The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/netpfil/ipfw/ip_fw_sockopt.c

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
    3  *
    4  * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa
    5  * Copyright (c) 2014 Yandex LLC
    6  * Copyright (c) 2014 Alexander V. Chernikov
    7  *
    8  * Supported by: Valeria Paoli
    9  *
   10  * Redistribution and use in source and binary forms, with or without
   11  * modification, are permitted provided that the following conditions
   12  * are met:
   13  * 1. Redistributions of source code must retain the above copyright
   14  *    notice, this list of conditions and the following disclaimer.
   15  * 2. Redistributions in binary form must reproduce the above copyright
   16  *    notice, this list of conditions and the following disclaimer in the
   17  *    documentation and/or other materials provided with the distribution.
   18  *
   19  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
   20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
   23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   29  * SUCH DAMAGE.
   30  */
   31 
   32 #include <sys/cdefs.h>
   33 __FBSDID("$FreeBSD$");
   34 
   35 /*
   36  * Control socket and rule management routines for ipfw.
   37  * Control is currently implemented via IP_FW3 setsockopt() code.
   38  */
   39 
   40 #include "opt_ipfw.h"
   41 #include "opt_inet.h"
   42 #ifndef INET
   43 #error IPFIREWALL requires INET.
   44 #endif /* INET */
   45 #include "opt_inet6.h"
   46 
   47 #include <sys/param.h>
   48 #include <sys/systm.h>
   49 #include <sys/malloc.h>
   50 #include <sys/mbuf.h>   /* struct m_tag used by nested headers */
   51 #include <sys/kernel.h>
   52 #include <sys/lock.h>
   53 #include <sys/priv.h>
   54 #include <sys/proc.h>
   55 #include <sys/rwlock.h>
   56 #include <sys/rmlock.h>
   57 #include <sys/socket.h>
   58 #include <sys/socketvar.h>
   59 #include <sys/sysctl.h>
   60 #include <sys/syslog.h>
   61 #include <sys/fnv_hash.h>
   62 #include <net/if.h>
   63 #include <net/route.h>
   64 #include <net/vnet.h>
   65 #include <vm/vm.h>
   66 #include <vm/vm_extern.h>
   67 
   68 #include <netinet/in.h>
   69 #include <netinet/ip_var.h> /* hooks */
   70 #include <netinet/ip_fw.h>
   71 
   72 #include <netpfil/ipfw/ip_fw_private.h>
   73 #include <netpfil/ipfw/ip_fw_table.h>
   74 
   75 #ifdef MAC
   76 #include <security/mac/mac_framework.h>
   77 #endif
   78 
   79 static int ipfw_ctl(struct sockopt *sopt);
   80 static int check_ipfw_rule_body(ipfw_insn *cmd, int cmd_len,
   81     struct rule_check_info *ci);
   82 static int check_ipfw_rule1(struct ip_fw_rule *rule, int size,
   83     struct rule_check_info *ci);
   84 static int check_ipfw_rule0(struct ip_fw_rule0 *rule, int size,
   85     struct rule_check_info *ci);
   86 static int rewrite_rule_uidx(struct ip_fw_chain *chain,
   87     struct rule_check_info *ci);
   88 
   89 #define NAMEDOBJ_HASH_SIZE      32
   90 
   91 struct namedobj_instance {
   92         struct namedobjects_head        *names;
   93         struct namedobjects_head        *values;
   94         uint32_t nn_size;               /* names hash size */
   95         uint32_t nv_size;               /* number hash size */
   96         u_long *idx_mask;               /* used items bitmask */
   97         uint32_t max_blocks;            /* number of "long" blocks in bitmask */
   98         uint32_t count;                 /* number of items */
   99         uint16_t free_off[IPFW_MAX_SETS];       /* first possible free offset */
  100         objhash_hash_f  *hash_f;
  101         objhash_cmp_f   *cmp_f;
  102 };
  103 #define BLOCK_ITEMS     (8 * sizeof(u_long))    /* Number of items for ffsl() */
  104 
  105 static uint32_t objhash_hash_name(struct namedobj_instance *ni,
  106     const void *key, uint32_t kopt);
  107 static uint32_t objhash_hash_idx(struct namedobj_instance *ni, uint32_t val);
  108 static int objhash_cmp_name(struct named_object *no, const void *name,
  109     uint32_t set);
  110 
  111 MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's");
  112 
  113 static int dump_config(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
  114     struct sockopt_data *sd);
  115 static int add_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
  116     struct sockopt_data *sd);
  117 static int del_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
  118     struct sockopt_data *sd);
  119 static int clear_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
  120     struct sockopt_data *sd);
  121 static int move_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
  122     struct sockopt_data *sd);
  123 static int manage_sets(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
  124     struct sockopt_data *sd);
  125 static int dump_soptcodes(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
  126     struct sockopt_data *sd);
  127 static int dump_srvobjects(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
  128     struct sockopt_data *sd);
  129 
  130 /* ctl3 handler data */
  131 struct mtx ctl3_lock;
  132 #define CTL3_LOCK_INIT()        mtx_init(&ctl3_lock, "ctl3_lock", NULL, MTX_DEF)
  133 #define CTL3_LOCK_DESTROY()     mtx_destroy(&ctl3_lock)
  134 #define CTL3_LOCK()             mtx_lock(&ctl3_lock)
  135 #define CTL3_UNLOCK()           mtx_unlock(&ctl3_lock)
  136 
  137 static struct ipfw_sopt_handler *ctl3_handlers;
  138 static size_t ctl3_hsize;
  139 static uint64_t ctl3_refct, ctl3_gencnt;
  140 #define CTL3_SMALLBUF   4096                    /* small page-size write buffer */
  141 #define CTL3_LARGEBUF   16 * 1024 * 1024        /* handle large rulesets */
  142 
  143 static int ipfw_flush_sopt_data(struct sockopt_data *sd);
  144 
  145 static struct ipfw_sopt_handler scodes[] = {
  146         { IP_FW_XGET,           0,      HDIR_GET,       dump_config },
  147         { IP_FW_XADD,           0,      HDIR_BOTH,      add_rules },
  148         { IP_FW_XDEL,           0,      HDIR_BOTH,      del_rules },
  149         { IP_FW_XZERO,          0,      HDIR_SET,       clear_rules },
  150         { IP_FW_XRESETLOG,      0,      HDIR_SET,       clear_rules },
  151         { IP_FW_XMOVE,          0,      HDIR_SET,       move_rules },
  152         { IP_FW_SET_SWAP,       0,      HDIR_SET,       manage_sets },
  153         { IP_FW_SET_MOVE,       0,      HDIR_SET,       manage_sets },
  154         { IP_FW_SET_ENABLE,     0,      HDIR_SET,       manage_sets },
  155         { IP_FW_DUMP_SOPTCODES, 0,      HDIR_GET,       dump_soptcodes },
  156         { IP_FW_DUMP_SRVOBJECTS,0,      HDIR_GET,       dump_srvobjects },
  157 };
  158 
  159 static int
  160 set_legacy_obj_kidx(struct ip_fw_chain *ch, struct ip_fw_rule0 *rule);
  161 static struct opcode_obj_rewrite *find_op_rw(ipfw_insn *cmd,
  162     uint16_t *puidx, uint8_t *ptype);
  163 static int ref_rule_objects(struct ip_fw_chain *ch, struct ip_fw *rule,
  164     struct rule_check_info *ci, struct obj_idx *oib, struct tid_info *ti);
  165 static int ref_opcode_object(struct ip_fw_chain *ch, ipfw_insn *cmd,
  166     struct tid_info *ti, struct obj_idx *pidx, int *unresolved);
  167 static void unref_rule_objects(struct ip_fw_chain *chain, struct ip_fw *rule);
  168 static void unref_oib_objects(struct ip_fw_chain *ch, ipfw_insn *cmd,
  169     struct obj_idx *oib, struct obj_idx *end);
  170 static int export_objhash_ntlv(struct namedobj_instance *ni, uint16_t kidx,
  171     struct sockopt_data *sd);
  172 
  173 /*
  174  * Opcode object rewriter variables
  175  */
  176 struct opcode_obj_rewrite *ctl3_rewriters;
  177 static size_t ctl3_rsize;
  178 
  179 /*
  180  * static variables followed by global ones
  181  */
  182 
  183 VNET_DEFINE_STATIC(uma_zone_t, ipfw_cntr_zone);
  184 #define V_ipfw_cntr_zone                VNET(ipfw_cntr_zone)
  185 
  186 void
  187 ipfw_init_counters(void)
  188 {
  189 
  190         V_ipfw_cntr_zone = uma_zcreate("IPFW counters",
  191             IPFW_RULE_CNTR_SIZE, NULL, NULL, NULL, NULL,
  192             UMA_ALIGN_PTR, UMA_ZONE_PCPU);
  193 }
  194 
  195 void
  196 ipfw_destroy_counters(void)
  197 {
  198 
  199         uma_zdestroy(V_ipfw_cntr_zone);
  200 }
  201 
  202 struct ip_fw *
  203 ipfw_alloc_rule(struct ip_fw_chain *chain, size_t rulesize)
  204 {
  205         struct ip_fw *rule;
  206 
  207         rule = malloc(rulesize, M_IPFW, M_WAITOK | M_ZERO);
  208         rule->cntr = uma_zalloc_pcpu(V_ipfw_cntr_zone, M_WAITOK | M_ZERO);
  209         rule->refcnt = 1;
  210 
  211         return (rule);
  212 }
  213 
  214 void
  215 ipfw_free_rule(struct ip_fw *rule)
  216 {
  217 
  218         /*
  219          * We don't release refcnt here, since this function
  220          * can be called without any locks held. The caller
  221          * must release reference under IPFW_UH_WLOCK, and then
  222          * call this function if refcount becomes 1.
  223          */
  224         if (rule->refcnt > 1)
  225                 return;
  226         uma_zfree_pcpu(V_ipfw_cntr_zone, rule->cntr);
  227         free(rule, M_IPFW);
  228 }
  229 
  230 /*
  231  * Find the smallest rule >= key, id.
  232  * We could use bsearch but it is so simple that we code it directly
  233  */
  234 int
  235 ipfw_find_rule(struct ip_fw_chain *chain, uint32_t key, uint32_t id)
  236 {
  237         int i, lo, hi;
  238         struct ip_fw *r;
  239 
  240         for (lo = 0, hi = chain->n_rules - 1; lo < hi;) {
  241                 i = (lo + hi) / 2;
  242                 r = chain->map[i];
  243                 if (r->rulenum < key)
  244                         lo = i + 1;     /* continue from the next one */
  245                 else if (r->rulenum > key)
  246                         hi = i;         /* this might be good */
  247                 else if (r->id < id)
  248                         lo = i + 1;     /* continue from the next one */
  249                 else /* r->id >= id */
  250                         hi = i;         /* this might be good */
  251         }
  252         return hi;
  253 }
  254 
  255 /*
  256  * Builds skipto cache on rule set @map.
  257  */
  258 static void
  259 update_skipto_cache(struct ip_fw_chain *chain, struct ip_fw **map)
  260 {
  261         int *smap, rulenum;
  262         int i, mi;
  263 
  264         IPFW_UH_WLOCK_ASSERT(chain);
  265 
  266         mi = 0;
  267         rulenum = map[mi]->rulenum;
  268         smap = chain->idxmap_back;
  269 
  270         if (smap == NULL)
  271                 return;
  272 
  273         for (i = 0; i < 65536; i++) {
  274                 smap[i] = mi;
  275                 /* Use the same rule index until i < rulenum */
  276                 if (i != rulenum || i == 65535)
  277                         continue;
  278                 /* Find next rule with num > i */
  279                 rulenum = map[++mi]->rulenum;
  280                 while (rulenum == i)
  281                         rulenum = map[++mi]->rulenum;
  282         }
  283 }
  284 
  285 /*
  286  * Swaps prepared (backup) index with current one.
  287  */
  288 static void
  289 swap_skipto_cache(struct ip_fw_chain *chain)
  290 {
  291         int *map;
  292 
  293         IPFW_UH_WLOCK_ASSERT(chain);
  294         IPFW_WLOCK_ASSERT(chain);
  295 
  296         map = chain->idxmap;
  297         chain->idxmap = chain->idxmap_back;
  298         chain->idxmap_back = map;
  299 }
  300 
  301 /*
  302  * Allocate and initialize skipto cache.
  303  */
  304 void
  305 ipfw_init_skipto_cache(struct ip_fw_chain *chain)
  306 {
  307         int *idxmap, *idxmap_back;
  308 
  309         idxmap = malloc(65536 * sizeof(int), M_IPFW, M_WAITOK | M_ZERO);
  310         idxmap_back = malloc(65536 * sizeof(int), M_IPFW, M_WAITOK);
  311 
  312         /*
  313          * Note we may be called at any time after initialization,
  314          * for example, on first skipto rule, so we need to
  315          * provide valid chain->idxmap on return
  316          */
  317 
  318         IPFW_UH_WLOCK(chain);
  319         if (chain->idxmap != NULL) {
  320                 IPFW_UH_WUNLOCK(chain);
  321                 free(idxmap, M_IPFW);
  322                 free(idxmap_back, M_IPFW);
  323                 return;
  324         }
  325 
  326         /* Set backup pointer first to permit building cache */
  327         chain->idxmap_back = idxmap_back;
  328         update_skipto_cache(chain, chain->map);
  329         IPFW_WLOCK(chain);
  330         /* It is now safe to set chain->idxmap ptr */
  331         chain->idxmap = idxmap;
  332         swap_skipto_cache(chain);
  333         IPFW_WUNLOCK(chain);
  334         IPFW_UH_WUNLOCK(chain);
  335 }
  336 
  337 /*
  338  * Destroys skipto cache.
  339  */
  340 void
  341 ipfw_destroy_skipto_cache(struct ip_fw_chain *chain)
  342 {
  343 
  344         if (chain->idxmap != NULL)
  345                 free(chain->idxmap, M_IPFW);
  346         if (chain->idxmap != NULL)
  347                 free(chain->idxmap_back, M_IPFW);
  348 }
  349 
  350 /*
  351  * allocate a new map, returns the chain locked. extra is the number
  352  * of entries to add or delete.
  353  */
  354 static struct ip_fw **
  355 get_map(struct ip_fw_chain *chain, int extra, int locked)
  356 {
  357 
  358         for (;;) {
  359                 struct ip_fw **map;
  360                 u_int i, mflags;
  361 
  362                 mflags = M_ZERO | ((locked != 0) ? M_NOWAIT : M_WAITOK);
  363 
  364                 i = chain->n_rules + extra;
  365                 map = malloc(i * sizeof(struct ip_fw *), M_IPFW, mflags);
  366                 if (map == NULL) {
  367                         printf("%s: cannot allocate map\n", __FUNCTION__);
  368                         return NULL;
  369                 }
  370                 if (!locked)
  371                         IPFW_UH_WLOCK(chain);
  372                 if (i >= chain->n_rules + extra) /* good */
  373                         return map;
  374                 /* otherwise we lost the race, free and retry */
  375                 if (!locked)
  376                         IPFW_UH_WUNLOCK(chain);
  377                 free(map, M_IPFW);
  378         }
  379 }
  380 
  381 /*
  382  * swap the maps. It is supposed to be called with IPFW_UH_WLOCK
  383  */
  384 static struct ip_fw **
  385 swap_map(struct ip_fw_chain *chain, struct ip_fw **new_map, int new_len)
  386 {
  387         struct ip_fw **old_map;
  388 
  389         IPFW_WLOCK(chain);
  390         chain->id++;
  391         chain->n_rules = new_len;
  392         old_map = chain->map;
  393         chain->map = new_map;
  394         swap_skipto_cache(chain);
  395         IPFW_WUNLOCK(chain);
  396         return old_map;
  397 }
  398 
  399 static void
  400 export_cntr1_base(struct ip_fw *krule, struct ip_fw_bcounter *cntr)
  401 {
  402         struct timeval boottime;
  403 
  404         cntr->size = sizeof(*cntr);
  405 
  406         if (krule->cntr != NULL) {
  407                 cntr->pcnt = counter_u64_fetch(krule->cntr);
  408                 cntr->bcnt = counter_u64_fetch(krule->cntr + 1);
  409                 cntr->timestamp = krule->timestamp;
  410         }
  411         if (cntr->timestamp > 0) {
  412                 getboottime(&boottime);
  413                 cntr->timestamp += boottime.tv_sec;
  414         }
  415 }
  416 
  417 static void
  418 export_cntr0_base(struct ip_fw *krule, struct ip_fw_bcounter0 *cntr)
  419 {
  420         struct timeval boottime;
  421 
  422         if (krule->cntr != NULL) {
  423                 cntr->pcnt = counter_u64_fetch(krule->cntr);
  424                 cntr->bcnt = counter_u64_fetch(krule->cntr + 1);
  425                 cntr->timestamp = krule->timestamp;
  426         }
  427         if (cntr->timestamp > 0) {
  428                 getboottime(&boottime);
  429                 cntr->timestamp += boottime.tv_sec;
  430         }
  431 }
  432 
  433 /*
  434  * Copies rule @urule from v1 userland format (current).
  435  * to kernel @krule.
  436  * Assume @krule is zeroed.
  437  */
  438 static void
  439 import_rule1(struct rule_check_info *ci)
  440 {
  441         struct ip_fw_rule *urule;
  442         struct ip_fw *krule;
  443 
  444         urule = (struct ip_fw_rule *)ci->urule;
  445         krule = (struct ip_fw *)ci->krule;
  446 
  447         /* copy header */
  448         krule->act_ofs = urule->act_ofs;
  449         krule->cmd_len = urule->cmd_len;
  450         krule->rulenum = urule->rulenum;
  451         krule->set = urule->set;
  452         krule->flags = urule->flags;
  453 
  454         /* Save rulenum offset */
  455         ci->urule_numoff = offsetof(struct ip_fw_rule, rulenum);
  456 
  457         /* Copy opcodes */
  458         memcpy(krule->cmd, urule->cmd, krule->cmd_len * sizeof(uint32_t));
  459 }
  460 
  461 /*
  462  * Export rule into v1 format (Current).
  463  * Layout:
  464  * [ ipfw_obj_tlv(IPFW_TLV_RULE_ENT)
  465  *     [ ip_fw_rule ] OR
  466  *     [ ip_fw_bcounter ip_fw_rule] (depends on rcntrs).
  467  * ]
  468  * Assume @data is zeroed.
  469  */
  470 static void
  471 export_rule1(struct ip_fw *krule, caddr_t data, int len, int rcntrs)
  472 {
  473         struct ip_fw_bcounter *cntr;
  474         struct ip_fw_rule *urule;
  475         ipfw_obj_tlv *tlv;
  476 
  477         /* Fill in TLV header */
  478         tlv = (ipfw_obj_tlv *)data;
  479         tlv->type = IPFW_TLV_RULE_ENT;
  480         tlv->length = len;
  481 
  482         if (rcntrs != 0) {
  483                 /* Copy counters */
  484                 cntr = (struct ip_fw_bcounter *)(tlv + 1);
  485                 urule = (struct ip_fw_rule *)(cntr + 1);
  486                 export_cntr1_base(krule, cntr);
  487         } else
  488                 urule = (struct ip_fw_rule *)(tlv + 1);
  489 
  490         /* copy header */
  491         urule->act_ofs = krule->act_ofs;
  492         urule->cmd_len = krule->cmd_len;
  493         urule->rulenum = krule->rulenum;
  494         urule->set = krule->set;
  495         urule->flags = krule->flags;
  496         urule->id = krule->id;
  497 
  498         /* Copy opcodes */
  499         memcpy(urule->cmd, krule->cmd, krule->cmd_len * sizeof(uint32_t));
  500 }
  501 
  502 /*
  503  * Copies rule @urule from FreeBSD8 userland format (v0)
  504  * to kernel @krule.
  505  * Assume @krule is zeroed.
  506  */
  507 static void
  508 import_rule0(struct rule_check_info *ci)
  509 {
  510         struct ip_fw_rule0 *urule;
  511         struct ip_fw *krule;
  512         int cmdlen, l;
  513         ipfw_insn *cmd;
  514         ipfw_insn_limit *lcmd;
  515         ipfw_insn_if *cmdif;
  516 
  517         urule = (struct ip_fw_rule0 *)ci->urule;
  518         krule = (struct ip_fw *)ci->krule;
  519 
  520         /* copy header */
  521         krule->act_ofs = urule->act_ofs;
  522         krule->cmd_len = urule->cmd_len;
  523         krule->rulenum = urule->rulenum;
  524         krule->set = urule->set;
  525         if ((urule->_pad & 1) != 0)
  526                 krule->flags |= IPFW_RULE_NOOPT;
  527 
  528         /* Save rulenum offset */
  529         ci->urule_numoff = offsetof(struct ip_fw_rule0, rulenum);
  530 
  531         /* Copy opcodes */
  532         memcpy(krule->cmd, urule->cmd, krule->cmd_len * sizeof(uint32_t));
  533 
  534         /*
  535          * Alter opcodes:
  536          * 1) convert tablearg value from 65535 to 0
  537          * 2) Add high bit to O_SETFIB/O_SETDSCP values (to make room
  538          *    for targ).
  539          * 3) convert table number in iface opcodes to u16
  540          * 4) convert old `nat global` into new 65535
  541          */
  542         l = krule->cmd_len;
  543         cmd = krule->cmd;
  544         cmdlen = 0;
  545 
  546         for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) {
  547                 cmdlen = F_LEN(cmd);
  548 
  549                 switch (cmd->opcode) {
  550                 /* Opcodes supporting tablearg */
  551                 case O_TAG:
  552                 case O_TAGGED:
  553                 case O_PIPE:
  554                 case O_QUEUE:
  555                 case O_DIVERT:
  556                 case O_TEE:
  557                 case O_SKIPTO:
  558                 case O_CALLRETURN:
  559                 case O_NETGRAPH:
  560                 case O_NGTEE:
  561                 case O_NAT:
  562                         if (cmd->arg1 == IP_FW_TABLEARG)
  563                                 cmd->arg1 = IP_FW_TARG;
  564                         else if (cmd->arg1 == 0)
  565                                 cmd->arg1 = IP_FW_NAT44_GLOBAL;
  566                         break;
  567                 case O_SETFIB:
  568                 case O_SETDSCP:
  569                         if (cmd->arg1 == IP_FW_TABLEARG)
  570                                 cmd->arg1 = IP_FW_TARG;
  571                         else
  572                                 cmd->arg1 |= 0x8000;
  573                         break;
  574                 case O_LIMIT:
  575                         lcmd = (ipfw_insn_limit *)cmd;
  576                         if (lcmd->conn_limit == IP_FW_TABLEARG)
  577                                 lcmd->conn_limit = IP_FW_TARG;
  578                         break;
  579                 /* Interface tables */
  580                 case O_XMIT:
  581                 case O_RECV:
  582                 case O_VIA:
  583                         /* Interface table, possibly */
  584                         cmdif = (ipfw_insn_if *)cmd;
  585                         if (cmdif->name[0] != '\1')
  586                                 break;
  587 
  588                         cmdif->p.kidx = (uint16_t)cmdif->p.glob;
  589                         break;
  590                 }
  591         }
  592 }
  593 
  594 /*
  595  * Copies rule @krule from kernel to FreeBSD8 userland format (v0)
  596  */
  597 static void
  598 export_rule0(struct ip_fw *krule, struct ip_fw_rule0 *urule, int len)
  599 {
  600         int cmdlen, l;
  601         ipfw_insn *cmd;
  602         ipfw_insn_limit *lcmd;
  603         ipfw_insn_if *cmdif;
  604 
  605         /* copy header */
  606         memset(urule, 0, len);
  607         urule->act_ofs = krule->act_ofs;
  608         urule->cmd_len = krule->cmd_len;
  609         urule->rulenum = krule->rulenum;
  610         urule->set = krule->set;
  611         if ((krule->flags & IPFW_RULE_NOOPT) != 0)
  612                 urule->_pad |= 1;
  613 
  614         /* Copy opcodes */
  615         memcpy(urule->cmd, krule->cmd, krule->cmd_len * sizeof(uint32_t));
  616 
  617         /* Export counters */
  618         export_cntr0_base(krule, (struct ip_fw_bcounter0 *)&urule->pcnt);
  619 
  620         /*
  621          * Alter opcodes:
  622          * 1) convert tablearg value from 0 to 65535
  623          * 2) Remove highest bit from O_SETFIB/O_SETDSCP values.
