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 <netinet6/ip6_var.h>
  76 #include <netinet/ip_icmp.h>
  77
  78 #ifdef INET6
  79 #include <netinet/ip6.h>
  80 #endif /* INET6 */
  81
  82 #ifdef ALTQ
  83 #include <altq/altq.h>
  84 #endif
  85
  86 #define PF_TABLES_MAX_REQUEST   65535 /* Maximum tables per request. */
  87
  88 static int               pfattach(void);
  89 static struct pf_pool   *pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
  90                             u_int8_t, u_int8_t, u_int8_t);
  91
  92 static void              pf_mv_pool(struct pf_palist *, struct pf_palist *);
  93 static void              pf_empty_pool(struct pf_palist *);
  94 static int               pfioctl(struct cdev *, u_long, caddr_t, int,
  95                             struct thread *);
  96 #ifdef ALTQ
  97 static int               pf_begin_altq(u_int32_t *);
  98 static int               pf_rollback_altq(u_int32_t);
  99 static int               pf_commit_altq(u_int32_t);
 100 static int               pf_enable_altq(struct pf_altq *);
 101 static int               pf_disable_altq(struct pf_altq *);
 102 static u_int32_t         pf_qname2qid(char *);
 103 static void              pf_qid_unref(u_int32_t);
 104 #endif /* ALTQ */
 105 static int               pf_begin_rules(u_int32_t *, int, const char *);
 106 static int               pf_rollback_rules(u_int32_t, int, char *);
 107 static int               pf_setup_pfsync_matching(struct pf_ruleset *);
 108 static void              pf_hash_rule(MD5_CTX *, struct pf_rule *);
 109 static void              pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
 110 static int               pf_commit_rules(u_int32_t, int, char *);
 111 static int               pf_addr_setup(struct pf_ruleset *,
 112                             struct pf_addr_wrap *, sa_family_t);
 113 static void              pf_addr_copyout(struct pf_addr_wrap *);
 114
 115 VNET_DEFINE(struct pf_rule,     pf_default_rule);
 116
 117 #ifdef ALTQ
 118 static VNET_DEFINE(int,         pf_altq_running);
 119 #define V_pf_altq_running       VNET(pf_altq_running)
 120 #endif
 121
 122 #define TAGID_MAX        50000
 123 struct pf_tagname {
 124         TAILQ_ENTRY(pf_tagname) entries;
 125         char                    name[PF_TAG_NAME_SIZE];
 126         uint16_t                tag;
 127         int                     ref;
 128 };
 129
 130 TAILQ_HEAD(pf_tags, pf_tagname);
 131 #define V_pf_tags               VNET(pf_tags)
 132 VNET_DEFINE(struct pf_tags, pf_tags);
 133 #define V_pf_qids               VNET(pf_qids)
 134 VNET_DEFINE(struct pf_tags, pf_qids);
 135 static MALLOC_DEFINE(M_PFTAG, "pf_tag", "pf(4) tag names");
 136 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
 137 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
 138
 139 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
 140 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
 141 #endif
 142
 143 static u_int16_t         tagname2tag(struct pf_tags *, char *);
 144 static u_int16_t         pf_tagname2tag(char *);
 145 static void              tag_unref(struct pf_tags *, u_int16_t);
 146
 147 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
 148
 149 struct cdev *pf_dev;
 150
 151 /*
 152  * XXX - These are new and need to be checked when moveing to a new version
 153  */
 154 static void              pf_clear_states(void);
 155 static int               pf_clear_tables(void);
 156 static void              pf_clear_srcnodes(struct pf_src_node *);
 157 static void              pf_kill_srcnodes(struct pfioc_src_node_kill *);
 158 static void              pf_tbladdr_copyout(struct pf_addr_wrap *);
 159
 160 /*
 161  * Wrapper functions for pfil(9) hooks
 162  */
 163 #ifdef INET
 164 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
 165     int dir, struct inpcb *inp);
 166 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
 167     int dir, struct inpcb *inp);
 168 #endif
 169 #ifdef INET6
 170 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
 171     int dir, struct inpcb *inp);
 172 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
 173     int dir, struct inpcb *inp);
 174 #endif
 175
 176 static int              hook_pf(void);
 177 static int              dehook_pf(void);
 178 static int              shutdown_pf(void);
 179 static int              pf_load(void);
 180 static int              pf_unload(void);
 181
 182 static struct cdevsw pf_cdevsw = {
 183         .d_ioctl =      pfioctl,
 184         .d_name =       PF_NAME,
 185         .d_version =    D_VERSION,
 186 };
 187
 188 static volatile VNET_DEFINE(int, pf_pfil_hooked);
 189 #define V_pf_pfil_hooked        VNET(pf_pfil_hooked)
 190 VNET_DEFINE(int,                pf_end_threads);
 191
 192 struct rwlock                   pf_rules_lock;
 193 struct sx                       pf_ioctl_lock;
 194
 195 /* pfsync */
 196 pfsync_state_import_t           *pfsync_state_import_ptr = NULL;
 197 pfsync_insert_state_t           *pfsync_insert_state_ptr = NULL;
 198 pfsync_update_state_t           *pfsync_update_state_ptr = NULL;
 199 pfsync_delete_state_t           *pfsync_delete_state_ptr = NULL;
 200 pfsync_clear_states_t           *pfsync_clear_states_ptr = NULL;
 201 pfsync_defer_t                  *pfsync_defer_ptr = NULL;
 202 /* pflog */
 203 pflog_packet_t                  *pflog_packet_ptr = NULL;
 204
 205 static int
 206 pfattach(void)
 207 {
 208         u_int32_t *my_timeout = V_pf_default_rule.timeout;
 209         int error;
 210
 211         if (IS_DEFAULT_VNET(curvnet))
 212                 pf_mtag_initialize();
 213         pf_initialize();
 214         pfr_initialize();
 215         pfi_initialize();
 216         pf_normalize_init();
 217
 218         V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
 219         V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
 220
 221         RB_INIT(&V_pf_anchors);
 222         pf_init_ruleset(&pf_main_ruleset);
 223
 224         /* default rule should never be garbage collected */
 225         V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
 226 #ifdef PF_DEFAULT_TO_DROP
 227         V_pf_default_rule.action = PF_DROP;
 228 #else
 229         V_pf_default_rule.action = PF_PASS;
 230 #endif
 231         V_pf_default_rule.nr = -1;
 232         V_pf_default_rule.rtableid = -1;
 233
 234         V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
 235         V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
 236         V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
 237
 238         /* initialize default timeouts */
 239         my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
 240         my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
 241         my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
 242         my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
 243         my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
 244         my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
 245         my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
 246         my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
 247         my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
 248         my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
 249         my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
 250         my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
 251         my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
 252         my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
 253         my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
 254         my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
 255         my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
 256         my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
 257         my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
 258         my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
 259
 260         bzero(&V_pf_status, sizeof(V_pf_status));
 261         V_pf_status.debug = PF_DEBUG_URGENT;
 262
 263         V_pf_pfil_hooked = 0;
 264
 265         /* XXX do our best to avoid a conflict */
 266         V_pf_status.hostid = arc4random();
 267
 268         for (int i = 0; i < PFRES_MAX; i++)
 269                 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
 270         for (int i = 0; i < LCNT_MAX; i++)
 271                 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
 272         for (int i = 0; i < FCNT_MAX; i++)
 273                 V_pf_status.fcounters[i] = counter_u64_alloc(M_WAITOK);
 274         for (int i = 0; i < SCNT_MAX; i++)
 275                 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
 276
 277         if ((error = kproc_create(pf_purge_thread, curvnet, NULL, 0, 0,
 278             "pf purge")) != 0)
 279                 /* XXXGL: leaked all above. */
 280                 return (error);
 281         if ((error = swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
 282             INTR_MPSAFE, &V_pf_swi_cookie)) != 0)
 283                 /* XXXGL: leaked all above. */
 284                 return (error);
 285
 286         return (0);
 287 }
 288
 289 static struct pf_pool *
 290 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
 291     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
 292     u_int8_t check_ticket)
 293 {
 294         struct pf_ruleset       *ruleset;
 295         struct pf_rule          *rule;
 296         int                      rs_num;
 297
 298         ruleset = pf_find_ruleset(anchor);
 299         if (ruleset == NULL)
 300                 return (NULL);
 301         rs_num = pf_get_ruleset_number(rule_action);
 302         if (rs_num >= PF_RULESET_MAX)
 303                 return (NULL);
 304         if (active) {
 305                 if (check_ticket && ticket !=
 306                     ruleset->rules[rs_num].active.ticket)
 307                         return (NULL);
 308                 if (r_last)
 309                         rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
 310                             pf_rulequeue);
 311                 else
 312                         rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
 313         } else {
 314                 if (check_ticket && ticket !=
 315                     ruleset->rules[rs_num].inactive.ticket)
 316                         return (NULL);
 317                 if (r_last)
 318                         rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
 319                             pf_rulequeue);
 320                 else
 321                         rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
 322         }
 323         if (!r_last) {
 324                 while ((rule != NULL) && (rule->nr != rule_number))
 325                         rule = TAILQ_NEXT(rule, entries);
 326         }
 327         if (rule == NULL)
 328                 return (NULL);
 329
 330         return (&rule->rpool);
 331 }
 332
 333 static void
 334 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
 335 {
 336         struct pf_pooladdr      *mv_pool_pa;
 337
 338         while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
 339                 TAILQ_REMOVE(poola, mv_pool_pa, entries);
 340                 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
 341         }
 342 }
 343
 344 static void
 345 pf_empty_pool(struct pf_palist *poola)
 346 {
 347         struct pf_pooladdr *pa;
 348
 349         while ((pa = TAILQ_FIRST(poola)) != NULL) {
 350                 switch (pa->addr.type) {
 351                 case PF_ADDR_DYNIFTL:
 352                         pfi_dynaddr_remove(pa->addr.p.dyn);
 353                         break;
 354                 case PF_ADDR_TABLE:
 355                         /* XXX: this could be unfinished pooladdr on pabuf */
 356                         if (pa->addr.p.tbl != NULL)
 357                                 pfr_detach_table(pa->addr.p.tbl);
 358                         break;
 359                 }
 360                 if (pa->kif)
 361                         pfi_kif_unref(pa->kif);
 362                 TAILQ_REMOVE(poola, pa, entries);
 363                 free(pa, M_PFRULE);
 364         }
 365 }
 366
 367 static void
 368 pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
 369 {
 370
 371         PF_RULES_WASSERT();
 372
 373         TAILQ_REMOVE(rulequeue, rule, entries);
 374
 375         PF_UNLNKDRULES_LOCK();
 376         rule->rule_flag |= PFRULE_REFS;
 377         TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
 378         PF_UNLNKDRULES_UNLOCK();
 379 }
 380
 381 void
 382 pf_free_rule(struct pf_rule *rule)
 383 {
 384
 385         PF_RULES_WASSERT();
 386
 387         if (rule->tag)
 388                 tag_unref(&V_pf_tags, rule->tag);
 389         if (rule->match_tag)
 390                 tag_unref(&V_pf_tags, rule->match_tag);
 391 #ifdef ALTQ
 392         if (rule->pqid != rule->qid)
 393                 pf_qid_unref(rule->pqid);
 394         pf_qid_unref(rule->qid);
 395 #endif
 396         switch (rule->src.addr.type) {
 397         case PF_ADDR_DYNIFTL:
 398                 pfi_dynaddr_remove(rule->src.addr.p.dyn);
 399                 break;
 400         case PF_ADDR_TABLE:
 401                 pfr_detach_table(rule->src.addr.p.tbl);
 402                 break;
 403         }
 404         switch (rule->dst.addr.type) {
 405         case PF_ADDR_DYNIFTL:
 406                 pfi_dynaddr_remove(rule->dst.addr.p.dyn);
 407                 break;
 408         case PF_ADDR_TABLE:
 409                 pfr_detach_table(rule->dst.addr.p.tbl);
 410                 break;
 411         }
 412         if (rule->overload_tbl)
 413                 pfr_detach_table(rule->overload_tbl);
 414         if (rule->kif)
 415                 pfi_kif_unref(rule->kif);
 416         pf_anchor_remove(rule);
 417         pf_empty_pool(&rule->rpool.list);
 418         counter_u64_free(rule->states_cur);
 419         counter_u64_free(rule->states_tot);
 420         counter_u64_free(rule->src_nodes);
 421         free(rule, M_PFRULE);
 422 }
 423
 424 static u_int16_t
 425 tagname2tag(struct pf_tags *head, char *tagname)
 426 {
 427         struct pf_tagname       *tag, *p = NULL;
 428         u_int16_t                new_tagid = 1;
 429
 430         PF_RULES_WASSERT();
 431
 432         TAILQ_FOREACH(tag, head, entries)
 433                 if (strcmp(tagname, tag->name) == 0) {
 434                         tag->ref++;
 435                         return (tag->tag);
 436                 }
 437
 438         /*
 439          * to avoid fragmentation, we do a linear search from the beginning
 440          * and take the first free slot we find. if there is none or the list
 441          * is empty, append a new entry at the end.
