sys/netpfil/pf/pf_ioctl.c

   1 /*-
   2  * Copyright (c) 2001 Daniel Hartmeier
   3  * Copyright (c) 2002,2003 Henning Brauer
   4  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
   5  * All rights reserved.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  *
  11  *    - Redistributions of source code must retain the above copyright
  12  *      notice, this list of conditions and the following disclaimer.
  13  *    - Redistributions in binary form must reproduce the above
  14  *      copyright notice, this list of conditions and the following
  15  *      disclaimer in the documentation and/or other materials provided
  16  *      with the distribution.
  17  *
  18  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  19  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  20  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  21  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
  22  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  23  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
  24  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  25  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
  26  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  28  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  29  * POSSIBILITY OF SUCH DAMAGE.
  30  *
  31  * Effort sponsored in part by the Defense Advanced Research Projects
  32  * Agency (DARPA) and Air Force Research Laboratory, Air Force
  33  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
  34  *
  35  *      $OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
  36  */
  37
  38 #include <sys/cdefs.h>
  39 __FBSDID("$FreeBSD$");
  40
  41 #include "opt_inet.h"
  42 #include "opt_inet6.h"
  43 #include "opt_bpf.h"
  44 #include "opt_pf.h"
  45
  46 #include <sys/param.h>
  47 #include <sys/bus.h>
  48 #include <sys/conf.h>
  49 #include <sys/endian.h>
  50 #include <sys/fcntl.h>
  51 #include <sys/filio.h>
  52 #include <sys/interrupt.h>
  53 #include <sys/jail.h>
  54 #include <sys/kernel.h>
  55 #include <sys/kthread.h>
  56 #include <sys/mbuf.h>
  57 #include <sys/module.h>
  58 #include <sys/proc.h>
  59 #include <sys/smp.h>
  60 #include <sys/socket.h>
  61 #include <sys/sysctl.h>
  62 #include <sys/md5.h>
  63 #include <sys/ucred.h>
  64
  65 #include <net/if.h>
  66 #include <net/route.h>
  67 #include <net/pfil.h>
  68 #include <net/pfvar.h>
  69 #include <net/if_pfsync.h>
  70 #include <net/if_pflog.h>
  71
  72 #include <netinet/in.h>
  73 #include <netinet/ip.h>
  74 #include <netinet/ip_var.h>
  75 #include <netinet/ip_icmp.h>
  76
  77 #ifdef INET6
  78 #include <netinet/ip6.h>
  79 #endif /* INET6 */
  80
  81 #ifdef ALTQ
  82 #include <altq/altq.h>
  83 #endif
  84
  85 static int               pfattach(void);
  86 static struct pf_pool   *pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
  87                             u_int8_t, u_int8_t, u_int8_t);
  88
  89 static void              pf_mv_pool(struct pf_palist *, struct pf_palist *);
  90 static void              pf_empty_pool(struct pf_palist *);
  91 static int               pfioctl(struct cdev *, u_long, caddr_t, int,
  92                             struct thread *);
  93 #ifdef ALTQ
  94 static int               pf_begin_altq(u_int32_t *);
  95 static int               pf_rollback_altq(u_int32_t);
  96 static int               pf_commit_altq(u_int32_t);
  97 static int               pf_enable_altq(struct pf_altq *);
  98 static int               pf_disable_altq(struct pf_altq *);
  99 static u_int32_t         pf_qname2qid(char *);
 100 static void              pf_qid_unref(u_int32_t);
 101 #endif /* ALTQ */
 102 static int               pf_begin_rules(u_int32_t *, int, const char *);
 103 static int               pf_rollback_rules(u_int32_t, int, char *);
 104 static int               pf_setup_pfsync_matching(struct pf_ruleset *);
 105 static void              pf_hash_rule(MD5_CTX *, struct pf_rule *);
 106 static void              pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
 107 static int               pf_commit_rules(u_int32_t, int, char *);
 108 static int               pf_addr_setup(struct pf_ruleset *,
 109                             struct pf_addr_wrap *, sa_family_t);
 110 static void              pf_addr_copyout(struct pf_addr_wrap *);
 111
 112 VNET_DEFINE(struct pf_rule,     pf_default_rule);
 113
 114 #ifdef ALTQ
 115 static VNET_DEFINE(int,         pf_altq_running);
 116 #define V_pf_altq_running       VNET(pf_altq_running)
 117 #endif
 118
 119 #define TAGID_MAX        50000
 120 struct pf_tagname {
 121         TAILQ_ENTRY(pf_tagname) entries;
 122         char                    name[PF_TAG_NAME_SIZE];
 123         uint16_t                tag;
 124         int                     ref;
 125 };
 126
 127 TAILQ_HEAD(pf_tags, pf_tagname);
 128 #define V_pf_tags               VNET(pf_tags)
 129 VNET_DEFINE(struct pf_tags, pf_tags);
 130 #define V_pf_qids               VNET(pf_qids)
 131 VNET_DEFINE(struct pf_tags, pf_qids);
 132 static MALLOC_DEFINE(M_PFTAG, "pf_tag", "pf(4) tag names");
 133 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
 134 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
 135
 136 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
 137 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
 138 #endif
 139
 140 static u_int16_t         tagname2tag(struct pf_tags *, char *);
 141 static u_int16_t         pf_tagname2tag(char *);
 142 static void              tag_unref(struct pf_tags *, u_int16_t);
 143
 144 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
 145
 146 struct cdev *pf_dev;
 147
 148 /*
 149  * XXX - These are new and need to be checked when moveing to a new version
 150  */
 151 static void              pf_clear_states(void);
 152 static int               pf_clear_tables(void);
 153 static void              pf_clear_srcnodes(struct pf_src_node *);
 154 static void              pf_kill_srcnodes(struct pfioc_src_node_kill *);
 155 static void              pf_tbladdr_copyout(struct pf_addr_wrap *);
 156
 157 /*
 158  * Wrapper functions for pfil(9) hooks
 159  */
 160 #ifdef INET
 161 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
 162     int dir, struct inpcb *inp);
 163 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
 164     int dir, struct inpcb *inp);
 165 #endif
 166 #ifdef INET6
 167 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
 168     int dir, struct inpcb *inp);
 169 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
 170     int dir, struct inpcb *inp);
 171 #endif
 172
 173 static int              hook_pf(void);
 174 static int              dehook_pf(void);
 175 static int              shutdown_pf(void);
 176 static int              pf_load(void);
 177 static int              pf_unload(void);
 178
 179 static struct cdevsw pf_cdevsw = {
 180         .d_ioctl =      pfioctl,
 181         .d_name =       PF_NAME,
 182         .d_version =    D_VERSION,
 183 };
 184
 185 static volatile VNET_DEFINE(int, pf_pfil_hooked);
 186 #define V_pf_pfil_hooked        VNET(pf_pfil_hooked)
 187 VNET_DEFINE(int,                pf_end_threads);
 188
 189 struct rwlock                   pf_rules_lock;
 190
 191 /* pfsync */
 192 pfsync_state_import_t           *pfsync_state_import_ptr = NULL;
 193 pfsync_insert_state_t           *pfsync_insert_state_ptr = NULL;
 194 pfsync_update_state_t           *pfsync_update_state_ptr = NULL;
 195 pfsync_delete_state_t           *pfsync_delete_state_ptr = NULL;
 196 pfsync_clear_states_t           *pfsync_clear_states_ptr = NULL;
 197 pfsync_defer_t                  *pfsync_defer_ptr = NULL;
 198 /* pflog */
 199 pflog_packet_t                  *pflog_packet_ptr = NULL;
 200
 201 static int
 202 pfattach(void)
 203 {
 204         u_int32_t *my_timeout = V_pf_default_rule.timeout;
 205         int error;
 206
 207         if (IS_DEFAULT_VNET(curvnet))
 208                 pf_mtag_initialize();
 209         pf_initialize();
 210         pfr_initialize();
 211         pfi_initialize();
 212         pf_normalize_init();
 213
 214         V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
 215         V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
 216
 217         RB_INIT(&V_pf_anchors);
 218         pf_init_ruleset(&pf_main_ruleset);
 219
 220         /* default rule should never be garbage collected */
 221         V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
 222 #ifdef PF_DEFAULT_TO_DROP
 223         V_pf_default_rule.action = PF_DROP;
 224 #else
 225         V_pf_default_rule.action = PF_PASS;
 226 #endif
 227         V_pf_default_rule.nr = -1;
 228         V_pf_default_rule.rtableid = -1;
 229
 230         V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
 231         V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
 232         V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
 233
 234         /* initialize default timeouts */
 235         my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
 236         my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
 237         my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
 238         my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
 239         my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
 240         my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
 241         my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
 242         my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
 243         my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
 244         my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
 245         my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
 246         my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
 247         my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
 248         my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
 249         my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
 250         my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
 251         my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
 252         my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
 253         my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
 254         my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
 255
 256         bzero(&V_pf_status, sizeof(V_pf_status));
 257         V_pf_status.debug = PF_DEBUG_URGENT;
 258
 259         V_pf_pfil_hooked = 0;
 260
 261         /* XXX do our best to avoid a conflict */
 262         V_pf_status.hostid = arc4random();
 263
 264         for (int i = 0; i < PFRES_MAX; i++)
 265                 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
 266         for (int i = 0; i < LCNT_MAX; i++)
 267                 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
 268         for (int i = 0; i < FCNT_MAX; i++)
 269                 V_pf_status.fcounters[i] = counter_u64_alloc(M_WAITOK);
 270         for (int i = 0; i < SCNT_MAX; i++)
 271                 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
 272
 273         if ((error = kproc_create(pf_purge_thread, curvnet, NULL, 0, 0,
 274             "pf purge")) != 0)
 275                 /* XXXGL: leaked all above. */
 276                 return (error);
 277         if ((error = swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
 278             INTR_MPSAFE, &V_pf_swi_cookie)) != 0)
 279                 /* XXXGL: leaked all above. */
 280                 return (error);
 281
 282         return (0);
 283 }
 284
 285 static struct pf_pool *
 286 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
 287     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
 288     u_int8_t check_ticket)
 289 {
 290         struct pf_ruleset       *ruleset;
 291         struct pf_rule          *rule;
 292         int                      rs_num;
 293
 294         ruleset = pf_find_ruleset(anchor);
 295         if (ruleset == NULL)
 296                 return (NULL);
 297         rs_num = pf_get_ruleset_number(rule_action);
 298         if (rs_num >= PF_RULESET_MAX)
 299                 return (NULL);
 300         if (active) {
 301                 if (check_ticket && ticket !=
 302                     ruleset->rules[rs_num].active.ticket)
 303                         return (NULL);
 304                 if (r_last)
 305                         rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
 306                             pf_rulequeue);
 307                 else
 308                         rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
 309         } else {
 310                 if (check_ticket && ticket !=
 311                     ruleset->rules[rs_num].inactive.ticket)
 312                         return (NULL);
 313                 if (r_last)
 314                         rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
 315                             pf_rulequeue);
 316                 else
 317                         rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
 318         }
 319         if (!r_last) {
 320                 while ((rule != NULL) && (rule->nr != rule_number))
 321                         rule = TAILQ_NEXT(rule, entries);
 322         }
 323         if (rule == NULL)
 324                 return (NULL);
 325
 326         return (&rule->rpool);
 327 }
 328
 329 static void
 330 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
 331 {
 332         struct pf_pooladdr      *mv_pool_pa;
 333
 334         while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
 335                 TAILQ_REMOVE(poola, mv_pool_pa, entries);
 336                 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
 337         }
 338 }
 339
 340 static void
 341 pf_empty_pool(struct pf_palist *poola)
 342 {
 343         struct pf_pooladdr *pa;
 344
 345         while ((pa = TAILQ_FIRST(poola)) != NULL) {
 346                 switch (pa->addr.type) {
 347                 case PF_ADDR_DYNIFTL:
 348                         pfi_dynaddr_remove(pa->addr.p.dyn);
 349                         break;
 350                 case PF_ADDR_TABLE:
 351                         pfr_detach_table(pa->addr.p.tbl);
 352                         break;
 353                 }
 354                 if (pa->kif)
 355                         pfi_kif_unref(pa->kif);
 356                 TAILQ_REMOVE(poola, pa, entries);
 357                 free(pa, M_PFRULE);
 358         }
 359 }
 360
 361 static void
 362 pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
 363 {
 364
 365         PF_RULES_WASSERT();
 366
 367         TAILQ_REMOVE(rulequeue, rule, entries);
 368
 369         PF_UNLNKDRULES_LOCK();
 370         rule->rule_flag |= PFRULE_REFS;
 371         TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
 372         PF_UNLNKDRULES_UNLOCK();
 373 }
 374
 375 void
 376 pf_free_rule(struct pf_rule *rule)
 377 {
 378
 379         PF_RULES_WASSERT();
 380
 381         if (rule->tag)
 382                 tag_unref(&V_pf_tags, rule->tag);
 383         if (rule->match_tag)
 384                 tag_unref(&V_pf_tags, rule->match_tag);
 385 #ifdef ALTQ
 386         if (rule->pqid != rule->qid)
 387                 pf_qid_unref(rule->pqid);
 388         pf_qid_unref(rule->qid);
 389 #endif
 390         switch (rule->src.addr.type) {
 391         case PF_ADDR_DYNIFTL:
 392                 pfi_dynaddr_remove(rule->src.addr.p.dyn);
 393                 break;
 394         case PF_ADDR_TABLE:
 395                 pfr_detach_table(rule->src.addr.p.tbl);
 396                 break;
 397         }
 398         switch (rule->dst.addr.type) {
 399         case PF_ADDR_DYNIFTL:
 400                 pfi_dynaddr_remove(rule->dst.addr.p.dyn);
 401                 break;
 402         case PF_ADDR_TABLE:
 403                 pfr_detach_table(rule->dst.addr.p.tbl);
 404                 break;
 405         }
 406         if (rule->overload_tbl)
 407                 pfr_detach_table(rule->overload_tbl);
 408         if (rule->kif)
 409                 pfi_kif_unref(rule->kif);
 410         pf_anchor_remove(rule);
 411         pf_empty_pool(&rule->rpool.list);
 412         counter_u64_free(rule->states_cur);
 413         counter_u64_free(rule->states_tot);
 414         counter_u64_free(rule->src_nodes);
 415         free(rule, M_PFRULE);
 416 }
 417
 418 static u_int16_t
 419 tagname2tag(struct pf_tags *head, char *tagname)
 420 {
 421         struct pf_tagname       *tag, *p = NULL;
 422         u_int16_t                new_tagid = 1;
 423
 424         PF_RULES_WASSERT();
 425
 426         TAILQ_FOREACH(tag, head, entries)
 427                 if (strcmp(tagname, tag->name) == 0) {
 428                         tag->ref++;
 429                         return (tag->tag);
 430                 }
 431
 432         /*
 433          * to avoid fragmentation, we do a linear search from the beginning
 434          * and take the first free slot we find. if there is none or the list
 435          * is empty, append a new entry at the end.
