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_ipdivert.h"
  38 #include "opt_ipdn.h"
  39 #include "opt_inet.h"
  40 #ifndef INET
  41 #error IPFIREWALL requires INET.
  42 #endif /* INET */
  43 #endif
  44 #include "opt_inet6.h"
  45 #include "opt_ipsec.h"
  46 #include "opt_mac.h"
  47
  48 #include <sys/param.h>
  49 #include <sys/systm.h>
  50 #include <sys/condvar.h>
  51 #include <sys/eventhandler.h>
  52 #include <sys/malloc.h>
  53 #include <sys/mbuf.h>
  54 #include <sys/kernel.h>
  55 #include <sys/lock.h>
  56 #include <sys/jail.h>
  57 #include <sys/module.h>
  58 #include <sys/priv.h>
  59 #include <sys/proc.h>
  60 #include <sys/rwlock.h>
  61 #include <sys/socket.h>
  62 #include <sys/socketvar.h>
  63 #include <sys/sysctl.h>
  64 #include <sys/syslog.h>
  65 #include <sys/ucred.h>
  66 #include <net/if.h>
  67 #include <net/radix.h>
  68 #include <net/route.h>
  69 #include <netinet/in.h>
  70 #include <netinet/in_systm.h>
  71 #include <netinet/in_var.h>
  72 #include <netinet/in_pcb.h>
  73 #include <netinet/ip.h>
  74 #include <netinet/ip_var.h>
  75 #include <netinet/ip_icmp.h>
  76 #include <netinet/ip_fw.h>
  77 #include <netinet/ip_divert.h>
  78 #include <netinet/ip_dummynet.h>
  79 #include <netinet/ip_carp.h>
  80 #include <netinet/pim.h>
  81 #include <netinet/tcp.h>
  82 #include <netinet/tcp_timer.h>
  83 #include <netinet/tcp_var.h>
  84 #include <netinet/tcpip.h>
  85 #include <netinet/udp.h>
  86 #include <netinet/udp_var.h>
  87 #include <netinet/sctp.h>
  88 #ifdef IPFIREWALL_NAT
  89 #include <netinet/libalias/alias.h>
  90 #include <netinet/libalias/alias_local.h>
  91 #endif
  92 #include <netgraph/ng_ipfw.h>
  93
  94 #include <altq/if_altq.h>
  95
  96 #include <netinet/ip6.h>
  97 #include <netinet/icmp6.h>
  98 #ifdef INET6
  99 #include <netinet6/scope6_var.h>
 100 #endif
 101
 102 #include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */
 103
 104 #include <machine/in_cksum.h>   /* XXX for in_cksum */
 105
 106 #include <security/mac/mac_framework.h>
 107
 108 /*
 109  * set_disable contains one bit per set value (0..31).
 110  * If the bit is set, all rules with the corresponding set
 111  * are disabled. Set RESVD_SET(31) is reserved for the default rule
 112  * and rules that are not deleted by the flush command,
 113  * and CANNOT be disabled.
 114  * Rules in set RESVD_SET can only be deleted explicitly.
 115  */
 116 static u_int32_t set_disable;
 117
 118 static int fw_verbose;
 119 static int verbose_limit;
 120
 121 static struct callout ipfw_timeout;
 122 static uma_zone_t ipfw_dyn_rule_zone;
 123 #define IPFW_DEFAULT_RULE       65535
 124
 125 /*
 126  * Data structure to cache our ucred related
 127  * information. This structure only gets used if
 128  * the user specified UID/GID based constraints in
 129  * a firewall rule.
 130  */
 131 struct ip_fw_ugid {
 132         gid_t           fw_groups[NGROUPS];
 133         int             fw_ngroups;
 134         uid_t           fw_uid;
 135         int             fw_prid;
 136 };
 137
 138 #define IPFW_TABLES_MAX         128
 139 struct ip_fw_chain {
 140         struct ip_fw    *rules;         /* list of rules */
 141         struct ip_fw    *reap;          /* list of rules to reap */
 142         LIST_HEAD(, cfg_nat) nat;       /* list of nat entries */
 143         struct radix_node_head *tables[IPFW_TABLES_MAX];
 144         struct rwlock   rwmtx;
 145 };
 146 #define IPFW_LOCK_INIT(_chain) \
 147         rw_init(&(_chain)->rwmtx, "IPFW static rules")
 148 #define IPFW_LOCK_DESTROY(_chain)       rw_destroy(&(_chain)->rwmtx)
 149 #define IPFW_WLOCK_ASSERT(_chain)       do {                            \
 150         rw_assert(&(_chain)->rwmtx, RA_WLOCKED);                                        \
 151         NET_ASSERT_GIANT();                                             \
 152 } while (0)
 153
 154 #define IPFW_RLOCK(p) rw_rlock(&(p)->rwmtx)
 155 #define IPFW_RUNLOCK(p) rw_runlock(&(p)->rwmtx)
 156 #define IPFW_WLOCK(p) rw_wlock(&(p)->rwmtx)
 157 #define IPFW_WUNLOCK(p) rw_wunlock(&(p)->rwmtx)
 158
 159 /*
 160  * list of rules for layer 3
 161  */
 162 static struct ip_fw_chain layer3_chain;
 163
 164 MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's");
 165 MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables");
 166
 167 struct table_entry {
 168         struct radix_node       rn[2];
 169         struct sockaddr_in      addr, mask;
 170         u_int32_t               value;
 171 };
 172
 173 static int fw_debug = 1;
 174 static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */
 175
 176 extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS);
 177
 178 #ifdef SYSCTL_NODE
 179 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 180 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable,
 181     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &fw_enable, 0,
 182     ipfw_chg_hook, "I", "Enable ipfw");
 183 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW,
 184     &autoinc_step, 0, "Rule number autincrement step");
 185 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
 186     CTLFLAG_RW | CTLFLAG_SECURE3,
 187     &fw_one_pass, 0,
 188     "Only do a single pass through ipfw when using dummynet(4)");
 189 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW,
 190     &fw_debug, 0, "Enable printing of debug ip_fw statements");
 191 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose,
 192     CTLFLAG_RW | CTLFLAG_SECURE3,
 193     &fw_verbose, 0, "Log matches to ipfw rules");
 194 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW,
 195     &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged");
 196
 197 /*
 198  * Description of dynamic rules.
 199  *
 200  * Dynamic rules are stored in lists accessed through a hash table
 201  * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can
 202  * be modified through the sysctl variable dyn_buckets which is
 203  * updated when the table becomes empty.
 204  *
 205  * XXX currently there is only one list, ipfw_dyn.
 206  *
 207  * When a packet is received, its address fields are first masked
 208  * with the mask defined for the rule, then hashed, then matched
 209  * against the entries in the corresponding list.
 210  * Dynamic rules can be used for different purposes:
 211  *  + stateful rules;
 212  *  + enforcing limits on the number of sessions;
 213  *  + in-kernel NAT (not implemented yet)
 214  *
 215  * The lifetime of dynamic rules is regulated by dyn_*_lifetime,
 216  * measured in seconds and depending on the flags.
 217  *
 218  * The total number of dynamic rules is stored in dyn_count.
 219  * The max number of dynamic rules is dyn_max. When we reach
 220  * the maximum number of rules we do not create anymore. This is
 221  * done to avoid consuming too much memory, but also too much
 222  * time when searching on each packet (ideally, we should try instead
 223  * to put a limit on the length of the list on each bucket...).
 224  *
 225  * Each dynamic rule holds a pointer to the parent ipfw rule so
 226  * we know what action to perform. Dynamic rules are removed when
 227  * the parent rule is deleted. XXX we should make them survive.
 228  *
 229  * There are some limitations with dynamic rules -- we do not
 230  * obey the 'randomized match', and we do not do multiple
 231  * passes through the firewall. XXX check the latter!!!
 232  */
 233 static ipfw_dyn_rule **ipfw_dyn_v = NULL;
 234 static u_int32_t dyn_buckets = 256; /* must be power of 2 */
 235 static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */
 236
 237 static struct mtx ipfw_dyn_mtx;         /* mutex guarding dynamic rules */
 238 #define IPFW_DYN_LOCK_INIT() \
 239         mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF)
 240 #define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx)
 241 #define IPFW_DYN_LOCK()         mtx_lock(&ipfw_dyn_mtx)
 242 #define IPFW_DYN_UNLOCK()       mtx_unlock(&ipfw_dyn_mtx)
 243 #define IPFW_DYN_LOCK_ASSERT()  mtx_assert(&ipfw_dyn_mtx, MA_OWNED)
 244
 245 /*
 246  * Timeouts for various events in handing dynamic rules.
 247  */
 248 static u_int32_t dyn_ack_lifetime = 300;
 249 static u_int32_t dyn_syn_lifetime = 20;
 250 static u_int32_t dyn_fin_lifetime = 1;
 251 static u_int32_t dyn_rst_lifetime = 1;
 252 static u_int32_t dyn_udp_lifetime = 10;
 253 static u_int32_t dyn_short_lifetime = 5;
 254
 255 /*
 256  * Keepalives are sent if dyn_keepalive is set. They are sent every
 257  * dyn_keepalive_period seconds, in the last dyn_keepalive_interval
 258  * seconds of lifetime of a rule.
 259  * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower
 260  * than dyn_keepalive_period.
 261  */
 262
 263 static u_int32_t dyn_keepalive_interval = 20;
 264 static u_int32_t dyn_keepalive_period = 5;
 265 static u_int32_t dyn_keepalive = 1;     /* do send keepalives */
 266
 267 static u_int32_t static_count;  /* # of static rules */
 268 static u_int32_t static_len;    /* size in bytes of static rules */
 269 static u_int32_t dyn_count;             /* # of dynamic rules */
 270 static u_int32_t dyn_max = 4096;        /* max # of dynamic rules */
 271
 272 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW,
 273     &dyn_buckets, 0, "Number of dyn. buckets");
 274 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD,
 275     &curr_dyn_buckets, 0, "Current Number of dyn. buckets");
 276 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD,
 277     &dyn_count, 0, "Number of dyn. rules");
 278 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW,
 279     &dyn_max, 0, "Max number of dyn. rules");
 280 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD,
 281     &static_count, 0, "Number of static rules");
 282 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW,
 283     &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks");
 284 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW,
 285     &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn");
 286 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW,
 287     &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin");
 288 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW,
 289     &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst");
 290 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW,
 291     &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP");
 292 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW,
 293     &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations");
 294 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW,
 295     &dyn_keepalive, 0, "Enable keepalives for dyn. rules");
 296
 297 #ifdef INET6
 298 /*
 299  * IPv6 specific variables
 300  */
 301 SYSCTL_DECL(_net_inet6_ip6);
 302
 303 static struct sysctl_ctx_list ip6_fw_sysctl_ctx;
 304 static struct sysctl_oid *ip6_fw_sysctl_tree;
 305 #endif /* INET6 */
 306 #endif /* SYSCTL_NODE */
 307
 308 #ifdef IPFIREWALL_NAT
 309 MODULE_DEPEND(ipfw, libalias, 1, 1, 1);
 310 #endif
 311 static int fw_deny_unknown_exthdrs = 1;
 312
 313
 314 /*
 315  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
 316  * Other macros just cast void * into the appropriate type
 317  */
 318 #define L3HDR(T, ip)    ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
 319 #define TCP(p)          ((struct tcphdr *)(p))
 320 #define SCTP(p)         ((struct sctphdr *)(p))
 321 #define UDP(p)          ((struct udphdr *)(p))
 322 #define ICMP(p)         ((struct icmphdr *)(p))
 323 #define ICMP6(p)        ((struct icmp6_hdr *)(p))
 324
 325 static __inline int
 326 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
 327 {
 328         int type = icmp->icmp_type;
 329
 330         return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
 331 }
 332
 333 #define TT      ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
 334     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
 335
 336 static int
 337 is_icmp_query(struct icmphdr *icmp)
 338 {
 339         int type = icmp->icmp_type;
 340
 341         return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
 342 }
 343 #undef TT
 344
 345 /*
 346  * The following checks use two arrays of 8 or 16 bits to store the
 347  * bits that we want set or clear, respectively. They are in the
 348  * low and high half of cmd->arg1 or cmd->d[0].
 349  *
 350  * We scan options and store the bits we find set. We succeed if
 351  *
 352  *      (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
 353  *
 354  * The code is sometimes optimized not to store additional variables.
 355  */
 356
 357 static int
 358 flags_match(ipfw_insn *cmd, u_int8_t bits)
 359 {
 360         u_char want_clear;
 361         bits = ~bits;
 362
 363         if ( ((cmd->arg1 & 0xff) & bits) != 0)
 364                 return 0; /* some bits we want set were clear */
 365         want_clear = (cmd->arg1 >> 8) & 0xff;
 366         if ( (want_clear & bits) != want_clear)
 367                 return 0; /* some bits we want clear were set */
 368         return 1;
 369 }
 370
 371 static int
 372 ipopts_match(struct ip *ip, ipfw_insn *cmd)
 373 {
 374         int optlen, bits = 0;
 375         u_char *cp = (u_char *)(ip + 1);
 376         int x = (ip->ip_hl << 2) - sizeof (struct ip);
 377
 378         for (; x > 0; x -= optlen, cp += optlen) {
 379                 int opt = cp[IPOPT_OPTVAL];
 380
 381                 if (opt == IPOPT_EOL)
 382                         break;
 383                 if (opt == IPOPT_NOP)
 384                         optlen = 1;
 385                 else {
 386                         optlen = cp[IPOPT_OLEN];
 387                         if (optlen <= 0 || optlen > x)
 388                                 return 0; /* invalid or truncated */
 389                 }
 390                 switch (opt) {
 391
 392                 default:
 393                         break;
 394
 395                 case IPOPT_LSRR:
 396                         bits |= IP_FW_IPOPT_LSRR;
 397                         break;
 398
 399                 case IPOPT_SSRR:
 400                         bits |= IP_FW_IPOPT_SSRR;
 401                         break;
 402
 403                 case IPOPT_RR:
 404                         bits |= IP_FW_IPOPT_RR;
 405                         break;
 406
 407                 case IPOPT_TS:
 408                         bits |= IP_FW_IPOPT_TS;
 409                         break;
 410                 }
 411         }
 412         return (flags_match(cmd, bits));
 413 }
 414
 415 static int
 416 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
 417 {
 418         int optlen, bits = 0;
 419         u_char *cp = (u_char *)(tcp + 1);
 420         int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
 421
 422         for (; x > 0; x -= optlen, cp += optlen) {
 423                 int opt = cp[0];
 424                 if (opt == TCPOPT_EOL)
 425                         break;
 426                 if (opt == TCPOPT_NOP)
 427                         optlen = 1;
 428                 else {
 429                         optlen = cp[1];
 430                         if (optlen <= 0)
 431                                 break;
 432                 }
 433
 434                 switch (opt) {
 435
 436                 default:
 437                         break;
 438
 439                 case TCPOPT_MAXSEG:
 440                         bits |= IP_FW_TCPOPT_MSS;
 441                         break;
 442
 443                 case TCPOPT_WINDOW:
 444                         bits |= IP_FW_TCPOPT_WINDOW;
 445                         break;
 446
 447                 case TCPOPT_SACK_PERMITTED:
 448                 case TCPOPT_SACK:
 449                         bits |= IP_FW_TCPOPT_SACK;
 450                         break;
 451
 452                 case TCPOPT_TIMESTAMP:
 453                         bits |= IP_FW_TCPOPT_TS;
 454                         break;
 455
 456                 }
 457         }
 458         return (flags_match(cmd, bits));
 459 }
 460
 461 static int
 462 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd)
 463 {
 464         if (ifp == NULL)        /* no iface with this packet, match fails */
 465                 return 0;
 466         /* Check by name or by IP address */
 467         if (cmd->name[0] != '\0') { /* match by name */
 468                 /* Check name */
 469                 if (cmd->p.glob) {
 470                         if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
 471                                 return(1);
 472                 } else {
 473                         if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
 474                                 return(1);
 475                 }
 476         } else {
 477                 struct ifaddr *ia;
 478
 479                 /* XXX lock? */
 480                 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
 481                         if (ia->ifa_addr->sa_family != AF_INET)
 482                                 continue;
 483                         if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
 484                             (ia->ifa_addr))->sin_addr.s_addr)
 485                                 return(1);      /* match */
 486                 }
 487         }
 488         return(0);      /* no match, fail ... */
 489 }
 490
 491 /*
 492  * The verify_path function checks if a route to the src exists and
 493  * if it is reachable via ifp (when provided).
 494  *
 495  * The 'verrevpath' option checks that the interface that an IP packet
 496  * arrives on is the same interface that traffic destined for the
 497  * packet's source address would be routed out of.  The 'versrcreach'
 498  * option just checks that the source address is reachable via any route
 499  * (except default) in the routing table.  These two are a measure to block
 500  * forged packets.  This is also commonly known as "anti-spoofing" or Unicast
 501  * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
 502  * is purposely reminiscent of the Cisco IOS command,
 503  *
 504  *   ip verify unicast reverse-path
 505  *   ip verify unicast source reachable-via any
 506  *
 507  * which implements the same functionality. But note that syntax is
 508  * misleading. The check may be performed on all IP packets whether unicast,
 509  * multicast, or broadcast.
 510  */
 511 static int
 512 verify_path(struct in_addr src, struct ifnet *ifp)
 513 {
 514         struct route ro;
 515         struct sockaddr_in *dst;
 516
 517         bzero(&ro, sizeof(ro));
 518
 519         dst = (struct sockaddr_in *)&(ro.ro_dst);
 520         dst->sin_family = AF_INET;
 521         dst->sin_len = sizeof(*dst);
 522         dst->sin_addr = src;
 523         rtalloc_ign(&ro, RTF_CLONING);
 524
 525         if (ro.ro_rt == NULL)
 526                 return 0;
 527
 528         /*
 529          * If ifp is provided, check for equality with rtentry.
 530          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 531          * in order to pass packets injected back by if_simloop():
 532          * if useloopback == 1 routing entry (via lo0) for our own address
 533          * may exist, so we need to handle routing assymetry.
 534          */
 535         if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
 536                 RTFREE(ro.ro_rt);
 537                 return 0;
 538         }
 539
 540         /* if no ifp provided, check if rtentry is not default route */
 541         if (ifp == NULL &&
 542              satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) {
 543                 RTFREE(ro.ro_rt);
 544                 return 0;
 545         }
 546
 547         /* or if this is a blackhole/reject route */
 548         if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
 549                 RTFREE(ro.ro_rt);
 550                 return 0;
 551         }
 552
 553         /* found valid route */
 554         RTFREE(ro.ro_rt);
 555         return 1;
 556 }
 557
 558 #ifdef INET6
 559 /*
 560  * ipv6 specific rules here...
 561  */
 562 static __inline int
 563 icmp6type_match (int type, ipfw_insn_u32 *cmd)
 564 {
 565         return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
 566 }
 567
 568 static int
 569 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
 570 {
 571         int i;
 572         for (i=0; i <= cmd->o.arg1; ++i )
 573                 if (curr_flow == cmd->d[i] )
 574                         return 1;
 575         return 0;
 576 }
 577
 578 /* support for IP6_*_ME opcodes */
 579 static int
 580 search_ip6_addr_net (struct in6_addr * ip6_addr)
 581 {
 582         struct ifnet *mdc;
 583         struct ifaddr *mdc2;
 584         struct in6_ifaddr *fdm;
 585         struct in6_addr copia;
 586
 587         TAILQ_FOREACH(mdc, &ifnet, if_link)
 588                 TAILQ_FOREACH(mdc2, &mdc->if_addrlist, ifa_list) {
 589                         if (mdc2->ifa_addr->sa_family == AF_INET6) {
 590                                 fdm = (struct in6_ifaddr *)mdc2;
 591                                 copia = fdm->ia_addr.sin6_addr;
 592                                 /* need for leaving scope_id in the sock_addr */
 593                                 in6_clearscope(&copia);
 594                                 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia))
 595                                         return 1;
 596                         }
 597                 }
 598         return 0;
 599 }
 600
 601 static int
 602 verify_path6(struct in6_addr *src, struct ifnet *ifp)
 603 {
 604         struct route_in6 ro;
 605         struct sockaddr_in6 *dst;
 606
 607         bzero(&ro, sizeof(ro));
 608
 609         dst = (struct sockaddr_in6 * )&(ro.ro_dst);
 610         dst->sin6_family = AF_INET6;
 611         dst->sin6_len = sizeof(*dst);
 612         dst->sin6_addr = *src;
 613         rtalloc_ign((struct route *)&ro, RTF_CLONING);
 614
 615         if (ro.ro_rt == NULL)
 616                 return 0;
 617
 618         /*
 619          * if ifp is provided, check for equality with rtentry
 620          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 621          * to support the case of sending packets to an address of our own.
 622          * (where the former interface is the first argument of if_simloop()
 623          *  (=ifp), the latter is lo0)
 624          */
 625         if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
 626                 RTFREE(ro.ro_rt);
 627                 return 0;
 628         }
 629
 630         /* if no ifp provided, check if rtentry is not default route */
 631         if (ifp == NULL &&
 632             IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) {
 633                 RTFREE(ro.ro_rt);
 634                 return 0;
 635         }
 636
 637         /* or if this is a blackhole/reject route */
 638         if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
 639                 RTFREE(ro.ro_rt);
 640                 return 0;
 641         }
 642
 643         /* found valid route */
 644         RTFREE(ro.ro_rt);
 645         return 1;
 646
 647 }
 648 static __inline int
 649 hash_packet6(struct ipfw_flow_id *id)
 650 {
 651         u_int32_t i;
 652         i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^
 653             (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^
 654             (id->src_ip6.__u6_addr.__u6_addr32[2]) ^
 655             (id->src_ip6.__u6_addr.__u6_addr32[3]) ^
 656             (id->dst_port) ^ (id->src_port);
 657         return i;
 658 }
 659
 660 static int
 661 is_icmp6_query(int icmp6_type)
 662 {
 663         if ((icmp6_type <= ICMP6_MAXTYPE) &&
 664             (icmp6_type == ICMP6_ECHO_REQUEST ||
 665             icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
 666             icmp6_type == ICMP6_WRUREQUEST ||
 667             icmp6_type == ICMP6_FQDN_QUERY ||
 668             icmp6_type == ICMP6_NI_QUERY))
 669                 return (1);
 670
 671         return (0);
 672 }
 673
 674 static void
 675 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
 676 {
 677         struct mbuf *m;
 678
 679         m = args->m;
 680         if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
 681                 struct tcphdr *tcp;
 682                 tcp_seq ack, seq;
 683                 int flags;
 684                 struct {
 685                         struct ip6_hdr ip6;
 686                         struct tcphdr th;
 687                 } ti;
 688                 tcp = (struct tcphdr *)((char *)ip6 + hlen);
 689
 690                 if ((tcp->th_flags & TH_RST) != 0) {
 691                         m_freem(m);
 692                         args->m = NULL;
 693                         return;
 694                 }
 695
 696                 ti.ip6 = *ip6;
 697                 ti.th = *tcp;
 698                 ti.th.th_seq = ntohl(ti.th.th_seq);
 699                 ti.th.th_ack = ntohl(ti.th.th_ack);
 700                 ti.ip6.ip6_nxt = IPPROTO_TCP;
 701
 702                 if (ti.th.th_flags & TH_ACK) {
 703                         ack = 0;
 704                         seq = ti.th.th_ack;
 705                         flags = TH_RST;
 706                 } else {
 707                         ack = ti.th.th_seq;
 708                         if ((m->m_flags & M_PKTHDR) != 0) {
 709                                 /*
 710                                  * total new data to ACK is:
 711                                  * total packet length,
 712                                  * minus the header length,
 713                                  * minus the tcp header length.