 442          */
 443
 444         /* new entry */
 445         if (!TAILQ_EMPTY(head))
 446                 for (p = TAILQ_FIRST(head); p != NULL &&
 447                     p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
 448                         new_tagid = p->tag + 1;
 449
 450         if (new_tagid > TAGID_MAX)
 451                 return (0);
 452
 453         /* allocate and fill new struct pf_tagname */
 454         tag = malloc(sizeof(*tag), M_PFTAG, M_NOWAIT|M_ZERO);
 455         if (tag == NULL)
 456                 return (0);
 457         strlcpy(tag->name, tagname, sizeof(tag->name));
 458         tag->tag = new_tagid;
 459         tag->ref++;
 460
 461         if (p != NULL)  /* insert new entry before p */
 462                 TAILQ_INSERT_BEFORE(p, tag, entries);
 463         else    /* either list empty or no free slot in between */
 464                 TAILQ_INSERT_TAIL(head, tag, entries);
 465
 466         return (tag->tag);
 467 }
 468
 469 static void
 470 tag_unref(struct pf_tags *head, u_int16_t tag)
 471 {
 472         struct pf_tagname       *p, *next;
 473
 474         PF_RULES_WASSERT();
 475
 476         for (p = TAILQ_FIRST(head); p != NULL; p = next) {
 477                 next = TAILQ_NEXT(p, entries);
 478                 if (tag == p->tag) {
 479                         if (--p->ref == 0) {
 480                                 TAILQ_REMOVE(head, p, entries);
 481                                 free(p, M_PFTAG);
 482                         }
 483                         break;
 484                 }
 485         }
 486 }
 487
 488 static u_int16_t
 489 pf_tagname2tag(char *tagname)
 490 {
 491         return (tagname2tag(&V_pf_tags, tagname));
 492 }
 493
 494 #ifdef ALTQ
 495 static u_int32_t
 496 pf_qname2qid(char *qname)
 497 {
 498         return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
 499 }
 500
 501 static void
 502 pf_qid_unref(u_int32_t qid)
 503 {
 504         tag_unref(&V_pf_qids, (u_int16_t)qid);
 505 }
 506
 507 static int
 508 pf_begin_altq(u_int32_t *ticket)
 509 {
 510         struct pf_altq  *altq;
 511         int              error = 0;
 512
 513         PF_RULES_WASSERT();
 514
 515         /* Purge the old altq list */
 516         while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
 517                 TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
 518                 if (altq->qname[0] == 0 &&
 519                     (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 520                         /* detach and destroy the discipline */
 521                         error = altq_remove(altq);
 522                 } else
 523                         pf_qid_unref(altq->qid);
 524                 free(altq, M_PFALTQ);
 525         }
 526         if (error)
 527                 return (error);
 528         *ticket = ++V_ticket_altqs_inactive;
 529         V_altqs_inactive_open = 1;
 530         return (0);
 531 }
 532
 533 static int
 534 pf_rollback_altq(u_int32_t ticket)
 535 {
 536         struct pf_altq  *altq;
 537         int              error = 0;
 538
 539         PF_RULES_WASSERT();
 540
 541         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 542                 return (0);
 543         /* Purge the old altq list */
 544         while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
 545                 TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
 546                 if (altq->qname[0] == 0 &&
 547                    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 548                         /* detach and destroy the discipline */
 549                         error = altq_remove(altq);
 550                 } else
 551                         pf_qid_unref(altq->qid);
 552                 free(altq, M_PFALTQ);
 553         }
 554         V_altqs_inactive_open = 0;
 555         return (error);
 556 }
 557
 558 static int
 559 pf_commit_altq(u_int32_t ticket)
 560 {
 561         struct pf_altqqueue     *old_altqs;
 562         struct pf_altq          *altq;
 563         int                      err, error = 0;
 564
 565         PF_RULES_WASSERT();
 566
 567         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 568                 return (EBUSY);
 569
 570         /* swap altqs, keep the old. */
 571         old_altqs = V_pf_altqs_active;
 572         V_pf_altqs_active = V_pf_altqs_inactive;
 573         V_pf_altqs_inactive = old_altqs;
 574         V_ticket_altqs_active = V_ticket_altqs_inactive;
 575
 576         /* Attach new disciplines */
 577         TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
 578         if (altq->qname[0] == 0 &&
 579            (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 580                         /* attach the discipline */
 581                         error = altq_pfattach(altq);
 582                         if (error == 0 && V_pf_altq_running)
 583                                 error = pf_enable_altq(altq);
 584                         if (error != 0)
 585                                 return (error);
 586                 }
 587         }
 588
 589         /* Purge the old altq list */
 590         while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
 591                 TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
 592                 if (altq->qname[0] == 0 &&
 593                     (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 594                         /* detach and destroy the discipline */
 595                         if (V_pf_altq_running)
 596                                 error = pf_disable_altq(altq);
 597                         err = altq_pfdetach(altq);
 598                         if (err != 0 && error == 0)
 599                                 error = err;
 600                         err = altq_remove(altq);
 601                         if (err != 0 && error == 0)
 602                                 error = err;
 603                 } else
 604                         pf_qid_unref(altq->qid);
 605                 free(altq, M_PFALTQ);
 606         }
 607
 608         V_altqs_inactive_open = 0;
 609         return (error);
 610 }
 611
 612 static int
 613 pf_enable_altq(struct pf_altq *altq)
 614 {
 615         struct ifnet            *ifp;
 616         struct tb_profile        tb;
 617         int                      error = 0;
 618
 619         if ((ifp = ifunit(altq->ifname)) == NULL)
 620                 return (EINVAL);
 621
 622         if (ifp->if_snd.altq_type != ALTQT_NONE)
 623                 error = altq_enable(&ifp->if_snd);
 624
 625         /* set tokenbucket regulator */
 626         if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
 627                 tb.rate = altq->ifbandwidth;
 628                 tb.depth = altq->tbrsize;
 629                 error = tbr_set(&ifp->if_snd, &tb);
 630         }
 631
 632         return (error);
 633 }
 634
 635 static int
 636 pf_disable_altq(struct pf_altq *altq)
 637 {
 638         struct ifnet            *ifp;
 639         struct tb_profile        tb;
 640         int                      error;
 641
 642         if ((ifp = ifunit(altq->ifname)) == NULL)
 643                 return (EINVAL);
 644
 645         /*
 646          * when the discipline is no longer referenced, it was overridden
 647          * by a new one.  if so, just return.
 648          */
 649         if (altq->altq_disc != ifp->if_snd.altq_disc)
 650                 return (0);
 651
 652         error = altq_disable(&ifp->if_snd);
 653
 654         if (error == 0) {
 655                 /* clear tokenbucket regulator */
 656                 tb.rate = 0;
 657                 error = tbr_set(&ifp->if_snd, &tb);
 658         }
 659
 660         return (error);
 661 }
 662
 663 void
 664 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
 665 {
 666         struct ifnet    *ifp1;
 667         struct pf_altq  *a1, *a2, *a3;
 668         u_int32_t        ticket;
 669         int              error = 0;
 670
 671         /* Interrupt userland queue modifications */
 672         if (V_altqs_inactive_open)
 673                 pf_rollback_altq(V_ticket_altqs_inactive);
 674
 675         /* Start new altq ruleset */
 676         if (pf_begin_altq(&ticket))
 677                 return;
 678
 679         /* Copy the current active set */
 680         TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
 681                 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
 682                 if (a2 == NULL) {
 683                         error = ENOMEM;
 684                         break;
 685                 }
 686                 bcopy(a1, a2, sizeof(struct pf_altq));
 687
 688                 if (a2->qname[0] != 0) {
 689                         if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
 690                                 error = EBUSY;
 691                                 free(a2, M_PFALTQ);
 692                                 break;
 693                         }
 694                         a2->altq_disc = NULL;
 695                         TAILQ_FOREACH(a3, V_pf_altqs_inactive, entries) {
 696                                 if (strncmp(a3->ifname, a2->ifname,
 697                                     IFNAMSIZ) == 0 && a3->qname[0] == 0) {
 698                                         a2->altq_disc = a3->altq_disc;
 699                                         break;
 700                                 }
 701                         }
 702                 }
 703                 /* Deactivate the interface in question */
 704                 a2->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
 705                 if ((ifp1 = ifunit(a2->ifname)) == NULL ||
 706                     (remove && ifp1 == ifp)) {
 707                         a2->local_flags |= PFALTQ_FLAG_IF_REMOVED;
 708                 } else {
 709                         error = altq_add(a2);
 710
 711                         if (ticket != V_ticket_altqs_inactive)
 712                                 error = EBUSY;
 713
 714                         if (error) {
 715                                 free(a2, M_PFALTQ);
 716                                 break;
 717                         }
 718                 }
 719
 720                 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
 721         }
 722
 723         if (error != 0)
 724                 pf_rollback_altq(ticket);
 725         else
 726                 pf_commit_altq(ticket);
 727 }
 728 #endif /* ALTQ */
 729
 730 static int
 731 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
 732 {
 733         struct pf_ruleset       *rs;
 734         struct pf_rule          *rule;
 735
 736         PF_RULES_WASSERT();
 737
 738         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 739                 return (EINVAL);
 740         rs = pf_find_or_create_ruleset(anchor);
 741         if (rs == NULL)
 742                 return (EINVAL);
 743         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 744                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 745                 rs->rules[rs_num].inactive.rcount--;
 746         }
 747         *ticket = ++rs->rules[rs_num].inactive.ticket;
 748         rs->rules[rs_num].inactive.open = 1;
 749         return (0);
 750 }
 751
 752 static int
 753 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
 754 {
 755         struct pf_ruleset       *rs;
 756         struct pf_rule          *rule;
 757
 758         PF_RULES_WASSERT();
 759
 760         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 761                 return (EINVAL);
 762         rs = pf_find_ruleset(anchor);
 763         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 764             rs->rules[rs_num].inactive.ticket != ticket)
 765                 return (0);
 766         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 767                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 768                 rs->rules[rs_num].inactive.rcount--;
 769         }
 770         rs->rules[rs_num].inactive.open = 0;
 771         return (0);
 772 }
 773
 774 #define PF_MD5_UPD(st, elm)                                             \
 775                 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
 776
 777 #define PF_MD5_UPD_STR(st, elm)                                         \
 778                 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
 779
 780 #define PF_MD5_UPD_HTONL(st, elm, stor) do {                            \
 781                 (stor) = htonl((st)->elm);                              \
 782                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
 783 } while (0)
 784
 785 #define PF_MD5_UPD_HTONS(st, elm, stor) do {                            \
 786                 (stor) = htons((st)->elm);                              \
 787                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
 788 } while (0)
 789
 790 static void
 791 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
 792 {
 793         PF_MD5_UPD(pfr, addr.type);
 794         switch (pfr->addr.type) {
 795                 case PF_ADDR_DYNIFTL:
 796                         PF_MD5_UPD(pfr, addr.v.ifname);
 797                         PF_MD5_UPD(pfr, addr.iflags);
 798                         break;
 799                 case PF_ADDR_TABLE:
 800                         PF_MD5_UPD(pfr, addr.v.tblname);
 801                         break;
 802                 case PF_ADDR_ADDRMASK:
 803                         /* XXX ignore af? */
 804                         PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
 805                         PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
 806                         break;
 807         }
 808
 809         PF_MD5_UPD(pfr, port[0]);
 810         PF_MD5_UPD(pfr, port[1]);
 811         PF_MD5_UPD(pfr, neg);
 812         PF_MD5_UPD(pfr, port_op);
 813 }
 814
 815 static void
 816 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
 817 {
 818         u_int16_t x;
 819         u_int32_t y;
 820
 821         pf_hash_rule_addr(ctx, &rule->src);
 822         pf_hash_rule_addr(ctx, &rule->dst);
 823         PF_MD5_UPD_STR(rule, label);
 824         PF_MD5_UPD_STR(rule, ifname);
 825         PF_MD5_UPD_STR(rule, match_tagname);
 826         PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
 827         PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
 828         PF_MD5_UPD_HTONL(rule, prob, y);
 829         PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
 830         PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
 831         PF_MD5_UPD(rule, uid.op);
 832         PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
 833         PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
 834         PF_MD5_UPD(rule, gid.op);
 835         PF_MD5_UPD_HTONL(rule, rule_flag, y);
 836         PF_MD5_UPD(rule, action);
 837         PF_MD5_UPD(rule, direction);
 838         PF_MD5_UPD(rule, af);
 839         PF_MD5_UPD(rule, quick);
 840         PF_MD5_UPD(rule, ifnot);
 841         PF_MD5_UPD(rule, match_tag_not);
 842         PF_MD5_UPD(rule, natpass);
 843         PF_MD5_UPD(rule, keep_state);
 844         PF_MD5_UPD(rule, proto);
 845         PF_MD5_UPD(rule, type);
 846         PF_MD5_UPD(rule, code);
 847         PF_MD5_UPD(rule, flags);
 848         PF_MD5_UPD(rule, flagset);
 849         PF_MD5_UPD(rule, allow_opts);
 850         PF_MD5_UPD(rule, rt);
 851         PF_MD5_UPD(rule, tos);
 852 }
 853
 854 static int
 855 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
 856 {
 857         struct pf_ruleset       *rs;
 858         struct pf_rule          *rule, **old_array;
 859         struct pf_rulequeue     *old_rules;
 860         int                      error;
 861         u_int32_t                old_rcount;
 862
 863         PF_RULES_WASSERT();
 864
 865         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 866                 return (EINVAL);
 867         rs = pf_find_ruleset(anchor);
 868         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 869             ticket != rs->rules[rs_num].inactive.ticket)
 870                 return (EBUSY);
 871
 872         /* Calculate checksum for the main ruleset */
 873         if (rs == &pf_main_ruleset) {
 874                 error = pf_setup_pfsync_matching(rs);
 875                 if (error != 0)
 876                         return (error);
 877         }
 878
 879         /* Swap rules, keep the old. */
 880         old_rules = rs->rules[rs_num].active.ptr;
 881         old_rcount = rs->rules[rs_num].active.rcount;
 882         old_array = rs->rules[rs_num].active.ptr_array;
 883
 884         rs->rules[rs_num].active.ptr =
 885             rs->rules[rs_num].inactive.ptr;
 886         rs->rules[rs_num].active.ptr_array =
 887             rs->rules[rs_num].inactive.ptr_array;
 888         rs->rules[rs_num].active.rcount =
 889             rs->rules[rs_num].inactive.rcount;
 890         rs->rules[rs_num].inactive.ptr = old_rules;
 891         rs->rules[rs_num].inactive.ptr_array = old_array;
 892         rs->rules[rs_num].inactive.rcount = old_rcount;
 893
 894         rs->rules[rs_num].active.ticket =
 895             rs->rules[rs_num].inactive.ticket;
 896         pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
 897
 898
 899         /* Purge the old rule list. */
 900         while ((rule = TAILQ_FIRST(old_rules)) != NULL)
 901                 pf_unlink_rule(old_rules, rule);
 902         if (rs->rules[rs_num].inactive.ptr_array)
 903                 free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
 904         rs->rules[rs_num].inactive.ptr_array = NULL;
 905         rs->rules[rs_num].inactive.rcount = 0;
 906         rs->rules[rs_num].inactive.open = 0;
 907         pf_remove_if_empty_ruleset(rs);
 908
 909         return (0);
 910 }
 911
 912 static int
 913 pf_setup_pfsync_matching(struct pf_ruleset *rs)
 914 {
 915         MD5_CTX                  ctx;
 916         struct pf_rule          *rule;
 917         int                      rs_cnt;
 918         u_int8_t                 digest[PF_MD5_DIGEST_LENGTH];
 919
 920         MD5Init(&ctx);
 921         for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
 922                 /* XXX PF_RULESET_SCRUB as well? */
 923                 if (rs_cnt == PF_RULESET_SCRUB)
 924                         continue;
 925
 926                 if (rs->rules[rs_cnt].inactive.ptr_array)
 927                         free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
 928                 rs->rules[rs_cnt].inactive.ptr_array = NULL;
 929
 930                 if (rs->rules[rs_cnt].inactive.rcount) {
 931                         rs->rules[rs_cnt].inactive.ptr_array =
 932                             malloc(sizeof(caddr_t) *
 933                             rs->rules[rs_cnt].inactive.rcount,
 934                             M_TEMP, M_NOWAIT);
 935
 936                         if (!rs->rules[rs_cnt].inactive.ptr_array)
 937                                 return (ENOMEM);
 938                 }
 939
 940                 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
 941                     entries) {
 942                         pf_hash_rule(&ctx, rule);
 943                         (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
 944                 }
 945         }
 946
 947         MD5Final(digest, &ctx);
 948         memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
 949         return (0);
 950 }
 951
 952 static int
 953 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
 954     sa_family_t af)
 955 {
 956         int error = 0;
 957
 958         switch (addr->type) {
 959         case PF_ADDR_TABLE:
 960                 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
 961                 if (addr->p.tbl == NULL)
 962                         error = ENOMEM;
 963                 break;
 964         case PF_ADDR_DYNIFTL:
 965                 error = pfi_dynaddr_setup(addr, af);
 966                 break;
 967         }
 968
 969         return (error);
 970 }
 971
 972 static void
 973 pf_addr_copyout(struct pf_addr_wrap *addr)
 974 {
 975
 976         switch (addr->type) {
 977         case PF_ADDR_DYNIFTL:
 978                 pfi_dynaddr_copyout(addr);
 979                 break;
 980         case PF_ADDR_TABLE:
 981                 pf_tbladdr_copyout(addr);
 982                 break;
 983         }
 984 }
 985
 986 static int
 987 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
 988 {
 989         int                      error = 0;
 990
 991         /* XXX keep in sync with switch() below */
 992         if (securelevel_gt(td->td_ucred, 2))
 993                 switch (cmd) {
 994                 case DIOCGETRULES:
 995                 case DIOCGETRULE:
 996                 case DIOCGETADDRS:
 997                 case DIOCGETADDR:
 998                 case DIOCGETSTATE:
 999                 case DIOCSETSTATUSIF:
1000                 case DIOCGETSTATUS:
1001                 case DIOCCLRSTATUS:
1002                 case DIOCNATLOOK:
1003                 case DIOCSETDEBUG:
1004                 case DIOCGETSTATES:
1005                 case DIOCGETTIMEOUT:
1006                 case DIOCCLRRULECTRS:
1007                 case DIOCGETLIMIT:
1008                 case DIOCGETALTQS:
1009                 case DIOCGETALTQ:
1010                 case DIOCGETQSTATS:
1011                 case DIOCGETRULESETS:
1012                 case DIOCGETRULESET:
1013                 case DIOCRGETTABLES:
1014                 case DIOCRGETTSTATS:
1015                 case DIOCRCLRTSTATS:
1016                 case DIOCRCLRADDRS:
1017                 case DIOCRADDADDRS:
1018                 case DIOCRDELADDRS:
1019                 case DIOCRSETADDRS:
1020                 case DIOCRGETADDRS:
1021                 case DIOCRGETASTATS:
1022                 case DIOCRCLRASTATS:
1023                 case DIOCRTSTADDRS:
1024                 case DIOCOSFPGET:
1025                 case DIOCGETSRCNODES:
1026                 case DIOCCLRSRCNODES:
1027                 case DIOCIGETIFACES:
1028                 case DIOCGIFSPEED:
1029                 case DIOCSETIFFLAG:
1030                 case DIOCCLRIFFLAG:
1031                         break;
1032                 case DIOCRCLRTABLES:
1033                 case DIOCRADDTABLES:
1034                 case DIOCRDELTABLES:
1035                 case DIOCRSETTFLAGS:
1036                         if (((struct pfioc_table *)addr)->pfrio_flags &
1037                             PFR_FLAG_DUMMY)
1038                                 break; /* dummy operation ok */
1039                         return (EPERM);
1040                 default:
1041                         return (EPERM);
1042                 }
1043
1044         if (!(flags & FWRITE))
1045                 switch (cmd) {
1046                 case DIOCGETRULES:
1047                 case DIOCGETADDRS:
1048                 case DIOCGETADDR:
1049                 case DIOCGETSTATE:
1050                 case DIOCGETSTATUS:
1051                 case DIOCGETSTATES:
1052                 case DIOCGETTIMEOUT:
1053                 case DIOCGETLIMIT:
1054                 case DIOCGETALTQS:
1055                 case DIOCGETALTQ:
1056                 case DIOCGETQSTATS:
1057                 case DIOCGETRULESETS:
1058                 case DIOCGETRULESET:
1059                 case DIOCNATLOOK:
1060                 case DIOCRGETTABLES:
1061                 case DIOCRGETTSTATS:
1062                 case DIOCRGETADDRS:
1063                 case DIOCRGETASTATS:
1064                 case DIOCRTSTADDRS:
1065                 case DIOCOSFPGET:
1066                 case DIOCGETSRCNODES:
1067                 case DIOCIGETIFACES:
1068                 case DIOCGIFSPEED:
1069                         break;
1070                 case DIOCRCLRTABLES:
1071                 case DIOCRADDTABLES:
1072                 case DIOCRDELTABLES:
1073                 case DIOCRCLRTSTATS:
1074                 case DIOCRCLRADDRS:
1075                 case DIOCRADDADDRS:
1076                 case DIOCRDELADDRS:
1077                 case DIOCRSETADDRS:
1078                 case DIOCRSETTFLAGS:
1079                         if (((struct pfioc_table *)addr)->pfrio_flags &
1080                             PFR_FLAG_DUMMY) {
1081                                 flags |= FWRITE; /* need write lock for dummy */
1082                                 break; /* dummy operation ok */
1083                         }
1084                         return (EACCES);
1085                 case DIOCGETRULE:
1086                         if (((struct pfioc_rule *)addr)->action ==
1087                             PF_GET_CLR_CNTR)
1088                                 return (EACCES);
1089                         break;
1090                 default:
1091                         return (EACCES);
1092                 }
1093
1094         CURVNET_SET(TD_TO_VNET(td));
1095
1096         switch (cmd) {
1097         case DIOCSTART:
1098                 sx_xlock(&pf_ioctl_lock);
1099                 if (V_pf_status.running)
1100                         error = EEXIST;
1101                 else {
1102                         int cpu;
1103
1104                         error = hook_pf();
1105                         if (error) {
1106                                 DPFPRINTF(PF_DEBUG_MISC,
1107                                     ("pf: pfil registration failed\n"));
1108                                 break;
1109                         }
1110                         V_pf_status.running = 1;
1111                         V_pf_status.since = time_second;
1112
1113                         CPU_FOREACH(cpu)
1114                                 V_pf_stateid[cpu] = time_second;
1115
1116                         DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1117                 }
1118                 break;
1119
1120         case DIOCSTOP:
1121                 sx_xlock(&pf_ioctl_lock);
1122                 if (!V_pf_status.running)
1123                         error = ENOENT;
1124                 else {
1125                         V_pf_status.running = 0;
1126                         error = dehook_pf();
1127                         if (error) {
1128                                 V_pf_status.running = 1;
1129                                 DPFPRINTF(PF_DEBUG_MISC,
1130                                     ("pf: pfil unregistration failed\n"));
1131                         }
1132                         V_pf_status.since = time_second;
1133                         DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1134                 }
1135                 break;
1136
1137         case DIOCADDRULE: {
1138                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1139                 struct pf_ruleset       *ruleset;
1140                 struct pf_rule          *rule, *tail;
1141                 struct pf_pooladdr      *pa;
1142                 struct pfi_kif          *kif = NULL;
1143                 int                      rs_num;
1144
1145                 if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1146                         error = EINVAL;
1147                         break;
1148                 }
1149 #ifndef INET
1150                 if (pr->rule.af == AF_INET) {
1151                         error = EAFNOSUPPORT;
1152                         break;
1153                 }
1154 #endif /* INET */
1155 #ifndef INET6
1156                 if (pr->rule.af == AF_INET6) {
1157                         error = EAFNOSUPPORT;
1158                         break;
1159                 }
1160 #endif /* INET6 */
1161
1162                 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1163                 bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1164                 if (rule->ifname[0])
1165                         kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1166                 rule->states_cur = counter_u64_alloc(M_WAITOK);
1167                 rule->states_tot = counter_u64_alloc(M_WAITOK);
1168                 rule->src_nodes = counter_u64_alloc(M_WAITOK);
1169                 rule->cuid = td->td_ucred->cr_ruid;
1170                 rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1171                 TAILQ_INIT(&rule->rpool.list);
1172
1173 #define ERROUT(x)       { error = (x); goto DIOCADDRULE_error; }
1174
1175                 PF_RULES_WLOCK();
1176                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1177                 ruleset = pf_find_ruleset(pr->anchor);
1178                 if (ruleset == NULL)
1179                         ERROUT(EINVAL);
1180                 rs_num = pf_get_ruleset_number(pr->rule.action);
1181                 if (rs_num >= PF_RULESET_MAX)
1182                         ERROUT(EINVAL);
1183                 if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1184                         DPFPRINTF(PF_DEBUG_MISC,
1185                             ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1186                             ruleset->rules[rs_num].inactive.ticket));
1187                         ERROUT(EBUSY);
1188                 }
1189                 if (pr->pool_ticket != V_ticket_pabuf) {
1190                         DPFPRINTF(PF_DEBUG_MISC,
1191                             ("pool_ticket: %d != %d\n", pr->pool_ticket,
1192                             V_ticket_pabuf));
1193                         ERROUT(EBUSY);
1194                 }
1195
1196                 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1197                     pf_rulequeue);
1198                 if (tail)
1199                         rule->nr = tail->nr + 1;
1200                 else
1201                         rule->nr = 0;
1202                 if (rule->ifname[0]) {
1203                         rule->kif = pfi_kif_attach(kif, rule->ifname);
1204                         pfi_kif_ref(rule->kif);
1205                 } else
1206                         rule->kif = NULL;
1207
1208                 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1209                         error = EBUSY;
1210
1211 #ifdef ALTQ
1212                 /* set queue IDs */
1213                 if (rule->qname[0] != 0) {
1214                         if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1215                                 error = EBUSY;
1216                         else if (rule->pqname[0] != 0) {
1217                                 if ((rule->pqid =
1218                                     pf_qname2qid(rule->pqname)) == 0)
1219                                         error = EBUSY;
1220                         } else
1221                                 rule->pqid = rule->qid;
1222                 }
1223 #endif
1224                 if (rule->tagname[0])
1225                         if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1226                                 error = EBUSY;
1227                 if (rule->match_tagname[0])
1228                         if ((rule->match_tag =
1229                             pf_tagname2tag(rule->match_tagname)) == 0)
1230                                 error = EBUSY;
1231                 if (rule->rt && !rule->direction)
1232                         error = EINVAL;
1233                 if (!rule->log)
1234                         rule->logif = 0;
1235                 if (rule->logif >= PFLOGIFS_MAX)
1236                         error = EINVAL;
1237                 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1238                         error = ENOMEM;
1239                 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1240                         error = ENOMEM;
1241                 if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1242                         error = EINVAL;
1243                 TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1244                         if (pa->addr.type == PF_ADDR_TABLE) {
1245                                 pa->addr.p.tbl = pfr_attach_table(ruleset,
1246                                     pa->addr.v.tblname);
1247                                 if (pa->addr.p.tbl == NULL)
1248                                         error = ENOMEM;
1249                         }
1250
1251                 if (rule->overload_tblname[0]) {
1252                         if ((rule->overload_tbl = pfr_attach_table(ruleset,
1253                             rule->overload_tblname)) == NULL)
1254                                 error = EINVAL;
1255                         else
1256                                 rule->overload_tbl->pfrkt_flags |=
1257                                     PFR_TFLAG_ACTIVE;
1258                 }
1259
1260                 pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1261                 if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1262                     (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1263                     (rule->rt > PF_FASTROUTE)) &&
1264                     (TAILQ_FIRST(&rule->rpool.list) == NULL))
1265                         error = EINVAL;
1266
1267                 if (error) {
1268                         pf_free_rule(rule);
1269                         PF_RULES_WUNLOCK();
1270                         break;
1271                 }
1272
1273                 rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1274                 rule->evaluations = rule->packets[0] = rule->packets[1] =
1275                     rule->bytes[0] = rule->bytes[1] = 0;
1276                 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1277                     rule, entries);
1278                 ruleset->rules[rs_num].inactive.rcount++;
1279                 PF_RULES_WUNLOCK();
1280                 break;
1281
1282 #undef ERROUT
1283 DIOCADDRULE_error:
1284                 PF_RULES_WUNLOCK();
1285                 counter_u64_free(rule->states_cur);
1286                 counter_u64_free(rule->states_tot);
1287                 counter_u64_free(rule->src_nodes);
1288                 free(rule, M_PFRULE);
1289                 if (kif)
1290                         free(kif, PFI_MTYPE);
1291                 break;
1292         }
1293
1294         case DIOCGETRULES: {
1295                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1296                 struct pf_ruleset       *ruleset;
1297                 struct pf_rule          *tail;
1298                 int                      rs_num;
1299
1300                 PF_RULES_WLOCK();
1301                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1302                 ruleset = pf_find_ruleset(pr->anchor);
1303                 if (ruleset == NULL) {
1304                         PF_RULES_WUNLOCK();
1305                         error = EINVAL;
1306                         break;
1307                 }
1308                 rs_num = pf_get_ruleset_number(pr->rule.action);
1309                 if (rs_num >= PF_RULESET_MAX) {
1310                         PF_RULES_WUNLOCK();
1311                         error = EINVAL;
1312                         break;
1313                 }
1314                 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1315                     pf_rulequeue);
1316                 if (tail)
1317                         pr->nr = tail->nr + 1;
1318                 else
1319                         pr->nr = 0;
1320                 pr->ticket = ruleset->rules[rs_num].active.ticket;
1321                 PF_RULES_WUNLOCK();
1322                 break;
1323         }
1324
1325         case DIOCGETRULE: {
1326                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1327                 struct pf_ruleset       *ruleset;
1328                 struct pf_rule          *rule;
1329                 int                      rs_num, i;
1330
1331                 PF_RULES_WLOCK();
1332                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1333                 ruleset = pf_find_ruleset(pr->anchor);
1334                 if (ruleset == NULL) {
1335                         PF_RULES_WUNLOCK();
1336                         error = EINVAL;
1337                         break;
1338                 }
1339                 rs_num = pf_get_ruleset_number(pr->rule.action);
1340                 if (rs_num >= PF_RULESET_MAX) {
1341                         PF_RULES_WUNLOCK();
1342                         error = EINVAL;
1343                         break;
1344                 }
1345                 if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1346                         PF_RULES_WUNLOCK();
1347                         error = EBUSY;
1348                         break;
1349                 }
1350                 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1351                 while ((rule != NULL) && (rule->nr != pr->nr))
1352                         rule = TAILQ_NEXT(rule, entries);
1353                 if (rule == NULL) {
1354                         PF_RULES_WUNLOCK();
1355                         error = EBUSY;
1356                         break;
1357                 }
1358                 bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1359                 pr->rule.u_states_cur = counter_u64_fetch(rule->states_cur);
1360                 pr->rule.u_states_tot = counter_u64_fetch(rule->states_tot);
1361                 pr->rule.u_src_nodes = counter_u64_fetch(rule->src_nodes);
1362                 if (pf_anchor_copyout(ruleset, rule, pr)) {
1363                         PF_RULES_WUNLOCK();
1364                         error = EBUSY;
1365                         break;
1366                 }
1367                 pf_addr_copyout(&pr->rule.src.addr);
1368                 pf_addr_copyout(&pr->rule.dst.addr);
1369                 for (i = 0; i < PF_SKIP_COUNT; ++i)
1370                         if (rule->skip[i].ptr == NULL)
1371                                 pr->rule.skip[i].nr = -1;
1372                         else
1373                                 pr->rule.skip[i].nr =
1374                                     rule->skip[i].ptr->nr;
1375
1376                 if (pr->action == PF_GET_CLR_CNTR) {
1377                         rule->evaluations = 0;
1378                         rule->packets[0] = rule->packets[1] = 0;
1379                         rule->bytes[0] = rule->bytes[1] = 0;
1380                         counter_u64_zero(rule->states_tot);
1381                 }
1382                 PF_RULES_WUNLOCK();
1383                 break;
1384         }
1385
1386         case DIOCCHANGERULE: {
1387                 struct pfioc_rule       *pcr = (struct pfioc_rule *)addr;
1388                 struct pf_ruleset       *ruleset;
1389                 struct pf_rule          *oldrule = NULL, *newrule = NULL;
1390                 struct pfi_kif          *kif = NULL;
1391                 struct pf_pooladdr      *pa;
1392                 u_int32_t                nr = 0;
1393                 int                      rs_num;
1394
1395                 if (pcr->action < PF_CHANGE_ADD_HEAD ||
1396                     pcr->action > PF_CHANGE_GET_TICKET) {
1397                         error = EINVAL;
1398                         break;
1399                 }
1400                 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1401                         error = EINVAL;
1402                         break;
1403                 }
1404
1405                 if (pcr->action != PF_CHANGE_REMOVE) {
1406 #ifndef INET
1407                         if (pcr->rule.af == AF_INET) {
1408                                 error = EAFNOSUPPORT;
1409                                 break;
1410                         }
1411 #endif /* INET */
1412 #ifndef INET6
1413                         if (pcr->rule.af == AF_INET6) {
1414                                 error = EAFNOSUPPORT;
1415                                 break;
1416                         }
1417 #endif /* INET6 */
1418                         newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1419                         bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1420                         if (newrule->ifname[0])
1421                                 kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1422                         newrule->states_cur = counter_u64_alloc(M_WAITOK);
1423                         newrule->states_tot = counter_u64_alloc(M_WAITOK);
1424                         newrule->src_nodes = counter_u64_alloc(M_WAITOK);
1425                         newrule->cuid = td->td_ucred->cr_ruid;
1426                         newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1427                         TAILQ_INIT(&newrule->rpool.list);
1428                 }
1429
1430 #define ERROUT(x)       { error = (x); goto DIOCCHANGERULE_error; }
1431
1432                 PF_RULES_WLOCK();
1433                 if (!(pcr->action == PF_CHANGE_REMOVE ||
1434                     pcr->action == PF_CHANGE_GET_TICKET) &&
1435                     pcr->pool_ticket != V_ticket_pabuf)
1436                         ERROUT(EBUSY);
1437
1438                 ruleset = pf_find_ruleset(pcr->anchor);
1439                 if (ruleset == NULL)
1440                         ERROUT(EINVAL);
1441
1442                 rs_num = pf_get_ruleset_number(pcr->rule.action);
1443                 if (rs_num >= PF_RULESET_MAX)
1444                         ERROUT(EINVAL);
1445
1446                 if (pcr->action == PF_CHANGE_GET_TICKET) {
1447                         pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1448                         ERROUT(0);
1449                 } else if (pcr->ticket !=
1450                             ruleset->rules[rs_num].active.ticket)
1451                                 ERROUT(EINVAL);
1452
1453                 if (pcr->action != PF_CHANGE_REMOVE) {
1454                         if (newrule->ifname[0]) {
1455                                 newrule->kif = pfi_kif_attach(kif,
1456                                     newrule->ifname);
1457                                 pfi_kif_ref(newrule->kif);
1458                         } else
1459                                 newrule->kif = NULL;
1460
1461                         if (newrule->rtableid > 0 &&
1462                             newrule->rtableid >= rt_numfibs)
1463                                 error = EBUSY;
1464
1465 #ifdef ALTQ
1466                         /* set queue IDs */
1467                         if (newrule->qname[0] != 0) {
1468                                 if ((newrule->qid =
1469                                     pf_qname2qid(newrule->qname)) == 0)
1470                                         error = EBUSY;
1471                                 else if (newrule->pqname[0] != 0) {
1472                                         if ((newrule->pqid =
1473                                             pf_qname2qid(newrule->pqname)) == 0)
1474                                                 error = EBUSY;
1475                                 } else
1476                                         newrule->pqid = newrule->qid;
1477                         }
1478 #endif /* ALTQ */
1479                         if (newrule->tagname[0])
1480                                 if ((newrule->tag =
1481                                     pf_tagname2tag(newrule->tagname)) == 0)
1482                                         error = EBUSY;
1483                         if (newrule->match_tagname[0])
1484                                 if ((newrule->match_tag = pf_tagname2tag(
1485                                     newrule->match_tagname)) == 0)
1486                                         error = EBUSY;
1487                         if (newrule->rt && !newrule->direction)
1488                                 error = EINVAL;
1489                         if (!newrule->log)
1490                                 newrule->logif = 0;
1491                         if (newrule->logif >= PFLOGIFS_MAX)
1492                                 error = EINVAL;
1493                         if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1494                                 error = ENOMEM;