 436          */
 437
 438         /* new entry */
 439         if (!TAILQ_EMPTY(head))
 440                 for (p = TAILQ_FIRST(head); p != NULL &&
 441                     p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
 442                         new_tagid = p->tag + 1;
 443
 444         if (new_tagid > TAGID_MAX)
 445                 return (0);
 446
 447         /* allocate and fill new struct pf_tagname */
 448         tag = malloc(sizeof(*tag), M_PFTAG, M_NOWAIT|M_ZERO);
 449         if (tag == NULL)
 450                 return (0);
 451         strlcpy(tag->name, tagname, sizeof(tag->name));
 452         tag->tag = new_tagid;
 453         tag->ref++;
 454
 455         if (p != NULL)  /* insert new entry before p */
 456                 TAILQ_INSERT_BEFORE(p, tag, entries);
 457         else    /* either list empty or no free slot in between */
 458                 TAILQ_INSERT_TAIL(head, tag, entries);
 459
 460         return (tag->tag);
 461 }
 462
 463 static void
 464 tag_unref(struct pf_tags *head, u_int16_t tag)
 465 {
 466         struct pf_tagname       *p, *next;
 467
 468         PF_RULES_WASSERT();
 469
 470         for (p = TAILQ_FIRST(head); p != NULL; p = next) {
 471                 next = TAILQ_NEXT(p, entries);
 472                 if (tag == p->tag) {
 473                         if (--p->ref == 0) {
 474                                 TAILQ_REMOVE(head, p, entries);
 475                                 free(p, M_PFTAG);
 476                         }
 477                         break;
 478                 }
 479         }
 480 }
 481
 482 static u_int16_t
 483 pf_tagname2tag(char *tagname)
 484 {
 485         return (tagname2tag(&V_pf_tags, tagname));
 486 }
 487
 488 #ifdef ALTQ
 489 static u_int32_t
 490 pf_qname2qid(char *qname)
 491 {
 492         return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
 493 }
 494
 495 static void
 496 pf_qid_unref(u_int32_t qid)
 497 {
 498         tag_unref(&V_pf_qids, (u_int16_t)qid);
 499 }
 500
 501 static int
 502 pf_begin_altq(u_int32_t *ticket)
 503 {
 504         struct pf_altq  *altq;
 505         int              error = 0;
 506
 507         PF_RULES_WASSERT();
 508
 509         /* Purge the old altq list */
 510         while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
 511                 TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
 512                 if (altq->qname[0] == 0 &&
 513                     (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 514                         /* detach and destroy the discipline */
 515                         error = altq_remove(altq);
 516                 } else
 517                         pf_qid_unref(altq->qid);
 518                 free(altq, M_PFALTQ);
 519         }
 520         if (error)
 521                 return (error);
 522         *ticket = ++V_ticket_altqs_inactive;
 523         V_altqs_inactive_open = 1;
 524         return (0);
 525 }
 526
 527 static int
 528 pf_rollback_altq(u_int32_t ticket)
 529 {
 530         struct pf_altq  *altq;
 531         int              error = 0;
 532
 533         PF_RULES_WASSERT();
 534
 535         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 536                 return (0);
 537         /* Purge the old altq list */
 538         while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
 539                 TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
 540                 if (altq->qname[0] == 0 &&
 541                    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 542                         /* detach and destroy the discipline */
 543                         error = altq_remove(altq);
 544                 } else
 545                         pf_qid_unref(altq->qid);
 546                 free(altq, M_PFALTQ);
 547         }
 548         V_altqs_inactive_open = 0;
 549         return (error);
 550 }
 551
 552 static int
 553 pf_commit_altq(u_int32_t ticket)
 554 {
 555         struct pf_altqqueue     *old_altqs;
 556         struct pf_altq          *altq;
 557         int                      err, error = 0;
 558
 559         PF_RULES_WASSERT();
 560
 561         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 562                 return (EBUSY);
 563
 564         /* swap altqs, keep the old. */
 565         old_altqs = V_pf_altqs_active;
 566         V_pf_altqs_active = V_pf_altqs_inactive;
 567         V_pf_altqs_inactive = old_altqs;
 568         V_ticket_altqs_active = V_ticket_altqs_inactive;
 569
 570         /* Attach new disciplines */
 571         TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
 572         if (altq->qname[0] == 0 &&
 573            (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 574                         /* attach the discipline */
 575                         error = altq_pfattach(altq);
 576                         if (error == 0 && V_pf_altq_running)
 577                                 error = pf_enable_altq(altq);
 578                         if (error != 0)
 579                                 return (error);
 580                 }
 581         }
 582
 583         /* Purge the old altq list */
 584         while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
 585                 TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
 586                 if (altq->qname[0] == 0 &&
 587                     (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 588                         /* detach and destroy the discipline */
 589                         if (V_pf_altq_running)
 590                                 error = pf_disable_altq(altq);
 591                         err = altq_pfdetach(altq);
 592                         if (err != 0 && error == 0)
 593                                 error = err;
 594                         err = altq_remove(altq);
 595                         if (err != 0 && error == 0)
 596                                 error = err;
 597                 } else
 598                         pf_qid_unref(altq->qid);
 599                 free(altq, M_PFALTQ);
 600         }
 601
 602         V_altqs_inactive_open = 0;
 603         return (error);
 604 }
 605
 606 static int
 607 pf_enable_altq(struct pf_altq *altq)
 608 {
 609         struct ifnet            *ifp;
 610         struct tb_profile        tb;
 611         int                      error = 0;
 612
 613         if ((ifp = ifunit(altq->ifname)) == NULL)
 614                 return (EINVAL);
 615
 616         if (ifp->if_snd.altq_type != ALTQT_NONE)
 617                 error = altq_enable(&ifp->if_snd);
 618
 619         /* set tokenbucket regulator */
 620         if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
 621                 tb.rate = altq->ifbandwidth;
 622                 tb.depth = altq->tbrsize;
 623                 error = tbr_set(&ifp->if_snd, &tb);
 624         }
 625
 626         return (error);
 627 }
 628
 629 static int
 630 pf_disable_altq(struct pf_altq *altq)
 631 {
 632         struct ifnet            *ifp;
 633         struct tb_profile        tb;
 634         int                      error;
 635
 636         if ((ifp = ifunit(altq->ifname)) == NULL)
 637                 return (EINVAL);
 638
 639         /*
 640          * when the discipline is no longer referenced, it was overridden
 641          * by a new one.  if so, just return.
 642          */
 643         if (altq->altq_disc != ifp->if_snd.altq_disc)
 644                 return (0);
 645
 646         error = altq_disable(&ifp->if_snd);
 647
 648         if (error == 0) {
 649                 /* clear tokenbucket regulator */
 650                 tb.rate = 0;
 651                 error = tbr_set(&ifp->if_snd, &tb);
 652         }
 653
 654         return (error);
 655 }
 656
 657 void
 658 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
 659 {
 660         struct ifnet    *ifp1;
 661         struct pf_altq  *a1, *a2, *a3;
 662         u_int32_t        ticket;
 663         int              error = 0;
 664
 665         /* Interrupt userland queue modifications */
 666         if (V_altqs_inactive_open)
 667                 pf_rollback_altq(V_ticket_altqs_inactive);
 668
 669         /* Start new altq ruleset */
 670         if (pf_begin_altq(&ticket))
 671                 return;
 672
 673         /* Copy the current active set */
 674         TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
 675                 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
 676                 if (a2 == NULL) {
 677                         error = ENOMEM;
 678                         break;
 679                 }
 680                 bcopy(a1, a2, sizeof(struct pf_altq));
 681
 682                 if (a2->qname[0] != 0) {
 683                         if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
 684                                 error = EBUSY;
 685                                 free(a2, M_PFALTQ);
 686                                 break;
 687                         }
 688                         a2->altq_disc = NULL;
 689                         TAILQ_FOREACH(a3, V_pf_altqs_inactive, entries) {
 690                                 if (strncmp(a3->ifname, a2->ifname,
 691                                     IFNAMSIZ) == 0 && a3->qname[0] == 0) {
 692                                         a2->altq_disc = a3->altq_disc;
 693                                         break;
 694                                 }
 695                         }
 696                 }
 697                 /* Deactivate the interface in question */
 698                 a2->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
 699                 if ((ifp1 = ifunit(a2->ifname)) == NULL ||
 700                     (remove && ifp1 == ifp)) {
 701                         a2->local_flags |= PFALTQ_FLAG_IF_REMOVED;
 702                 } else {
 703                         error = altq_add(a2);
 704
 705                         if (ticket != V_ticket_altqs_inactive)
 706                                 error = EBUSY;
 707
 708                         if (error) {
 709                                 free(a2, M_PFALTQ);
 710                                 break;
 711                         }
 712                 }
 713
 714                 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
 715         }
 716
 717         if (error != 0)
 718                 pf_rollback_altq(ticket);
 719         else
 720                 pf_commit_altq(ticket);
 721 }
 722 #endif /* ALTQ */
 723
 724 static int
 725 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
 726 {
 727         struct pf_ruleset       *rs;
 728         struct pf_rule          *rule;
 729
 730         PF_RULES_WASSERT();
 731
 732         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 733                 return (EINVAL);
 734         rs = pf_find_or_create_ruleset(anchor);
 735         if (rs == NULL)
 736                 return (EINVAL);
 737         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 738                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 739                 rs->rules[rs_num].inactive.rcount--;
 740         }
 741         *ticket = ++rs->rules[rs_num].inactive.ticket;
 742         rs->rules[rs_num].inactive.open = 1;
 743         return (0);
 744 }
 745
 746 static int
 747 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
 748 {
 749         struct pf_ruleset       *rs;
 750         struct pf_rule          *rule;
 751
 752         PF_RULES_WASSERT();
 753
 754         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 755                 return (EINVAL);
 756         rs = pf_find_ruleset(anchor);
 757         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 758             rs->rules[rs_num].inactive.ticket != ticket)
 759                 return (0);
 760         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 761                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 762                 rs->rules[rs_num].inactive.rcount--;
 763         }
 764         rs->rules[rs_num].inactive.open = 0;
 765         return (0);
 766 }
 767
 768 #define PF_MD5_UPD(st, elm)                                             \
 769                 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
 770
 771 #define PF_MD5_UPD_STR(st, elm)                                         \
 772                 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
 773
 774 #define PF_MD5_UPD_HTONL(st, elm, stor) do {                            \
 775                 (stor) = htonl((st)->elm);                              \
 776                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
 777 } while (0)
 778
 779 #define PF_MD5_UPD_HTONS(st, elm, stor) do {                            \
 780                 (stor) = htons((st)->elm);                              \
 781                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
 782 } while (0)
 783
 784 static void
 785 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
 786 {
 787         PF_MD5_UPD(pfr, addr.type);
 788         switch (pfr->addr.type) {
 789                 case PF_ADDR_DYNIFTL:
 790                         PF_MD5_UPD(pfr, addr.v.ifname);
 791                         PF_MD5_UPD(pfr, addr.iflags);
 792                         break;
 793                 case PF_ADDR_TABLE:
 794                         PF_MD5_UPD(pfr, addr.v.tblname);
 795                         break;
 796                 case PF_ADDR_ADDRMASK:
 797                         /* XXX ignore af? */
 798                         PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
 799                         PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
 800                         break;
 801         }
 802
 803         PF_MD5_UPD(pfr, port[0]);
 804         PF_MD5_UPD(pfr, port[1]);
 805         PF_MD5_UPD(pfr, neg);
 806         PF_MD5_UPD(pfr, port_op);
 807 }
 808
 809 static void
 810 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
 811 {
 812         u_int16_t x;
 813         u_int32_t y;
 814
 815         pf_hash_rule_addr(ctx, &rule->src);
 816         pf_hash_rule_addr(ctx, &rule->dst);
 817         PF_MD5_UPD_STR(rule, label);
 818         PF_MD5_UPD_STR(rule, ifname);
 819         PF_MD5_UPD_STR(rule, match_tagname);
 820         PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
 821         PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
 822         PF_MD5_UPD_HTONL(rule, prob, y);
 823         PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
 824         PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
 825         PF_MD5_UPD(rule, uid.op);
 826         PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
 827         PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
 828         PF_MD5_UPD(rule, gid.op);
 829         PF_MD5_UPD_HTONL(rule, rule_flag, y);
 830         PF_MD5_UPD(rule, action);
 831         PF_MD5_UPD(rule, direction);
 832         PF_MD5_UPD(rule, af);
 833         PF_MD5_UPD(rule, quick);
 834         PF_MD5_UPD(rule, ifnot);
 835         PF_MD5_UPD(rule, match_tag_not);
 836         PF_MD5_UPD(rule, natpass);
 837         PF_MD5_UPD(rule, keep_state);
 838         PF_MD5_UPD(rule, proto);
 839         PF_MD5_UPD(rule, type);
 840         PF_MD5_UPD(rule, code);
 841         PF_MD5_UPD(rule, flags);
 842         PF_MD5_UPD(rule, flagset);
 843         PF_MD5_UPD(rule, allow_opts);
 844         PF_MD5_UPD(rule, rt);
 845         PF_MD5_UPD(rule, tos);
 846 }
 847
 848 static int
 849 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
 850 {
 851         struct pf_ruleset       *rs;
 852         struct pf_rule          *rule, **old_array;
 853         struct pf_rulequeue     *old_rules;
 854         int                      error;
 855         u_int32_t                old_rcount;
 856
 857         PF_RULES_WASSERT();
 858
 859         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 860                 return (EINVAL);
 861         rs = pf_find_ruleset(anchor);
 862         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 863             ticket != rs->rules[rs_num].inactive.ticket)
 864                 return (EBUSY);
 865
 866         /* Calculate checksum for the main ruleset */
 867         if (rs == &pf_main_ruleset) {
 868                 error = pf_setup_pfsync_matching(rs);
 869                 if (error != 0)
 870                         return (error);
 871         }
 872
 873         /* Swap rules, keep the old. */
 874         old_rules = rs->rules[rs_num].active.ptr;
 875         old_rcount = rs->rules[rs_num].active.rcount;
 876         old_array = rs->rules[rs_num].active.ptr_array;
 877
 878         rs->rules[rs_num].active.ptr =
 879             rs->rules[rs_num].inactive.ptr;
 880         rs->rules[rs_num].active.ptr_array =
 881             rs->rules[rs_num].inactive.ptr_array;
 882         rs->rules[rs_num].active.rcount =
 883             rs->rules[rs_num].inactive.rcount;
 884         rs->rules[rs_num].inactive.ptr = old_rules;
 885         rs->rules[rs_num].inactive.ptr_array = old_array;
 886         rs->rules[rs_num].inactive.rcount = old_rcount;
 887
 888         rs->rules[rs_num].active.ticket =
 889             rs->rules[rs_num].inactive.ticket;
 890         pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
 891
 892
 893         /* Purge the old rule list. */
 894         while ((rule = TAILQ_FIRST(old_rules)) != NULL)
 895                 pf_unlink_rule(old_rules, rule);
 896         if (rs->rules[rs_num].inactive.ptr_array)
 897                 free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
 898         rs->rules[rs_num].inactive.ptr_array = NULL;
 899         rs->rules[rs_num].inactive.rcount = 0;
 900         rs->rules[rs_num].inactive.open = 0;
 901         pf_remove_if_empty_ruleset(rs);
 902
 903         return (0);
 904 }
 905
 906 static int
 907 pf_setup_pfsync_matching(struct pf_ruleset *rs)
 908 {
 909         MD5_CTX                  ctx;
 910         struct pf_rule          *rule;
 911         int                      rs_cnt;
 912         u_int8_t                 digest[PF_MD5_DIGEST_LENGTH];
 913
 914         MD5Init(&ctx);
 915         for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
 916                 /* XXX PF_RULESET_SCRUB as well? */
 917                 if (rs_cnt == PF_RULESET_SCRUB)
 918                         continue;
 919
 920                 if (rs->rules[rs_cnt].inactive.ptr_array)
 921                         free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
 922                 rs->rules[rs_cnt].inactive.ptr_array = NULL;
 923
 924                 if (rs->rules[rs_cnt].inactive.rcount) {
 925                         rs->rules[rs_cnt].inactive.ptr_array =
 926                             malloc(sizeof(caddr_t) *
 927                             rs->rules[rs_cnt].inactive.rcount,
 928                             M_TEMP, M_NOWAIT);
 929
 930                         if (!rs->rules[rs_cnt].inactive.ptr_array)
 931                                 return (ENOMEM);
 932                 }
 933
 934                 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
 935                     entries) {
 936                         pf_hash_rule(&ctx, rule);
 937                         (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
 938                 }
 939         }
 940
 941         MD5Final(digest, &ctx);
 942         memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
 943         return (0);
 944 }
 945
 946 static int
 947 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
 948     sa_family_t af)
 949 {
 950         int error = 0;
 951
 952         switch (addr->type) {
 953         case PF_ADDR_TABLE:
 954                 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
 955                 if (addr->p.tbl == NULL)
 956                         error = ENOMEM;
 957                 break;
 958         case PF_ADDR_DYNIFTL:
 959                 error = pfi_dynaddr_setup(addr, af);
 960                 break;
 961         }
 962
 963         return (error);
 964 }
 965
 966 static void
 967 pf_addr_copyout(struct pf_addr_wrap *addr)
 968 {
 969
 970         switch (addr->type) {
 971         case PF_ADDR_DYNIFTL:
 972                 pfi_dynaddr_copyout(addr);
 973                 break;
 974         case PF_ADDR_TABLE:
 975                 pf_tbladdr_copyout(addr);
 976                 break;
 977         }
 978 }
 979
 980 static int
 981 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
 982 {
 983         int                      error = 0;
 984
 985         /* XXX keep in sync with switch() below */
 986         if (securelevel_gt(td->td_ucred, 2))
 987                 switch (cmd) {
 988                 case DIOCGETRULES:
 989                 case DIOCGETRULE:
 990                 case DIOCGETADDRS:
 991                 case DIOCGETADDR:
 992                 case DIOCGETSTATE:
 993                 case DIOCSETSTATUSIF:
 994                 case DIOCGETSTATUS:
 995                 case DIOCCLRSTATUS:
 996                 case DIOCNATLOOK:
 997                 case DIOCSETDEBUG:
 998                 case DIOCGETSTATES:
 999                 case DIOCGETTIMEOUT:
1000                 case DIOCCLRRULECTRS:
1001                 case DIOCGETLIMIT:
1002                 case DIOCGETALTQS:
1003                 case DIOCGETALTQ:
1004                 case DIOCGETQSTATS:
1005                 case DIOCGETRULESETS:
1006                 case DIOCGETRULESET:
1007                 case DIOCRGETTABLES:
1008                 case DIOCRGETTSTATS:
1009                 case DIOCRCLRTSTATS:
1010                 case DIOCRCLRADDRS:
1011                 case DIOCRADDADDRS:
1012                 case DIOCRDELADDRS:
1013                 case DIOCRSETADDRS:
1014                 case DIOCRGETADDRS:
1015                 case DIOCRGETASTATS:
1016                 case DIOCRCLRASTATS:
1017                 case DIOCRTSTADDRS:
1018                 case DIOCOSFPGET:
1019                 case DIOCGETSRCNODES:
1020                 case DIOCCLRSRCNODES:
1021                 case DIOCIGETIFACES:
1022                 case DIOCGIFSPEED:
1023                 case DIOCSETIFFLAG:
1024                 case DIOCCLRIFFLAG:
1025                         break;
1026                 case DIOCRCLRTABLES:
1027                 case DIOCRADDTABLES:
1028                 case DIOCRDELTABLES:
1029                 case DIOCRSETTFLAGS:
1030                         if (((struct pfioc_table *)addr)->pfrio_flags &
1031                             PFR_FLAG_DUMMY)
1032                                 break; /* dummy operation ok */
1033                         return (EPERM);
1034                 default:
1035                         return (EPERM);
1036                 }
1037
1038         if (!(flags & FWRITE))
1039                 switch (cmd) {
1040                 case DIOCGETRULES:
1041                 case DIOCGETADDRS:
1042                 case DIOCGETADDR:
1043                 case DIOCGETSTATE:
1044                 case DIOCGETSTATUS:
1045                 case DIOCGETSTATES:
1046                 case DIOCGETTIMEOUT:
1047                 case DIOCGETLIMIT:
1048                 case DIOCGETALTQS:
1049                 case DIOCGETALTQ:
1050                 case DIOCGETQSTATS:
1051                 case DIOCGETRULESETS:
1052                 case DIOCGETRULESET:
1053                 case DIOCNATLOOK:
1054                 case DIOCRGETTABLES:
1055                 case DIOCRGETTSTATS:
1056                 case DIOCRGETADDRS:
1057                 case DIOCRGETASTATS:
1058                 case DIOCRTSTADDRS:
1059                 case DIOCOSFPGET:
1060                 case DIOCGETSRCNODES:
1061                 case DIOCIGETIFACES:
1062                 case DIOCGIFSPEED:
1063                         break;
1064                 case DIOCRCLRTABLES:
1065                 case DIOCRADDTABLES:
1066                 case DIOCRDELTABLES:
1067                 case DIOCRCLRTSTATS:
1068                 case DIOCRCLRADDRS:
1069                 case DIOCRADDADDRS:
1070                 case DIOCRDELADDRS:
1071                 case DIOCRSETADDRS:
1072                 case DIOCRSETTFLAGS:
1073                         if (((struct pfioc_table *)addr)->pfrio_flags &
1074                             PFR_FLAG_DUMMY) {
1075                                 flags |= FWRITE; /* need write lock for dummy */
1076                                 break; /* dummy operation ok */
1077                         }
1078                         return (EACCES);
1079                 case DIOCGETRULE:
1080                         if (((struct pfioc_rule *)addr)->action ==
1081                             PF_GET_CLR_CNTR)
1082                                 return (EACCES);
1083                         break;
1084                 default:
1085                         return (EACCES);
1086                 }
1087
1088         CURVNET_SET(TD_TO_VNET(td));
1089
1090         switch (cmd) {
1091         case DIOCSTART:
1092                 PF_RULES_WLOCK();
1093                 if (V_pf_status.running)
1094                         error = EEXIST;
1095                 else {
1096                         int cpu;
1097
1098                         PF_RULES_WUNLOCK();
1099                         error = hook_pf();
1100                         if (error) {
1101                                 DPFPRINTF(PF_DEBUG_MISC,
1102                                     ("pf: pfil registration failed\n"));
1103                                 break;
1104                         }
1105                         PF_RULES_WLOCK();
1106                         V_pf_status.running = 1;
1107                         V_pf_status.since = time_second;
1108
1109                         CPU_FOREACH(cpu)
1110                                 V_pf_stateid[cpu] = time_second;
1111
1112                         DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1113                 }
1114                 PF_RULES_WUNLOCK();
1115                 break;
1116
1117         case DIOCSTOP:
1118                 PF_RULES_WLOCK();
1119                 if (!V_pf_status.running)
1120                         error = ENOENT;
1121                 else {
1122                         V_pf_status.running = 0;
1123                         PF_RULES_WUNLOCK();
1124                         error = dehook_pf();
1125                         if (error) {
1126                                 V_pf_status.running = 1;
1127                                 DPFPRINTF(PF_DEBUG_MISC,
1128                                     ("pf: pfil unregistration failed\n"));
1129                         }
1130                         PF_RULES_WLOCK();
1131                         V_pf_status.since = time_second;
1132                         DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1133                 }
1134                 PF_RULES_WUNLOCK();
1135                 break;
1136
1137         case DIOCADDRULE: {
1138                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1139                 struct pf_ruleset       *ruleset;
1140                 struct pf_rule          *rule, *tail;
1141                 struct pf_pooladdr      *pa;
1142                 struct pfi_kif          *kif = NULL;
1143                 int                      rs_num;
1144
1145                 if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1146                         error = EINVAL;
1147                         break;
1148                 }
1149 #ifndef INET
1150                 if (pr->rule.af == AF_INET) {
1151                         error = EAFNOSUPPORT;
1152                         break;
1153                 }
1154 #endif /* INET */
1155 #ifndef INET6
1156                 if (pr->rule.af == AF_INET6) {
1157                         error = EAFNOSUPPORT;
1158                         break;
1159                 }
1160 #endif /* INET6 */
1161
1162                 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1163                 bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1164                 if (rule->ifname[0])
1165                         kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1166                 rule->states_cur = counter_u64_alloc(M_WAITOK);
1167                 rule->states_tot = counter_u64_alloc(M_WAITOK);
1168                 rule->src_nodes = counter_u64_alloc(M_WAITOK);
1169                 rule->cuid = td->td_ucred->cr_ruid;
1170                 rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1171                 TAILQ_INIT(&rule->rpool.list);
1172
1173 #define ERROUT(x)       { error = (x); goto DIOCADDRULE_error; }
1174
1175                 PF_RULES_WLOCK();
1176                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1177                 ruleset = pf_find_ruleset(pr->anchor);
1178                 if (ruleset == NULL)
1179                         ERROUT(EINVAL);
1180                 rs_num = pf_get_ruleset_number(pr->rule.action);
1181                 if (rs_num >= PF_RULESET_MAX)
1182                         ERROUT(EINVAL);
1183                 if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1184                         DPFPRINTF(PF_DEBUG_MISC,
1185                             ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1186                             ruleset->rules[rs_num].inactive.ticket));
1187                         ERROUT(EBUSY);
1188                 }
1189                 if (pr->pool_ticket != V_ticket_pabuf) {
1190                         DPFPRINTF(PF_DEBUG_MISC,
1191                             ("pool_ticket: %d != %d\n", pr->pool_ticket,
1192                             V_ticket_pabuf));
1193                         ERROUT(EBUSY);
1194                 }
1195
1196                 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1197                     pf_rulequeue);
1198                 if (tail)
1199                         rule->nr = tail->nr + 1;
1200                 else
1201                         rule->nr = 0;
1202                 if (rule->ifname[0]) {
1203                         rule->kif = pfi_kif_attach(kif, rule->ifname);
1204                         pfi_kif_ref(rule->kif);
1205                 } else
1206                         rule->kif = NULL;
1207
1208                 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1209                         error = EBUSY;
1210
1211 #ifdef ALTQ
1212                 /* set queue IDs */
1213                 if (rule->qname[0] != 0) {
1214                         if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1215                                 error = EBUSY;
1216                         else if (rule->pqname[0] != 0) {
1217                                 if ((rule->pqid =
1218                                     pf_qname2qid(rule->pqname)) == 0)
1219                                         error = EBUSY;
1220                         } else
1221                                 rule->pqid = rule->qid;
1222                 }
1223 #endif
1224                 if (rule->tagname[0])
1225                         if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1226                                 error = EBUSY;
1227                 if (rule->match_tagname[0])
1228                         if ((rule->match_tag =
1229                             pf_tagname2tag(rule->match_tagname)) == 0)
1230                                 error = EBUSY;
1231                 if (rule->rt && !rule->direction)
1232                         error = EINVAL;
1233                 if (!rule->log)
1234                         rule->logif = 0;
1235                 if (rule->logif >= PFLOGIFS_MAX)
1236                         error = EINVAL;
1237                 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1238                         error = ENOMEM;
1239                 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1240                         error = ENOMEM;
1241                 if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1242                         error = EINVAL;
1243                 TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1244                         if (pa->addr.type == PF_ADDR_TABLE) {
1245                                 pa->addr.p.tbl = pfr_attach_table(ruleset,
1246                                     pa->addr.v.tblname);
1247                                 if (pa->addr.p.tbl == NULL)
1248                                         error = ENOMEM;
1249                         }
1250
1251                 if (rule->overload_tblname[0]) {
1252                         if ((rule->overload_tbl = pfr_attach_table(ruleset,
1253                             rule->overload_tblname)) == NULL)
1254                                 error = EINVAL;
1255                         else
1256                                 rule->overload_tbl->pfrkt_flags |=
1257                                     PFR_TFLAG_ACTIVE;
1258                 }
1259
1260                 pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1261                 if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1262                     (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1263                     (rule->rt > PF_FASTROUTE)) &&
1264                     (TAILQ_FIRST(&rule->rpool.list) == NULL))
1265                         error = EINVAL;
1266
1267                 if (error) {
1268                         pf_free_rule(rule);
1269                         PF_RULES_WUNLOCK();
1270                         break;
1271                 }
1272
1273                 rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1274                 rule->evaluations = rule->packets[0] = rule->packets[1] =
1275                     rule->bytes[0] = rule->bytes[1] = 0;
1276                 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1277                     rule, entries);
1278                 ruleset->rules[rs_num].inactive.rcount++;
1279                 PF_RULES_WUNLOCK();
1280                 break;
1281
1282 #undef ERROUT
1283 DIOCADDRULE_error:
1284                 PF_RULES_WUNLOCK();
1285                 counter_u64_free(rule->states_cur);
1286                 counter_u64_free(rule->states_tot);
1287                 counter_u64_free(rule->src_nodes);
1288                 free(rule, M_PFRULE);
1289                 if (kif)
1290                         free(kif, PFI_MTYPE);
1291                 break;
1292         }
1293
1294         case DIOCGETRULES: {
1295                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1296                 struct pf_ruleset       *ruleset;
1297                 struct pf_rule          *tail;
1298                 int                      rs_num;
1299
1300                 PF_RULES_WLOCK();
1301                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1302                 ruleset = pf_find_ruleset(pr->anchor);
1303                 if (ruleset == NULL) {
1304                         PF_RULES_WUNLOCK();
1305                         error = EINVAL;
1306                         break;
1307                 }
1308                 rs_num = pf_get_ruleset_number(pr->rule.action);
1309                 if (rs_num >= PF_RULESET_MAX) {
1310                         PF_RULES_WUNLOCK();
1311                         error = EINVAL;
1312                         break;
1313                 }
1314                 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1315                     pf_rulequeue);
1316                 if (tail)
1317                         pr->nr = tail->nr + 1;
1318                 else
1319                         pr->nr = 0;
1320                 pr->ticket = ruleset->rules[rs_num].active.ticket;
1321                 PF_RULES_WUNLOCK();
1322                 break;
1323         }
1324
1325         case DIOCGETRULE: {
1326                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1327                 struct pf_ruleset       *ruleset;
1328                 struct pf_rule          *rule;
1329                 int                      rs_num, i;
1330
1331                 PF_RULES_WLOCK();
1332                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1333                 ruleset = pf_find_ruleset(pr->anchor);
1334                 if (ruleset == NULL) {
1335                         PF_RULES_WUNLOCK();
1336                         error = EINVAL;
1337                         break;
1338                 }
1339                 rs_num = pf_get_ruleset_number(pr->rule.action);
1340                 if (rs_num >= PF_RULESET_MAX) {
1341                         PF_RULES_WUNLOCK();
1342                         error = EINVAL;
1343                         break;
1344                 }
1345                 if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1346                         PF_RULES_WUNLOCK();
1347                         error = EBUSY;
1348                         break;
1349                 }
1350                 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1351                 while ((rule != NULL) && (rule->nr != pr->nr))
1352                         rule = TAILQ_NEXT(rule, entries);
1353                 if (rule == NULL) {
1354                         PF_RULES_WUNLOCK();
1355                         error = EBUSY;
1356                         break;
1357                 }
1358                 bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1359                 pr->rule.u_states_cur = counter_u64_fetch(rule->states_cur);
1360                 pr->rule.u_states_tot = counter_u64_fetch(rule->states_tot);
1361                 pr->rule.u_src_nodes = counter_u64_fetch(rule->src_nodes);
1362                 if (pf_anchor_copyout(ruleset, rule, pr)) {
1363                         PF_RULES_WUNLOCK();
1364                         error = EBUSY;
1365                         break;
1366                 }
1367                 pf_addr_copyout(&pr->rule.src.addr);
1368                 pf_addr_copyout(&pr->rule.dst.addr);
1369                 for (i = 0; i < PF_SKIP_COUNT; ++i)
1370                         if (rule->skip[i].ptr == NULL)
1371                                 pr->rule.skip[i].nr = -1;
1372                         else
1373                                 pr->rule.skip[i].nr =
1374                                     rule->skip[i].ptr->nr;
1375
1376                 if (pr->action == PF_GET_CLR_CNTR) {
1377                         rule->evaluations = 0;
1378                         rule->packets[0] = rule->packets[1] = 0;
1379                         rule->bytes[0] = rule->bytes[1] = 0;
1380                         counter_u64_zero(rule->states_tot);
1381                 }
1382                 PF_RULES_WUNLOCK();
1383                 break;
1384         }
1385
1386         case DIOCCHANGERULE: {
1387                 struct pfioc_rule       *pcr = (struct pfioc_rule *)addr;
1388                 struct pf_ruleset       *ruleset;
1389                 struct pf_rule          *oldrule = NULL, *newrule = NULL;
1390                 struct pfi_kif          *kif = NULL;
1391                 struct pf_pooladdr      *pa;
1392                 u_int32_t                nr = 0;
1393                 int                      rs_num;
1394
1395                 if (pcr->action < PF_CHANGE_ADD_HEAD ||
1396                     pcr->action > PF_CHANGE_GET_TICKET) {
1397                         error = EINVAL;
1398                         break;
1399                 }
1400                 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1401                         error = EINVAL;
1402                         break;
1403                 }
1404
1405                 if (pcr->action != PF_CHANGE_REMOVE) {
1406 #ifndef INET
1407                         if (pcr->rule.af == AF_INET) {
1408                                 error = EAFNOSUPPORT;
1409                                 break;
1410                         }
1411 #endif /* INET */
1412 #ifndef INET6
1413                         if (pcr->rule.af == AF_INET6) {
1414                                 error = EAFNOSUPPORT;
1415                                 break;
1416                         }
1417 #endif /* INET6 */
1418                         newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1419                         bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1420                         if (newrule->ifname[0])
1421                                 kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1422                         newrule->states_cur = counter_u64_alloc(M_WAITOK);
1423                         newrule->states_tot = counter_u64_alloc(M_WAITOK);
1424                         newrule->src_nodes = counter_u64_alloc(M_WAITOK);
1425                         newrule->cuid = td->td_ucred->cr_ruid;
1426                         newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1427                         TAILQ_INIT(&newrule->rpool.list);
1428                 }
1429
1430 #define ERROUT(x)       { error = (x); goto DIOCCHANGERULE_error; }
1431
1432                 PF_RULES_WLOCK();
1433                 if (!(pcr->action == PF_CHANGE_REMOVE ||
1434                     pcr->action == PF_CHANGE_GET_TICKET) &&
1435                     pcr->pool_ticket != V_ticket_pabuf)
1436                         ERROUT(EBUSY);
1437
1438                 ruleset = pf_find_ruleset(pcr->anchor);
1439                 if (ruleset == NULL)
1440                         ERROUT(EINVAL);
1441
1442                 rs_num = pf_get_ruleset_number(pcr->rule.action);
1443                 if (rs_num >= PF_RULESET_MAX)
1444                         ERROUT(EINVAL);
1445
1446                 if (pcr->action == PF_CHANGE_GET_TICKET) {
1447                         pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1448                         ERROUT(0);
1449                 } else if (pcr->ticket !=
1450                             ruleset->rules[rs_num].active.ticket)
1451                                 ERROUT(EINVAL);
1452
1453                 if (pcr->action != PF_CHANGE_REMOVE) {
1454                         if (newrule->ifname[0]) {
1455                                 newrule->kif = pfi_kif_attach(kif,
1456                                     newrule->ifname);
1457                                 pfi_kif_ref(newrule->kif);
1458                         } else
1459                                 newrule->kif = NULL;
1460
1461                         if (newrule->rtableid > 0 &&
1462                             newrule->rtableid >= rt_numfibs)
1463                                 error = EBUSY;
1464
1465 #ifdef ALTQ
1466                         /* set queue IDs */
1467                         if (newrule->qname[0] != 0) {
1468                                 if ((newrule->qid =
