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
  26 #include <sys/cdefs.h>
  27 __FBSDID("$FreeBSD$");
  28
  29 #define        DEB(x)
  30 #define        DDB(x) x
  31
  32 /*
  33  * Implement IP packet firewall (new version)
  34  */
  35
  36 #if !defined(KLD_MODULE)
  37 #include "opt_ipfw.h"
  38 #include "opt_ipdivert.h"
  39 #include "opt_ipdn.h"
  40 #include "opt_inet.h"
  41 #ifndef INET
  42 #error IPFIREWALL requires INET.
  43 #endif /* INET */
  44 #endif
  45 #include "opt_inet6.h"
  46 #include "opt_ipsec.h"
  47 #include "opt_mac.h"
  48
  49 #include <sys/param.h>
  50 #include <sys/systm.h>
  51 #include <sys/condvar.h>
  52 #include <sys/eventhandler.h>
  53 #include <sys/malloc.h>
  54 #include <sys/mbuf.h>
  55 #include <sys/kernel.h>
  56 #include <sys/lock.h>
  57 #include <sys/jail.h>
  58 #include <sys/module.h>
  59 #include <sys/priv.h>
  60 #include <sys/proc.h>
  61 #include <sys/rwlock.h>
  62 #include <sys/socket.h>
  63 #include <sys/socketvar.h>
  64 #include <sys/sysctl.h>
  65 #include <sys/syslog.h>
  66 #include <sys/ucred.h>
  67 #include <net/if.h>
  68 #include <net/radix.h>
  69 #include <net/route.h>
  70 #include <net/pf_mtag.h>
  71 #include <netinet/in.h>
  72 #include <netinet/in_systm.h>
  73 #include <netinet/in_var.h>
  74 #include <netinet/in_pcb.h>
  75 #include <netinet/ip.h>
  76 #include <netinet/ip_var.h>
  77 #include <netinet/ip_icmp.h>
  78 #include <netinet/ip_fw.h>
  79 #include <netinet/ip_divert.h>
  80 #include <netinet/ip_dummynet.h>
  81 #include <netinet/ip_carp.h>
  82 #include <netinet/pim.h>
  83 #include <netinet/tcp.h>
  84 #include <netinet/tcp_timer.h>
  85 #include <netinet/tcp_var.h>
  86 #include <netinet/tcpip.h>
  87 #include <netinet/udp.h>
  88 #include <netinet/udp_var.h>
  89 #include <netinet/sctp.h>
  90 #ifdef IPFIREWALL_NAT
  91 #include <netinet/libalias/alias.h>
  92 #include <netinet/libalias/alias_local.h>
  93 #endif
  94 #include <netgraph/ng_ipfw.h>
  95
  96 #include <altq/if_altq.h>
  97
  98 #include <netinet/ip6.h>
  99 #include <netinet/icmp6.h>
 100 #ifdef INET6
 101 #include <netinet6/scope6_var.h>
 102 #endif
 103
 104 #include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */
 105
 106 #include <machine/in_cksum.h>   /* XXX for in_cksum */
 107
 108 #include <security/mac/mac_framework.h>
 109
 110 /*
 111  * set_disable contains one bit per set value (0..31).
 112  * If the bit is set, all rules with the corresponding set
 113  * are disabled. Set RESVD_SET(31) is reserved for the default rule
 114  * and rules that are not deleted by the flush command,
 115  * and CANNOT be disabled.
 116  * Rules in set RESVD_SET can only be deleted explicitly.
 117  */
 118 static u_int32_t set_disable;
 119
 120 static int fw_verbose;
 121 static int verbose_limit;
 122
 123 static struct callout ipfw_timeout;
 124 static uma_zone_t ipfw_dyn_rule_zone;
 125 #define IPFW_DEFAULT_RULE       65535
 126
 127 /*
 128  * Data structure to cache our ucred related
 129  * information. This structure only gets used if
 130  * the user specified UID/GID based constraints in
 131  * a firewall rule.
 132  */
 133 struct ip_fw_ugid {
 134         gid_t           fw_groups[NGROUPS];
 135         int             fw_ngroups;
 136         uid_t           fw_uid;
 137         int             fw_prid;
 138 };
 139
 140 #define IPFW_TABLES_MAX         128
 141 struct ip_fw_chain {
 142         struct ip_fw    *rules;         /* list of rules */
 143         struct ip_fw    *reap;          /* list of rules to reap */
 144         LIST_HEAD(, cfg_nat) nat;       /* list of nat entries */
 145         struct radix_node_head *tables[IPFW_TABLES_MAX];
 146         struct rwlock   rwmtx;
 147 };
 148 #define IPFW_LOCK_INIT(_chain) \
 149         rw_init(&(_chain)->rwmtx, "IPFW static rules")
 150 #define IPFW_LOCK_DESTROY(_chain)       rw_destroy(&(_chain)->rwmtx)
 151 #define IPFW_WLOCK_ASSERT(_chain)       rw_assert(&(_chain)->rwmtx, RA_WLOCKED)
 152
 153 #define IPFW_RLOCK(p) rw_rlock(&(p)->rwmtx)
 154 #define IPFW_RUNLOCK(p) rw_runlock(&(p)->rwmtx)
 155 #define IPFW_WLOCK(p) rw_wlock(&(p)->rwmtx)
 156 #define IPFW_WUNLOCK(p) rw_wunlock(&(p)->rwmtx)
 157
 158 /*
 159  * list of rules for layer 3
 160  */
 161 static struct ip_fw_chain layer3_chain;
 162
 163 MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's");
 164 MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables");
 165
 166 struct table_entry {
 167         struct radix_node       rn[2];
 168         struct sockaddr_in      addr, mask;
 169         u_int32_t               value;
 170 };
 171
 172 static int fw_debug = 1;
 173 static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */
 174
 175 extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS);
 176
 177 #ifdef SYSCTL_NODE
 178 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 179 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable,
 180     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &fw_enable, 0,
 181     ipfw_chg_hook, "I", "Enable ipfw");
 182 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW,
 183     &autoinc_step, 0, "Rule number autincrement step");
 184 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
 185     CTLFLAG_RW | CTLFLAG_SECURE3,
 186     &fw_one_pass, 0,
 187     "Only do a single pass through ipfw when using dummynet(4)");
 188 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW,
 189     &fw_debug, 0, "Enable printing of debug ip_fw statements");
 190 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose,
 191     CTLFLAG_RW | CTLFLAG_SECURE3,
 192     &fw_verbose, 0, "Log matches to ipfw rules");
 193 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW,
 194     &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged");
 195
 196 /*
 197  * Description of dynamic rules.
 198  *
 199  * Dynamic rules are stored in lists accessed through a hash table
 200  * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can
 201  * be modified through the sysctl variable dyn_buckets which is
 202  * updated when the table becomes empty.
 203  *
 204  * XXX currently there is only one list, ipfw_dyn.
 205  *
 206  * When a packet is received, its address fields are first masked
 207  * with the mask defined for the rule, then hashed, then matched
 208  * against the entries in the corresponding list.
 209  * Dynamic rules can be used for different purposes:
 210  *  + stateful rules;
 211  *  + enforcing limits on the number of sessions;
 212  *  + in-kernel NAT (not implemented yet)
 213  *
 214  * The lifetime of dynamic rules is regulated by dyn_*_lifetime,
 215  * measured in seconds and depending on the flags.
 216  *
 217  * The total number of dynamic rules is stored in dyn_count.
 218  * The max number of dynamic rules is dyn_max. When we reach
 219  * the maximum number of rules we do not create anymore. This is
 220  * done to avoid consuming too much memory, but also too much
 221  * time when searching on each packet (ideally, we should try instead
 222  * to put a limit on the length of the list on each bucket...).
 223  *
 224  * Each dynamic rule holds a pointer to the parent ipfw rule so
 225  * we know what action to perform. Dynamic rules are removed when
 226  * the parent rule is deleted. XXX we should make them survive.
 227  *
 228  * There are some limitations with dynamic rules -- we do not
 229  * obey the 'randomized match', and we do not do multiple
 230  * passes through the firewall. XXX check the latter!!!
 231  */
 232 static ipfw_dyn_rule **ipfw_dyn_v = NULL;
 233 static u_int32_t dyn_buckets = 256; /* must be power of 2 */
 234 static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */
 235
 236 static struct mtx ipfw_dyn_mtx;         /* mutex guarding dynamic rules */
 237 #define IPFW_DYN_LOCK_INIT() \
 238         mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF)
 239 #define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx)
 240 #define IPFW_DYN_LOCK()         mtx_lock(&ipfw_dyn_mtx)
 241 #define IPFW_DYN_UNLOCK()       mtx_unlock(&ipfw_dyn_mtx)
 242 #define IPFW_DYN_LOCK_ASSERT()  mtx_assert(&ipfw_dyn_mtx, MA_OWNED)
 243
 244 /*
 245  * Timeouts for various events in handing dynamic rules.
 246  */
 247 static u_int32_t dyn_ack_lifetime = 300;
 248 static u_int32_t dyn_syn_lifetime = 20;
 249 static u_int32_t dyn_fin_lifetime = 1;
 250 static u_int32_t dyn_rst_lifetime = 1;
 251 static u_int32_t dyn_udp_lifetime = 10;
 252 static u_int32_t dyn_short_lifetime = 5;
 253
 254 /*
 255  * Keepalives are sent if dyn_keepalive is set. They are sent every
 256  * dyn_keepalive_period seconds, in the last dyn_keepalive_interval
 257  * seconds of lifetime of a rule.
 258  * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower
 259  * than dyn_keepalive_period.
 260  */
 261
 262 static u_int32_t dyn_keepalive_interval = 20;
 263 static u_int32_t dyn_keepalive_period = 5;
 264 static u_int32_t dyn_keepalive = 1;     /* do send keepalives */
 265
 266 static u_int32_t static_count;  /* # of static rules */
 267 static u_int32_t static_len;    /* size in bytes of static rules */
 268 static u_int32_t dyn_count;             /* # of dynamic rules */
 269 static u_int32_t dyn_max = 4096;        /* max # of dynamic rules */
 270
 271 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW,
 272     &dyn_buckets, 0, "Number of dyn. buckets");
 273 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD,
 274     &curr_dyn_buckets, 0, "Current Number of dyn. buckets");
 275 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD,
 276     &dyn_count, 0, "Number of dyn. rules");
 277 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW,
 278     &dyn_max, 0, "Max number of dyn. rules");
 279 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD,
 280     &static_count, 0, "Number of static rules");
 281 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW,
 282     &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks");
 283 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW,
 284     &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn");
 285 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW,
 286     &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin");
 287 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW,
 288     &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst");
 289 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW,
 290     &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP");
 291 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW,
 292     &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations");
 293 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW,
 294     &dyn_keepalive, 0, "Enable keepalives for dyn. rules");
 295
 296 #ifdef INET6
 297 /*
 298  * IPv6 specific variables
 299  */
 300 SYSCTL_DECL(_net_inet6_ip6);
 301
 302 static struct sysctl_ctx_list ip6_fw_sysctl_ctx;
 303 static struct sysctl_oid *ip6_fw_sysctl_tree;
 304 #endif /* INET6 */
 305 #endif /* SYSCTL_NODE */
 306
 307 #ifdef IPFIREWALL_NAT
 308 MODULE_DEPEND(ipfw, libalias, 1, 1, 1);
 309 #endif
 310 static int fw_deny_unknown_exthdrs = 1;
 311
 312
 313 /*
 314  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
 315  * Other macros just cast void * into the appropriate type
 316  */
 317 #define L3HDR(T, ip)    ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
 318 #define TCP(p)          ((struct tcphdr *)(p))
 319 #define SCTP(p)         ((struct sctphdr *)(p))
 320 #define UDP(p)          ((struct udphdr *)(p))
 321 #define ICMP(p)         ((struct icmphdr *)(p))
 322 #define ICMP6(p)        ((struct icmp6_hdr *)(p))
 323
 324 static __inline int
 325 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
 326 {
 327         int type = icmp->icmp_type;
 328
 329         return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
 330 }
 331
 332 #define TT      ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
 333     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
 334
 335 static int
 336 is_icmp_query(struct icmphdr *icmp)
 337 {
 338         int type = icmp->icmp_type;
 339
 340         return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
 341 }
 342 #undef TT
 343
 344 /*
 345  * The following checks use two arrays of 8 or 16 bits to store the
 346  * bits that we want set or clear, respectively. They are in the
 347  * low and high half of cmd->arg1 or cmd->d[0].
 348  *
 349  * We scan options and store the bits we find set. We succeed if
 350  *
 351  *      (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
 352  *
 353  * The code is sometimes optimized not to store additional variables.
 354  */
 355
 356 static int
 357 flags_match(ipfw_insn *cmd, u_int8_t bits)
 358 {
 359         u_char want_clear;
 360         bits = ~bits;
 361
 362         if ( ((cmd->arg1 & 0xff) & bits) != 0)
 363                 return 0; /* some bits we want set were clear */
 364         want_clear = (cmd->arg1 >> 8) & 0xff;
 365         if ( (want_clear & bits) != want_clear)
 366                 return 0; /* some bits we want clear were set */
 367         return 1;
 368 }
 369
 370 static int
 371 ipopts_match(struct ip *ip, ipfw_insn *cmd)
 372 {
 373         int optlen, bits = 0;
 374         u_char *cp = (u_char *)(ip + 1);
 375         int x = (ip->ip_hl << 2) - sizeof (struct ip);
 376
 377         for (; x > 0; x -= optlen, cp += optlen) {
 378                 int opt = cp[IPOPT_OPTVAL];
 379
 380                 if (opt == IPOPT_EOL)
 381                         break;
 382                 if (opt == IPOPT_NOP)
 383                         optlen = 1;
 384                 else {
 385                         optlen = cp[IPOPT_OLEN];
 386                         if (optlen <= 0 || optlen > x)
 387                                 return 0; /* invalid or truncated */
 388                 }
 389                 switch (opt) {
 390
 391                 default:
 392                         break;
 393
 394                 case IPOPT_LSRR:
 395                         bits |= IP_FW_IPOPT_LSRR;
 396                         break;
 397
 398                 case IPOPT_SSRR:
 399                         bits |= IP_FW_IPOPT_SSRR;
 400                         break;
 401
 402                 case IPOPT_RR:
 403                         bits |= IP_FW_IPOPT_RR;
 404                         break;
 405
 406                 case IPOPT_TS:
 407                         bits |= IP_FW_IPOPT_TS;
 408                         break;
 409                 }
 410         }
 411         return (flags_match(cmd, bits));
 412 }
 413
 414 static int
 415 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
 416 {
 417         int optlen, bits = 0;
 418         u_char *cp = (u_char *)(tcp + 1);
 419         int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
 420
 421         for (; x > 0; x -= optlen, cp += optlen) {
 422                 int opt = cp[0];
 423                 if (opt == TCPOPT_EOL)
 424                         break;
 425                 if (opt == TCPOPT_NOP)
 426                         optlen = 1;
 427                 else {
 428                         optlen = cp[1];
 429                         if (optlen <= 0)
 430                                 break;
 431                 }
 432
 433                 switch (opt) {
 434
 435                 default:
 436                         break;
 437
 438                 case TCPOPT_MAXSEG:
 439                         bits |= IP_FW_TCPOPT_MSS;
 440                         break;
 441
 442                 case TCPOPT_WINDOW:
 443                         bits |= IP_FW_TCPOPT_WINDOW;
 444                         break;
 445
 446                 case TCPOPT_SACK_PERMITTED:
 447                 case TCPOPT_SACK:
 448                         bits |= IP_FW_TCPOPT_SACK;
 449                         break;
 450
 451                 case TCPOPT_TIMESTAMP:
 452                         bits |= IP_FW_TCPOPT_TS;
 453                         break;
 454
 455                 }
 456         }
 457         return (flags_match(cmd, bits));
 458 }
 459
 460 static int
 461 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd)
 462 {
 463         if (ifp == NULL)        /* no iface with this packet, match fails */
 464                 return 0;
 465         /* Check by name or by IP address */
 466         if (cmd->name[0] != '\0') { /* match by name */
 467                 /* Check name */
 468                 if (cmd->p.glob) {
 469                         if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
 470                                 return(1);
 471                 } else {
 472                         if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
 473                                 return(1);
 474                 }
 475         } else {
 476                 struct ifaddr *ia;
 477
 478                 /* XXX lock? */
 479                 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
 480                         if (ia->ifa_addr->sa_family != AF_INET)
 481                                 continue;
 482                         if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
 483                             (ia->ifa_addr))->sin_addr.s_addr)
 484                                 return(1);      /* match */
 485                 }
 486         }
 487         return(0);      /* no match, fail ... */
 488 }
 489
 490 /*
 491  * The verify_path function checks if a route to the src exists and
 492  * if it is reachable via ifp (when provided).
