulp, struct udphdr);
2178                                 dst_port = UDP(ulp)->uh_dport;
2179                                 src_port = UDP(ulp)->uh_sport;
2180                                 break;
2181
2182                         case IPPROTO_HOPOPTS:   /* RFC 2460 */
2183                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
2184                                 ext_hd |= EXT_HOPOPTS;
2185                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
2186                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
2187                                 ulp = NULL;
2188                                 break;
2189
2190                         case IPPROTO_ROUTING:   /* RFC 2460 */
2191                                 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
2192                                 if (((struct ip6_rthdr *)ulp)->ip6r_type != 0) {
2193                                         printf("IPFW2: IPV6 - Unknown Routing "
2194                                             "Header type(%d)\n",
2195                                             ((struct ip6_rthdr *)ulp)->ip6r_type);
2196                                         if (fw_deny_unknown_exthdrs)
2197                                             return (IP_FW_DENY);
2198                                         break;
2199                                 }
2200                                 ext_hd |= EXT_ROUTING;
2201                                 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
2202                                 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
2203                                 ulp = NULL;
2204                                 break;
2205
2206                         case IPPROTO_FRAGMENT:  /* RFC 2460 */
2207                                 PULLUP_TO(hlen, ulp, struct ip6_frag);
2208                                 ext_hd |= EXT_FRAGMENT;
2209                                 hlen += sizeof (struct ip6_frag);
2210                                 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
2211                                 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
2212                                         IP6F_OFF_MASK;
2213                                 /* Add IP6F_MORE_FRAG for offset of first
2214                                  * fragment to be != 0. */
2215                                 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg &
2216                                         IP6F_MORE_FRAG;
2217                                 if (offset == 0) {
2218                                         printf("IPFW2: IPV6 - Invalid Fragment "
2219                                             "Header\n");
2220                                         if (fw_deny_unknown_exthdrs)
2221                                             return (IP_FW_DENY);
2222                                         break;
2223                                 }
2224                                 args->f_id.frag_id6 =
2225                                     ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
2226                                 ulp = NULL;
2227                                 break;
2228
2229                         case IPPROTO_DSTOPTS:   /* RFC 2460 */
2230                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
2231                                 ext_hd |= EXT_DSTOPTS;
2232                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
2233                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
2234                                 ulp = NULL;
2235                                 break;
2236
2237                         case IPPROTO_AH:        /* RFC 2402 */
2238                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
2239                                 ext_hd |= EXT_AH;
2240                                 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
2241                                 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
2242                                 ulp = NULL;
2243                                 break;
2244
2245                         case IPPROTO_ESP:       /* RFC 2406 */
2246                                 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
2247                                 /* Anything past Seq# is variable length and
2248                                  * data past this ext. header is encrypted. */
2249                                 ext_hd |= EXT_ESP;
2250                                 break;
2251
2252                         case IPPROTO_NONE:      /* RFC 2460 */
2253                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
2254                                 /* Packet ends here. if ip6e_len!=0 octets
2255                                  * must be ignored. */
2256                                 break;
2257
2258                         case IPPROTO_OSPFIGP:
2259                                 /* XXX OSPF header check? */
2260                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
2261                                 break;
2262
2263                         default:
2264                                 printf("IPFW2: IPV6 - Unknown Extension "
2265                                     "Header(%d), ext_hd=%x\n", proto, ext_hd);
2266                                 if (fw_deny_unknown_exthdrs)
2267                                     return (IP_FW_DENY);
2268                                 break;
2269                         } /*switch */
2270                 }
2271                 args->f_id.src_ip6 = mtod(m,struct ip6_hdr *)->ip6_src;
2272                 args->f_id.dst_ip6 = mtod(m,struct ip6_hdr *)->ip6_dst;
2273                 args->f_id.src_ip = 0;
2274                 args->f_id.dst_ip = 0;
2275                 args->f_id.flow_id6 = ntohl(mtod(m, struct ip6_hdr *)->ip6_flow);
2276         } else if (pktlen >= sizeof(struct ip) &&
2277             (args->eh == NULL || ntohs(args->eh->ether_type) == ETHERTYPE_IP) &&
2278             mtod(m, struct ip *)->ip_v == 4) {
2279                 is_ipv4 = 1;
2280                 ip = mtod(m, struct ip *);
2281                 hlen = ip->ip_hl << 2;
2282                 args->f_id.addr_type = 4;
2283
2284                 /*
2285                  * Collect parameters into local variables for faster matching.
2286                  */
2287                 proto = ip->ip_p;
2288                 src_ip = ip->ip_src;
2289                 dst_ip = ip->ip_dst;
2290                 if (args->eh != NULL) { /* layer 2 packets are as on the wire */
2291                         offset = ntohs(ip->ip_off) & IP_OFFMASK;
2292                         ip_len = ntohs(ip->ip_len);
2293                 } else {
2294                         offset = ip->ip_off & IP_OFFMASK;
2295                         ip_len = ip->ip_len;
2296                 }
2297                 pktlen = ip_len < pktlen ? ip_len : pktlen;
2298
2299                 if (offset == 0) {
2300                         switch (proto) {
2301                         case IPPROTO_TCP:
2302                                 PULLUP_TO(hlen, ulp, struct tcphdr);
2303                                 dst_port = TCP(ulp)->th_dport;
2304                                 src_port = TCP(ulp)->th_sport;
2305                                 args->f_id.flags = TCP(ulp)->th_flags;
2306                                 break;
2307
2308                         case IPPROTO_UDP:
2309                                 PULLUP_TO(hlen, ulp, struct udphdr);
2310                                 dst_port = UDP(ulp)->uh_dport;
2311                                 src_port = UDP(ulp)->uh_sport;
2312                                 break;
2313
2314                         case IPPROTO_ICMP:
2315                                 PULLUP_TO(hlen, ulp, struct icmphdr);
2316                                 args->f_id.flags = ICMP(ulp)->icmp_type;
2317                                 break;
2318
2319                         default:
2320                                 break;
2321                         }
2322                 }
2323
2324                 args->f_id.src_ip = ntohl(src_ip.s_addr);
2325                 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
2326         }
2327 #undef PULLUP_TO
2328         if (proto) { /* we may have port numbers, store them */
2329                 args->f_id.proto = proto;
2330                 args->f_id.src_port = src_port = ntohs(src_port);
2331                 args->f_id.dst_port = dst_port = ntohs(dst_port);
2332         }
2333
2334         IPFW_RLOCK(chain);
2335         mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL);
2336         if (args->rule) {
2337                 /*
2338                  * Packet has already been tagged. Look for the next rule
2339                  * to restart processing.
2340                  *
2341                  * If fw_one_pass != 0 then just accept it.
2342                  * XXX should not happen here, but optimized out in
2343                  * the caller.
2344                  */
2345                 if (fw_one_pass) {
2346                         IPFW_RUNLOCK(chain);
2347                         return (IP_FW_PASS);
2348                 }
2349
2350                 f = args->rule->next_rule;
2351                 if (f == NULL)
2352                         f = lookup_next_rule(args->rule);
2353         } else {
2354                 /*
2355                  * Find the starting rule. It can be either the first
2356                  * one, or the one after divert_rule if asked so.
2357                  */
2358                 int skipto = mtag ? divert_cookie(mtag) : 0;
2359
2360                 f = chain->rules;
2361                 if (args->eh == NULL && skipto != 0) {
2362                         if (skipto >= IPFW_DEFAULT_RULE) {
2363                                 IPFW_RUNLOCK(chain);
2364                                 return (IP_FW_DENY); /* invalid */
2365                         }
2366                         while (f && f->rulenum <= skipto)
2367                                 f = f->next;
2368                         if (f == NULL) {        /* drop packet */
2369                                 IPFW_RUNLOCK(chain);
2370                                 return (IP_FW_DENY);
2371                         }
2372                 }
2373         }
2374         /* reset divert rule to avoid confusion later */
2375         if (mtag) {
2376                 divinput_flags = divert_info(mtag) &
2377                     (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG);
2378                 m_tag_delete(m, mtag);
2379         }
2380
2381         /*
2382          * Now scan the rules, and parse microinstructions for each rule.
2383          */
2384         for (; f; f = f->next) {
2385                 ipfw_insn *cmd;
2386                 uint32_t tablearg = 0;
2387                 int l, cmdlen, skip_or; /* skip rest of OR block */
2388
2389 again:
2390                 if (set_disable & (1 << f->set) )
2391                         continue;
2392
2393                 skip_or = 0;
2394                 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
2395                     l -= cmdlen, cmd += cmdlen) {
2396                         int match;
2397
2398                         /*
2399                          * check_body is a jump target used when we find a
2400                          * CHECK_STATE, and need to jump to the body of
2401                          * the target rule.
2402                          */
2403
2404 check_body:
2405                         cmdlen = F_LEN(cmd);
2406                         /*
2407                          * An OR block (insn_1 || .. || insn_n) has the
2408                          * F_OR bit set in all but the last instruction.
2409                          * The first match will set "skip_or", and cause
2410                          * the following instructions to be skipped until
2411                          * past the one with the F_OR bit clear.
2412                          */
2413                         if (skip_or) {          /* skip this instruction */
2414                                 if ((cmd->len & F_OR) == 0)
2415                                         skip_or = 0;    /* next one is good */
2416                                 continue;
2417                         }
2418                         match = 0; /* set to 1 if we succeed */
2419
2420                         switch (cmd->opcode) {
2421                         /*
2422                          * The first set of opcodes compares the packet's
2423                          * fields with some pattern, setting 'match' if a
2424                          * match is found. At the end of the loop there is
2425                          * logic to deal with F_NOT and F_OR flags associated
2426                          * with the opcode.
