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         pf_initialize();
 208         pfr_initialize();
 209         pfi_initialize();
 210         pf_normalize_init();
 211
 212         V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
 213         V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
 214
 215         RB_INIT(&V_pf_anchors);
 216         pf_init_ruleset(&pf_main_ruleset);
 217
 218         /* default rule should never be garbage collected */
 219         V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
 220 #ifdef PF_DEFAULT_TO_DROP
 221         V_pf_default_rule.action = PF_DROP;
 222 #else
 223         V_pf_default_rule.action = PF_PASS;
 224 #endif
 225         V_pf_default_rule.nr = -1;
 226         V_pf_default_rule.rtableid = -1;
 227
 228         /* initialize default timeouts */
 229         my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
 230         my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
 231         my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
 232         my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
 233         my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
 234         my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
 235         my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
 236         my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
 237         my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
 238         my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
 239         my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
 240         my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
 241         my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
 242         my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
 243         my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
 244         my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
 245         my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
 246         my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
 247         my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
 248         my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
 249
 250         bzero(&V_pf_status, sizeof(V_pf_status));
 251         V_pf_status.debug = PF_DEBUG_URGENT;
 252
 253         V_pf_pfil_hooked = 0;
 254
 255         /* XXX do our best to avoid a conflict */
 256         V_pf_status.hostid = arc4random();
 257
 258         if ((error = kproc_create(pf_purge_thread, curvnet, NULL, 0, 0,
 259             "pf purge")) != 0)
 260                 /* XXXGL: leaked all above. */
 261                 return (error);
 262         if ((error = swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
 263             INTR_MPSAFE, &V_pf_swi_cookie)) != 0)
 264                 /* XXXGL: leaked all above. */
 265                 return (error);
 266
 267         return (0);
 268 }
 269
 270 static struct pf_pool *
 271 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
 272     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
 273     u_int8_t check_ticket)
 274 {
 275         struct pf_ruleset       *ruleset;
 276         struct pf_rule          *rule;
 277         int                      rs_num;
 278
 279         ruleset = pf_find_ruleset(anchor);
 280         if (ruleset == NULL)
 281                 return (NULL);
 282         rs_num = pf_get_ruleset_number(rule_action);
 283         if (rs_num >= PF_RULESET_MAX)
 284                 return (NULL);
 285         if (active) {
 286                 if (check_ticket && ticket !=
 287                     ruleset->rules[rs_num].active.ticket)
 288                         return (NULL);
 289                 if (r_last)
 290                         rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
 291                             pf_rulequeue);
 292                 else
 293                         rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
 294         } else {
 295                 if (check_ticket && ticket !=
 296                     ruleset->rules[rs_num].inactive.ticket)
 297                         return (NULL);
 298                 if (r_last)
 299                         rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
 300                             pf_rulequeue);
 301                 else
 302                         rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
 303         }
 304         if (!r_last) {
 305                 while ((rule != NULL) && (rule->nr != rule_number))
 306                         rule = TAILQ_NEXT(rule, entries);
 307         }
 308         if (rule == NULL)
 309                 return (NULL);
 310
 311         return (&rule->rpool);
 312 }
 313
 314 static void
 315 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
 316 {
 317         struct pf_pooladdr      *mv_pool_pa;
 318
 319         while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
 320                 TAILQ_REMOVE(poola, mv_pool_pa, entries);
 321                 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
 322         }
 323 }
 324
 325 static void
 326 pf_empty_pool(struct pf_palist *poola)
 327 {
 328         struct pf_pooladdr *pa;
 329
 330         while ((pa = TAILQ_FIRST(poola)) != NULL) {
 331                 switch (pa->addr.type) {
 332                 case PF_ADDR_DYNIFTL:
 333                         pfi_dynaddr_remove(pa->addr.p.dyn);
 334                         break;
 335                 case PF_ADDR_TABLE:
 336                         pfr_detach_table(pa->addr.p.tbl);
 337                         break;
 338                 }
 339                 if (pa->kif)
 340                         pfi_kif_unref(pa->kif);
 341                 TAILQ_REMOVE(poola, pa, entries);
 342                 free(pa, M_PFRULE);
 343         }
 344 }
 345
 346 static void
 347 pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
 348 {
 349
 350         PF_RULES_WASSERT();
 351
 352         TAILQ_REMOVE(rulequeue, rule, entries);
 353
 354         PF_UNLNKDRULES_LOCK();
 355         rule->rule_flag |= PFRULE_REFS;
 356         TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
 357         PF_UNLNKDRULES_UNLOCK();
 358 }
 359
 360 void
 361 pf_free_rule(struct pf_rule *rule)
 362 {
 363
 364         PF_RULES_WASSERT();
 365
 366         if (rule->tag)
 367                 tag_unref(&V_pf_tags, rule->tag);
 368         if (rule->match_tag)
 369                 tag_unref(&V_pf_tags, rule->match_tag);
 370 #ifdef ALTQ
 371         if (rule->pqid != rule->qid)
 372                 pf_qid_unref(rule->pqid);
 373         pf_qid_unref(rule->qid);
 374 #endif
 375         switch (rule->src.addr.type) {
 376         case PF_ADDR_DYNIFTL:
 377                 pfi_dynaddr_remove(rule->src.addr.p.dyn);
 378                 break;
 379         case PF_ADDR_TABLE:
 380                 pfr_detach_table(rule->src.addr.p.tbl);
 381                 break;
 382         }
 383         switch (rule->dst.addr.type) {
 384         case PF_ADDR_DYNIFTL:
 385                 pfi_dynaddr_remove(rule->dst.addr.p.dyn);
 386                 break;
 387         case PF_ADDR_TABLE:
 388                 pfr_detach_table(rule->dst.addr.p.tbl);
 389                 break;
 390         }
 391         if (rule->overload_tbl)
 392                 pfr_detach_table(rule->overload_tbl);
 393         if (rule->kif)
 394                 pfi_kif_unref(rule->kif);
 395         pf_anchor_remove(rule);
 396         pf_empty_pool(&rule->rpool.list);
 397         free(rule, M_PFRULE);
 398 }
 399
 400 static u_int16_t
 401 tagname2tag(struct pf_tags *head, char *tagname)
 402 {
 403         struct pf_tagname       *tag, *p = NULL;
 404         u_int16_t                new_tagid = 1;
 405
 406         PF_RULES_WASSERT();
 407
 408         TAILQ_FOREACH(tag, head, entries)
 409                 if (strcmp(tagname, tag->name) == 0) {
 410                         tag->ref++;
 411                         return (tag->tag);
 412                 }
 413
 414         /*
 415          * to avoid fragmentation, we do a linear search from the beginning
 416          * and take the first free slot we find. if there is none or the list
 417          * is empty, append a new entry at the end.
 418          */
 419
 420         /* new entry */
 421         if (!TAILQ_EMPTY(head))
 422                 for (p = TAILQ_FIRST(head); p != NULL &&
 423                     p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
 424                         new_tagid = p->tag + 1;
 425
 426         if (new_tagid > TAGID_MAX)
 427                 return (0);
 428
 429         /* allocate and fill new struct pf_tagname */
 430         tag = malloc(sizeof(*tag), M_PFTAG, M_NOWAIT|M_ZERO);
 431         if (tag == NULL)
 432                 return (0);
 433         strlcpy(tag->name, tagname, sizeof(tag->name));
 434         tag->tag = new_tagid;
 435         tag->ref++;
 436
 437         if (p != NULL)  /* insert new entry before p */
 438                 TAILQ_INSERT_BEFORE(p, tag, entries);
 439         else    /* either list empty or no free slot in between */
 440                 TAILQ_INSERT_TAIL(head, tag, entries);
 441
 442         return (tag->tag);
 443 }
 444
 445 static void
 446 tag_unref(struct pf_tags *head, u_int16_t tag)
 447 {
 448         struct pf_tagname       *p, *next;
 449
 450         PF_RULES_WASSERT();
 451
 452         for (p = TAILQ_FIRST(head); p != NULL; p = next) {
 453                 next = TAILQ_NEXT(p, entries);
 454                 if (tag == p->tag) {
 455                         if (--p->ref == 0) {
 456                                 TAILQ_REMOVE(head, p, entries);
 457                                 free(p, M_PFTAG);
 458                         }
 459                         break;
 460                 }
 461         }
 462 }
 463
 464 static u_int16_t
 465 pf_tagname2tag(char *tagname)
 466 {
 467         return (tagname2tag(&V_pf_tags, tagname));
 468 }
 469
 470 #ifdef ALTQ
 471 static u_int32_t
 472 pf_qname2qid(char *qname)
 473 {
 474         return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
 475 }
 476
 477 static void
 478 pf_qid_unref(u_int32_t qid)
 479 {
 480         tag_unref(&V_pf_qids, (u_int16_t)qid);
 481 }
 482
 483 static int
 484 pf_begin_altq(u_int32_t *ticket)
 485 {
 486         struct pf_altq  *altq;
 487         int              error = 0;
 488
 489         PF_RULES_WASSERT();
 490
 491         /* Purge the old altq list */
 492         while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
 493                 TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
 494                 if (altq->qname[0] == 0 &&
 495                     (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 496                         /* detach and destroy the discipline */
 497                         error = altq_remove(altq);
 498                 } else
 499                         pf_qid_unref(altq->qid);
 500                 free(altq, M_PFALTQ);
 501         }
 502         if (error)
 503                 return (error);
 504         *ticket = ++V_ticket_altqs_inactive;
 505         V_altqs_inactive_open = 1;
 506         return (0);
 507 }
 508
 509 static int
 510 pf_rollback_altq(u_int32_t ticket)
 511 {
 512         struct pf_altq  *altq;
 513         int              error = 0;
 514
 515         PF_RULES_WASSERT();
 516
 517         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 518                 return (0);
 519         /* Purge the old altq list */
 520         while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
 521                 TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
 522                 if (altq->qname[0] == 0 &&
 523                    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 524                         /* detach and destroy the discipline */
 525                         error = altq_remove(altq);
 526                 } else
 527                         pf_qid_unref(altq->qid);
 528                 free(altq, M_PFALTQ);
 529         }
 530         V_altqs_inactive_open = 0;
 531         return (error);
 532 }
 533
 534 static int
 535 pf_commit_altq(u_int32_t ticket)
 536 {
 537         struct pf_altqqueue     *old_altqs;
 538         struct pf_altq          *altq;
 539         int                      err, error = 0;
 540
 541         PF_RULES_WASSERT();
 542
 543         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 544                 return (EBUSY);
 545
 546         /* swap altqs, keep the old. */
 547         old_altqs = V_pf_altqs_active;
 548         V_pf_altqs_active = V_pf_altqs_inactive;
 549         V_pf_altqs_inactive = old_altqs;
 550         V_ticket_altqs_active = V_ticket_altqs_inactive;
 551
 552         /* Attach new disciplines */
 553         TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
 554         if (altq->qname[0] == 0 &&
 555            (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 556                         /* attach the discipline */
 557                         error = altq_pfattach(altq);
 558                         if (error == 0 && V_pf_altq_running)
 559                                 error = pf_enable_altq(altq);
 560                         if (error != 0)
 561                                 return (error);
 562                 }
 563         }
 564
 565         /* Purge the old altq list */
 566         while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
 567                 TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
 568                 if (altq->qname[0] == 0 &&
 569                     (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 570                         /* detach and destroy the discipline */
 571                         if (V_pf_altq_running)
 572                                 error = pf_disable_altq(altq);
 573                         err = altq_pfdetach(altq);
 574                         if (err != 0 && error == 0)
 575                                 error = err;
 576                         err = altq_remove(altq);
 577                         if (err != 0 && error == 0)
 578                                 error = err;
 579                 } else
 580                         pf_qid_unref(altq->qid);
 581                 free(altq, M_PFALTQ);
 582         }
 583
 584         V_altqs_inactive_open = 0;
 585         return (error);
 586 }
 587
 588 static int
 589 pf_enable_altq(struct pf_altq *altq)
 590 {
 591         struct ifnet            *ifp;
 592         struct tb_profile        tb;
 593         int                      error = 0;
 594
 595         if ((ifp = ifunit(altq->ifname)) == NULL)
 596                 return (EINVAL);
 597
 598         if (ifp->if_snd.altq_type != ALTQT_NONE)
 599                 error = altq_enable(&ifp->if_snd);
 600
 601         /* set tokenbucket regulator */
 602         if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
 603                 tb.rate = altq->ifbandwidth;
 604                 tb.depth = altq->tbrsize;
 605                 error = tbr_set(&ifp->if_snd, &tb);
 606         }
 607
 608         return (error);
 609 }
 610
 611 static int
 612 pf_disable_altq(struct pf_altq *altq)
 613 {
 614         struct ifnet            *ifp;
 615         struct tb_profile        tb;
 616         int                      error;
 617
 618         if ((ifp = ifunit(altq->ifname)) == NULL)
 619                 return (EINVAL);
 620
 621         /*
 622          * when the discipline is no longer referenced, it was overridden
 623          * by a new one.  if so, just return.
