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/lock.h>
  57 #include <sys/mbuf.h>
  58 #include <sys/module.h>
  59 #include <sys/proc.h>
  60 #include <sys/rwlock.h>
  61 #include <sys/smp.h>
  62 #include <sys/socket.h>
  63 #include <sys/sysctl.h>
  64 #include <sys/md5.h>
  65 #include <sys/ucred.h>
  66
  67 #include <net/if.h>
  68 #include <net/if_var.h>
  69 #include <net/vnet.h>
  70 #include <net/route.h>
  71 #include <net/pfil.h>
  72 #include <net/pfvar.h>
  73 #include <net/if_pfsync.h>
  74 #include <net/if_pflog.h>
  75
  76 #include <netinet/in.h>
  77 #include <netinet/ip.h>
  78 #include <netinet/ip_var.h>
  79 #include <netinet6/ip6_var.h>
  80 #include <netinet/ip_icmp.h>
  81
  82 #ifdef INET6
  83 #include <netinet/ip6.h>
  84 #endif /* INET6 */
  85
  86 #ifdef ALTQ
  87 #include <net/altq/altq.h>
  88 #endif
  89
  90 static int               pfattach(void);
  91 static struct pf_pool   *pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
  92                             u_int8_t, u_int8_t, u_int8_t);
  93
  94 static void              pf_mv_pool(struct pf_palist *, struct pf_palist *);
  95 static void              pf_empty_pool(struct pf_palist *);
  96 static int               pfioctl(struct cdev *, u_long, caddr_t, int,
  97                             struct thread *);
  98 #ifdef ALTQ
  99 static int               pf_begin_altq(u_int32_t *);
 100 static int               pf_rollback_altq(u_int32_t);
 101 static int               pf_commit_altq(u_int32_t);
 102 static int               pf_enable_altq(struct pf_altq *);
 103 static int               pf_disable_altq(struct pf_altq *);
 104 static u_int32_t         pf_qname2qid(char *);
 105 static void              pf_qid_unref(u_int32_t);
 106 #endif /* ALTQ */
 107 static int               pf_begin_rules(u_int32_t *, int, const char *);
 108 static int               pf_rollback_rules(u_int32_t, int, char *);
 109 static int               pf_setup_pfsync_matching(struct pf_ruleset *);
 110 static void              pf_hash_rule(MD5_CTX *, struct pf_rule *);
 111 static void              pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
 112 static int               pf_commit_rules(u_int32_t, int, char *);
 113 static int               pf_addr_setup(struct pf_ruleset *,
 114                             struct pf_addr_wrap *, sa_family_t);
 115 static void              pf_addr_copyout(struct pf_addr_wrap *);
 116
 117 VNET_DEFINE(struct pf_rule,     pf_default_rule);
 118
 119 #ifdef ALTQ
 120 static VNET_DEFINE(int,         pf_altq_running);
 121 #define V_pf_altq_running       VNET(pf_altq_running)
 122 #endif
 123
 124 #define TAGID_MAX        50000
 125 struct pf_tagname {
 126         TAILQ_ENTRY(pf_tagname) entries;
 127         char                    name[PF_TAG_NAME_SIZE];
 128         uint16_t                tag;
 129         int                     ref;
 130 };
 131
 132 TAILQ_HEAD(pf_tags, pf_tagname);
 133 #define V_pf_tags               VNET(pf_tags)
 134 VNET_DEFINE(struct pf_tags, pf_tags);
 135 #define V_pf_qids               VNET(pf_qids)
 136 VNET_DEFINE(struct pf_tags, pf_qids);
 137 static MALLOC_DEFINE(M_PFTAG, "pf_tag", "pf(4) tag names");
 138 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
 139 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
 140
 141 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
 142 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
 143 #endif
 144
 145 static u_int16_t         tagname2tag(struct pf_tags *, char *);
 146 static u_int16_t         pf_tagname2tag(char *);
 147 static void              tag_unref(struct pf_tags *, u_int16_t);
 148
 149 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
 150
 151 struct cdev *pf_dev;
 152
 153 /*
 154  * XXX - These are new and need to be checked when moveing to a new version
 155  */
 156 static void              pf_clear_states(void);
 157 static int               pf_clear_tables(void);
 158 static void              pf_clear_srcnodes(struct pf_src_node *);
 159 static void              pf_kill_srcnodes(struct pfioc_src_node_kill *);
 160 static void              pf_tbladdr_copyout(struct pf_addr_wrap *);
 161
 162 /*
 163  * Wrapper functions for pfil(9) hooks
 164  */
 165 #ifdef INET
 166 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
 167     int dir, struct inpcb *inp);
 168 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
 169     int dir, struct inpcb *inp);
 170 #endif
 171 #ifdef INET6
 172 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
 173     int dir, struct inpcb *inp);
 174 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
 175     int dir, struct inpcb *inp);
 176 #endif
 177
 178 static int              hook_pf(void);
 179 static int              dehook_pf(void);
 180 static int              shutdown_pf(void);
 181 static int              pf_load(void);
 182 static int              pf_unload(void);
 183
 184 static struct cdevsw pf_cdevsw = {
 185         .d_ioctl =      pfioctl,
 186         .d_name =       PF_NAME,
 187         .d_version =    D_VERSION,
 188 };
 189
 190 static volatile VNET_DEFINE(int, pf_pfil_hooked);
 191 #define V_pf_pfil_hooked        VNET(pf_pfil_hooked)
 192 VNET_DEFINE(int,                pf_end_threads);
 193
 194 struct rwlock                   pf_rules_lock;
 195 struct sx                       pf_ioctl_lock;
 196
 197 /* pfsync */
 198 pfsync_state_import_t           *pfsync_state_import_ptr = NULL;
 199 pfsync_insert_state_t           *pfsync_insert_state_ptr = NULL;
 200 pfsync_update_state_t           *pfsync_update_state_ptr = NULL;
 201 pfsync_delete_state_t           *pfsync_delete_state_ptr = NULL;
 202 pfsync_clear_states_t           *pfsync_clear_states_ptr = NULL;
 203 pfsync_defer_t                  *pfsync_defer_ptr = NULL;
 204 /* pflog */
 205 pflog_packet_t                  *pflog_packet_ptr = NULL;
 206
 207 static int
 208 pfattach(void)
 209 {
 210         u_int32_t *my_timeout = V_pf_default_rule.timeout;
 211         int error;
 212
 213         if (IS_DEFAULT_VNET(curvnet))
 214                 pf_mtag_initialize();
 215         pf_initialize();
 216         pfr_initialize();
 217         pfi_initialize();
 218         pf_normalize_init();
 219
 220         V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
 221         V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
 222
 223         RB_INIT(&V_pf_anchors);
 224         pf_init_ruleset(&pf_main_ruleset);
 225
 226         /* default rule should never be garbage collected */
 227         V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
 228 #ifdef PF_DEFAULT_TO_DROP
 229         V_pf_default_rule.action = PF_DROP;
 230 #else
 231         V_pf_default_rule.action = PF_PASS;
 232 #endif
 233         V_pf_default_rule.nr = -1;
 234         V_pf_default_rule.rtableid = -1;
 235
 236         V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
 237         V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
 238         V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
 239
 240         /* initialize default timeouts */
 241         my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
 242         my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
 243         my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
 244         my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
 245         my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
 246         my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
 247         my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
 248         my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
 249         my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
 250         my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
 251         my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
 252         my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
 253         my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
 254         my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
 255         my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
 256         my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
 257         my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
 258         my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
 259         my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
 260         my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
 261
 262         bzero(&V_pf_status, sizeof(V_pf_status));
 263         V_pf_status.debug = PF_DEBUG_URGENT;
 264
 265         V_pf_pfil_hooked = 0;
 266
 267         /* XXX do our best to avoid a conflict */
 268         V_pf_status.hostid = arc4random();
 269
 270         for (int i = 0; i < PFRES_MAX; i++)
 271                 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
 272         for (int i = 0; i < LCNT_MAX; i++)
 273                 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
 274         for (int i = 0; i < FCNT_MAX; i++)
 275                 V_pf_status.fcounters[i] = counter_u64_alloc(M_WAITOK);
 276         for (int i = 0; i < SCNT_MAX; i++)
 277                 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
 278
 279         if ((error = kproc_create(pf_purge_thread, curvnet, NULL, 0, 0,
 280             "pf purge")) != 0)
 281                 /* XXXGL: leaked all above. */
 282                 return (error);
 283         if ((error = swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
 284             INTR_MPSAFE, &V_pf_swi_cookie)) != 0)
 285                 /* XXXGL: leaked all above. */
 286                 return (error);
 287
 288         return (0);
 289 }
 290
 291 static struct pf_pool *
 292 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
 293     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
 294     u_int8_t check_ticket)
 295 {
 296         struct pf_ruleset       *ruleset;
 297         struct pf_rule          *rule;
 298         int                      rs_num;
 299
 300         ruleset = pf_find_ruleset(anchor);
 301         if (ruleset == NULL)
 302                 return (NULL);
 303         rs_num = pf_get_ruleset_number(rule_action);
 304         if (rs_num >= PF_RULESET_MAX)
 305                 return (NULL);
 306         if (active) {
 307                 if (check_ticket && ticket !=
 308                     ruleset->rules[rs_num].active.ticket)
 309                         return (NULL);
 310                 if (r_last)
 311                         rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
 312                             pf_rulequeue);
 313                 else
 314                         rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
 315         } else {
 316                 if (check_ticket && ticket !=
 317                     ruleset->rules[rs_num].inactive.ticket)
 318                         return (NULL);
 319                 if (r_last)
 320                         rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
 321                             pf_rulequeue);
 322                 else
 323                         rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
 324         }
 325         if (!r_last) {
 326                 while ((rule != NULL) && (rule->nr != rule_number))
 327                         rule = TAILQ_NEXT(rule, entries);
 328         }
 329         if (rule == NULL)
 330                 return (NULL);
 331
 332         return (&rule->rpool);
 333 }
 334
 335 static void
 336 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
 337 {
 338         struct pf_pooladdr      *mv_pool_pa;
 339
 340         while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
 341                 TAILQ_REMOVE(poola, mv_pool_pa, entries);
 342                 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
 343         }
 344 }
 345
 346 static void
 347 pf_empty_pool(struct pf_palist *poola)
 348 {
 349         struct pf_pooladdr *pa;
 350
 351         while ((pa = TAILQ_FIRST(poola)) != NULL) {
 352                 switch (pa->addr.type) {
 353                 case PF_ADDR_DYNIFTL:
 354                         pfi_dynaddr_remove(pa->addr.p.dyn);
 355                         break;
 356                 case PF_ADDR_TABLE:
 357                         /* XXX: this could be unfinished pooladdr on pabuf */
 358                         if (pa->addr.p.tbl != NULL)
 359                                 pfr_detach_table(pa->addr.p.tbl);
 360                         break;
 361                 }
 362                 if (pa->kif)
 363                         pfi_kif_unref(pa->kif);
 364                 TAILQ_REMOVE(poola, pa, entries);
 365                 free(pa, M_PFRULE);
 366         }
 367 }
 368
 369 static void
 370 pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
 371 {
 372
 373         PF_RULES_WASSERT();
 374
 375         TAILQ_REMOVE(rulequeue, rule, entries);
 376
 377         PF_UNLNKDRULES_LOCK();
 378         rule->rule_flag |= PFRULE_REFS;
 379         TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
 380         PF_UNLNKDRULES_UNLOCK();
 381 }
 382
 383 void
 384 pf_free_rule(struct pf_rule *rule)
 385 {
 386
 387         PF_RULES_WASSERT();
 388
 389         if (rule->tag)
 390                 tag_unref(&V_pf_tags, rule->tag);
 391         if (rule->match_tag)
 392                 tag_unref(&V_pf_tags, rule->match_tag);
 393 #ifdef ALTQ
 394         if (rule->pqid != rule->qid)
 395                 pf_qid_unref(rule->pqid);
 396         pf_qid_unref(rule->qid);
 397 #endif
 398         switch (rule->src.addr.type) {
 399         case PF_ADDR_DYNIFTL:
 400                 pfi_dynaddr_remove(rule->src.addr.p.dyn);
 401                 break;
 402         case PF_ADDR_TABLE:
 403                 pfr_detach_table(rule->src.addr.p.tbl);
 404                 break;
 405         }
 406         switch (rule->dst.addr.type) {
 407         case PF_ADDR_DYNIFTL:
 408                 pfi_dynaddr_remove(rule->dst.addr.p.dyn);
 409                 break;
 410         case PF_ADDR_TABLE:
 411                 pfr_detach_table(rule->dst.addr.p.tbl);
 412                 break;
 413         }
 414         if (rule->overload_tbl)
 415                 pfr_detach_table(rule->overload_tbl);
 416         if (rule->kif)
 417                 pfi_kif_unref(rule->kif);
 418         pf_anchor_remove(rule);
 419         pf_empty_pool(&rule->rpool.list);
 420         counter_u64_free(rule->states_cur);
 421         counter_u64_free(rule->states_tot);
 422         counter_u64_free(rule->src_nodes);
 423         free(rule, M_PFRULE);
 424 }
 425
 426 static u_int16_t
 427 tagname2tag(struct pf_tags *head, char *tagname)
 428 {
 429         struct pf_tagname       *tag, *p = NULL;
 430         u_int16_t                new_tagid = 1;
 431
 432         PF_RULES_WASSERT();
 433
 434         TAILQ_FOREACH(tag, head, entries)
 435                 if (strcmp(tagname, tag->name) == 0) {
 436                         tag->ref++;
 437                         return (tag->tag);
 438                 }
 439
 440         /*
 441          * to avoid fragmentation, we do a linear search from the beginning
 442          * and take the first free slot we find. if there is none or the list
 443          * is empty, append a new entry at the end.