  624          * 3) convert table number in iface opcodes to int
  625          */
  626         l = urule->cmd_len;
  627         cmd = urule->cmd;
  628         cmdlen = 0;
  629 
  630         for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) {
  631                 cmdlen = F_LEN(cmd);
  632 
  633                 switch (cmd->opcode) {
  634                 /* Opcodes supporting tablearg */
  635                 case O_TAG:
  636                 case O_TAGGED:
  637                 case O_PIPE:
  638                 case O_QUEUE:
  639                 case O_DIVERT:
  640                 case O_TEE:
  641                 case O_SKIPTO:
  642                 case O_CALLRETURN:
  643                 case O_NETGRAPH:
  644                 case O_NGTEE:
  645                 case O_NAT:
  646                         if (cmd->arg1 == IP_FW_TARG)
  647                                 cmd->arg1 = IP_FW_TABLEARG;
  648                         else if (cmd->arg1 == IP_FW_NAT44_GLOBAL)
  649                                 cmd->arg1 = 0;
  650                         break;
  651                 case O_SETFIB:
  652                 case O_SETDSCP:
  653                         if (cmd->arg1 == IP_FW_TARG)
  654                                 cmd->arg1 = IP_FW_TABLEARG;
  655                         else
  656                                 cmd->arg1 &= ~0x8000;
  657                         break;
  658                 case O_LIMIT:
  659                         lcmd = (ipfw_insn_limit *)cmd;
  660                         if (lcmd->conn_limit == IP_FW_TARG)
  661                                 lcmd->conn_limit = IP_FW_TABLEARG;
  662                         break;
  663                 /* Interface tables */
  664                 case O_XMIT:
  665                 case O_RECV:
  666                 case O_VIA:
  667                         /* Interface table, possibly */
  668                         cmdif = (ipfw_insn_if *)cmd;
  669                         if (cmdif->name[0] != '\1')
  670                                 break;
  671 
  672                         cmdif->p.glob = cmdif->p.kidx;
  673                         break;
  674                 }
  675         }
  676 }
  677 
  678 /*
  679  * Add new rule(s) to the list possibly creating rule number for each.
  680  * Update the rule_number in the input struct so the caller knows it as well.
  681  * Must be called without IPFW_UH held
  682  */
  683 static int
  684 commit_rules(struct ip_fw_chain *chain, struct rule_check_info *rci, int count)
  685 {
  686         int error, i, insert_before, tcount;
  687         uint16_t rulenum, *pnum;
  688         struct rule_check_info *ci;
  689         struct ip_fw *krule;
  690         struct ip_fw **map;     /* the new array of pointers */
  691 
  692         /* Check if we need to do table/obj index remap */
  693         tcount = 0;
  694         for (ci = rci, i = 0; i < count; ci++, i++) {
  695                 if (ci->object_opcodes == 0)
  696                         continue;
  697 
  698                 /*
  699                  * Rule has some object opcodes.
  700                  * We need to find (and create non-existing)
  701                  * kernel objects, and reference existing ones.
  702                  */
  703                 error = rewrite_rule_uidx(chain, ci);
  704                 if (error != 0) {
  705                         /*
  706                          * rewrite failed, state for current rule
  707                          * has been reverted. Check if we need to
  708                          * revert more.
  709                          */
  710                         if (tcount > 0) {
  711                                 /*
  712                                  * We have some more table rules
  713                                  * we need to rollback.
  714                                  */
  715 
  716                                 IPFW_UH_WLOCK(chain);
  717                                 while (ci != rci) {
  718                                         ci--;
  719                                         if (ci->object_opcodes == 0)
  720                                                 continue;
  721                                         unref_rule_objects(chain,ci->krule);
  722                                 }
  723                                 IPFW_UH_WUNLOCK(chain);
  724                         }
  725 
  726                         return (error);
  727                 }
  728 
  729                 tcount++;
  730         }
  731 
  732         /* get_map returns with IPFW_UH_WLOCK if successful */
  733         map = get_map(chain, count, 0 /* not locked */);
  734         if (map == NULL) {
  735                 if (tcount > 0) {
  736                         /* Unbind tables */
  737                         IPFW_UH_WLOCK(chain);
  738                         for (ci = rci, i = 0; i < count; ci++, i++) {
  739                                 if (ci->object_opcodes == 0)
  740                                         continue;
  741 
  742                                 unref_rule_objects(chain, ci->krule);
  743                         }
  744                         IPFW_UH_WUNLOCK(chain);
  745                 }
  746 
  747                 return (ENOSPC);
  748         }
  749 
  750         if (V_autoinc_step < 1)
  751                 V_autoinc_step = 1;
  752         else if (V_autoinc_step > 1000)
  753                 V_autoinc_step = 1000;
  754 
  755         /* FIXME: Handle count > 1 */
  756         ci = rci;
  757         krule = ci->krule;
  758         rulenum = krule->rulenum;
  759 
  760         /* find the insertion point, we will insert before */
  761         insert_before = rulenum ? rulenum + 1 : IPFW_DEFAULT_RULE;
  762         i = ipfw_find_rule(chain, insert_before, 0);
  763         /* duplicate first part */
  764         if (i > 0)
  765                 bcopy(chain->map, map, i * sizeof(struct ip_fw *));
  766         map[i] = krule;
  767         /* duplicate remaining part, we always have the default rule */
  768         bcopy(chain->map + i, map + i + 1,
  769                 sizeof(struct ip_fw *) *(chain->n_rules - i));
  770         if (rulenum == 0) {
  771                 /* Compute rule number and write it back */
  772                 rulenum = i > 0 ? map[i-1]->rulenum : 0;
  773                 if (rulenum < IPFW_DEFAULT_RULE - V_autoinc_step)
  774                         rulenum += V_autoinc_step;
  775                 krule->rulenum = rulenum;
  776                 /* Save number to userland rule */
  777                 pnum = (uint16_t *)((caddr_t)ci->urule + ci->urule_numoff);
  778                 *pnum = rulenum;
  779         }
  780 
  781         krule->id = chain->id + 1;
  782         update_skipto_cache(chain, map);
  783         map = swap_map(chain, map, chain->n_rules + 1);
  784         chain->static_len += RULEUSIZE0(krule);
  785         IPFW_UH_WUNLOCK(chain);
  786         if (map)
  787                 free(map, M_IPFW);
  788         return (0);
  789 }
  790 
  791 int
  792 ipfw_add_protected_rule(struct ip_fw_chain *chain, struct ip_fw *rule,
  793     int locked)
  794 {
  795         struct ip_fw **map;
  796 
  797         map = get_map(chain, 1, locked);
  798         if (map == NULL)
  799                 return (ENOMEM);
  800         if (chain->n_rules > 0)
  801                 bcopy(chain->map, map,
  802                     chain->n_rules * sizeof(struct ip_fw *));
  803         map[chain->n_rules] = rule;
  804         rule->rulenum = IPFW_DEFAULT_RULE;
  805         rule->set = RESVD_SET;
  806         rule->id = chain->id + 1;
  807         /* We add rule in the end of chain, no need to update skipto cache */
  808         map = swap_map(chain, map, chain->n_rules + 1);
  809         chain->static_len += RULEUSIZE0(rule);
  810         IPFW_UH_WUNLOCK(chain);
  811         free(map, M_IPFW);
  812         return (0);
  813 }
  814 
  815 /*
  816  * Adds @rule to the list of rules to reap
  817  */
  818 void
  819 ipfw_reap_add(struct ip_fw_chain *chain, struct ip_fw **head,
  820     struct ip_fw *rule)
  821 {
  822 
  823         IPFW_UH_WLOCK_ASSERT(chain);
  824 
  825         /* Unlink rule from everywhere */
  826         unref_rule_objects(chain, rule);
  827 
  828         rule->next = *head;
  829         *head = rule;
  830 }
  831 
  832 /*
  833  * Reclaim storage associated with a list of rules.  This is
  834  * typically the list created using remove_rule.
  835  * A NULL pointer on input is handled correctly.
  836  */
  837 void
  838 ipfw_reap_rules(struct ip_fw *head)
  839 {
  840         struct ip_fw *rule;
  841 
  842         while ((rule = head) != NULL) {
  843                 head = head->next;
  844                 ipfw_free_rule(rule);
  845         }
  846 }
  847 
  848 /*
  849  * Rules to keep are
  850  *      (default || reserved || !match_set || !match_number)
  851  * where
  852  *   default ::= (rule->rulenum == IPFW_DEFAULT_RULE)
  853  *      // the default rule is always protected
  854  *
  855  *   reserved ::= (cmd == 0 && n == 0 && rule->set == RESVD_SET)
  856  *      // RESVD_SET is protected only if cmd == 0 and n == 0 ("ipfw flush")
  857  *
  858  *   match_set ::= (cmd == 0 || rule->set == set)
  859  *      // set number is ignored for cmd == 0
  860  *
  861  *   match_number ::= (cmd == 1 || n == 0 || n == rule->rulenum)
  862  *      // number is ignored for cmd == 1 or n == 0
  863  *
  864  */
  865 int
  866 ipfw_match_range(struct ip_fw *rule, ipfw_range_tlv *rt)
  867 {
  868 
  869         /* Don't match default rule for modification queries */
  870         if (rule->rulenum == IPFW_DEFAULT_RULE &&
  871             (rt->flags & IPFW_RCFLAG_DEFAULT) == 0)
  872                 return (0);
  873 
  874         /* Don't match rules in reserved set for flush requests */
  875         if ((rt->flags & IPFW_RCFLAG_ALL) != 0 && rule->set == RESVD_SET)
  876                 return (0);
  877 
  878         /* If we're filtering by set, don't match other sets */
  879         if ((rt->flags & IPFW_RCFLAG_SET) != 0 && rule->set != rt->set)
  880                 return (0);
  881 
  882         if ((rt->flags & IPFW_RCFLAG_RANGE) != 0 &&
  883             (rule->rulenum < rt->start_rule || rule->rulenum > rt->end_rule))
  884                 return (0);
  885 
  886         return (1);
  887 }
  888 
  889 struct manage_sets_args {
  890         uint16_t        set;
  891         uint8_t         new_set;
  892 };
  893 
  894 static int
  895 swap_sets_cb(struct namedobj_instance *ni, struct named_object *no,
  896     void *arg)
  897 {
  898         struct manage_sets_args *args;
  899 
  900         args = (struct manage_sets_args *)arg;
  901         if (no->set == (uint8_t)args->set)
  902                 no->set = args->new_set;
  903         else if (no->set == args->new_set)
  904                 no->set = (uint8_t)args->set;
  905         return (0);
  906 }
  907 
  908 static int
  909 move_sets_cb(struct namedobj_instance *ni, struct named_object *no,
  910     void *arg)
  911 {
  912         struct manage_sets_args *args;
  913 
  914         args = (struct manage_sets_args *)arg;
  915         if (no->set == (uint8_t)args->set)
  916                 no->set = args->new_set;
  917         return (0);
  918 }
  919 
  920 static int
  921 test_sets_cb(struct namedobj_instance *ni, struct named_object *no,
  922     void *arg)
  923 {
  924         struct manage_sets_args *args;
  925 
  926         args = (struct manage_sets_args *)arg;
  927         if (no->set != (uint8_t)args->set)
  928                 return (0);
  929         if (ipfw_objhash_lookup_name_type(ni, args->new_set,
  930             no->etlv, no->name) != NULL)
  931                 return (EEXIST);
  932         return (0);
  933 }
  934 
  935 /*
  936  * Generic function to handler moving and swapping sets.
  937  */
  938 int
  939 ipfw_obj_manage_sets(struct namedobj_instance *ni, uint16_t type,
  940     uint16_t set, uint8_t new_set, enum ipfw_sets_cmd cmd)
  941 {
  942         struct manage_sets_args args;
  943         struct named_object *no;
  944 
  945         args.set = set;
  946         args.new_set = new_set;
  947         switch (cmd) {
  948         case SWAP_ALL:
  949                 return (ipfw_objhash_foreach_type(ni, swap_sets_cb,
  950                     &args, type));
  951         case TEST_ALL:
  952                 return (ipfw_objhash_foreach_type(ni, test_sets_cb,
  953                     &args, type));
  954         case MOVE_ALL:
  955                 return (ipfw_objhash_foreach_type(ni, move_sets_cb,
  956                     &args, type));
  957         case COUNT_ONE:
  958                 /*
  959                  * @set used to pass kidx.
  960                  * When @new_set is zero - reset object counter,
  961                  * otherwise increment it.
  962                  */
  963                 no = ipfw_objhash_lookup_kidx(ni, set);
  964                 if (new_set != 0)
  965                         no->ocnt++;
  966                 else
  967                         no->ocnt = 0;
  968                 return (0);
  969         case TEST_ONE:
  970                 /* @set used to pass kidx */
  971                 no = ipfw_objhash_lookup_kidx(ni, set);
  972                 /*
  973                  * First check number of references:
  974                  * when it differs, this mean other rules are holding
  975                  * reference to given object, so it is not possible to
  976                  * change its set. Note that refcnt may account references
  977                  * to some going-to-be-added rules. Since we don't know
  978                  * their numbers (and even if they will be added) it is
  979                  * perfectly OK to return error here.
  980                  */
  981                 if (no->ocnt != no->refcnt)
  982                         return (EBUSY);
  983                 if (ipfw_objhash_lookup_name_type(ni, new_set, type,
  984                     no->name) != NULL)
  985                         return (EEXIST);
  986                 return (0);
  987         case MOVE_ONE:
  988                 /* @set used to pass kidx */
  989                 no = ipfw_objhash_lookup_kidx(ni, set);
  990                 no->set = new_set;
  991                 return (0);
  992         }
  993         return (EINVAL);
  994 }
  995 
  996 /*
  997  * Delete rules matching range @rt.
  998  * Saves number of deleted rules in @ndel.
  999  *
 1000  * Returns 0 on success.
 1001  */
 1002 static int
 1003 delete_range(struct ip_fw_chain *chain, ipfw_range_tlv *rt, int *ndel)
 1004 {
 1005         struct ip_fw *reap, *rule, **map;
 1006         int end, start;
 1007         int i, n, ndyn, ofs;
 1008 
 1009         reap = NULL;
 1010         IPFW_UH_WLOCK(chain);   /* arbitrate writers */
 1011 
 1012         /*
 1013          * Stage 1: Determine range to inspect.
 1014          * Range is half-inclusive, e.g [start, end).
 1015          */
 1016         start = 0;
 1017         end = chain->n_rules - 1;
 1018 
 1019         if ((rt->flags & IPFW_RCFLAG_RANGE) != 0) {
 1020                 start = ipfw_find_rule(chain, rt->start_rule, 0);
 1021 
 1022                 if (rt->end_rule >= IPFW_DEFAULT_RULE)
 1023                         rt->end_rule = IPFW_DEFAULT_RULE - 1;
 1024                 end = ipfw_find_rule(chain, rt->end_rule, UINT32_MAX);
 1025         }
 1026 
 1027         if (rt->flags & IPFW_RCFLAG_DYNAMIC) {
 1028                 /*
 1029                  * Requested deleting only for dynamic states.
 1030                  */
 1031                 *ndel = 0;
 1032                 ipfw_expire_dyn_states(chain, rt);
 1033                 IPFW_UH_WUNLOCK(chain);
 1034                 return (0);
 1035         }
 1036 
 1037         /* Allocate new map of the same size */
 1038         map = get_map(chain, 0, 1 /* locked */);
 1039         if (map == NULL) {
 1040                 IPFW_UH_WUNLOCK(chain);
 1041                 return (ENOMEM);
 1042         }
 1043 
 1044         n = 0;
 1045         ndyn = 0;
 1046         ofs = start;
 1047         /* 1. bcopy the initial part of the map */
 1048         if (start > 0)
 1049                 bcopy(chain->map, map, start * sizeof(struct ip_fw *));
 1050         /* 2. copy active rules between start and end */
 1051         for (i = start; i < end; i++) {
 1052                 rule = chain->map[i];
 1053                 if (ipfw_match_range(rule, rt) == 0) {
 1054                         map[ofs++] = rule;
 1055                         continue;
 1056                 }
 1057 
 1058                 n++;
 1059                 if (ipfw_is_dyn_rule(rule) != 0)
 1060                         ndyn++;
 1061         }
 1062         /* 3. copy the final part of the map */
 1063         bcopy(chain->map + end, map + ofs,
 1064                 (chain->n_rules - end) * sizeof(struct ip_fw *));
 1065         /* 4. recalculate skipto cache */
 1066         update_skipto_cache(chain, map);
 1067         /* 5. swap the maps (under UH_WLOCK + WHLOCK) */
 1068         map = swap_map(chain, map, chain->n_rules - n);
 1069         /* 6. Remove all dynamic states originated by deleted rules */
 1070         if (ndyn > 0)
 1071                 ipfw_expire_dyn_states(chain, rt);
 1072         /* 7. now remove the rules deleted from the old map */
 1073         for (i = start; i < end; i++) {
 1074                 rule = map[i];
 1075                 if (ipfw_match_range(rule, rt) == 0)
 1076                         continue;
 1077                 chain->static_len -= RULEUSIZE0(rule);
 1078                 ipfw_reap_add(chain, &reap, rule);
 1079         }
 1080         IPFW_UH_WUNLOCK(chain);
 1081 
 1082         ipfw_reap_rules(reap);
 1083         if (map != NULL)
 1084                 free(map, M_IPFW);
 1085         *ndel = n;
 1086         return (0);
 1087 }
 1088 
 1089 static int
 1090 move_objects(struct ip_fw_chain *ch, ipfw_range_tlv *rt)
 1091 {
 1092         struct opcode_obj_rewrite *rw;
 1093         struct ip_fw *rule;
 1094         ipfw_insn *cmd;
 1095         int cmdlen, i, l, c;
 1096         uint16_t kidx;
 1097 
 1098         IPFW_UH_WLOCK_ASSERT(ch);
 1099 
 1100         /* Stage 1: count number of references by given rules */
 1101         for (c = 0, i = 0; i < ch->n_rules - 1; i++) {
 1102                 rule = ch->map[i];
 1103                 if (ipfw_match_range(rule, rt) == 0)
 1104                         continue;
 1105                 if (rule->set == rt->new_set) /* nothing to do */
 1106                         continue;
 1107                 /* Search opcodes with named objects */
 1108                 for (l = rule->cmd_len, cmdlen = 0, cmd = rule->cmd;
 1109                     l > 0; l -= cmdlen, cmd += cmdlen) {
 1110                         cmdlen = F_LEN(cmd);
 1111                         rw = find_op_rw(cmd, &kidx, NULL);
 1112                         if (rw == NULL || rw->manage_sets == NULL)
 1113                                 continue;
 1114                         /*
 1115                          * When manage_sets() returns non-zero value to
 1116                          * COUNT_ONE command, consider this as an object
 1117                          * doesn't support sets (e.g. disabled with sysctl).
 1118                          * So, skip checks for this object.
 1119                          */
 1120                         if (rw->manage_sets(ch, kidx, 1, COUNT_ONE) != 0)
 1121                                 continue;
 1122                         c++;
 1123                 }
 1124         }
 1125         if (c == 0) /* No objects found */
 1126                 return (0);
 1127         /* Stage 2: verify "ownership" */
 1128         for (c = 0, i = 0; (i < ch->n_rules - 1) && c == 0; i++) {
 1129                 rule = ch->map[i];
 1130                 if (ipfw_match_range(rule, rt) == 0)
 1131                         continue;
 1132                 if (rule->set == rt->new_set) /* nothing to do */
 1133                         continue;
 1134                 /* Search opcodes with named objects */
 1135                 for (l = rule->cmd_len, cmdlen = 0, cmd = rule->cmd;
 1136                     l > 0 && c == 0; l -= cmdlen, cmd += cmdlen) {
 1137                         cmdlen = F_LEN(cmd);
 1138                         rw = find_op_rw(cmd, &kidx, NULL);
 1139                         if (rw == NULL || rw->manage_sets == NULL)
 1140                                 continue;
 1141                         /* Test for ownership and conflicting names */
 1142                         c = rw->manage_sets(ch, kidx,
 1143                             (uint8_t)rt->new_set, TEST_ONE);
 1144                 }
 1145         }
 1146         /* Stage 3: change set and cleanup */
 1147         for (i = 0; i < ch->n_rules - 1; i++) {
 1148                 rule = ch->map[i];
 1149                 if (ipfw_match_range(rule, rt) == 0)
 1150                         continue;
 1151                 if (rule->set == rt->new_set) /* nothing to do */
 1152                         continue;
 1153                 /* Search opcodes with named objects */
 1154                 for (l = rule->cmd_len, cmdlen = 0, cmd = rule->cmd;
 1155                     l > 0; l -= cmdlen, cmd += cmdlen) {
 1156                         cmdlen = F_LEN(cmd);
 1157                         rw = find_op_rw(cmd, &kidx, NULL);
 1158                         if (rw == NULL || rw->manage_sets == NULL)
 1159                                 continue;
 1160                         /* cleanup object counter */
 1161                         rw->manage_sets(ch, kidx,
 1162                             0 /* reset counter */, COUNT_ONE);
 1163                         if (c != 0)
 1164                                 continue;
 1165                         /* change set */
 1166                         rw->manage_sets(ch, kidx,
 1167                             (uint8_t)rt->new_set, MOVE_ONE);
 1168                 }
 1169         }
 1170         return (c);
 1171 }
 1172 
 1173 /*
 1174  * Changes set of given rule rannge @rt
 1175  * with each other.