 714                                  */
 715                                 ack += m->m_pkthdr.len - hlen
 716                                         - (ti.th.th_off << 2);
 717                         } else if (ip6->ip6_plen) {
 718                                 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) -
 719                                     hlen - (ti.th.th_off << 2);
 720                         } else {
 721                                 m_freem(m);
 722                                 return;
 723                         }
 724                         if (tcp->th_flags & TH_SYN)
 725                                 ack++;
 726                         seq = 0;
 727                         flags = TH_RST|TH_ACK;
 728                 }
 729                 bcopy(&ti, ip6, sizeof(ti));
 730                 /*
 731                  * m is only used to recycle the mbuf
 732                  * The data in it is never read so we don't need
 733                  * to correct the offsets or anything
 734                  */
 735                 tcp_respond(NULL, ip6, tcp, m, ack, seq, flags);
 736         } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
 737 #if 0
 738                 /*
 739                  * Unlike above, the mbufs need to line up with the ip6 hdr,
 740                  * as the contents are read. We need to m_adj() the
 741                  * needed amount.
 742                  * The mbuf will however be thrown away so we can adjust it.
 743                  * Remember we did an m_pullup on it already so we
 744                  * can make some assumptions about contiguousness.
 745                  */
 746                 if (args->L3offset)
 747                         m_adj(m, args->L3offset);
 748 #endif
 749                 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
 750         } else
 751                 m_freem(m);
 752
 753         args->m = NULL;
 754 }
 755
 756 #endif /* INET6 */
 757
 758 static u_int64_t norule_counter;        /* counter for ipfw_log(NULL...) */
 759
 760 #define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0
 761 #define SNP(buf) buf, sizeof(buf)
 762
 763 /*
 764  * We enter here when we have a rule with O_LOG.
 765  * XXX this function alone takes about 2Kbytes of code!
 766  */
 767 static void
 768 ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args,
 769     struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg,
 770     struct ip *ip)
 771 {
 772         struct ether_header *eh = args->eh;
 773         char *action;
 774         int limit_reached = 0;
 775         char action2[40], proto[128], fragment[32];
 776
 777         fragment[0] = '\0';
 778         proto[0] = '\0';
 779
 780         if (f == NULL) {        /* bogus pkt */
 781                 if (verbose_limit != 0 && norule_counter >= verbose_limit)
 782                         return;
 783                 norule_counter++;
 784                 if (norule_counter == verbose_limit)
 785                         limit_reached = verbose_limit;
 786                 action = "Refuse";
 787         } else {        /* O_LOG is the first action, find the real one */
 788                 ipfw_insn *cmd = ACTION_PTR(f);
 789                 ipfw_insn_log *l = (ipfw_insn_log *)cmd;
 790
 791                 if (l->max_log != 0 && l->log_left == 0)
 792                         return;
 793                 l->log_left--;
 794                 if (l->log_left == 0)
 795                         limit_reached = l->max_log;
 796                 cmd += F_LEN(cmd);      /* point to first action */
 797                 if (cmd->opcode == O_ALTQ) {
 798                         ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
 799
 800                         snprintf(SNPARGS(action2, 0), "Altq %d",
 801                                 altq->qid);
 802                         cmd += F_LEN(cmd);
 803                 }
 804                 if (cmd->opcode == O_PROB)
 805                         cmd += F_LEN(cmd);
 806
 807                 if (cmd->opcode == O_TAG)
 808                         cmd += F_LEN(cmd);
 809
 810                 action = action2;
 811                 switch (cmd->opcode) {
 812                 case O_DENY:
 813                         action = "Deny";
 814                         break;
 815
 816                 case O_REJECT:
 817                         if (cmd->arg1==ICMP_REJECT_RST)
 818                                 action = "Reset";
 819                         else if (cmd->arg1==ICMP_UNREACH_HOST)
 820                                 action = "Reject";
 821                         else
 822                                 snprintf(SNPARGS(action2, 0), "Unreach %d",
 823                                         cmd->arg1);
 824                         break;
 825
 826                 case O_UNREACH6:
 827                         if (cmd->arg1==ICMP6_UNREACH_RST)
 828                                 action = "Reset";
 829                         else
 830                                 snprintf(SNPARGS(action2, 0), "Unreach %d",
 831                                         cmd->arg1);
 832                         break;
 833
 834                 case O_ACCEPT:
 835                         action = "Accept";
 836                         break;
 837                 case O_COUNT:
 838                         action = "Count";
 839                         break;
 840                 case O_DIVERT:
 841                         snprintf(SNPARGS(action2, 0), "Divert %d",
 842                                 cmd->arg1);
 843                         break;
 844                 case O_TEE:
 845                         snprintf(SNPARGS(action2, 0), "Tee %d",
 846                                 cmd->arg1);
 847                         break;
 848                 case O_SKIPTO:
 849                         snprintf(SNPARGS(action2, 0), "SkipTo %d",
 850                                 cmd->arg1);
 851                         break;
 852                 case O_PIPE:
 853                         snprintf(SNPARGS(action2, 0), "Pipe %d",
 854                                 cmd->arg1);
 855                         break;
 856                 case O_QUEUE:
 857                         snprintf(SNPARGS(action2, 0), "Queue %d",
 858                                 cmd->arg1);
 859                         break;
 860                 case O_FORWARD_IP: {
 861                         ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd;
 862                         int len;
 863                         struct in_addr dummyaddr;
 864                         if (sa->sa.sin_addr.s_addr == INADDR_ANY)
 865                                 dummyaddr.s_addr = htonl(tablearg);
 866                         else
 867                                 dummyaddr.s_addr = sa->sa.sin_addr.s_addr;
 868
 869                         len = snprintf(SNPARGS(action2, 0), "Forward to %s",
 870                                 inet_ntoa(dummyaddr));
 871
 872                         if (sa->sa.sin_port)
 873                                 snprintf(SNPARGS(action2, len), ":%d",
 874                                     sa->sa.sin_port);
 875                         }
 876                         break;
 877                 case O_NETGRAPH:
 878                         snprintf(SNPARGS(action2, 0), "Netgraph %d",
 879                                 cmd->arg1);
 880                         break;
 881                 case O_NGTEE:
 882                         snprintf(SNPARGS(action2, 0), "Ngtee %d",
 883                                 cmd->arg1);
 884                         break;
 885                 case O_NAT:
 886                         action = "Nat";
 887                         break;
 888                 default:
 889                         action = "UNKNOWN";
 890                         break;
 891                 }
 892         }
 893
 894         if (hlen == 0) {        /* non-ip */
 895                 snprintf(SNPARGS(proto, 0), "MAC");
 896
 897         } else {
 898                 int len;
 899                 char src[48], dst[48];
 900                 struct icmphdr *icmp;
 901                 struct tcphdr *tcp;
 902                 struct udphdr *udp;
 903 #ifdef INET6
 904                 struct ip6_hdr *ip6 = NULL;
 905                 struct icmp6_hdr *icmp6;
 906 #endif
 907                 src[0] = '\0';
 908                 dst[0] = '\0';
 909 #ifdef INET6
 910                 if (IS_IP6_FLOW_ID(&(args->f_id))) {
 911                         char ip6buf[INET6_ADDRSTRLEN];
 912                         snprintf(src, sizeof(src), "[%s]",
 913                             ip6_sprintf(ip6buf, &args->f_id.src_ip6));
 914                         snprintf(dst, sizeof(dst), "[%s]",
 915                             ip6_sprintf(ip6buf, &args->f_id.dst_ip6));
 916
 917                         ip6 = (struct ip6_hdr *)ip;
 918                         tcp = (struct tcphdr *)(((char *)ip) + hlen);
 919                         udp = (struct udphdr *)(((char *)ip) + hlen);
 920                 } else
 921 #endif
 922                 {
 923                         tcp = L3HDR(struct tcphdr, ip);
 924                         udp = L3HDR(struct udphdr, ip);
 925
 926                         inet_ntoa_r(ip->ip_src, src);
 927                         inet_ntoa_r(ip->ip_dst, dst);
 928                 }
 929
 930                 switch (args->f_id.proto) {
 931                 case IPPROTO_TCP:
 932                         len = snprintf(SNPARGS(proto, 0), "TCP %s", src);
 933                         if (offset == 0)
 934                                 snprintf(SNPARGS(proto, len), ":%d %s:%d",
 935                                     ntohs(tcp->th_sport),
 936                                     dst,
 937                                     ntohs(tcp->th_dport));
 938                         else
 939                                 snprintf(SNPARGS(proto, len), " %s", dst);
 940                         break;
 941
 942                 case IPPROTO_UDP:
 943                         len = snprintf(SNPARGS(proto, 0), "UDP %s", src);
 944                         if (offset == 0)
 945                                 snprintf(SNPARGS(proto, len), ":%d %s:%d",
 946                                     ntohs(udp->uh_sport),
 947                                     dst,
 948                                     ntohs(udp->uh_dport));
 949                         else
 950                                 snprintf(SNPARGS(proto, len), " %s", dst);
 951                         break;
 952
 953                 case IPPROTO_ICMP:
 954                         icmp = L3HDR(struct icmphdr, ip);
 955                         if (offset == 0)
 956                                 len = snprintf(SNPARGS(proto, 0),
 957                                     "ICMP:%u.%u ",
 958                                     icmp->icmp_type, icmp->icmp_code);
 959                         else
 960                                 len = snprintf(SNPARGS(proto, 0), "ICMP ");
 961                         len += snprintf(SNPARGS(proto, len), "%s", src);
 962                         snprintf(SNPARGS(proto, len), " %s", dst);
 963                         break;
 964 #ifdef INET6
 965                 case IPPROTO_ICMPV6:
 966                         icmp6 = (struct icmp6_hdr *)(((char *)ip) + hlen);
 967                         if (offset == 0)
 968                                 len = snprintf(SNPARGS(proto, 0),
 969                                     "ICMPv6:%u.%u ",
 970                                     icmp6->icmp6_type, icmp6->icmp6_code);
 971                         else
 972                                 len = snprintf(SNPARGS(proto, 0), "ICMPv6 ");
 973                         len += snprintf(SNPARGS(proto, len), "%s", src);
 974                         snprintf(SNPARGS(proto, len), " %s", dst);
 975                         break;
 976 #endif
 977                 default:
 978                         len = snprintf(SNPARGS(proto, 0), "P:%d %s",
 979                             args->f_id.proto, src);
 980                         snprintf(SNPARGS(proto, len), " %s", dst);
 981                         break;
 982                 }
 983
 984 #ifdef INET6
 985                 if (IS_IP6_FLOW_ID(&(args->f_id))) {
 986                         if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG))
 987                                 snprintf(SNPARGS(fragment, 0),
 988                                     " (frag %08x:%d@%d%s)",
 989                                     args->f_id.frag_id6,
 990                                     ntohs(ip6->ip6_plen) - hlen,
 991                                     ntohs(offset & IP6F_OFF_MASK) << 3,
 992                                     (offset & IP6F_MORE_FRAG) ? "+" : "");
 993                 } else
 994 #endif
 995                 {
 996                         int ip_off, ip_len;
 997                         if (eh != NULL) { /* layer 2 packets are as on the wire */
 998                                 ip_off = ntohs(ip->ip_off);
 999                                 ip_len = ntohs(ip->ip_len);
1000                         } else {
1001                                 ip_off = ip->ip_off;
1002                                 ip_len = ip->ip_len;
1003                         }
1004                         if (ip_off & (IP_MF | IP_OFFMASK))
1005                                 snprintf(SNPARGS(fragment, 0),
1006                                     " (frag %d:%d@%d%s)",
1007                                     ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2),
1008                                     offset << 3,
1009                                     (ip_off & IP_MF) ? "+" : "");
1010                 }
1011         }
1012         if (oif || m->m_pkthdr.rcvif)
1013                 log(LOG_SECURITY | LOG_INFO,
1014                     "ipfw: %d %s %s %s via %s%s\n",
1015                     f ? f->rulenum : -1,
1016                     action, proto, oif ? "out" : "in",
1017                     oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname,
1018                     fragment);
1019         else
1020                 log(LOG_SECURITY | LOG_INFO,
1021                     "ipfw: %d %s %s [no if info]%s\n",
1022                     f ? f->rulenum : -1,
1023                     action, proto, fragment);
1024         if (limit_reached)
1025                 log(LOG_SECURITY | LOG_NOTICE,
1026                     "ipfw: limit %d reached on entry %d\n",
1027                     limit_reached, f ? f->rulenum : -1);
1028 }
1029
1030 /*
1031  * IMPORTANT: the hash function for dynamic rules must be commutative
1032  * in source and destination (ip,port), because rules are bidirectional
1033  * and we want to find both in the same bucket.
1034  */
1035 static __inline int
1036 hash_packet(struct ipfw_flow_id *id)
1037 {
1038         u_int32_t i;
1039
1040 #ifdef INET6
1041         if (IS_IP6_FLOW_ID(id))
1042                 i = hash_packet6(id);
1043         else
1044 #endif /* INET6 */
1045         i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port);
1046         i &= (curr_dyn_buckets - 1);
1047         return i;
1048 }
1049
1050 /**
1051  * unlink a dynamic rule from a chain. prev is a pointer to
1052  * the previous one, q is a pointer to the rule to delete,
1053  * head is a pointer to the head of the queue.
1054  * Modifies q and potentially also head.
1055  */
1056 #define UNLINK_DYN_RULE(prev, head, q) {                                \
1057         ipfw_dyn_rule *old_q = q;                                       \
1058                                                                         \
1059         /* remove a refcount to the parent */                           \
1060         if (q->dyn_type == O_LIMIT)                                     \
1061                 q->parent->count--;                                     \
1062         DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\
1063                 (q->id.src_ip), (q->id.src_port),                       \
1064                 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); )      \
1065         if (prev != NULL)                                               \
1066                 prev->next = q = q->next;                               \
1067         else                                                            \
1068                 head = q = q->next;                                     \
1069         dyn_count--;                                                    \
1070         uma_zfree(ipfw_dyn_rule_zone, old_q); }
1071
1072 #define TIME_LEQ(a,b)       ((int)((a)-(b)) <= 0)
1073
1074 /**
1075  * Remove dynamic rules pointing to "rule", or all of them if rule == NULL.
1076  *
1077  * If keep_me == NULL, rules are deleted even if not expired,
1078  * otherwise only expired rules are removed.
1079  *
1080  * The value of the second parameter is also used to point to identify
1081  * a rule we absolutely do not want to remove (e.g. because we are
1082  * holding a reference to it -- this is the case with O_LIMIT_PARENT
1083  * rules). The pointer is only used for comparison, so any non-null
1084  * value will do.
1085  */
1086 static void
1087 remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me)
1088 {
1089         static u_int32_t last_remove = 0;
1090
1091 #define FORCE (keep_me == NULL)
1092
1093         ipfw_dyn_rule *prev, *q;
1094         int i, pass = 0, max_pass = 0;
1095
1096         IPFW_DYN_LOCK_ASSERT();
1097
1098         if (ipfw_dyn_v == NULL || dyn_count == 0)
1099                 return;
1100         /* do not expire more than once per second, it is useless */
1101         if (!FORCE && last_remove == time_uptime)
1102                 return;
1103         last_remove = time_uptime;
1104
1105         /*
1106          * because O_LIMIT refer to parent rules, during the first pass only
1107          * remove child and mark any pending LIMIT_PARENT, and remove
1108          * them in a second pass.
1109          */
1110 next_pass:
1111         for (i = 0 ; i < curr_dyn_buckets ; i++) {
1112                 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) {
1113                         /*
1114                          * Logic can become complex here, so we split tests.
1115                          */
1116                         if (q == keep_me)
1117                                 goto next;
1118                         if (rule != NULL && rule != q->rule)
1119                                 goto next; /* not the one we are looking for */
1120                         if (q->dyn_type == O_LIMIT_PARENT) {
1121                                 /*
1122                                  * handle parent in the second pass,
1123                                  * record we need one.
1124                                  */
1125                                 max_pass = 1;
1126                                 if (pass == 0)
1127                                         goto next;
1128                                 if (FORCE && q->count != 0 ) {
1129                                         /* XXX should not happen! */
1130                                         printf("ipfw: OUCH! cannot remove rule,"
1131                                              " count %d\n", q->count);
1132                                 }
1133                         } else {
1134                                 if (!FORCE &&
1135                                     !TIME_LEQ( q->expire, time_uptime ))
1136                                         goto next;
1137                         }
1138              if (q->dyn_type != O_LIMIT_PARENT || !q->count) {
1139                      UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
1140                      continue;
1141              }
1142 next:
1143                         prev=q;
1144                         q=q->next;
1145                 }
1146         }
1147         if (pass++ < max_pass)
1148                 goto next_pass;
1149 }
1150
1151
1152 /**
1153  * lookup a dynamic rule.
1154  */
1155 static ipfw_dyn_rule *
1156 lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction,
1157     struct tcphdr *tcp)
1158 {
1159         /*
1160          * stateful ipfw extensions.
1161          * Lookup into dynamic session queue
1162          */
1163 #define MATCH_REVERSE   0
1164 #define MATCH_FORWARD   1
1165 #define MATCH_NONE      2
1166 #define MATCH_UNKNOWN   3
1167         int i, dir = MATCH_NONE;
1168         ipfw_dyn_rule *prev, *q=NULL;
1169
1170         IPFW_DYN_LOCK_ASSERT();
1171
1172         if (ipfw_dyn_v == NULL)
1173                 goto done;      /* not found */
1174         i = hash_packet( pkt );
1175         for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) {
1176                 if (q->dyn_type == O_LIMIT_PARENT && q->count)
1177                         goto next;
1178                 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */
1179                         UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
1180                         continue;
1181                 }
1182                 if (pkt->proto == q->id.proto &&
1183                     q->dyn_type != O_LIMIT_PARENT) {
1184                         if (IS_IP6_FLOW_ID(pkt)) {
1185                             if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6),
1186                                 &(q->id.src_ip6)) &&
1187                             IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6),
1188                                 &(q->id.dst_ip6)) &&
1189                             pkt->src_port == q->id.src_port &&
1190                             pkt->dst_port == q->id.dst_port ) {
1191                                 dir = MATCH_FORWARD;
1192                                 break;
1193                             }
1194                             if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6),
1195                                     &(q->id.dst_ip6)) &&
1196                                 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6),
1197                                     &(q->id.src_ip6)) &&
1198                                 pkt->src_port == q->id.dst_port &&
1199                                 pkt->dst_port == q->id.src_port ) {
1200                                     dir = MATCH_REVERSE;
1201                                     break;
1202                             }
1203                         } else {
1204                             if (pkt->src_ip == q->id.src_ip &&
1205                                 pkt->dst_ip == q->id.dst_ip &&
1206                                 pkt->src_port == q->id.src_port &&
1207                                 pkt->dst_port == q->id.dst_port ) {
1208                                     dir = MATCH_FORWARD;
1209                                     break;
1210                             }
1211                             if (pkt->src_ip == q->id.dst_ip &&
1212                                 pkt->dst_ip == q->id.src_ip &&
1213                                 pkt->src_port == q->id.dst_port &&
1214                                 pkt->dst_port == q->id.src_port ) {
1215                                     dir = MATCH_REVERSE;
1216                                     break;
1217                             }
1218                         }
1219                 }
1220 next:
1221                 prev = q;
1222                 q = q->next;
1223         }
1224         if (q == NULL)
1225                 goto done; /* q = NULL, not found */
1226
1227         if ( prev != NULL) { /* found and not in front */
1228                 prev->next = q->next;
1229                 q->next = ipfw_dyn_v[i];
1230                 ipfw_dyn_v[i] = q;
1231         }
1232         if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */
1233                 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST);
1234
1235 #define BOTH_SYN        (TH_SYN | (TH_SYN << 8))
1236 #define BOTH_FIN        (TH_FIN | (TH_FIN << 8))
1237                 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8);
1238                 switch (q->state) {
1239                 case TH_SYN:                            /* opening */
1240                         q->expire = time_uptime + dyn_syn_lifetime;
1241                         break;
1242
1243                 case BOTH_SYN:                  /* move to established */
1244                 case BOTH_SYN | TH_FIN :        /* one side tries to close */
1245                 case BOTH_SYN | (TH_FIN << 8) :
1246                         if (tcp) {
1247 #define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0)
1248                             u_int32_t ack = ntohl(tcp->th_ack);
1249                             if (dir == MATCH_FORWARD) {
1250                                 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd))
1251                                     q->ack_fwd = ack;
1252                                 else { /* ignore out-of-sequence */
1253                                     break;
1254                                 }
1255                             } else {
1256                                 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev))
1257                                     q->ack_rev = ack;
1258                                 else { /* ignore out-of-sequence */
1259                                     break;
1260                                 }
1261                             }
1262                         }
1263                         q->expire = time_uptime + dyn_ack_lifetime;
1264                         break;
1265
1266                 case BOTH_SYN | BOTH_FIN:       /* both sides closed */
1267                         if (dyn_fin_lifetime >= dyn_keepalive_period)
1268                                 dyn_fin_lifetime = dyn_keepalive_period - 1;
1269                         q->expire = time_uptime + dyn_fin_lifetime;
1270                         break;
1271
1272                 default:
1273 #if 0
1274                         /*
1275                          * reset or some invalid combination, but can also
1276                          * occur if we use keep-state the wrong way.
1277                          */
1278                         if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0)
1279                                 printf("invalid state: 0x%x\n", q->state);
1280 #endif
1281                         if (dyn_rst_lifetime >= dyn_keepalive_period)
1282                                 dyn_rst_lifetime = dyn_keepalive_period - 1;
1283                         q->expire = time_uptime + dyn_rst_lifetime;
1284                         break;
1285                 }
1286         } else if (pkt->proto == IPPROTO_UDP) {
1287                 q->expire = time_uptime + dyn_udp_lifetime;
1288         } else {
1289                 /* other protocols */
1290                 q->expire = time_uptime + dyn_short_lifetime;
1291         }
1292 done:
1293         if (match_direction)
1294                 *match_direction = dir;
1295         return q;
1296 }
1297
1298 static ipfw_dyn_rule *
1299 lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction,
1300     struct tcphdr *tcp)
1301 {
1302         ipfw_dyn_rule *q;
1303
1304         IPFW_DYN_LOCK();
1305         q = lookup_dyn_rule_locked(pkt, match_direction, tcp);
1306         if (q == NULL)
1307                 IPFW_DYN_UNLOCK();
1308         /* NB: return table locked when q is not NULL */
1309         return q;
1310 }
1311
1312 static void
1313 realloc_dynamic_table(void)
1314 {
1315         IPFW_DYN_LOCK_ASSERT();
1316
1317         /*
1318          * Try reallocation, make sure we have a power of 2 and do
1319          * not allow more than 64k entries. In case of overflow,
1320          * default to 1024.