 493  *
 494  * The 'verrevpath' option checks that the interface that an IP packet
 495  * arrives on is the same interface that traffic destined for the
 496  * packet's source address would be routed out of.  The 'versrcreach'
 497  * option just checks that the source address is reachable via any route
 498  * (except default) in the routing table.  These two are a measure to block
 499  * forged packets.  This is also commonly known as "anti-spoofing" or Unicast
 500  * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
 501  * is purposely reminiscent of the Cisco IOS command,
 502  *
 503  *   ip verify unicast reverse-path
 504  *   ip verify unicast source reachable-via any
 505  *
 506  * which implements the same functionality. But note that syntax is
 507  * misleading. The check may be performed on all IP packets whether unicast,
 508  * multicast, or broadcast.
 509  */
 510 static int
 511 verify_path(struct in_addr src, struct ifnet *ifp)
 512 {
 513         struct route ro;
 514         struct sockaddr_in *dst;
 515
 516         bzero(&ro, sizeof(ro));
 517
 518         dst = (struct sockaddr_in *)&(ro.ro_dst);
 519         dst->sin_family = AF_INET;
 520         dst->sin_len = sizeof(*dst);
 521         dst->sin_addr = src;
 522         rtalloc_ign(&ro, RTF_CLONING);
 523
 524         if (ro.ro_rt == NULL)
 525                 return 0;
 526
 527         /*
 528          * If ifp is provided, check for equality with rtentry.
 529          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 530          * in order to pass packets injected back by if_simloop():
 531          * if useloopback == 1 routing entry (via lo0) for our own address
 532          * may exist, so we need to handle routing assymetry.
 533          */
 534         if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
 535                 RTFREE(ro.ro_rt);
 536                 return 0;
 537         }
 538
 539         /* if no ifp provided, check if rtentry is not default route */
 540         if (ifp == NULL &&
 541              satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) {
 542                 RTFREE(ro.ro_rt);
 543                 return 0;
 544         }
 545
 546         /* or if this is a blackhole/reject route */
 547         if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
 548                 RTFREE(ro.ro_rt);
 549                 return 0;
 550         }
 551
 552         /* found valid route */
 553         RTFREE(ro.ro_rt);
 554         return 1;
 555 }
 556
 557 #ifdef INET6
 558 /*
 559  * ipv6 specific rules here...
 560  */
 561 static __inline int
 562 icmp6type_match (int type, ipfw_insn_u32 *cmd)
 563 {
 564         return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
 565 }
 566
 567 static int
 568 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
 569 {
 570         int i;
 571         for (i=0; i <= cmd->o.arg1; ++i )
 572                 if (curr_flow == cmd->d[i] )
 573                         return 1;
 574         return 0;
 575 }
 576
 577 /* support for IP6_*_ME opcodes */
 578 static int
 579 search_ip6_addr_net (struct in6_addr * ip6_addr)
 580 {
 581         struct ifnet *mdc;
 582         struct ifaddr *mdc2;
 583         struct in6_ifaddr *fdm;
 584         struct in6_addr copia;
 585
 586         TAILQ_FOREACH(mdc, &ifnet, if_link)
 587                 TAILQ_FOREACH(mdc2, &mdc->if_addrlist, ifa_list) {
 588                         if (mdc2->ifa_addr->sa_family == AF_INET6) {
 589                                 fdm = (struct in6_ifaddr *)mdc2;
 590                                 copia = fdm->ia_addr.sin6_addr;
 591                                 /* need for leaving scope_id in the sock_addr */
 592                                 in6_clearscope(&copia);
 593                                 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia))
 594                                         return 1;
 595                         }
 596                 }
 597         return 0;
 598 }
 599
 600 static int
 601 verify_path6(struct in6_addr *src, struct ifnet *ifp)
 602 {
 603         struct route_in6 ro;
 604         struct sockaddr_in6 *dst;
 605
 606         bzero(&ro, sizeof(ro));
 607
 608         dst = (struct sockaddr_in6 * )&(ro.ro_dst);
 609         dst->sin6_family = AF_INET6;
 610         dst->sin6_len = sizeof(*dst);
 611         dst->sin6_addr = *src;
 612         rtalloc_ign((struct route *)&ro, RTF_CLONING);
 613
 614         if (ro.ro_rt == NULL)
 615                 return 0;
 616
 617         /*
 618          * if ifp is provided, check for equality with rtentry
 619          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 620          * to support the case of sending packets to an address of our own.
 621          * (where the former interface is the first argument of if_simloop()
 622          *  (=ifp), the latter is lo0)
 623          */
 624         if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
 625                 RTFREE(ro.ro_rt);
 626                 return 0;
 627         }
 628
 629         /* if no ifp provided, check if rtentry is not default route */
 630         if (ifp == NULL &&
 631             IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) {
 632                 RTFREE(ro.ro_rt);
 633                 return 0;
 634         }
 635
 636         /* or if this is a blackhole/reject route */
 637         if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
 638                 RTFREE(ro.ro_rt);
 639                 return 0;
 640         }
 641
 642         /* found valid route */
 643         RTFREE(ro.ro_rt);
 644         return 1;
 645
 646 }
 647 static __inline int
 648 hash_packet6(struct ipfw_flow_id *id)
 649 {
 650         u_int32_t i;
 651         i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^
 652             (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^
 653             (id->src_ip6.__u6_addr.__u6_addr32[2]) ^
 654             (id->src_ip6.__u6_addr.__u6_addr32[3]) ^
 655             (id->dst_port) ^ (id->src_port);
 656         return i;
 657 }
 658
 659 static int
 660 is_icmp6_query(int icmp6_type)
 661 {
 662         if ((icmp6_type <= ICMP6_MAXTYPE) &&
 663             (icmp6_type == ICMP6_ECHO_REQUEST ||
 664             icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
 665             icmp6_type == ICMP6_WRUREQUEST ||
 666             icmp6_type == ICMP6_FQDN_QUERY ||
 667             icmp6_type == ICMP6_NI_QUERY))
 668                 return (1);
 669
 670         return (0);
 671 }
 672
 673 static void
 674 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
 675 {
 676         struct mbuf *m;
 677
 678         m = args->m;
 679         if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
 680                 struct tcphdr *tcp;
 681                 tcp_seq ack, seq;
 682                 int flags;
 683                 struct {
 684                         struct ip6_hdr ip6;
 685                         struct tcphdr th;
 686                 } ti;
 687                 tcp = (struct tcphdr *)((char *)ip6 + hlen);
 688
 689                 if ((tcp->th_flags & TH_RST) != 0) {
 690                         m_freem(m);
 691                         args->m = NULL;
 692                         return;
 693                 }
 694
 695                 ti.ip6 = *ip6;
 696                 ti.th = *tcp;
 697                 ti.th.th_seq = ntohl(ti.th.th_seq);
 698                 ti.th.th_ack = ntohl(ti.th.th_ack);
 699                 ti.ip6.ip6_nxt = IPPROTO_TCP;
 700
 701                 if (ti.th.th_flags & TH_ACK) {
 702                         ack = 0;
 703                         seq = ti.th.th_ack;
 704                         flags = TH_RST;
 705                 } else {
 706                         ack = ti.th.th_seq;
 707                         if ((m->m_flags & M_PKTHDR) != 0) {
 708                                 /*
 709                                  * total new data to ACK is:
 710                                  * total packet length,
 711                                  * minus the header length,
 712                                  * minus the tcp header length.
 713                                  */
 714                                 ack += m->m_pkthdr.len - hlen
 715                                         - (ti.th.th_off << 2);
 716                         } else if (ip6->ip6_plen) {
 717                                 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) -
 718                                     hlen - (ti.th.th_off << 2);
 719                         } else {
 720                                 m_freem(m);
 721                                 return;
 722                         }
 723                         if (tcp->th_flags & TH_SYN)
 724                                 ack++;
 725                         seq = 0;
 726                         flags = TH_RST|TH_ACK;
 727                 }
 728                 bcopy(&ti, ip6, sizeof(ti));
 729                 /*
 730                  * m is only used to recycle the mbuf
 731                  * The data in it is never read so we don't need
 732                  * to correct the offsets or anything
 733                  */
 734                 tcp_respond(NULL, ip6, tcp, m, ack, seq, flags);
 735         } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
 736 #if 0
 737                 /*
 738                  * Unlike above, the mbufs need to line up with the ip6 hdr,
 739                  * as the contents are read. We need to m_adj() the
 740                  * needed amount.
 741                  * The mbuf will however be thrown away so we can adjust it.
 742                  * Remember we did an m_pullup on it already so we
 743                  * can make some assumptions about contiguousness.
 744                  */
 745                 if (args->L3offset)
 746                         m_adj(m, args->L3offset);
 747 #endif
 748                 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
 749         } else
 750                 m_freem(m);
 751
 752         args->m = NULL;
 753 }
 754
 755 #endif /* INET6 */
 756
 757 static u_int64_t norule_counter;        /* counter for ipfw_log(NULL...) */
 758
 759 #define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0
 760 #define SNP(buf) buf, sizeof(buf)
 761
 762 /*
 763  * We enter here when we have a rule with O_LOG.
 764  * XXX this function alone takes about 2Kbytes of code!
 765  */
 766 static void
 767 ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args,
 768     struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg,
 769     struct ip *ip)
 770 {
 771         struct ether_header *eh = args->eh;
 772         char *action;
 773         int limit_reached = 0;
 774         char action2[40], proto[128], fragment[32];
 775
 776         fragment[0] = '\0';
 777         proto[0] = '\0';
 778
 779         if (f == NULL) {        /* bogus pkt */
 780                 if (verbose_limit != 0 && norule_counter >= verbose_limit)
 781                         return;
 782                 norule_counter++;
 783                 if (norule_counter == verbose_limit)
 784                         limit_reached = verbose_limit;
 785                 action = "Refuse";
 786         } else {        /* O_LOG is the first action, find the real one */
 787                 ipfw_insn *cmd = ACTION_PTR(f);
 788                 ipfw_insn_log *l = (ipfw_insn_log *)cmd;
 789
 790                 if (l->max_log != 0 && l->log_left == 0)
 791                         return;
 792                 l->log_left--;
 793                 if (l->log_left == 0)
 794                         limit_reached = l->max_log;
 795                 cmd += F_LEN(cmd);      /* point to first action */
 796                 if (cmd->opcode == O_ALTQ) {
 797                         ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
 798
 799                         snprintf(SNPARGS(action2, 0), "Altq %d",
 800                                 altq->qid);
 801                         cmd += F_LEN(cmd);
 802                 }
 803                 if (cmd->opcode == O_PROB)
 804                         cmd += F_LEN(cmd);
 805
 806                 if (cmd->opcode == O_TAG)
 807                         cmd += F_LEN(cmd);
 808
 809                 action = action2;
 810                 switch (cmd->opcode) {
 811                 case O_DENY:
 812                         action = "Deny";
 813                         break;
 814
 815                 case O_REJECT:
 816                         if (cmd->arg1==ICMP_REJECT_RST)
 817                                 action = "Reset";
 818                         else if (cmd->arg1==ICMP_UNREACH_HOST)
 819                                 action = "Reject";
 820                         else
 821                                 snprintf(SNPARGS(action2, 0), "Unreach %d",
 822                                         cmd->arg1);
 823                         break;
 824
 825                 case O_UNREACH6:
 826                         if (cmd->arg1==ICMP6_UNREACH_RST)
 827                                 action = "Reset";
 828                         else
 829                                 snprintf(SNPARGS(action2, 0), "Unreach %d",
 830                                         cmd->arg1);
 831                         break;
 832
 833                 case O_ACCEPT:
 834                         action = "Accept";
 835                         break;
 836                 case O_COUNT:
 837                         action = "Count";
 838                         break;
 839                 case O_DIVERT:
 840                         snprintf(SNPARGS(action2, 0), "Divert %d",
 841                                 cmd->arg1);
 842                         break;
 843                 case O_TEE:
 844                         snprintf(SNPARGS(action2, 0), "Tee %d",
 845                                 cmd->arg1);
 846                         break;
 847                 case O_SKIPTO:
 848                         snprintf(SNPARGS(action2, 0), "SkipTo %d",
 849                                 cmd->arg1);
 850                         break;
 851                 case O_PIPE:
 852                         snprintf(SNPARGS(action2, 0), "Pipe %d",
 853                                 cmd->arg1);
 854                         break;
 855                 case O_QUEUE:
 856                         snprintf(SNPARGS(action2, 0), "Queue %d",
 857                                 cmd->arg1);
 858                         break;
 859                 case O_FORWARD_IP: {
 860                         ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd;
 861                         int len;
 862                         struct in_addr dummyaddr;
 863                         if (sa->sa.sin_addr.s_addr == INADDR_ANY)
 864                                 dummyaddr.s_addr = htonl(tablearg);
 865                         else
 866                                 dummyaddr.s_addr = sa->sa.sin_addr.s_addr;
 867
 868                         len = snprintf(SNPARGS(action2, 0), "Forward to %s",
 869                                 inet_ntoa(dummyaddr));
 870
 871                         if (sa->sa.sin_port)
 872                                 snprintf(SNPARGS(action2, len), ":%d",
 873                                     sa->sa.sin_port);
 874                         }
 875                         break;
 876                 case O_NETGRAPH:
 877                         snprintf(SNPARGS(action2, 0), "Netgraph %d",
 878                                 cmd->arg1);
 879                         break;
 880                 case O_NGTEE:
 881                         snprintf(SNPARGS(action2, 0), "Ngtee %d",
 882                                 cmd->arg1);
 883                         break;
 884                 case O_NAT:
 885                         action = "Nat";
 886                         break;
 887                 default:
 888                         action = "UNKNOWN";
 889                         break;
 890                 }
 891         }
 892
 893         if (hlen == 0) {        /* non-ip */
 894                 snprintf(SNPARGS(proto, 0), "MAC");
 895
 896         } else {
 897                 int len;
 898                 char src[48], dst[48];
 899                 struct icmphdr *icmp;
 900                 struct tcphdr *tcp;
 901                 struct udphdr *udp;
 902 #ifdef INET6
 903                 struct ip6_hdr *ip6 = NULL;
 904                 struct icmp6_hdr *icmp6;
 905 #endif
 906                 src[0] = '\0';
 907                 dst[0] = '\0';
 908 #ifdef INET6
 909                 if (IS_IP6_FLOW_ID(&(args->f_id))) {
 910                         char ip6buf[INET6_ADDRSTRLEN];
 911                         snprintf(src, sizeof(src), "[%s]",
 912                             ip6_sprintf(ip6buf, &args->f_id.src_ip6));
 913                         snprintf(dst, sizeof(dst), "[%s]",
 914                             ip6_sprintf(ip6buf, &args->f_id.dst_ip6));
 915
 916                         ip6 = (struct ip6_hdr *)ip;
 917                         tcp = (struct tcphdr *)(((char *)ip) + hlen);
 918                         udp = (struct udphdr *)(((char *)ip) + hlen);
 919                 } else
 920 #endif
 921                 {
 922                         tcp = L3HDR(struct tcphdr, ip);
 923                         udp = L3HDR(struct udphdr, ip);
 924
 925                         inet_ntoa_r(ip->ip_src, src);
 926                         inet_ntoa_r(ip->ip_dst, dst);
 927                 }
 928
 929                 switch (args->f_id.proto) {
 930                 case IPPROTO_TCP:
 931                         len = snprintf(SNPARGS(proto, 0), "TCP %s", src);
 932                         if (offset == 0)
 933                                 snprintf(SNPARGS(proto, len), ":%d %s:%d",
 934                                     ntohs(tcp->th_sport),
 935                                     dst,
 936                                     ntohs(tcp->th_dport));
 937                         else
 938                                 snprintf(SNPARGS(proto, len), " %s", dst);
 939                         break;
 940
 941                 case IPPROTO_UDP:
 942                         len = snprintf(SNPARGS(proto, 0), "UDP %s", src);
 943                         if (offset == 0)
 944                                 snprintf(SNPARGS(proto, len), ":%d %s:%d",
 945                                     ntohs(udp->uh_sport),
 946                                     dst,
 947                                     ntohs(udp->uh_dport));
 948                         else
 949                                 snprintf(SNPARGS(proto, len), " %s", dst);
 950                         break;
 951
 952                 case IPPROTO_ICMP:
 953                         icmp = L3HDR(struct icmphdr, ip);
 954                         if (offset == 0)
 955                                 len = snprintf(SNPARGS(proto, 0),
 956                                     "ICMP:%u.%u ",
 957                                     icmp->icmp_type, icmp->icmp_code);
 958                         else
 959                                 len = snprintf(SNPARGS(proto, 0), "ICMP ");
 960                         len += snprintf(SNPARGS(proto, len), "%s", src);
 961                         snprintf(SNPARGS(proto, len), " %s", dst);
 962                         break;
 963 #ifdef INET6
 964                 case IPPROTO_ICMPV6:
 965                         icmp6 = (struct icmp6_hdr *)(((char *)ip) + hlen);
 966                         if (offset == 0)
 967                                 len = snprintf(SNPARGS(proto, 0),
 968                                     "ICMPv6:%u.%u ",
 969                                     icmp6->icmp6_type, icmp6->icmp6_code);
 970                         else
 971                                 len = snprintf(SNPARGS(proto, 0), "ICMPv6 ");
 972                         len += snprintf(SNPARGS(proto, len), "%s", src);
 973                         snprintf(SNPARGS(proto, len), " %s", dst);
 974                         break;
 975 #endif
 976                 default:
 977                         len = snprintf(SNPARGS(proto, 0), "P:%d %s",
 978                             args->f_id.proto, src);
 979                         snprintf(SNPARGS(proto, len), " %s", dst);
 980                         break;
 981                 }
 982
 983 #ifdef INET6
 984                 if (IS_IP6_FLOW_ID(&(args->f_id))) {
 985                         if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG))
 986                                 snprintf(SNPARGS(fragment, 0),
 987                                     " (frag %08x:%d@%d%s)",
 988                                     args->f_id.frag_id6,
 989                                     ntohs(ip6->ip6_plen) - hlen,
 990                                     ntohs(offset & IP6F_OFF_MASK) << 3,
 991                                     (offset & IP6F_MORE_FRAG) ? "+" : "");
 992                 } else
 993 #endif
 994                 {
 995                         int ip_off, ip_len;
 996                         if (eh != NULL) { /* layer 2 packets are as on the wire */
 997                                 ip_off = ntohs(ip->ip_off);
 998                                 ip_len = ntohs(ip->ip_len);
 999                         } else {
1000                                 ip_off = ip->ip_off;
1001                                 ip_len = ip->ip_len;
1002                         }
1003                         if (ip_off & (IP_MF | IP_OFFMASK))
1004                                 snprintf(SNPARGS(fragment, 0),
1005                                     " (frag %d:%d@%d%s)",
1006                                     ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2),
1007                                     offset << 3,
1008                                     (ip_off & IP_MF) ? "+" : "");
1009                 }
1010         }
1011         if (oif || m->m_pkthdr.rcvif)
1012                 log(LOG_SECURITY | LOG_INFO,
1013                     "ipfw: %d %s %s %s via %s%s\n",
1014                     f ? f->rulenum : -1,
1015                     action, proto, oif ? "out" : "in",
1016                     oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname,
1017                     fragment);
1018         else
1019                 log(LOG_SECURITY | LOG_INFO,
1020                     "ipfw: %d %s %s [no if info]%s\n",
1021                     f ? f->rulenum : -1,
1022                     action, proto, fragment);
1023         if (limit_reached)
1024                 log(LOG_SECURITY | LOG_NOTICE,
1025                     "ipfw: limit %d reached on entry %d\n",
1026                     limit_reached, f ? f->rulenum : -1);
1027 }
1028
1029 /*
1030  * IMPORTANT: the hash function for dynamic rules must be commutative
1031  * in source and destination (ip,port), because rules are bidirectional
1032  * and we want to find both in the same bucket.