 1176  *
 1177  * Returns 0 on success.
 1178  */
 1179 static int
 1180 move_range(struct ip_fw_chain *chain, ipfw_range_tlv *rt)
 1181 {
 1182         struct ip_fw *rule;
 1183         int i;
 1184 
 1185         IPFW_UH_WLOCK(chain);
 1186 
 1187         /*
 1188          * Move rules with matching paramenerts to a new set.
 1189          * This one is much more complex. We have to ensure
 1190          * that all referenced tables (if any) are referenced
 1191          * by given rule subset only. Otherwise, we can't move
 1192          * them to new set and have to return error.
 1193          */
 1194         if ((i = move_objects(chain, rt)) != 0) {
 1195                 IPFW_UH_WUNLOCK(chain);
 1196                 return (i);
 1197         }
 1198 
 1199         /* XXX: We have to do swap holding WLOCK */
 1200         for (i = 0; i < chain->n_rules; i++) {
 1201                 rule = chain->map[i];
 1202                 if (ipfw_match_range(rule, rt) == 0)
 1203                         continue;
 1204                 rule->set = rt->new_set;
 1205         }
 1206 
 1207         IPFW_UH_WUNLOCK(chain);
 1208 
 1209         return (0);
 1210 }
 1211 
 1212 /*
 1213  * Returns pointer to action instruction, skips all possible rule
 1214  * modifiers like O_LOG, O_TAG, O_ALTQ.
 1215  */
 1216 ipfw_insn *
 1217 ipfw_get_action(struct ip_fw *rule)
 1218 {
 1219         ipfw_insn *cmd;
 1220         int l, cmdlen;
 1221 
 1222         cmd = ACTION_PTR(rule);
 1223         l = rule->cmd_len - rule->act_ofs;
 1224         while (l > 0) {
 1225                 switch (cmd->opcode) {
 1226                 case O_ALTQ:
 1227                 case O_LOG:
 1228                 case O_TAG:
 1229                         break;
 1230                 default:
 1231                         return (cmd);
 1232                 }
 1233                 cmdlen = F_LEN(cmd);
 1234                 l -= cmdlen;
 1235                 cmd += cmdlen;
 1236         }
 1237         panic("%s: rule (%p) has not action opcode", __func__, rule);
 1238         return (NULL);
 1239 }
 1240 
 1241 /*
 1242  * Clear counters for a specific rule.
 1243  * Normally run under IPFW_UH_RLOCK, but these are idempotent ops
 1244  * so we only care that rules do not disappear.
 1245  */
 1246 static void
 1247 clear_counters(struct ip_fw *rule, int log_only)
 1248 {
 1249         ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule);
 1250 
 1251         if (log_only == 0)
 1252                 IPFW_ZERO_RULE_COUNTER(rule);
 1253         if (l->o.opcode == O_LOG)
 1254                 l->log_left = l->max_log;
 1255 }
 1256 
 1257 /*
 1258  * Flushes rules counters and/or log values on matching range.
 1259  *
 1260  * Returns number of items cleared.
 1261  */
 1262 static int
 1263 clear_range(struct ip_fw_chain *chain, ipfw_range_tlv *rt, int log_only)
 1264 {
 1265         struct ip_fw *rule;
 1266         int num;
 1267         int i;
 1268 
 1269         num = 0;
 1270         rt->flags |= IPFW_RCFLAG_DEFAULT;
 1271 
 1272         IPFW_UH_WLOCK(chain);   /* arbitrate writers */
 1273         for (i = 0; i < chain->n_rules; i++) {
 1274                 rule = chain->map[i];
 1275                 if (ipfw_match_range(rule, rt) == 0)
 1276                         continue;
 1277                 clear_counters(rule, log_only);
 1278                 num++;
 1279         }
 1280         IPFW_UH_WUNLOCK(chain);
 1281 
 1282         return (num);
 1283 }
 1284 
 1285 static int
 1286 check_range_tlv(ipfw_range_tlv *rt)
 1287 {
 1288 
 1289         if (rt->head.length != sizeof(*rt))
 1290                 return (1);
 1291         if (rt->start_rule > rt->end_rule)
 1292                 return (1);
 1293         if (rt->set >= IPFW_MAX_SETS || rt->new_set >= IPFW_MAX_SETS)
 1294                 return (1);
 1295 
 1296         if ((rt->flags & IPFW_RCFLAG_USER) != rt->flags)
 1297                 return (1);
 1298 
 1299         return (0);
 1300 }
 1301 
 1302 /*
 1303  * Delete rules matching specified parameters
 1304  * Data layout (v0)(current):
 1305  * Request: [ ipfw_obj_header ipfw_range_tlv ]
 1306  * Reply: [ ipfw_obj_header ipfw_range_tlv ]
 1307  *
 1308  * Saves number of deleted rules in ipfw_range_tlv->new_set.
 1309  *
 1310  * Returns 0 on success.
 1311  */
 1312 static int
 1313 del_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
 1314     struct sockopt_data *sd)
 1315 {
 1316         ipfw_range_header *rh;
 1317         int error, ndel;
 1318 
 1319         if (sd->valsize != sizeof(*rh))
 1320                 return (EINVAL);
 1321 
 1322         rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize);
 1323 
 1324         if (check_range_tlv(&rh->range) != 0)
 1325                 return (EINVAL);
 1326 
 1327         ndel = 0;
 1328         if ((error = delete_range(chain, &rh->range, &ndel)) != 0)
 1329                 return (error);
 1330 
 1331         /* Save number of rules deleted */
 1332         rh->range.new_set = ndel;
 1333         return (0);
 1334 }
 1335 
 1336 /*
 1337  * Move rules/sets matching specified parameters
 1338  * Data layout (v0)(current):
 1339  * Request: [ ipfw_obj_header ipfw_range_tlv ]
 1340  *
 1341  * Returns 0 on success.
 1342  */
 1343 static int
 1344 move_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
 1345     struct sockopt_data *sd)
 1346 {
 1347         ipfw_range_header *rh;
 1348 
 1349         if (sd->valsize != sizeof(*rh))
 1350                 return (EINVAL);
 1351 
 1352         rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize);
 1353 
 1354         if (check_range_tlv(&rh->range) != 0)
 1355                 return (EINVAL);
 1356 
 1357         return (move_range(chain, &rh->range));
 1358 }
 1359 
 1360 /*
 1361  * Clear rule accounting data matching specified parameters
 1362  * Data layout (v0)(current):
 1363  * Request: [ ipfw_obj_header ipfw_range_tlv ]
 1364  * Reply: [ ipfw_obj_header ipfw_range_tlv ]
 1365  *
 1366  * Saves number of cleared rules in ipfw_range_tlv->new_set.
 1367  *
 1368  * Returns 0 on success.
 1369  */
 1370 static int
 1371 clear_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
 1372     struct sockopt_data *sd)
 1373 {
 1374         ipfw_range_header *rh;
 1375         int log_only, num;
 1376         char *msg;
 1377 
 1378         if (sd->valsize != sizeof(*rh))
 1379                 return (EINVAL);
 1380 
 1381         rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize);
 1382 
 1383         if (check_range_tlv(&rh->range) != 0)
 1384                 return (EINVAL);
 1385 
 1386         log_only = (op3->opcode == IP_FW_XRESETLOG);
 1387 
 1388         num = clear_range(chain, &rh->range, log_only);
 1389 
 1390         if (rh->range.flags & IPFW_RCFLAG_ALL)
 1391                 msg = log_only ? "All logging counts reset" :
 1392                     "Accounting cleared";
 1393         else
 1394                 msg = log_only ? "logging count reset" : "cleared";
 1395 
 1396         if (V_fw_verbose) {
 1397                 int lev = LOG_SECURITY | LOG_NOTICE;
 1398                 log(lev, "ipfw: %s.\n", msg);
 1399         }
 1400 
 1401         /* Save number of rules cleared */
 1402         rh->range.new_set = num;
 1403         return (0);
 1404 }
 1405 
 1406 static void
 1407 enable_sets(struct ip_fw_chain *chain, ipfw_range_tlv *rt)
 1408 {
 1409         uint32_t v_set;
 1410 
 1411         IPFW_UH_WLOCK_ASSERT(chain);
 1412 
 1413         /* Change enabled/disabled sets mask */
 1414         v_set = (V_set_disable | rt->set) & ~rt->new_set;
 1415         v_set &= ~(1 << RESVD_SET); /* set RESVD_SET always enabled */
 1416         IPFW_WLOCK(chain);
 1417         V_set_disable = v_set;
 1418         IPFW_WUNLOCK(chain);
 1419 }
 1420 
 1421 static int
 1422 swap_sets(struct ip_fw_chain *chain, ipfw_range_tlv *rt, int mv)
 1423 {
 1424         struct opcode_obj_rewrite *rw;
 1425         struct ip_fw *rule;
 1426         int i;
 1427 
 1428         IPFW_UH_WLOCK_ASSERT(chain);
 1429 
 1430         if (rt->set == rt->new_set) /* nothing to do */
 1431                 return (0);
 1432 
 1433         if (mv != 0) {
 1434                 /*
 1435                  * Berfore moving the rules we need to check that
 1436                  * there aren't any conflicting named objects.
 1437                  */
 1438                 for (rw = ctl3_rewriters;
 1439                     rw < ctl3_rewriters + ctl3_rsize; rw++) {
 1440                         if (rw->manage_sets == NULL)
 1441                                 continue;
 1442                         i = rw->manage_sets(chain, (uint8_t)rt->set,
 1443                             (uint8_t)rt->new_set, TEST_ALL);
 1444                         if (i != 0)
 1445                                 return (EEXIST);
 1446                 }
 1447         }
 1448         /* Swap or move two sets */
 1449         for (i = 0; i < chain->n_rules - 1; i++) {
 1450                 rule = chain->map[i];
 1451                 if (rule->set == (uint8_t)rt->set)
 1452                         rule->set = (uint8_t)rt->new_set;
 1453                 else if (rule->set == (uint8_t)rt->new_set && mv == 0)
 1454                         rule->set = (uint8_t)rt->set;
 1455         }
 1456         for (rw = ctl3_rewriters; rw < ctl3_rewriters + ctl3_rsize; rw++) {
 1457                 if (rw->manage_sets == NULL)
 1458                         continue;
 1459                 rw->manage_sets(chain, (uint8_t)rt->set,
 1460                     (uint8_t)rt->new_set, mv != 0 ? MOVE_ALL: SWAP_ALL);
 1461         }
 1462         return (0);
 1463 }
 1464 
 1465 /*
 1466  * Swaps or moves set
 1467  * Data layout (v0)(current):
 1468  * Request: [ ipfw_obj_header ipfw_range_tlv ]
 1469  *
 1470  * Returns 0 on success.
 1471  */
 1472 static int
 1473 manage_sets(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
 1474     struct sockopt_data *sd)
 1475 {
 1476         ipfw_range_header *rh;
 1477         int ret;
 1478 
 1479         if (sd->valsize != sizeof(*rh))
 1480                 return (EINVAL);
 1481 
 1482         rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize);
 1483 
 1484         if (rh->range.head.length != sizeof(ipfw_range_tlv))
 1485                 return (1);
 1486         /* enable_sets() expects bitmasks. */
 1487         if (op3->opcode != IP_FW_SET_ENABLE &&
 1488             (rh->range.set >= IPFW_MAX_SETS ||
 1489             rh->range.new_set >= IPFW_MAX_SETS))
 1490                 return (EINVAL);
 1491 
 1492         ret = 0;
 1493         IPFW_UH_WLOCK(chain);
 1494         switch (op3->opcode) {
 1495         case IP_FW_SET_SWAP:
 1496         case IP_FW_SET_MOVE:
 1497                 ret = swap_sets(chain, &rh->range,
 1498                     op3->opcode == IP_FW_SET_MOVE);
 1499                 break;
 1500         case IP_FW_SET_ENABLE:
 1501                 enable_sets(chain, &rh->range);
 1502                 break;
 1503         }
 1504         IPFW_UH_WUNLOCK(chain);
 1505 
 1506         return (ret);
 1507 }
 1508 
 1509 /**
 1510  * Remove all rules with given number, or do set manipulation.
 1511  * Assumes chain != NULL && *chain != NULL.
 1512  *
 1513  * The argument is an uint32_t. The low 16 bit are the rule or set number;
 1514  * the next 8 bits are the new set; the top 8 bits indicate the command:
 1515  *
 1516  *      0       delete rules numbered "rulenum"
 1517  *      1       delete rules in set "rulenum"
 1518  *      2       move rules "rulenum" to set "new_set"
 1519  *      3       move rules from set "rulenum" to set "new_set"
 1520  *      4       swap sets "rulenum" and "new_set"
 1521  *      5       delete rules "rulenum" and set "new_set"
 1522  */
 1523 static int
 1524 del_entry(struct ip_fw_chain *chain, uint32_t arg)
 1525 {
 1526         uint32_t num;   /* rule number or old_set */
 1527         uint8_t cmd, new_set;
 1528         int do_del, ndel;
 1529         int error = 0;
 1530         ipfw_range_tlv rt;
 1531 
 1532         num = arg & 0xffff;
 1533         cmd = (arg >> 24) & 0xff;
 1534         new_set = (arg >> 16) & 0xff;
 1535 
 1536         if (cmd > 5 || new_set > RESVD_SET)
 1537                 return EINVAL;
 1538         if (cmd == 0 || cmd == 2 || cmd == 5) {
 1539                 if (num >= IPFW_DEFAULT_RULE)
 1540                         return EINVAL;
 1541         } else {
 1542                 if (num > RESVD_SET)    /* old_set */
 1543                         return EINVAL;
 1544         }
 1545 
 1546         /* Convert old requests into new representation */
 1547         memset(&rt, 0, sizeof(rt));
 1548         rt.start_rule = num;
 1549         rt.end_rule = num;
 1550         rt.set = num;
 1551         rt.new_set = new_set;
 1552         do_del = 0;
 1553 
 1554         switch (cmd) {
 1555         case 0: /* delete rules numbered "rulenum" */
 1556                 if (num == 0)
 1557                         rt.flags |= IPFW_RCFLAG_ALL;
 1558                 else
 1559                         rt.flags |= IPFW_RCFLAG_RANGE;
 1560                 do_del = 1;
 1561                 break;
 1562         case 1: /* delete rules in set "rulenum" */
 1563                 rt.flags |= IPFW_RCFLAG_SET;
 1564                 do_del = 1;
 1565                 break;
 1566         case 5: /* delete rules "rulenum" and set "new_set" */
 1567                 rt.flags |= IPFW_RCFLAG_RANGE | IPFW_RCFLAG_SET;
 1568                 rt.set = new_set;
 1569                 rt.new_set = 0;
 1570                 do_del = 1;
 1571                 break;
 1572         case 2: /* move rules "rulenum" to set "new_set" */
 1573                 rt.flags |= IPFW_RCFLAG_RANGE;
 1574                 break;
 1575         case 3: /* move rules from set "rulenum" to set "new_set" */
 1576                 IPFW_UH_WLOCK(chain);
 1577                 error = swap_sets(chain, &rt, 1);
 1578                 IPFW_UH_WUNLOCK(chain);
 1579                 return (error);
 1580         case 4: /* swap sets "rulenum" and "new_set" */
 1581                 IPFW_UH_WLOCK(chain);
 1582                 error = swap_sets(chain, &rt, 0);
 1583                 IPFW_UH_WUNLOCK(chain);
 1584                 return (error);
 1585         default:
 1586                 return (ENOTSUP);
 1587         }
 1588 
 1589         if (do_del != 0) {
 1590                 if ((error = delete_range(chain, &rt, &ndel)) != 0)
 1591                         return (error);
 1592 
 1593                 if (ndel == 0 && (cmd != 1 && num != 0))
 1594                         return (EINVAL);
 1595 
 1596                 return (0);
 1597         }
 1598 
 1599         return (move_range(chain, &rt));
 1600 }
 1601 
 1602 /**
 1603  * Reset some or all counters on firewall rules.
 1604  * The argument `arg' is an u_int32_t. The low 16 bit are the rule number,
 1605  * the next 8 bits are the set number, the top 8 bits are the command:
 1606  *      0       work with rules from all set's;
 1607  *      1       work with rules only from specified set.
 1608  * Specified rule number is zero if we want to clear all entries.
 1609  * log_only is 1 if we only want to reset logs, zero otherwise.
 1610  */
 1611 static int
 1612 zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only)
 1613 {
 1614         struct ip_fw *rule;
 1615         char *msg;
 1616         int i;
 1617 
 1618         uint16_t rulenum = arg & 0xffff;
 1619         uint8_t set = (arg >> 16) & 0xff;
 1620         uint8_t cmd = (arg >> 24) & 0xff;
 1621 
 1622         if (cmd > 1)
 1623                 return (EINVAL);
 1624         if (cmd == 1 && set > RESVD_SET)
 1625                 return (EINVAL);
 1626 
 1627         IPFW_UH_RLOCK(chain);
 1628         if (rulenum == 0) {
 1629                 V_norule_counter = 0;
 1630                 for (i = 0; i < chain->n_rules; i++) {
 1631                         rule = chain->map[i];
 1632                         /* Skip rules not in our set. */
 1633                         if (cmd == 1 && rule->set != set)
 1634                                 continue;
 1635                         clear_counters(rule, log_only);
 1636                 }
 1637                 msg = log_only ? "All logging counts reset" :
 1638                     "Accounting cleared";
 1639         } else {
 1640                 int cleared = 0;
 1641                 for (i = 0; i < chain->n_rules; i++) {
 1642                         rule = chain->map[i];
 1643                         if (rule->rulenum == rulenum) {
 1644                                 if (cmd == 0 || rule->set == set)
 1645                                         clear_counters(rule, log_only);
 1646                                 cleared = 1;
 1647                         }
 1648                         if (rule->rulenum > rulenum)
 1649                                 break;
 1650                 }
 1651                 if (!cleared) { /* we did not find any matching rules */
 1652                         IPFW_UH_RUNLOCK(chain);
 1653                         return (EINVAL);
 1654                 }
 1655                 msg = log_only ? "logging count reset" : "cleared";
 1656         }
 1657         IPFW_UH_RUNLOCK(chain);
 1658 
 1659         if (V_fw_verbose) {
 1660                 int lev = LOG_SECURITY | LOG_NOTICE;
 1661 
 1662                 if (rulenum)
 1663                         log(lev, "ipfw: Entry %d %s.\n", rulenum, msg);
 1664                 else
 1665                         log(lev, "ipfw: %s.\n", msg);
 1666         }
 1667         return (0);
 1668 }
 1669 
 1670 /*
 1671  * Check rule head in FreeBSD11 format
 1672  *
 1673  */
 1674 static int
 1675 check_ipfw_rule1(struct ip_fw_rule *rule, int size,
 1676     struct rule_check_info *ci)
 1677 {
 1678         int l;
 1679 
 1680         if (size < sizeof(*rule)) {
 1681                 printf("ipfw: rule too short\n");
 1682                 return (EINVAL);
 1683         }
 1684 
 1685         /* Check for valid cmd_len */
 1686         l = roundup2(RULESIZE(rule), sizeof(uint64_t));
 1687         if (l != size) {
 1688                 printf("ipfw: size mismatch (have %d want %d)\n", size, l);
 1689                 return (EINVAL);
 1690         }
 1691         if (rule->act_ofs >= rule->cmd_len) {
 1692                 printf("ipfw: bogus action offset (%u > %u)\n",
 1693                     rule->act_ofs, rule->cmd_len - 1);
 1694                 return (EINVAL);
 1695         }
 1696 
 1697         if (rule->rulenum > IPFW_DEFAULT_RULE - 1)
 1698                 return (EINVAL);
 1699 
 1700         return (check_ipfw_rule_body(rule->cmd, rule->cmd_len, ci));
 1701 }
 1702 
 1703 /*
 1704  * Check rule head in FreeBSD8 format
 1705  *
 1706  */
 1707 static int
 1708 check_ipfw_rule0(struct ip_fw_rule0 *rule, int size,
 1709     struct rule_check_info *ci)
 1710 {
 1711         int l;
 1712 
 1713         if (size < sizeof(*rule)) {
 1714                 printf("ipfw: rule too short\n");
 1715                 return (EINVAL);
 1716         }
 1717 
 1718         /* Check for valid cmd_len */
 1719         l = sizeof(*rule) + rule->cmd_len * 4 - 4;
 1720         if (l != size) {
 1721                 printf("ipfw: size mismatch (have %d want %d)\n", size, l);
 1722                 return (EINVAL);
 1723         }
 1724         if (rule->act_ofs >= rule->cmd_len) {
 1725                 printf("ipfw: bogus action offset (%u > %u)\n",
 1726                     rule->act_ofs, rule->cmd_len - 1);
 1727                 return (EINVAL);
 1728         }
 1729 
 1730         if (rule->rulenum > IPFW_DEFAULT_RULE - 1)
 1731                 return (EINVAL);
 1732 
 1733         return (check_ipfw_rule_body(rule->cmd, rule->cmd_len, ci));
 1734 }
 1735 
 1736 static int
 1737 check_ipfw_rule_body(ipfw_insn *cmd, int cmd_len, struct rule_check_info *ci)
 1738 {
 1739         int cmdlen, l;
 1740         int have_action;
 1741 
 1742         have_action = 0;
 1743 
 1744         /*
 1745          * Now go for the individual checks. Very simple ones, basically only
 1746          * instruction sizes.