2427                          */
2428                         case O_NOP:
2429                                 match = 1;
2430                                 break;
2431
2432                         case O_FORWARD_MAC:
2433                                 printf("ipfw: opcode %d unimplemented\n",
2434                                     cmd->opcode);
2435                                 break;
2436
2437                         case O_GID:
2438                         case O_UID:
2439                         case O_JAIL:
2440                                 /*
2441                                  * We only check offset == 0 && proto != 0,
2442                                  * as this ensures that we have a
2443                                  * packet with the ports info.
2444                                  */
2445                                 if (offset!=0)
2446                                         break;
2447                                 if (is_ipv6) /* XXX to be fixed later */
2448                                         break;
2449                                 if (proto == IPPROTO_TCP ||
2450                                     proto == IPPROTO_UDP)
2451                                         match = check_uidgid(
2452                                                     (ipfw_insn_u32 *)cmd,
2453                                                     proto, oif,
2454                                                     dst_ip, dst_port,
2455                                                     src_ip, src_port, &fw_ugid_cache,
2456                                                     &ugid_lookup, args->inp);
2457                                 break;
2458
2459                         case O_RECV:
2460                                 match = iface_match(m->m_pkthdr.rcvif,
2461                                     (ipfw_insn_if *)cmd);
2462                                 break;
2463
2464                         case O_XMIT:
2465                                 match = iface_match(oif, (ipfw_insn_if *)cmd);
2466                                 break;
2467
2468                         case O_VIA:
2469                                 match = iface_match(oif ? oif :
2470                                     m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd);
2471                                 break;
2472
2473                         case O_MACADDR2:
2474                                 if (args->eh != NULL) { /* have MAC header */
2475                                         u_int32_t *want = (u_int32_t *)
2476                                                 ((ipfw_insn_mac *)cmd)->addr;
2477                                         u_int32_t *mask = (u_int32_t *)
2478                                                 ((ipfw_insn_mac *)cmd)->mask;
2479                                         u_int32_t *hdr = (u_int32_t *)args->eh;
2480
2481                                         match =
2482                                             ( want[0] == (hdr[0] & mask[0]) &&
2483                                               want[1] == (hdr[1] & mask[1]) &&
2484                                               want[2] == (hdr[2] & mask[2]) );
2485                                 }
2486                                 break;
2487
2488                         case O_MAC_TYPE:
2489                                 if (args->eh != NULL) {
2490                                         u_int16_t t =
2491                                             ntohs(args->eh->ether_type);
2492                                         u_int16_t *p =
2493                                             ((ipfw_insn_u16 *)cmd)->ports;
2494                                         int i;
2495
2496                                         for (i = cmdlen - 1; !match && i>0;
2497                                             i--, p += 2)
2498                                                 match = (t>=p[0] && t<=p[1]);
2499                                 }
2500                                 break;
2501
2502                         case O_FRAG:
2503                                 match = (offset != 0);
2504                                 break;
2505
2506                         case O_IN:      /* "out" is "not in" */
2507                                 match = (oif == NULL);
2508                                 break;
2509
2510                         case O_LAYER2:
2511                                 match = (args->eh != NULL);
2512                                 break;
2513
2514                         case O_DIVERTED:
2515                                 match = (cmd->arg1 & 1 && divinput_flags &
2516                                     IP_FW_DIVERT_LOOPBACK_FLAG) ||
2517                                         (cmd->arg1 & 2 && divinput_flags &
2518                                     IP_FW_DIVERT_OUTPUT_FLAG);
2519                                 break;
2520
2521                         case O_PROTO:
2522                                 /*
2523                                  * We do not allow an arg of 0 so the
2524                                  * check of "proto" only suffices.
2525                                  */
2526                                 match = (proto == cmd->arg1);
2527                                 break;
2528
2529                         case O_IP_SRC:
2530                                 match = is_ipv4 &&
2531                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2532                                     src_ip.s_addr);
2533                                 break;
2534
2535                         case O_IP_SRC_LOOKUP:
2536                         case O_IP_DST_LOOKUP:
2537                                 if (is_ipv4) {
2538                                     uint32_t a =
2539                                         (cmd->opcode == O_IP_DST_LOOKUP) ?
2540                                             dst_ip.s_addr : src_ip.s_addr;
2541                                     uint32_t v;
2542
2543                                     match = lookup_table(chain, cmd->arg1, a,
2544                                         &v);
2545                                     if (!match)
2546                                         break;
2547                                     if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
2548                                         match =
2549                                             ((ipfw_insn_u32 *)cmd)->d[0] == v;
2550                                     else
2551                                         tablearg = v;
2552                                 }
2553                                 break;
2554
2555                         case O_IP_SRC_MASK:
2556                         case O_IP_DST_MASK:
2557                                 if (is_ipv4) {
2558                                     uint32_t a =
2559                                         (cmd->opcode == O_IP_DST_MASK) ?
2560                                             dst_ip.s_addr : src_ip.s_addr;
2561                                     uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2562                                     int i = cmdlen-1;
2563
2564                                     for (; !match && i>0; i-= 2, p+= 2)
2565                                         match = (p[0] == (a & p[1]));
2566                                 }
2567                                 break;
2568
2569                         case O_IP_SRC_ME:
2570                                 if (is_ipv4) {
2571                                         struct ifnet *tif;
2572
2573                                         INADDR_TO_IFP(src_ip, tif);
2574                                         match = (tif != NULL);
2575                                 }
2576                                 break;
2577
2578                         case O_IP_DST_SET:
2579                         case O_IP_SRC_SET:
2580                                 if (is_ipv4) {
2581                                         u_int32_t *d = (u_int32_t *)(cmd+1);
2582                                         u_int32_t addr =
2583                                             cmd->opcode == O_IP_DST_SET ?
2584                                                 args->f_id.dst_ip :
2585                                                 args->f_id.src_ip;
2586
2587                                             if (addr < d[0])
2588                                                     break;
2589                                             addr -= d[0]; /* subtract base */
2590                                             match = (addr < cmd->arg1) &&
2591                                                 ( d[ 1 + (addr>>5)] &
2592                                                   (1<<(addr & 0x1f)) );
2593                                 }
2594                                 break;
2595
2596                         case O_IP_DST:
2597                                 match = is_ipv4 &&
2598                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2599                                     dst_ip.s_addr);
2600                                 break;
2601
2602                         case O_IP_DST_ME:
2603                                 if (is_ipv4) {
2604                                         struct ifnet *tif;
2605
2606                                         INADDR_TO_IFP(dst_ip, tif);
2607                                         match = (tif != NULL);
2608                                 }
2609                                 break;
2610
2611                         case O_IP_SRCPORT:
2612                         case O_IP_DSTPORT:
2613                                 /*
2614                                  * offset == 0 && proto != 0 is enough
2615                                  * to guarantee that we have a
2616                                  * packet with port info.
2617                                  */
2618                                 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
2619                                     && offset == 0) {
2620                                         u_int16_t x =
2621                                             (cmd->opcode == O_IP_SRCPORT) ?
2622                                                 src_port : dst_port ;
2623                                         u_int16_t *p =
2624                                             ((ipfw_insn_u16 *)cmd)->ports;
2625                                         int i;
2626
2627                                         for (i = cmdlen - 1; !match && i>0;
2628                                             i--, p += 2)
2629                                                 match = (x>=p[0] && x<=p[1]);
2630                                 }
2631                                 break;
2632
2633                         case O_ICMPTYPE:
2634                                 match = (offset == 0 && proto==IPPROTO_ICMP &&
2635                                     icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
2636                                 break;
2637
2638 #ifdef INET6
2639                         case O_ICMP6TYPE:
2640                                 match = is_ipv6 && offset == 0 &&
2641                                     proto==IPPROTO_ICMPV6 &&
2642                                     icmp6type_match(
2643                                         ICMP6(ulp)->icmp6_type,
2644                                         (ipfw_insn_u32 *)cmd);
2645                                 break;
2646 #endif /* INET6 */
2647
2648                         case O_IPOPT:
2649                                 match = (is_ipv4 &&
2650                                     ipopts_match(mtod(m, struct ip *), cmd) );
2651                                 break;
2652
2653                         case O_IPVER:
2654                                 match = (is_ipv4 &&
2655                                     cmd->arg1 == mtod(m, struct ip *)->ip_v);
2656                                 break;
2657
2658                         case O_IPID:
2659                         case O_IPLEN:
2660                         case O_IPTTL:
2661                                 if (is_ipv4) {  /* only for IP packets */
2662                                     uint16_t x;
2663                                     uint16_t *p;
2664                                     int i;
2665
2666                                     if (cmd->opcode == O_IPLEN)
2667                                         x = ip_len;
2668                                     else if (cmd->opcode == O_IPTTL)
2669                                         x = mtod(m, struct ip *)->ip_ttl;
2670                                     else /* must be IPID */
2671                                         x = ntohs(mtod(m, struct ip *)->ip_id);
2672                                     if (cmdlen == 1) {
2673                                         match = (cmd->arg1 == x);
2674                                         break;
2675                                     }
2676                                     /* otherwise we have ranges */
2677                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2678                                     i = cmdlen - 1;
2679                                     for (; !match && i>0; i--, p += 2)
2680                                         match = (x >= p[0] && x <= p[1]);
2681                                 }
2682                                 break;
2683
2684                         case O_IPPRECEDENCE:
2685                                 match = (is_ipv4 &&
2686                                     (cmd->arg1 == (mtod(m, struct ip *)->ip_tos & 0xe0)) );
2687                                 break;
2688
2689                         case O_IPTOS:
2690                                 match = (is_ipv4 &&
2691                                     flags_match(cmd, mtod(m, struct ip *)->ip_tos));
2692                                 break;
2693
2694                         case O_TCPDATALEN:
2695                                 if (proto == IPPROTO_TCP && offset == 0) {
2696                                     struct tcphdr *tcp;
2697                                     uint16_t x;
2698                                     uint16_t *p;
2699                                     int i;
2700
2701                                     tcp = TCP(ulp);
2702                                     x = ip_len -
2703                                         ((ip->ip_hl + tcp->th_off) << 2);
2704                                     if (cmdlen == 1) {
2705                                         match = (cmd->arg1 == x);
2706                                         break;
2707                                     }
2708                                     /* otherwise we have ranges */
2709                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2710                                     i = cmdlen - 1;
2711                                     for (; !match && i>0; i--, p += 2)
2712                                         match = (x >= p[0] && x <= p[1]);
2713                                 }
2714                                 break;
2715
2716                         case O_TCPFLAGS:
2717                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2718                                     flags_match(cmd, TCP(ulp)->th_flags));
2719                                 break;
2720
2721                         case O_TCPOPTS:
2722                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2723                                     tcpopts_match(TCP(ulp), cmd));
2724                                 break;
2725
2726                         case O_TCPSEQ:
2727                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2728                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
2729                                         TCP(ulp)->th_seq);
2730                                 break;
2731
2732                         case O_TCPACK:
2733                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2734                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
2735                                         TCP(ulp)->th_ack);
2736                                 break;
2737
2738                         case O_TCPWIN:
2739                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2740                                     cmd->arg1 == TCP(ulp)->th_win);
2741                                 break;
2742
2743                         case O_ESTAB:
2744                                 /* reject packets which have SYN only */
2745                                 /* XXX should i also check for TH_ACK ? */
2746                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2747                                     (TCP(ulp)->th_flags &
2748                                      (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
2749                                 break;
2750
2751                         case O_ALTQ: {
2752                                 struct altq_tag *at;
2753                                 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
2754
2755                                 match = 1;
2756                                 mtag = m_tag_find(m, PACKET_TAG_PF_QID, NULL);
2757                                 if (mtag != NULL)
2758                                         break;
2759                                 mtag = m_tag_get(PACKET_TAG_PF_QID,
2760                                                 sizeof(struct altq_tag),
2761                                                 M_NOWAIT);
2762                                 if (mtag == NULL) {
2763                                         /*
2764                                          * Let the packet fall back to the
2765                                          * default ALTQ.