1469                                     pf_qname2qid(newrule->qname)) == 0)
1470                                         error = EBUSY;
1471                                 else if (newrule->pqname[0] != 0) {
1472                                         if ((newrule->pqid =
1473                                             pf_qname2qid(newrule->pqname)) == 0)
1474                                                 error = EBUSY;
1475                                 } else
1476                                         newrule->pqid = newrule->qid;
1477                         }
1478 #endif /* ALTQ */
1479                         if (newrule->tagname[0])
1480                                 if ((newrule->tag =
1481                                     pf_tagname2tag(newrule->tagname)) == 0)
1482                                         error = EBUSY;
1483                         if (newrule->match_tagname[0])
1484                                 if ((newrule->match_tag = pf_tagname2tag(
1485                                     newrule->match_tagname)) == 0)
1486                                         error = EBUSY;
1487                         if (newrule->rt && !newrule->direction)
1488                                 error = EINVAL;
1489                         if (!newrule->log)
1490                                 newrule->logif = 0;
1491                         if (newrule->logif >= PFLOGIFS_MAX)
1492                                 error = EINVAL;
1493                         if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1494                                 error = ENOMEM;
1495                         if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1496                                 error = ENOMEM;
1497                         if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1498                                 error = EINVAL;
1499                         TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1500                                 if (pa->addr.type == PF_ADDR_TABLE) {
1501                                         pa->addr.p.tbl =
1502                                             pfr_attach_table(ruleset,
1503                                             pa->addr.v.tblname);
1504                                         if (pa->addr.p.tbl == NULL)
1505                                                 error = ENOMEM;
1506                                 }
1507
1508                         if (newrule->overload_tblname[0]) {
1509                                 if ((newrule->overload_tbl = pfr_attach_table(
1510                                     ruleset, newrule->overload_tblname)) ==
1511                                     NULL)
1512                                         error = EINVAL;
1513                                 else
1514                                         newrule->overload_tbl->pfrkt_flags |=
1515                                             PFR_TFLAG_ACTIVE;
1516                         }
1517
1518                         pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1519                         if (((((newrule->action == PF_NAT) ||
1520                             (newrule->action == PF_RDR) ||
1521                             (newrule->action == PF_BINAT) ||
1522                             (newrule->rt > PF_FASTROUTE)) &&
1523                             !newrule->anchor)) &&
1524                             (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1525                                 error = EINVAL;
1526
1527                         if (error) {
1528                                 pf_free_rule(newrule);
1529                                 PF_RULES_WUNLOCK();
1530                                 break;
1531                         }
1532
1533                         newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1534                         newrule->evaluations = 0;
1535                         newrule->packets[0] = newrule->packets[1] = 0;
1536                         newrule->bytes[0] = newrule->bytes[1] = 0;
1537                 }
1538                 pf_empty_pool(&V_pf_pabuf);
1539
1540                 if (pcr->action == PF_CHANGE_ADD_HEAD)
1541                         oldrule = TAILQ_FIRST(
1542                             ruleset->rules[rs_num].active.ptr);
1543                 else if (pcr->action == PF_CHANGE_ADD_TAIL)
1544                         oldrule = TAILQ_LAST(
1545                             ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1546                 else {
1547                         oldrule = TAILQ_FIRST(
1548                             ruleset->rules[rs_num].active.ptr);
1549                         while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1550                                 oldrule = TAILQ_NEXT(oldrule, entries);
1551                         if (oldrule == NULL) {
1552                                 if (newrule != NULL)
1553                                         pf_free_rule(newrule);
1554                                 PF_RULES_WUNLOCK();
1555                                 error = EINVAL;
1556                                 break;
1557                         }
1558                 }
1559
1560                 if (pcr->action == PF_CHANGE_REMOVE) {
1561                         pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1562                             oldrule);
1563                         ruleset->rules[rs_num].active.rcount--;
1564                 } else {
1565                         if (oldrule == NULL)
1566                                 TAILQ_INSERT_TAIL(
1567                                     ruleset->rules[rs_num].active.ptr,
1568                                     newrule, entries);
1569                         else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1570                             pcr->action == PF_CHANGE_ADD_BEFORE)
1571                                 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1572                         else
1573                                 TAILQ_INSERT_AFTER(
1574                                     ruleset->rules[rs_num].active.ptr,
1575                                     oldrule, newrule, entries);
1576                         ruleset->rules[rs_num].active.rcount++;
1577                 }
1578
1579                 nr = 0;
1580                 TAILQ_FOREACH(oldrule,
1581                     ruleset->rules[rs_num].active.ptr, entries)
1582                         oldrule->nr = nr++;
1583
1584                 ruleset->rules[rs_num].active.ticket++;
1585
1586                 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1587                 pf_remove_if_empty_ruleset(ruleset);
1588
1589                 PF_RULES_WUNLOCK();
1590                 break;
1591
1592 #undef ERROUT
1593 DIOCCHANGERULE_error:
1594                 PF_RULES_WUNLOCK();
1595                 if (newrule != NULL) {
1596                         counter_u64_free(newrule->states_cur);
1597                         counter_u64_free(newrule->states_tot);
1598                         counter_u64_free(newrule->src_nodes);
1599                         free(newrule, M_PFRULE);
1600                 }
1601                 if (kif != NULL)
1602                         free(kif, PFI_MTYPE);
1603                 break;
1604         }
1605
1606         case DIOCCLRSTATES: {
1607                 struct pf_state         *s;
1608                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1609                 u_int                    i, killed = 0;
1610
1611                 for (i = 0; i <= V_pf_hashmask; i++) {
1612                         struct pf_idhash *ih = &V_pf_idhash[i];
1613
1614 relock_DIOCCLRSTATES:
1615                         PF_HASHROW_LOCK(ih);
1616                         LIST_FOREACH(s, &ih->states, entry)
1617                                 if (!psk->psk_ifname[0] ||
1618                                     !strcmp(psk->psk_ifname,
1619                                     s->kif->pfik_name)) {
1620                                         /*
1621                                          * Don't send out individual
1622                                          * delete messages.
1623                                          */
1624                                         s->state_flags |= PFSTATE_NOSYNC;
1625                                         pf_unlink_state(s, PF_ENTER_LOCKED);
1626                                         killed++;
1627                                         goto relock_DIOCCLRSTATES;
1628                                 }
1629                         PF_HASHROW_UNLOCK(ih);
1630                 }
1631                 psk->psk_killed = killed;
1632                 if (pfsync_clear_states_ptr != NULL)
1633                         pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
1634                 break;
1635         }
1636
1637         case DIOCKILLSTATES: {
1638                 struct pf_state         *s;
1639                 struct pf_state_key     *sk;
1640                 struct pf_addr          *srcaddr, *dstaddr;
1641                 u_int16_t                srcport, dstport;
1642                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1643                 u_int                    i, killed = 0;
1644
1645                 if (psk->psk_pfcmp.id) {
1646                         if (psk->psk_pfcmp.creatorid == 0)
1647                                 psk->psk_pfcmp.creatorid = V_pf_status.hostid;
1648                         if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
1649                             psk->psk_pfcmp.creatorid))) {
1650                                 pf_unlink_state(s, PF_ENTER_LOCKED);
1651                                 psk->psk_killed = 1;
1652                         }
1653                         break;
1654                 }
1655
1656                 for (i = 0; i <= V_pf_hashmask; i++) {
1657                         struct pf_idhash *ih = &V_pf_idhash[i];
1658
1659 relock_DIOCKILLSTATES:
1660                         PF_HASHROW_LOCK(ih);
1661                         LIST_FOREACH(s, &ih->states, entry) {
1662                                 sk = s->key[PF_SK_WIRE];
1663                                 if (s->direction == PF_OUT) {
1664                                         srcaddr = &sk->addr[1];
1665                                         dstaddr = &sk->addr[0];
1666                                         srcport = sk->port[0];
1667                                         dstport = sk->port[0];
1668                                 } else {
1669                                         srcaddr = &sk->addr[0];
1670                                         dstaddr = &sk->addr[1];
1671                                         srcport = sk->port[0];
1672                                         dstport = sk->port[0];
1673                                 }
1674
1675                                 if ((!psk->psk_af || sk->af == psk->psk_af)
1676                                     && (!psk->psk_proto || psk->psk_proto ==
1677                                     sk->proto) &&
1678                                     PF_MATCHA(psk->psk_src.neg,
1679                                     &psk->psk_src.addr.v.a.addr,
1680                                     &psk->psk_src.addr.v.a.mask,
1681                                     srcaddr, sk->af) &&
1682                                     PF_MATCHA(psk->psk_dst.neg,
1683                                     &psk->psk_dst.addr.v.a.addr,
1684                                     &psk->psk_dst.addr.v.a.mask,
1685                                     dstaddr, sk->af) &&
1686                                     (psk->psk_src.port_op == 0 ||
1687                                     pf_match_port(psk->psk_src.port_op,
1688                                     psk->psk_src.port[0], psk->psk_src.port[1],
1689                                     srcport)) &&
1690                                     (psk->psk_dst.port_op == 0 ||
1691                                     pf_match_port(psk->psk_dst.port_op,
1692                                     psk->psk_dst.port[0], psk->psk_dst.port[1],
1693                                     dstport)) &&
1694                                     (!psk->psk_label[0] ||
1695                                     (s->rule.ptr->label[0] &&
1696                                     !strcmp(psk->psk_label,
1697                                     s->rule.ptr->label))) &&
1698                                     (!psk->psk_ifname[0] ||
1699                                     !strcmp(psk->psk_ifname,
1700                                     s->kif->pfik_name))) {
1701                                         pf_unlink_state(s, PF_ENTER_LOCKED);
1702                                         killed++;
1703                                         goto relock_DIOCKILLSTATES;
1704                                 }
1705                         }
1706                         PF_HASHROW_UNLOCK(ih);
1707                 }
1708                 psk->psk_killed = killed;
1709                 break;
1710         }
1711
1712         case DIOCADDSTATE: {
1713                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
1714                 struct pfsync_state     *sp = &ps->state;
1715
1716                 if (sp->timeout >= PFTM_MAX) {
1717                         error = EINVAL;
1718                         break;
1719                 }
1720                 if (pfsync_state_import_ptr != NULL) {
1721                         PF_RULES_RLOCK();
1722                         error = pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
1723                         PF_RULES_RUNLOCK();
1724                 } else
1725                         error = EOPNOTSUPP;
1726                 break;
1727         }
1728
1729         case DIOCGETSTATE: {
1730                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
1731                 struct pf_state         *s;
1732
1733                 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
1734                 if (s == NULL) {
1735                         error = ENOENT;
1736                         break;
1737                 }
1738
1739                 pfsync_state_export(&ps->state, s);
1740                 PF_STATE_UNLOCK(s);
1741                 break;
1742         }
1743
1744         case DIOCGETSTATES: {
1745                 struct pfioc_states     *ps = (struct pfioc_states *)addr;
1746                 struct pf_state         *s;
1747                 struct pfsync_state     *pstore, *p;
1748                 int i, nr;
1749
1750                 if (ps->ps_len == 0) {
1751                         nr = uma_zone_get_cur(V_pf_state_z);
1752                         ps->ps_len = sizeof(struct pfsync_state) * nr;
1753                         break;
1754                 }
1755
1756                 p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK);
1757                 nr = 0;
1758
1759                 for (i = 0; i <= V_pf_hashmask; i++) {
1760                         struct pf_idhash *ih = &V_pf_idhash[i];
1761
1762                         PF_HASHROW_LOCK(ih);
1763                         LIST_FOREACH(s, &ih->states, entry) {
1764
1765                                 if (s->timeout == PFTM_UNLINKED)
1766                                         continue;
1767
1768                                 if ((nr+1) * sizeof(*p) > ps->ps_len) {
1769                                         PF_HASHROW_UNLOCK(ih);
1770                                         goto DIOCGETSTATES_full;
1771                                 }
1772                                 pfsync_state_export(p, s);
1773                                 p++;
1774                                 nr++;
1775                         }
1776                         PF_HASHROW_UNLOCK(ih);
1777                 }
1778 DIOCGETSTATES_full:
1779                 error = copyout(pstore, ps->ps_states,
1780                     sizeof(struct pfsync_state) * nr);
1781                 if (error) {
1782                         free(pstore, M_TEMP);
1783                         break;
1784                 }
1785                 ps->ps_len = sizeof(struct pfsync_state) * nr;
1786                 free(pstore, M_TEMP);
1787
1788                 break;
1789         }
1790
1791         case DIOCGETSTATUS: {
1792                 struct pf_status *s = (struct pf_status *)addr;
1793
1794                 PF_RULES_RLOCK();
1795                 s->running = V_pf_status.running;
1796                 s->since   = V_pf_status.since;
1797                 s->debug   = V_pf_status.debug;
1798                 s->hostid  = V_pf_status.hostid;
1799                 s->states  = V_pf_status.states;
1800                 s->src_nodes = V_pf_status.src_nodes;
1801
1802                 for (int i = 0; i < PFRES_MAX; i++)
1803                         s->counters[i] =
1804                             counter_u64_fetch(V_pf_status.counters[i]);
1805                 for (int i = 0; i < LCNT_MAX; i++)
1806                         s->lcounters[i] =
1807                             counter_u64_fetch(V_pf_status.lcounters[i]);
1808                 for (int i = 0; i < FCNT_MAX; i++)
1809                         s->fcounters[i] =
1810                             counter_u64_fetch(V_pf_status.fcounters[i]);
1811                 for (int i = 0; i < SCNT_MAX; i++)
1812                         s->scounters[i] =
1813                             counter_u64_fetch(V_pf_status.scounters[i]);
1814
1815                 bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
1816                 bcopy(V_pf_status.pf_chksum, s->pf_chksum,
1817                     PF_MD5_DIGEST_LENGTH);
1818
1819                 pfi_update_status(s->ifname, s);
1820                 PF_RULES_RUNLOCK();
1821                 break;
1822         }
1823
1824         case DIOCSETSTATUSIF: {
1825                 struct pfioc_if *pi = (struct pfioc_if *)addr;
1826
1827                 if (pi->ifname[0] == 0) {
1828                         bzero(V_pf_status.ifname, IFNAMSIZ);
1829                         break;
1830                 }
1831                 PF_RULES_WLOCK();
1832                 strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
1833                 PF_RULES_WUNLOCK();
1834                 break;
1835         }
1836
1837         case DIOCCLRSTATUS: {
1838                 PF_RULES_WLOCK();
1839                 for (int i = 0; i < PFRES_MAX; i++)
1840                         counter_u64_zero(V_pf_status.counters[i]);
1841                 for (int i = 0; i < FCNT_MAX; i++)
1842                         counter_u64_zero(V_pf_status.fcounters[i]);
1843                 for (int i = 0; i < SCNT_MAX; i++)
1844                         counter_u64_zero(V_pf_status.scounters[i]);
1845                 V_pf_status.since = time_second;
1846                 if (*V_pf_status.ifname)
1847                         pfi_update_status(V_pf_status.ifname, NULL);
1848                 PF_RULES_WUNLOCK();
1849                 break;
1850         }
1851
1852         case DIOCNATLOOK: {
1853                 struct pfioc_natlook    *pnl = (struct pfioc_natlook *)addr;
1854                 struct pf_state_key     *sk;
1855                 struct pf_state         *state;
1856                 struct pf_state_key_cmp  key;
1857                 int                      m = 0, direction = pnl->direction;
1858                 int                      sidx, didx;
1859
1860                 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
1861                 sidx = (direction == PF_IN) ? 1 : 0;
1862                 didx = (direction == PF_IN) ? 0 : 1;
1863
1864                 if (!pnl->proto ||
1865                     PF_AZERO(&pnl->saddr, pnl->af) ||
1866                     PF_AZERO(&pnl->daddr, pnl->af) ||
1867                     ((pnl->proto == IPPROTO_TCP ||
1868                     pnl->proto == IPPROTO_UDP) &&
1869                     (!pnl->dport || !pnl->sport)))
1870                         error = EINVAL;
1871                 else {
1872                         bzero(&key, sizeof(key));
1873                         key.af = pnl->af;
1874                         key.proto = pnl->proto;
1875                         PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1876                         key.port[sidx] = pnl->sport;
1877                         PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1878                         key.port[didx] = pnl->dport;
1879
1880                         state = pf_find_state_all(&key, direction, &m);
1881
1882                         if (m > 1)
1883                                 error = E2BIG;  /* more than one state */
1884                         else if (state != NULL) {
1885                                 /* XXXGL: not locked read */
1886                                 sk = state->key[sidx];
1887                                 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1888                                 pnl->rsport = sk->port[sidx];
1889                                 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1890                                 pnl->rdport = sk->port[didx];
1891                         } else
1892                                 error = ENOENT;
1893                 }
1894                 break;
1895         }
1896
1897         case DIOCSETTIMEOUT: {