 1747          */
 1748         for (l = cmd_len; l > 0 ; l -= cmdlen, cmd += cmdlen) {
 1749                 cmdlen = F_LEN(cmd);
 1750                 if (cmdlen > l) {
 1751                         printf("ipfw: opcode %d size truncated\n",
 1752                             cmd->opcode);
 1753                         return EINVAL;
 1754                 }
 1755                 switch (cmd->opcode) {
 1756                 case O_PROBE_STATE:
 1757                 case O_KEEP_STATE:
 1758                         if (cmdlen != F_INSN_SIZE(ipfw_insn))
 1759                                 goto bad_size;
 1760                         ci->object_opcodes++;
 1761                         break;
 1762                 case O_PROTO:
 1763                 case O_IP_SRC_ME:
 1764                 case O_IP_DST_ME:
 1765                 case O_LAYER2:
 1766                 case O_IN:
 1767                 case O_FRAG:
 1768                 case O_DIVERTED:
 1769                 case O_IPOPT:
 1770                 case O_IPTOS:
 1771                 case O_IPPRECEDENCE:
 1772                 case O_IPVER:
 1773                 case O_SOCKARG:
 1774                 case O_TCPFLAGS:
 1775                 case O_TCPOPTS:
 1776                 case O_ESTAB:
 1777                 case O_VERREVPATH:
 1778                 case O_VERSRCREACH:
 1779                 case O_ANTISPOOF:
 1780                 case O_IPSEC:
 1781 #ifdef INET6
 1782                 case O_IP6_SRC_ME:
 1783                 case O_IP6_DST_ME:
 1784                 case O_EXT_HDR:
 1785                 case O_IP6:
 1786 #endif
 1787                 case O_IP4:
 1788                 case O_TAG:
 1789                 case O_SKIP_ACTION:
 1790                         if (cmdlen != F_INSN_SIZE(ipfw_insn))
 1791                                 goto bad_size;
 1792                         break;
 1793 
 1794                 case O_EXTERNAL_ACTION:
 1795                         if (cmd->arg1 == 0 ||
 1796                             cmdlen != F_INSN_SIZE(ipfw_insn)) {
 1797                                 printf("ipfw: invalid external "
 1798                                     "action opcode\n");
 1799                                 return (EINVAL);
 1800                         }
 1801                         ci->object_opcodes++;
 1802                         /*
 1803                          * Do we have O_EXTERNAL_INSTANCE or O_EXTERNAL_DATA
 1804                          * opcode?
 1805                          */
 1806                         if (l != cmdlen) {
 1807                                 l -= cmdlen;
 1808                                 cmd += cmdlen;
 1809                                 cmdlen = F_LEN(cmd);
 1810                                 if (cmd->opcode == O_EXTERNAL_DATA)
 1811                                         goto check_action;
 1812                                 if (cmd->opcode != O_EXTERNAL_INSTANCE) {
 1813                                         printf("ipfw: invalid opcode "
 1814                                             "next to external action %u\n",
 1815                                             cmd->opcode);
 1816                                         return (EINVAL);
 1817                                 }
 1818                                 if (cmd->arg1 == 0 ||
 1819                                     cmdlen != F_INSN_SIZE(ipfw_insn)) {
 1820                                         printf("ipfw: invalid external "
 1821                                             "action instance opcode\n");
 1822                                         return (EINVAL);
 1823                                 }
 1824                                 ci->object_opcodes++;
 1825                         }
 1826                         goto check_action;
 1827 
 1828                 case O_FIB:
 1829                         if (cmdlen != F_INSN_SIZE(ipfw_insn))
 1830                                 goto bad_size;
 1831                         if (cmd->arg1 >= rt_numfibs) {
 1832                                 printf("ipfw: invalid fib number %d\n",
 1833                                         cmd->arg1);
 1834                                 return EINVAL;
 1835                         }
 1836                         break;
 1837 
 1838                 case O_SETFIB:
 1839                         if (cmdlen != F_INSN_SIZE(ipfw_insn))
 1840                                 goto bad_size;
 1841                         if ((cmd->arg1 != IP_FW_TARG) &&
 1842                             ((cmd->arg1 & 0x7FFF) >= rt_numfibs)) {
 1843                                 printf("ipfw: invalid fib number %d\n",
 1844                                         cmd->arg1 & 0x7FFF);
 1845                                 return EINVAL;
 1846                         }
 1847                         goto check_action;
 1848 
 1849                 case O_UID:
 1850                 case O_GID:
 1851                 case O_JAIL:
 1852                 case O_IP_SRC:
 1853                 case O_IP_DST:
 1854                 case O_TCPSEQ:
 1855                 case O_TCPACK:
 1856                 case O_PROB:
 1857                 case O_ICMPTYPE:
 1858                         if (cmdlen != F_INSN_SIZE(ipfw_insn_u32))
 1859                                 goto bad_size;
 1860                         break;
 1861 
 1862                 case O_LIMIT:
 1863                         if (cmdlen != F_INSN_SIZE(ipfw_insn_limit))
 1864                                 goto bad_size;
 1865                         ci->object_opcodes++;
 1866                         break;
 1867 
 1868                 case O_LOG:
 1869                         if (cmdlen != F_INSN_SIZE(ipfw_insn_log))
 1870                                 goto bad_size;
 1871 
 1872                         ((ipfw_insn_log *)cmd)->log_left =
 1873                             ((ipfw_insn_log *)cmd)->max_log;
 1874 
 1875                         break;
 1876 
 1877                 case O_IP_SRC_MASK:
 1878                 case O_IP_DST_MASK:
 1879                         /* only odd command lengths */
 1880                         if ((cmdlen & 1) == 0)
 1881                                 goto bad_size;
 1882                         break;
 1883 
 1884                 case O_IP_SRC_SET:
 1885                 case O_IP_DST_SET:
 1886                         if (cmd->arg1 == 0 || cmd->arg1 > 256) {
 1887                                 printf("ipfw: invalid set size %d\n",
 1888                                         cmd->arg1);
 1889                                 return EINVAL;
 1890                         }
 1891                         if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
 1892                             (cmd->arg1+31)/32 )
 1893                                 goto bad_size;
 1894                         break;
 1895 
 1896                 case O_IP_SRC_LOOKUP:
 1897                         if (cmdlen > F_INSN_SIZE(ipfw_insn_u32))
 1898                                 goto bad_size;
 1899                 case O_IP_DST_LOOKUP:
 1900                         if (cmd->arg1 >= V_fw_tables_max) {
 1901                                 printf("ipfw: invalid table number %d\n",
 1902                                     cmd->arg1);
 1903                                 return (EINVAL);
 1904                         }
 1905                         if (cmdlen != F_INSN_SIZE(ipfw_insn) &&
 1906                             cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1 &&
 1907                             cmdlen != F_INSN_SIZE(ipfw_insn_u32))
 1908                                 goto bad_size;
 1909                         ci->object_opcodes++;
 1910                         break;
 1911                 case O_IP_FLOW_LOOKUP:
 1912                 case O_MAC_DST_LOOKUP:
 1913                 case O_MAC_SRC_LOOKUP:
 1914                         if (cmd->arg1 >= V_fw_tables_max) {
 1915                                 printf("ipfw: invalid table number %d\n",
 1916                                     cmd->arg1);
 1917                                 return (EINVAL);
 1918                         }
 1919                         if (cmdlen != F_INSN_SIZE(ipfw_insn) &&
 1920                             cmdlen != F_INSN_SIZE(ipfw_insn_u32))
 1921                                 goto bad_size;
 1922                         ci->object_opcodes++;
 1923                         break;
 1924                 case O_MACADDR2:
 1925                         if (cmdlen != F_INSN_SIZE(ipfw_insn_mac))
 1926                                 goto bad_size;
 1927                         break;
 1928 
 1929                 case O_NOP:
 1930                 case O_IPID:
 1931                 case O_IPTTL:
 1932                 case O_IPLEN:
 1933                 case O_TCPDATALEN:
 1934                 case O_TCPMSS:
 1935                 case O_TCPWIN:
 1936                 case O_TAGGED:
 1937                         if (cmdlen < 1 || cmdlen > 31)
 1938                                 goto bad_size;
 1939                         break;
 1940 
 1941                 case O_DSCP:
 1942                         if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1)
 1943                                 goto bad_size;
 1944                         break;
 1945 
 1946                 case O_MAC_TYPE:
 1947                 case O_IP_SRCPORT:
 1948                 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */
 1949                         if (cmdlen < 2 || cmdlen > 31)
 1950                                 goto bad_size;
 1951                         break;
 1952 
 1953                 case O_RECV:
 1954                 case O_XMIT:
 1955                 case O_VIA:
 1956                         if (cmdlen != F_INSN_SIZE(ipfw_insn_if))
 1957                                 goto bad_size;
 1958                         ci->object_opcodes++;
 1959                         break;
 1960 
 1961                 case O_ALTQ:
 1962                         if (cmdlen != F_INSN_SIZE(ipfw_insn_altq))
 1963                                 goto bad_size;
 1964                         break;
 1965 
 1966                 case O_PIPE:
 1967                 case O_QUEUE:
 1968                         if (cmdlen != F_INSN_SIZE(ipfw_insn))
 1969                                 goto bad_size;
 1970                         goto check_action;
 1971 
 1972                 case O_FORWARD_IP:
 1973                         if (cmdlen != F_INSN_SIZE(ipfw_insn_sa))
 1974                                 goto bad_size;
 1975                         goto check_action;
 1976 #ifdef INET6
 1977                 case O_FORWARD_IP6:
 1978                         if (cmdlen != F_INSN_SIZE(ipfw_insn_sa6))
 1979                                 goto bad_size;
 1980                         goto check_action;
 1981 #endif /* INET6 */
 1982 
 1983                 case O_DIVERT:
 1984                 case O_TEE:
 1985                         if (ip_divert_ptr == NULL)
 1986                                 return EINVAL;
 1987                         else
 1988                                 goto check_size;
 1989                 case O_NETGRAPH:
 1990                 case O_NGTEE:
 1991                         if (ng_ipfw_input_p == NULL)
 1992                                 return EINVAL;
 1993                         else
 1994                                 goto check_size;
 1995                 case O_NAT:
 1996                         if (!IPFW_NAT_LOADED)
 1997                                 return EINVAL;
 1998                         if (cmdlen != F_INSN_SIZE(ipfw_insn_nat))
 1999                                 goto bad_size;          
 2000                         goto check_action;
 2001                 case O_CHECK_STATE:
 2002                         ci->object_opcodes++;
 2003                         goto check_size;
 2004                 case O_REJECT:
 2005                         /* "unreach needfrag" has variable len. */
 2006                         if ((cmdlen == F_INSN_SIZE(ipfw_insn) ||
 2007                             cmdlen == F_INSN_SIZE(ipfw_insn_u16)))
 2008                                 goto check_action;
 2009                         /* FALLTHROUGH */
 2010                 case O_FORWARD_MAC: /* XXX not implemented yet */
 2011                 case O_COUNT:
 2012                 case O_ACCEPT:
 2013                 case O_DENY:
 2014                 case O_SETDSCP:
 2015 #ifdef INET6
 2016                 case O_UNREACH6:
 2017 #endif
 2018                 case O_SKIPTO:
 2019                 case O_REASS:
 2020                 case O_CALLRETURN:
 2021 check_size:
 2022                         if (cmdlen != F_INSN_SIZE(ipfw_insn))
 2023                                 goto bad_size;
 2024 check_action:
 2025                         if (have_action) {
 2026                                 printf("ipfw: opcode %d, multiple actions"
 2027                                         " not allowed\n",
 2028                                         cmd->opcode);
 2029                                 return (EINVAL);
 2030                         }
 2031                         have_action = 1;
 2032                         if (l != cmdlen) {
 2033                                 printf("ipfw: opcode %d, action must be"
 2034                                         " last opcode\n",
 2035                                         cmd->opcode);
 2036                                 return (EINVAL);
 2037                         }
 2038                         break;
 2039 #ifdef INET6
 2040                 case O_IP6_SRC:
 2041                 case O_IP6_DST:
 2042                         if (cmdlen != F_INSN_SIZE(struct in6_addr) +
 2043                             F_INSN_SIZE(ipfw_insn))
 2044                                 goto bad_size;
 2045                         break;
 2046 
 2047                 case O_FLOW6ID:
 2048                         if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
 2049                             ((ipfw_insn_u32 *)cmd)->o.arg1)
 2050                                 goto bad_size;
 2051                         break;
 2052 
 2053                 case O_IP6_SRC_MASK:
 2054                 case O_IP6_DST_MASK:
 2055                         if ( !(cmdlen & 1) || cmdlen > 127)
 2056                                 goto bad_size;
 2057                         break;
 2058                 case O_ICMP6TYPE:
 2059                         if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) )
 2060                                 goto bad_size;
 2061                         break;
 2062 #endif
 2063 
 2064                 default:
 2065                         switch (cmd->opcode) {
 2066 #ifndef INET6
 2067                         case O_IP6_SRC_ME:
 2068                         case O_IP6_DST_ME:
 2069                         case O_EXT_HDR:
 2070                         case O_IP6:
 2071                         case O_UNREACH6:
 2072                         case O_IP6_SRC:
 2073                         case O_IP6_DST:
 2074                         case O_FLOW6ID:
 2075                         case O_IP6_SRC_MASK:
 2076                         case O_IP6_DST_MASK:
 2077                         case O_ICMP6TYPE:
 2078                                 printf("ipfw: no IPv6 support in kernel\n");
 2079                                 return (EPROTONOSUPPORT);
 2080 #endif
 2081                         default:
 2082                                 printf("ipfw: opcode %d, unknown opcode\n",
 2083                                         cmd->opcode);
 2084                                 return (EINVAL);
 2085                         }
 2086                 }
 2087         }
 2088         if (have_action == 0) {
 2089                 printf("ipfw: missing action\n");
 2090                 return (EINVAL);
 2091         }
 2092         return 0;
 2093 
 2094 bad_size:
 2095         printf("ipfw: opcode %d size %d wrong\n",
 2096                 cmd->opcode, cmdlen);
 2097         return (EINVAL);
 2098 }
 2099 
 2100 /*
 2101  * Translation of requests for compatibility with FreeBSD 7.2/8.
 2102  * a static variable tells us if we have an old client from userland,
 2103  * and if necessary we translate requests and responses between the
 2104  * two formats.
 2105  */
 2106 static int is7 = 0;
 2107 
 2108 struct ip_fw7 {
 2109         struct ip_fw7   *next;          /* linked list of rules     */
 2110         struct ip_fw7   *next_rule;     /* ptr to next [skipto] rule    */
 2111         /* 'next_rule' is used to pass up 'set_disable' status      */
 2112 
 2113         uint16_t        act_ofs;        /* offset of action in 32-bit units */
 2114         uint16_t        cmd_len;        /* # of 32-bit words in cmd */
 2115         uint16_t        rulenum;        /* rule number          */
 2116         uint8_t         set;            /* rule set (0..31)     */
 2117         // #define RESVD_SET   31  /* set for default and persistent rules */
 2118         uint8_t         _pad;           /* padding          */
 2119         // uint32_t        id;             /* rule id, only in v.8 */
 2120         /* These fields are present in all rules.           */
 2121         uint64_t        pcnt;           /* Packet counter       */
 2122         uint64_t        bcnt;           /* Byte counter         */
 2123         uint32_t        timestamp;      /* tv_sec of last match     */
 2124 
 2125         ipfw_insn       cmd[1];         /* storage for commands     */
 2126 };
 2127 
 2128 static int convert_rule_to_7(struct ip_fw_rule0 *rule);
 2129 static int convert_rule_to_8(struct ip_fw_rule0 *rule);
 2130 
 2131 #ifndef RULESIZE7
 2132 #define RULESIZE7(rule)  (sizeof(struct ip_fw7) + \
 2133         ((struct ip_fw7 *)(rule))->cmd_len * 4 - 4)
 2134 #endif
 2135 
 2136 /*
 2137  * Copy the static and dynamic rules to the supplied buffer
 2138  * and return the amount of space actually used.
 2139  * Must be run under IPFW_UH_RLOCK
 2140  */
 2141 static size_t
 2142 ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space)
 2143 {
 2144         char *bp = buf;
 2145         char *ep = bp + space;
 2146         struct ip_fw *rule;
 2147         struct ip_fw_rule0 *dst;
 2148         struct timeval boottime;
 2149         int error, i, l, warnflag;
 2150         time_t  boot_seconds;
 2151 
 2152         warnflag = 0;
 2153 
 2154         getboottime(&boottime);
 2155         boot_seconds = boottime.tv_sec;
 2156         for (i = 0; i < chain->n_rules; i++) {
 2157                 rule = chain->map[i];
 2158 
 2159                 if (is7) {
 2160                     /* Convert rule to FreeBSd 7.2 format */
 2161                     l = RULESIZE7(rule);
 2162                     if (bp + l + sizeof(uint32_t) <= ep) {
 2163                         bcopy(rule, bp, l + sizeof(uint32_t));
 2164                         error = set_legacy_obj_kidx(chain,
 2165                             (struct ip_fw_rule0 *)bp);
 2166                         if (error != 0)
 2167                                 return (0);
 2168                         error = convert_rule_to_7((struct ip_fw_rule0 *) bp);
 2169                         if (error)
 2170                                 return 0; /*XXX correct? */
 2171                         /*
 2172                          * XXX HACK. Store the disable mask in the "next"
 2173                          * pointer in a wild attempt to keep the ABI the same.
 2174                          * Why do we do this on EVERY rule?
 2175                          */
 2176                         bcopy(&V_set_disable,
 2177                                 &(((struct ip_fw7 *)bp)->next_rule),
 2178                                 sizeof(V_set_disable));
 2179                         if (((struct ip_fw7 *)bp)->timestamp)
 2180                             ((struct ip_fw7 *)bp)->timestamp += boot_seconds;
 2181                         bp += l;
 2182                     }
 2183                     continue; /* go to next rule */
 2184                 }
 2185 
 2186                 l = RULEUSIZE0(rule);
 2187                 if (bp + l > ep) { /* should not happen */
 2188                         printf("overflow dumping static rules\n");
 2189                         break;
 2190                 }
 2191                 dst = (struct ip_fw_rule0 *)bp;
 2192                 export_rule0(rule, dst, l);
 2193                 error = set_legacy_obj_kidx(chain, dst);
 2194 
 2195                 /*
 2196                  * XXX HACK. Store the disable mask in the "next"
 2197                  * pointer in a wild attempt to keep the ABI the same.
 2198                  * Why do we do this on EVERY rule?
 2199                  *
 2200                  * XXX: "ipfw set show" (ab)uses IP_FW_GET to read disabled mask
 2201                  * so we need to fail _after_ saving at least one mask.
 2202                  */
 2203                 bcopy(&V_set_disable, &dst->next_rule, sizeof(V_set_disable));
 2204                 if (dst->timestamp)
 2205                         dst->timestamp += boot_seconds;
 2206                 bp += l;
 2207 
 2208                 if (error != 0) {
 2209                         if (error == 2) {
 2210                                 /* Non-fatal table rewrite error. */
 2211                                 warnflag = 1;
 2212                                 continue;
 2213                         }
 2214                         printf("Stop on rule %d. Fail to convert table\n",
 2215                             rule->rulenum);
 2216                         break;
 2217                 }
 2218         }
 2219         if (warnflag != 0)
 2220                 printf("ipfw: process %s is using legacy interfaces,"
 2221                     " consider rebuilding\n", "");
 2222         ipfw_get_dynamic(chain, &bp, ep); /* protected by the dynamic lock */
 2223         return (bp - (char *)buf);
 2224 }
 2225 
 2226 struct dump_args {
 2227         uint32_t        b;      /* start rule */
 2228         uint32_t        e;      /* end rule */
 2229         uint32_t        rcount; /* number of rules */
 2230         uint32_t        rsize;  /* rules size */
 2231         uint32_t        tcount; /* number of tables */
 2232         int             rcounters;      /* counters */
 2233         uint32_t        *bmask; /* index bitmask of used named objects */
 2234 };
 2235 
 2236 void
 2237 ipfw_export_obj_ntlv(struct named_object *no, ipfw_obj_ntlv *ntlv)
 2238 {
 2239 
 2240         ntlv->head.type = no->etlv;
 2241         ntlv->head.length = sizeof(*ntlv);
 2242         ntlv->idx = no->kidx;
 2243         strlcpy(ntlv->name, no->name, sizeof(ntlv->name));
 2244 }
 2245 
 2246 /*
 2247  * Export named object info in instance @ni, identified by @kidx
 2248  * to ipfw_obj_ntlv. TLV is allocated from @sd space.
 2249  *
 2250  * Returns 0 on success.
 2251  */
 2252 static int
 2253 export_objhash_ntlv(struct namedobj_instance *ni, uint16_t kidx,
 2254     struct sockopt_data *sd)
 2255 {
 2256         struct named_object *no;
 2257         ipfw_obj_ntlv *ntlv;
 2258 
 2259         no = ipfw_objhash_lookup_kidx(ni, kidx);
 2260         KASSERT(no != NULL, ("invalid object kernel index passed"));
 2261 
 2262         ntlv = (ipfw_obj_ntlv *)ipfw_get_sopt_space(sd, sizeof(*ntlv));
 2263         if (ntlv == NULL)
 2264                 return (ENOMEM);
 2265 
 2266         ipfw_export_obj_ntlv(no, ntlv);
 2267         return (0);
 2268 }
 2269 
 2270 static int
 2271 export_named_objects(struct namedobj_instance *ni, struct dump_args *da,
 2272     struct sockopt_data *sd)
 2273 {
 2274         int error, i;
 2275 
 2276         for (i = 0; i < IPFW_TABLES_MAX && da->tcount > 0; i++) {
 2277                 if ((da->bmask[i / 32] & (1 << (i % 32))) == 0)
 2278                         continue;
 2279                 if ((error = export_objhash_ntlv(ni, i, sd)) != 0)
 2280                         return (error);
 2281                 da->tcount--;
 2282         }
 2283         return (0);
 2284 }
 2285 
 2286 static int
 2287 dump_named_objects(struct ip_fw_chain *ch, struct dump_args *da,
 2288     struct sockopt_data *sd)
 2289 {
 2290         ipfw_obj_ctlv *ctlv;
 2291         int error;
 2292 
 2293         MPASS(da->tcount > 0);
 2294         /* Header first */
 2295         ctlv = (ipfw_obj_ctlv *)ipfw_get_sopt_space(sd, sizeof(*ctlv));
 2296         if (ctlv == NULL)
 2297                 return (ENOMEM);
 2298         ctlv->head.type = IPFW_TLV_TBLNAME_LIST;
 2299         ctlv->head.length = da->tcount * sizeof(ipfw_obj_ntlv) +
 2300             sizeof(*ctlv);
 2301         ctlv->count = da->tcount;
 2302         ctlv->objsize = sizeof(ipfw_obj_ntlv);
 2303 
 2304         /* Dump table names first (if any) */
 2305         error = export_named_objects(ipfw_get_table_objhash(ch), da, sd);
 2306         if (error != 0)
 2307                 return (error);
 2308         /* Then dump another named objects */
 2309         da->bmask += IPFW_TABLES_MAX / 32;
 2310         return (export_named_objects(CHAIN_TO_SRV(ch), da, sd));
 2311 }
 2312 
 2313 /*
 2314  * Dumps static rules with table TLVs in buffer @sd.
 2315  *
 2316  * Returns 0 on success.