2766                                          */
2767                                         break;
2768                                 }
2769                                 at = (struct altq_tag *)(mtag+1);
2770                                 at->qid = altq->qid;
2771                                 if (is_ipv4)
2772                                         at->af = AF_INET;
2773                                 else
2774                                         at->af = AF_LINK;
2775                                 at->hdr = ip;
2776                                 m_tag_prepend(m, mtag);
2777                                 break;
2778                         }
2779
2780                         case O_LOG:
2781                                 if (fw_verbose)
2782                                         ipfw_log(f, hlen, args, m, oif, offset);
2783                                 match = 1;
2784                                 break;
2785
2786                         case O_PROB:
2787                                 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
2788                                 break;
2789
2790                         case O_VERREVPATH:
2791                                 /* Outgoing packets automatically pass/match */
2792                                 match = ((oif != NULL) ||
2793                                     (m->m_pkthdr.rcvif == NULL) ||
2794                                     (
2795 #ifdef INET6
2796                                     is_ipv6 ?
2797                                         verify_path6(&(args->f_id.src_ip6),
2798                                             m->m_pkthdr.rcvif) :
2799 #endif
2800                                     verify_path(src_ip, m->m_pkthdr.rcvif)));
2801                                 break;
2802
2803                         case O_VERSRCREACH:
2804                                 /* Outgoing packets automatically pass/match */
2805                                 match = (hlen > 0 && ((oif != NULL) ||
2806 #ifdef INET6
2807                                     is_ipv6 ?
2808                                         verify_path6(&(args->f_id.src_ip6),
2809                                             NULL) :
2810 #endif
2811                                     verify_path(src_ip, NULL)));
2812                                 break;
2813
2814                         case O_ANTISPOOF:
2815                                 /* Outgoing packets automatically pass/match */
2816                                 if (oif == NULL && hlen > 0 &&
2817                                     (  (is_ipv4 && in_localaddr(src_ip))
2818 #ifdef INET6
2819                                     || (is_ipv6 &&
2820                                         in6_localaddr(&(args->f_id.src_ip6)))
2821 #endif
2822                                     ))
2823                                         match =
2824 #ifdef INET6
2825                                             is_ipv6 ? verify_path6(
2826                                                 &(args->f_id.src_ip6),
2827                                                 m->m_pkthdr.rcvif) :
2828 #endif
2829                                             verify_path(src_ip,
2830                                                 m->m_pkthdr.rcvif);
2831                                 else
2832                                         match = 1;
2833                                 break;
2834
2835                         case O_IPSEC:
2836 #ifdef FAST_IPSEC
2837                                 match = (m_tag_find(m,
2838                                     PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
2839 #endif
2840 #ifdef IPSEC
2841                                 match = (ipsec_getnhist(m) != 0);
2842 #endif
2843                                 /* otherwise no match */
2844                                 break;
2845
2846 #ifdef INET6
2847                         case O_IP6_SRC:
2848                                 match = is_ipv6 &&
2849                                     IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
2850                                     &((ipfw_insn_ip6 *)cmd)->addr6);
2851                                 break;
2852
2853                         case O_IP6_DST:
2854                                 match = is_ipv6 &&
2855                                 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
2856                                     &((ipfw_insn_ip6 *)cmd)->addr6);
2857                                 break;
2858                         case O_IP6_SRC_MASK:
2859                                 if (is_ipv6) {
2860                                         ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd;
2861                                         struct in6_addr p = args->f_id.src_ip6;
2862
2863                                         APPLY_MASK(&p, &te->mask6);
2864                                         match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p);
2865                                 }
2866                                 break;
2867
2868                         case O_IP6_DST_MASK:
2869                                 if (is_ipv6) {
2870                                         ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd;
2871                                         struct in6_addr p = args->f_id.dst_ip6;
2872
2873                                         APPLY_MASK(&p, &te->mask6);
2874                                         match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p);
2875                                 }
2876                                 break;
2877
2878                         case O_IP6_SRC_ME:
2879                                 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6);
2880                                 break;
2881
2882                         case O_IP6_DST_ME:
2883                                 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6);
2884                                 break;
2885
2886                         case O_FLOW6ID:
2887                                 match = is_ipv6 &&
2888                                     flow6id_match(args->f_id.flow_id6,
2889                                     (ipfw_insn_u32 *) cmd);
2890                                 break;
2891
2892                         case O_EXT_HDR:
2893                                 match = is_ipv6 &&
2894                                     (ext_hd & ((ipfw_insn *) cmd)->arg1);
2895                                 break;
2896
2897                         case O_IP6:
2898                                 match = is_ipv6;
2899                                 break;
2900 #endif
2901
2902                         case O_IP4:
2903                                 match = is_ipv4;
2904                                 break;
2905
2906                         case O_TAG: {
2907                                 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
2908                                     tablearg : cmd->arg1;
2909
2910                                 /* Packet is already tagged with this tag? */
2911                                 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
2912
2913                                 /* We have `untag' action when F_NOT flag is
2914                                  * present. And we must remove this mtag from
2915                                  * mbuf and reset `match' to zero (`match' will
2916                                  * be inversed later).
2917                                  * Otherwise we should allocate new mtag and
2918                                  * push it into mbuf.
2919                                  */
2920                                 if (cmd->len & F_NOT) { /* `untag' action */
2921                                         if (mtag != NULL)
2922                                                 m_tag_delete(m, mtag);
2923                                 } else if (mtag == NULL) {
2924                                         if ((mtag = m_tag_alloc(MTAG_IPFW,
2925                                             tag, 0, M_NOWAIT)) != NULL)
2926                                                 m_tag_prepend(m, mtag);
2927                                 }
2928                                 match = (cmd->len & F_NOT) ? 0: 1;
2929                                 break;
2930                         }
2931
2932                         case O_TAGGED: {
2933                                 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
2934                                     tablearg : cmd->arg1;
2935
2936                                 if (cmdlen == 1) {
2937                                         match = m_tag_locate(m, MTAG_IPFW,
2938                                             tag, NULL) != NULL;
2939                                         break;
2940                                 }
2941
2942                                 /* we have ranges */
2943                                 for (mtag = m_tag_first(m);
2944                                     mtag != NULL && !match;
2945                                     mtag = m_tag_next(m, mtag)) {
2946                                         uint16_t *p;
2947                                         int i;
2948
2949                                         if (mtag->m_tag_cookie != MTAG_IPFW)
2950                                                 continue;
2951
2952                                         p = ((ipfw_insn_u16 *)cmd)->ports;
2953                                         i = cmdlen - 1;
2954                                         for(; !match && i > 0; i--, p += 2)
2955                                                 match =
2956                                                     mtag->m_tag_id >= p[0] &&
2957                                                     mtag->m_tag_id <= p[1];
2958                                 }
2959                                 break;
2960                         }
2961
2962                         /*
2963                          * The second set of opcodes represents 'actions',
2964                          * i.e. the terminal part of a rule once the packet
2965                          * matches all previous patterns.
2966                          * Typically there is only one action for each rule,
2967                          * and the opcode is stored at the end of the rule
2968                          * (but there are exceptions -- see below).
2969                          *
2970                          * In general, here we set retval and terminate the
2971                          * outer loop (would be a 'break 3' in some language,
2972                          * but we need to do a 'goto done').