 624          */
 625         if (altq->altq_disc != ifp->if_snd.altq_disc)
 626                 return (0);
 627
 628         error = altq_disable(&ifp->if_snd);
 629
 630         if (error == 0) {
 631                 /* clear tokenbucket regulator */
 632                 tb.rate = 0;
 633                 error = tbr_set(&ifp->if_snd, &tb);
 634         }
 635
 636         return (error);
 637 }
 638
 639 void
 640 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
 641 {
 642         struct ifnet    *ifp1;
 643         struct pf_altq  *a1, *a2, *a3;
 644         u_int32_t        ticket;
 645         int              error = 0;
 646
 647         /* Interrupt userland queue modifications */
 648         if (V_altqs_inactive_open)
 649                 pf_rollback_altq(V_ticket_altqs_inactive);
 650
 651         /* Start new altq ruleset */
 652         if (pf_begin_altq(&ticket))
 653                 return;
 654
 655         /* Copy the current active set */
 656         TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
 657                 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
 658                 if (a2 == NULL) {
 659                         error = ENOMEM;
 660                         break;
 661                 }
 662                 bcopy(a1, a2, sizeof(struct pf_altq));
 663
 664                 if (a2->qname[0] != 0) {
 665                         if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
 666                                 error = EBUSY;
 667                                 free(a2, M_PFALTQ);
 668                                 break;
 669                         }
 670                         a2->altq_disc = NULL;
 671                         TAILQ_FOREACH(a3, V_pf_altqs_inactive, entries) {
 672                                 if (strncmp(a3->ifname, a2->ifname,
 673                                     IFNAMSIZ) == 0 && a3->qname[0] == 0) {
 674                                         a2->altq_disc = a3->altq_disc;
 675                                         break;
 676                                 }
 677                         }
 678                 }
 679                 /* Deactivate the interface in question */
 680                 a2->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
 681                 if ((ifp1 = ifunit(a2->ifname)) == NULL ||
 682                     (remove && ifp1 == ifp)) {
 683                         a2->local_flags |= PFALTQ_FLAG_IF_REMOVED;
 684                 } else {
 685                         error = altq_add(a2);
 686
 687                         if (ticket != V_ticket_altqs_inactive)
 688                                 error = EBUSY;
 689
 690                         if (error) {
 691                                 free(a2, M_PFALTQ);
 692                                 break;
 693                         }
 694                 }
 695
 696                 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
 697         }
 698
 699         if (error != 0)
 700                 pf_rollback_altq(ticket);
 701         else
 702                 pf_commit_altq(ticket);
 703 }
 704 #endif /* ALTQ */
 705
 706 static int
 707 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
 708 {
 709         struct pf_ruleset       *rs;
 710         struct pf_rule          *rule;
 711
 712         PF_RULES_WASSERT();
 713
 714         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 715                 return (EINVAL);
 716         rs = pf_find_or_create_ruleset(anchor);
 717         if (rs == NULL)
 718                 return (EINVAL);
 719         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 720                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 721                 rs->rules[rs_num].inactive.rcount--;
 722         }
 723         *ticket = ++rs->rules[rs_num].inactive.ticket;
 724         rs->rules[rs_num].inactive.open = 1;
 725         return (0);
 726 }
 727
 728 static int
 729 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
 730 {
 731         struct pf_ruleset       *rs;
 732         struct pf_rule          *rule;
 733
 734         PF_RULES_WASSERT();
 735
 736         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 737                 return (EINVAL);
 738         rs = pf_find_ruleset(anchor);
 739         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 740             rs->rules[rs_num].inactive.ticket != ticket)
 741                 return (0);
 742         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 743                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 744                 rs->rules[rs_num].inactive.rcount--;
 745         }
 746         rs->rules[rs_num].inactive.open = 0;
 747         return (0);
 748 }
 749
 750 #define PF_MD5_UPD(st, elm)                                             \
 751                 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
 752
 753 #define PF_MD5_UPD_STR(st, elm)                                         \
 754                 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
 755
 756 #define PF_MD5_UPD_HTONL(st, elm, stor) do {                            \
 757                 (stor) = htonl((st)->elm);                              \
 758                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
 759 } while (0)
 760
 761 #define PF_MD5_UPD_HTONS(st, elm, stor) do {                            \
 762                 (stor) = htons((st)->elm);                              \
 763                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
 764 } while (0)
 765
 766 static void
 767 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
 768 {
 769         PF_MD5_UPD(pfr, addr.type);
 770         switch (pfr->addr.type) {
 771                 case PF_ADDR_DYNIFTL:
 772                         PF_MD5_UPD(pfr, addr.v.ifname);
 773                         PF_MD5_UPD(pfr, addr.iflags);
 774                         break;
 775                 case PF_ADDR_TABLE:
 776                         PF_MD5_UPD(pfr, addr.v.tblname);
 777                         break;
 778                 case PF_ADDR_ADDRMASK:
 779                         /* XXX ignore af? */
 780                         PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
 781                         PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
 782                         break;
 783         }
 784
 785         PF_MD5_UPD(pfr, port[0]);
 786         PF_MD5_UPD(pfr, port[1]);
 787         PF_MD5_UPD(pfr, neg);
 788         PF_MD5_UPD(pfr, port_op);
 789 }
 790
 791 static void
 792 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
 793 {
 794         u_int16_t x;
 795         u_int32_t y;
 796
 797         pf_hash_rule_addr(ctx, &rule->src);
 798         pf_hash_rule_addr(ctx, &rule->dst);
 799         PF_MD5_UPD_STR(rule, label);
 800         PF_MD5_UPD_STR(rule, ifname);
 801         PF_MD5_UPD_STR(rule, match_tagname);
 802         PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
 803         PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
 804         PF_MD5_UPD_HTONL(rule, prob, y);
 805         PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
 806         PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
 807         PF_MD5_UPD(rule, uid.op);
 808         PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
 809         PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
 810         PF_MD5_UPD(rule, gid.op);
 811         PF_MD5_UPD_HTONL(rule, rule_flag, y);
 812         PF_MD5_UPD(rule, action);
 813         PF_MD5_UPD(rule, direction);
 814         PF_MD5_UPD(rule, af);
 815         PF_MD5_UPD(rule, quick);
 816         PF_MD5_UPD(rule, ifnot);
 817         PF_MD5_UPD(rule, match_tag_not);
 818         PF_MD5_UPD(rule, natpass);
 819         PF_MD5_UPD(rule, keep_state);
 820         PF_MD5_UPD(rule, proto);
 821         PF_MD5_UPD(rule, type);
 822         PF_MD5_UPD(rule, code);
 823         PF_MD5_UPD(rule, flags);
 824         PF_MD5_UPD(rule, flagset);
 825         PF_MD5_UPD(rule, allow_opts);
 826         PF_MD5_UPD(rule, rt);
 827         PF_MD5_UPD(rule, tos);
 828 }
 829
 830 static int
 831 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
 832 {
 833         struct pf_ruleset       *rs;
 834         struct pf_rule          *rule, **old_array;
 835         struct pf_rulequeue     *old_rules;
 836         int                      error;
 837         u_int32_t                old_rcount;
 838
 839         PF_RULES_WASSERT();
 840
 841         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 842                 return (EINVAL);
 843         rs = pf_find_ruleset(anchor);
 844         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 845             ticket != rs->rules[rs_num].inactive.ticket)
 846                 return (EBUSY);
 847
 848         /* Calculate checksum for the main ruleset */
 849         if (rs == &pf_main_ruleset) {
 850                 error = pf_setup_pfsync_matching(rs);
 851                 if (error != 0)
 852                         return (error);
 853         }
 854
 855         /* Swap rules, keep the old. */
 856         old_rules = rs->rules[rs_num].active.ptr;
 857         old_rcount = rs->rules[rs_num].active.rcount;
 858         old_array = rs->rules[rs_num].active.ptr_array;
 859
 860         rs->rules[rs_num].active.ptr =
 861             rs->rules[rs_num].inactive.ptr;
 862         rs->rules[rs_num].active.ptr_array =
 863             rs->rules[rs_num].inactive.ptr_array;
 864         rs->rules[rs_num].active.rcount =
 865             rs->rules[rs_num].inactive.rcount;
 866         rs->rules[rs_num].inactive.ptr = old_rules;
 867         rs->rules[rs_num].inactive.ptr_array = old_array;
 868         rs->rules[rs_num].inactive.rcount = old_rcount;
 869
 870         rs->rules[rs_num].active.ticket =
 871             rs->rules[rs_num].inactive.ticket;
 872         pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
 873
 874
 875         /* Purge the old rule list. */
 876         while ((rule = TAILQ_FIRST(old_rules)) != NULL)
 877                 pf_unlink_rule(old_rules, rule);
 878         if (rs->rules[rs_num].inactive.ptr_array)
 879                 free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
 880         rs->rules[rs_num].inactive.ptr_array = NULL;
 881         rs->rules[rs_num].inactive.rcount = 0;
 882         rs->rules[rs_num].inactive.open = 0;
 883         pf_remove_if_empty_ruleset(rs);
 884
 885         return (0);
 886 }
 887
 888 static int
 889 pf_setup_pfsync_matching(struct pf_ruleset *rs)
 890 {
 891         MD5_CTX                  ctx;
 892         struct pf_rule          *rule;
 893         int                      rs_cnt;
 894         u_int8_t                 digest[PF_MD5_DIGEST_LENGTH];
 895
 896         MD5Init(&ctx);
 897         for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
 898                 /* XXX PF_RULESET_SCRUB as well? */
 899                 if (rs_cnt == PF_RULESET_SCRUB)
 900                         continue;
 901
 902                 if (rs->rules[rs_cnt].inactive.ptr_array)
 903                         free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
 904                 rs->rules[rs_cnt].inactive.ptr_array = NULL;
 905
 906                 if (rs->rules[rs_cnt].inactive.rcount) {
 907                         rs->rules[rs_cnt].inactive.ptr_array =
 908                             malloc(sizeof(caddr_t) *
 909                             rs->rules[rs_cnt].inactive.rcount,
 910                             M_TEMP, M_NOWAIT);
 911
 912                         if (!rs->rules[rs_cnt].inactive.ptr_array)
 913                                 return (ENOMEM);
 914                 }
 915
 916                 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
 917                     entries) {
 918                         pf_hash_rule(&ctx, rule);
 919                         (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
 920                 }
 921         }
 922
 923         MD5Final(digest, &ctx);
 924         memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
 925         return (0);
 926 }
 927
 928 static int
 929 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
 930     sa_family_t af)
 931 {
 932         int error = 0;
 933
 934         switch (addr->type) {
 935         case PF_ADDR_TABLE:
 936                 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
 937                 if (addr->p.tbl == NULL)
 938                         error = ENOMEM;
 939                 break;
 940         case PF_ADDR_DYNIFTL:
 941                 error = pfi_dynaddr_setup(addr, af);
 942                 break;
 943         }
 944
 945         return (error);
 946 }
 947
 948 static void
 949 pf_addr_copyout(struct pf_addr_wrap *addr)
 950 {
 951
 952         switch (addr->type) {
 953         case PF_ADDR_DYNIFTL:
 954                 pfi_dynaddr_copyout(addr);
 955                 break;
 956         case PF_ADDR_TABLE:
 957                 pf_tbladdr_copyout(addr);
 958                 break;
 959         }
 960 }
 961
 962 static int
 963 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
 964 {
 965         int                      error = 0;
 966
 967         /* XXX keep in sync with switch() below */
 968         if (securelevel_gt(td->td_ucred, 2))
 969                 switch (cmd) {
 970                 case DIOCGETRULES:
 971                 case DIOCGETRULE:
 972                 case DIOCGETADDRS:
 973                 case DIOCGETADDR:
 974                 case DIOCGETSTATE:
 975                 case DIOCSETSTATUSIF:
 976                 case DIOCGETSTATUS:
 977                 case DIOCCLRSTATUS:
 978                 case DIOCNATLOOK:
 979                 case DIOCSETDEBUG:
 980                 case DIOCGETSTATES:
 981                 case DIOCGETTIMEOUT:
 982                 case DIOCCLRRULECTRS:
 983                 case DIOCGETLIMIT:
 984                 case DIOCGETALTQS:
 985                 case DIOCGETALTQ:
 986                 case DIOCGETQSTATS:
 987                 case DIOCGETRULESETS:
 988                 case DIOCGETRULESET:
 989                 case DIOCRGETTABLES:
 990                 case DIOCRGETTSTATS:
 991                 case DIOCRCLRTSTATS:
 992                 case DIOCRCLRADDRS:
 993                 case DIOCRADDADDRS:
 994                 case DIOCRDELADDRS:
 995                 case DIOCRSETADDRS:
 996                 case DIOCRGETADDRS:
 997                 case DIOCRGETASTATS:
 998                 case DIOCRCLRASTATS:
 999                 case DIOCRTSTADDRS:
1000                 case DIOCOSFPGET:
1001                 case DIOCGETSRCNODES:
1002                 case DIOCCLRSRCNODES:
1003                 case DIOCIGETIFACES:
1004                 case DIOCGIFSPEED:
1005                 case DIOCSETIFFLAG:
1006                 case DIOCCLRIFFLAG:
1007                         break;
1008                 case DIOCRCLRTABLES:
1009                 case DIOCRADDTABLES:
1010                 case DIOCRDELTABLES:
1011                 case DIOCRSETTFLAGS:
1012                         if (((struct pfioc_table *)addr)->pfrio_flags &
1013                             PFR_FLAG_DUMMY)
1014                                 break; /* dummy operation ok */
1015                         return (EPERM);
1016                 default:
1017                         return (EPERM);
1018                 }
1019
1020         if (!(flags & FWRITE))
1021                 switch (cmd) {
1022                 case DIOCGETRULES:
1023                 case DIOCGETADDRS:
1024                 case DIOCGETADDR:
1025                 case DIOCGETSTATE:
1026                 case DIOCGETSTATUS:
1027                 case DIOCGETSTATES:
1028                 case DIOCGETTIMEOUT:
1029                 case DIOCGETLIMIT:
1030                 case DIOCGETALTQS:
1031                 case DIOCGETALTQ:
1032                 case DIOCGETQSTATS:
1033                 case DIOCGETRULESETS:
1034                 case DIOCGETRULESET:
1035                 case DIOCNATLOOK:
1036                 case DIOCRGETTABLES:
1037                 case DIOCRGETTSTATS:
1038                 case DIOCRGETADDRS:
1039                 case DIOCRGETASTATS:
1040                 case DIOCRTSTADDRS:
1041                 case DIOCOSFPGET:
1042                 case DIOCGETSRCNODES:
1043                 case DIOCIGETIFACES:
1044                 case DIOCGIFSPEED:
1045                         break;
1046                 case DIOCRCLRTABLES:
1047                 case DIOCRADDTABLES:
1048                 case DIOCRDELTABLES:
1049                 case DIOCRCLRTSTATS:
1050                 case DIOCRCLRADDRS:
1051                 case DIOCRADDADDRS:
1052                 case DIOCRDELADDRS:
1053                 case DIOCRSETADDRS:
1054                 case DIOCRSETTFLAGS:
1055                         if (((struct pfioc_table *)addr)->pfrio_flags &
1056                             PFR_FLAG_DUMMY) {
1057                                 flags |= FWRITE; /* need write lock for dummy */
1058                                 break; /* dummy operation ok */
1059                         }
1060                         return (EACCES);
1061                 case DIOCGETRULE:
1062                         if (((struct pfioc_rule *)addr)->action ==
1063                             PF_GET_CLR_CNTR)
1064                                 return (EACCES);
1065                         break;
1066                 default:
1067                         return (EACCES);
1068                 }
1069
1070         CURVNET_SET(TD_TO_VNET(td));
1071
1072         switch (cmd) {
1073         case DIOCSTART:
1074                 PF_RULES_WLOCK();
1075                 if (V_pf_status.running)
1076                         error = EEXIST;
1077                 else {
1078                         int cpu;
1079
1080                         PF_RULES_WUNLOCK();
1081                         error = hook_pf();
1082                         if (error) {
1083                                 DPFPRINTF(PF_DEBUG_MISC,
1084                                     ("pf: pfil registration failed\n"));
1085                                 break;
1086                         }
1087                         PF_RULES_WLOCK();
1088                         V_pf_status.running = 1;
1089                         V_pf_status.since = time_second;
1090
1091                         CPU_FOREACH(cpu)
1092                                 V_pf_stateid[cpu] = time_second;
1093
1094                         DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1095                 }
1096                 PF_RULES_WUNLOCK();
1097                 break;
1098
1099         case DIOCSTOP:
1100                 PF_RULES_WLOCK();
1101                 if (!V_pf_status.running)
1102                         error = ENOENT;
1103                 else {
1104                         V_pf_status.running = 0;
1105                         PF_RULES_WUNLOCK();
1106                         error = dehook_pf();
1107                         if (error) {
1108                                 V_pf_status.running = 1;
1109                                 DPFPRINTF(PF_DEBUG_MISC,
1110                                     ("pf: pfil unregistration failed\n"));
1111                         }
1112                         PF_RULES_WLOCK();
1113                         V_pf_status.since = time_second;
1114                         DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1115                 }
1116                 PF_RULES_WUNLOCK();
1117                 break;
1118
1119         case DIOCADDRULE: {
1120                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1121                 struct pf_ruleset       *ruleset;
1122                 struct pf_rule          *rule, *tail;
1123                 struct pf_pooladdr      *pa;
1124                 struct pfi_kif          *kif = NULL;
1125                 int                      rs_num;
1126
1127                 if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1128                         error = EINVAL;
1129                         break;
1130                 }
1131 #ifndef INET
1132                 if (pr->rule.af == AF_INET) {
1133                         error = EAFNOSUPPORT;
1134                         break;
1135                 }
1136 #endif /* INET */
1137 #ifndef INET6
1138                 if (pr->rule.af == AF_INET6) {
1139                         error = EAFNOSUPPORT;
1140                         break;
1141                 }
1142 #endif /* INET6 */
1143
1144                 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1145                 bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1146                 if (rule->ifname[0])
1147                         kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1148                 rule->cuid = td->td_ucred->cr_ruid;
1149                 rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1150                 TAILQ_INIT(&rule->rpool.list);
1151
1152 #define ERROUT(x)       { error = (x); goto DIOCADDRULE_error; }
1153
1154                 PF_RULES_WLOCK();
1155                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1156                 ruleset = pf_find_ruleset(pr->anchor);
1157                 if (ruleset == NULL)
1158                         ERROUT(EINVAL);