 444          */
 445
 446         /* new entry */
 447         if (!TAILQ_EMPTY(head))
 448                 for (p = TAILQ_FIRST(head); p != NULL &&
 449                     p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
 450                         new_tagid = p->tag + 1;
 451
 452         if (new_tagid > TAGID_MAX)
 453                 return (0);
 454
 455         /* allocate and fill new struct pf_tagname */
 456         tag = malloc(sizeof(*tag), M_PFTAG, M_NOWAIT|M_ZERO);
 457         if (tag == NULL)
 458                 return (0);
 459         strlcpy(tag->name, tagname, sizeof(tag->name));
 460         tag->tag = new_tagid;
 461         tag->ref++;
 462
 463         if (p != NULL)  /* insert new entry before p */
 464                 TAILQ_INSERT_BEFORE(p, tag, entries);
 465         else    /* either list empty or no free slot in between */
 466                 TAILQ_INSERT_TAIL(head, tag, entries);
 467
 468         return (tag->tag);
 469 }
 470
 471 static void
 472 tag_unref(struct pf_tags *head, u_int16_t tag)
 473 {
 474         struct pf_tagname       *p, *next;
 475
 476         PF_RULES_WASSERT();
 477
 478         for (p = TAILQ_FIRST(head); p != NULL; p = next) {
 479                 next = TAILQ_NEXT(p, entries);
 480                 if (tag == p->tag) {
 481                         if (--p->ref == 0) {
 482                                 TAILQ_REMOVE(head, p, entries);
 483                                 free(p, M_PFTAG);
 484                         }
 485                         break;
 486                 }
 487         }
 488 }
 489
 490 static u_int16_t
 491 pf_tagname2tag(char *tagname)
 492 {
 493         return (tagname2tag(&V_pf_tags, tagname));
 494 }
 495
 496 #ifdef ALTQ
 497 static u_int32_t
 498 pf_qname2qid(char *qname)
 499 {
 500         return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
 501 }
 502
 503 static void
 504 pf_qid_unref(u_int32_t qid)
 505 {
 506         tag_unref(&V_pf_qids, (u_int16_t)qid);
 507 }
 508
 509 static int
 510 pf_begin_altq(u_int32_t *ticket)
 511 {
 512         struct pf_altq  *altq;
 513         int              error = 0;
 514
 515         PF_RULES_WASSERT();
 516
 517         /* Purge the old altq list */
 518         while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
 519                 TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
 520                 if (altq->qname[0] == 0 &&
 521                     (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 522                         /* detach and destroy the discipline */
 523                         error = altq_remove(altq);
 524                 } else
 525                         pf_qid_unref(altq->qid);
 526                 free(altq, M_PFALTQ);
 527         }
 528         if (error)
 529                 return (error);
 530         *ticket = ++V_ticket_altqs_inactive;
 531         V_altqs_inactive_open = 1;
 532         return (0);
 533 }
 534
 535 static int
 536 pf_rollback_altq(u_int32_t ticket)
 537 {
 538         struct pf_altq  *altq;
 539         int              error = 0;
 540
 541         PF_RULES_WASSERT();
 542
 543         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 544                 return (0);
 545         /* Purge the old altq list */
 546         while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
 547                 TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
 548                 if (altq->qname[0] == 0 &&
 549                    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 550                         /* detach and destroy the discipline */
 551                         error = altq_remove(altq);
 552                 } else
 553                         pf_qid_unref(altq->qid);
 554                 free(altq, M_PFALTQ);
 555         }
 556         V_altqs_inactive_open = 0;
 557         return (error);
 558 }
 559
 560 static int
 561 pf_commit_altq(u_int32_t ticket)
 562 {
 563         struct pf_altqqueue     *old_altqs;
 564         struct pf_altq          *altq;
 565         int                      err, error = 0;
 566
 567         PF_RULES_WASSERT();
 568
 569         if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
 570                 return (EBUSY);
 571
 572         /* swap altqs, keep the old. */
 573         old_altqs = V_pf_altqs_active;
 574         V_pf_altqs_active = V_pf_altqs_inactive;
 575         V_pf_altqs_inactive = old_altqs;
 576         V_ticket_altqs_active = V_ticket_altqs_inactive;
 577
 578         /* Attach new disciplines */
 579         TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
 580         if (altq->qname[0] == 0 &&
 581            (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 582                         /* attach the discipline */
 583                         error = altq_pfattach(altq);
 584                         if (error == 0 && V_pf_altq_running)
 585                                 error = pf_enable_altq(altq);
 586                         if (error != 0)
 587                                 return (error);
 588                 }
 589         }
 590
 591         /* Purge the old altq list */
 592         while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
 593                 TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
 594                 if (altq->qname[0] == 0 &&
 595                     (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
 596                         /* detach and destroy the discipline */
 597                         if (V_pf_altq_running)
 598                                 error = pf_disable_altq(altq);
 599                         err = altq_pfdetach(altq);
 600                         if (err != 0 && error == 0)
 601                                 error = err;
 602                         err = altq_remove(altq);
 603                         if (err != 0 && error == 0)
 604                                 error = err;
 605                 } else
 606                         pf_qid_unref(altq->qid);
 607                 free(altq, M_PFALTQ);
 608         }
 609
 610         V_altqs_inactive_open = 0;
 611         return (error);
 612 }
 613
 614 static int
 615 pf_enable_altq(struct pf_altq *altq)
 616 {
 617         struct ifnet            *ifp;
 618         struct tb_profile        tb;
 619         int                      error = 0;
 620
 621         if ((ifp = ifunit(altq->ifname)) == NULL)
 622                 return (EINVAL);
 623
 624         if (ifp->if_snd.altq_type != ALTQT_NONE)
 625                 error = altq_enable(&ifp->if_snd);
 626
 627         /* set tokenbucket regulator */
 628         if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
 629                 tb.rate = altq->ifbandwidth;
 630                 tb.depth = altq->tbrsize;
 631                 error = tbr_set(&ifp->if_snd, &tb);
 632         }
 633
 634         return (error);
 635 }
 636
 637 static int
 638 pf_disable_altq(struct pf_altq *altq)
 639 {
 640         struct ifnet            *ifp;
 641         struct tb_profile        tb;
 642         int                      error;
 643
 644         if ((ifp = ifunit(altq->ifname)) == NULL)
 645                 return (EINVAL);
 646
 647         /*
 648          * when the discipline is no longer referenced, it was overridden
 649          * by a new one.  if so, just return.
 650          */
 651         if (altq->altq_disc != ifp->if_snd.altq_disc)
 652                 return (0);
 653
 654         error = altq_disable(&ifp->if_snd);
 655
 656         if (error == 0) {
 657                 /* clear tokenbucket regulator */
 658                 tb.rate = 0;
 659                 error = tbr_set(&ifp->if_snd, &tb);
 660         }
 661
 662         return (error);
 663 }
 664
 665 void
 666 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
 667 {
 668         struct ifnet    *ifp1;
 669         struct pf_altq  *a1, *a2, *a3;
 670         u_int32_t        ticket;
 671         int              error = 0;
 672
 673         /* Interrupt userland queue modifications */
 674         if (V_altqs_inactive_open)
 675                 pf_rollback_altq(V_ticket_altqs_inactive);
 676
 677         /* Start new altq ruleset */
 678         if (pf_begin_altq(&ticket))
 679                 return;
 680
 681         /* Copy the current active set */
 682         TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
 683                 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
 684                 if (a2 == NULL) {
 685                         error = ENOMEM;
 686                         break;
 687                 }
 688                 bcopy(a1, a2, sizeof(struct pf_altq));
 689
 690                 if (a2->qname[0] != 0) {
 691                         if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
 692                                 error = EBUSY;
 693                                 free(a2, M_PFALTQ);
 694                                 break;
 695                         }
 696                         a2->altq_disc = NULL;
 697                         TAILQ_FOREACH(a3, V_pf_altqs_inactive, entries) {
 698                                 if (strncmp(a3->ifname, a2->ifname,
 699                                     IFNAMSIZ) == 0 && a3->qname[0] == 0) {
 700                                         a2->altq_disc = a3->altq_disc;
 701                                         break;
 702                                 }
 703                         }
 704                 }
 705                 /* Deactivate the interface in question */
 706                 a2->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
 707                 if ((ifp1 = ifunit(a2->ifname)) == NULL ||
 708                     (remove && ifp1 == ifp)) {
 709                         a2->local_flags |= PFALTQ_FLAG_IF_REMOVED;
 710                 } else {
 711                         error = altq_add(a2);
 712
 713                         if (ticket != V_ticket_altqs_inactive)
 714                                 error = EBUSY;
 715
 716                         if (error) {
 717                                 free(a2, M_PFALTQ);
 718                                 break;
 719                         }
 720                 }
 721
 722                 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
 723         }
 724
 725         if (error != 0)
 726                 pf_rollback_altq(ticket);
 727         else
 728                 pf_commit_altq(ticket);
 729 }
 730 #endif /* ALTQ */
 731
 732 static int
 733 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
 734 {
 735         struct pf_ruleset       *rs;
 736         struct pf_rule          *rule;
 737
 738         PF_RULES_WASSERT();
 739
 740         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 741                 return (EINVAL);
 742         rs = pf_find_or_create_ruleset(anchor);
 743         if (rs == NULL)
 744                 return (EINVAL);
 745         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 746                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 747                 rs->rules[rs_num].inactive.rcount--;
 748         }
 749         *ticket = ++rs->rules[rs_num].inactive.ticket;
 750         rs->rules[rs_num].inactive.open = 1;
 751         return (0);
 752 }
 753
 754 static int
 755 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
 756 {
 757         struct pf_ruleset       *rs;
 758         struct pf_rule          *rule;
 759
 760         PF_RULES_WASSERT();
 761
 762         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 763                 return (EINVAL);
 764         rs = pf_find_ruleset(anchor);
 765         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 766             rs->rules[rs_num].inactive.ticket != ticket)
 767                 return (0);
 768         while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
 769                 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
 770                 rs->rules[rs_num].inactive.rcount--;
 771         }
 772         rs->rules[rs_num].inactive.open = 0;
 773         return (0);
 774 }
 775
 776 #define PF_MD5_UPD(st, elm)                                             \
 777                 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
 778
 779 #define PF_MD5_UPD_STR(st, elm)                                         \
 780                 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
 781
 782 #define PF_MD5_UPD_HTONL(st, elm, stor) do {                            \
 783                 (stor) = htonl((st)->elm);                              \
 784                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
 785 } while (0)
 786
 787 #define PF_MD5_UPD_HTONS(st, elm, stor) do {                            \
 788                 (stor) = htons((st)->elm);                              \
 789                 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
 790 } while (0)
 791
 792 static void
 793 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
 794 {
 795         PF_MD5_UPD(pfr, addr.type);
 796         switch (pfr->addr.type) {
 797                 case PF_ADDR_DYNIFTL:
 798                         PF_MD5_UPD(pfr, addr.v.ifname);
 799                         PF_MD5_UPD(pfr, addr.iflags);
 800                         break;
 801                 case PF_ADDR_TABLE:
 802                         PF_MD5_UPD(pfr, addr.v.tblname);
 803                         break;
 804                 case PF_ADDR_ADDRMASK:
 805                         /* XXX ignore af? */
 806                         PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
 807                         PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
 808                         break;
 809         }
 810
 811         PF_MD5_UPD(pfr, port[0]);
 812         PF_MD5_UPD(pfr, port[1]);
 813         PF_MD5_UPD(pfr, neg);
 814         PF_MD5_UPD(pfr, port_op);
 815 }
 816
 817 static void
 818 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
 819 {
 820         u_int16_t x;
 821         u_int32_t y;
 822
 823         pf_hash_rule_addr(ctx, &rule->src);
 824         pf_hash_rule_addr(ctx, &rule->dst);
 825         PF_MD5_UPD_STR(rule, label);
 826         PF_MD5_UPD_STR(rule, ifname);
 827         PF_MD5_UPD_STR(rule, match_tagname);
 828         PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
 829         PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
 830         PF_MD5_UPD_HTONL(rule, prob, y);
 831         PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
 832         PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
 833         PF_MD5_UPD(rule, uid.op);
 834         PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
 835         PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
 836         PF_MD5_UPD(rule, gid.op);
 837         PF_MD5_UPD_HTONL(rule, rule_flag, y);
 838         PF_MD5_UPD(rule, action);
 839         PF_MD5_UPD(rule, direction);
 840         PF_MD5_UPD(rule, af);
 841         PF_MD5_UPD(rule, quick);
 842         PF_MD5_UPD(rule, ifnot);
 843         PF_MD5_UPD(rule, match_tag_not);
 844         PF_MD5_UPD(rule, natpass);
 845         PF_MD5_UPD(rule, keep_state);
 846         PF_MD5_UPD(rule, proto);
 847         PF_MD5_UPD(rule, type);
 848         PF_MD5_UPD(rule, code);
 849         PF_MD5_UPD(rule, flags);
 850         PF_MD5_UPD(rule, flagset);
 851         PF_MD5_UPD(rule, allow_opts);
 852         PF_MD5_UPD(rule, rt);
 853         PF_MD5_UPD(rule, tos);
 854 }
 855
 856 static int
 857 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
 858 {
 859         struct pf_ruleset       *rs;
 860         struct pf_rule          *rule, **old_array;
 861         struct pf_rulequeue     *old_rules;
 862         int                      error;
 863         u_int32_t                old_rcount;
 864
 865         PF_RULES_WASSERT();
 866
 867         if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
 868                 return (EINVAL);
 869         rs = pf_find_ruleset(anchor);
 870         if (rs == NULL || !rs->rules[rs_num].inactive.open ||
 871             ticket != rs->rules[rs_num].inactive.ticket)
 872                 return (EBUSY);
 873
 874         /* Calculate checksum for the main ruleset */
 875         if (rs == &pf_main_ruleset) {
 876                 error = pf_setup_pfsync_matching(rs);
 877                 if (error != 0)
 878                         return (error);
 879         }
 880
 881         /* Swap rules, keep the old. */
 882         old_rules = rs->rules[rs_num].active.ptr;
 883         old_rcount = rs->rules[rs_num].active.rcount;
 884         old_array = rs->rules[rs_num].active.ptr_array;
 885
 886         rs->rules[rs_num].active.ptr =
 887             rs->rules[rs_num].inactive.ptr;
 888         rs->rules[rs_num].active.ptr_array =
 889             rs->rules[rs_num].inactive.ptr_array;
 890         rs->rules[rs_num].active.rcount =
 891             rs->rules[rs_num].inactive.rcount;
 892         rs->rules[rs_num].inactive.ptr = old_rules;
 893         rs->rules[rs_num].inactive.ptr_array = old_array;
 894         rs->rules[rs_num].inactive.rcount = old_rcount;
 895
 896         rs->rules[rs_num].active.ticket =
 897             rs->rules[rs_num].inactive.ticket;
 898         pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
 899
 900
 901         /* Purge the old rule list. */
 902         while ((rule = TAILQ_FIRST(old_rules)) != NULL)
 903                 pf_unlink_rule(old_rules, rule);
 904         if (rs->rules[rs_num].inactive.ptr_array)
 905                 free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
 906         rs->rules[rs_num].inactive.ptr_array = NULL;
 907         rs->rules[rs_num].inactive.rcount = 0;
 908         rs->rules[rs_num].inactive.open = 0;
 909         pf_remove_if_empty_ruleset(rs);
 910
 911         return (0);
 912 }
 913
 914 static int
 915 pf_setup_pfsync_matching(struct pf_ruleset *rs)
 916 {
 917         MD5_CTX                  ctx;
 918         struct pf_rule          *rule;
 919         int                      rs_cnt;
 920         u_int8_t                 digest[PF_MD5_DIGEST_LENGTH];
 921
 922         MD5Init(&ctx);
 923         for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
 924                 /* XXX PF_RULESET_SCRUB as well? */
 925                 if (rs_cnt == PF_RULESET_SCRUB)
 926                         continue;
 927
 928                 if (rs->rules[rs_cnt].inactive.ptr_array)
 929                         free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
 930                 rs->rules[rs_cnt].inactive.ptr_array = NULL;
 931
 932                 if (rs->rules[rs_cnt].inactive.rcount) {
 933                         rs->rules[rs_cnt].inactive.ptr_array =
 934                             malloc(sizeof(caddr_t) *
 935                             rs->rules[rs_cnt].inactive.rcount,
 936                             M_TEMP, M_NOWAIT);
 937
 938                         if (!rs->rules[rs_cnt].inactive.ptr_array)
 939                                 return (ENOMEM);
 940                 }
 941
 942                 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
 943                     entries) {
 944                         pf_hash_rule(&ctx, rule);
 945                         (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
 946                 }
 947         }
 948
 949         MD5Final(digest, &ctx);
 950         memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
 951         return (0);
 952 }
 953
 954 static int
 955 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
 956     sa_family_t af)
 957 {
 958         int error = 0;
 959
 960         switch (addr->type) {
 961         case PF_ADDR_TABLE:
 962                 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
 963                 if (addr->p.tbl == NULL)
 964                         error = ENOMEM;
 965                 break;
 966         case PF_ADDR_DYNIFTL:
 967                 error = pfi_dynaddr_setup(addr, af);
 968                 break;
 969         }
 970
 971         return (error);
 972 }
 973
 974 static void
 975 pf_addr_copyout(struct pf_addr_wrap *addr)
 976 {
 977
 978         switch (addr->type) {
 979         case PF_ADDR_DYNIFTL:
 980                 pfi_dynaddr_copyout(addr);
 981                 break;
 982         case PF_ADDR_TABLE:
 983                 pf_tbladdr_copyout(addr);
 984                 break;
 985         }
 986 }
 987
 988 static int
 989 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
 990 {
 991         int                      error = 0;
 992
 993         /* XXX keep in sync with switch() below */
 994         if (securelevel_gt(td->td_ucred, 2))
 995                 switch (cmd) {
 996                 case DIOCGETRULES:
 997                 case DIOCGETRULE:
 998                 case DIOCGETADDRS:
 999                 case DIOCGETADDR:
1000                 case DIOCGETSTATE:
1001                 case DIOCSETSTATUSIF:
1002                 case DIOCGETSTATUS:
1003                 case DIOCCLRSTATUS:
1004                 case DIOCNATLOOK:
1005                 case DIOCSETDEBUG:
1006                 case DIOCGETSTATES:
1007                 case DIOCGETTIMEOUT:
1008                 case DIOCCLRRULECTRS:
1009                 case DIOCGETLIMIT:
1010                 case DIOCGETALTQS:
1011                 case DIOCGETALTQ:
1012                 case DIOCGETQSTATS:
1013                 case DIOCGETRULESETS:
1014                 case DIOCGETRULESET:
1015                 case DIOCRGETTABLES:
1016                 case DIOCRGETTSTATS:
1017                 case DIOCRCLRTSTATS:
1018                 case DIOCRCLRADDRS:
1019                 case DIOCRADDADDRS:
1020                 case DIOCRDELADDRS:
1021                 case DIOCRSETADDRS:
1022                 case DIOCRGETADDRS:
1023                 case DIOCRGETASTATS:
1024                 case DIOCRCLRASTATS:
1025                 case DIOCRTSTADDRS:
1026                 case DIOCOSFPGET:
1027                 case DIOCGETSRCNODES:
1028                 case DIOCCLRSRCNODES:
1029                 case DIOCIGETIFACES:
1030                 case DIOCGIFSPEED:
1031                 case DIOCSETIFFLAG:
1032                 case DIOCCLRIFFLAG:
1033                         break;
1034                 case DIOCRCLRTABLES:
1035                 case DIOCRADDTABLES:
1036                 case DIOCRDELTABLES:
1037                 case DIOCRSETTFLAGS:
1038                         if (((struct pfioc_table *)addr)->pfrio_flags &
1039                             PFR_FLAG_DUMMY)
1040                                 break; /* dummy operation ok */
1041                         return (EPERM);
1042                 default:
1043                         return (EPERM);
1044                 }
1045
1046         if (!(flags & FWRITE))
1047                 switch (cmd) {
1048                 case DIOCGETRULES:
1049                 case DIOCGETADDRS:
1050                 case DIOCGETADDR:
1051                 case DIOCGETSTATE:
1052                 case DIOCGETSTATUS:
1053                 case DIOCGETSTATES:
1054                 case DIOCGETTIMEOUT:
1055                 case DIOCGETLIMIT:
1056                 case DIOCGETALTQS:
1057                 case DIOCGETALTQ:
1058                 case DIOCGETQSTATS:
1059                 case DIOCGETRULESETS:
1060                 case DIOCGETRULESET:
1061                 case DIOCNATLOOK:
1062                 case DIOCRGETTABLES:
1063                 case DIOCRGETTSTATS:
1064                 case DIOCRGETADDRS:
1065                 case DIOCRGETASTATS:
1066                 case DIOCRTSTADDRS:
1067                 case DIOCOSFPGET:
1068                 case DIOCGETSRCNODES:
1069                 case DIOCIGETIFACES:
1070                 case DIOCGIFSPEED:
1071                         break;
1072                 case DIOCRCLRTABLES:
1073                 case DIOCRADDTABLES:
1074                 case DIOCRDELTABLES:
1075                 case DIOCRCLRTSTATS:
1076                 case DIOCRCLRADDRS:
1077                 case DIOCRADDADDRS:
1078                 case DIOCRDELADDRS:
1079                 case DIOCRSETADDRS:
1080                 case DIOCRSETTFLAGS:
1081                         if (((struct pfioc_table *)addr)->pfrio_flags &
1082                             PFR_FLAG_DUMMY) {
1083                                 flags |= FWRITE; /* need write lock for dummy */
1084                                 break; /* dummy operation ok */
1085                         }
1086                         return (EACCES);
1087                 case DIOCGETRULE:
1088                         if (((struct pfioc_rule *)addr)->action ==
1089                             PF_GET_CLR_CNTR)
1090                                 return (EACCES);
1091                         break;
1092                 default:
1093                         return (EACCES);
1094                 }
1095
1096         CURVNET_SET(TD_TO_VNET(td));
1097
1098         switch (cmd) {
1099         case DIOCSTART:
1100                 sx_xlock(&pf_ioctl_lock);
1101                 if (V_pf_status.running)
1102                         error = EEXIST;
1103                 else {
1104                         int cpu;
1105
1106                         error = hook_pf();
1107                         if (error) {
1108                                 DPFPRINTF(PF_DEBUG_MISC,
1109                                     ("pf: pfil registration failed\n"));
1110                                 break;
1111                         }
1112                         V_pf_status.running = 1;
1113                         V_pf_status.since = time_second;
1114
1115                         CPU_FOREACH(cpu)
1116                                 V_pf_stateid[cpu] = time_second;
1117
1118                         DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1119                 }
1120                 break;
1121
1122         case DIOCSTOP:
1123                 sx_xlock(&pf_ioctl_lock);
1124                 if (!V_pf_status.running)
1125                         error = ENOENT;
1126                 else {
1127                         V_pf_status.running = 0;
1128                         error = dehook_pf();
1129                         if (error) {
1130                                 V_pf_status.running = 1;
1131                                 DPFPRINTF(PF_DEBUG_MISC,
1132                                     ("pf: pfil unregistration failed\n"));
1133                         }
1134                         V_pf_status.since = time_second;
1135                         DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1136                 }
1137                 break;
1138
1139         case DIOCADDRULE: {
1140                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1141                 struct pf_ruleset       *ruleset;
1142                 struct pf_rule          *rule, *tail;
1143                 struct pf_pooladdr      *pa;
1144                 struct pfi_kif          *kif = NULL;
1145                 int                      rs_num;
1146
1147                 if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1148                         error = EINVAL;
1149                         break;
1150                 }
1151 #ifndef INET
1152                 if (pr->rule.af == AF_INET) {
1153                         error = EAFNOSUPPORT;
1154                         break;
1155                 }
1156 #endif /* INET */
1157 #ifndef INET6
1158                 if (pr->rule.af == AF_INET6) {
1159                         error = EAFNOSUPPORT;
1160                         break;
1161                 }
1162 #endif /* INET6 */
1163
1164                 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1165                 bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1166                 if (rule->ifname[0])
1167                         kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1168                 rule->states_cur = counter_u64_alloc(M_WAITOK);
1169                 rule->states_tot = counter_u64_alloc(M_WAITOK);
1170                 rule->src_nodes = counter_u64_alloc(M_WAITOK);
1171                 rule->cuid = td->td_ucred->cr_ruid;
1172                 rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1173                 TAILQ_INIT(&rule->rpool.list);
1174
1175 #define ERROUT(x)       { error = (x); goto DIOCADDRULE_error; }
1176
1177                 PF_RULES_WLOCK();
1178                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1179                 ruleset = pf_find_ruleset(pr->anchor);
1180                 if (ruleset == NULL)
1181                         ERROUT(EINVAL);
1182                 rs_num = pf_get_ruleset_number(pr->rule.action);
1183                 if (rs_num >= PF_RULESET_MAX)
1184                         ERROUT(EINVAL);
1185                 if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1186                         DPFPRINTF(PF_DEBUG_MISC,
1187                             ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1188                             ruleset->rules[rs_num].inactive.ticket));
1189                         ERROUT(EBUSY);
1190                 }
1191                 if (pr->pool_ticket != V_ticket_pabuf) {
1192                         DPFPRINTF(PF_DEBUG_MISC,
1193                             ("pool_ticket: %d != %d\n", pr->pool_ticket,
1194                             V_ticket_pabuf));
1195                         ERROUT(EBUSY);
1196                 }
1197
1198                 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1199                     pf_rulequeue);
1200                 if (tail)
1201                         rule->nr = tail->nr + 1;
1202                 else
1203                         rule->nr = 0;
1204                 if (rule->ifname[0]) {
1205                         rule->kif = pfi_kif_attach(kif, rule->ifname);
1206                         pfi_kif_ref(rule->kif);
1207                 } else
1208                         rule->kif = NULL;
1209
1210                 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1211                         error = EBUSY;
1212
1213 #ifdef ALTQ
1214                 /* set queue IDs */
1215                 if (rule->qname[0] != 0) {
1216                         if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1217                                 error = EBUSY;
1218                         else if (rule->pqname[0] != 0) {
1219                                 if ((rule->pqid =
1220                                     pf_qname2qid(rule->pqname)) == 0)
1221                                         error = EBUSY;
1222                         } else
1223                                 rule->pqid = rule->qid;
1224                 }
1225 #endif
1226                 if (rule->tagname[0])
1227                         if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1228                                 error = EBUSY;
1229                 if (rule->match_tagname[0])
1230                         if ((rule->match_tag =
1231                             pf_tagname2tag(rule->match_tagname)) == 0)
1232                                 error = EBUSY;
1233                 if (rule->rt && !rule->direction)
1234                         error = EINVAL;
1235                 if (!rule->log)
1236                         rule->logif = 0;
1237                 if (rule->logif >= PFLOGIFS_MAX)
1238                         error = EINVAL;
1239                 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1240                         error = ENOMEM;
1241                 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1242                         error = ENOMEM;
1243                 if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1244                         error = EINVAL;
1245                 if (rule->scrub_flags & PFSTATE_SETPRIO &&
1246                     (rule->set_prio[0] > PF_PRIO_MAX ||
1247                     rule->set_prio[1] > PF_PRIO_MAX))
1248                         error = EINVAL;
1249                 TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1250                         if (pa->addr.type == PF_ADDR_TABLE) {
1251                                 pa->addr.p.tbl = pfr_attach_table(ruleset,
1252                                     pa->addr.v.tblname);
1253                                 if (pa->addr.p.tbl == NULL)
1254                                         error = ENOMEM;
1255                         }
1256
1257                 if (rule->overload_tblname[0]) {
1258                         if ((rule->overload_tbl = pfr_attach_table(ruleset,
1259                             rule->overload_tblname)) == NULL)
1260                                 error = EINVAL;
1261                         else
1262                                 rule->overload_tbl->pfrkt_flags |=
1263                                     PFR_TFLAG_ACTIVE;
1264                 }
1265
1266                 pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1267                 if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1268                     (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1269                     (rule->rt > PF_FASTROUTE)) &&
1270                     (TAILQ_FIRST(&rule->rpool.list) == NULL))
1271                         error = EINVAL;
1272
1273                 if (error) {
1274                         pf_free_rule(rule);
1275                         PF_RULES_WUNLOCK();
1276                         break;
1277                 }
1278
1279                 rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1280                 rule->evaluations = rule->packets[0] = rule->packets[1] =
1281                     rule->bytes[0] = rule->bytes[1] = 0;
1282                 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1283                     rule, entries);
1284                 ruleset->rules[rs_num].inactive.rcount++;
1285                 PF_RULES_WUNLOCK();
1286                 break;
1287
1288 #undef ERROUT
1289 DIOCADDRULE_error:
1290                 PF_RULES_WUNLOCK();
1291                 counter_u64_free(rule->states_cur);
1292                 counter_u64_free(rule->states_tot);
1293                 counter_u64_free(rule->src_nodes);
1294                 free(rule, M_PFRULE);
1295                 if (kif)
1296                         free(kif, PFI_MTYPE);
1297                 break;
1298         }
1299
1300         case DIOCGETRULES: {
1301                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1302                 struct pf_ruleset       *ruleset;
1303                 struct pf_rule          *tail;
1304                 int                      rs_num;
1305
1306                 PF_RULES_WLOCK();
1307                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1308                 ruleset = pf_find_ruleset(pr->anchor);
1309                 if (ruleset == NULL) {
1310                         PF_RULES_WUNLOCK();
1311                         error = EINVAL;
1312                         break;
1313                 }
1314                 rs_num = pf_get_ruleset_number(pr->rule.action);
1315                 if (rs_num >= PF_RULESET_MAX) {
1316                         PF_RULES_WUNLOCK();
1317                         error = EINVAL;
1318                         break;
1319                 }
1320                 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1321                     pf_rulequeue);
1322                 if (tail)
1323                         pr->nr = tail->nr + 1;
1324                 else
1325                         pr->nr = 0;
1326                 pr->ticket = ruleset->rules[rs_num].active.ticket;
1327                 PF_RULES_WUNLOCK();
1328                 break;
1329         }
1330
1331         case DIOCGETRULE: {
1332                 struct pfioc_rule       *pr = (struct pfioc_rule *)addr;
1333                 struct pf_ruleset       *ruleset;
1334                 struct pf_rule          *rule;
1335                 int                      rs_num, i;
1336
1337                 PF_RULES_WLOCK();
1338                 pr->anchor[sizeof(pr->anchor) - 1] = 0;
1339                 ruleset = pf_find_ruleset(pr->anchor);
1340                 if (ruleset == NULL) {
1341                         PF_RULES_WUNLOCK();
1342                         error = EINVAL;
1343                         break;
1344                 }
1345                 rs_num = pf_get_ruleset_number(pr->rule.action);
1346                 if (rs_num >= PF_RULESET_MAX) {
1347                         PF_RULES_WUNLOCK();
1348                         error = EINVAL;
1349                         break;
1350                 }
1351                 if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1352                         PF_RULES_WUNLOCK();
1353                         error = EBUSY;
1354                         break;
1355                 }
1356                 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1357                 while ((rule != NULL) && (rule->nr != pr->nr))
1358                         rule = TAILQ_NEXT(rule, entries);
1359                 if (rule == NULL) {
1360                         PF_RULES_WUNLOCK();
1361                         error = EBUSY;
1362                         break;
1363                 }
1364                 bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1365                 pr->rule.u_states_cur = counter_u64_fetch(rule->states_cur);
1366                 pr->rule.u_states_tot = counter_u64_fetch(rule->states_tot);
1367                 pr->rule.u_src_nodes = counter_u64_fetch(rule->src_nodes);
1368                 if (pf_anchor_copyout(ruleset, rule, pr)) {
1369                         PF_RULES_WUNLOCK();
1370                         error = EBUSY;
1371                         break;
1372                 }
1373                 pf_addr_copyout(&pr->rule.src.addr);
1374                 pf_addr_copyout(&pr->rule.dst.addr);
1375                 for (i = 0; i < PF_SKIP_COUNT; ++i)
1376                         if (rule->skip[i].ptr == NULL)
1377                                 pr->rule.skip[i].nr = -1;
1378                         else
1379                                 pr->rule.skip[i].nr =
1380                                     rule->skip[i].ptr->nr;
1381
1382                 if (pr->action == PF_GET_CLR_CNTR) {
1383                         rule->evaluations = 0;
1384                         rule->packets[0] = rule->packets[1] = 0;
1385                         rule->bytes[0] = rule->bytes[1] = 0;
1386                         counter_u64_zero(rule->states_tot);
1387                 }
1388                 PF_RULES_WUNLOCK();
1389                 break;
1390         }
1391
1392         case DIOCCHANGERULE: {
1393                 struct pfioc_rule       *pcr = (struct pfioc_rule *)addr;
1394                 struct pf_ruleset       *ruleset;
1395                 struct pf_rule          *oldrule = NULL, *newrule = NULL;
1396                 struct pfi_kif          *kif = NULL;
1397                 struct pf_pooladdr      *pa;
1398                 u_int32_t                nr = 0;
1399                 int                      rs_num;
1400
1401                 if (pcr->action < PF_CHANGE_ADD_HEAD ||
1402                     pcr->action > PF_CHANGE_GET_TICKET) {
1403                         error = EINVAL;
1404                         break;
1405                 }
1406                 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1407                         error = EINVAL;
1408                         break;
1409                 }
1410
1411                 if (pcr->action != PF_CHANGE_REMOVE) {
1412 #ifndef INET
1413                         if (pcr->rule.af == AF_INET) {
1414                                 error = EAFNOSUPPORT;
1415                                 break;
1416                         }
1417 #endif /* INET */
1418 #ifndef INET6
1419                         if (pcr->rule.af == AF_INET6) {
1420                                 error = EAFNOSUPPORT;
1421                                 break;
1422                         }
1423 #endif /* INET6 */
1424                         newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1425                         bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1426                         if (newrule->ifname[0])
1427                                 kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1428                         newrule->states_cur = counter_u64_alloc(M_WAITOK);
1429                         newrule->states_tot = counter_u64_alloc(M_WAITOK);
1430                         newrule->src_nodes = counter_u64_alloc(M_WAITOK);
1431                         newrule->cuid = td->td_ucred->cr_ruid;
1432                         newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1433                         TAILQ_INIT(&newrule->rpool.list);
1434                 }
1435
1436 #define ERROUT(x)       { error = (x); goto DIOCCHANGERULE_error; }
1437
1438                 PF_RULES_WLOCK();
1439                 if (!(pcr->action == PF_CHANGE_REMOVE ||
1440                     pcr->action == PF_CHANGE_GET_TICKET) &&
1441                     pcr->pool_ticket != V_ticket_pabuf)
1442                         ERROUT(EBUSY);
1443
1444                 ruleset = pf_find_ruleset(pcr->anchor);
1445                 if (ruleset == NULL)
1446                         ERROUT(EINVAL);
1447
1448                 rs_num = pf_get_ruleset_number(pcr->rule.action);
1449                 if (rs_num >= PF_RULESET_MAX)
1450                         ERROUT(EINVAL);
1451
1452                 if (pcr->action == PF_CHANGE_GET_TICKET) {
1453                         pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1454                         ERROUT(0);
1455                 } else if (pcr->ticket !=
1456                             ruleset->rules[rs_num].active.ticket)
1457                                 ERROUT(EINVAL);
1458
1459                 if (pcr->action != PF_CHANGE_REMOVE) {
1460                         if (newrule->ifname[0]) {
1461                                 newrule->kif = pfi_kif_attach(kif,
1462                                     newrule->ifname);
1463                                 pfi_kif_ref(newrule->kif);
1464                         } else
1465                                 newrule->kif = NULL;
1466
1467                         if (newrule->rtableid > 0 &&
1468                             newrule->rtableid >= rt_numfibs)
1469                                 error = EBUSY;
1470
1471 #ifdef ALTQ
1472                         /* set queue IDs */
1473                         if (newrule->qname[0] != 0) {
1474                                 if ((newrule->qid =
1475                                     pf_qname2qid(newrule->qname)) == 0)