1033  */
1034 static __inline int
1035 hash_packet(struct ipfw_flow_id *id)
1036 {
1037         u_int32_t i;
1038
1039 #ifdef INET6
1040         if (IS_IP6_FLOW_ID(id))
1041                 i = hash_packet6(id);
1042         else
1043 #endif /* INET6 */
1044         i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port);
1045         i &= (curr_dyn_buckets - 1);
1046         return i;
1047 }
1048
1049 /**
1050  * unlink a dynamic rule from a chain. prev is a pointer to
1051  * the previous one, q is a pointer to the rule to delete,
1052  * head is a pointer to the head of the queue.
1053  * Modifies q and potentially also head.
1054  */
1055 #define UNLINK_DYN_RULE(prev, head, q) {                                \
1056         ipfw_dyn_rule *old_q = q;                                       \
1057                                                                         \
1058         /* remove a refcount to the parent */                           \
1059         if (q->dyn_type == O_LIMIT)                                     \
1060                 q->parent->count--;                                     \
1061         DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\
1062                 (q->id.src_ip), (q->id.src_port),                       \
1063                 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); )      \
1064         if (prev != NULL)                                               \
1065                 prev->next = q = q->next;                               \
1066         else                                                            \
1067                 head = q = q->next;                                     \
1068         dyn_count--;                                                    \
1069         uma_zfree(ipfw_dyn_rule_zone, old_q); }
1070
1071 #define TIME_LEQ(a,b)       ((int)((a)-(b)) <= 0)
1072
1073 /**
1074  * Remove dynamic rules pointing to "rule", or all of them if rule == NULL.
1075  *
1076  * If keep_me == NULL, rules are deleted even if not expired,
1077  * otherwise only expired rules are removed.
1078  *
1079  * The value of the second parameter is also used to point to identify
1080  * a rule we absolutely do not want to remove (e.g. because we are
1081  * holding a reference to it -- this is the case with O_LIMIT_PARENT
1082  * rules). The pointer is only used for comparison, so any non-null
1083  * value will do.
1084  */
1085 static void
1086 remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me)
1087 {
1088         static u_int32_t last_remove = 0;
1089
1090 #define FORCE (keep_me == NULL)
1091
1092         ipfw_dyn_rule *prev, *q;
1093         int i, pass = 0, max_pass = 0;
1094
1095         IPFW_DYN_LOCK_ASSERT();
1096
1097         if (ipfw_dyn_v == NULL || dyn_count == 0)
1098                 return;
1099         /* do not expire more than once per second, it is useless */
1100         if (!FORCE && last_remove == time_uptime)
1101                 return;
1102         last_remove = time_uptime;
1103
1104         /*
1105          * because O_LIMIT refer to parent rules, during the first pass only
1106          * remove child and mark any pending LIMIT_PARENT, and remove
1107          * them in a second pass.
1108          */
1109 next_pass:
1110         for (i = 0 ; i < curr_dyn_buckets ; i++) {
1111                 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) {
1112                         /*
1113                          * Logic can become complex here, so we split tests.
1114                          */
1115                         if (q == keep_me)
1116                                 goto next;
1117                         if (rule != NULL && rule != q->rule)
1118                                 goto next; /* not the one we are looking for */
1119                         if (q->dyn_type == O_LIMIT_PARENT) {
1120                                 /*
1121                                  * handle parent in the second pass,
1122                                  * record we need one.
1123                                  */
1124                                 max_pass = 1;
1125                                 if (pass == 0)
1126                                         goto next;
1127                                 if (FORCE && q->count != 0 ) {
1128                                         /* XXX should not happen! */
1129                                         printf("ipfw: OUCH! cannot remove rule,"
1130                                              " count %d\n", q->count);
1131                                 }
1132                         } else {
1133                                 if (!FORCE &&
1134                                     !TIME_LEQ( q->expire, time_uptime ))
1135                                         goto next;
1136                         }
1137              if (q->dyn_type != O_LIMIT_PARENT || !q->count) {
1138                      UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
1139                      continue;
1140              }
1141 next:
1142                         prev=q;
1143                         q=q->next;
1144                 }
1145         }
1146         if (pass++ < max_pass)
1147                 goto next_pass;
1148 }
1149
1150
1151 /**
1152  * lookup a dynamic rule.
1153  */
1154 static ipfw_dyn_rule *
1155 lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction,
1156     struct tcphdr *tcp)
1157 {
1158         /*
1159          * stateful ipfw extensions.
1160          * Lookup into dynamic session queue
1161          */
1162 #define MATCH_REVERSE   0
1163 #define MATCH_FORWARD   1
1164 #define MATCH_NONE      2
1165 #define MATCH_UNKNOWN   3
1166         int i, dir = MATCH_NONE;
1167         ipfw_dyn_rule *prev, *q=NULL;
1168
1169         IPFW_DYN_LOCK_ASSERT();
1170
1171         if (ipfw_dyn_v == NULL)
1172                 goto done;      /* not found */
1173         i = hash_packet( pkt );
1174         for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) {
1175                 if (q->dyn_type == O_LIMIT_PARENT && q->count)
1176                         goto next;
1177                 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */
1178                         UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
1179                         continue;
1180                 }
1181                 if (pkt->proto == q->id.proto &&
1182                     q->dyn_type != O_LIMIT_PARENT) {
1183                         if (IS_IP6_FLOW_ID(pkt)) {
1184                             if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6),
1185                                 &(q->id.src_ip6)) &&
1186                             IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6),
1187                                 &(q->id.dst_ip6)) &&
1188                             pkt->src_port == q->id.src_port &&
1189                             pkt->dst_port == q->id.dst_port ) {
1190                                 dir = MATCH_FORWARD;
1191                                 break;
1192                             }
1193                             if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6),
1194                                     &(q->id.dst_ip6)) &&
1195                                 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6),
1196                                     &(q->id.src_ip6)) &&
1197                                 pkt->src_port == q->id.dst_port &&
1198                                 pkt->dst_port == q->id.src_port ) {
1199                                     dir = MATCH_REVERSE;
1200                                     break;
1201                             }
1202                         } else {
1203                             if (pkt->src_ip == q->id.src_ip &&
1204                                 pkt->dst_ip == q->id.dst_ip &&
1205                                 pkt->src_port == q->id.src_port &&
1206                                 pkt->dst_port == q->id.dst_port ) {
1207                                     dir = MATCH_FORWARD;
1208                                     break;
1209                             }
1210                             if (pkt->src_ip == q->id.dst_ip &&
1211                                 pkt->dst_ip == q->id.src_ip &&
1212                                 pkt->src_port == q->id.dst_port &&
1213                                 pkt->dst_port == q->id.src_port ) {
1214                                     dir = MATCH_REVERSE;
1215                                     break;
1216                             }
1217                         }
1218                 }
1219 next:
1220                 prev = q;
1221                 q = q->next;
1222         }
1223         if (q == NULL)
1224                 goto done; /* q = NULL, not found */
1225
1226         if ( prev != NULL) { /* found and not in front */
1227                 prev->next = q->next;
1228                 q->next = ipfw_dyn_v[i];
1229                 ipfw_dyn_v[i] = q;
1230         }
1231         if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */
1232                 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST);
1233
1234 #define BOTH_SYN        (TH_SYN | (TH_SYN << 8))
1235 #define BOTH_FIN        (TH_FIN | (TH_FIN << 8))
1236                 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8);
1237                 switch (q->state) {
1238                 case TH_SYN:                            /* opening */
1239                         q->expire = time_uptime + dyn_syn_lifetime;
1240                         break;
1241
1242                 case BOTH_SYN:                  /* move to established */
1243                 case BOTH_SYN | TH_FIN :        /* one side tries to close */
1244                 case BOTH_SYN | (TH_FIN << 8) :
1245                         if (tcp) {
1246 #define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0)
1247                             u_int32_t ack = ntohl(tcp->th_ack);
1248                             if (dir == MATCH_FORWARD) {
1249                                 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd))
1250                                     q->ack_fwd = ack;
1251                                 else { /* ignore out-of-sequence */
1252                                     break;
1253                                 }
1254                             } else {
1255                                 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev))
1256                                     q->ack_rev = ack;
1257                                 else { /* ignore out-of-sequence */
1258                                     break;
1259                                 }
1260                             }
1261                         }
1262                         q->expire = time_uptime + dyn_ack_lifetime;
1263                         break;
1264
1265                 case BOTH_SYN | BOTH_FIN:       /* both sides closed */
1266                         if (dyn_fin_lifetime >= dyn_keepalive_period)
1267                                 dyn_fin_lifetime = dyn_keepalive_period - 1;
1268                         q->expire = time_uptime + dyn_fin_lifetime;
1269                         break;
1270
1271                 default:
1272 #if 0
1273                         /*
1274                          * reset or some invalid combination, but can also
1275                          * occur if we use keep-state the wrong way.
1276                          */
1277                         if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0)
1278                                 printf("invalid state: 0x%x\n", q->state);
1279 #endif
1280                         if (dyn_rst_lifetime >= dyn_keepalive_period)
1281                                 dyn_rst_lifetime = dyn_keepalive_period - 1;
1282                         q->expire = time_uptime + dyn_rst_lifetime;
1283                         break;
1284                 }
1285         } else if (pkt->proto == IPPROTO_UDP) {
1286                 q->expire = time_uptime + dyn_udp_lifetime;
1287         } else {
1288                 /* other protocols */
1289                 q->expire = time_uptime + dyn_short_lifetime;
1290         }
1291 done:
1292         if (match_direction)
1293                 *match_direction = dir;
1294         return q;
1295 }
1296
1297 static ipfw_dyn_rule *
1298 lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction,
1299     struct tcphdr *tcp)
1300 {
1301         ipfw_dyn_rule *q;
1302
1303         IPFW_DYN_LOCK();
1304         q = lookup_dyn_rule_locked(pkt, match_direction, tcp);
1305         if (q == NULL)
1306                 IPFW_DYN_UNLOCK();
1307         /* NB: return table locked when q is not NULL */
1308         return q;
1309 }
1310
1311 static void
1312 realloc_dynamic_table(void)
1313 {
1314         IPFW_DYN_LOCK_ASSERT();
1315
1316         /*
1317          * Try reallocation, make sure we have a power of 2 and do
1318          * not allow more than 64k entries. In case of overflow,
1319          * default to 1024.
1320          */
1321
1322         if (dyn_buckets > 65536)
1323                 dyn_buckets = 1024;
1324         if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */
1325                 dyn_buckets = curr_dyn_buckets; /* reset */
1326                 return;
1327         }
1328         curr_dyn_buckets = dyn_buckets;
1329         if (ipfw_dyn_v != NULL)
1330                 free(ipfw_dyn_v, M_IPFW);
1331         for (;;) {
1332                 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *),
1333                        M_IPFW, M_NOWAIT | M_ZERO);
1334                 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2)
1335                         break;
1336                 curr_dyn_buckets /= 2;
1337         }
1338 }
1339
1340 /**
1341  * Install state of type 'type' for a dynamic session.
1342  * The hash table contains two type of rules:
1343  * - regular rules (O_KEEP_STATE)
1344  * - rules for sessions with limited number of sess per user
1345  *   (O_LIMIT). When they are created, the parent is
1346  *   increased by 1, and decreased on delete. In this case,
1347  *   the third parameter is the parent rule and not the chain.
1348  * - "parent" rules for the above (O_LIMIT_PARENT).