1321          */
1322
1323         if (dyn_buckets > 65536)
1324                 dyn_buckets = 1024;
1325         if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */
1326                 dyn_buckets = curr_dyn_buckets; /* reset */
1327                 return;
1328         }
1329         curr_dyn_buckets = dyn_buckets;
1330         if (ipfw_dyn_v != NULL)
1331                 free(ipfw_dyn_v, M_IPFW);
1332         for (;;) {
1333                 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *),
1334                        M_IPFW, M_NOWAIT | M_ZERO);
1335                 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2)
1336                         break;
1337                 curr_dyn_buckets /= 2;
1338         }
1339 }
1340
1341 /**
1342  * Install state of type 'type' for a dynamic session.
1343  * The hash table contains two type of rules:
1344  * - regular rules (O_KEEP_STATE)
1345  * - rules for sessions with limited number of sess per user
1346  *   (O_LIMIT). When they are created, the parent is
1347  *   increased by 1, and decreased on delete. In this case,
1348  *   the third parameter is the parent rule and not the chain.
1349  * - "parent" rules for the above (O_LIMIT_PARENT).
1350  */
1351 static ipfw_dyn_rule *
1352 add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule)
1353 {
1354         ipfw_dyn_rule *r;
1355         int i;
1356
1357         IPFW_DYN_LOCK_ASSERT();
1358
1359         if (ipfw_dyn_v == NULL ||
1360             (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) {
1361                 realloc_dynamic_table();
1362                 if (ipfw_dyn_v == NULL)
1363                         return NULL; /* failed ! */
1364         }
1365         i = hash_packet(id);
1366
1367         r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO);
1368         if (r == NULL) {
1369                 printf ("ipfw: sorry cannot allocate state\n");
1370                 return NULL;
1371         }
1372
1373         /* increase refcount on parent, and set pointer */
1374         if (dyn_type == O_LIMIT) {
1375                 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule;
1376                 if ( parent->dyn_type != O_LIMIT_PARENT)
1377                         panic("invalid parent");
1378                 parent->count++;
1379                 r->parent = parent;
1380                 rule = parent->rule;
1381         }
1382
1383         r->id = *id;
1384         r->expire = time_uptime + dyn_syn_lifetime;
1385         r->rule = rule;
1386         r->dyn_type = dyn_type;
1387         r->pcnt = r->bcnt = 0;
1388         r->count = 0;
1389
1390         r->bucket = i;
1391         r->next = ipfw_dyn_v[i];
1392         ipfw_dyn_v[i] = r;
1393         dyn_count++;
1394         DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n",
1395            dyn_type,
1396            (r->id.src_ip), (r->id.src_port),
1397            (r->id.dst_ip), (r->id.dst_port),
1398            dyn_count ); )
1399         return r;
1400 }
1401
1402 /**
1403  * lookup dynamic parent rule using pkt and rule as search keys.
1404  * If the lookup fails, then install one.
1405  */
1406 static ipfw_dyn_rule *
1407 lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule)
1408 {
1409         ipfw_dyn_rule *q;
1410         int i;
1411
1412         IPFW_DYN_LOCK_ASSERT();
1413
1414         if (ipfw_dyn_v) {
1415                 int is_v6 = IS_IP6_FLOW_ID(pkt);
1416                 i = hash_packet( pkt );
1417                 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next)
1418                         if (q->dyn_type == O_LIMIT_PARENT &&
1419                             rule== q->rule &&
1420                             pkt->proto == q->id.proto &&
1421                             pkt->src_port == q->id.src_port &&
1422                             pkt->dst_port == q->id.dst_port &&
1423                             (
1424                                 (is_v6 &&
1425                                  IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6),
1426                                         &(q->id.src_ip6)) &&
1427                                  IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6),
1428                                         &(q->id.dst_ip6))) ||
1429                                 (!is_v6 &&
1430                                  pkt->src_ip == q->id.src_ip &&
1431                                  pkt->dst_ip == q->id.dst_ip)
1432                             )
1433                         ) {
1434                                 q->expire = time_uptime + dyn_short_lifetime;
1435                                 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);)
1436                                 return q;
1437                         }
1438         }
1439         return add_dyn_rule(pkt, O_LIMIT_PARENT, rule);
1440 }
1441
1442 /**
1443  * Install dynamic state for rule type cmd->o.opcode
1444  *
1445  * Returns 1 (failure) if state is not installed because of errors or because
1446  * session limitations are enforced.
1447  */
1448 static int
1449 install_state(struct ip_fw *rule, ipfw_insn_limit *cmd,
1450     struct ip_fw_args *args, uint32_t tablearg)
1451 {
1452         static int last_log;
1453         ipfw_dyn_rule *q;
1454         struct in_addr da;
1455         char src[48], dst[48];
1456
1457         src[0] = '\0';
1458         dst[0] = '\0';
1459
1460         DEB(
1461         printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n",
1462             __func__, cmd->o.opcode,
1463             (args->f_id.src_ip), (args->f_id.src_port),
1464             (args->f_id.dst_ip), (args->f_id.dst_port));
1465         )
1466
1467         IPFW_DYN_LOCK();
1468
1469         q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL);
1470
1471         if (q != NULL) {        /* should never occur */
1472                 if (last_log != time_uptime) {
1473                         last_log = time_uptime;
1474                         printf("ipfw: %s: entry already present, done\n",
1475                             __func__);
1476                 }
1477                 IPFW_DYN_UNLOCK();
1478                 return (0);
1479         }
1480
1481         if (dyn_count >= dyn_max)
1482                 /* Run out of slots, try to remove any expired rule. */
1483                 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1);
1484
1485         if (dyn_count >= dyn_max) {
1486                 if (last_log != time_uptime) {
1487                         last_log = time_uptime;
1488                         printf("ipfw: %s: Too many dynamic rules\n", __func__);
1489                 }
1490                 IPFW_DYN_UNLOCK();
1491                 return (1);     /* cannot install, notify caller */
1492         }
1493
1494         switch (cmd->o.opcode) {
1495         case O_KEEP_STATE:      /* bidir rule */
1496                 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule);
1497                 break;
1498
1499         case O_LIMIT: {         /* limit number of sessions */
1500                 struct ipfw_flow_id id;
1501                 ipfw_dyn_rule *parent;
1502                 uint32_t conn_limit;
1503                 uint16_t limit_mask = cmd->limit_mask;
1504
1505                 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ?
1506                     tablearg : cmd->conn_limit;
1507
1508                 DEB(
1509                 if (cmd->conn_limit == IP_FW_TABLEARG)
1510                         printf("ipfw: %s: O_LIMIT rule, conn_limit: %u "
1511                             "(tablearg)\n", __func__, conn_limit);
1512                 else
1513                         printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n",
1514                             __func__, conn_limit);
1515                 )
1516
1517                 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0;
1518                 id.proto = args->f_id.proto;
1519                 id.addr_type = args->f_id.addr_type;
1520
1521                 if (IS_IP6_FLOW_ID (&(args->f_id))) {
1522                         if (limit_mask & DYN_SRC_ADDR)
1523                                 id.src_ip6 = args->f_id.src_ip6;
1524                         if (limit_mask & DYN_DST_ADDR)
1525                                 id.dst_ip6 = args->f_id.dst_ip6;
1526                 } else {
1527                         if (limit_mask & DYN_SRC_ADDR)
1528                                 id.src_ip = args->f_id.src_ip;
1529                         if (limit_mask & DYN_DST_ADDR)
1530                                 id.dst_ip = args->f_id.dst_ip;
1531                 }
1532                 if (limit_mask & DYN_SRC_PORT)
1533                         id.src_port = args->f_id.src_port;
1534                 if (limit_mask & DYN_DST_PORT)
1535                         id.dst_port = args->f_id.dst_port;
1536                 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) {
1537                         printf("ipfw: %s: add parent failed\n", __func__);
1538                         IPFW_DYN_UNLOCK();
1539                         return (1);
1540                 }
1541
1542                 if (parent->count >= conn_limit) {
1543                         /* See if we can remove some expired rule. */
1544                         remove_dyn_rule(rule, parent);
1545                         if (parent->count >= conn_limit) {
1546                                 if (fw_verbose && last_log != time_uptime) {
1547                                         last_log = time_uptime;
1548 #ifdef INET6
1549                                         /*
1550                                          * XXX IPv6 flows are not
1551                                          * supported yet.
1552                                          */
1553                                         if (IS_IP6_FLOW_ID(&(args->f_id))) {
1554                                                 char ip6buf[INET6_ADDRSTRLEN];
1555                                                 snprintf(src, sizeof(src),
1556                                                     "[%s]", ip6_sprintf(ip6buf,
1557                                                         &args->f_id.src_ip6));
1558                                                 snprintf(dst, sizeof(dst),
1559                                                     "[%s]", ip6_sprintf(ip6buf,
1560                                                         &args->f_id.dst_ip6));
1561                                         } else
1562 #endif
1563                                         {
1564                                                 da.s_addr =
1565                                                     htonl(args->f_id.src_ip);
1566                                                 inet_ntoa_r(da, src);
1567                                                 da.s_addr =
1568                                                     htonl(args->f_id.dst_ip);
1569                                                 inet_ntoa_r(da, dst);
1570                                         }
1571                                         log(LOG_SECURITY | LOG_DEBUG,
1572                                             "%s %s:%u -> %s:%u, %s\n",
1573                                             "drop session",
1574                                             src, (args->f_id.src_port),
1575                                             dst, (args->f_id.dst_port),
1576                                             "too many entries");
1577                                 }
1578                                 IPFW_DYN_UNLOCK();
1579                                 return (1);
1580                         }
1581                 }
1582                 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent);
1583                 break;
1584         }
1585         default:
1586                 printf("ipfw: %s: unknown dynamic rule type %u\n",
1587                     __func__, cmd->o.opcode);
1588                 IPFW_DYN_UNLOCK();
1589                 return (1);
1590         }
1591
1592         /* XXX just set lifetime */
1593         lookup_dyn_rule_locked(&args->f_id, NULL, NULL);
1594
1595         IPFW_DYN_UNLOCK();
1596         return (0);
1597 }
1598
1599 /*
1600  * Generate a TCP packet, containing either a RST or a keepalive.
1601  * When flags & TH_RST, we are sending a RST packet, because of a
1602  * "reset" action matched the packet.
1603  * Otherwise we are sending a keepalive, and flags & TH_
1604  * The 'replyto' mbuf is the mbuf being replied to, if any, and is required
1605  * so that MAC can label the reply appropriately.
1606  */
1607 static struct mbuf *
1608 send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq,
1609     u_int32_t ack, int flags)
1610 {
1611         struct mbuf *m;
1612         struct ip *ip;
1613         struct tcphdr *tcp;
1614
1615         MGETHDR(m, M_DONTWAIT, MT_DATA);
1616         if (m == 0)
1617                 return (NULL);
1618         m->m_pkthdr.rcvif = (struct ifnet *)0;
1619
1620 #ifdef MAC
1621         if (replyto != NULL)
1622                 mac_create_mbuf_netlayer(replyto, m);
1623         else
1624                 mac_create_mbuf_from_firewall(m);
1625 #else
1626         (void)replyto;          /* don't warn about unused arg */
1627 #endif
1628
1629         m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr);
1630         m->m_data += max_linkhdr;
1631
1632         ip = mtod(m, struct ip *);
1633         bzero(ip, m->m_len);
1634         tcp = (struct tcphdr *)(ip + 1); /* no IP options */
1635         ip->ip_p = IPPROTO_TCP;
1636         tcp->th_off = 5;
1637         /*
1638          * Assume we are sending a RST (or a keepalive in the reverse
1639          * direction), swap src and destination addresses and ports.
1640          */
1641         ip->ip_src.s_addr = htonl(id->dst_ip);
1642         ip->ip_dst.s_addr = htonl(id->src_ip);
1643         tcp->th_sport = htons(id->dst_port);
1644         tcp->th_dport = htons(id->src_port);
1645         if (flags & TH_RST) {   /* we are sending a RST */
1646                 if (flags & TH_ACK) {
1647                         tcp->th_seq = htonl(ack);
1648                         tcp->th_ack = htonl(0);
1649                         tcp->th_flags = TH_RST;
1650                 } else {
1651                         if (flags & TH_SYN)
1652                                 seq++;
1653                         tcp->th_seq = htonl(0);
1654                         tcp->th_ack = htonl(seq);
1655                         tcp->th_flags = TH_RST | TH_ACK;
1656                 }
1657         } else {
1658                 /*
1659                  * We are sending a keepalive. flags & TH_SYN determines
1660                  * the direction, forward if set, reverse if clear.
1661                  * NOTE: seq and ack are always assumed to be correct
1662                  * as set by the caller. This may be confusing...
1663                  */
1664                 if (flags & TH_SYN) {
1665                         /*
1666                          * we have to rewrite the correct addresses!
1667                          */
1668                         ip->ip_dst.s_addr = htonl(id->dst_ip);
1669                         ip->ip_src.s_addr = htonl(id->src_ip);
1670                         tcp->th_dport = htons(id->dst_port);
1671                         tcp->th_sport = htons(id->src_port);
1672                 }
1673                 tcp->th_seq = htonl(seq);
1674                 tcp->th_ack = htonl(ack);
1675                 tcp->th_flags = TH_ACK;
1676         }
1677         /*
1678          * set ip_len to the payload size so we can compute
1679          * the tcp checksum on the pseudoheader
1680          * XXX check this, could save a couple of words ?
1681          */
1682         ip->ip_len = htons(sizeof(struct tcphdr));
1683         tcp->th_sum = in_cksum(m, m->m_pkthdr.len);
1684         /*
1685          * now fill fields left out earlier
1686          */
1687         ip->ip_ttl = ip_defttl;
1688         ip->ip_len = m->m_pkthdr.len;
1689         m->m_flags |= M_SKIP_FIREWALL;
1690         return (m);
1691 }
1692
1693 /*
1694  * sends a reject message, consuming the mbuf passed as an argument.
1695  */
1696 static void
1697 send_reject(struct ip_fw_args *args, int code, int ip_len, struct ip *ip)
1698 {
1699
1700 #if 0
1701         /* XXX When ip is not guaranteed to be at mtod() we will
1702          * need to account for this */
1703          * The mbuf will however be thrown away so we can adjust it.
1704          * Remember we did an m_pullup on it already so we
1705          * can make some assumptions about contiguousness.
1706          */
1707         if (args->L3offset)
1708                 m_adj(m, args->L3offset);
1709 #endif
1710         if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
1711                 /* We need the IP header in host order for icmp_error(). */
1712                 if (args->eh != NULL) {
1713                         ip->ip_len = ntohs(ip->ip_len);
1714                         ip->ip_off = ntohs(ip->ip_off);
1715                 }
1716                 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
1717         } else if (args->f_id.proto == IPPROTO_TCP) {
1718                 struct tcphdr *const tcp =
1719                     L3HDR(struct tcphdr, mtod(args->m, struct ip *));
1720                 if ( (tcp->th_flags & TH_RST) == 0) {
1721                         struct mbuf *m;
1722                         m = send_pkt(args->m, &(args->f_id),
1723                                 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
1724                                 tcp->th_flags | TH_RST);
1725                         if (m != NULL)
1726                                 ip_output(m, NULL, NULL, 0, NULL, NULL);
1727                 }
1728                 m_freem(args->m);
1729         } else
1730                 m_freem(args->m);
1731         args->m = NULL;
1732 }
1733
1734 /**
1735  *
1736  * Given an ip_fw *, lookup_next_rule will return a pointer
1737  * to the next rule, which can be either the jump
1738  * target (for skipto instructions) or the next one in the list (in
1739  * all other cases including a missing jump target).
1740  * The result is also written in the "next_rule" field of the rule.
1741  * Backward jumps are not allowed, so start looking from the next
1742  * rule...
1743  *
1744  * This never returns NULL -- in case we do not have an exact match,
1745  * the next rule is returned. When the ruleset is changed,
1746  * pointers are flushed so we are always correct.
1747  */
1748
1749 static struct ip_fw *
1750 lookup_next_rule(struct ip_fw *me)
1751 {
1752         struct ip_fw *rule = NULL;
1753         ipfw_insn *cmd;
1754
1755         /* look for action, in case it is a skipto */
1756         cmd = ACTION_PTR(me);
1757         if (cmd->opcode == O_LOG)
1758                 cmd += F_LEN(cmd);
1759         if (cmd->opcode == O_ALTQ)
1760                 cmd += F_LEN(cmd);
1761         if (cmd->opcode == O_TAG)
1762                 cmd += F_LEN(cmd);
1763         if ( cmd->opcode == O_SKIPTO )
1764                 for (rule = me->next; rule ; rule = rule->next)
1765                         if (rule->rulenum >= cmd->arg1)
1766                                 break;
1767         if (rule == NULL)                       /* failure or not a skipto */
1768                 rule = me->next;
1769         me->next_rule = rule;
1770         return rule;
1771 }
1772
1773 static int
1774 add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr,
1775     uint8_t mlen, uint32_t value)
1776 {
1777         struct radix_node_head *rnh;
1778         struct table_entry *ent;
1779
1780         if (tbl >= IPFW_TABLES_MAX)
1781                 return (EINVAL);
1782         rnh = ch->tables[tbl];
1783         ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO);
1784         if (ent == NULL)
1785                 return (ENOMEM);
1786         ent->value = value;
1787         ent->addr.sin_len = ent->mask.sin_len = 8;
1788         ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0);
1789         ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr;
1790         IPFW_WLOCK(&layer3_chain);
1791         if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) ==
1792             NULL) {
1793                 IPFW_WUNLOCK(&layer3_chain);
1794                 free(ent, M_IPFW_TBL);
1795                 return (EEXIST);
1796         }
1797         IPFW_WUNLOCK(&layer3_chain);
1798         return (0);
1799 }
1800
1801 static int
1802 del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr,
1803     uint8_t mlen)
1804 {
1805         struct radix_node_head *rnh;
1806         struct table_entry *ent;
1807         struct sockaddr_in sa, mask;
1808
1809         if (tbl >= IPFW_TABLES_MAX)
1810                 return (EINVAL);
1811         rnh = ch->tables[tbl];
1812         sa.sin_len = mask.sin_len = 8;
1813         mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0);
1814         sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr;
1815         IPFW_WLOCK(ch);
1816         ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh);
1817         if (ent == NULL) {
1818                 IPFW_WUNLOCK(ch);
1819                 return (ESRCH);
1820         }
1821         IPFW_WUNLOCK(ch);
1822         free(ent, M_IPFW_TBL);
1823         return (0);
1824 }
1825
1826 static int
1827 flush_table_entry(struct radix_node *rn, void *arg)
1828 {
1829         struct radix_node_head * const rnh = arg;
1830         struct table_entry *ent;
1831
1832         ent = (struct table_entry *)
1833             rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh);
1834         if (ent != NULL)
1835                 free(ent, M_IPFW_TBL);
1836         return (0);
1837 }
1838
1839 static int
1840 flush_table(struct ip_fw_chain *ch, uint16_t tbl)
1841 {
1842         struct radix_node_head *rnh;
1843
1844         IPFW_WLOCK_ASSERT(ch);
1845
1846         if (tbl >= IPFW_TABLES_MAX)
1847                 return (EINVAL);
1848         rnh = ch->tables[tbl];
1849         KASSERT(rnh != NULL, ("NULL IPFW table"));
1850         rnh->rnh_walktree(rnh, flush_table_entry, rnh);
1851         return (0);
1852 }
1853
1854 static void
1855 flush_tables(struct ip_fw_chain *ch)
1856 {
1857         uint16_t tbl;
1858
1859         IPFW_WLOCK_ASSERT(ch);
1860
1861         for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++)
1862                 flush_table(ch, tbl);
1863 }
1864
1865 static int
1866 init_tables(struct ip_fw_chain *ch)
1867 {
1868         int i;
1869         uint16_t j;
1870
1871         for (i = 0; i < IPFW_TABLES_MAX; i++) {
1872                 if (!rn_inithead((void **)&ch->tables[i], 32)) {
1873                         for (j = 0; j < i; j++) {
1874                                 (void) flush_table(ch, j);
1875                         }
1876                         return (ENOMEM);
1877                 }
1878         }
1879         return (0);
1880 }
1881
1882 static int
1883 lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr,
1884     uint32_t *val)
1885 {
1886         struct radix_node_head *rnh;
1887         struct table_entry *ent;
1888         struct sockaddr_in sa;
1889
1890         if (tbl >= IPFW_TABLES_MAX)
1891                 return (0);
1892         rnh = ch->tables[tbl];
1893         sa.sin_len = 8;
1894         sa.sin_addr.s_addr = addr;
1895         ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh));
1896         if (ent != NULL) {
1897                 *val = ent->value;
1898                 return (1);
1899         }
1900         return (0);
1901 }
1902
1903 static int
1904 count_table_entry(struct radix_node *rn, void *arg)
1905 {
1906         u_int32_t * const cnt = arg;
1907
1908         (*cnt)++;
1909         return (0);
1910 }
1911
1912 static int
1913 count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt)
1914 {
1915         struct radix_node_head *rnh;
1916
1917         if (tbl >= IPFW_TABLES_MAX)
1918                 return (EINVAL);
1919         rnh = ch->tables[tbl];
1920         *cnt = 0;
1921         rnh->rnh_walktree(rnh, count_table_entry, cnt);
1922         return (0);
1923 }
1924
1925 static int
1926 dump_table_entry(struct radix_node *rn, void *arg)
1927 {
1928         struct table_entry * const n = (struct table_entry *)rn;
1929         ipfw_table * const tbl = arg;
1930         ipfw_table_entry *ent;
1931
1932         if (tbl->cnt == tbl->size)
1933                 return (1);
1934         ent = &tbl->ent[tbl->cnt];
1935         ent->tbl = tbl->tbl;
1936         if (in_nullhost(n->mask.sin_addr))
1937                 ent->masklen = 0;
1938         else
1939                 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr));
1940         ent->addr = n->addr.sin_addr.s_addr;
1941         ent->value = n->value;
1942         tbl->cnt++;
1943         return (0);
1944 }
1945
1946 static int
1947 dump_table(struct ip_fw_chain *ch, ipfw_table *tbl)
1948 {
1949         struct radix_node_head *rnh;
1950
1951         if (tbl->tbl >= IPFW_TABLES_MAX)
1952                 return (EINVAL);
1953         rnh = ch->tables[tbl->tbl];
1954         tbl->cnt = 0;
1955         rnh->rnh_walktree(rnh, dump_table_entry, tbl);
1956         return (0);
1957 }
1958
1959 static void
1960 fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp)
1961 {
1962         struct ucred *cr;
1963
1964         if (inp->inp_socket != NULL) {
1965                 cr = inp->inp_socket->so_cred;
1966                 ugp->fw_prid = jailed(cr) ?