1476                                         error = EBUSY;
1477                                 else if (newrule->pqname[0] != 0) {
1478                                         if ((newrule->pqid =
1479                                             pf_qname2qid(newrule->pqname)) == 0)
1480                                                 error = EBUSY;
1481                                 } else
1482                                         newrule->pqid = newrule->qid;
1483                         }
1484 #endif /* ALTQ */
1485                         if (newrule->tagname[0])
1486                                 if ((newrule->tag =
1487                                     pf_tagname2tag(newrule->tagname)) == 0)
1488                                         error = EBUSY;
1489                         if (newrule->match_tagname[0])
1490                                 if ((newrule->match_tag = pf_tagname2tag(
1491                                     newrule->match_tagname)) == 0)
1492                                         error = EBUSY;
1493                         if (newrule->rt && !newrule->direction)
1494                                 error = EINVAL;
1495                         if (!newrule->log)
1496                                 newrule->logif = 0;
1497                         if (newrule->logif >= PFLOGIFS_MAX)
1498                                 error = EINVAL;
1499                         if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1500                                 error = ENOMEM;
1501                         if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1502                                 error = ENOMEM;
1503                         if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1504                                 error = EINVAL;
1505                         TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1506                                 if (pa->addr.type == PF_ADDR_TABLE) {
1507                                         pa->addr.p.tbl =
1508                                             pfr_attach_table(ruleset,
1509                                             pa->addr.v.tblname);
1510                                         if (pa->addr.p.tbl == NULL)
1511                                                 error = ENOMEM;
1512                                 }
1513
1514                         if (newrule->overload_tblname[0]) {
1515                                 if ((newrule->overload_tbl = pfr_attach_table(
1516                                     ruleset, newrule->overload_tblname)) ==
1517                                     NULL)
1518                                         error = EINVAL;
1519                                 else
1520                                         newrule->overload_tbl->pfrkt_flags |=
1521                                             PFR_TFLAG_ACTIVE;
1522                         }
1523
1524                         pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1525                         if (((((newrule->action == PF_NAT) ||
1526                             (newrule->action == PF_RDR) ||
1527                             (newrule->action == PF_BINAT) ||
1528                             (newrule->rt > PF_FASTROUTE)) &&
1529                             !newrule->anchor)) &&
1530                             (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1531                                 error = EINVAL;
1532
1533                         if (error) {
1534                                 pf_free_rule(newrule);
1535                                 PF_RULES_WUNLOCK();
1536                                 break;
1537                         }
1538
1539                         newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1540                         newrule->evaluations = 0;
1541                         newrule->packets[0] = newrule->packets[1] = 0;
1542                         newrule->bytes[0] = newrule->bytes[1] = 0;
1543                 }
1544                 pf_empty_pool(&V_pf_pabuf);
1545
1546                 if (pcr->action == PF_CHANGE_ADD_HEAD)
1547                         oldrule = TAILQ_FIRST(
1548                             ruleset->rules[rs_num].active.ptr);
1549                 else if (pcr->action == PF_CHANGE_ADD_TAIL)
1550                         oldrule = TAILQ_LAST(
1551                             ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1552                 else {
1553                         oldrule = TAILQ_FIRST(
1554                             ruleset->rules[rs_num].active.ptr);
1555                         while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1556                                 oldrule = TAILQ_NEXT(oldrule, entries);
1557                         if (oldrule == NULL) {
1558                                 if (newrule != NULL)
1559                                         pf_free_rule(newrule);
1560                                 PF_RULES_WUNLOCK();
1561                                 error = EINVAL;
1562                                 break;
1563                         }
1564                 }
1565
1566                 if (pcr->action == PF_CHANGE_REMOVE) {
1567                         pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1568                             oldrule);
1569                         ruleset->rules[rs_num].active.rcount--;
1570                 } else {
1571                         if (oldrule == NULL)
1572                                 TAILQ_INSERT_TAIL(
1573                                     ruleset->rules[rs_num].active.ptr,
1574                                     newrule, entries);
1575                         else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1576                             pcr->action == PF_CHANGE_ADD_BEFORE)
1577                                 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1578                         else
1579                                 TAILQ_INSERT_AFTER(
1580                                     ruleset->rules[rs_num].active.ptr,
1581                                     oldrule, newrule, entries);
1582                         ruleset->rules[rs_num].active.rcount++;
1583                 }
1584
1585                 nr = 0;
1586                 TAILQ_FOREACH(oldrule,
1587                     ruleset->rules[rs_num].active.ptr, entries)
1588                         oldrule->nr = nr++;
1589
1590                 ruleset->rules[rs_num].active.ticket++;
1591
1592                 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1593                 pf_remove_if_empty_ruleset(ruleset);
1594
1595                 PF_RULES_WUNLOCK();
1596                 break;
1597
1598 #undef ERROUT
1599 DIOCCHANGERULE_error:
1600                 PF_RULES_WUNLOCK();
1601                 if (newrule != NULL) {
1602                         counter_u64_free(newrule->states_cur);
1603                         counter_u64_free(newrule->states_tot);
1604                         counter_u64_free(newrule->src_nodes);
1605                         free(newrule, M_PFRULE);
1606                 }
1607                 if (kif != NULL)
1608                         free(kif, PFI_MTYPE);
1609                 break;
1610         }
1611
1612         case DIOCCLRSTATES: {
1613                 struct pf_state         *s;
1614                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1615                 u_int                    i, killed = 0;
1616
1617                 for (i = 0; i <= pf_hashmask; i++) {
1618                         struct pf_idhash *ih = &V_pf_idhash[i];
1619
1620 relock_DIOCCLRSTATES:
1621                         PF_HASHROW_LOCK(ih);
1622                         LIST_FOREACH(s, &ih->states, entry)
1623                                 if (!psk->psk_ifname[0] ||
1624                                     !strcmp(psk->psk_ifname,
1625                                     s->kif->pfik_name)) {
1626                                         /*
1627                                          * Don't send out individual
1628                                          * delete messages.
1629                                          */
1630                                         s->state_flags |= PFSTATE_NOSYNC;
1631                                         pf_unlink_state(s, PF_ENTER_LOCKED);
1632                                         killed++;
1633                                         goto relock_DIOCCLRSTATES;
1634                                 }
1635                         PF_HASHROW_UNLOCK(ih);
1636                 }
1637                 psk->psk_killed = killed;
1638                 if (pfsync_clear_states_ptr != NULL)
1639                         pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
1640                 break;
1641         }
1642
1643         case DIOCKILLSTATES: {
1644                 struct pf_state         *s;
1645                 struct pf_state_key     *sk;
1646                 struct pf_addr          *srcaddr, *dstaddr;
1647                 u_int16_t                srcport, dstport;
1648                 struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1649                 u_int                    i, killed = 0;
1650
1651                 if (psk->psk_pfcmp.id) {
1652                         if (psk->psk_pfcmp.creatorid == 0)
1653                                 psk->psk_pfcmp.creatorid = V_pf_status.hostid;
1654                         if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
1655                             psk->psk_pfcmp.creatorid))) {
1656                                 pf_unlink_state(s, PF_ENTER_LOCKED);
1657                                 psk->psk_killed = 1;
1658                         }
1659                         break;
1660                 }
1661
1662                 for (i = 0; i <= pf_hashmask; i++) {
1663                         struct pf_idhash *ih = &V_pf_idhash[i];
1664
1665 relock_DIOCKILLSTATES:
1666                         PF_HASHROW_LOCK(ih);
1667                         LIST_FOREACH(s, &ih->states, entry) {
1668                                 sk = s->key[PF_SK_WIRE];
1669                                 if (s->direction == PF_OUT) {
1670                                         srcaddr = &sk->addr[1];
1671                                         dstaddr = &sk->addr[0];
1672                                         srcport = sk->port[1];
1673                                         dstport = sk->port[0];
1674                                 } else {
1675                                         srcaddr = &sk->addr[0];
1676                                         dstaddr = &sk->addr[1];
1677                                         srcport = sk->port[0];
1678                                         dstport = sk->port[1];
1679                                 }
1680
1681                                 if ((!psk->psk_af || sk->af == psk->psk_af)
1682                                     && (!psk->psk_proto || psk->psk_proto ==
1683                                     sk->proto) &&
1684                                     PF_MATCHA(psk->psk_src.neg,
1685                                     &psk->psk_src.addr.v.a.addr,
1686                                     &psk->psk_src.addr.v.a.mask,
1687                                     srcaddr, sk->af) &&
1688                                     PF_MATCHA(psk->psk_dst.neg,
1689                                     &psk->psk_dst.addr.v.a.addr,
1690                                     &psk->psk_dst.addr.v.a.mask,
1691                                     dstaddr, sk->af) &&
1692                                     (psk->psk_src.port_op == 0 ||
1693                                     pf_match_port(psk->psk_src.port_op,
1694                                     psk->psk_src.port[0], psk->psk_src.port[1],
1695                                     srcport)) &&
1696                                     (psk->psk_dst.port_op == 0 ||
1697                                     pf_match_port(psk->psk_dst.port_op,
1698                                     psk->psk_dst.port[0], psk->psk_dst.port[1],
1699                                     dstport)) &&
1700                                     (!psk->psk_label[0] ||
1701                                     (s->rule.ptr->label[0] &&
1702                                     !strcmp(psk->psk_label,
1703                                     s->rule.ptr->label))) &&
1704                                     (!psk->psk_ifname[0] ||
1705                                     !strcmp(psk->psk_ifname,
1706                                     s->kif->pfik_name))) {
1707                                         pf_unlink_state(s, PF_ENTER_LOCKED);
1708                                         killed++;
1709                                         goto relock_DIOCKILLSTATES;
1710                                 }
1711                         }
1712                         PF_HASHROW_UNLOCK(ih);
1713                 }
1714                 psk->psk_killed = killed;
1715                 break;
1716         }
1717
1718         case DIOCADDSTATE: {
1719                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
1720                 struct pfsync_state     *sp = &ps->state;
1721
1722                 if (sp->timeout >= PFTM_MAX) {
1723                         error = EINVAL;
1724                         break;
1725                 }
1726                 if (pfsync_state_import_ptr != NULL) {
1727                         PF_RULES_RLOCK();
1728                         error = pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
1729                         PF_RULES_RUNLOCK();
1730                 } else
1731                         error = EOPNOTSUPP;
1732                 break;
1733         }
1734
1735         case DIOCGETSTATE: {
1736                 struct pfioc_state      *ps = (struct pfioc_state *)addr;
1737                 struct pf_state         *s;
1738
1739                 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
1740                 if (s == NULL) {
1741                         error = ENOENT;
1742                         break;
1743                 }
1744
1745                 pfsync_state_export(&ps->state, s);
1746                 PF_STATE_UNLOCK(s);
1747                 break;
1748         }
1749
1750         case DIOCGETSTATES: {
1751                 struct pfioc_states     *ps = (struct pfioc_states *)addr;
1752                 struct pf_state         *s;
1753                 struct pfsync_state     *pstore, *p;
1754                 int i, nr;
1755
1756                 if (ps->ps_len == 0) {
1757                         nr = uma_zone_get_cur(V_pf_state_z);
1758                         ps->ps_len = sizeof(struct pfsync_state) * nr;
1759                         break;
1760                 }
1761
1762                 p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK);
1763                 nr = 0;
1764
1765                 for (i = 0; i <= pf_hashmask; i++) {
1766                         struct pf_idhash *ih = &V_pf_idhash[i];
1767
1768                         PF_HASHROW_LOCK(ih);
1769                         LIST_FOREACH(s, &ih->states, entry) {
1770
1771                                 if (s->timeout == PFTM_UNLINKED)
1772                                         continue;
1773
1774                                 if ((nr+1) * sizeof(*p) > ps->ps_len) {
1775                                         PF_HASHROW_UNLOCK(ih);
1776                                         goto DIOCGETSTATES_full;
1777                                 }
1778                                 pfsync_state_export(p, s);
1779                                 p++;
1780                                 nr++;
1781                         }
1782                         PF_HASHROW_UNLOCK(ih);
1783                 }
1784 DIOCGETSTATES_full:
1785                 error = copyout(pstore, ps->ps_states,
1786                     sizeof(struct pfsync_state) * nr);
1787                 if (error) {
1788                         free(pstore, M_TEMP);
1789                         break;
1790                 }
1791                 ps->ps_len = sizeof(struct pfsync_state) * nr;
1792                 free(pstore, M_TEMP);
1793
1794                 break;
1795         }
1796
1797         case DIOCGETSTATUS: {
1798                 struct pf_status *s = (struct pf_status *)addr;
1799
1800                 PF_RULES_RLOCK();
1801                 s->running = V_pf_status.running;
1802                 s->since   = V_pf_status.since;
1803                 s->debug   = V_pf_status.debug;
1804                 s->hostid  = V_pf_status.hostid;
1805                 s->states  = V_pf_status.states;
1806                 s->src_nodes = V_pf_status.src_nodes;
1807
1808                 for (int i = 0; i < PFRES_MAX; i++)
1809                         s->counters[i] =
1810                             counter_u64_fetch(V_pf_status.counters[i]);
1811                 for (int i = 0; i < LCNT_MAX; i++)
1812                         s->lcounters[i] =
1813                             counter_u64_fetch(V_pf_status.lcounters[i]);
1814                 for (int i = 0; i < FCNT_MAX; i++)
1815                         s->fcounters[i] =
1816                             counter_u64_fetch(V_pf_status.fcounters[i]);
1817                 for (int i = 0; i < SCNT_MAX; i++)
1818                         s->scounters[i] =
1819                             counter_u64_fetch(V_pf_status.scounters[i]);
1820
1821                 bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
1822                 bcopy(V_pf_status.pf_chksum, s->pf_chksum,
1823                     PF_MD5_DIGEST_LENGTH);
1824
1825                 pfi_update_status(s->ifname, s);
1826                 PF_RULES_RUNLOCK();
1827                 break;
1828         }
1829
1830         case DIOCSETSTATUSIF: {
1831                 struct pfioc_if *pi = (struct pfioc_if *)addr;
1832
1833                 if (pi->ifname[0] == 0) {
1834                         bzero(V_pf_status.ifname, IFNAMSIZ);
1835                         break;
1836                 }
1837                 PF_RULES_WLOCK();
1838                 strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
1839                 PF_RULES_WUNLOCK();
1840                 break;
1841         }
1842
1843         case DIOCCLRSTATUS: {
1844                 PF_RULES_WLOCK();
1845                 for (int i = 0; i < PFRES_MAX; i++)
1846                         counter_u64_zero(V_pf_status.counters[i]);
1847                 for (int i = 0; i < FCNT_MAX; i++)
1848                         counter_u64_zero(V_pf_status.fcounters[i]);
1849                 for (int i = 0; i < SCNT_MAX; i++)
1850                         counter_u64_zero(V_pf_status.scounters[i]);
1851                 V_pf_status.since = time_second;
1852                 if (*V_pf_status.ifname)
1853                         pfi_update_status(V_pf_status.ifname, NULL);
1854                 PF_RULES_WUNLOCK();
1855                 break;
1856         }
1857
1858         case DIOCNATLOOK: {
1859                 struct pfioc_natlook    *pnl = (struct pfioc_natlook *)addr;
1860                 struct pf_state_key     *sk;
1861                 struct pf_state         *state;
1862                 struct pf_state_key_cmp  key;
1863                 int                      m = 0, direction = pnl->direction;
1864                 int                      sidx, didx;
1865
1866                 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
1867                 sidx = (direction == PF_IN) ? 1 : 0;
1868                 didx = (direction == PF_IN) ? 0 : 1;
1869
1870                 if (!pnl->proto ||
1871                     PF_AZERO(&pnl->saddr, pnl->af) ||
1872                     PF_AZERO(&pnl->daddr, pnl->af) ||
1873                     ((pnl->proto == IPPROTO_TCP ||
1874                     pnl->proto == IPPROTO_UDP) &&
1875                     (!pnl->dport || !pnl->sport)))
1876                         error = EINVAL;
1877                 else {
1878                         bzero(&key, sizeof(key));
1879                         key.af = pnl->af;
1880                         key.proto = pnl->proto;
1881                         PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1882                         key.port[sidx] = pnl->sport;
1883                         PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1884                         key.port[didx] = pnl->dport;
1885
1886                         state = pf_find_state_all(&key, direction, &m);
1887
1888                         if (m > 1)
1889                                 error = E2BIG;  /* more than one state */
1890                         else if (state != NULL) {
1891                                 /* XXXGL: not locked read */
1892                                 sk = state->key[sidx];
1893                                 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1894                                 pnl->rsport = sk->port[sidx];
1895                                 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1896                                 pnl->rdport = sk->port[didx];
1897                         } else
1898                                 error = ENOENT;
1899                 }
1900                 break;
1901         }
1902
1903         case DIOCSETTIMEOUT: {
1904                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