1898                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
1899                 int              old;
1900
1901                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1902                     pt->seconds < 0) {
1903                         error = EINVAL;
1904                         break;
1905                 }
1906                 PF_RULES_WLOCK();
1907                 old = V_pf_default_rule.timeout[pt->timeout];
1908                 if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1909                         pt->seconds = 1;
1910                 V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
1911                 if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1912                         wakeup(pf_purge_thread);
1913                 pt->seconds = old;
1914                 PF_RULES_WUNLOCK();
1915                 break;
1916         }
1917
1918         case DIOCGETTIMEOUT: {
1919                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
1920
1921                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1922                         error = EINVAL;
1923                         break;
1924                 }
1925                 PF_RULES_RLOCK();
1926                 pt->seconds = V_pf_default_rule.timeout[pt->timeout];
1927                 PF_RULES_RUNLOCK();
1928                 break;
1929         }
1930
1931         case DIOCGETLIMIT: {
1932                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
1933
1934                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1935                         error = EINVAL;
1936                         break;
1937                 }
1938                 PF_RULES_RLOCK();
1939                 pl->limit = V_pf_limits[pl->index].limit;
1940                 PF_RULES_RUNLOCK();
1941                 break;
1942         }
1943
1944         case DIOCSETLIMIT: {
1945                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
1946                 int                      old_limit;
1947
1948                 PF_RULES_WLOCK();
1949                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1950                     V_pf_limits[pl->index].zone == NULL) {
1951                         PF_RULES_WUNLOCK();
1952                         error = EINVAL;
1953                         break;
1954                 }
1955                 uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
1956                 old_limit = V_pf_limits[pl->index].limit;
1957                 V_pf_limits[pl->index].limit = pl->limit;
1958                 pl->limit = old_limit;
1959                 PF_RULES_WUNLOCK();
1960                 break;
1961         }
1962
1963         case DIOCSETDEBUG: {
1964                 u_int32_t       *level = (u_int32_t *)addr;
1965
1966                 PF_RULES_WLOCK();
1967                 V_pf_status.debug = *level;
1968                 PF_RULES_WUNLOCK();
1969                 break;
1970         }
1971
1972         case DIOCCLRRULECTRS: {
1973                 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1974                 struct pf_ruleset       *ruleset = &pf_main_ruleset;
1975                 struct pf_rule          *rule;
1976
1977                 PF_RULES_WLOCK();
1978                 TAILQ_FOREACH(rule,
1979                     ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1980                         rule->evaluations = 0;
1981                         rule->packets[0] = rule->packets[1] = 0;
1982                         rule->bytes[0] = rule->bytes[1] = 0;
1983                 }
1984                 PF_RULES_WUNLOCK();
1985                 break;
1986         }
1987
1988         case DIOCGIFSPEED: {
1989                 struct pf_ifspeed       *psp = (struct pf_ifspeed *)addr;
1990                 struct pf_ifspeed       ps;
1991                 struct ifnet            *ifp;
1992
1993                 if (psp->ifname[0] != 0) {
1994                         /* Can we completely trust user-land? */
1995                         strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1996                         ifp = ifunit(ps.ifname);
1997                         if (ifp != NULL)
1998                                 psp->baudrate = ifp->if_baudrate;
1999                         else
2000                                 error = EINVAL;
2001                 } else
2002                         error = EINVAL;
2003                 break;
2004         }
2005
2006 #ifdef ALTQ
2007         case DIOCSTARTALTQ: {
2008                 struct pf_altq          *altq;
2009
2010                 PF_RULES_WLOCK();
2011                 /* enable all altq interfaces on active list */
2012                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2013                         if (altq->qname[0] == 0 && (altq->local_flags &
2014                             PFALTQ_FLAG_IF_REMOVED) == 0) {
2015                                 error = pf_enable_altq(altq);
2016                                 if (error != 0)
2017                                         break;
2018                         }
2019                 }
2020                 if (error == 0)
2021                         V_pf_altq_running = 1;
2022                 PF_RULES_WUNLOCK();
2023                 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2024                 break;
2025         }
2026
2027         case DIOCSTOPALTQ: {
2028                 struct pf_altq          *altq;
2029
2030                 PF_RULES_WLOCK();
2031                 /* disable all altq interfaces on active list */
2032                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2033                         if (altq->qname[0] == 0 && (altq->local_flags &
2034                             PFALTQ_FLAG_IF_REMOVED) == 0) {
2035                                 error = pf_disable_altq(altq);
2036                                 if (error != 0)
2037                                         break;
2038                         }
2039                 }
2040                 if (error == 0)
2041                         V_pf_altq_running = 0;
2042                 PF_RULES_WUNLOCK();
2043                 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2044                 break;
2045         }
2046
2047         case DIOCADDALTQ: {
2048                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2049                 struct pf_altq          *altq, *a;
2050                 struct ifnet            *ifp;
2051
2052                 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK);
2053                 bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2054                 altq->local_flags = 0;
2055
2056                 PF_RULES_WLOCK();
2057                 if (pa->ticket != V_ticket_altqs_inactive) {
2058                         PF_RULES_WUNLOCK();
2059                         free(altq, M_PFALTQ);
2060                         error = EBUSY;
2061                         break;
2062                 }
2063
2064                 /*
2065                  * if this is for a queue, find the discipline and
2066                  * copy the necessary fields
2067                  */
2068                 if (altq->qname[0] != 0) {
2069                         if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2070                                 PF_RULES_WUNLOCK();
2071                                 error = EBUSY;
2072                                 free(altq, M_PFALTQ);
2073                                 break;
2074                         }
2075                         altq->altq_disc = NULL;
2076                         TAILQ_FOREACH(a, V_pf_altqs_inactive, entries) {
2077                                 if (strncmp(a->ifname, altq->ifname,
2078                                     IFNAMSIZ) == 0 && a->qname[0] == 0) {
2079                                         altq->altq_disc = a->altq_disc;
2080                                         break;
2081                                 }
2082                         }
2083                 }
2084
2085                 if ((ifp = ifunit(altq->ifname)) == NULL)
2086                         altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2087                 else
2088                         error = altq_add(altq);
2089
2090                 if (error) {
2091                         PF_RULES_WUNLOCK();
2092                         free(altq, M_PFALTQ);
2093                         break;
2094                 }
2095
2096                 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2097                 bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2098                 PF_RULES_WUNLOCK();
2099                 break;
2100         }
2101
2102         case DIOCGETALTQS: {
2103                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2104                 struct pf_altq          *altq;
2105
2106                 PF_RULES_RLOCK();
2107                 pa->nr = 0;
2108                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2109                         pa->nr++;
2110                 pa->ticket = V_ticket_altqs_active;
2111                 PF_RULES_RUNLOCK();
2112                 break;
2113         }
2114
2115         case DIOCGETALTQ: {
2116                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2117                 struct pf_altq          *altq;
2118                 u_int32_t                nr;
2119
2120                 PF_RULES_RLOCK();
2121                 if (pa->ticket != V_ticket_altqs_active) {
2122                         PF_RULES_RUNLOCK();
2123                         error = EBUSY;
2124                         break;
2125                 }
2126                 nr = 0;
2127                 altq = TAILQ_FIRST(V_pf_altqs_active);
2128                 while ((altq != NULL) && (nr < pa->nr)) {
2129                         altq = TAILQ_NEXT(altq, entries);
2130                         nr++;
2131                 }
2132                 if (altq == NULL) {
2133                         PF_RULES_RUNLOCK();
2134                         error = EBUSY;
2135                         break;
2136                 }
2137                 bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2138                 PF_RULES_RUNLOCK();
2139                 break;
2140         }
2141
2142         case DIOCCHANGEALTQ:
2143                 /* CHANGEALTQ not supported yet! */
2144                 error = ENODEV;
2145                 break;
2146
2147         case DIOCGETQSTATS: {
2148                 struct pfioc_qstats     *pq = (struct pfioc_qstats *)addr;
2149                 struct pf_altq          *altq;
2150                 u_int32_t                nr;
2151                 int                      nbytes;
2152
2153                 PF_RULES_RLOCK();
2154                 if (pq->ticket != V_ticket_altqs_active) {
2155                         PF_RULES_RUNLOCK();
2156                         error = EBUSY;
2157                         break;
2158                 }
2159                 nbytes = pq->nbytes;
2160                 nr = 0;
2161                 altq = TAILQ_FIRST(V_pf_altqs_active);
2162                 while ((altq != NULL) && (nr < pq->nr)) {
2163                         altq = TAILQ_NEXT(altq, entries);
2164                         nr++;
2165                 }
2166                 if (altq == NULL) {
2167                         PF_RULES_RUNLOCK();
2168                         error = EBUSY;
2169                         break;
2170                 }
2171
2172                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2173                         PF_RULES_RUNLOCK();
2174                         error = ENXIO;
2175                         break;
2176                 }
2177                 PF_RULES_RUNLOCK();
2178                 error = altq_getqstats(altq, pq->buf, &nbytes);
2179                 if (error == 0) {
2180                         pq->scheduler = altq->scheduler;
2181                         pq->nbytes = nbytes;
2182                 }
2183                 break;
2184         }
2185 #endif /* ALTQ */
2186
2187         case DIOCBEGINADDRS: {
2188                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2189
2190                 PF_RULES_WLOCK();
2191                 pf_empty_pool(&V_pf_pabuf);
2192                 pp->ticket = ++V_ticket_pabuf;
2193                 PF_RULES_WUNLOCK();
2194                 break;
2195         }
2196
2197         case DIOCADDADDR: {
2198                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2199                 struct pf_pooladdr      *pa;
2200                 struct pfi_kif          *kif = NULL;
2201
2202 #ifndef INET
2203                 if (pp->af == AF_INET) {
2204                         error = EAFNOSUPPORT;
2205                         break;
2206                 }
2207 #endif /* INET */
2208 #ifndef INET6
2209                 if (pp->af == AF_INET6) {
2210                         error = EAFNOSUPPORT;
2211                         break;
2212                 }
2213 #endif /* INET6 */
2214                 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2215                     pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2216                     pp->addr.addr.type != PF_ADDR_TABLE) {
2217                         error = EINVAL;
2218                         break;
2219                 }
2220                 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2221                 bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2222                 if (pa->ifname[0])
2223                         kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2224                 PF_RULES_WLOCK();
2225                 if (pp->ticket != V_ticket_pabuf) {
2226                         PF_RULES_WUNLOCK();
2227                         if (pa->ifname[0])
2228                                 free(kif, PFI_MTYPE);
2229                         free(pa, M_PFRULE);
2230                         error = EBUSY;
2231                         break;
2232                 }
2233                 if (pa->ifname[0]) {
2234                         pa->kif = pfi_kif_attach(kif, pa->ifname);
2235                         pfi_kif_ref(pa->kif);
2236                 } else
2237                         pa->kif = NULL;
2238                 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2239                     pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2240                         if (pa->ifname[0])
2241                                 pfi_kif_unref(pa->kif);
2242                         PF_RULES_WUNLOCK();
2243                         free(pa, M_PFRULE);
2244                         break;
2245                 }
2246                 TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2247                 PF_RULES_WUNLOCK();
2248                 break;
2249         }
2250
2251         case DIOCGETADDRS: {
2252                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2253                 struct pf_pool          *pool;
2254                 struct pf_pooladdr      *pa;
2255
2256                 PF_RULES_RLOCK();
2257                 pp->nr = 0;
2258                 pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2259                     pp->r_num, 0, 1, 0);
2260                 if (pool == NULL) {
2261                         PF_RULES_RUNLOCK();
2262                         error = EBUSY;
2263                         break;
2264                 }
2265                 TAILQ_FOREACH(pa, &pool->list, entries)
2266                         pp->nr++;
2267                 PF_RULES_RUNLOCK();
2268                 break;
2269         }
2270
2271         case DIOCGETADDR: {
2272                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2273                 struct pf_pool          *pool;
2274                 struct pf_pooladdr      *pa;
2275                 u_int32_t                nr = 0;
2276
2277                 PF_RULES_RLOCK();
2278                 pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2279                     pp->r_num, 0, 1, 1);
2280                 if (pool == NULL) {
2281                         PF_RULES_RUNLOCK();
2282                         error = EBUSY;
2283                         break;
2284                 }
2285                 pa = TAILQ_FIRST(&pool->list);
2286                 while ((pa != NULL) && (nr < pp->nr)) {
2287                         pa = TAILQ_NEXT(pa, entries);
2288                         nr++;
2289                 }
2290                 if (pa == NULL) {
2291                         PF_RULES_RUNLOCK();
2292                         error = EBUSY;
2293                         break;
2294                 }
2295                 bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2296                 pf_addr_copyout(&pp->addr.addr);
2297                 PF_RULES_RUNLOCK();
2298                 break;
2299         }
2300
2301         case DIOCCHANGEADDR: {
2302                 struct pfioc_pooladdr   *pca = (struct pfioc_pooladdr *)addr;
2303                 struct pf_pool          *pool;
2304                 struct pf_pooladdr      *oldpa = NULL, *newpa = NULL;
2305                 struct pf_ruleset       *ruleset;
2306                 struct pfi_kif          *kif = NULL;
2307
2308                 if (pca->action < PF_CHANGE_ADD_HEAD ||
2309                     pca->action > PF_CHANGE_REMOVE) {
2310                         error = EINVAL;
2311                         break;
2312                 }
2313                 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2314                     pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2315                     pca->addr.addr.type != PF_ADDR_TABLE) {
2316                         error = EINVAL;
2317                         break;
2318                 }
2319
2320                 if (pca->action != PF_CHANGE_REMOVE) {
2321 #ifndef INET
2322                         if (pca->af == AF_INET) {
2323                                 error = EAFNOSUPPORT;
2324                                 break;
2325                         }
2326 #endif /* INET */
2327 #ifndef INET6
2328                         if (pca->af == AF_INET6) {
2329                                 error = EAFNOSUPPORT;
2330                                 break;
2331                         }
2332 #endif /* INET6 */
2333                         newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2334                         bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2335                         if (newpa->ifname[0])
2336                                 kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2337                         newpa->kif = NULL;
2338                 }
2339
2340 #define ERROUT(x)       { error = (x); goto DIOCCHANGEADDR_error; }
2341                 PF_RULES_WLOCK();
2342                 ruleset = pf_find_ruleset(pca->anchor);
2343                 if (ruleset == NULL)
2344                         ERROUT(EBUSY);
2345
2346                 pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2347                     pca->r_num, pca->r_last, 1, 1);
2348                 if (pool == NULL)
2349                         ERROUT(EBUSY);
2350
2351                 if (pca->action != PF_CHANGE_REMOVE) {
2352                         if (newpa->ifname[0]) {
2353                                 newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2354                                 pfi_kif_ref(newpa->kif);
2355                                 kif = NULL;
2356                         }
2357
2358                         switch (newpa->addr.type) {
2359                         case PF_ADDR_DYNIFTL:
2360                                 error = pfi_dynaddr_setup(&newpa->addr,
2361                                     pca->af);
2362                                 break;
2363                         case PF_ADDR_TABLE:
2364                                 newpa->addr.p.tbl = pfr_attach_table(ruleset,
2365                                     newpa->addr.v.tblname);
2366                                 if (newpa->addr.p.tbl == NULL)
2367                                         error = ENOMEM;
2368                                 break;
2369                         }
2370                         if (error)
2371                                 goto DIOCCHANGEADDR_error;
2372                 }
2373
2374                 switch (pca->action) {
2375                 case PF_CHANGE_ADD_HEAD:
2376                         oldpa = TAILQ_FIRST(&pool->list);
2377                         break;
2378                 case PF_CHANGE_ADD_TAIL:
2379                         oldpa = TAILQ_LAST(&pool->list, pf_palist);
2380                         break;
2381                 default:
2382                         oldpa = TAILQ_FIRST(&pool->list);
2383                         for (int i = 0; oldpa && i < pca->nr; i++)
2384                                 oldpa = TAILQ_NEXT(oldpa, entries);
2385
2386                         if (oldpa == NULL)
2387                                 ERROUT(EINVAL);
2388                 }
2389
2390                 if (pca->action == PF_CHANGE_REMOVE) {
2391                         TAILQ_REMOVE(&pool->list, oldpa, entries);
2392                         switch (oldpa->addr.type) {
2393                         case PF_ADDR_DYNIFTL:
2394                                 pfi_dynaddr_remove(oldpa->addr.p.dyn);