1495                         if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1496                                 error = ENOMEM;
1497                         if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1498                                 error = EINVAL;
1499                         TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1500                                 if (pa->addr.type == PF_ADDR_TABLE) {
1501                                         pa->addr.p.tbl =
1502                                             pfr_attach_table(ruleset,
1503                                             pa->addr.v.tblname);
1504                                         if (pa->addr.p.tbl == NULL)
1505                                                 error = ENOMEM;
1506                                 }
1507
1508                         if (newrule->overload_tblname[0]) {
1509                                 if ((newrule->overload_tbl = pfr_attach_table(
1510                                     ruleset, newrule->overload_tblname)) ==
1511                                     NULL)
1512                                         error = EINVAL;
1513                                 else
1514                                         newrule->overload_tbl->pfrkt_flags |=
1515                                             PFR_TFLAG_ACTIVE;
1516                         }
1517
1518                         pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1519                         if (((((newrule->action == PF_NAT) ||
1520                             (newrule->action == PF_RDR) ||
1521                             (newrule->action == PF_BINAT) ||
1522                             (newrule->rt > PF_FASTROUTE)) &&
1523                             !newrule->anchor)) &&
1524                             (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1525                                 error = EINVAL;
1526
1527                         if (error) {
1528                                 pf_free_rule(newrule);
1529                                 PF_RULES_WUNLOCK();
1530                                 break;
1531                         }
1532
1533                         newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1534                         newrule->evaluations = 0;
1535                         newrule->packets[0] = newrule->packets[1] = 0;
1536                         newrule->bytes[0] = newrule->bytes[1] = 0;
1537                 }
1538                 pf_empty_pool(&V_pf_pabuf);
1539
1540                 if (pcr->action == PF_CHANGE_ADD_HEAD)
1541                         oldrule = TAILQ_FIRST(
1542                             ruleset->rules[rs_num].active.ptr);
1543                 else if (pcr->action == PF_CHANGE_ADD_TAIL)
1544                         oldrule = TAILQ_LAST(
1545                             ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1546                 else {
1547                         oldrule = TAILQ_FIRST(
1548                             ruleset->rules[rs_num].active.ptr);
1549                         while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1550                                 oldrule = TAILQ_NEXT(oldrule, entries);
1551                         if (oldrule == NULL) {
1552                                 if (newrule != NULL)
1553                                         pf_free_rule(newrule);
1554                                 PF_RULES_WUNLOCK();
1555                                 error = EINVAL;
1556                                 break;
1557                         }
1558                 }
1559
1560                 if (pcr->action == PF_CHANGE_REMOVE) {
1561                         pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1562                             oldrule);
1563                         ruleset->rules[rs_num].active.rcount--;
1564                 } else {
1565                         if (oldrule == NULL)
1566                                 TAILQ_INSERT_TAIL(
1567                                     ruleset->rules[rs_num].active.ptr,
1568                                     newrule, entries);
1569                         else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1570                             pcr->action == PF_CHANGE_ADD_BEFORE)
1571                                 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1572                         else
1573                                 TAILQ_INSERT_AFTER(
1574                                     ruleset->rules[rs_num].active.ptr,
1575                                     oldrule, newrule, entries);
1576                         ruleset->rules[rs_num].active.rcount++;
1577                 }
1578
1579                 nr = 0;
1580                 TAILQ_FOREACH(oldrule,
1581                     ruleset->rules[rs_num].active.ptr, entries)
1582                         oldrule->nr = nr++;
1583
1584                 ruleset->rules[rs_num].active.ticket++;
1585
1586                 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1587                 pf_remove_if_empty_ruleset(ruleset);
1588
1589                 PF_RULES_WUNLOCK();
1590                 break;
1591
1592 #undef ERROUT
1593 DIOCCHANGERULE_error:
1594                 PF_RULES_WUNLOCK();
1595                 if (newrule != NULL) {
1596                         counter_u64_free(newrule->states_cur);
1597                         counter_u64_free(newrule->states_tot);
1598                         counter_u64_free(newrule->src_nodes);
1599                         free(newrule, M_PFRULE);
1600                 }
1601                 if (kif != NULL)
1602                         free(kif, PFI_MTYPE);
1603                 break;
1604         }
1605
1606         case DIOCCLRSTATES: {
1607                 struct pf_state         *s;
1608                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1609                 u_int                    i, killed = 0;
1610
1611                 for (i = 0; i <= pf_hashmask; i++) {
1612                         struct pf_idhash *ih = &V_pf_idhash[i];
1613
1614 relock_DIOCCLRSTATES:
1615                         PF_HASHROW_LOCK(ih);
1616                         LIST_FOREACH(s, &ih->states, entry)
1617                                 if (!psk->psk_ifname[0] ||
1618                                     !strcmp(psk->psk_ifname,
1619                                     s->kif->pfik_name)) {
1620                                         /*
1621                                          * Don't send out individual
1622                                          * delete messages.
1623                                          */
1624                                         s->state_flags |= PFSTATE_NOSYNC;
1625                                         pf_unlink_state(s, PF_ENTER_LOCKED);
1626                                         killed++;
1627                                         goto relock_DIOCCLRSTATES;
1628                                 }
1629                         PF_HASHROW_UNLOCK(ih);
1630                 }
1631                 psk->psk_killed = killed;
1632                 if (pfsync_clear_states_ptr != NULL)
1633                         pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
1634                 break;
1635         }
1636
1637         case DIOCKILLSTATES: {
1638                 struct pf_state         *s;
1639                 struct pf_state_key     *sk;
1640                 struct pf_addr          *srcaddr, *dstaddr;
1641                 u_int16_t                srcport, dstport;
1642                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1643                 u_int                    i, killed = 0;
1644
1645                 if (psk->psk_pfcmp.id) {
1646                         if (psk->psk_pfcmp.creatorid == 0)
1647                                 psk->psk_pfcmp.creatorid = V_pf_status.hostid;
1648                         if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
1649                             psk->psk_pfcmp.creatorid))) {
1650                                 pf_unlink_state(s, PF_ENTER_LOCKED);
1651                                 psk->psk_killed = 1;
1652                         }
1653                         break;
1654                 }
1655
1656                 for (i = 0; i <= pf_hashmask; i++) {
1657                         struct pf_idhash *ih = &V_pf_idhash[i];
1658
1659 relock_DIOCKILLSTATES:
1660                         PF_HASHROW_LOCK(ih);
1661                         LIST_FOREACH(s, &ih->states, entry) {
1662                                 sk = s->key[PF_SK_WIRE];
1663                                 if (s->direction == PF_OUT) {
1664                                         srcaddr = &sk->addr[1];
1665                                         dstaddr = &sk->addr[0];
1666                                         srcport = sk->port[1];
1667                                         dstport = sk->port[0];
1668                                 } else {
1669                                         srcaddr = &sk->addr[0];
1670                                         dstaddr = &sk->addr[1];
1671                                         srcport = sk->port[0];
1672                                         dstport = sk->port[1];
1673                                 }
1674
1675                                 if ((!psk->psk_af || sk->af == psk->psk_af)
1676                                     && (!psk->psk_proto || psk->psk_proto ==
1677                                     sk->proto) &&
1678                                     PF_MATCHA(psk->psk_src.neg,
1679                                     &psk->psk_src.addr.v.a.addr,
1680                                     &psk->psk_src.addr.v.a.mask,
1681                                     srcaddr, sk->af) &&
1682                                     PF_MATCHA(psk->psk_dst.neg,
1683                                     &psk->psk_dst.addr.v.a.addr,
1684                                     &psk->psk_dst.addr.v.a.mask,
1685                                     dstaddr, sk->af) &&
1686                                     (psk->psk_src.port_op == 0 ||
1687                                     pf_match_port(psk->psk_src.port_op,
1688                                     psk->psk_src.port[0], psk->psk_src.port[1],
1689                                     srcport)) &&
1690                                     (psk->psk_dst.port_op == 0 ||
1691                                     pf_match_port(psk->psk_dst.port_op,
1692                                     psk->psk_dst.port[0], psk->psk_dst.port[1],
1693                                     dstport)) &&
1694                                     (!psk->psk_label[0] ||
1695                                     (s->rule.ptr->label[0] &&
1696                                     !strcmp(psk->psk_label,
1697                                     s->rule.ptr->label))) &&
1698                                     (!psk->psk_ifname[0] ||
1699                                     !strcmp(psk->psk_ifname,
1700                                     s->kif->pfik_name))) {
1701                                         pf_unlink_state(s, PF_ENTER_LOCKED);
1702                                         killed++;
1703                                         goto relock_DIOCKILLSTATES;
1704                                 }
1705                         }
1706                         PF_HASHROW_UNLOCK(ih);
1707                 }
1708                 psk->psk_killed = killed;
1709                 break;
1710         }
1711
1712         case DIOCADDSTATE: {
1713                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
1714                 struct pfsync_state     *sp = &ps->state;
1715
1716                 if (sp->timeout >= PFTM_MAX) {
1717                         error = EINVAL;
1718                         break;
1719                 }
1720                 if (pfsync_state_import_ptr != NULL) {
1721                         PF_RULES_RLOCK();
1722                         error = pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
1723                         PF_RULES_RUNLOCK();
1724                 } else
1725                         error = EOPNOTSUPP;
1726                 break;
1727         }
1728
1729         case DIOCGETSTATE: {
1730                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
1731                 struct pf_state         *s;
1732
1733                 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
1734                 if (s == NULL) {
1735                         error = ENOENT;
1736                         break;
1737                 }
1738
1739                 pfsync_state_export(&ps->state, s);
1740                 PF_STATE_UNLOCK(s);
1741                 break;
1742         }
1743
1744         case DIOCGETSTATES: {
1745                 struct pfioc_states     *ps = (struct pfioc_states *)addr;
1746                 struct pf_state         *s;
1747                 struct pfsync_state     *pstore, *p;
1748                 int i, nr;
1749
1750                 if (ps->ps_len == 0) {
1751                         nr = uma_zone_get_cur(V_pf_state_z);
1752                         ps->ps_len = sizeof(struct pfsync_state) * nr;
1753                         break;
1754                 }
1755
1756                 p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK);
1757                 nr = 0;
1758
1759                 for (i = 0; i <= pf_hashmask; i++) {
1760                         struct pf_idhash *ih = &V_pf_idhash[i];
1761
1762                         PF_HASHROW_LOCK(ih);
1763                         LIST_FOREACH(s, &ih->states, entry) {
1764
1765                                 if (s->timeout == PFTM_UNLINKED)
1766                                         continue;
1767
1768                                 if ((nr+1) * sizeof(*p) > ps->ps_len) {
1769                                         PF_HASHROW_UNLOCK(ih);
1770                                         goto DIOCGETSTATES_full;
1771                                 }
1772                                 pfsync_state_export(p, s);
1773                                 p++;
1774                                 nr++;
1775                         }
1776                         PF_HASHROW_UNLOCK(ih);
1777                 }
1778 DIOCGETSTATES_full:
1779                 error = copyout(pstore, ps->ps_states,
1780                     sizeof(struct pfsync_state) * nr);
1781                 if (error) {
1782                         free(pstore, M_TEMP);
1783                         break;
1784                 }
1785                 ps->ps_len = sizeof(struct pfsync_state) * nr;
1786                 free(pstore, M_TEMP);
1787
1788                 break;
1789         }
1790
1791         case DIOCGETSTATUS: {
1792                 struct pf_status *s = (struct pf_status *)addr;
1793
1794                 PF_RULES_RLOCK();
1795                 s->running = V_pf_status.running;
1796                 s->since   = V_pf_status.since;
1797                 s->debug   = V_pf_status.debug;
1798                 s->hostid  = V_pf_status.hostid;
1799                 s->states  = V_pf_status.states;
1800                 s->src_nodes = V_pf_status.src_nodes;
1801
1802                 for (int i = 0; i < PFRES_MAX; i++)
1803                         s->counters[i] =
1804                             counter_u64_fetch(V_pf_status.counters[i]);
1805                 for (int i = 0; i < LCNT_MAX; i++)
1806                         s->lcounters[i] =
1807                             counter_u64_fetch(V_pf_status.lcounters[i]);
1808                 for (int i = 0; i < FCNT_MAX; i++)
1809                         s->fcounters[i] =
1810                             counter_u64_fetch(V_pf_status.fcounters[i]);
1811                 for (int i = 0; i < SCNT_MAX; i++)
1812                         s->scounters[i] =
1813                             counter_u64_fetch(V_pf_status.scounters[i]);
1814
1815                 bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
1816                 bcopy(V_pf_status.pf_chksum, s->pf_chksum,
1817                     PF_MD5_DIGEST_LENGTH);
1818
1819                 pfi_update_status(s->ifname, s);
1820                 PF_RULES_RUNLOCK();
1821                 break;
1822         }
1823
1824         case DIOCSETSTATUSIF: {
1825                 struct pfioc_if *pi = (struct pfioc_if *)addr;
1826
1827                 if (pi->ifname[0] == 0) {
1828                         bzero(V_pf_status.ifname, IFNAMSIZ);
1829                         break;
1830                 }
1831                 PF_RULES_WLOCK();
1832                 strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
1833                 PF_RULES_WUNLOCK();
1834                 break;
1835         }
1836
1837         case DIOCCLRSTATUS: {
1838                 PF_RULES_WLOCK();
1839                 for (int i = 0; i < PFRES_MAX; i++)
1840                         counter_u64_zero(V_pf_status.counters[i]);
1841                 for (int i = 0; i < FCNT_MAX; i++)
1842                         counter_u64_zero(V_pf_status.fcounters[i]);
1843                 for (int i = 0; i < SCNT_MAX; i++)
1844                         counter_u64_zero(V_pf_status.scounters[i]);
1845                 V_pf_status.since = time_second;
1846                 if (*V_pf_status.ifname)
1847                         pfi_update_status(V_pf_status.ifname, NULL);
1848                 PF_RULES_WUNLOCK();
1849                 break;
1850         }
1851
1852         case DIOCNATLOOK: {
1853                 struct pfioc_natlook    *pnl = (struct pfioc_natlook *)addr;
1854                 struct pf_state_key     *sk;
1855                 struct pf_state         *state;
1856                 struct pf_state_key_cmp  key;
1857                 int                      m = 0, direction = pnl->direction;
1858                 int                      sidx, didx;
1859
1860                 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
1861                 sidx = (direction == PF_IN) ? 1 : 0;
1862                 didx = (direction == PF_IN) ? 0 : 1;
1863
1864                 if (!pnl->proto ||
1865                     PF_AZERO(&pnl->saddr, pnl->af) ||
1866                     PF_AZERO(&pnl->daddr, pnl->af) ||
1867                     ((pnl->proto == IPPROTO_TCP ||
1868                     pnl->proto == IPPROTO_UDP) &&
1869                     (!pnl->dport || !pnl->sport)))
1870                         error = EINVAL;
1871                 else {
1872                         bzero(&key, sizeof(key));
1873                         key.af = pnl->af;
1874                         key.proto = pnl->proto;
1875                         PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1876                         key.port[sidx] = pnl->sport;
1877                         PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1878                         key.port[didx] = pnl->dport;
1879
1880                         state = pf_find_state_all(&key, direction, &m);
1881
1882                         if (m > 1)
1883                                 error = E2BIG;  /* more than one state */
1884                         else if (state != NULL) {
1885                                 /* XXXGL: not locked read */
1886                                 sk = state->key[sidx];
1887                                 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1888                                 pnl->rsport = sk->port[sidx];
1889                                 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1890                                 pnl->rdport = sk->port[didx];
1891                         } else
1892                                 error = ENOENT;
1893                 }
1894                 break;
1895         }
1896
1897         case DIOCSETTIMEOUT: {
1898                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
1899                 int              old;
1900
1901                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1902                     pt->seconds < 0) {
1903                         error = EINVAL;
1904                         break;
1905                 }
1906                 PF_RULES_WLOCK();
1907                 old = V_pf_default_rule.timeout[pt->timeout];
1908                 if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1909                         pt->seconds = 1;
1910                 V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
1911                 if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1912                         wakeup(pf_purge_thread);
1913                 pt->seconds = old;
1914                 PF_RULES_WUNLOCK();
1915                 break;
1916         }
1917
1918         case DIOCGETTIMEOUT: {
1919                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
1920
1921                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1922                         error = EINVAL;
1923                         break;
1924                 }
1925                 PF_RULES_RLOCK();
1926                 pt->seconds = V_pf_default_rule.timeout[pt->timeout];
1927                 PF_RULES_RUNLOCK();
1928                 break;
1929         }
1930
1931         case DIOCGETLIMIT: {
1932                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
1933
1934                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1935                         error = EINVAL;
1936                         break;
1937                 }
1938                 PF_RULES_RLOCK();
1939                 pl->limit = V_pf_limits[pl->index].limit;
1940                 PF_RULES_RUNLOCK();
1941                 break;
1942         }
1943
1944         case DIOCSETLIMIT: {
1945                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
1946                 int                      old_limit;
1947
1948                 PF_RULES_WLOCK();
1949                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1950                     V_pf_limits[pl->index].zone == NULL) {