 2317  */
 2318 static int
 2319 dump_static_rules(struct ip_fw_chain *chain, struct dump_args *da,
 2320     struct sockopt_data *sd)
 2321 {
 2322         ipfw_obj_ctlv *ctlv;
 2323         struct ip_fw *krule;
 2324         caddr_t dst;
 2325         int i, l;
 2326 
 2327         /* Dump rules */
 2328         ctlv = (ipfw_obj_ctlv *)ipfw_get_sopt_space(sd, sizeof(*ctlv));
 2329         if (ctlv == NULL)
 2330                 return (ENOMEM);
 2331         ctlv->head.type = IPFW_TLV_RULE_LIST;
 2332         ctlv->head.length = da->rsize + sizeof(*ctlv);
 2333         ctlv->count = da->rcount;
 2334 
 2335         for (i = da->b; i < da->e; i++) {
 2336                 krule = chain->map[i];
 2337 
 2338                 l = RULEUSIZE1(krule) + sizeof(ipfw_obj_tlv);
 2339                 if (da->rcounters != 0)
 2340                         l += sizeof(struct ip_fw_bcounter);
 2341                 dst = (caddr_t)ipfw_get_sopt_space(sd, l);
 2342                 if (dst == NULL)
 2343                         return (ENOMEM);
 2344 
 2345                 export_rule1(krule, dst, l, da->rcounters);
 2346         }
 2347 
 2348         return (0);
 2349 }
 2350 
 2351 int
 2352 ipfw_mark_object_kidx(uint32_t *bmask, uint16_t etlv, uint16_t kidx)
 2353 {
 2354         uint32_t bidx;
 2355 
 2356         /*
 2357          * Maintain separate bitmasks for table and non-table objects.
 2358          */
 2359         bidx = (etlv == IPFW_TLV_TBL_NAME) ? 0: IPFW_TABLES_MAX / 32;
 2360         bidx += kidx / 32;
 2361         if ((bmask[bidx] & (1 << (kidx % 32))) != 0)
 2362                 return (0);
 2363 
 2364         bmask[bidx] |= 1 << (kidx % 32);
 2365         return (1);
 2366 }
 2367 
 2368 /*
 2369  * Marks every object index used in @rule with bit in @bmask.
 2370  * Used to generate bitmask of referenced tables/objects for given ruleset
 2371  * or its part.
 2372  */
 2373 static void
 2374 mark_rule_objects(struct ip_fw_chain *ch, struct ip_fw *rule,
 2375     struct dump_args *da)
 2376 {
 2377         struct opcode_obj_rewrite *rw;
 2378         ipfw_insn *cmd;
 2379         int cmdlen, l;
 2380         uint16_t kidx;
 2381         uint8_t subtype;
 2382 
 2383         l = rule->cmd_len;
 2384         cmd = rule->cmd;
 2385         cmdlen = 0;
 2386         for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) {
 2387                 cmdlen = F_LEN(cmd);
 2388 
 2389                 rw = find_op_rw(cmd, &kidx, &subtype);
 2390                 if (rw == NULL)
 2391                         continue;
 2392 
 2393                 if (ipfw_mark_object_kidx(da->bmask, rw->etlv, kidx))
 2394                         da->tcount++;
 2395         }
 2396 }
 2397 
 2398 /*
 2399  * Dumps requested objects data
 2400  * Data layout (version 0)(current):
 2401  * Request: [ ipfw_cfg_lheader ] + IPFW_CFG_GET_* flags
 2402  *   size = ipfw_cfg_lheader.size
 2403  * Reply: [ ipfw_cfg_lheader 
 2404  *   [ ipfw_obj_ctlv(IPFW_TLV_TBL_LIST) ipfw_obj_ntlv x N ] (optional)
 2405  *   [ ipfw_obj_ctlv(IPFW_TLV_RULE_LIST)
 2406  *     ipfw_obj_tlv(IPFW_TLV_RULE_ENT) [ ip_fw_bcounter (optional) ip_fw_rule ]
 2407  *   ] (optional)
 2408  *   [ ipfw_obj_ctlv(IPFW_TLV_STATE_LIST) ipfw_obj_dyntlv x N ] (optional)
 2409  * ]
 2410  * * NOTE IPFW_TLV_STATE_LIST has the single valid field: objsize.
 2411  * The rest (size, count) are set to zero and needs to be ignored.
 2412  *
 2413  * Returns 0 on success.
 2414  */
 2415 static int
 2416 dump_config(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
 2417     struct sockopt_data *sd)
 2418 {
 2419         struct dump_args da;
 2420         ipfw_cfg_lheader *hdr;
 2421         struct ip_fw *rule;
 2422         size_t sz, rnum;
 2423         uint32_t hdr_flags, *bmask;
 2424         int error, i;
 2425 
 2426         hdr = (ipfw_cfg_lheader *)ipfw_get_sopt_header(sd, sizeof(*hdr));
 2427         if (hdr == NULL)
 2428                 return (EINVAL);
 2429 
 2430         error = 0;
 2431         bmask = NULL;
 2432         memset(&da, 0, sizeof(da));
 2433         /*
 2434          * Allocate needed state.
 2435          * Note we allocate 2xspace mask, for table & srv
 2436          */
 2437         if (hdr->flags & (IPFW_CFG_GET_STATIC | IPFW_CFG_GET_STATES))
 2438                 da.bmask = bmask = malloc(
 2439                     sizeof(uint32_t) * IPFW_TABLES_MAX * 2 / 32, M_TEMP,
 2440                     M_WAITOK | M_ZERO);
 2441         IPFW_UH_RLOCK(chain);
 2442 
 2443         /*
 2444          * STAGE 1: Determine size/count for objects in range.
 2445          * Prepare used tables bitmask.
 2446          */
 2447         sz = sizeof(ipfw_cfg_lheader);
 2448         da.e = chain->n_rules;
 2449 
 2450         if (hdr->end_rule != 0) {
 2451                 /* Handle custom range */
 2452                 if ((rnum = hdr->start_rule) > IPFW_DEFAULT_RULE)
 2453                         rnum = IPFW_DEFAULT_RULE;
 2454                 da.b = ipfw_find_rule(chain, rnum, 0);
 2455                 rnum = (hdr->end_rule < IPFW_DEFAULT_RULE) ?
 2456                     hdr->end_rule + 1: IPFW_DEFAULT_RULE;
 2457                 da.e = ipfw_find_rule(chain, rnum, UINT32_MAX) + 1;
 2458         }
 2459 
 2460         if (hdr->flags & IPFW_CFG_GET_STATIC) {
 2461                 for (i = da.b; i < da.e; i++) {
 2462                         rule = chain->map[i];
 2463                         da.rsize += RULEUSIZE1(rule) + sizeof(ipfw_obj_tlv);
 2464                         da.rcount++;
 2465                         /* Update bitmask of used objects for given range */
 2466                         mark_rule_objects(chain, rule, &da);
 2467                 }
 2468                 /* Add counters if requested */
 2469                 if (hdr->flags & IPFW_CFG_GET_COUNTERS) {
 2470                         da.rsize += sizeof(struct ip_fw_bcounter) * da.rcount;
 2471                         da.rcounters = 1;
 2472                 }
 2473                 sz += da.rsize + sizeof(ipfw_obj_ctlv);
 2474         }
 2475 
 2476         if (hdr->flags & IPFW_CFG_GET_STATES) {
 2477                 sz += sizeof(ipfw_obj_ctlv) +
 2478                     ipfw_dyn_get_count(bmask, &i) * sizeof(ipfw_obj_dyntlv);
 2479                 da.tcount += i;
 2480         }
 2481 
 2482         if (da.tcount > 0)
 2483                 sz += da.tcount * sizeof(ipfw_obj_ntlv) +
 2484                     sizeof(ipfw_obj_ctlv);
 2485 
 2486         /*
 2487          * Fill header anyway.
 2488          * Note we have to save header fields to stable storage
 2489          * buffer inside @sd can be flushed after dumping rules
 2490          */
 2491         hdr->size = sz;
 2492         hdr->set_mask = ~V_set_disable;
 2493         hdr_flags = hdr->flags;
 2494         hdr = NULL;
 2495 
 2496         if (sd->valsize < sz) {
 2497                 error = ENOMEM;
 2498                 goto cleanup;
 2499         }
 2500 
 2501         /* STAGE2: Store actual data */
 2502         if (da.tcount > 0) {
 2503                 error = dump_named_objects(chain, &da, sd);
 2504                 if (error != 0)
 2505                         goto cleanup;
 2506         }
 2507 
 2508         if (hdr_flags & IPFW_CFG_GET_STATIC) {
 2509                 error = dump_static_rules(chain, &da, sd);
 2510                 if (error != 0)
 2511                         goto cleanup;
 2512         }
 2513 
 2514         if (hdr_flags & IPFW_CFG_GET_STATES)
 2515                 error = ipfw_dump_states(chain, sd);
 2516 
 2517 cleanup:
 2518         IPFW_UH_RUNLOCK(chain);
 2519 
 2520         if (bmask != NULL)
 2521                 free(bmask, M_TEMP);
 2522 
 2523         return (error);
 2524 }
 2525 
 2526 int
 2527 ipfw_check_object_name_generic(const char *name)
 2528 {
 2529         int nsize;
 2530 
 2531         nsize = sizeof(((ipfw_obj_ntlv *)0)->name);
 2532         if (strnlen(name, nsize) == nsize)
 2533                 return (EINVAL);
 2534         if (name[0] == '\0')
 2535                 return (EINVAL);
 2536         return (0);
 2537 }
 2538 
 2539 /*
 2540  * Creates non-existent objects referenced by rule.
 2541  *
 2542  * Return 0 on success.
 2543  */
 2544 int
 2545 create_objects_compat(struct ip_fw_chain *ch, ipfw_insn *cmd,
 2546     struct obj_idx *oib, struct obj_idx *pidx, struct tid_info *ti)
 2547 {
 2548         struct opcode_obj_rewrite *rw;
 2549         struct obj_idx *p;
 2550         uint16_t kidx;
 2551         int error;
 2552 
 2553         /*
 2554          * Compatibility stuff: do actual creation for non-existing,
 2555          * but referenced objects.
 2556          */
 2557         for (p = oib; p < pidx; p++) {
 2558                 if (p->kidx != 0)
 2559                         continue;
 2560 
 2561                 ti->uidx = p->uidx;
 2562                 ti->type = p->type;
 2563                 ti->atype = 0;
 2564 
 2565                 rw = find_op_rw(cmd + p->off, NULL, NULL);
 2566                 KASSERT(rw != NULL, ("Unable to find handler for op %d",
 2567                     (cmd + p->off)->opcode));
 2568 
 2569                 if (rw->create_object == NULL)
 2570                         error = EOPNOTSUPP;
 2571                 else
 2572                         error = rw->create_object(ch, ti, &kidx);
 2573                 if (error == 0) {
 2574                         p->kidx = kidx;
 2575                         continue;
 2576                 }
 2577 
 2578                 /*
 2579                  * Error happened. We have to rollback everything.
 2580                  * Drop all already acquired references.
 2581                  */
 2582                 IPFW_UH_WLOCK(ch);
 2583                 unref_oib_objects(ch, cmd, oib, pidx);
 2584                 IPFW_UH_WUNLOCK(ch);
 2585 
 2586                 return (error);
 2587         }
 2588 
 2589         return (0);
 2590 }
 2591 
 2592 /*
 2593  * Compatibility function for old ipfw(8) binaries.
 2594  * Rewrites table/nat kernel indices with userland ones.
 2595  * Convert tables matching '/^\d+$/' to their atoi() value.
 2596  * Use number 65535 for other tables.
 2597  *
 2598  * Returns 0 on success.
 2599  */
 2600 static int
 2601 set_legacy_obj_kidx(struct ip_fw_chain *ch, struct ip_fw_rule0 *rule)
 2602 {
 2603         struct opcode_obj_rewrite *rw;
 2604         struct named_object *no;
 2605         ipfw_insn *cmd;
 2606         char *end;
 2607         long val;
 2608         int cmdlen, error, l;
 2609         uint16_t kidx, uidx;
 2610         uint8_t subtype;
 2611 
 2612         error = 0;
 2613 
 2614         l = rule->cmd_len;
 2615         cmd = rule->cmd;
 2616         cmdlen = 0;
 2617         for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) {
 2618                 cmdlen = F_LEN(cmd);
 2619 
 2620                 /* Check if is index in given opcode */
 2621                 rw = find_op_rw(cmd, &kidx, &subtype);
 2622                 if (rw == NULL)
 2623                         continue;
 2624 
 2625                 /* Try to find referenced kernel object */
 2626                 no = rw->find_bykidx(ch, kidx);
 2627                 if (no == NULL)
 2628                         continue;
 2629 
 2630                 val = strtol(no->name, &end, 10);
 2631                 if (*end == '\0' && val < 65535) {
 2632                         uidx = val;
 2633                 } else {
 2634                         /*
 2635                          * We are called via legacy opcode.
 2636                          * Save error and show table as fake number
 2637                          * not to make ipfw(8) hang.
 2638                          */
 2639                         uidx = 65535;
 2640                         error = 2;
 2641                 }
 2642 
 2643                 rw->update(cmd, uidx);
 2644         }
 2645 
 2646         return (error);
 2647 }
 2648 
 2649 /*
 2650  * Unreferences all already-referenced objects in given @cmd rule,
 2651  * using information in @oib.
 2652  *
 2653  * Used to rollback partially converted rule on error.
 2654  */
 2655 static void
 2656 unref_oib_objects(struct ip_fw_chain *ch, ipfw_insn *cmd, struct obj_idx *oib,
 2657     struct obj_idx *end)
 2658 {
 2659         struct opcode_obj_rewrite *rw;
 2660         struct named_object *no;
 2661         struct obj_idx *p;
 2662 
 2663         IPFW_UH_WLOCK_ASSERT(ch);
 2664 
 2665         for (p = oib; p < end; p++) {
 2666                 if (p->kidx == 0)
 2667                         continue;
 2668 
 2669                 rw = find_op_rw(cmd + p->off, NULL, NULL);
 2670                 KASSERT(rw != NULL, ("Unable to find handler for op %d",
 2671                     (cmd + p->off)->opcode));
 2672 
 2673                 /* Find & unref by existing idx */
 2674                 no = rw->find_bykidx(ch, p->kidx);
 2675                 KASSERT(no != NULL, ("Ref'd object %d disappeared", p->kidx));
 2676                 no->refcnt--;
 2677         }
 2678 }
 2679 
 2680 /*
 2681  * Remove references from every object used in @rule.
 2682  * Used at rule removal code.
 2683  */
 2684 static void
 2685 unref_rule_objects(struct ip_fw_chain *ch, struct ip_fw *rule)
 2686 {
 2687         struct opcode_obj_rewrite *rw;
 2688         struct named_object *no;
 2689         ipfw_insn *cmd;
 2690         int cmdlen, l;
 2691         uint16_t kidx;
 2692         uint8_t subtype;
 2693 
 2694         IPFW_UH_WLOCK_ASSERT(ch);
 2695 
 2696         l = rule->cmd_len;
 2697         cmd = rule->cmd;
 2698         cmdlen = 0;
 2699         for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) {
 2700                 cmdlen = F_LEN(cmd);
 2701 
 2702                 rw = find_op_rw(cmd, &kidx, &subtype);
 2703                 if (rw == NULL)
 2704                         continue;
 2705                 no = rw->find_bykidx(ch, kidx);
 2706 
 2707                 KASSERT(no != NULL, ("object id %d not found", kidx));
 2708                 KASSERT(no->subtype == subtype,
 2709                     ("wrong type %d (%d) for object id %d",
 2710                     no->subtype, subtype, kidx));
 2711                 KASSERT(no->refcnt > 0, ("refcount for object %d is %d",
 2712                     kidx, no->refcnt));
 2713 
 2714                 if (no->refcnt == 1 && rw->destroy_object != NULL)
 2715                         rw->destroy_object(ch, no);
 2716                 else
 2717                         no->refcnt--;
 2718         }
 2719 }
 2720 
 2721 /*
 2722  * Find and reference object (if any) stored in instruction @cmd.
 2723  *
 2724  * Saves object info in @pidx, sets
 2725  *  - @unresolved to 1 if object should exists but not found
 2726  *
 2727  * Returns non-zero value in case of error.
 2728  */
 2729 static int
 2730 ref_opcode_object(struct ip_fw_chain *ch, ipfw_insn *cmd, struct tid_info *ti,
 2731     struct obj_idx *pidx, int *unresolved)
 2732 {
 2733         struct named_object *no;
 2734         struct opcode_obj_rewrite *rw;
 2735         int error;
 2736 
 2737         /* Check if this opcode is candidate for rewrite */
 2738         rw = find_op_rw(cmd, &ti->uidx, &ti->type);
 2739         if (rw == NULL)
 2740                 return (0);
 2741 
 2742         /* Need to rewrite. Save necessary fields */
 2743         pidx->uidx = ti->uidx;
 2744         pidx->type = ti->type;
 2745 
 2746         /* Try to find referenced kernel object */
 2747         error = rw->find_byname(ch, ti, &no);
 2748         if (error != 0)
 2749                 return (error);
 2750         if (no == NULL) {
 2751                 /*
 2752                  * Report about unresolved object for automaic
 2753                  * creation.
 2754                  */
 2755                 *unresolved = 1;
 2756                 return (0);
 2757         }
 2758 
 2759         /*
 2760          * Object is already exist.
 2761          * Its subtype should match with expected value.
 2762          */
 2763         if (ti->type != no->subtype)
 2764                 return (EINVAL);
 2765 
 2766         /* Bump refcount and update kidx. */
 2767         no->refcnt++;
 2768         rw->update(cmd, no->kidx);
 2769         return (0);
 2770 }
 2771 
 2772 /*
 2773  * Finds and bumps refcount for objects referenced by given @rule.
 2774  * Auto-creates non-existing tables.
 2775  * Fills in @oib array with userland/kernel indexes.
 2776  *
 2777  * Returns 0 on success.
 2778  */
 2779 static int
 2780 ref_rule_objects(struct ip_fw_chain *ch, struct ip_fw *rule,
 2781     struct rule_check_info *ci, struct obj_idx *oib, struct tid_info *ti)
 2782 {
 2783         struct obj_idx *pidx;
 2784         ipfw_insn *cmd;
 2785         int cmdlen, error, l, unresolved;
 2786 
 2787         pidx = oib;
 2788         l = rule->cmd_len;
 2789         cmd = rule->cmd;
 2790         cmdlen = 0;
 2791         error = 0;
 2792 
 2793         IPFW_UH_WLOCK(ch);
 2794 
 2795         /* Increase refcount on each existing referenced table. */
 2796         for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) {
 2797                 cmdlen = F_LEN(cmd);
 2798                 unresolved = 0;
 2799 
 2800                 error = ref_opcode_object(ch, cmd, ti, pidx, &unresolved);
 2801                 if (error != 0)
 2802                         break;
 2803                 /*
 2804                  * Compatibility stuff for old clients:
 2805                  * prepare to automaitcally create non-existing objects.
 2806                  */
 2807                 if (unresolved != 0) {
 2808                         pidx->off = rule->cmd_len - l;
 2809                         pidx++;
 2810                 }
 2811         }
 2812 
 2813         if (error != 0) {
 2814                 /* Unref everything we have already done */
 2815                 unref_oib_objects(ch, rule->cmd, oib, pidx);
 2816                 IPFW_UH_WUNLOCK(ch);
 2817                 return (error);
 2818         }
 2819         IPFW_UH_WUNLOCK(ch);
 2820 
 2821         /* Perform auto-creation for non-existing objects */
 2822         if (pidx != oib)
 2823                 error = create_objects_compat(ch, rule->cmd, oib, pidx, ti);
 2824 
 2825         /* Calculate real number of dynamic objects */
 2826         ci->object_opcodes = (uint16_t)(pidx - oib);
 2827 
 2828         return (error);
 2829 }
 2830 
 2831 /*
 2832  * Checks is opcode is referencing table of appropriate type.
 2833  * Adds reference count for found table if true.
 2834  * Rewrites user-supplied opcode values with kernel ones.
 2835  *
 2836  * Returns 0 on success and appropriate error code otherwise.
 2837  */
 2838 static int
 2839 rewrite_rule_uidx(struct ip_fw_chain *chain, struct rule_check_info *ci)
 2840 {
 2841         int error;
 2842         ipfw_insn *cmd;
 2843         struct obj_idx *p, *pidx_first, *pidx_last;
 2844         struct tid_info ti;
 2845 
 2846         /*
 2847          * Prepare an array for storing opcode indices.
 2848          * Use stack allocation by default.
 2849          */
 2850         if (ci->object_opcodes <= (sizeof(ci->obuf)/sizeof(ci->obuf[0]))) {
 2851                 /* Stack */
 2852                 pidx_first = ci->obuf;
 2853         } else
 2854                 pidx_first = malloc(
 2855                     ci->object_opcodes * sizeof(struct obj_idx),
 2856                     M_IPFW, M_WAITOK | M_ZERO);
 2857 
 2858         error = 0;
 2859         memset(&ti, 0, sizeof(ti));
 2860 
 2861         /* Use set rule is assigned to. */
 2862         ti.set = ci->krule->set;
 2863         if (ci->ctlv != NULL) {
 2864                 ti.tlvs = (void *)(ci->ctlv + 1);
 2865                 ti.tlen = ci->ctlv->head.length - sizeof(ipfw_obj_ctlv);
 2866         }
 2867 
 2868         /* Reference all used tables and other objects */
 2869         error = ref_rule_objects(chain, ci->krule, ci, pidx_first, &ti);
 2870         if (error != 0)
 2871                 goto free;
 2872         /*
 2873          * Note that ref_rule_objects() might have updated ci->object_opcodes
 2874          * to reflect actual number of object opcodes.
 2875          */
 2876 
 2877         /* Perform rewrite of remaining opcodes */
 2878         p = pidx_first;
 2879         pidx_last = pidx_first + ci->object_opcodes;
 2880         for (p = pidx_first; p < pidx_last; p++) {
 2881                 cmd = ci->krule->cmd + p->off;
 2882                 update_opcode_kidx(cmd, p->kidx);
 2883         }
 2884 
 2885 free:
 2886         if (pidx_first != ci->obuf)
 2887                 free(pidx_first, M_IPFW);
 2888 
 2889         return (error);
 2890 }
 2891 
 2892 /*
 2893  * Adds one or more rules to ipfw @chain.
 2894  * Data layout (version 0)(current):
 2895  * Request:
 2896  * [
 2897  *   ip_fw3_opheader
 2898  *   [ ipfw_obj_ctlv(IPFW_TLV_TBL_LIST) ipfw_obj_ntlv x N ] (optional *1)
 2899  *   [ ipfw_obj_ctlv(IPFW_TLV_RULE_LIST) ip_fw x N ] (*2) (*3)
 2900  * ]
 2901  * Reply:
 2902  * [
 2903  *   ip_fw3_opheader
 2904  *   [ ipfw_obj_ctlv(IPFW_TLV_TBL_LIST) ipfw_obj_ntlv x N ] (optional)
 2905  *   [ ipfw_obj_ctlv(IPFW_TLV_RULE_LIST) ip_fw x N ]
 2906  * ]
 2907  *
 2908  * Rules in reply are modified to store their actual ruleset number.