1905                 int              old;
1906
1907                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1908                     pt->seconds < 0) {
1909                         error = EINVAL;
1910                         break;
1911                 }
1912                 PF_RULES_WLOCK();
1913                 old = V_pf_default_rule.timeout[pt->timeout];
1914                 if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1915                         pt->seconds = 1;
1916                 V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
1917                 if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1918                         wakeup(pf_purge_thread);
1919                 pt->seconds = old;
1920                 PF_RULES_WUNLOCK();
1921                 break;
1922         }
1923
1924         case DIOCGETTIMEOUT: {
1925                 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
1926
1927                 if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1928                         error = EINVAL;
1929                         break;
1930                 }
1931                 PF_RULES_RLOCK();
1932                 pt->seconds = V_pf_default_rule.timeout[pt->timeout];
1933                 PF_RULES_RUNLOCK();
1934                 break;
1935         }
1936
1937         case DIOCGETLIMIT: {
1938                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
1939
1940                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1941                         error = EINVAL;
1942                         break;
1943                 }
1944                 PF_RULES_RLOCK();
1945                 pl->limit = V_pf_limits[pl->index].limit;
1946                 PF_RULES_RUNLOCK();
1947                 break;
1948         }
1949
1950         case DIOCSETLIMIT: {
1951                 struct pfioc_limit      *pl = (struct pfioc_limit *)addr;
1952                 int                      old_limit;
1953
1954                 PF_RULES_WLOCK();
1955                 if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1956                     V_pf_limits[pl->index].zone == NULL) {
1957                         PF_RULES_WUNLOCK();
1958                         error = EINVAL;
1959                         break;
1960                 }
1961                 uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
1962                 old_limit = V_pf_limits[pl->index].limit;
1963                 V_pf_limits[pl->index].limit = pl->limit;
1964                 pl->limit = old_limit;
1965                 PF_RULES_WUNLOCK();
1966                 break;
1967         }
1968
1969         case DIOCSETDEBUG: {
1970                 u_int32_t       *level = (u_int32_t *)addr;
1971
1972                 PF_RULES_WLOCK();
1973                 V_pf_status.debug = *level;
1974                 PF_RULES_WUNLOCK();
1975                 break;
1976         }
1977
1978         case DIOCCLRRULECTRS: {
1979                 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1980                 struct pf_ruleset       *ruleset = &pf_main_ruleset;
1981                 struct pf_rule          *rule;
1982
1983                 PF_RULES_WLOCK();
1984                 TAILQ_FOREACH(rule,
1985                     ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1986                         rule->evaluations = 0;
1987                         rule->packets[0] = rule->packets[1] = 0;
1988                         rule->bytes[0] = rule->bytes[1] = 0;
1989                 }
1990                 PF_RULES_WUNLOCK();
1991                 break;
1992         }
1993
1994         case DIOCGIFSPEED: {
1995                 struct pf_ifspeed       *psp = (struct pf_ifspeed *)addr;
1996                 struct pf_ifspeed       ps;
1997                 struct ifnet            *ifp;
1998
1999                 if (psp->ifname[0] != 0) {
2000                         /* Can we completely trust user-land? */
2001                         strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2002                         ifp = ifunit(ps.ifname);
2003                         if (ifp != NULL)
2004                                 psp->baudrate = ifp->if_baudrate;
2005                         else
2006                                 error = EINVAL;
2007                 } else
2008                         error = EINVAL;
2009                 break;
2010         }
2011
2012 #ifdef ALTQ
2013         case DIOCSTARTALTQ: {
2014                 struct pf_altq          *altq;
2015
2016                 PF_RULES_WLOCK();
2017                 /* enable all altq interfaces on active list */
2018                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2019                         if (altq->qname[0] == 0 && (altq->local_flags &
2020                             PFALTQ_FLAG_IF_REMOVED) == 0) {
2021                                 error = pf_enable_altq(altq);
2022                                 if (error != 0)
2023                                         break;
2024                         }
2025                 }
2026                 if (error == 0)
2027                         V_pf_altq_running = 1;
2028                 PF_RULES_WUNLOCK();
2029                 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2030                 break;
2031         }
2032
2033         case DIOCSTOPALTQ: {
2034                 struct pf_altq          *altq;
2035
2036                 PF_RULES_WLOCK();
2037                 /* disable all altq interfaces on active list */
2038                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2039                         if (altq->qname[0] == 0 && (altq->local_flags &
2040                             PFALTQ_FLAG_IF_REMOVED) == 0) {
2041                                 error = pf_disable_altq(altq);
2042                                 if (error != 0)
2043                                         break;
2044                         }
2045                 }
2046                 if (error == 0)
2047                         V_pf_altq_running = 0;
2048                 PF_RULES_WUNLOCK();
2049                 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2050                 break;
2051         }
2052
2053         case DIOCADDALTQ: {
2054                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2055                 struct pf_altq          *altq, *a;
2056                 struct ifnet            *ifp;
2057
2058                 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK);
2059                 bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2060                 altq->local_flags = 0;
2061
2062                 PF_RULES_WLOCK();
2063                 if (pa->ticket != V_ticket_altqs_inactive) {
2064                         PF_RULES_WUNLOCK();
2065                         free(altq, M_PFALTQ);
2066                         error = EBUSY;
2067                         break;
2068                 }
2069
2070                 /*
2071                  * if this is for a queue, find the discipline and
2072                  * copy the necessary fields
2073                  */
2074                 if (altq->qname[0] != 0) {
2075                         if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2076                                 PF_RULES_WUNLOCK();
2077                                 error = EBUSY;
2078                                 free(altq, M_PFALTQ);
2079                                 break;
2080                         }
2081                         altq->altq_disc = NULL;
2082                         TAILQ_FOREACH(a, V_pf_altqs_inactive, entries) {
2083                                 if (strncmp(a->ifname, altq->ifname,
2084                                     IFNAMSIZ) == 0 && a->qname[0] == 0) {
2085                                         altq->altq_disc = a->altq_disc;
2086                                         break;
2087                                 }
2088                         }
2089                 }
2090
2091                 if ((ifp = ifunit(altq->ifname)) == NULL)
2092                         altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2093                 else
2094                         error = altq_add(altq);
2095
2096                 if (error) {
2097                         PF_RULES_WUNLOCK();
2098                         free(altq, M_PFALTQ);
2099                         break;
2100                 }
2101
2102                 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2103                 bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2104                 PF_RULES_WUNLOCK();
2105                 break;
2106         }
2107
2108         case DIOCGETALTQS: {
2109                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2110                 struct pf_altq          *altq;
2111
2112                 PF_RULES_RLOCK();
2113                 pa->nr = 0;
2114                 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2115                         pa->nr++;
2116                 pa->ticket = V_ticket_altqs_active;
2117                 PF_RULES_RUNLOCK();
2118                 break;
2119         }
2120
2121         case DIOCGETALTQ: {
2122                 struct pfioc_altq       *pa = (struct pfioc_altq *)addr;
2123                 struct pf_altq          *altq;
2124                 u_int32_t                nr;
2125
2126                 PF_RULES_RLOCK();
2127                 if (pa->ticket != V_ticket_altqs_active) {
2128                         PF_RULES_RUNLOCK();
2129                         error = EBUSY;
2130                         break;
2131                 }
2132                 nr = 0;
2133                 altq = TAILQ_FIRST(V_pf_altqs_active);
2134                 while ((altq != NULL) && (nr < pa->nr)) {
2135                         altq = TAILQ_NEXT(altq, entries);
2136                         nr++;
2137                 }
2138                 if (altq == NULL) {
2139                         PF_RULES_RUNLOCK();
2140                         error = EBUSY;
2141                         break;
2142                 }
2143                 bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2144                 PF_RULES_RUNLOCK();
2145                 break;
2146         }
2147
2148         case DIOCCHANGEALTQ:
2149                 /* CHANGEALTQ not supported yet! */
2150                 error = ENODEV;
2151                 break;
2152
2153         case DIOCGETQSTATS: {
2154                 struct pfioc_qstats     *pq = (struct pfioc_qstats *)addr;
2155                 struct pf_altq          *altq;
2156                 u_int32_t                nr;
2157                 int                      nbytes;
2158
2159                 PF_RULES_RLOCK();
2160                 if (pq->ticket != V_ticket_altqs_active) {
2161                         PF_RULES_RUNLOCK();
2162                         error = EBUSY;
2163                         break;
2164                 }
2165                 nbytes = pq->nbytes;
2166                 nr = 0;
2167                 altq = TAILQ_FIRST(V_pf_altqs_active);
2168                 while ((altq != NULL) && (nr < pq->nr)) {
2169                         altq = TAILQ_NEXT(altq, entries);
2170                         nr++;
2171                 }
2172                 if (altq == NULL) {
2173                         PF_RULES_RUNLOCK();
2174                         error = EBUSY;
2175                         break;
2176                 }
2177
2178                 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2179                         PF_RULES_RUNLOCK();
2180                         error = ENXIO;
2181                         break;
2182                 }
2183                 PF_RULES_RUNLOCK();
2184                 error = altq_getqstats(altq, pq->buf, &nbytes);
2185                 if (error == 0) {
2186                         pq->scheduler = altq->scheduler;
2187                         pq->nbytes = nbytes;
2188                 }
2189                 break;
2190         }
2191 #endif /* ALTQ */
2192
2193         case DIOCBEGINADDRS: {
2194                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2195
2196                 PF_RULES_WLOCK();
2197                 pf_empty_pool(&V_pf_pabuf);
2198                 pp->ticket = ++V_ticket_pabuf;
2199                 PF_RULES_WUNLOCK();
2200                 break;
2201         }
2202
2203         case DIOCADDADDR: {
2204                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2205                 struct pf_pooladdr      *pa;
2206                 struct pfi_kif          *kif = NULL;
2207
2208 #ifndef INET
2209                 if (pp->af == AF_INET) {
2210                         error = EAFNOSUPPORT;
2211                         break;
2212                 }
2213 #endif /* INET */
2214 #ifndef INET6
2215                 if (pp->af == AF_INET6) {
2216                         error = EAFNOSUPPORT;
2217                         break;
2218                 }
2219 #endif /* INET6 */
2220                 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2221                     pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2222                     pp->addr.addr.type != PF_ADDR_TABLE) {
2223                         error = EINVAL;
2224                         break;
2225                 }
2226                 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2227                 bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2228                 if (pa->ifname[0])
2229                         kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2230                 PF_RULES_WLOCK();
2231                 if (pp->ticket != V_ticket_pabuf) {
2232                         PF_RULES_WUNLOCK();
2233                         if (pa->ifname[0])
2234                                 free(kif, PFI_MTYPE);
2235                         free(pa, M_PFRULE);
2236                         error = EBUSY;
2237                         break;
2238                 }
2239                 if (pa->ifname[0]) {
2240                         pa->kif = pfi_kif_attach(kif, pa->ifname);
2241                         pfi_kif_ref(pa->kif);
2242                 } else
2243                         pa->kif = NULL;
2244                 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2245                     pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2246                         if (pa->ifname[0])
2247                                 pfi_kif_unref(pa->kif);
2248                         PF_RULES_WUNLOCK();
2249                         free(pa, M_PFRULE);
2250                         break;
2251                 }
2252                 TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2253                 PF_RULES_WUNLOCK();
2254                 break;
2255         }
2256
2257         case DIOCGETADDRS: {
2258                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2259                 struct pf_pool          *pool;
2260                 struct pf_pooladdr      *pa;
2261
2262                 PF_RULES_RLOCK();
2263                 pp->nr = 0;
2264                 pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2265                     pp->r_num, 0, 1, 0);
2266                 if (pool == NULL) {
2267                         PF_RULES_RUNLOCK();
2268                         error = EBUSY;
2269                         break;
2270                 }
2271                 TAILQ_FOREACH(pa, &pool->list, entries)
2272                         pp->nr++;
2273                 PF_RULES_RUNLOCK();
2274                 break;
2275         }
2276
2277         case DIOCGETADDR: {
2278                 struct pfioc_pooladdr   *pp = (struct pfioc_pooladdr *)addr;
2279                 struct pf_pool          *pool;
2280                 struct pf_pooladdr      *pa;
2281                 u_int32_t                nr = 0;
2282
2283                 PF_RULES_RLOCK();
2284                 pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2285                     pp->r_num, 0, 1, 1);
2286                 if (pool == NULL) {
2287                         PF_RULES_RUNLOCK();
2288                         error = EBUSY;
2289                         break;
2290                 }
2291                 pa = TAILQ_FIRST(&pool->list);
2292                 while ((pa != NULL) && (nr < pp->nr)) {
2293                         pa = TAILQ_NEXT(pa, entries);
2294                         nr++;
2295                 }
2296                 if (pa == NULL) {
2297                         PF_RULES_RUNLOCK();
2298                         error = EBUSY;
2299                         break;
2300                 }
2301                 bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2302                 pf_addr_copyout(&pp->addr.addr);
2303                 PF_RULES_RUNLOCK();
2304                 break;
2305         }
2306
2307         case DIOCCHANGEADDR: {
2308                 struct pfioc_pooladdr   *pca = (struct pfioc_pooladdr *)addr;
2309                 struct pf_pool          *pool;
2310                 struct pf_pooladdr      *oldpa = NULL, *newpa = NULL;
2311                 struct pf_ruleset       *ruleset;
2312                 struct pfi_kif          *kif = NULL;
2313
2314                 if (pca->action < PF_CHANGE_ADD_HEAD ||
2315                     pca->action > PF_CHANGE_REMOVE) {
2316                         error = EINVAL;
2317                         break;
2318                 }
2319                 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2320                     pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2321                     pca->addr.addr.type != PF_ADDR_TABLE) {
2322                         error = EINVAL;
2323                         break;
2324                 }
2325
2326                 if (pca->action != PF_CHANGE_REMOVE) {
2327 #ifndef INET
2328                         if (pca->af == AF_INET) {
2329                                 error = EAFNOSUPPORT;
2330                                 break;
2331                         }
2332 #endif /* INET */
2333 #ifndef INET6
2334                         if (pca->af == AF_INET6) {
2335                                 error = EAFNOSUPPORT;
2336                                 break;
2337                         }
2338 #endif /* INET6 */
2339                         newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2340                         bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2341                         if (newpa->ifname[0])
2342                                 kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2343                         newpa->kif = NULL;
2344                 }
2345
2346 #define ERROUT(x)       { error = (x); goto DIOCCHANGEADDR_error; }
2347                 PF_RULES_WLOCK();
2348                 ruleset = pf_find_ruleset(pca->anchor);
2349                 if (ruleset == NULL)
2350                         ERROUT(EBUSY);
2351
2352                 pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2353                     pca->r_num, pca->r_last, 1, 1);
2354                 if (pool == NULL)
2355                         ERROUT(EBUSY);
2356
2357                 if (pca->action != PF_CHANGE_REMOVE) {
2358                         if (newpa->ifname[0]) {
2359                                 newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2360                                 pfi_kif_ref(newpa->kif);
2361                                 kif = NULL;
2362                         }
2363
2364                         switch (newpa->addr.type) {
2365                         case PF_ADDR_DYNIFTL:
2366                                 error = pfi_dynaddr_setup(&newpa->addr,
2367                                     pca->af);
2368                                 break;
2369                         case PF_ADDR_TABLE:
2370                                 newpa->addr.p.tbl = pfr_attach_table(ruleset,
2371                                     newpa->addr.v.tblname);
2372                                 if (newpa->addr.p.tbl == NULL)
2373                                         error = ENOMEM;
2374                                 break;
2375                         }
2376                         if (error)
2377                                 goto DIOCCHANGEADDR_error;
2378                 }
2379
2380                 switch (pca->action) {
2381                 case PF_CHANGE_ADD_HEAD:
2382                         oldpa = TAILQ_FIRST(&pool->list);
2383                         break;
2384                 case PF_CHANGE_ADD_TAIL:
2385                         oldpa = TAILQ_LAST(&pool->list, pf_palist);
2386                         break;
2387                 default:
2388                         oldpa = TAILQ_FIRST(&pool->list);
2389                         for (int i = 0; oldpa && i < pca->nr; i++)
2390                                 oldpa = TAILQ_NEXT(oldpa, entries);
2391
2392                         if (oldpa == NULL)
2393                                 ERROUT(EINVAL);
2394                 }
2395
2396                 if (pca->action == PF_CHANGE_REMOVE) {
2397                         TAILQ_REMOVE(&pool->list, oldpa, entries);
2398                         switch (oldpa->addr.type) {
2399                         case PF_ADDR_DYNIFTL:
2400                                 pfi_dynaddr_remove(oldpa->addr.p.dyn);
2401                                 break;