1951                         PF_RULES_WUNLOCK();
1952                         error = EINVAL;
1953                         break;
1954                 }
1955                 uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
1956                 old_limit = V_pf_limits[pl->index].limit;
1957                 V_pf_limits[pl->index].limit = pl->limit;
1958                 pl->limit = old_limit;
1959                 PF_RULES_WUNLOCK();
1960                 break;
1961         }
1962
1963         case DIOCSETDEBUG: {
1964                 u_int32_t       *level = (u_int32_t *)addr;
1965
1966                 PF_RULES_WLOCK();
1967                 V_pf_status.debug = *level;
1968                 PF_RULES_WUNLOCK();
1969                 break;
1970         }
1971
1972         case DIOCCLRRULECTRS: {
1973                 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1974                 struct pf_ruleset       *ruleset = &pf_main_ruleset;
1975                 struct pf_rule          *rule;
1976
1977                 PF_RULES_WLOCK();
1978                 TAILQ_FOREACH(rule,
1979                     ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1980                         rule->evaluations = 0;
1981                         rule->packets[0] = rule->packets[1] = 0;
1982                         rule->bytes[0] = rule->bytes[1] = 0;
1983                 }
1984                 PF_RULES_WUNLOCK();
1985                 break;
1986         }
1987
1988         case DIOCGIFSPEED: {
1989                 struct pf_ifspeed       *psp = (struct pf_ifspeed *)addr;
1990                 struct pf_ifspeed       ps;
1991                 struct ifnet            *ifp;
1992
1993                 if (psp->ifname[0] != 0) {
1994                         /* Can we completely trust user-land? */
1995                         strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1996                         ifp = ifunit(ps.ifname);
1997                         if (ifp != NULL)
1998                                 psp->baudrate = ifp->if_baudrate;
1999                         else
2000                                 error = EINVAL;
2001                 } else
2002                         error = EINVAL;
2003                 break;
2004         }
2005
2006 #ifdef ALTQ
2007         case DIOCSTARTALTQ: {
2008                 struct pf_altq          *altq;
2009
2010                 PF_RULES_WLOCK();
2011                 /* enable all altq interfaces on active list */
2012                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2013                         if (altq->qname[0] == 0 && (altq->local_flags &
2014                             PFALTQ_FLAG_IF_REMOVED) == 0) {
2015                                 error = pf_enable_altq(altq);
2016                                 if (error != 0)
2017                                         break;
2018                         }
2019                 }
2020                 if (error == 0)
2021                         V_pf_altq_running = 1;
2022                 PF_RULES_WUNLOCK();
2023                 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2024                 break;
2025         }
2026
2027         case DIOCSTOPALTQ: {
2028                 struct pf_altq          *altq;
2029
2030                 PF_RULES_WLOCK();
2031                 /* disable all altq interfaces on active list */
2032                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2033                         if (altq->qname[0] == 0 && (altq->local_flags &
2034                             PFALTQ_FLAG_IF_REMOVED) == 0) {
2035                                 error = pf_disable_altq(altq);
2036                                 if (error != 0)
2037                                         break;
2038                         }
2039                 }
2040                 if (error == 0)
2041                         V_pf_altq_running = 0;
2042                 PF_RULES_WUNLOCK();
2043                 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2044                 break;
2045         }
2046
2047         case DIOCADDALTQ: {
2048                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2049                 struct pf_altq          *altq, *a;
2050                 struct ifnet            *ifp;
2051
2052                 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK);
2053                 bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2054                 altq->local_flags = 0;
2055
2056                 PF_RULES_WLOCK();
2057                 if (pa->ticket != V_ticket_altqs_inactive) {
2058                         PF_RULES_WUNLOCK();
2059                         free(altq, M_PFALTQ);
2060                         error = EBUSY;
2061                         break;
2062                 }
2063
2064                 /*
2065                  * if this is for a queue, find the discipline and
2066                  * copy the necessary fields
2067                  */
2068                 if (altq->qname[0] != 0) {
2069                         if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2070                                 PF_RULES_WUNLOCK();
2071                                 error = EBUSY;
2072                                 free(altq, M_PFALTQ);
2073                                 break;
2074                         }
2075                         altq->altq_disc = NULL;
2076                         TAILQ_FOREACH(a, V_pf_altqs_inactive, entries) {
2077                                 if (strncmp(a->ifname, altq->ifname,
2078                                     IFNAMSIZ) == 0 && a->qname[0] == 0) {
2079                                         altq->altq_disc = a->altq_disc;
2080                                         break;
2081                                 }
2082                         }
2083                 }
2084
2085                 if ((ifp = ifunit(altq->ifname)) == NULL)
2086                         altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2087                 else
2088                         error = altq_add(altq);
2089
2090                 if (error) {
2091                         PF_RULES_WUNLOCK();
2092                         free(altq, M_PFALTQ);
2093                         break;
2094                 }
2095
2096                 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2097                 bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2098                 PF_RULES_WUNLOCK();
2099                 break;
2100         }
2101
2102         case DIOCGETALTQS: {
2103                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2104                 struct pf_altq          *altq;
2105
2106                 PF_RULES_RLOCK();
2107                 pa->nr = 0;
2108                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2109                         pa->nr++;
2110                 pa->ticket = V_ticket_altqs_active;
2111                 PF_RULES_RUNLOCK();
2112                 break;
2113         }
2114
2115         case DIOCGETALTQ: {
2116                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2117                 struct pf_altq          *altq;
2118                 u_int32_t                nr;
2119
2120                 PF_RULES_RLOCK();
2121                 if (pa->ticket != V_ticket_altqs_active) {
2122                         PF_RULES_RUNLOCK();
2123                         error = EBUSY;
2124                         break;
2125                 }
2126                 nr = 0;
2127                 altq = TAILQ_FIRST(V_pf_altqs_active);
2128                 while ((altq != NULL) && (nr < pa->nr)) {
2129                         altq = TAILQ_NEXT(altq, entries);
2130                         nr++;
2131                 }
2132                 if (altq == NULL) {
2133                         PF_RULES_RUNLOCK();
2134                         error = EBUSY;
2135                         break;
2136                 }
2137                 bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2138                 PF_RULES_RUNLOCK();
2139                 break;
2140         }
2141
2142         case DIOCCHANGEALTQ:
2143                 /* CHANGEALTQ not supported yet! */
2144                 error = ENODEV;
2145                 break;
2146
2147         case DIOCGETQSTATS: {
2148                 struct pfioc_qstats     *pq = (struct pfioc_qstats *)addr;
2149                 struct pf_altq          *altq;
2150                 u_int32_t                nr;
2151                 int                      nbytes;
2152
2153                 PF_RULES_RLOCK();
2154                 if (pq->ticket != V_ticket_altqs_active) {
2155                         PF_RULES_RUNLOCK();
2156                         error = EBUSY;
2157                         break;
2158                 }
2159                 nbytes = pq->nbytes;
2160                 nr = 0;
2161                 altq = TAILQ_FIRST(V_pf_altqs_active);
2162                 while ((altq != NULL) && (nr < pq->nr)) {
2163                         altq = TAILQ_NEXT(altq, entries);
2164                         nr++;
2165                 }
2166                 if (altq == NULL) {
2167                         PF_RULES_RUNLOCK();
2168                         error = EBUSY;
2169                         break;
2170                 }
2171
2172                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2173                         PF_RULES_RUNLOCK();
2174                         error = ENXIO;
2175                         break;
2176                 }
2177                 PF_RULES_RUNLOCK();
2178                 error = altq_getqstats(altq, pq->buf, &nbytes);
2179                 if (error == 0) {
2180                         pq->scheduler = altq->scheduler;
2181                         pq->nbytes = nbytes;
2182                 }
2183                 break;
2184         }
2185 #endif /* ALTQ */
2186
2187         case DIOCBEGINADDRS: {
2188                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2189
2190                 PF_RULES_WLOCK();
2191                 pf_empty_pool(&V_pf_pabuf);
2192                 pp->ticket = ++V_ticket_pabuf;
2193                 PF_RULES_WUNLOCK();
2194                 break;
2195         }
2196
2197         case DIOCADDADDR: {
2198                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2199                 struct pf_pooladdr      *pa;
2200                 struct pfi_kif          *kif = NULL;
2201
2202 #ifndef INET
2203                 if (pp->af == AF_INET) {
2204                         error = EAFNOSUPPORT;
2205                         break;
2206                 }
2207 #endif /* INET */
2208 #ifndef INET6
2209                 if (pp->af == AF_INET6) {
2210                         error = EAFNOSUPPORT;
2211                         break;
2212                 }
2213 #endif /* INET6 */
2214                 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2215                     pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2216                     pp->addr.addr.type != PF_ADDR_TABLE) {
2217                         error = EINVAL;
2218                         break;
2219                 }
2220                 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2221                 bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2222                 if (pa->ifname[0])
2223                         kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2224                 PF_RULES_WLOCK();
2225                 if (pp->ticket != V_ticket_pabuf) {
2226                         PF_RULES_WUNLOCK();
2227                         if (pa->ifname[0])
2228                                 free(kif, PFI_MTYPE);
2229                         free(pa, M_PFRULE);
2230                         error = EBUSY;
2231                         break;
2232                 }
2233                 if (pa->ifname[0]) {
2234                         pa->kif = pfi_kif_attach(kif, pa->ifname);
2235                         pfi_kif_ref(pa->kif);
2236                 } else
2237                         pa->kif = NULL;
2238                 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2239                     pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2240                         if (pa->ifname[0])
2241                                 pfi_kif_unref(pa->kif);
2242                         PF_RULES_WUNLOCK();
2243                         free(pa, M_PFRULE);
2244                         break;
2245                 }
2246                 TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2247                 PF_RULES_WUNLOCK();
2248                 break;
2249         }
2250
2251         case DIOCGETADDRS: {
2252                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2253                 struct pf_pool          *pool;
2254                 struct pf_pooladdr      *pa;
2255
2256                 PF_RULES_RLOCK();
2257                 pp->nr = 0;
2258                 pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2259                     pp->r_num, 0, 1, 0);
2260                 if (pool == NULL) {
2261                         PF_RULES_RUNLOCK();
2262                         error = EBUSY;
2263                         break;
2264                 }
2265                 TAILQ_FOREACH(pa, &pool->list, entries)
2266                         pp->nr++;
2267                 PF_RULES_RUNLOCK();
2268                 break;
2269         }
2270
2271         case DIOCGETADDR: {
2272                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2273                 struct pf_pool          *pool;
2274                 struct pf_pooladdr      *pa;
2275                 u_int32_t                nr = 0;
2276
2277                 PF_RULES_RLOCK();
2278                 pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2279                     pp->r_num, 0, 1, 1);
2280                 if (pool == NULL) {
2281                         PF_RULES_RUNLOCK();
2282                         error = EBUSY;
2283                         break;
2284                 }
2285                 pa = TAILQ_FIRST(&pool->list);
2286                 while ((pa != NULL) && (nr < pp->nr)) {
2287                         pa = TAILQ_NEXT(pa, entries);
2288                         nr++;
2289                 }
2290                 if (pa == NULL) {
2291                         PF_RULES_RUNLOCK();
2292                         error = EBUSY;
2293                         break;
2294                 }
2295                 bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2296                 pf_addr_copyout(&pp->addr.addr);
2297                 PF_RULES_RUNLOCK();
2298                 break;
2299         }
2300
2301         case DIOCCHANGEADDR: {
2302                 struct pfioc_pooladdr   *pca = (struct pfioc_pooladdr *)addr;
2303                 struct pf_pool          *pool;
2304                 struct pf_pooladdr      *oldpa = NULL, *newpa = NULL;
2305                 struct pf_ruleset       *ruleset;
2306                 struct pfi_kif          *kif = NULL;
2307
2308                 if (pca->action < PF_CHANGE_ADD_HEAD ||
2309                     pca->action > PF_CHANGE_REMOVE) {
2310                         error = EINVAL;
2311                         break;
2312                 }
2313                 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2314                     pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2315                     pca->addr.addr.type != PF_ADDR_TABLE) {
2316                         error = EINVAL;
2317                         break;
2318                 }
2319
2320                 if (pca->action != PF_CHANGE_REMOVE) {
2321 #ifndef INET
2322                         if (pca->af == AF_INET) {
2323                                 error = EAFNOSUPPORT;
2324                                 break;
2325                         }
2326 #endif /* INET */
2327 #ifndef INET6
2328                         if (pca->af == AF_INET6) {
2329                                 error = EAFNOSUPPORT;
2330                                 break;
2331                         }
2332 #endif /* INET6 */
2333                         newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2334                         bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2335                         if (newpa->ifname[0])
2336                                 kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2337                         newpa->kif = NULL;
2338                 }
2339
2340 #define ERROUT(x)       { error = (x); goto DIOCCHANGEADDR_error; }
2341                 PF_RULES_WLOCK();
2342                 ruleset = pf_find_ruleset(pca->anchor);
2343                 if (ruleset == NULL)
2344                         ERROUT(EBUSY);
2345
2346                 pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2347                     pca->r_num, pca->r_last, 1, 1);
2348                 if (pool == NULL)
2349                         ERROUT(EBUSY);
2350
2351                 if (pca->action != PF_CHANGE_REMOVE) {
2352                         if (newpa->ifname[0]) {
2353                                 newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2354                                 pfi_kif_ref(newpa->kif);
2355                                 kif = NULL;
2356                         }
2357
2358                         switch (newpa->addr.type) {
2359                         case PF_ADDR_DYNIFTL:
2360                                 error = pfi_dynaddr_setup(&newpa->addr,
2361                                     pca->af);
2362                                 break;
2363                         case PF_ADDR_TABLE:
2364                                 newpa->addr.p.tbl = pfr_attach_table(ruleset,
2365                                     newpa->addr.v.tblname);
2366                                 if (newpa->addr.p.tbl == NULL)
2367                                         error = ENOMEM;
2368                                 break;
2369                         }
2370                         if (error)
2371                                 goto DIOCCHANGEADDR_error;
2372                 }
2373
2374                 switch (pca->action) {
2375                 case PF_CHANGE_ADD_HEAD:
2376                         oldpa = TAILQ_FIRST(&pool->list);
2377                         break;
2378                 case PF_CHANGE_ADD_TAIL:
2379                         oldpa = TAILQ_LAST(&pool->list, pf_palist);
2380                         break;
2381                 default:
2382                         oldpa = TAILQ_FIRST(&pool->list);
2383                         for (int i = 0; oldpa && i < pca->nr; i++)
2384                                 oldpa = TAILQ_NEXT(oldpa, entries);
2385
2386                         if (oldpa == NULL)
2387                                 ERROUT(EINVAL);
2388                 }
2389
2390                 if (pca->action == PF_CHANGE_REMOVE) {
2391                         TAILQ_REMOVE(&pool->list, oldpa, entries);
2392                         switch (oldpa->addr.type) {
2393                         case PF_ADDR_DYNIFTL:
2394                                 pfi_dynaddr_remove(oldpa->addr.p.dyn);
2395                                 break;
2396                         case PF_ADDR_TABLE:
2397                                 pfr_detach_table(oldpa->addr.p.tbl);
2398                                 break;
2399                         }
2400                         if (oldpa->kif)
2401                                 pfi_kif_unref(oldpa->kif);
2402                         free(oldpa, M_PFRULE);
2403                 } else {
2404                         if (oldpa == NULL)
2405                                 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2406                         else if (pca->action == PF_CHANGE_ADD_HEAD ||
2407                             pca->action == PF_CHANGE_ADD_BEFORE)
2408                                 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2409                         else
2410                                 TAILQ_INSERT_AFTER(&pool->list, oldpa,
2411                                     newpa, entries);
2412                 }
2413
2414                 pool->cur = TAILQ_FIRST(&pool->list);
2415                 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
2416                 PF_RULES_WUNLOCK();
2417                 break;
2418
2419 #undef ERROUT
2420 DIOCCHANGEADDR_error:
2421                 if (newpa->kif)
2422                         pfi_kif_unref(newpa->kif);
2423                 PF_RULES_WUNLOCK();
2424                 if (newpa != NULL)
2425                         free(newpa, M_PFRULE);
2426                 if (kif != NULL)
2427                         free(kif, PFI_MTYPE);
2428                 break;
2429         }
2430
2431         case DIOCGETRULESETS: {
2432                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
2433                 struct pf_ruleset       *ruleset;
2434                 struct pf_anchor        *anchor;
2435
2436                 PF_RULES_RLOCK();
2437                 pr->path[sizeof(pr->path) - 1] = 0;
2438                 if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2439                         PF_RULES_RUNLOCK();
2440                         error = ENOENT;
2441                         break;
2442                 }
2443                 pr->nr = 0;
2444                 if (ruleset->anchor == NULL) {
2445                         /* XXX kludge for pf_main_ruleset */
2446                         RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2447                                 if (anchor->parent == NULL)
2448                                         pr->nr++;
2449                 } else {
2450                         RB_FOREACH(anchor, pf_anchor_node,
2451                             &ruleset->anchor->children)
2452                                 pr->nr++;
2453                 }
2454                 PF_RULES_RUNLOCK();
2455                 break;