1967                     cr->cr_prison->pr_id : -1;
1968                 ugp->fw_uid = cr->cr_uid;
1969                 ugp->fw_ngroups = cr->cr_ngroups;
1970                 bcopy(cr->cr_groups, ugp->fw_groups,
1971                     sizeof(ugp->fw_groups));
1972         }
1973 }
1974
1975 static int
1976 check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif,
1977     struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip,
1978     u_int16_t src_port, struct ip_fw_ugid *ugp, int *lookup,
1979     struct inpcb *inp)
1980 {
1981         struct inpcbinfo *pi;
1982         int wildcard;
1983         struct inpcb *pcb;
1984         int match;
1985         gid_t *gp;
1986
1987         /*
1988          * Check to see if the UDP or TCP stack supplied us with
1989          * the PCB. If so, rather then holding a lock and looking
1990          * up the PCB, we can use the one that was supplied.
1991          */
1992         if (inp && *lookup == 0) {
1993                 INP_LOCK_ASSERT(inp);
1994                 if (inp->inp_socket != NULL) {
1995                         fill_ugid_cache(inp, ugp);
1996                         *lookup = 1;
1997                 }
1998         }
1999         /*
2000          * If we have already been here and the packet has no
2001          * PCB entry associated with it, then we can safely
2002          * assume that this is a no match.
2003          */
2004         if (*lookup == -1)
2005                 return (0);
2006         if (proto == IPPROTO_TCP) {
2007                 wildcard = 0;
2008                 pi = &tcbinfo;
2009         } else if (proto == IPPROTO_UDP) {
2010                 wildcard = INPLOOKUP_WILDCARD;
2011                 pi = &udbinfo;
2012         } else
2013                 return 0;
2014         match = 0;
2015         if (*lookup == 0) {
2016                 INP_INFO_RLOCK(pi);
2017                 pcb =  (oif) ?
2018                         in_pcblookup_hash(pi,
2019                                 dst_ip, htons(dst_port),
2020                                 src_ip, htons(src_port),
2021                                 wildcard, oif) :
2022                         in_pcblookup_hash(pi,
2023                                 src_ip, htons(src_port),
2024                                 dst_ip, htons(dst_port),
2025                                 wildcard, NULL);
2026                 if (pcb != NULL) {
2027                         INP_LOCK(pcb);
2028                         if (pcb->inp_socket != NULL) {
2029                                 fill_ugid_cache(pcb, ugp);
2030                                 *lookup = 1;
2031                         }
2032                         INP_UNLOCK(pcb);
2033                 }
2034                 INP_INFO_RUNLOCK(pi);
2035                 if (*lookup == 0) {
2036                         /*
2037                          * If the lookup did not yield any results, there
2038                          * is no sense in coming back and trying again. So
2039                          * we can set lookup to -1 and ensure that we wont
2040                          * bother the pcb system again.
2041                          */
2042                         *lookup = -1;
2043                         return (0);
2044                 }
2045         }
2046         if (insn->o.opcode == O_UID)
2047                 match = (ugp->fw_uid == (uid_t)insn->d[0]);
2048         else if (insn->o.opcode == O_GID) {
2049                 for (gp = ugp->fw_groups;
2050                         gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++)
2051                         if (*gp == (gid_t)insn->d[0]) {
2052                                 match = 1;
2053                                 break;
2054                         }
2055         } else if (insn->o.opcode == O_JAIL)
2056                 match = (ugp->fw_prid == (int)insn->d[0]);
2057         return match;
2058 }
2059
2060 #ifdef IPFIREWALL_NAT
2061 static eventhandler_tag ifaddr_event_tag;
2062
2063 static void
2064 ifaddr_change(void *arg __unused, struct ifnet *ifp)
2065 {
2066         struct cfg_nat *ptr;
2067         struct ifaddr *ifa;
2068
2069         IPFW_WLOCK(&layer3_chain);
2070         /* Check every nat entry... */
2071         LIST_FOREACH(ptr, &layer3_chain.nat, _next) {
2072                 /* ...using nic 'ifp->if_xname' as dynamic alias address. */
2073                 if (strncmp(ptr->if_name, ifp->if_xname, IF_NAMESIZE) == 0) {
2074                         mtx_lock(&ifp->if_addr_mtx);
2075                         TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
2076                                 if (ifa->ifa_addr == NULL)
2077                                         continue;
2078                                 if (ifa->ifa_addr->sa_family != AF_INET)
2079                                         continue;
2080                                 ptr->ip = ((struct sockaddr_in *)
2081                                     (ifa->ifa_addr))->sin_addr;
2082                                 LibAliasSetAddress(ptr->lib, ptr->ip);
2083                         }
2084                         mtx_unlock(&ifp->if_addr_mtx);
2085                 }
2086         }
2087         IPFW_WUNLOCK(&layer3_chain);
2088 }
2089
2090 static void
2091 flush_nat_ptrs(const int i)
2092 {
2093         struct ip_fw *rule;
2094
2095         IPFW_WLOCK_ASSERT(&layer3_chain);
2096         for (rule = layer3_chain.rules; rule; rule = rule->next) {
2097                 ipfw_insn_nat *cmd = (ipfw_insn_nat *)ACTION_PTR(rule);
2098                 if (cmd->o.opcode != O_NAT)
2099                         continue;
2100                 if (cmd->nat != NULL && cmd->nat->id == i)
2101                         cmd->nat = NULL;
2102         }
2103 }
2104
2105 static struct cfg_nat *
2106 lookup_nat(const int i)
2107 {
2108         struct cfg_nat *ptr;
2109
2110         LIST_FOREACH(ptr, &layer3_chain.nat, _next)
2111                 if (ptr->id == i)
2112                         return(ptr);
2113         return (NULL);
2114 }
2115
2116 #define HOOK_NAT(b, p) do {                                     \
2117         IPFW_WLOCK_ASSERT(&layer3_chain);                       \
2118         LIST_INSERT_HEAD(b, p, _next);                          \
2119 } while (0)
2120
2121 #define UNHOOK_NAT(p) do {                                      \
2122         IPFW_WLOCK_ASSERT(&layer3_chain);                       \
2123         LIST_REMOVE(p, _next);                                  \
2124 } while (0)
2125
2126 #define HOOK_REDIR(b, p) do {                                   \
2127         LIST_INSERT_HEAD(b, p, _next);                          \
2128 } while (0)
2129
2130 #define HOOK_SPOOL(b, p) do {                                   \
2131         LIST_INSERT_HEAD(b, p, _next);                          \
2132 } while (0)
2133
2134 static void
2135 del_redir_spool_cfg(struct cfg_nat *n, struct redir_chain *head)
2136 {
2137         struct cfg_redir *r, *tmp_r;
2138         struct cfg_spool *s, *tmp_s;
2139         int i, num;
2140
2141         LIST_FOREACH_SAFE(r, head, _next, tmp_r) {
2142                 num = 1; /* Number of alias_link to delete. */
2143                 switch (r->mode) {
2144                 case REDIR_PORT:
2145                         num = r->pport_cnt;
2146                         /* FALLTHROUGH */
2147                 case REDIR_ADDR:
2148                 case REDIR_PROTO:
2149                         /* Delete all libalias redirect entry. */
2150                         for (i = 0; i < num; i++)
2151                                 LibAliasRedirectDelete(n->lib, r->alink[i]);
2152                         /* Del spool cfg if any. */
2153                         LIST_FOREACH_SAFE(s, &r->spool_chain, _next, tmp_s) {
2154                                 LIST_REMOVE(s, _next);
2155                                 free(s, M_IPFW);
2156                         }
2157                         free(r->alink, M_IPFW);
2158                         LIST_REMOVE(r, _next);
2159                         free(r, M_IPFW);
2160                         break;
2161                 default:
2162                         printf("unknown redirect mode: %u\n", r->mode);
2163                         /* XXX - panic?!?!? */
2164                         break;
2165                 }
2166         }
2167 }
2168
2169 static int
2170 add_redir_spool_cfg(char *buf, struct cfg_nat *ptr)
2171 {
2172         struct cfg_redir *r, *ser_r;
2173         struct cfg_spool *s, *ser_s;
2174         int cnt, off, i;
2175         char *panic_err;
2176
2177         for (cnt = 0, off = 0; cnt < ptr->redir_cnt; cnt++) {
2178                 ser_r = (struct cfg_redir *)&buf[off];
2179                 r = malloc(SOF_REDIR, M_IPFW, M_WAITOK | M_ZERO);
2180                 memcpy(r, ser_r, SOF_REDIR);
2181                 LIST_INIT(&r->spool_chain);
2182                 off += SOF_REDIR;
2183                 r->alink = malloc(sizeof(struct alias_link *) * r->pport_cnt,
2184                     M_IPFW, M_WAITOK | M_ZERO);
2185                 switch (r->mode) {
2186                 case REDIR_ADDR:
2187                         r->alink[0] = LibAliasRedirectAddr(ptr->lib, r->laddr,
2188                             r->paddr);
2189                         break;
2190                 case REDIR_PORT:
2191                         for (i = 0 ; i < r->pport_cnt; i++) {
2192                                 /* If remotePort is all ports, set it to 0. */
2193                                 u_short remotePortCopy = r->rport + i;
2194                                 if (r->rport_cnt == 1 && r->rport == 0)
2195                                         remotePortCopy = 0;
2196                                 r->alink[i] = LibAliasRedirectPort(ptr->lib,
2197                                     r->laddr, htons(r->lport + i), r->raddr,
2198                                     htons(remotePortCopy), r->paddr,
2199                                     htons(r->pport + i), r->proto);
2200                                 if (r->alink[i] == NULL) {
2201                                         r->alink[0] = NULL;
2202                                         break;
2203                                 }
2204                         }
2205                         break;
2206                 case REDIR_PROTO:
2207                         r->alink[0] = LibAliasRedirectProto(ptr->lib ,r->laddr,
2208                             r->raddr, r->paddr, r->proto);
2209                         break;
2210                 default:
2211                         printf("unknown redirect mode: %u\n", r->mode);
2212                         break;
2213                 }
2214                 if (r->alink[0] == NULL) {
2215                         panic_err = "LibAliasRedirect* returned NULL";
2216                         goto bad;
2217                 } else /* LSNAT handling. */
2218                         for (i = 0; i < r->spool_cnt; i++) {
2219                                 ser_s = (struct cfg_spool *)&buf[off];
2220                                 s = malloc(SOF_REDIR, M_IPFW,
2221                                     M_WAITOK | M_ZERO);
2222                                 memcpy(s, ser_s, SOF_SPOOL);
2223                                 LibAliasAddServer(ptr->lib, r->alink[0],
2224                                     s->addr, htons(s->port));
2225                                 off += SOF_SPOOL;
2226                                 /* Hook spool entry. */
2227                                 HOOK_SPOOL(&r->spool_chain, s);
2228                         }
2229                 /* And finally hook this redir entry. */
2230                 HOOK_REDIR(&ptr->redir_chain, r);
2231         }
2232         return (1);
2233 bad:
2234         /* something really bad happened: panic! */
2235         panic("%s\n", panic_err);
2236 }
2237 #endif
2238
2239 /*
2240  * The main check routine for the firewall.
2241  *
2242  * All arguments are in args so we can modify them and return them
2243  * back to the caller.
2244  *
2245  * Parameters:
2246  *
2247  *      args->m (in/out) The packet; we set to NULL when/if we nuke it.
2248  *              Starts with the IP header.
2249  *      args->eh (in)   Mac header if present, or NULL for layer3 packet.
2250  *      args->L3offset  Number of bytes bypassed if we came from L2.
2251  *                      e.g. often sizeof(eh)  ** NOTYET **
2252  *      args->oif       Outgoing interface, or NULL if packet is incoming.
2253  *              The incoming interface is in the mbuf. (in)
2254  *      args->divert_rule (in/out)
2255  *              Skip up to the first rule past this rule number;
2256  *              upon return, non-zero port number for divert or tee.
2257  *
2258  *      args->rule      Pointer to the last matching rule (in/out)
2259  *      args->next_hop  Socket we are forwarding to (out).
2260  *      args->f_id      Addresses grabbed from the packet (out)
2261  *      args->cookie    a cookie depending on rule action
2262  *
2263  * Return value:
2264  *
2265  *      IP_FW_PASS      the packet must be accepted
2266  *      IP_FW_DENY      the packet must be dropped
2267  *      IP_FW_DIVERT    divert packet, port in m_tag
2268  *      IP_FW_TEE       tee packet, port in m_tag
2269  *      IP_FW_DUMMYNET  to dummynet, pipe in args->cookie
2270  *      IP_FW_NETGRAPH  into netgraph, cookie args->cookie
2271  *
2272  */
2273 int
2274 ipfw_chk(struct ip_fw_args *args)
2275 {
2276         /*
2277          * Local variables holding state during the processing of a packet:
2278          *
2279          * IMPORTANT NOTE: to speed up the processing of rules, there
2280          * are some assumption on the values of the variables, which
2281          * are documented here. Should you change them, please check
2282          * the implementation of the various instructions to make sure
2283          * that they still work.
2284          *
2285          * args->eh     The MAC header. It is non-null for a layer2
2286          *      packet, it is NULL for a layer-3 packet.
2287          * **notyet**
2288          * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
2289          *
2290          * m | args->m  Pointer to the mbuf, as received from the caller.
2291          *      It may change if ipfw_chk() does an m_pullup, or if it
2292          *      consumes the packet because it calls send_reject().
2293          *      XXX This has to change, so that ipfw_chk() never modifies
2294          *      or consumes the buffer.
2295          * ip   is the beginning of the ip(4 or 6) header.
2296          *      Calculated by adding the L3offset to the start of data.
2297          *      (Until we start using L3offset, the packet is
2298          *      supposed to start with the ip header).
2299          */
2300         struct mbuf *m = args->m;
2301         struct ip *ip = mtod(m, struct ip *);
2302
2303         /*
2304          * For rules which contain uid/gid or jail constraints, cache
2305          * a copy of the users credentials after the pcb lookup has been
2306          * executed. This will speed up the processing of rules with
2307          * these types of constraints, as well as decrease contention
2308          * on pcb related locks.
2309          */
2310         struct ip_fw_ugid fw_ugid_cache;
2311         int ugid_lookup = 0;
2312
2313         /*
2314          * divinput_flags       If non-zero, set to the IP_FW_DIVERT_*_FLAG
2315          *      associated with a packet input on a divert socket.  This
2316          *      will allow to distinguish traffic and its direction when
2317          *      it originates from a divert socket.
2318          */
2319         u_int divinput_flags = 0;
2320
2321         /*
2322          * oif | args->oif      If NULL, ipfw_chk has been called on the
2323          *      inbound path (ether_input, ip_input).
2324          *      If non-NULL, ipfw_chk has been called on the outbound path
2325          *      (ether_output, ip_output).
2326          */
2327         struct ifnet *oif = args->oif;
2328
2329         struct ip_fw *f = NULL;         /* matching rule */
2330         int retval = 0;
2331
2332         /*
2333          * hlen The length of the IP header.
2334          */
2335         u_int hlen = 0;         /* hlen >0 means we have an IP pkt */
2336
2337         /*
2338          * offset       The offset of a fragment. offset != 0 means that
2339          *      we have a fragment at this offset of an IPv4 packet.
2340          *      offset == 0 means that (if this is an IPv4 packet)
2341          *      this is the first or only fragment.
2342          *      For IPv6 offset == 0 means there is no Fragment Header.
2343          *      If offset != 0 for IPv6 always use correct mask to
2344          *      get the correct offset because we add IP6F_MORE_FRAG
2345          *      to be able to dectect the first fragment which would
2346          *      otherwise have offset = 0.
2347          */
2348         u_short offset = 0;
2349
2350         /*
2351          * Local copies of addresses. They are only valid if we have
2352          * an IP packet.
2353          *
2354          * proto        The protocol. Set to 0 for non-ip packets,
2355          *      or to the protocol read from the packet otherwise.
2356          *      proto != 0 means that we have an IPv4 packet.
2357          *
2358          * src_port, dst_port   port numbers, in HOST format. Only
2359          *      valid for TCP and UDP packets.
2360          *
2361          * src_ip, dst_ip       ip addresses, in NETWORK format.
2362          *      Only valid for IPv4 packets.
2363          */
2364         u_int8_t proto;
2365         u_int16_t src_port = 0, dst_port = 0;   /* NOTE: host format    */
2366         struct in_addr src_ip, dst_ip;          /* NOTE: network format */
2367         u_int16_t ip_len=0;
2368         int pktlen;
2369         u_int16_t       etype = 0;      /* Host order stored ether type */
2370
2371         /*
2372          * dyn_dir = MATCH_UNKNOWN when rules unchecked,
2373          *      MATCH_NONE when checked and not matched (q = NULL),
2374          *      MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
2375          */
2376         int dyn_dir = MATCH_UNKNOWN;
2377         ipfw_dyn_rule *q = NULL;
2378         struct ip_fw_chain *chain = &layer3_chain;
2379         struct m_tag *mtag;
2380
2381         /*
2382          * We store in ulp a pointer to the upper layer protocol header.
2383          * In the ipv4 case this is easy to determine from the header,
2384          * but for ipv6 we might have some additional headers in the middle.
2385          * ulp is NULL if not found.
2386          */
2387         void *ulp = NULL;               /* upper layer protocol pointer. */
2388         /* XXX ipv6 variables */
2389         int is_ipv6 = 0;
2390         u_int16_t ext_hd = 0;   /* bits vector for extension header filtering */
2391         /* end of ipv6 variables */
2392         int is_ipv4 = 0;
2393
2394         if (m->m_flags & M_SKIP_FIREWALL)
2395                 return (IP_FW_PASS);    /* accept */
2396
2397         pktlen = m->m_pkthdr.len;
2398         proto = args->f_id.proto = 0;   /* mark f_id invalid */
2399                 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
2400
2401 /*
2402  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
2403  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
2404  * pointer might become stale after other pullups (but we never use it
2405  * this way).
2406  */
2407 #define PULLUP_TO(len, p, T)                                            \
2408 do {                                                                    \
2409         int x = (len) + sizeof(T);                                      \
2410         if ((m)->m_len < x) {                                           \
2411                 args->m = m = m_pullup(m, x);                           \
2412                 if (m == NULL)                                          \
2413                         goto pullup_failed;                             \
2414         }                                                               \
2415         p = (mtod(m, char *) + (len));                                  \
2416 } while (0)
2417
2418         /*
2419          * if we have an ether header,
2420          */
2421         if (args->eh)
2422                 etype = ntohs(args->eh->ether_type);
2423
2424         /* Identify IP packets and fill up variables. */
2425         if (pktlen >= sizeof(struct ip6_hdr) &&
2426             (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
2427                 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
2428                 is_ipv6 = 1;
2429                 args->f_id.addr_type = 6;
2430                 hlen = sizeof(struct ip6_hdr);
2431                 proto = ip6->ip6_nxt;
2432
2433                 /* Search extension headers to find upper layer protocols */
2434                 while (ulp == NULL) {
2435                         switch (proto) {
2436                         case IPPROTO_ICMPV6:
2437                                 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
2438                                 args->f_id.flags = ICMP6(ulp)->icmp6_type;
2439                                 break;
2440
2441                         case IPPROTO_TCP:
2442                                 PULLUP_TO(hlen, ulp, struct tcphdr);
2443                                 dst_port = TCP(ulp)->th_dport;
2444                                 src_port = TCP(ulp)->th_sport;
2445                                 args->f_id.flags = TCP(ulp)->th_flags;
2446                                 break;
2447
2448                         case IPPROTO_SCTP:
2449                                 PULLUP_TO(hlen, ulp, struct sctphdr);
2450                                 src_port = SCTP(ulp)->src_port;
2451                                 dst_port = SCTP(ulp)->dest_port;
2452                                 break;
2453
2454                         case IPPROTO_UDP:
2455                                 PULLUP_TO(hlen, ulp, struct udphdr);
2456                                 dst_port = UDP(ulp)->uh_dport;
2457                                 src_port = UDP(ulp)->uh_sport;
2458                                 break;
2459
2460                         case IPPROTO_HOPOPTS:   /* RFC 2460 */
2461                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
2462                                 ext_hd |= EXT_HOPOPTS;
2463                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
2464                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
2465                                 ulp = NULL;
2466                                 break;
2467
2468                         case IPPROTO_ROUTING:   /* RFC 2460 */
2469                                 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
2470                                 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
2471                                 case 0:
2472                                         ext_hd |= EXT_RTHDR0;
2473                                         break;
2474                                 case 2:
2475                                         ext_hd |= EXT_RTHDR2;
2476                                         break;
2477                                 default:
2478                                         printf("IPFW2: IPV6 - Unknown Routing "
2479                                             "Header type(%d)\n",
2480                                             ((struct ip6_rthdr *)ulp)->ip6r_type);
2481                                         if (fw_deny_unknown_exthdrs)
2482                                             return (IP_FW_DENY);
2483                                         break;
2484                                 }
2485                                 ext_hd |= EXT_ROUTING;
2486                                 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
2487                                 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
2488                                 ulp = NULL;
2489                                 break;
2490
2491                         case IPPROTO_FRAGMENT:  /* RFC 2460 */
2492                                 PULLUP_TO(hlen, ulp, struct ip6_frag);
2493                                 ext_hd |= EXT_FRAGMENT;
2494                                 hlen += sizeof (struct ip6_frag);
2495                                 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
2496                                 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
2497                                         IP6F_OFF_MASK;
2498                                 /* Add IP6F_MORE_FRAG for offset of first
2499                                  * fragment to be != 0. */
2500                                 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg &
2501                                         IP6F_MORE_FRAG;
2502                                 if (offset == 0) {
2503                                         printf("IPFW2: IPV6 - Invalid Fragment "
2504                                             "Header\n");
2505                                         if (fw_deny_unknown_exthdrs)
2506                                             return (IP_FW_DENY);
2507                                         break;
2508                                 }
2509                                 args->f_id.frag_id6 =
2510                                     ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
2511                                 ulp = NULL;
2512                                 break;
2513
2514                         case IPPROTO_DSTOPTS:   /* RFC 2460 */
2515                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
2516                                 ext_hd |= EXT_DSTOPTS;
2517                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
2518                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
2519                                 ulp = NULL;
2520                                 break;
2521
2522                         case IPPROTO_AH:        /* RFC 2402 */
2523                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
2524                                 ext_hd |= EXT_AH;
2525                                 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
2526                                 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
2527                                 ulp = NULL;
2528                                 break;
2529
2530                         case IPPROTO_ESP:       /* RFC 2406 */
2531                                 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
2532                                 /* Anything past Seq# is variable length and
2533                                  * data past this ext. header is encrypted. */
2534                                 ext_hd |= EXT_ESP;
2535                                 break;
2536
2537                         case IPPROTO_NONE:      /* RFC 2460 */
2538                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
2539                                 /* Packet ends here. if ip6e_len!=0 octets
2540                                  * must be ignored. */
2541                                 break;
2542
2543                         case IPPROTO_OSPFIGP:
2544                                 /* XXX OSPF header check? */
2545                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
2546                                 break;
2547
2548                         case IPPROTO_PIM:
2549                                 /* XXX PIM header check? */
2550                                 PULLUP_TO(hlen, ulp, struct pim);
2551                                 break;
2552
2553                         case IPPROTO_CARP:
2554                                 PULLUP_TO(hlen, ulp, struct carp_header);
2555                                 if (((struct carp_header *)ulp)->carp_version !=
2556                                     CARP_VERSION)
2557                                         return (IP_FW_DENY);
2558                                 if (((struct carp_header *)ulp)->carp_type !=
2559                                     CARP_ADVERTISEMENT)
2560                                         return (IP_FW_DENY);
2561                                 break;
2562
2563                         case IPPROTO_IPV6:      /* RFC 2893 */
2564                                 PULLUP_TO(hlen, ulp, struct ip6_hdr);
2565                                 break;
2566
2567                         case IPPROTO_IPV4:      /* RFC 2893 */
2568                                 PULLUP_TO(hlen, ulp, struct ip);
2569                                 break;
2570
2571                         default:
2572                                 printf("IPFW2: IPV6 - Unknown Extension "
2573                                     "Header(%d), ext_hd=%x\n", proto, ext_hd);
2574                                 if (fw_deny_unknown_exthdrs)
2575                                     return (IP_FW_DENY);
2576                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
2577                                 break;
2578                         } /*switch */
2579                 }
2580                 ip = mtod(m, struct ip *);
2581                 ip6 = (struct ip6_hdr *)ip;
2582                 args->f_id.src_ip6 = ip6->ip6_src;
2583                 args->f_id.dst_ip6 = ip6->ip6_dst;
2584                 args->f_id.src_ip = 0;
2585                 args->f_id.dst_ip = 0;
2586                 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
2587         } else if (pktlen >= sizeof(struct ip) &&
2588             (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
2589                 is_ipv4 = 1;
2590                 hlen = ip->ip_hl << 2;
2591                 args->f_id.addr_type = 4;
2592
2593                 /*
2594                  * Collect parameters into local variables for faster matching.