1349  */
1350 static ipfw_dyn_rule *
1351 add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule)
1352 {
1353         ipfw_dyn_rule *r;
1354         int i;
1355
1356         IPFW_DYN_LOCK_ASSERT();
1357
1358         if (ipfw_dyn_v == NULL ||
1359             (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) {
1360                 realloc_dynamic_table();
1361                 if (ipfw_dyn_v == NULL)
1362                         return NULL; /* failed ! */
1363         }
1364         i = hash_packet(id);
1365
1366         r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO);
1367         if (r == NULL) {
1368                 printf ("ipfw: sorry cannot allocate state\n");
1369                 return NULL;
1370         }
1371
1372         /* increase refcount on parent, and set pointer */
1373         if (dyn_type == O_LIMIT) {
1374                 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule;
1375                 if ( parent->dyn_type != O_LIMIT_PARENT)
1376                         panic("invalid parent");
1377                 parent->count++;
1378                 r->parent = parent;
1379                 rule = parent->rule;
1380         }
1381
1382         r->id = *id;
1383         r->expire = time_uptime + dyn_syn_lifetime;
1384         r->rule = rule;
1385         r->dyn_type = dyn_type;
1386         r->pcnt = r->bcnt = 0;
1387         r->count = 0;
1388
1389         r->bucket = i;
1390         r->next = ipfw_dyn_v[i];
1391         ipfw_dyn_v[i] = r;
1392         dyn_count++;
1393         DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n",
1394            dyn_type,
1395            (r->id.src_ip), (r->id.src_port),
1396            (r->id.dst_ip), (r->id.dst_port),
1397            dyn_count ); )
1398         return r;
1399 }
1400
1401 /**
1402  * lookup dynamic parent rule using pkt and rule as search keys.
1403  * If the lookup fails, then install one.
1404  */
1405 static ipfw_dyn_rule *
1406 lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule)
1407 {
1408         ipfw_dyn_rule *q;
1409         int i;
1410
1411         IPFW_DYN_LOCK_ASSERT();
1412
1413         if (ipfw_dyn_v) {
1414                 int is_v6 = IS_IP6_FLOW_ID(pkt);
1415                 i = hash_packet( pkt );
1416                 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next)
1417                         if (q->dyn_type == O_LIMIT_PARENT &&
1418                             rule== q->rule &&
1419                             pkt->proto == q->id.proto &&
1420                             pkt->src_port == q->id.src_port &&
1421                             pkt->dst_port == q->id.dst_port &&
1422                             (
1423                                 (is_v6 &&
1424                                  IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6),
1425                                         &(q->id.src_ip6)) &&
1426                                  IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6),
1427                                         &(q->id.dst_ip6))) ||
1428                                 (!is_v6 &&
1429                                  pkt->src_ip == q->id.src_ip &&
1430                                  pkt->dst_ip == q->id.dst_ip)
1431                             )
1432                         ) {
1433                                 q->expire = time_uptime + dyn_short_lifetime;
1434                                 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);)
1435                                 return q;
1436                         }
1437         }
1438         return add_dyn_rule(pkt, O_LIMIT_PARENT, rule);
1439 }
1440
1441 /**
1442  * Install dynamic state for rule type cmd->o.opcode
1443  *
1444  * Returns 1 (failure) if state is not installed because of errors or because
1445  * session limitations are enforced.
1446  */
1447 static int
1448 install_state(struct ip_fw *rule, ipfw_insn_limit *cmd,
1449     struct ip_fw_args *args, uint32_t tablearg)
1450 {
1451         static int last_log;
1452         ipfw_dyn_rule *q;
1453         struct in_addr da;
1454         char src[48], dst[48];
1455
1456         src[0] = '\0';
1457         dst[0] = '\0';
1458
1459         DEB(
1460         printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n",
1461             __func__, cmd->o.opcode,
1462             (args->f_id.src_ip), (args->f_id.src_port),
1463             (args->f_id.dst_ip), (args->f_id.dst_port));
1464         )
1465
1466         IPFW_DYN_LOCK();
1467
1468         q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL);
1469
1470         if (q != NULL) {        /* should never occur */
1471                 if (last_log != time_uptime) {
1472                         last_log = time_uptime;
1473                         printf("ipfw: %s: entry already present, done\n",
1474                             __func__);
1475                 }
1476                 IPFW_DYN_UNLOCK();
1477                 return (0);
1478         }
1479
1480         if (dyn_count >= dyn_max)
1481                 /* Run out of slots, try to remove any expired rule. */
1482                 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1);
1483
1484         if (dyn_count >= dyn_max) {
1485                 if (last_log != time_uptime) {
1486                         last_log = time_uptime;
1487                         printf("ipfw: %s: Too many dynamic rules\n", __func__);
1488                 }
1489                 IPFW_DYN_UNLOCK();
1490                 return (1);     /* cannot install, notify caller */
1491         }
1492
1493         switch (cmd->o.opcode) {
1494         case O_KEEP_STATE:      /* bidir rule */
1495                 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule);
1496                 break;
1497
1498         case O_LIMIT: {         /* limit number of sessions */
1499                 struct ipfw_flow_id id;
1500                 ipfw_dyn_rule *parent;
1501                 uint32_t conn_limit;
1502                 uint16_t limit_mask = cmd->limit_mask;
1503
1504                 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ?
1505                     tablearg : cmd->conn_limit;
1506
1507                 DEB(
1508                 if (cmd->conn_limit == IP_FW_TABLEARG)
1509                         printf("ipfw: %s: O_LIMIT rule, conn_limit: %u "
1510                             "(tablearg)\n", __func__, conn_limit);
1511                 else
1512                         printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n",
1513                             __func__, conn_limit);
1514                 )
1515
1516                 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0;
1517                 id.proto = args->f_id.proto;
1518                 id.addr_type = args->f_id.addr_type;
1519
1520                 if (IS_IP6_FLOW_ID (&(args->f_id))) {
1521                         if (limit_mask & DYN_SRC_ADDR)
1522                                 id.src_ip6 = args->f_id.src_ip6;
1523                         if (limit_mask & DYN_DST_ADDR)
1524                                 id.dst_ip6 = args->f_id.dst_ip6;
1525                 } else {
1526                         if (limit_mask & DYN_SRC_ADDR)
1527                                 id.src_ip = args->f_id.src_ip;
1528                         if (limit_mask & DYN_DST_ADDR)
1529                                 id.dst_ip = args->f_id.dst_ip;
1530                 }
1531                 if (limit_mask & DYN_SRC_PORT)
1532                         id.src_port = args->f_id.src_port;
1533                 if (limit_mask & DYN_DST_PORT)
1534                         id.dst_port = args->f_id.dst_port;
1535                 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) {
1536                         printf("ipfw: %s: add parent failed\n", __func__);
1537                         IPFW_DYN_UNLOCK();
1538                         return (1);
1539                 }
1540
1541                 if (parent->count >= conn_limit) {
1542                         /* See if we can remove some expired rule. */
1543                         remove_dyn_rule(rule, parent);
1544                         if (parent->count >= conn_limit) {
1545                                 if (fw_verbose && last_log != time_uptime) {
1546                                         last_log = time_uptime;
1547 #ifdef INET6
1548                                         /*
1549                                          * XXX IPv6 flows are not
1550                                          * supported yet.
1551                                          */
1552                                         if (IS_IP6_FLOW_ID(&(args->f_id))) {
1553                                                 char ip6buf[INET6_ADDRSTRLEN];
1554                                                 snprintf(src, sizeof(src),
1555                                                     "[%s]", ip6_sprintf(ip6buf,
1556                                                         &args->f_id.src_ip6));
1557                                                 snprintf(dst, sizeof(dst),
1558                                                     "[%s]", ip6_sprintf(ip6buf,
1559                                                         &args->f_id.dst_ip6));
1560                                         } else
1561 #endif
1562                                         {
1563                                                 da.s_addr =
1564                                                     htonl(args->f_id.src_ip);
1565                                                 inet_ntoa_r(da, src);
1566                                                 da.s_addr =
1567                                                     htonl(args->f_id.dst_ip);
1568                                                 inet_ntoa_r(da, dst);
1569                                         }
1570                                         log(LOG_SECURITY | LOG_DEBUG,
1571                                             "ipfw: %d %s %s:%u -> %s:%u, %s\n",
1572                                             parent->rule->rulenum,
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_mbuf_create_netlayer(replyto, m);
1623         else
1624                 mac_mbuf_create_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 pf_mtag *at;
3062                                 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
3063
3064                                 match = 1;
3065                                 at = pf_find_mtag(m);
3066                                 if (at != NULL && at->qid != 0)
3067                                         break;
3068                                 at = pf_get_mtag(m);
3069                                 if (at == NULL) {
3070                                         /*
3071                                          * Let the packet fall back to the
3072                                          * default ALTQ.
3073                                          */
3074                                         break;
3075                                 }
3076                                 at->qid = altq->qid;
3077                                 if (is_ipv4)
3078                                         at->af = AF_INET;
3079                                 else
3080                                         at->af = AF_LINK;
3081                                 at->hdr = ip;
3082                                 break;
3083                         }
3084
3085                         case O_LOG:
3086                                 if (fw_verbose)
3087                                         ipfw_log(f, hlen, args, m,
3088                                             oif, offset, tablearg, ip);
3089                                 match = 1;
3090                                 break;
3091
3092                         case O_PROB:
3093                                 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
3094                                 break;
3095
3096                         case O_VERREVPATH:
3097                                 /* Outgoing packets automatically pass/match */
3098                                 match = ((oif != NULL) ||
3099                                     (m->m_pkthdr.rcvif == NULL) ||
3100                                     (
3101 #ifdef INET6
3102                                     is_ipv6 ?
3103                                         verify_path6(&(args->f_id.src_ip6),
3104                                             m->m_pkthdr.rcvif) :
3105 #endif
3106                                     verify_path(src_ip, m->m_pkthdr.rcvif)));
3107                                 break;
3108
3109                         case O_VERSRCREACH:
3110                                 /* Outgoing packets automatically pass/match */
3111                                 match = (hlen > 0 && ((oif != NULL) ||
3112 #ifdef INET6
3113                                     is_ipv6 ?
3114                                         verify_path6(&(args->f_id.src_ip6),
3115                                             NULL) :
3116 #endif
3117                                     verify_path(src_ip, NULL)));
3118                                 break;
3119
3120                         case O_ANTISPOOF:
3121                                 /* Outgoing packets automatically pass/match */
3122                                 if (oif == NULL && hlen > 0 &&
3123                                     (  (is_ipv4 && in_localaddr(src_ip))
3124 #ifdef INET6
3125                                     || (is_ipv6 &&
3126                                         in6_localaddr(&(args->f_id.src_ip6)))
3127 #endif
3128                                     ))
3129                                         match =
3130 #ifdef INET6
3131                                             is_ipv6 ? verify_path6(
3132                                                 &(args->f_id.src_ip6),
3133                                                 m->m_pkthdr.rcvif) :
3134 #endif
3135                                             verify_path(src_ip,
3136                                                 m->m_pkthdr.rcvif);
3137                                 else
3138                                         match = 1;
3139                                 break;
3140
3141                         case O_IPSEC:
3142 #ifdef IPSEC
3143                                 match = (m_tag_find(m,
3144                                     PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
3145 #endif
3146                                 /* otherwise no match */
3147                                 break;
3148
3149 #ifdef INET6
3150                         case O_IP6_SRC:
3151                                 match = is_ipv6 &&
3152                                     IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
3153                                     &((ipfw_insn_ip6 *)cmd)->addr6);
3154                                 break;
3155
3156                         case O_IP6_DST:
3157                                 match = is_ipv6 &&
3158                                 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
3159                                     &((ipfw_insn_ip6 *)cmd)->addr6);
3160                                 break;
3161                         case O_IP6_SRC_MASK:
3162                         case O_IP6_DST_MASK:
3163                                 if (is_ipv6) {
3164                                         int i = cmdlen - 1;
3165                                         struct in6_addr p;
3166                                         struct in6_addr *d =
3167                                             &((ipfw_insn_ip6 *)cmd)->addr6;
3168
3169                                         for (; !match && i > 0; d += 2,
3170                                             i -= F_INSN_SIZE(struct in6_addr)
3171                                             * 2) {
3172                                                 p = (cmd->opcode ==
3173                                                     O_IP6_SRC_MASK) ?
3174                                                     args->f_id.src_ip6:
3175                                                     args->f_id.dst_ip6;
3176                                                 APPLY_MASK(&p, &d[1]);
3177                                                 match =
3178                                                     IN6_ARE_ADDR_EQUAL(&d[0],
3179                                                     &p);
3180                                         }
3181                                 }
3182                                 break;
3183
3184                         case O_IP6_SRC_ME:
3185                                 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6);
3186                                 break;
3187
3188                         case O_IP6_DST_ME:
3189                                 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6);
3190                                 break;
3191
3192                         case O_FLOW6ID:
3193                                 match = is_ipv6 &&
3194                                     flow6id_match(args->f_id.flow_id6,
3195                                     (ipfw_insn_u32 *) cmd);
3196                                 break;
3197
3198                         case O_EXT_HDR:
3199                                 match = is_ipv6 &&
3200                                     (ext_hd & ((ipfw_insn *) cmd)->arg1);
3201                                 break;
3202
3203                         case O_IP6:
3204                                 match = is_ipv6;
3205                                 break;
3206 #endif
3207
3208                         case O_IP4:
3209                                 match = is_ipv4;
3210                                 break;
3211
3212                         case O_TAG: {
3213                                 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
3214                                     tablearg : cmd->arg1;
3215
3216                                 /* Packet is already tagged with this tag? */
3217                                 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
3218
3219                                 /* We have `untag' action when F_NOT flag is
3220                                  * present. And we must remove this mtag from
3221                                  * mbuf and reset `match' to zero (`match' will
3222                                  * be inversed later).
3223                                  * Otherwise we should allocate new mtag and
3224                                  * push it into mbuf.
3225                                  */
3226                                 if (cmd->len & F_NOT) { /* `untag' action */
3227                                         if (mtag != NULL)
3228                                                 m_tag_delete(m, mtag);
3229                                 } else if (mtag == NULL) {
3230                                         if ((mtag = m_tag_alloc(MTAG_IPFW,
3231                                             tag, 0, M_NOWAIT)) != NULL)
3232                                                 m_tag_prepend(m, mtag);
3233                                 }
3234                                 match = (cmd->len & F_NOT) ? 0: 1;
3235                                 break;
3236                         }
3237
3238                         case O_TAGGED: {
3239                                 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
3240                                     tablearg : cmd->arg1;
3241
3242                                 if (cmdlen == 1) {
3243                                         match = m_tag_locate(m, MTAG_IPFW,
3244                                             tag, NULL) != NULL;
3245                                         break;
3246                                 }
3247
3248                                 /* we have ranges */
3249                                 for (mtag = m_tag_first(m);
3250                                     mtag != NULL && !match;
3251                                     mtag = m_tag_next(m, mtag)) {
3252                                         uint16_t *p;
3253                                         int i;
3254
3255                                         if (mtag->m_tag_cookie != MTAG_IPFW)
3256                                                 continue;
3257
3258                                         p = ((ipfw_insn_u16 *)cmd)->ports;
3259                                         i = cmdlen - 1;
3260                                         for(; !match && i > 0; i--, p += 2)
3261                                                 match =
3262                                                     mtag->m_tag_id >= p[0] &&
3263                                                     mtag->m_tag_id <= p[1];
3264                                 }
3265                                 break;
3266                         }
3267
3268                         /*
3269                          * The second set of opcodes represents 'actions',
3270                          * i.e. the terminal part of a rule once the packet
3271                          * matches all previous patterns.
3272                          * Typically there is only one action for each rule,
3273                          * and the opcode is stored at the end of the rule
3274                          * (but there are exceptions -- see below).
3275                          *
3276                          * In general, here we set retval and terminate the
3277                          * outer loop (would be a 'break 3' in some language,
3278                          * but we need to do a 'goto done').
3279                          *
3280                          * Exceptions:
3281                          * O_COUNT and O_SKIPTO actions:
3282                          *   instead of terminating, we jump to the next rule
3283                          *   ('goto next_rule', equivalent to a 'break 2'),
3284                          *   or to the SKIPTO target ('goto again' after
3285                          *   having set f, cmd and l), respectively.
3286                          *
3287                          * O_TAG, O_LOG and O_ALTQ action parameters:
3288                          *   perform some action and set match = 1;
3289                          *
3290                          * O_LIMIT and O_KEEP_STATE: these opcodes are
3291                          *   not real 'actions', and are stored right
3292                          *   before the 'action' part of the rule.