2456         }
2457
2458         case DIOCGETRULESET: {
2459                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
2460                 struct pf_ruleset       *ruleset;
2461                 struct pf_anchor        *anchor;
2462                 u_int32_t                nr = 0;
2463
2464                 PF_RULES_RLOCK();
2465                 pr->path[sizeof(pr->path) - 1] = 0;
2466                 if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2467                         PF_RULES_RUNLOCK();
2468                         error = ENOENT;
2469                         break;
2470                 }
2471                 pr->name[0] = 0;
2472                 if (ruleset->anchor == NULL) {
2473                         /* XXX kludge for pf_main_ruleset */
2474                         RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2475                                 if (anchor->parent == NULL && nr++ == pr->nr) {
2476                                         strlcpy(pr->name, anchor->name,
2477                                             sizeof(pr->name));
2478                                         break;
2479                                 }
2480                 } else {
2481                         RB_FOREACH(anchor, pf_anchor_node,
2482                             &ruleset->anchor->children)
2483                                 if (nr++ == pr->nr) {
2484                                         strlcpy(pr->name, anchor->name,
2485                                             sizeof(pr->name));
2486                                         break;
2487                                 }
2488                 }
2489                 if (!pr->name[0])
2490                         error = EBUSY;
2491                 PF_RULES_RUNLOCK();
2492                 break;
2493         }
2494
2495         case DIOCRCLRTABLES: {
2496                 struct pfioc_table *io = (struct pfioc_table *)addr;
2497
2498                 if (io->pfrio_esize != 0) {
2499                         error = ENODEV;
2500                         break;
2501                 }
2502                 PF_RULES_WLOCK();
2503                 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2504                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
2505                 PF_RULES_WUNLOCK();
2506                 break;
2507         }
2508
2509         case DIOCRADDTABLES: {
2510                 struct pfioc_table *io = (struct pfioc_table *)addr;
2511                 struct pfr_table *pfrts;
2512                 size_t totlen;
2513
2514                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2515                         error = ENODEV;
2516                         break;
2517                 }
2518
2519                 if (io->pfrio_size < 0 || io->pfrio_size > PF_TABLES_MAX_REQUEST) {
2520                         error = ENOMEM;
2521                         break;
2522                 }
2523
2524                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2525                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2526                     M_TEMP, M_WAITOK);
2527                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2528                 if (error) {
2529                         free(pfrts, M_TEMP);
2530                         break;
2531                 }
2532                 PF_RULES_WLOCK();
2533                 error = pfr_add_tables(pfrts, io->pfrio_size,
2534                     &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2535                 PF_RULES_WUNLOCK();
2536                 free(pfrts, M_TEMP);
2537                 break;
2538         }
2539
2540         case DIOCRDELTABLES: {
2541                 struct pfioc_table *io = (struct pfioc_table *)addr;
2542                 struct pfr_table *pfrts;
2543                 size_t totlen;
2544
2545                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2546                         error = ENODEV;
2547                         break;
2548                 }
2549
2550                 if (io->pfrio_size < 0 || io->pfrio_size > PF_TABLES_MAX_REQUEST) {
2551                         error = ENOMEM;
2552                         break;
2553                 }
2554
2555                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2556                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2557                     M_TEMP, M_WAITOK);
2558                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2559                 if (error) {
2560                         free(pfrts, M_TEMP);
2561                         break;
2562                 }
2563                 PF_RULES_WLOCK();
2564                 error = pfr_del_tables(pfrts, io->pfrio_size,
2565                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2566                 PF_RULES_WUNLOCK();
2567                 free(pfrts, M_TEMP);
2568                 break;
2569         }
2570
2571         case DIOCRGETTABLES: {
2572                 struct pfioc_table *io = (struct pfioc_table *)addr;
2573                 struct pfr_table *pfrts;
2574                 size_t totlen;
2575
2576                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2577                         error = ENODEV;
2578                         break;
2579                 }
2580                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2581                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2582                     M_TEMP, M_WAITOK);
2583                 if (! pfrts) {
2584                         error = ENOMEM;
2585                         break;
2586                 }
2587                 PF_RULES_RLOCK();
2588                 error = pfr_get_tables(&io->pfrio_table, pfrts,
2589                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2590                 PF_RULES_RUNLOCK();
2591                 if (error == 0)
2592                         error = copyout(pfrts, io->pfrio_buffer, totlen);
2593                 free(pfrts, M_TEMP);
2594                 break;
2595         }
2596
2597         case DIOCRGETTSTATS: {
2598                 struct pfioc_table *io = (struct pfioc_table *)addr;
2599                 struct pfr_tstats *pfrtstats;
2600                 size_t totlen;
2601
2602                 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2603                         error = ENODEV;
2604                         break;
2605                 }
2606                 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
2607                 pfrtstats = mallocarray(io->pfrio_size,
2608                     sizeof(struct pfr_tstats), M_TEMP, M_WAITOK);
2609                 if (! pfrtstats) {
2610                         error = ENOMEM;
2611                         break;
2612                 }
2613                 PF_RULES_WLOCK();
2614                 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
2615                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2616                 PF_RULES_WUNLOCK();
2617                 if (error == 0)
2618                         error = copyout(pfrtstats, io->pfrio_buffer, totlen);
2619                 free(pfrtstats, M_TEMP);
2620                 break;
2621         }
2622
2623         case DIOCRCLRTSTATS: {
2624                 struct pfioc_table *io = (struct pfioc_table *)addr;
2625                 struct pfr_table *pfrts;
2626                 size_t totlen;
2627
2628                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2629                         error = ENODEV;
2630                         break;
2631                 }
2632                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2633                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2634                     M_TEMP, M_WAITOK);
2635                 if (! pfrts) {
2636                         error = ENOMEM;
2637                         break;
2638                 }
2639                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2640                 if (error) {
2641                         free(pfrts, M_TEMP);
2642                         break;
2643                 }
2644                 PF_RULES_WLOCK();
2645                 error = pfr_clr_tstats(pfrts, io->pfrio_size,
2646                     &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2647                 PF_RULES_WUNLOCK();
2648                 free(pfrts, M_TEMP);
2649                 break;
2650         }
2651
2652         case DIOCRSETTFLAGS: {
2653                 struct pfioc_table *io = (struct pfioc_table *)addr;
2654                 struct pfr_table *pfrts;
2655                 size_t totlen;
2656
2657                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2658                         error = ENODEV;
2659                         break;
2660                 }
2661                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2662                 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2663                     M_TEMP, M_WAITOK);
2664                 if (! pfrts) {
2665                         error = ENOMEM;
2666                         break;
2667                 }
2668                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2669                 if (error) {
2670                         free(pfrts, M_TEMP);
2671                         break;
2672                 }
2673                 PF_RULES_WLOCK();
2674                 error = pfr_set_tflags(pfrts, io->pfrio_size,
2675                     io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2676                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2677                 PF_RULES_WUNLOCK();
2678                 free(pfrts, M_TEMP);
2679                 break;
2680         }
2681
2682         case DIOCRCLRADDRS: {
2683                 struct pfioc_table *io = (struct pfioc_table *)addr;
2684
2685                 if (io->pfrio_esize != 0) {
2686                         error = ENODEV;
2687                         break;
2688                 }
2689                 PF_RULES_WLOCK();
2690                 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2691                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
2692                 PF_RULES_WUNLOCK();
2693                 break;
2694         }
2695
2696         case DIOCRADDADDRS: {
2697                 struct pfioc_table *io = (struct pfioc_table *)addr;
2698                 struct pfr_addr *pfras;
2699                 size_t totlen;
2700
2701                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2702                         error = ENODEV;
2703                         break;
2704                 }
2705                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2706                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
2707                     M_TEMP, M_WAITOK);
2708                 if (! pfras) {
2709                         error = ENOMEM;
2710                         break;
2711                 }
2712                 error = copyin(io->pfrio_buffer, pfras, totlen);
2713                 if (error) {
2714                         free(pfras, M_TEMP);
2715                         break;
2716                 }
2717                 PF_RULES_WLOCK();
2718                 error = pfr_add_addrs(&io->pfrio_table, pfras,
2719                     io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2720                     PFR_FLAG_USERIOCTL);
2721                 PF_RULES_WUNLOCK();
2722                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2723                         error = copyout(pfras, io->pfrio_buffer, totlen);
2724                 free(pfras, M_TEMP);
2725                 break;
2726         }
2727
2728         case DIOCRDELADDRS: {
2729                 struct pfioc_table *io = (struct pfioc_table *)addr;
2730                 struct pfr_addr *pfras;
2731                 size_t totlen;
2732
2733                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2734                         error = ENODEV;
2735                         break;
2736                 }
2737                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2738                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
2739                     M_TEMP, M_WAITOK);
2740                 if (! pfras) {
2741                         error = ENOMEM;
2742                         break;
2743                 }
2744                 error = copyin(io->pfrio_buffer, pfras, totlen);
2745                 if (error) {
2746                         free(pfras, M_TEMP);
2747                         break;
2748                 }
2749                 PF_RULES_WLOCK();
2750                 error = pfr_del_addrs(&io->pfrio_table, pfras,
2751                     io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2752                     PFR_FLAG_USERIOCTL);
2753                 PF_RULES_WUNLOCK();
2754                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2755                         error = copyout(pfras, io->pfrio_buffer, totlen);
2756                 free(pfras, M_TEMP);
2757                 break;
2758         }
2759
2760         case DIOCRSETADDRS: {
2761                 struct pfioc_table *io = (struct pfioc_table *)addr;
2762                 struct pfr_addr *pfras;
2763                 size_t totlen, count;
2764
2765                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2766                         error = ENODEV;
2767                         break;
2768                 }
2769                 count = max(io->pfrio_size, io->pfrio_size2);
2770                 totlen = count * sizeof(struct pfr_addr);
2771                 pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
2772                     M_WAITOK);
2773                 if (! pfras) {
2774                         error = ENOMEM;
2775                         break;
2776                 }
2777                 error = copyin(io->pfrio_buffer, pfras, totlen);
2778                 if (error) {
2779                         free(pfras, M_TEMP);
2780                         break;
2781                 }
2782                 PF_RULES_WLOCK();
2783                 error = pfr_set_addrs(&io->pfrio_table, pfras,
2784                     io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2785                     &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2786                     PFR_FLAG_USERIOCTL, 0);
2787                 PF_RULES_WUNLOCK();
2788                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2789                         error = copyout(pfras, io->pfrio_buffer, totlen);
2790                 free(pfras, M_TEMP);
2791                 break;
2792         }
2793
2794         case DIOCRGETADDRS: {
2795                 struct pfioc_table *io = (struct pfioc_table *)addr;
2796                 struct pfr_addr *pfras;
2797                 size_t totlen;
2798
2799                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2800                         error = ENODEV;
2801                         break;
2802                 }
2803                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2804                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
2805                     M_TEMP, M_WAITOK);
2806                 if (! pfras) {
2807                         error = ENOMEM;
2808                         break;
2809                 }
2810                 PF_RULES_RLOCK();
2811                 error = pfr_get_addrs(&io->pfrio_table, pfras,
2812                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2813                 PF_RULES_RUNLOCK();
2814                 if (error == 0)
2815                         error = copyout(pfras, io->pfrio_buffer, totlen);
2816                 free(pfras, M_TEMP);
2817                 break;
2818         }
2819
2820         case DIOCRGETASTATS: {
2821                 struct pfioc_table *io = (struct pfioc_table *)addr;
2822                 struct pfr_astats *pfrastats;
2823                 size_t totlen;
2824
2825                 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2826                         error = ENODEV;
2827                         break;
2828                 }
2829                 totlen = io->pfrio_size * sizeof(struct pfr_astats);
2830                 pfrastats = mallocarray(io->pfrio_size,
2831                     sizeof(struct pfr_astats), M_TEMP, M_WAITOK);
2832                 if (! pfrastats) {
2833                         error = ENOMEM;
2834                         break;
2835                 }
2836                 PF_RULES_RLOCK();
2837                 error = pfr_get_astats(&io->pfrio_table, pfrastats,
2838                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2839                 PF_RULES_RUNLOCK();
2840                 if (error == 0)
2841                         error = copyout(pfrastats, io->pfrio_buffer, totlen);
2842                 free(pfrastats, M_TEMP);
2843                 break;
2844         }
2845
2846         case DIOCRCLRASTATS: {
2847                 struct pfioc_table *io = (struct pfioc_table *)addr;
2848                 struct pfr_addr *pfras;
2849                 size_t totlen;
2850
2851                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2852                         error = ENODEV;
2853                         break;
2854                 }
2855                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2856                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
2857                     M_TEMP, M_WAITOK);
2858                 if (! pfras) {
2859                         error = ENOMEM;
2860                         break;
2861                 }
2862                 error = copyin(io->pfrio_buffer, pfras, totlen);
2863                 if (error) {
2864                         free(pfras, M_TEMP);
2865                         break;
2866                 }
2867                 PF_RULES_WLOCK();
2868                 error = pfr_clr_astats(&io->pfrio_table, pfras,
2869                     io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2870                     PFR_FLAG_USERIOCTL);
2871                 PF_RULES_WUNLOCK();
2872                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2873                         error = copyout(pfras, io->pfrio_buffer, totlen);
2874                 free(pfras, M_TEMP);
2875                 break;
2876         }
2877
2878         case DIOCRTSTADDRS: {
2879                 struct pfioc_table *io = (struct pfioc_table *)addr;
2880                 struct pfr_addr *pfras;
2881                 size_t totlen;
2882
2883                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2884                         error = ENODEV;
2885                         break;
2886                 }
2887                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2888                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
2889                     M_TEMP, M_WAITOK);
2890                 if (! pfras) {
2891                         error = ENOMEM;
2892                         break;
2893                 }
2894                 error = copyin(io->pfrio_buffer, pfras, totlen);
2895                 if (error) {
2896                         free(pfras, M_TEMP);
2897                         break;
2898                 }
2899                 PF_RULES_RLOCK();
2900                 error = pfr_tst_addrs(&io->pfrio_table, pfras,
2901                     io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2902                     PFR_FLAG_USERIOCTL);
2903                 PF_RULES_RUNLOCK();
2904                 if (error == 0)
2905                         error = copyout(pfras, io->pfrio_buffer, totlen);
2906                 free(pfras, M_TEMP);
2907                 break;
2908         }
2909
2910         case DIOCRINADEFINE: {
2911                 struct pfioc_table *io = (struct pfioc_table *)addr;
2912                 struct pfr_addr *pfras;
2913                 size_t totlen;
2914
2915                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2916                         error = ENODEV;
2917                         break;
2918                 }
2919                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2920                 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
2921                     M_TEMP, M_WAITOK);
2922                 if (! pfras) {
2923                         error = ENOMEM;
2924                         break;
2925                 }
2926                 error = copyin(io->pfrio_buffer, pfras, totlen);
2927                 if (error) {
2928                         free(pfras, M_TEMP);
2929                         break;
2930                 }
2931                 PF_RULES_WLOCK();
2932                 error = pfr_ina_define(&io->pfrio_table, pfras,
2933                     io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2934                     io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2935                 PF_RULES_WUNLOCK();
2936                 free(pfras, M_TEMP);
2937                 break;
2938         }
2939
2940         case DIOCOSFPADD: {
2941                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2942                 PF_RULES_WLOCK();
2943                 error = pf_osfp_add(io);
2944                 PF_RULES_WUNLOCK();
2945                 break;
2946         }
2947
2948         case DIOCOSFPGET: {
2949                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2950                 PF_RULES_RLOCK();
2951                 error = pf_osfp_get(io);
2952                 PF_RULES_RUNLOCK();
2953                 break;
2954         }
2955
2956         case DIOCXBEGIN: {
2957                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2958                 struct pfioc_trans_e    *ioes, *ioe;
2959                 size_t                   totlen;
2960                 int                      i;
2961
2962                 if (io->esize != sizeof(*ioe)) {