 2909  *
 2910  * (*1) TLVs inside IPFW_TLV_TBL_LIST needs to be sorted ascending
 2911  * according to their idx field and there has to be no duplicates.
 2912  * (*2) Numbered rules inside IPFW_TLV_RULE_LIST needs to be sorted ascending.
 2913  * (*3) Each ip_fw structure needs to be aligned to u64 boundary.
 2914  *
 2915  * Returns 0 on success.
 2916  */
 2917 static int
 2918 add_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
 2919     struct sockopt_data *sd)
 2920 {
 2921         ipfw_obj_ctlv *ctlv, *rtlv, *tstate;
 2922         ipfw_obj_ntlv *ntlv;
 2923         int clen, error, idx;
 2924         uint32_t count, read;
 2925         struct ip_fw_rule *r;
 2926         struct rule_check_info rci, *ci, *cbuf;
 2927         int i, rsize;
 2928 
 2929         op3 = (ip_fw3_opheader *)ipfw_get_sopt_space(sd, sd->valsize);
 2930         ctlv = (ipfw_obj_ctlv *)(op3 + 1);
 2931 
 2932         read = sizeof(ip_fw3_opheader);
 2933         rtlv = NULL;
 2934         tstate = NULL;
 2935         cbuf = NULL;
 2936         memset(&rci, 0, sizeof(struct rule_check_info));
 2937 
 2938         if (read + sizeof(*ctlv) > sd->valsize)
 2939                 return (EINVAL);
 2940 
 2941         if (ctlv->head.type == IPFW_TLV_TBLNAME_LIST) {
 2942                 clen = ctlv->head.length;
 2943                 /* Check size and alignment */
 2944                 if (clen > sd->valsize || clen < sizeof(*ctlv))
 2945                         return (EINVAL);
 2946                 if ((clen % sizeof(uint64_t)) != 0)
 2947                         return (EINVAL);
 2948 
 2949                 /*
 2950                  * Some table names or other named objects.
 2951                  * Check for validness.
 2952                  */
 2953                 count = (ctlv->head.length - sizeof(*ctlv)) / sizeof(*ntlv);
 2954                 if (ctlv->count != count || ctlv->objsize != sizeof(*ntlv))
 2955                         return (EINVAL);
 2956 
 2957                 /*
 2958                  * Check each TLV.
 2959                  * Ensure TLVs are sorted ascending and
 2960                  * there are no duplicates.
 2961                  */
 2962                 idx = -1;
 2963                 ntlv = (ipfw_obj_ntlv *)(ctlv + 1);
 2964                 while (count > 0) {
 2965                         if (ntlv->head.length != sizeof(ipfw_obj_ntlv))
 2966                                 return (EINVAL);
 2967 
 2968                         error = ipfw_check_object_name_generic(ntlv->name);
 2969                         if (error != 0)
 2970                                 return (error);
 2971 
 2972                         if (ntlv->idx <= idx)
 2973                                 return (EINVAL);
 2974 
 2975                         idx = ntlv->idx;
 2976                         count--;
 2977                         ntlv++;
 2978                 }
 2979 
 2980                 tstate = ctlv;
 2981                 read += ctlv->head.length;
 2982                 ctlv = (ipfw_obj_ctlv *)((caddr_t)ctlv + ctlv->head.length);
 2983         }
 2984 
 2985         if (read + sizeof(*ctlv) > sd->valsize)
 2986                 return (EINVAL);
 2987 
 2988         if (ctlv->head.type == IPFW_TLV_RULE_LIST) {
 2989                 clen = ctlv->head.length;
 2990                 if (clen + read > sd->valsize || clen < sizeof(*ctlv))
 2991                         return (EINVAL);
 2992                 if ((clen % sizeof(uint64_t)) != 0)
 2993                         return (EINVAL);
 2994 
 2995                 /*
 2996                  * TODO: Permit adding multiple rules at once
 2997                  */
 2998                 if (ctlv->count != 1)
 2999                         return (ENOTSUP);
 3000 
 3001                 clen -= sizeof(*ctlv);
 3002 
 3003                 if (ctlv->count > clen / sizeof(struct ip_fw_rule))
 3004                         return (EINVAL);
 3005 
 3006                 /* Allocate state for each rule or use stack */
 3007                 if (ctlv->count == 1) {
 3008                         memset(&rci, 0, sizeof(struct rule_check_info));
 3009                         cbuf = &rci;
 3010                 } else
 3011                         cbuf = malloc(ctlv->count * sizeof(*ci), M_TEMP,
 3012                             M_WAITOK | M_ZERO);
 3013                 ci = cbuf;
 3014 
 3015                 /*
 3016                  * Check each rule for validness.
 3017                  * Ensure numbered rules are sorted ascending
 3018                  * and properly aligned
 3019                  */
 3020                 idx = 0;
 3021                 r = (struct ip_fw_rule *)(ctlv + 1);
 3022                 count = 0;
 3023                 error = 0;
 3024                 while (clen > 0) {
 3025                         rsize = roundup2(RULESIZE(r), sizeof(uint64_t));
 3026                         if (rsize > clen || ctlv->count <= count) {
 3027                                 error = EINVAL;
 3028                                 break;
 3029                         }
 3030 
 3031                         ci->ctlv = tstate;
 3032                         error = check_ipfw_rule1(r, rsize, ci);
 3033                         if (error != 0)
 3034                                 break;
 3035 
 3036                         /* Check sorting */
 3037                         if (r->rulenum != 0 && r->rulenum < idx) {
 3038                                 printf("rulenum %d idx %d\n", r->rulenum, idx);
 3039                                 error = EINVAL;
 3040                                 break;
 3041                         }
 3042                         idx = r->rulenum;
 3043 
 3044                         ci->urule = (caddr_t)r;
 3045 
 3046                         rsize = roundup2(rsize, sizeof(uint64_t));
 3047                         clen -= rsize;
 3048                         r = (struct ip_fw_rule *)((caddr_t)r + rsize);
 3049                         count++;
 3050                         ci++;
 3051                 }
 3052 
 3053                 if (ctlv->count != count || error != 0) {
 3054                         if (cbuf != &rci)
 3055                                 free(cbuf, M_TEMP);
 3056                         return (EINVAL);
 3057                 }
 3058 
 3059                 rtlv = ctlv;
 3060                 read += ctlv->head.length;
 3061                 ctlv = (ipfw_obj_ctlv *)((caddr_t)ctlv + ctlv->head.length);
 3062         }
 3063 
 3064         if (read != sd->valsize || rtlv == NULL || rtlv->count == 0) {
 3065                 if (cbuf != NULL && cbuf != &rci)
 3066                         free(cbuf, M_TEMP);
 3067                 return (EINVAL);
 3068         }
 3069 
 3070         /*
 3071          * Passed rules seems to be valid.
 3072          * Allocate storage and try to add them to chain.
 3073          */
 3074         for (i = 0, ci = cbuf; i < rtlv->count; i++, ci++) {
 3075                 clen = RULEKSIZE1((struct ip_fw_rule *)ci->urule);
 3076                 ci->krule = ipfw_alloc_rule(chain, clen);
 3077                 import_rule1(ci);
 3078         }
 3079 
 3080         if ((error = commit_rules(chain, cbuf, rtlv->count)) != 0) {
 3081                 /* Free allocate krules */
 3082                 for (i = 0, ci = cbuf; i < rtlv->count; i++, ci++)
 3083                         ipfw_free_rule(ci->krule);
 3084         }
 3085 
 3086         if (cbuf != NULL && cbuf != &rci)
 3087                 free(cbuf, M_TEMP);
 3088 
 3089         return (error);
 3090 }
 3091 
 3092 /*
 3093  * Lists all sopts currently registered.
 3094  * Data layout (v0)(current):
 3095  * Request: [ ipfw_obj_lheader ], size = ipfw_obj_lheader.size
 3096  * Reply: [ ipfw_obj_lheader ipfw_sopt_info x N ]
 3097  *
 3098  * Returns 0 on success
 3099  */
 3100 static int
 3101 dump_soptcodes(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
 3102     struct sockopt_data *sd)
 3103 {
 3104         struct _ipfw_obj_lheader *olh;
 3105         ipfw_sopt_info *i;
 3106         struct ipfw_sopt_handler *sh;
 3107         uint32_t count, n, size;
 3108 
 3109         olh = (struct _ipfw_obj_lheader *)ipfw_get_sopt_header(sd,sizeof(*olh));
 3110         if (olh == NULL)
 3111                 return (EINVAL);
 3112         if (sd->valsize < olh->size)
 3113                 return (EINVAL);
 3114 
 3115         CTL3_LOCK();
 3116         count = ctl3_hsize;
 3117         size = count * sizeof(ipfw_sopt_info) + sizeof(ipfw_obj_lheader);
 3118 
 3119         /* Fill in header regadless of buffer size */
 3120         olh->count = count;
 3121         olh->objsize = sizeof(ipfw_sopt_info);
 3122 
 3123         if (size > olh->size) {
 3124                 olh->size = size;
 3125                 CTL3_UNLOCK();
 3126                 return (ENOMEM);
 3127         }
 3128         olh->size = size;
 3129 
 3130         for (n = 1; n <= count; n++) {
 3131                 i = (ipfw_sopt_info *)ipfw_get_sopt_space(sd, sizeof(*i));
 3132                 KASSERT(i != NULL, ("previously checked buffer is not enough"));
 3133                 sh = &ctl3_handlers[n];
 3134                 i->opcode = sh->opcode;
 3135                 i->version = sh->version;
 3136                 i->refcnt = sh->refcnt;
 3137         }
 3138         CTL3_UNLOCK();
 3139 
 3140         return (0);
 3141 }
 3142 
 3143 /*
 3144  * Compares two opcodes.
 3145  * Used both in qsort() and bsearch().
 3146  *
 3147  * Returns 0 if match is found.
 3148  */
 3149 static int
 3150 compare_opcodes(const void *_a, const void *_b)
 3151 {
 3152         const struct opcode_obj_rewrite *a, *b;
 3153 
 3154         a = (const struct opcode_obj_rewrite *)_a;
 3155         b = (const struct opcode_obj_rewrite *)_b;
 3156 
 3157         if (a->opcode < b->opcode)
 3158                 return (-1);
 3159         else if (a->opcode > b->opcode)
 3160                 return (1);
 3161 
 3162         return (0);
 3163 }
 3164 
 3165 /*
 3166  * XXX: Rewrite bsearch()
 3167  */
 3168 static int
 3169 find_op_rw_range(uint16_t op, struct opcode_obj_rewrite **plo,
 3170     struct opcode_obj_rewrite **phi)
 3171 {
 3172         struct opcode_obj_rewrite *ctl3_max, *lo, *hi, h, *rw;
 3173 
 3174         memset(&h, 0, sizeof(h));
 3175         h.opcode = op;
 3176 
 3177         rw = (struct opcode_obj_rewrite *)bsearch(&h, ctl3_rewriters,
 3178             ctl3_rsize, sizeof(h), compare_opcodes);
 3179         if (rw == NULL)
 3180                 return (1);
 3181 
 3182         /* Find the first element matching the same opcode */
 3183         lo = rw;
 3184         for ( ; lo > ctl3_rewriters && (lo - 1)->opcode == op; lo--)
 3185                 ;
 3186 
 3187         /* Find the last element matching the same opcode */
 3188         hi = rw;
 3189         ctl3_max = ctl3_rewriters + ctl3_rsize;
 3190         for ( ; (hi + 1) < ctl3_max && (hi + 1)->opcode == op; hi++)
 3191                 ;
 3192 
 3193         *plo = lo;
 3194         *phi = hi;
 3195 
 3196         return (0);
 3197 }
 3198 
 3199 /*
 3200  * Finds opcode object rewriter based on @code.
 3201  *
 3202  * Returns pointer to handler or NULL.
 3203  */
 3204 static struct opcode_obj_rewrite *
 3205 find_op_rw(ipfw_insn *cmd, uint16_t *puidx, uint8_t *ptype)
 3206 {
 3207         struct opcode_obj_rewrite *rw, *lo, *hi;
 3208         uint16_t uidx;
 3209         uint8_t subtype;
 3210 
 3211         if (find_op_rw_range(cmd->opcode, &lo, &hi) != 0)
 3212                 return (NULL);
 3213 
 3214         for (rw = lo; rw <= hi; rw++) {
 3215                 if (rw->classifier(cmd, &uidx, &subtype) == 0) {
 3216                         if (puidx != NULL)
 3217                                 *puidx = uidx;
 3218                         if (ptype != NULL)
 3219                                 *ptype = subtype;
 3220                         return (rw);
 3221                 }
 3222         }
 3223 
 3224         return (NULL);
 3225 }
 3226 int
 3227 classify_opcode_kidx(ipfw_insn *cmd, uint16_t *puidx)
 3228 {
 3229 
 3230         if (find_op_rw(cmd, puidx, NULL) == NULL)
 3231                 return (1);
 3232         return (0);
 3233 }
 3234 
 3235 void
 3236 update_opcode_kidx(ipfw_insn *cmd, uint16_t idx)
 3237 {
 3238         struct opcode_obj_rewrite *rw;
 3239 
 3240         rw = find_op_rw(cmd, NULL, NULL);
 3241         KASSERT(rw != NULL, ("No handler to update opcode %d", cmd->opcode));
 3242         rw->update(cmd, idx);
 3243 }
 3244 
 3245 void
 3246 ipfw_init_obj_rewriter(void)
 3247 {
 3248 
 3249         ctl3_rewriters = NULL;
 3250         ctl3_rsize = 0;
 3251 }
 3252 
 3253 void
 3254 ipfw_destroy_obj_rewriter(void)
 3255 {
 3256 
 3257         if (ctl3_rewriters != NULL)
 3258                 free(ctl3_rewriters, M_IPFW);
 3259         ctl3_rewriters = NULL;
 3260         ctl3_rsize = 0;
 3261 }
 3262 
 3263 /*
 3264  * Adds one or more opcode object rewrite handlers to the global array.
 3265  * Function may sleep.
 3266  */
 3267 void
 3268 ipfw_add_obj_rewriter(struct opcode_obj_rewrite *rw, size_t count)
 3269 {
 3270         size_t sz;
 3271         struct opcode_obj_rewrite *tmp;
 3272 
 3273         CTL3_LOCK();
 3274 
 3275         for (;;) {
 3276                 sz = ctl3_rsize + count;
 3277                 CTL3_UNLOCK();
 3278                 tmp = malloc(sizeof(*rw) * sz, M_IPFW, M_WAITOK | M_ZERO);
 3279                 CTL3_LOCK();
 3280                 if (ctl3_rsize + count <= sz)
 3281                         break;
 3282 
 3283                 /* Retry */
 3284                 free(tmp, M_IPFW);
 3285         }
 3286 
 3287         /* Merge old & new arrays */
 3288         sz = ctl3_rsize + count;
 3289         memcpy(tmp, ctl3_rewriters, ctl3_rsize * sizeof(*rw));
 3290         memcpy(&tmp[ctl3_rsize], rw, count * sizeof(*rw));
 3291         qsort(tmp, sz, sizeof(*rw), compare_opcodes);
 3292         /* Switch new and free old */
 3293         if (ctl3_rewriters != NULL)
 3294                 free(ctl3_rewriters, M_IPFW);
 3295         ctl3_rewriters = tmp;
 3296         ctl3_rsize = sz;
 3297 
 3298         CTL3_UNLOCK();
 3299 }
 3300 
 3301 /*
 3302  * Removes one or more object rewrite handlers from the global array.
 3303  */
 3304 int
 3305 ipfw_del_obj_rewriter(struct opcode_obj_rewrite *rw, size_t count)
 3306 {
 3307         size_t sz;
 3308         struct opcode_obj_rewrite *ctl3_max, *ktmp, *lo, *hi;
 3309         int i;
 3310 
 3311         CTL3_LOCK();
 3312 
 3313         for (i = 0; i < count; i++) {
 3314                 if (find_op_rw_range(rw[i].opcode, &lo, &hi) != 0)
 3315                         continue;
 3316 
 3317                 for (ktmp = lo; ktmp <= hi; ktmp++) {
 3318                         if (ktmp->classifier != rw[i].classifier)
 3319                                 continue;
 3320 
 3321                         ctl3_max = ctl3_rewriters + ctl3_rsize;
 3322                         sz = (ctl3_max - (ktmp + 1)) * sizeof(*ktmp);
 3323                         memmove(ktmp, ktmp + 1, sz);
 3324                         ctl3_rsize--;
 3325                         break;
 3326                 }
 3327         }
 3328 
 3329         if (ctl3_rsize == 0) {
 3330                 if (ctl3_rewriters != NULL)
 3331                         free(ctl3_rewriters, M_IPFW);
 3332                 ctl3_rewriters = NULL;
 3333         }
 3334 
 3335         CTL3_UNLOCK();
 3336 
 3337         return (0);
 3338 }
 3339 
 3340 static int
 3341 export_objhash_ntlv_internal(struct namedobj_instance *ni,
 3342     struct named_object *no, void *arg)
 3343 {
 3344         struct sockopt_data *sd;
 3345         ipfw_obj_ntlv *ntlv;
 3346 
 3347         sd = (struct sockopt_data *)arg;
 3348         ntlv = (ipfw_obj_ntlv *)ipfw_get_sopt_space(sd, sizeof(*ntlv));
 3349         if (ntlv == NULL)
 3350                 return (ENOMEM);
 3351         ipfw_export_obj_ntlv(no, ntlv);
 3352         return (0);
 3353 }
 3354 
 3355 /*
 3356  * Lists all service objects.
 3357  * Data layout (v0)(current):
 3358  * Request: [ ipfw_obj_lheader ] size = ipfw_obj_lheader.size
 3359  * Reply: [ ipfw_obj_lheader [ ipfw_obj_ntlv x N ] (optional) ]
 3360  * Returns 0 on success
 3361  */
 3362 static int
 3363 dump_srvobjects(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
 3364     struct sockopt_data *sd)
 3365 {
 3366         ipfw_obj_lheader *hdr;
 3367         int count;
 3368 
 3369         hdr = (ipfw_obj_lheader *)ipfw_get_sopt_header(sd, sizeof(*hdr));
 3370         if (hdr == NULL)
 3371                 return (EINVAL);
 3372 
 3373         IPFW_UH_RLOCK(chain);
 3374         count = ipfw_objhash_count(CHAIN_TO_SRV(chain));
 3375         hdr->size = sizeof(ipfw_obj_lheader) + count * sizeof(ipfw_obj_ntlv);
 3376         if (sd->valsize < hdr->size) {
 3377                 IPFW_UH_RUNLOCK(chain);
 3378                 return (ENOMEM);
 3379         }
 3380         hdr->count = count;
 3381         hdr->objsize = sizeof(ipfw_obj_ntlv);
 3382         if (count > 0)
 3383                 ipfw_objhash_foreach(CHAIN_TO_SRV(chain),
 3384                     export_objhash_ntlv_internal, sd);
 3385         IPFW_UH_RUNLOCK(chain);
 3386         return (0);
 3387 }
 3388 
 3389 /*
 3390  * Compares two sopt handlers (code, version and handler ptr).
 3391  * Used both as qsort() and bsearch().
 3392  * Does not compare handler for latter case.
 3393  *
 3394  * Returns 0 if match is found.
 3395  */
 3396 static int
 3397 compare_sh(const void *_a, const void *_b)
 3398 {
 3399         const struct ipfw_sopt_handler *a, *b;
 3400 
 3401         a = (const struct ipfw_sopt_handler *)_a;
 3402         b = (const struct ipfw_sopt_handler *)_b;
 3403 
 3404         if (a->opcode < b->opcode)
 3405                 return (-1);
 3406         else if (a->opcode > b->opcode)
 3407                 return (1);
 3408 
 3409         if (a->version < b->version)
 3410                 return (-1);
 3411         else if (a->version > b->version)
 3412                 return (1);
 3413 
 3414         /* bsearch helper */
 3415         if (a->handler == NULL)
 3416                 return (0);
 3417 
 3418         if ((uintptr_t)a->handler < (uintptr_t)b->handler)
 3419                 return (-1);
 3420         else if ((uintptr_t)a->handler > (uintptr_t)b->handler)
 3421                 return (1);
 3422 
 3423         return (0);
 3424 }
 3425 
 3426 /*
 3427  * Finds sopt handler based on @code and @version.
 3428  *
 3429  * Returns pointer to handler or NULL.
 3430  */
 3431 static struct ipfw_sopt_handler *
 3432 find_sh(uint16_t code, uint8_t version, sopt_handler_f *handler)
 3433 {
 3434         struct ipfw_sopt_handler *sh, h;
 3435 
 3436         memset(&h, 0, sizeof(h));
 3437         h.opcode = code;
 3438         h.version = version;
 3439         h.handler = handler;
 3440 
 3441         sh = (struct ipfw_sopt_handler *)bsearch(&h, ctl3_handlers,
 3442             ctl3_hsize, sizeof(h), compare_sh);
 3443 
 3444         return (sh);
 3445 }
 3446 
 3447 static int
 3448 find_ref_sh(uint16_t opcode, uint8_t version, struct ipfw_sopt_handler *psh)
 3449 {
 3450         struct ipfw_sopt_handler *sh;
 3451 
 3452         CTL3_LOCK();
 3453         if ((sh = find_sh(opcode, version, NULL)) == NULL) {
 3454                 CTL3_UNLOCK();
 3455                 printf("ipfw: ipfw_ctl3 invalid option %d""v""%d\n",
 3456                     opcode, version);
 3457                 return (EINVAL);
 3458         }
 3459         sh->refcnt++;
 3460         ctl3_refct++;
 3461         /* Copy handler data to requested buffer */
 3462         *psh = *sh; 
 3463         CTL3_UNLOCK();
 3464 
 3465         return (0);
 3466 }
 3467 
 3468 static void
 3469 find_unref_sh(struct ipfw_sopt_handler *psh)
 3470 {
 3471         struct ipfw_sopt_handler *sh;
 3472 
 3473         CTL3_LOCK();
 3474         sh = find_sh(psh->opcode, psh->version, NULL);
 3475         KASSERT(sh != NULL, ("ctl3 handler disappeared"));
 3476         sh->refcnt--;
 3477         ctl3_refct--;
 3478         CTL3_UNLOCK();
 3479 }
 3480 
 3481 void
 3482 ipfw_init_sopt_handler(void)
 3483 {
 3484 
 3485         CTL3_LOCK_INIT();
 3486         IPFW_ADD_SOPT_HANDLER(1, scodes);
 3487 }
 3488 
 3489 void
 3490 ipfw_destroy_sopt_handler(void)
 3491 {
 3492 
 3493         IPFW_DEL_SOPT_HANDLER(1, scodes);
 3494         CTL3_LOCK_DESTROY();
 3495 }
 3496 
 3497 /*
 3498  * Adds one or more sockopt handlers to the global array.
 3499  * Function may sleep.