1159                 rs_num = pf_get_ruleset_number(pr->rule.action);
1160                 if (rs_num >= PF_RULESET_MAX)
1161                         ERROUT(EINVAL);
1162                 if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1163                         DPFPRINTF(PF_DEBUG_MISC,
1164                             ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1165                             ruleset->rules[rs_num].inactive.ticket));
1166                         ERROUT(EBUSY);
1167                 }
1168                 if (pr->pool_ticket != V_ticket_pabuf) {
1169                         DPFPRINTF(PF_DEBUG_MISC,
1170                             ("pool_ticket: %d != %d\n", pr->pool_ticket,
1171                             V_ticket_pabuf));
1172                         ERROUT(EBUSY);
1173                 }
1174
1175                 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1176                     pf_rulequeue);
1177                 if (tail)
1178                         rule->nr = tail->nr + 1;
1179                 else
1180                         rule->nr = 0;
1181                 if (rule->ifname[0]) {
1182                         rule->kif = pfi_kif_attach(kif, rule->ifname);
1183                         pfi_kif_ref(rule->kif);
1184                 } else
1185                         rule->kif = NULL;
1186
1187                 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1188                         error = EBUSY;
1189
1190 #ifdef ALTQ
1191                 /* set queue IDs */
1192                 if (rule->qname[0] != 0) {
1193                         if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1194                                 error = EBUSY;
1195                         else if (rule->pqname[0] != 0) {
1196                                 if ((rule->pqid =
1197                                     pf_qname2qid(rule->pqname)) == 0)
1198                                         error = EBUSY;
1199                         } else
1200                                 rule->pqid = rule->qid;
1201                 }
1202 #endif
1203                 if (rule->tagname[0])
1204                         if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1205                                 error = EBUSY;
1206                 if (rule->match_tagname[0])
1207                         if ((rule->match_tag =
1208                             pf_tagname2tag(rule->match_tagname)) == 0)
1209                                 error = EBUSY;
1210                 if (rule->rt && !rule->direction)
1211                         error = EINVAL;
1212                 if (!rule->log)
1213                         rule->logif = 0;
1214                 if (rule->logif >= PFLOGIFS_MAX)
1215                         error = EINVAL;
1216                 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1217                         error = ENOMEM;
1218                 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1219                         error = ENOMEM;
1220                 if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1221                         error = EINVAL;
1222                 TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1223                         if (pa->addr.type == PF_ADDR_TABLE) {
1224                                 pa->addr.p.tbl = pfr_attach_table(ruleset,
1225                                     pa->addr.v.tblname);
1226                                 if (pa->addr.p.tbl == NULL)
1227                                         error = ENOMEM;
1228                         }
1229
1230                 if (rule->overload_tblname[0]) {
1231                         if ((rule->overload_tbl = pfr_attach_table(ruleset,
1232                             rule->overload_tblname)) == NULL)
1233                                 error = EINVAL;
1234                         else
1235                                 rule->overload_tbl->pfrkt_flags |=
1236                                     PFR_TFLAG_ACTIVE;
1237                 }
1238
1239                 pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1240                 if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1241                     (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1242                     (rule->rt > PF_FASTROUTE)) &&
1243                     (TAILQ_FIRST(&rule->rpool.list) == NULL))
1244                         error = EINVAL;
1245
1246                 if (error) {
1247                         pf_free_rule(rule);
1248                         PF_RULES_WUNLOCK();
1249                         break;
1250                 }
1251
1252                 rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1253                 rule->evaluations = rule->packets[0] = rule->packets[1] =
1254                     rule->bytes[0] = rule->bytes[1] = 0;
1255                 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1256                     rule, entries);
1257                 ruleset->rules[rs_num].inactive.rcount++;
1258                 PF_RULES_WUNLOCK();
1259                 break;
1260
1261 #undef ERROUT
1262 DIOCADDRULE_error:
1263                 PF_RULES_WUNLOCK();
1264                 free(rule, M_PFRULE);
1265                 if (kif)
1266                         free(kif, PFI_MTYPE);
1267                 break;
1268         }
1269
1270         case DIOCGETRULES: {
1271                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1272                 struct pf_ruleset       *ruleset;
1273                 struct pf_rule          *tail;
1274                 int                      rs_num;
1275
1276                 PF_RULES_WLOCK();
1277                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1278                 ruleset = pf_find_ruleset(pr->anchor);
1279                 if (ruleset == NULL) {
1280                         PF_RULES_WUNLOCK();
1281                         error = EINVAL;
1282                         break;
1283                 }
1284                 rs_num = pf_get_ruleset_number(pr->rule.action);
1285                 if (rs_num >= PF_RULESET_MAX) {
1286                         PF_RULES_WUNLOCK();
1287                         error = EINVAL;
1288                         break;
1289                 }
1290                 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1291                     pf_rulequeue);
1292                 if (tail)
1293                         pr->nr = tail->nr + 1;
1294                 else
1295                         pr->nr = 0;
1296                 pr->ticket = ruleset->rules[rs_num].active.ticket;
1297                 PF_RULES_WUNLOCK();
1298                 break;
1299         }
1300
1301         case DIOCGETRULE: {
1302                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1303                 struct pf_ruleset       *ruleset;
1304                 struct pf_rule          *rule;
1305                 int                      rs_num, i;
1306
1307                 PF_RULES_WLOCK();
1308                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1309                 ruleset = pf_find_ruleset(pr->anchor);
1310                 if (ruleset == NULL) {
1311                         PF_RULES_WUNLOCK();
1312                         error = EINVAL;
1313                         break;
1314                 }
1315                 rs_num = pf_get_ruleset_number(pr->rule.action);
1316                 if (rs_num >= PF_RULESET_MAX) {
1317                         PF_RULES_WUNLOCK();
1318                         error = EINVAL;
1319                         break;
1320                 }
1321                 if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1322                         PF_RULES_WUNLOCK();
1323                         error = EBUSY;
1324                         break;
1325                 }
1326                 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1327                 while ((rule != NULL) && (rule->nr != pr->nr))
1328                         rule = TAILQ_NEXT(rule, entries);
1329                 if (rule == NULL) {
1330                         PF_RULES_WUNLOCK();
1331                         error = EBUSY;
1332                         break;
1333                 }
1334                 bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1335                 if (pf_anchor_copyout(ruleset, rule, pr)) {
1336                         PF_RULES_WUNLOCK();
1337                         error = EBUSY;
1338                         break;
1339                 }
1340                 pf_addr_copyout(&pr->rule.src.addr);
1341                 pf_addr_copyout(&pr->rule.dst.addr);
1342                 for (i = 0; i < PF_SKIP_COUNT; ++i)
1343                         if (rule->skip[i].ptr == NULL)
1344                                 pr->rule.skip[i].nr = -1;
1345                         else
1346                                 pr->rule.skip[i].nr =
1347                                     rule->skip[i].ptr->nr;
1348
1349                 if (pr->action == PF_GET_CLR_CNTR) {
1350                         rule->evaluations = 0;
1351                         rule->packets[0] = rule->packets[1] = 0;
1352                         rule->bytes[0] = rule->bytes[1] = 0;
1353                         rule->states_tot = 0;
1354                 }
1355                 PF_RULES_WUNLOCK();
1356                 break;
1357         }
1358
1359         case DIOCCHANGERULE: {
1360                 struct pfioc_rule       *pcr = (struct pfioc_rule *)addr;
1361                 struct pf_ruleset       *ruleset;
1362                 struct pf_rule          *oldrule = NULL, *newrule = NULL;
1363                 struct pfi_kif          *kif = NULL;
1364                 struct pf_pooladdr      *pa;
1365                 u_int32_t                nr = 0;
1366                 int                      rs_num;
1367
1368                 if (pcr->action < PF_CHANGE_ADD_HEAD ||
1369                     pcr->action > PF_CHANGE_GET_TICKET) {
1370                         error = EINVAL;
1371                         break;
1372                 }
1373                 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1374                         error = EINVAL;
1375                         break;
1376                 }
1377
1378                 if (pcr->action != PF_CHANGE_REMOVE) {
1379 #ifndef INET
1380                         if (pcr->rule.af == AF_INET) {
1381                                 error = EAFNOSUPPORT;
1382                                 break;
1383                         }
1384 #endif /* INET */
1385 #ifndef INET6
1386                         if (pcr->rule.af == AF_INET6) {
1387                                 error = EAFNOSUPPORT;
1388                                 break;
1389                         }
1390 #endif /* INET6 */
1391                         newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1392                         bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1393                         newrule->cuid = td->td_ucred->cr_ruid;
1394                         newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1395                         TAILQ_INIT(&newrule->rpool.list);
1396                         /* Initialize refcounting. */
1397                         newrule->states_cur = 0;
1398                         newrule->entries.tqe_prev = NULL;
1399
1400                         if (newrule->ifname[0])
1401                                 kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1402                 }
1403
1404 #define ERROUT(x)       { error = (x); goto DIOCCHANGERULE_error; }
1405
1406                 PF_RULES_WLOCK();
1407                 if (!(pcr->action == PF_CHANGE_REMOVE ||
1408                     pcr->action == PF_CHANGE_GET_TICKET) &&
1409                     pcr->pool_ticket != V_ticket_pabuf)
1410                         ERROUT(EBUSY);
1411
1412                 ruleset = pf_find_ruleset(pcr->anchor);
1413                 if (ruleset == NULL)
1414                         ERROUT(EINVAL);
1415
1416                 rs_num = pf_get_ruleset_number(pcr->rule.action);
1417                 if (rs_num >= PF_RULESET_MAX)
1418                         ERROUT(EINVAL);
1419
1420                 if (pcr->action == PF_CHANGE_GET_TICKET) {
1421                         pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1422                         ERROUT(0);
1423                 } else if (pcr->ticket !=
1424                             ruleset->rules[rs_num].active.ticket)
1425                                 ERROUT(EINVAL);
1426
1427                 if (pcr->action != PF_CHANGE_REMOVE) {
1428                         if (newrule->ifname[0]) {
1429                                 newrule->kif = pfi_kif_attach(kif,
1430                                     newrule->ifname);
1431                                 pfi_kif_ref(newrule->kif);
1432                         } else
1433                                 newrule->kif = NULL;
1434
1435                         if (newrule->rtableid > 0 &&
1436                             newrule->rtableid >= rt_numfibs)
1437                                 error = EBUSY;
1438
1439 #ifdef ALTQ
1440                         /* set queue IDs */
1441                         if (newrule->qname[0] != 0) {
1442                                 if ((newrule->qid =
1443                                     pf_qname2qid(newrule->qname)) == 0)
1444                                         error = EBUSY;
1445                                 else if (newrule->pqname[0] != 0) {
1446                                         if ((newrule->pqid =
1447                                             pf_qname2qid(newrule->pqname)) == 0)
1448                                                 error = EBUSY;
1449                                 } else
1450                                         newrule->pqid = newrule->qid;
1451                         }
1452 #endif /* ALTQ */
1453                         if (newrule->tagname[0])
1454                                 if ((newrule->tag =
1455                                     pf_tagname2tag(newrule->tagname)) == 0)
1456                                         error = EBUSY;
1457                         if (newrule->match_tagname[0])
1458                                 if ((newrule->match_tag = pf_tagname2tag(
1459                                     newrule->match_tagname)) == 0)
1460                                         error = EBUSY;
1461                         if (newrule->rt && !newrule->direction)
1462                                 error = EINVAL;
1463                         if (!newrule->log)
1464                                 newrule->logif = 0;
1465                         if (newrule->logif >= PFLOGIFS_MAX)
1466                                 error = EINVAL;
1467                         if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1468                                 error = ENOMEM;
1469                         if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1470                                 error = ENOMEM;
1471                         if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1472                                 error = EINVAL;
1473                         TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1474                                 if (pa->addr.type == PF_ADDR_TABLE) {
1475                                         pa->addr.p.tbl =
1476                                             pfr_attach_table(ruleset,
1477                                             pa->addr.v.tblname);
1478                                         if (pa->addr.p.tbl == NULL)
1479                                                 error = ENOMEM;
1480                                 }
1481
1482                         if (newrule->overload_tblname[0]) {
1483                                 if ((newrule->overload_tbl = pfr_attach_table(
1484                                     ruleset, newrule->overload_tblname)) ==
1485                                     NULL)
1486                                         error = EINVAL;
1487                                 else
1488                                         newrule->overload_tbl->pfrkt_flags |=
1489                                             PFR_TFLAG_ACTIVE;
1490                         }
1491
1492                         pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1493                         if (((((newrule->action == PF_NAT) ||
1494                             (newrule->action == PF_RDR) ||
1495                             (newrule->action == PF_BINAT) ||
1496                             (newrule->rt > PF_FASTROUTE)) &&
1497                             !newrule->anchor)) &&
1498                             (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1499                                 error = EINVAL;
1500
1501                         if (error) {
1502                                 pf_free_rule(newrule);
1503                                 PF_RULES_WUNLOCK();
1504                                 break;
1505                         }
1506
1507                         newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1508                         newrule->evaluations = 0;
1509                         newrule->packets[0] = newrule->packets[1] = 0;
1510                         newrule->bytes[0] = newrule->bytes[1] = 0;
1511                 }
1512                 pf_empty_pool(&V_pf_pabuf);
1513
1514                 if (pcr->action == PF_CHANGE_ADD_HEAD)
1515                         oldrule = TAILQ_FIRST(
1516                             ruleset->rules[rs_num].active.ptr);
1517                 else if (pcr->action == PF_CHANGE_ADD_TAIL)
1518                         oldrule = TAILQ_LAST(
1519                             ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1520                 else {
1521                         oldrule = TAILQ_FIRST(
1522                             ruleset->rules[rs_num].active.ptr);
1523                         while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1524                                 oldrule = TAILQ_NEXT(oldrule, entries);
1525                         if (oldrule == NULL) {
1526                                 if (newrule != NULL)
1527                                         pf_free_rule(newrule);
1528                                 PF_RULES_WUNLOCK();
1529                                 error = EINVAL;
1530                                 break;
1531                         }
1532                 }
1533
1534                 if (pcr->action == PF_CHANGE_REMOVE) {
1535                         pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1536                             oldrule);
1537                         ruleset->rules[rs_num].active.rcount--;
1538                 } else {
1539                         if (oldrule == NULL)
1540                                 TAILQ_INSERT_TAIL(
1541                                     ruleset->rules[rs_num].active.ptr,
1542                                     newrule, entries);
1543                         else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1544                             pcr->action == PF_CHANGE_ADD_BEFORE)
1545                                 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1546                         else
1547                                 TAILQ_INSERT_AFTER(
1548                                     ruleset->rules[rs_num].active.ptr,
1549                                     oldrule, newrule, entries);
1550                         ruleset->rules[rs_num].active.rcount++;
1551                 }
1552
1553                 nr = 0;
1554                 TAILQ_FOREACH(oldrule,
1555                     ruleset->rules[rs_num].active.ptr, entries)
1556                         oldrule->nr = nr++;
1557
1558                 ruleset->rules[rs_num].active.ticket++;
1559
1560                 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1561                 pf_remove_if_empty_ruleset(ruleset);
1562
1563                 PF_RULES_WUNLOCK();
1564                 break;
1565
1566 #undef ERROUT
1567 DIOCCHANGERULE_error:
1568                 PF_RULES_WUNLOCK();
1569                 if (newrule != NULL)
1570                         free(newrule, M_PFRULE);
1571                 if (kif != NULL)
1572                         free(kif, PFI_MTYPE);
1573                 break;
1574         }
1575
1576         case DIOCCLRSTATES: {
1577                 struct pf_state         *s;
1578                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1579                 u_int                    i, killed = 0;
1580
1581                 for (i = 0; i <= V_pf_hashmask; i++) {
1582                         struct pf_idhash *ih = &V_pf_idhash[i];
1583
1584 relock_DIOCCLRSTATES:
1585                         PF_HASHROW_LOCK(ih);
1586                         LIST_FOREACH(s, &ih->states, entry)
1587                                 if (!psk->psk_ifname[0] ||
1588                                     !strcmp(psk->psk_ifname,
1589                                     s->kif->pfik_name)) {
1590                                         /*
1591                                          * Don't send out individual
1592                                          * delete messages.