2595                  */
2596                 proto = ip->ip_p;
2597                 src_ip = ip->ip_src;
2598                 dst_ip = ip->ip_dst;
2599                 if (args->eh != NULL) { /* layer 2 packets are as on the wire */
2600                         offset = ntohs(ip->ip_off) & IP_OFFMASK;
2601                         ip_len = ntohs(ip->ip_len);
2602                 } else {
2603                         offset = ip->ip_off & IP_OFFMASK;
2604                         ip_len = ip->ip_len;
2605                 }
2606                 pktlen = ip_len < pktlen ? ip_len : pktlen;
2607
2608                 if (offset == 0) {
2609                         switch (proto) {
2610                         case IPPROTO_TCP:
2611                                 PULLUP_TO(hlen, ulp, struct tcphdr);
2612                                 dst_port = TCP(ulp)->th_dport;
2613                                 src_port = TCP(ulp)->th_sport;
2614                                 args->f_id.flags = TCP(ulp)->th_flags;
2615                                 break;
2616
2617                         case IPPROTO_UDP:
2618                                 PULLUP_TO(hlen, ulp, struct udphdr);
2619                                 dst_port = UDP(ulp)->uh_dport;
2620                                 src_port = UDP(ulp)->uh_sport;
2621                                 break;
2622
2623                         case IPPROTO_ICMP:
2624                                 PULLUP_TO(hlen, ulp, struct icmphdr);
2625                                 args->f_id.flags = ICMP(ulp)->icmp_type;
2626                                 break;
2627
2628                         default:
2629                                 break;
2630                         }
2631                 }
2632
2633                 ip = mtod(m, struct ip *);
2634                 args->f_id.src_ip = ntohl(src_ip.s_addr);
2635                 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
2636         }
2637 #undef PULLUP_TO
2638         if (proto) { /* we may have port numbers, store them */
2639                 args->f_id.proto = proto;
2640                 args->f_id.src_port = src_port = ntohs(src_port);
2641                 args->f_id.dst_port = dst_port = ntohs(dst_port);
2642         }
2643
2644         IPFW_RLOCK(chain);
2645         mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL);
2646         if (args->rule) {
2647                 /*
2648                  * Packet has already been tagged. Look for the next rule
2649                  * to restart processing.
2650                  *
2651                  * If fw_one_pass != 0 then just accept it.
2652                  * XXX should not happen here, but optimized out in
2653                  * the caller.
2654                  */
2655                 if (fw_one_pass) {
2656                         IPFW_RUNLOCK(chain);
2657                         return (IP_FW_PASS);
2658                 }
2659
2660                 f = args->rule->next_rule;
2661                 if (f == NULL)
2662                         f = lookup_next_rule(args->rule);
2663         } else {
2664                 /*
2665                  * Find the starting rule. It can be either the first
2666                  * one, or the one after divert_rule if asked so.
2667                  */
2668                 int skipto = mtag ? divert_cookie(mtag) : 0;
2669
2670                 f = chain->rules;
2671                 if (args->eh == NULL && skipto != 0) {
2672                         if (skipto >= IPFW_DEFAULT_RULE) {
2673                                 IPFW_RUNLOCK(chain);
2674                                 return (IP_FW_DENY); /* invalid */
2675                         }
2676                         while (f && f->rulenum <= skipto)
2677                                 f = f->next;
2678                         if (f == NULL) {        /* drop packet */
2679                                 IPFW_RUNLOCK(chain);
2680                                 return (IP_FW_DENY);
2681                         }
2682                 }
2683         }
2684         /* reset divert rule to avoid confusion later */
2685         if (mtag) {
2686                 divinput_flags = divert_info(mtag) &
2687                     (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG);
2688                 m_tag_delete(m, mtag);
2689         }
2690
2691         /*
2692          * Now scan the rules, and parse microinstructions for each rule.
2693          */
2694         for (; f; f = f->next) {
2695                 ipfw_insn *cmd;
2696                 uint32_t tablearg = 0;
2697                 int l, cmdlen, skip_or; /* skip rest of OR block */
2698
2699 again:
2700                 if (set_disable & (1 << f->set) )
2701                         continue;
2702
2703                 skip_or = 0;
2704                 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
2705                     l -= cmdlen, cmd += cmdlen) {
2706                         int match;
2707
2708                         /*
2709                          * check_body is a jump target used when we find a
2710                          * CHECK_STATE, and need to jump to the body of
2711                          * the target rule.
2712                          */
2713
2714 check_body:
2715                         cmdlen = F_LEN(cmd);
2716                         /*
2717                          * An OR block (insn_1 || .. || insn_n) has the
2718                          * F_OR bit set in all but the last instruction.
2719                          * The first match will set "skip_or", and cause
2720                          * the following instructions to be skipped until
2721                          * past the one with the F_OR bit clear.
2722                          */
2723                         if (skip_or) {          /* skip this instruction */
2724                                 if ((cmd->len & F_OR) == 0)
2725                                         skip_or = 0;    /* next one is good */
2726                                 continue;
2727                         }
2728                         match = 0; /* set to 1 if we succeed */
2729
2730                         switch (cmd->opcode) {
2731                         /*
2732                          * The first set of opcodes compares the packet's
2733                          * fields with some pattern, setting 'match' if a
2734                          * match is found. At the end of the loop there is
2735                          * logic to deal with F_NOT and F_OR flags associated
2736                          * with the opcode.
2737                          */
2738                         case O_NOP:
2739                                 match = 1;
2740                                 break;
2741
2742                         case O_FORWARD_MAC:
2743                                 printf("ipfw: opcode %d unimplemented\n",
2744                                     cmd->opcode);
2745                                 break;
2746
2747                         case O_GID:
2748                         case O_UID:
2749                         case O_JAIL:
2750                                 /*
2751                                  * We only check offset == 0 && proto != 0,
2752                                  * as this ensures that we have a
2753                                  * packet with the ports info.
2754                                  */
2755                                 if (offset!=0)
2756                                         break;
2757                                 if (is_ipv6) /* XXX to be fixed later */
2758                                         break;
2759                                 if (proto == IPPROTO_TCP ||
2760                                     proto == IPPROTO_UDP)
2761                                         match = check_uidgid(
2762                                                     (ipfw_insn_u32 *)cmd,
2763                                                     proto, oif,
2764                                                     dst_ip, dst_port,
2765                                                     src_ip, src_port, &fw_ugid_cache,
2766                                                     &ugid_lookup, args->inp);
2767                                 break;
2768
2769                         case O_RECV:
2770                                 match = iface_match(m->m_pkthdr.rcvif,
2771                                     (ipfw_insn_if *)cmd);
2772                                 break;
2773
2774                         case O_XMIT:
2775                                 match = iface_match(oif, (ipfw_insn_if *)cmd);
2776                                 break;
2777
2778                         case O_VIA:
2779                                 match = iface_match(oif ? oif :
2780                                     m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd);
2781                                 break;
2782
2783                         case O_MACADDR2:
2784                                 if (args->eh != NULL) { /* have MAC header */
2785                                         u_int32_t *want = (u_int32_t *)
2786                                                 ((ipfw_insn_mac *)cmd)->addr;
2787                                         u_int32_t *mask = (u_int32_t *)
2788                                                 ((ipfw_insn_mac *)cmd)->mask;
2789                                         u_int32_t *hdr = (u_int32_t *)args->eh;
2790
2791                                         match =
2792                                             ( want[0] == (hdr[0] & mask[0]) &&
2793                                               want[1] == (hdr[1] & mask[1]) &&
2794                                               want[2] == (hdr[2] & mask[2]) );
2795                                 }
2796                                 break;
2797
2798                         case O_MAC_TYPE:
2799                                 if (args->eh != NULL) {
2800                                         u_int16_t *p =
2801                                             ((ipfw_insn_u16 *)cmd)->ports;
2802                                         int i;
2803
2804                                         for (i = cmdlen - 1; !match && i>0;
2805                                             i--, p += 2)
2806                                                 match = (etype >= p[0] &&
2807                                                     etype <= p[1]);
2808                                 }
2809                                 break;
2810
2811                         case O_FRAG:
2812                                 match = (offset != 0);
2813                                 break;
2814
2815                         case O_IN:      /* "out" is "not in" */
2816                                 match = (oif == NULL);
2817                                 break;
2818
2819                         case O_LAYER2:
2820                                 match = (args->eh != NULL);
2821                                 break;
2822
2823                         case O_DIVERTED:
2824                                 match = (cmd->arg1 & 1 && divinput_flags &
2825                                     IP_FW_DIVERT_LOOPBACK_FLAG) ||
2826                                         (cmd->arg1 & 2 && divinput_flags &
2827                                     IP_FW_DIVERT_OUTPUT_FLAG);
2828                                 break;
2829
2830                         case O_PROTO:
2831                                 /*
2832                                  * We do not allow an arg of 0 so the
2833                                  * check of "proto" only suffices.
2834                                  */
2835                                 match = (proto == cmd->arg1);
2836                                 break;
2837
2838                         case O_IP_SRC:
2839                                 match = is_ipv4 &&
2840                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2841                                     src_ip.s_addr);
2842                                 break;
2843
2844                         case O_IP_SRC_LOOKUP:
2845                         case O_IP_DST_LOOKUP:
2846                                 if (is_ipv4) {
2847                                     uint32_t a =
2848                                         (cmd->opcode == O_IP_DST_LOOKUP) ?
2849                                             dst_ip.s_addr : src_ip.s_addr;
2850                                     uint32_t v;
2851
2852                                     match = lookup_table(chain, cmd->arg1, a,
2853                                         &v);
2854                                     if (!match)
2855                                         break;
2856                                     if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
2857                                         match =
2858                                             ((ipfw_insn_u32 *)cmd)->d[0] == v;
2859                                     else
2860                                         tablearg = v;
2861                                 }
2862                                 break;
2863
2864                         case O_IP_SRC_MASK:
2865                         case O_IP_DST_MASK:
2866                                 if (is_ipv4) {
2867                                     uint32_t a =
2868                                         (cmd->opcode == O_IP_DST_MASK) ?
2869                                             dst_ip.s_addr : src_ip.s_addr;
2870                                     uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2871                                     int i = cmdlen-1;
2872
2873                                     for (; !match && i>0; i-= 2, p+= 2)
2874                                         match = (p[0] == (a & p[1]));
2875                                 }
2876                                 break;
2877
2878                         case O_IP_SRC_ME:
2879                                 if (is_ipv4) {
2880                                         struct ifnet *tif;
2881
2882                                         INADDR_TO_IFP(src_ip, tif);
2883                                         match = (tif != NULL);
2884                                 }
2885                                 break;
2886
2887                         case O_IP_DST_SET:
2888                         case O_IP_SRC_SET:
2889                                 if (is_ipv4) {
2890                                         u_int32_t *d = (u_int32_t *)(cmd+1);
2891                                         u_int32_t addr =
2892                                             cmd->opcode == O_IP_DST_SET ?
2893                                                 args->f_id.dst_ip :
2894                                                 args->f_id.src_ip;
2895
2896                                             if (addr < d[0])
2897                                                     break;
2898                                             addr -= d[0]; /* subtract base */
2899                                             match = (addr < cmd->arg1) &&
2900                                                 ( d[ 1 + (addr>>5)] &
2901                                                   (1<<(addr & 0x1f)) );
2902                                 }
2903                                 break;
2904
2905                         case O_IP_DST:
2906                                 match = is_ipv4 &&
2907                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2908                                     dst_ip.s_addr);
2909                                 break;
2910
2911                         case O_IP_DST_ME:
2912                                 if (is_ipv4) {
2913                                         struct ifnet *tif;
2914
2915                                         INADDR_TO_IFP(dst_ip, tif);
2916                                         match = (tif != NULL);
2917                                 }
2918                                 break;
2919
2920                         case O_IP_SRCPORT:
2921                         case O_IP_DSTPORT:
2922                                 /*
2923                                  * offset == 0 && proto != 0 is enough
2924                                  * to guarantee that we have a
2925                                  * packet with port info.
2926                                  */
2927                                 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
2928                                     && offset == 0) {
2929                                         u_int16_t x =
2930                                             (cmd->opcode == O_IP_SRCPORT) ?
2931                                                 src_port : dst_port ;
2932                                         u_int16_t *p =
2933                                             ((ipfw_insn_u16 *)cmd)->ports;
2934                                         int i;
2935
2936                                         for (i = cmdlen - 1; !match && i>0;
2937                                             i--, p += 2)
2938                                                 match = (x>=p[0] && x<=p[1]);
2939                                 }
2940                                 break;
2941
2942                         case O_ICMPTYPE:
2943                                 match = (offset == 0 && proto==IPPROTO_ICMP &&
2944                                     icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
2945                                 break;
2946
2947 #ifdef INET6
2948                         case O_ICMP6TYPE:
2949                                 match = is_ipv6 && offset == 0 &&
2950                                     proto==IPPROTO_ICMPV6 &&
2951                                     icmp6type_match(
2952                                         ICMP6(ulp)->icmp6_type,
2953                                         (ipfw_insn_u32 *)cmd);
2954                                 break;
2955 #endif /* INET6 */
2956
2957                         case O_IPOPT:
2958                                 match = (is_ipv4 &&
2959                                     ipopts_match(ip, cmd) );
2960                                 break;
2961
2962                         case O_IPVER:
2963                                 match = (is_ipv4 &&
2964                                     cmd->arg1 == ip->ip_v);
2965                                 break;
2966
2967                         case O_IPID:
2968                         case O_IPLEN:
2969                         case O_IPTTL:
2970                                 if (is_ipv4) {  /* only for IP packets */
2971                                     uint16_t x;
2972                                     uint16_t *p;
2973                                     int i;
2974
2975                                     if (cmd->opcode == O_IPLEN)
2976                                         x = ip_len;
2977                                     else if (cmd->opcode == O_IPTTL)
2978                                         x = ip->ip_ttl;
2979                                     else /* must be IPID */
2980                                         x = ntohs(ip->ip_id);
2981                                     if (cmdlen == 1) {
2982                                         match = (cmd->arg1 == x);
2983                                         break;
2984                                     }
2985                                     /* otherwise we have ranges */
2986                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2987                                     i = cmdlen - 1;
2988                                     for (; !match && i>0; i--, p += 2)
2989                                         match = (x >= p[0] && x <= p[1]);
2990                                 }
2991                                 break;
2992
2993                         case O_IPPRECEDENCE:
2994                                 match = (is_ipv4 &&
2995                                     (cmd->arg1 == (ip->ip_tos & 0xe0)) );
2996                                 break;
2997
2998                         case O_IPTOS:
2999                                 match = (is_ipv4 &&
3000                                     flags_match(cmd, ip->ip_tos));
3001                                 break;
3002
3003                         case O_TCPDATALEN:
3004                                 if (proto == IPPROTO_TCP && offset == 0) {
3005                                     struct tcphdr *tcp;
3006                                     uint16_t x;
3007                                     uint16_t *p;
3008                                     int i;
3009
3010                                     tcp = TCP(ulp);
3011                                     x = ip_len -
3012                                         ((ip->ip_hl + tcp->th_off) << 2);
3013                                     if (cmdlen == 1) {
3014                                         match = (cmd->arg1 == x);
3015                                         break;
3016                                     }
3017                                     /* otherwise we have ranges */
3018                                     p = ((ipfw_insn_u16 *)cmd)->ports;
3019                                     i = cmdlen - 1;
3020                                     for (; !match && i>0; i--, p += 2)
3021                                         match = (x >= p[0] && x <= p[1]);
3022                                 }
3023                                 break;
3024
3025                         case O_TCPFLAGS:
3026                                 match = (proto == IPPROTO_TCP && offset == 0 &&
3027                                     flags_match(cmd, TCP(ulp)->th_flags));
3028                                 break;
3029
3030                         case O_TCPOPTS:
3031                                 match = (proto == IPPROTO_TCP && offset == 0 &&
3032                                     tcpopts_match(TCP(ulp), cmd));
3033                                 break;
3034
3035                         case O_TCPSEQ:
3036                                 match = (proto == IPPROTO_TCP && offset == 0 &&
3037                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
3038                                         TCP(ulp)->th_seq);
3039                                 break;
3040
3041                         case O_TCPACK:
3042                                 match = (proto == IPPROTO_TCP && offset == 0 &&
3043                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
3044                                         TCP(ulp)->th_ack);
3045                                 break;
3046
3047                         case O_TCPWIN:
3048                                 match = (proto == IPPROTO_TCP && offset == 0 &&
3049                                     cmd->arg1 == TCP(ulp)->th_win);
3050                                 break;
3051
3052                         case O_ESTAB:
3053                                 /* reject packets which have SYN only */
3054                                 /* XXX should i also check for TH_ACK ? */
3055                                 match = (proto == IPPROTO_TCP && offset == 0 &&
3056                                     (TCP(ulp)->th_flags &
3057                                      (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
3058                                 break;
3059
3060                         case O_ALTQ: {
3061                                 struct altq_tag *at;
3062                                 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
3063
3064                                 match = 1;
3065                                 mtag = m_tag_find(m, PACKET_TAG_PF_QID, NULL);
3066                                 if (mtag != NULL)
3067                                         break;
3068                                 mtag = m_tag_get(PACKET_TAG_PF_QID,
3069                                                 sizeof(struct altq_tag),
3070                                                 M_NOWAIT);
3071                                 if (mtag == NULL) {
3072                                         /*
3073                                          * Let the packet fall back to the
3074                                          * default ALTQ.
3075                                          */
3076                                         break;
3077                                 }
3078                                 at = (struct altq_tag *)(mtag+1);
3079                                 at->qid = altq->qid;
3080                                 if (is_ipv4)
3081                                         at->af = AF_INET;
3082                                 else
3083                                         at->af = AF_LINK;
3084                                 at->hdr = ip;
3085                                 m_tag_prepend(m, mtag);
3086                                 break;
3087                         }
3088
3089                         case O_LOG:
3090                                 if (fw_verbose)
3091                                         ipfw_log(f, hlen, args, m,
3092                                             oif, offset, tablearg, ip);
3093                                 match = 1;
3094                                 break;
3095
3096                         case O_PROB:
3097                                 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
3098                                 break;
3099
3100                         case O_VERREVPATH:
3101                                 /* Outgoing packets automatically pass/match */
3102                                 match = ((oif != NULL) ||
3103                                     (m->m_pkthdr.rcvif == NULL) ||
3104                                     (
3105 #ifdef INET6
3106                                     is_ipv6 ?
3107                                         verify_path6(&(args->f_id.src_ip6),
3108                                             m->m_pkthdr.rcvif) :
3109 #endif
3110                                     verify_path(src_ip, m->m_pkthdr.rcvif)));
3111                                 break;
3112
3113                         case O_VERSRCREACH:
3114                                 /* Outgoing packets automatically pass/match */
3115                                 match = (hlen > 0 && ((oif != NULL) ||
3116 #ifdef INET6
3117                                     is_ipv6 ?
3118                                         verify_path6(&(args->f_id.src_ip6),
3119                                             NULL) :
3120 #endif
3121                                     verify_path(src_ip, NULL)));
3122                                 break;
3123
3124                         case O_ANTISPOOF:
3125                                 /* Outgoing packets automatically pass/match */
3126                                 if (oif == NULL && hlen > 0 &&
3127                                     (  (is_ipv4 && in_localaddr(src_ip))
3128 #ifdef INET6
3129                                     || (is_ipv6 &&
3130                                         in6_localaddr(&(args->f_id.src_ip6)))
3131 #endif
3132                                     ))
3133                                         match =
3134 #ifdef INET6
3135                                             is_ipv6 ? verify_path6(
3136                                                 &(args->f_id.src_ip6),
3137                                                 m->m_pkthdr.rcvif) :
3138 #endif
3139                                             verify_path(src_ip,
3140                                                 m->m_pkthdr.rcvif);
3141                                 else
3142                                         match = 1;
3143                                 break;
3144
3145                         case O_IPSEC:
3146 #ifdef FAST_IPSEC
3147                                 match = (m_tag_find(m,
3148                                     PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
3149 #endif
3150                                 /* otherwise no match */
3151                                 break;
3152
3153 #ifdef INET6
3154                         case O_IP6_SRC:
3155                                 match = is_ipv6 &&
3156                                     IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
3157                                     &((ipfw_insn_ip6 *)cmd)->addr6);
3158                                 break;
3159
3160                         case O_IP6_DST:
3161                                 match = is_ipv6 &&
3162                                 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
3163                                     &((ipfw_insn_ip6 *)cmd)->addr6);
3164                                 break;
3165                         case O_IP6_SRC_MASK:
3166                         case O_IP6_DST_MASK:
3167                                 if (is_ipv6) {
3168                                         int i = cmdlen - 1;
3169                                         struct in6_addr p;
3170                                         struct in6_addr *d =
3171                                             &((ipfw_insn_ip6 *)cmd)->addr6;
3172
3173                                         for (; !match && i > 0; d += 2,
3174                                             i -= F_INSN_SIZE(struct in6_addr)
3175                                             * 2) {
3176                                                 p = (cmd->opcode ==
3177                                                     O_IP6_SRC_MASK) ?
3178                                                     args->f_id.src_ip6:
3179                                                     args->f_id.dst_ip6;
3180                                                 APPLY_MASK(&p, &d[1]);
3181                                                 match =
3182                                                     IN6_ARE_ADDR_EQUAL(&d[0],
3183                                                     &p);
3184                                         }
3185                                 }
3186                                 break;
3187
3188                         case O_IP6_SRC_ME:
3189                                 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6);
3190                                 break;
3191
3192                         case O_IP6_DST_ME:
3193                                 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6);
3194                                 break;
3195
3196                         case O_FLOW6ID:
3197                                 match = is_ipv6 &&
3198                                     flow6id_match(args->f_id.flow_id6,
3199                                     (ipfw_insn_u32 *) cmd);
3200                                 break;
3201
3202                         case O_EXT_HDR:
3203                                 match = is_ipv6 &&
3204                                     (ext_hd & ((ipfw_insn *) cmd)->arg1);
3205                                 break;
3206
3207                         case O_IP6:
3208                                 match = is_ipv6;
3209                                 break;
3210 #endif
3211
3212                         case O_IP4:
3213                                 match = is_ipv4;
3214                                 break;
3215
3216                         case O_TAG: {
3217                                 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
3218                                     tablearg : cmd->arg1;
3219
3220                                 /* Packet is already tagged with this tag? */
3221                                 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
3222
3223                                 /* We have `untag' action when F_NOT flag is
3224                                  * present. And we must remove this mtag from
3225                                  * mbuf and reset `match' to zero (`match' will
3226                                  * be inversed later).