 3500  */
 3501 void
 3502 ipfw_add_sopt_handler(struct ipfw_sopt_handler *sh, size_t count)
 3503 {
 3504         size_t sz;
 3505         struct ipfw_sopt_handler *tmp;
 3506 
 3507         CTL3_LOCK();
 3508 
 3509         for (;;) {
 3510                 sz = ctl3_hsize + count;
 3511                 CTL3_UNLOCK();
 3512                 tmp = malloc(sizeof(*sh) * sz, M_IPFW, M_WAITOK | M_ZERO);
 3513                 CTL3_LOCK();
 3514                 if (ctl3_hsize + count <= sz)
 3515                         break;
 3516 
 3517                 /* Retry */
 3518                 free(tmp, M_IPFW);
 3519         }
 3520 
 3521         /* Merge old & new arrays */
 3522         sz = ctl3_hsize + count;
 3523         memcpy(tmp, ctl3_handlers, ctl3_hsize * sizeof(*sh));
 3524         memcpy(&tmp[ctl3_hsize], sh, count * sizeof(*sh));
 3525         qsort(tmp, sz, sizeof(*sh), compare_sh);
 3526         /* Switch new and free old */
 3527         if (ctl3_handlers != NULL)
 3528                 free(ctl3_handlers, M_IPFW);
 3529         ctl3_handlers = tmp;
 3530         ctl3_hsize = sz;
 3531         ctl3_gencnt++;
 3532 
 3533         CTL3_UNLOCK();
 3534 }
 3535 
 3536 /*
 3537  * Removes one or more sockopt handlers from the global array.
 3538  */
 3539 int
 3540 ipfw_del_sopt_handler(struct ipfw_sopt_handler *sh, size_t count)
 3541 {
 3542         size_t sz;
 3543         struct ipfw_sopt_handler *tmp, *h;
 3544         int i;
 3545 
 3546         CTL3_LOCK();
 3547 
 3548         for (i = 0; i < count; i++) {
 3549                 tmp = &sh[i];
 3550                 h = find_sh(tmp->opcode, tmp->version, tmp->handler);
 3551                 if (h == NULL)
 3552                         continue;
 3553 
 3554                 sz = (ctl3_handlers + ctl3_hsize - (h + 1)) * sizeof(*h);
 3555                 memmove(h, h + 1, sz);
 3556                 ctl3_hsize--;
 3557         }
 3558 
 3559         if (ctl3_hsize == 0) {
 3560                 if (ctl3_handlers != NULL)
 3561                         free(ctl3_handlers, M_IPFW);
 3562                 ctl3_handlers = NULL;
 3563         }
 3564 
 3565         ctl3_gencnt++;
 3566 
 3567         CTL3_UNLOCK();
 3568 
 3569         return (0);
 3570 }
 3571 
 3572 /*
 3573  * Writes data accumulated in @sd to sockopt buffer.
 3574  * Zeroes internal @sd buffer.
 3575  */
 3576 static int
 3577 ipfw_flush_sopt_data(struct sockopt_data *sd)
 3578 {
 3579         struct sockopt *sopt;
 3580         int error;
 3581         size_t sz;
 3582 
 3583         sz = sd->koff;
 3584         if (sz == 0)
 3585                 return (0);
 3586 
 3587         sopt = sd->sopt;
 3588 
 3589         if (sopt->sopt_dir == SOPT_GET) {
 3590                 error = copyout(sd->kbuf, sopt->sopt_val, sz);
 3591                 if (error != 0)
 3592                         return (error);
 3593         }
 3594 
 3595         memset(sd->kbuf, 0, sd->ksize);
 3596         sd->ktotal += sz;
 3597         sd->koff = 0;
 3598         if (sd->ktotal + sd->ksize < sd->valsize)
 3599                 sd->kavail = sd->ksize;
 3600         else
 3601                 sd->kavail = sd->valsize - sd->ktotal;
 3602 
 3603         /* Update sopt buffer data */
 3604         sopt->sopt_valsize = sd->ktotal;
 3605         sopt->sopt_val = sd->sopt_val + sd->ktotal;
 3606 
 3607         return (0);
 3608 }
 3609 
 3610 /*
 3611  * Ensures that @sd buffer has contiguous @neeeded number of
 3612  * bytes.
 3613  *
 3614  * Returns pointer to requested space or NULL.
 3615  */
 3616 caddr_t
 3617 ipfw_get_sopt_space(struct sockopt_data *sd, size_t needed)
 3618 {
 3619         int error;
 3620         caddr_t addr;
 3621 
 3622         if (sd->kavail < needed) {
 3623                 /*
 3624                  * Flush data and try another time.
 3625                  */
 3626                 error = ipfw_flush_sopt_data(sd);
 3627 
 3628                 if (sd->kavail < needed || error != 0)
 3629                         return (NULL);
 3630         }
 3631 
 3632         addr = sd->kbuf + sd->koff;
 3633         sd->koff += needed;
 3634         sd->kavail -= needed;
 3635         return (addr);
 3636 }
 3637 
 3638 /*
 3639  * Requests @needed contiguous bytes from @sd buffer.
 3640  * Function is used to notify subsystem that we are
 3641  * interesed in first @needed bytes (request header)
 3642  * and the rest buffer can be safely zeroed.
 3643  *
 3644  * Returns pointer to requested space or NULL.
 3645  */
 3646 caddr_t
 3647 ipfw_get_sopt_header(struct sockopt_data *sd, size_t needed)
 3648 {
 3649         caddr_t addr;
 3650 
 3651         if ((addr = ipfw_get_sopt_space(sd, needed)) == NULL)
 3652                 return (NULL);
 3653 
 3654         if (sd->kavail > 0)
 3655                 memset(sd->kbuf + sd->koff, 0, sd->kavail);
 3656 
 3657         return (addr);
 3658 }
 3659 
 3660 /*
 3661  * New sockopt handler.
 3662  */
 3663 int
 3664 ipfw_ctl3(struct sockopt *sopt)
 3665 {
 3666         int error, locked;
 3667         size_t size, valsize;
 3668         struct ip_fw_chain *chain;
 3669         char xbuf[256];
 3670         struct sockopt_data sdata;
 3671         struct ipfw_sopt_handler h;
 3672         ip_fw3_opheader *op3 = NULL;
 3673 
 3674         error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW);
 3675         if (error != 0)
 3676                 return (error);
 3677 
 3678         if (sopt->sopt_name != IP_FW3)
 3679                 return (ipfw_ctl(sopt));
 3680 
 3681         chain = &V_layer3_chain;
 3682         error = 0;
 3683 
 3684         /* Save original valsize before it is altered via sooptcopyin() */
 3685         valsize = sopt->sopt_valsize;
 3686         memset(&sdata, 0, sizeof(sdata));
 3687         /* Read op3 header first to determine actual operation */
 3688         op3 = (ip_fw3_opheader *)xbuf;
 3689         error = sooptcopyin(sopt, op3, sizeof(*op3), sizeof(*op3));
 3690         if (error != 0)
 3691                 return (error);
 3692         sopt->sopt_valsize = valsize;
 3693 
 3694         /*
 3695          * Find and reference command.
 3696          */
 3697         error = find_ref_sh(op3->opcode, op3->version, &h);
 3698         if (error != 0)
 3699                 return (error);
 3700 
 3701         /*
 3702          * Disallow modifications in really-really secure mode, but still allow
 3703          * the logging counters to be reset.
 3704          */
 3705         if ((h.dir & HDIR_SET) != 0 && h.opcode != IP_FW_XRESETLOG) {
 3706                 error = securelevel_ge(sopt->sopt_td->td_ucred, 3);
 3707                 if (error != 0) {
 3708                         find_unref_sh(&h);
 3709                         return (error);
 3710                 }
 3711         }
 3712 
 3713         /*
 3714          * Fill in sockopt_data structure that may be useful for
 3715          * IP_FW3 get requests.
 3716          */
 3717         locked = 0;
 3718         if (valsize <= sizeof(xbuf)) {
 3719                 /* use on-stack buffer */
 3720                 sdata.kbuf = xbuf;
 3721                 sdata.ksize = sizeof(xbuf);
 3722                 sdata.kavail = valsize;
 3723         } else {
 3724                 /*
 3725                  * Determine opcode type/buffer size:
 3726                  * allocate sliding-window buf for data export or
 3727                  * contiguous buffer for special ops.
 3728                  */
 3729                 if ((h.dir & HDIR_SET) != 0) {
 3730                         /* Set request. Allocate contigous buffer. */
 3731                         if (valsize > CTL3_LARGEBUF) {
 3732                                 find_unref_sh(&h);
 3733                                 return (EFBIG);
 3734                         }
 3735 
 3736                         size = valsize;
 3737                 } else {
 3738                         /* Get request. Allocate sliding window buffer */
 3739                         size = (valsize<CTL3_SMALLBUF) ? valsize:CTL3_SMALLBUF;
 3740 
 3741                         if (size < valsize) {
 3742                                 /* We have to wire user buffer */
 3743                                 error = vslock(sopt->sopt_val, valsize);
 3744                                 if (error != 0)
 3745                                         return (error);
 3746                                 locked = 1;
 3747                         }
 3748                 }
 3749 
 3750                 sdata.kbuf = malloc(size, M_TEMP, M_WAITOK | M_ZERO);
 3751                 sdata.ksize = size;
 3752                 sdata.kavail = size;
 3753         }
 3754 
 3755         sdata.sopt = sopt;
 3756         sdata.sopt_val = sopt->sopt_val;
 3757         sdata.valsize = valsize;
 3758 
 3759         /*
 3760          * Copy either all request (if valsize < bsize_max)
 3761          * or first bsize_max bytes to guarantee most consumers
 3762          * that all necessary data has been copied).
 3763          * Anyway, copy not less than sizeof(ip_fw3_opheader).
 3764          */
 3765         if ((error = sooptcopyin(sopt, sdata.kbuf, sdata.ksize,
 3766             sizeof(ip_fw3_opheader))) != 0)
 3767                 return (error);
 3768         op3 = (ip_fw3_opheader *)sdata.kbuf;
 3769 
 3770         /* Finally, run handler */
 3771         error = h.handler(chain, op3, &sdata);
 3772         find_unref_sh(&h);
 3773 
 3774         /* Flush state and free buffers */
 3775         if (error == 0)
 3776                 error = ipfw_flush_sopt_data(&sdata);
 3777         else
 3778                 ipfw_flush_sopt_data(&sdata);
 3779 
 3780         if (locked != 0)
 3781                 vsunlock(sdata.sopt_val, valsize);
 3782 
 3783         /* Restore original pointer and set number of bytes written */
 3784         sopt->sopt_val = sdata.sopt_val;
 3785         sopt->sopt_valsize = sdata.ktotal;
 3786         if (sdata.kbuf != xbuf)
 3787                 free(sdata.kbuf, M_TEMP);
 3788 
 3789         return (error);
 3790 }
 3791 
 3792 /**
 3793  * {set|get}sockopt parser.
 3794  */
 3795 int
 3796 ipfw_ctl(struct sockopt *sopt)
 3797 {
 3798 #define RULE_MAXSIZE    (512*sizeof(u_int32_t))
 3799         int error;
 3800         size_t size;
 3801         struct ip_fw *buf;
 3802         struct ip_fw_rule0 *rule;
 3803         struct ip_fw_chain *chain;
 3804         u_int32_t rulenum[2];
 3805         uint32_t opt;
 3806         struct rule_check_info ci;
 3807         IPFW_RLOCK_TRACKER;
 3808 
 3809         chain = &V_layer3_chain;
 3810         error = 0;
 3811 
 3812         opt = sopt->sopt_name;
 3813 
 3814         /*
 3815          * Disallow modifications in really-really secure mode, but still allow
 3816          * the logging counters to be reset.
 3817          */
 3818         if (opt == IP_FW_ADD ||
 3819             (sopt->sopt_dir == SOPT_SET && opt != IP_FW_RESETLOG)) {
 3820                 error = securelevel_ge(sopt->sopt_td->td_ucred, 3);
 3821                 if (error != 0)
 3822                         return (error);
 3823         }
 3824 
 3825         switch (opt) {
 3826         case IP_FW_GET:
 3827                 /*
 3828                  * pass up a copy of the current rules. Static rules
 3829                  * come first (the last of which has number IPFW_DEFAULT_RULE),
 3830                  * followed by a possibly empty list of dynamic rule.
 3831                  * The last dynamic rule has NULL in the "next" field.
 3832                  *
 3833                  * Note that the calculated size is used to bound the
 3834                  * amount of data returned to the user.  The rule set may
 3835                  * change between calculating the size and returning the
 3836                  * data in which case we'll just return what fits.
 3837                  */
 3838                 for (;;) {
 3839                         int len = 0, want;
 3840 
 3841                         size = chain->static_len;
 3842                         size += ipfw_dyn_len();
 3843                         if (size >= sopt->sopt_valsize)
 3844                                 break;
 3845                         buf = malloc(size, M_TEMP, M_WAITOK | M_ZERO);
 3846                         IPFW_UH_RLOCK(chain);
 3847                         /* check again how much space we need */
 3848                         want = chain->static_len + ipfw_dyn_len();
 3849                         if (size >= want)
 3850                                 len = ipfw_getrules(chain, buf, size);
 3851                         IPFW_UH_RUNLOCK(chain);
 3852                         if (size >= want)
 3853                                 error = sooptcopyout(sopt, buf, len);
 3854                         free(buf, M_TEMP);
 3855                         if (size >= want)
 3856                                 break;
 3857                 }
 3858                 break;
 3859 
 3860         case IP_FW_FLUSH:
 3861                 /* locking is done within del_entry() */
 3862                 error = del_entry(chain, 0); /* special case, rule=0, cmd=0 means all */
 3863                 break;
 3864 
 3865         case IP_FW_ADD:
 3866                 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK);
 3867                 error = sooptcopyin(sopt, rule, RULE_MAXSIZE,
 3868                         sizeof(struct ip_fw7) );
 3869 
 3870                 memset(&ci, 0, sizeof(struct rule_check_info));
 3871 
 3872                 /*
 3873                  * If the size of commands equals RULESIZE7 then we assume
 3874                  * a FreeBSD7.2 binary is talking to us (set is7=1).
 3875                  * is7 is persistent so the next 'ipfw list' command
 3876                  * will use this format.
 3877                  * NOTE: If wrong version is guessed (this can happen if
 3878                  *       the first ipfw command is 'ipfw [pipe] list')
 3879                  *       the ipfw binary may crash or loop infinitly...
 3880                  */
 3881                 size = sopt->sopt_valsize;
 3882                 if (size == RULESIZE7(rule)) {
 3883                     is7 = 1;
 3884                     error = convert_rule_to_8(rule);
 3885                     if (error) {
 3886                         free(rule, M_TEMP);
 3887                         return error;
 3888                     }
 3889                     size = RULESIZE(rule);
 3890                 } else
 3891                     is7 = 0;
 3892                 if (error == 0)
 3893                         error = check_ipfw_rule0(rule, size, &ci);
 3894                 if (error == 0) {
 3895                         /* locking is done within add_rule() */
 3896                         struct ip_fw *krule;
 3897                         krule = ipfw_alloc_rule(chain, RULEKSIZE0(rule));
 3898                         ci.urule = (caddr_t)rule;
 3899                         ci.krule = krule;
 3900                         import_rule0(&ci);
 3901                         error = commit_rules(chain, &ci, 1);
 3902                         if (error != 0)
 3903                                 ipfw_free_rule(ci.krule);
 3904                         else if (sopt->sopt_dir == SOPT_GET) {
 3905                                 if (is7) {
 3906                                         error = convert_rule_to_7(rule);
 3907                                         size = RULESIZE7(rule);
 3908                                         if (error) {
 3909                                                 free(rule, M_TEMP);
 3910                                                 return error;
 3911                                         }
 3912                                 }
 3913                                 error = sooptcopyout(sopt, rule, size);
 3914                         }
 3915                 }
 3916                 free(rule, M_TEMP);
 3917                 break;
 3918 
 3919         case IP_FW_DEL:
 3920                 /*
 3921                  * IP_FW_DEL is used for deleting single rules or sets,
 3922                  * and (ab)used to atomically manipulate sets. Argument size
 3923                  * is used to distinguish between the two:
 3924                  *    sizeof(u_int32_t)
 3925                  *      delete single rule or set of rules,
 3926                  *      or reassign rules (or sets) to a different set.
 3927                  *    2*sizeof(u_int32_t)
 3928                  *      atomic disable/enable sets.
 3929                  *      first u_int32_t contains sets to be disabled,
 3930                  *      second u_int32_t contains sets to be enabled.
 3931                  */
 3932                 error = sooptcopyin(sopt, rulenum,
 3933                         2*sizeof(u_int32_t), sizeof(u_int32_t));
 3934                 if (error)
 3935                         break;
 3936                 size = sopt->sopt_valsize;
 3937                 if (size == sizeof(u_int32_t) && rulenum[0] != 0) {
 3938                         /* delete or reassign, locking done in del_entry() */
 3939                         error = del_entry(chain, rulenum[0]);
 3940                 } else if (size == 2*sizeof(u_int32_t)) { /* set enable/disable */
 3941                         IPFW_UH_WLOCK(chain);
 3942                         V_set_disable =
 3943                             (V_set_disable | rulenum[0]) & ~rulenum[1] &
 3944                             ~(1<<RESVD_SET); /* set RESVD_SET always enabled */
 3945                         IPFW_UH_WUNLOCK(chain);
 3946                 } else
 3947                         error = EINVAL;
 3948                 break;
 3949 
 3950         case IP_FW_ZERO:
 3951         case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */
 3952                 rulenum[0] = 0;
 3953                 if (sopt->sopt_val != 0) {
 3954                     error = sooptcopyin(sopt, rulenum,
 3955                             sizeof(u_int32_t), sizeof(u_int32_t));
 3956                     if (error)
 3957                         break;
 3958                 }
 3959                 error = zero_entry(chain, rulenum[0],
 3960                         sopt->sopt_name == IP_FW_RESETLOG);
 3961                 break;
 3962 
 3963         /*--- TABLE opcodes ---*/
 3964         case IP_FW_TABLE_ADD:
 3965         case IP_FW_TABLE_DEL:
 3966                 {
 3967                         ipfw_table_entry ent;
 3968                         struct tentry_info tei;
 3969                         struct tid_info ti;
 3970                         struct table_value v;
 3971 
 3972                         error = sooptcopyin(sopt, &ent,
 3973                             sizeof(ent), sizeof(ent));
 3974                         if (error)
 3975                                 break;
 3976 
 3977                         memset(&tei, 0, sizeof(tei));
 3978                         tei.paddr = &ent.addr;
 3979                         tei.subtype = AF_INET;
 3980                         tei.masklen = ent.masklen;
 3981                         ipfw_import_table_value_legacy(ent.value, &v);
 3982                         tei.pvalue = &v;
 3983                         memset(&ti, 0, sizeof(ti));
 3984                         ti.uidx = ent.tbl;
 3985                         ti.type = IPFW_TABLE_CIDR;
 3986 
 3987                         error = (opt == IP_FW_TABLE_ADD) ?