2963                         error = ENODEV;
2964                         break;
2965                 }
2966                 totlen = sizeof(struct pfioc_trans_e) * io->size;
2967                 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
2968                     M_TEMP, M_WAITOK);
2969                 if (! ioes) {
2970                         error = ENOMEM;
2971                         break;
2972                 }
2973                 error = copyin(io->array, ioes, totlen);
2974                 if (error) {
2975                         free(ioes, M_TEMP);
2976                         break;
2977                 }
2978                 PF_RULES_WLOCK();
2979                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2980                         switch (ioe->rs_num) {
2981 #ifdef ALTQ
2982                         case PF_RULESET_ALTQ:
2983                                 if (ioe->anchor[0]) {
2984                                         PF_RULES_WUNLOCK();
2985                                         free(ioes, M_TEMP);
2986                                         error = EINVAL;
2987                                         goto fail;
2988                                 }
2989                                 if ((error = pf_begin_altq(&ioe->ticket))) {
2990                                         PF_RULES_WUNLOCK();
2991                                         free(ioes, M_TEMP);
2992                                         goto fail;
2993                                 }
2994                                 break;
2995 #endif /* ALTQ */
2996                         case PF_RULESET_TABLE:
2997                             {
2998                                 struct pfr_table table;
2999
3000                                 bzero(&table, sizeof(table));
3001                                 strlcpy(table.pfrt_anchor, ioe->anchor,
3002                                     sizeof(table.pfrt_anchor));
3003                                 if ((error = pfr_ina_begin(&table,
3004                                     &ioe->ticket, NULL, 0))) {
3005                                         PF_RULES_WUNLOCK();
3006                                         free(ioes, M_TEMP);
3007                                         goto fail;
3008                                 }
3009                                 break;
3010                             }
3011                         default:
3012                                 if ((error = pf_begin_rules(&ioe->ticket,
3013                                     ioe->rs_num, ioe->anchor))) {
3014                                         PF_RULES_WUNLOCK();
3015                                         free(ioes, M_TEMP);
3016                                         goto fail;
3017                                 }
3018                                 break;
3019                         }
3020                 }
3021                 PF_RULES_WUNLOCK();
3022                 error = copyout(ioes, io->array, totlen);
3023                 free(ioes, M_TEMP);
3024                 break;
3025         }
3026
3027         case DIOCXROLLBACK: {
3028                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
3029                 struct pfioc_trans_e    *ioe, *ioes;
3030                 size_t                   totlen;
3031                 int                      i;
3032
3033                 if (io->esize != sizeof(*ioe)) {
3034                         error = ENODEV;
3035                         break;
3036                 }
3037                 totlen = sizeof(struct pfioc_trans_e) * io->size;
3038                 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3039                     M_TEMP, M_WAITOK);
3040                 if (! ioes) {
3041                         error = ENOMEM;
3042                         break;
3043                 }
3044                 error = copyin(io->array, ioes, totlen);
3045                 if (error) {
3046                         free(ioes, M_TEMP);
3047                         break;
3048                 }
3049                 PF_RULES_WLOCK();
3050                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3051                         switch (ioe->rs_num) {
3052 #ifdef ALTQ
3053                         case PF_RULESET_ALTQ:
3054                                 if (ioe->anchor[0]) {
3055                                         PF_RULES_WUNLOCK();
3056                                         free(ioes, M_TEMP);
3057                                         error = EINVAL;
3058                                         goto fail;
3059                                 }
3060                                 if ((error = pf_rollback_altq(ioe->ticket))) {
3061                                         PF_RULES_WUNLOCK();
3062                                         free(ioes, M_TEMP);
3063                                         goto fail; /* really bad */
3064                                 }
3065                                 break;
3066 #endif /* ALTQ */
3067                         case PF_RULESET_TABLE:
3068                             {
3069                                 struct pfr_table table;
3070
3071                                 bzero(&table, sizeof(table));
3072                                 strlcpy(table.pfrt_anchor, ioe->anchor,
3073                                     sizeof(table.pfrt_anchor));
3074                                 if ((error = pfr_ina_rollback(&table,
3075                                     ioe->ticket, NULL, 0))) {
3076                                         PF_RULES_WUNLOCK();
3077                                         free(ioes, M_TEMP);
3078                                         goto fail; /* really bad */
3079                                 }
3080                                 break;
3081                             }
3082                         default:
3083                                 if ((error = pf_rollback_rules(ioe->ticket,
3084                                     ioe->rs_num, ioe->anchor))) {
3085                                         PF_RULES_WUNLOCK();
3086                                         free(ioes, M_TEMP);
3087                                         goto fail; /* really bad */
3088                                 }
3089                                 break;
3090                         }
3091                 }
3092                 PF_RULES_WUNLOCK();
3093                 free(ioes, M_TEMP);
3094                 break;
3095         }
3096
3097         case DIOCXCOMMIT: {
3098                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
3099                 struct pfioc_trans_e    *ioe, *ioes;
3100                 struct pf_ruleset       *rs;
3101                 size_t                   totlen;
3102                 int                      i;
3103
3104                 if (io->esize != sizeof(*ioe)) {
3105                         error = ENODEV;
3106                         break;
3107                 }
3108
3109                 if (io->size < 0 ||
3110                     WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3111                         error = EINVAL;
3112                         break;
3113                 }
3114
3115                 totlen = sizeof(struct pfioc_trans_e) * io->size;
3116                 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3117                     M_TEMP, M_NOWAIT);
3118                 if (ioes == NULL) {
3119                         error = ENOMEM;
3120                         break;
3121                 }
3122                 error = copyin(io->array, ioes, totlen);
3123                 if (error) {
3124                         free(ioes, M_TEMP);
3125                         break;
3126                 }
3127                 PF_RULES_WLOCK();
3128                 /* First makes sure everything will succeed. */
3129                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3130                         switch (ioe->rs_num) {
3131 #ifdef ALTQ
3132                         case PF_RULESET_ALTQ:
3133                                 if (ioe->anchor[0]) {
3134                                         PF_RULES_WUNLOCK();
3135                                         free(ioes, M_TEMP);
3136                                         error = EINVAL;
3137                                         goto fail;
3138                                 }
3139                                 if (!V_altqs_inactive_open || ioe->ticket !=
3140                                     V_ticket_altqs_inactive) {
3141                                         PF_RULES_WUNLOCK();
3142                                         free(ioes, M_TEMP);
3143                                         error = EBUSY;
3144                                         goto fail;
3145                                 }
3146                                 break;
3147 #endif /* ALTQ */
3148                         case PF_RULESET_TABLE:
3149                                 rs = pf_find_ruleset(ioe->anchor);
3150                                 if (rs == NULL || !rs->topen || ioe->ticket !=
3151                                     rs->tticket) {
3152                                         PF_RULES_WUNLOCK();
3153                                         free(ioes, M_TEMP);
3154                                         error = EBUSY;
3155                                         goto fail;
3156                                 }
3157                                 break;
3158                         default:
3159                                 if (ioe->rs_num < 0 || ioe->rs_num >=
3160                                     PF_RULESET_MAX) {
3161                                         PF_RULES_WUNLOCK();
3162                                         free(ioes, M_TEMP);
3163                                         error = EINVAL;
3164                                         goto fail;
3165                                 }
3166                                 rs = pf_find_ruleset(ioe->anchor);
3167                                 if (rs == NULL ||
3168                                     !rs->rules[ioe->rs_num].inactive.open ||
3169                                     rs->rules[ioe->rs_num].inactive.ticket !=
3170                                     ioe->ticket) {
3171                                         PF_RULES_WUNLOCK();
3172                                         free(ioes, M_TEMP);
3173                                         error = EBUSY;
3174                                         goto fail;
3175                                 }
3176                                 break;
3177                         }
3178                 }
3179                 /* Now do the commit - no errors should happen here. */
3180                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3181                         switch (ioe->rs_num) {
3182 #ifdef ALTQ
3183                         case PF_RULESET_ALTQ:
3184                                 if ((error = pf_commit_altq(ioe->ticket))) {
3185                                         PF_RULES_WUNLOCK();
3186                                         free(ioes, M_TEMP);
3187                                         goto fail; /* really bad */
3188                                 }
3189                                 break;
3190 #endif /* ALTQ */
3191                         case PF_RULESET_TABLE:
3192                             {
3193                                 struct pfr_table table;
3194
3195                                 bzero(&table, sizeof(table));
3196                                 strlcpy(table.pfrt_anchor, ioe->anchor,
3197                                     sizeof(table.pfrt_anchor));
3198                                 if ((error = pfr_ina_commit(&table,
3199                                     ioe->ticket, NULL, NULL, 0))) {
3200                                         PF_RULES_WUNLOCK();
3201                                         free(ioes, M_TEMP);
3202                                         goto fail; /* really bad */
3203                                 }
3204                                 break;
3205                             }
3206                         default:
3207                                 if ((error = pf_commit_rules(ioe->ticket,
3208                                     ioe->rs_num, ioe->anchor))) {
3209                                         PF_RULES_WUNLOCK();
3210                                         free(ioes, M_TEMP);
3211                                         goto fail; /* really bad */
3212                                 }
3213                                 break;
3214                         }
3215                 }
3216                 PF_RULES_WUNLOCK();
3217                 free(ioes, M_TEMP);
3218                 break;
3219         }
3220
3221         case DIOCGETSRCNODES: {
3222                 struct pfioc_src_nodes  *psn = (struct pfioc_src_nodes *)addr;
3223                 struct pf_srchash       *sh;
3224                 struct pf_src_node      *n, *p, *pstore;
3225                 uint32_t                 i, nr = 0;
3226
3227                 if (psn->psn_len == 0) {
3228                         for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3229                             i++, sh++) {
3230                                 PF_HASHROW_LOCK(sh);
3231                                 LIST_FOREACH(n, &sh->nodes, entry)
3232                                         nr++;
3233                                 PF_HASHROW_UNLOCK(sh);
3234                         }
3235                         psn->psn_len = sizeof(struct pf_src_node) * nr;
3236                         break;
3237                 }
3238
3239                 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK);
3240                 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3241                     i++, sh++) {
3242                     PF_HASHROW_LOCK(sh);
3243                     LIST_FOREACH(n, &sh->nodes, entry) {
3244                         int     secs = time_uptime, diff;
3245
3246                         if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3247                                 break;
3248
3249                         bcopy(n, p, sizeof(struct pf_src_node));
3250                         if (n->rule.ptr != NULL)
3251                                 p->rule.nr = n->rule.ptr->nr;
3252                         p->creation = secs - p->creation;
3253                         if (p->expire > secs)
3254                                 p->expire -= secs;
3255                         else
3256                                 p->expire = 0;
3257
3258                         /* Adjust the connection rate estimate. */
3259                         diff = secs - n->conn_rate.last;
3260                         if (diff >= n->conn_rate.seconds)
3261                                 p->conn_rate.count = 0;
3262                         else
3263                                 p->conn_rate.count -=
3264                                     n->conn_rate.count * diff /
3265                                     n->conn_rate.seconds;
3266                         p++;
3267                         nr++;
3268                     }
3269                     PF_HASHROW_UNLOCK(sh);
3270                 }
3271                 error = copyout(pstore, psn->psn_src_nodes,
3272                     sizeof(struct pf_src_node) * nr);
3273                 if (error) {
3274                         free(pstore, M_TEMP);
3275                         break;
3276                 }
3277                 psn->psn_len = sizeof(struct pf_src_node) * nr;
3278                 free(pstore, M_TEMP);
3279                 break;
3280         }
3281
3282         case DIOCCLRSRCNODES: {
3283
3284                 pf_clear_srcnodes(NULL);
3285                 pf_purge_expired_src_nodes();
3286                 break;
3287         }
3288
3289         case DIOCKILLSRCNODES:
3290                 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
3291                 break;
3292
3293         case DIOCSETHOSTID: {
3294                 u_int32_t       *hostid = (u_int32_t *)addr;
3295
3296                 PF_RULES_WLOCK();
3297                 if (*hostid == 0)
3298                         V_pf_status.hostid = arc4random();
3299                 else
3300                         V_pf_status.hostid = *hostid;
3301                 PF_RULES_WUNLOCK();
3302                 break;
3303         }
3304
3305         case DIOCOSFPFLUSH:
3306                 PF_RULES_WLOCK();
3307                 pf_osfp_flush();
3308                 PF_RULES_WUNLOCK();
3309                 break;
3310
3311         case DIOCIGETIFACES: {
3312                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3313                 struct pfi_kif *ifstore;
3314                 size_t bufsiz;
3315
3316                 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3317                         error = ENODEV;
3318                         break;
3319                 }
3320
3321                 if (io->pfiio_size < 0 ||
3322                     WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
3323                         error = EINVAL;
3324                         break;
3325                 }
3326
3327                 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3328                 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
3329                     M_TEMP, M_NOWAIT);
3330                 if (ifstore == NULL) {
3331                         error = ENOMEM;
3332                         break;
3333                 }
3334
3335                 PF_RULES_RLOCK();
3336                 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3337                 PF_RULES_RUNLOCK();
3338                 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3339                 free(ifstore, M_TEMP);
3340                 break;
3341         }
3342
3343         case DIOCSETIFFLAG: {
3344                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3345
3346                 PF_RULES_WLOCK();
3347                 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3348                 PF_RULES_WUNLOCK();
3349                 break;
3350         }
3351
3352         case DIOCCLRIFFLAG: {
3353                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3354
3355                 PF_RULES_WLOCK();
3356                 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3357                 PF_RULES_WUNLOCK();
3358                 break;
3359         }
3360
3361         default:
3362                 error = ENODEV;
3363                 break;
3364         }
3365 fail:
3366         if (sx_xlocked(&pf_ioctl_lock))
3367                 sx_xunlock(&pf_ioctl_lock);
3368         CURVNET_RESTORE();
3369
3370         return (error);
3371 }
3372
3373 void
3374 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3375 {
3376         bzero(sp, sizeof(struct pfsync_state));
3377
3378         /* copy from state key */
3379         sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3380         sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3381         sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3382         sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3383         sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3384         sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3385         sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3386         sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3387         sp->proto = st->key[PF_SK_WIRE]->proto;
3388         sp->af = st->key[PF_SK_WIRE]->af;
3389
3390         /* copy from state */
3391         strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3392         bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3393         sp->creation = htonl(time_uptime - st->creation);
3394         sp->expire = pf_state_expires(st);
3395         if (sp->expire <= time_uptime)
3396                 sp->expire = htonl(0);
3397         else
3398                 sp->expire = htonl(sp->expire - time_uptime);
3399
3400         sp->direction = st->direction;
3401         sp->log = st->log;
3402         sp->timeout = st->timeout;
3403         sp->state_flags = st->state_flags;
3404         if (st->src_node)
3405                 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3406         if (st->nat_src_node)
3407                 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3408
3409         sp->id = st->id;
3410         sp->creatorid = st->creatorid;
3411         pf_state_peer_hton(&st->src, &sp->src);
3412         pf_state_peer_hton(&st->dst, &sp->dst);
3413
3414         if (st->rule.ptr == NULL)
3415                 sp->rule = htonl(-1);
3416         else
3417                 sp->rule = htonl(st->rule.ptr->nr);
3418         if (st->anchor.ptr == NULL)
3419                 sp->anchor = htonl(-1);
3420         else
3421                 sp->anchor = htonl(st->anchor.ptr->nr);
3422         if (st->nat_rule.ptr == NULL)
3423                 sp->nat_rule = htonl(-1);
3424         else
3425                 sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3426
3427         pf_state_counter_hton(st->packets[0], sp->packets[0]);
3428         pf_state_counter_hton(st->packets[1], sp->packets[1]);
3429         pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3430         pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3431
3432 }
3433
3434 static void
3435 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3436 {
3437         struct pfr_ktable *kt;
3438
3439         KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3440
3441         kt = aw->p.tbl;
3442         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3443                 kt = kt->pfrkt_root;
3444         aw->p.tbl = NULL;
3445         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3446                 kt->pfrkt_cnt : -1;
3447 }
3448
3449 /*
3450  * XXX - Check for version missmatch!!!