3293                          *   These opcodes try to install an entry in the
3294                          *   state tables; if successful, we continue with
3295                          *   the next opcode (match=1; break;), otherwise
3296                          *   the packet *   must be dropped
3297                          *   ('goto done' after setting retval);
3298                          *
3299                          * O_PROBE_STATE and O_CHECK_STATE: these opcodes
3300                          *   cause a lookup of the state table, and a jump
3301                          *   to the 'action' part of the parent rule
3302                          *   ('goto check_body') if an entry is found, or
3303                          *   (CHECK_STATE only) a jump to the next rule if
3304                          *   the entry is not found ('goto next_rule').
3305                          *   The result of the lookup is cached to make
3306                          *   further instances of these opcodes are
3307                          *   effectively NOPs.
3308                          */
3309                         case O_LIMIT:
3310                         case O_KEEP_STATE:
3311                                 if (install_state(f,
3312                                     (ipfw_insn_limit *)cmd, args, tablearg)) {
3313                                         retval = IP_FW_DENY;
3314                                         goto done; /* error/limit violation */
3315                                 }
3316                                 match = 1;
3317                                 break;
3318
3319                         case O_PROBE_STATE:
3320                         case O_CHECK_STATE:
3321                                 /*
3322                                  * dynamic rules are checked at the first
3323                                  * keep-state or check-state occurrence,
3324                                  * with the result being stored in dyn_dir.
3325                                  * The compiler introduces a PROBE_STATE
3326                                  * instruction for us when we have a
3327                                  * KEEP_STATE (because PROBE_STATE needs
3328                                  * to be run first).
3329                                  */
3330                                 if (dyn_dir == MATCH_UNKNOWN &&
3331                                     (q = lookup_dyn_rule(&args->f_id,
3332                                      &dyn_dir, proto == IPPROTO_TCP ?
3333                                         TCP(ulp) : NULL))
3334                                         != NULL) {
3335                                         /*
3336                                          * Found dynamic entry, update stats
3337                                          * and jump to the 'action' part of
3338                                          * the parent rule.
3339                                          */
3340                                         q->pcnt++;
3341                                         q->bcnt += pktlen;
3342                                         f = q->rule;
3343                                         cmd = ACTION_PTR(f);
3344                                         l = f->cmd_len - f->act_ofs;
3345                                         IPFW_DYN_UNLOCK();
3346                                         goto check_body;
3347                                 }
3348                                 /*
3349                                  * Dynamic entry not found. If CHECK_STATE,
3350                                  * skip to next rule, if PROBE_STATE just
3351                                  * ignore and continue with next opcode.
3352                                  */
3353                                 if (cmd->opcode == O_CHECK_STATE)
3354                                         goto next_rule;
3355                                 match = 1;
3356                                 break;
3357
3358                         case O_ACCEPT:
3359                                 retval = 0;     /* accept */
3360                                 goto done;
3361
3362                         case O_PIPE:
3363                         case O_QUEUE:
3364                                 args->rule = f; /* report matching rule */
3365                                 if (cmd->arg1 == IP_FW_TABLEARG)
3366                                         args->cookie = tablearg;
3367                                 else
3368                                         args->cookie = cmd->arg1;
3369                                 retval = IP_FW_DUMMYNET;
3370                                 goto done;
3371
3372                         case O_DIVERT:
3373                         case O_TEE: {
3374                                 struct divert_tag *dt;
3375
3376                                 if (args->eh) /* not on layer 2 */
3377                                         break;
3378                                 mtag = m_tag_get(PACKET_TAG_DIVERT,
3379                                                 sizeof(struct divert_tag),
3380                                                 M_NOWAIT);
3381                                 if (mtag == NULL) {
3382                                         /* XXX statistic */
3383                                         /* drop packet */
3384                                         IPFW_RUNLOCK(chain);
3385                                         return (IP_FW_DENY);
3386                                 }
3387                                 dt = (struct divert_tag *)(mtag+1);
3388                                 dt->cookie = f->rulenum;
3389                                 if (cmd->arg1 == IP_FW_TABLEARG)
3390                                         dt->info = tablearg;
3391                                 else
3392                                         dt->info = cmd->arg1;
3393                                 m_tag_prepend(m, mtag);
3394                                 retval = (cmd->opcode == O_DIVERT) ?
3395                                     IP_FW_DIVERT : IP_FW_TEE;
3396                                 goto done;
3397                         }
3398
3399                         case O_COUNT:
3400                         case O_SKIPTO:
3401                                 f->pcnt++;      /* update stats */
3402                                 f->bcnt += pktlen;
3403                                 f->timestamp = time_uptime;
3404                                 if (cmd->opcode == O_COUNT)
3405                                         goto next_rule;
3406                                 /* handle skipto */
3407                                 if (f->next_rule == NULL)
3408                                         lookup_next_rule(f);
3409                                 f = f->next_rule;
3410                                 goto again;
3411
3412                         case O_REJECT:
3413                                 /*
3414                                  * Drop the packet and send a reject notice
3415                                  * if the packet is not ICMP (or is an ICMP
3416                                  * query), and it is not multicast/broadcast.
3417                                  */
3418                                 if (hlen > 0 && is_ipv4 && offset == 0 &&
3419                                     (proto != IPPROTO_ICMP ||
3420                                      is_icmp_query(ICMP(ulp))) &&
3421                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
3422                                     !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
3423                                         send_reject(args, cmd->arg1, ip_len, ip);
3424                                         m = args->m;
3425                                 }
3426                                 /* FALLTHROUGH */
3427 #ifdef INET6
3428                         case O_UNREACH6:
3429                                 if (hlen > 0 && is_ipv6 &&
3430                                     ((offset & IP6F_OFF_MASK) == 0) &&
3431                                     (proto != IPPROTO_ICMPV6 ||
3432                                      (is_icmp6_query(args->f_id.flags) == 1)) &&
3433                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
3434                                     !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
3435                                         send_reject6(
3436                                             args, cmd->arg1, hlen,
3437                                             (struct ip6_hdr *)ip);
3438                                         m = args->m;
3439                                 }
3440                                 /* FALLTHROUGH */
3441 #endif
3442                         case O_DENY:
3443                                 retval = IP_FW_DENY;
3444                                 goto done;
3445
3446                         case O_FORWARD_IP: {
3447                                 struct sockaddr_in *sa;
3448                                 sa = &(((ipfw_insn_sa *)cmd)->sa);
3449                                 if (args->eh)   /* not valid on layer2 pkts */
3450                                         break;
3451                                 if (!q || dyn_dir == MATCH_FORWARD) {
3452                                         if (sa->sin_addr.s_addr == INADDR_ANY) {
3453                                                 bcopy(sa, &args->hopstore,
3454                                                         sizeof(*sa));
3455                                                 args->hopstore.sin_addr.s_addr =
3456                                                     htonl(tablearg);
3457                                                 args->next_hop =
3458                                                     &args->hopstore;
3459                                         } else {
3460                                                 args->next_hop = sa;
3461                                         }
3462                                 }
3463                                 retval = IP_FW_PASS;
3464                             }
3465                             goto done;
3466
3467                         case O_NETGRAPH:
3468                         case O_NGTEE:
3469                                 args->rule = f; /* report matching rule */
3470                                 if (cmd->arg1 == IP_FW_TABLEARG)
3471                                         args->cookie = tablearg;
3472                                 else
3473                                         args->cookie = cmd->arg1;
3474                                 retval = (cmd->opcode == O_NETGRAPH) ?
3475                                     IP_FW_NETGRAPH : IP_FW_NGTEE;
3476                                 goto done;
3477
3478 #ifdef IPFIREWALL_NAT
3479                         case O_NAT: {
3480                                 struct cfg_nat *t;
3481                                 struct mbuf *mcl;
3482                                 /* XXX - libalias duct tape */
3483                                 int ldt;
3484                                 char *c;
3485
3486                                 ldt = 0;
3487                                 args->rule = f; /* Report matching rule. */
3488                                 retval = 0;
3489                                 t = ((ipfw_insn_nat *)cmd)->nat;
3490                                 if (t == NULL) {
3491                                         t = lookup_nat(cmd->arg1);
3492                                         if (t == NULL) {
3493                                                 retval = IP_FW_DENY;
3494                                                 goto done;
3495                                         } else
3496                                                 ((ipfw_insn_nat *)cmd)->nat =
3497                                                     t;
3498                                 }
3499                                 if ((mcl = m_megapullup(m, m->m_pkthdr.len)) ==
3500                                     NULL)
3501                                         goto badnat;
3502                                 ip = mtod(mcl, struct ip *);
3503                                 if (args->eh == NULL) {
3504                                         ip->ip_len = htons(ip->ip_len);
3505                                         ip->ip_off = htons(ip->ip_off);
3506                                 }
3507
3508                                 /*
3509                                  * XXX - Libalias checksum offload 'duct tape':
3510                                  *
3511                                  * locally generated packets have only
3512                                  * pseudo-header checksum calculated
3513                                  * and libalias will screw it[1], so
3514                                  * mark them for later fix.  Moreover
3515                                  * there are cases when libalias
3516                                  * modify tcp packet data[2], mark it
3517                                  * for later fix too.
3518                                  *
3519                                  * [1] libalias was never meant to run
3520                                  * in kernel, so it doesn't have any
3521                                  * knowledge about checksum
3522                                  * offloading, and it expects a packet
3523                                  * with a full internet
3524                                  * checksum. Unfortunately, packets
3525                                  * generated locally will have just the
3526                                  * pseudo header calculated, and when
3527                                  * libalias tries to adjust the
3528                                  * checksum it will actually screw it.
3529                                  *
3530                                  * [2] when libalias modify tcp's data
3531                                  * content, full TCP checksum has to
3532                                  * be recomputed: the problem is that
3533                                  * libalias doesn't have any idea
3534                                  * about checksum offloading To
3535                                  * workaround this, we do not do
3536                                  * checksumming in LibAlias, but only
3537                                  * mark the packets in th_x2 field. If
3538                                  * we receive a marked packet, we
3539                                  * calculate correct checksum for it
3540                                  * aware of offloading.  Why such a
3541                                  * terrible hack instead of
3542                                  * recalculating checksum for each
3543                                  * packet?  Because the previous
3544                                  * checksum was not checked!
3545                                  * Recalculating checksums for EVERY
3546                                  * packet will hide ALL transmission
3547                                  * errors. Yes, marked packets still
3548                                  * suffer from this problem. But,
3549                                  * sigh, natd(8) has this problem,
3550                                  * too.
3551                                  *
3552                                  * TODO: -make libalias mbuf aware (so
3553                                  * it can handle delayed checksum and tso)
3554                                  */
3555
3556                                 if (mcl->m_pkthdr.rcvif == NULL &&
3557                                     mcl->m_pkthdr.csum_flags &
3558                                     CSUM_DELAY_DATA)
3559                                         ldt = 1;
3560
3561                                 c = mtod(mcl, char *);
3562                                 if (oif == NULL)
3563                                         retval = LibAliasIn(t->lib, c,
3564                                             MCLBYTES);
3565                                 else
3566                                         retval = LibAliasOut(t->lib, c,
3567                                             MCLBYTES);
3568                                 if (retval != PKT_ALIAS_OK) {
3569                                         /* XXX - should i add some logging? */
3570                                         m_free(mcl);
3571                                 badnat:
3572                                         args->m = NULL;
3573                                         retval = IP_FW_DENY;
3574                                         goto done;
3575                                 }
3576                                 mcl->m_pkthdr.len = mcl->m_len =
3577                                     ntohs(ip->ip_len);
3578
3579                                 /*
3580                                  * XXX - libalias checksum offload
3581                                  * 'duct tape' (see above)
3582                                  */
3583
3584                                 if ((ip->ip_off & htons(IP_OFFMASK)) == 0 &&
3585                                     ip->ip_p == IPPROTO_TCP) {
3586                                         struct tcphdr   *th;
3587
3588                                         th = (struct tcphdr *)(ip + 1);
3589                                         if (th->th_x2)
3590                                                 ldt = 1;
3591                                 }
3592
3593                                 if (ldt) {
3594                                         struct tcphdr   *th;
3595                                         struct udphdr   *uh;
3596                                         u_short cksum;
3597
3598                                         ip->ip_len = ntohs(ip->ip_len);
3599                                         cksum = in_pseudo(
3600                                                 ip->ip_src.s_addr,
3601                                                 ip->ip_dst.s_addr,
3602                                                 htons(ip->ip_p + ip->ip_len -
3603                                                     (ip->ip_hl << 2))
3604                                                 );
3605
3606                                         switch (ip->ip_p) {
3607                                         case IPPROTO_TCP:
3608                                                 th = (struct tcphdr *)(ip + 1);
3609                                                 /*
3610                                                  * Maybe it was set in
3611                                                  * libalias...
3612                                                  */
3613                                                 th->th_x2 = 0;
3614                                                 th->th_sum = cksum;
3615                                                 mcl->m_pkthdr.csum_data =
3616                                                     offsetof(struct tcphdr,
3617                                                     th_sum);
3618                                                 break;
3619                                         case IPPROTO_UDP:
3620                                                 uh = (struct udphdr *)(ip + 1);
3621                                                 uh->uh_sum = cksum;
3622                                                 mcl->m_pkthdr.csum_data =
3623                                                     offsetof(struct udphdr,
3624                                                     uh_sum);
3625                                                 break;
3626                                         }
3627                                         /*
3628                                          * No hw checksum offloading: do it
3629                                          * by ourself.
3630                                          */
3631                                         if ((mcl->m_pkthdr.csum_flags &
3632                                              CSUM_DELAY_DATA) == 0) {
3633                                                 in_delayed_cksum(mcl);
3634                                                 mcl->m_pkthdr.csum_flags &=
3635                                                     ~CSUM_DELAY_DATA;
3636                                         }
3637                                         ip->ip_len = htons(ip->ip_len);
3638                                 }
3639
3640                                 if (args->eh == NULL) {
3641                                         ip->ip_len = ntohs(ip->ip_len);
3642                                         ip->ip_off = ntohs(ip->ip_off);
3643                                 }
3644
3645                                 args->m = mcl;
3646                                 retval = IP_FW_NAT;
3647                                 goto done;
3648                         }
3649 #endif
3650
3651                         default:
3652                                 panic("-- unknown opcode %d\n", cmd->opcode);
3653                         } /* end of switch() on opcodes */
3654
3655                         if (cmd->len & F_NOT)
3656                                 match = !match;
3657
3658                         if (match) {
3659                                 if (cmd->len & F_OR)
3660                                         skip_or = 1;
3661                         } else {
3662                                 if (!(cmd->len & F_OR)) /* not an OR block, */
3663                                         break;          /* try next rule    */
3664                         }
3665
3666                 }       /* end of inner for, scan opcodes */
3667
3668 next_rule:;             /* try next rule                */
3669
3670         }               /* end of outer for, scan rules */
3671         printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3672         IPFW_RUNLOCK(chain);
3673         return (IP_FW_DENY);
3674
3675 done:
3676         /* Update statistics */
3677         f->pcnt++;
3678         f->bcnt += pktlen;
3679         f->timestamp = time_uptime;
3680         IPFW_RUNLOCK(chain);
3681         return (retval);
3682
3683 pullup_failed:
3684         if (fw_verbose)
3685                 printf("ipfw: pullup failed\n");
3686         return (IP_FW_DENY);
3687 }
3688
3689 /*
3690  * When a rule is added/deleted, clear the next_rule pointers in all rules.
3691  * These will be reconstructed on the fly as packets are matched.
3692  */
3693 static void
3694 flush_rule_ptrs(struct ip_fw_chain *chain)
3695 {
3696         struct ip_fw *rule;
3697
3698         IPFW_WLOCK_ASSERT(chain);
3699
3700         for (rule = chain->rules; rule; rule = rule->next)
3701                 rule->next_rule = NULL;
3702 }
3703
3704 /*
3705  * Add a new rule to the list. Copy the rule into a malloc'ed area, then
3706  * possibly create a rule number and add the rule to the list.
3707  * Update the rule_number in the input struct so the caller knows it as well.