1593                                          */
1594                                         s->state_flags |= PFSTATE_NOSYNC;
1595                                         pf_unlink_state(s, PF_ENTER_LOCKED);
1596                                         killed++;
1597                                         goto relock_DIOCCLRSTATES;
1598                                 }
1599                         PF_HASHROW_UNLOCK(ih);
1600                 }
1601                 psk->psk_killed = killed;
1602                 if (pfsync_clear_states_ptr != NULL)
1603                         pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
1604                 break;
1605         }
1606
1607         case DIOCKILLSTATES: {
1608                 struct pf_state         *s;
1609                 struct pf_state_key     *sk;
1610                 struct pf_addr          *srcaddr, *dstaddr;
1611                 u_int16_t                srcport, dstport;
1612                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1613                 u_int                    i, killed = 0;
1614
1615                 if (psk->psk_pfcmp.id) {
1616                         if (psk->psk_pfcmp.creatorid == 0)
1617                                 psk->psk_pfcmp.creatorid = V_pf_status.hostid;
1618                         if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
1619                             psk->psk_pfcmp.creatorid))) {
1620                                 pf_unlink_state(s, PF_ENTER_LOCKED);
1621                                 psk->psk_killed = 1;
1622                         }
1623                         break;
1624                 }
1625
1626                 for (i = 0; i <= V_pf_hashmask; i++) {
1627                         struct pf_idhash *ih = &V_pf_idhash[i];
1628
1629 relock_DIOCKILLSTATES:
1630                         PF_HASHROW_LOCK(ih);
1631                         LIST_FOREACH(s, &ih->states, entry) {
1632                                 sk = s->key[PF_SK_WIRE];
1633                                 if (s->direction == PF_OUT) {
1634                                         srcaddr = &sk->addr[1];
1635                                         dstaddr = &sk->addr[0];
1636                                         srcport = sk->port[0];
1637                                         dstport = sk->port[0];
1638                                 } else {
1639                                         srcaddr = &sk->addr[0];
1640                                         dstaddr = &sk->addr[1];
1641                                         srcport = sk->port[0];
1642                                         dstport = sk->port[0];
1643                                 }
1644
1645                                 if ((!psk->psk_af || sk->af == psk->psk_af)
1646                                     && (!psk->psk_proto || psk->psk_proto ==
1647                                     sk->proto) &&
1648                                     PF_MATCHA(psk->psk_src.neg,
1649                                     &psk->psk_src.addr.v.a.addr,
1650                                     &psk->psk_src.addr.v.a.mask,
1651                                     srcaddr, sk->af) &&
1652                                     PF_MATCHA(psk->psk_dst.neg,
1653                                     &psk->psk_dst.addr.v.a.addr,
1654                                     &psk->psk_dst.addr.v.a.mask,
1655                                     dstaddr, sk->af) &&
1656                                     (psk->psk_src.port_op == 0 ||
1657                                     pf_match_port(psk->psk_src.port_op,
1658                                     psk->psk_src.port[0], psk->psk_src.port[1],
1659                                     srcport)) &&
1660                                     (psk->psk_dst.port_op == 0 ||
1661                                     pf_match_port(psk->psk_dst.port_op,
1662                                     psk->psk_dst.port[0], psk->psk_dst.port[1],
1663                                     dstport)) &&
1664                                     (!psk->psk_label[0] ||
1665                                     (s->rule.ptr->label[0] &&
1666                                     !strcmp(psk->psk_label,
1667                                     s->rule.ptr->label))) &&
1668                                     (!psk->psk_ifname[0] ||
1669                                     !strcmp(psk->psk_ifname,
1670                                     s->kif->pfik_name))) {
1671                                         pf_unlink_state(s, PF_ENTER_LOCKED);
1672                                         killed++;
1673                                         goto relock_DIOCKILLSTATES;
1674                                 }
1675                         }
1676                         PF_HASHROW_UNLOCK(ih);
1677                 }
1678                 psk->psk_killed = killed;
1679                 break;
1680         }
1681
1682         case DIOCADDSTATE: {
1683                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
1684                 struct pfsync_state     *sp = &ps->state;
1685
1686                 if (sp->timeout >= PFTM_MAX) {
1687                         error = EINVAL;
1688                         break;
1689                 }
1690                 if (pfsync_state_import_ptr != NULL) {
1691                         PF_RULES_RLOCK();
1692                         error = pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
1693                         PF_RULES_RUNLOCK();
1694                 } else
1695                         error = EOPNOTSUPP;
1696                 break;
1697         }
1698
1699         case DIOCGETSTATE: {
1700                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
1701                 struct pf_state         *s;
1702
1703                 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
1704                 if (s == NULL) {
1705                         error = ENOENT;
1706                         break;
1707                 }
1708
1709                 pfsync_state_export(&ps->state, s);
1710                 PF_STATE_UNLOCK(s);
1711                 break;
1712         }
1713
1714         case DIOCGETSTATES: {
1715                 struct pfioc_states     *ps = (struct pfioc_states *)addr;
1716                 struct pf_state         *s;
1717                 struct pfsync_state     *pstore, *p;
1718                 int i, nr;
1719
1720                 if (ps->ps_len == 0) {
1721                         nr = uma_zone_get_cur(V_pf_state_z);
1722                         ps->ps_len = sizeof(struct pfsync_state) * nr;
1723                         break;
1724                 }
1725
1726                 p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK);
1727                 nr = 0;
1728
1729                 for (i = 0; i <= V_pf_hashmask; i++) {
1730                         struct pf_idhash *ih = &V_pf_idhash[i];
1731
1732                         PF_HASHROW_LOCK(ih);
1733                         LIST_FOREACH(s, &ih->states, entry) {
1734
1735                                 if (s->timeout == PFTM_UNLINKED)
1736                                         continue;
1737
1738                                 if ((nr+1) * sizeof(*p) > ps->ps_len) {
1739                                         PF_HASHROW_UNLOCK(ih);
1740                                         goto DIOCGETSTATES_full;
1741                                 }
1742                                 pfsync_state_export(p, s);
1743                                 p++;
1744                                 nr++;
1745                         }
1746                         PF_HASHROW_UNLOCK(ih);
1747                 }
1748 DIOCGETSTATES_full:
1749                 error = copyout(pstore, ps->ps_states,
1750                     sizeof(struct pfsync_state) * nr);
1751                 if (error) {
1752                         free(pstore, M_TEMP);
1753                         break;
1754                 }
1755                 ps->ps_len = sizeof(struct pfsync_state) * nr;
1756                 free(pstore, M_TEMP);
1757
1758                 break;
1759         }
1760
1761         case DIOCGETSTATUS: {
1762                 struct pf_status *s = (struct pf_status *)addr;
1763                 PF_RULES_RLOCK();
1764                 bcopy(&V_pf_status, s, sizeof(struct pf_status));
1765                 pfi_update_status(s->ifname, s);
1766                 PF_RULES_RUNLOCK();
1767                 break;
1768         }
1769
1770         case DIOCSETSTATUSIF: {
1771                 struct pfioc_if *pi = (struct pfioc_if *)addr;
1772
1773                 if (pi->ifname[0] == 0) {
1774                         bzero(V_pf_status.ifname, IFNAMSIZ);
1775                         break;
1776                 }
1777                 PF_RULES_WLOCK();
1778                 strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
1779                 PF_RULES_WUNLOCK();
1780                 break;
1781         }
1782
1783         case DIOCCLRSTATUS: {
1784                 PF_RULES_WLOCK();
1785                 bzero(V_pf_status.counters, sizeof(V_pf_status.counters));
1786                 bzero(V_pf_status.fcounters, sizeof(V_pf_status.fcounters));
1787                 bzero(V_pf_status.scounters, sizeof(V_pf_status.scounters));
1788                 V_pf_status.since = time_second;
1789                 if (*V_pf_status.ifname)
1790                         pfi_update_status(V_pf_status.ifname, NULL);
1791                 PF_RULES_WUNLOCK();
1792                 break;
1793         }
1794
1795         case DIOCNATLOOK: {
1796                 struct pfioc_natlook    *pnl = (struct pfioc_natlook *)addr;
1797                 struct pf_state_key     *sk;
1798                 struct pf_state         *state;
1799                 struct pf_state_key_cmp  key;
1800                 int                      m = 0, direction = pnl->direction;
1801                 int                      sidx, didx;
1802
1803                 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
1804                 sidx = (direction == PF_IN) ? 1 : 0;
1805                 didx = (direction == PF_IN) ? 0 : 1;
1806
1807                 if (!pnl->proto ||
1808                     PF_AZERO(&pnl->saddr, pnl->af) ||
1809                     PF_AZERO(&pnl->daddr, pnl->af) ||
1810                     ((pnl->proto == IPPROTO_TCP ||
1811                     pnl->proto == IPPROTO_UDP) &&
1812                     (!pnl->dport || !pnl->sport)))
1813                         error = EINVAL;
1814                 else {
1815                         bzero(&key, sizeof(key));
1816                         key.af = pnl->af;
1817                         key.proto = pnl->proto;
1818                         PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1819                         key.port[sidx] = pnl->sport;
1820                         PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1821                         key.port[didx] = pnl->dport;
1822
1823                         state = pf_find_state_all(&key, direction, &m);
1824
1825                         if (m > 1)
1826                                 error = E2BIG;  /* more than one state */
1827                         else if (state != NULL) {
1828                                 /* XXXGL: not locked read */
1829                                 sk = state->key[sidx];
1830                                 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1831                                 pnl->rsport = sk->port[sidx];
1832                                 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1833                                 pnl->rdport = sk->port[didx];
1834                         } else
1835                                 error = ENOENT;
1836                 }
1837                 break;
1838         }
1839
1840         case DIOCSETTIMEOUT: {
1841                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
1842                 int              old;
1843
1844                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1845                     pt->seconds < 0) {
1846                         error = EINVAL;
1847                         break;
1848                 }
1849                 PF_RULES_WLOCK();
1850                 old = V_pf_default_rule.timeout[pt->timeout];
1851                 if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1852                         pt->seconds = 1;
1853                 V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
1854                 if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1855                         wakeup(pf_purge_thread);
1856                 pt->seconds = old;
1857                 PF_RULES_WUNLOCK();
1858                 break;
1859         }
1860
1861         case DIOCGETTIMEOUT: {
1862                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
1863
1864                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1865                         error = EINVAL;
1866                         break;
1867                 }
1868                 PF_RULES_RLOCK();
1869                 pt->seconds = V_pf_default_rule.timeout[pt->timeout];
1870                 PF_RULES_RUNLOCK();
1871                 break;
1872         }
1873
1874         case DIOCGETLIMIT: {
1875                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
1876
1877                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1878                         error = EINVAL;
1879                         break;
1880                 }
1881                 PF_RULES_RLOCK();
1882                 pl->limit = V_pf_limits[pl->index].limit;
1883                 PF_RULES_RUNLOCK();
1884                 break;
1885         }
1886
1887         case DIOCSETLIMIT: {
1888                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
1889                 int                      old_limit;
1890
1891                 PF_RULES_WLOCK();
1892                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1893                     V_pf_limits[pl->index].zone == NULL) {
1894                         PF_RULES_WUNLOCK();
1895                         error = EINVAL;
1896                         break;
1897                 }
1898                 uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
1899                 old_limit = V_pf_limits[pl->index].limit;
1900                 V_pf_limits[pl->index].limit = pl->limit;
1901                 pl->limit = old_limit;
1902                 PF_RULES_WUNLOCK();
1903                 break;
1904         }
1905
1906         case DIOCSETDEBUG: {
1907                 u_int32_t       *level = (u_int32_t *)addr;
1908
1909                 PF_RULES_WLOCK();
1910                 V_pf_status.debug = *level;
1911                 PF_RULES_WUNLOCK();
1912                 break;
1913         }
1914
1915         case DIOCCLRRULECTRS: {
1916                 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1917                 struct pf_ruleset       *ruleset = &pf_main_ruleset;
1918                 struct pf_rule          *rule;
1919
1920                 PF_RULES_WLOCK();
1921                 TAILQ_FOREACH(rule,
1922                     ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1923                         rule->evaluations = 0;
1924                         rule->packets[0] = rule->packets[1] = 0;
1925                         rule->bytes[0] = rule->bytes[1] = 0;
1926                 }
1927                 PF_RULES_WUNLOCK();
1928                 break;
1929         }
1930
1931         case DIOCGIFSPEED: {
1932                 struct pf_ifspeed       *psp = (struct pf_ifspeed *)addr;
1933                 struct pf_ifspeed       ps;
1934                 struct ifnet            *ifp;
1935
1936                 if (psp->ifname[0] != 0) {
1937                         /* Can we completely trust user-land? */
1938                         strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1939                         ifp = ifunit(ps.ifname);
1940                         if (ifp != NULL)
1941                                 psp->baudrate = ifp->if_baudrate;
1942                         else
1943                                 error = EINVAL;
1944                 } else
1945                         error = EINVAL;
1946                 break;
1947         }
1948
1949 #ifdef ALTQ
1950         case DIOCSTARTALTQ: {
1951                 struct pf_altq          *altq;
1952
1953                 PF_RULES_WLOCK();
1954                 /* enable all altq interfaces on active list */
1955                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1956                         if (altq->qname[0] == 0 && (altq->local_flags &
1957                             PFALTQ_FLAG_IF_REMOVED) == 0) {
1958                                 error = pf_enable_altq(altq);
1959                                 if (error != 0)
1960                                         break;
1961                         }
1962                 }
1963                 if (error == 0)
1964                         V_pf_altq_running = 1;
1965                 PF_RULES_WUNLOCK();
1966                 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
1967                 break;
1968         }
1969
1970         case DIOCSTOPALTQ: {
1971                 struct pf_altq          *altq;
1972
1973                 PF_RULES_WLOCK();
1974                 /* disable all altq interfaces on active list */
1975                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1976                         if (altq->qname[0] == 0 && (altq->local_flags &
1977                             PFALTQ_FLAG_IF_REMOVED) == 0) {
1978                                 error = pf_disable_altq(altq);
1979                                 if (error != 0)
1980                                         break;
1981                         }
1982                 }
1983                 if (error == 0)
1984                         V_pf_altq_running = 0;
1985                 PF_RULES_WUNLOCK();
1986                 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
1987                 break;
1988         }
1989
1990         case DIOCADDALTQ: {
1991                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
1992                 struct pf_altq          *altq, *a;
1993                 struct ifnet            *ifp;
1994
1995                 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK);
1996                 bcopy(&pa->altq, altq, sizeof(struct pf_altq));
1997                 altq->local_flags = 0;
1998
1999                 PF_RULES_WLOCK();
2000                 if (pa->ticket != V_ticket_altqs_inactive) {
2001                         PF_RULES_WUNLOCK();
2002                         free(altq, M_PFALTQ);
2003                         error = EBUSY;
2004                         break;
2005                 }
2006
2007                 /*
2008                  * if this is for a queue, find the discipline and
2009                  * copy the necessary fields
2010                  */