2973                          *
2974                          * Exceptions:
2975                          * O_COUNT and O_SKIPTO actions:
2976                          *   instead of terminating, we jump to the next rule
2977                          *   ('goto next_rule', equivalent to a 'break 2'),
2978                          *   or to the SKIPTO target ('goto again' after
2979                          *   having set f, cmd and l), respectively.
2980                          *
2981                          * O_TAG, O_LOG and O_ALTQ action parameters:
2982                          *   perform some action and set match = 1;
2983                          *
2984                          * O_LIMIT and O_KEEP_STATE: these opcodes are
2985                          *   not real 'actions', and are stored right
2986                          *   before the 'action' part of the rule.
2987                          *   These opcodes try to install an entry in the
2988                          *   state tables; if successful, we continue with
2989                          *   the next opcode (match=1; break;), otherwise
2990                          *   the packet *   must be dropped
2991                          *   ('goto done' after setting retval);
2992                          *
2993                          * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2994                          *   cause a lookup of the state table, and a jump
2995                          *   to the 'action' part of the parent rule
2996                          *   ('goto check_body') if an entry is found, or
2997                          *   (CHECK_STATE only) a jump to the next rule if
2998                          *   the entry is not found ('goto next_rule').
2999                          *   The result of the lookup is cached to make
3000                          *   further instances of these opcodes are
3001                          *   effectively NOPs.
3002                          */
3003                         case O_LIMIT:
3004                         case O_KEEP_STATE:
3005                                 if (install_state(f,
3006                                     (ipfw_insn_limit *)cmd, args, tablearg)) {
3007                                         retval = IP_FW_DENY;
3008                                         goto done; /* error/limit violation */
3009                                 }
3010                                 match = 1;
3011                                 break;
3012
3013                         case O_PROBE_STATE:
3014                         case O_CHECK_STATE:
3015                                 /*
3016                                  * dynamic rules are checked at the first
3017                                  * keep-state or check-state occurrence,
3018                                  * with the result being stored in dyn_dir.
3019                                  * The compiler introduces a PROBE_STATE
3020                                  * instruction for us when we have a
3021                                  * KEEP_STATE (because PROBE_STATE needs
3022                                  * to be run first).
3023                                  */
3024                                 if (dyn_dir == MATCH_UNKNOWN &&
3025                                     (q = lookup_dyn_rule(&args->f_id,
3026                                      &dyn_dir, proto == IPPROTO_TCP ?
3027                                         TCP(ulp) : NULL))
3028                                         != NULL) {
3029                                         /*
3030                                          * Found dynamic entry, update stats
3031                                          * and jump to the 'action' part of
3032                                          * the parent rule.
3033                                          */
3034                                         q->pcnt++;
3035                                         q->bcnt += pktlen;
3036                                         f = q->rule;
3037                                         cmd = ACTION_PTR(f);
3038                                         l = f->cmd_len - f->act_ofs;
3039                                         IPFW_DYN_UNLOCK();
3040                                         goto check_body;
3041                                 }
3042                                 /*
3043                                  * Dynamic entry not found. If CHECK_STATE,
3044                                  * skip to next rule, if PROBE_STATE just
3045                                  * ignore and continue with next opcode.
3046                                  */
3047                                 if (cmd->opcode == O_CHECK_STATE)
3048                                         goto next_rule;
3049                                 match = 1;
3050                                 break;
3051
3052                         case O_ACCEPT:
3053                                 retval = 0;     /* accept */
3054                                 goto done;
3055
3056                         case O_PIPE:
3057                         case O_QUEUE:
3058                                 args->rule = f; /* report matching rule */
3059                                 if (cmd->arg1 == IP_FW_TABLEARG)
3060                                         args->cookie = tablearg;
3061                                 else
3062                                         args->cookie = cmd->arg1;
3063                                 retval = IP_FW_DUMMYNET;
3064                                 goto done;
3065
3066                         case O_DIVERT:
3067                         case O_TEE: {
3068                                 struct divert_tag *dt;
3069
3070                                 if (args->eh) /* not on layer 2 */
3071                                         break;
3072                                 mtag = m_tag_get(PACKET_TAG_DIVERT,
3073                                                 sizeof(struct divert_tag),
3074                                                 M_NOWAIT);
3075                                 if (mtag == NULL) {
3076                                         /* XXX statistic */
3077                                         /* drop packet */
3078                                         IPFW_RUNLOCK(chain);
3079                                         return (IP_FW_DENY);
3080                                 }
3081                                 dt = (struct divert_tag *)(mtag+1);
3082                                 dt->cookie = f->rulenum;
3083                                 if (cmd->arg1 == IP_FW_TABLEARG)
3084                                         dt->info = tablearg;
3085                                 else
3086                                         dt->info = cmd->arg1;
3087                                 m_tag_prepend(m, mtag);
3088                                 retval = (cmd->opcode == O_DIVERT) ?
3089                                     IP_FW_DIVERT : IP_FW_TEE;
3090                                 goto done;
3091                         }
3092
3093                         case O_COUNT:
3094                         case O_SKIPTO:
3095                                 f->pcnt++;      /* update stats */
3096                                 f->bcnt += pktlen;
3097                                 f->timestamp = time_uptime;
3098                                 if (cmd->opcode == O_COUNT)
3099                                         goto next_rule;
3100                                 /* handle skipto */
3101                                 if (f->next_rule == NULL)
3102                                         lookup_next_rule(f);
3103                                 f = f->next_rule;
3104                                 goto again;
3105
3106                         case O_REJECT:
3107                                 /*
3108                                  * Drop the packet and send a reject notice
3109                                  * if the packet is not ICMP (or is an ICMP
3110                                  * query), and it is not multicast/broadcast.
3111                                  */
3112                                 if (hlen > 0 && is_ipv4 && offset == 0 &&
3113                                     (proto != IPPROTO_ICMP ||
3114                                      is_icmp_query(ICMP(ulp))) &&
3115                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
3116                                     !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
3117                                         send_reject(args, cmd->arg1,
3118                                             offset,ip_len);
3119                                         m = args->m;
3120                                 }
3121                                 /* FALLTHROUGH */
3122 #ifdef INET6
3123                         case O_UNREACH6:
3124                                 if (hlen > 0 && is_ipv6 &&
3125                                     (proto != IPPROTO_ICMPV6 ||
3126                                      (is_icmp6_query(args->f_id.flags) == 1)) &&
3127                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
3128                                     !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
3129                                         send_reject6(args, cmd->arg1,
3130                                             offset, hlen);
3131                                         m = args->m;
3132                                 }
3133                                 /* FALLTHROUGH */
3134 #endif
3135                         case O_DENY:
3136                                 retval = IP_FW_DENY;
3137                                 goto done;
3138
3139                         case O_FORWARD_IP:
3140                                 if (args->eh)   /* not valid on layer2 pkts */
3141                                         break;
3142                                 if (!q || dyn_dir == MATCH_FORWARD)
3143                                         args->next_hop =
3144                                             &((ipfw_insn_sa *)cmd)->sa;
3145                                 retval = IP_FW_PASS;
3146                                 goto done;
3147
3148                         case O_NETGRAPH:
3149                         case O_NGTEE:
3150                                 args->rule = f; /* report matching rule */
3151                                 if (cmd->arg1 == IP_FW_TABLEARG)
3152                                         args->cookie = tablearg;
3153                                 else
3154                                         args->cookie = cmd->arg1;
3155                                 retval = (cmd->opcode == O_NETGRAPH) ?
3156                                     IP_FW_NETGRAPH : IP_FW_NGTEE;
3157                                 goto done;
3158
3159                         default:
3160                                 panic("-- unknown opcode %d\n", cmd->opcode);
3161                         } /* end of switch() on opcodes */
3162
3163                         if (cmd->len & F_NOT)
3164                                 match = !match;
3165
3166                         if (match) {
3167                                 if (cmd->len & F_OR)
3168                                         skip_or = 1;
3169                         } else {
3170                                 if (!(cmd->len & F_OR)) /* not an OR block, */
3171                                         break;          /* try next rule    */
3172                         }
3173
3174                 }       /* end of inner for, scan opcodes */
3175
3176 next_rule:;             /* try next rule                */
3177
3178         }               /* end of outer for, scan rules */
3179         printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3180         IPFW_RUNLOCK(chain);
3181         return (IP_FW_DENY);
3182
3183 done:
3184         /* Update statistics */
3185         f->pcnt++;
3186         f->bcnt += pktlen;
3187         f->timestamp = time_uptime;
3188         IPFW_RUNLOCK(chain);
3189         return (retval);
3190
3191 pullup_failed:
3192         if (fw_verbose)
3193                 printf("ipfw: pullup failed\n");
3194         return (IP_FW_DENY);
3195 }
3196
3197 /*
3198  * When a rule is added/deleted, clear the next_rule pointers in all rules.
3199  * These will be reconstructed on the fly as packets are matched.
3200  */
3201 static void
3202 flush_rule_ptrs(struct ip_fw_chain *chain)
3203 {
3204         struct ip_fw *rule;
3205
3206         IPFW_WLOCK_ASSERT(chain);
3207
3208         for (rule = chain->rules; rule; rule = rule->next)
3209                 rule->next_rule = NULL;
3210 }
3211
3212 /*
3213  * Add a new rule to the list. Copy the rule into a malloc'ed area, then
3214  * possibly create a rule number and add the rule to the list.
3215  * Update the rule_number in the input struct so the caller knows it as well.