3227                                  * Otherwise we should allocate new mtag and
3228                                  * push it into mbuf.
3229                                  */
3230                                 if (cmd->len & F_NOT) { /* `untag' action */
3231                                         if (mtag != NULL)
3232                                                 m_tag_delete(m, mtag);
3233                                 } else if (mtag == NULL) {
3234                                         if ((mtag = m_tag_alloc(MTAG_IPFW,
3235                                             tag, 0, M_NOWAIT)) != NULL)
3236                                                 m_tag_prepend(m, mtag);
3237                                 }
3238                                 match = (cmd->len & F_NOT) ? 0: 1;
3239                                 break;
3240                         }
3241
3242                         case O_TAGGED: {
3243                                 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
3244                                     tablearg : cmd->arg1;
3245
3246                                 if (cmdlen == 1) {
3247                                         match = m_tag_locate(m, MTAG_IPFW,
3248                                             tag, NULL) != NULL;
3249                                         break;
3250                                 }
3251
3252                                 /* we have ranges */
3253                                 for (mtag = m_tag_first(m);
3254                                     mtag != NULL && !match;
3255                                     mtag = m_tag_next(m, mtag)) {
3256                                         uint16_t *p;
3257                                         int i;
3258
3259                                         if (mtag->m_tag_cookie != MTAG_IPFW)
3260                                                 continue;
3261
3262                                         p = ((ipfw_insn_u16 *)cmd)->ports;
3263                                         i = cmdlen - 1;
3264                                         for(; !match && i > 0; i--, p += 2)
3265                                                 match =
3266                                                     mtag->m_tag_id >= p[0] &&
3267                                                     mtag->m_tag_id <= p[1];
3268                                 }
3269                                 break;
3270                         }
3271
3272                         /*
3273                          * The second set of opcodes represents 'actions',
3274                          * i.e. the terminal part of a rule once the packet
3275                          * matches all previous patterns.
3276                          * Typically there is only one action for each rule,
3277                          * and the opcode is stored at the end of the rule
3278                          * (but there are exceptions -- see below).
3279                          *
3280                          * In general, here we set retval and terminate the
3281                          * outer loop (would be a 'break 3' in some language,
3282                          * but we need to do a 'goto done').
3283                          *
3284                          * Exceptions:
3285                          * O_COUNT and O_SKIPTO actions:
3286                          *   instead of terminating, we jump to the next rule
3287                          *   ('goto next_rule', equivalent to a 'break 2'),
3288                          *   or to the SKIPTO target ('goto again' after
3289                          *   having set f, cmd and l), respectively.
3290                          *
3291                          * O_TAG, O_LOG and O_ALTQ action parameters:
3292                          *   perform some action and set match = 1;
3293                          *
3294                          * O_LIMIT and O_KEEP_STATE: these opcodes are
3295                          *   not real 'actions', and are stored right
3296                          *   before the 'action' part of the rule.
3297                          *   These opcodes try to install an entry in the
3298                          *   state tables; if successful, we continue with
3299                          *   the next opcode (match=1; break;), otherwise
3300                          *   the packet *   must be dropped
3301                          *   ('goto done' after setting retval);
3302                          *
3303                          * O_PROBE_STATE and O_CHECK_STATE: these opcodes
3304                          *   cause a lookup of the state table, and a jump
3305                          *   to the 'action' part of the parent rule
3306                          *   ('goto check_body') if an entry is found, or
3307                          *   (CHECK_STATE only) a jump to the next rule if
3308                          *   the entry is not found ('goto next_rule').
3309                          *   The result of the lookup is cached to make
3310                          *   further instances of these opcodes are
3311                          *   effectively NOPs.
3312                          */
3313                         case O_LIMIT:
3314                         case O_KEEP_STATE:
3315                                 if (install_state(f,
3316                                     (ipfw_insn_limit *)cmd, args, tablearg)) {
3317                                         retval = IP_FW_DENY;
3318                                         goto done; /* error/limit violation */
3319                                 }
3320                                 match = 1;
3321                                 break;
3322
3323                         case O_PROBE_STATE:
3324                         case O_CHECK_STATE:
3325                                 /*
3326                                  * dynamic rules are checked at the first
3327                                  * keep-state or check-state occurrence,
3328                                  * with the result being stored in dyn_dir.
3329                                  * The compiler introduces a PROBE_STATE
3330                                  * instruction for us when we have a
3331                                  * KEEP_STATE (because PROBE_STATE needs
3332                                  * to be run first).
3333                                  */
3334                                 if (dyn_dir == MATCH_UNKNOWN &&
3335                                     (q = lookup_dyn_rule(&args->f_id,
3336                                      &dyn_dir, proto == IPPROTO_TCP ?
3337                                         TCP(ulp) : NULL))
3338                                         != NULL) {
3339                                         /*
3340                                          * Found dynamic entry, update stats
3341                                          * and jump to the 'action' part of
3342                                          * the parent rule.
3343                                          */
3344                                         q->pcnt++;
3345                                         q->bcnt += pktlen;
3346                                         f = q->rule;
3347                                         cmd = ACTION_PTR(f);
3348                                         l = f->cmd_len - f->act_ofs;
3349                                         IPFW_DYN_UNLOCK();
3350                                         goto check_body;
3351                                 }
3352                                 /*
3353                                  * Dynamic entry not found. If CHECK_STATE,
3354                                  * skip to next rule, if PROBE_STATE just
3355                                  * ignore and continue with next opcode.
3356                                  */
3357                                 if (cmd->opcode == O_CHECK_STATE)
3358                                         goto next_rule;
3359                                 match = 1;
3360                                 break;
3361
3362                         case O_ACCEPT:
3363                                 retval = 0;     /* accept */
3364                                 goto done;
3365
3366                         case O_PIPE:
3367                         case O_QUEUE:
3368                                 args->rule = f; /* report matching rule */
3369                                 if (cmd->arg1 == IP_FW_TABLEARG)
3370                                         args->cookie = tablearg;
3371                                 else
3372                                         args->cookie = cmd->arg1;
3373                                 retval = IP_FW_DUMMYNET;
3374                                 goto done;
3375
3376                         case O_DIVERT:
3377                         case O_TEE: {
3378                                 struct divert_tag *dt;
3379
3380                                 if (args->eh) /* not on layer 2 */
3381                                         break;
3382                                 mtag = m_tag_get(PACKET_TAG_DIVERT,
3383                                                 sizeof(struct divert_tag),
3384                                                 M_NOWAIT);
3385                                 if (mtag == NULL) {
3386                                         /* XXX statistic */
3387                                         /* drop packet */
3388                                         IPFW_RUNLOCK(chain);
3389                                         return (IP_FW_DENY);
3390                                 }
3391                                 dt = (struct divert_tag *)(mtag+1);
3392                                 dt->cookie = f->rulenum;
3393                                 if (cmd->arg1 == IP_FW_TABLEARG)
3394                                         dt->info = tablearg;
3395                                 else
3396                                         dt->info = cmd->arg1;
3397                                 m_tag_prepend(m, mtag);
3398                                 retval = (cmd->opcode == O_DIVERT) ?
3399                                     IP_FW_DIVERT : IP_FW_TEE;
3400                                 goto done;
3401                         }
3402
3403                         case O_COUNT:
3404                         case O_SKIPTO:
3405                                 f->pcnt++;      /* update stats */
3406                                 f->bcnt += pktlen;
3407                                 f->timestamp = time_uptime;
3408                                 if (cmd->opcode == O_COUNT)
3409                                         goto next_rule;
3410                                 /* handle skipto */
3411                                 if (f->next_rule == NULL)
3412                                         lookup_next_rule(f);
3413                                 f = f->next_rule;
3414                                 goto again;
3415
3416                         case O_REJECT:
3417                                 /*
3418                                  * Drop the packet and send a reject notice
3419                                  * if the packet is not ICMP (or is an ICMP
3420                                  * query), and it is not multicast/broadcast.
3421                                  */
3422                                 if (hlen > 0 && is_ipv4 && offset == 0 &&
3423                                     (proto != IPPROTO_ICMP ||
3424                                      is_icmp_query(ICMP(ulp))) &&
3425                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
3426                                     !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
3427                                         send_reject(args, cmd->arg1, ip_len, ip);
3428                                         m = args->m;
3429                                 }
3430                                 /* FALLTHROUGH */
3431 #ifdef INET6
3432                         case O_UNREACH6:
3433                                 if (hlen > 0 && is_ipv6 &&
3434                                     ((offset & IP6F_OFF_MASK) == 0) &&
3435                                     (proto != IPPROTO_ICMPV6 ||
3436                                      (is_icmp6_query(args->f_id.flags) == 1)) &&
3437                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
3438                                     !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
3439                                         send_reject6(
3440                                             args, cmd->arg1, hlen,
3441                                             (struct ip6_hdr *)ip);
3442                                         m = args->m;
3443                                 }
3444                                 /* FALLTHROUGH */
3445 #endif
3446                         case O_DENY:
3447                                 retval = IP_FW_DENY;
3448                                 goto done;
3449
3450                         case O_FORWARD_IP: {
3451                                 struct sockaddr_in *sa;
3452                                 sa = &(((ipfw_insn_sa *)cmd)->sa);
3453                                 if (args->eh)   /* not valid on layer2 pkts */
3454                                         break;
3455                                 if (!q || dyn_dir == MATCH_FORWARD) {
3456                                         if (sa->sin_addr.s_addr == INADDR_ANY) {
3457                                                 bcopy(sa, &args->hopstore,
3458                                                         sizeof(*sa));
3459                                                 args->hopstore.sin_addr.s_addr =
3460                                                     htonl(tablearg);
3461                                                 args->next_hop =
3462                                                     &args->hopstore;
3463                                         } else {
3464                                                 args->next_hop = sa;
3465                                         }
3466                                 }
3467                                 retval = IP_FW_PASS;
3468                             }
3469                             goto done;
3470
3471                         case O_NETGRAPH:
3472                         case O_NGTEE:
3473                                 args->rule = f; /* report matching rule */
3474                                 if (cmd->arg1 == IP_FW_TABLEARG)
3475                                         args->cookie = tablearg;
3476                                 else
3477                                         args->cookie = cmd->arg1;
3478                                 retval = (cmd->opcode == O_NETGRAPH) ?
3479                                     IP_FW_NETGRAPH : IP_FW_NGTEE;
3480                                 goto done;
3481
3482 #ifdef IPFIREWALL_NAT
3483                         case O_NAT: {
3484                                 struct cfg_nat *t;
3485                                 struct mbuf *mcl;
3486                                 /* XXX - libalias duct tape */
3487                                 int ldt;
3488                                 char *c;
3489
3490                                 ldt = 0;
3491                                 args->rule = f; /* Report matching rule. */
3492                                 retval = 0;
3493                                 t = ((ipfw_insn_nat *)cmd)->nat;
3494                                 if (t == NULL) {
3495                                         t = lookup_nat(cmd->arg1);
3496                                         if (t == NULL) {
3497                                                 retval = IP_FW_DENY;
3498                                                 goto done;
3499                                         } else
3500                                                 ((ipfw_insn_nat *)cmd)->nat =
3501                                                     t;
3502                                 }
3503                                 if ((mcl = m_megapullup(m, m->m_pkthdr.len)) ==
3504                                     NULL)
3505                                         goto badnat;
3506                                 ip = mtod(mcl, struct ip *);
3507                                 if (args->eh == NULL) {
3508                                         ip->ip_len = htons(ip->ip_len);
3509                                         ip->ip_off = htons(ip->ip_off);
3510                                 }
3511
3512                                 /*
3513                                  * XXX - Libalias checksum offload 'duct tape':
3514                                  *
3515                                  * locally generated packets have only
3516                                  * pseudo-header checksum calculated
3517                                  * and libalias will screw it[1], so
3518                                  * mark them for later fix.  Moreover
3519                                  * there are cases when libalias
3520                                  * modify tcp packet data[2], mark it
3521                                  * for later fix too.
3522                                  *
3523                                  * [1] libalias was never meant to run
3524                                  * in kernel, so it doesn't have any
3525                                  * knowledge about checksum
3526                                  * offloading, and it expects a packet
3527                                  * with a full internet
3528                                  * checksum. Unfortunately, packets
3529                                  * generated locally will have just the
3530                                  * pseudo header calculated, and when
3531                                  * libalias tries to adjust the
3532                                  * checksum it will actually screw it.
3533                                  *
3534                                  * [2] when libalias modify tcp's data
3535                                  * content, full TCP checksum has to
3536                                  * be recomputed: the problem is that
3537                                  * libalias doesn't have any idea
3538                                  * about checksum offloading To
3539                                  * workaround this, we do not do
3540                                  * checksumming in LibAlias, but only
3541                                  * mark the packets in th_x2 field. If
3542                                  * we receive a marked packet, we
3543                                  * calculate correct checksum for it
3544                                  * aware of offloading.  Why such a
3545                                  * terrible hack instead of
3546                                  * recalculating checksum for each
3547                                  * packet?  Because the previous
3548                                  * checksum was not checked!
3549                                  * Recalculating checksums for EVERY
3550                                  * packet will hide ALL transmission
3551                                  * errors. Yes, marked packets still
3552                                  * suffer from this problem. But,
3553                                  * sigh, natd(8) has this problem,
3554                                  * too.
3555                                  *
3556                                  * TODO: -make libalias mbuf aware (so
3557                                  * it can handle delayed checksum and tso)
3558                                  */
3559
3560                                 if (mcl->m_pkthdr.rcvif == NULL &&
3561                                     mcl->m_pkthdr.csum_flags &
3562                                     CSUM_DELAY_DATA)
3563                                         ldt = 1;
3564
3565                                 c = mtod(mcl, char *);
3566                                 if (oif == NULL)
3567                                         retval = LibAliasIn(t->lib, c,
3568                                             MCLBYTES);
3569                                 else
3570                                         retval = LibAliasOut(t->lib, c,
3571                                             MCLBYTES);
3572                                 if (retval != PKT_ALIAS_OK) {
3573                                         /* XXX - should i add some logging? */
3574                                         m_free(mcl);
3575                                 badnat:
3576                                         args->m = NULL;
3577                                         retval = IP_FW_DENY;
3578                                         goto done;
3579                                 }
3580                                 mcl->m_pkthdr.len = mcl->m_len =
3581                                     ntohs(ip->ip_len);
3582
3583                                 /*
3584                                  * XXX - libalias checksum offload
3585                                  * 'duct tape' (see above)
3586                                  */
3587
3588                                 if ((ip->ip_off & htons(IP_OFFMASK)) == 0 &&
3589                                     ip->ip_p == IPPROTO_TCP) {
3590                                         struct tcphdr   *th;
3591
3592                                         th = (struct tcphdr *)(ip + 1);
3593                                         if (th->th_x2)
3594                                                 ldt = 1;
3595                                 }
3596
3597                                 if (ldt) {
3598                                         struct tcphdr   *th;
3599                                         struct udphdr   *uh;
3600                                         u_short cksum;
3601
3602                                         ip->ip_len = ntohs(ip->ip_len);
3603                                         cksum = in_pseudo(
3604                                                 ip->ip_src.s_addr,
3605                                                 ip->ip_dst.s_addr,
3606                                                 htons(ip->ip_p + ip->ip_len -
3607                                                     (ip->ip_hl << 2))
3608                                                 );
3609
3610                                         switch (ip->ip_p) {
3611                                         case IPPROTO_TCP:
3612                                                 th = (struct tcphdr *)(ip + 1);
3613                                                 /*
3614                                                  * Maybe it was set in
3615                                                  * libalias...
3616                                                  */
3617                                                 th->th_x2 = 0;
3618                                                 th->th_sum = cksum;
3619                                                 mcl->m_pkthdr.csum_data =
3620                                                     offsetof(struct tcphdr,
3621                                                     th_sum);
3622                                                 break;
3623                                         case IPPROTO_UDP:
3624                                                 uh = (struct udphdr *)(ip + 1);
3625                                                 uh->uh_sum = cksum;
3626                                                 mcl->m_pkthdr.csum_data =
3627                                                     offsetof(struct udphdr,
3628                                                     uh_sum);
3629                                                 break;
3630                                         }
3631                                         /*
3632                                          * No hw checksum offloading: do it
3633                                          * by ourself.
3634                                          */
3635                                         if ((mcl->m_pkthdr.csum_flags &
3636                                              CSUM_DELAY_DATA) == 0) {
3637                                                 in_delayed_cksum(mcl);
3638                                                 mcl->m_pkthdr.csum_flags &=
3639                                                     ~CSUM_DELAY_DATA;
3640                                         }
3641                                         ip->ip_len = htons(ip->ip_len);
3642                                 }
3643
3644                                 if (args->eh == NULL) {
3645                                         ip->ip_len = ntohs(ip->ip_len);
3646                                         ip->ip_off = ntohs(ip->ip_off);
3647                                 }
3648
3649                                 args->m = mcl;
3650                                 retval = IP_FW_NAT;
3651                                 goto done;
3652                         }
3653 #endif
3654
3655                         default:
3656                                 panic("-- unknown opcode %d\n", cmd->opcode);
3657                         } /* end of switch() on opcodes */
3658
3659                         if (cmd->len & F_NOT)
3660                                 match = !match;
3661
3662                         if (match) {
3663                                 if (cmd->len & F_OR)
3664                                         skip_or = 1;
3665                         } else {
3666                                 if (!(cmd->len & F_OR)) /* not an OR block, */
3667                                         break;          /* try next rule    */
3668                         }
3669
3670                 }       /* end of inner for, scan opcodes */
3671
3672 next_rule:;             /* try next rule                */
3673
3674         }               /* end of outer for, scan rules */
3675         printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3676         IPFW_RUNLOCK(chain);
3677         return (IP_FW_DENY);
3678
3679 done:
3680         /* Update statistics */
3681         f->pcnt++;
3682         f->bcnt += pktlen;
3683         f->timestamp = time_uptime;
3684         IPFW_RUNLOCK(chain);
3685         return (retval);
3686
3687 pullup_failed:
3688         if (fw_verbose)
3689                 printf("ipfw: pullup failed\n");
3690         return (IP_FW_DENY);
3691 }
3692
3693 /*
3694  * When a rule is added/deleted, clear the next_rule pointers in all rules.
3695  * These will be reconstructed on the fly as packets are matched.
3696  */
3697 static void
3698 flush_rule_ptrs(struct ip_fw_chain *chain)
3699 {
3700         struct ip_fw *rule;
3701
3702         IPFW_WLOCK_ASSERT(chain);
3703
3704         for (rule = chain->rules; rule; rule = rule->next)
3705                 rule->next_rule = NULL;
3706 }
3707
3708 /*
3709  * Add a new rule to the list. Copy the rule into a malloc'ed area, then
3710  * possibly create a rule number and add the rule to the list.
3711  * Update the rule_number in the input struct so the caller knows it as well.
3712  */
3713 static int
3714 add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule)
3715 {
3716         struct ip_fw *rule, *f, *prev;
3717         int l = RULESIZE(input_rule);
3718
3719         if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE)
3720                 return (EINVAL);
3721
3722         rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO);
3723         if (rule == NULL)
3724                 return (ENOSPC);
3725
3726         bcopy(input_rule, rule, l);
3727
3728         rule->next = NULL;
3729         rule->next_rule = NULL;
3730
3731         rule->pcnt = 0;
3732         rule->bcnt = 0;
3733         rule->timestamp = 0;
3734
3735         IPFW_WLOCK(chain);
3736
3737         if (chain->rules == NULL) {     /* default rule */
3738                 chain->rules = rule;
3739                 goto done;
3740         }
3741
3742         /*
3743          * If rulenum is 0, find highest numbered rule before the
3744          * default rule, and add autoinc_step
3745          */
3746         if (autoinc_step < 1)
3747                 autoinc_step = 1;
3748         else if (autoinc_step > 1000)
3749                 autoinc_step = 1000;
3750         if (rule->rulenum == 0) {
3751                 /*
3752                  * locate the highest numbered rule before default
3753                  */
3754                 for (f = chain->rules; f; f = f->next) {
3755                         if (f->rulenum == IPFW_DEFAULT_RULE)
3756                                 break;
3757                         rule->rulenum = f->rulenum;
3758                 }
3759                 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step)
3760                         rule->rulenum += autoinc_step;
3761                 input_rule->rulenum = rule->rulenum;
3762         }
3763
3764         /*
3765          * Now insert the new rule in the right place in the sorted list.
3766          */
3767         for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) {
3768                 if (f->rulenum > rule->rulenum) { /* found the location */
3769                         if (prev) {
3770                                 rule->next = f;
3771                                 prev->next = rule;
3772                         } else { /* head insert */
3773                                 rule->next = chain->rules;
3774                                 chain->rules = rule;
3775                         }
3776                         break;
3777                 }
3778         }
3779         flush_rule_ptrs(chain);
3780 done:
3781         static_count++;
3782         static_len += l;
3783         IPFW_WUNLOCK(chain);
3784         DEB(printf("ipfw: installed rule %d, static count now %d\n",
3785                 rule->rulenum, static_count);)
3786         return (0);
3787 }
3788
3789 /**
3790  * Remove a static rule (including derived * dynamic rules)
3791  * and place it on the ``reap list'' for later reclamation.
3792  * The caller is in charge of clearing rule pointers to avoid
3793  * dangling pointers.
3794  * @return a pointer to the next entry.
3795  * Arguments are not checked, so they better be correct.