2402                         case PF_ADDR_TABLE:
2403                                 pfr_detach_table(oldpa->addr.p.tbl);
2404                                 break;
2405                         }
2406                         if (oldpa->kif)
2407                                 pfi_kif_unref(oldpa->kif);
2408                         free(oldpa, M_PFRULE);
2409                 } else {
2410                         if (oldpa == NULL)
2411                                 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2412                         else if (pca->action == PF_CHANGE_ADD_HEAD ||
2413                             pca->action == PF_CHANGE_ADD_BEFORE)
2414                                 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2415                         else
2416                                 TAILQ_INSERT_AFTER(&pool->list, oldpa,
2417                                     newpa, entries);
2418                 }
2419
2420                 pool->cur = TAILQ_FIRST(&pool->list);
2421                 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
2422                 PF_RULES_WUNLOCK();
2423                 break;
2424
2425 #undef ERROUT
2426 DIOCCHANGEADDR_error:
2427                 if (newpa->kif)
2428                         pfi_kif_unref(newpa->kif);
2429                 PF_RULES_WUNLOCK();
2430                 if (newpa != NULL)
2431                         free(newpa, M_PFRULE);
2432                 if (kif != NULL)
2433                         free(kif, PFI_MTYPE);
2434                 break;
2435         }
2436
2437         case DIOCGETRULESETS: {
2438                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
2439                 struct pf_ruleset       *ruleset;
2440                 struct pf_anchor        *anchor;
2441
2442                 PF_RULES_RLOCK();
2443                 pr->path[sizeof(pr->path) - 1] = 0;
2444                 if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2445                         PF_RULES_RUNLOCK();
2446                         error = ENOENT;
2447                         break;
2448                 }
2449                 pr->nr = 0;
2450                 if (ruleset->anchor == NULL) {
2451                         /* XXX kludge for pf_main_ruleset */
2452                         RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2453                                 if (anchor->parent == NULL)
2454                                         pr->nr++;
2455                 } else {
2456                         RB_FOREACH(anchor, pf_anchor_node,
2457                             &ruleset->anchor->children)
2458                                 pr->nr++;
2459                 }
2460                 PF_RULES_RUNLOCK();
2461                 break;
2462         }
2463
2464         case DIOCGETRULESET: {
2465                 struct pfioc_ruleset    *pr = (struct pfioc_ruleset *)addr;
2466                 struct pf_ruleset       *ruleset;
2467                 struct pf_anchor        *anchor;
2468                 u_int32_t                nr = 0;
2469
2470                 PF_RULES_RLOCK();
2471                 pr->path[sizeof(pr->path) - 1] = 0;
2472                 if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2473                         PF_RULES_RUNLOCK();
2474                         error = ENOENT;
2475                         break;
2476                 }
2477                 pr->name[0] = 0;
2478                 if (ruleset->anchor == NULL) {
2479                         /* XXX kludge for pf_main_ruleset */
2480                         RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2481                                 if (anchor->parent == NULL && nr++ == pr->nr) {
2482                                         strlcpy(pr->name, anchor->name,
2483                                             sizeof(pr->name));
2484                                         break;
2485                                 }
2486                 } else {
2487                         RB_FOREACH(anchor, pf_anchor_node,
2488                             &ruleset->anchor->children)
2489                                 if (nr++ == pr->nr) {
2490                                         strlcpy(pr->name, anchor->name,
2491                                             sizeof(pr->name));
2492                                         break;
2493                                 }
2494                 }
2495                 if (!pr->name[0])
2496                         error = EBUSY;
2497                 PF_RULES_RUNLOCK();
2498                 break;
2499         }
2500
2501         case DIOCRCLRTABLES: {
2502                 struct pfioc_table *io = (struct pfioc_table *)addr;
2503
2504                 if (io->pfrio_esize != 0) {
2505                         error = ENODEV;
2506                         break;
2507                 }
2508                 PF_RULES_WLOCK();
2509                 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2510                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
2511                 PF_RULES_WUNLOCK();
2512                 break;
2513         }
2514
2515         case DIOCRADDTABLES: {
2516                 struct pfioc_table *io = (struct pfioc_table *)addr;
2517                 struct pfr_table *pfrts;
2518                 size_t totlen;
2519
2520                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2521                         error = ENODEV;
2522                         break;
2523                 }
2524                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2525                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2526                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2527                 if (error) {
2528                         free(pfrts, M_TEMP);
2529                         break;
2530                 }
2531                 PF_RULES_WLOCK();
2532                 error = pfr_add_tables(pfrts, io->pfrio_size,
2533                     &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2534                 PF_RULES_WUNLOCK();
2535                 free(pfrts, M_TEMP);
2536                 break;
2537         }
2538
2539         case DIOCRDELTABLES: {
2540                 struct pfioc_table *io = (struct pfioc_table *)addr;
2541                 struct pfr_table *pfrts;
2542                 size_t totlen;
2543
2544                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2545                         error = ENODEV;
2546                         break;
2547                 }
2548                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2549                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2550                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2551                 if (error) {
2552                         free(pfrts, M_TEMP);
2553                         break;
2554                 }
2555                 PF_RULES_WLOCK();
2556                 error = pfr_del_tables(pfrts, io->pfrio_size,
2557                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2558                 PF_RULES_WUNLOCK();
2559                 free(pfrts, M_TEMP);
2560                 break;
2561         }
2562
2563         case DIOCRGETTABLES: {
2564                 struct pfioc_table *io = (struct pfioc_table *)addr;
2565                 struct pfr_table *pfrts;
2566                 size_t totlen;
2567
2568                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2569                         error = ENODEV;
2570                         break;
2571                 }
2572                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2573                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2574                 PF_RULES_RLOCK();
2575                 error = pfr_get_tables(&io->pfrio_table, pfrts,
2576                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2577                 PF_RULES_RUNLOCK();
2578                 if (error == 0)
2579                         error = copyout(pfrts, io->pfrio_buffer, totlen);
2580                 free(pfrts, M_TEMP);
2581                 break;
2582         }
2583
2584         case DIOCRGETTSTATS: {
2585                 struct pfioc_table *io = (struct pfioc_table *)addr;
2586                 struct pfr_tstats *pfrtstats;
2587                 size_t totlen;
2588
2589                 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2590                         error = ENODEV;
2591                         break;
2592                 }
2593                 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
2594                 pfrtstats = malloc(totlen, M_TEMP, M_WAITOK);
2595                 PF_RULES_WLOCK();
2596                 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
2597                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2598                 PF_RULES_WUNLOCK();
2599                 if (error == 0)
2600                         error = copyout(pfrtstats, io->pfrio_buffer, totlen);
2601                 free(pfrtstats, M_TEMP);
2602                 break;
2603         }
2604
2605         case DIOCRCLRTSTATS: {
2606                 struct pfioc_table *io = (struct pfioc_table *)addr;
2607                 struct pfr_table *pfrts;
2608                 size_t totlen;
2609
2610                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2611                         error = ENODEV;
2612                         break;
2613                 }
2614                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2615                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2616                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2617                 if (error) {
2618                         free(pfrts, M_TEMP);
2619                         break;
2620                 }
2621                 PF_RULES_WLOCK();
2622                 error = pfr_clr_tstats(pfrts, io->pfrio_size,
2623                     &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2624                 PF_RULES_WUNLOCK();
2625                 free(pfrts, M_TEMP);
2626                 break;
2627         }
2628
2629         case DIOCRSETTFLAGS: {
2630                 struct pfioc_table *io = (struct pfioc_table *)addr;
2631                 struct pfr_table *pfrts;
2632                 size_t totlen;
2633
2634                 if (io->pfrio_esize != sizeof(struct pfr_table)) {
2635                         error = ENODEV;
2636                         break;
2637                 }
2638                 totlen = io->pfrio_size * sizeof(struct pfr_table);
2639                 pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2640                 error = copyin(io->pfrio_buffer, pfrts, totlen);
2641                 if (error) {
2642                         free(pfrts, M_TEMP);
2643                         break;
2644                 }
2645                 PF_RULES_WLOCK();
2646                 error = pfr_set_tflags(pfrts, io->pfrio_size,
2647                     io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2648                     &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2649                 PF_RULES_WUNLOCK();
2650                 free(pfrts, M_TEMP);
2651                 break;
2652         }
2653
2654         case DIOCRCLRADDRS: {
2655                 struct pfioc_table *io = (struct pfioc_table *)addr;
2656
2657                 if (io->pfrio_esize != 0) {
2658                         error = ENODEV;
2659                         break;
2660                 }
2661                 PF_RULES_WLOCK();
2662                 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2663                     io->pfrio_flags | PFR_FLAG_USERIOCTL);
2664                 PF_RULES_WUNLOCK();
2665                 break;
2666         }
2667
2668         case DIOCRADDADDRS: {
2669                 struct pfioc_table *io = (struct pfioc_table *)addr;
2670                 struct pfr_addr *pfras;
2671                 size_t totlen;
2672
2673                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2674                         error = ENODEV;
2675                         break;
2676                 }
2677                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2678                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2679                 error = copyin(io->pfrio_buffer, pfras, totlen);
2680                 if (error) {
2681                         free(pfras, M_TEMP);
2682                         break;
2683                 }
2684                 PF_RULES_WLOCK();
2685                 error = pfr_add_addrs(&io->pfrio_table, pfras,
2686                     io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2687                     PFR_FLAG_USERIOCTL);
2688                 PF_RULES_WUNLOCK();
2689                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2690                         error = copyout(pfras, io->pfrio_buffer, totlen);
2691                 free(pfras, M_TEMP);
2692                 break;
2693         }
2694
2695         case DIOCRDELADDRS: {
2696                 struct pfioc_table *io = (struct pfioc_table *)addr;
2697                 struct pfr_addr *pfras;
2698                 size_t totlen;
2699
2700                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2701                         error = ENODEV;
2702                         break;
2703                 }
2704                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2705                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2706                 error = copyin(io->pfrio_buffer, pfras, totlen);
2707                 if (error) {
2708                         free(pfras, M_TEMP);
2709                         break;
2710                 }
2711                 PF_RULES_WLOCK();
2712                 error = pfr_del_addrs(&io->pfrio_table, pfras,
2713                     io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2714                     PFR_FLAG_USERIOCTL);
2715                 PF_RULES_WUNLOCK();
2716                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2717                         error = copyout(pfras, io->pfrio_buffer, totlen);
2718                 free(pfras, M_TEMP);
2719                 break;
2720         }
2721
2722         case DIOCRSETADDRS: {
2723                 struct pfioc_table *io = (struct pfioc_table *)addr;
2724                 struct pfr_addr *pfras;
2725                 size_t totlen, count;
2726
2727                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2728                         error = ENODEV;
2729                         break;
2730                 }
2731                 count = max(io->pfrio_size, io->pfrio_size2);
2732                 totlen = count * sizeof(struct pfr_addr);
2733                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2734                 error = copyin(io->pfrio_buffer, pfras, totlen);
2735                 if (error) {
2736                         free(pfras, M_TEMP);
2737                         break;
2738                 }
2739                 PF_RULES_WLOCK();
2740                 error = pfr_set_addrs(&io->pfrio_table, pfras,
2741                     io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2742                     &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2743                     PFR_FLAG_USERIOCTL, 0);
2744                 PF_RULES_WUNLOCK();
2745                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2746                         error = copyout(pfras, io->pfrio_buffer, totlen);
2747                 free(pfras, M_TEMP);
2748                 break;
2749         }
2750
2751         case DIOCRGETADDRS: {
2752                 struct pfioc_table *io = (struct pfioc_table *)addr;
2753                 struct pfr_addr *pfras;
2754                 size_t totlen;
2755
2756                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2757                         error = ENODEV;
2758                         break;
2759                 }
2760                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2761                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2762                 PF_RULES_RLOCK();
2763                 error = pfr_get_addrs(&io->pfrio_table, pfras,
2764                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2765                 PF_RULES_RUNLOCK();
2766                 if (error == 0)
2767                         error = copyout(pfras, io->pfrio_buffer, totlen);
2768                 free(pfras, M_TEMP);
2769                 break;
2770         }
2771
2772         case DIOCRGETASTATS: {
2773                 struct pfioc_table *io = (struct pfioc_table *)addr;
2774                 struct pfr_astats *pfrastats;
2775                 size_t totlen;
2776
2777                 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2778                         error = ENODEV;
2779                         break;
2780                 }
2781                 totlen = io->pfrio_size * sizeof(struct pfr_astats);
2782                 pfrastats = malloc(totlen, M_TEMP, M_WAITOK);
2783                 PF_RULES_RLOCK();
2784                 error = pfr_get_astats(&io->pfrio_table, pfrastats,
2785                     &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2786                 PF_RULES_RUNLOCK();
2787                 if (error == 0)
2788                         error = copyout(pfrastats, io->pfrio_buffer, totlen);
2789                 free(pfrastats, M_TEMP);
2790                 break;
2791         }
2792
2793         case DIOCRCLRASTATS: {
2794                 struct pfioc_table *io = (struct pfioc_table *)addr;
2795                 struct pfr_addr *pfras;
2796                 size_t totlen;
2797
2798                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2799                         error = ENODEV;
2800                         break;
2801                 }
2802                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2803                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2804                 error = copyin(io->pfrio_buffer, pfras, totlen);
2805                 if (error) {
2806                         free(pfras, M_TEMP);
2807                         break;
2808                 }
2809                 PF_RULES_WLOCK();
2810                 error = pfr_clr_astats(&io->pfrio_table, pfras,
2811                     io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2812                     PFR_FLAG_USERIOCTL);
2813                 PF_RULES_WUNLOCK();
2814                 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2815                         error = copyout(pfras, io->pfrio_buffer, totlen);
2816                 free(pfras, M_TEMP);
2817                 break;
2818         }
2819
2820         case DIOCRTSTADDRS: {
2821                 struct pfioc_table *io = (struct pfioc_table *)addr;
2822                 struct pfr_addr *pfras;
2823                 size_t totlen;
2824
2825                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2826                         error = ENODEV;
2827                         break;
2828                 }
2829                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2830                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2831                 error = copyin(io->pfrio_buffer, pfras, totlen);
2832                 if (error) {
2833                         free(pfras, M_TEMP);
2834                         break;
2835                 }
2836                 PF_RULES_RLOCK();
2837                 error = pfr_tst_addrs(&io->pfrio_table, pfras,
2838                     io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2839                     PFR_FLAG_USERIOCTL);
2840                 PF_RULES_RUNLOCK();
2841                 if (error == 0)
2842                         error = copyout(pfras, io->pfrio_buffer, totlen);
2843                 free(pfras, M_TEMP);
2844                 break;
2845         }
2846
2847         case DIOCRINADEFINE: {
2848                 struct pfioc_table *io = (struct pfioc_table *)addr;
2849                 struct pfr_addr *pfras;
2850                 size_t totlen;
2851
2852                 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2853                         error = ENODEV;
2854                         break;
2855                 }
2856                 totlen = io->pfrio_size * sizeof(struct pfr_addr);
2857                 pfras = malloc(totlen, M_TEMP, M_WAITOK);
2858                 error = copyin(io->pfrio_buffer, pfras, totlen);
2859                 if (error) {
2860                         free(pfras, M_TEMP);
2861                         break;
2862                 }
2863                 PF_RULES_WLOCK();
2864                 error = pfr_ina_define(&io->pfrio_table, pfras,
2865                     io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2866                     io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2867                 PF_RULES_WUNLOCK();
2868                 free(pfras, M_TEMP);
2869                 break;
2870         }
2871
2872         case DIOCOSFPADD: {
2873                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2874                 PF_RULES_WLOCK();
2875                 error = pf_osfp_add(io);
2876                 PF_RULES_WUNLOCK();