3451  */
3452 static void
3453 pf_clear_states(void)
3454 {
3455         struct pf_state *s;
3456         u_int i;
3457
3458         for (i = 0; i <= pf_hashmask; i++) {
3459                 struct pf_idhash *ih = &V_pf_idhash[i];
3460 relock:
3461                 PF_HASHROW_LOCK(ih);
3462                 LIST_FOREACH(s, &ih->states, entry) {
3463                         s->timeout = PFTM_PURGE;
3464                         /* Don't send out individual delete messages. */
3465                         s->sync_state = PFSTATE_NOSYNC;
3466                         pf_unlink_state(s, PF_ENTER_LOCKED);
3467                         goto relock;
3468                 }
3469                 PF_HASHROW_UNLOCK(ih);
3470         }
3471 }
3472
3473 static int
3474 pf_clear_tables(void)
3475 {
3476         struct pfioc_table io;
3477         int error;
3478
3479         bzero(&io, sizeof(io));
3480
3481         error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3482             io.pfrio_flags);
3483
3484         return (error);
3485 }
3486
3487 static void
3488 pf_clear_srcnodes(struct pf_src_node *n)
3489 {
3490         struct pf_state *s;
3491         int i;
3492
3493         for (i = 0; i <= pf_hashmask; i++) {
3494                 struct pf_idhash *ih = &V_pf_idhash[i];
3495
3496                 PF_HASHROW_LOCK(ih);
3497                 LIST_FOREACH(s, &ih->states, entry) {
3498                         if (n == NULL || n == s->src_node)
3499                                 s->src_node = NULL;
3500                         if (n == NULL || n == s->nat_src_node)
3501                                 s->nat_src_node = NULL;
3502                 }
3503                 PF_HASHROW_UNLOCK(ih);
3504         }
3505
3506         if (n == NULL) {
3507                 struct pf_srchash *sh;
3508
3509                 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3510                     i++, sh++) {
3511                         PF_HASHROW_LOCK(sh);
3512                         LIST_FOREACH(n, &sh->nodes, entry) {
3513                                 n->expire = 1;
3514                                 n->states = 0;
3515                         }
3516                         PF_HASHROW_UNLOCK(sh);
3517                 }
3518         } else {
3519                 /* XXX: hash slot should already be locked here. */
3520                 n->expire = 1;
3521                 n->states = 0;
3522         }
3523 }
3524
3525 static void
3526 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
3527 {
3528         struct pf_src_node_list  kill;
3529
3530         LIST_INIT(&kill);
3531         for (int i = 0; i <= pf_srchashmask; i++) {
3532                 struct pf_srchash *sh = &V_pf_srchash[i];
3533                 struct pf_src_node *sn, *tmp;
3534
3535                 PF_HASHROW_LOCK(sh);
3536                 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
3537                         if (PF_MATCHA(psnk->psnk_src.neg,
3538                               &psnk->psnk_src.addr.v.a.addr,
3539                               &psnk->psnk_src.addr.v.a.mask,
3540                               &sn->addr, sn->af) &&
3541                             PF_MATCHA(psnk->psnk_dst.neg,
3542                               &psnk->psnk_dst.addr.v.a.addr,
3543                               &psnk->psnk_dst.addr.v.a.mask,
3544                               &sn->raddr, sn->af)) {
3545                                 pf_unlink_src_node(sn);
3546                                 LIST_INSERT_HEAD(&kill, sn, entry);
3547                                 sn->expire = 1;
3548                         }
3549                 PF_HASHROW_UNLOCK(sh);
3550         }
3551
3552         for (int i = 0; i <= pf_hashmask; i++) {
3553                 struct pf_idhash *ih = &V_pf_idhash[i];
3554                 struct pf_state *s;
3555
3556                 PF_HASHROW_LOCK(ih);
3557                 LIST_FOREACH(s, &ih->states, entry) {
3558                         if (s->src_node && s->src_node->expire == 1)
3559                                 s->src_node = NULL;
3560                         if (s->nat_src_node && s->nat_src_node->expire == 1)
3561                                 s->nat_src_node = NULL;
3562                 }
3563                 PF_HASHROW_UNLOCK(ih);
3564         }
3565
3566         psnk->psnk_killed = pf_free_src_nodes(&kill);
3567 }
3568
3569 /*
3570  * XXX - Check for version missmatch!!!
3571  */
3572
3573 /*
3574  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3575  */
3576 static int
3577 shutdown_pf(void)
3578 {
3579         int error = 0;
3580         u_int32_t t[5];
3581         char nn = '\0';
3582
3583         V_pf_status.running = 0;
3584
3585         counter_u64_free(V_pf_default_rule.states_cur);
3586         counter_u64_free(V_pf_default_rule.states_tot);
3587         counter_u64_free(V_pf_default_rule.src_nodes);
3588
3589         for (int i = 0; i < PFRES_MAX; i++)
3590                 counter_u64_free(V_pf_status.counters[i]);
3591         for (int i = 0; i < LCNT_MAX; i++)
3592                 counter_u64_free(V_pf_status.lcounters[i]);
3593         for (int i = 0; i < FCNT_MAX; i++)
3594                 counter_u64_free(V_pf_status.fcounters[i]);
3595         for (int i = 0; i < SCNT_MAX; i++)
3596                 counter_u64_free(V_pf_status.scounters[i]);
3597
3598         do {
3599                 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3600                     != 0) {
3601                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3602                         break;
3603                 }
3604                 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3605                     != 0) {
3606                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3607                         break;          /* XXX: rollback? */
3608                 }
3609                 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3610                     != 0) {
3611                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3612                         break;          /* XXX: rollback? */
3613                 }
3614                 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3615                     != 0) {
3616                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3617                         break;          /* XXX: rollback? */
3618                 }
3619                 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3620                     != 0) {
3621                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3622                         break;          /* XXX: rollback? */
3623                 }
3624
3625                 /* XXX: these should always succeed here */
3626                 pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3627                 pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3628                 pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3629                 pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3630                 pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3631
3632                 if ((error = pf_clear_tables()) != 0)
3633                         break;
3634
3635 #ifdef ALTQ
3636                 if ((error = pf_begin_altq(&t[0])) != 0) {
3637                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3638                         break;
3639                 }
3640                 pf_commit_altq(t[0]);
3641 #endif
3642
3643                 pf_clear_states();
3644
3645                 pf_clear_srcnodes(NULL);
3646
3647                 /* status does not use malloced mem so no need to cleanup */
3648                 /* fingerprints and interfaces have thier own cleanup code */
3649         } while(0);
3650
3651         return (error);
3652 }
3653
3654 #ifdef INET
3655 static int
3656 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3657     struct inpcb *inp)
3658 {
3659         int chk;
3660
3661         chk = pf_test(PF_IN, ifp, m, inp);
3662         if (chk && *m) {
3663                 m_freem(*m);
3664                 *m = NULL;
3665         }
3666
3667         if (chk != PF_PASS)
3668                 return (EACCES);
3669         return (0);
3670 }
3671
3672 static int
3673 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3674     struct inpcb *inp)
3675 {
3676         int chk;
3677
3678         chk = pf_test(PF_OUT, ifp, m, inp);
3679         if (chk && *m) {
3680                 m_freem(*m);
3681                 *m = NULL;
3682         }
3683
3684         if (chk != PF_PASS)
3685                 return (EACCES);
3686         return (0);
3687 }
3688 #endif
3689
3690 #ifdef INET6
3691 static int
3692 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3693     struct inpcb *inp)
3694 {
3695         int chk;
3696
3697         /*
3698          * In case of loopback traffic IPv6 uses the real interface in
3699          * order to support scoped addresses. In order to support stateful
3700          * filtering we have change this to lo0 as it is the case in IPv4.
3701          */
3702         CURVNET_SET(ifp->if_vnet);
3703         chk = pf_test6(PF_IN, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
3704         CURVNET_RESTORE();
3705         if (chk && *m) {
3706                 m_freem(*m);
3707                 *m = NULL;
3708         }
3709         if (chk != PF_PASS)
3710                 return (EACCES);
3711         return (0);
3712 }
3713
3714 static int
3715 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3716     struct inpcb *inp)
3717 {
3718         int chk;
3719
3720         CURVNET_SET(ifp->if_vnet);
3721         chk = pf_test6(PF_OUT, ifp, m, inp);
3722         CURVNET_RESTORE();
3723         if (chk && *m) {
3724                 m_freem(*m);
3725                 *m = NULL;
3726         }
3727         if (chk != PF_PASS)
3728                 return (EACCES);
3729         return (0);
3730 }
3731 #endif /* INET6 */
3732
3733 static int
3734 hook_pf(void)
3735 {
3736 #ifdef INET
3737         struct pfil_head *pfh_inet;
3738 #endif
3739 #ifdef INET6
3740         struct pfil_head *pfh_inet6;
3741 #endif
3742
3743         if (V_pf_pfil_hooked)
3744                 return (0);
3745
3746 #ifdef INET
3747         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3748         if (pfh_inet == NULL)
3749                 return (ESRCH); /* XXX */
3750         pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
3751         pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
3752 #endif
3753 #ifdef INET6
3754         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3755         if (pfh_inet6 == NULL) {
3756 #ifdef INET
3757                 pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3758                     pfh_inet);
3759                 pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3760                     pfh_inet);
3761 #endif
3762                 return (ESRCH); /* XXX */
3763         }
3764         pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
3765         pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
3766 #endif
3767
3768         V_pf_pfil_hooked = 1;
3769         return (0);
3770 }
3771
3772 static int
3773 dehook_pf(void)
3774 {
3775 #ifdef INET
3776         struct pfil_head *pfh_inet;
3777 #endif
3778 #ifdef INET6
3779         struct pfil_head *pfh_inet6;
3780 #endif
3781
3782         if (V_pf_pfil_hooked == 0)
3783                 return (0);
3784
3785 #ifdef INET
3786         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3787         if (pfh_inet == NULL)
3788                 return (ESRCH); /* XXX */
3789         pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3790             pfh_inet);
3791         pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3792             pfh_inet);
3793 #endif
3794 #ifdef INET6
3795         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3796         if (pfh_inet6 == NULL)
3797                 return (ESRCH); /* XXX */
3798         pfil_remove_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
3799             pfh_inet6);
3800         pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
3801             pfh_inet6);
3802 #endif
3803
3804         V_pf_pfil_hooked = 0;
3805         return (0);
3806 }
3807
3808 static int
3809 pf_load(void)
3810 {
3811         int error;
3812
3813         VNET_ITERATOR_DECL(vnet_iter);
3814
3815         VNET_LIST_RLOCK();
3816         VNET_FOREACH(vnet_iter) {
3817                 CURVNET_SET(vnet_iter);
3818                 V_pf_pfil_hooked = 0;
3819                 V_pf_end_threads = 0;
3820                 TAILQ_INIT(&V_pf_tags);
3821                 TAILQ_INIT(&V_pf_qids);
3822                 CURVNET_RESTORE();
3823         }
3824         VNET_LIST_RUNLOCK();
3825
3826         rw_init(&pf_rules_lock, "pf rulesets");
3827         sx_init(&pf_ioctl_lock, "pf ioctl");
3828
3829         pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
3830         if ((error = pfattach()) != 0)
3831                 return (error);
3832
3833         return (0);
3834 }
3835
3836 static int
3837 pf_unload(void)
3838 {
3839         int error = 0;
3840
3841         V_pf_status.running = 0;
3842         swi_remove(V_pf_swi_cookie);
3843         error = dehook_pf();
3844         if (error) {
3845                 /*
3846                  * Should not happen!
3847                  * XXX Due to error code ESRCH, kldunload will show
3848                  * a message like 'No such process'.
3849                  */
3850                 printf("%s : pfil unregisteration fail\n", __FUNCTION__);
3851                 return error;
3852         }
3853         PF_RULES_WLOCK();
3854         shutdown_pf();
3855         V_pf_end_threads = 1;
3856         while (V_pf_end_threads < 2) {
3857                 wakeup_one(pf_purge_thread);
3858                 rw_sleep(pf_purge_thread, &pf_rules_lock, 0, "pftmo", 0);
3859         }
3860         PF_RULES_WUNLOCK();
3861         pf_normalize_cleanup();
3862         pfi_cleanup();
3863         pfr_cleanup();
3864         pf_osfp_flush();
3865         pf_cleanup();
3866         if (IS_DEFAULT_VNET(curvnet))
3867                 pf_mtag_cleanup();
3868         destroy_dev(pf_dev);
3869         rw_destroy(&pf_rules_lock);
3870         sx_destroy(&pf_ioctl_lock);
3871
3872         return (error);
3873 }
3874
3875 static int
3876 pf_modevent(module_t mod, int type, void *data)
3877 {
3878         int error = 0;
3879
3880         switch(type) {
3881         case MOD_LOAD:
3882                 error = pf_load();
3883                 break;
3884         case MOD_QUIESCE:
3885                 /*
3886                  * Module should not be unloaded due to race conditions.
3887                  */
3888                 error = EBUSY;
3889                 break;
3890         case MOD_UNLOAD:
3891                 error = pf_unload();
3892                 break;
3893         default:
3894                 error = EINVAL;
3895                 break;
3896         }
3897
3898         return (error);
3899 }
3900
3901 static moduledata_t pf_mod = {
3902         "pf",
3903         pf_modevent,
3904         0
3905 };
3906
3907 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3908 MODULE_VERSION(pf, PF_MODVER);