 3988                             add_table_entry(chain, &ti, &tei, 0, 1) :
 3989                             del_table_entry(chain, &ti, &tei, 0, 1);
 3990                 }
 3991                 break;
 3992 
 3993         case IP_FW_TABLE_FLUSH:
 3994                 {
 3995                         u_int16_t tbl;
 3996                         struct tid_info ti;
 3997 
 3998                         error = sooptcopyin(sopt, &tbl,
 3999                             sizeof(tbl), sizeof(tbl));
 4000                         if (error)
 4001                                 break;
 4002                         memset(&ti, 0, sizeof(ti));
 4003                         ti.uidx = tbl;
 4004                         error = flush_table(chain, &ti);
 4005                 }
 4006                 break;
 4007 
 4008         case IP_FW_TABLE_GETSIZE:
 4009                 {
 4010                         u_int32_t tbl, cnt;
 4011                         struct tid_info ti;
 4012 
 4013                         if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl),
 4014                             sizeof(tbl))))
 4015                                 break;
 4016                         memset(&ti, 0, sizeof(ti));
 4017                         ti.uidx = tbl;
 4018                         IPFW_RLOCK(chain);
 4019                         error = ipfw_count_table(chain, &ti, &cnt);
 4020                         IPFW_RUNLOCK(chain);
 4021                         if (error)
 4022                                 break;
 4023                         error = sooptcopyout(sopt, &cnt, sizeof(cnt));
 4024                 }
 4025                 break;
 4026 
 4027         case IP_FW_TABLE_LIST:
 4028                 {
 4029                         ipfw_table *tbl;
 4030                         struct tid_info ti;
 4031 
 4032                         if (sopt->sopt_valsize < sizeof(*tbl)) {
 4033                                 error = EINVAL;
 4034                                 break;
 4035                         }
 4036                         size = sopt->sopt_valsize;
 4037                         tbl = malloc(size, M_TEMP, M_WAITOK);
 4038                         error = sooptcopyin(sopt, tbl, size, sizeof(*tbl));
 4039                         if (error) {
 4040                                 free(tbl, M_TEMP);
 4041                                 break;
 4042                         }
 4043                         tbl->size = (size - sizeof(*tbl)) /
 4044                             sizeof(ipfw_table_entry);
 4045                         memset(&ti, 0, sizeof(ti));
 4046                         ti.uidx = tbl->tbl;
 4047                         IPFW_RLOCK(chain);
 4048                         error = ipfw_dump_table_legacy(chain, &ti, tbl);
 4049                         IPFW_RUNLOCK(chain);
 4050                         if (error) {
 4051                                 free(tbl, M_TEMP);
 4052                                 break;
 4053                         }
 4054                         error = sooptcopyout(sopt, tbl, size);
 4055                         free(tbl, M_TEMP);
 4056                 }
 4057                 break;
 4058 
 4059         /*--- NAT operations are protected by the IPFW_LOCK ---*/
 4060         case IP_FW_NAT_CFG:
 4061                 if (IPFW_NAT_LOADED)
 4062                         error = ipfw_nat_cfg_ptr(sopt);
 4063                 else {
 4064                         printf("IP_FW_NAT_CFG: %s\n",
 4065                             "ipfw_nat not present, please load it");
 4066                         error = EINVAL;
 4067                 }
 4068                 break;
 4069 
 4070         case IP_FW_NAT_DEL:
 4071                 if (IPFW_NAT_LOADED)
 4072                         error = ipfw_nat_del_ptr(sopt);
 4073                 else {
 4074                         printf("IP_FW_NAT_DEL: %s\n",
 4075                             "ipfw_nat not present, please load it");
 4076                         error = EINVAL;
 4077                 }
 4078                 break;
 4079 
 4080         case IP_FW_NAT_GET_CONFIG:
 4081                 if (IPFW_NAT_LOADED)
 4082                         error = ipfw_nat_get_cfg_ptr(sopt);
 4083                 else {
 4084                         printf("IP_FW_NAT_GET_CFG: %s\n",
 4085                             "ipfw_nat not present, please load it");
 4086                         error = EINVAL;
 4087                 }
 4088                 break;
 4089 
 4090         case IP_FW_NAT_GET_LOG:
 4091                 if (IPFW_NAT_LOADED)
 4092                         error = ipfw_nat_get_log_ptr(sopt);
 4093                 else {
 4094                         printf("IP_FW_NAT_GET_LOG: %s\n",
 4095                             "ipfw_nat not present, please load it");
 4096                         error = EINVAL;
 4097                 }
 4098                 break;
 4099 
 4100         default:
 4101                 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name);
 4102                 error = EINVAL;
 4103         }
 4104 
 4105         return (error);
 4106 #undef RULE_MAXSIZE
 4107 }
 4108 #define RULE_MAXSIZE    (256*sizeof(u_int32_t))
 4109 
 4110 /* Functions to convert rules 7.2 <==> 8.0 */
 4111 static int
 4112 convert_rule_to_7(struct ip_fw_rule0 *rule)
 4113 {
 4114         /* Used to modify original rule */
 4115         struct ip_fw7 *rule7 = (struct ip_fw7 *)rule;
 4116         /* copy of original rule, version 8 */
 4117         struct ip_fw_rule0 *tmp;
 4118 
 4119         /* Used to copy commands */
 4120         ipfw_insn *ccmd, *dst;
 4121         int ll = 0, ccmdlen = 0;
 4122 
 4123         tmp = malloc(RULE_MAXSIZE, M_TEMP, M_NOWAIT | M_ZERO);
 4124         if (tmp == NULL) {
 4125                 return 1; //XXX error
 4126         }
 4127         bcopy(rule, tmp, RULE_MAXSIZE);
 4128 
 4129         /* Copy fields */
 4130         //rule7->_pad = tmp->_pad;
 4131         rule7->set = tmp->set;
 4132         rule7->rulenum = tmp->rulenum;
 4133         rule7->cmd_len = tmp->cmd_len;
 4134         rule7->act_ofs = tmp->act_ofs;
 4135         rule7->next_rule = (struct ip_fw7 *)tmp->next_rule;
 4136         rule7->cmd_len = tmp->cmd_len;
 4137         rule7->pcnt = tmp->pcnt;
 4138         rule7->bcnt = tmp->bcnt;
 4139         rule7->timestamp = tmp->timestamp;
 4140 
 4141         /* Copy commands */
 4142         for (ll = tmp->cmd_len, ccmd = tmp->cmd, dst = rule7->cmd ;
 4143                         ll > 0 ; ll -= ccmdlen, ccmd += ccmdlen, dst += ccmdlen) {
 4144                 ccmdlen = F_LEN(ccmd);
 4145 
 4146                 bcopy(ccmd, dst, F_LEN(ccmd)*sizeof(uint32_t));
 4147 
 4148                 if (dst->opcode > O_NAT)
 4149                         /* O_REASS doesn't exists in 7.2 version, so
 4150                          * decrement opcode if it is after O_REASS
 4151                          */
 4152                         dst->opcode--;
 4153 
 4154                 if (ccmdlen > ll) {
 4155                         printf("ipfw: opcode %d size truncated\n",
 4156                                 ccmd->opcode);
 4157                         return EINVAL;
 4158                 }
 4159         }
 4160         free(tmp, M_TEMP);
 4161 
 4162         return 0;
 4163 }
 4164 
 4165 static int
 4166 convert_rule_to_8(struct ip_fw_rule0 *rule)
 4167 {
 4168         /* Used to modify original rule */
 4169         struct ip_fw7 *rule7 = (struct ip_fw7 *) rule;
 4170 
 4171         /* Used to copy commands */
 4172         ipfw_insn *ccmd, *dst;
 4173         int ll = 0, ccmdlen = 0;
 4174 
 4175         /* Copy of original rule */
 4176         struct ip_fw7 *tmp = malloc(RULE_MAXSIZE, M_TEMP, M_NOWAIT | M_ZERO);
 4177         if (tmp == NULL) {
 4178                 return 1; //XXX error
 4179         }
 4180 
 4181         bcopy(rule7, tmp, RULE_MAXSIZE);
 4182 
 4183         for (ll = tmp->cmd_len, ccmd = tmp->cmd, dst = rule->cmd ;
 4184                         ll > 0 ; ll -= ccmdlen, ccmd += ccmdlen, dst += ccmdlen) {
 4185                 ccmdlen = F_LEN(ccmd);
 4186                 
 4187                 bcopy(ccmd, dst, F_LEN(ccmd)*sizeof(uint32_t));
 4188 
 4189                 if (dst->opcode > O_NAT)
 4190                         /* O_REASS doesn't exists in 7.2 version, so
 4191                          * increment opcode if it is after O_REASS
 4192                          */
 4193                         dst->opcode++;
 4194 
 4195                 if (ccmdlen > ll) {
 4196                         printf("ipfw: opcode %d size truncated\n",
 4197                             ccmd->opcode);
 4198                         return EINVAL;
 4199                 }
 4200         }
 4201 
 4202         rule->_pad = tmp->_pad;
 4203         rule->set = tmp->set;
 4204         rule->rulenum = tmp->rulenum;
 4205         rule->cmd_len = tmp->cmd_len;
 4206         rule->act_ofs = tmp->act_ofs;
 4207         rule->next_rule = (struct ip_fw *)tmp->next_rule;
 4208         rule->cmd_len = tmp->cmd_len;
 4209         rule->id = 0; /* XXX see if is ok = 0 */
 4210         rule->pcnt = tmp->pcnt;
 4211         rule->bcnt = tmp->bcnt;
 4212         rule->timestamp = tmp->timestamp;
 4213 
 4214         free (tmp, M_TEMP);
 4215         return 0;
 4216 }
 4217 
 4218 /*
 4219  * Named object api
 4220  *
 4221  */
 4222 
 4223 void
 4224 ipfw_init_srv(struct ip_fw_chain *ch)
 4225 {
 4226 
 4227         ch->srvmap = ipfw_objhash_create(IPFW_OBJECTS_DEFAULT);
 4228         ch->srvstate = malloc(sizeof(void *) * IPFW_OBJECTS_DEFAULT,
 4229             M_IPFW, M_WAITOK | M_ZERO);
 4230 }
 4231 
 4232 void
 4233 ipfw_destroy_srv(struct ip_fw_chain *ch)
 4234 {
 4235 
 4236         free(ch->srvstate, M_IPFW);
 4237         ipfw_objhash_destroy(ch->srvmap);
 4238 }
 4239 
 4240 /*
 4241  * Allocate new bitmask which can be used to enlarge/shrink
 4242  * named instance index.
 4243  */
 4244 void
 4245 ipfw_objhash_bitmap_alloc(uint32_t items, void **idx, int *pblocks)
 4246 {
 4247         size_t size;
 4248         int max_blocks;
 4249         u_long *idx_mask;
 4250 
 4251         KASSERT((items % BLOCK_ITEMS) == 0,
 4252            ("bitmask size needs to power of 2 and greater or equal to %zu",
 4253             BLOCK_ITEMS));
 4254 
 4255         max_blocks = items / BLOCK_ITEMS;
 4256         size = items / 8;
 4257         idx_mask = malloc(size * IPFW_MAX_SETS, M_IPFW, M_WAITOK);
 4258         /* Mark all as free */
 4259         memset(idx_mask, 0xFF, size * IPFW_MAX_SETS);
 4260         *idx_mask &= ~(u_long)1; /* Skip index 0 */
 4261 
 4262         *idx = idx_mask;
 4263         *pblocks = max_blocks;
 4264 }
 4265 
 4266 /*
 4267  * Copy current bitmask index to new one.
 4268  */
 4269 void
 4270 ipfw_objhash_bitmap_merge(struct namedobj_instance *ni, void **idx, int *blocks)
 4271 {
 4272         int old_blocks, new_blocks;
 4273         u_long *old_idx, *new_idx;
 4274         int i;
 4275 
 4276         old_idx = ni->idx_mask;
 4277         old_blocks = ni->max_blocks;
 4278         new_idx = *idx;
 4279         new_blocks = *blocks;
 4280 
 4281         for (i = 0; i < IPFW_MAX_SETS; i++) {
 4282                 memcpy(&new_idx[new_blocks * i], &old_idx[old_blocks * i],
 4283                     old_blocks * sizeof(u_long));
 4284         }
 4285 }
 4286 
 4287 /*
 4288  * Swaps current @ni index with new one.
 4289  */
 4290 void
 4291 ipfw_objhash_bitmap_swap(struct namedobj_instance *ni, void **idx, int *blocks)
 4292 {
 4293         int old_blocks;
 4294         u_long *old_idx;
 4295 
 4296         old_idx = ni->idx_mask;
 4297         old_blocks = ni->max_blocks;
 4298 
 4299         ni->idx_mask = *idx;
 4300         ni->max_blocks = *blocks;
 4301 
 4302         /* Save old values */
 4303         *idx = old_idx;
 4304         *blocks = old_blocks;
 4305 }
 4306 
 4307 void
 4308 ipfw_objhash_bitmap_free(void *idx, int blocks)
 4309 {
 4310 
 4311         free(idx, M_IPFW);
 4312 }
 4313 
 4314 /*
 4315  * Creates named hash instance.
 4316  * Must be called without holding any locks.
 4317  * Return pointer to new instance.
 4318  */
 4319 struct namedobj_instance *
 4320 ipfw_objhash_create(uint32_t items)
 4321 {
 4322         struct namedobj_instance *ni;
 4323         int i;
 4324         size_t size;
 4325 
 4326         size = sizeof(struct namedobj_instance) +
 4327             sizeof(struct namedobjects_head) * NAMEDOBJ_HASH_SIZE +
 4328             sizeof(struct namedobjects_head) * NAMEDOBJ_HASH_SIZE;
 4329 
 4330         ni = malloc(size, M_IPFW, M_WAITOK | M_ZERO);
 4331         ni->nn_size = NAMEDOBJ_HASH_SIZE;
 4332         ni->nv_size = NAMEDOBJ_HASH_SIZE;
 4333 
 4334         ni->names = (struct namedobjects_head *)(ni +1);
 4335         ni->values = &ni->names[ni->nn_size];
 4336 
 4337         for (i = 0; i < ni->nn_size; i++)
 4338                 TAILQ_INIT(&ni->names[i]);
 4339 
 4340         for (i = 0; i < ni->nv_size; i++)
 4341                 TAILQ_INIT(&ni->values[i]);
 4342 
 4343         /* Set default hashing/comparison functions */
 4344         ni->hash_f = objhash_hash_name;
 4345         ni->cmp_f = objhash_cmp_name;
 4346 
 4347         /* Allocate bitmask separately due to possible resize */
 4348         ipfw_objhash_bitmap_alloc(items, (void*)&ni->idx_mask, &ni->max_blocks);
 4349 
 4350         return (ni);
 4351 }
 4352 
 4353 void
 4354 ipfw_objhash_destroy(struct namedobj_instance *ni)
 4355 {
 4356 
 4357         free(ni->idx_mask, M_IPFW);
 4358         free(ni, M_IPFW);
 4359 }
 4360 
 4361 void
 4362 ipfw_objhash_set_funcs(struct namedobj_instance *ni, objhash_hash_f *hash_f,
 4363     objhash_cmp_f *cmp_f)
 4364 {
 4365 
 4366         ni->hash_f = hash_f;
 4367         ni->cmp_f = cmp_f;
 4368 }
 4369 
 4370 static uint32_t
 4371 objhash_hash_name(struct namedobj_instance *ni, const void *name, uint32_t set)
 4372 {
 4373 
 4374         return (fnv_32_str((const char *)name, FNV1_32_INIT));
 4375 }
 4376 
 4377 static int
 4378 objhash_cmp_name(struct named_object *no, const void *name, uint32_t set)
 4379 {
 4380 
 4381         if ((strcmp(no->name, (const char *)name) == 0) && (no->set == set))
 4382                 return (0);
 4383 
 4384         return (1);
 4385 }
 4386 
 4387 static uint32_t
 4388 objhash_hash_idx(struct namedobj_instance *ni, uint32_t val)
 4389 {
 4390         uint32_t v;
 4391 
 4392         v = val % (ni->nv_size - 1);
 4393 
 4394         return (v);
 4395 }
 4396 
 4397 struct named_object *
 4398 ipfw_objhash_lookup_name(struct namedobj_instance *ni, uint32_t set,
 4399     const char *name)
 4400 {
 4401         struct named_object *no;
 4402         uint32_t hash;
 4403 
 4404         hash = ni->hash_f(ni, name, set) % ni->nn_size;
 4405 
 4406         TAILQ_FOREACH(no, &ni->names[hash], nn_next) {
 4407                 if (ni->cmp_f(no, name, set) == 0)
 4408                         return (no);
 4409         }
 4410 
 4411         return (NULL);
 4412 }
 4413 
 4414 /*
 4415  * Find named object by @uid.
 4416  * Check @tlvs for valid data inside.
 4417  *
 4418  * Returns pointer to found TLV or NULL.
 4419  */
 4420 ipfw_obj_ntlv *
 4421 ipfw_find_name_tlv_type(void *tlvs, int len, uint16_t uidx, uint32_t etlv)
 4422 {
 4423         ipfw_obj_ntlv *ntlv;
 4424         uintptr_t pa, pe;
 4425         int l;
 4426 
 4427         pa = (uintptr_t)tlvs;
 4428         pe = pa + len;
 4429         l = 0;
 4430         for (; pa < pe; pa += l) {
 4431                 ntlv = (ipfw_obj_ntlv *)pa;
 4432                 l = ntlv->head.length;
 4433 
 4434                 if (l != sizeof(*ntlv))
 4435                         return (NULL);
 4436 
 4437                 if (ntlv->idx != uidx)
 4438                         continue;
 4439                 /*
 4440                  * When userland has specified zero TLV type, do
 4441                  * not compare it with eltv. In some cases userland
 4442                  * doesn't know what type should it have. Use only
 4443                  * uidx and name for search named_object.
 4444                  */
 4445                 if (ntlv->head.type != 0 &&
 4446                     ntlv->head.type != (uint16_t)etlv)
 4447                         continue;
 4448 
 4449                 if (ipfw_check_object_name_generic(ntlv->name) != 0)
 4450                         return (NULL);
 4451 
 4452                 return (ntlv);
 4453         }
 4454 
 4455         return (NULL);
 4456 }
 4457 
 4458 /*
 4459  * Finds object config based on either legacy index
 4460  * or name in ntlv.
 4461  * Note @ti structure contains unchecked data from userland.
 4462  *
 4463  * Returns 0 in success and fills in @pno with found config
 4464  */
 4465 int
 4466 ipfw_objhash_find_type(struct namedobj_instance *ni, struct tid_info *ti,
 4467     uint32_t etlv, struct named_object **pno)
 4468 {
 4469         char *name;
 4470         ipfw_obj_ntlv *ntlv;
 4471         uint32_t set;
 4472 
 4473         if (ti->tlvs == NULL)
 4474                 return (EINVAL);
 4475 
 4476         ntlv = ipfw_find_name_tlv_type(ti->tlvs, ti->tlen, ti->uidx, etlv);
 4477         if (ntlv == NULL)
 4478                 return (EINVAL);
 4479         name = ntlv->name;
 4480 
 4481         /*
 4482          * Use set provided by @ti instead of @ntlv one.
 4483          * This is needed due to different sets behavior
 4484          * controlled by V_fw_tables_sets.
 4485          */
 4486         set = ti->set;
 4487         *pno = ipfw_objhash_lookup_name(ni, set, name);
 4488         if (*pno == NULL)
 4489                 return (ESRCH);
 4490         return (0);
 4491 }
 4492 
 4493 /*
 4494  * Find named object by name, considering also its TLV type.
 4495  */
 4496 struct named_object *
 4497 ipfw_objhash_lookup_name_type(struct namedobj_instance *ni, uint32_t set,
 4498     uint32_t type, const char *name)
 4499 {
 4500         struct named_object *no;
 4501         uint32_t hash;
 4502 
 4503         hash = ni->hash_f(ni, name, set) % ni->nn_size;
 4504 
 4505         TAILQ_FOREACH(no, &ni->names[hash], nn_next) {
 4506                 if (ni->cmp_f(no, name, set) == 0 &&
 4507                     no->etlv == (uint16_t)type)
 4508                         return (no);
 4509         }
 4510 
 4511         return (NULL);
 4512 }
 4513 
 4514 struct named_object *
 4515 ipfw_objhash_lookup_kidx(struct namedobj_instance *ni, uint16_t kidx)
 4516 {
 4517         struct named_object *no;
 4518         uint32_t hash;
 4519 
 4520         hash = objhash_hash_idx(ni, kidx);
 4521 
 4522         TAILQ_FOREACH(no, &ni->values[hash], nv_next) {
 4523                 if (no->kidx == kidx)
 4524                         return (no);
 4525         }
 4526 
 4527         return (NULL);
 4528 }
 4529 
 4530 int
 4531 ipfw_objhash_same_name(struct namedobj_instance *ni, struct named_object *a,
 4532     struct named_object *b)
 4533 {
 4534 
 4535         if ((strcmp(a->name, b->name) == 0) && a->set == b->set)
 4536                 return (1);
 4537 
 4538         return (0);
 4539 }
 4540 
 4541 void
 4542 ipfw_objhash_add(struct namedobj_instance *ni, struct named_object *no)
 4543 {
 4544         uint32_t hash;
 4545 
 4546         hash = ni->hash_f(ni, no->name, no->set) % ni->nn_size;
 4547         TAILQ_INSERT_HEAD(&ni->names[hash], no, nn_next);
 4548 
 4549         hash = objhash_hash_idx(ni, no->kidx);
 4550         TAILQ_INSERT_HEAD(&ni->values[hash], no, nv_next);
 4551 
 4552         ni->count++;
 4553 }
 4554 
 4555 void
 4556 ipfw_objhash_del(struct namedobj_instance *ni, struct named_object *no)
 4557 {
 4558         uint32_t hash;
 4559 
 4560         hash = ni->hash_f(ni, no->name, no->set) % ni->nn_size;
 4561         TAILQ_REMOVE(&ni->names[hash], no, nn_next);
 4562 
 4563         hash = objhash_hash_idx(ni, no->kidx);
 4564         TAILQ_REMOVE(&ni->values[hash], no, nv_next);
 4565 
 4566         ni->count--;
 4567 }
 4568 
 4569 uint32_t
 4570 ipfw_objhash_count(struct namedobj_instance *ni)
 4571 {
 4572 
 4573         return (ni->count);
 4574 }
 4575 
 4576 uint32_t
 4577 ipfw_objhash_count_type(struct namedobj_instance *ni, uint16_t type)
 4578 {
 4579         struct named_object *no;
 4580         uint32_t count;
 4581         int i;
 4582 
 4583         count = 0;
 4584         for (i = 0; i < ni->nn_size; i++) {
 4585                 TAILQ_FOREACH(no, &ni->names[i], nn_next) {
 4586                         if (no->etlv == type)
 4587                                 count++;
 4588                 }
 4589         }
 4590         return (count);
 4591 }
 4592 
 4593 /*
 4594  * Runs @func for each found named object.
 4595  * It is safe to delete objects from callback
 4596  */
 4597 int
 4598 ipfw_objhash_foreach(struct namedobj_instance *ni, objhash_cb_t *f, void *arg)
 4599 {
 4600         struct named_object *no, *no_tmp;
 4601         int i, ret;
 4602 
 4603         for (i = 0; i < ni->nn_size; i++) {
 4604                 TAILQ_FOREACH_SAFE(no, &ni->names[i], nn_next, no_tmp) {
 4605                         ret = f(ni, no, arg);
 4606                         if (ret != 0)
 4607                                 return (ret);
 4608                 }
 4609         }
 4610         return (0);
 4611 }
 4612 
 4613 /*
 4614  * Runs @f for each found named object with type @type.
 4615  * It is safe to delete objects from callback
 4616  */
 4617 int
 4618 ipfw_objhash_foreach_type(struct namedobj_instance *ni, objhash_cb_t *f,
 4619     void *arg, uint16_t type)
 4620 {
 4621         struct named_object *no, *no_tmp;
 4622         int i, ret;
 4623 
 4624         for (i = 0; i < ni->nn_size; i++) {
 4625                 TAILQ_FOREACH_SAFE(no, &ni->names[i], nn_next, no_tmp) {
 4626                         if (no->etlv != type)
 4627                                 continue;
 4628                         ret = f(ni, no, arg);
 4629                         if (ret != 0)
 4630                                 return (ret);
 4631                 }
 4632         }
 4633         return (0);
 4634 }
 4635 
 4636 /*
 4637  * Removes index from given set.
 4638  * Returns 0 on success.
 4639  */
 4640 int
 4641 ipfw_objhash_free_idx(struct namedobj_instance *ni, uint16_t idx)
 4642 {
 4643         u_long *mask;
 4644         int i, v;
 4645 
 4646         i = idx / BLOCK_ITEMS;
 4647         v = idx % BLOCK_ITEMS;
 4648 
 4649         if (i >= ni->max_blocks)
 4650                 return (1);
 4651 
 4652         mask = &ni->idx_mask[i];
 4653 
 4654         if ((*mask & ((u_long)1 << v)) != 0)
 4655                 return (1);
 4656 
 4657         /* Mark as free */
 4658         *mask |= (u_long)1 << v;
 4659 
 4660         /* Update free offset */
 4661         if (ni->free_off[0] > i)
 4662                 ni->free_off[0] = i;
 4663 
 4664         return (0);
 4665 }
 4666 
 4667 /*
 4668  * Allocate new index in given instance and stores in in @pidx.
 4669  * Returns 0 on success.
 4670  */
 4671 int
 4672 ipfw_objhash_alloc_idx(void *n, uint16_t *pidx)
 4673 {
 4674         struct namedobj_instance *ni;
 4675         u_long *mask;
 4676         int i, off, v;
 4677 
 4678         ni = (struct namedobj_instance *)n;
 4679 
 4680         off = ni->free_off[0];
 4681         mask = &ni->idx_mask[off];
 4682 
 4683         for (i = off; i < ni->max_blocks; i++, mask++) {
 4684                 if ((v = ffsl(*mask)) == 0)
 4685                         continue;
 4686 
 4687                 /* Mark as busy */
 4688                 *mask &= ~ ((u_long)1 << (v - 1));
 4689 
 4690                 ni->free_off[0] = i;
 4691                 
 4692                 v = BLOCK_ITEMS * i + v - 1;
 4693 
 4694                 *pidx = v;
 4695                 return (0);
 4696         }
 4697 
 4698         return (1);
 4699 }
 4700 
 4701 /* end of file */

Cache object: 163217f06edd60139dbda84acafd1d59


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.