3216  */
3217 static int
3218 add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule)
3219 {
3220         struct ip_fw *rule, *f, *prev;
3221         int l = RULESIZE(input_rule);
3222
3223         if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE)
3224                 return (EINVAL);
3225
3226         rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO);
3227         if (rule == NULL)
3228                 return (ENOSPC);
3229
3230         bcopy(input_rule, rule, l);
3231
3232         rule->next = NULL;
3233         rule->next_rule = NULL;
3234
3235         rule->pcnt = 0;
3236         rule->bcnt = 0;
3237         rule->timestamp = 0;
3238
3239         IPFW_WLOCK(chain);
3240
3241         if (chain->rules == NULL) {     /* default rule */
3242                 chain->rules = rule;
3243                 goto done;
3244         }
3245
3246         /*
3247          * If rulenum is 0, find highest numbered rule before the
3248          * default rule, and add autoinc_step
3249          */
3250         if (autoinc_step < 1)
3251                 autoinc_step = 1;
3252         else if (autoinc_step > 1000)
3253                 autoinc_step = 1000;
3254         if (rule->rulenum == 0) {
3255                 /*
3256                  * locate the highest numbered rule before default
3257                  */
3258                 for (f = chain->rules; f; f = f->next) {
3259                         if (f->rulenum == IPFW_DEFAULT_RULE)
3260                                 break;
3261                         rule->rulenum = f->rulenum;
3262                 }
3263                 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step)
3264                         rule->rulenum += autoinc_step;
3265                 input_rule->rulenum = rule->rulenum;
3266         }
3267
3268         /*
3269          * Now insert the new rule in the right place in the sorted list.
3270          */
3271         for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) {
3272                 if (f->rulenum > rule->rulenum) { /* found the location */
3273                         if (prev) {
3274                                 rule->next = f;
3275                                 prev->next = rule;
3276                         } else { /* head insert */
3277                                 rule->next = chain->rules;
3278                                 chain->rules = rule;
3279                         }
3280                         break;
3281                 }
3282         }
3283         flush_rule_ptrs(chain);
3284 done:
3285         static_count++;
3286         static_len += l;
3287         IPFW_WUNLOCK(chain);
3288         DEB(printf("ipfw: installed rule %d, static count now %d\n",
3289                 rule->rulenum, static_count);)
3290         return (0);
3291 }
3292
3293 /**
3294  * Remove a static rule (including derived * dynamic rules)
3295  * and place it on the ``reap list'' for later reclamation.
3296  * The caller is in charge of clearing rule pointers to avoid
3297  * dangling pointers.
3298  * @return a pointer to the next entry.
3299  * Arguments are not checked, so they better be correct.
3300  */
3301 static struct ip_fw *
3302 remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev)
3303 {
3304         struct ip_fw *n;
3305         int l = RULESIZE(rule);
3306
3307         IPFW_WLOCK_ASSERT(chain);
3308
3309         n = rule->next;
3310         IPFW_DYN_LOCK();
3311         remove_dyn_rule(rule, NULL /* force removal */);
3312         IPFW_DYN_UNLOCK();
3313         if (prev == NULL)
3314                 chain->rules = n;
3315         else
3316                 prev->next = n;
3317         static_count--;
3318         static_len -= l;
3319
3320         rule->next = chain->reap;
3321         chain->reap = rule;
3322
3323         return n;
3324 }
3325
3326 /**
3327  * Reclaim storage associated with a list of rules.  This is
3328  * typically the list created using remove_rule.
3329  */
3330 static void
3331 reap_rules(struct ip_fw *head)
3332 {
3333         struct ip_fw *rule;
3334
3335         while ((rule = head) != NULL) {
3336                 head = head->next;
3337                 if (DUMMYNET_LOADED)
3338                         ip_dn_ruledel_ptr(rule);
3339                 free(rule, M_IPFW);
3340         }
3341 }
3342
3343 /*
3344  * Remove all rules from a chain (except rules in set RESVD_SET
3345  * unless kill_default = 1).  The caller is responsible for
3346  * reclaiming storage for the rules left in chain->reap.
3347  */
3348 static void
3349 free_chain(struct ip_fw_chain *chain, int kill_default)
3350 {
3351         struct ip_fw *prev, *rule;
3352
3353         IPFW_WLOCK_ASSERT(chain);
3354
3355         flush_rule_ptrs(chain); /* more efficient to do outside the loop */
3356         for (prev = NULL, rule = chain->rules; rule ; )
3357                 if (kill_default || rule->set != RESVD_SET)
3358                         rule = remove_rule(chain, rule, prev);
3359                 else {
3360                         prev = rule;
3361                         rule = rule->next;
3362                 }
3363 }
3364
3365 /**
3366  * Remove all rules with given number, and also do set manipulation.
3367  * Assumes chain != NULL && *chain != NULL.
3368  *
3369  * The argument is an u_int32_t. The low 16 bit are the rule or set number,
3370  * the next 8 bits are the new set, the top 8 bits are the command:
3371  *
3372  *      0       delete rules with given number
3373  *      1       delete rules with given set number
3374  *      2       move rules with given number to new set
3375  *      3       move rules with given set number to new set
3376  *      4       swap sets with given numbers
3377  */
3378 static int
3379 del_entry(struct ip_fw_chain *chain, u_int32_t arg)
3380 {
3381         struct ip_fw *prev = NULL, *rule;
3382         u_int16_t rulenum;      /* rule or old_set */
3383         u_int8_t cmd, new_set;
3384
3385         rulenum = arg & 0xffff;
3386         cmd = (arg >> 24) & 0xff;
3387         new_set = (arg >> 16) & 0xff;
3388
3389         if (cmd > 4)
3390                 return EINVAL;
3391         if (new_set > RESVD_SET)
3392                 return EINVAL;
3393         if (cmd == 0 || cmd == 2) {
3394                 if (rulenum >= IPFW_DEFAULT_RULE)
3395                         return EINVAL;
3396         } else {
3397                 if (rulenum > RESVD_SET)        /* old_set */
3398                         return EINVAL;
3399         }
3400
3401         IPFW_WLOCK(chain);
3402         rule = chain->rules;
3403         chain->reap = NULL;
3404         switch (cmd) {
3405         case 0: /* delete rules with given number */
3406                 /*
3407                  * locate first rule to delete
3408                  */
3409                 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next)
3410                         ;
3411                 if (rule->rulenum != rulenum) {
3412                         IPFW_WUNLOCK(chain);
3413                         return EINVAL;
3414                 }
3415
3416                 /*
3417                  * flush pointers outside the loop, then delete all matching
3418                  * rules. prev remains the same throughout the cycle.
3419                  */
3420                 flush_rule_ptrs(chain);
3421                 while (rule->rulenum == rulenum)
3422                         rule = remove_rule(chain, rule, prev);
3423                 break;
3424
3425         case 1: /* delete all rules with given set number */
3426                 flush_rule_ptrs(chain);
3427                 rule = chain->rules;
3428                 while (rule->rulenum < IPFW_DEFAULT_RULE)
3429                         if (rule->set == rulenum)
3430                                 rule = remove_rule(chain, rule, prev);
3431                         else {
3432                                 prev = rule;
3433                                 rule = rule->next;
3434                         }
3435                 break;
3436
3437         case 2: /* move rules with given number to new set */
3438                 rule = chain->rules;
3439                 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
3440                         if (rule->rulenum == rulenum)
3441                                 rule->set = new_set;
3442                 break;
3443
3444         case 3: /* move rules with given set number to new set */
3445                 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
3446                         if (rule->set == rulenum)
3447                                 rule->set = new_set;
3448                 break;
3449
3450         case 4: /* swap two sets */
3451                 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
3452                         if (rule->set == rulenum)
3453                                 rule->set = new_set;
3454                         else if (rule->set == new_set)
3455                                 rule->set = rulenum;
3456                 break;
3457         }
3458         /*
3459          * Look for rules to reclaim.  We grab the list before
3460          * releasing the lock then reclaim them w/o the lock to
3461          * avoid a LOR with dummynet.
3462          */
3463         rule = chain->reap;
3464         chain->reap = NULL;
3465         IPFW_WUNLOCK(chain);
3466         if (rule)
3467                 reap_rules(rule);
3468         return 0;
3469 }
3470
3471 /*
3472  * Clear counters for a specific rule.
3473  * The enclosing "table" is assumed locked.
3474  */
3475 static void
3476 clear_counters(struct ip_fw *rule, int log_only)
3477 {
3478         ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule);
3479
3480         if (log_only == 0) {
3481                 rule->bcnt = rule->pcnt = 0;
3482                 rule->timestamp = 0;
3483         }
3484         if (l->o.opcode == O_LOG)
3485                 l->log_left = l->max_log;
3486 }
3487
3488 /**
3489  * Reset some or all counters on firewall rules.
3490  * @arg frwl is null to clear all entries, or contains a specific
3491  * rule number.
3492  * @arg log_only is 1 if we only want to reset logs, zero otherwise.
3493  */
3494 static int
3495 zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only)
3496 {
3497         struct ip_fw *rule;
3498         char *msg;
3499
3500         IPFW_WLOCK(chain);
3501         if (rulenum == 0) {
3502                 norule_counter = 0;
3503                 for (rule = chain->rules; rule; rule = rule->next)
3504                         clear_counters(rule, log_only);
3505                 msg = log_only ? "ipfw: All logging counts reset.\n" :
3506                                 "ipfw: Accounting cleared.\n";
3507         } else {
3508                 int cleared = 0;
3509                 /*
3510                  * We can have multiple rules with the same number, so we
3511                  * need to clear them all.