3708  */
3709 static int
3710 add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule)
3711 {
3712         struct ip_fw *rule, *f, *prev;
3713         int l = RULESIZE(input_rule);
3714
3715         if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE)
3716                 return (EINVAL);
3717
3718         rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO);
3719         if (rule == NULL)
3720                 return (ENOSPC);
3721
3722         bcopy(input_rule, rule, l);
3723
3724         rule->next = NULL;
3725         rule->next_rule = NULL;
3726
3727         rule->pcnt = 0;
3728         rule->bcnt = 0;
3729         rule->timestamp = 0;
3730
3731         IPFW_WLOCK(chain);
3732
3733         if (chain->rules == NULL) {     /* default rule */
3734                 chain->rules = rule;
3735                 goto done;
3736         }
3737
3738         /*
3739          * If rulenum is 0, find highest numbered rule before the
3740          * default rule, and add autoinc_step
3741          */
3742         if (autoinc_step < 1)
3743                 autoinc_step = 1;
3744         else if (autoinc_step > 1000)
3745                 autoinc_step = 1000;
3746         if (rule->rulenum == 0) {
3747                 /*
3748                  * locate the highest numbered rule before default
3749                  */
3750                 for (f = chain->rules; f; f = f->next) {
3751                         if (f->rulenum == IPFW_DEFAULT_RULE)
3752                                 break;
3753                         rule->rulenum = f->rulenum;
3754                 }
3755                 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step)
3756                         rule->rulenum += autoinc_step;
3757                 input_rule->rulenum = rule->rulenum;
3758         }
3759
3760         /*
3761          * Now insert the new rule in the right place in the sorted list.
3762          */
3763         for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) {
3764                 if (f->rulenum > rule->rulenum) { /* found the location */
3765                         if (prev) {
3766                                 rule->next = f;
3767                                 prev->next = rule;
3768                         } else { /* head insert */
3769                                 rule->next = chain->rules;
3770                                 chain->rules = rule;
3771                         }
3772                         break;
3773                 }
3774         }
3775         flush_rule_ptrs(chain);
3776 done:
3777         static_count++;
3778         static_len += l;
3779         IPFW_WUNLOCK(chain);
3780         DEB(printf("ipfw: installed rule %d, static count now %d\n",
3781                 rule->rulenum, static_count);)
3782         return (0);
3783 }
3784
3785 /**
3786  * Remove a static rule (including derived * dynamic rules)
3787  * and place it on the ``reap list'' for later reclamation.
3788  * The caller is in charge of clearing rule pointers to avoid
3789  * dangling pointers.
3790  * @return a pointer to the next entry.
3791  * Arguments are not checked, so they better be correct.
3792  */
3793 static struct ip_fw *
3794 remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule,
3795     struct ip_fw *prev)
3796 {
3797         struct ip_fw *n;
3798         int l = RULESIZE(rule);
3799
3800         IPFW_WLOCK_ASSERT(chain);
3801
3802         n = rule->next;
3803         IPFW_DYN_LOCK();
3804         remove_dyn_rule(rule, NULL /* force removal */);
3805         IPFW_DYN_UNLOCK();
3806         if (prev == NULL)
3807                 chain->rules = n;
3808         else
3809                 prev->next = n;
3810         static_count--;
3811         static_len -= l;
3812
3813         rule->next = chain->reap;
3814         chain->reap = rule;
3815
3816         return n;
3817 }
3818
3819 /**
3820  * Reclaim storage associated with a list of rules.  This is
3821  * typically the list created using remove_rule.
3822  */
3823 static void
3824 reap_rules(struct ip_fw *head)
3825 {
3826         struct ip_fw *rule;
3827
3828         while ((rule = head) != NULL) {
3829                 head = head->next;
3830                 if (DUMMYNET_LOADED)
3831                         ip_dn_ruledel_ptr(rule);
3832                 free(rule, M_IPFW);
3833         }
3834 }
3835
3836 /*
3837  * Remove all rules from a chain (except rules in set RESVD_SET
3838  * unless kill_default = 1).  The caller is responsible for
3839  * reclaiming storage for the rules left in chain->reap.
3840  */
3841 static void
3842 free_chain(struct ip_fw_chain *chain, int kill_default)
3843 {
3844         struct ip_fw *prev, *rule;
3845
3846         IPFW_WLOCK_ASSERT(chain);
3847
3848         flush_rule_ptrs(chain); /* more efficient to do outside the loop */
3849         for (prev = NULL, rule = chain->rules; rule ; )
3850                 if (kill_default || rule->set != RESVD_SET)
3851                         rule = remove_rule(chain, rule, prev);
3852                 else {
3853                         prev = rule;
3854                         rule = rule->next;
3855                 }
3856 }
3857
3858 /**
3859  * Remove all rules with given number, and also do set manipulation.
3860  * Assumes chain != NULL && *chain != NULL.
3861  *
3862  * The argument is an u_int32_t. The low 16 bit are the rule or set number,
3863  * the next 8 bits are the new set, the top 8 bits are the command:
3864  *
3865  *      0       delete rules with given number
3866  *      1       delete rules with given set number
3867  *      2       move rules with given number to new set
3868  *      3       move rules with given set number to new set
3869  *      4       swap sets with given numbers
3870  *      5       delete rules with given number and with given set number
3871  */
3872 static int
3873 del_entry(struct ip_fw_chain *chain, u_int32_t arg)
3874 {
3875         struct ip_fw *prev = NULL, *rule;
3876         u_int16_t rulenum;      /* rule or old_set */
3877         u_int8_t cmd, new_set;
3878
3879         rulenum = arg & 0xffff;
3880         cmd = (arg >> 24) & 0xff;
3881         new_set = (arg >> 16) & 0xff;
3882
3883         if (cmd > 5 || new_set > RESVD_SET)
3884                 return EINVAL;
3885         if (cmd == 0 || cmd == 2 || cmd == 5) {
3886                 if (rulenum >= IPFW_DEFAULT_RULE)
3887                         return EINVAL;
3888         } else {
3889                 if (rulenum > RESVD_SET)        /* old_set */
3890                         return EINVAL;
3891         }
3892
3893         IPFW_WLOCK(chain);
3894         rule = chain->rules;
3895         chain->reap = NULL;
3896         switch (cmd) {
3897         case 0: /* delete rules with given number */
3898                 /*
3899                  * locate first rule to delete
3900                  */
3901                 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next)
3902                         ;
3903                 if (rule->rulenum != rulenum) {
3904                         IPFW_WUNLOCK(chain);
3905                         return EINVAL;
3906                 }
3907
3908                 /*
3909                  * flush pointers outside the loop, then delete all matching
3910                  * rules. prev remains the same throughout the cycle.
3911                  */
3912                 flush_rule_ptrs(chain);
3913                 while (rule->rulenum == rulenum)
3914                         rule = remove_rule(chain, rule, prev);
3915                 break;
3916
3917         case 1: /* delete all rules with given set number */
3918                 flush_rule_ptrs(chain);
3919                 rule = chain->rules;
3920                 while (rule->rulenum < IPFW_DEFAULT_RULE)
3921                         if (rule->set == rulenum)
3922                                 rule = remove_rule(chain, rule, prev);
3923                         else {
3924                                 prev = rule;
3925                                 rule = rule->next;
3926                         }
3927                 break;
3928
3929         case 2: /* move rules with given number to new set */
3930                 rule = chain->rules;
3931                 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
3932                         if (rule->rulenum == rulenum)
3933                                 rule->set = new_set;
3934                 break;
3935
3936         case 3: /* move rules with given set number to new set */
3937                 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
3938                         if (rule->set == rulenum)
3939                                 rule->set = new_set;
3940                 break;
3941
3942         case 4: /* swap two sets */
3943                 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
3944                         if (rule->set == rulenum)
3945                                 rule->set = new_set;
3946                         else if (rule->set == new_set)
3947                                 rule->set = rulenum;
3948                 break;
3949         case 5: /* delete rules with given number and with given set number.
3950                  * rulenum - given rule number;
3951                  * new_set - given set number.
3952                  */
3953                 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next)
3954                         ;
3955                 if (rule->rulenum != rulenum) {
3956                         IPFW_WUNLOCK(chain);
3957                         return (EINVAL);
3958                 }
3959                 flush_rule_ptrs(chain);
3960                 while (rule->rulenum == rulenum) {
3961                         if (rule->set == new_set)
3962                                 rule = remove_rule(chain, rule, prev);
3963                         else {
3964                                 prev = rule;
3965                                 rule = rule->next;
3966                         }
3967                 }
3968         }
3969         /*
3970          * Look for rules to reclaim.  We grab the list before
3971          * releasing the lock then reclaim them w/o the lock to
3972          * avoid a LOR with dummynet.
3973          */
3974         rule = chain->reap;
3975         chain->reap = NULL;
3976         IPFW_WUNLOCK(chain);
3977         if (rule)
3978                 reap_rules(rule);
3979         return 0;
3980 }
3981
3982 /*
3983  * Clear counters for a specific rule.
3984  * The enclosing "table" is assumed locked.
3985  */
3986 static void
3987 clear_counters(struct ip_fw *rule, int log_only)
3988 {
3989         ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule);
3990
3991         if (log_only == 0) {
3992                 rule->bcnt = rule->pcnt = 0;
3993                 rule->timestamp = 0;
3994         }
3995         if (l->o.opcode == O_LOG)
3996                 l->log_left = l->max_log;
3997 }
3998
3999 /**
4000  * Reset some or all counters on firewall rules.
4001  * The argument `arg' is an u_int32_t. The low 16 bit are the rule number,
4002  * the next 8 bits are the set number, the top 8 bits are the command:
4003  *      0       work with rules from all set's;
4004  *      1       work with rules only from specified set.
4005  * Specified rule number is zero if we want to clear all entries.
4006  * log_only is 1 if we only want to reset logs, zero otherwise.
4007  */
4008 static int
4009 zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only)
4010 {
4011         struct ip_fw *rule;
4012         char *msg;
4013
4014         uint16_t rulenum = arg & 0xffff;
4015         uint8_t set = (arg >> 16) & 0xff;
4016         uint8_t cmd = (arg >> 24) & 0xff;
4017
4018         if (cmd > 1)
4019                 return (EINVAL);
4020         if (cmd == 1 && set > RESVD_SET)
4021                 return (EINVAL);
4022
4023         IPFW_WLOCK(chain);
4024         if (rulenum == 0) {
4025                 norule_counter = 0;
4026                 for (rule = chain->rules; rule; rule = rule->next) {
4027                         /* Skip rules from another set. */
4028                         if (cmd == 1 && rule->set != set)
4029                                 continue;
4030                         clear_counters(rule, log_only);
4031                 }
4032                 msg = log_only ? "ipfw: All logging counts reset.\n" :
4033                     "ipfw: Accounting cleared.\n";
4034         } else {
4035                 int cleared = 0;
4036                 /*
4037                  * We can have multiple rules with the same number, so we
4038                  * need to clear them all.
4039                  */
4040                 for (rule = chain->rules; rule; rule = rule->next)
4041                         if (rule->rulenum == rulenum) {
4042                                 while (rule && rule->rulenum == rulenum) {
4043                                         if (cmd == 0 || rule->set == set)
4044                                                 clear_counters(rule, log_only);
4045                                         rule = rule->next;
4046                                 }
4047                                 cleared = 1;
4048                                 break;
4049                         }
4050                 if (!cleared) { /* we did not find any matching rules */
4051                         IPFW_WUNLOCK(chain);
4052                         return (EINVAL);
4053                 }
4054                 msg = log_only ? "ipfw: Entry %d logging count reset.\n" :
4055                     "ipfw: Entry %d cleared.\n";
4056         }
4057         IPFW_WUNLOCK(chain);
4058
4059         if (fw_verbose)
4060                 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum);
4061         return (0);
4062 }
4063
4064 /*
4065  * Check validity of the structure before insert.
4066  * Fortunately rules are simple, so this mostly need to check rule sizes.
4067  */
4068 static int
4069 check_ipfw_struct(struct ip_fw *rule, int size)
4070 {
4071         int l, cmdlen = 0;
4072         int have_action=0;
4073         ipfw_insn *cmd;
4074
4075         if (size < sizeof(*rule)) {
4076                 printf("ipfw: rule too short\n");
4077                 return (EINVAL);
4078         }
4079         /* first, check for valid size */
4080         l = RULESIZE(rule);
4081         if (l != size) {
4082                 printf("ipfw: size mismatch (have %d want %d)\n", size, l);
4083                 return (EINVAL);
4084         }
4085         if (rule->act_ofs >= rule->cmd_len) {
4086                 printf("ipfw: bogus action offset (%u > %u)\n",
4087                     rule->act_ofs, rule->cmd_len - 1);
4088                 return (EINVAL);
4089         }
4090         /*
4091          * Now go for the individual checks. Very simple ones, basically only
4092          * instruction sizes.