2395                                 break;
2396                         case PF_ADDR_TABLE:
2397                                 pfr_detach_table(oldpa->addr.p.tbl);
2398                                 break;
2399                         }
2400                         if (oldpa->kif)
2401                                 pfi_kif_unref(oldpa->kif);
2402                         free(oldpa, M_PFRULE);
2403                 } else {
2404                         if (oldpa == NULL)
2405                                 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2406                         else if (pca->action == PF_CHANGE_ADD_HEAD ||
2407                             pca->action == PF_CHANGE_ADD_BEFORE)
2408                                 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2409                         else
2410                                 TAILQ_INSERT_AFTER(&pool->list, oldpa,
2411                                     newpa, entries);
2412                 }
2413
2414                 pool->cur = TAILQ_FIRST(&pool->list);
2415                 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
2416                 PF_RULES_WUNLOCK();
2417                 break;
2418
2419 #undef ERROUT
2420 DIOCCHANGEADDR_error:
2421                 if (newpa->kif)
2422                         pfi_kif_unref(newpa->kif);
2423                 PF_RULES_WUNLOCK();
2424                 if (newpa != NULL)
2425                         free(newpa, M_PFRULE);
2426                 if (kif != NULL)
2427                         free(kif, PFI_MTYPE);
2428                 break;
2429         }
2430
2431         case DIOCGETRULESETS: {
2432                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
2433                 struct pf_ruleset       *ruleset;
2434                 struct pf_anchor        *anchor;
2435
2436                 PF_RULES_RLOCK();
2437                 pr->path[sizeof(pr->path) - 1] = 0;
2438                 if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2439                         PF_RULES_RUNLOCK();
2440                         error = ENOENT;
2441                         break;
2442                 }
2443                 pr->nr = 0;
2444                 if (ruleset->anchor == NULL) {
2445                         /* XXX kludge for pf_main_ruleset */
2446                         RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2447                                 if (anchor->parent == NULL)
2448                                         pr->nr++;
2449                 } else {
2450                         RB_FOREACH(anchor, pf_anchor_node,
2451                             &ruleset->anchor->children)
2452                                 pr->nr++;
2453                 }
2454                 PF_RULES_RUNLOCK();
2455                 break;
2456         }
2457
2458         case DIOCGETRULESET: {
2459                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
2460                 struct pf_ruleset       *ruleset;
2461                 struct pf_anchor        *anchor;
2462                 u_int32_t                nr = 0;
2463
2464                 PF_RULES_RLOCK();
2465                 pr->path[sizeof(pr->path) - 1] = 0;
2466                 if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2467                         PF_RULES_RUNLOCK();
2468                         error = ENOENT;
2469                         break;
2470                 }
2471                 pr->name[0] = 0;
2472                 if (ruleset->anchor == NULL) {
2473                         /* XXX kludge for pf_main_ruleset */
2474                         RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2475                                 if (anchor->parent == NULL && nr++ == pr->nr) {
2476                                         strlcpy(pr->name, anchor->name,
2477                                             sizeof(pr->name));
2478                                         break;
2479                                 }
2480                 } else {
2481                         RB_FOREACH(anchor, pf_anchor_node,
2482                             &ruleset->anchor->children)
2483                                 if (nr++ == pr->nr) {
2484                                         strlcpy(pr->name, anchor->name,
2485                                             sizeof(pr->name));
2486                                         break;
2487                                 }
2488                 }
2489                 if (!pr->name[0])
2490                         error = EBUSY;
2491                 PF_RULES_RUNLOCK();
2492                 break;
2493         }
2494
2495         case DIOCRCLRTABLES: {
2496                 struct pfioc_table *io = (struct pfioc_table *)addr;
2497
2498                 if (io->pfrio_esize != 0) {
2499                         error = ENODEV;
2500                         break;
2501                 }
2502                 PF_RULES_WLOCK();
2503                 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2504                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
2505                 PF_RULES_WUNLOCK();
2506                 break;
2507         }
2508
2509         case DIOCRADDTABLES: {
2510                 struct pfioc_table *io = (struct pfioc_table *)addr;
2511                 struct pfr_table *pfrts;
2512                 size_t totlen;
2513
2514                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2515                         error = ENODEV;
2516                         break;
2517                 }
2518                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2519                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2520                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2521                 if (error) {
2522                         free(pfrts, M_TEMP);
2523                         break;
2524                 }
2525                 PF_RULES_WLOCK();
2526                 error = pfr_add_tables(pfrts, io->pfrio_size,
2527                     &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2528                 PF_RULES_WUNLOCK();
2529                 free(pfrts, M_TEMP);
2530                 break;
2531         }
2532
2533         case DIOCRDELTABLES: {
2534                 struct pfioc_table *io = (struct pfioc_table *)addr;
2535                 struct pfr_table *pfrts;
2536                 size_t totlen;
2537
2538                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2539                         error = ENODEV;
2540                         break;
2541                 }
2542                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2543                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2544                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2545                 if (error) {
2546                         free(pfrts, M_TEMP);
2547                         break;
2548                 }
2549                 PF_RULES_WLOCK();
2550                 error = pfr_del_tables(pfrts, io->pfrio_size,
2551                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2552                 PF_RULES_WUNLOCK();
2553                 free(pfrts, M_TEMP);
2554                 break;
2555         }
2556
2557         case DIOCRGETTABLES: {
2558                 struct pfioc_table *io = (struct pfioc_table *)addr;
2559                 struct pfr_table *pfrts;
2560                 size_t totlen;
2561
2562                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2563                         error = ENODEV;
2564                         break;
2565                 }
2566                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2567                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2568                 PF_RULES_RLOCK();
2569                 error = pfr_get_tables(&io->pfrio_table, pfrts,
2570                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2571                 PF_RULES_RUNLOCK();
2572                 if (error == 0)
2573                         error = copyout(pfrts, io->pfrio_buffer, totlen);
2574                 free(pfrts, M_TEMP);
2575                 break;
2576         }
2577
2578         case DIOCRGETTSTATS: {
2579                 struct pfioc_table *io = (struct pfioc_table *)addr;
2580                 struct pfr_tstats *pfrtstats;
2581                 size_t totlen;
2582
2583                 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2584                         error = ENODEV;
2585                         break;
2586                 }
2587                 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
2588                 pfrtstats = malloc(totlen, M_TEMP, M_WAITOK);
2589                 PF_RULES_WLOCK();
2590                 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
2591                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2592                 PF_RULES_WUNLOCK();
2593                 if (error == 0)
2594                         error = copyout(pfrtstats, io->pfrio_buffer, totlen);
2595                 free(pfrtstats, M_TEMP);
2596                 break;
2597         }
2598
2599         case DIOCRCLRTSTATS: {
2600                 struct pfioc_table *io = (struct pfioc_table *)addr;
2601                 struct pfr_table *pfrts;
2602                 size_t totlen;
2603
2604                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2605                         error = ENODEV;
2606                         break;
2607                 }
2608                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2609                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2610                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2611                 if (error) {
2612                         free(pfrts, M_TEMP);
2613                         break;
2614                 }
2615                 PF_RULES_WLOCK();
2616                 error = pfr_clr_tstats(pfrts, io->pfrio_size,
2617                     &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2618                 PF_RULES_WUNLOCK();
2619                 free(pfrts, M_TEMP);
2620                 break;
2621         }
2622
2623         case DIOCRSETTFLAGS: {
2624                 struct pfioc_table *io = (struct pfioc_table *)addr;
2625                 struct pfr_table *pfrts;
2626                 size_t totlen;
2627
2628                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2629                         error = ENODEV;
2630                         break;
2631                 }
2632                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2633                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2634                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2635                 if (error) {
2636                         free(pfrts, M_TEMP);
2637                         break;
2638                 }
2639                 PF_RULES_WLOCK();
2640                 error = pfr_set_tflags(pfrts, io->pfrio_size,
2641                     io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2642                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2643                 PF_RULES_WUNLOCK();
2644                 free(pfrts, M_TEMP);
2645                 break;
2646         }
2647
2648         case DIOCRCLRADDRS: {
2649                 struct pfioc_table *io = (struct pfioc_table *)addr;
2650
2651                 if (io->pfrio_esize != 0) {
2652                         error = ENODEV;
2653                         break;
2654                 }
2655                 PF_RULES_WLOCK();
2656                 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2657                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
2658                 PF_RULES_WUNLOCK();
2659                 break;
2660         }
2661
2662         case DIOCRADDADDRS: {
2663                 struct pfioc_table *io = (struct pfioc_table *)addr;
2664                 struct pfr_addr *pfras;
2665                 size_t totlen;
2666
2667                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2668                         error = ENODEV;
2669                         break;
2670                 }
2671                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2672                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2673                 error = copyin(io->pfrio_buffer, pfras, totlen);
2674                 if (error) {
2675                         free(pfras, M_TEMP);
2676                         break;
2677                 }
2678                 PF_RULES_WLOCK();
2679                 error = pfr_add_addrs(&io->pfrio_table, pfras,
2680                     io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2681                     PFR_FLAG_USERIOCTL);
2682                 PF_RULES_WUNLOCK();
2683                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2684                         error = copyout(pfras, io->pfrio_buffer, totlen);
2685                 free(pfras, M_TEMP);
2686                 break;
2687         }
2688
2689         case DIOCRDELADDRS: {
2690                 struct pfioc_table *io = (struct pfioc_table *)addr;
2691                 struct pfr_addr *pfras;
2692                 size_t totlen;
2693
2694                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2695                         error = ENODEV;
2696                         break;
2697                 }
2698                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2699                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2700                 error = copyin(io->pfrio_buffer, pfras, totlen);
2701                 if (error) {
2702                         free(pfras, M_TEMP);
2703                         break;
2704                 }
2705                 PF_RULES_WLOCK();
2706                 error = pfr_del_addrs(&io->pfrio_table, pfras,
2707                     io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2708                     PFR_FLAG_USERIOCTL);
2709                 PF_RULES_WUNLOCK();
2710                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2711                         error = copyout(pfras, io->pfrio_buffer, totlen);
2712                 free(pfras, M_TEMP);
2713                 break;
2714         }
2715
2716         case DIOCRSETADDRS: {
2717                 struct pfioc_table *io = (struct pfioc_table *)addr;
2718                 struct pfr_addr *pfras;
2719                 size_t totlen;
2720
2721                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2722                         error = ENODEV;
2723                         break;
2724                 }
2725                 totlen = (io->pfrio_size + io->pfrio_size2) *
2726                     sizeof(struct pfr_addr);
2727                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2728                 error = copyin(io->pfrio_buffer, pfras, totlen);
2729                 if (error) {
2730                         free(pfras, M_TEMP);
2731                         break;
2732                 }
2733                 PF_RULES_WLOCK();
2734                 error = pfr_set_addrs(&io->pfrio_table, pfras,
2735                     io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2736                     &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2737                     PFR_FLAG_USERIOCTL, 0);
2738                 PF_RULES_WUNLOCK();
2739                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2740                         error = copyout(pfras, io->pfrio_buffer, totlen);
2741                 free(pfras, M_TEMP);
2742                 break;
2743         }
2744
2745         case DIOCRGETADDRS: {
2746                 struct pfioc_table *io = (struct pfioc_table *)addr;
2747                 struct pfr_addr *pfras;
2748                 size_t totlen;
2749
2750                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2751                         error = ENODEV;
2752                         break;
2753                 }
2754                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2755                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2756                 PF_RULES_RLOCK();
2757                 error = pfr_get_addrs(&io->pfrio_table, pfras,
2758                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2759                 PF_RULES_RUNLOCK();
2760                 if (error == 0)
2761                         error = copyout(pfras, io->pfrio_buffer, totlen);
2762                 free(pfras, M_TEMP);
2763                 break;
2764         }
2765
2766         case DIOCRGETASTATS: {
2767                 struct pfioc_table *io = (struct pfioc_table *)addr;
2768                 struct pfr_astats *pfrastats;
2769                 size_t totlen;
2770
2771                 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2772                         error = ENODEV;
2773                         break;
2774                 }
2775                 totlen = io->pfrio_size * sizeof(struct pfr_astats);
2776                 pfrastats = malloc(totlen, M_TEMP, M_WAITOK);
2777                 PF_RULES_RLOCK();
2778                 error = pfr_get_astats(&io->pfrio_table, pfrastats,
2779                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2780                 PF_RULES_RUNLOCK();
2781                 if (error == 0)
2782                         error = copyout(pfrastats, io->pfrio_buffer, totlen);
2783                 free(pfrastats, M_TEMP);
2784                 break;
2785         }
2786
2787         case DIOCRCLRASTATS: {
2788                 struct pfioc_table *io = (struct pfioc_table *)addr;
2789                 struct pfr_addr *pfras;
2790                 size_t totlen;
2791
2792                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2793                         error = ENODEV;
2794                         break;
2795                 }
2796                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2797                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2798                 error = copyin(io->pfrio_buffer, pfras, totlen);
2799                 if (error) {
2800                         free(pfras, M_TEMP);
2801                         break;
2802                 }
2803                 PF_RULES_WLOCK();
2804                 error = pfr_clr_astats(&io->pfrio_table, pfras,
2805                     io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2806                     PFR_FLAG_USERIOCTL);
2807                 PF_RULES_WUNLOCK();
2808                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2809                         error = copyout(pfras, io->pfrio_buffer, totlen);
2810                 free(pfras, M_TEMP);
2811                 break;
2812         }
2813
2814         case DIOCRTSTADDRS: {
2815                 struct pfioc_table *io = (struct pfioc_table *)addr;
2816                 struct pfr_addr *pfras;
2817                 size_t totlen;
2818
2819                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2820                         error = ENODEV;
2821                         break;
2822                 }
2823                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2824                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2825                 error = copyin(io->pfrio_buffer, pfras, totlen);
2826                 if (error) {
2827                         free(pfras, M_TEMP);
2828                         break;
2829                 }
2830                 PF_RULES_RLOCK();
2831                 error = pfr_tst_addrs(&io->pfrio_table, pfras,
2832                     io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2833                     PFR_FLAG_USERIOCTL);
2834                 PF_RULES_RUNLOCK();
2835                 if (error == 0)
2836                         error = copyout(pfras, io->pfrio_buffer, totlen);
2837                 free(pfras, M_TEMP);
2838                 break;
2839         }
2840
2841         case DIOCRINADEFINE: {
2842                 struct pfioc_table *io = (struct pfioc_table *)addr;
2843                 struct pfr_addr *pfras;
2844                 size_t totlen;
2845
2846                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2847                         error = ENODEV;
2848                         break;
2849                 }
2850                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2851                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2852                 error = copyin(io->pfrio_buffer, pfras, totlen);
2853                 if (error) {
2854                         free(pfras, M_TEMP);
2855                         break;
2856                 }
2857                 PF_RULES_WLOCK();
2858                 error = pfr_ina_define(&io->pfrio_table, pfras,
2859                     io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2860                     io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2861                 PF_RULES_WUNLOCK();
2862                 free(pfras, M_TEMP);
2863                 break;
2864         }
2865
2866         case DIOCOSFPADD: {
2867                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2868                 PF_RULES_WLOCK();