2011                 if (altq->qname[0] != 0) {
2012                         if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2013                                 PF_RULES_WUNLOCK();
2014                                 error = EBUSY;
2015                                 free(altq, M_PFALTQ);
2016                                 break;
2017                         }
2018                         altq->altq_disc = NULL;
2019                         TAILQ_FOREACH(a, V_pf_altqs_inactive, entries) {
2020                                 if (strncmp(a->ifname, altq->ifname,
2021                                     IFNAMSIZ) == 0 && a->qname[0] == 0) {
2022                                         altq->altq_disc = a->altq_disc;
2023                                         break;
2024                                 }
2025                         }
2026                 }
2027
2028                 if ((ifp = ifunit(altq->ifname)) == NULL)
2029                         altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2030                 else
2031                         error = altq_add(altq);
2032
2033                 if (error) {
2034                         PF_RULES_WUNLOCK();
2035                         free(altq, M_PFALTQ);
2036                         break;
2037                 }
2038
2039                 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2040                 bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2041                 PF_RULES_WUNLOCK();
2042                 break;
2043         }
2044
2045         case DIOCGETALTQS: {
2046                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2047                 struct pf_altq          *altq;
2048
2049                 PF_RULES_RLOCK();
2050                 pa->nr = 0;
2051                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2052                         pa->nr++;
2053                 pa->ticket = V_ticket_altqs_active;
2054                 PF_RULES_RUNLOCK();
2055                 break;
2056         }
2057
2058         case DIOCGETALTQ: {
2059                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2060                 struct pf_altq          *altq;
2061                 u_int32_t                nr;
2062
2063                 PF_RULES_RLOCK();
2064                 if (pa->ticket != V_ticket_altqs_active) {
2065                         PF_RULES_RUNLOCK();
2066                         error = EBUSY;
2067                         break;
2068                 }
2069                 nr = 0;
2070                 altq = TAILQ_FIRST(V_pf_altqs_active);
2071                 while ((altq != NULL) && (nr < pa->nr)) {
2072                         altq = TAILQ_NEXT(altq, entries);
2073                         nr++;
2074                 }
2075                 if (altq == NULL) {
2076                         PF_RULES_RUNLOCK();
2077                         error = EBUSY;
2078                         break;
2079                 }
2080                 bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2081                 PF_RULES_RUNLOCK();
2082                 break;
2083         }
2084
2085         case DIOCCHANGEALTQ:
2086                 /* CHANGEALTQ not supported yet! */
2087                 error = ENODEV;
2088                 break;
2089
2090         case DIOCGETQSTATS: {
2091                 struct pfioc_qstats     *pq = (struct pfioc_qstats *)addr;
2092                 struct pf_altq          *altq;
2093                 u_int32_t                nr;
2094                 int                      nbytes;
2095
2096                 PF_RULES_RLOCK();
2097                 if (pq->ticket != V_ticket_altqs_active) {
2098                         PF_RULES_RUNLOCK();
2099                         error = EBUSY;
2100                         break;
2101                 }
2102                 nbytes = pq->nbytes;
2103                 nr = 0;
2104                 altq = TAILQ_FIRST(V_pf_altqs_active);
2105                 while ((altq != NULL) && (nr < pq->nr)) {
2106                         altq = TAILQ_NEXT(altq, entries);
2107                         nr++;
2108                 }
2109                 if (altq == NULL) {
2110                         PF_RULES_RUNLOCK();
2111                         error = EBUSY;
2112                         break;
2113                 }
2114
2115                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2116                         PF_RULES_RUNLOCK();
2117                         error = ENXIO;
2118                         break;
2119                 }
2120                 PF_RULES_RUNLOCK();
2121                 error = altq_getqstats(altq, pq->buf, &nbytes);
2122                 if (error == 0) {
2123                         pq->scheduler = altq->scheduler;
2124                         pq->nbytes = nbytes;
2125                 }
2126                 break;
2127         }
2128 #endif /* ALTQ */
2129
2130         case DIOCBEGINADDRS: {
2131                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2132
2133                 PF_RULES_WLOCK();
2134                 pf_empty_pool(&V_pf_pabuf);
2135                 pp->ticket = ++V_ticket_pabuf;
2136                 PF_RULES_WUNLOCK();
2137                 break;
2138         }
2139
2140         case DIOCADDADDR: {
2141                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2142                 struct pf_pooladdr      *pa;
2143                 struct pfi_kif          *kif = NULL;
2144
2145 #ifndef INET
2146                 if (pp->af == AF_INET) {
2147                         error = EAFNOSUPPORT;
2148                         break;
2149                 }
2150 #endif /* INET */
2151 #ifndef INET6
2152                 if (pp->af == AF_INET6) {
2153                         error = EAFNOSUPPORT;
2154                         break;
2155                 }
2156 #endif /* INET6 */
2157                 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2158                     pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2159                     pp->addr.addr.type != PF_ADDR_TABLE) {
2160                         error = EINVAL;
2161                         break;
2162                 }
2163                 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2164                 bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2165                 if (pa->ifname[0])
2166                         kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2167                 PF_RULES_WLOCK();
2168                 if (pp->ticket != V_ticket_pabuf) {
2169                         PF_RULES_WUNLOCK();
2170                         if (pa->ifname[0])
2171                                 free(kif, PFI_MTYPE);
2172                         free(pa, M_PFRULE);
2173                         error = EBUSY;
2174                         break;
2175                 }
2176                 if (pa->ifname[0]) {
2177                         pa->kif = pfi_kif_attach(kif, pa->ifname);
2178                         pfi_kif_ref(pa->kif);
2179                 } else
2180                         pa->kif = NULL;
2181                 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2182                     pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2183                         if (pa->ifname[0])
2184                                 pfi_kif_unref(pa->kif);
2185                         PF_RULES_WUNLOCK();
2186                         free(pa, M_PFRULE);
2187                         break;
2188                 }
2189                 TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2190                 PF_RULES_WUNLOCK();
2191                 break;
2192         }
2193
2194         case DIOCGETADDRS: {
2195                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2196                 struct pf_pool          *pool;
2197                 struct pf_pooladdr      *pa;
2198
2199                 PF_RULES_RLOCK();
2200                 pp->nr = 0;
2201                 pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2202                     pp->r_num, 0, 1, 0);
2203                 if (pool == NULL) {
2204                         PF_RULES_RUNLOCK();
2205                         error = EBUSY;
2206                         break;
2207                 }
2208                 TAILQ_FOREACH(pa, &pool->list, entries)
2209                         pp->nr++;
2210                 PF_RULES_RUNLOCK();
2211                 break;
2212         }
2213
2214         case DIOCGETADDR: {
2215                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2216                 struct pf_pool          *pool;
2217                 struct pf_pooladdr      *pa;
2218                 u_int32_t                nr = 0;
2219
2220                 PF_RULES_RLOCK();
2221                 pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2222                     pp->r_num, 0, 1, 1);
2223                 if (pool == NULL) {
2224                         PF_RULES_RUNLOCK();
2225                         error = EBUSY;
2226                         break;
2227                 }
2228                 pa = TAILQ_FIRST(&pool->list);
2229                 while ((pa != NULL) && (nr < pp->nr)) {
2230                         pa = TAILQ_NEXT(pa, entries);
2231                         nr++;
2232                 }
2233                 if (pa == NULL) {
2234                         PF_RULES_RUNLOCK();
2235                         error = EBUSY;
2236                         break;
2237                 }
2238                 bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2239                 pf_addr_copyout(&pp->addr.addr);
2240                 PF_RULES_RUNLOCK();
2241                 break;
2242         }
2243
2244         case DIOCCHANGEADDR: {
2245                 struct pfioc_pooladdr   *pca = (struct pfioc_pooladdr *)addr;
2246                 struct pf_pool          *pool;
2247                 struct pf_pooladdr      *oldpa = NULL, *newpa = NULL;
2248                 struct pf_ruleset       *ruleset;
2249                 struct pfi_kif          *kif = NULL;
2250
2251                 if (pca->action < PF_CHANGE_ADD_HEAD ||
2252                     pca->action > PF_CHANGE_REMOVE) {
2253                         error = EINVAL;
2254                         break;
2255                 }
2256                 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2257                     pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2258                     pca->addr.addr.type != PF_ADDR_TABLE) {
2259                         error = EINVAL;
2260                         break;
2261                 }
2262
2263                 if (pca->action != PF_CHANGE_REMOVE) {
2264 #ifndef INET
2265                         if (pca->af == AF_INET) {
2266                                 error = EAFNOSUPPORT;
2267                                 break;
2268                         }
2269 #endif /* INET */
2270 #ifndef INET6
2271                         if (pca->af == AF_INET6) {
2272                                 error = EAFNOSUPPORT;
2273                                 break;
2274                         }
2275 #endif /* INET6 */
2276                         newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2277                         bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2278                         if (newpa->ifname[0])
2279                                 kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2280                 }
2281
2282 #define ERROUT(x)       { error = (x); goto DIOCCHANGEADDR_error; }
2283                 PF_RULES_WLOCK();
2284                 ruleset = pf_find_ruleset(pca->anchor);
2285                 if (ruleset == NULL)
2286                         ERROUT(EBUSY);
2287
2288                 pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2289                     pca->r_num, pca->r_last, 1, 1);
2290                 if (pool == NULL)
2291                         ERROUT(EBUSY);
2292
2293                 if (pca->action != PF_CHANGE_REMOVE) {
2294                         if (newpa->ifname[0]) {
2295                                 newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2296                                 pfi_kif_ref(newpa->kif);
2297                         } else
2298                                 newpa->kif = NULL;
2299
2300                         switch (newpa->addr.type) {
2301                         case PF_ADDR_DYNIFTL:
2302                                 error = pfi_dynaddr_setup(&newpa->addr,
2303                                     pca->af);
2304                                 break;
2305                         case PF_ADDR_TABLE:
2306                                 newpa->addr.p.tbl = pfr_attach_table(ruleset,
2307                                     newpa->addr.v.tblname);
2308                                 if (newpa->addr.p.tbl == NULL)
2309                                         error = ENOMEM;
2310                                 break;
2311                         }
2312                         if (error) {
2313                                 if (newpa->kif)
2314                                         pfi_kif_unref(newpa->kif);
2315                                 PF_RULES_WUNLOCK();
2316                                 free(newpa, M_PFRULE);
2317                                 break;
2318                         }
2319                 }
2320
2321                 if (pca->action == PF_CHANGE_ADD_HEAD)
2322                         oldpa = TAILQ_FIRST(&pool->list);
2323                 else if (pca->action == PF_CHANGE_ADD_TAIL)
2324                         oldpa = TAILQ_LAST(&pool->list, pf_palist);
2325                 else {
2326                         int     i = 0;
2327
2328                         oldpa = TAILQ_FIRST(&pool->list);
2329                         while ((oldpa != NULL) && (i < pca->nr)) {
2330                                 oldpa = TAILQ_NEXT(oldpa, entries);
2331                                 i++;
2332                         }
2333                         if (oldpa == NULL) {
2334                                 PF_RULES_WUNLOCK();
2335                                 error = EINVAL;
2336                                 break;
2337                         }
2338                 }
2339
2340                 if (pca->action == PF_CHANGE_REMOVE) {
2341                         TAILQ_REMOVE(&pool->list, oldpa, entries);
2342                         switch (oldpa->addr.type) {
2343                         case PF_ADDR_DYNIFTL:
2344                                 pfi_dynaddr_remove(oldpa->addr.p.dyn);
2345                                 break;
2346                         case PF_ADDR_TABLE:
2347                                 pfr_detach_table(oldpa->addr.p.tbl);
2348                                 break;
2349                         }
2350                         if (oldpa->kif)
2351                                 pfi_kif_unref(oldpa->kif);
2352                         free(oldpa, M_PFRULE);
2353                 } else {
2354                         if (oldpa == NULL)
2355                                 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2356                         else if (pca->action == PF_CHANGE_ADD_HEAD ||
2357                             pca->action == PF_CHANGE_ADD_BEFORE)
2358                                 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2359                         else
2360                                 TAILQ_INSERT_AFTER(&pool->list, oldpa,
2361                                     newpa, entries);
2362                 }
2363
2364                 pool->cur = TAILQ_FIRST(&pool->list);
2365                 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2366                     pca->af);
2367                 PF_RULES_WUNLOCK();
2368                 break;
2369
2370 #undef ERROUT
2371 DIOCCHANGEADDR_error:
2372                 PF_RULES_WUNLOCK();
2373                 if (newpa != NULL)
2374                         free(newpa, M_PFRULE);
2375                 if (kif != NULL)
2376                         free(kif, PFI_MTYPE);
2377                 break;
2378         }
2379
2380         case DIOCGETRULESETS: {
2381                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
2382                 struct pf_ruleset       *ruleset;
2383                 struct pf_anchor        *anchor;
2384
2385                 PF_RULES_RLOCK();
2386                 pr->path[sizeof(pr->path) - 1] = 0;
2387                 if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2388                         PF_RULES_RUNLOCK();
2389                         error = ENOENT;
2390                         break;
2391                 }
2392                 pr->nr = 0;
2393                 if (ruleset->anchor == NULL) {
2394                         /* XXX kludge for pf_main_ruleset */
2395                         RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2396                                 if (anchor->parent == NULL)
2397                                         pr->nr++;
2398                 } else {
2399                         RB_FOREACH(anchor, pf_anchor_node,
2400                             &ruleset->anchor->children)
2401                                 pr->nr++;
2402                 }
2403                 PF_RULES_RUNLOCK();
2404                 break;
2405         }
2406
2407         case DIOCGETRULESET: {
2408                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
2409                 struct pf_ruleset       *ruleset;
2410                 struct pf_anchor        *anchor;
2411                 u_int32_t                nr = 0;
2412
2413                 PF_RULES_RLOCK();
2414                 pr->path[sizeof(pr->path) - 1] = 0;
2415                 if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2416                         PF_RULES_RUNLOCK();
2417                         error = ENOENT;
2418                         break;
2419                 }
2420                 pr->name[0] = 0;
2421                 if (ruleset->anchor == NULL) {
2422                         /* XXX kludge for pf_main_ruleset */
2423                         RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2424                                 if (anchor->parent == NULL && nr++ == pr->nr) {
2425                                         strlcpy(pr->name, anchor->name,
2426                                             sizeof(pr->name));
2427                                         break;
2428                                 }
2429                 } else {
2430                         RB_FOREACH(anchor, pf_anchor_node,
2431                             &ruleset->anchor->children)
2432                                 if (nr++ == pr->nr) {
2433                                         strlcpy(pr->name, anchor->name,
2434                                             sizeof(pr->name));
2435                                         break;
2436                                 }
2437                 }
2438                 if (!pr->name[0])