3796  */
3797 static struct ip_fw *
3798 remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule,
3799     struct ip_fw *prev)
3800 {
3801         struct ip_fw *n;
3802         int l = RULESIZE(rule);
3803
3804         IPFW_WLOCK_ASSERT(chain);
3805
3806         n = rule->next;
3807         IPFW_DYN_LOCK();
3808         remove_dyn_rule(rule, NULL /* force removal */);
3809         IPFW_DYN_UNLOCK();
3810         if (prev == NULL)
3811                 chain->rules = n;
3812         else
3813                 prev->next = n;
3814         static_count--;
3815         static_len -= l;
3816
3817         rule->next = chain->reap;
3818         chain->reap = rule;
3819
3820         return n;
3821 }
3822
3823 /**
3824  * Reclaim storage associated with a list of rules.  This is
3825  * typically the list created using remove_rule.
3826  */
3827 static void
3828 reap_rules(struct ip_fw *head)
3829 {
3830         struct ip_fw *rule;
3831
3832         while ((rule = head) != NULL) {
3833                 head = head->next;
3834                 if (DUMMYNET_LOADED)
3835                         ip_dn_ruledel_ptr(rule);
3836                 free(rule, M_IPFW);
3837         }
3838 }
3839
3840 /*
3841  * Remove all rules from a chain (except rules in set RESVD_SET
3842  * unless kill_default = 1).  The caller is responsible for
3843  * reclaiming storage for the rules left in chain->reap.
3844  */
3845 static void
3846 free_chain(struct ip_fw_chain *chain, int kill_default)
3847 {
3848         struct ip_fw *prev, *rule;
3849
3850         IPFW_WLOCK_ASSERT(chain);
3851
3852         flush_rule_ptrs(chain); /* more efficient to do outside the loop */
3853         for (prev = NULL, rule = chain->rules; rule ; )
3854                 if (kill_default || rule->set != RESVD_SET)
3855                         rule = remove_rule(chain, rule, prev);
3856                 else {
3857                         prev = rule;
3858                         rule = rule->next;
3859                 }
3860 }
3861
3862 /**
3863  * Remove all rules with given number, and also do set manipulation.
3864  * Assumes chain != NULL && *chain != NULL.
3865  *
3866  * The argument is an u_int32_t. The low 16 bit are the rule or set number,
3867  * the next 8 bits are the new set, the top 8 bits are the command:
3868  *
3869  *      0       delete rules with given number
3870  *      1       delete rules with given set number
3871  *      2       move rules with given number to new set
3872  *      3       move rules with given set number to new set
3873  *      4       swap sets with given numbers
3874  *      5       delete rules with given number and with given set number
3875  */
3876 static int
3877 del_entry(struct ip_fw_chain *chain, u_int32_t arg)
3878 {
3879         struct ip_fw *prev = NULL, *rule;
3880         u_int16_t rulenum;      /* rule or old_set */
3881         u_int8_t cmd, new_set;
3882
3883         rulenum = arg & 0xffff;
3884         cmd = (arg >> 24) & 0xff;
3885         new_set = (arg >> 16) & 0xff;
3886
3887         if (cmd > 5 || new_set > RESVD_SET)
3888                 return EINVAL;
3889         if (cmd == 0 || cmd == 2 || cmd == 5) {
3890                 if (rulenum >= IPFW_DEFAULT_RULE)
3891                         return EINVAL;
3892         } else {
3893                 if (rulenum > RESVD_SET)        /* old_set */
3894                         return EINVAL;
3895         }
3896
3897         IPFW_WLOCK(chain);
3898         rule = chain->rules;
3899         chain->reap = NULL;
3900         switch (cmd) {
3901         case 0: /* delete rules with given number */
3902                 /*
3903                  * locate first rule to delete
3904                  */
3905                 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next)
3906                         ;
3907                 if (rule->rulenum != rulenum) {
3908                         IPFW_WUNLOCK(chain);
3909                         return EINVAL;
3910                 }
3911
3912                 /*
3913                  * flush pointers outside the loop, then delete all matching
3914                  * rules. prev remains the same throughout the cycle.
3915                  */
3916                 flush_rule_ptrs(chain);
3917                 while (rule->rulenum == rulenum)
3918                         rule = remove_rule(chain, rule, prev);
3919                 break;
3920
3921         case 1: /* delete all rules with given set number */
3922                 flush_rule_ptrs(chain);
3923                 rule = chain->rules;
3924                 while (rule->rulenum < IPFW_DEFAULT_RULE)
3925                         if (rule->set == rulenum)
3926                                 rule = remove_rule(chain, rule, prev);
3927                         else {
3928                                 prev = rule;
3929                                 rule = rule->next;
3930                         }
3931                 break;
3932
3933         case 2: /* move rules with given number to new set */
3934                 rule = chain->rules;
3935                 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
3936                         if (rule->rulenum == rulenum)
3937                                 rule->set = new_set;
3938                 break;
3939
3940         case 3: /* move rules with given set number to new set */
3941                 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
3942                         if (rule->set == rulenum)
3943                                 rule->set = new_set;
3944                 break;
3945
3946         case 4: /* swap two sets */
3947                 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
3948                         if (rule->set == rulenum)
3949                                 rule->set = new_set;
3950                         else if (rule->set == new_set)
3951                                 rule->set = rulenum;
3952                 break;
3953         case 5: /* delete rules with given number and with given set number.
3954                  * rulenum - given rule number;
3955                  * new_set - given set number.
3956                  */
3957                 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next)
3958                         ;
3959                 if (rule->rulenum != rulenum) {
3960                         IPFW_WUNLOCK(chain);
3961                         return (EINVAL);
3962                 }
3963                 flush_rule_ptrs(chain);
3964                 while (rule->rulenum == rulenum) {
3965                         if (rule->set == new_set)
3966                                 rule = remove_rule(chain, rule, prev);
3967                         else {
3968                                 prev = rule;
3969                                 rule = rule->next;
3970                         }
3971                 }
3972         }
3973         /*
3974          * Look for rules to reclaim.  We grab the list before
3975          * releasing the lock then reclaim them w/o the lock to
3976          * avoid a LOR with dummynet.
3977          */
3978         rule = chain->reap;
3979         chain->reap = NULL;
3980         IPFW_WUNLOCK(chain);
3981         if (rule)
3982                 reap_rules(rule);
3983         return 0;
3984 }
3985
3986 /*
3987  * Clear counters for a specific rule.
3988  * The enclosing "table" is assumed locked.
3989  */
3990 static void
3991 clear_counters(struct ip_fw *rule, int log_only)
3992 {
3993         ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule);
3994
3995         if (log_only == 0) {
3996                 rule->bcnt = rule->pcnt = 0;
3997                 rule->timestamp = 0;
3998         }
3999         if (l->o.opcode == O_LOG)
4000                 l->log_left = l->max_log;
4001 }
4002
4003 /**
4004  * Reset some or all counters on firewall rules.
4005  * The argument `arg' is an u_int32_t. The low 16 bit are the rule number,
4006  * the next 8 bits are the set number, the top 8 bits are the command:
4007  *      0       work with rules from all set's;
4008  *      1       work with rules only from specified set.
4009  * Specified rule number is zero if we want to clear all entries.
4010  * log_only is 1 if we only want to reset logs, zero otherwise.
4011  */
4012 static int
4013 zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only)
4014 {
4015         struct ip_fw *rule;
4016         char *msg;
4017
4018         uint16_t rulenum = arg & 0xffff;
4019         uint8_t set = (arg >> 16) & 0xff;
4020         uint8_t cmd = (arg >> 24) & 0xff;
4021
4022         if (cmd > 1)
4023                 return (EINVAL);
4024         if (cmd == 1 && set > RESVD_SET)
4025                 return (EINVAL);
4026
4027         IPFW_WLOCK(chain);
4028         if (rulenum == 0) {
4029                 norule_counter = 0;
4030                 for (rule = chain->rules; rule; rule = rule->next) {
4031                         /* Skip rules from another set. */
4032                         if (cmd == 1 && rule->set != set)
4033                                 continue;
4034                         clear_counters(rule, log_only);
4035                 }
4036                 msg = log_only ? "ipfw: All logging counts reset.\n" :
4037                     "ipfw: Accounting cleared.\n";
4038         } else {
4039                 int cleared = 0;
4040                 /*
4041                  * We can have multiple rules with the same number, so we
4042                  * need to clear them all.
4043                  */
4044                 for (rule = chain->rules; rule; rule = rule->next)
4045                         if (rule->rulenum == rulenum) {
4046                                 while (rule && rule->rulenum == rulenum) {
4047                                         if (cmd == 0 || rule->set == set)
4048                                                 clear_counters(rule, log_only);
4049                                         rule = rule->next;
4050                                 }
4051                                 cleared = 1;
4052                                 break;
4053                         }
4054                 if (!cleared) { /* we did not find any matching rules */
4055                         IPFW_WUNLOCK(chain);
4056                         return (EINVAL);
4057                 }
4058                 msg = log_only ? "ipfw: Entry %d logging count reset.\n" :
4059                     "ipfw: Entry %d cleared.\n";
4060         }
4061         IPFW_WUNLOCK(chain);
4062
4063         if (fw_verbose)
4064                 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum);
4065         return (0);
4066 }
4067
4068 /*
4069  * Check validity of the structure before insert.
4070  * Fortunately rules are simple, so this mostly need to check rule sizes.
4071  */
4072 static int
4073 check_ipfw_struct(struct ip_fw *rule, int size)
4074 {
4075         int l, cmdlen = 0;
4076         int have_action=0;
4077         ipfw_insn *cmd;
4078
4079         if (size < sizeof(*rule)) {
4080                 printf("ipfw: rule too short\n");
4081                 return (EINVAL);
4082         }
4083         /* first, check for valid size */
4084         l = RULESIZE(rule);
4085         if (l != size) {
4086                 printf("ipfw: size mismatch (have %d want %d)\n", size, l);
4087                 return (EINVAL);
4088         }
4089         if (rule->act_ofs >= rule->cmd_len) {
4090                 printf("ipfw: bogus action offset (%u > %u)\n",
4091                     rule->act_ofs, rule->cmd_len - 1);
4092                 return (EINVAL);
4093         }
4094         /*
4095          * Now go for the individual checks. Very simple ones, basically only
4096          * instruction sizes.
4097          */
4098         for (l = rule->cmd_len, cmd = rule->cmd ;
4099                         l > 0 ; l -= cmdlen, cmd += cmdlen) {
4100                 cmdlen = F_LEN(cmd);
4101                 if (cmdlen > l) {
4102                         printf("ipfw: opcode %d size truncated\n",
4103                             cmd->opcode);
4104                         return EINVAL;
4105                 }
4106                 DEB(printf("ipfw: opcode %d\n", cmd->opcode);)
4107                 switch (cmd->opcode) {
4108                 case O_PROBE_STATE:
4109                 case O_KEEP_STATE:
4110                 case O_PROTO:
4111                 case O_IP_SRC_ME:
4112                 case O_IP_DST_ME:
4113                 case O_LAYER2:
4114                 case O_IN:
4115                 case O_FRAG:
4116                 case O_DIVERTED:
4117                 case O_IPOPT:
4118                 case O_IPTOS:
4119                 case O_IPPRECEDENCE:
4120                 case O_IPVER:
4121                 case O_TCPWIN:
4122                 case O_TCPFLAGS:
4123                 case O_TCPOPTS:
4124                 case O_ESTAB:
4125                 case O_VERREVPATH:
4126                 case O_VERSRCREACH:
4127                 case O_ANTISPOOF:
4128                 case O_IPSEC:
4129 #ifdef INET6
4130                 case O_IP6_SRC_ME:
4131                 case O_IP6_DST_ME:
4132                 case O_EXT_HDR:
4133                 case O_IP6:
4134 #endif
4135                 case O_IP4:
4136                 case O_TAG:
4137                         if (cmdlen != F_INSN_SIZE(ipfw_insn))
4138                                 goto bad_size;
4139                         break;
4140
4141                 case O_UID:
4142                 case O_GID:
4143                 case O_JAIL:
4144                 case O_IP_SRC:
4145                 case O_IP_DST:
4146                 case O_TCPSEQ:
4147                 case O_TCPACK:
4148                 case O_PROB:
4149                 case O_ICMPTYPE:
4150                         if (cmdlen != F_INSN_SIZE(ipfw_insn_u32))
4151                                 goto bad_size;
4152                         break;
4153
4154                 case O_LIMIT:
4155                         if (cmdlen != F_INSN_SIZE(ipfw_insn_limit))
4156                                 goto bad_size;
4157                         break;
4158
4159                 case O_LOG:
4160                         if (cmdlen != F_INSN_SIZE(ipfw_insn_log))
4161                                 goto bad_size;
4162
4163                         ((ipfw_insn_log *)cmd)->log_left =
4164                             ((ipfw_insn_log *)cmd)->max_log;
4165
4166                         break;
4167
4168                 case O_IP_SRC_MASK:
4169                 case O_IP_DST_MASK:
4170                         /* only odd command lengths */
4171                         if ( !(cmdlen & 1) || cmdlen > 31)
4172                                 goto bad_size;
4173                         break;
4174
4175                 case O_IP_SRC_SET:
4176                 case O_IP_DST_SET:
4177                         if (cmd->arg1 == 0 || cmd->arg1 > 256) {
4178                                 printf("ipfw: invalid set size %d\n",
4179                                         cmd->arg1);
4180                                 return EINVAL;
4181                         }
4182                         if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
4183                             (cmd->arg1+31)/32 )
4184                                 goto bad_size;
4185                         break;
4186
4187                 case O_IP_SRC_LOOKUP:
4188                 case O_IP_DST_LOOKUP:
4189                         if (cmd->arg1 >= IPFW_TABLES_MAX) {
4190                                 printf("ipfw: invalid table number %d\n",
4191                                     cmd->arg1);
4192                                 return (EINVAL);
4193                         }
4194                         if (cmdlen != F_INSN_SIZE(ipfw_insn) &&
4195                             cmdlen != F_INSN_SIZE(ipfw_insn_u32))
4196                                 goto bad_size;
4197                         break;
4198
4199                 case O_MACADDR2:
4200                         if (cmdlen != F_INSN_SIZE(ipfw_insn_mac))
4201                                 goto bad_size;
4202                         break;
4203
4204                 case O_NOP:
4205                 case O_IPID:
4206                 case O_IPTTL:
4207                 case O_IPLEN:
4208                 case O_TCPDATALEN:
4209                 case O_TAGGED:
4210                         if (cmdlen < 1 || cmdlen > 31)
4211                                 goto bad_size;
4212                         break;
4213
4214                 case O_MAC_TYPE:
4215                 case O_IP_SRCPORT:
4216                 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */
4217                         if (cmdlen < 2 || cmdlen > 31)
4218                                 goto bad_size;
4219                         break;
4220
4221                 case O_RECV:
4222                 case O_XMIT:
4223                 case O_VIA:
4224                         if (cmdlen != F_INSN_SIZE(ipfw_insn_if))
4225                                 goto bad_size;
4226                         break;
4227
4228                 case O_ALTQ:
4229                         if (cmdlen != F_INSN_SIZE(ipfw_insn_altq))
4230                                 goto bad_size;
4231                         break;
4232
4233                 case O_PIPE:
4234                 case O_QUEUE:
4235                         if (cmdlen != F_INSN_SIZE(ipfw_insn))
4236                                 goto bad_size;
4237                         goto check_action;
4238
4239                 case O_FORWARD_IP:
4240 #ifdef  IPFIREWALL_FORWARD
4241                         if (cmdlen != F_INSN_SIZE(ipfw_insn_sa))
4242                                 goto bad_size;
4243                         goto check_action;
4244 #else
4245                         return EINVAL;
4246 #endif
4247
4248                 case O_DIVERT:
4249                 case O_TEE:
4250                         if (ip_divert_ptr == NULL)
4251                                 return EINVAL;
4252                         else
4253                                 goto check_size;
4254                 case O_NETGRAPH:
4255                 case O_NGTEE:
4256                         if (!NG_IPFW_LOADED)
4257                                 return EINVAL;
4258                         else
4259                                 goto check_size;
4260                 case O_NAT:
4261 #ifdef IPFIREWALL_NAT
4262                         if (cmdlen != F_INSN_SIZE(ipfw_insn_nat))
4263                                 goto bad_size;
4264                         goto check_action;
4265 #else
4266                         return EINVAL;
4267 #endif
4268                 case O_FORWARD_MAC: /* XXX not implemented yet */
4269                 case O_CHECK_STATE:
4270                 case O_COUNT:
4271                 case O_ACCEPT:
4272                 case O_DENY:
4273                 case O_REJECT:
4274 #ifdef INET6
4275                 case O_UNREACH6:
4276 #endif
4277                 case O_SKIPTO:
4278 check_size:
4279                         if (cmdlen != F_INSN_SIZE(ipfw_insn))
4280                                 goto bad_size;
4281 check_action:
4282                         if (have_action) {
4283                                 printf("ipfw: opcode %d, multiple actions"
4284                                         " not allowed\n",
4285                                         cmd->opcode);
4286                                 return EINVAL;
4287                         }
4288                         have_action = 1;
4289                         if (l != cmdlen) {
4290                                 printf("ipfw: opcode %d, action must be"
4291                                         " last opcode\n",
4292                                         cmd->opcode);
4293                                 return EINVAL;
4294                         }
4295                         break;
4296 #ifdef INET6
4297                 case O_IP6_SRC:
4298                 case O_IP6_DST:
4299                         if (cmdlen != F_INSN_SIZE(struct in6_addr) +
4300                             F_INSN_SIZE(ipfw_insn))
4301                                 goto bad_size;
4302                         break;
4303
4304                 case O_FLOW6ID:
4305                         if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
4306                             ((ipfw_insn_u32 *)cmd)->o.arg1)
4307                                 goto bad_size;
4308                         break;
4309
4310                 case O_IP6_SRC_MASK:
4311                 case O_IP6_DST_MASK:
4312                         if ( !(cmdlen & 1) || cmdlen > 127)
4313                                 goto bad_size;
4314                         break;
4315                 case O_ICMP6TYPE:
4316                         if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) )
4317                                 goto bad_size;
4318                         break;
4319 #endif
4320
4321                 default:
4322                         switch (cmd->opcode) {
4323 #ifndef INET6
4324                         case O_IP6_SRC_ME:
4325                         case O_IP6_DST_ME:
4326                         case O_EXT_HDR:
4327                         case O_IP6:
4328                         case O_UNREACH6:
4329                         case O_IP6_SRC:
4330                         case O_IP6_DST:
4331                         case O_FLOW6ID:
4332                         case O_IP6_SRC_MASK:
4333                         case O_IP6_DST_MASK:
4334                         case O_ICMP6TYPE:
4335                                 printf("ipfw: no IPv6 support in kernel\n");
4336                                 return EPROTONOSUPPORT;
4337 #endif
4338                         default:
4339                                 printf("ipfw: opcode %d, unknown opcode\n",
4340                                         cmd->opcode);
4341                                 return EINVAL;
4342                         }
4343                 }
4344         }
4345         if (have_action == 0) {
4346                 printf("ipfw: missing action\n");
4347                 return EINVAL;
4348         }
4349         return 0;
4350
4351 bad_size:
4352         printf("ipfw: opcode %d size %d wrong\n",
4353                 cmd->opcode, cmdlen);
4354         return EINVAL;
4355 }
4356
4357 /*
4358  * Copy the static and dynamic rules to the supplied buffer
4359  * and return the amount of space actually used.
4360  */
4361 static size_t
4362 ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space)
4363 {
4364         char *bp = buf;
4365         char *ep = bp + space;
4366         struct ip_fw *rule;
4367         int i;
4368         time_t  boot_seconds;
4369
4370         boot_seconds = boottime.tv_sec;
4371         /* XXX this can take a long time and locking will block packet flow */
4372         IPFW_RLOCK(chain);
4373         for (rule = chain->rules; rule ; rule = rule->next) {
4374                 /*
4375                  * Verify the entry fits in the buffer in case the
4376                  * rules changed between calculating buffer space and
4377                  * now.  This would be better done using a generation
4378                  * number but should suffice for now.
4379                  */
4380                 i = RULESIZE(rule);
4381                 if (bp + i <= ep) {
4382                         bcopy(rule, bp, i);
4383                         /*
4384                          * XXX HACK. Store the disable mask in the "next" pointer
4385                          * in a wild attempt to keep the ABI the same.
4386                          * Why do we do this on EVERY rule?
4387                          */
4388                         bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule),
4389                             sizeof(set_disable));
4390                         if (((struct ip_fw *)bp)->timestamp)
4391                                 ((struct ip_fw *)bp)->timestamp += boot_seconds;
4392                         bp += i;
4393                 }
4394         }
4395         IPFW_RUNLOCK(chain);
4396         if (ipfw_dyn_v) {
4397                 ipfw_dyn_rule *p, *last = NULL;
4398
4399                 IPFW_DYN_LOCK();
4400                 for (i = 0 ; i < curr_dyn_buckets; i++)
4401                         for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) {
4402                                 if (bp + sizeof *p <= ep) {
4403                                         ipfw_dyn_rule *dst =
4404                                                 (ipfw_dyn_rule *)bp;
4405                                         bcopy(p, dst, sizeof *p);
4406                                         bcopy(&(p->rule->rulenum), &(dst->rule),
4407                                             sizeof(p->rule->rulenum));
4408                                         /*
4409                                          * store set number into high word of
4410                                          * dst->rule pointer.
4411                                          */
4412                                         bcopy(&(p->rule->set), &dst->rule +
4413                                             sizeof(p->rule->rulenum),
4414                                             sizeof(p->rule->set));
4415                                         /*
4416                                          * store a non-null value in "next".
4417                                          * The userland code will interpret a
4418                                          * NULL here as a marker
4419                                          * for the last dynamic rule.
4420                                          */
4421                                         bcopy(&dst, &dst->next, sizeof(dst));
4422                                         last = dst;
4423                                         dst->expire =
4424                                             TIME_LEQ(dst->expire, time_uptime) ?
4425                                                 0 : dst->expire - time_uptime ;
4426                                         bp += sizeof(ipfw_dyn_rule);
4427                                 }
4428                         }
4429                 IPFW_DYN_UNLOCK();
4430                 if (last != NULL) /* mark last dynamic rule */
4431                         bzero(&last->next, sizeof(last));
4432         }
4433         return (bp - (char *)buf);
4434 }
4435
4436
4437 /**
4438  * {set|get}sockopt parser.
4439  */
4440 static int
4441 ipfw_ctl(struct sockopt *sopt)
4442 {
4443 #define RULE_MAXSIZE    (256*sizeof(u_int32_t))
4444         int error;
4445         size_t size;
4446         struct ip_fw *buf, *rule;
4447         u_int32_t rulenum[2];
4448
4449         error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW);
4450         if (error)
4451                 return (error);
4452
4453         /*
4454          * Disallow modifications in really-really secure mode, but still allow
4455          * the logging counters to be reset.