2869                 error = pf_osfp_add(io);
2870                 PF_RULES_WUNLOCK();
2871                 break;
2872         }
2873
2874         case DIOCOSFPGET: {
2875                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2876                 PF_RULES_RLOCK();
2877                 error = pf_osfp_get(io);
2878                 PF_RULES_RUNLOCK();
2879                 break;
2880         }
2881
2882         case DIOCXBEGIN: {
2883                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2884                 struct pfioc_trans_e    *ioes, *ioe;
2885                 size_t                   totlen;
2886                 int                      i;
2887
2888                 if (io->esize != sizeof(*ioe)) {
2889                         error = ENODEV;
2890                         break;
2891                 }
2892                 totlen = sizeof(struct pfioc_trans_e) * io->size;
2893                 ioes = malloc(totlen, M_TEMP, M_WAITOK);
2894                 error = copyin(io->array, ioes, totlen);
2895                 if (error) {
2896                         free(ioes, M_TEMP);
2897                         break;
2898                 }
2899                 PF_RULES_WLOCK();
2900                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2901                         switch (ioe->rs_num) {
2902 #ifdef ALTQ
2903                         case PF_RULESET_ALTQ:
2904                                 if (ioe->anchor[0]) {
2905                                         PF_RULES_WUNLOCK();
2906                                         free(ioes, M_TEMP);
2907                                         error = EINVAL;
2908                                         goto fail;
2909                                 }
2910                                 if ((error = pf_begin_altq(&ioe->ticket))) {
2911                                         PF_RULES_WUNLOCK();
2912                                         free(ioes, M_TEMP);
2913                                         goto fail;
2914                                 }
2915                                 break;
2916 #endif /* ALTQ */
2917                         case PF_RULESET_TABLE:
2918                             {
2919                                 struct pfr_table table;
2920
2921                                 bzero(&table, sizeof(table));
2922                                 strlcpy(table.pfrt_anchor, ioe->anchor,
2923                                     sizeof(table.pfrt_anchor));
2924                                 if ((error = pfr_ina_begin(&table,
2925                                     &ioe->ticket, NULL, 0))) {
2926                                         PF_RULES_WUNLOCK();
2927                                         free(ioes, M_TEMP);
2928                                         goto fail;
2929                                 }
2930                                 break;
2931                             }
2932                         default:
2933                                 if ((error = pf_begin_rules(&ioe->ticket,
2934                                     ioe->rs_num, ioe->anchor))) {
2935                                         PF_RULES_WUNLOCK();
2936                                         free(ioes, M_TEMP);
2937                                         goto fail;
2938                                 }
2939                                 break;
2940                         }
2941                 }
2942                 PF_RULES_WUNLOCK();
2943                 error = copyout(ioes, io->array, totlen);
2944                 free(ioes, M_TEMP);
2945                 break;
2946         }
2947
2948         case DIOCXROLLBACK: {
2949                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2950                 struct pfioc_trans_e    *ioe, *ioes;
2951                 size_t                   totlen;
2952                 int                      i;
2953
2954                 if (io->esize != sizeof(*ioe)) {
2955                         error = ENODEV;
2956                         break;
2957                 }
2958                 totlen = sizeof(struct pfioc_trans_e) * io->size;
2959                 ioes = malloc(totlen, M_TEMP, M_WAITOK);
2960                 error = copyin(io->array, ioes, totlen);
2961                 if (error) {
2962                         free(ioes, M_TEMP);
2963                         break;
2964                 }
2965                 PF_RULES_WLOCK();
2966                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2967                         switch (ioe->rs_num) {
2968 #ifdef ALTQ
2969                         case PF_RULESET_ALTQ:
2970                                 if (ioe->anchor[0]) {
2971                                         PF_RULES_WUNLOCK();
2972                                         free(ioes, M_TEMP);
2973                                         error = EINVAL;
2974                                         goto fail;
2975                                 }
2976                                 if ((error = pf_rollback_altq(ioe->ticket))) {
2977                                         PF_RULES_WUNLOCK();
2978                                         free(ioes, M_TEMP);
2979                                         goto fail; /* really bad */
2980                                 }
2981                                 break;
2982 #endif /* ALTQ */
2983                         case PF_RULESET_TABLE:
2984                             {
2985                                 struct pfr_table table;
2986
2987                                 bzero(&table, sizeof(table));
2988                                 strlcpy(table.pfrt_anchor, ioe->anchor,
2989                                     sizeof(table.pfrt_anchor));
2990                                 if ((error = pfr_ina_rollback(&table,
2991                                     ioe->ticket, NULL, 0))) {
2992                                         PF_RULES_WUNLOCK();
2993                                         free(ioes, M_TEMP);
2994                                         goto fail; /* really bad */
2995                                 }
2996                                 break;
2997                             }
2998                         default:
2999                                 if ((error = pf_rollback_rules(ioe->ticket,
3000                                     ioe->rs_num, ioe->anchor))) {
3001                                         PF_RULES_WUNLOCK();
3002                                         free(ioes, M_TEMP);
3003                                         goto fail; /* really bad */
3004                                 }
3005                                 break;
3006                         }
3007                 }
3008                 PF_RULES_WUNLOCK();
3009                 free(ioes, M_TEMP);
3010                 break;
3011         }
3012
3013         case DIOCXCOMMIT: {
3014                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
3015                 struct pfioc_trans_e    *ioe, *ioes;
3016                 struct pf_ruleset       *rs;
3017                 size_t                   totlen;
3018                 int                      i;
3019
3020                 if (io->esize != sizeof(*ioe)) {
3021                         error = ENODEV;
3022                         break;
3023                 }
3024                 totlen = sizeof(struct pfioc_trans_e) * io->size;
3025                 ioes = malloc(totlen, M_TEMP, M_WAITOK);
3026                 error = copyin(io->array, ioes, totlen);
3027                 if (error) {
3028                         free(ioes, M_TEMP);
3029                         break;
3030                 }
3031                 PF_RULES_WLOCK();
3032                 /* First makes sure everything will succeed. */
3033                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3034                         switch (ioe->rs_num) {
3035 #ifdef ALTQ
3036                         case PF_RULESET_ALTQ:
3037                                 if (ioe->anchor[0]) {
3038                                         PF_RULES_WUNLOCK();
3039                                         free(ioes, M_TEMP);
3040                                         error = EINVAL;
3041                                         goto fail;
3042                                 }
3043                                 if (!V_altqs_inactive_open || ioe->ticket !=
3044                                     V_ticket_altqs_inactive) {
3045                                         PF_RULES_WUNLOCK();
3046                                         free(ioes, M_TEMP);
3047                                         error = EBUSY;
3048                                         goto fail;
3049                                 }
3050                                 break;
3051 #endif /* ALTQ */
3052                         case PF_RULESET_TABLE:
3053                                 rs = pf_find_ruleset(ioe->anchor);
3054                                 if (rs == NULL || !rs->topen || ioe->ticket !=
3055                                     rs->tticket) {
3056                                         PF_RULES_WUNLOCK();
3057                                         free(ioes, M_TEMP);
3058                                         error = EBUSY;
3059                                         goto fail;
3060                                 }
3061                                 break;
3062                         default:
3063                                 if (ioe->rs_num < 0 || ioe->rs_num >=
3064                                     PF_RULESET_MAX) {
3065                                         PF_RULES_WUNLOCK();
3066                                         free(ioes, M_TEMP);
3067                                         error = EINVAL;
3068                                         goto fail;
3069                                 }
3070                                 rs = pf_find_ruleset(ioe->anchor);
3071                                 if (rs == NULL ||
3072                                     !rs->rules[ioe->rs_num].inactive.open ||
3073                                     rs->rules[ioe->rs_num].inactive.ticket !=
3074                                     ioe->ticket) {
3075                                         PF_RULES_WUNLOCK();
3076                                         free(ioes, M_TEMP);
3077                                         error = EBUSY;
3078                                         goto fail;
3079                                 }
3080                                 break;
3081                         }
3082                 }
3083                 /* Now do the commit - no errors should happen here. */
3084                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3085                         switch (ioe->rs_num) {
3086 #ifdef ALTQ
3087                         case PF_RULESET_ALTQ:
3088                                 if ((error = pf_commit_altq(ioe->ticket))) {
3089                                         PF_RULES_WUNLOCK();
3090                                         free(ioes, M_TEMP);
3091                                         goto fail; /* really bad */
3092                                 }
3093                                 break;
3094 #endif /* ALTQ */
3095                         case PF_RULESET_TABLE:
3096                             {
3097                                 struct pfr_table table;
3098
3099                                 bzero(&table, sizeof(table));
3100                                 strlcpy(table.pfrt_anchor, ioe->anchor,
3101                                     sizeof(table.pfrt_anchor));
3102                                 if ((error = pfr_ina_commit(&table,
3103                                     ioe->ticket, NULL, NULL, 0))) {
3104                                         PF_RULES_WUNLOCK();
3105                                         free(ioes, M_TEMP);
3106                                         goto fail; /* really bad */
3107                                 }
3108                                 break;
3109                             }
3110                         default:
3111                                 if ((error = pf_commit_rules(ioe->ticket,
3112                                     ioe->rs_num, ioe->anchor))) {
3113                                         PF_RULES_WUNLOCK();
3114                                         free(ioes, M_TEMP);
3115                                         goto fail; /* really bad */
3116                                 }
3117                                 break;
3118                         }
3119                 }
3120                 PF_RULES_WUNLOCK();
3121                 free(ioes, M_TEMP);
3122                 break;
3123         }
3124
3125         case DIOCGETSRCNODES: {
3126                 struct pfioc_src_nodes  *psn = (struct pfioc_src_nodes *)addr;
3127                 struct pf_srchash       *sh;
3128                 struct pf_src_node      *n, *p, *pstore;
3129                 uint32_t                 i, nr = 0;
3130
3131                 if (psn->psn_len == 0) {
3132                         for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
3133                             i++, sh++) {
3134                                 PF_HASHROW_LOCK(sh);
3135                                 LIST_FOREACH(n, &sh->nodes, entry)
3136                                         nr++;
3137                                 PF_HASHROW_UNLOCK(sh);
3138                         }
3139                         psn->psn_len = sizeof(struct pf_src_node) * nr;
3140                         break;
3141                 }
3142
3143                 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK);
3144                 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
3145                     i++, sh++) {
3146                     PF_HASHROW_LOCK(sh);
3147                     LIST_FOREACH(n, &sh->nodes, entry) {
3148                         int     secs = time_uptime, diff;
3149
3150                         if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3151                                 break;
3152
3153                         bcopy(n, p, sizeof(struct pf_src_node));
3154                         if (n->rule.ptr != NULL)
3155                                 p->rule.nr = n->rule.ptr->nr;
3156                         p->creation = secs - p->creation;
3157                         if (p->expire > secs)
3158                                 p->expire -= secs;
3159                         else
3160                                 p->expire = 0;
3161
3162                         /* Adjust the connection rate estimate. */
3163                         diff = secs - n->conn_rate.last;
3164                         if (diff >= n->conn_rate.seconds)
3165                                 p->conn_rate.count = 0;
3166                         else
3167                                 p->conn_rate.count -=
3168                                     n->conn_rate.count * diff /
3169                                     n->conn_rate.seconds;
3170                         p++;
3171                         nr++;
3172                     }
3173                     PF_HASHROW_UNLOCK(sh);
3174                 }
3175                 error = copyout(pstore, psn->psn_src_nodes,
3176                     sizeof(struct pf_src_node) * nr);
3177                 if (error) {
3178                         free(pstore, M_TEMP);
3179                         break;
3180                 }
3181                 psn->psn_len = sizeof(struct pf_src_node) * nr;
3182                 free(pstore, M_TEMP);
3183                 break;
3184         }
3185
3186         case DIOCCLRSRCNODES: {
3187
3188                 pf_clear_srcnodes(NULL);
3189                 pf_purge_expired_src_nodes();
3190                 break;
3191         }
3192
3193         case DIOCKILLSRCNODES:
3194                 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
3195                 break;
3196
3197         case DIOCSETHOSTID: {
3198                 u_int32_t       *hostid = (u_int32_t *)addr;
3199
3200                 PF_RULES_WLOCK();
3201                 if (*hostid == 0)
3202                         V_pf_status.hostid = arc4random();
3203                 else
3204                         V_pf_status.hostid = *hostid;
3205                 PF_RULES_WUNLOCK();
3206                 break;
3207         }
3208
3209         case DIOCOSFPFLUSH:
3210                 PF_RULES_WLOCK();
3211                 pf_osfp_flush();
3212                 PF_RULES_WUNLOCK();
3213                 break;
3214
3215         case DIOCIGETIFACES: {
3216                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3217                 struct pfi_kif *ifstore;
3218                 size_t bufsiz;
3219
3220                 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3221                         error = ENODEV;
3222                         break;
3223                 }
3224
3225                 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3226                 ifstore = malloc(bufsiz, M_TEMP, M_WAITOK);
3227                 PF_RULES_RLOCK();
3228                 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3229                 PF_RULES_RUNLOCK();
3230                 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3231                 free(ifstore, M_TEMP);
3232                 break;
3233         }
3234
3235         case DIOCSETIFFLAG: {
3236                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3237
3238                 PF_RULES_WLOCK();
3239                 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3240                 PF_RULES_WUNLOCK();
3241                 break;
3242         }
3243
3244         case DIOCCLRIFFLAG: {
3245                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3246
3247                 PF_RULES_WLOCK();
3248                 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3249                 PF_RULES_WUNLOCK();
3250                 break;
3251         }
3252
3253         default:
3254                 error = ENODEV;
3255                 break;
3256         }
3257 fail:
3258         CURVNET_RESTORE();
3259
3260         return (error);
3261 }
3262
3263 void
3264 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3265 {
3266         bzero(sp, sizeof(struct pfsync_state));
3267
3268         /* copy from state key */
3269         sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3270         sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3271         sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3272         sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3273         sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3274         sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3275         sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3276         sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3277         sp->proto = st->key[PF_SK_WIRE]->proto;
3278         sp->af = st->key[PF_SK_WIRE]->af;
3279
3280         /* copy from state */
3281         strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3282         bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3283         sp->creation = htonl(time_uptime - st->creation);
3284         sp->expire = pf_state_expires(st);
3285         if (sp->expire <= time_uptime)
3286                 sp->expire = htonl(0);
3287         else
3288                 sp->expire = htonl(sp->expire - time_uptime);
3289
3290         sp->direction = st->direction;
3291         sp->log = st->log;
3292         sp->timeout = st->timeout;
3293         sp->state_flags = st->state_flags;
3294         if (st->src_node)
3295                 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3296         if (st->nat_src_node)
3297                 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3298
3299         sp->id = st->id;
3300         sp->creatorid = st->creatorid;
3301         pf_state_peer_hton(&st->src, &sp->src);
3302         pf_state_peer_hton(&st->dst, &sp->dst);
3303
3304         if (st->rule.ptr == NULL)
3305                 sp->rule = htonl(-1);
3306         else
3307                 sp->rule = htonl(st->rule.ptr->nr);
3308         if (st->anchor.ptr == NULL)
3309                 sp->anchor = htonl(-1);
3310         else
3311                 sp->anchor = htonl(st->anchor.ptr->nr);
3312         if (st->nat_rule.ptr == NULL)
3313                 sp->nat_rule = htonl(-1);
3314         else
3315                 sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3316
3317         pf_state_counter_hton(st->packets[0], sp->packets[0]);
3318         pf_state_counter_hton(st->packets[1], sp->packets[1]);
3319         pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3320         pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3321
3322 }
3323
3324 static void
3325 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3326 {
3327         struct pfr_ktable *kt;
3328
3329         KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3330
3331         kt = aw->p.tbl;
3332         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3333                 kt = kt->pfrkt_root;
3334         aw->p.tbl = NULL;
3335         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3336                 kt->pfrkt_cnt : -1;
3337 }
3338
3339 /*
3340  * XXX - Check for version missmatch!!!