4093          */
4094         for (l = rule->cmd_len, cmd = rule->cmd ;
4095                         l > 0 ; l -= cmdlen, cmd += cmdlen) {
4096                 cmdlen = F_LEN(cmd);
4097                 if (cmdlen > l) {
4098                         printf("ipfw: opcode %d size truncated\n",
4099                             cmd->opcode);
4100                         return EINVAL;
4101                 }
4102                 DEB(printf("ipfw: opcode %d\n", cmd->opcode);)
4103                 switch (cmd->opcode) {
4104                 case O_PROBE_STATE:
4105                 case O_KEEP_STATE:
4106                 case O_PROTO:
4107                 case O_IP_SRC_ME:
4108                 case O_IP_DST_ME:
4109                 case O_LAYER2:
4110                 case O_IN:
4111                 case O_FRAG:
4112                 case O_DIVERTED:
4113                 case O_IPOPT:
4114                 case O_IPTOS:
4115                 case O_IPPRECEDENCE:
4116                 case O_IPVER:
4117                 case O_TCPWIN:
4118                 case O_TCPFLAGS:
4119                 case O_TCPOPTS:
4120                 case O_ESTAB:
4121                 case O_VERREVPATH:
4122                 case O_VERSRCREACH:
4123                 case O_ANTISPOOF:
4124                 case O_IPSEC:
4125 #ifdef INET6
4126                 case O_IP6_SRC_ME:
4127                 case O_IP6_DST_ME:
4128                 case O_EXT_HDR:
4129                 case O_IP6:
4130 #endif
4131                 case O_IP4:
4132                 case O_TAG:
4133                         if (cmdlen != F_INSN_SIZE(ipfw_insn))
4134                                 goto bad_size;
4135                         break;
4136
4137                 case O_UID:
4138                 case O_GID:
4139                 case O_JAIL:
4140                 case O_IP_SRC:
4141                 case O_IP_DST:
4142                 case O_TCPSEQ:
4143                 case O_TCPACK:
4144                 case O_PROB:
4145                 case O_ICMPTYPE:
4146                         if (cmdlen != F_INSN_SIZE(ipfw_insn_u32))
4147                                 goto bad_size;
4148                         break;
4149
4150                 case O_LIMIT:
4151                         if (cmdlen != F_INSN_SIZE(ipfw_insn_limit))
4152                                 goto bad_size;
4153                         break;
4154
4155                 case O_LOG:
4156                         if (cmdlen != F_INSN_SIZE(ipfw_insn_log))
4157                                 goto bad_size;
4158
4159                         ((ipfw_insn_log *)cmd)->log_left =
4160                             ((ipfw_insn_log *)cmd)->max_log;
4161
4162                         break;
4163
4164                 case O_IP_SRC_MASK:
4165                 case O_IP_DST_MASK:
4166                         /* only odd command lengths */
4167                         if ( !(cmdlen & 1) || cmdlen > 31)
4168                                 goto bad_size;
4169                         break;
4170
4171                 case O_IP_SRC_SET:
4172                 case O_IP_DST_SET:
4173                         if (cmd->arg1 == 0 || cmd->arg1 > 256) {
4174                                 printf("ipfw: invalid set size %d\n",
4175                                         cmd->arg1);
4176                                 return EINVAL;
4177                         }
4178                         if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
4179                             (cmd->arg1+31)/32 )
4180                                 goto bad_size;
4181                         break;
4182
4183                 case O_IP_SRC_LOOKUP:
4184                 case O_IP_DST_LOOKUP:
4185                         if (cmd->arg1 >= IPFW_TABLES_MAX) {
4186                                 printf("ipfw: invalid table number %d\n",
4187                                     cmd->arg1);
4188                                 return (EINVAL);
4189                         }
4190                         if (cmdlen != F_INSN_SIZE(ipfw_insn) &&
4191                             cmdlen != F_INSN_SIZE(ipfw_insn_u32))
4192                                 goto bad_size;
4193                         break;
4194
4195                 case O_MACADDR2:
4196                         if (cmdlen != F_INSN_SIZE(ipfw_insn_mac))
4197                                 goto bad_size;
4198                         break;
4199
4200                 case O_NOP:
4201                 case O_IPID:
4202                 case O_IPTTL:
4203                 case O_IPLEN:
4204                 case O_TCPDATALEN:
4205                 case O_TAGGED:
4206                         if (cmdlen < 1 || cmdlen > 31)
4207                                 goto bad_size;
4208                         break;
4209
4210                 case O_MAC_TYPE:
4211                 case O_IP_SRCPORT:
4212                 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */
4213                         if (cmdlen < 2 || cmdlen > 31)
4214                                 goto bad_size;
4215                         break;
4216
4217                 case O_RECV:
4218                 case O_XMIT:
4219                 case O_VIA:
4220                         if (cmdlen != F_INSN_SIZE(ipfw_insn_if))
4221                                 goto bad_size;
4222                         break;
4223
4224                 case O_ALTQ:
4225                         if (cmdlen != F_INSN_SIZE(ipfw_insn_altq))
4226                                 goto bad_size;
4227                         break;
4228
4229                 case O_PIPE:
4230                 case O_QUEUE:
4231                         if (cmdlen != F_INSN_SIZE(ipfw_insn))
4232                                 goto bad_size;
4233                         goto check_action;
4234
4235                 case O_FORWARD_IP:
4236 #ifdef  IPFIREWALL_FORWARD
4237                         if (cmdlen != F_INSN_SIZE(ipfw_insn_sa))
4238                                 goto bad_size;
4239                         goto check_action;
4240 #else
4241                         return EINVAL;
4242 #endif
4243
4244                 case O_DIVERT:
4245                 case O_TEE:
4246                         if (ip_divert_ptr == NULL)
4247                                 return EINVAL;
4248                         else
4249                                 goto check_size;
4250                 case O_NETGRAPH:
4251                 case O_NGTEE:
4252                         if (!NG_IPFW_LOADED)
4253                                 return EINVAL;
4254                         else
4255                                 goto check_size;
4256                 case O_NAT:
4257 #ifdef IPFIREWALL_NAT
4258                         if (cmdlen != F_INSN_SIZE(ipfw_insn_nat))
4259                                 goto bad_size;
4260                         goto check_action;
4261 #else
4262                         return EINVAL;
4263 #endif
4264                 case O_FORWARD_MAC: /* XXX not implemented yet */
4265                 case O_CHECK_STATE:
4266                 case O_COUNT:
4267                 case O_ACCEPT:
4268                 case O_DENY:
4269                 case O_REJECT:
4270 #ifdef INET6
4271                 case O_UNREACH6:
4272 #endif
4273                 case O_SKIPTO:
4274 check_size:
4275                         if (cmdlen != F_INSN_SIZE(ipfw_insn))
4276                                 goto bad_size;
4277 check_action:
4278                         if (have_action) {
4279                                 printf("ipfw: opcode %d, multiple actions"
4280                                         " not allowed\n",
4281                                         cmd->opcode);
4282                                 return EINVAL;
4283                         }
4284                         have_action = 1;
4285                         if (l != cmdlen) {
4286                                 printf("ipfw: opcode %d, action must be"
4287                                         " last opcode\n",
4288                                         cmd->opcode);
4289                                 return EINVAL;
4290                         }
4291                         break;
4292 #ifdef INET6
4293                 case O_IP6_SRC:
4294                 case O_IP6_DST:
4295                         if (cmdlen != F_INSN_SIZE(struct in6_addr) +
4296                             F_INSN_SIZE(ipfw_insn))
4297                                 goto bad_size;
4298                         break;
4299
4300                 case O_FLOW6ID:
4301                         if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
4302                             ((ipfw_insn_u32 *)cmd)->o.arg1)
4303                                 goto bad_size;
4304                         break;
4305
4306                 case O_IP6_SRC_MASK:
4307                 case O_IP6_DST_MASK:
4308                         if ( !(cmdlen & 1) || cmdlen > 127)
4309                                 goto bad_size;
4310                         break;
4311                 case O_ICMP6TYPE:
4312                         if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) )
4313                                 goto bad_size;
4314                         break;
4315 #endif
4316
4317                 default:
4318                         switch (cmd->opcode) {
4319 #ifndef INET6
4320                         case O_IP6_SRC_ME:
4321                         case O_IP6_DST_ME:
4322                         case O_EXT_HDR:
4323                         case O_IP6:
4324                         case O_UNREACH6:
4325                         case O_IP6_SRC:
4326                         case O_IP6_DST:
4327                         case O_FLOW6ID:
4328                         case O_IP6_SRC_MASK:
4329                         case O_IP6_DST_MASK:
4330                         case O_ICMP6TYPE:
4331                                 printf("ipfw: no IPv6 support in kernel\n");
4332                                 return EPROTONOSUPPORT;
4333 #endif
4334                         default:
4335                                 printf("ipfw: opcode %d, unknown opcode\n",
4336                                         cmd->opcode);
4337                                 return EINVAL;
4338                         }
4339                 }
4340         }
4341         if (have_action == 0) {
4342                 printf("ipfw: missing action\n");
4343                 return EINVAL;
4344         }
4345         return 0;
4346
4347 bad_size:
4348         printf("ipfw: opcode %d size %d wrong\n",
4349                 cmd->opcode, cmdlen);
4350         return EINVAL;
4351 }
4352
4353 /*
4354  * Copy the static and dynamic rules to the supplied buffer
4355  * and return the amount of space actually used.
4356  */
4357 static size_t
4358 ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space)
4359 {
4360         char *bp = buf;
4361         char *ep = bp + space;
4362         struct ip_fw *rule;
4363         int i;
4364         time_t  boot_seconds;
4365
4366         boot_seconds = boottime.tv_sec;
4367         /* XXX this can take a long time and locking will block packet flow */
4368         IPFW_RLOCK(chain);
4369         for (rule = chain->rules; rule ; rule = rule->next) {
4370                 /*
4371                  * Verify the entry fits in the buffer in case the
4372                  * rules changed between calculating buffer space and
4373                  * now.  This would be better done using a generation
4374                  * number but should suffice for now.
4375                  */
4376                 i = RULESIZE(rule);
4377                 if (bp + i <= ep) {
4378                         bcopy(rule, bp, i);
4379                         /*
4380                          * XXX HACK. Store the disable mask in the "next" pointer
4381                          * in a wild attempt to keep the ABI the same.
4382                          * Why do we do this on EVERY rule?
4383                          */
4384                         bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule),
4385                             sizeof(set_disable));
4386                         if (((struct ip_fw *)bp)->timestamp)
4387                                 ((struct ip_fw *)bp)->timestamp += boot_seconds;
4388                         bp += i;
4389                 }
4390         }
4391         IPFW_RUNLOCK(chain);
4392         if (ipfw_dyn_v) {
4393                 ipfw_dyn_rule *p, *last = NULL;
4394
4395                 IPFW_DYN_LOCK();
4396                 for (i = 0 ; i < curr_dyn_buckets; i++)
4397                         for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) {
4398                                 if (bp + sizeof *p <= ep) {
4399                                         ipfw_dyn_rule *dst =
4400                                                 (ipfw_dyn_rule *)bp;
4401                                         bcopy(p, dst, sizeof *p);
4402                                         bcopy(&(p->rule->rulenum), &(dst->rule),
4403                                             sizeof(p->rule->rulenum));
4404                                         /*
4405                                          * store set number into high word of
4406                                          * dst->rule pointer.
4407                                          */
4408                                         bcopy(&(p->rule->set),
4409                                             (char *)&dst->rule +
4410                                             sizeof(p->rule->rulenum),
4411                                             sizeof(p->rule->set));
4412                                         /*
4413                                          * store a non-null value in "next".
4414                                          * The userland code will interpret a
4415                                          * NULL here as a marker
4416                                          * for the last dynamic rule.
4417                                          */
4418                                         bcopy(&dst, &dst->next, sizeof(dst));
4419                                         last = dst;
4420                                         dst->expire =
4421                                             TIME_LEQ(dst->expire, time_uptime) ?
4422                                                 0 : dst->expire - time_uptime ;
4423                                         bp += sizeof(ipfw_dyn_rule);
4424                                 }
4425                         }
4426                 IPFW_DYN_UNLOCK();
4427                 if (last != NULL) /* mark last dynamic rule */
4428                         bzero(&last->next, sizeof(last));
4429         }
4430         return (bp - (char *)buf);
4431 }
4432
4433
4434 /**
4435  * {set|get}sockopt parser.
4436  */
4437 static int
4438 ipfw_ctl(struct sockopt *sopt)
4439 {
4440 #define RULE_MAXSIZE    (256*sizeof(u_int32_t))
4441         int error;
4442         size_t size;
4443         struct ip_fw *buf, *rule;
4444         u_int32_t rulenum[2];
4445
4446         error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW);
4447         if (error)
4448                 return (error);
4449
4450         /*
4451          * Disallow modifications in really-really secure mode, but still allow
4452          * the logging counters to be reset.
4453          */
4454         if (sopt->sopt_name == IP_FW_ADD ||
4455             (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) {
4456                 error = securelevel_ge(sopt->sopt_td->td_ucred, 3);
4457                 if (error)
4458                         return (error);
4459         }
4460
4461         error = 0;
4462
4463         switch (sopt->sopt_name) {
4464         case IP_FW_GET:
4465                 /*
4466                  * pass up a copy of the current rules. Static rules
4467                  * come first (the last of which has number IPFW_DEFAULT_RULE),
4468                  * followed by a possibly empty list of dynamic rule.
4469                  * The last dynamic rule has NULL in the "next" field.
4470                  *
4471                  * Note that the calculated size is used to bound the
4472                  * amount of data returned to the user.  The rule set may
4473                  * change between calculating the size and returning the
4474                  * data in which case we'll just return what fits.
4475                  */
4476                 size = static_len;      /* size of static rules */
4477                 if (ipfw_dyn_v)         /* add size of dyn.rules */
4478                         size += (dyn_count * sizeof(ipfw_dyn_rule));
4479
4480                 /*
4481                  * XXX todo: if the user passes a short length just to know
4482                  * how much room is needed, do not bother filling up the
4483                  * buffer, just jump to the sooptcopyout.
4484                  */
4485                 buf = malloc(size, M_TEMP, M_WAITOK);
4486                 error = sooptcopyout(sopt, buf,
4487                                 ipfw_getrules(&layer3_chain, buf, size));
4488                 free(buf, M_TEMP);
4489                 break;
4490
4491         case IP_FW_FLUSH:
4492                 /*
4493                  * Normally we cannot release the lock on each iteration.
4494                  * We could do it here only because we start from the head all
4495                  * the times so there is no risk of missing some entries.
4496                  * On the other hand, the risk is that we end up with
4497                  * a very inconsistent ruleset, so better keep the lock
4498                  * around the whole cycle.
4499                  *
4500                  * XXX this code can be improved by resetting the head of
4501                  * the list to point to the default rule, and then freeing
4502                  * the old list without the need for a lock.
4503                  */
4504
4505                 IPFW_WLOCK(&layer3_chain);
4506                 layer3_chain.reap = NULL;
4507                 free_chain(&layer3_chain, 0 /* keep default rule */);
4508                 rule = layer3_chain.reap;
4509                 layer3_chain.reap = NULL;
4510                 IPFW_WUNLOCK(&layer3_chain);
4511                 if (rule != NULL)
4512                         reap_rules(rule);
4513                 break;
4514
4515         case IP_FW_ADD:
4516                 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK);
4517                 error = sooptcopyin(sopt, rule, RULE_MAXSIZE,
4518                         sizeof(struct ip_fw) );
4519                 if (error == 0)
4520                         error = check_ipfw_struct(rule, sopt->sopt_valsize);
4521                 if (error == 0) {
4522                         error = add_rule(&layer3_chain, rule);
4523                         size = RULESIZE(rule);
4524                         if (!error && sopt->sopt_dir == SOPT_GET)
4525                                 error = sooptcopyout(sopt, rule, size);
4526                 }
4527                 free(rule, M_TEMP);
4528                 break;
4529
4530         case IP_FW_DEL:
4531                 /*
4532                  * IP_FW_DEL is used for deleting single rules or sets,
4533                  * and (ab)used to atomically manipulate sets. Argument size
4534                  * is used to distinguish between the two:
4535                  *    sizeof(u_int32_t)
4536                  *      delete single rule or set of rules,
4537                  *      or reassign rules (or sets) to a different set.
4538                  *    2*sizeof(u_int32_t)
4539                  *      atomic disable/enable sets.
4540                  *      first u_int32_t contains sets to be disabled,
4541                  *      second u_int32_t contains sets to be enabled.
4542                  */
4543                 error = sooptcopyin(sopt, rulenum,
4544                         2*sizeof(u_int32_t), sizeof(u_int32_t));
4545                 if (error)
4546                         break;
4547                 size = sopt->sopt_valsize;
4548                 if (size == sizeof(u_int32_t))  /* delete or reassign */
4549                         error = del_entry(&layer3_chain, rulenum[0]);
4550                 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */
4551                         set_disable =
4552                             (set_disable | rulenum[0]) & ~rulenum[1] &
4553                             ~(1<<RESVD_SET); /* set RESVD_SET always enabled */
4554                 else
4555                         error = EINVAL;
4556                 break;
4557
4558         case IP_FW_ZERO:
4559         case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */
4560                 rulenum[0] = 0;
4561                 if (sopt->sopt_val != 0) {
4562                     error = sooptcopyin(sopt, rulenum,
4563                             sizeof(u_int32_t), sizeof(u_int32_t));
4564                     if (error)
4565                         break;
4566                 }
4567                 error = zero_entry(&layer3_chain, rulenum[0],
4568                         sopt->sopt_name == IP_FW_RESETLOG);
4569                 break;
4570
4571         case IP_FW_TABLE_ADD:
4572                 {
4573                         ipfw_table_entry ent;
4574
4575                         error = sooptcopyin(sopt, &ent,
4576                             sizeof(ent), sizeof(ent));
4577                         if (error)
4578                                 break;
4579                         error = add_table_entry(&layer3_chain, ent.tbl,
4580                             ent.addr, ent.masklen, ent.value);
4581                 }
4582                 break;
4583
4584         case IP_FW_TABLE_DEL:
4585                 {
4586                         ipfw_table_entry ent;
4587
4588                         error = sooptcopyin(sopt, &ent,
4589                             sizeof(ent), sizeof(ent));
4590                         if (error)
4591                                 break;
4592                         error = del_table_entry(&layer3_chain, ent.tbl,
4593                             ent.addr, ent.masklen);
4594                 }
4595                 break;
4596
4597         case IP_FW_TABLE_FLUSH:
4598                 {
4599                         u_int16_t tbl;
4600
4601                         error = sooptcopyin(sopt, &tbl,
4602                             sizeof(tbl), sizeof(tbl));
4603                         if (error)
4604                                 break;
4605                         IPFW_WLOCK(&layer3_chain);
4606                         error = flush_table(&layer3_chain, tbl);
4607                         IPFW_WUNLOCK(&layer3_chain);
4608                 }
4609                 break;
4610
4611         case IP_FW_TABLE_GETSIZE:
4612                 {
4613                         u_int32_t tbl, cnt;
4614
4615                         if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl),
4616                             sizeof(tbl))))
4617                                 break;
4618                         IPFW_RLOCK(&layer3_chain);
4619                         error = count_table(&layer3_chain, tbl, &cnt);
4620                         IPFW_RUNLOCK(&layer3_chain);
4621                         if (error)
4622                                 break;
4623                         error = sooptcopyout(sopt, &cnt, sizeof(cnt));
4624                 }
4625                 break;
4626
4627         case IP_FW_TABLE_LIST:
4628                 {
4629                         ipfw_table *tbl;
4630
4631                         if (sopt->sopt_valsize < sizeof(*tbl)) {
4632                                 error = EINVAL;
4633                                 break;
4634                         }
4635                         size = sopt->sopt_valsize;
4636                         tbl = malloc(size, M_TEMP, M_WAITOK);
4637                         error = sooptcopyin(sopt, tbl, size, sizeof(*tbl));
4638                         if (error) {
4639                                 free(tbl, M_TEMP);
4640                                 break;
4641                         }
4642                         tbl->size = (size - sizeof(*tbl)) /
4643                             sizeof(ipfw_table_entry);
4644                         IPFW_RLOCK(&layer3_chain);
4645                         error = dump_table(&layer3_chain, tbl);
4646                         IPFW_RUNLOCK(&layer3_chain);
4647                         if (error) {
4648                                 free(tbl, M_TEMP);
4649                                 break;
4650                         }
4651                         error = sooptcopyout(sopt, tbl, size);
4652                         free(tbl, M_TEMP);
4653                 }
4654                 break;
4655
4656 #ifdef IPFIREWALL_NAT
4657         case IP_FW_NAT_CFG:
4658         {
4659                 struct cfg_nat *ptr, *ser_n;
4660                 char *buf;
4661
4662                 buf = malloc(NAT_BUF_LEN, M_IPFW, M_WAITOK | M_ZERO);
4663                 error = sooptcopyin(sopt, buf, NAT_BUF_LEN,
4664                     sizeof(struct cfg_nat));
4665                 ser_n = (struct cfg_nat *)buf;
4666
4667                 /*
4668                  * Find/create nat rule.