3512                  */
3513                 for (rule = chain->rules; rule; rule = rule->next)
3514                         if (rule->rulenum == rulenum) {
3515                                 while (rule && rule->rulenum == rulenum) {
3516                                         clear_counters(rule, log_only);
3517                                         rule = rule->next;
3518                                 }
3519                                 cleared = 1;
3520                                 break;
3521                         }
3522                 if (!cleared) { /* we did not find any matching rules */
3523                         IPFW_WUNLOCK(chain);
3524                         return (EINVAL);
3525                 }
3526                 msg = log_only ? "ipfw: Entry %d logging count reset.\n" :
3527                                 "ipfw: Entry %d cleared.\n";
3528         }
3529         IPFW_WUNLOCK(chain);
3530
3531         if (fw_verbose)
3532                 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum);
3533         return (0);
3534 }
3535
3536 /*
3537  * Check validity of the structure before insert.
3538  * Fortunately rules are simple, so this mostly need to check rule sizes.
3539  */
3540 static int
3541 check_ipfw_struct(struct ip_fw *rule, int size)
3542 {
3543         int l, cmdlen = 0;
3544         int have_action=0;
3545         ipfw_insn *cmd;
3546
3547         if (size < sizeof(*rule)) {
3548                 printf("ipfw: rule too short\n");
3549                 return (EINVAL);
3550         }
3551         /* first, check for valid size */
3552         l = RULESIZE(rule);
3553         if (l != size) {
3554                 printf("ipfw: size mismatch (have %d want %d)\n", size, l);
3555                 return (EINVAL);
3556         }
3557         if (rule->act_ofs >= rule->cmd_len) {
3558                 printf("ipfw: bogus action offset (%u > %u)\n",
3559                     rule->act_ofs, rule->cmd_len - 1);
3560                 return (EINVAL);
3561         }
3562         /*
3563          * Now go for the individual checks. Very simple ones, basically only
3564          * instruction sizes.
3565          */
3566         for (l = rule->cmd_len, cmd = rule->cmd ;
3567                         l > 0 ; l -= cmdlen, cmd += cmdlen) {
3568                 cmdlen = F_LEN(cmd);
3569                 if (cmdlen > l) {
3570                         printf("ipfw: opcode %d size truncated\n",
3571                             cmd->opcode);
3572                         return EINVAL;
3573                 }
3574                 DEB(printf("ipfw: opcode %d\n", cmd->opcode);)
3575                 switch (cmd->opcode) {
3576                 case O_PROBE_STATE:
3577                 case O_KEEP_STATE:
3578                 case O_PROTO:
3579                 case O_IP_SRC_ME:
3580                 case O_IP_DST_ME:
3581                 case O_LAYER2:
3582                 case O_IN:
3583                 case O_FRAG:
3584                 case O_DIVERTED:
3585                 case O_IPOPT:
3586                 case O_IPTOS:
3587                 case O_IPPRECEDENCE:
3588                 case O_IPVER:
3589                 case O_TCPWIN:
3590                 case O_TCPFLAGS:
3591                 case O_TCPOPTS:
3592                 case O_ESTAB:
3593                 case O_VERREVPATH:
3594                 case O_VERSRCREACH:
3595                 case O_ANTISPOOF:
3596                 case O_IPSEC:
3597 #ifdef INET6
3598                 case O_IP6_SRC_ME:
3599                 case O_IP6_DST_ME:
3600                 case O_EXT_HDR:
3601                 case O_IP6:
3602 #endif
3603                 case O_IP4:
3604                 case O_TAG:
3605                         if (cmdlen != F_INSN_SIZE(ipfw_insn))
3606                                 goto bad_size;
3607                         break;
3608
3609                 case O_UID:
3610                 case O_GID:
3611                 case O_JAIL:
3612                 case O_IP_SRC:
3613                 case O_IP_DST:
3614                 case O_TCPSEQ:
3615                 case O_TCPACK:
3616                 case O_PROB:
3617                 case O_ICMPTYPE:
3618                         if (cmdlen != F_INSN_SIZE(ipfw_insn_u32))
3619                                 goto bad_size;
3620                         break;
3621
3622                 case O_LIMIT:
3623                         if (cmdlen != F_INSN_SIZE(ipfw_insn_limit))
3624                                 goto bad_size;
3625                         break;
3626
3627                 case O_LOG:
3628                         if (cmdlen != F_INSN_SIZE(ipfw_insn_log))
3629                                 goto bad_size;
3630
3631                         ((ipfw_insn_log *)cmd)->log_left =
3632                             ((ipfw_insn_log *)cmd)->max_log;
3633
3634                         break;
3635
3636                 case O_IP_SRC_MASK:
3637                 case O_IP_DST_MASK:
3638                         /* only odd command lengths */
3639                         if ( !(cmdlen & 1) || cmdlen > 31)
3640                                 goto bad_size;
3641                         break;
3642
3643                 case O_IP_SRC_SET:
3644                 case O_IP_DST_SET:
3645                         if (cmd->arg1 == 0 || cmd->arg1 > 256) {
3646                                 printf("ipfw: invalid set size %d\n",
3647                                         cmd->arg1);
3648                                 return EINVAL;
3649                         }
3650                         if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
3651                             (cmd->arg1+31)/32 )
3652                                 goto bad_size;
3653                         break;
3654
3655                 case O_IP_SRC_LOOKUP:
3656                 case O_IP_DST_LOOKUP:
3657                         if (cmd->arg1 >= IPFW_TABLES_MAX) {
3658                                 printf("ipfw: invalid table number %d\n",
3659                                     cmd->arg1);
3660                                 return (EINVAL);
3661                         }
3662                         if (cmdlen != F_INSN_SIZE(ipfw_insn) &&
3663                             cmdlen != F_INSN_SIZE(ipfw_insn_u32))
3664                                 goto bad_size;
3665                         break;
3666
3667                 case O_MACADDR2:
3668                         if (cmdlen != F_INSN_SIZE(ipfw_insn_mac))
3669                                 goto bad_size;
3670                         break;
3671
3672                 case O_NOP:
3673                 case O_IPID:
3674                 case O_IPTTL:
3675                 case O_IPLEN:
3676                 case O_TCPDATALEN:
3677                 case O_TAGGED:
3678                         if (cmdlen < 1 || cmdlen > 31)
3679                                 goto bad_size;
3680                         break;
3681
3682                 case O_MAC_TYPE:
3683                 case O_IP_SRCPORT:
3684                 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */
3685                         if (cmdlen < 2 || cmdlen > 31)
3686                                 goto bad_size;
3687                         break;
3688
3689                 case O_RECV:
3690                 case O_XMIT:
3691                 case O_VIA:
3692                         if (cmdlen != F_INSN_SIZE(ipfw_insn_if))
3693                                 goto bad_size;
3694                         break;
3695
3696                 case O_ALTQ:
3697                         if (cmdlen != F_INSN_SIZE(ipfw_insn_altq))
3698                                 goto bad_size;
3699                         break;
3700
3701                 case O_PIPE:
3702                 case O_QUEUE:
3703                         if (cmdlen != F_INSN_SIZE(ipfw_insn))
3704                                 goto bad_size;
3705                         goto check_action;
3706
3707                 case O_FORWARD_IP:
3708 #ifdef  IPFIREWALL_FORWARD
3709                         if (cmdlen != F_INSN_SIZE(ipfw_insn_sa))
3710                                 goto bad_size;
3711                         goto check_action;
3712 #else
3713                         return EINVAL;
3714 #endif
3715
3716                 case O_DIVERT:
3717                 case O_TEE:
3718                         if (ip_divert_ptr == NULL)
3719                                 return EINVAL;
3720                         else
3721                                 goto check_size;
3722                 case O_NETGRAPH:
3723                 case O_NGTEE:
3724                         if (!NG_IPFW_LOADED)
3725                                 return EINVAL;
3726                         else
3727                                 goto check_size;
3728                 case O_FORWARD_MAC: /* XXX not implemented yet */
3729                 case O_CHECK_STATE:
3730                 case O_COUNT:
3731                 case O_ACCEPT:
3732                 case O_DENY:
3733                 case O_REJECT:
3734 #ifdef INET6
3735                 case O_UNREACH6:
3736 #endif
3737                 case O_SKIPTO:
3738 check_size:
3739                         if (cmdlen != F_INSN_SIZE(ipfw_insn))
3740                                 goto bad_size;
3741 check_action:
3742                         if (have_action) {
3743                                 printf("ipfw: opcode %d, multiple actions"
3744                                         " not allowed\n",
3745                                         cmd->opcode);
3746                                 return EINVAL;
3747                         }
3748                         have_action = 1;
3749                         if (l != cmdlen) {
3750                                 printf("ipfw: opcode %d, action must be"
3751                                         " last opcode\n",
3752                                         cmd->opcode);
3753                                 return EINVAL;
3754                         }
3755                         break;
3756 #ifdef INET6
3757                 case O_IP6_SRC:
3758                 case O_IP6_DST:
3759                         if (cmdlen != F_INSN_SIZE(struct in6_addr) +
3760                             F_INSN_SIZE(ipfw_insn))
3761                                 goto bad_size;
3762                         break;
3763
3764                 case O_FLOW6ID:
3765                         if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
3766                             ((ipfw_insn_u32 *)cmd)->o.arg1)
3767                                 goto bad_size;
3768                         break;
3769
3770                 case O_IP6_SRC_MASK:
3771                 case O_IP6_DST_MASK:
3772                         if ( !(cmdlen & 1) || cmdlen > 127)
3773                                 goto bad_size;
3774                         break;
3775                 case O_ICMP6TYPE:
3776                         if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) )
3777                                 goto bad_size;
3778                         break;
3779 #endif
3780
3781                 default:
3782                         switch (cmd->opcode) {
3783 #ifndef INET6
3784                         case O_IP6_SRC_ME:
3785                         case O_IP6_DST_ME:
3786                         case O_EXT_HDR:
3787                         case O_IP6:
3788                         case O_UNREACH6:
3789                         case O_IP6_SRC:
3790                         case O_IP6_DST:
3791                         case O_FLOW6ID:
3792                         case O_IP6_SRC_MASK:
3793                         case O_IP6_DST_MASK:
3794                         case O_ICMP6TYPE:
3795                                 printf("ipfw: no IPv6 support in kernel\n");
3796                                 return EPROTONOSUPPORT;
3797 #endif
3798                         default:
3799                                 printf("ipfw: opcode %d, unknown opcode\n",
3800                                         cmd->opcode);
3801                                 return EINVAL;
3802                         }
3803                 }
3804         }
3805         if (have_action == 0) {
3806                 printf("ipfw: missing action\n");
3807                 return EINVAL;
3808         }
3809         return 0;
3810
3811 bad_size:
3812         printf("ipfw: opcode %d size %d wrong\n",
3813                 cmd->opcode, cmdlen);
3814         return EINVAL;
3815 }
3816
3817 /*
3818  * Copy the static and dynamic rules to the supplied buffer
3819  * and return the amount of space actually used.