2439                         error = EBUSY;
2440                 PF_RULES_RUNLOCK();
2441                 break;
2442         }
2443
2444         case DIOCRCLRTABLES: {
2445                 struct pfioc_table *io = (struct pfioc_table *)addr;
2446
2447                 if (io->pfrio_esize != 0) {
2448                         error = ENODEV;
2449                         break;
2450                 }
2451                 PF_RULES_WLOCK();
2452                 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2453                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
2454                 PF_RULES_WUNLOCK();
2455                 break;
2456         }
2457
2458         case DIOCRADDTABLES: {
2459                 struct pfioc_table *io = (struct pfioc_table *)addr;
2460                 struct pfr_table *pfrts;
2461                 size_t totlen;
2462
2463                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2464                         error = ENODEV;
2465                         break;
2466                 }
2467                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2468                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2469                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2470                 if (error) {
2471                         free(pfrts, M_TEMP);
2472                         break;
2473                 }
2474                 PF_RULES_WLOCK();
2475                 error = pfr_add_tables(pfrts, io->pfrio_size,
2476                     &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2477                 PF_RULES_WUNLOCK();
2478                 free(pfrts, M_TEMP);
2479                 break;
2480         }
2481
2482         case DIOCRDELTABLES: {
2483                 struct pfioc_table *io = (struct pfioc_table *)addr;
2484                 struct pfr_table *pfrts;
2485                 size_t totlen;
2486
2487                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2488                         error = ENODEV;
2489                         break;
2490                 }
2491                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2492                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2493                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2494                 if (error) {
2495                         free(pfrts, M_TEMP);
2496                         break;
2497                 }
2498                 PF_RULES_WLOCK();
2499                 error = pfr_del_tables(pfrts, io->pfrio_size,
2500                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2501                 PF_RULES_WUNLOCK();
2502                 free(pfrts, M_TEMP);
2503                 break;
2504         }
2505
2506         case DIOCRGETTABLES: {
2507                 struct pfioc_table *io = (struct pfioc_table *)addr;
2508                 struct pfr_table *pfrts;
2509                 size_t totlen;
2510
2511                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2512                         error = ENODEV;
2513                         break;
2514                 }
2515                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2516                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2517                 PF_RULES_RLOCK();
2518                 error = pfr_get_tables(&io->pfrio_table, pfrts,
2519                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2520                 PF_RULES_RUNLOCK();
2521                 if (error == 0)
2522                         error = copyout(pfrts, io->pfrio_buffer, totlen);
2523                 free(pfrts, M_TEMP);
2524                 break;
2525         }
2526
2527         case DIOCRGETTSTATS: {
2528                 struct pfioc_table *io = (struct pfioc_table *)addr;
2529                 struct pfr_tstats *pfrtstats;
2530                 size_t totlen;
2531
2532                 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2533                         error = ENODEV;
2534                         break;
2535                 }
2536                 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
2537                 pfrtstats = malloc(totlen, M_TEMP, M_WAITOK);
2538                 PF_RULES_WLOCK();
2539                 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
2540                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2541                 PF_RULES_WUNLOCK();
2542                 if (error == 0)
2543                         error = copyout(pfrtstats, io->pfrio_buffer, totlen);
2544                 free(pfrtstats, M_TEMP);
2545                 break;
2546         }
2547
2548         case DIOCRCLRTSTATS: {
2549                 struct pfioc_table *io = (struct pfioc_table *)addr;
2550                 struct pfr_table *pfrts;
2551                 size_t totlen;
2552
2553                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2554                         error = ENODEV;
2555                         break;
2556                 }
2557                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2558                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2559                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2560                 if (error) {
2561                         free(pfrts, M_TEMP);
2562                         break;
2563                 }
2564                 PF_RULES_WLOCK();
2565                 error = pfr_clr_tstats(pfrts, io->pfrio_size,
2566                     &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2567                 PF_RULES_WUNLOCK();
2568                 free(pfrts, M_TEMP);
2569                 break;
2570         }
2571
2572         case DIOCRSETTFLAGS: {
2573                 struct pfioc_table *io = (struct pfioc_table *)addr;
2574                 struct pfr_table *pfrts;
2575                 size_t totlen;
2576
2577                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2578                         error = ENODEV;
2579                         break;
2580                 }
2581                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2582                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2583                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2584                 if (error) {
2585                         free(pfrts, M_TEMP);
2586                         break;
2587                 }
2588                 PF_RULES_WLOCK();
2589                 error = pfr_set_tflags(pfrts, io->pfrio_size,
2590                     io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2591                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2592                 PF_RULES_WUNLOCK();
2593                 free(pfrts, M_TEMP);
2594                 break;
2595         }
2596
2597         case DIOCRCLRADDRS: {
2598                 struct pfioc_table *io = (struct pfioc_table *)addr;
2599
2600                 if (io->pfrio_esize != 0) {
2601                         error = ENODEV;
2602                         break;
2603                 }
2604                 PF_RULES_WLOCK();
2605                 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2606                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
2607                 PF_RULES_WUNLOCK();
2608                 break;
2609         }
2610
2611         case DIOCRADDADDRS: {
2612                 struct pfioc_table *io = (struct pfioc_table *)addr;
2613                 struct pfr_addr *pfras;
2614                 size_t totlen;
2615
2616                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2617                         error = ENODEV;
2618                         break;
2619                 }
2620                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2621                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2622                 error = copyin(io->pfrio_buffer, pfras, totlen);
2623                 if (error) {
2624                         free(pfras, M_TEMP);
2625                         break;
2626                 }
2627                 PF_RULES_WLOCK();
2628                 error = pfr_add_addrs(&io->pfrio_table, pfras,
2629                     io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2630                     PFR_FLAG_USERIOCTL);
2631                 PF_RULES_WUNLOCK();
2632                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2633                         error = copyout(pfras, io->pfrio_buffer, totlen);
2634                 free(pfras, M_TEMP);
2635                 break;
2636         }
2637
2638         case DIOCRDELADDRS: {
2639                 struct pfioc_table *io = (struct pfioc_table *)addr;
2640                 struct pfr_addr *pfras;
2641                 size_t totlen;
2642
2643                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2644                         error = ENODEV;
2645                         break;
2646                 }
2647                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2648                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2649                 error = copyin(io->pfrio_buffer, pfras, totlen);
2650                 if (error) {
2651                         free(pfras, M_TEMP);
2652                         break;
2653                 }
2654                 PF_RULES_WLOCK();
2655                 error = pfr_del_addrs(&io->pfrio_table, pfras,
2656                     io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2657                     PFR_FLAG_USERIOCTL);
2658                 PF_RULES_WUNLOCK();
2659                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2660                         error = copyout(pfras, io->pfrio_buffer, totlen);
2661                 free(pfras, M_TEMP);
2662                 break;
2663         }
2664
2665         case DIOCRSETADDRS: {
2666                 struct pfioc_table *io = (struct pfioc_table *)addr;
2667                 struct pfr_addr *pfras;
2668                 size_t totlen;
2669
2670                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2671                         error = ENODEV;
2672                         break;
2673                 }
2674                 totlen = (io->pfrio_size + io->pfrio_size2) *
2675                     sizeof(struct pfr_addr);
2676                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2677                 error = copyin(io->pfrio_buffer, pfras, totlen);
2678                 if (error) {
2679                         free(pfras, M_TEMP);
2680                         break;
2681                 }
2682                 PF_RULES_WLOCK();
2683                 error = pfr_set_addrs(&io->pfrio_table, pfras,
2684                     io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2685                     &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2686                     PFR_FLAG_USERIOCTL, 0);
2687                 PF_RULES_WUNLOCK();
2688                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2689                         error = copyout(pfras, io->pfrio_buffer, totlen);
2690                 free(pfras, M_TEMP);
2691                 break;
2692         }
2693
2694         case DIOCRGETADDRS: {
2695                 struct pfioc_table *io = (struct pfioc_table *)addr;
2696                 struct pfr_addr *pfras;
2697                 size_t totlen;
2698
2699                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2700                         error = ENODEV;
2701                         break;
2702                 }
2703                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2704                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2705                 PF_RULES_RLOCK();
2706                 error = pfr_get_addrs(&io->pfrio_table, pfras,
2707                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2708                 PF_RULES_RUNLOCK();
2709                 if (error == 0)
2710                         error = copyout(pfras, io->pfrio_buffer, totlen);
2711                 free(pfras, M_TEMP);
2712                 break;
2713         }
2714
2715         case DIOCRGETASTATS: {
2716                 struct pfioc_table *io = (struct pfioc_table *)addr;
2717                 struct pfr_astats *pfrastats;
2718                 size_t totlen;
2719
2720                 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2721                         error = ENODEV;
2722                         break;
2723                 }
2724                 totlen = io->pfrio_size * sizeof(struct pfr_astats);
2725                 pfrastats = malloc(totlen, M_TEMP, M_WAITOK);
2726                 PF_RULES_RLOCK();
2727                 error = pfr_get_astats(&io->pfrio_table, pfrastats,
2728                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2729                 PF_RULES_RUNLOCK();
2730                 if (error == 0)
2731                         error = copyout(pfrastats, io->pfrio_buffer, totlen);
2732                 free(pfrastats, M_TEMP);
2733                 break;
2734         }
2735
2736         case DIOCRCLRASTATS: {
2737                 struct pfioc_table *io = (struct pfioc_table *)addr;
2738                 struct pfr_addr *pfras;
2739                 size_t totlen;
2740
2741                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2742                         error = ENODEV;
2743                         break;
2744                 }
2745                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2746                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2747                 error = copyin(io->pfrio_buffer, pfras, totlen);
2748                 if (error) {
2749                         free(pfras, M_TEMP);
2750                         break;
2751                 }
2752                 PF_RULES_WLOCK();
2753                 error = pfr_clr_astats(&io->pfrio_table, pfras,
2754                     io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2755                     PFR_FLAG_USERIOCTL);
2756                 PF_RULES_WUNLOCK();
2757                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2758                         error = copyout(pfras, io->pfrio_buffer, totlen);
2759                 free(pfras, M_TEMP);
2760                 break;
2761         }
2762
2763         case DIOCRTSTADDRS: {
2764                 struct pfioc_table *io = (struct pfioc_table *)addr;
2765                 struct pfr_addr *pfras;
2766                 size_t totlen;
2767
2768                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2769                         error = ENODEV;
2770                         break;
2771                 }
2772                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2773                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2774                 error = copyin(io->pfrio_buffer, pfras, totlen);
2775                 if (error) {
2776                         free(pfras, M_TEMP);
2777                         break;
2778                 }
2779                 PF_RULES_RLOCK();
2780                 error = pfr_tst_addrs(&io->pfrio_table, pfras,
2781                     io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2782                     PFR_FLAG_USERIOCTL);
2783                 PF_RULES_RUNLOCK();
2784                 if (error == 0)
2785                         error = copyout(pfras, io->pfrio_buffer, totlen);
2786                 free(pfras, M_TEMP);
2787                 break;
2788         }
2789
2790         case DIOCRINADEFINE: {
2791                 struct pfioc_table *io = (struct pfioc_table *)addr;
2792                 struct pfr_addr *pfras;
2793                 size_t totlen;
2794
2795                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2796                         error = ENODEV;
2797                         break;
2798                 }
2799                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2800                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2801                 error = copyin(io->pfrio_buffer, pfras, totlen);
2802                 if (error) {
2803                         free(pfras, M_TEMP);
2804                         break;
2805                 }
2806                 PF_RULES_WLOCK();
2807                 error = pfr_ina_define(&io->pfrio_table, pfras,
2808                     io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2809                     io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2810                 PF_RULES_WUNLOCK();
2811                 free(pfras, M_TEMP);
2812                 break;
2813         }
2814
2815         case DIOCOSFPADD: {
2816                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2817                 PF_RULES_WLOCK();
2818                 error = pf_osfp_add(io);
2819                 PF_RULES_WUNLOCK();
2820                 break;
2821         }
2822
2823         case DIOCOSFPGET: {
2824                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2825                 PF_RULES_RLOCK();
2826                 error = pf_osfp_get(io);
2827                 PF_RULES_RUNLOCK();
2828                 break;
2829         }
2830
2831         case DIOCXBEGIN: {
2832                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2833                 struct pfioc_trans_e    *ioes, *ioe;
2834                 size_t                   totlen;
2835                 int                      i;
2836
2837                 if (io->esize != sizeof(*ioe)) {
2838                         error = ENODEV;
2839                         break;
2840                 }
2841                 totlen = sizeof(struct pfioc_trans_e) * io->size;
2842                 ioes = malloc(totlen, M_TEMP, M_WAITOK);
2843                 error = copyin(io->array, ioes, totlen);
2844                 if (error) {
2845                         free(ioes, M_TEMP);
2846                         break;
2847                 }
2848                 PF_RULES_WLOCK();
2849                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2850                         switch (ioe->rs_num) {
2851 #ifdef ALTQ
2852                         case PF_RULESET_ALTQ:
2853                                 if (ioe->anchor[0]) {
2854                                         PF_RULES_WUNLOCK();
2855                                         free(ioes, M_TEMP);
2856                                         error = EINVAL;
2857                                         goto fail;
2858                                 }
2859                                 if ((error = pf_begin_altq(&ioe->ticket))) {
2860                                         PF_RULES_WUNLOCK();
2861                                         free(ioes, M_TEMP);
2862                                         goto fail;
2863                                 }
2864                                 break;
2865 #endif /* ALTQ */
2866                         case PF_RULESET_TABLE:
2867                             {