3341  */
3342 static void
3343 pf_clear_states(void)
3344 {
3345         struct pf_state *s;
3346         u_int i;
3347
3348         for (i = 0; i <= V_pf_hashmask; i++) {
3349                 struct pf_idhash *ih = &V_pf_idhash[i];
3350 relock:
3351                 PF_HASHROW_LOCK(ih);
3352                 LIST_FOREACH(s, &ih->states, entry) {
3353                         s->timeout = PFTM_PURGE;
3354                         /* Don't send out individual delete messages. */
3355                         s->sync_state = PFSTATE_NOSYNC;
3356                         pf_unlink_state(s, PF_ENTER_LOCKED);
3357                         goto relock;
3358                 }
3359                 PF_HASHROW_UNLOCK(ih);
3360         }
3361 }
3362
3363 static int
3364 pf_clear_tables(void)
3365 {
3366         struct pfioc_table io;
3367         int error;
3368
3369         bzero(&io, sizeof(io));
3370
3371         error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3372             io.pfrio_flags);
3373
3374         return (error);
3375 }
3376
3377 static void
3378 pf_clear_srcnodes(struct pf_src_node *n)
3379 {
3380         struct pf_state *s;
3381         int i;
3382
3383         for (i = 0; i <= V_pf_hashmask; i++) {
3384                 struct pf_idhash *ih = &V_pf_idhash[i];
3385
3386                 PF_HASHROW_LOCK(ih);
3387                 LIST_FOREACH(s, &ih->states, entry) {
3388                         if (n == NULL || n == s->src_node)
3389                                 s->src_node = NULL;
3390                         if (n == NULL || n == s->nat_src_node)
3391                                 s->nat_src_node = NULL;
3392                 }
3393                 PF_HASHROW_UNLOCK(ih);
3394         }
3395
3396         if (n == NULL) {
3397                 struct pf_srchash *sh;
3398
3399                 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
3400                     i++, sh++) {
3401                         PF_HASHROW_LOCK(sh);
3402                         LIST_FOREACH(n, &sh->nodes, entry) {
3403                                 n->expire = 1;
3404                                 n->states = 0;
3405                         }
3406                         PF_HASHROW_UNLOCK(sh);
3407                 }
3408         } else {
3409                 /* XXX: hash slot should already be locked here. */
3410                 n->expire = 1;
3411                 n->states = 0;
3412         }
3413 }
3414
3415 static void
3416 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
3417 {
3418         struct pf_src_node_list  kill;
3419
3420         LIST_INIT(&kill);
3421         for (int i = 0; i <= V_pf_srchashmask; i++) {
3422                 struct pf_srchash *sh = &V_pf_srchash[i];
3423                 struct pf_src_node *sn, *tmp;
3424
3425                 PF_HASHROW_LOCK(sh);
3426                 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
3427                         if (PF_MATCHA(psnk->psnk_src.neg,
3428                               &psnk->psnk_src.addr.v.a.addr,
3429                               &psnk->psnk_src.addr.v.a.mask,
3430                               &sn->addr, sn->af) &&
3431                             PF_MATCHA(psnk->psnk_dst.neg,
3432                               &psnk->psnk_dst.addr.v.a.addr,
3433                               &psnk->psnk_dst.addr.v.a.mask,
3434                               &sn->raddr, sn->af)) {
3435                                 pf_unlink_src_node_locked(sn);
3436                                 LIST_INSERT_HEAD(&kill, sn, entry);
3437                                 sn->expire = 1;
3438                         }
3439                 PF_HASHROW_UNLOCK(sh);
3440         }
3441
3442         for (int i = 0; i <= V_pf_hashmask; i++) {
3443                 struct pf_idhash *ih = &V_pf_idhash[i];
3444                 struct pf_state *s;
3445
3446                 PF_HASHROW_LOCK(ih);
3447                 LIST_FOREACH(s, &ih->states, entry) {
3448                         if (s->src_node && s->src_node->expire == 1) {
3449 #ifdef INVARIANTS
3450                                 s->src_node->states--;
3451 #endif
3452                                 s->src_node = NULL;
3453                         }
3454                         if (s->nat_src_node && s->nat_src_node->expire == 1) {
3455 #ifdef INVARIANTS
3456                                 s->nat_src_node->states--;
3457 #endif
3458                                 s->nat_src_node = NULL;
3459                         }
3460                 }
3461                 PF_HASHROW_UNLOCK(ih);
3462         }
3463
3464         psnk->psnk_killed = pf_free_src_nodes(&kill);
3465 }
3466
3467 /*
3468  * XXX - Check for version missmatch!!!
3469  */
3470
3471 /*
3472  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3473  */
3474 static int
3475 shutdown_pf(void)
3476 {
3477         int error = 0;
3478         u_int32_t t[5];
3479         char nn = '\0';
3480
3481         V_pf_status.running = 0;
3482
3483         counter_u64_free(V_pf_default_rule.states_cur);
3484         counter_u64_free(V_pf_default_rule.states_tot);
3485         counter_u64_free(V_pf_default_rule.src_nodes);
3486
3487         for (int i = 0; i < PFRES_MAX; i++)
3488                 counter_u64_free(V_pf_status.counters[i]);
3489         for (int i = 0; i < LCNT_MAX; i++)
3490                 counter_u64_free(V_pf_status.lcounters[i]);
3491         for (int i = 0; i < FCNT_MAX; i++)
3492                 counter_u64_free(V_pf_status.fcounters[i]);
3493         for (int i = 0; i < SCNT_MAX; i++)
3494                 counter_u64_free(V_pf_status.scounters[i]);
3495
3496         do {
3497                 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3498                     != 0) {
3499                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3500                         break;
3501                 }
3502                 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3503                     != 0) {
3504                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3505                         break;          /* XXX: rollback? */
3506                 }
3507                 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3508                     != 0) {
3509                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3510                         break;          /* XXX: rollback? */
3511                 }
3512                 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3513                     != 0) {
3514                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3515                         break;          /* XXX: rollback? */
3516                 }
3517                 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3518                     != 0) {
3519                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3520                         break;          /* XXX: rollback? */
3521                 }
3522
3523                 /* XXX: these should always succeed here */
3524                 pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3525                 pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3526                 pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3527                 pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3528                 pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3529
3530                 if ((error = pf_clear_tables()) != 0)
3531                         break;
3532
3533 #ifdef ALTQ
3534                 if ((error = pf_begin_altq(&t[0])) != 0) {
3535                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3536                         break;
3537                 }
3538                 pf_commit_altq(t[0]);
3539 #endif
3540
3541                 pf_clear_states();
3542
3543                 pf_clear_srcnodes(NULL);
3544
3545                 /* status does not use malloced mem so no need to cleanup */
3546                 /* fingerprints and interfaces have thier own cleanup code */
3547         } while(0);
3548
3549         return (error);
3550 }
3551
3552 #ifdef INET
3553 static int
3554 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3555     struct inpcb *inp)
3556 {
3557         int chk;
3558
3559         chk = pf_test(PF_IN, ifp, m, inp);
3560         if (chk && *m) {
3561                 m_freem(*m);
3562                 *m = NULL;
3563         }
3564
3565         return (chk);
3566 }
3567
3568 static int
3569 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3570     struct inpcb *inp)
3571 {
3572         int chk;
3573
3574         chk = pf_test(PF_OUT, ifp, m, inp);
3575         if (chk && *m) {
3576                 m_freem(*m);
3577                 *m = NULL;
3578         }
3579
3580         return (chk);
3581 }
3582 #endif
3583
3584 #ifdef INET6
3585 static int
3586 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3587     struct inpcb *inp)
3588 {
3589         int chk;
3590
3591         /*
3592          * In case of loopback traffic IPv6 uses the real interface in
3593          * order to support scoped addresses. In order to support stateful
3594          * filtering we have change this to lo0 as it is the case in IPv4.
3595          */
3596         CURVNET_SET(ifp->if_vnet);
3597         chk = pf_test6(PF_IN, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
3598         CURVNET_RESTORE();
3599         if (chk && *m) {
3600                 m_freem(*m);
3601                 *m = NULL;
3602         }
3603         return chk;
3604 }
3605
3606 static int
3607 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3608     struct inpcb *inp)
3609 {
3610         int chk;
3611
3612         CURVNET_SET(ifp->if_vnet);
3613         chk = pf_test6(PF_OUT, ifp, m, inp);
3614         CURVNET_RESTORE();
3615         if (chk && *m) {
3616                 m_freem(*m);
3617                 *m = NULL;
3618         }
3619         return chk;
3620 }
3621 #endif /* INET6 */
3622
3623 static int
3624 hook_pf(void)
3625 {
3626 #ifdef INET
3627         struct pfil_head *pfh_inet;
3628 #endif
3629 #ifdef INET6
3630         struct pfil_head *pfh_inet6;
3631 #endif
3632
3633         if (V_pf_pfil_hooked)
3634                 return (0);
3635
3636 #ifdef INET
3637         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3638         if (pfh_inet == NULL)
3639                 return (ESRCH); /* XXX */
3640         pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
3641         pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
3642 #endif
3643 #ifdef INET6
3644         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3645         if (pfh_inet6 == NULL) {
3646 #ifdef INET
3647                 pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3648                     pfh_inet);
3649                 pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3650                     pfh_inet);
3651 #endif
3652                 return (ESRCH); /* XXX */
3653         }
3654         pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
3655         pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
3656 #endif
3657
3658         V_pf_pfil_hooked = 1;
3659         return (0);
3660 }
3661
3662 static int
3663 dehook_pf(void)
3664 {
3665 #ifdef INET
3666         struct pfil_head *pfh_inet;
3667 #endif
3668 #ifdef INET6
3669         struct pfil_head *pfh_inet6;
3670 #endif
3671
3672         if (V_pf_pfil_hooked == 0)
3673                 return (0);
3674
3675 #ifdef INET
3676         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3677         if (pfh_inet == NULL)
3678                 return (ESRCH); /* XXX */
3679         pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3680             pfh_inet);
3681         pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3682             pfh_inet);
3683 #endif
3684 #ifdef INET6
3685         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3686         if (pfh_inet6 == NULL)
3687                 return (ESRCH); /* XXX */
3688         pfil_remove_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
3689             pfh_inet6);
3690         pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
3691             pfh_inet6);
3692 #endif
3693
3694         V_pf_pfil_hooked = 0;
3695         return (0);
3696 }
3697
3698 static int
3699 pf_load(void)
3700 {
3701         int error;
3702
3703         VNET_ITERATOR_DECL(vnet_iter);
3704
3705         VNET_LIST_RLOCK();
3706         VNET_FOREACH(vnet_iter) {
3707                 CURVNET_SET(vnet_iter);
3708                 V_pf_pfil_hooked = 0;
3709                 V_pf_end_threads = 0;
3710                 TAILQ_INIT(&V_pf_tags);
3711                 TAILQ_INIT(&V_pf_qids);
3712                 CURVNET_RESTORE();
3713         }
3714         VNET_LIST_RUNLOCK();
3715
3716         rw_init(&pf_rules_lock, "pf rulesets");
3717
3718         pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
3719         if ((error = pfattach()) != 0)
3720                 return (error);
3721
3722         return (0);
3723 }
3724
3725 static int
3726 pf_unload(void)
3727 {
3728         int error = 0;
3729
3730         PF_RULES_WLOCK();
3731         V_pf_status.running = 0;
3732         PF_RULES_WUNLOCK();
3733         swi_remove(V_pf_swi_cookie);
3734         error = dehook_pf();
3735         if (error) {
3736                 /*
3737                  * Should not happen!
3738                  * XXX Due to error code ESRCH, kldunload will show
3739                  * a message like 'No such process'.
3740                  */
3741                 printf("%s : pfil unregisteration fail\n", __FUNCTION__);
3742                 return error;
3743         }
3744         PF_RULES_WLOCK();
3745         shutdown_pf();
3746         V_pf_end_threads = 1;
3747         while (V_pf_end_threads < 2) {
3748                 wakeup_one(pf_purge_thread);
3749                 rw_sleep(pf_purge_thread, &pf_rules_lock, 0, "pftmo", 0);
3750         }
3751         pf_normalize_cleanup();
3752         pfi_cleanup();
3753         pfr_cleanup();
3754         pf_osfp_flush();
3755         pf_cleanup();
3756         if (IS_DEFAULT_VNET(curvnet))
3757                 pf_mtag_cleanup();
3758         PF_RULES_WUNLOCK();
3759         destroy_dev(pf_dev);
3760         rw_destroy(&pf_rules_lock);
3761
3762         return (error);
3763 }
3764
3765 static int
3766 pf_modevent(module_t mod, int type, void *data)
3767 {
3768         int error = 0;
3769
3770         switch(type) {
3771         case MOD_LOAD:
3772                 error = pf_load();
3773                 break;
3774         case MOD_QUIESCE:
3775                 /*
3776                  * Module should not be unloaded due to race conditions.
3777                  */
3778                 error = EBUSY;
3779                 break;
3780         case MOD_UNLOAD:
3781                 error = pf_unload();
3782                 break;
3783         default:
3784                 error = EINVAL;
3785                 break;
3786         }
3787
3788         return (error);
3789 }
3790
3791 static moduledata_t pf_mod = {
3792         "pf",
3793         pf_modevent,
3794         0
3795 };
3796
3797 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3798 MODULE_VERSION(pf, PF_MODVER);