3820  */
3821 static size_t
3822 ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space)
3823 {
3824         char *bp = buf;
3825         char *ep = bp + space;
3826         struct ip_fw *rule;
3827         int i;
3828
3829         /* XXX this can take a long time and locking will block packet flow */
3830         IPFW_RLOCK(chain);
3831         for (rule = chain->rules; rule ; rule = rule->next) {
3832                 /*
3833                  * Verify the entry fits in the buffer in case the
3834                  * rules changed between calculating buffer space and
3835                  * now.  This would be better done using a generation
3836                  * number but should suffice for now.
3837                  */
3838                 i = RULESIZE(rule);
3839                 if (bp + i <= ep) {
3840                         bcopy(rule, bp, i);
3841                         bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule),
3842                             sizeof(set_disable));
3843                         bp += i;
3844                 }
3845         }
3846         IPFW_RUNLOCK(chain);
3847         if (ipfw_dyn_v) {
3848                 ipfw_dyn_rule *p, *last = NULL;
3849
3850                 IPFW_DYN_LOCK();
3851                 for (i = 0 ; i < curr_dyn_buckets; i++)
3852                         for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) {
3853                                 if (bp + sizeof *p <= ep) {
3854                                         ipfw_dyn_rule *dst =
3855                                                 (ipfw_dyn_rule *)bp;
3856                                         bcopy(p, dst, sizeof *p);
3857                                         bcopy(&(p->rule->rulenum), &(dst->rule),
3858                                             sizeof(p->rule->rulenum));
3859                                         /*
3860                                          * store a non-null value in "next".
3861                                          * The userland code will interpret a
3862                                          * NULL here as a marker
3863                                          * for the last dynamic rule.
3864                                          */
3865                                         bcopy(&dst, &dst->next, sizeof(dst));
3866                                         last = dst;
3867                                         dst->expire =
3868                                             TIME_LEQ(dst->expire, time_uptime) ?
3869                                                 0 : dst->expire - time_uptime ;
3870                                         bp += sizeof(ipfw_dyn_rule);
3871                                 }
3872                         }
3873                 IPFW_DYN_UNLOCK();
3874                 if (last != NULL) /* mark last dynamic rule */
3875                         bzero(&last->next, sizeof(last));
3876         }
3877         return (bp - (char *)buf);
3878 }
3879
3880
3881 /**
3882  * {set|get}sockopt parser.
3883  */
3884 static int
3885 ipfw_ctl(struct sockopt *sopt)
3886 {
3887 #define RULE_MAXSIZE    (256*sizeof(u_int32_t))
3888         int error, rule_num;
3889         size_t size;
3890         struct ip_fw *buf, *rule;
3891         u_int32_t rulenum[2];
3892
3893         error = suser(sopt->sopt_td);
3894         if (error)
3895                 return (error);
3896
3897         /*
3898          * Disallow modifications in really-really secure mode, but still allow
3899          * the logging counters to be reset.
3900          */
3901         if (sopt->sopt_name == IP_FW_ADD ||
3902             (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) {
3903                 error = securelevel_ge(sopt->sopt_td->td_ucred, 3);
3904                 if (error)
3905                         return (error);
3906         }
3907
3908         error = 0;
3909
3910         switch (sopt->sopt_name) {
3911         case IP_FW_GET:
3912                 /*
3913                  * pass up a copy of the current rules. Static rules
3914                  * come first (the last of which has number IPFW_DEFAULT_RULE),
3915                  * followed by a possibly empty list of dynamic rule.
3916                  * The last dynamic rule has NULL in the "next" field.
3917                  *
3918                  * Note that the calculated size is used to bound the
3919                  * amount of data returned to the user.  The rule set may
3920                  * change between calculating the size and returning the
3921                  * data in which case we'll just return what fits.
3922                  */
3923                 size = static_len;      /* size of static rules */
3924                 if (ipfw_dyn_v)         /* add size of dyn.rules */
3925                         size += (dyn_count * sizeof(ipfw_dyn_rule));
3926
3927                 /*
3928                  * XXX todo: if the user passes a short length just to know
3929                  * how much room is needed, do not bother filling up the
3930                  * buffer, just jump to the sooptcopyout.
3931                  */
3932                 buf = malloc(size, M_TEMP, M_WAITOK);
3933                 error = sooptcopyout(sopt, buf,
3934                                 ipfw_getrules(&layer3_chain, buf, size));
3935                 free(buf, M_TEMP);
3936                 break;
3937
3938         case IP_FW_FLUSH:
3939                 /*
3940                  * Normally we cannot release the lock on each iteration.
3941                  * We could do it here only because we start from the head all
3942                  * the times so there is no risk of missing some entries.
3943                  * On the other hand, the risk is that we end up with
3944                  * a very inconsistent ruleset, so better keep the lock
3945                  * around the whole cycle.
3946                  *
3947                  * XXX this code can be improved by resetting the head of
3948                  * the list to point to the default rule, and then freeing
3949                  * the old list without the need for a lock.
3950                  */
3951
3952                 IPFW_WLOCK(&layer3_chain);
3953                 layer3_chain.reap = NULL;
3954                 free_chain(&layer3_chain, 0 /* keep default rule */);
3955                 rule = layer3_chain.reap, layer3_chain.reap = NULL;
3956                 IPFW_WUNLOCK(&layer3_chain);
3957                 if (layer3_chain.reap != NULL)
3958                         reap_rules(rule);
3959                 break;
3960
3961         case IP_FW_ADD:
3962                 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK);
3963                 error = sooptcopyin(sopt, rule, RULE_MAXSIZE,
3964                         sizeof(struct ip_fw) );
3965                 if (error == 0)
3966                         error = check_ipfw_struct(rule, sopt->sopt_valsize);
3967                 if (error == 0) {
3968                         error = add_rule(&layer3_chain, rule);
3969                         size = RULESIZE(rule);
3970                         if (!error && sopt->sopt_dir == SOPT_GET)
3971                                 error = sooptcopyout(sopt, rule, size);
3972                 }
3973                 free(rule, M_TEMP);
3974                 break;
3975
3976         case IP_FW_DEL:
3977                 /*
3978                  * IP_FW_DEL is used for deleting single rules or sets,
3979                  * and (ab)used to atomically manipulate sets. Argument size
3980                  * is used to distinguish between the two:
3981                  *    sizeof(u_int32_t)
3982                  *      delete single rule or set of rules,
3983                  *      or reassign rules (or sets) to a different set.
3984                  *    2*sizeof(u_int32_t)
3985                  *      atomic disable/enable sets.
3986                  *      first u_int32_t contains sets to be disabled,
3987                  *      second u_int32_t contains sets to be enabled.