2868                                 struct pfr_table table;
2869
2870                                 bzero(&table, sizeof(table));
2871                                 strlcpy(table.pfrt_anchor, ioe->anchor,
2872                                     sizeof(table.pfrt_anchor));
2873                                 if ((error = pfr_ina_begin(&table,
2874                                     &ioe->ticket, NULL, 0))) {
2875                                         PF_RULES_WUNLOCK();
2876                                         free(ioes, M_TEMP);
2877                                         goto fail;
2878                                 }
2879                                 break;
2880                             }
2881                         default:
2882                                 if ((error = pf_begin_rules(&ioe->ticket,
2883                                     ioe->rs_num, ioe->anchor))) {
2884                                         PF_RULES_WUNLOCK();
2885                                         free(ioes, M_TEMP);
2886                                         goto fail;
2887                                 }
2888                                 break;
2889                         }
2890                 }
2891                 PF_RULES_WUNLOCK();
2892                 error = copyout(ioes, io->array, totlen);
2893                 free(ioes, M_TEMP);
2894                 break;
2895         }
2896
2897         case DIOCXROLLBACK: {
2898                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2899                 struct pfioc_trans_e    *ioe, *ioes;
2900                 size_t                   totlen;
2901                 int                      i;
2902
2903                 if (io->esize != sizeof(*ioe)) {
2904                         error = ENODEV;
2905                         break;
2906                 }
2907                 totlen = sizeof(struct pfioc_trans_e) * io->size;
2908                 ioes = malloc(totlen, M_TEMP, M_WAITOK);
2909                 error = copyin(io->array, ioes, totlen);
2910                 if (error) {
2911                         free(ioes, M_TEMP);
2912                         break;
2913                 }
2914                 PF_RULES_WLOCK();
2915                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2916                         switch (ioe->rs_num) {
2917 #ifdef ALTQ
2918                         case PF_RULESET_ALTQ:
2919                                 if (ioe->anchor[0]) {
2920                                         PF_RULES_WUNLOCK();
2921                                         free(ioes, M_TEMP);
2922                                         error = EINVAL;
2923                                         goto fail;
2924                                 }
2925                                 if ((error = pf_rollback_altq(ioe->ticket))) {
2926                                         PF_RULES_WUNLOCK();
2927                                         free(ioes, M_TEMP);
2928                                         goto fail; /* really bad */
2929                                 }
2930                                 break;
2931 #endif /* ALTQ */
2932                         case PF_RULESET_TABLE:
2933                             {
2934                                 struct pfr_table table;
2935
2936                                 bzero(&table, sizeof(table));
2937                                 strlcpy(table.pfrt_anchor, ioe->anchor,
2938                                     sizeof(table.pfrt_anchor));
2939                                 if ((error = pfr_ina_rollback(&table,
2940                                     ioe->ticket, NULL, 0))) {
2941                                         PF_RULES_WUNLOCK();
2942                                         free(ioes, M_TEMP);
2943                                         goto fail; /* really bad */
2944                                 }
2945                                 break;
2946                             }
2947                         default:
2948                                 if ((error = pf_rollback_rules(ioe->ticket,
2949                                     ioe->rs_num, ioe->anchor))) {
2950                                         PF_RULES_WUNLOCK();
2951                                         free(ioes, M_TEMP);
2952                                         goto fail; /* really bad */
2953                                 }
2954                                 break;
2955                         }
2956                 }
2957                 PF_RULES_WUNLOCK();
2958                 free(ioes, M_TEMP);
2959                 break;
2960         }
2961
2962         case DIOCXCOMMIT: {
2963                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2964                 struct pfioc_trans_e    *ioe, *ioes;
2965                 struct pf_ruleset       *rs;
2966                 size_t                   totlen;
2967                 int                      i;
2968
2969                 if (io->esize != sizeof(*ioe)) {
2970                         error = ENODEV;
2971                         break;
2972                 }
2973                 totlen = sizeof(struct pfioc_trans_e) * io->size;
2974                 ioes = malloc(totlen, M_TEMP, M_WAITOK);
2975                 error = copyin(io->array, ioes, totlen);
2976                 if (error) {
2977                         free(ioes, M_TEMP);
2978                         break;
2979                 }
2980                 PF_RULES_WLOCK();
2981                 /* First makes sure everything will succeed. */
2982                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2983                         switch (ioe->rs_num) {
2984 #ifdef ALTQ
2985                         case PF_RULESET_ALTQ:
2986                                 if (ioe->anchor[0]) {
2987                                         PF_RULES_WUNLOCK();
2988                                         free(ioes, M_TEMP);
2989                                         error = EINVAL;
2990                                         goto fail;
2991                                 }
2992                                 if (!V_altqs_inactive_open || ioe->ticket !=
2993                                     V_ticket_altqs_inactive) {
2994                                         PF_RULES_WUNLOCK();
2995                                         free(ioes, M_TEMP);
2996                                         error = EBUSY;
2997                                         goto fail;
2998                                 }
2999                                 break;
3000 #endif /* ALTQ */
3001                         case PF_RULESET_TABLE:
3002                                 rs = pf_find_ruleset(ioe->anchor);
3003                                 if (rs == NULL || !rs->topen || ioe->ticket !=
3004                                     rs->tticket) {
3005                                         PF_RULES_WUNLOCK();
3006                                         free(ioes, M_TEMP);
3007                                         error = EBUSY;
3008                                         goto fail;
3009                                 }
3010                                 break;
3011                         default:
3012                                 if (ioe->rs_num < 0 || ioe->rs_num >=
3013                                     PF_RULESET_MAX) {
3014                                         PF_RULES_WUNLOCK();
3015                                         free(ioes, M_TEMP);
3016                                         error = EINVAL;
3017                                         goto fail;
3018                                 }
3019                                 rs = pf_find_ruleset(ioe->anchor);
3020                                 if (rs == NULL ||
3021                                     !rs->rules[ioe->rs_num].inactive.open ||
3022                                     rs->rules[ioe->rs_num].inactive.ticket !=
3023                                     ioe->ticket) {
3024                                         PF_RULES_WUNLOCK();
3025                                         free(ioes, M_TEMP);
3026                                         error = EBUSY;
3027                                         goto fail;
3028                                 }
3029                                 break;
3030                         }
3031                 }
3032                 /* Now do the commit - no errors should happen here. */
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 ((error = pf_commit_altq(ioe->ticket))) {
3038                                         PF_RULES_WUNLOCK();
3039                                         free(ioes, M_TEMP);
3040                                         goto fail; /* really bad */
3041                                 }
3042                                 break;
3043 #endif /* ALTQ */
3044                         case PF_RULESET_TABLE:
3045                             {
3046                                 struct pfr_table table;
3047
3048                                 bzero(&table, sizeof(table));
3049                                 strlcpy(table.pfrt_anchor, ioe->anchor,
3050                                     sizeof(table.pfrt_anchor));
3051                                 if ((error = pfr_ina_commit(&table,
3052                                     ioe->ticket, NULL, NULL, 0))) {
3053                                         PF_RULES_WUNLOCK();
3054                                         free(ioes, M_TEMP);
3055                                         goto fail; /* really bad */
3056                                 }
3057                                 break;
3058                             }
3059                         default:
3060                                 if ((error = pf_commit_rules(ioe->ticket,
3061                                     ioe->rs_num, ioe->anchor))) {
3062                                         PF_RULES_WUNLOCK();
3063                                         free(ioes, M_TEMP);
3064                                         goto fail; /* really bad */
3065                                 }
3066                                 break;
3067                         }
3068                 }
3069                 PF_RULES_WUNLOCK();
3070                 free(ioes, M_TEMP);
3071                 break;
3072         }
3073
3074         case DIOCGETSRCNODES: {
3075                 struct pfioc_src_nodes  *psn = (struct pfioc_src_nodes *)addr;
3076                 struct pf_srchash       *sh;
3077                 struct pf_src_node      *n, *p, *pstore;
3078                 uint32_t                 i, nr = 0;
3079
3080                 if (psn->psn_len == 0) {
3081                         for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
3082                             i++, sh++) {
3083                                 PF_HASHROW_LOCK(sh);
3084                                 LIST_FOREACH(n, &sh->nodes, entry)
3085                                         nr++;
3086                                 PF_HASHROW_UNLOCK(sh);
3087                         }
3088                         psn->psn_len = sizeof(struct pf_src_node) * nr;
3089                         break;
3090                 }
3091
3092                 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK);
3093                 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
3094                     i++, sh++) {
3095                     PF_HASHROW_LOCK(sh);
3096                     LIST_FOREACH(n, &sh->nodes, entry) {
3097                         int     secs = time_uptime, diff;
3098
3099                         if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3100                                 break;
3101
3102                         bcopy(n, p, sizeof(struct pf_src_node));
3103                         if (n->rule.ptr != NULL)
3104                                 p->rule.nr = n->rule.ptr->nr;
3105                         p->creation = secs - p->creation;
3106                         if (p->expire > secs)
3107                                 p->expire -= secs;
3108                         else
3109                                 p->expire = 0;
3110
3111                         /* Adjust the connection rate estimate. */
3112                         diff = secs - n->conn_rate.last;
3113                         if (diff >= n->conn_rate.seconds)
3114                                 p->conn_rate.count = 0;
3115                         else
3116                                 p->conn_rate.count -=
3117                                     n->conn_rate.count * diff /
3118                                     n->conn_rate.seconds;
3119                         p++;
3120                         nr++;
3121                     }
3122                     PF_HASHROW_UNLOCK(sh);
3123                 }
3124                 error = copyout(pstore, psn->psn_src_nodes,
3125                     sizeof(struct pf_src_node) * nr);
3126                 if (error) {
3127                         free(pstore, M_TEMP);
3128                         break;
3129                 }
3130                 psn->psn_len = sizeof(struct pf_src_node) * nr;
3131                 free(pstore, M_TEMP);
3132                 break;
3133         }
3134
3135         case DIOCCLRSRCNODES: {
3136
3137                 pf_clear_srcnodes(NULL);
3138                 pf_purge_expired_src_nodes();
3139                 V_pf_status.src_nodes = 0;
3140                 break;
3141         }
3142
3143         case DIOCKILLSRCNODES:
3144                 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
3145                 break;
3146
3147         case DIOCSETHOSTID: {
3148                 u_int32_t       *hostid = (u_int32_t *)addr;
3149
3150                 PF_RULES_WLOCK();
3151                 if (*hostid == 0)
3152                         V_pf_status.hostid = arc4random();
3153                 else
3154                         V_pf_status.hostid = *hostid;
3155                 PF_RULES_WUNLOCK();
3156                 break;
3157         }
3158
3159         case DIOCOSFPFLUSH:
3160                 PF_RULES_WLOCK();
3161                 pf_osfp_flush();
3162                 PF_RULES_WUNLOCK();
3163                 break;
3164
3165         case DIOCIGETIFACES: {
3166                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3167                 struct pfi_kif *ifstore;
3168                 size_t bufsiz;
3169
3170                 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3171                         error = ENODEV;
3172                         break;
3173                 }
3174
3175                 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3176                 ifstore = malloc(bufsiz, M_TEMP, M_WAITOK);
3177                 PF_RULES_RLOCK();
3178                 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3179                 PF_RULES_RUNLOCK();
3180                 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3181                 free(ifstore, M_TEMP);
3182                 break;
3183         }
3184
3185         case DIOCSETIFFLAG: {
3186                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3187
3188                 PF_RULES_WLOCK();
3189                 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3190                 PF_RULES_WUNLOCK();
3191                 break;
3192         }
3193
3194         case DIOCCLRIFFLAG: {
3195                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3196
3197                 PF_RULES_WLOCK();
3198                 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3199                 PF_RULES_WUNLOCK();
3200                 break;
3201         }
3202
3203         default:
3204                 error = ENODEV;
3205                 break;
3206         }
3207 fail:
3208         CURVNET_RESTORE();
3209
3210         return (error);
3211 }
3212
3213 void
3214 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3215 {
3216         bzero(sp, sizeof(struct pfsync_state));
3217
3218         /* copy from state key */
3219         sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3220         sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3221         sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3222         sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3223         sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3224         sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3225         sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3226         sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3227         sp->proto = st->key[PF_SK_WIRE]->proto;
3228         sp->af = st->key[PF_SK_WIRE]->af;
3229
3230         /* copy from state */
3231         strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3232         bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3233         sp->creation = htonl(time_uptime - st->creation);
3234         sp->expire = pf_state_expires(st);
3235         if (sp->expire <= time_uptime)
3236                 sp->expire = htonl(0);
3237         else
3238                 sp->expire = htonl(sp->expire - time_uptime);
3239
3240         sp->direction = st->direction;
3241         sp->log = st->log;
3242         sp->timeout = st->timeout;
3243         sp->state_flags = st->state_flags;
3244         if (st->src_node)
3245                 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3246         if (st->nat_src_node)
3247                 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3248
3249         sp->id = st->id;
3250         sp->creatorid = st->creatorid;
3251         pf_state_peer_hton(&st->src, &sp->src);
3252         pf_state_peer_hton(&st->dst, &sp->dst);
3253
3254         if (st->rule.ptr == NULL)
3255                 sp->rule = htonl(-1);
3256         else
3257                 sp->rule = htonl(st->rule.ptr->nr);
3258         if (st->anchor.ptr == NULL)
3259                 sp->anchor = htonl(-1);
3260         else
3261                 sp->anchor = htonl(st->anchor.ptr->nr);
3262         if (st->nat_rule.ptr == NULL)
3263                 sp->nat_rule = htonl(-1);
3264         else
3265                 sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3266
3267         pf_state_counter_hton(st->packets[0], sp->packets[0]);
3268         pf_state_counter_hton(st->packets[1], sp->packets[1]);
3269         pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3270         pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3271
3272 }
3273
3274 static void
3275 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3276 {
3277         struct pfr_ktable *kt;
3278
3279         KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3280
3281         kt = aw->p.tbl;
3282         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3283                 kt = kt->pfrkt_root;
3284         aw->p.tbl = NULL;
3285         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3286                 kt->pfrkt_cnt : -1;
3287 }
3288
3289 /*
3290  * XXX - Check for version missmatch!!!