2877                 break;
2878         }
2879
2880         case DIOCOSFPGET: {
2881                 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2882                 PF_RULES_RLOCK();
2883                 error = pf_osfp_get(io);
2884                 PF_RULES_RUNLOCK();
2885                 break;
2886         }
2887
2888         case DIOCXBEGIN: {
2889                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2890                 struct pfioc_trans_e    *ioes, *ioe;
2891                 size_t                   totlen;
2892                 int                      i;
2893
2894                 if (io->esize != sizeof(*ioe)) {
2895                         error = ENODEV;
2896                         break;
2897                 }
2898                 totlen = sizeof(struct pfioc_trans_e) * io->size;
2899                 ioes = malloc(totlen, M_TEMP, M_WAITOK);
2900                 error = copyin(io->array, ioes, totlen);
2901                 if (error) {
2902                         free(ioes, M_TEMP);
2903                         break;
2904                 }
2905                 PF_RULES_WLOCK();
2906                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2907                         switch (ioe->rs_num) {
2908 #ifdef ALTQ
2909                         case PF_RULESET_ALTQ:
2910                                 if (ioe->anchor[0]) {
2911                                         PF_RULES_WUNLOCK();
2912                                         free(ioes, M_TEMP);
2913                                         error = EINVAL;
2914                                         goto fail;
2915                                 }
2916                                 if ((error = pf_begin_altq(&ioe->ticket))) {
2917                                         PF_RULES_WUNLOCK();
2918                                         free(ioes, M_TEMP);
2919                                         goto fail;
2920                                 }
2921                                 break;
2922 #endif /* ALTQ */
2923                         case PF_RULESET_TABLE:
2924                             {
2925                                 struct pfr_table table;
2926
2927                                 bzero(&table, sizeof(table));
2928                                 strlcpy(table.pfrt_anchor, ioe->anchor,
2929                                     sizeof(table.pfrt_anchor));
2930                                 if ((error = pfr_ina_begin(&table,
2931                                     &ioe->ticket, NULL, 0))) {
2932                                         PF_RULES_WUNLOCK();
2933                                         free(ioes, M_TEMP);
2934                                         goto fail;
2935                                 }
2936                                 break;
2937                             }
2938                         default:
2939                                 if ((error = pf_begin_rules(&ioe->ticket,
2940                                     ioe->rs_num, ioe->anchor))) {
2941                                         PF_RULES_WUNLOCK();
2942                                         free(ioes, M_TEMP);
2943                                         goto fail;
2944                                 }
2945                                 break;
2946                         }
2947                 }
2948                 PF_RULES_WUNLOCK();
2949                 error = copyout(ioes, io->array, totlen);
2950                 free(ioes, M_TEMP);
2951                 break;
2952         }
2953
2954         case DIOCXROLLBACK: {
2955                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
2956                 struct pfioc_trans_e    *ioe, *ioes;
2957                 size_t                   totlen;
2958                 int                      i;
2959
2960                 if (io->esize != sizeof(*ioe)) {
2961                         error = ENODEV;
2962                         break;
2963                 }
2964                 totlen = sizeof(struct pfioc_trans_e) * io->size;
2965                 ioes = malloc(totlen, M_TEMP, M_WAITOK);
2966                 error = copyin(io->array, ioes, totlen);
2967                 if (error) {
2968                         free(ioes, M_TEMP);
2969                         break;
2970                 }
2971                 PF_RULES_WLOCK();
2972                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2973                         switch (ioe->rs_num) {
2974 #ifdef ALTQ
2975                         case PF_RULESET_ALTQ:
2976                                 if (ioe->anchor[0]) {
2977                                         PF_RULES_WUNLOCK();
2978                                         free(ioes, M_TEMP);
2979                                         error = EINVAL;
2980                                         goto fail;
2981                                 }
2982                                 if ((error = pf_rollback_altq(ioe->ticket))) {
2983                                         PF_RULES_WUNLOCK();
2984                                         free(ioes, M_TEMP);
2985                                         goto fail; /* really bad */
2986                                 }
2987                                 break;
2988 #endif /* ALTQ */
2989                         case PF_RULESET_TABLE:
2990                             {
2991                                 struct pfr_table table;
2992
2993                                 bzero(&table, sizeof(table));
2994                                 strlcpy(table.pfrt_anchor, ioe->anchor,
2995                                     sizeof(table.pfrt_anchor));
2996                                 if ((error = pfr_ina_rollback(&table,
2997                                     ioe->ticket, NULL, 0))) {
2998                                         PF_RULES_WUNLOCK();
2999                                         free(ioes, M_TEMP);
3000                                         goto fail; /* really bad */
3001                                 }
3002                                 break;
3003                             }
3004                         default:
3005                                 if ((error = pf_rollback_rules(ioe->ticket,
3006                                     ioe->rs_num, ioe->anchor))) {
3007                                         PF_RULES_WUNLOCK();
3008                                         free(ioes, M_TEMP);
3009                                         goto fail; /* really bad */
3010                                 }
3011                                 break;
3012                         }
3013                 }
3014                 PF_RULES_WUNLOCK();
3015                 free(ioes, M_TEMP);
3016                 break;
3017         }
3018
3019         case DIOCXCOMMIT: {
3020                 struct pfioc_trans      *io = (struct pfioc_trans *)addr;
3021                 struct pfioc_trans_e    *ioe, *ioes;
3022                 struct pf_ruleset       *rs;
3023                 size_t                   totlen;
3024                 int                      i;
3025
3026                 if (io->esize != sizeof(*ioe)) {
3027                         error = ENODEV;
3028                         break;
3029                 }
3030                 totlen = sizeof(struct pfioc_trans_e) * io->size;
3031                 ioes = malloc(totlen, M_TEMP, M_WAITOK);
3032                 error = copyin(io->array, ioes, totlen);
3033                 if (error) {
3034                         free(ioes, M_TEMP);
3035                         break;
3036                 }
3037                 PF_RULES_WLOCK();
3038                 /* First makes sure everything will succeed. */
3039                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3040                         switch (ioe->rs_num) {
3041 #ifdef ALTQ
3042                         case PF_RULESET_ALTQ:
3043                                 if (ioe->anchor[0]) {
3044                                         PF_RULES_WUNLOCK();
3045                                         free(ioes, M_TEMP);
3046                                         error = EINVAL;
3047                                         goto fail;
3048                                 }
3049                                 if (!V_altqs_inactive_open || ioe->ticket !=
3050                                     V_ticket_altqs_inactive) {
3051                                         PF_RULES_WUNLOCK();
3052                                         free(ioes, M_TEMP);
3053                                         error = EBUSY;
3054                                         goto fail;
3055                                 }
3056                                 break;
3057 #endif /* ALTQ */
3058                         case PF_RULESET_TABLE:
3059                                 rs = pf_find_ruleset(ioe->anchor);
3060                                 if (rs == NULL || !rs->topen || ioe->ticket !=
3061                                     rs->tticket) {
3062                                         PF_RULES_WUNLOCK();
3063                                         free(ioes, M_TEMP);
3064                                         error = EBUSY;
3065                                         goto fail;
3066                                 }
3067                                 break;
3068                         default:
3069                                 if (ioe->rs_num < 0 || ioe->rs_num >=
3070                                     PF_RULESET_MAX) {
3071                                         PF_RULES_WUNLOCK();
3072                                         free(ioes, M_TEMP);
3073                                         error = EINVAL;
3074                                         goto fail;
3075                                 }
3076                                 rs = pf_find_ruleset(ioe->anchor);
3077                                 if (rs == NULL ||
3078                                     !rs->rules[ioe->rs_num].inactive.open ||
3079                                     rs->rules[ioe->rs_num].inactive.ticket !=
3080                                     ioe->ticket) {
3081                                         PF_RULES_WUNLOCK();
3082                                         free(ioes, M_TEMP);
3083                                         error = EBUSY;
3084                                         goto fail;
3085                                 }
3086                                 break;
3087                         }
3088                 }
3089                 /* Now do the commit - no errors should happen here. */
3090                 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3091                         switch (ioe->rs_num) {
3092 #ifdef ALTQ
3093                         case PF_RULESET_ALTQ:
3094                                 if ((error = pf_commit_altq(ioe->ticket))) {
3095                                         PF_RULES_WUNLOCK();
3096                                         free(ioes, M_TEMP);
3097                                         goto fail; /* really bad */
3098                                 }
3099                                 break;
3100 #endif /* ALTQ */
3101                         case PF_RULESET_TABLE:
3102                             {
3103                                 struct pfr_table table;
3104
3105                                 bzero(&table, sizeof(table));
3106                                 strlcpy(table.pfrt_anchor, ioe->anchor,
3107                                     sizeof(table.pfrt_anchor));
3108                                 if ((error = pfr_ina_commit(&table,
3109                                     ioe->ticket, NULL, NULL, 0))) {
3110                                         PF_RULES_WUNLOCK();
3111                                         free(ioes, M_TEMP);
3112                                         goto fail; /* really bad */
3113                                 }
3114                                 break;
3115                             }
3116                         default:
3117                                 if ((error = pf_commit_rules(ioe->ticket,
3118                                     ioe->rs_num, ioe->anchor))) {
3119                                         PF_RULES_WUNLOCK();
3120                                         free(ioes, M_TEMP);
3121                                         goto fail; /* really bad */
3122                                 }
3123                                 break;
3124                         }
3125                 }
3126                 PF_RULES_WUNLOCK();
3127                 free(ioes, M_TEMP);
3128                 break;
3129         }
3130
3131         case DIOCGETSRCNODES: {
3132                 struct pfioc_src_nodes  *psn = (struct pfioc_src_nodes *)addr;
3133                 struct pf_srchash       *sh;
3134                 struct pf_src_node      *n, *p, *pstore;
3135                 uint32_t                 i, nr = 0;
3136
3137                 if (psn->psn_len == 0) {
3138                         for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3139                             i++, sh++) {
3140                                 PF_HASHROW_LOCK(sh);
3141                                 LIST_FOREACH(n, &sh->nodes, entry)
3142                                         nr++;
3143                                 PF_HASHROW_UNLOCK(sh);
3144                         }
3145                         psn->psn_len = sizeof(struct pf_src_node) * nr;
3146                         break;
3147                 }
3148
3149                 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK);
3150                 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3151                     i++, sh++) {
3152                     PF_HASHROW_LOCK(sh);
3153                     LIST_FOREACH(n, &sh->nodes, entry) {
3154                         int     secs = time_uptime, diff;
3155
3156                         if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3157                                 break;
3158
3159                         bcopy(n, p, sizeof(struct pf_src_node));
3160                         if (n->rule.ptr != NULL)
3161                                 p->rule.nr = n->rule.ptr->nr;
3162                         p->creation = secs - p->creation;
3163                         if (p->expire > secs)
3164                                 p->expire -= secs;
3165                         else
3166                                 p->expire = 0;
3167
3168                         /* Adjust the connection rate estimate. */
3169                         diff = secs - n->conn_rate.last;
3170                         if (diff >= n->conn_rate.seconds)
3171                                 p->conn_rate.count = 0;
3172                         else
3173                                 p->conn_rate.count -=
3174                                     n->conn_rate.count * diff /
3175                                     n->conn_rate.seconds;
3176                         p++;
3177                         nr++;
3178                     }
3179                     PF_HASHROW_UNLOCK(sh);
3180                 }
3181                 error = copyout(pstore, psn->psn_src_nodes,
3182                     sizeof(struct pf_src_node) * nr);
3183                 if (error) {
3184                         free(pstore, M_TEMP);
3185                         break;
3186                 }
3187                 psn->psn_len = sizeof(struct pf_src_node) * nr;
3188                 free(pstore, M_TEMP);
3189                 break;
3190         }
3191
3192         case DIOCCLRSRCNODES: {
3193
3194                 pf_clear_srcnodes(NULL);
3195                 pf_purge_expired_src_nodes();
3196                 break;
3197         }
3198
3199         case DIOCKILLSRCNODES:
3200                 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
3201                 break;
3202
3203         case DIOCSETHOSTID: {
3204                 u_int32_t       *hostid = (u_int32_t *)addr;
3205
3206                 PF_RULES_WLOCK();
3207                 if (*hostid == 0)
3208                         V_pf_status.hostid = arc4random();
3209                 else
3210                         V_pf_status.hostid = *hostid;
3211                 PF_RULES_WUNLOCK();
3212                 break;
3213         }
3214
3215         case DIOCOSFPFLUSH:
3216                 PF_RULES_WLOCK();
3217                 pf_osfp_flush();
3218                 PF_RULES_WUNLOCK();
3219                 break;
3220
3221         case DIOCIGETIFACES: {
3222                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3223                 struct pfi_kif *ifstore;
3224                 size_t bufsiz;
3225
3226                 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3227                         error = ENODEV;
3228                         break;
3229                 }
3230
3231                 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3232                 ifstore = malloc(bufsiz, M_TEMP, M_WAITOK);
3233                 PF_RULES_RLOCK();
3234                 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3235                 PF_RULES_RUNLOCK();
3236                 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3237                 free(ifstore, M_TEMP);
3238                 break;
3239         }
3240
3241         case DIOCSETIFFLAG: {
3242                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3243
3244                 PF_RULES_WLOCK();
3245                 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3246                 PF_RULES_WUNLOCK();
3247                 break;
3248         }
3249
3250         case DIOCCLRIFFLAG: {
3251                 struct pfioc_iface *io = (struct pfioc_iface *)addr;
3252
3253                 PF_RULES_WLOCK();
3254                 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3255                 PF_RULES_WUNLOCK();
3256                 break;
3257         }
3258
3259         default:
3260                 error = ENODEV;
3261                 break;
3262         }
3263 fail:
3264         if (sx_xlocked(&pf_ioctl_lock))
3265                 sx_xunlock(&pf_ioctl_lock);
3266         CURVNET_RESTORE();
3267
3268         return (error);
3269 }
3270
3271 void
3272 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3273 {
3274         bzero(sp, sizeof(struct pfsync_state));
3275
3276         /* copy from state key */
3277         sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3278         sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3279         sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3280         sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3281         sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3282         sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3283         sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3284         sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3285         sp->proto = st->key[PF_SK_WIRE]->proto;
3286         sp->af = st->key[PF_SK_WIRE]->af;
3287
3288         /* copy from state */
3289         strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3290         bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3291         sp->creation = htonl(time_uptime - st->creation);
3292         sp->expire = pf_state_expires(st);
3293         if (sp->expire <= time_uptime)
3294                 sp->expire = htonl(0);
3295         else
3296                 sp->expire = htonl(sp->expire - time_uptime);
3297
3298         sp->direction = st->direction;
3299         sp->log = st->log;
3300         sp->timeout = st->timeout;
3301         sp->state_flags = st->state_flags;
3302         if (st->src_node)
3303                 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3304         if (st->nat_src_node)
3305                 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3306
3307         sp->id = st->id;
3308         sp->creatorid = st->creatorid;
3309         pf_state_peer_hton(&st->src, &sp->src);
3310         pf_state_peer_hton(&st->dst, &sp->dst);
3311
3312         if (st->rule.ptr == NULL)
3313                 sp->rule = htonl(-1);
3314         else
3315                 sp->rule = htonl(st->rule.ptr->nr);
3316         if (st->anchor.ptr == NULL)
3317                 sp->anchor = htonl(-1);
3318         else
3319                 sp->anchor = htonl(st->anchor.ptr->nr);
3320         if (st->nat_rule.ptr == NULL)
3321                 sp->nat_rule = htonl(-1);
3322         else
3323                 sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3324
3325         pf_state_counter_hton(st->packets[0], sp->packets[0]);
3326         pf_state_counter_hton(st->packets[1], sp->packets[1]);
3327         pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3328         pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3329
3330 }
3331
3332 static void
3333 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3334 {
3335         struct pfr_ktable *kt;
3336
3337         KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3338
3339         kt = aw->p.tbl;
3340         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3341                 kt = kt->pfrkt_root;
3342         aw->p.tbl = NULL;
3343         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3344                 kt->pfrkt_cnt : -1;
3345 }
3346
3347 /*
3348  * XXX - Check for version missmatch!!!