3988                  */
3989                 error = sooptcopyin(sopt, rulenum,
3990                         2*sizeof(u_int32_t), sizeof(u_int32_t));
3991                 if (error)
3992                         break;
3993                 size = sopt->sopt_valsize;
3994                 if (size == sizeof(u_int32_t))  /* delete or reassign */
3995                         error = del_entry(&layer3_chain, rulenum[0]);
3996                 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */
3997                         set_disable =
3998                             (set_disable | rulenum[0]) & ~rulenum[1] &
3999                             ~(1<<RESVD_SET); /* set RESVD_SET always enabled */
4000                 else
4001                         error = EINVAL;
4002                 break;
4003
4004         case IP_FW_ZERO:
4005         case IP_FW_RESETLOG: /* argument is an int, the rule number */
4006                 rule_num = 0;
4007                 if (sopt->sopt_val != 0) {
4008                     error = sooptcopyin(sopt, &rule_num,
4009                             sizeof(int), sizeof(int));
4010                     if (error)
4011                         break;
4012                 }
4013                 error = zero_entry(&layer3_chain, rule_num,
4014                         sopt->sopt_name == IP_FW_RESETLOG);
4015                 break;
4016
4017         case IP_FW_TABLE_ADD:
4018                 {
4019                         ipfw_table_entry ent;
4020
4021                         error = sooptcopyin(sopt, &ent,
4022                             sizeof(ent), sizeof(ent));
4023                         if (error)
4024                                 break;
4025                         error = add_table_entry(&layer3_chain, ent.tbl,
4026                             ent.addr, ent.masklen, ent.value);
4027                 }
4028                 break;
4029
4030         case IP_FW_TABLE_DEL:
4031                 {
4032                         ipfw_table_entry ent;
4033
4034                         error = sooptcopyin(sopt, &ent,
4035                             sizeof(ent), sizeof(ent));
4036                         if (error)
4037                                 break;
4038                         error = del_table_entry(&layer3_chain, ent.tbl,
4039                             ent.addr, ent.masklen);
4040                 }
4041                 break;
4042
4043         case IP_FW_TABLE_FLUSH:
4044                 {
4045                         u_int16_t tbl;
4046
4047                         error = sooptcopyin(sopt, &tbl,
4048                             sizeof(tbl), sizeof(tbl));
4049                         if (error)
4050                                 break;
4051                         IPFW_WLOCK(&layer3_chain);
4052                         error = flush_table(&layer3_chain, tbl);
4053                         IPFW_WUNLOCK(&layer3_chain);
4054                 }
4055                 break;
4056
4057         case IP_FW_TABLE_GETSIZE:
4058                 {
4059                         u_int32_t tbl, cnt;
4060
4061                         if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl),
4062                             sizeof(tbl))))
4063                                 break;
4064                         IPFW_RLOCK(&layer3_chain);
4065                         error = count_table(&layer3_chain, tbl, &cnt);
4066                         IPFW_RUNLOCK(&layer3_chain);
4067                         if (error)
4068                                 break;
4069                         error = sooptcopyout(sopt, &cnt, sizeof(cnt));
4070                 }
4071                 break;
4072
4073         case IP_FW_TABLE_LIST:
4074                 {
4075                         ipfw_table *tbl;
4076
4077                         if (sopt->sopt_valsize < sizeof(*tbl)) {
4078                                 error = EINVAL;
4079                                 break;
4080                         }
4081                         size = sopt->sopt_valsize;
4082                         tbl = malloc(size, M_TEMP, M_WAITOK);
4083                         if (tbl == NULL) {
4084                                 error = ENOMEM;
4085                                 break;
4086                         }
4087                         error = sooptcopyin(sopt, tbl, size, sizeof(*tbl));
4088                         if (error) {
4089                                 free(tbl, M_TEMP);
4090                                 break;
4091                         }
4092                         tbl->size = (size - sizeof(*tbl)) /
4093                             sizeof(ipfw_table_entry);
4094                         IPFW_RLOCK(&layer3_chain);
4095                         error = dump_table(&layer3_chain, tbl);
4096                         IPFW_RUNLOCK(&layer3_chain);
4097                         if (error) {
4098                                 free(tbl, M_TEMP);
4099                                 break;
4100                         }
4101                         error = sooptcopyout(sopt, tbl, size);
4102                         free(tbl, M_TEMP);
4103                 }
4104                 break;
4105
4106         default:
4107                 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name);
4108                 error = EINVAL;
4109         }
4110
4111         return (error);
4112 #undef RULE_MAXSIZE
4113 }
4114
4115 /**
4116  * dummynet needs a reference to the default rule, because rules can be
4117  * deleted while packets hold a reference to them. When this happens,
4118  * dummynet changes the reference to the default rule (it could well be a
4119  * NULL pointer, but this way we do not need to check for the special
4120  * case, plus here he have info on the default behaviour).
4121  */
4122 struct ip_fw *ip_fw_default_rule;
4123
4124 /*
4125  * This procedure is only used to handle keepalives. It is invoked
4126  * every dyn_keepalive_period
4127  */
4128 static void
4129 ipfw_tick(void * __unused unused)
4130 {
4131         struct mbuf *m0, *m, *mnext, **mtailp;
4132         int i;
4133         ipfw_dyn_rule *q;
4134
4135         if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0)
4136                 goto done;
4137
4138         /*
4139          * We make a chain of packets to go out here -- not deferring
4140          * until after we drop the IPFW dynamic rule lock would result
4141          * in a lock order reversal with the normal packet input -> ipfw
4142          * call stack.
4143          */
4144         m0 = NULL;
4145         mtailp = &m0;
4146         IPFW_DYN_LOCK();
4147         for (i = 0 ; i < curr_dyn_buckets ; i++) {
4148                 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) {
4149                         if (q->dyn_type == O_LIMIT_PARENT)
4150                                 continue;
4151                         if (q->id.proto != IPPROTO_TCP)
4152                                 continue;
4153                         if ( (q->state & BOTH_SYN) != BOTH_SYN)
4154                                 continue;
4155                         if (TIME_LEQ( time_uptime+dyn_keepalive_interval,
4156                             q->expire))
4157                                 continue;       /* too early */
4158                         if (TIME_LEQ(q->expire, time_uptime))
4159                                 continue;       /* too late, rule expired */
4160
4161                         *mtailp = send_pkt(&(q->id), q->ack_rev - 1,
4162                                 q->ack_fwd, TH_SYN);
4163                         if (*mtailp != NULL)
4164                                 mtailp = &(*mtailp)->m_nextpkt;
4165                         *mtailp = send_pkt(&(q->id), q->ack_fwd - 1,
4166                                 q->ack_rev, 0);
4167                         if (*mtailp != NULL)
4168                                 mtailp = &(*mtailp)->m_nextpkt;
4169                 }
4170         }
4171         IPFW_DYN_UNLOCK();
4172         for (m = mnext = m0; m != NULL; m = mnext) {
4173                 mnext = m->m_nextpkt;
4174                 m->m_nextpkt = NULL;
4175                 ip_output(m, NULL, NULL, 0, NULL, NULL);
4176         }
4177 done:
4178         callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL);
4179 }
4180
4181 int
4182 ipfw_init(void)
4183 {
4184         struct ip_fw default_rule;
4185         int error;
4186
4187 #ifdef INET6
4188         /* Setup IPv6 fw sysctl tree. */
4189         sysctl_ctx_init(&ip6_fw_sysctl_ctx);
4190         ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx,
4191             SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw",
4192             CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall");
4193         SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree),
4194             OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3,
4195             &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6");
4196         SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree),
4197             OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE,
4198             &fw_deny_unknown_exthdrs, 0,
4199             "Deny packets with unknown IPv6 Extension Headers");
4200 #endif
4201
4202         layer3_chain.rules = NULL;
4203         IPFW_LOCK_INIT(&layer3_chain);
4204         ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule zone",
4205             sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL,
4206             UMA_ALIGN_PTR, 0);
4207         IPFW_DYN_LOCK_INIT();
4208         callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE);
4209
4210         bzero(&default_rule, sizeof default_rule);
4211
4212         default_rule.act_ofs = 0;
4213         default_rule.rulenum = IPFW_DEFAULT_RULE;
4214         default_rule.cmd_len = 1;
4215         default_rule.set = RESVD_SET;
4216
4217         default_rule.cmd[0].len = 1;
4218         default_rule.cmd[0].opcode =
4219 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
4220                                 1 ? O_ACCEPT :
4221 #endif
4222                                 O_DENY;
4223
4224         error = add_rule(&layer3_chain, &default_rule);
4225         if (error != 0) {
4226                 printf("ipfw2: error %u initializing default rule "
4227                         "(support disabled)\n", error);
4228                 IPFW_DYN_LOCK_DESTROY();
4229                 IPFW_LOCK_DESTROY(&layer3_chain);
4230                 uma_zdestroy(ipfw_dyn_rule_zone);
4231                 return (error);
4232         }
4233
4234         ip_fw_default_rule = layer3_chain.rules;
4235         printf("ipfw2 "
4236 #ifdef INET6
4237                 "(+ipv6) "
4238 #endif
4239                 "initialized, divert %s, "
4240                 "rule-based forwarding "
4241 #ifdef IPFIREWALL_FORWARD
4242                 "enabled, "
4243 #else
4244                 "disabled, "
4245 #endif
4246                 "default to %s, logging ",
4247 #ifdef IPDIVERT
4248                 "enabled",
4249 #else
4250                 "loadable",
4251 #endif
4252                 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny");
4253
4254 #ifdef IPFIREWALL_VERBOSE
4255         fw_verbose = 1;
4256 #endif
4257 #ifdef IPFIREWALL_VERBOSE_LIMIT
4258         verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
4259 #endif
4260         if (fw_verbose == 0)
4261                 printf("disabled\n");
4262         else if (verbose_limit == 0)
4263                 printf("unlimited\n");
4264         else
4265                 printf("limited to %d packets/entry by default\n",
4266                     verbose_limit);
4267
4268         error = init_tables(&layer3_chain);
4269         if (error) {
4270                 IPFW_DYN_LOCK_DESTROY();
4271                 IPFW_LOCK_DESTROY(&layer3_chain);
4272                 uma_zdestroy(ipfw_dyn_rule_zone);
4273                 return (error);
4274         }
4275         ip_fw_ctl_ptr = ipfw_ctl;
4276         ip_fw_chk_ptr = ipfw_chk;
4277         callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL);
4278
4279         return (0);
4280 }
4281
4282 void
4283 ipfw_destroy(void)
4284 {
4285         struct ip_fw *reap;
4286
4287         ip_fw_chk_ptr = NULL;
4288         ip_fw_ctl_ptr = NULL;
4289         callout_drain(&ipfw_timeout);
4290         IPFW_WLOCK(&layer3_chain);
4291         flush_tables(&layer3_chain);
4292         layer3_chain.reap = NULL;
4293         free_chain(&layer3_chain, 1 /* kill default rule */);
4294         reap = layer3_chain.reap, layer3_chain.reap = NULL;
4295         IPFW_WUNLOCK(&layer3_chain);
4296         if (reap != NULL)
4297                 reap_rules(reap);
4298         IPFW_DYN_LOCK_DESTROY();
4299         uma_zdestroy(ipfw_dyn_rule_zone);
4300         IPFW_LOCK_DESTROY(&layer3_chain);
4301
4302 #ifdef INET6
4303         /* Free IPv6 fw sysctl tree. */
4304         sysctl_ctx_free(&ip6_fw_sysctl_ctx);
4305 #endif
4306
4307         printf("IP firewall unloaded\n");
4308 }