4669                  */
4670                 IPFW_WLOCK(&layer3_chain);
4671                 ptr = lookup_nat(ser_n->id);
4672                 if (ptr == NULL) {
4673                         /* New rule: allocate and init new instance. */
4674                         ptr = malloc(sizeof(struct cfg_nat),
4675                             M_IPFW, M_NOWAIT | M_ZERO);
4676                         if (ptr == NULL) {
4677                                 IPFW_WUNLOCK(&layer3_chain);
4678                                 free(buf, M_IPFW);
4679                                 return (ENOSPC);
4680                         }
4681                         ptr->lib = LibAliasInit(NULL);
4682                         if (ptr->lib == NULL) {
4683                                 IPFW_WUNLOCK(&layer3_chain);
4684                                 free(ptr, M_IPFW);
4685                                 free(buf, M_IPFW);
4686                                 return (EINVAL);
4687                         }
4688                         LIST_INIT(&ptr->redir_chain);
4689                 } else {
4690                         /* Entry already present: temporarly unhook it. */
4691                         UNHOOK_NAT(ptr);
4692                         flush_nat_ptrs(ser_n->id);
4693                 }
4694                 IPFW_WUNLOCK(&layer3_chain);
4695
4696                 /*
4697                  * Basic nat configuration.
4698                  */
4699                 ptr->id = ser_n->id;
4700                 /*
4701                  * XXX - what if this rule doesn't nat any ip and just
4702                  * redirect?
4703                  * do we set aliasaddress to 0.0.0.0?
4704                  */
4705                 ptr->ip = ser_n->ip;
4706                 ptr->redir_cnt = ser_n->redir_cnt;
4707                 ptr->mode = ser_n->mode;
4708                 LibAliasSetMode(ptr->lib, ser_n->mode, ser_n->mode);
4709                 LibAliasSetAddress(ptr->lib, ptr->ip);
4710                 memcpy(ptr->if_name, ser_n->if_name, IF_NAMESIZE);
4711
4712                 /*
4713                  * Redir and LSNAT configuration.
4714                  */
4715                 /* Delete old cfgs. */
4716                 del_redir_spool_cfg(ptr, &ptr->redir_chain);
4717                 /* Add new entries. */
4718                 add_redir_spool_cfg(&buf[(sizeof(struct cfg_nat))], ptr);
4719                 free(buf, M_IPFW);
4720                 IPFW_WLOCK(&layer3_chain);
4721                 HOOK_NAT(&layer3_chain.nat, ptr);
4722                 IPFW_WUNLOCK(&layer3_chain);
4723         }
4724         break;
4725
4726         case IP_FW_NAT_DEL:
4727         {
4728                 struct cfg_nat *ptr;
4729                 int i;
4730
4731                 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
4732                 IPFW_WLOCK(&layer3_chain);
4733                 ptr = lookup_nat(i);
4734                 if (ptr == NULL) {
4735                         error = EINVAL;
4736                         IPFW_WUNLOCK(&layer3_chain);
4737                         break;
4738                 }
4739                 UNHOOK_NAT(ptr);
4740                 flush_nat_ptrs(i);
4741                 IPFW_WUNLOCK(&layer3_chain);
4742                 del_redir_spool_cfg(ptr, &ptr->redir_chain);
4743                 LibAliasUninit(ptr->lib);
4744                 free(ptr, M_IPFW);
4745         }
4746         break;
4747
4748         case IP_FW_NAT_GET_CONFIG:
4749         {
4750                 uint8_t *data;
4751                 struct cfg_nat *n;
4752                 struct cfg_redir *r;
4753                 struct cfg_spool *s;
4754                 int nat_cnt, off;
4755
4756                 nat_cnt = 0;
4757                 off = sizeof(nat_cnt);
4758
4759                 data = malloc(NAT_BUF_LEN, M_IPFW, M_WAITOK | M_ZERO);
4760                 IPFW_RLOCK(&layer3_chain);
4761                 /* Serialize all the data. */
4762                 LIST_FOREACH(n, &layer3_chain.nat, _next) {
4763                         nat_cnt++;
4764                         if (off + SOF_NAT < NAT_BUF_LEN) {
4765                                 bcopy(n, &data[off], SOF_NAT);
4766                                 off += SOF_NAT;
4767                                 LIST_FOREACH(r, &n->redir_chain, _next) {
4768                                         if (off + SOF_REDIR < NAT_BUF_LEN) {
4769                                                 bcopy(r, &data[off],
4770                                                     SOF_REDIR);
4771                                                 off += SOF_REDIR;
4772                                                 LIST_FOREACH(s, &r->spool_chain,
4773                                                     _next) {
4774                                                         if (off + SOF_SPOOL <
4775                                                             NAT_BUF_LEN) {
4776                                                                 bcopy(s,
4777                                                                     &data[off],
4778                                                                     SOF_SPOOL);
4779                                                                 off +=
4780                                                                     SOF_SPOOL;
4781                                                         } else
4782                                                                 goto nospace;
4783                                                 }
4784                                         } else
4785                                                 goto nospace;
4786                                 }
4787                         } else
4788                                 goto nospace;
4789                 }
4790                 bcopy(&nat_cnt, data, sizeof(nat_cnt));
4791                 IPFW_RUNLOCK(&layer3_chain);
4792                 error = sooptcopyout(sopt, data, NAT_BUF_LEN);
4793                 free(data, M_IPFW);
4794                 break;
4795         nospace:
4796                 IPFW_RUNLOCK(&layer3_chain);
4797                 printf("serialized data buffer not big enough:"
4798                     "please increase NAT_BUF_LEN\n");
4799                 free(data, M_IPFW);
4800         }
4801         break;
4802
4803         case IP_FW_NAT_GET_LOG:
4804         {
4805                 uint8_t *data;
4806                 struct cfg_nat *ptr;
4807                 int i, size, cnt, sof;
4808
4809                 data = NULL;
4810                 sof = LIBALIAS_BUF_SIZE;
4811                 cnt = 0;
4812
4813                 IPFW_RLOCK(&layer3_chain);
4814                 size = i = 0;
4815                 LIST_FOREACH(ptr, &layer3_chain.nat, _next) {
4816                         if (ptr->lib->logDesc == NULL)
4817                                 continue;
4818                         cnt++;
4819                         size = cnt * (sof + sizeof(int));
4820                         data = realloc(data, size, M_IPFW, M_NOWAIT | M_ZERO);
4821                         if (data == NULL) {
4822                                 IPFW_RUNLOCK(&layer3_chain);
4823                                 return (ENOSPC);
4824                         }
4825                         bcopy(&ptr->id, &data[i], sizeof(int));
4826                         i += sizeof(int);
4827                         bcopy(ptr->lib->logDesc, &data[i], sof);
4828                         i += sof;
4829                 }
4830                 IPFW_RUNLOCK(&layer3_chain);
4831                 error = sooptcopyout(sopt, data, size);
4832                 free(data, M_IPFW);
4833         }
4834         break;
4835 #endif
4836
4837         default:
4838                 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name);
4839                 error = EINVAL;
4840         }
4841
4842         return (error);
4843 #undef RULE_MAXSIZE
4844 }
4845
4846 /**
4847  * dummynet needs a reference to the default rule, because rules can be
4848  * deleted while packets hold a reference to them. When this happens,
4849  * dummynet changes the reference to the default rule (it could well be a
4850  * NULL pointer, but this way we do not need to check for the special
4851  * case, plus here he have info on the default behaviour).
4852  */
4853 struct ip_fw *ip_fw_default_rule;
4854
4855 /*
4856  * This procedure is only used to handle keepalives. It is invoked
4857  * every dyn_keepalive_period
4858  */
4859 static void
4860 ipfw_tick(void * __unused unused)
4861 {
4862         struct mbuf *m0, *m, *mnext, **mtailp;
4863         int i;
4864         ipfw_dyn_rule *q;
4865
4866         if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0)
4867                 goto done;
4868
4869         /*
4870          * We make a chain of packets to go out here -- not deferring
4871          * until after we drop the IPFW dynamic rule lock would result
4872          * in a lock order reversal with the normal packet input -> ipfw
4873          * call stack.
4874          */
4875         m0 = NULL;
4876         mtailp = &m0;
4877         IPFW_DYN_LOCK();
4878         for (i = 0 ; i < curr_dyn_buckets ; i++) {
4879                 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) {
4880                         if (q->dyn_type == O_LIMIT_PARENT)
4881                                 continue;
4882                         if (q->id.proto != IPPROTO_TCP)
4883                                 continue;
4884                         if ( (q->state & BOTH_SYN) != BOTH_SYN)
4885                                 continue;
4886                         if (TIME_LEQ( time_uptime+dyn_keepalive_interval,
4887                             q->expire))
4888                                 continue;       /* too early */
4889                         if (TIME_LEQ(q->expire, time_uptime))
4890                                 continue;       /* too late, rule expired */
4891
4892                         *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1,
4893                                 q->ack_fwd, TH_SYN);
4894                         if (*mtailp != NULL)
4895                                 mtailp = &(*mtailp)->m_nextpkt;
4896                         *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1,
4897                                 q->ack_rev, 0);
4898                         if (*mtailp != NULL)
4899                                 mtailp = &(*mtailp)->m_nextpkt;
4900                 }
4901         }
4902         IPFW_DYN_UNLOCK();
4903         for (m = mnext = m0; m != NULL; m = mnext) {
4904                 mnext = m->m_nextpkt;
4905                 m->m_nextpkt = NULL;
4906                 ip_output(m, NULL, NULL, 0, NULL, NULL);
4907         }
4908 done:
4909         callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL);
4910 }
4911
4912 int
4913 ipfw_init(void)
4914 {
4915         struct ip_fw default_rule;
4916         int error;
4917
4918 #ifdef INET6
4919         /* Setup IPv6 fw sysctl tree. */
4920         sysctl_ctx_init(&ip6_fw_sysctl_ctx);
4921         ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx,
4922             SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw",
4923             CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall");
4924         SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree),
4925             OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3,
4926             &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6");
4927         SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree),
4928             OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE,
4929             &fw_deny_unknown_exthdrs, 0,
4930             "Deny packets with unknown IPv6 Extension Headers");
4931 #endif
4932
4933         layer3_chain.rules = NULL;
4934         IPFW_LOCK_INIT(&layer3_chain);
4935         ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule",
4936             sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL,
4937             UMA_ALIGN_PTR, 0);
4938         IPFW_DYN_LOCK_INIT();
4939         callout_init(&ipfw_timeout, CALLOUT_MPSAFE);
4940
4941         bzero(&default_rule, sizeof default_rule);
4942
4943         default_rule.act_ofs = 0;
4944         default_rule.rulenum = IPFW_DEFAULT_RULE;
4945         default_rule.cmd_len = 1;
4946         default_rule.set = RESVD_SET;
4947
4948         default_rule.cmd[0].len = 1;
4949         default_rule.cmd[0].opcode =
4950 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
4951                                 1 ? O_ACCEPT :
4952 #endif
4953                                 O_DENY;
4954
4955         error = add_rule(&layer3_chain, &default_rule);
4956         if (error != 0) {
4957                 printf("ipfw2: error %u initializing default rule "
4958                         "(support disabled)\n", error);
4959                 IPFW_DYN_LOCK_DESTROY();
4960                 IPFW_LOCK_DESTROY(&layer3_chain);
4961                 uma_zdestroy(ipfw_dyn_rule_zone);
4962                 return (error);
4963         }
4964
4965         ip_fw_default_rule = layer3_chain.rules;
4966         printf("ipfw2 "
4967 #ifdef INET6
4968                 "(+ipv6) "
4969 #endif
4970                 "initialized, divert %s, "
4971                 "rule-based forwarding "
4972 #ifdef IPFIREWALL_FORWARD
4973                 "enabled, "
4974 #else
4975                 "disabled, "
4976 #endif
4977                 "default to %s, logging ",
4978 #ifdef IPDIVERT
4979                 "enabled",
4980 #else
4981                 "loadable",
4982 #endif
4983                 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny");
4984
4985 #ifdef IPFIREWALL_VERBOSE
4986         fw_verbose = 1;
4987 #endif
4988 #ifdef IPFIREWALL_VERBOSE_LIMIT
4989         verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
4990 #endif
4991         if (fw_verbose == 0)
4992                 printf("disabled\n");
4993         else if (verbose_limit == 0)
4994                 printf("unlimited\n");
4995         else
4996                 printf("limited to %d packets/entry by default\n",
4997                     verbose_limit);
4998
4999         error = init_tables(&layer3_chain);
5000         if (error) {
5001                 IPFW_DYN_LOCK_DESTROY();
5002                 IPFW_LOCK_DESTROY(&layer3_chain);
5003                 uma_zdestroy(ipfw_dyn_rule_zone);
5004                 return (error);
5005         }
5006         ip_fw_ctl_ptr = ipfw_ctl;
5007         ip_fw_chk_ptr = ipfw_chk;
5008         callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL);
5009 #ifdef IPFIREWALL_NAT
5010         LIST_INIT(&layer3_chain.nat);
5011         ifaddr_event_tag = EVENTHANDLER_REGISTER(ifaddr_event, ifaddr_change,
5012             NULL, EVENTHANDLER_PRI_ANY);
5013 #endif
5014         return (0);
5015 }
5016
5017 void
5018 ipfw_destroy(void)
5019 {
5020         struct ip_fw *reap;
5021 #ifdef IPFIREWALL_NAT
5022         struct cfg_nat *ptr, *ptr_temp;
5023 #endif
5024
5025         ip_fw_chk_ptr = NULL;
5026         ip_fw_ctl_ptr = NULL;
5027         callout_drain(&ipfw_timeout);
5028         IPFW_WLOCK(&layer3_chain);
5029         flush_tables(&layer3_chain);
5030 #ifdef IPFIREWALL_NAT
5031         LIST_FOREACH_SAFE(ptr, &layer3_chain.nat, _next, ptr_temp) {
5032                 LIST_REMOVE(ptr, _next);
5033                 del_redir_spool_cfg(ptr, &ptr->redir_chain);
5034                 LibAliasUninit(ptr->lib);
5035                 free(ptr, M_IPFW);
5036         }
5037         EVENTHANDLER_DEREGISTER(ifaddr_event, ifaddr_event_tag);
5038 #endif
5039         layer3_chain.reap = NULL;
5040         free_chain(&layer3_chain, 1 /* kill default rule */);
5041         reap = layer3_chain.reap, layer3_chain.reap = NULL;
5042         IPFW_WUNLOCK(&layer3_chain);
5043         if (reap != NULL)
5044                 reap_rules(reap);
5045         IPFW_DYN_LOCK_DESTROY();
5046         uma_zdestroy(ipfw_dyn_rule_zone);
5047         if (ipfw_dyn_v != NULL)
5048                 free(ipfw_dyn_v, M_IPFW);
5049         IPFW_LOCK_DESTROY(&layer3_chain);
5050
5051 #ifdef INET6
5052         /* Free IPv6 fw sysctl tree. */
5053         sysctl_ctx_free(&ip6_fw_sysctl_ctx);
5054 #endif
5055
5056         printf("IP firewall unloaded\n");
5057 }