3349  */
3350 static void
3351 pf_clear_states(void)
3352 {
3353         struct pf_state *s;
3354         u_int i;
3355
3356         for (i = 0; i <= pf_hashmask; i++) {
3357                 struct pf_idhash *ih = &V_pf_idhash[i];
3358 relock:
3359                 PF_HASHROW_LOCK(ih);
3360                 LIST_FOREACH(s, &ih->states, entry) {
3361                         s->timeout = PFTM_PURGE;
3362                         /* Don't send out individual delete messages. */
3363                         s->sync_state = PFSTATE_NOSYNC;
3364                         pf_unlink_state(s, PF_ENTER_LOCKED);
3365                         goto relock;
3366                 }
3367                 PF_HASHROW_UNLOCK(ih);
3368         }
3369 }
3370
3371 static int
3372 pf_clear_tables(void)
3373 {
3374         struct pfioc_table io;
3375         int error;
3376
3377         bzero(&io, sizeof(io));
3378
3379         error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3380             io.pfrio_flags);
3381
3382         return (error);
3383 }
3384
3385 static void
3386 pf_clear_srcnodes(struct pf_src_node *n)
3387 {
3388         struct pf_state *s;
3389         int i;
3390
3391         for (i = 0; i <= pf_hashmask; i++) {
3392                 struct pf_idhash *ih = &V_pf_idhash[i];
3393
3394                 PF_HASHROW_LOCK(ih);
3395                 LIST_FOREACH(s, &ih->states, entry) {
3396                         if (n == NULL || n == s->src_node)
3397                                 s->src_node = NULL;
3398                         if (n == NULL || n == s->nat_src_node)
3399                                 s->nat_src_node = NULL;
3400                 }
3401                 PF_HASHROW_UNLOCK(ih);
3402         }
3403
3404         if (n == NULL) {
3405                 struct pf_srchash *sh;
3406
3407                 for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3408                     i++, sh++) {
3409                         PF_HASHROW_LOCK(sh);
3410                         LIST_FOREACH(n, &sh->nodes, entry) {
3411                                 n->expire = 1;
3412                                 n->states = 0;
3413                         }
3414                         PF_HASHROW_UNLOCK(sh);
3415                 }
3416         } else {
3417                 /* XXX: hash slot should already be locked here. */
3418                 n->expire = 1;
3419                 n->states = 0;
3420         }
3421 }
3422
3423 static void
3424 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
3425 {
3426         struct pf_src_node_list  kill;
3427
3428         LIST_INIT(&kill);
3429         for (int i = 0; i <= pf_srchashmask; i++) {
3430                 struct pf_srchash *sh = &V_pf_srchash[i];
3431                 struct pf_src_node *sn, *tmp;
3432
3433                 PF_HASHROW_LOCK(sh);
3434                 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
3435                         if (PF_MATCHA(psnk->psnk_src.neg,
3436                               &psnk->psnk_src.addr.v.a.addr,
3437                               &psnk->psnk_src.addr.v.a.mask,
3438                               &sn->addr, sn->af) &&
3439                             PF_MATCHA(psnk->psnk_dst.neg,
3440                               &psnk->psnk_dst.addr.v.a.addr,
3441                               &psnk->psnk_dst.addr.v.a.mask,
3442                               &sn->raddr, sn->af)) {
3443                                 pf_unlink_src_node(sn);
3444                                 LIST_INSERT_HEAD(&kill, sn, entry);
3445                                 sn->expire = 1;
3446                         }
3447                 PF_HASHROW_UNLOCK(sh);
3448         }
3449
3450         for (int i = 0; i <= pf_hashmask; i++) {
3451                 struct pf_idhash *ih = &V_pf_idhash[i];
3452                 struct pf_state *s;
3453
3454                 PF_HASHROW_LOCK(ih);
3455                 LIST_FOREACH(s, &ih->states, entry) {
3456                         if (s->src_node && s->src_node->expire == 1)
3457                                 s->src_node = NULL;
3458                         if (s->nat_src_node && s->nat_src_node->expire == 1)
3459                                 s->nat_src_node = NULL;
3460                 }
3461                 PF_HASHROW_UNLOCK(ih);
3462         }
3463
3464         psnk->psnk_killed = pf_free_src_nodes(&kill);
3465 }
3466
3467 /*
3468  * XXX - Check for version missmatch!!!
3469  */
3470
3471 /*
3472  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3473  */
3474 static int
3475 shutdown_pf(void)
3476 {
3477         int error = 0;
3478         u_int32_t t[5];
3479         char nn = '\0';
3480
3481         V_pf_status.running = 0;
3482
3483         counter_u64_free(V_pf_default_rule.states_cur);
3484         counter_u64_free(V_pf_default_rule.states_tot);
3485         counter_u64_free(V_pf_default_rule.src_nodes);
3486
3487         for (int i = 0; i < PFRES_MAX; i++)
3488                 counter_u64_free(V_pf_status.counters[i]);
3489         for (int i = 0; i < LCNT_MAX; i++)
3490                 counter_u64_free(V_pf_status.lcounters[i]);
3491         for (int i = 0; i < FCNT_MAX; i++)
3492                 counter_u64_free(V_pf_status.fcounters[i]);
3493         for (int i = 0; i < SCNT_MAX; i++)
3494                 counter_u64_free(V_pf_status.scounters[i]);
3495
3496         do {
3497                 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3498                     != 0) {
3499                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3500                         break;
3501                 }
3502                 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3503                     != 0) {
3504                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3505                         break;          /* XXX: rollback? */
3506                 }
3507                 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3508                     != 0) {
3509                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3510                         break;          /* XXX: rollback? */
3511                 }
3512                 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3513                     != 0) {
3514                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3515                         break;          /* XXX: rollback? */
3516                 }
3517                 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3518                     != 0) {
3519                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3520                         break;          /* XXX: rollback? */
3521                 }
3522
3523                 /* XXX: these should always succeed here */
3524                 pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3525                 pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3526                 pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3527                 pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3528                 pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3529
3530                 if ((error = pf_clear_tables()) != 0)
3531                         break;
3532
3533 #ifdef ALTQ
3534                 if ((error = pf_begin_altq(&t[0])) != 0) {
3535                         DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3536                         break;
3537                 }
3538                 pf_commit_altq(t[0]);
3539 #endif
3540
3541                 pf_clear_states();
3542
3543                 pf_clear_srcnodes(NULL);
3544
3545                 /* status does not use malloced mem so no need to cleanup */
3546                 /* fingerprints and interfaces have their own cleanup code */
3547         } while(0);
3548
3549         return (error);
3550 }
3551
3552 #ifdef INET
3553 static int
3554 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3555     struct inpcb *inp)
3556 {
3557         int chk;
3558
3559         chk = pf_test(PF_IN, ifp, m, inp);
3560         if (chk && *m) {
3561                 m_freem(*m);
3562                 *m = NULL;
3563         }
3564
3565         return (chk);
3566 }
3567
3568 static int
3569 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3570     struct inpcb *inp)
3571 {
3572         int chk;
3573
3574         chk = pf_test(PF_OUT, ifp, m, inp);
3575         if (chk && *m) {
3576                 m_freem(*m);
3577                 *m = NULL;
3578         }
3579
3580         return (chk);
3581 }
3582 #endif
3583
3584 #ifdef INET6
3585 static int
3586 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3587     struct inpcb *inp)
3588 {
3589         int chk;
3590
3591         /*
3592          * In case of loopback traffic IPv6 uses the real interface in
3593          * order to support scoped addresses. In order to support stateful
3594          * filtering we have change this to lo0 as it is the case in IPv4.
3595          */
3596         CURVNET_SET(ifp->if_vnet);
3597         chk = pf_test6(PF_IN, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
3598         CURVNET_RESTORE();
3599         if (chk && *m) {
3600                 m_freem(*m);
3601                 *m = NULL;
3602         }
3603         return chk;
3604 }
3605
3606 static int
3607 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3608     struct inpcb *inp)
3609 {
3610         int chk;
3611
3612         CURVNET_SET(ifp->if_vnet);
3613         chk = pf_test6(PF_OUT, ifp, m, inp);
3614         CURVNET_RESTORE();
3615         if (chk && *m) {
3616                 m_freem(*m);
3617                 *m = NULL;
3618         }
3619         return chk;
3620 }
3621 #endif /* INET6 */
3622
3623 static int
3624 hook_pf(void)
3625 {
3626 #ifdef INET
3627         struct pfil_head *pfh_inet;
3628 #endif
3629 #ifdef INET6
3630         struct pfil_head *pfh_inet6;
3631 #endif
3632
3633         if (V_pf_pfil_hooked)
3634                 return (0);
3635
3636 #ifdef INET
3637         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3638         if (pfh_inet == NULL)
3639                 return (ESRCH); /* XXX */
3640         pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
3641         pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
3642 #endif
3643 #ifdef INET6
3644         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3645         if (pfh_inet6 == NULL) {
3646 #ifdef INET
3647                 pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3648                     pfh_inet);
3649                 pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3650                     pfh_inet);
3651 #endif
3652                 return (ESRCH); /* XXX */
3653         }
3654         pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
3655         pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
3656 #endif
3657
3658         V_pf_pfil_hooked = 1;
3659         return (0);
3660 }
3661
3662 static int
3663 dehook_pf(void)
3664 {
3665 #ifdef INET
3666         struct pfil_head *pfh_inet;
3667 #endif
3668 #ifdef INET6
3669         struct pfil_head *pfh_inet6;
3670 #endif
3671
3672         if (V_pf_pfil_hooked == 0)
3673                 return (0);
3674
3675 #ifdef INET
3676         pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3677         if (pfh_inet == NULL)
3678                 return (ESRCH); /* XXX */
3679         pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3680             pfh_inet);
3681         pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3682             pfh_inet);
3683 #endif
3684 #ifdef INET6
3685         pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3686         if (pfh_inet6 == NULL)
3687                 return (ESRCH); /* XXX */
3688         pfil_remove_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
3689             pfh_inet6);
3690         pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
3691             pfh_inet6);
3692 #endif
3693
3694         V_pf_pfil_hooked = 0;
3695         return (0);
3696 }
3697
3698 static int
3699 pf_load(void)
3700 {
3701         int error;
3702
3703         VNET_ITERATOR_DECL(vnet_iter);
3704
3705         VNET_LIST_RLOCK();
3706         VNET_FOREACH(vnet_iter) {
3707                 CURVNET_SET(vnet_iter);
3708                 V_pf_pfil_hooked = 0;
3709                 V_pf_end_threads = 0;
3710                 TAILQ_INIT(&V_pf_tags);
3711                 TAILQ_INIT(&V_pf_qids);
3712                 CURVNET_RESTORE();
3713         }
3714         VNET_LIST_RUNLOCK();
3715
3716         rw_init(&pf_rules_lock, "pf rulesets");
3717         sx_init(&pf_ioctl_lock, "pf ioctl");
3718
3719         pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
3720         if ((error = pfattach()) != 0)
3721                 return (error);
3722
3723         return (0);
3724 }
3725
3726 static int
3727 pf_unload(void)
3728 {
3729         int error = 0;
3730
3731         V_pf_status.running = 0;
3732         swi_remove(V_pf_swi_cookie);
3733         error = dehook_pf();
3734         if (error) {
3735                 /*
3736                  * Should not happen!
3737                  * XXX Due to error code ESRCH, kldunload will show
3738                  * a message like 'No such process'.
3739                  */
3740                 printf("%s : pfil unregisteration fail\n", __FUNCTION__);
3741                 return error;
3742         }
3743         PF_RULES_WLOCK();
3744         shutdown_pf();
3745         V_pf_end_threads = 1;
3746         while (V_pf_end_threads < 2) {
3747                 wakeup_one(pf_purge_thread);
3748                 rw_sleep(pf_purge_thread, &pf_rules_lock, 0, "pftmo", 0);
3749         }
3750         PF_RULES_WUNLOCK();
3751         pf_normalize_cleanup();
3752         pfi_cleanup();
3753         pfr_cleanup();
3754         pf_osfp_flush();
3755         pf_cleanup();
3756         if (IS_DEFAULT_VNET(curvnet))
3757                 pf_mtag_cleanup();
3758         destroy_dev(pf_dev);
3759         rw_destroy(&pf_rules_lock);
3760         sx_destroy(&pf_ioctl_lock);
3761
3762         return (error);
3763 }
3764
3765 static int
3766 pf_modevent(module_t mod, int type, void *data)
3767 {
3768         int error = 0;
3769
3770         switch(type) {
3771         case MOD_LOAD:
3772                 error = pf_load();
3773                 break;
3774         case MOD_QUIESCE:
3775                 /*
3776                  * Module should not be unloaded due to race conditions.
3777                  */
3778                 error = EBUSY;
3779                 break;
3780         case MOD_UNLOAD:
3781                 error = pf_unload();
3782                 break;
3783         default:
3784                 error = EINVAL;
3785                 break;
3786         }
3787
3788         return (error);
3789 }
3790
3791 static moduledata_t pf_mod = {
3792         "pf",
3793         pf_modevent,
3794         0
3795 };
3796
3797 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_FIRST);
3798 MODULE_VERSION(pf, PF_MODVER);