4456          */
4457         if (sopt->sopt_name == IP_FW_ADD ||
4458             (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) {
4459                 error = securelevel_ge(sopt->sopt_td->td_ucred, 3);
4460                 if (error)
4461                         return (error);
4462         }
4463
4464         error = 0;
4465
4466         switch (sopt->sopt_name) {
4467         case IP_FW_GET:
4468                 /*
4469                  * pass up a copy of the current rules. Static rules
4470                  * come first (the last of which has number IPFW_DEFAULT_RULE),
4471                  * followed by a possibly empty list of dynamic rule.
4472                  * The last dynamic rule has NULL in the "next" field.
4473                  *
4474                  * Note that the calculated size is used to bound the
4475                  * amount of data returned to the user.  The rule set may
4476                  * change between calculating the size and returning the
4477                  * data in which case we'll just return what fits.
4478                  */
4479                 size = static_len;      /* size of static rules */
4480                 if (ipfw_dyn_v)         /* add size of dyn.rules */
4481                         size += (dyn_count * sizeof(ipfw_dyn_rule));
4482
4483                 /*
4484                  * XXX todo: if the user passes a short length just to know
4485                  * how much room is needed, do not bother filling up the
4486                  * buffer, just jump to the sooptcopyout.
4487                  */
4488                 buf = malloc(size, M_TEMP, M_WAITOK);
4489                 error = sooptcopyout(sopt, buf,
4490                                 ipfw_getrules(&layer3_chain, buf, size));
4491                 free(buf, M_TEMP);
4492                 break;
4493
4494         case IP_FW_FLUSH:
4495                 /*
4496                  * Normally we cannot release the lock on each iteration.
4497                  * We could do it here only because we start from the head all
4498                  * the times so there is no risk of missing some entries.
4499                  * On the other hand, the risk is that we end up with
4500                  * a very inconsistent ruleset, so better keep the lock
4501                  * around the whole cycle.
4502                  *
4503                  * XXX this code can be improved by resetting the head of
4504                  * the list to point to the default rule, and then freeing
4505                  * the old list without the need for a lock.
4506                  */
4507
4508                 IPFW_WLOCK(&layer3_chain);
4509                 layer3_chain.reap = NULL;
4510                 free_chain(&layer3_chain, 0 /* keep default rule */);
4511                 rule = layer3_chain.reap;
4512                 layer3_chain.reap = NULL;
4513                 IPFW_WUNLOCK(&layer3_chain);
4514                 if (rule != NULL)
4515                         reap_rules(rule);
4516                 break;
4517
4518         case IP_FW_ADD:
4519                 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK);
4520                 error = sooptcopyin(sopt, rule, RULE_MAXSIZE,
4521                         sizeof(struct ip_fw) );
4522                 if (error == 0)
4523                         error = check_ipfw_struct(rule, sopt->sopt_valsize);
4524                 if (error == 0) {
4525                         error = add_rule(&layer3_chain, rule);
4526                         size = RULESIZE(rule);
4527                         if (!error && sopt->sopt_dir == SOPT_GET)
4528                                 error = sooptcopyout(sopt, rule, size);
4529                 }
4530                 free(rule, M_TEMP);
4531                 break;
4532
4533         case IP_FW_DEL:
4534                 /*
4535                  * IP_FW_DEL is used for deleting single rules or sets,
4536                  * and (ab)used to atomically manipulate sets. Argument size
4537                  * is used to distinguish between the two:
4538                  *    sizeof(u_int32_t)
4539                  *      delete single rule or set of rules,
4540                  *      or reassign rules (or sets) to a different set.
4541                  *    2*sizeof(u_int32_t)
4542                  *      atomic disable/enable sets.
4543                  *      first u_int32_t contains sets to be disabled,
4544                  *      second u_int32_t contains sets to be enabled.
4545                  */
4546                 error = sooptcopyin(sopt, rulenum,
4547                         2*sizeof(u_int32_t), sizeof(u_int32_t));
4548                 if (error)
4549                         break;
4550                 size = sopt->sopt_valsize;
4551                 if (size == sizeof(u_int32_t))  /* delete or reassign */
4552                         error = del_entry(&layer3_chain, rulenum[0]);
4553                 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */
4554                         set_disable =
4555                             (set_disable | rulenum[0]) & ~rulenum[1] &
4556                             ~(1<<RESVD_SET); /* set RESVD_SET always enabled */
4557                 else
4558                         error = EINVAL;
4559                 break;
4560
4561         case IP_FW_ZERO:
4562         case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */
4563                 rulenum[0] = 0;
4564                 if (sopt->sopt_val != 0) {
4565                     error = sooptcopyin(sopt, rulenum,
4566                             sizeof(u_int32_t), sizeof(u_int32_t));
4567                     if (error)
4568                         break;
4569                 }
4570                 error = zero_entry(&layer3_chain, rulenum[0],
4571                         sopt->sopt_name == IP_FW_RESETLOG);
4572                 break;
4573
4574         case IP_FW_TABLE_ADD:
4575                 {
4576                         ipfw_table_entry ent;
4577
4578                         error = sooptcopyin(sopt, &ent,
4579                             sizeof(ent), sizeof(ent));
4580                         if (error)
4581                                 break;
4582                         error = add_table_entry(&layer3_chain, ent.tbl,
4583                             ent.addr, ent.masklen, ent.value);
4584                 }
4585                 break;
4586
4587         case IP_FW_TABLE_DEL:
4588                 {
4589                         ipfw_table_entry ent;
4590
4591                         error = sooptcopyin(sopt, &ent,
4592                             sizeof(ent), sizeof(ent));
4593                         if (error)
4594                                 break;
4595                         error = del_table_entry(&layer3_chain, ent.tbl,
4596                             ent.addr, ent.masklen);
4597                 }
4598                 break;
4599
4600         case IP_FW_TABLE_FLUSH:
4601                 {
4602                         u_int16_t tbl;
4603
4604                         error = sooptcopyin(sopt, &tbl,
4605                             sizeof(tbl), sizeof(tbl));
4606                         if (error)
4607                                 break;
4608                         IPFW_WLOCK(&layer3_chain);
4609                         error = flush_table(&layer3_chain, tbl);
4610                         IPFW_WUNLOCK(&layer3_chain);
4611                 }
4612                 break;
4613
4614         case IP_FW_TABLE_GETSIZE:
4615                 {
4616                         u_int32_t tbl, cnt;
4617
4618                         if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl),
4619                             sizeof(tbl))))
4620                                 break;
4621                         IPFW_RLOCK(&layer3_chain);
4622                         error = count_table(&layer3_chain, tbl, &cnt);
4623                         IPFW_RUNLOCK(&layer3_chain);
4624                         if (error)
4625                                 break;
4626                         error = sooptcopyout(sopt, &cnt, sizeof(cnt));
4627                 }
4628                 break;
4629
4630         case IP_FW_TABLE_LIST:
4631                 {
4632                         ipfw_table *tbl;
4633
4634                         if (sopt->sopt_valsize < sizeof(*tbl)) {
4635                                 error = EINVAL;
4636                                 break;
4637                         }
4638                         size = sopt->sopt_valsize;
4639                         tbl = malloc(size, M_TEMP, M_WAITOK);
4640                         error = sooptcopyin(sopt, tbl, size, sizeof(*tbl));
4641                         if (error) {
4642                                 free(tbl, M_TEMP);
4643                                 break;
4644                         }
4645                         tbl->size = (size - sizeof(*tbl)) /
4646                             sizeof(ipfw_table_entry);
4647                         IPFW_RLOCK(&layer3_chain);
4648                         error = dump_table(&layer3_chain, tbl);
4649                         IPFW_RUNLOCK(&layer3_chain);
4650                         if (error) {
4651                                 free(tbl, M_TEMP);
4652                                 break;
4653                         }
4654                         error = sooptcopyout(sopt, tbl, size);
4655                         free(tbl, M_TEMP);
4656                 }
4657                 break;
4658
4659 #ifdef IPFIREWALL_NAT
4660         case IP_FW_NAT_CFG:
4661         {
4662                 struct cfg_nat *ptr, *ser_n;
4663                 char *buf;
4664
4665                 buf = malloc(NAT_BUF_LEN, M_IPFW, M_WAITOK | M_ZERO);
4666                 error = sooptcopyin(sopt, buf, NAT_BUF_LEN,
4667                     sizeof(struct cfg_nat));
4668                 ser_n = (struct cfg_nat *)buf;
4669
4670                 /*
4671                  * Find/create nat rule.
4672                  */
4673                 IPFW_WLOCK(&layer3_chain);
4674                 ptr = lookup_nat(ser_n->id);
4675                 if (ptr == NULL) {
4676                         /* New rule: allocate and init new instance. */
4677                         ptr = malloc(sizeof(struct cfg_nat),
4678                             M_IPFW, M_NOWAIT | M_ZERO);
4679                         if (ptr == NULL) {
4680                                 IPFW_WUNLOCK(&layer3_chain);
4681                                 free(buf, M_IPFW);
4682                                 return (ENOSPC);
4683                         }
4684                         ptr->lib = LibAliasInit(NULL);
4685                         if (ptr->lib == NULL) {
4686                                 IPFW_WUNLOCK(&layer3_chain);
4687                                 free(ptr, M_IPFW);
4688                                 free(buf, M_IPFW);
4689                                 return (EINVAL);
4690                         }
4691                         LIST_INIT(&ptr->redir_chain);
4692                 } else {
4693                         /* Entry already present: temporarly unhook it. */
4694                         UNHOOK_NAT(ptr);
4695                         flush_nat_ptrs(ser_n->id);
4696                 }
4697                 IPFW_WUNLOCK(&layer3_chain);
4698
4699                 /*
4700                  * Basic nat configuration.
4701                  */
4702                 ptr->id = ser_n->id;
4703                 /*
4704                  * XXX - what if this rule doesn't nat any ip and just
4705                  * redirect?
4706                  * do we set aliasaddress to 0.0.0.0?
4707                  */
4708                 ptr->ip = ser_n->ip;
4709                 ptr->redir_cnt = ser_n->redir_cnt;
4710                 ptr->mode = ser_n->mode;
4711                 LibAliasSetMode(ptr->lib, ser_n->mode, ser_n->mode);
4712                 LibAliasSetAddress(ptr->lib, ptr->ip);
4713                 memcpy(ptr->if_name, ser_n->if_name, IF_NAMESIZE);
4714
4715                 /*
4716                  * Redir and LSNAT configuration.
4717                  */
4718                 /* Delete old cfgs. */
4719                 del_redir_spool_cfg(ptr, &ptr->redir_chain);
4720                 /* Add new entries. */
4721                 add_redir_spool_cfg(&buf[(sizeof(struct cfg_nat))], ptr);
4722                 free(buf, M_IPFW);
4723                 IPFW_WLOCK(&layer3_chain);
4724                 HOOK_NAT(&layer3_chain.nat, ptr);
4725                 IPFW_WUNLOCK(&layer3_chain);
4726         }
4727         break;
4728
4729         case IP_FW_NAT_DEL:
4730         {
4731                 struct cfg_nat *ptr;
4732                 int i;
4733
4734                 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
4735                 IPFW_WLOCK(&layer3_chain);
4736                 ptr = lookup_nat(i);
4737                 if (ptr == NULL) {
4738                         error = EINVAL;
4739                         IPFW_WUNLOCK(&layer3_chain);
4740                         break;
4741                 }
4742                 UNHOOK_NAT(ptr);
4743                 flush_nat_ptrs(i);
4744                 IPFW_WUNLOCK(&layer3_chain);
4745                 del_redir_spool_cfg(ptr, &ptr->redir_chain);
4746                 LibAliasUninit(ptr->lib);
4747                 free(ptr, M_IPFW);
4748         }
4749         break;
4750
4751         case IP_FW_NAT_GET_CONFIG:
4752         {
4753                 uint8_t *data;
4754                 struct cfg_nat *n;
4755                 struct cfg_redir *r;
4756                 struct cfg_spool *s;
4757                 int nat_cnt, off;
4758
4759                 nat_cnt = 0;
4760                 off = sizeof(nat_cnt);
4761
4762                 data = malloc(NAT_BUF_LEN, M_IPFW, M_WAITOK | M_ZERO);
4763                 IPFW_RLOCK(&layer3_chain);
4764                 /* Serialize all the data. */
4765                 LIST_FOREACH(n, &layer3_chain.nat, _next) {
4766                         nat_cnt++;
4767                         if (off + SOF_NAT < NAT_BUF_LEN) {
4768                                 bcopy(n, &data[off], SOF_NAT);
4769                                 off += SOF_NAT;
4770                                 LIST_FOREACH(r, &n->redir_chain, _next) {
4771                                         if (off + SOF_REDIR < NAT_BUF_LEN) {
4772                                                 bcopy(r, &data[off],
4773                                                     SOF_REDIR);
4774                                                 off += SOF_REDIR;
4775                                                 LIST_FOREACH(s, &r->spool_chain,
4776                                                     _next) {
4777                                                         if (off + SOF_SPOOL <
4778                                                             NAT_BUF_LEN) {
4779                                                                 bcopy(s,
4780                                                                     &data[off],
4781                                                                     SOF_SPOOL);
4782                                                                 off +=
4783                                                                     SOF_SPOOL;
4784                                                         } else
4785                                                                 goto nospace;
4786                                                 }
4787                                         } else
4788                                                 goto nospace;
4789                                 }
4790                         } else
4791                                 goto nospace;
4792                 }
4793                 bcopy(&nat_cnt, data, sizeof(nat_cnt));
4794                 IPFW_RUNLOCK(&layer3_chain);
4795                 error = sooptcopyout(sopt, data, NAT_BUF_LEN);
4796                 free(data, M_IPFW);
4797                 break;
4798         nospace:
4799                 IPFW_RUNLOCK(&layer3_chain);
4800                 printf("serialized data buffer not big enough:"
4801                     "please increase NAT_BUF_LEN\n");
4802                 free(data, M_IPFW);
4803         }
4804         break;
4805
4806         case IP_FW_NAT_GET_LOG:
4807         {
4808                 uint8_t *data;
4809                 struct cfg_nat *ptr;
4810                 int i, size, cnt, sof;
4811
4812                 data = NULL;
4813                 sof = LIBALIAS_BUF_SIZE;
4814                 cnt = 0;
4815
4816                 IPFW_RLOCK(&layer3_chain);
4817                 size = i = 0;
4818                 LIST_FOREACH(ptr, &layer3_chain.nat, _next) {
4819                         if (ptr->lib->logDesc == NULL)
4820                                 continue;
4821                         cnt++;
4822                         size = cnt * (sof + sizeof(int));
4823                         data = realloc(data, size, M_IPFW, M_NOWAIT | M_ZERO);
4824                         if (data == NULL) {
4825                                 IPFW_RUNLOCK(&layer3_chain);
4826                                 return (ENOSPC);
4827                         }
4828                         bcopy(&ptr->id, &data[i], sizeof(int));
4829                         i += sizeof(int);
4830                         bcopy(ptr->lib->logDesc, &data[i], sof);
4831                         i += sof;
4832                 }
4833                 IPFW_RUNLOCK(&layer3_chain);
4834                 error = sooptcopyout(sopt, data, size);
4835                 free(data, M_IPFW);
4836         }
4837         break;
4838 #endif
4839
4840         default:
4841                 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name);
4842                 error = EINVAL;
4843         }
4844
4845         return (error);
4846 #undef RULE_MAXSIZE
4847 }
4848
4849 /**
4850  * dummynet needs a reference to the default rule, because rules can be
4851  * deleted while packets hold a reference to them. When this happens,
4852  * dummynet changes the reference to the default rule (it could well be a
4853  * NULL pointer, but this way we do not need to check for the special
4854  * case, plus here he have info on the default behaviour).
4855  */
4856 struct ip_fw *ip_fw_default_rule;
4857
4858 /*
4859  * This procedure is only used to handle keepalives. It is invoked
4860  * every dyn_keepalive_period
4861  */
4862 static void
4863 ipfw_tick(void * __unused unused)
4864 {
4865         struct mbuf *m0, *m, *mnext, **mtailp;
4866         int i;
4867         ipfw_dyn_rule *q;
4868
4869         if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0)
4870                 goto done;
4871
4872         /*
4873          * We make a chain of packets to go out here -- not deferring
4874          * until after we drop the IPFW dynamic rule lock would result
4875          * in a lock order reversal with the normal packet input -> ipfw
4876          * call stack.
4877          */
4878         m0 = NULL;
4879         mtailp = &m0;
4880         IPFW_DYN_LOCK();
4881         for (i = 0 ; i < curr_dyn_buckets ; i++) {
4882                 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) {
4883                         if (q->dyn_type == O_LIMIT_PARENT)
4884                                 continue;
4885                         if (q->id.proto != IPPROTO_TCP)
4886                                 continue;
4887                         if ( (q->state & BOTH_SYN) != BOTH_SYN)
4888                                 continue;
4889                         if (TIME_LEQ( time_uptime+dyn_keepalive_interval,
4890                             q->expire))
4891                                 continue;       /* too early */
4892                         if (TIME_LEQ(q->expire, time_uptime))
4893                                 continue;       /* too late, rule expired */
4894
4895                         *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1,
4896                                 q->ack_fwd, TH_SYN);
4897                         if (*mtailp != NULL)
4898                                 mtailp = &(*mtailp)->m_nextpkt;
4899                         *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1,
4900                                 q->ack_rev, 0);
4901                         if (*mtailp != NULL)
4902                                 mtailp = &(*mtailp)->m_nextpkt;
4903                 }
4904         }
4905         IPFW_DYN_UNLOCK();
4906         for (m = mnext = m0; m != NULL; m = mnext) {
4907                 mnext = m->m_nextpkt;
4908                 m->m_nextpkt = NULL;
4909                 ip_output(m, NULL, NULL, 0, NULL, NULL);
4910         }
4911 done:
4912         callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL);
4913 }
4914
4915 int
4916 ipfw_init(void)
4917 {
4918         struct ip_fw default_rule;
4919         int error;
4920
4921 #ifdef INET6
4922         /* Setup IPv6 fw sysctl tree. */
4923         sysctl_ctx_init(&ip6_fw_sysctl_ctx);
4924         ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx,
4925             SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw",
4926             CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall");
4927         SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree),
4928             OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3,
4929             &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6");
4930         SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree),
4931             OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE,
4932             &fw_deny_unknown_exthdrs, 0,
4933             "Deny packets with unknown IPv6 Extension Headers");
4934 #endif
4935
4936         layer3_chain.rules = NULL;
4937         IPFW_LOCK_INIT(&layer3_chain);
4938         ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule",
4939             sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL,
4940             UMA_ALIGN_PTR, 0);
4941         IPFW_DYN_LOCK_INIT();
4942         callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE);
4943
4944         bzero(&default_rule, sizeof default_rule);
4945
4946         default_rule.act_ofs = 0;
4947         default_rule.rulenum = IPFW_DEFAULT_RULE;
4948         default_rule.cmd_len = 1;
4949         default_rule.set = RESVD_SET;
4950
4951         default_rule.cmd[0].len = 1;
4952         default_rule.cmd[0].opcode =
4953 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
4954                                 1 ? O_ACCEPT :
4955 #endif
4956                                 O_DENY;
4957
4958         error = add_rule(&layer3_chain, &default_rule);
4959         if (error != 0) {
4960                 printf("ipfw2: error %u initializing default rule "
4961                         "(support disabled)\n", error);
4962                 IPFW_DYN_LOCK_DESTROY();
4963                 IPFW_LOCK_DESTROY(&layer3_chain);
4964                 uma_zdestroy(ipfw_dyn_rule_zone);
4965                 return (error);
4966         }
4967
4968         ip_fw_default_rule = layer3_chain.rules;
4969         printf("ipfw2 "
4970 #ifdef INET6
4971                 "(+ipv6) "
4972 #endif
4973                 "initialized, divert %s, "
4974                 "rule-based forwarding "
4975 #ifdef IPFIREWALL_FORWARD
4976                 "enabled, "
4977 #else
4978                 "disabled, "
4979 #endif
4980                 "default to %s, logging ",
4981 #ifdef IPDIVERT
4982                 "enabled",
4983 #else
4984                 "loadable",
4985 #endif
4986                 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny");
4987
4988 #ifdef IPFIREWALL_VERBOSE
4989         fw_verbose = 1;
4990 #endif
4991 #ifdef IPFIREWALL_VERBOSE_LIMIT
4992         verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
4993 #endif
4994         if (fw_verbose == 0)
4995                 printf("disabled\n");
4996         else if (verbose_limit == 0)
4997                 printf("unlimited\n");
4998         else
4999                 printf("limited to %d packets/entry by default\n",
5000                     verbose_limit);
5001
5002         error = init_tables(&layer3_chain);
5003         if (error) {
5004                 IPFW_DYN_LOCK_DESTROY();
5005                 IPFW_LOCK_DESTROY(&layer3_chain);
5006                 uma_zdestroy(ipfw_dyn_rule_zone);
5007                 return (error);
5008         }
5009         ip_fw_ctl_ptr = ipfw_ctl;
5010         ip_fw_chk_ptr = ipfw_chk;
5011         callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL);
5012 #ifdef IPFIREWALL_NAT
5013         LIST_INIT(&layer3_chain.nat);
5014         ifaddr_event_tag = EVENTHANDLER_REGISTER(ifaddr_event, ifaddr_change,
5015             NULL, EVENTHANDLER_PRI_ANY);
5016 #endif
5017         return (0);
5018 }
5019
5020 void
5021 ipfw_destroy(void)
5022 {
5023         struct ip_fw *reap;
5024 #ifdef IPFIREWALL_NAT
5025         struct cfg_nat *ptr, *ptr_temp;
5026 #endif
5027
5028         ip_fw_chk_ptr = NULL;
5029         ip_fw_ctl_ptr = NULL;
5030         callout_drain(&ipfw_timeout);
5031         IPFW_WLOCK(&layer3_chain);
5032         flush_tables(&layer3_chain);
5033 #ifdef IPFIREWALL_NAT
5034         LIST_FOREACH_SAFE(ptr, &layer3_chain.nat, _next, ptr_temp) {
5035                 LIST_REMOVE(ptr, _next);
5036                 del_redir_spool_cfg(ptr, &ptr->redir_chain);
5037                 LibAliasUninit(ptr->lib);
5038                 free(ptr, M_IPFW);
5039         }
5040         EVENTHANDLER_DEREGISTER(ifaddr_event, ifaddr_event_tag);
5041 #endif
5042         layer3_chain.reap = NULL;
5043         free_chain(&layer3_chain, 1 /* kill default rule */);
5044         reap = layer3_chain.reap, layer3_chain.reap = NULL;
5045         IPFW_WUNLOCK(&layer3_chain);
5046         if (reap != NULL)
5047                 reap_rules(reap);
5048         IPFW_DYN_LOCK_DESTROY();
5049         uma_zdestroy(ipfw_dyn_rule_zone);
5050         IPFW_LOCK_DESTROY(&layer3_chain);
5051
5052 #ifdef INET6
5053         /* Free IPv6 fw sysctl tree. */
5054         sysctl_ctx_free(&ip6_fw_sysctl_ctx);
5055 #endif
5056
5057         printf("IP firewall unloaded\n");
5058 }