3291  */
3292 static void
3293 pf_clear_states(void)
3294 {
3295         struct pf_state *s;
3296         u_int i;
3297
3298         for (i = 0; i <= V_pf_hashmask; i++) {
3299                 struct pf_idhash *ih = &V_pf_idhash[i];
3300 relock:
3301                 PF_HASHROW_LOCK(ih);
3302                 LIST_FOREACH(s, &ih->states, entry) {
3303                         s->timeout = PFTM_PURGE;
3304                         /* Don't send out individual delete messages. */
3305                         s->sync_state = PFSTATE_NOSYNC;
3306                         pf_unlink_state(s, PF_ENTER_LOCKED);
3307                         goto relock;
3308                 }
3309                 PF_HASHROW_UNLOCK(ih);
3310         }
3311 }
3312
3313 static int
3314 pf_clear_tables(void)
3315 {
3316         struct pfioc_table io;
3317         int error;
3318
3319         bzero(&io, sizeof(io));
3320
3321         error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3322             io.pfrio_flags);
3323
3324         return (error);
3325 }
3326
3327 static void
3328 pf_clear_srcnodes(struct pf_src_node *n)
3329 {
3330         struct pf_state *s;
3331         int i;
3332
3333         for (i = 0; i <= V_pf_hashmask; i++) {
3334                 struct pf_idhash *ih = &V_pf_idhash[i];
3335
3336                 PF_HASHROW_LOCK(ih);
3337                 LIST_FOREACH(s, &ih->states, entry) {
3338                         if (n == NULL || n == s->src_node)
3339                                 s->src_node = NULL;
3340                         if (n == NULL || n == s->nat_src_node)
3341                                 s->nat_src_node = NULL;
3342                 }
3343                 PF_HASHROW_UNLOCK(ih);
3344         }
3345
3346         if (n == NULL) {
3347                 struct pf_srchash *sh;
3348
3349                 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
3350                     i++, sh++) {
3351                         PF_HASHROW_LOCK(sh);
3352                         LIST_FOREACH(n, &sh->nodes, entry) {
3353                                 n->expire = 1;
3354                                 n->states = 0;
3355                         }
3356                         PF_HASHROW_UNLOCK(sh);
3357                 }
3358         } else {
3359                 /* XXX: hash slot should already be locked here. */
3360                 n->expire = 1;
3361                 n->states = 0;
3362         }
3363 }
3364
3365 static void
3366 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
3367 {
3368         struct pf_src_node_list  kill;
3369
3370         LIST_INIT(&kill);
3371         for (int i = 0; i <= V_pf_srchashmask; i++) {
3372                 struct pf_srchash *sh = &V_pf_srchash[i];
3373                 struct pf_src_node *sn, *tmp;
3374
3375                 PF_HASHROW_LOCK(sh);
3376                 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
3377                         if (PF_MATCHA(psnk->psnk_src.neg,
3378                               &psnk->psnk_src.addr.v.a.addr,
3379                               &psnk->psnk_src.addr.v.a.mask,
3380                               &sn->addr, sn->af) &&
3381                             PF_MATCHA(psnk->psnk_dst.neg,
3382                               &psnk->psnk_dst.addr.v.a.addr,
3383                               &psnk->psnk_dst.addr.v.a.mask,
3384                               &sn->raddr, sn->af)) {
3385                                 pf_unlink_src_node_locked(sn);
3386                                 LIST_INSERT_HEAD(&kill, sn, entry);
3387                                 sn->expire = 1;
3388                         }
3389                 PF_HASHROW_UNLOCK(sh);
3390         }
3391
3392         for (int i = 0; i <= V_pf_hashmask; i++) {
3393                 struct pf_idhash *ih = &V_pf_idhash[i];
3394                 struct pf_state *s;
3395
3396                 PF_HASHROW_LOCK(ih);
3397                 LIST_FOREACH(s, &ih->states, entry) {
3398                         if (s->src_node && s->src_node->expire == 1) {
3399 #ifdef INVARIANTS
3400                                 s->src_node->states--;
3401 #endif
3402                                 s->src_node = NULL;
3403                         }
3404                         if (s->nat_src_node && s->nat_src_node->expire == 1) {
3405 #ifdef INVARIANTS
3406                                 s->nat_src_node->states--;
3407 #endif
3408                                 s->nat_src_node = NULL;
3409                         }
3410                 }
3411                 PF_HASHROW_UNLOCK(ih);
3412         }
3413
3414         psnk->psnk_killed = pf_free_src_nodes(&kill);
3415 }
3416
3417 /*
3418  * XXX - Check for version missmatch!!!
3419  */
3420
3421 /*
3422  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3423  */
3424 static int
3425 shutdown_pf(void)
3426 {
3427         int error = 0;
3428         u_int32_t t[5];
3429         char nn = '\0';
3430
3431         V_pf_status.running = 0;
3432         do {
3433                 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3434                     != 0) {
3435                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3436                         break;
3437                 }
3438                 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3439                     != 0) {
3440                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3441                         break;          /* XXX: rollback? */
3442                 }
3443                 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3444                     != 0) {
3445                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3446                         break;          /* XXX: rollback? */
3447                 }
3448                 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3449                     != 0) {
3450                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3451                         break;          /* XXX: rollback? */
3452                 }
3453                 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3454                     != 0) {
3455                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3456                         break;          /* XXX: rollback? */
3457                 }
3458
3459                 /* XXX: these should always succeed here */
3460                 pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3461                 pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3462                 pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3463                 pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3464                 pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3465
3466                 if ((error = pf_clear_tables()) != 0)
3467                         break;
3468
3469 #ifdef ALTQ
3470                 if ((error = pf_begin_altq(&t[0])) != 0) {
3471                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3472                         break;
3473                 }
3474                 pf_commit_altq(t[0]);
3475 #endif
3476
3477                 pf_clear_states();
3478
3479                 pf_clear_srcnodes(NULL);
3480
3481                 /* status does not use malloced mem so no need to cleanup */
3482                 /* fingerprints and interfaces have thier own cleanup code */
3483         } while(0);
3484
3485         return (error);
3486 }
3487
3488 #ifdef INET
3489 static int
3490 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3491     struct inpcb *inp)
3492 {
3493         int chk;
3494
3495         chk = pf_test(PF_IN, ifp, m, inp);
3496         if (chk && *m) {
3497                 m_freem(*m);
3498                 *m = NULL;
3499         }
3500
3501         return (chk);
3502 }
3503
3504 static int
3505 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3506     struct inpcb *inp)
3507 {
3508         int chk;
3509
3510         /* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3511         if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3512                 in_delayed_cksum(*m);
3513                 (*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3514         }
3515
3516         chk = pf_test(PF_OUT, ifp, m, inp);
3517         if (chk && *m) {
3518                 m_freem(*m);
3519                 *m = NULL;
3520         }
3521
3522         return (chk);
3523 }
3524 #endif
3525
3526 #ifdef INET6
3527 static int
3528 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3529     struct inpcb *inp)
3530 {
3531         int chk;
3532
3533         /*
3534          * In case of loopback traffic IPv6 uses the real interface in
3535          * order to support scoped addresses. In order to support stateful
3536          * filtering we have change this to lo0 as it is the case in IPv4.
3537          */
3538         CURVNET_SET(ifp->if_vnet);
3539         chk = pf_test6(PF_IN, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
3540         CURVNET_RESTORE();
3541         if (chk && *m) {
3542                 m_freem(*m);
3543                 *m = NULL;
3544         }
3545         return chk;
3546 }
3547
3548 static int
3549 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3550     struct inpcb *inp)
3551 {
3552         int chk;
3553
3554         /* We need a proper CSUM before we start (s. OpenBSD ip_output) */
3555         if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3556 #ifdef INET
3557                 /* XXX-BZ copy&paste error from r126261? */
3558                 in_delayed_cksum(*m);
3559 #endif
3560                 (*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3561         }
3562         CURVNET_SET(ifp->if_vnet);
3563         chk = pf_test6(PF_OUT, ifp, m, inp);
3564         CURVNET_RESTORE();
3565         if (chk && *m) {
3566                 m_freem(*m);
3567                 *m = NULL;
3568         }
3569         return chk;
3570 }
3571 #endif /* INET6 */
3572
3573 static int
3574 hook_pf(void)
3575 {
3576 #ifdef INET
3577         struct pfil_head *pfh_inet;
3578 #endif
3579 #ifdef INET6
3580         struct pfil_head *pfh_inet6;
3581 #endif
3582
3583         if (V_pf_pfil_hooked)
3584                 return (0);
3585
3586 #ifdef INET
3587         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3588         if (pfh_inet == NULL)
3589                 return (ESRCH); /* XXX */
3590         pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
3591         pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
3592 #endif
3593 #ifdef INET6
3594         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3595         if (pfh_inet6 == NULL) {
3596 #ifdef INET
3597                 pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3598                     pfh_inet);
3599                 pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3600                     pfh_inet);
3601 #endif
3602                 return (ESRCH); /* XXX */
3603         }
3604         pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
3605         pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
3606 #endif
3607
3608         V_pf_pfil_hooked = 1;
3609         return (0);
3610 }
3611
3612 static int
3613 dehook_pf(void)
3614 {
3615 #ifdef INET
3616         struct pfil_head *pfh_inet;
3617 #endif
3618 #ifdef INET6
3619         struct pfil_head *pfh_inet6;
3620 #endif
3621
3622         if (V_pf_pfil_hooked == 0)
3623                 return (0);
3624
3625 #ifdef INET
3626         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3627         if (pfh_inet == NULL)
3628                 return (ESRCH); /* XXX */
3629         pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3630             pfh_inet);
3631         pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3632             pfh_inet);
3633 #endif
3634 #ifdef INET6
3635         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3636         if (pfh_inet6 == NULL)
3637                 return (ESRCH); /* XXX */
3638         pfil_remove_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
3639             pfh_inet6);
3640         pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
3641             pfh_inet6);
3642 #endif
3643
3644         V_pf_pfil_hooked = 0;
3645         return (0);
3646 }
3647
3648 static int
3649 pf_load(void)
3650 {
3651         int error;
3652
3653         VNET_ITERATOR_DECL(vnet_iter);
3654
3655         VNET_LIST_RLOCK();
3656         VNET_FOREACH(vnet_iter) {
3657                 CURVNET_SET(vnet_iter);
3658                 V_pf_pfil_hooked = 0;
3659                 V_pf_end_threads = 0;
3660                 TAILQ_INIT(&V_pf_tags);
3661                 TAILQ_INIT(&V_pf_qids);
3662                 CURVNET_RESTORE();
3663         }
3664         VNET_LIST_RUNLOCK();
3665
3666         rw_init(&pf_rules_lock, "pf rulesets");
3667
3668         pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
3669         if ((error = pfattach()) != 0)
3670                 return (error);
3671
3672         return (0);
3673 }
3674
3675 static int
3676 pf_unload(void)
3677 {
3678         int error = 0;
3679
3680         PF_RULES_WLOCK();
3681         V_pf_status.running = 0;
3682         PF_RULES_WUNLOCK();
3683         swi_remove(V_pf_swi_cookie);
3684         error = dehook_pf();
3685         if (error) {
3686                 /*
3687                  * Should not happen!
3688                  * XXX Due to error code ESRCH, kldunload will show
3689                  * a message like 'No such process'.
3690                  */
3691                 printf("%s : pfil unregisteration fail\n", __FUNCTION__);
3692                 return error;
3693         }
3694         PF_RULES_WLOCK();
3695         shutdown_pf();
3696         V_pf_end_threads = 1;
3697         while (V_pf_end_threads < 2) {
3698                 wakeup_one(pf_purge_thread);
3699                 rw_sleep(pf_purge_thread, &pf_rules_lock, 0, "pftmo", 0);
3700         }
3701         pf_normalize_cleanup();
3702         pfi_cleanup();
3703         pfr_cleanup();
3704         pf_osfp_flush();
3705         pf_cleanup();
3706         PF_RULES_WUNLOCK();
3707         destroy_dev(pf_dev);
3708         rw_destroy(&pf_rules_lock);
3709
3710         return (error);
3711 }
3712
3713 static int
3714 pf_modevent(module_t mod, int type, void *data)
3715 {
3716         int error = 0;
3717
3718         switch(type) {
3719         case MOD_LOAD:
3720                 error = pf_load();
3721                 break;
3722         case MOD_QUIESCE:
3723                 /*
3724                  * Module should not be unloaded due to race conditions.
3725                  */
3726                 error = EBUSY;
3727                 break;
3728         case MOD_UNLOAD:
3729                 error = pf_unload();
3730                 break;
3731         default:
3732                 error = EINVAL;
3733                 break;
3734         }
3735
3736         return (error);
3737 }
3738
3739 static moduledata_t pf_mod = {
3740         "pf",
3741         pf_modevent,
3742         0
3743 };
3744
3745 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3746 MODULE_VERSION(pf, PF_MODVER);